IMO 1982
Read the problems at the official link. The 1982 IMO problems are Β© the International Mathematical Olympiad, so we don't reproduce them here. Open the 1982 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 Almost additive function pinned down by a few values
- Hint 1. Find $f(1)$ and $f(3)$, then show $f(3n)\ge n$ for all $n$.
- Hint 2. If $f(3n)>n$ for some $n\le3333$, what happens to $f(9999)$? Then use $f(3x)\ge3f(x)$.
Our solution (not official)
Answer: $f(1982)=660$
Key idea: $f$ is superadditive, so $f(3n)\ge n$. The value at $9999$ forces equality up to $9999$, and superadditivity squeezes $f(1982)$ between $660$ and $660.67$.
The condition says $f(m+n)\ge f(m)+f(n)$, with excess at most $1$.
- $0=f(2)\ge2f(1)$, so $f(1)=0$.
- $f(3)\in\{f(2)+f(1),\,f(2)+f(1)+1\}=\{0,1\}$ and $f(3)>0$, so $f(3)=1$.
- $f(3(n+1))\ge f(3n)+f(3)$, so by induction $f(3n)\ge n$.
$f(3n)=n$ for $n\le3333$. If $f(3n)\ge n+1$ for some $n<3333$, then $$f(9999)\ge f(3n)+f(9999-3n)\ge(n+1)+(3333-n)=3334,$$ a contradiction. So $f(3n)=n$ for these $n$, in particular $f(5946)=1982$ and $f(1980)=660$.
Squeeze. From superadditivity, $f(5946)\ge3f(1982)$, so $f(1982)\le\frac{1982}3<661$. And $f(1982)\ge f(1980)+f(2)=660$. So $f(1982)=660$.
(Such a function exists: $f(n)=\lfloor n/3\rfloor$ satisfies every condition.)
How we checked it: script: floor(n/3) meets all conditions; the argument above is exact
P2 Reflected incircle touch points and side midpoints concur
- Hint 1. The reflection of a touch point in an angle bisector stays on the incircle. Where?
- Hint 2. Show that $S_1S_2$ is parallel to the side $A_1A_2$. Then the triangles $S_1S_2S_3$ and $M_1M_2M_3$ have parallel sides.
Our solution (not official)
Key idea: the reflected points lie on the incircle and form a triangle whose sides are parallel to those of the medial triangle. Two triangles with parallel sides and different sizes are related by a homothety, and its center is the common point.
Let $I$ and $r$ be the incenter and inradius. The incircle is symmetric in each bisector $A_iI$, so each $S_i$ is on the incircle. Describe points of the incircle by the direction angle of the ray from $I$: say $T_i$ has angle $\theta_i$.
Positions of the $S_i$. The bisector $A_1I$ bisects the angle $T_2IT_3$ (the two touch points on the sides from $A_1$ are mirror images in it), so it has direction $\frac{\theta_2+\theta_3}2$. Reflecting in it sends the angle $\theta$ to $\theta_2+\theta_3-\theta$. So $S_1$ has angle $\theta_2+\theta_3-\theta_1$, and likewise $S_2$ has $\theta_1+\theta_3-\theta_2$ and $S_3$ has $\theta_1+\theta_2-\theta_3$.
Parallel sides. A chord between angles $\psi$ and $\psi'$ is perpendicular to the direction $\frac{\psi+\psi'}2$. For $S_1S_2$ this is $\theta_3$: the chord $S_1S_2$ is perpendicular to $IT_3$, so it is parallel to side $A_1A_2$, which is tangent at $T_3$. (The points $S_1$, $S_2$ are different: otherwise $2\theta_1\equiv2\theta_2$, making $T_1$, $T_2$ equal or opposite, and two sides parallel.) The midline $M_1M_2$ is also parallel to $A_1A_2$. In the same way $S_2S_3\parallel M_2M_3$ and $S_3S_1\parallel M_3M_1$.
Homothety. Two triangles with corresponding sides parallel are related by a translation or by a homothety. The medial triangle has circumradius $\frac R2$, while $S_1S_2S_3$ has circumradius $r$. Euler’s inequality $R\ge2r$, with equality only for an equilateral triangle, gives $\frac R2>r$. So the map is not a translation but a homothety, and the lines joining corresponding vertices, $M_iS_i$, pass through its center. (Each $M_i\ne S_i$: the incircle meets side $a_i$ only at $T_i$, and $S_i=T_i$ would put $T_i$ on the bisector from $A_i$, making the triangle isosceles.)
How we checked it: script: random non-isosceles triangles; the three lines meet within 1e-14, and each S_i is on the incircle
P3 Decreasing sequence: a sum of squared ratios nears four
- Hint 1. AMβGM: $\frac{x_i^2}{x_{i+1}}+4x_{i+1}\ge4x_i$. Sum this up.
- Hint 2. If the sequence does not tend to $0$, each term $\frac{x_i^2}{x_{i+1}}\ge x_i$ stays large. For (b), try a geometric sequence.
Our solution (not official)
Key idea: AMβGM gives a telescoping lower bound $4(x_0-x_n)$. Either $x_n$ gets small, or the terms themselves stay large.
Let $S_n=\sum_{i=0}^{n-1}\frac{x_i^2}{x_{i+1}}$.
(a) By AMβGM, $\frac{x_i^2}{x_{i+1}}+4x_{i+1}\ge2\sqrt{4x_i^2}=4x_i$. Summing for $i=0,\dots,n-1$, $$S_n\ge4\sum_{i=0}^{n-1}x_i-4\sum_{i=1}^{n}x_i=4(x_0-x_n)=4-4x_n .$$
- If some $x_n\le\frac1{4000}$, then $S_n\ge4-0.001=3.999$.
- Otherwise all $x_i>\frac1{4000}$, and since $x_{i+1}\le x_i$, each term $\frac{x_i^2}{x_{i+1}}\ge x_i>\frac1{4000}$. Then $S_{16000}>4$.
(b) Take $x_n=2^{-n}$. Then $\frac{x_i^2}{x_{i+1}}=\frac{2^{-2i}}{2^{-i-1}}=2^{1-i}$, and $$S_n=\sum_{i=0}^{n-1}2^{1-i}=4\left(1-2^{-n}\right)<4 .$$
How we checked it: script: exact sums for the example equal 4(1 β 2^(βn)), below 4
P4 Cubic form: solutions come in threes; one value impossible
- Hint 1. Look for a linear change of variables $(x,y)\mapsto(y-x,-x)$ that keeps the form unchanged. What is its order?
- Hint 2. For $2891$, work modulo $9$. Show the form modulo $9$ depends only on $x,y$ modulo $3$.
Our solution (not official)
Key idea: a linear map of order $3$ preserves the form and moves every nonzero solution, so solutions come in triples. Modulo $9$ the form takes only a few values, and $2891$ is not among them.
Let $F(x,y)=x^3-3xy^2+y^3$ and $\tau(x,y)=(y-x,-x)$.
$\tau$ preserves $F$. $$F(y-x,-x)=(y-x)^3-3(y-x)x^2-x^3=y^3-3y^2x+3yx^2-x^3-3x^2y+3x^3-x^3=F(x,y).$$ $\tau$ has order $3$: $\tau^2(x,y)=(-y,x-y)$ and $\tau^3(x,y)=(x,y)$. Its only fixed point is $(0,0)$ (from $y-x=x$ and $-x=y$). So if $F(x,y)=n>0$, then $(x,y)\ne(0,0)$, and $(x,y)$, $\tau(x,y)$, $\tau^2(x,y)$ are three different solutions (an orbit of size $3$, since its size divides $3$ and is not $1$).
No solutions for $2891$. Modulo $9$: $(x+3s)^3\equiv x^3$ and $3(x+3s)y^2\equiv3xy^2$, and the same for $y$. So $F\bmod9$ depends only on $x,y\bmod3$. Since $2891\equiv2\pmod 9$, and $F\equiv x^3+y^3\equiv x+y\pmod3$, we need $x+y\equiv2\pmod3$, that is, $(x,y)\equiv(0,2),(2,0)$ or $(1,1)\pmod3$. These give $$F(0,2)=8,\qquad F(2,0)=8,\qquad F(1,1)=-1\equiv8\pmod9 .$$ Never $2$. So $F(x,y)=2891$ has no integer solutions.
How we checked it: script: F mod 9 takes only the values 0, 1, 3, 6, 8; no solution with |x|, |y| β€ 300; sympy confirms the invariance
P5 Regular hexagon: two diagonal points collinear with a vertex
- Hint 1. Put the hexagon’s center at the origin with unit circumradius and write $M$ and $N$ in coordinates.
- Hint 2. Collinearity of $B,M,N$ is a $2\times2$ determinant equal to $0$.
Our solution (not official)
Answer: $r=\dfrac1{\sqrt3}$
Key idea: with coordinates, collinearity is one determinant, and it reduces to $3r^2=1$.
Take the hexagon with center at the origin and vertices on the unit circle: $$A=(1,0),\ B=\left(\tfrac12,\tfrac{\sqrt3}2\right),\ C=\left(-\tfrac12,\tfrac{\sqrt3}2\right),\ E=\left(-\tfrac12,-\tfrac{\sqrt3}2\right).$$ Then $$M=A+r(C-A)=\left(1-\tfrac32r,\ \tfrac{\sqrt3}2r\right),\qquad N=C+r(E-C)=\left(-\tfrac12,\ \tfrac{\sqrt3}2-\sqrt3\,r\right).$$ Relative to $B$: $M-B=\left(\tfrac12-\tfrac32r,\ \tfrac{\sqrt3}2(r-1)\right)$ and $N-B=\left(-1,\ -\sqrt3\,r\right)$. They are parallel when $$\left(\tfrac12-\tfrac32r\right)(-\sqrt3\,r)-\tfrac{\sqrt3}2(r-1)(-1)=\frac{\sqrt3}2\left(3r^2-1\right)=0 .$$ With $0<r<1$, this gives $r=\frac1{\sqrt3}$.
How we checked it: script: sympy solves the collinearity condition symbolically
P6 Path covering a square’s boundary nearly returns on itself
- Hint 1. Pick a corner $W$ and a moment when the path is near $W$. If some point of the two sides away from $W$ is approached both before and after that moment, you are done: the detour to $W$ and back is long.
- Hint 2. Which corner? Avoid the corner the path reaches first and the corner it leaves last. Then use that the two far sides form a connected set.
Our solution (not official)
Key idea: find a moment when the path is near a corner $W$, while some point $P$ on the two sides away from $W$ is approached both before and after that moment. Going from near $P$ to near $W$ and back costs at least $2\cdot 99$.
Let the path have length $\ell$ and parametrize it by arc length, $L(t)$ for $0\le t\le\ell$, from $A_0$ to $A_n$. The path does not meet itself and $A_0\ne A_n$, so each of its points has exactly one parameter. The part between $L(s)$ and $L(t)$ is $L([s,t])$, of length $|t-s|$.
For a corner $V$ let $f(V)$ and $g(V)$ be the first and the last time $t$ with $|L(t)-V|\le\frac12$. They exist: the set of such times is closed and, by the hypothesis, not empty.
Choosing the corner. Let $V_1$ be a corner with the smallest $f$, and $V_2$ a corner with the largest $g$. Pick a corner $W$ different from both (there are four corners), and let $\tau=f(W)$. Then
- $f(V_1)\le f(W)=\tau$: the path is near the corner $V_1\ne W$ at some time $\le\tau$;
- $g(V_2)\ge g(W)\ge\tau$: the path is near the corner $V_2\ne W$ at some time $\ge\tau$.
The far sides. Let $K$ be the union of the two sides of the square that do not contain $W$. It is a connected broken line through the other three corners, and each of its points is at distance at least $100$ from $W$. Put $$K_-=\{P\in K:\ |P-L(t)|\le\tfrac12\ \text{for some } t\le\tau\},\qquad K_+=\{P\in K:\ |P-L(t)|\le\tfrac12\ \text{for some } t\ge\tau\}.$$ Each is closed (the points of $K$ within $\frac12$ of a compact piece of the path). By the hypothesis $K_-\cup K_+=K$. Also $V_1\in K_-$ and $V_2\in K_+$, so neither is empty. A connected set cannot be split into two disjoint nonempty closed sets, so some point $P$ lies in both.
The two points. Choose $t_1\le\tau\le t_2$ with $|L(t_1)-P|\le\frac12$ and $|L(t_2)-P|\le\frac12$, and let $X=L(t_1)$, $Y=L(t_2)$. Then $XY\le1$. Since $|L(\tau)-W|\le\frac12$ and $PW\ge100$, $$|X-L(\tau)|\ge100-\tfrac12-\tfrac12=99,\qquad |L(\tau)-Y|\ge99.$$ The part of $L$ from $X$ to $Y$ passes through $L(\tau)$, and a path is at least as long as the straight segments joining its points in order. So that part has length at least $99+99=198$.
How we checked it: proof checked step by step (no numerical answer to test)