IMO 1977
Read the problems at the official link. The 1977 IMO problems are Β© the International Mathematical Olympiad, so we don't reproduce them here. Open the 1977 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 Square with inner equilateral triangles hides a regular dodecagon
- Hint 1. Use coordinates with the square’s center at the origin. A quarter turn about the center maps the whole figure to itself.
- Hint 2. So you only need two of the twelve points: one midpoint between two apexes, and one midpoint between an apex and a corner.
Our solution (not official)
Key idea: the figure has quarter-turn symmetry, so two representative midpoints suffice. Both are at the same distance from the center, at angles $135^\circ$ and $\pm15^\circ$; turning them by quarter turns fills in every multiple of $30^\circ$ plus $15^\circ$.
Let the square have side $2$ and center $O$ at the origin: $A=(-1,-1)$, $B=(1,-1)$, $C=(1,1)$, $D=(-1,1)$. An inner equilateral triangle on a side of length $2$ has height $\sqrt3$, so $$K=(0,\sqrt3-1),\quad L=(1-\sqrt3,0),\quad M=(0,1-\sqrt3),\quad N=(\sqrt3-1,0).$$ A quarter turn about $O$ maps $A\to B\to C\to D\to A$ and so $K\to L\to M\to N\to K$. It maps the set of twelve midpoints to itself.
The midpoint of $KL$ is $\left(\frac{1-\sqrt3}2,\frac{\sqrt3-1}2\right)$: at angle $135^\circ$ and distance $\rho$ with $\rho^2=\frac{(\sqrt3-1)^2}{2}=2-\sqrt3$.
The midpoints of $AK$ and $BK$ are $\left(\mp\frac12,\frac{\sqrt3-2}2\right)$. Their squared distance from $O$ is $\frac14+\frac{(2-\sqrt3)^2}{4}=\frac{8-4\sqrt3}{4}=2-\sqrt3=\rho^2$. Their angles: the slope magnitude is $2-\sqrt3=\tan15^\circ$, so they sit at $195^\circ$ and $345^\circ$.
All twelve. Applying quarter turns:
- the four apex-apex midpoints are at $45^\circ,135^\circ,225^\circ,315^\circ$;
- the eight apex-corner midpoints are at $195^\circ,345^\circ$ plus multiples of $90^\circ$, that is, $15^\circ,105^\circ,195^\circ,285^\circ$ and $75^\circ,165^\circ,255^\circ,345^\circ$.
Together these are the twelve angles $15^\circ+30^\circ k$, all at distance $\rho$ from $O$: the vertices of a regular dodecagon.
How we checked it: script: the twelve points all have radius 0.5176 and angles 15Β°, 45Β°, β¦, 345Β°
P2 Window sums of seven negative, eleven positive: longest sequence
- Hint 1. With $17$ terms, arrange $77$ of them in a $7\times11$ table so that rows and columns are windows. Add it up two ways.
- Hint 2. For $16$ terms, try a pattern of $5$s and $-13$s.
Our solution (not official)
Answer: $16$
Key idea: a table whose rows are $11$-windows and whose columns are $7$-windows has a total that must be both positive and negative.
At most $16$. Suppose there are $17$ terms $x_1,\dots,x_{17}$. Fill a $7\times11$ table with $x_{i+j-1}$ in row $i$, column $j$; the largest index is $7+11-1=17$. Each row is $11$ consecutive terms, so the total is positive. Each column is $7$ consecutive terms, so the total is negative. Contradiction.
Sixteen terms work: $$5,\ 5,\ -13,\ 5,\ 5,\ 5,\ -13,\ 5,\ 5,\ -13,\ 5,\ 5,\ 5,\ -13,\ 5,\ 5 .$$ Every $7$-window contains exactly two $-13$s and five $5$s, so its sum is $25-26=-1$. Every $11$-window contains exactly three $-13$s and eight $5$s, so its sum is $40-39=1$. (The $-13$s sit at positions $3,7,10,14$, and one checks each window directly.)
How we checked it: script: all 10 seven-windows of the example are negative and all 6 eleven-windows are positive
P3 Numbers one more than multiples of n: non-unique factoring
- Hint 1. Any element of the set is at least $n+1$, so a product of two elements is at least $(n+1)^2$. Small elements are therefore automatically indecomposable.
- Hint 2. Try $(n-1)^2$, $(2n-1)^2$ and $(n-1)(2n-1)$. Two values of $n$ need a separate example.
Our solution (not official)
Key idea: with $u=n-1$ and $w=2n-1$, the numbers $u^2$, $w^2$, $uw$ all lie in $V_n$, are usually indecomposable, and $u^2\cdot w^2=uw\cdot uw$ gives two different factorizations.
Write $V=V_n$. Its smallest element is $n+1$, and the only element below $2n+1$ is $n+1$.
Since $u\equiv-1$ and $w\equiv-1\pmod n$, the numbers $u^2$, $w^2$ and $uw$ are all $\equiv1\pmod n$, and they exceed $1$, so they lie in $V$.
- $u^2$ is indecomposable: a product of two elements is at least $(n+1)^2>u^2$.
- $w^2$: in a factorization $w^2=pq$ with $p\le q$, we get $p\le w=2n-1$, so $p=n+1$. That needs $n+1\mid(2n-1)^2$. Since $2n-1\equiv-3\pmod{n+1}$, this means $n+1\mid9$, so $n=8$ (as $n>2$).
- $uw$: in a factorization $uw=pq$ with $p\le q$, we get $p\le\sqrt{uw}<\sqrt2\,n<2n+1$, so $p=n+1$. That needs $n+1\mid(n-1)(2n-1)\equiv(-2)(-3)=6$, so $n=5$.
So for $n\ne5,8$, all three are indecomposable, and $$r=(n-1)^2(2n-1)^2=u^2\cdot w^2=uw\cdot uw$$ has two different factorizations (as $u^2\ne uw$).
The two exceptions.
- $n=5$: $V=\{6,11,16,21,26,\dots\}$ and $336=16\cdot21=6\cdot56$. Here $6,16,21$ are below $6^2=36$, so indecomposable, and $56$ is not divisible by $6$, the only element up to $\sqrt{56}$.
- $n=8$: $V=\{9,17,25,33,\dots\}$ and $1089=33\cdot33=9\cdot121$. Here $9$ and $33$ are below $81$, and $121$ is not divisible by $9$, the only element up to $11$.
How we checked it: script: for every n from 3 to 300 except 5 and 8 the three numbers are indecomposable; the two exceptions were found by search
P4 Nonnegative trigonometric polynomial bounds its coefficients
- Hint 1. Write the two parts as $r\cos(\theta-\alpha)$ and $R\cos(2\theta-\beta)$.
- Hint 2. Add $f(\theta)$ and $f(\theta+\pi)$ to kill the first part. Add $f(\theta)$ and $f(\theta+\frac\pi2)$ to kill the second.
Our solution (not official)
Key idea: averaging $f$ at two well-chosen points cancels one harmonic and leaves a clean bound on the other.
Write $a\cos\theta+b\sin\theta=r\cos(\theta-\alpha)$ with $r=\sqrt{a^2+b^2}$, and $A\cos2\theta+B\sin2\theta=R\cos(2\theta-\beta)$ with $R=\sqrt{A^2+B^2}$. So $$f(\theta)=1-r\cos(\theta-\alpha)-R\cos(2\theta-\beta)\ge0\quad\text{for all }\theta .$$
$R\le1$. Shifting $\theta$ by $\pi$ flips the sign of the first cosine and keeps the second, so $f(\theta)+f(\theta+\pi)=2-2R\cos(2\theta-\beta)$. At $\theta=\frac\beta2$ this is $2-2R\ge0$.
$r\le\sqrt2$. Shifting $\theta$ by $\frac\pi2$ flips the sign of the second cosine, and $\cos(\varphi+\frac\pi2)=-\sin\varphi$. With $\varphi=\theta-\alpha$, $$f(\theta)+f(\theta+\tfrac\pi2)=2-r(\cos\varphi-\sin\varphi)=2-\sqrt2\,r\cos(\varphi+\tfrac\pi4).$$ At $\varphi=-\frac\pi4$ this is $2-\sqrt2\,r\ge0$, so $r\le\sqrt2$.
Hence $a^2+b^2\le2$ and $A^2+B^2\le1$.
How we checked it: script: random coefficients that keep f β₯ 0 on a fine grid all satisfy both bounds
P5 Quotient and remainder of a sum of squares
- Hint 1. Since $\frac{a^2+b^2}{a+b}\ge\frac{a+b}2$, the remainder is small compared to the quotient. Show $q=44$.
- Hint 2. Then complete the squares: $(a-22)^2+(b-22)^2=1009$.
Our solution (not official)
Answer: $(a,b)=(50,37),(50,7),(37,50),(7,50)$
Key idea: the remainder is less than $a+b$, which is less than about $2q$. So $q^2$ is almost $1977$, giving $q=44$, and the rest is a sum of two squares.
We have $a^2+b^2=q(a+b)+r$ with $0\le r<a+b$. Since $a^2+b^2\ge\frac{(a+b)^2}2$, $$\frac{a+b}2\le\frac{a^2+b^2}{a+b}<q+1,\qquad\text{so } a+b\le 2q+1 .$$ Then $1977-q^2=r<a+b\le2q+1$, so $(q+1)^2>1977$ and $q\ge44$. Also $q^2\le1977$ gives $q\le44$. So $q=44$ and $r=1977-1936=41$.
Now $a^2+b^2=44(a+b)+41$, that is, $$(a-22)^2+(b-22)^2=41+2\cdot22^2=1009 .$$ Checking $x=0,1,\dots,31$, the only way to write $1009$ as a sum of two squares is $15^2+28^2$. So $\{|a-22|,|b-22|\}=\{15,28\}$. Since $a,b>0$, $|a-22|=28$ forces $a=50$, and $|a-22|=15$ gives $a=37$ or $7$. This yields $(50,37),(50,7),(37,50),(7,50)$. In each case $r=41<a+b$, so $q$ and $r$ really are the quotient and remainder.
How we checked it: script: brute force over 1 β€ a, b < 500 finds exactly these four pairs
P6 Self-map with f(n+1) beyond f(f(n)) is the identity
- Hint 1. Prove by induction on $k$: every $n\ge k$ has $f(n)\ge k$.
- Hint 2. So $f(n)\ge n$. Use this to show that $f$ is increasing, and then compare $f(n+1)$ with $f(f(n))$ again.
Our solution (not official)
Key idea: a short induction shows $f(n)\ge n$. Then $f$ is increasing, and an increasing $f$ with $f(n+1)>f(f(n))$ must have $f(n)<n+1$.
Step 1: $f(n)\ge n$. We show for each $k\ge1$: if $n\ge k$ then $f(n)\ge k$. For $k=1$ this is clear. Assume it for $k$ and let $n\ge k+1$. Then $n-1\ge k$, so $f(n-1)\ge k$, and applying the assumption to $f(n-1)$ gives $f(f(n-1))\ge k$. So $f(n)>f(f(n-1))\ge k$, that is, $f(n)\ge k+1$. Taking $k=n$ gives $f(n)\ge n$.
Step 2: $f$ is increasing. By Step 1 applied to $f(n)$, $f(f(n))\ge f(n)$. So $f(n+1)>f(f(n))\ge f(n)$.
Step 3: $f(n)\le n$. If $f(n)\ge n+1$, then since $f$ is increasing, $f(f(n))\ge f(n+1)$, contradicting $f(n+1)>f(f(n))$.
So $f(n)=n$ for every $n$.
How we checked it: algebra only: induction