IMO 1978
Read the problems at the official link. The 1978 IMO problems are Β© the International Mathematical Olympiad, so we don't reproduce them here. Open the 1978 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 Two powers of 1978 with the same last three digits
- Hint 1. The difference is $1978^m\left(1978^{n-m}-1\right)$. The second factor is odd. What does $8\mid$ the difference say about $m$?
- Hint 2. For the factor $125$, find the order of $1978\equiv103\pmod{125}$.
Our solution (not official)
Answer: $m=3$, $n=103$, so $m+n=106$
Key idea: split $1000=8\cdot125$. The power of $2$ fixes $m\ge3$, and the order of $1978$ modulo $125$ fixes $n-m$.
We need $1000\mid1978^n-1978^m=1978^m\left(1978^{n-m}-1\right)$.
The factor $8$. The second factor is odd, so $8\mid1978^m$. As $1978=2\cdot989$ with $989$ odd, this means $m\ge3$.
The factor $125$. $1978^m$ is prime to $5$, so we need $1978^{n-m}\equiv1\pmod{125}$. Now $1978\equiv103\equiv-22\pmod{125}$. The order of $-22$ divides $\varphi(125)=100$. Modulo $5$, $-22\equiv3$, whose order is $4$; so the order modulo $125$ is a multiple of $4$ dividing $100$: one of $4,20,100$. Compute: $$(-22)^2=484\equiv-16,\qquad (-22)^4\equiv256\equiv6,\qquad (-22)^{20}\equiv6^5=7776\equiv26\pmod{125}.$$ Neither is $1$, so the order is $100$, and $100\mid n-m$.
The smallest sum is $m=3$, $n=3+100=103$, giving $m+n=106$.
How we checked it: script: search over n < 400 confirms the minimum m + n = 106
P2 Box from three perpendicular chords: locus of far corner
- Hint 1. Use vectors from the center $O$. Write $U=P+u$, $V=P+v$, $W=P+w$ with $u,v,w$ pairwise perpendicular, so $Q=P+u+v+w$.
- Hint 2. Add the three equations $|P+u|^2=R^2$ (and the other two) and compare with $|Q|^2$. For the converse, the far corner sees each neighbor of $P$ at a right angle.
Our solution (not official)
Answer: The sphere with the same center $O$ and radius $\sqrt{3R^2-2\,OP^2}$
Key idea: adding the three sphere equations gives $|Q|^2$ directly. For the converse, build the box one edge at a time, using the intermediate value theorem twice.
Put the origin at the center $O$, let $R$ be the radius, and write $p=OP<R$.
Every $Q$ lies on the sphere. Write $U=P+u$, $V=P+v$, $W=P+w$ with $u,v,w$ pairwise perpendicular, so $Q=P+u+v+w$. Since $U$ is on the sphere, $p^2+2P\cdot u+|u|^2=R^2$, and similarly for $v,w$. The cross terms $u\cdot v$ etc. vanish, so $$|Q|^2=p^2+2P\cdot(u+v+w)+|u|^2+|v|^2+|w|^2=p^2+3(R^2-p^2)=3R^2-2p^2 .$$
Every point of that sphere occurs. Take $Q$ with $|Q|^2=3R^2-2p^2$; note $|Q|>R$ because $p<R$.
- The first edge. The sphere $\Sigma$ with diameter $PQ$ contains $P$ (inside the big sphere) and $Q$ (outside it). It is connected, so it contains a point $U$ with $|U|=R$, and $U\ne P,Q$. Since $U\in\Sigma$, the angle $PUQ$ is right. Let $u=U-P$ and $d=Q-U$; then $d\perp u$ and $d\ne0$.
- The other two edges. In the plane through $P$ perpendicular to $u$, take the circle with diameter from $P$ to $P+d$. For a point $X=P+v$ on it, let $X^*=P+(d-v)$ be the opposite end of the diameter through $X$. By Thales, $v\perp d-v$, so $$|X|^2+|X^*|^2=2p^2+2P\cdot d+|d|^2 .$$ Expanding $|Q|^2=|U+d|^2=R^2+2P\cdot d+|d|^2$ (using $U=P+u$ and $u\perp d$) and comparing with $|Q|^2=3R^2-2p^2$ shows that the right side equals $2R^2$. So $h(X)=|X|^2-R^2$ satisfies $h(X^*)=-h(X)$ on this circle. Since $h(P)<0<h(P+d)$, the continuous function $h$ has a zero at some $X$ other than $P$ and $P+d$. Then $X^*$ is also on the big sphere.
- The box. Take $V=X$ and $W=X^*$: the vectors $u$, $v$, $w=d-v$ are pairwise perpendicular and nonzero, all three endpoints lie on the sphere, and $P+u+v+w=U+d=Q$.
So the locus is the whole sphere centered at $O$ with radius $\sqrt{3R^2-2\,OP^2}$.
How we checked it: script: 20,000 random triads give |OQ| equal to the formula within 2e-15, and the far corners spread over the whole sphere
P3 Complementary sequences with g(n) = f(f(n)) + 1
- Hint 1. Count the integers below $g(n)$. Show that $g(n)=f(n)+n$.
- Hint 2. The conditions determine the sequences one step at a time. Check that $f(n)=\lfloor n\varphi\rfloor$ with the golden ratio $\varphi$ satisfies them.
Our solution (not official)
Answer: $f(240)=388$
Key idea: counting gives $g(n)=f(n)+n$, so the pair is determined step by step. The golden-ratio Beatty sequences satisfy the conditions, so they are the pair.
Step 1: $g(n)=f(n)+n$. The integers $1,\dots,g(n)-1$ consist of $g(1),\dots,g(n-1)$ and all values of $f$ below $g(n)=f(f(n))+1$, which are $f(1),\dots,f(f(n))$. So $g(n)-1=(n-1)+f(n)$.
Step 2: uniqueness. Decide, for $k=1,2,3,\dots$ in turn, whether $k$ is an $f$-value or a $g$-value. Suppose this is settled for $1,\dots,k-1$, which are $f(1),\dots,f(i)$ and $g(1),\dots,g(j)$. Then $k$ is either $f(i+1)$ or $g(j+1)$, and $g(j+1)=f(j+1)+j+1$ by Step 1.
- If $j+1\le i$, then $f(j+1)$ is known, and $k=g(j+1)$ exactly when $k=f(j+1)+j+1$.
- If $j+1>i$, then $g(j+1)>f(j+1)\ge f(i+1)\ge k$, so $k=f(i+1)$.
Either way the answer is forced, so at most one pair of sequences satisfies the conditions.
Step 3: the golden pair works. Let $\varphi=\frac{1+\sqrt5}2$, so $\varphi^2=\varphi+1$ and $\frac1\varphi+\frac1{\varphi^2}=1$. Put $f(n)=\lfloor n\varphi\rfloor$ and $g(n)=\lfloor n\varphi^2\rfloor=f(n)+n$.
- They split the positive integers. Both are strictly increasing. The number of terms of either sequence that are at most $N$ is $\left\lfloor\frac{N+1}{\varphi}\right\rfloor+\left\lfloor\frac{N+1}{\varphi^2}\right\rfloor=N$, because the two fractions are irrational and add up to the integer $N+1$. Since this count is $N$ for every $N$, each positive integer is hit exactly once.
- The relation. Let $m=f(n)=n\varphi-\varepsilon$ with $0<\varepsilon<1$. Then $$m\varphi=n\varphi^2-\varepsilon\varphi=n\varphi+n-\varepsilon\varphi=m+n-\frac{\varepsilon}{\varphi},$$ using $\varphi-1=\frac1\varphi$. Since $0<\frac\varepsilon\varphi<1$, $f(f(n))=m+n-1=g(n)-1$.
So $f(240)=\lfloor240\varphi\rfloor=\lfloor388.33\ldots\rfloor=388$.
How we checked it: script: building the sequences greedily from the conditions gives f(n) = floor(nΟ) for every n computed, and f(240) = 388
P4 Isosceles triangle: tangent circle chord passes through incenter
- Hint 1. Everything is symmetric in the axis through $A$. Where does the small circle touch the circumcircle?
- Hint 2. Let $2\alpha$ be the apex angle and $R$ the circumradius. Compute both the distance from $A$ to the midpoint of $PQ$ and the distance $AI$.
Our solution (not official)
Key idea: both points lie on the symmetry axis, so it is enough to compute their distances from $A$. Both come out as $2R(1-\sin\alpha)$.
Let $\angle A=2\alpha$, $R$ the circumradius, and $\ell$ the axis of symmetry through $A$. The small circle (center $K$, radius $\rho$) is tangent to both legs, so $K$ is on $\ell$ and $AK=\frac\rho{\sin\alpha}$. Both centers lie on $\ell$, so the circles touch at a point of $\ell$. Being inside, the small circle touches at the far end of the diameter from $A$: $$\frac{\rho}{\sin\alpha}+\rho=2R,\qquad \rho=\frac{2R\sin\alpha}{1+\sin\alpha}.$$
Midpoint of $PQ$. It lies on $\ell$ at distance $AP\cos\alpha$ from $A$, with $AP=AK\cos\alpha=\frac{\rho\cos\alpha}{\sin\alpha}$. So $$AP\cos\alpha=\frac{\rho\cos^2\alpha}{\sin\alpha}=\frac{2R\cos^2\alpha}{1+\sin\alpha}=2R(1-\sin\alpha).$$
Incenter. By the law of sines, $AB=2R\sin C=2R\cos\alpha$. The altitude from $A$ is $AB\cos\alpha=2R\cos^2\alpha$ and half the base is $2R\cos\alpha\sin\alpha$. So the area is $4R^2\sin\alpha\cos^3\alpha$ and the semiperimeter is $2R\cos\alpha(1+\sin\alpha)$, which gives the inradius $$r=\frac{2R\sin\alpha\cos^2\alpha}{1+\sin\alpha}=2R\sin\alpha(1-\sin\alpha).$$ The incenter is on $\ell$ with $AI=\frac{r}{\sin\alpha}=2R(1-\sin\alpha)$.
Both points are on $\ell$, on the same side of $A$, at the same distance. So they coincide.
How we checked it: script: random apex angles; the two distances agree to 1e-15
P5 Distinct integers over squares beat the harmonic sum
- Hint 1. Sort $a_1,\dots,a_n$ increasingly as $b_1<\dots<b_n$. Why is $b_k\ge k$?
- Hint 2. Which pairing of the $a$’s with the weights $\frac1{k^2}$ makes the sum smallest?
Our solution (not official)
Key idea: rearrange so that the smallest numbers get the largest weights; that only lowers the sum, and then each number is at least its position.
Let $b_1<b_2<\dots<b_n$ be $a_1,\dots,a_n$ in increasing order. They are distinct positive integers, so $b_k\ge k$. The weights $\frac1{k^2}$ decrease. By the rearrangement inequality, the sum $\sum c_k/k^2$ over orderings $c$ of these numbers is smallest when the numbers increase. So $$\sum_{k=1}^n\frac{a_k}{k^2}\ge\sum_{k=1}^n\frac{b_k}{k^2}\ge\sum_{k=1}^n\frac{k}{k^2}=\sum_{k=1}^n\frac1k .$$ (The rearrangement step: if $c_i>c_j$ with $i<j$, swapping them changes the sum by $(c_j-c_i)\left(\frac1{i^2}-\frac1{j^2}\right)<0$.)
How we checked it: script: exact check over all choices of distinct values up to 8 for n β€ 6
P6 Six groups of 1 to 1978: a sum inside one
- Hint 1. Call a group bad if no member is the sum of two members of it (or twice one). In a bad group, the difference of two members never lies in the group.
- Hint 2. Take the largest group, form differences with its largest member, and repeat inside the largest group of those differences. The groups run out.
Our solution (not official)
Key idea: repeatedly take differences from a largest element. Differences of differences are still differences in all earlier groups, so they are banned from every group used so far, and with six groups the list runs dry.
Suppose no group has a member that is the sum of two members of the same group (possibly the same member twice). Then for $x>y$ in one group, $x-y$ is not in that group, since $y+(x-y)=x$.
A useful fact. If numbers $t_1<\dots<t_m$ all have the form $t_i=K-s_i$ for one fixed $K$, then $t_j-t_i=s_i-s_j$. So a difference of two $t$’s is also a difference of two $s$’s.
The chain.
- Some group $G_1$ has at least $\lceil1978/6\rceil=330$ members. Subtract each of the others from the largest: this gives $329$ distinct numbers in $[1,1977]$, each a difference of two members of $G_1$, so none is in $G_1$.
- Among these $329$ numbers, at least $\lceil329/5\rceil=66$ lie in one other group $G_2$. Subtract each from the largest of them: $65$ numbers. By the useful fact, each is a difference of two members of $G_2$ and also of two members of $G_1$. So none lies in $G_1$ or $G_2$.
- At least $\lceil65/4\rceil=17$ of them lie in one group $G_3$. Subtracting from the largest gives $16$ numbers that avoid $G_1,G_2,G_3$.
- At least $\lceil16/3\rceil=6$ lie in one group $G_4$, giving $5$ numbers that avoid $G_1,\dots,G_4$.
- At least $\lceil5/2\rceil=3$ lie in one group $G_5$, giving $2$ numbers that avoid $G_1,\dots,G_5$. Both lie in $G_6$.
- Their difference is a difference of two members of each of $G_1,\dots,G_6$, so it lies in no group at all. But it is an integer between $1$ and $1977$.
This contradiction shows that some member’s number is the sum of two numbers from its own country, or twice one of them.
How we checked it: algebra only: the counting chain 1978 β 330 β 66 β 17 β 6 β 3 β 2