IMO 1975
Read the problems at the official link. The 1975 IMO problems are Β© the International Mathematical Olympiad, so we don't reproduce them here. Open the 1975 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 Sorted pairing minimizes the sum of squared gaps
- Hint 1. Expand the squares. Only the cross terms $\sum x_iz_i$ depend on the order.
- Hint 2. If $z_i<z_j$ for some $i<j$, what does swapping them do to $\sum x_iz_i$?
Our solution (not official)
Key idea: after expanding, the claim is the rearrangement inequality, and swapping an out-of-order pair never hurts.
Since $z$ is a rearrangement of $y$, $\sum z_i^2=\sum y_i^2$. Expanding both sides, the claim becomes $$\sum_i x_iy_i\ge\sum_i x_iz_i .$$ Suppose $z$ is not sorted in decreasing order, so $z_i<z_j$ for some $i<j$. Swap them. The sum $\sum x_kz_k$ changes by $$x_iz_j+x_jz_i-x_iz_i-x_jz_j=(x_i-x_j)(z_j-z_i)\ge0,$$ because $x_i\ge x_j$. Each swap removes at least one out-of-order pair, so after finitely many swaps we reach the decreasing order, which is $y$. The sum never decreased along the way, so $\sum x_iz_i\le\sum x_iy_i$.
How we checked it: algebra only: exchange argument
P2 Increasing sequence: terms as combinations of two earlier terms
- Hint 1. Sort the later terms by their remainders modulo $a_p$.
- Hint 2. Infinitely many later terms share one remainder. Compare each of them with the first one.
Our solution (not official)
Key idea: pigeonhole on remainders modulo $a_p$; two terms with the same remainder differ by a multiple of $a_p$.
Fix $p$. The terms $a_{p+1},a_{p+2},\dots$ are infinitely many, and there are only $a_p$ possible remainders modulo $a_p$. So some remainder is shared by infinitely many of them. Let $a_q$ (with $q>p$) be the first such term. Every later term $a_m$ in the same class satisfies $a_m>a_q$ and $a_m\equiv a_q\pmod{a_p}$, so $$a_m=x\,a_p+1\cdot a_q\quad\text{with }x=\frac{a_m-a_q}{a_p}\ge1 .$$ There are infinitely many such $a_m$, each with $y=1$ and $q>p$.
How we checked it: algebra only: pigeonhole
P3 Three outer triangles with 15, 30, 45 degree angles
- Hint 1. Use complex numbers. Each of $P,Q,R$ is a fixed complex combination of two vertices. Note that the side ratio $\frac{\sin30^\circ}{\sin105^\circ}$ equals $\frac1{2\cos15^\circ}$.
- Hint 2. The claim says $P-R=\pm i(Q-R)$. This is linear in $A,B,C$, so check the coefficient of each vertex.
Our solution (not official)
Key idea: with complex numbers, the claim “$RP$ is $RQ$ turned through $90^\circ$” is a linear identity in $A,B,C$. Its coefficients vanish because of one exact value of $\cos15^\circ$.
Take $A,B,C$ counterclockwise; for the other orientation, reflect the whole figure. Then the outside of each side is on the right as we walk around the triangle. Let $$k=\frac{1}{2\cos15^\circ}=\frac{\sin30^\circ}{\sin105^\circ}.$$
The three apexes. In triangle $BCP$ the angles are $45^\circ$ at $B$, $30^\circ$ at $C$ and $105^\circ$ at $P$. By the law of sines $BP=k\cdot BC$, and $P$ is reached from $B$ by turning $\overrightarrow{BC}$ by $45^\circ$ clockwise (to the outside). Treating $Q$ (from $A$, toward $C$) and $R$ (base $AB$, base angles $15^\circ$, so $AR=k\cdot AB$) the same way: $$P=B+u(C-B),\quad Q=A+v(C-A),\quad R=A+w(B-A),$$ with $u=k\,e^{-i\pi/4}$, $v=k\,e^{i\pi/4}$, $w=k\,e^{-i\pi/12}$.
The identity. We show $(P-R)+i(Q-R)=0$. The left side is $\alpha A+\beta B+\gamma C$ for some constants.
- Coefficient of $C$: $\gamma=u+iv=k\left(e^{-i\pi/4}+e^{3i\pi/4}\right)=0$.
- Coefficient of $B$: $\beta=1-u-(1+i)w$. Here $(1+i)w=\sqrt2\,k\,e^{i\pi/6}$, so $$u+(1+i)w=k\left(e^{-i\pi/4}+\sqrt2e^{i\pi/6}\right)=k\left(\tfrac{\sqrt2}2+\tfrac{\sqrt6}2\right)=k\cdot2\cos15^\circ=1,$$ since the imaginary parts $-\frac{\sqrt2}2+\frac{\sqrt2}2$ cancel and $\cos15^\circ=\frac{\sqrt6+\sqrt2}4$. So $\beta=0$.
- Coefficient of $A$: if $A=B=C$, then $P=Q=R$ and the expression is $0$, so $\alpha+\beta+\gamma=0$, hence $\alpha=0$.
So $P-R=-i(Q-R)$: the vector $RP$ is $RQ$ turned through $90^\circ$. That is, $RP=RQ$ and $\angle QRP=90^\circ$.
How we checked it: script: random triangles; angles of the constructed triangles checked from the definition, and |(PβR)+i(QβR)| < 1e-15
P4 Repeated digit sums of a huge power
- Hint 1. Bound the number of digits of the power, then the sizes of $A$, $B$ and the final digit sum.
- Hint 2. A digit sum keeps the remainder modulo $9$.
Our solution (not official)
Answer: $7$
Key idea: size bounds make the final answer small, and remainders modulo $9$ pin it down.
Let $N=4444^{4444}$.
Sizes. $N<10000^{4444}=10^{17776}$, so $N$ has at most $17776$ digits and $A\le9\cdot17776=159984$. A number up to $159984$ has at most six digits, the first of them at most $1$, so $B\le1+5\cdot9=46$. A number up to $46$ has digit sum at most $3+9=12$.
Modulo $9$. A number and its digit sum leave the same remainder modulo $9$. Now $4444\equiv16\equiv7\pmod9$, and $7^3=343\equiv1\pmod 9$. Since $4444=3\cdot1481+1$, $$N\equiv7^{4444}\equiv7\pmod9 .$$ So the digit sum of $B$ is a positive number at most $12$ that is $\equiv7\pmod 9$. It is $7$.
How we checked it: script: computed exactly with big integers: A = 72601, B = 16, digit sum of B = 7
P5 Many points on a circle with rational distances
- Hint 1. A chord of the unit circle subtending the angle $2\alpha$ at the center has length $2|\sin\alpha|$.
- Hint 2. Use an angle $\theta$ with $\cos\theta=\frac45$, $\sin\theta=\frac35$, and its multiples. Why are the points distinct?
Our solution (not official)
Answer: Yes, such points exist
Key idea: multiples of an angle with rational sine and cosine have rational sines, and chord lengths are sines of half the central angle.
Let $\theta$ be the angle with $\cos\theta=\frac45$, $\sin\theta=\frac35$. By the addition formulas, $\cos k\theta$ and $\sin k\theta$ are rational for every integer $k$. Take the points $$Z_k=(\cos2k\theta,\ \sin2k\theta),\qquad k=1,\dots,1975 .$$ The chord $Z_jZ_k$ subtends the central angle $2(j-k)\theta$, so its length is $2|\sin(j-k)\theta|$, a rational number.
The points are distinct. It is enough that $\sin m\theta\ne0$ for $m\ge1$. Write $(4+3i)^m=x_m+iy_m$; then $\sin m\theta=y_m/5^m$. From $x_{m+1}=4x_m-3y_m$ and $y_{m+1}=3x_m+4y_m$, modulo $5$ we get $x_{m+1}\equiv-x_m+2y_m$ and $y_{m+1}\equiv-2x_m-y_m$. Starting from $(x_1,y_1)=(4,3)$, which has $y_1\equiv2x_1$, induction shows $$y_m\equiv2x_m,\qquad x_{m+1}\equiv3x_m\pmod5 .$$ Indeed, if $y\equiv2x$ then $x'\equiv3x$ and $y'\equiv-4x\equiv x\equiv2\cdot3x$. So $x_m\equiv4\cdot3^{m-1}\not\equiv0$, hence $y_m\not\equiv0\pmod5$, and in particular $y_m\ne0$.
How we checked it: script: y_m is not divisible by 5 for m < 3000
P6 Homogeneous two-variable polynomial with a cyclic vanishing sum
- Hint 1. Put $s=a+b+c$. Each term has the form $P(s-t,t)$. Study $f(t)=P(1-t,t)$.
- Hint 2. Show that $f(a)+f(b)=f\!\left(a+b-\frac13\right)$, so a shifted $f$ is additive. An additive polynomial is linear.
Our solution (not official)
Answer: $P(x,y)=(x-2y)(x+y)^{n-1}$
Key idea: on the line $x+y=1$ the condition makes $P$ an additive function (after a shift), hence linear. Homogeneity then rebuilds $P$ everywhere.
Note that $(b+c)+a=(c+a)+b=(a+b)+c=s$, where $s=a+b+c$. Let $f(t)=P(1-t,t)$, a polynomial in $t$. Taking $s=1$, condition (ii) says $$f(a)+f(b)+f(c)=0\qquad\text{whenever }a+b+c=1 .$$
- With $a=b=c=\frac13$: $f(\frac13)=0$.
- With $c=\frac13$: $f(a)+f(b)=0$ whenever $a+b=\frac23$.
- So for any $a,b$: $f(1-a-b)=-f\left(a+b-\frac13\right)$, and the condition becomes $f(a)+f(b)=f\left(a+b-\frac13\right)$.
Let $g(u)=f(u+\frac13)$. Then $g(u)+g(v)=g(u+v)$ for all $u,v$. Comparing the coefficients of $g(2u)=2g(u)$ shows that only the degree-$1$ coefficient can be nonzero, so $g(u)=\lambda u$ and $f(t)=\lambda(t-\frac13)$. From $f(0)=P(1,0)=1$ we get $\lambda=-3$, so $$P(1-t,t)=1-3t .$$ For $x+y\ne0$, homogeneity gives $$P(x,y)=(x+y)^nP\!\left(\tfrac{x}{x+y},\tfrac{y}{x+y}\right)=(x+y)^n\left(1-\tfrac{3y}{x+y}\right)=(x-2y)(x+y)^{n-1}.$$ Two polynomials that agree off the line $x+y=0$ agree everywhere.
Check. $(x-2y)(x+y)^{n-1}$ is homogeneous of degree $n$ and equals $1$ at $(1,0)$. With $s=a+b+c$, the term $P(s-a,a)=(s-3a)s^{n-1}$, and the three terms add up to $(3s-3s)s^{n-1}=0$.
How we checked it: script: sympy verifies conditions (i)β(iii) for n = 1 to 6