IMO 2001
Read the problems at the official link. The 2001 IMO problems are © the International Mathematical Olympiad, so we don't reproduce them here. Open the 2001 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 “Report a problem” at the bottom of the page.
P1 Circumcenter, altitude foot, and an angle budget
- Hint 1. In triangle $OPC$, the angle at $C$ is $90^\circ-\angle A$. So the goal is the same as $\angle COP<\angle OCP$. Which sides does that compare?
- Hint 2. Show $CP<\tfrac R2$, where $R$ is the circumradius. Write $CP=2R\sin B\cos C$ and use a product-to-sum formula.
Our solution (not official)
Key idea: the claim is a comparison of two angles in triangle $OPC$, so it is a comparison of two sides: we need $CP<OP$. The angle condition makes $CP$ less than half the circumradius, and that is enough.
Write $A,B,C$ for the angles and $R$ for the circumradius. The triangle is acute, so $O$ is inside it, and $\angle OCB=90^\circ-A$ (triangle $OBC$ is isosceles with apex angle $2A$).
Reduce to sides. In triangle $OPC$ the angle at $C$ is $\angle OCP=90^\circ-A$. The claim $A+\angle COP<90^\circ$ says $\angle COP<\angle OCP$. In a triangle the larger angle faces the larger side, so this is the same as $CP<OP$.
$CP$ is short. From the right triangle $APC$, $CP=AC\cos C=2R\sin B\cos C$. A product-to-sum formula gives $$CP=R\big(\sin(B+C)-\sin(C-B)\big)=R\big(\sin A-\sin(C-B)\big).$$ We know $30^\circ\le C-B<90^\circ$, so $\sin(C-B)\ge\tfrac12$. Also $\sin A<1$ because $A<90^\circ$. So $CP<R\big(1-\tfrac12\big)=\tfrac R2$.
Finish. By the triangle inequality in $OPC$, $OP\ge OC-CP=R-CP>\tfrac R2>CP$. So $CP<OP$, and therefore $\angle CAB+\angle COP<90^\circ$.
How we checked it: script: 20,000 random acute triangles with C ≥ B + 30°; the largest value of A + ∠COP found was 89.98°
P2 Square-root fractions bounded below by one
- Hint 1. Look for an exponent $r$ with $\dfrac{a}{\sqrt{a^2+8bc}}\ge\dfrac{a^r}{a^r+b^r+c^r}$. Summing such bounds gives exactly $1$.
- Hint 2. $r=\tfrac43$ works. After squaring, you need $(a^{4/3}+b^{4/3}+c^{4/3})^2-a^{8/3}\ge 8a^{2/3}bc$. Factor the left side and use AM–GM on each factor.
Our solution (not official)
Key idea: compare each term with a fraction of the form $\dfrac{a^r}{a^r+b^r+c^r}$. These three fractions add up to exactly $1$, and $r=\tfrac43$ makes each comparison true.
Let $x=a^{4/3}$, $y=b^{4/3}$, $z=c^{4/3}$. We claim $$\frac{a}{\sqrt{a^2+8bc}}\ \ge\ \frac{x}{x+y+z}.$$ Both sides are positive. Cross-multiplying and squaring, this is $(x+y+z)^2\ge a^{2/3}(a^2+8bc)=x^2+8a^{2/3}bc$, that is, $$(y+z)(2x+y+z)\ \ge\ 8\,a^{2/3}bc .$$ By AM–GM, $y+z\ge2\sqrt{yz}=2\,b^{2/3}c^{2/3}$, and $$2x+y+z=x+x+y+z\ge4\sqrt[4]{x^2yz}=4\,a^{2/3}b^{1/3}c^{1/3}.$$ Multiplying, $(y+z)(2x+y+z)\ge8\,a^{2/3}bc$, which proves the claim.
Writing the same bound for the other two terms and adding, $$\sum\frac{a}{\sqrt{a^2+8bc}}\ \ge\ \frac{x+y+z}{x+y+z}=1 .$$ Equality holds when $a=b=c$.
How we checked it: script: 100,000 random positive triples; the smallest sum found was 1.00002
P3 Girls, boys, shared problems: a popular problem must exist
- Hint 1. Suppose not: every problem was solved by at most two girls or by at most two boys. Make a $21\times21$ table of girl–boy pairs.
- Hint 2. Fix one boy. If $11$ girls share a “few girls” problem with him, how many of his problems must be of that kind? What then happens to the other girls?
Our solution (not official)
Key idea: count girl–boy pairs. Each boy can share a “few girls” problem with at most $10$ girls, and each girl can share a “few boys” problem with at most $10$ boys. That covers at most $420$ of the $441$ pairs.
Suppose, for contradiction, that every problem has at most two solvers of one sex or the other. Call a problem G-light if at most two girls solved it, and B-light if at most two boys solved it. Every problem is G-light or B-light (or both).
Columns. Fix a boy $b$. Suppose at least $11$ girls share a G-light problem with $b$. Each G-light problem reaches at most $2$ girls, so $b$ solved at least $6$ G-light problems. He solved at most $6$ problems in all, so all of his problems are G-light. But every one of the $21$ girls shares some problem with $b$, and his $6$ problems reach at most $12$ girls. That is a contradiction. So each boy shares a G-light problem with at most $10$ girls.
Rows. In the same way, each girl shares a B-light problem with at most $10$ boys.
Count. There are $21\cdot21=441$ girl–boy pairs. Each pair shares a problem, which is G-light or B-light. By the column bound, at most $21\cdot10=210$ pairs share a G-light problem. By the row bound, at most $210$ pairs share a B-light problem. That covers at most $420<441$ pairs, a contradiction.
So the assumption fails: some problem has three or more solvers of each sex.
How we checked it: logic: the counting is exact; 441 pairs against at most 420 covered
P4 Weighted sums over all permutations: two must collide
- Hint 1. Suppose the $n!$ sums are all different mod $n!$. Then they are $0,1,\dots,n!-1$ in some order. What is their total mod $n!$?
- Hint 2. Compute the total directly: each number $1,\dots,n$ sits in each position $(n-1)!$ times.
Our solution (not official)
Key idea: add up all $n!$ sums in two ways. If they were all different mod $n!$, the total would be $\tfrac{n!}2$ mod $n!$. A direct count shows the total is $0$ mod $n!$.
Suppose no two of the $n!$ numbers $S(a)$ are congruent mod $n!$. Then their remainders are exactly $0,1,\dots,n!-1$, so $$\sum_a S(a)\equiv\frac{(n!-1)\,n!}{2}\pmod{n!}.$$ Since $n!$ is even and $n!-1$ is odd, this is $\tfrac{n!}2\cdot(\text{odd})\equiv\tfrac{n!}2\pmod{n!}$, which is not $0$.
Now compute directly. For each position $i$, each value $1,\dots,n$ appears in that position in exactly $(n-1)!$ permutations. So $$\sum_a S(a)=\sum_{i=1}^n k_i\,(n-1)!\,(1+2+\dots+n)=n!\cdot\frac{n+1}{2}\sum_i k_i .$$ Since $n$ is odd, $\tfrac{n+1}2$ is an integer, so the total is a multiple of $n!$. This contradicts the first count. Hence two different permutations $b\ne c$ have $n!\mid S(b)-S(c)$.
How we checked it: script: n = 3 and 5 with random integer weights; a collision mod n! always appears
P5 Two angle bisectors and a pair of equal sums
- Hint 1. Fix $AB=1$ and let $\angle B=x$. Use the law of sines in triangles $ABP$ and $ABQ$ to write every length in terms of $x$.
- Hint 2. Put $t=x/4$. Sum-to-product formulas turn the equation into $\cos(3t+30^\circ)\,(2\cos2t-1)=0$.
Our solution (not official)
Answer: $\angle B=80^\circ$, $\angle C=40^\circ$ (and $\angle A=60^\circ$)
Key idea: turn the length condition into one trigonometric equation in $\angle B$. It factors completely.
Let $\angle B=x$, so $\angle C=120^\circ-x$ and $0<x<120^\circ$. Scale so that $AB=1$.
Lengths. In triangle $ABP$ the angles are $30^\circ$ at $A$, $x$ at $B$, and $150^\circ-x$ at $P$. The law of sines gives $$BP=\frac{\sin30^\circ}{\sin(150^\circ-x)}=\frac{1}{2\sin(30^\circ+x)} .$$ In triangle $ABQ$ the angles are $60^\circ$ at $A$, $\tfrac x2$ at $B$, and $120^\circ-\tfrac x2$ at $Q$, so $$AQ+QB=\frac{\sin\frac x2+\sin60^\circ}{\sin\big(120^\circ-\frac x2\big)} .$$
Simplify. Put $t=\tfrac x4$, so $0<t<30^\circ$. Sum-to-product gives $\sin\tfrac x2+\sin60^\circ=2\sin(t+30^\circ)\cos(t-30^\circ)$, and $\sin(120^\circ-2t)=\sin(60^\circ+2t)=2\sin(30^\circ+t)\cos(30^\circ+t)$. So $$AQ+QB=\frac{\cos(t-30^\circ)}{\cos(t+30^\circ)}=1+\frac{\cos(t-30^\circ)-\cos(t+30^\circ)}{\cos(t+30^\circ)}=1+\frac{\sin t}{\cos(t+30^\circ)} .$$ The condition $AB+BP=AQ+QB$ becomes $$\frac{1}{2\sin(30^\circ+4t)}=\frac{\sin t}{\cos(t+30^\circ)}\iff \cos(t+30^\circ)=2\sin t\,\sin(30^\circ+4t)=\cos(30^\circ+3t)-\cos(30^\circ+5t).$$ Since $\cos(t+30^\circ)+\cos(5t+30^\circ)=2\cos(3t+30^\circ)\cos2t$, this is $$\cos(3t+30^\circ)\,\big(2\cos2t-1\big)=0 .$$
Solve. For $0<t<30^\circ$: $2\cos2t=1$ needs $t=30^\circ$, which is excluded. And $\cos(3t+30^\circ)=0$ with $30^\circ<3t+30^\circ<120^\circ$ needs $3t+30^\circ=90^\circ$, so $t=20^\circ$. Hence $\angle B=80^\circ$ and $\angle C=40^\circ$. Every step was an equivalence, so this triangle does satisfy the condition.
How we checked it: script: scanned ∠B from 0° to 120° in steps of 0.01°; the only root of the length equation is 80°
P6 A quadratic constraint that makes a sum of products composite
- Hint 1. Expand the condition. It says $a^2-ac+c^2=b^2+bd+d^2$. Call this common value $K$.
- Hint 2. Show $(ab+cd)(ad+bc)=(ac+bd)K$. Then compare the sizes of $ab+cd$, $ac+bd$ and $ad+bc$.
Our solution (not official)
Key idea: the condition makes $(ab+cd)(ad+bc)$ equal to $(ac+bd)$ times a positive integer. Since $ab+cd$ is the largest of the three numbers involved, it can’t be prime.
Rewrite the condition. The right side is $(b+d)^2-(a-c)^2$. So $ac+bd=b^2+2bd+d^2-a^2+2ac-c^2$, which rearranges to $$a^2-ac+c^2=b^2+bd+d^2=:K>0 .$$
An identity. Expand: $$(ab+cd)(ad+bc)=ac(b^2+d^2)+bd(a^2+c^2).$$ Using $b^2+d^2=K-bd$ and $a^2+c^2=K+ac$, this is $ac(K-bd)+bd(K+ac)=(ac+bd)K$.
Sizes. Since $a>b>c>d>0$: $$(ab+cd)-(ac+bd)=(a-d)(b-c)>0,\qquad (ac+bd)-(ad+bc)=(a-b)(c-d)>0 .$$ So $ab+cd>ac+bd>ad+bc>0$.
Finish. Suppose $p=ab+cd$ is prime. It divides $(ac+bd)K$. Since $0<ac+bd<p$, the prime $p$ doesn’t divide $ac+bd$, so $p\mid K$. Write $K=pm$ with $m\ge1$. Dividing the identity by $p$ gives $ad+bc=(ac+bd)\,m\ge ac+bd$. That contradicts $ad+bc<ac+bd$. So $ab+cd$ is not prime.
How we checked it: script: all 732 solutions with a < 80 satisfy the identity, and none has ab + cd prime