Library · Between the Challenge and the Olympiad · Chapter 6

Inequalities and Optimisation

Revised Report an error
On this page
  1. Problems
  2. Solutions
  3. Solution: PRMO 2012, Q2
  4. Solution: IOQM 2020, Q5
  5. Solution: IOQM 2025 Part SEP, Q7
  6. Solution: PRMO 2014, Q4
  7. Solution: PRMO 2017, Q7
  8. Solution: IOQM 2024, Q6
  9. Solution: IOQM 2025 Part SEP, Q28
  10. Solution: IOQM 2026, Q3
  11. Solution: PRMO 2015 Part A, Q9
  12. Solution: PRMO 2017, Q15
  13. Solution: PRMO 2017, Q18
  14. Solution: PRMO 2019, Q13
  15. Solution: IOQM 2021 Part A, Q8
  16. Solution: IOQM 2023, Q1
  17. Solution: IOQM 2024, Q15
  18. Solution: PRMO 2018, Q18
  19. Solution: PRMO 2018, Q23
  20. Solution: PRMO 2019, Q22
  21. Solution: PRMO 2019, Q26
  22. Solution: IOQM 2022, Q15
  23. Solution: IOQM 2024, Q11

Problems

Problem 1

A triangle with perimeter 7 has integer side lengths. What is the maximum possible area of such a triangle?

Problem 2

Find the number of integer solutions to ∣∣x∣−2020∣<5\left| |x| - 2020 \right| < 5.

Problem 3

The area of an integer-sided rectangle is 2020. What is the minimum possible value of its perimeter?

Problem 4

In a triangle with integer side lengths, one side is three times as long as a second side, and the length of the third side is 1717. What is the greatest possible perimeter of the triangle?

Problem 5

Find the number of positive integers nn, such that n+n+1<11\sqrt{n} + \sqrt{n+1} < 11.

Problem 6

Find the number of triples of real numbers (a,b,c)(a, b, c) such that a20+b20+c20=a24+b24+c24=1a^{20} + b^{20} + c^{20} = a^{24} + b^{24} + c^{24} = 1.

Problem 7

The sum of two real numbers is a positive integer nn and the sum of their squares is n+1012n + 1012. Find the maximum possible value of nn.

Problem 8

Find the largest integer nn such that a square of side length nn is contained in a circular disc of area 1000.

the square fits when its diagonal fits
the square fits when its diagonal fits

Problem 9

If x1,x2,…,x49x_1, x_2, \ldots, x_{49} are non-zero integers such that ∑i=149xi=0\sum_{i=1}^{49} x_i = 0, then what is the minimum possible value of ∑i=149xi2\sum_{i=1}^{49} x_i^2?

Problem 10

A 2×32 \times 3 rectangle and a 3×43 \times 4 rectangle are contained within a square without overlapping at any interior point, and the sides of the square are parallel to the sides of the two given rectangles. What is the smallest possible area of the square?

Problem 11

Integers 1,2,3,…,n1, 2, 3, \ldots, n, where n>2n > 2, are written on a board. Two numbers m,km, k such that 1<m<n1 < m < n, 1<k<n1 < k < n are removed and the average of the remaining numbers is found to be 17. What is the maximum sum of the two removed numbers?

Problem 12

If the real numbers x,y,zx, y, z are such that x2+4y2+16z2=48x^2 + 4y^2 + 16z^2 = 48 and xy+4yz+2zx=24xy + 4yz + 2zx = 24, what is the value of x2+y2+z2x^2 + y^2 + z^2?

Problem 13

Each of the numbers x1,x2,…,x101x_1, x_2, \ldots, x_{101} is ±1\pm 1. What is the smallest positive value of ∑1≤i<j≤101xixj\displaystyle\sum_{1 \leq i < j \leq 101} x_i x_j?

Problem 14

For any real number tt, let ⌊t⌋\lfloor t \rfloor denote the largest integer ≤t\leq t. Suppose that NN is the greatest integer such that ⌊⌊⌊N⌋⌋⌋=4\left\lfloor \sqrt{\left\lfloor \sqrt{\left\lfloor \sqrt{N} \right\rfloor} \right\rfloor} \right\rfloor = 4 Find the sum of digits of NN.

Problem 15

Let nn be a positive integer such that 1≤n≤10001 \leq n \leq 1000. Let MnM_n be the number of integers in the set Xn={4n+1,4n+2,…,4n+1000}X_n = \{\sqrt{4n+1}, \sqrt{4n+2}, \ldots, \sqrt{4n+1000}\}. Let a=max⁡{Mn:1≤n≤1000}, and b=min⁡{Mn:1≤n≤1000}.a = \max\{M_n : 1 \leq n \leq 1000\}, \text{ and } b = \min\{M_n : 1 \leq n \leq 1000\}. Find a−ba - b.

Problem 16

Let XX be the set consisting of twenty positive integers n,n+2,…,n+38n, n+2, \ldots, n+38. The smallest value of nn for which any three numbers a,b,c∈Xa, b, c \in X, not necessarily distinct, form the sides of an acute-angled triangle is:

Problem 17

Four sides and a diagonal of a quadrilateral are of lengths 10,20,28,50,7510, 20, 28, 50, 75, not necessarily in that order. Which amongst them is the only possible length of the diagonal?

four sides and one diagonal
four sides and one diagonal

Problem 18

If a,b,c≥4a, b, c \geq 4 are integers, not all equal, and 4abc=(a+3)(b+3)(c+3)4abc = (a+3)(b+3)(c+3), then what is the value of a+b+ca + b + c?

Problem 19

What is the largest positive integer nn such that a2b29+c31+b2c29+a31+c2a29+b31≥n(a+b+c)\frac{a^2}{\frac{b}{29} + \frac{c}{31}} + \frac{b^2}{\frac{c}{29} + \frac{a}{31}} + \frac{c^2}{\frac{a}{29} + \frac{b}{31}} \geq n(a+b+c) holds for all positive real numbers a,b,ca, b, c.

Problem 20

What is the greatest integer not exceeding the sum ∑n=115991n\sum_{n=1}^{1599} \frac{1}{\sqrt{n}}?

Problem 21

Positive integers x,y,zx, y, z satisfy xy+z=160xy + z = 160. Compute the smallest possible value of x+yzx + yz.

Problem 22

Let x,yx, y be real numbers such that xy=1xy = 1. Let TT and tt be the largest and the smallest values of the expression (x+y)2−(x−y)−2(x+y)2+(x−y)−2.\frac{(x+y)^2 - (x-y) - 2}{(x+y)^2 + (x-y) - 2}. If T+tT + t can be expressed in the form mn\dfrac{m}{n} where m,nm, n are nonzero integers with GCD(m,n)=1\mathrm{GCD}(m,n) = 1, find the value of m+nm + n.

Problem 23

The positive real numbers a,b,ca, b, c satisfy: a2b+1+2b3c+1+3ca+1=1\frac{a}{2b+1} + \frac{2b}{3c+1} + \frac{3c}{a+1} = 1 1a+1+12b+1+13c+1=2\frac{1}{a+1} + \frac{1}{2b+1} + \frac{1}{3c+1} = 2 What is the value of 1a+1b+1c\dfrac{1}{a} + \dfrac{1}{b} + \dfrac{1}{c}?

Problem 24

MTAIMTAI is a parallelogram of area 4041\dfrac{40}{41} square units such that MI=1/MTMI = 1/MT. If dd is the least possible length of the diagonal MAMA, and d2=abd^2 = \dfrac{a}{b}, where a,ba, b are positive integers with gcd⁡(a,b)=1\gcd(a, b) = 1, find ∣a−b∣|a - b|.

image

Solutions

Solution: PRMO 2012, Q2

Key idea

Only two integer triangles have perimeter 77. Split each isosceles triangle into two right triangles to compare its base and height.

image

The sides are positive integers summing to 77, and there are exactly four ways to write 77 as a sum of three of them, up to order: (1,1,5),(1,2,4),(1,3,3),(2,2,3).(1,1,5), \qquad (1,2,4), \qquad (1,3,3), \qquad (2,2,3). The first two fail the triangle inequality, since 1+1<51 + 1 < 5 and 1+2<41 + 2 < 4. The other two are genuine triangles.

For (1,3,3)(1,3,3), take the side of length 11 as the base. The altitude bisects it, so Pythagoras gives the height hh: h2=32−(12)2=354,area=12⋅1⋅352=354.h^2 = 3^2 - \left(\tfrac12\right)^2 = \tfrac{35}{4}, \qquad \text{area} = \tfrac12 \cdot 1 \cdot \tfrac{\sqrt{35}}2 = \tfrac{\sqrt{35}}4. For (2,2,3)(2,2,3), take the side of length 33 as the base. Its altitude gives h2=22−(32)2=74,area=12⋅3⋅72=374.h^2 = 2^2 - \left(\tfrac32\right)^2 = \tfrac74, \qquad \text{area} = \tfrac12 \cdot 3 \cdot \tfrac{\sqrt7}2 = \tfrac{3\sqrt7}4. No decimals are needed to compare them: 374=634\tfrac{3\sqrt7}{4} = \tfrac{\sqrt{63}}{4}, and 63>3563 > 35. So the maximum area is 374\tfrac{3\sqrt7}{4}.

This is one of several answers on this paper that are not whole numbers. The Mumbai regional papers of this period did not use the two-digit answer format that the national PRMO adopted later, and the questions were written accordingly.

Answer 37/43\sqrt7/4

Solution: IOQM 2020, Q5

Key idea

Peel the two absolute values one at a time, working from the outside in, and count what the inner quantity is allowed to be before worrying about xx itself.

Nested absolute values look forbidding, but they unwrap one layer at a time. The outer condition ∣∣x∣−2020∣<5\bigl| |x| - 2020 \bigr| < 5 says that the quantity ∣x∣−2020|x| - 2020 lies strictly between −5-5 and 55, which is to say that ∣x∣|x| itself lies strictly between 20152015 and 20252025.

Since ∣x∣|x| is a whole number whenever xx is, the possibilities for ∣x∣|x| are the integers from 20162016 up to 20242024 inclusive, and there are 2024−2016+1=92024 - 2016 + 1 = 9 of them. It is only now, at the last step, that we return to xx. Each admissible value vv of ∣x∣|x| arises from exactly two integers, x=vx = v and x=−vx = -v, and these are genuinely different because vv is not zero.

The count is therefore 9×2=189 \times 2 = 18.

Answer 18

Solution: IOQM 2025 Part SEP, Q7

Key idea

With the area fixed, the perimeter shrinks as the rectangle gets closer to square, so among the integer factorisations of 2020 the best is the most balanced one.

An integer-sided rectangle of area 2020 has sides dd and 20/d20/d for a divisor dd of 2020, and its perimeter is 2(d+20/d)2(d + 20/d). The three shapes are 1×20 (perimeter 42),2×10 (24),4×5 (18).1 \times 20 \ (\text{perimeter } 42), \qquad 2 \times 10 \ (24), \qquad 4 \times 5 \ (18).

So the minimum is 1818. The pattern is worth naming: for a fixed product the sum d+20/dd + 20/d is smallest when the two factors are closest together, which is the discrete shadow of the fact that among all rectangles of a given area the square has the least perimeter.

Answer 18

Solution: PRMO 2014, Q4

Key idea

For sides aa, 3a3a and 1717, the triangle inequalities give a lower and an upper bound on aa. The upper bound decides the greatest perimeter.

Let the two proportional sides be aa and 3a3a. All three triangle inequalities must hold. The inequality 3a+17>a3a+17>a is automatic for positive aa. The other two give a+3a>17⟹a≥5,a+3a>17 \quad\Longrightarrow\quad a\ge5, a+17>3a⟹2a<17⟹a≤8.a+17>3a \quad\Longrightarrow\quad 2a<17 \quad\Longrightarrow\quad a\le8. Thus the allowed integer values are 5,6,7,85,6,7,8.

The perimeter 4a+174a + 17 increases with aa, so take a=8a = 8: the sides are 88, 2424 and 1717, and indeed 8+17=25>248 + 17 = 25 > 24. The perimeter is 8+24+17=49.8 + 24 + 17 = 49.

Answer 49

Solution: PRMO 2017, Q7

Key idea

Isolating one square root and squaring twice turns the inequality into the linear condition 484n<14400484n < 14400.

The left side increases with nn, so the answer is a count of an initial run. Rearranged, the condition n+n+1<11\sqrt n + \sqrt{n+1} < 11 says n+1<11−n\sqrt{n+1} < 11 - \sqrt n, which needs n<11\sqrt n < 11 and then squares to n+1<121−22n+n,22n<120,n<6011.n + 1 < 121 - 22\sqrt n + n, \qquad 22\sqrt n < 120, \qquad \sqrt n < \frac{60}{11}. Squaring once more, n<3600121.n < \frac{3600}{121}. That bound lies between 2929 and 3030, since 29⋅121=3509<3600<3630=30⋅12129 \cdot 121 = 3509 < 3600 < 3630 = 30 \cdot 121.

So the condition holds exactly for n=1,2,…,29n = 1, 2, \ldots, 29, which is 2929 positive integers. The two boundary cases confirm it without decimals. Squaring, 29+30<11\sqrt{29} + \sqrt{30} < 11 says 59+2870<12159 + 2\sqrt{870} < 121, that is 870<31\sqrt{870} < 31, and 870<961870 < 961; while 30+31>11\sqrt{30} + \sqrt{31} > 11 says 61+2930>12161 + 2\sqrt{930} > 121, that is 930>30\sqrt{930} > 30, and 930>900930 > 900.

Answer 29

Solution: IOQM 2024, Q6

Key idea

Both sums equal 11, so subtracting them gives ∑a20(a4−1)=0\sum a^{20}(a^4 - 1) = 0; the first equation forces every ∣a∣≤1|a| \le 1, which makes every term of that sum non-positive, so each term vanishes.

From a20+b20+c20=1a^{20} + b^{20} + c^{20} = 1 and every twentieth power being non-negative, each of a20,b20,c20a^{20}, b^{20}, c^{20} is at most 11, so ∣a∣≤1,∣b∣≤1,∣c∣≤1.|a| \le 1, \qquad |b| \le 1, \qquad |c| \le 1.

Now subtract the first equation from the second: a20(a4−1)+b20(b4−1)+c20(c4−1)=0.a^{20}(a^4 - 1) + b^{20}(b^4-1) + c^{20}(c^4-1) = 0. Each bracket is at most 00, because ∣a∣≤1|a| \le 1 gives a4≤1a^4 \le 1, and each factor a20a^{20} is at least 00. A sum of three non-positive numbers is zero only if all three are zero, so for each variable either a20=0a^{20} = 0 or a4=1a^4 = 1, that is a∈{0,1,−1},a \in \{0, 1, -1\}, and likewise for bb and cc.

Finally a20+b20+c20=1a^{20} + b^{20} + c^{20} = 1 says exactly one of the three is non-zero, and that one may be 11 or −1-1. Choosing which variable is non-zero gives 33 ways and choosing its sign gives 22, so there are 3⋅2=63 \cdot 2 = 6 triples.

Answer 6

Solution: IOQM 2025 Part SEP, Q7

Key idea

For real numbers, (x+y)2≤2(x2+y2)(x+y)^2 \le 2(x^2+y^2), which turns the two conditions into a quadratic inequality in nn alone.

Let the numbers be xx and yy with x+y=nx + y = n and x2+y2=n+1012x^2 + y^2 = n + 1012. Since (x−y)2≥0(x-y)^2 \ge 0, (x+y)2≤2(x2+y2),that isn2≤2n+2024.(x+y)^2 \le 2(x^2+y^2), \qquad \text{that is} \qquad n^2 \le 2n + 2024. So n2−2n−2024≤0n^2 - 2n - 2024 \le 0. The roots of n2−2n−2024n^2 - 2n - 2024 are n=2±4+80962=1±45,n = \frac{2 \pm \sqrt{4 + 8096}}{2} = 1 \pm 45, using 8100=9028100 = 90^2, so the inequality holds exactly for −44≤n≤46-44 \le n \le 46.

The largest possible nn is therefore 4646, and it is attained: equality in (x−y)2≥0(x-y)^2 \ge 0 means x=y=23x = y = 23, and indeed 232+232=1058=46+101223^2 + 23^2 = 1058 = 46 + 1012.

Answer 46

Solution: IOQM 2025 Part SEP, Q28

Key idea

A square fits inside a disc exactly when its diagonal is at most the diameter, so n2≤2Rn\sqrt2 \le 2R with πR2=1000\pi R^2 = 1000.

The side, diagonal and radius determine whether the square fits.
The side, diagonal and radius determine whether the square fits.

A square of side nn fits inside a circular disc of radius RR precisely when its circumscribed circle is no larger, that is when its diagonal n2n\sqrt2 is at most the diameter 2R2R. So n≤R2=2R2=2000π,n \le R\sqrt2 = \sqrt{2R^2} = \sqrt{\frac{2000}{\pi}}, using πR2=1000\pi R^2 = 1000.

No decimal is needed. From 3<π<2273 < \pi < \tfrac{22}{7}, 625<200022/7=1400022<2000π<20003<676,625 < \frac{2000}{22/7} = \frac{14000}{22} < \frac{2000}{\pi} < \frac{2000}{3} < 676, and 625=252625 = 25^2 while 676=262676 = 26^2. So the largest integer is n=25.n = 25.

Answer 25

Solution: IOQM 2026, Q3

Key idea

Forty-nine odd numbers cannot add up to zero, so the xix_i cannot all be ±1\pm 1, and one of them has to be at least 22 in size.

Each xix_i is a non-zero integer, so xi2≥1x_i^2 \ge 1, and the sum of squares is at least 4949. That bound would need every xix_i to be 11 or −1-1. But then all forty-nine are odd, and a sum of an odd number of odd numbers is odd, so it cannot be 00.

Hence some xix_i has ∣xi∣≥2|x_i| \ge 2 and contributes at least 44, while the other forty-eight contribute at least 11 each: ∑i=149xi2≥4+48=52.\sum_{i=1}^{49} x_i^2 \ge 4 + 48 = 52.

The bound is reached. Take one 22, twenty-three 11s and twenty-five −1-1s. That is 4949 numbers, their sum is 2+23−25=02 + 23 - 25 = 0, and their squares add up to 4+48=524 + 48 = 52.

Answer 52

Solution: PRMO 2015 Part A, Q9

Key idea

Once the 3×43 \times 4 rectangle is placed, the two strips it leaves behind are each too thin for the 2×32 \times 3 unless the square has side at least 55, and side 55 can be achieved.

A side- 5 square holds both rectangles, with the coordinate bounds used below.
A side-55 square holds both rectangles, with the coordinate bounds used below.

A construction first, so that we know what we are aiming to prove is best possible. In a square of side 55, put the 3×43 \times 4 rectangle in a bottom corner with its side of length 44 horizontal, so it occupies the region 0≤x≤40 \le x \le 4, 0≤y≤30 \le y \le 3. The band above it, 0≤x≤50 \le x \le 5 and 3≤y≤53 \le y \le 5, measures 55 by 22 and comfortably holds the 2×32 \times 3 rectangle laid with its side of length 33 horizontal. Both fit, so a square of side 55 and area 2525 suffices.

Now suppose a square of side ss holds both, and let us see why s<5s < 5 is impossible. Certainly s≥4s \ge 4, since the 3×43 \times 4 rectangle has a side of length 44 and the sides are parallel to the square’s. Assume 4≤s<54 \le s < 5, and take the 3×43 \times 4 rectangle to have its side of length 44 horizontal, which we may do by rotating the whole picture if necessary.

Consider the horizontal direction. The big rectangle occupies an interval of length 44 out of the available s<5s < 5, so the space strictly to its left and strictly to its right totals s−4<1s - 4 < 1. The 2×32 \times 3 rectangle has both sides at least 22, so it cannot fit into either of those slivers, and therefore its horizontal extent must overlap that of the 3×43 \times 4 rectangle.

Two rectangles with parallel sides that overlap horizontally and do not share interior points must be separated vertically. So the 2×32 \times 3 rectangle lies entirely above or entirely below the 3×43 \times 4 rectangle, and the vertical room left over is s−3<2s - 3 < 2. That is less than either side of the 2×32 \times 3 rectangle, so no placement exists. Hence s≥5s \ge 5 and the least possible area is 2525.

Answer 25

Solution: PRMO 2017, Q15

Key idea

The removed sum is 12n(n+1)−17(n−2)\tfrac12 n(n+1) - 17(n-2), and the requirement that it lie between the smallest and largest possible pair sums traps nn in {32,33,34}\{32, 33, 34\}.

If the two removed numbers are mm and kk, the n−2n - 2 that remain have total 17(n−2)17(n-2), so m+k=n(n+1)2−17(n−2)=n2−33n+682.m + k = \frac{n(n+1)}{2} - 17(n-2) = \frac{n^2 - 33n + 68}{2}. Call this S(n)S(n); it is always an integer, since nn and n−33n - 33 have opposite parities.

Two constraints pin down nn. Both removed numbers lie strictly between 11 and nn, so 2+3=5≤m+k≤(n−1)+(n−2)=2n−3.2 + 3 = 5 \le m + k \le (n-1) + (n-2) = 2n - 3.

The upper bound gives n2−33n+68≤4n−6n^2 - 33n + 68 \le 4n - 6, that is n2−37n+74≤0n^2 - 37n + 74 \le 0. Test integers rather than take a square root: at n=34n = 34 the value is −28-28 and at n=35n = 35 it is 44, so n≤34n \le 34. The lower bound gives n2−33n+58≥0n^2 - 33n + 58 \ge 0, where at n=31n = 31 the value is −4-4 and at n=32n = 32 it is 2626, so n≥32n \ge 32 on the branch that matters (the other branch, n≤1n \le 1, is far too small). Hence n∈{32,33,34}n \in \{32, 33, 34\}, and S(32)=18,S(33)=34,S(34)=51.S(32) = 18, \qquad S(33) = 34, \qquad S(34) = 51.

The largest is 5151, and it is attainable: with n=34n = 34 remove 1818 and 3333, both strictly between 11 and 3434, and the remaining 3232 numbers average 595−5132=54432=17\tfrac{595 - 51}{32} = \tfrac{544}{32} = 17.

Answer 51

Solution: PRMO 2017, Q18

Key idea

Substituting u=xu = x, v=2yv = 2y, w=4zw = 4z turns the two conditions into u2+v2+w2=uv+vw+wuu^2+v^2+w^2 = uv+vw+wu, which forces u=v=wu = v = w.

The coefficients 1,4,161,4,16 are squares: scaling the variables by 1,2,41,2,4 makes the first condition a sum of three ordinary squares. The mixed terms also acquire one common coefficient. With that in mind, set u=xu = x, v=2yv = 2y and w=4zw = 4z. The first condition becomes u2+v2+w2=48.u^2 + v^2 + w^2 = 48. For the second, note that xy=uv2xy = \tfrac{uv}{2}, 4yz=vw24yz = \tfrac{vw}{2} and 2zx=wu22zx = \tfrac{wu}{2}, so it becomes uv+vw+wu2=24,uv+vw+wu=48.\frac{uv + vw + wu}{2} = 24, \qquad uv + vw + wu = 48.

The two right-hand sides are equal, and that is the whole problem: u2+v2+w2−uv−vw−wu=0,u^2 + v^2 + w^2 - uv - vw - wu = 0, which is the same as 12[(u−v)2+(v−w)2+(w−u)2]=0.\tfrac12\left[(u-v)^2 + (v-w)^2 + (w-u)^2\right] = 0. A sum of squares vanishes only when each term does, so u=v=wu = v = w. Then 3u2=483u^2 = 48 gives u=±4u = \pm4, and x=u=±4,y=v2=±2,z=w4=±1,x = u = \pm 4, \qquad y = \frac{v}{2} = \pm 2, \qquad z = \frac{w}{4} = \pm 1, with the same sign throughout. Either way x2+y2+z2=16+4+1=21.x^2 + y^2 + z^2 = 16 + 4 + 1 = 21.

The coefficients 1,4,161, 4, 16 and the pairing xy,4yz,2zxxy, 4yz, 2zx were designed to make that substitution work; spotting u,v,wu, v, w is the entire difficulty.

Answer 21

Solution: PRMO 2019, Q13

Key idea

The sum of pairwise products is 12((∑xi)2−101)\tfrac12\bigl((\sum x_i)^2 - 101\bigr), and ∑xi\sum x_i is forced to be odd, so the smallest positive value comes from the smallest odd square above 101101.

Square the total. Since each xi2=1x_i^2 = 1, (∑i=1101xi) ⁣2=∑i=1101xi2+2∑i<jxixj=101+2S,\left(\sum_{i=1}^{101} x_i\right)^{\!2} = \sum_{i=1}^{101} x_i^2 + 2\sum_{i<j} x_i x_j = 101 + 2S, so, writing s=∑xis = \sum x_i, we have S=(s2−101)/2S = (s^2 - 101)/2.

Now ss is a sum of 101101 terms each ±1\pm1, so s≡101≡1(mod2)s \equiv 101 \equiv 1 \pmod 2: it is odd. For SS to be positive we need s2>101s^2 > 101, and the smallest odd square exceeding 101101 is 112=12111^2 = 121. That gives S=121−1012=10.S = \frac{121 - 101}{2} = 10.

It remains to check that s=11s = 11 is achievable, and it is: take pp of the values to be +1+1 and 101−p101 - p to be −1-1, so s=2p−101s = 2p - 101, and p=56p = 56 gives s=11s = 11. So the smallest positive value of the sum is 1010.

Answer 10

Solution: IOQM 2021 Part A, Q8

Key idea

Unwrap the three floors from the outside in, at each stage asking for the largest input that still gives the required output.

We want the largest NN with ⌊⌊⌊N⌋⌋⌋=4\bigl\lfloor \sqrt{\lfloor \sqrt{\lfloor \sqrt N \rfloor} \rfloor} \bigr\rfloor = 4, and the way in is to name the intermediate quantities and work outwards to inwards.

Write k=⌊N⌋k = \lfloor \sqrt N \rfloor and m=⌊k⌋m = \lfloor \sqrt k \rfloor, so the condition is ⌊m⌋=4\lfloor \sqrt m \rfloor = 4. That holds exactly when 16≤m≤2416 \leq m \leq 24, since 42=164^2 = 16 and 52=255^2 = 25. We want NN as large as possible, so take mm as large as possible, m=24m = 24.

Next, ⌊k⌋=24\lfloor \sqrt k \rfloor = 24 holds exactly when 576≤k≤624576 \leq k \leq 624, so the largest useful kk is 624624. Finally ⌊N⌋=624\lfloor \sqrt N \rfloor = 624 holds exactly when 6242≤N≤6252−1624^2 \leq N \leq 625^2 - 1, so the largest NN is N=6252−1=390625−1=390624.N = 625^2 - 1 = 390625 - 1 = 390624.

It is worth checking the chain forwards, since three nested floors are easy to slip on: ⌊390624⌋=624\lfloor \sqrt{390624} \rfloor = 624, then ⌊624⌋=24\lfloor \sqrt{624} \rfloor = 24, then ⌊24⌋=4\lfloor \sqrt{24} \rfloor = 4. The sum of the digits of NN is 3+9+0+6+2+4=243 + 9 + 0 + 6 + 2 + 4 = 24.

Answer 24

Solution: IOQM 2023, Q1

Key idea

The set XnX_n contains the square roots of a block of 10001000 consecutive integers, so MnM_n counts the integers in the interval from 4n+1\sqrt{4n+1} to 4n+1000\sqrt{4n+1000}, and the length of that interval shrinks steadily as nn grows.

An entry 4n+k\sqrt{4n+k} is an integer exactly when 4n+k4n+k is a perfect square, so MnM_n counts the integers jj with 4n+1≤j≤4n+1000.\sqrt{4n+1} \le j \le \sqrt{4n+1000}. The number of integers in an interval is controlled by its length, and here the length is ℓ(n)=4n+1000−4n+1=9994n+1000+4n+1,\ell(n) = \sqrt{4n+1000} - \sqrt{4n+1} = \frac{999}{\sqrt{4n+1000} + \sqrt{4n+1}}, where the second form comes from multiplying and dividing by the conjugate. Written that way the behaviour is plain: ℓ(n)\ell(n) decreases as nn increases, so the longest interval is at n=1n = 1 and the shortest at n=1000n = 1000.

For a small picture, the interval from 0.40.4 to 4.84.8 has length 4.44.4 and contains the four integers 1,2,3,41,2,3,4:

Interval length controls the number of integer points, but the endpoints matter too.
Interval length controls the number of integer points, but the endpoints matter too.

In general an interval of length ℓ\ell contains at least ⌊ℓ⌋\lfloor\ell\rfloor integers and at most ⌊ℓ⌋+1\lfloor\ell\rfloor+1. For the upper bound, qq integer points span at least q−1q-1 units, so q−1≤ℓq-1\le\ell. For the lower bound, the first integer at or after the left endpoint is less than one unit from it. That integer and the next ⌊ℓ⌋−1\lfloor\ell\rfloor-1 integers all fit in the interval. If ℓ<1\ell<1, the asserted lower bound is just zero.

For the maximum, take n=1n = 1. The interval runs from 5\sqrt5 to 1004\sqrt{1004}, and the integers in it are 33 through 3131, so M1=29M_1 = 29. No nn does better. For n≥2n \ge 2, ℓ(n)≤ℓ(2)=1008−9<32−3=29,\ell(n) \le \ell(2) = \sqrt{1008} - \sqrt9 < 32 - 3 = 29, using 1008<10241008 < 1024, and an interval shorter than 2929 holds at most 2929 integers. So a=29a = 29.

For the minimum, take n=1000n = 1000. The interval runs from 4001\sqrt{4001} to 5000\sqrt{5000}, and the integers it contains are exactly 6464 through 7070, since 632=3969<4001≤4096=642,63^2 = 3969 < 4001 \le 4096 = 64^2, 702=4900≤5000<5041=712.70^2 = 4900 \le 5000 < 5041 = 71^2. So M1000=7M_{1000} = 7. No nn can do worse, because ℓ(n)≥ℓ(1000)\ell(n) \ge \ell(1000) for every n≤1000n \le 1000 and ℓ(1000)≥7\ell(1000) \ge 7. That last claim says 5000≥7+4001\sqrt{5000} \ge 7 + \sqrt{4001}, which on squaring becomes 950≥144001950 \ge 14\sqrt{4001}, and squaring again gives 902500≥784196902500 \ge 784196. An interval of length at least 77 always contains at least 77 integers, so b=7b = 7.

Hence a−b=29−7=22a - b = 29 - 7 = 22.

Answer 22

Solution: IOQM 2024, Q15

Key idea

The hardest triple is the one with two shortest sides and the longest side, so everything turns on the single inequality n2+n2>(n+38)2n^2 + n^2 > (n+38)^2.

The largest angle is opposite the largest side w .
The largest angle is opposite the largest side ww.

The requirement is that every choice of three elements of XX, repeats allowed, forms an acute triangle, so one bad triple is enough to disqualify a value of nn. For sides u≤v≤wu\le v\le w, the angle θ\theta opposite ww is the largest. By the cosine rule, cos⁡θ=u2+v2−w22uv.\cos\theta=\frac{u^2+v^2-w^2}{2uv}. All three angles are acute exactly when this largest one is acute, or cos⁡θ>0\cos\theta>0. Thus the triangle is acute exactly when u2+v2>w2u^2+v^2>w^2; and the condition also has to include u+v>wu + v > w, which follows from it here. Among the numbers of XX, the inequality is hardest to satisfy when uu and vv are as small as possible and ww as large as possible, and since a,b,ca, b, c need not be distinct that worst case is u=v=n,w=n+38.u = v = n, \qquad w = n + 38. If that case is acute then so is every other, because increasing uu or vv only helps and decreasing ww only helps.

So the condition is 2n2>(n+38)2,that isn2−76n−1444>0.2n^2 > (n+38)^2, \qquad \text{that is} \qquad n^2 - 76n - 1444 > 0. The positive root of n2−76n−1444=0n^2 - 76n - 1444 = 0 is n=76+5776+57762=38+382,n = \frac{76 + \sqrt{5776 + 5776}}{2} = 38 + 38\sqrt2, and this lies between 9191 and 9292 without any decimal work: 382<5438\sqrt2 < 54 because 2⋅382=2888<2916=5422 \cdot 38^2 = 2888 < 2916 = 54^2, and 382>5338\sqrt2 > 53 because 2888>2809=5322888 > 2809 = 53^2. So the inequality holds exactly for n≥92n \ge 92. (It is worth pausing on the neat form 38(1+2)38(1 + \sqrt2), which is what the number 3838 was chosen to produce.)

Hence the smallest value is n=92n = 92.

Answer 92

Solution: IOQM 2025 Part SEP, Q7

Key idea

The diagonal splits the quadrilateral into two triangles, so it must be smaller than the sum and larger than the difference of each of the two pairs it is paired with, and only one of the five lengths can manage that.

The same diagonal d must satisfy the triangle inequalities with p,q and with r,s .
The same diagonal dd must satisfy the triangle inequalities with p,qp,q and with r,sr,s.

If dd is the diagonal, the other four lengths split into two pairs, one pair forming a triangle with dd on each side of it. So we need to split {10,20,28,50,75}∖{d}\{10, 20, 28, 50, 75\} \setminus \{d\} into pairs {p,q}\{p,q\} and {r,s}\{r,s\} with ∣p−q∣<d<p+qand∣r−s∣<d<r+s.|p - q| < d < p+q \qquad \text{and} \qquad |r-s| < d < r+s.

There are only three ways to split four lengths into two pairs, so each candidate is a short check.

  • d=75d = 75: from {10,20,28,50}\{10,20,28,50\} every splitting leaves a pair of sum 3030, 3838 or 4848, all below 7575. Fails.

  • d=50d = 50: from {10,20,28,75}\{10,20,28,75\} the three splittings leave a pair of sum 3030, 3838 or 4848 again, all below 5050. Fails.

  • d=20d = 20: from {10,28,50,75}\{10,28,50,75\} each splitting contains a pair differing by at least 2222, namely one of {50,75}\{50,75\}, {10,50}\{10,50\} or {10,75}\{10,75\}. Fails.

  • d=10d = 10: from {20,28,50,75}\{20,28,50,75\} every splitting contains a pair differing by at least 2222. Fails.

  • d=28d = 28: pair the rest as {10,20}\{10, 20\} and {50,75}\{50, 75\}. Then 10<28<3010 < 28 < 30 and 25<28<12525 < 28 < 125, both satisfied.

So the diagonal can only be 2828.

Answer 28

Solution: PRMO 2018, Q18

Key idea

Dividing by abcabc turns the equation into (1+3/a)(1+3/b)(1+3/c)=4(1+3/a)(1+3/b)(1+3/c) = 4, and since each factor shrinks as its variable grows, the smallest variable is squeezed into {4,5}\{4,5\}.

Divide both sides by abcabc: (1+3a)(1+3b)(1+3c)=4.\left(1 + \frac3a\right)\left(1 + \frac3b\right)\left(1 + \frac3c\right) = 4. Each factor decreases as its variable increases, which is what makes the search finite. Order the variables so that a≤b≤ca \le b \le c; then the largest factor is the one with aa, and (1+3a)3≥4.\left(1 + \frac3a\right)^3 \ge 4. No cube root is needed: if a≥6a \ge 6 then 1+3a≤321 + \tfrac3a \le \tfrac32 and (32)3=278<4\left(\tfrac32\right)^3 = \tfrac{27}{8} < 4, which fails. With a≥4a \ge 4 given, that leaves a∈{4,5}a \in \{4, 5\}.

a=4a = 4

The remaining two factors must multiply to 4÷74=1674 \div \tfrac74 = \tfrac{16}{7}. Since bb is the smaller of the two, (1+3b)2≥167\left(1 + \tfrac3b\right)^2 \ge \tfrac{16}{7}. Again no root is needed: b≥6b \ge 6 would give (1+3b)2≤94<167\left(1+\tfrac3b\right)^2 \le \tfrac94 < \tfrac{16}{7}, since 9⋅7=63<64=16⋅49 \cdot 7 = 63 < 64 = 16 \cdot 4. So b∈{4,5}b \in \{4,5\}.

  • b=4b = 4: then 1+3/c=167÷74=64491 + 3/c = \tfrac{16}{7} \div \tfrac74 = \tfrac{64}{49}, so c=495c = \tfrac{49}{5}, not an integer.

  • b=5b = 5: then 1+3/c=167÷85=1071 + 3/c = \tfrac{16}{7} \div \tfrac85 = \tfrac{10}{7}, so 3/c=3/73/c = 3/7 and c=7c = 7.

a=5a = 5

Now the other two factors multiply to 4÷85=524 \div \tfrac85 = \tfrac52, and b≥5b \ge 5 with (1+3b)2≥52\left(1+\tfrac3b\right)^2 \ge \tfrac52 forces b≤5b \le 5, since b≥6b \ge 6 would give at most 94<52\tfrac94 < \tfrac52. So b=5b = 5. Then 1+3/c=52÷85=25161 + 3/c = \tfrac52 \div \tfrac85 = \tfrac{25}{16}, giving c=163c = \tfrac{16}{3}, not an integer.

So up to order the only solution is {a,b,c}={4,5,7}\{a,b,c\} = \{4,5,7\}, which is not all equal as required, and indeed 4⋅4⋅5⋅7=560=7⋅8⋅104 \cdot 4 \cdot 5 \cdot 7 = 560 = 7 \cdot 8 \cdot 10. Hence a+b+c=16.a + b + c = 16.

Answer 16

Solution: PRMO 2018, Q23

Key idea

Cauchy-Schwarz in the form ∑a2/x≥(∑a)2/∑x\sum a^2/x \ge (\sum a)^2/\sum x applies exactly here, and the three denominators add to (a+b+c)(1/29+1/31)(a+b+c)(1/29 + 1/31), so the best constant is 899/60899/60.

The left-hand side is a sum of squares over positive quantities, and there is an inequality that handles exactly that shape: for positive x,y,zx, y, z, a2x+b2y+c2z≥(a+b+c)2x+y+z.\frac{a^2}{x} + \frac{b^2}{y} + \frac{c^2}{z} \ge \frac{(a+b+c)^2}{x+y+z}.

It is worth seeing where that comes from, because two terms already contain the whole idea. Clearing denominators in a2x+b2y≥(a+b)2x+y\frac{a^2}{x} + \frac{b^2}{y} \ge \frac{(a+b)^2}{x+y} turns it into (a2y+b2x)(x+y)≥(a+b)2xy\left(a^2y + b^2x\right)(x+y) \ge (a+b)^2xy, and expanding both sides leaves a2y2+b2x2≥2abxya^2y^2 + b^2x^2 \ge 2abxy, which is (ay−bx)2≥0(ay - bx)^2 \ge 0. Applying the two-term version twice, first to the last pair and then to what is left, a2x+b2y+c2z ≥ a2x+(b+c)2y+z ≥ (a+b+c)2x+y+z,\frac{a^2}{x} + \frac{b^2}{y} + \frac{c^2}{z} \ \ge\ \frac{a^2}{x} + \frac{(b+c)^2}{y+z} \ \ge\ \frac{(a+b+c)^2}{x+y+z}, which is the three-term form. It is Cauchy-Schwarz wearing a convenient disguise, and equality needs ay=bxay = bx at each step, that is a:b:ca : b : c matching x:y:zx : y : z.

Here the three denominators are b29+c31\tfrac{b}{29} + \tfrac{c}{31}, c29+a31\tfrac{c}{29} + \tfrac{a}{31} and a29+b31\tfrac{a}{29} + \tfrac{b}{31}, and every one of aa, bb, cc appears once with each coefficient, so they add to (a+b+c)(129+131)=(a+b+c)⋅60899.(a+b+c)\left(\frac{1}{29} + \frac{1}{31}\right) = (a+b+c)\cdot\frac{60}{899}. Therefore LHS≥(a+b+c)2⋅89960(a+b+c)=89960(a+b+c),\text{LHS} \ge \frac{(a+b+c)^2 \cdot 899}{60(a+b+c)} = \frac{899}{60}(a+b+c), and 899/60=14.98…899/60 = 14.98\ldots

Equality, by the condition noted above, needs a:b:ca : b : c to match the denominators, which happens at a=b=ca = b = c; there the left-hand side is exactly 89960(a+b+c)\tfrac{899}{60}(a+b+c). So no constant larger than 899/60899/60 can work, and the largest integer that does is n=14.n = 14.

Answer 14

Solution: PRMO 2019, Q22

Key idea

1/n1/\sqrt n sits between the two telescoping differences 2(n+1−n)2(\sqrt{n+1} - \sqrt n) and 2(n−n−1)2(\sqrt n - \sqrt{n-1}), and 1599+1=16001599 + 1 = 1600 is a perfect square, which is why the bound is sharp enough to name the integer part.

We want nearby terms that cancel when added. The difference of adjacent square roots is promising: n+1−n=1n+1+n.\sqrt{n+1}-\sqrt n = \frac1{\sqrt{n+1}+\sqrt n}. Its denominator is close to 2n2\sqrt n, so twice this difference should be close to 1/n1/\sqrt n. Using the next root gives one side of the bound and using the previous root gives the other:

2(n+1−n)=2n+1+n<22n=1n,2\left(\sqrt{n+1} - \sqrt n\right) = \frac{2}{\sqrt{n+1} + \sqrt n} < \frac{2}{2\sqrt n} = \frac{1}{\sqrt n}, 2(n−n−1)=2n+n−1>22n=1n(n≥2).2\left(\sqrt{n} - \sqrt{n-1}\right) = \frac{2}{\sqrt{n} + \sqrt{n-1}} > \frac{2}{2\sqrt n} = \frac{1}{\sqrt n} \qquad (n \ge 2). Both are strict, and both telescope when summed.

Write SS for the sum in question. From below, using the first estimate for every term, S>∑n=115992(n+1−n)=2(1600−1)=2(40−1)=78.S > \sum_{n=1}^{1599} 2\left(\sqrt{n+1} - \sqrt n\right) = 2\left(\sqrt{1600} - 1\right) = 2(40 - 1) = 78. From above, keeping the first term separate because the second estimate needs n≥2n \ge 2, S<1+∑n=215992(n−n−1)=1+2(1599−1)=21599−1,S < 1 + \sum_{n=2}^{1599} 2\left(\sqrt n - \sqrt{n-1}\right) = 1 + 2\left(\sqrt{1599} - 1\right) = 2\sqrt{1599} - 1, and that is smaller than 2×40−1=792 \times 40 - 1 = 79, because 1599<16001599 < 1600.

So 78<S<7978 < S < 79, and the greatest integer not exceeding SS is 7878.

The number 15991599 was chosen with care, though not quite as tightly as it first looks. One term fewer breaks the argument: the lower bound would read 2(1599−1)2(\sqrt{1599} - 1), which is below 7878. One term more is still fine, since the upper bound becomes exactly 7979 and the inequality is strict. Two terms more does break it, the upper bound passing 7979. So the window is narrow, and the reason 16001600 appears in the lower estimate is exactly that the sum runs to 15991599.

Answer 78

Solution: PRMO 2019, Q26

Key idea

For fixed y≥2y \ge 2 the quantity x+yzx + yz falls as xx rises, so xx takes its largest value ⌊159/y⌋\lfloor 159/y \rfloor, and writing 159=qy+s159 = qy + s collapses the whole expression to y(1+s)+qy(1+s) + q.

Since z≥1z \ge 1, the constraint gives xy≤159xy \le 159. Substituting z=160−xyz = 160 - xy, x+yz=x+y(160−xy)=160y−x(y2−1).x + yz = x + y(160 - xy) = 160y - x(y^2 - 1). For y=1y = 1 this is just x+z=160x + z = 160. For y≥2y \ge 2 the coefficient y2−1y^2 - 1 is positive, so the expression decreases as xx grows, and xx should be as large as the constraint allows: x=⌊159y⌋.x = \left\lfloor \frac{159}{y} \right\rfloor.

Write 159=qy+s159 = qy + s with 0≤s<y0 \le s < y, so that x=qx = q and z=160−qy=s+1z = 160 - qy = s + 1. Then x+yz=160y−q(y2−1)=160y−y(qy)+q,x + yz = 160y - q(y^2 - 1) = 160y - y(qy) + q, and qy=159−sqy = 159 - s, so the whole thing collapses to y(1+s)+qy(1+s) + q. Because xx is forced for each yy, every row below is a single evaluation of this formula rather than a search over xx. Running yy from 11 to 2424:

y12345678x+yz16083565556506483y910111213141516x+yz8011580616495160265y1718192021222324x+yz128296160407280139512390\small \begin{array}{r|rrrrrrrr} y & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 \\ \hline x + yz & 160 & 83 & 56 & 55 & 56 & \mathbf{50} & 64 & 83 \\[6pt] y & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 \\ \hline x + yz & 80 & 115 & 80 & 61 & 64 & 95 & 160 & 265 \\[6pt] y & 17 & 18 & 19 & 20 & 21 & 22 & 23 & 24 \\ \hline x + yz & 128 & 296 & 160 & 407 & 280 & 139 & 512 & 390 \end{array}

Larger yy need no table. If s≥1s \ge 1 then y(1+s)+q≥2y+q≥2×25+1=51y(1+s) + q \ge 2y + q \ge 2 \times 25 + 1 = 51 for y≥25y \ge 25. If s=0s = 0 then yy divides 159=3×53159 = 3 \times 53, so yy is 5353 or 159159, giving 53+3=5653 + 3 = 56 and 159+1=160159 + 1 = 160. Every value beyond the table exceeds 5050.

The minimum is therefore 5050, at y=6y = 6, x=26x = 26, z=4z = 4: indeed 26×6+4=16026 \times 6 + 4 = 160 and 26+6×4=5026 + 6 \times 4 = 50.

Why y=6y = 6 and not some neighbour? Rearranging the constraint gives the identity x+yz=160+(y−1)(z−x)x + yz = 160 + (y-1)(z - x), so we want (y−1)(x−z)(y-1)(x-z) large. Raising yy helps the first factor and hurts the second, and y=6y = 6 divides 159159 almost exactly, which keeps zz small while xx is still 2626.

Answer 50

Solution: IOQM 2022, Q15

Key idea

With s=x+ys = x+y and d=x−yd = x-y, the constraint xy=1xy = 1 says exactly s2−d2=4s^2 - d^2 = 4, which removes ss entirely and leaves a one-variable function of dd.

Set s=x+ys = x + y and d=x−yd = x - y. Then s2−d2=4xy=4s^2 - d^2 = 4xy = 4, so s2=d2+4s^2 = d^2 + 4 and in particular s2−2=d2+2s^2 - 2 = d^2 + 2. Substituting into the expression, (x+y)2−(x−y)−2(x+y)2+(x−y)−2=d2−d+2d2+d+2.\frac{(x+y)^2 - (x-y) - 2}{(x+y)^2 + (x-y) - 2} = \frac{d^2 - d + 2}{d^2 + d + 2}. The two variables have become one, and dd ranges over all real numbers, since for any dd we may take s=±d2+4s = \pm\sqrt{d^2+4} and recover a genuine pair with xy=1xy = 1.

To find the range, ask which values kk are attained. The denominator never vanishes, since d2+d+2=(d+12)2+74>0,d^2 + d + 2 = \left(d + \tfrac12\right)^2 + \tfrac74 > 0, so clearing it is safe. Setting the expression equal to kk and clearing the denominator, (1−k)d2−(1+k)d+(2−2k)=0,(1-k)d^2 - (1+k)d + (2 - 2k) = 0, For k=1k=1 this is the linear equation −2d=0-2d=0, so that value is attained at d=0d=0. For k≠1k\ne1 the equation is quadratic, and a real dd exists exactly when its discriminant is non-negative: (1+k)2−4(1−k)(2−2k)=(1+k)2−8(1−k)2≥0.(1+k)^2 - 4(1-k)(2-2k) = (1+k)^2 - 8(1-k)^2 \geq 0. Expanding, 1+2k+k2−8+16k−8k2≥01 + 2k + k^2 - 8 + 16k - 8k^2 \geq 0, that is 7k2−18k+7≤07k^2 - 18k + 7 \leq 0.

A quadratic is negative between its roots, so the attained values fill the closed interval between them, and TT and tt are precisely those roots. We do not even need to compute them separately, because the sum of the roots of 7k2−18k+77k^2 - 18k + 7 is 187\tfrac{18}{7} by Vieta.

So T+t=187T + t = \tfrac{18}{7}, giving m=18m = 18, n=7n = 7 and m+n=25m + n = 25.

Answer 25

Solution: IOQM 2024, Q11

Key idea

Adding the two equations makes each fraction complete itself: the sum becomes a+12b+1+2b+13c+1+3c+1a+1=3\frac{a+1}{2b+1} + \frac{2b+1}{3c+1} + \frac{3c+1}{a+1} = 3, three positive numbers with product 11, so all three are 11.

Write x=ax = a, y=2by = 2b, z=3cz = 3c, all positive. The two conditions are xy+1+yz+1+zx+1=1,1x+1+1y+1+1z+1=2.\frac{x}{y+1} + \frac{y}{z+1} + \frac{z}{x+1} = 1, \qquad \frac{1}{x+1} + \frac{1}{y+1} + \frac{1}{z+1} = 2.

Add them, pairing each fraction of the first with the fraction of the second that shares its denominator: x+1y+1+y+1z+1+z+1x+1=3.\frac{x+1}{y+1} + \frac{y+1}{z+1} + \frac{z+1}{x+1} = 3. The three terms are positive and their product is 11, so by the arithmetic-geometric mean

inequality their sum is at least 33, with equality only when all three equal 11. Equality is exactly what we have, so x+1=y+1=z+1,that isx=y=z.x + 1 = y + 1 = z + 1, \qquad \text{that is} \qquad x = y = z.

Substituting into the first equation, 3xx+1=1\dfrac{3x}{x+1} = 1, so x=12x = \tfrac12. Unwinding, a=12,2b=12⇒b=14,3c=12⇒c=16,a = \tfrac12, \qquad 2b = \tfrac12 \Rightarrow b = \tfrac14, \qquad 3c = \tfrac12 \Rightarrow c = \tfrac16, and therefore 1a+1b+1c=2+4+6=12.\frac1a + \frac1b + \frac1c = 2 + 4 + 6 = 12.

Answer 12

Solution: IOQM 2025 Part SEP, Q7

Key idea

With MI=1/MTMI = 1/MT the product of the two side lengths is 11, so the area alone gives the sine of the angle between them, and the diagonal is minimised by equal sides (a rhombus) and the obtuse angle.

Adjacent sides p,q and their included angle \theta .
Adjacent sides p,qp,q and their included angle θ\theta.

Write p=MTp = MT and q=MIq = MI, so that pq=1pq = 1, and let θ\theta be the angle between those two sides at MM. The area of a parallelogram is the product of two adjacent sides times the sine of the angle between them, so pqsin⁡θ=sin⁡θ=4041,cos⁡θ=±941.pq \sin\theta = \sin\theta = \frac{40}{41}, \qquad \cos\theta = \pm\frac{9}{41}.

The diagonal MAMA is the sum of the two side vectors, so by the cosine rule MA2=p2+q2+2pqcos⁡θ=p2+1p2+2cos⁡θ.MA^2 = p^2 + q^2 + 2pq\cos\theta = p^2 + \frac{1}{p^2} + 2\cos\theta. Two independent choices now make this as small as possible. First, p2+1/p2≥2p^2 + 1/p^2 \ge 2 by the arithmetic-geometric mean inequality, with equality when p=1p = 1. Second, cos⁡θ\cos\theta should be the negative value −941-\tfrac{9}{41}, which is the obtuse choice.

Hence d2=2−1841=82−1841=6441,d^2 = 2 - \frac{18}{41} = \frac{82 - 18}{41} = \frac{64}{41}, and since gcd⁡(64,41)=1\gcd(64,41) = 1 we read off a=64a = 64, b=41b = 41 and ∣a−b∣=23|a - b| = 23.

Answer 23

Report an error on this page

Reports are stored by Netlify. See the privacy note.