🇮🇳 INMO · 2016

INMO 2016

Indian National Mathematical Olympiad 2016 — 6 problems with solutions

INMO 2016

Indian National Mathematical Olympiad 2016. 6 problems. Problems sourced from MathNet (CC BY 4.0). Solutions are expert-authored.


Problem 1

Algebra
Problem

Let mm and nn be positive integers such that the equation x2mx+n=0x^{2}-m x+n=0 has real roots α\alpha and β\beta. Prove that α\alpha and β\beta are integers if and only if [mα]+[mβ][m \alpha]+[m \beta] is the square of an integer. (Here [x][x] denotes the largest integer not exceeding xx.)

View Solution

If α\alpha and β\beta are both integers, then

[mα]+[mβ]=mα+mβ=m(α+β)=m2[m \alpha]+[m \beta]=m \alpha+m \beta=m(\alpha+\beta)=m^{2}

This proves one implication.

Observe that α+β=m\alpha+\beta=m and αβ=n\alpha \beta=n. We use the property of integer function: x1<[x]xx-1<[x] \leq x for any real number xx. Thus

m22=m(α+β)2=mα1+mβ1<[mα]+[mβ]m(α+β)=m2.m^{2}-2=m(\alpha+\beta)-2=m \alpha-1+m \beta-1<[m \alpha]+[m \beta] \leq m(\alpha+\beta)=m^{2}.

Since mm and nn are positive integers, both α\alpha and β\beta must be positive. If m2m \geq 2, we observe that there is no square between m22m^{2}-2 and m2m^{2}. Hence, either m=1m=1 or [mα]+[mβ]=m2[m \alpha]+[m \beta]=m^{2}. If m=1m=1, then α+β=1\alpha+\beta=1 implies that both α\alpha and β\beta are positive reals smaller than 1. Hence n=αβn=\alpha \beta cannot be a positive integer. We conclude that [mα]+[mβ]=m2[m \alpha]+[m \beta]=m^{2}.

Putting m=α+βm=\alpha+\beta in this relation, we get

[α2+n]+[β2+n]=(α+β)2\left[\alpha^{2}+n\right]+\left[\beta^{2}+n\right]= (\alpha+\beta)^{2}

Using [x+k]=[x]+k[x+k]=[x]+k for any real number xx and integer kk, this reduces to

[α2]+[β2]=α2+β2\left[\alpha^{2}\right]+\left[\beta^{2}\right]=\alpha^{2}+\beta^{2}

This shows that α2\alpha^{2} and β2\beta^{2} are both integers. On the other hand,

α2β2=(α+β)(αβ)=m(αβ)\alpha^{2}-\beta^{2}=(\alpha+\beta)(\alpha-\beta)=m(\alpha-\beta)

Thus

(αβ)=α2β2m(\alpha-\beta)=\frac{\alpha^{2}-\beta^{2}}{m}

is a rational number. Since α+β=m\alpha+\beta=m is a rational number, it follows that both α\alpha and β\beta are rational numbers. However, both α2\alpha^{2} and β2\beta^{2} are integers. Hence each of α\alpha and β\beta is an integer.


Problem 2

Problem

Let σ=(a1,a2,a3,,an)\sigma=\left(a_{1}, a_{2}, a_{3}, \ldots, a_{n}\right) be a permutation of (1,2,3,,n)(1,2,3, \ldots, n). A pair (ai,aj)\left(a_{i}, a_{j}\right) is said to correspond to an inversion of σ\sigma, if i<ji<j but ai>aja_{i}>a_{j}. (Example: In the permutation (2,4,5,3,1)(2,4,5,3,1), there are 6 inversions corresponding to the pairs (2,1)(2,1), (4,3),(4,1),(5,3),(5,1),(3,1)(4,3),(4,1),(5,3),(5,1),(3,1).) How many permutations of (1,2,3,n)(1,2,3, \ldots n), (n3)(n \geq 3), have exactly two inversions?

View Solution

In a permutation of (1,2,3,,n)(1,2,3, \ldots, n), two inversions can occur in only one of the following two ways: (A) Two disjoint consecutive pairs are interchanged:

(1,2,3,,j1,j,j+1,,k1,k,k+1,k+2,,n)(1,2,,j1,j+1,j,j+2,,k1,k+1,k,k+2,,n)\begin{aligned} & (1,2,3, \ldots, j-1, j, j+1, \ldots, k-1, k, k+1, k+2, \ldots, n) \\ & \quad \longrightarrow (1,2, \ldots, j-1, j+1, j, j+2, \ldots, k-1, k+1, k, k+2, \ldots, n) \end{aligned}

(B) Each block of three consecutive integers can be permuted in any of the following 2 ways:

(1,2,3,,k,k+1,k+2,,n)(1,2,,k+2,k,k+1,,n)(1,2,3,,k,k+1,k+2,,n)(1,2,,k+1,k+2,k,,n)\begin{aligned} & (1,2,3, \ldots, k, k+1, k+2, \ldots, n) \longrightarrow (1,2, \ldots, k+2, k, k+1, \ldots, n) \\ & (1,2,3, \ldots, k, k+1, k+2, \ldots, n) \longrightarrow (1,2, \ldots, k+1, k+2, k, \ldots, n) \end{aligned}

Consider case (A). For j=1j=1, there are n3n-3 possible values of kk; for j=2j=2, there are n4n-4 possibilities for kk and so on. Thus the number of permutations with two inversions of this type is

1+2++(n3)=(n3)(n2)21+2+\cdots+(n-3)=\frac{(n-3)(n-2)}{2}

In case (B), we see that there are n2n-2 permutations of each type, since kk can take values from 11 to n2n-2. Hence we get 2(n2)2(n-2) permutations of this type. Finally, the number of permutations with two inversions is

(n3)(n2)2+2(n2)=(n+1)(n2)2\frac{(n-3)(n-2)}{2}+2(n-2)=\frac{(n+1)(n-2)}{2}

Problem 3

Geometry
Problem

Let ABCABC be a triangle in which AB=ACAB = AC. Let DD be the mid-point of BCBC and PP be a point on ADAD. Suppose EE is the foot of perpendicular from PP on ACAC. If APPD=BPPE=λ\frac{AP}{PD} = \frac{BP}{PE} = \lambda, BDAD=m\frac{BD}{AD} = m and z=m2(1+λ)z = m^2(1+\lambda), prove that

z2(λ3λ22)z+1=0z^2 - (\lambda^3 - \lambda^2 - 2)z + 1 = 0

Hence show that λ2\lambda \geq 2 and λ=2\lambda = 2 if and only if ABCABC is equilateral.

Note

This problem originally includes a geometry diagram. The diagram is omitted in this version.

View Solution

Let AD=hAD = h, PD=yPD = y and BD=DC=aBD = DC = a. We observe that BP2=a2+y2BP^2 = a^2 + y^2. Moreover, PE=PAsinDAC=(hy)DCAC=a(hy)bPE = PA \sin \angle DAC = (h - y) \frac{DC}{AC} = \frac{a(h - y)}{b}, where b=AC=ABb = AC = AB. Using AP/PD=(hy)/yAP / PD = (h - y) / y, we obtain y=h/(1+λ)y = h / (1 + \lambda). Thus

λ2=BP2PE2=(a2+y2)b2(hy)2a2\lambda^2 = \frac{BP^2}{PE^2} = \frac{(a^2 + y^2) b^2}{(h - y)^2 a^2}

But (hy)=λy=λh/(1+λ)(h - y) = \lambda y = \lambda h / (1 + \lambda) and b2=a2+h2b^2 = a^2 + h^2. Thus we obtain

λ4=(a2(1+λ)2+h2)(a2+h2)a2h2\lambda^4 = \frac{(a^2(1 + \lambda)^2 + h^2)(a^2 + h^2)}{a^2 h^2}

Using m=a/hm = a / h and z=m2(1+λ)z = m^2(1 + \lambda), this simplifies to

z2z(λ3λ22)+1=0z^2 - z(\lambda^3 - \lambda^2 - 2) + 1 = 0

Dividing by zz, this gives

z+1z=λ3λ22z + \frac{1}{z} = \lambda^3 - \lambda^2 - 2

However z+(1/z)2z + (1 / z) \geq 2 for any positive real number zz. Thus λ3λ240\lambda^3 - \lambda^2 - 4 \geq 0. This may be written in the form (λ2)(λ2+λ+2)0(\lambda - 2)(\lambda^2 + \lambda + 2) \geq 0. But λ2+λ+2>0\lambda^2 + \lambda + 2 > 0. (For example, one may check that its discriminant is negative.) Hence λ2\lambda \geq 2. If λ=2\lambda = 2, then z+(1/z)=2z + (1 / z) = 2 and hence z=1z = 1. This gives m2=1/3m^2 = 1 / 3 or tan(A/2)=m=1/3\tan (A / 2) = m = 1 / \sqrt{3}. Thus A=60A = 60^\circ and hence ABCABC is equilateral.

Conversely, if triangle ABCABC is equilateral, then m=tan(A/2)=1/3m = \tan (A / 2) = 1 / \sqrt{3} and hence z=(1+λ)/3z = (1 + \lambda) / 3. Substituting this in the equation satisfied by zz, we obtain

(1+λ)23(1+λ)(λ3λ22)+9=0(1 + \lambda)^2 - 3(1 + \lambda)(\lambda^3 - \lambda^2 - 2) + 9 = 0

This may be written in the form (λ2)(3λ3+6λ2+8λ+8)=0(\lambda - 2)(3\lambda^3 + 6\lambda^2 + 8\lambda + 8) = 0. Here the second factor is positive because λ>0\lambda > 0. We conclude that λ=2\lambda = 2.


Problem 4

Algebra Geometry
Problem

If x,y,zx, y, z are positive real numbers, prove that

(x+y+z)2(yz+zx+xy)23(y2+yz+z2)(z2+zx+x2)(x2+xy+y2)(x+y+z)^{2}(y z+z x+x y)^{2} \leq 3\left(y^{2}+y z+z^{2}\right)\left(z^{2}+z x+x^{2}\right)\left(x^{2}+x y+y^{2}\right)
View Solution

Solution 1.

We begin with the observation that

x2+xy+y2=34(x+y)2+14(xy)234(x+y)2x^{2}+x y+y^{2}=\frac{3}{4}(x+y)^{2}+\frac{1}{4}(x-y)^{2} \geq \frac{3}{4}(x+y)^{2}

and similar bounds for y2+yz+z2,z2+zx+x2y^{2}+y z+z^{2}, z^{2}+z x+x^{2}. Thus

3(x2+xy+y2)(y2+yz+z2)(z2+zx+x2)8164(x+y)2(y+z)2(z+x)23\left(x^{2}+x y+y^{2}\right)\left(y^{2}+y z+z^{2}\right)\left(z^{2}+z x+x^{2}\right) \geq \frac{81}{64}(x+y)^{2}(y+z)^{2}(z+x)^{2}

Thus it is sufficient to prove that

(x+y+z)(xy+yz+zx)98(x+y)(y+z)(z+x)(x+y+z)(x y+y z+z x) \leq \frac{9}{8}(x+y)(y+z)(z+x)

Equivalently, we need to prove that

8(x+y+z)(xy+yz+zx)9(x+y)(y+z)(z+x)8(x+y+z)(x y+y z+z x) \leq 9(x+y)(y+z)(z+x)

However, we note that

(x+y)(y+z)(z+x)=(x+y+z)(yz+zx+xy)xyz(x+y)(y+z)(z+x)=(x+y+z)(y z+z x+x y)-x y z

Thus the required inequality takes the form

(x+y)(y+z)(z+x)8xyz(x+y)(y+z)(z+x) \geq 8 x y z

This follows from AM-GM inequalities;

x+y2xy,y+z2yz,z+x2zxx+y \geq 2 \sqrt{x y}, \quad y+z \geq 2 \sqrt{y z}, \quad z+x \geq 2 \sqrt{z x}
View Solution

Solution 2.

Let us introduce x+y=c,y+z=ax+y=c, y+z=a and z+x=bz+x=b. Then a,b,ca, b, c are the sides of a triangle. If s=(a+b+c)/2s=(a+b+c) / 2, then it is easy to calculate x=sa,y=sb,z=scx=s-a, y=s-b, z=s-c and x+y+z=sx+y+z=s. We also observe that x2+xy+y2=(x+y)2xy=c214(c+ab)(c+ba)=34c2+14(ab)234c2x^{2}+x y+y^{2}=(x+y)^{2}-x y=c^{2}-\frac{1}{4}(c+a-b)(c+b-a)=\frac{3}{4} c^{2}+\frac{1}{4}(a-b)^{2} \geq \frac{3}{4} c^{2}. Moreover, xy+yz+zx=(sa)(sb)+(sb)(sc)+(sc)(sa)x y+y z+z x=(s-a)(s-b)+(s-b)(s-c)+(s-c)(s-a). Thus it is sufficient to prove that

s(sa)(sb)98abcs \sum(s-a)(s-b) \leq \frac{9}{8} a b c

But, (sa)(sb)=r(4R+r)\sum(s-a)(s-b)=r(4 R+r), where r,Rr, R are respectively the in-radius, the circum-radius of the triangle whose sides are a,b,ca, b, c, and abc=4Rrsa b c=4 R r s. Thus the inequality reduces to

r(4R+r)92Rrr(4 R+r) \leq \frac{9}{2} R r

This is simply 2rR2 r \leq R. This follows from IO2=R(R2r)I O^{2}=R(R-2 r), where II is the incentre and OO the circumcentre.

View Solution

Solution 3.

If we set x=λa,y=λb,z=λcx=\lambda a, y=\lambda b, z=\lambda c, then the inequality changes to

(a+b+c)2(ab+bc+ca)23(a2+ab+b2)(b2+bc+c2)(c2+ca+a2)(a+b+c)^{2}(a b+b c+c a)^{2} \leq 3\left(a^{2}+a b+b^{2}\right)\left(b^{2}+b c+c^{2}\right)\left(c^{2}+c a+a^{2}\right)

This shows that we may assume x+y+z=1x+y+z=1. Let α=xy+yz+zx\alpha=x y+y z+z x. We see that

x2+xy+y2=(x+y)2xy=(x+y)(1z)xy=x+yα=1zα\begin{aligned} x^{2}+x y+y^{2} & =(x+y)^{2}-x y \\ & =(x+y)(1-z)-x y \\ & =x+y-\alpha=1-z-\alpha \end{aligned}

Thus

(x2+xy+y2)=(1αz)(1αx)(1αy)=(1α)3(1α)2+(1α)αxyz=α2α3xyz\begin{aligned} \prod\left(x^{2}+x y+y^{2}\right) & =(1-\alpha-z)(1-\alpha-x)(1-\alpha-y) \\ & =(1-\alpha)^{3}-(1-\alpha)^{2}+(1-\alpha) \alpha-x y z \\ & =\alpha^{2}-\alpha^{3}-x y z \end{aligned}

Thus we need to prove that α23(α2α3xyz)\alpha^{2} \leq 3\left(\alpha^{2}-\alpha^{3}-x y z\right). This reduces to

3xyzα2(23α)3 x y z \leq \alpha^{2}(2-3 \alpha)

However

3α=3(xy+yz+zx)(x+y+z)2=13 \alpha=3(x y+y z+z x) \leq(x+y+z)^{2}=1

so that 23α12-3 \alpha \geq 1. Thus it suffices to prove that 3xyzα23 x y z \leq \alpha^{2}. But

α23xyz=(xy+yz+zx)23xyz(x+y+z)=cyclic x2y2xyz(x+y+z)=12cyclic (xyyz)20\begin{aligned} \alpha^{2}-3 x y z & =(x y+y z+z x)^{2}-3 x y z(x+y+z) \\ & =\sum_{\text {cyclic }} x^{2} y^{2}-x y z(x+y+z) \\ & =\frac{1}{2} \sum_{\text {cyclic }}(x y-y z)^{2} \geq 0 \end{aligned}

Problem 5

Geometry
Problem

Let ABCABC be a triangle, II its in-centre; A1,B1,C1A_{1}, B_{1}, C_{1} be the reflections of II in BC,CA,ABBC, CA, AB respectively. Suppose the circum-circle of triangle A1B1C1A_{1}B_{1}C_{1} passes through AA. Prove that B1,C1,I,I1B_{1}, C_{1}, I, I_{1} are concyclic, where I1I_{1} is the in-centre of triangle A1B1C1A_{1}B_{1}C_{1}.

Note

This problem originally includes a geometry diagram. The diagram is omitted in this version.

View Solution

Note that IA1=IB1=IC1=2rIA_{1} = IB_{1} = IC_{1} = 2r, where rr is the in-radius of the triangle ABCABC. Hence II is the circum-centre of the triangle A1B1C1A_{1}B_{1}C_{1}.

Let KK be the point of intersection of IB1IB_{1} and ACAC. Then IK=rIK = r, IA=2rIA = 2r and IKA=90\angle IKA = 90^{\circ}. It follows that IAK=30\angle IAK = 30^{\circ} and hence IAB1=60\angle IAB_{1} = 60^{\circ}. Thus AIB1AIB_{1} is an equilateral triangle. Similarly triangle AIC1AIC_{1} is also equilateral. We hence obtain AB1=AC1=AI=IB1=IC1=2rAB_{1} = AC_{1} = AI = IB_{1} = IC_{1} = 2r.

We also observe that B1IC1=120\angle B_{1}IC_{1} = 120^{\circ} and IB1AC1IB_{1}AC_{1} is a rhombus. Thus B1AC1=120\angle B_{1}AC_{1} = 120^{\circ} and by concyclicity A1=60\angle A_{1} = 60^{\circ}. Since AB1=AC1AB_{1} = AC_{1}, AA is the midpoint of the arc B1AC1B_{1}AC_{1}. It follows that A1AA_{1}A bisects A1\angle A_{1} and I1I_{1} lies on the line A1AA_{1}A. This implies that

B1I1C1=90+A1/2=90+30=120\angle B_{1}I_{1}C_{1} = 90^{\circ} + \angle A_{1}/2 = 90^{\circ} + 30^{\circ} = 120^{\circ}

Since B1IC1=120\angle B_{1}IC_{1} = 120^{\circ}, we conclude that B1,I,I1,C1B_{1}, I, I_{1}, C_{1} are concyclic. (Further AA is the centre.)


Problem 6

Number Theory Algebra
Problem

Find all triples (p,x,y)(p, x, y) such that px=y4+4p^{x} = y^{4} + 4, where pp is a prime and x,yx, y are natural numbers.

View Solution

We begin with the standard factorisation

y4+4=(y22y+2)(y2+2y+2)y^{4}+4 = \left(y^{2}-2y+2\right)\left(y^{2}+2y+2\right)

Thus we have y22y+2=pmy^{2}-2y+2 = p^{m} and y2+2y+2=pny^{2}+2y+2 = p^{n} for some positive integers mm and nn such that m+n=xm+n = x. Since y22y+2<y2+2y+2y^{2}-2y+2 < y^{2}+2y+2, we have m<nm < n so that pmp^{m} divides pnp^{n}. Thus y22y+2y^{2}-2y+2 divides y2+2y+2y^{2}+2y+2. Writing y2+2y+2=y22y+2+4yy^{2}+2y+2 = y^{2}-2y+2 + 4y, we infer that y22y+2y^{2}-2y+2 divides 4y4y and hence y22y+2y^{2}-2y+2 divides 4y24y^{2}. But

4y2=4(y22y+2)+8(y1)4y^{2} = 4\left(y^{2}-2y+2\right) + 8(y-1)

Thus y22y+2y^{2}-2y+2 divides 8(y1)8(y-1). Since y22y+2y^{2}-2y+2 divides both 4y4y and 8(y1)8(y-1), we conclude that it also divides 88. This gives y22y+2=1,2,4y^{2}-2y+2 = 1, 2, 4 or 88. If y22y+2=1y^{2}-2y+2 = 1, then y=1y = 1 and y4+4=5y^{4}+4 = 5, giving p=5p = 5 and x=1x = 1. If y22y+2=2y^{2}-2y+2 = 2, then y22y=0y^{2}-2y = 0 giving y=2y = 2. But then y4+4=20y^{4}+4 = 20 is not the power of a prime. The equations y22y+2=4y^{2}-2y+2 = 4 and y22y+2=8y^{2}-2y+2 = 8 have no integer solutions. We conclude that (p,x,y)=(5,1,1)(p, x, y) = (5, 1, 1) is the only solution.

Alternatively, using y22y+2=pmy^{2}-2y+2 = p^{m} and y2+2y+2=pny^{2}+2y+2 = p^{n}, we may get

4y=pm(pnm1)4y = p^{m}\left(p^{n-m}-1\right)

If m>0m > 0, then pp divides 44 or yy. If pp divides 44, then p=2p = 2. If pp divides yy, then y22y+2=pmy^{2}-2y+2 = p^{m} shows that pp divides 22 and hence p=2p = 2. But then 2x=y4+42^{x} = y^{4}+4, which shows that yy is even. Taking y=2zy = 2z, we get 2x2=4z4+12^{x-2} = 4z^{4}+1. This implies that z=0z = 0 and hence y=0y = 0, which is a contradiction. Thus m=0m = 0 and y22y+2=1y^{2}-2y+2 = 1. This gives y=1y = 1 and hence p=5,x=1p = 5, x = 1.