Olympiad Maths Prep

Library / /28 of 28

Algebra Difficulty 8.0 Shortlist Prove it Ukraine

Find all polynomials P(x)P(x) with real coefficients, such that for every distinct natural x,y,z,tx, y, z, t satisfying x2+y2+z2=2t2x^2 + y^2 + z^2 = 2t^2 and GCD{x,y,z,t}=1\text{GCD}\{x, y, z, t\} = 1, the following equality holds:
2P2(t)+2P(xy+yz+zx)=P2(x+y+z). 2P^2(t) + 2P(xy + yz + zx) = P^2(x + y + z).

Solutions — 2

Solution 1

Take t=5kt = 5k and z=7kz = 7k, then we want the equality x2+y2=k2x^2 + y^2 = k^2 to hold. Consider two series of solutions: x=n21x = n^2 - 1, y=2ny = 2n, k=n2+1k = n^2 + 1 and x=3n2+4n+1x = 3n^2 + 4n + 1, y=2n(2n+1)y = 2n(2n+1), k=5n2+4n+1k = 5n^2 + 4n + 1. It is evident that in both cases (x,y,z,t)=1(x, y, z, t) = 1 (here and on (x,y,z,t)(x, y, z, t) denotes GCD{x,y,z,t}\text{GCD}\{x, y, z, t\}) for even nn. Consider the first series and denote Q=2P2(t)+2P(xy+yz+zx)P2(x+y+z)Q = 2P^2(t) + 2P(xy + yz + zx) - P^2(x + y + z) and P(x)=amxm+am1xm1++a1x+a0P(x) = a_m x^m + a_{m-1} x^{m-1} + \dots + a_1 x + a_0, am0a_m \neq 0. Then polynomial Q1(n)=2P2(5(n2+1))+2P(7n4+16n312n7)P2(8n2+2n+6)Q_1(n) = 2P^2(5(n^2+1)) + 2P(7n^4 + 16n^3 - 12n - 7) - P^2(8n^2 + 2n + 6) equals to zero for every n=2sn = 2s with sNs \in \mathbb{N}, whence Q1(n)=0Q_1(n) = 0. Consider the coefficient at x4mx^{4m} in Q1(x)Q_1(x): on the one hand it is equal to 252mam2+27mam82mam22 \cdot 5^{2m} a_m^2 + 2 \cdot 7^m a_m - 8^{2m} a_m^2 and on the other 00, therefore 252mam2+27mam82mam2=02 \cdot 5^{2m} a_m^2 + 2 \cdot 7^m a_m - 8^{2m} a_m^2 = 0. Analogously we can obtain that Q2(n)=2P2(5(5n2+4n+1))+2P(257n4+T2(n))P2(42n2+34n+8)Q_2(n) = 2P^2(5(5n^2+4n+1)) + 2P(257n^4 + T_2(n)) - P^2(42n^2 + 34n + 8), with deg(T2)3\deg(T_2) \le 3, is equal to zero and thus the coefficient at x4mx^{4m} in this polynomial also equals zero: 2252mam2+2257mam422mam2=02 \cdot 25^{2m} a_m^2 + 2 \cdot 257^m a_m - 42^{2m} a_m^2 = 0. So, am=27m64m225m=2257m422m2252ma_m = \frac{2 \cdot 7^m}{64^m - 2 \cdot 25^m} = \frac{2 \cdot 257^m}{42^{2m} - 2 \cdot 25^{2m}}, whence 27m(422m2252m)=(64m225m)2257m2 \cdot 7^m \cdot (42^{2m} - 2 \cdot 25^{2m}) = (64^m - 2 \cdot 25^m) \cdot 2 \cdot 257^m. For m2m \ge 2 we have: (64m225m)2257m>(64m225m)2252m(64^m - 2 \cdot 25^m) \cdot 2 \cdot 257^m > (64^m - 2 \cdot 25^m) \cdot 2 \cdot 252^m, and so 422m2252m>36m(64m2252m)42^{2m} - 2 \cdot 25^{2m} > 36^m (64^m - 2 \cdot 25^{2m}). 72m>72m2252m36m>64m225m=(72+15)m225m>72m+m72m215225m225m>m49m115249m11549m11549257^{2m} > 7^{2m} - \frac{2 \cdot 25^{2m}}{36^m} > 64^m - 2 \cdot 25^m = (7^2 + 15)^m - 2 \cdot 25^m > 7^{2m} + m \cdot 7^{2m-2} \cdot 15 - 2 \cdot 25^m \Rightarrow 2 \cdot 25^m > m \cdot 49^{m-1} \cdot 15 \ge 2 \cdot 49^{m-1} \cdot 15 \Rightarrow \frac{49^{m-1}}{15} \ge \frac{49}{25}, which is evidently false. Thus, m1m \le 1. For m=1m = 1, we have P(x)=a1x+a0P(x) = a_1 x + a_0 and a1=1a_1 = 1, i.e. P(x)=x+a0P(x) = x + a_0. Substitute P(x)P(x) into the initial equality and take (x,y,z,t)(x, y, z, t) from the first series: (n21,2n,7(n2+1),5(n2+1))(n^2 - 1, 2n, 7(n^2 + 1), 5(n^2 + 1)). After some easy transformations we will have: 4a0(5n2+5)+2a02+2a02a0(8n2+2n+6)a02=04a_0(5n^2 + 5) + 2a_0^2 + 2a_0 - 2a_0(8n^2 + 2n + 6) - a_0^2 = 0, which implies that a0=0a_0 = 0 or that a0a_0 depends on nn, which is impossible. So for m=1m = 1 we have solution P(x)=xP(x) = x. Now if n=0n = 0 then a02+2a0=0a_0^2 + 2a_0 = 0, and we obtain two more answers: P(x)=0P(x) = 0 and P(x)=2P(x) = -2. An easy check shows that these solutions meet all the requirements.

Solution 2

Consider the equation x2+y2+z2=2t2x^2 + y^2 + z^2 = 2t^2 and make use of the method of Diophantine secants: (xt)2+(yt)2+(zt)2=2\left(\frac{x}{t}\right)^2 + \left(\frac{y}{t}\right)^2 + \left(\frac{z}{t}\right)^2 = 2, i.e. α2+β2+γ2=2\alpha^2 + \beta^2 + \gamma^2 = 2, where α,β,γQ+\alpha, \beta, \gamma \in \mathbb{Q}^+. Evidently, α=0\alpha = 0, β=γ=1\beta = \gamma = 1 is a solution of the last equation. Now take β=k1α+1\beta = -k_1\alpha + 1, γ=k2α+1\gamma = -k_2\alpha + 1, with k1,k2Qk_1, k_2 \in \mathbb{Q}. Then α2+(1k1α)2+(1k2α)2=2α(k12α+k22α2k12k2+α)=0\alpha^2 + (1-k_1\alpha)^2 + (1-k_2\alpha)^2 = 2 \Leftrightarrow \alpha(k_1^2\alpha + k_2^2\alpha - 2k_1 - 2k_2 + \alpha) = 0. If α0\alpha \neq 0, we have α=2(k1+k2)1+k12+k22β=1+k22k122k1k21+k12+k22\alpha = \frac{2(k_1+k_2)}{1+k_1^2+k_2^2} \Rightarrow \beta = \frac{1+k_2^2-k_1^2-2k_1k_2}{1+k_1^2+k_2^2}, γ=1+k12k222k1k21+k12+k22\gamma = \frac{1+k_1^2-k_2^2-2k_1k_2}{1+k_1^2+k_2^2}.
Take k1=npk_1 = \frac{n}{p}, k2=mpk_2 = \frac{m}{p}, n,m,pNn, m, p \in \mathbb{N}. Then α=2(m+n)pp2+n2+m2\alpha = \frac{2(m+n)p}{p^2+n^2+m^2}, β=p2+m2n22mnp2+n2+m2\beta = \frac{p^2+m^2-n^2-2mn}{p^2+n^2+m^2}, γ=p2+n2m22mnp2+n2+m2\gamma = \frac{p^2+n^2-m^2-2mn}{p^2+n^2+m^2}. Thus, we see that x=2(m+n)px = 2(m+n)p, y=p2+m2n22mny = p^2+m^2-n^2-2mn, z=p2+n2m22mnz = p^2+n^2-m^2-2mn, t=p2+m2+n2t = p^2+m^2+n^2, with n,m,pNn, m, p \in \mathbb{N} are the solutions of the equation x2+y2+z2=2t2x^2 + y^2 + z^2 = 2t^2.
First consider m=2m = 2, n=1n = 1. Then x1=6px_1 = 6p, y1=p21y_1 = p^2-1, z1=p27z_1 = p^2-7, t1=p2+5t_1 = p^2+5. It is evident that for p6p \neq 6 we have (x1,y1,z1,t1)=1(x_1, y_1, z_1, t_1) = 1. Then substituting these (x1,y1,z1,t1)(x_1, y_1, z_1, t_1) into the problem statement we obtain that Q=2P2(t)+2P(xy+yz+zx)P2(x+y+z)=0Q = 2P^2(t) + 2P(xy + yz + zx) - P^2(x + y + z) = 0, which means that Q1(p)=2P2(p2+5)+2P(p4+12p38p248p+7)P2(2p2+6p8)=0Q_1(p) = 2P^2(p^2+5) + 2P(p^4+12p^3-8p^2-48p+7) - P^2(2p^2+6p-8) = 0. From the problem conditions it follows that all numbers pp of the form p=6kp=6k should be the roots of Q1Q_1, thus Q10Q_1 \equiv 0.
Let P(x)=anxn+an1xn1++a1x+a0P(x) = a_n x^n + a_{n-1} x^{n-1} + \dots + a_1 x + a_0, with an0a_n \neq 0. Then Q1(p)=2(an(p2+5)n++a0)2+2(an(p4+12p38p248p+7)n++a0)2(an(2p2+6p8)n++a0)=(2an2+2an22nan2)p4n+R1(p)Q_1(p) = 2(a_n(p^2+5)^n + \dots + a_0)^2 + 2(a_n(p^4+12p^3-8p^2-48p+7)^n + \dots + a_0)^2 - (a_n(2p^2+6p-8)^n + \dots + a_0) = (2a_n^2 + 2a_n - 2^{2n}a_n^2)p^{4n} + R_1(p), where degR14n1\deg R_1 \le 4n-1. Since Q1(p)0Q_1(p) \equiv 0 and an0a_n \neq 0 we have that
2an+222nan=0.(1) 2a_n + 2 - 2^{2n}a_n = 0. \quad (1)
Consider p=2mp = 2m, n=1n = 1. Then x2=4m(m+1)x_2 = 4m(m+1), y2=5m22m1y_2 = 5m^2 - 2m - 1, z2=3m22m+1z_2 = 3m^2 - 2m + 1, t2=5m2+1t_2 = 5m^2 + 1. For mm large enough we have t2>y2>x2>z2t_2 > y_2 > x_2 > z_2, and so these numbers are distinct. If mm is even, (x2,z2)=(4m(m+1),3m22m+1)=(4(m+1),3m22m+1)=(m+1,3m22m+1)=(m+1,5m1)=(m+1,6)(x_2, z_2) = (4m(m+1), 3m^2 - 2m + 1) = (4(m+1), 3m^2 - 2m + 1) = (m+1, 3m^2 - 2m + 1) = (m+1, 5m-1) = (m+1, 6), thus for m6m \neq 6 (x2,y2,z2,t2)=1(x_2, y_2, z_2, t_2) = 1. If we follow the same lines as for the first series (x,y,z,t)(x, y, z, t), we will obtain that Q2(m)=2P2(5m2+1)+2P(47m4+T(m))P2(12m2)=0Q_2(m) = 2P^2(5m^2+1) + 2P(47m^4 + T(m)) - P^2(12m^2) = 0, where degT3\deg T \le 3. Let us compute the coefficient at m4nm^{4n} in Q2(m)Q_2(m) - it's equal to 2an252n+2an47n122nan2=02a_n^2 \cdot 5^{2n} + 2a_n \cdot 47^n - 12^{2n}a_n^2 = 0. Since an0a_n \ne 0 we have that:
2an52n+247n122nan=0.(2) 2a_n \cdot 5^{2n} + 2 \cdot 47^n - 12^{2n} a_n = 0. \qquad (2)
Now from (1) and (2) we obtain that an=122n1a_n = \frac{1}{2^{2n-1}} and (22n11)247n=122n252n(2^{2n-1}-1) \cdot 2 \cdot 47^n = 12^{2n} - 2 \cdot 5^{2n}, whence 252n+(447)n=144n+247n2 \cdot 5^{2n} + (4 \cdot 47)^n = 144^n + 2 \cdot 47^n. It's easy to see that the last equality is impossible for arbitrary natural n2n \ge 2, since 252n+(447)n>188n=(144+44)n144n+44144n1>144n+247n2 \cdot 5^{2n} + (4 \cdot 47)^n > 188^n = (144 + 44)^n \ge 144^n + 44 \cdot 144^{n-1} > 144^n + 2 \cdot 47^n.
So degP1\deg P \le 1. First assume that degP=1\deg P = 1. Then the equality an=122n1a_n = \frac{1}{2^{2n-1}} implies that a1=1a_1 = 1, i.e. that P(x)=x+a0P(x) = x + a_0. Substitute this polynomial into the initial equality:
2P2(t)+2P(xy+yz+zx)=P2(x+y+z)2(t2+2a0t+a02)+2(xy+yz+zx)+2a0=(x+y+z)2+2(x+y+z)a0+a024a0t+a02+2a0=2(x+y+z)a0. 2P^2(t) + 2P(xy + yz + zx) = P^2(x + y + z) \Rightarrow \\ 2(t^2 + 2a_0t + a_0^2) + 2(xy + yz + zx) + 2a_0 = (x + y + z)^2 + 2(x + y + z)a_0 + a_0^2 \Leftrightarrow \\ 4a_0t + a_0^2 + 2a_0 = 2(x + y + z)a_0.
If a0=0a_0 = 0 then P(x)=xP(x) = x, otherwise 4t+a0+2=2(x+y+z)4t + a_0 + 2 = 2(x + y + z). Substitute here the known series x1=6px_1 = 6p, y1=p21y_1 = p^2 - 1, z1=p27z_1 = p^2 - 7, t1=p2+5t_1 = p^2 + 5 with p=6p = 6. Then 4(p2+5)+a0+2=2(2p2+6p8)4(p^2 + 5) + a_0 + 2 = 2(2p^2 + 6p - 8), and we get that the constant a0a_0 depends on pp, which is impossible.
If degP=0\deg P = 0 then P=a0=constP = a_0 = \text{const}. From the problem conditions we obtain that 2a02+2a0=a02a0=22a_0^2 + 2a_0 = a_0^2 \Rightarrow a_0 = -2 also a0=0a_0 = 0, which means that P(x)=0P(x) = 0 or P(x)=2P(x) = -2. An easy check shows that all three polynomials obtained meet all the requirements.

Looking for a route rather than an archive? The track puts 2,000 problems in a working order, from AMC 10 level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement and solution reproduced as published; topic and difficulty added by this site.