On the polynomial equation P ( Q ( x 1 , … , x m ) ) = Q ( P ( x 1 ) , … , P ( x m ) ) P(Q(x_{1},...,x_{m}))=Q(P(x_{1}),...,P(x_{m}))
Arseny Mingajev · 2024
Read On the polynomial equation P ( Q ( x 1 , … , x m ) ) = Q ( P ( x 1 ) , … , P ( x m ) ) P(Q(x_{1},...,x_{m}))=Q(P(x_{1}),...,P(x_{m})) by Arseny Mingajev (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-10465.txt.
Introduction
Arseny Mingajev solves a polynomial nesting puzzle in a 2024 paper. The equation compares two ways of nesting polynomials. In one order, several inputs are combined with Q and then P is applied. In the other, P acts on each input first, and Q combines the results.
Over the complex numbers, Mingajev proves that once P has degree above one, solutions are scarce. The only working pairs are ones that become simple monomials after a straight-line change of variables, a so-called affine conjugation. The result pins down a rigid corner of polynomial dynamics where two operations commute in this nested sense.
Catalog record
| CREATOR | Arseny Mingajev |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 17 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-10465
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
On the polynomial equation P ( Q ( x 1 , … , x m ) ) = Q ( P ( x 1 ) , … , P ( x m ) ) P(Q(x_{1},...,x_{m}))=Q(P(x_{1}),...,P(x_{m})) Arseny Mingajev Abstract . We consider the equation P ( Q ( x 1 , … , x ν ) ) = Q ( P ( x 1 ) , … , P ( x ν ) ) P(Q(x_{1},...,x_{\nu}))=Q(P(x_{1}),...,P(x_{\nu})) in polynomials Q ∈ ℂ [ x 1 , … , x ν ] Q\in\mathbb{C}[x_{1},\dotsc,x_{\nu}] , P ∈ ℂ [ x ] P\in\mathbb{C}[x] , and prove that if deg ( P ) > 1 \deg(P)>1 , then it is only solvable in polynomials that are affinely conjugate to monomials. 1. Introduction In 1922 and 1923, Gaston Julia [ 3 ] and Joseph Ritt [ 4 ] proved that, if two polynomials in one variable P(x) and Q(x) commute with each other, then one of three statements is true: (i) (Ritt, [ 4 ] ) P and Q are both affinely conjugate to monomials; (ii) (Ritt, [ 4 ] ) P and Q are both affinely conjugate to Chebyshev polynomials; (iii) (Julia, [ 3 ] ) P^{\circ\nu}=Q^{\circ\mu} for some positive integers \nu,\mu. An elementary proof of this theorem was published by Block and Thielman in a 1951 paper, [ 1 ] . The results of Ritt and Julia can be interpreted as results about commuting rational maps \mathbb{CP}^{1}\to\mathbb{CP}^{1} . The commutativity of rational maps \mathbb{CP}^{N}\to\mathbb{CP}^{N} is still a problem that is being investigated; see [ 2 ] and the references therein. In this paper, we consider another multivariable version of this problem. For each natural number \nu>1 we denote by \mathscr{S}_{\nu} the subset of \mathbb{C}[x]\times\mathbb{C}[x_{1},\dotsc,x_{\nu}] consisting of polynomials P\in\mathbb{C}[x] and Q\in\mathbb{C}[x_{1},\dotsc,x_{\nu}] , \deg P,\deg Q,\geq 1 , satisfying the functional equation P\big(\,Q(x_{1},\dotsc,x_{\nu})\,\big)=Q\big(\,P(x_{1}),\dotsc,P(x_{\nu})\,\big) (1.1) The main goal of this paper is to describe the structure of \mathscr{S}_{\nu} for \nu\geq 2 . A few elementary observations are in order. Denote by \Aff the group of affine bijections \mathbb{C}\ni z\mapsto az+b\in\mathbb{C},\;\;a,b\in\mathbb{C},\;\;a\neq 0. For any natural number n , the group \Aff acts on \mathbb{C}[z_{1},\dotsc,z_{n}] by conjugation, \Aff\times\mathbb{C}[z_{1},\dotsc,z_{\nu}]\ni(\alpha,R)\mapsto C_{\alpha}(R)\in\mathbb{C}[z_{1},\dotsc,z_{\nu}], \;C_{\alpha}(R)(z_{1},\dotsc,z_{\nu}]=\alpha\circ R\big(\,\alpha^{-1}(z_{1}),\dotsc\alpha^{-1}(z_{\nu})\,\big). An elementary computation shows that \big(\,P,Q\,\big)\in\mathscr{S}_{\nu}\iff\big(\,C_{\alpha}(P),C_{\alpha}(Q)\,\big)\in\mathscr{S}_{\nu},\;\;\forall\alpha\in\Aff. In other words, \mathscr{S}_{\nu} is invariant with respect to the action of \Aff . Lemma 1.1 . Any polynomial P(x)\in\mathbb{C}[x] , \deg P>0 is affinely conjugate to a polynomial vanishing at 0 . Moreover, if \deg P>1 , then P is affinely conjugate to a polynomial with leading coefficient 1 and vanishing at 0 . Proof. Let \lambda\in\Aff , \lambda(z)=z-r , where r is a root of the equation p(x)-z=0 . Then \lambda^{-1}(0)=r and C_{\lambda}(P)(0)=P\big(\,\lambda^{-1}(0)\,\big)-r=P(r)-r=0. This proves the first part. Suppose now that d=\deg P>1 and P(0)=0 . If \alpha(z)=cz , c\neq 0 , then C_{\alpha}(P)(z)=cP(z/c) . If p_{d} is the leading coefficient of P , then the leading coefficient of C_{\alpha}(P) is c^{1-d}p_{d} . Since d>1 we can find c such that c^{1-d}p_{d}=1 . \hbox to0.0pt{$\sqcap$\hss}\sqcup For any multi-index \alpha\in\mathbb{Z}_{\geq 0}^{\nu} we set |\alpha|=\alpha_{1}+\cdots+\alpha_{\nu},\;\;x^{\alpha}:=x_{1}^{\alpha_{1}}\cdots x_{\nu}^{\alpha_{\nu}}. We can now state our main result. Theorem 1.2 . Let P\in\mathbb{C}[x] , Q\in\mathbb{C}[x_{1},\dotsc,x_{\nu}] , \nu\geq 2 , \deg P=n>1 . If (P,Q)\in\mathscr{S}_{\nu} and \forall i\in\{1,\dotsc,\nu\},\;\;Q\not\in\mathbb{C}[z_{i}], (1.2) then there exists an affine transformation \lambda\in\Aff , a constant c\in\mathbb{C} , and a nonzero multi-index \alpha\in\mathbb{Z}_{\geq 0}^{\nu} such that C_{\lambda}(P)(x)=x^{n},\;\;C_{\lambda}(Q)(x_{1},\dotsc,x_{\nu})=cx^{\alpha},\;\;c^{n-1}=1. Remark 1.3 . (a) Observe that if (P,Q)\in\mathscr{S}_{\nu} are nontrivial monomials Q=cx^{\alpha} and P(x)=x^{n} , then the condition c^{n-1}=1 is automatically satisfied. (b) The condition ( 1.2 ) is necessary. Indeed , if Q depended only on one variable, then according to [ 4 ] the equation ( 1.1 ) has nonmonomial solutions. \hbox to0.0pt{$\sqcap$\hss}\sqcup The proof will be by induction. 2. The two-variable case Let P(x)=\sum_{k=0}^{n}a_{k}x^{k}. We can view Q(x,y) either as a polynomial in \mathbb{C}[y][x] or as a polynomial in \mathbb{C}[x][y] . In the first case we can express Q as Q(x,y)=\sum_{i=0}^{m}Q_{i}(y)x^{i}. and in the second case we can express Q as Q(x,y)=\sum_{i=0}^{\ell}R_{i}(x)y^{i}. Lemma 2.1 . Under the conditions of the main theorem 1.2 , if (P,Q)\in\mathscr{S}_{2} , then the leading coefficients Q_{m}(y) and R_{\ell}(x) are non-constant. Proof. We argue by contradiction. In this case Q(x,y) can be expressed as Q(x,y)=f(x)+g(y)+xyq(x,y), where the highest power of x in q(x,y) is less than \deg(f)-1 , and the highest power of y in q(x,y) is less than \deg(g)-1 . Moreover \min(\deg(f),\deg(g))>1 . The commutation equation ( 1.1 ) can be rewritten as P(f(x)+g(y)+xyq(x,y))=f(P(x))+g(P(y))+P(x)P(y)q(P(x),P(y)). By the statement of the theorem 1.2 , either g(y)\neq 0 or q(x,y)\neq 0 . Suppose first that q\neq 0 . If \deg(f)=N , the term with the highest power of x in xyq(x,y) is of the form q_{1}(y)x^{N-a},a>0 . The highest power of x with a non-constant coefficient in P\big(\,f(x)+g(y)+xyq(x,y)\,\big) is N(n-1)+N-a=Nn-a , n=\deg P . However, only P(x)P(y)q(P(x),P(y))\in\mathbb{C}[y][x] has coefficients that are nonconstant as functions of y . The highest p such that the coefficient C_{p}[y] of x^{p} is non-constant coefficients n(N-a)=nN-na . Since n>1 , this is impossible. If q(x,y)=0 , then ( 1.1 ) reads P(f(x)+g(y))=f(P(x))+g(P(y)). This is not solvable because \deg P>1 . Indeed P(f(x)+g(y)) contains terms divisible by xy , and f(P(x))+g(P(y)) does not. An analogous proof works for powers of y . \hbox to0.0pt{$\sqcap$\hss}\sqcup The equality P(Q(x,y))=Q(P(x),P(y)) can be rewritten as \sum_{k=0}^{n}a_{k}\Big(\,\sum_{i=0}^{m}Q_{i}(y)x^{i}\,\Big)^{k}=\sum_{i=0}^{m}Q_{i}(P(y))P(x)^{i}. Equating the terms with the highest power of x we deduce Q_{m}(y)^{n}=a_{n}^{m-1}Q_{m}(P(y)). (2.1) Denote by \mathscr{R} the set roots of Q_{m}(y) . For r\in\mathscr{R} we denote by m(r) the multiplicity of r . Then Q_{m}(y)=q_{m}\prod_{r\in\mathscr{R}}(y-r)^{m(r)},\;\;q_{m}\neq 0, and ( 2.1 ) implies that q_{m}^{n}\prod_{r\in\mathscr{R}}(y-r)^{nm(r)}=a_{n}^{m-1}\underbrace{\prod_{r\in\mathscr{R}}\big(\,P(y)-r\,\big)^{m(r)}}_{R(y)}. (2.2) Lemma 2.2 . The polynomial Q_{m} has a single root of multiplicity m , i.e., |\mathscr{R}|=1 . Moreover, if r_{0} is this unique root of Q_{m} , then P(y)=(y-r_{0})^{n}+r_{0} . Proof. The equality ( 2.2 ) shows that there exists a map \phi:\mathscr{R}\to\mathscr{R} such that P(r)=\phi(r),\;\;\forall r\in\mathscr{R}. (2.3) On the other hand, for any r\in\mathscr{R} , the equality ( 2.2 ) implies that the nonempty set P^{-1}(r) is contained in \mathscr{R} . Indeed, if P(y_{0})=r , then R(y_{0})=0 so Q_{m}(y_{0})=0 and thus y_{0}\in\mathscr{R} . Hence \phi is surjective. Since \mathscr{R} is finite, \phi is also bijective. We claim that for any r\in\mathscr{R} we have P^{-1}(r)=\{\phi^{-1}(r)\} . Set r^{\prime}=\phi^{-1}(r) . The equality ( 2.3 ) shows that \phi^{-1}(r)\in P^{-1}(r) . Conversely, for any r^{\prime\prime}\in P^{-1}(r) we have P(r^{\prime\prime})=r=\phi(r^{\prime\prime}) and since \phi is bijective we deduce r^{\prime\prime}=\phi^{-1}(r)=r^{\prime} . We conclude that for any r\in\mathscr{R} we have P(y)=\big(\,y-\phi^{-1}(r)\,\big)^{n}+r. (2.4) Let r_{1},r_{2}\in\mathscr{R} . Set s_{j}:=\phi^{-1}(r_{j}) , j=1,2 . Then (y-s_{1})^{n}+r_{1}=P(y)=(y-s_{2})^{n}+r_{2},\;\;\forall y\in\mathbb{C}. The Newton binomial formula implies that y^{n}-ns_{1}y^{n-1}+\cdots+(-1)^{n}s_{1}^{n}+r_{1}=y^{n}-ns_{2}y^{n-1}+\cdots+(-1)^{n}s_{2}^{n}+r_{2}. We have n-1\geq 1 . Equating the coefficients of y^{n-1} in the above equality we deduce s_{1}=s_{2} . Since \phi is bijective we conclude that r_{1}=r_{2} so that \mathscr{R} consists of a single root r_{0} . The equality ( 2.4 ) shows that P(y)=(y-r_{0})^{n}+r_{0} . \hbox to0.0pt{$\sqcap$\hss}\sqcup The above discussion shows that if (P,Q)\in\mathscr{S}_{2} , and \deg P>1 then P is affinely conjugate to the monomial x^{n} . The equation ( 1.1 ) becomes Q(x^{n},y^{n})=Q(x,y)^{n}. (2.5) To tackle the two-variable case, we must first learn how to solve the much simpler one-variable case Q(x)^{n}=c^{n}\prod_{k=1}^{p}(x-x_{k})^{n}=Q(x^{n})=c\prod_{k=1}^{p}(x^{n}-x_{k}). For n>1 , the polynomials \prod_{k=1}^{p}(x-x_{k})^{n},\prod_{k=1}^{p}(x^{n}-x_{k}) must have the same roots. Each non-zero