Ivan Shishkin, Rye (1878)

Problems/Sequence and seriesUnreviewed

No linear recurrence for the primes

by Sequoia·translated by visitor·
51
Difficulty scaleÉchelle de difficulté

This score reflects both the level of the required concepts and the difficulty of the solution.Ce score tient compte à la fois du niveau des notions nécessaires et de la difficulté de la résolution.

  1. 110First steps / middle schoolPremiers pas / collège
  2. 1125Beginner / high schoolDébutant / lycée
  3. 2650Intermediate / undergraduateIntermédiaire / licence
  4. 5170Advanced / graduateAvancé / master
  5. 7190Expert / specializedExpert / spécialisé
  6. 91100Research levelNiveau recherche
These levels are approximate guides.Ces niveaux sont des repères approximatifs.
·
English
EnglishFrançais
Unreviewed. This problem has not been reviewed by trusted users yet.

Let (pn)n\left(p_n\right)_n denote the sequence of primes, with p0=2,p1=3p_0=2, p_1=3, and so on. Can the sequence of primes be written as a linear recurrence with constant coefficients?

In other words, do there exist an integer dNd \in \mathbb{N}^* and real numbers α0,,αd1\alpha_0, \ldots, \alpha_{d-1} such that
nN,pn+d=αd1pn+d1++α1pn+1+α0pn?\forall n \in \mathbb{N}, \quad p_{n+d}=\alpha_{d-1} p_{n+d-1}+\cdots+\alpha_1 p_{n+1}+\alpha_0 p_n ?

I solved itMark it doneAdd to my listKeep it in your list

Hints

1

Hint 1

Open this only if you want a small nudge before looking at the solutions.

Available in Français; this hint has not been translated into the current language yet.

Solutions

2
Reveal solutionsAre you sure? Give it a try first.

Solution by SequoiaFR

Discussions0 useful votes

Si cette récurrence existe, alors on peut la résoudre et obtenir l’existence de polynômes complexes Q1,,QrQ_1,\dots, Q_r ainsi que de complexes α1,,αr\alpha_1,\dots,\alpha_r tels que nN,pn=i=1rQi(n)αin\forall n\in\N, p_n=\sum_{i=1}^r Q_i(n) \alpha_i^n. Ce résultat est issu du théorème sur les suites récurrentes à coefficients constants.

Ainsi, pour n+n\to+\infty, on a que pnp_n serait équivalent à quelque chose de la forme cnajJαjnc\,n^a \sum_{j\in J}\alpha_j^ncc est une constante complexe, JJ l’ensemble des αi\alpha_i de module maximal et aa le degré maximal des polynômes Qj,jJQ_j, j\in J.

Mais on sait pourtant que pnn+nln(n)p_n\underset{n\to+\infty}{\sim} n\ln(n) par le théorème des nombres premiers. Ce qui contredit ce qu’on vient de dire par croissance comparée (le log est négligeable que toutes les puissances de nn ou exponentielles).

Solution by visitor

Discussions0 useful votes

Following Phil Caldero

Theorem. Let K\mathbb K be a field of characteristic zero. The sequence (pn)n1(p_n)_{n\ge 1} of prime numbers satisfies no relation
a0pn+d=a1pn+d1++adpn(nn0)a_0\,p_{n+d}=a_1\,p_{n+d-1}+\dots+a_d\,p_n\qquad (n\ge n_0)with d1d\ge 1, a0,,adKa_0,\dots,a_d\in\mathbb K and a00a_0\neq 0.

Lemma 1 (descent). If a sequence with values in a subfield FKF\subset\mathbb K satisfies such a relation with coefficients in K\mathbb K, it satisfies one of the same order with coefficients in FF.

Proof. Put rn=(pn+d1,,pn)Fdr_n=(p_{n+d-1},\dots,p_n)\in F^d and bn=pn+db_n=p_{n+d}. By assumption there is aKda\in\mathbb K^d with rna=bnr_n\cdot a=b_n for nn0n\ge n_0. Choose rn1,,rnkr_{n_1},\dots,r_{n_k} a basis of the subspace of FdF^d spanned by the rnr_n. The finite system rnjX=bnjr_{n_j}\cdot X=b_{n_j} has coefficients in FF and a solution in Kd\mathbb K^d; since rank is unchanged by field extension, it has a solution aFda'\in F^d. If rn=jcjrnjr_n=\sum_j c_j r_{n_j} with cjFc_j\in F, then rna=jcjbnj=jcj(rnja)=rna=bnr_n\cdot a'=\sum_j c_j b_{n_j}=\sum_j c_j\,(r_{n_j}\cdot a)=r_n\cdot a=b_n. \square

Lemma 2 (periodicity modulo qq). Let (un)(u_n) be a sequence of integers and a0,,adZa_0,\dots,a_d\in\mathbb Z with a0ad0a_0a_d\ne 0 such that a0un+d=i=1daiun+dia_0u_{n+d}=\sum_{i=1}^d a_iu_{n+d-i} for nn0n\ge n_0. For every prime qq not dividing a0ada_0a_d, there is T1T\ge 1 such that un+Tun(modq)u_{n+T}\equiv u_n \pmod q for all nn0n\ge n_0.

Proof. Modulo qq, a0a_0 is invertible and the vector vn=(un,,un+d1)Fqdv_n=(u_n,\dots,u_{n+d-1})\in\mathbb F_q^d satisfies vn+1=Avnv_{n+1}=Av_n, where AA is the companion matrix with entries aˉi=aia01\bar a_i=a_ia_0^{-1}. Since detA=±aˉd0\det A=\pm\bar a_d\neq 0, AA lies in the finite group GLd(Fq)\mathrm{GL}_d(\mathbb F_q), so AT=IA^T=I for T=GLd(Fq)T=|\mathrm{GL}_d(\mathbb F_q)|, and vn+T=vnv_{n+T}=v_n for nn0n\ge n_0. \square

Proof of the theorem. Suppose not, and take a relation of minimal order d1d\ge 1, valid for nn0n\ge n_0. Since QK\mathbb Q\subset\mathbb K and pnQp_n\in\mathbb Q, Lemma 1 lets us assume the aia_i are rational, then integers after clearing denominators. We have ad0a_d\neq 0: otherwise the relation, shifted by one index, would have order d1d-1. The integer a0ada_0a_d has only finitely many prime divisors, so there is kn0k\ge n_0 such that q=pkq=p_k does not divide a0ada_0a_d. By Lemma 2, pk+Tpk=q0(modq)p_{k+T}\equiv p_k=q\equiv 0\pmod q. Thus qq divides the prime pk+Tp_{k+T}, hence pk+T=q=pkp_{k+T}=q=p_k, contradicting the strict monotonicity of (pn)(p_n). \square

Remarks. The proof uses only two properties of (pn)(p_n): every term is prime, and terms with distinct indices are distinct. It therefore applies to any injective sequence of primes. The characteristic-zero assumption is necessary: over F2\mathbb F_2, the sequence (0,1,1,1,)(0,1,1,1,\dots) satisfies un+2=un+1u_{n+2}=u_{n+1}.

Report

For an unclear, ambiguous, or possibly incorrect statement, please use the Discussion tab on the right. Report content that needs moderator intervention, such as dangerous, clearly non-mathematical, or plagiarized content.