Polynomial Continued Fractions for Algebraic Numbers
Henri Cohen · 2025
Read Polynomial Continued Fractions for Algebraic Numbers by Henri Cohen (2025) 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-2502-19575.txt.
Introduction
Henri Cohen proves this 2025 result in number theory about continued fractions, the nested divisions used to express numbers. For square roots, the pattern is a repeating cycle. Cohen shows that any real cubic algebraic number, a root of a degree-three equation, has a continued fraction whose terms follow polynomial rules. The short paper also notes some wider versions of the claim.
Catalog record
| CREATOR | Henri Cohen |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 19 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-19575
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Polynomial Continued Fractions for Algebraic Numbers Henri Cohen Abstract Our main result is that any real cubic algebraic number has a continued fraction expansion with polynomial coefficients. Some generalizations are mentioned. 1 Introduction Let \alpha be a real algebraic number. If \alpha is rational, it has a finite simple continued fraction expansion, and if \alpha is quadratic it has a simple continued fraction expansion which is ultimately periodic. The problem of studying the simple continued fraction expansion of higher degree algebraic numbers, in particular cubic ones, has attracted a lot of attention, and one can reasonably say that nothing is known. For instance one does not know if the coefficients can be bounded, or if they are always unbounded, etc… Our goal is not to attack this notoriously difficult problem, but to give a positive answer for certain algebraic numbers (in particular cubic ones) if one generalizes the type of continued fraction that one considers. Definition 1.1 1. A generalized continued fraction (abbreviated CF) is an expression of the type a(0)+b(0)/(a(1)+b(1)/(a(2)+b(2)/(a(3)+\cdots)))\;, where a(n) and b(n) are rational numbers, b(n)\neq 0 for all n , and a(n)\neq 0 except for a finite number of n . 2. Such a CF is said to be of polynomial type if a(n) and b(n) are polynomials for n sufficiently large. Remarks. 1. A simple CF corresponds by definition to b(n)=1 for all n , a(0)\in\mathbb{Z} , and a(n)\in\mathbb{Z}_{\geq 1} for n\geq 1 . 2. By simplifying, it is clear that we can suppress all a(n)=0 and assume that a(n)\neq 0 for all n\geq 1 . Notation: If a(n) is a polynomial A(n) in n for n>n_{a} and b(n) is a polynomial B(n) in n for n>n_{b} , we will denote the above CF of polynomial type by the very practical notation ((a(0),a(1),\dotsc,a(n_{a}),A(n)),(b(0),b(1),\dotsc,b(n_{b}),B(n))) Our main result is the following: Theorem 1.2 Any real cubic algebraic number has a continued fraction expansion of polynomial type. Note that it has been known probably since the 19th century that some cubic irrationalities have such an expansion, for instance the smallest positive root of x^{3}-x+t=0 when t is sufficiently small, but the point of the theorem is that it is true for all cubic irrationalities. The proof given here is quite simple, but involves a few subtleties. 2 Preliminary Lemmas Before giving the proof of this theorem, we need a series of lemmas. Definition 2.1 We will say that a real number u is of continued fraction type if it is equal to the limit of a convergent continued fraction ((a(n)),(b(n))) of polynomial type in the above sense, and we denote by {\cal C}\subset\mathbb{R} the set of such constants. Lemma 2.2 If \left(\begin{smallmatrix}{a}&{b}\\ {c}&{d}\end{smallmatrix}\right)\in\GL_{2}(\mathbb{Q}) and u\in{\cal C} we have (au+b)/(cu+d)\in{\cal C} . Proof. Immediate: if u=a(0)+b(0)/(a(1)+b(1)/(a(2)+\cdots)) , then if c\neq 0 we have \displaystyle\dfrac{au+b}{cu+d} \displaystyle=\dfrac{a}{c}+\dfrac{-(ad-bc)/c^{2}}{u+d/c} \displaystyle=(a/c)+(-(ad-bc)/c^{2})/(a(0)+d/c+b(0)/(a(1)+b(1)/(a(2)+\cdots)))\;, so is of the form a^{\prime}(0)+b^{\prime}(0)/(a^{\prime}(1)+\cdots) with a^{\prime}(0)=a/c , a^{\prime}(1)=a(0)+d/c , a^{\prime}(n)=a(n-1) for n\geq 2 , b^{\prime}(0)=-(ad-bc/c^{2}) , and b^{\prime}(n)=b(n-1) for n\geq 1 . If c=0 the result is even more trivial since (au+b)/(cu+d)=(a/d)u+(b/d) . \sqcap \sqcup Lemma 2.3 Let K=\mathbb{Q}(u) be a cubic number field. Any element v\in K\setminus\mathbb{Q} is of the form v=(au+b)/(cu+d) for some \left(\begin{smallmatrix}{a}&{b}\\ {c}&{d}\end{smallmatrix}\right)\in\GL_{2}(\mathbb{Q}) . Proof. The four elements \{uv,v,u,1\} belong to K which is a \mathbb{Q} -vector space of dimension 3 , so are linearly dependent, and since v\notin\mathbb{Q} the corresponding 2\times 2 matrix is invertible, proving the lemma. \sqcap \sqcup It follows from these two lemmas that to prove our theorem it is sufficient to prove that at least one non-rational element of a cubic number field is in {\cal C} . We now set the following definition: Definition 2.4 Let P be an irreducible polynomial of degree exactly equal to 3 , and write P(x)=ax^{3}+bx^{2}+cx+d . We define the ratio of P and denote by r(P) the quantity r(P)=\dfrac{4(3ac-b^{2})^{3}}{(27a^{2}d-9abc+2b^{3})^{2}} By extension, if u is an algebraic number of degree 3 we will denote by r(u) the ratio of the characteristic polynomial of u . For instance, if P(x)=x^{3}+ax+b then we simply have r(P)=4a^{3}/(27b^{2}) , and in particular if P(x)=x^{3}-cx+c we have r(P)=-4c/27 . It is immediate to check that if A and B are rational with A\neq 0 we have r(Au+B)=r(u) . Lemma 2.5 If \disc(P)>0 we have r(P)<-1 , and if \disc(P)<0 we have r(P)>-1 . Proof. Indeed, we immediately check that r(P)+1=-\dfrac{27a^{2}\disc(P)}{(27a^{2}d-9abc+2b^{3})^{2}}\;, and the result follows. \sqcap \sqcup Lemma 2.6 If K is a cubic number field then |r(u)| is unbounded for u\in K . Proof. Let K=\mathbb{Q}(\alpha) with \alpha a root of a cubic polynomial which we may assume to be x^{3}+ax+b , and set u=\alpha^{2}+D\alpha+2a/3 . The characteristic polynomial of u is of the form x^{3}+ex+f (i.e., again with no term in x^{2} ), with e=aD^{2}+3bD-a^{2}/3 and f=bD^{3}-(2a^{2}/3)D^{2}-abD-(b^{2}+2a^{3}/27) , so r(u)=4e^{3}/(27f^{2}) . We compute that the resultant of these two polynomials in D is equal to -\disc(\alpha)^{3}/729 , hence is nonzero, so it follows that if we choose a rational value for D close to a real root of the cubic polynomial f , |f| will be as small as we want while |e| will be bounded from below, so r(u) is unbounded as desired. \sqcap \sqcup Corollary 2.7 Let K be a cubic number field. There exists a defining equation for K of the form x^{3}-cx+c=0 (so that c>27/4 if K is totally real and c<27/4 otherwise), and |c| can be chosen as large as one wants. Proof. Clear: let u be chosen so that |r(u)| is large. If the characteristic polynomial of u is x^{3}+a_{2}x^{2}+a_{1}x+a_{0} , one checks that v=((9a_{2}^{2}-27a_{1})/(27a_{0}+2a_{2}^{3}-9a_{1}a_{2}))(u+a_{2}/3) has a characteristic polynomial of the form x^{3}-cx+c , so |c|=(27/4)|r(v)|=(27/4)|r(u)| is as large as we like. \sqcap \sqcup Lemma 2.8 if c>27/4 , the equation x^{3}-cx+c=0 has three real roots \beta_{1} , \beta_{2} , \beta_{3} which are such that \beta_{1}<-3<1<\beta_{2}<3/2<\beta_{3} . Proof. Trivial and left to the reader. \sqcap \sqcup Lemma 2.9 Let c>27/4 , c\neq 27/2 , and set \phi_{c}(x)=3c(x-3)/((2c-27)x) . Then the \phi_{c}(\beta_{i}) are the three roots of x^{3}-Cx+C with C=27c^{2}/(2c-27)^{2} , and if we denote by \gamma_{1}<-3<1<\gamma_{2}<3/2<\gamma_{3} these three roots, we have: • If c<27/2 : \phi_{c}(\beta_{1})=\gamma_{1} , \phi_{c}(\beta_{2})=\gamma_{3} , and \phi_{c}(\beta_{3})=\gamma_{2} . • If c>27/2 : \phi_{c}(\beta_{1})=\gamma_{3} , \phi_{c}(\beta_{2})=\gamma_{1} , and \phi_{c}(\beta_{3})=\gamma_{2} . Proof. This is a simple direct verification and left to the reader. Note that since in our application P will be irreducible, we cannot have c=27/2 since in that case x^{3}-cx+c is reducible. \sqcap \sqcup The main tool that we are going to use to prove our main theorem is the following crucial lemma, which is a direct consequence of the Lagrange inversion formula: Lemma 2.10 1. Assume that |c|>27/4 . The series S(c)=\sum_{n\geq 0}\dfrac{1}{2n+1}\binom{3n}{n}c^{-n} converges to a root \beta=S(c) of x^{3}-cx+c=0 . 2. More precisely, if c<-27/4 , \beta is its unique real root, while if c>27/4 , \beta=\beta_{2} is its unique root such that 1<\beta<3/2 . 3. More generally, let P=x^{3}+ax+b . If |r(P)|=|4a^{3}/(27b^{2})|>1 , one of the roots \alpha of P=0 is given by \alpha=(-b/a)\sum_{n\geq 0}\dfrac{1}{2n+1}\binom{3n}{n}(-b^{2}/a^{3})^{n}\;, and \alpha is the unique root of P if 4a^{3}/(27b^{2})>1 , and otherwise \alpha is the unique root of P such that |b/a|<|\alpha|<(3/2)|b/a| . Proof. (1). This is well-known and is a simple application of Lagrange’s inversion formula. Note that the simplicitly