Skip to main content
Log in

Irrationality measures for cubic irrationals whose conjugates lie on a curve

  • Published:
Mathematische Zeitschrift Aims and scope Submit manuscript

Abstract

We describe an unexpected connection between bounded height in families of finitely generated subgroups in tori problems and irrationality. Our method allow us to recover effective irrationality measures for some values of algebraic functions

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Similar content being viewed by others

Notes

  1. A similar requirement was worked by Bombieri and Van der Poorten for a different pourpose.

  2. i. e. better than the bound \({\mu _{\mathrm{eff}}}(\xi )\le 3\) given by Liouville’s Theorem

  3. Equivalently, as suggested by Waldschmidt, we can argue on the non zero resultant \(\mathrm{Res}_x(P,Q)\in {{\mathbb {C}}}[t]\) between \(Q(t,x)=Q_0+Q_1x+Q_2x^2\) and the minimal polynomial P of g over \({{\mathbb {C}}}[t]\).

References

  1. Alladi, K., Robinson, M.L.: Legendre polynomials and irrationality. J. Reine Angew. Math. 318, 137–155 (1980)

    MathSciNet  MATH  Google Scholar 

  2. Amoroso, F., Masser, D., Zannier, U.: Bounded height problems and Silverman specialization theorem. Duke Math J. 166(13), 2599–2642 (2017)

    Article  MathSciNet  Google Scholar 

  3. Baker, A.: Rational approximations to certain algebraic numbers. Proc. Lond. Math. Soc. 3(14), 385–398 (1964)

    Article  MathSciNet  Google Scholar 

  4. Beukers, F., Schlickewei, F.: The equation x + y = 1 in finitely generated groups. Acta Arith. 78(2), 189–199 (1996)

    Article  MathSciNet  Google Scholar 

  5. Beukers, F., Tijdeman, R.: On the multiplicities of binary complex recurrences. Compos. Math. 51(2), 193–213 (1984)

    MathSciNet  MATH  Google Scholar 

  6. Bombieri, E.: On the Thue–Siegel–Dyson theorem. Acta Math. 148, 255–296 (1982)

    Article  MathSciNet  Google Scholar 

  7. Bombieri, E., Cohen, P.: Siegel’s lemma, Padé approximations and Jacobians. Ann. Scuola Norm. Sup. Pise Cl. Sci. (5) 25(1–2), 155–178 (1997)

    MATH  Google Scholar 

  8. Bugeaud, Y.: Effective irrationality measures for real and p-adic roots of rational numbers close to 1, with an application to parametric families of Thue-Mahler equations. Math. Proc. Camb. Philos. Soc. 164(1), 99–108 (2018)

    Article  MathSciNet  Google Scholar 

  9. Chudnovsky, G.V.: The Thue–Siegel–Roth theorem for values of algebraic functions. Proc. Jpn. Acad. Ser. A Math. Sci. 59(6), 281–284 (1983)

    Article  MathSciNet  Google Scholar 

  10. David, S., Hirata-Kohno, N.: Linear forms in elliptic logarithms. J. Reine Angew. Math. 628, 37–89 (2009)

    MathSciNet  MATH  Google Scholar 

  11. Decker, W., Greuel, G.-M., Pfister, G., Schönemann, H.: Singular 4-1-1—A computer algebra system for polynomial computations (2018). http://www.singular.uni-kl.de

  12. Feldman, N.I.: An effective power sharpening of a theorem of Liouville. Izv. Akad. Nauk SSSR Ser. Mat. 35, 973–990 (1971)

    MathSciNet  Google Scholar 

  13. Masser, D.: Auxiliary polynomials in number theory. Camb. Tracts Math. 207, xviii+348 (2016)

    MathSciNet  MATH  Google Scholar 

  14. Zudilin, V.V.: On the irrationality measure of values of G-functions. Izv. Ross. Akad. Nauk Ser. Mat. 60(1), 87–114 (1996). translation in Izv. Math. 60 (1996), no. 1, 91–118

    Article  MathSciNet  Google Scholar 

Download references

Acknowledgements

The authors would like to express their gratitude to Yann Bugeaud, Sinnou David, Michel Waldschmidt and Wadim Zudilin for helpful conversations. We kindly thank Pietro Corvaja for reading a preliminary version of this paper and David Masser for his suggestions on the behavior of \(c({{\mathcal {C}}})\). We also thank the anonymous referee for carefully reading and many helpful suggestions.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to F. Amoroso.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix

Appendix

The aim of this appendix is to provide a complete proof of the bound (5.13) announced in the last section.

Let g be a cubic algebraic function, regular at 0 with rational Taylor coefficients. We choose a determination of the conjugates \(g_1=g\), \(g_2\), \(g_3\) of g over \(\overline{{\mathbb {Q}}}(t)\), which we assume defined (and analytic) in the closed disk \(\vert t\vert \le e^{-c_0}\), for some \(c_0\ge 1\).

Proposition 6.1

There exists \(C=C(g)\ge c_0\ge 1\) such that the following holds. Let \({\varepsilon }_1\), \({\varepsilon }_2\), \({\varepsilon }_3\), \({\varepsilon }_4\in (0,1)\) be such that \({\varepsilon }_2+{\varepsilon }_3+{\varepsilon }_4<1/2\). Let \(a\in {{\mathbb {Z}}}\), \(b\in {{\mathbb {N}}}\) coprime, with

$$\begin{aligned} \log b\ge {\left\{ \begin{array}{ll} \displaystyle \log \vert a\vert +\frac{C}{{\varepsilon }_2{\varepsilon }_3},&{}(a)\\ \displaystyle \frac{\log \vert a\vert +C/{\varepsilon }_3}{1/2-{\varepsilon }_2-{\varepsilon }_3-{\varepsilon }_4},&{}(b)\\ \displaystyle \frac{{\varepsilon }_2C^2}{{\varepsilon }_1^2{\varepsilon }_4^4}.&{}(c) \end{array}\right. } \end{aligned}$$
(6.1)

Let also

$$\begin{aligned} \kappa =\frac{(1-{\varepsilon }_3)\log (b/\vert a\vert )}{(1+{\varepsilon }_2)\log b +C/{\varepsilon }_3}. \end{aligned}$$
(6.2)

Then g(a/b) is irrational with effective irrationality measure

$$\begin{aligned} \le \frac{\kappa }{\kappa -1/2}+{\varepsilon }_1 \le 2+{\varepsilon }_1+\frac{1}{{\varepsilon }_4}\left( {\varepsilon }_2+{\varepsilon }_3+\frac{\log \vert a\vert +C/{\varepsilon }_3}{\log b}\right) . \end{aligned}$$

Proof

We first make some elementary remarks on the parameters. By definition (6.2) of \(\kappa \) we have

$$\begin{aligned} \begin{aligned} 1-\kappa&=\frac{({\varepsilon }_2+{\varepsilon }_3)\log b+(1-{\varepsilon }_3)\log \vert a\vert +C/{\varepsilon }_3}{(1+{\varepsilon }_2)\log b +C/{\varepsilon }_3}\\&\le {\varepsilon }_2+{\varepsilon }_3+\frac{\log \vert a\vert +C/{\varepsilon }_3}{\log b}. \end{aligned} \end{aligned}$$
(6.3)

Thus \(1-\kappa >0\) and \(1-\kappa \le \frac{1}{2}-{\varepsilon }_4\) by (6.1b), i. e.

$$\begin{aligned} 1/2+{\varepsilon }_4\le \kappa <1 \end{aligned}$$
(6.4)

Let now \(t_0=a/b\). Then, by (6.1a), \(\log b\ge \log \vert a\vert + C\) and

$$\begin{aligned} \vert t_0\vert \le \frac{1}{2}e^{-c_0}, \end{aligned}$$
(6.5)

assuming \(C\ge c_0+\log 2\). Remark also that \(H(t_0)=b\). We denote by \(c_1\), ..., \(c_9\) positive constants depending only on g. By the functorial properties of the height,

$$\begin{aligned} H(\xi )\le e^{c_1}b^d \end{aligned}$$
(6.6)

where \(\xi =g(t_0)\) and where where \(d\ge 1\) is the geometric degree of the algebraic function g (i. e. the degree of its polar divisor).

Let N be a positive integer and \(\delta \in [1/(N+1),3)\). We consider, in the terminology of [7], a \((N,\delta )\)-Padé approximant at 0 for \((1,g,g^2)\), i. e. a non-zero vector of polynomials \((Q_0,Q_1,Q_2)\in {{\mathbb {Q}}}[t]\) of degree at most N such that

$$\begin{aligned} f=Q_0+Q_1g+Q_2g^2 \end{aligned}$$

vanishes at 0 with multiplicity \(M\ge [(3-\delta )(N+1)]\). By Theorem 2 of [7] we can find a such vector of polynomials satisfying moreover

$$\begin{aligned} \max _jh(Q_j)\le \frac{c_2(3-\delta )^2(N+1)}{\delta }. \end{aligned}$$

We choose

$$\begin{aligned} N:=[6d/{\varepsilon }_2]+1\quad {{\mathrm{and}}}\quad \delta :=\frac{3({\varepsilon }_3 N+1)-1}{N+1}. \end{aligned}$$

Since \({\varepsilon }_3\le 1\) we have \(\delta \in [1/(N+1),3)\) as needed. Moreover

$$\begin{aligned} M\ge (3-\delta )(N+1)-1=3(1-{\varepsilon }_3)N \end{aligned}$$
(6.7)

and

$$\begin{aligned} \delta \ge \frac{2({\varepsilon }_3 N+1)-1}{N+1}=\frac{2{\varepsilon }_3 N+1}{N+1}\ge {\varepsilon }_3. \end{aligned}$$

Thus

$$\begin{aligned} \max _jh(Q_j)\le \frac{c_3N}{{\varepsilon }_3}. \end{aligned}$$
(6.8)

Remark that this implies

$$\begin{aligned} \vert f\vert _R\le e^{c_4N/{\varepsilon }_3}\;\hbox {on the disk } \vert t\vert \le R:=e^{-c_0}\le 1. \end{aligned}$$
(6.9)

Let \(\theta :=f(t_0)\). We want first show that \(\theta \) is non-zero. Let us assume the contrary. Let \(f(t)=\sum _{k=0}^{\infty } f_kt^k\) and \(g(t)^j=\sum _{k=0}^\infty a_{j,k}t^k\) (\(j=1\), 2) be the expansions of f and of g, \(g^2\) around \(t=0\). Let also \(a_{0,k}=1\) if \(k=0\) and \(a_{0,k}=0\) otherwise. Writing \(Q_j(t)=\sum _{l=0}^N q_{j,l}t^l\) for \(j=0,1,2\) we have

$$\begin{aligned} f_M=\sum _{j=0}^2\sum _{l=0}^N q_{jl}a_{j,M-l}\ne 0 \end{aligned}$$

since f vanishes at 0 with multiplicity exactly M. Let v be a place of \({{\mathbb {Q}}}\). The Local Eisenstein Theorem [7, Corollary p.161] gives \(\vert a_{j,k}\vert _v\le c'(v)R_v^{2k-1}\) for some \(c'(v)\), \(R_v\) depending on v and on g, with moreover \(c'(v)=1\) if v is finite. Thus, by (6.8)

$$\begin{aligned} h(f_M)\le c_5M+\max _jh(Q_j)\le c_5M+c_3N/{\varepsilon }_3 \end{aligned}$$

and, by Liouville’s inequality,

$$\begin{aligned} \log \vert f_M\vert \ge -c_5M-c_3N/{\varepsilon }_3. \end{aligned}$$

On the other hand, \(f_Mt_0^M=-\sum _{j>M}f_jt_0^j\) since we are assuming \(f(t_0)=0\). Let as before \(R:=e^{-c_0}\le 1\). Using Cauchy’s estimates \(\vert f_j\vert \le \vert f\vert _R R^{-j}\), we get

$$\begin{aligned} \vert f_M\vert&\le \vert t_0\vert ^{-M}\sum _{j=M+1}^\infty \vert f\vert _R (\vert t_0\vert /R)^j =\vert t_0\vert ^{-M}\vert f\vert _R \frac{(\vert t_0\vert /R)^{M+1}}{1-\vert t_0\vert /R}\\&\le 2(1/R)^{M+1}e^{c_4N/{\varepsilon }_3}\vert t_0\vert \end{aligned}$$

by (6.9) and since \(\vert t_0\vert \le \vert R\vert /2\) by (6.5). Comparing the lower and the upper bound for \(\vert f_M\vert \) we find

$$\begin{aligned} \log (1/\vert t_0\vert )\le \log (2/R)+(\log (1/R)+c_5)M+(c_3+c_4)\frac{N}{{\varepsilon }_3}. \end{aligned}$$

We still need an (elementary) zero’s estimate to bound M. Since the \(Q_j\)’s are polynomials of degree \(\le N\), the polar divisor of f has degree \(\le c_6N\). ThusFootnote 3\(M\le c_6N\). Inserting this bound in the last displayed formula we get

$$\begin{aligned} \log (1/\vert t_0\vert )\le \log (2/R)+(\log (1/R)+c_5)c_6N +(c_3+c_4)\frac{N}{{\varepsilon }_3}<\frac{c_7}{{\varepsilon }_2{\varepsilon }_3} \end{aligned}$$

since \(N\le 6d/{\varepsilon }_2+1\) by our choice. This contradicts (6.1a) provided that \(C\ge c_7\). The proof of \(\theta \ne 0\) is concluded.

We now prove an upper bound for \({H_0}(\theta )\) and for \(\Vert \theta \Vert \). Inequality (6.8) implies

$$\begin{aligned} H(Q_0(t_0):Q_1(t_0):Q_2(t_0))\le e^{c_8N/{\varepsilon }_3} b^N. \end{aligned}$$
(6.10)

Using Schwarz lemma in the disk \(\vert t\vert \le R:=e^{-c_0}\le 1\) and inequalities (6.9), (6.7), we get

$$\begin{aligned} \vert \theta \vert \le \vert f\vert _R(\vert t_0\vert /R)^M \le e^{c_4N/{\varepsilon }_3}(\vert t_0\vert /R)^{3(1-{\varepsilon }_3)N}. \end{aligned}$$
(6.11)

By (3.3) of Lemma 3.4, by (6.6), (6.10), (6.11), and since \(N\ge 6d{\varepsilon }_2\), we have

$$\begin{aligned} {H_0}(\theta )\le 3H(\xi )^6H(Q_0(t_0):Q_1(t_0):Q_2(t_0))\le e^{c_9N/{\varepsilon }_3}b^{(1+{\varepsilon }_2)N}, \end{aligned}$$
(6.12)

and

$$\begin{aligned} \begin{aligned} \Vert \theta \Vert&\le H(\xi )^6H(Q_0(t_0):Q_1(t_0):Q_2(t_0))\vert \theta \vert \\&\le e^{c_9N/{\varepsilon }_3}b^{(1+{\varepsilon }_2)N}\vert t_0\vert ^{3(1-{\varepsilon }_3)N}. \end{aligned} \end{aligned}$$
(6.13)

We are quite in position to apply Theorem 5.2. Assuming \(C\ge c_9\) and taking into account (6.13), (6.2), (6.4), we have

$$\begin{aligned} -\frac{1}{N}\log \Vert \theta \Vert&\ge 3(1-{\varepsilon }_3)\log (b/\vert a\vert )-(1+{\varepsilon }_2)\log b-C/{\varepsilon }_3\\&=((1+{\varepsilon }_2)\log b+C/{\varepsilon }_3)(3\kappa -1)>0. \end{aligned}$$

Liouville’s inequality (Lemma 3.3) implies that \(\theta \) is irrational and thus \(\xi \) is irrational as well, since \(\theta \in {{\mathbb {Q}}}(\xi )\). If \(\xi \) is a quadratic irrational our result is trivial, since its effective irrationality measure is 2 which is \(\le \frac{\kappa }{\kappa -1/2}\) by (6.4). Thus we may assume that \(\xi \) is a cubic irrational. Let \({{\mathbb {K}}}={{\mathbb {Q}}}(\xi )\), denote by \(\sigma _1=\mathrm{Id},\sigma _2,\sigma _3\) the immersions \({{\mathbb {K}}}\hookrightarrow {{\mathbb {C}}}\) and put \(\theta _j=f_j(t_0)\).

We apply Theorem 5.2 with \({{\mathcal {C}}}\subseteq {{\mathbb {P}}}_2\) be the projective curve parametrized by \((f_1:f_2:f_3)\), and \({\varepsilon }={\varepsilon }_1\). Let \(c=c({{\mathcal {C}}})\) be the constant appearing in that theorem. Set \(C:=\max (c_0+\log 2,c_7,c_9,4c)\),

$$\begin{aligned} \hbar :=((1+{\varepsilon }_2)\log b+C/{\varepsilon }_3)N,\qquad \lambda :=3\kappa -1. \end{aligned}$$

Inequality (6.4) implies

$$\begin{aligned} \frac{(\lambda -1/2)^2}{8(\lambda +1)}= & {} \frac{3(\kappa -1/2)^2}{8\kappa }\\\ge & {} \frac{{\varepsilon }_4^2}{4}. \end{aligned}$$

Thus, by (6.1c),

$$\begin{aligned} c^2\min \Big (1,\frac{(\lambda -1/2)^2}{8(\lambda +1)}{\varepsilon }_1\Big )^{-2} \le \frac{C^2}{{\varepsilon }_1^2{\varepsilon }_4^4}\le \frac{\log b}{{\varepsilon }_2}\le N\log b\le \hbar . \end{aligned}$$

Assumption (5.2) is satisfied. Moreover, by (6.12) we have \(h_0(\theta )\le \hbar \), and, by (6.13) and (6.2),

$$\begin{aligned} -\frac{\log \Vert \theta \Vert }{\hbar } \ge \frac{3(1-{\varepsilon }_3)\log (b/\vert a\vert )-(1+{\varepsilon }_2)\log b-C/{\varepsilon }_3}{(1+{\varepsilon }_2)\log b+C/{\varepsilon }_3}=3\kappa -1=\lambda . \end{aligned}$$

Thus (5.3) is also satisfied. By Theorem 5.2, \(\xi \) has irrationality measure

$$\begin{aligned} \le \frac{\lambda +1}{\lambda -1/2}+{\varepsilon }_1 =\frac{\kappa }{\kappa -1/2}+{\varepsilon }_1. \end{aligned}$$

We finally remark that, by (6.4) and (6.3),

$$\begin{aligned} \frac{\kappa }{\kappa -1/2}+{\varepsilon }_1 =2+\frac{1-\kappa }{\kappa -1/2}+{\varepsilon }_1 \le 2+{\varepsilon }_1+{\varepsilon }_4^{-1}\left( {\varepsilon }_2+{\varepsilon }_3+\frac{\log \vert a\vert +C/{\varepsilon }_3}{\log b}\right) . \end{aligned}$$

\(\square \)

From proposition (6.1), we easily deduce the bound (5.13) for the irrationality measure of the values of a cubic algebraic function.

Corollary 6.2

Let g be a cubic algebraic function, regular at 0 with rational Taylor coefficients. Then there exists \(C=C(g)\ge 1\) such that the following holds. Let \(a\in {{\mathbb {Z}}}\), \(b\in {{\mathbb {N}}}\) coprime and \({\varepsilon }\in (0,4/7)\) such that

$$\begin{aligned} \vert a\vert \le b^{1/2-{\varepsilon }} \end{aligned}$$
(6.14)

and

$$\begin{aligned} \log b\ge 2^{10}C^3/{\varepsilon }^4. \end{aligned}$$
(6.15)

Then g(a/b) is irrational with effective irrationality measure

$$\begin{aligned} \le 2+\frac{1}{{\varepsilon }}\left( 9\sqrt{\frac{C}{\log b}}+2\frac{\log \vert a\vert }{\log b}\right) . \end{aligned}$$

Proof

Let \(C=C(g)\) be the constant appearing in proposition (6.1). We let for short

$$\begin{aligned} {\varepsilon }'=\sqrt{\frac{C}{\log b}}. \end{aligned}$$

By (6.15) we have

$$\begin{aligned} {\varepsilon }'\le \sqrt{\frac{C}{2^{10}C^3/{\varepsilon }^4}}=\frac{{\varepsilon }^2}{32C}\le \frac{{\varepsilon }}{8}. \end{aligned}$$
(6.16)

We apply the proposition 6.1 choosing

$$\begin{aligned} {\varepsilon }_1={\varepsilon }'/{\varepsilon },\quad {\varepsilon }_2=2{\varepsilon }',\quad {\varepsilon }_3={\varepsilon }',\quad {\varepsilon }_4={\varepsilon }/2 \end{aligned}$$

which is an admissible choice since \({\varepsilon }_1\le 1/8<1\) by (6.16) and

$$\begin{aligned} {\varepsilon }_2+{\varepsilon }_3+{\varepsilon }_4=3{\varepsilon }'+{\varepsilon }/2\le 7{\varepsilon }'/8<1/2 \end{aligned}$$

again by (6.16) and by the assumption \({\varepsilon }<4/7\).

By the choice of \({\varepsilon }'\) and by (6.14) we have

$$\begin{aligned} \log \vert a\vert +\frac{C}{{\varepsilon }_2{\varepsilon }_3}=\log \vert a\vert +\frac{C}{2{\varepsilon }'^2}=\log \vert a\vert +\frac{1}{2}\log b\le \log b. \end{aligned}$$

Moreover, by (6.14) and since \(1/2-{\varepsilon }+{\varepsilon }'\le 1/2-3{\varepsilon }'-{\varepsilon }/2\) by (6.16),

$$\begin{aligned} \frac{1}{\log b}\cdot \frac{\log \vert a\vert +C/{\varepsilon }_3}{1/2-{\varepsilon }_2-{\varepsilon }_3-{\varepsilon }_4} =\frac{\frac{\log \vert a\vert }{\log b}+{\varepsilon }'}{1/2-3{\varepsilon }'-{\varepsilon }/2} \le \frac{1/2-{\varepsilon }+{\varepsilon }'}{1/2-3{\varepsilon }'-{\varepsilon }/2}\le 1. \end{aligned}$$

Finally, by the choice of \({\varepsilon }'\) and by (6.15),

$$\begin{aligned} \frac{1}{\log b}\cdot \frac{{\varepsilon }_2C^2}{{\varepsilon }_1^2{\varepsilon }_4^4} =\frac{1}{\log b}\cdot \frac{2{\varepsilon }'C^2}{({\varepsilon }'/{\varepsilon })^2({\varepsilon }/2)^4} =\frac{32C^2}{{\varepsilon }'{\varepsilon }^2\log b} =\frac{32C^{3/2}}{{\varepsilon }^2\sqrt{\log b}} \le 1. \end{aligned}$$

The last three displayed lines show that (6.1) is satisfied. Proposition 6.1 asserts that g(a/b) is irrational with effective irrationality measure

$$\begin{aligned}&\le 2+{\varepsilon }_1+\frac{1}{{\varepsilon }_4}\left( {\varepsilon }_2+{\varepsilon }_3+\frac{\log \vert a\vert +C/{\varepsilon }_3}{\log b}\right) \\&= 2+\frac{{\varepsilon }'}{{\varepsilon }}+\frac{2}{{\varepsilon }}\left( 4{\varepsilon }'+\frac{\log \vert a\vert }{\log b}\right) \\&= 2+9\frac{{\varepsilon }'}{{\varepsilon }}+\frac{2}{{\varepsilon }}\frac{\log \vert a\vert }{\log b}\\&= 2+\frac{1}{{\varepsilon }}\left( 9\sqrt{\frac{C}{\log b}}+2\frac{\log \vert a\vert }{\log b}\right) . \end{aligned}$$

\(\square \)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Amoroso, F., Zannier, U. Irrationality measures for cubic irrationals whose conjugates lie on a curve. Math. Z. 299, 1767–1788 (2021). https://doi.org/10.1007/s00209-021-02732-8

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00209-021-02732-8

Navigation