Greenberg’s conjecture for real quadratic number fields

Pietro Mercuri, Maurizio Paoluzi, René Schoof · 2025

Read Greenberg’s conjecture for real quadratic number fields by Pietro Mercuri, Maurizio Paoluzi, René Schoof (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-2503-00819.txt.

Introduction

Pietro Mercuri, Maurizio Paoluzi, and René Schoof present a 2025 paper testing Greenberg's conjecture for real quadratic number fields. The conjecture concerns towers of number fields built by adjoining roots of unity step by step. At each level, a group called the 3-class group measures arithmetic complexity. Greenberg predicted that for real quadratic fields this group stays bounded as the tower climbs, rather than growing forever.

The team computed these groups for all real quadratic fields with discriminant under one hundred thousand. In every case the group's order stayed bounded as the level rose. That uniform agreement supports the conjecture across a wide sample. The work is computational evidence, not a proof, but it covers far more fields than hand checks could. It strengthens confidence in a long-standing guess about these towers.

Catalog record

CREATORPietro Mercuri, Maurizio Paoluzi, René Schoof
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-00819

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Greenberg’s conjecture for real quadratic number fields Pietro Mercuri, Maurizio Paoluzi, René Schoof Abstract . We compute the 3 3 -class groups A n A_{n} of the fields F n F_{n} in the cyclotomic 𝐙 3 \mathbf{Z}_{3} -extensions of the real quadratic fields of discriminant f < 100,000 f<100,000 . In all cases the orders of A n A_{n} remain bounded as n n goes to infinity. This is in agreement with Greenberg’s conjecture. 2020 Mathematics Subject Classification 11R23,11R11,11R27,11R29 1. Introduction Let F be a totally real number field and let p be a prime. Let F=F_{0}\,\,\subset\,\,F_{1}\,\,\subset\,\,F_{2}\,\,\subset\,\,\ldots denote the cyclotomic \mathbf{Z}_{p} -extension of F . The p -class group A_{n} is the p -part of the ideal class group of the ring of integers of F_{n} . In his 1971 thesis Ralph Greenberg conjectured that \#A_{n} remains bounded as n\rightarrow\infty . See [ 5 , 6 ] and [ 7 , Conjecture (3.4)] . This is the so-called “ \lambda=0 ”-conjecture of Iwasawa theory. In this note we report on a computation for the prime p=3 involving the 30394 real quadratic fields \mathbf{Q}(\sqrt{f}) of discriminant f<100,000 . See [ 3 , 8 , 9 ] for earlier computations. As a consequence we obtain the following result. Theorem 1.1 . Greenberg’s conjecture is true for p=3 and the real quadratic fields of discriminant f<100,000 . For each of the real quadratic fields with discriminant f in the range of our compuation we have computed a certain Galois module C(f) , the finiteness of which is equivalent to Greenberg’s conjecture. In this introduction we define the module C(f) . In the subsequent sections we explain our computation and its results. See [ 9 ] for the algebraic properties of C(f) in the case f\not\equiv 1\pmod{3} . The slightly different case f\equiv 1\pmod{3} is discussed in detail in [ 11 ] and in Section 4 of this paper. Let F=\mathbf{Q}(\sqrt{f}) be a real quadratic field of discriminant f . The Galois module C(f) is defined in terms of cyclotomic units as follows. For k\geq 1 let \zeta_{k} denote a primitive k -th root of unity. For n\geq 0 the n -th layer in the cyclotomic \mathbf{Z}_{3} -extension of F is F_{n}=\mathbf{Q}(\sqrt{f},\zeta_{3^{n+1}}+\zeta_{3^{n+1}}^{-1}). The field F_{n} is a subfield of the cyclotomic field \mathbf{Q}(\zeta_{3^{n+1}f}) . It is a cyclic degree 3^{n} extension of F_{0}=\mathbf{Q}(\sqrt{f}) . Its ring of integers O_{n} contains cyclotomic units. See [ 12 , Section 4] . The 3 -part of the quotient of the unit group O_{n}^{*} by the subgroup generated by the cyclotomic units is a finite group denoted by B_{n} . It is known that the groups A_{n} and B_{n} have the same cardinality [ 12 , Theorems 4.1 and 5.3] . Therefore Greenberg’s conjecture is true for the field F if and only if \#B_{n} remains bounded as n\rightarrow\infty . When the discriminant f is not congruent to 1\pmod{3} , we let C_{n} denote the dual of the group B_{n} for n\geq 0 . When f\equiv 1\pmod{3} , we let C_{n} denote the dual of the group \tilde{B}_{n} . Here \tilde{B}_{n} sits in an exact sequence of the form 0\longrightarrow\tilde{B}_{n}\longrightarrow B_{n}\mathop{\longrightarrow}\limits^{\phi_{n}}\mathbf{Z}_{3}/\log_{3}\eta_{0}\mathbf{Z}_{3}. where for \epsilon\in O_{n}^{*} we put \phi_{n}(\epsilon)={1\over{3^{n}}}\log_{3}(N_{n}(\epsilon)) . Here N_{n}\colon F_{n}^{*}\rightarrow\mathbf{Q}(\sqrt{f})^{*} is the norm map. Since the 3 -adic logarithm of a generator \eta_{0} of the group of cyclotomic units in \mathbf{Q}(\sqrt{f}) is not zero, the rightmost group is a finite cyclic group. It follows that [B_{n}:\tilde{B}_{n}] and hence the quotient \#B_{n}/\#C_{n} is bounded independently of n . Therefore Greenberg’s conjecture is true if and only if \#C_{n} remains bounded as n\rightarrow\infty . By [ 9 , Lemma 2.1] and Section 4 , the natural maps B_{m}\rightarrow B_{n} are injective and the natural maps C_{n}\rightarrow C_{m} are surjective for n\geq m . Let C(f) denote the projective limit of the C_{n} . Then C(f) is a Galois module and hence in the usual way a module over the Iwasawa algebra \Lambda\cong\mathbf{Z}_{3}[[T]] . It follows from the structure of the cyclotomic units that C(f) is a cyclic \Lambda -module. See [ 9 , Theorem 2.4] and Section 4 . In other words, we have C(f)\,=\,\mathop{\rm lim}\limits_{\leftarrow}C_{n}\,\,\cong\Lambda/J,\qquad\text{for some ideal $J\subset\Lambda$.} The vanishing of the Iwasawa \mu -invariant of \mathbf{Q}(\sqrt{f}) means that J contains a monic polynomial and hence that C(f) is a finitely generated \mathbf{Z}_{3} -module. See [ 2 ] . Greenberg’s conjecture affirms that C(f) is actually finite , so that C_{n}=C(f) for sufficiently large n . We have computed the Galois modules C(f) for f<100,000 . It took about two weeks on a workstation with Intel processor i5. We found the following, which is equivalent to Theorem 1.1 . Theorem 1.2 . For p=3 and for all discriminants f<100,000 the module C(f) is finite. In most cases the module C(f) is actually zero. Indeed, for only 3359 out of the 30394 real quadratic fields considered, C(f) is not zero and, equivalently, J is a proper \Lambda -ideal. This is about 11 % of all cases. Of these, 2118 have J equal to the maximal ideal (3,T) of \Lambda . In these cases C(f) has order 3 . For the remaining 1241 discriminants the module C(f) is strictly larger. This is approximately 4 % of all cases. Rather than listing each ideal J , we indicate in Sections 3 and 5 how often ideals of a certain type appear in our computation. The full list of ideals may be of interest in itself and is available on GitHub [ 15 ] . In Sections 3 and 5 we also single out some discriminants for which the ideal J has a remarkable shape. 2. The case f\nequiv 1\pmod{3} In this section we give a brief description of the algorithm in the case where the discriminant f is congruent to 0 or 2 modulo 3 . This case is discussed in detail in [ 9 ] . Let F=\mathbf{Q}(\sqrt{f}) be a real quadratic field of discriminant f . Put f^{\prime}=f/3 when f\equiv 0\pmod{3} and f^{\prime}=f when f\equiv 2\pmod{3} . For n\geq 0 the n -th layer F_{n} in the cyclotomic tower of F is a subfield of \mathbf{Q}(\zeta_{3^{n+1}f^{\prime}}) . The cyclotomic unit 1-\zeta_{3^{n+1}f^{\prime}} is contained in \mathbf{Q}(\zeta_{3^{n+1}f^{\prime}}) . Put \eta_{n}=\mathop{\rm Norm}\limits_{\mathbf{Q}(\zeta_{3^{n+1}f^{\prime}})/F_{n}}(1-\zeta_{3^{n+1}f^{\prime}})^{\sigma-1}. Here \sigma is the non-trivial automorphism in {\rm Gal}(F_{n}/\mathbf{Q}_{n})\cong{\rm Gal}(\mathbf{Q}(\sqrt{f})/\mathbf{Q}) . In [ 9 ] it is explained that the Galois module generated by \eta_{n} is free of rank 1 over \mathbf{Z}[G_{n}] . Here G_{n} denotes {\rm Gal}(F_{n}/F_{0}) . This implies that the Galois module C(f) described in the introduction is a cyclic module over the Iwasawa algebra \Lambda=\mathop{\rm lim}\limits_{\leftarrow}\mathbf{Z}_{3}[G_{n}]\cong\mathbf{Z}_{3}[[T]] . So we have C(f)=\Lambda/J for some \Lambda -ideal J . For n\geq 0 we put \omega_{n}(T)=(1+T)^{p^{n}}-1 and we write (\omega_{n}) for the \Lambda -ideal generated by it. In [ 9 ] it is explained that in this case we have C_{n}=C(f)/\omega_{n}C(f)=\Lambda/(J+(\omega_{n})),\qquad\text{ for all\penalty\ $n\geq 0$.} The Galois module C(f) is finite if and only if \omega_{n}C(f)=0 and hence C(f)=C_{n} for some n\geq 0 . By Nakayama’s lemma this happens if and only if J+(\omega_{n})=J+(\omega_{n+1}) for some n\geq 0 . This observation leads to the following algorithm. For n=0,1,2,\ldots , we compute the shrinking ideals J+(\omega_{n}) until we find that J+(\omega_{n})=J+(\omega_{n+1}) . Our method for computing the ideals J+(\omega_{n})

Read the complete text