Optimal bounds for B\"uchi's problem in modular arithmetic II
Pablo S\'aez, Xavier Vidaux, Maxim Vsemirnov

TL;DR
This paper establishes upper bounds on the length of sequences of squares generated by quadratic polynomials over modular rings, extending B"uchi's problem to higher powers of primes.
Contribution
It provides the first bounds for B"uchi's problem in modular arithmetic for prime powers, generalizing previous results for prime moduli.
Findings
Existence of a uniform bound M for sequences of squares from quadratic polynomials over Z/p^sZ.
Reduction to polynomials with invertible dominant coefficients simplifies the analysis.
The bounds depend only on p, s, and the polynomial's properties.
Abstract
Given a prime and an integer , we show that there exists an integer such that for any quadratic polynomial with coefficients in the ring of integers modulo , such that is not a square, if a sequence is a sequence of squares, then is at most . We obtain this result by reducing to the case where has an invertible dominant coefficient.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAlgebraic Geometry and Number Theory · Analytic Number Theory Research · Coding theory and cryptography
Büchi’s problem in modular arithmetic for arbitrary quadratic polynomials
Pablo Sáez, Xavier Vidaux and Maxim Vsemirnov
Abstract
Given a prime and an integer , we show that there exists an integer such that for any quadratic polynomial with coefficients in the ring of integers modulo , such that is not a square, if a sequence is a sequence of squares, then is at most . We also provide some explicit formulas for the optimal .00footnotetext: The three authors have been partially supported by the first author Fondecyt research projects 1130134 and 1170315, Chile. The third author is partially supported by the Government of the Russian Federation (grant 14.Z50.31.0030).
Keywords: Büchi sequence, Hensley sequence
MSC 2010 Classification: 11B50, 11B83
1 Introduction
We are interested in the following question:
Question 1.1**.**
Given an integer and a quadratic polynomial over , consider the sequence
[TABLE]
modulo . How long can this sequence be if every element of it is a square modulo but itself is not a square modulo ?
The same question can be considered for any commutative ring with unit instead of a quotient of . For , it was first asked by R. Büchi in the early seventies, and was motivated by a decision problem in Logic — see [Lip90] and [Maz94].
In the case of modular arithmetic, Question 1.1 was first addressed by D. Hensley in [He, Thm. 3.1]. He showed that in the particular case where is an odd prime number and is of the form , then is strictly less than — nevertheless he does not give any explicit formula for the largest possible as and vary. In [SVV15], we deal with the case where, for an odd given , is invertible modulo — see Theorem 2.3 below for the case of prime powers, and [SVV15, Section 5] for general . In the present paper, we solve the problem for any given prime power and any , by reducing it to the case where is invertible — see Theorem 2.2 below.
To give a taste of our main result without introducing too many technicalities, here we state a corollary.
Theorem 1.2**.**
Let be a prime . Assume that is a non-zero square modulo , and are positive even integers, and . As varies within the set of non-square quadratic polynomials with this restriction on , the largest possible such that each of , …, is a square modulo , is
[TABLE]
(i.e.: there are sequences of this length, and no larger ones).
Let us give a concrete example. For any odd prime , modulo , the polynomial is not the square of a polynomial, because is not a square, but by Hensel’s lemma, it is easy to see that is a square modulo for , so the length is reached. Note that and are not squares modulo .
Though many analogous results exist in the literature over different type of rings , they always assume that the polynomial is monic, with just one exception: Natalia Garcia-Fritz [Ga17, Thm. 1.6, Cor. 1.7 and the comments that follow] does not put restrictions on (unconditionally if is a function field of a curve over , and assuming the Bombieri-Lang conjecture when ). There is some literature on sequences of squares whose second difference is an arbitrary element of , which corresponds essentially to considering a quadratic with an arbitrary dominant coefficient. Symmetric sequences of that kind were considered by Allison [All86], Bremner [Bre03], Browkin and Brzeziński [BB06], and Gonzalez-Jimenez and Xarles [GoX11].
Analogues of Question 1.1 have been considered for most classical rings (but in the case of number fields, under some well-known conjectures, like Bombieri-Lang for surfaces, or some version of ABC). Relevant results in positive characteristic can be found in [Pa11] (the analogue of Büchi’s problem for any power over fields with a prime number of elements), and in [PaW15] (over rings of functions), who generalize previous results in [PhV06, PhV10, ShV10, AW11, AHW13]. For a general survey on Büchi’s problem and its extensions to other structures and higher powers, see [PaPhV10].
2 Preliminaries and Main result
If is an integer, will denote its residue class modulo (we may use the brackets notation for polynomials and for sequences as well), and if is a prime, will stand for the usual order at of , with the convention , so that for every integer we have, if and only if .
From now on, we will only consider sequences over which satisfy the two following conditions for some odd integer :
- (C1)
, …, are squares modulo . 2. (C2)
is not the square of a polynomial modulo .
Following [SVV15], sequences satisfying (C1) are called -Büchi sequences modulo , and they are called non-trivial if (C2) is also satisfied (we may say just “Büchi” instead of “-Büchi” when it is clear what is). We should immediately point out that in [SVV15] we consider -Büchi sequences as trivial when is the square of a polynomial of degree at most . But indeed, when is invertible this makes no difference with condition (C2), as shown by the following proposition, which will be proved at the beginning of the next section.
Proposition 2.1**.**
Let be an odd prime and be a positive integer. Let . The following statements are equivalent:
The polynomial is the square of a polynomial modulo . 2. 2.
Either and is a square, or is the square of a polynomial modulo whose degree is at most one.
For odd , let us write for
[TABLE]
(with the convention if all the sequences in the set are trivial), and
[TABLE]
Also, we will write
[TABLE]
Note that when is prime, we trivially have
[TABLE]
Here “” stands for “maximal length” and “” stands for “optimal bound”. The reason to use both concepts is that the proofs are done in terms of maximal lengths but the formulas that we need from [SVV15] are nicer in terms of the optimal bound (the reader will see the point in Theorem 2.3 below).
We can now state our main theorem.
Theorem 2.2**.**
Let be a prime and be a positive integer. Let , with unless . Write and let be such that when , and otherwise. Assume . We have
[TABLE]
In this paper, we deal only with prime power modulus. The case of a general modulus can be reduced to the case of powers of primes following the strategy described in [SVV15, Section 5.1]. Any Büchi sequence modulo can be glued from Büchi sequences modulo the using the Chinese Remainder Theorem. The only subtle point to take care of is that one has to check that the resulting sequence modulo is non trivial, so there are various cases to consider, which result in an elementary but cumbersome analysis — we leave the details to the reader.
While Theorem 2.2 is a natural extension of what we did in our previous work, it was not clear from the beginning what the right statement should be (for example, we were surprised when we discovered that the concept of triviality had to be kept unchanged). We have tried to write the proof in the most uniform possible way, instead of doing the obvious case by case analysis.
In order to have a global picture of the situation, and for later references, we resume in a single theorem what we knew in the case where is invertible. In [SVV15], given not divisible by , we had defined as
[TABLE]
and . Note that when is invertible modulo an odd , then every polynomial can be written in a unique way in the form modulo , so the notation in [SVV15] is compatible with the present one.
Theorem 2.3**.**
([SVV15, Thm. 1.7, Thm. 1.8, Lem. 2.13])* Let be an odd prime number, and and be integers. Assume that is invertible.*
If is a non-square and , then , where is any quadratic non-residue modulo . 2. 2.
If is a non-zero square and is even, then . 3. 3.
If is a non-zero square and is odd, then . 4. 4.
We have . 5. 5.
For any , we have . 6. 6.
For any and , we have .
We get Theorem 1.2 by first applying Theorem 2.2 and then Theorem 2.3, items 2 and 4. For other cases, it is clear how similar corollaries can be obtained.
3 Reduction to the case when is invertible or is
We will frequently use the following well known fact.
Lemma 3.1**.**
Let be an odd prime number. If is a non-zero square modulo for some , then is a square modulo for any .
Next, we prove Proposition 2.1.
Proof of Proposition 2.1.
We first prove that 2 implies 1. Assume
[TABLE]
and is a square. If is identically [math] modulo , then the claim is trivial, so we may assume that is not [math] modulo . We have
[TABLE]
for some , not divisible by , and such that . The Taylor series modulo of the square root of is actually a polynomial, since denominators are powers of and numerators have increasing order at .
We now prove that 1 implies 2. Assume (indeed, for , the claim is trivial as is an integral domain). Let be such that . We can assume . Let be the largest integer such that and , so that . We have . If , there is nothing to prove, so we can assume
[TABLE]
hence . Let be such that and . So has degree at most , and since , we deduce that has degree at most (because we are now over the integral domain ). So we have
[TABLE]
for some and some . If is the zero polynomial, then we are done. Otherwise choose as large as possible, so that has at least one coefficient not divisible by , namely, . We have then
[TABLE]
Case 1: Assume that divides , so that does not divide . We have then for some , hence for some , so
[TABLE]
Case 2: Assume that does not divide . Write , so that
[TABLE]
hence
[TABLE]
If , then we are done (since has degree ), so it remains to consider the case where
[TABLE]
which will turn out to be impossible. Multiplying both sides of Equation (1) by
[TABLE]
we obtain:
[TABLE]
since . Hence we have
[TABLE]
since . We now compare the coefficients of on both sides, where is the degree of . Let be the coefficient of at (so is the dominant coefficient of ), and let be the constant term of . The coefficient of modulo on the left hand side is the coefficient of
[TABLE]
modulo (the terms that are not written in Equation (2) will have order at least , and in the term all monomials have degree at most ), which is
[TABLE]
(here the term comes from Equation (1)). On the other hand, the right hand side of Equation (2) gives
[TABLE]
so we have
[TABLE]
which is a contradiction since divides neither nor . ∎
Note that if any of or is invertible modulo , then the condition
[TABLE]
is never satisfied. We now prove a sequence of lemmas that will imply our main theorem.
Lemma 3.2**.**
Let be an odd prime and . Let . Write for each . Assume that and is a non-square. If is a Büchi sequence modulo , then .
Proof.
Note that . Also note that (because is a non-square, so in particular it is not ). Write , where , so that is invertible modulo . We assume and will get a contradiction. Let be such that . From the hypothesis of the lemma, we have
[TABLE]
hence, recalling that , is a square modulo . Since is non-zero modulo , also is non-zero modulo . So is a square modulo by Lemma 3.1, which contradicts our hypothesis on . ∎
Lemma 3.3**.**
Let be an odd prime and . Let . If is invertible modulo and is not invertible modulo , then
[TABLE]
Proof.
We first prove the “” inequality. Let and be integers. Write and assume that is a non-trivial Büchi sequence modulo . Modulo , since is not invertible, we have . Write . Since is a square modulo for each , it is a square modulo , so is a Büchi sequence modulo . Since is invertible, is a non-trivial Büchi sequence modulo , hence is at most .
We now prove the other inequality. Let be such that is a Büchi sequence of length (note that this is always finite by item 4 of Theorem 2.3). Consider
[TABLE]
If in the sequence there is no , then is a non-zero square modulo for any , hence it is a square modulo by Lemma 3.1. Assume that there is some such that is congruent to [math] modulo , so that is congruent to modulo (there can be at most one such ). In that case, consider instead
[TABLE]
so that is actually , hence a square modulo , and as before, when , is a non-zero square modulo . In both cases, is a Büchi sequence modulo .
We now prove that is a non-trivial Büchi sequence modulo . It is enough to prove that is not a square modulo . Indeed, we have , and the latter is not a square by definition of , so is not even a square modulo . ∎
Next comes the key lemma for having a uniform proof of Theorem 2.2.
Lemma 3.4**.**
Let be an odd prime and . Let , with unless . Assume that not both and are [math]. Write and . For , let be such that (if , take ). Write .
If is even, then we have:
[TABLE]
(where reads as [math] if ). 2. 2.
The sides of the equation in item 1 are infinite if and only if , and . 3. 3.
If is odd, then we have:
[TABLE] 4. 4.
If and is odd, then .
Proof.
We first prove items 3 and 4, together with the “” inequality in item 1. Let and be integers, and write and (with if ). Assume that is a non-trivial Büchi sequence modulo , where . In particular, by Proposition 2.1 the polynomial is not the square of a linear polynomial modulo , and we have , unless is a non-square. If and is a non-square, we have by Lemma 3.2, so we may assume . Write , where
[TABLE]
We can now complete the proof of item 3. Assume is odd. In that case, if is a square, then . If , we have
[TABLE]
hence can be [math] modulo for at most one value of , hence . If , we have
[TABLE]
so can be [math] modulo for at most two values of , hence .
We now turn to item 1. Assume is even. Since is not the square of a linear polynomial modulo , also is not the square of a linear polynomial modulo . Moreover, since , we have
[TABLE]
hence is a non-trivial Büchi sequence modulo by Proposition 2.1, so we have
[TABLE]
We now prove “” in item 1 (so in particular, we assume that is even). First note that the claim is trivial when (which is the case in particular when ). Let
[TABLE]
be such that is a non-trivial Büchi sequence of length ( may be ).
Consider
[TABLE]
For any , since is a square modulo and is even, also is a square modulo , so is a Büchi sequence modulo .
We now prove that is a non-trivial Büchi sequence modulo . Since is a non-trivial Büchi sequence modulo , is not the square of a linear polynomial modulo , hence also, since is even, is not the square of a linear polynomial modulo . Moreover, by Proposition 2.1, either , in which case , or is not a square, in which case is not a square. So is a non-trivial Büchi sequence modulo .
We prove item 2. If , then the right-hand side is finite by Lemma 3.3. Otherwise it is an immediate consequence of Theorem 2.3 applied to modulus (observe that the only case where is infinite is in item 5). ∎
Corollary 3.5**.**
Let be an odd prime and . Let , with , unless . Write and let be such that (if , take ). We have
[TABLE]
Proof.
First note that the claim is trivial when . If and is even, this is just Lemma 3.4. Assume that and is even. In particular, since , cannot be [math]. We have
[TABLE]
(recalling the convention that if ), where the first equality comes from Lemma 3.4, and the second equality comes from Lemma 3.3 (which can be applied because is not invertible modulo , but is invertible modulo since ). ∎
Lemma 3.6**.**
Let be a prime and . Let , with unless . Write . We have:
If , then
[TABLE] 2. 2.
If is odd, then
[TABLE] 3. 3.
If is even, then
[TABLE]
Proof.
For the first item, just note that for any non-zero which is coprime with , the function defines a non-trivial Büchi sequence of length . Indeed, if , then , hence , but , which contradicts the fact that is coprime with , and the constant term is [math], hence has order than the order of the other coefficients. We conclude by Proposition 2.1 that it is a non-trivial sequence.
For items 2 and 3, choose with and , so that and (so they are squares modulo any ). Since , is divisible by , hence is even and . Moreover, is a non-trivial sequence because, one the one hand we have , and on the other hand it is not the square of a linear polynomial modulo . If it were, then we would have
[TABLE]
which is impossible, since modulo the right-hand side is a constant polynomial (because divides , hence also ), while the left-hand side is a non-constant polynomial since is not divisible by . ∎
We conclude this work with the proof of our main theorem.
Proof of Theorem 2.2.
If , then we have (recalling the convention — see the introduction)
[TABLE]
where the second and third equalities come from item 1 of Lemma 3.6 and Lemma 3.4, and the fourth equality comes from Corollary 3.5.
If is odd, then we have (using again Lemmas 3.6, Lemma 3.4 and Corollary 3.5):
[TABLE]
If is even, then we have
[TABLE]
∎
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[All 86] D. Allison, On square values of quadratics , Math. Proc. Cambridge Philos. Soc. 99 , no. 3, 381–383 (1986).
- 2[AHW 13] T. T. H. An, H.-L. Huang and J. T.-Z. Wang, Generalized Büchi’s problem for algebraic functions and meromorphic functions , Math. Z. 273 , no. 1-2, 95–122 (2013).
- 3[AW 11] Ta Thi Hoai An and J. Tzu-Yueh Wang, Hensley’s problem for complex and non-Archimedean meromorphic functions , Journal of Mathematical Analysis and Applications, 381-2 , 661-677 (2011).
- 4[Bre 03] A. Bremner, On square values of quadratics , Acta Arith. 108 , 95–111 (2003).
- 5[BB 06] J. Browkin and J. Brzeziński, On sequences of squares with constant second differences , Canad. Math. Bull. 49-4 , 481-491 (2006).
- 6[Ga 17] N. Garcia-Fritz, Quadratic sequences of powers and Mohanty’s conjecture , Int. J. Number Theory 14 , no. 2, 479–507 (2018).
- 7[Go X 11] E. González-Jiménez and X. Xarles, On symmetric square values of quadratic polynomials , Acta Arith. 149 , no. 2, 145–159 (2011).
- 8[He] D. Hensley, Sequences of squares with second difference of two and a problem of logic , unpublished (1980-1983).
