A Conjecture on Primes in Arithmetic Progressions and Geometric Intervals

Abstract Dirichlet’s theorem on primes in arithmetic progressions states that for any positive integer q and any coprime integer a, there are infinitely many primes in the arithmetic progression a + nq ( ). Unfortunately, the theorem does not predict the gap between those primes. Several renowned open questions such as the size of Linnik’s constant try to say more about the distribution of such primes. This manuscript postulates a weak, but explicit, generalization of Linnik’s theorem, namely that each geometric interval contains a prime from each coprime congruence class modulo . Subsequently, it proves the conjecture theoretically for all sufficiently large t, as well as computationally for all sufficiently small q. Finally, the impact of this conjecture on a result of Pomerance related to Carmichael’s conjecture is outlined.


AN EXPLICIT GENERALIZATION OF LINNIK'S THEOREM. Dirichlet's theorem on primes in arithmetic progressions
is the direct generalization of Euclid's theorem on the infinity of primes to the setting of arithmetic progressions. Sadly, both of those theorems are purely existential and do not hint at the exact distribution of primes; a fact that has already puzzled mathematicians for several centuries. A first milestone was achieved in 1896 when the Prime Number Theorem describing the asymptotic distribution of primes was independently proven by Hadamard and De la Vallée Poussin. Shortly after, De la Vallée Poussin generalized his findings to primes in arithmetic progressions, indicating that the primes are evenly distributed among the coprime congruence classes. Over the last century, many mathematicians improved the asymptotic limits for the distribution of primes and outlined explicit computational bounds. Whereas the picture slowly clarifies for primes in general, there are not many explicit results for primes in arithmetic progressions. In fact, their study turns out to be extremely challenging if short intervals are involved. One example of this is the strenuous quest for predicting the size of the smallest prime in an arithmetic progression. An amazing result in this direction was achieved by Linnik in 1944 with the following theorem ( [10,11]). Theorem 1 (Linnik). There are absolute constants C and L such that for any q ∈ N ≥2 and any 1 ≤ a ≤ q − 1 with gcd(a, q) = 1, the smallest prime p 0 ≡ a (mod q) does not exceed Cq L .
Although Linnik did not write out his predicted constants C and L, his development showed them to be effectively computable. Today, the infimum over all possible absolute constants L for which Theorem 1 holds is known as Linnik's constant.

December 2022]
A CONJECTURE ON PRIMES L ≤ 5 [14] and under GRH one may conclude that any constant L > 2 is admissible [6]. Nonetheless, aligned with Schinzel's conjecture H 2 [13], the detailed survey [9] conjectures the upper bound L ≤ 2 and [3] shows that this actually holds for almost all moduli. After discussing the smallest prime in an arithmetic progression, one may go on and ask about the subsequent primes. Setting C = 1 and varying only the exponent in Theorem 1 leads to the following conjecture.

Conjecture 2.
For any positive integers q ≥ 2, 1 ≤ a ≤ q − 1 with gcd(q, a) = 1, and t ≥ 1, there exists a prime p such that q t ≤ p < q t+1 and p ≡ a (mod q). ( Of course, if primes smaller than q are allowed and t = 1, one recovers the conjecture on the size of Linnik's constant. Consequently, it is not surprising that Conjecture 2 is extremely difficult to prove for small exponents. However, the picture changes when t increases. Indeed, the expansion of these geometric intervals is faster than the decrease of the occurrence of primes. Thus, for sufficiently large t, the interval [q t , q t+1 ] will contain many primes of the desired form. One could even shrink the considered intervals and still conclude that they contain one prime from each congruence class. Whereas this is the main focus of most recent asymptotic results, Conjecture 2 is meant to be fully explicit without any hidden constants. Although it does not reflect the optimal interval length for the existential conclusion, it gives a practical range for retrieving such primes. Furthermore, it is deeply connected with other classical results such as Bertrand's postulate, which can be obtained by considering q = 2.

Remark 3.
For the sake of clarity, the upcoming development relies on explicit results only.

A PARTIAL PROOF FOR CONJECTURE 2.
Recent explicit results on the distribution of primes in arithmetic progressions allow a partial proof of Conjecture 2. Indeed, the article [2], accompanied by an enormous computational data set, gives explicit bounds for the number of primes in bounded arithmetic progressions. where the sum runs over all primes p ≤ x congruent to a modulo q. Then there exist explicit constants c θ (q) and x θ (q) such that Moreover, c θ (q) and x θ (q) satisfy c θ (q) ≤ c 0 (q) and x θ (q) ≤ x 0 (q), where c 0 (q) = Theorem 4 gives rise to an explicit lower bound for the exponents for which Conjecture 2 will hold, that is Lemma 5. Let q ≥ 3, 1 ≤ a ≤ q − 1 with gcd(a, q) = 1 and t ≥ T θ (q). Then there exists a prime p such that q t ≤ p < q t+1 and p ≡ a (mod q).
Although T θ (q) varies with q, Lemma 5 implies that for a fixed q ≥ 3 Conjecture 2 holds except maybe for finitely many exponents, namely for 0 < t < T θ (q). Using the explicit lists of constants c θ (q) and x θ (q) from [2] published at http://www.nt.math.ubc.ca/BeMaObRe/, one can computationally verify the conjecture for those finitely many missing exponents. A verification for all 3 ≤ q ≤ 45000 can be found at https://files.uni.lu/jim.barthel/PrimesInAP/ and leads to the following conclusion. Lemma 6. Let 3 ≤ q ≤ 45000, 1 ≤ a ≤ q − 1 with gcd(q, a) = 1, and 1 ≤ t ≤ T θ (q). Then there exists a prime p such that q t ≤ p < q t+1 and p ≡ a (mod q).
Using SageMath and its built-in Pari-GP function isprime(), the verification program computed for each modulus 3 ≤ q ≤ 45000 and each coprime remainder 1 ≤ a ≤ q − 1 an explicit list L(q, a) of primes p 1 , . . . , p T θ (q) −1 verifying and p i ≡ a (mod q) for all i ∈ {1, . . . , T θ (q) − 1}. These lists of primes are freely accessible and can also be used for other statistical analyses of primes in arithmetic progressions. Simultaneously, this verification guarantees correctness of the conjectured size of Linnik's constant for all q ≤ 45000.  where the sum runs over all primes p ≤ x congruent to a modulo q. Assuming the Extended Riemann Hypothesis, [1,Theorem 8.8.18] shows that Using those conditional bounds, one can show that Conjecture 2 holds for all t ≥ 3 when q ≥ 2, and even for t = 2 when q is sufficiently large (i.e., q ≥ 17386763).
Thus, under the Extended Riemann Hypothesis, only the conjectured bound L ≤ 2 for Linnik's constant remains to be proven.

A CONSEQUENCE FOR CARMICHAEL'S CONJECTURE.
Conjecture 2 is indirectly linked to Carmichael's totient function conjecture ( [4,5]). Before explaining this link, a short summary of this conjecture is required.

Conjecture 8 (Carmichael).
Let φ denote the Euler totient function. For a given number n, the equation φ(x) = n either has no solution at all or it has at least two solutions.
Correctness of the conjecture has already been proven for all x ≤ 10 10 10 [8], but the existence of a counterexample has not yet been ruled out for larger values. The only known attempt for constructing a counterexample has been developed by Pomerance in [12]. His construction relies on the following observation.

Theorem 9 (Pomerance)
. If x is a natural number such that for every prime p, (p − 1)|φ(x) implies p 2 |x, then φ(x) = φ(y) has only the trivial solution y = x.
However, Pomerance also points out that likely such a counterexample does not exist.
In particular, he shows that if the following conjecture holds, then there cannot be a counterexample as it would necessarily be divisible by every single prime. Proof. Assume Conjecture 2 holds, and let k ∈ N ≥2 . In order to prove that (p k − 1)| k−1 i=1 p i (p i − 1), it is sufficient to show that v q (p k − 1) ≤ v q ( k−1 i=1 p i (p i − 1)) for all q ∈ {p 1 , p 2 , . . . , p k−1 } where v q : N → N denotes the q-adic valuation map defined by v q (0) = ∞ and v q (n) = max{v ∈ N : q v |n}. Indeed, any prime divisor of (p k − 1) is smaller than p k and the valuation map outlines its highest power, granting an easy comparison of divisors. Accordingly, let q ∈ {p 1 , p 2 , . . . , p k−1 }, then q t ≤ p k − 1 < q t+1 for some t ∈ N. Thus v q (p k − 1) ≤ t. If t = 0, then the claim is trivial. If t > 0,

Conjecture 10 (Pomerance
where the last inequality follows from Conjecture 2's claim that for all t ∈ {1, . . . , t − 1} there is a prime p such that q t ≤ p < q t +1 and p ≡ 1 (mod q).