arXiv:1809.10813v4 [math.NT] 20 Jul 2020 hoe 1 Theorem integers. of divisors the where the ic hn R)hsbcm nw sRbnsieult.Teeaet are There [CNS11]. inequality. largest the Robin’s is as 5040 which known of become (RI), to has counterexamples (RI) then, Since rm ubr and number, prime ∗ † n18,Rbngv neuvln ttmn fteReanHypoth Riemann the of statement equivalent an gave Robin 1984, In oi’ nqaiyhsbe rvnfrvrosifiiefmle fintege of families infinite various for proven been has inequality Robin’s upre yAsrla eerhCuclDsoeyPoetDP1 Project Discovery Council Research Australian by Supported upre yAsrla CDsoeyPoetD10092adE and DP160100932 Project Discovery RC Australian by Supported t fe nees ealthat Recall integers. -free ing n18,Rbnsoe htteReanHptei for Hypothesis Riemann the that showed Robin 1984, In 1fe nees eetn hs eut ocvr2-rei 20-free cover to results 5-fre these extend integers: We power-free integers. of 11-free families infinite various for σ ( n σ ) ( n h nvriyo e ot ae abra Australia Canberra, Wales South New of University The stesmo iiosfnto and function divisors of sum the is Rbn[Rob84]) (Robin ) e < oi’ nqaiyfr2-reintegers 20-free for inequality Robin’s γ n t -full o log log tews.I 07 hi,Lciroo,Mre n Sol´e [CLMS07 and Moree, Lichiardopol, Choie, 2007, In otherwise. nvriyo rso,Bitl UK Bristol, Bristol, of University n . o all for h imn yohssi rei n nyi o all for if only and if true is Hypothesis Riemann The [email protected] [email protected] n colo Mathematics of School scalled is etme 7 2021 17, September ai onPlatt John David colo Science of School σ > n hmsMorrill Thomas ( n ) 00 oi’ nqaiyhssnebe proven been since has inequality Robin’s 5040. e < Abstract t -free γ 1 n o log log γ if steElrMshrn constant. Euler–Mascheroni the is n sntdvsbeb the by divisible not is n, ∗ † nees -reitgr,and integers, 7-free integers, e ζ ntegers. seuvln odemonstrat- to equivalent is 60100932 SCGatEP/K034383/1. Grant PSRC ssfor esis et-i known wenty-six t s nparticular in rs, hpwro any of power th ζ > n involving 5040 (RI) , ] showed that (RI) holds for all 5-free integers greater than 5040. Then, in 2012, Planat and Sol´e[SP12] improved this result to (RI) for 7-free integers greater than 5040, which was followed by Broughan and Trudgian [BT15] with (RI) for 11-free integers greater than 5040 in 2015. By updating Broughan and Trudgian’s work, we prove our main theorem. Theorem 2. Robin’s inequality holds for 20-free integers greater than 5040. Since there are no 20-full integers less than 5041, we may give a cleaner statement for Robin’s theorem. Corollary 3. The is true if and only if (RI) holds for all 20-full integers.

1 A bound for t-free integers

Sol´eand Planat [SP12] introduced the generalized Dedekind Ψ function t 1 (t 1) 1 p− Ψ (n) := n (1 + p− + + p− − )= n − . t ··· 1 p 1 p n p n − Y| Y| − Since 1 a σ(n)= n (1 + p− + + p− ), ··· pa n Y|| we see that σ(n) Ψ (n), provided that n is t-free. Thus, we study the function ≤ t Ψ (n) R (n) := t . t n log log n

By Proposition 2 of [SP12], it is sufficient to consider Rt only at the numbers n pn#= k=1 pk where pk is the kth prime. Compare this to the role of colossally abundant numbers in (RI) by Robin [Rob84]. UsingQ equation (2) of Broughan and Trudgian [BT15], we have for n 2 ≥ −t 1 p t 1 p # − −1 n p pn 1 p p>pn (1 p− )− 1 1 Rt(pn#) = ≤ − = − (1 p− )− pn# log log pn# ζ(t) log ϑ(pn) − Q Q p pn Y≤ where ϑ(x) is the Chebyshev function p x log p. ≤ In Sections 2 and 3, we construct two non-increasing functions, gB(w; t) and g (w; t) ∞ such that for some constants x , B weP have for x p B 0 0 ≤ n ≤ g (p ; t) R (p #) exp( γ) B n ≥ t n − and for pn > B g (pn; t) Rt(pn#) exp( γ). ∞ ≥ − For a given t 2, if we can show that all t-free numbers 5040 < n p # satisfy (RI), that ≥ ≤ k gB(pk; t) < 1 and that g (B; t) < 1, then we are done. ∞ 2 2 Deriving gB(pn; t) We start with some lemmas.

Lemma 4. Let ρ be a non-trivial zero of the with positive imaginary part 3 1012. Then ρ =1/2. ≤ · ℜ Proof. See Theorem 1 of [PT20].

Lemma 5. Let B =2.169 1025. Then we have · 1 ϑ(x) x √x log2 x for 599 x B. | − | ≤ 8π ≤ ≤ Proof. Given that one knows Riemann Hypothesis to height T , [B¨ut16] tells us that we may use Schoenfeld’s bounds from [Sch76] but restricted to B such that

B 4.92 T. slog B ≤

Using T =3 1012 from Lemma 4 we find B =2.169 1025 is admissible. · · Lemma 6. Let log x 55. Then ≥ 10 ϑ(x) x 1.388 10− x +1.4262√x | − | ≤ · or 10 ϑ(x) x 1.405 10− x. | − | ≤ · Proof. From Table 1 of [Dus18] we have for x> exp(55)

10 ψ(x) x 1.388 10− x | − | ≤ · so that by Theorem 13 of [RS62] we get, again for x> exp(55), that

10 ϑ(x) x 1.388 10− x +1.4262√x. | − | ≤ · The second bound follows trivially.

Lemma 7. Take B as above and define

∞ (ϑ(t) t)(1 + log t) C1 = − dt. t2 log2 t BZ

9 Then C 2.645 10− . 1 ≤ ·

3 Proof. We split the integral at X0 = exp(2000), apply Lemma 6 and consider

X0 X0 10 1 + log t 10 dt 10 1.405 10− dt 1.430 10− 5.055 10− . · t log2 t ≤ · t log t ≤ · BZ BZ For the tail of the integral, we use ϑ(x) x 30.3x log1.52 x exp( 0.8 log x) | − | ≤ − from Corollary 1 of [PT20a], valid for x X . We can thenp majorise the tail with ≥ 0 ∞ log t exp( 0.8√log t) 30.3 − dt t XZ0 9 which is less than 2.139 10− . ·

Lemma 8. Take B, C as above and let 599 x B. Then 1 ≤ ≤

1 exp( γ) 1.02 log x (log x + 3)√B (log B + 3)√x 1 − exp + + C1 + − . − p ≥ log x (x 1) log x 8π√x 4π√xB p x   ! Y≤ − Proof. Let M be the Meissel-Mertens constant M = γ + (log(1 1/p)+1/p). − p X Then by 4.20 of [RS62] we have

1 ϑ(x) x ∞ ϑ(t) t (1 + log t) log log x M | − | + | − | dt. p − − ≤ x log x 2 2 p x t log t ≤ Zx X Since 599 x B we can use Lemma 5 to bound the first term with ≤ ≤ log x . 8π√x We can split the integral at B and over the range [B, ) use the bound from Lemma 7. This ∞ leaves the range [x, B] where we can use Lemma 5 and a straightforward integration yields a contribution of (log x + 3)√B (log B + 3)√x − . 4π√xB We then simply follow the method used to prove Theorem 5.9 of [Dus18] with our bounds in place of η (k + 2)η k + k . k logk x (k + 1) logk+1 x

4 We also need Lemma 2 of [SP12].

Lemma 9 (Sol´eand Planat [SP12]). For n 2, ≥ 1 exp(2/p ). 1 p t ≤ n p>pn − Y − Putting all this together, we have the following.

Lemma 10. Define

. log pn (log pn+3)√B (log B+3)√pn exp 2 + 1 02 + + C + − log p pn (pn 1) log pn 8π√pn 1 4π√pnB n − gB(pn; t)= 2 .  ζ(t) log p √pn log pn  n − 8π   25 Then for t 2 and 599 pn B = 2.169 10 we have gB(pn; t) non-increasing in n and R (p #) ≥exp(γ)g (p ;≤t). ≤ · t n ≤ B n

3 Deriving g (pn; t) ∞ We will need a further bound.

Theorem 11. For x 767135587, ≥ p 1.02 1 5 eγ log x exp + + . p 1 ≤ (x 1) log x 6 log3 x 8 log4 x p x   Y≤ − −

Proof. This is the last display on page 245 of [Dus18] with k = 3 so that ηk =0.5. We can now deduce

Theorem 12. Define

2 1.02 1 5 exp + + 3 + 4 log p pn (pn 1) log pn 6 log pn 8 log pn n g (p ; t)= − . n 10 ∞ ζ(t) log pn 1.338 10 pn 1.4262√pn − · − − Then for t 2 and log p 55 we have  ≥ n ≥ γ Rt(pn#) e g (pn; t) ≤ ∞

and g (pn; t) is non-increasing in n. ∞

5 4 Computations

The proof rests on Briggs’ work [Bri06] on the colossally abundant numbers, which implies 10 (RI) for 5040 < n 10(10 ). We extend this result with the following Theorem: ≤ 13.114 85 Theorem 13. Robin’s inequality holds for all 5040 < n 10(10 ). ≤ Proof. We implemented Brigg’s algorithm from [Bri06] but using extended precision (100 bits) and interval arithmetic to carefully manage rounding errors. The final n checked was

29996208012611# 7662961# 44 293# 3 271# 666# 233# 109# 61# · · · · · · · 37# 23# 19# (13#)2 (7#)4 (5#)3 (3#)10 219. · · · · · · · ·

Corollary 14. Robin’s inequality holds for all 13# n 29996208012611#. ≤ ≤ We are now in a position to prove Theorem 2. We find that

gB(29996208012611;20) < 1 and g (B;20) < 1 ∞ and the result follows.

5 Comments

In terms of going further with this method, we observe that both

gB(29996208012611;21) > 1 and g (B;21) > 1 ∞ so one would need improvements in both. We only pause to note that one of the inputs to Dusart’s unconditional bounds that feed into g is again the height to which the Riemann ∞ Hypothesis is known1, so the improvements from Lemma 4 could be incorporated. Finally, we observe that if Rt(pn#) could be shown to be decreasing in n, then our lives would have been much easier.

6 Acknowledgements

The authors would like to thank Pierre Dusart and Keith Briggs for helpful conversations and Keith Briggs for sharing his code.

1Dusart uses T 2445999556030. ≥ 6 References

[B¨ut16] Jan B¨uthe. Estimating π(x) and related functions under partial RH assumptions. Math. Comp., 85(301):2483–2498, 2016.

[Bri06] Keith Briggs. Abundant numbers and the Riemann hypothesis. Experiment. Math., 15(2):251–256, 2006.

[BT15] Kevin A. Broughan and Tim Trudgian. Robin’s inequality for 11-free integers. Integers, 15:Paper No. A12, 5, 2015.

[CLMS07] YoungJu Choie, Nicolas Lichiardopol, Pieter Moree, and Patrick Sol´e. On Robin’s criterion for the Riemann hypothesis. J. Th´eor. Nombres Bordeaux, 19(2):357– 372, 2007.

[CNS11] Geoffrey Caveney, Jean-Louis Nicolas, and Jonathan Sondow. Robin’s theorem, primes, and a new elementary reformulation and the Riemann hypothesis. Inte- gers, 11:A33, 10, 2011.

[Dus18] Pierre Dusart. Explicit estimates of some functions over primes. Ramanujan J., 45(1):227–251, 2018.

[PT20] David J. Platt and Tim Trudgian The Riemann hypothesis is true up to3 1012. · To appear.

[PT20a] David J. Platt and Tim Trudgian The error term in the Theorem. To appear.

[RS62] J. Barclay Rosser and Lowell Schoenfeld. Approximate formulas for some func- tions of prime numbers. III. J. Math., 6, 64–94, 1962.

[Sch76] Lowell Schoenfeld. Sharper Bounds for the Chebyshev Functions θ(x) and ψ(x). II Math. Comp., 30(134):337–360, 1976.

[Rob84] G. Robin. Grandes valeurs de la fonction somme des diviseurs et hypoth`ese de Riemann. J. Math. Pures Appl. (9), 63(2):187–213, 1984.

[SP12] Patrick Sol´eand Michel Planat. The Robin inequality for 7-free integers. Integers, 12(2):301–309, 2012.

7