arxiv: 2604.06848 · v1 · submitted 2026-04-08 · 🧮 math.NT · math.PR

Recognition: no theorem link

A Hal\'{a}sz-type asymptotic formula for logarithmic means and its consequences

Alexander P. Mangerel, Oleksiy Klurman

Authors on Pith no claims yet

Pith reviewed 2026-05-10 18:28 UTC · model grok-4.3

classification 🧮 math.NT math.PR

keywords multiplicative functionslogarithmic meansHalász theoremLiouville functionrandom multiplicative functionsasymptotic formulasconverse theorems

0 comments

The pith

A sharp Halász-type asymptotic formula governs the logarithmic means of 1-bounded multiplicative functions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proves an asymptotic formula for the logarithmic mean L_f(x) of any multiplicative function f with |f| ≤ 1. The formula is sharp for many functions of interest. From this the authors deduce that any completely multiplicative f taking values in [-1,1] satisfies L_f(x) > -c/(log x)^{1-2/π} for a positive constant c and all x ≥ 3. The same formula yields an exponentially small bound on the probability that a random Rademacher completely multiplicative function has negative mean, and it supports a converse theorem characterizing when |L_f(x)| is small.

Core claim

We establish an asymptotic formula for the logarithmic mean value of a 1-bounded multiplicative function that is sharp in many cases of interest. In particular, if f is a completely multiplicative function taking values in [-1,1] then there is a constant c > 0 such that for every x ≥ 3, L_f(x) := sum_{n≤x} f(n)/n > -c/(log x)^{1-2/π}. The exponent can be improved to -1+o(1) unless f behaves like the Liouville function in a precise sense. For a Rademacher random completely multiplicative function the probability that L_f(x) is negative is O(exp(-x^c)) for some c in (0,1). A converse theorem for small |L_f(x)| is also obtained and shown to be essentially best possible.

What carries the argument

Halász-type asymptotic formula for the logarithmic mean L_f(x) = sum_{n≤x} f(n)/n of 1-bounded multiplicative functions f.

If this is right

Completely multiplicative functions valued in [-1,1] obey the lower bound L_f(x) > -c/(log x)^{1-2/π} for x ≥ 3.
The exponent improves to -1+o(1) whenever f does not behave like the Liouville function.
Rademacher random completely multiplicative functions have negative logarithmic mean with probability O(exp(-x^c)) for some c in (0,1).
Small values of |L_f(x)| are characterized by a converse theorem that is essentially sharp.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The formula may extend to ordinary means of multiplicative functions and yield analogous bounds there.
The random probability result suggests that sign oscillations of random multiplicative functions are controlled on the logarithmic scale.
Converse theorems of this type could classify arithmetic functions whose partial sums or means remain small.

Load-bearing premise

That a Halász-type asymptotic formula holds sharply for the logarithmic means of 1-bounded multiplicative functions.

What would settle it

A 1-bounded multiplicative function f for which L_f(x) falls below -C/(log x)^{1-2/π} along a sequence of x tending to infinity for arbitrarily large C.

read the original abstract

We establish an asymptotic formula for the logarithmic mean value of a 1-bounded multiplicative function that is sharp in many cases of interest. We derive from it a variety of applications, making progress on several old problems. As a first application, we show that if $f$ is a completely multiplicative function taking values in $[-1,1]$ then there is a constant $c > 0$ such that for every $x \geq 3$, $$ L_f(x) := \sum_{n \leq x} \frac{f(n)}{n} > -\frac{c}{(\log x)^{1-2/\pi}}, $$ thus significantly improving on a 20-year-old result of Granville and Soundararajan. We also show that the exponent of $\log x$ in this result can be improved to $-1+o(1)$, as long as $f$ does not ``behave like'' the Liouville function $\lambda$ in a precise sense. As a second application, we show that for a Rademacher random completely multiplicative function $\mathbf{f}$, the probability that $L_{\mathbf{f}}(x)$ is negative is $O(\exp(-x^c))$ for some $c \in (0,1)$, thus establishing a previously conjectured bound. Finally, we obtain a converse theorem for small absolute values $|L_f(x)|$, and construct examples $f$ that show that it is (essentially) best possible.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Klurman and Mangerel give a Halász-type asymptotic for logarithmic means that sharpens the negativity bound to (log x) to the power -(1-2/π) and settles the random tail conjecture.

read the letter

The main point is that Klurman and Mangerel have found a Halász-type asymptotic formula for the logarithmic mean of 1-bounded multiplicative functions. This formula is sharp enough in many cases to give new bounds on how negative the mean can get. They use it to show that for completely multiplicative f taking values in [-1,1], the sum L_f(x) stays above -c over (log x) to the power 1 minus 2 over pi. That is a clear improvement on the Granville-Soundararajan bound from twenty years ago. They also get a stronger almost 1 over log x bound when f does not mimic the Liouville function too closely. In the random case with Rademacher signs, the chance that L_f(x) goes negative drops exponentially fast, like exp of minus x to some positive power less than 1. They close with a converse result on small values of the mean, backed by examples that show it cannot be improved much. The paper does a good job turning the asymptotic into concrete applications without extra assumptions. The exponent 1-2/pi comes from a standard optimization in the pretentious distance, and having it explicit is useful. The random probability bound settles what was only a conjecture before. One minor soft spot is that the exponential decay in the random case has an unspecified c less than 1, so it is not the strongest possible tail, but it is still much better than previous work. The main asymptotic seems to rest on established Halász machinery, so no circularity there. This paper is aimed at people in analytic number theory who care about multiplicative functions and their means. A reader who knows the classical Halász theorem and the Granville-Soundararajan work will see the progress right away and can use the new formula in their own arguments. It deserves serious referee attention because the results are specific and address real open questions in the field. I recommend sending it out for peer review.

Referee Report

0 major / 4 minor

Summary. The manuscript establishes a Halász-type asymptotic formula for the logarithmic mean L_f(x) := sum_{n≤x} f(n)/n of 1-bounded multiplicative functions f, claimed to be sharp in many cases of interest. From this formula the authors derive three main applications: (i) for completely multiplicative f with values in [-1,1], L_f(x) > -c/(log x)^{1-2/π} for x≥3 and some c>0, improving Granville-Soundararajan; (ii) for a Rademacher random completely multiplicative function f, P(L_f(x)<0) = O(exp(-x^c)) for some c∈(0,1); (iii) a converse theorem characterizing when |L_f(x)| is small, together with examples showing the result is essentially sharp.

Significance. If the central asymptotic holds with the stated sharpness, the work supplies a useful new tool for controlling logarithmic means of multiplicative functions and yields concrete progress on longstanding questions. The explicit exponent 1-2/π, obtained via optimization over phases or pretentious distances, is a non-trivial quantitative improvement; the random-sign probability bound resolves a prior conjecture, and the converse theorem with matching examples adds to the structural theory. The derivation appears parameter-free beyond the constant c and produces falsifiable predictions.

minor comments (4)

The definition of the logarithmic mean L_f(x) appears only in the abstract and the displayed inequality; it should be restated verbatim in the introduction and in the statement of the main asymptotic theorem for reader convenience.
The phrase 'does not behave like the Liouville function λ in a precise sense' (used to obtain the improved exponent -1+o(1)) is introduced without an immediate formal definition; this notion should be defined explicitly at the first occurrence, preferably with a displayed condition on the pretentious distance or correlation.
The constant c>0 in the lower bound is asserted to exist but no indication is given whether it is effective or how its value depends on the function f; a brief remark on effectivity (or lack thereof) would strengthen the statement.
The abstract claims the asymptotic is 'sharp in many cases of interest'; the introduction or the section containing the main theorem should list at least two or three concrete families (e.g., Liouville, character sums, or random signs) where sharpness is verified, with a forward reference to the relevant proposition or example.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of our manuscript, recognition of its significance, and recommendation for minor revision. We are pleased that the sharpness of the asymptotic formula, the improvement on the Granville-Soundararajan bound, the resolution of the random-sign conjecture, and the converse theorem are viewed favorably.

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained

full rationale

The paper's central contribution is a new asymptotic formula for the logarithmic mean of 1-bounded multiplicative functions, established via standard analytic number theory methods building on Halász's theorem and related external results. This formula is not defined in terms of its own consequences, nor are any predictions fitted to subsets of the same data. The derived lower bound with exponent 1-2/π, the random multiplicative function probability bound, and the converse theorem all follow as direct applications once the main formula is available, without load-bearing self-citations, imported uniqueness theorems, or smuggled ansatzes. The structure is independent of the paper's own outputs and relies on externally verifiable prior theorems.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The claim rests on extending standard tools in analytic number theory to logarithmic means; no new entities invented and only an existential constant c as free parameter.

free parameters (1)

c
Existence of a positive constant c in the lower bound inequality, not explicitly computed.

axioms (2)

domain assumption Halász-type asymptotic formulas for multiplicative functions
The main result is a Halász-type formula, relying on such prior theorems.
domain assumption Complete multiplicativity and boundedness by 1
Assumed for the applications to L_f(x) and random f.

pith-pipeline@v0.9.0 · 5573 in / 1446 out tokens · 102887 ms · 2026-05-10T18:28:21.269108+00:00 · methodology

discussion (0)

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Small values of signed harmonic sums and logarithmic means of multiplicative functions
math.NT 2026-05 unverdicted novelty 7.0

Signed sequences and completely multiplicative functions are constructed with small signed harmonic sums and logarithmic means decaying like exp(-c N^{1/3}/(log N)^{1/3}) infinitely often.

Reference graph

Works this paper leans on

25 extracted references · 1 canonical work pages · cited by 1 Pith paper

[1]

Angelo and M.W

R. Angelo and M.W. Xu. On a Tur´ an conjecture and random multiplicative functions.Q.J. Math., 74(2):767–777, 2022

2022
[2]

Goldmakher

L. Goldmakher. Multiplicative mimicry and improvements of the P´ olya-Vinogradov inequality.Algebra Number Theory, 6:123–163, 2012

2012
[3]

Granville, A.J

A. Granville, A.J. Harper, and K. Soundararajan. A new proof of Hal´ asz’ theorem and its consequences. Compositio Math., 155(1):126–163, 2019

2019
[4]

Granville, D

A. Granville, D. Koukoulopoulos, and K. Matom¨ aki. When the sieve works.Duke Math. J., 164(10):1935– 1969, 2015

1935
[5]

Granville and A.P

A. Granville and A.P. Mangerel. Three conjectures about character sums.Math. Zeit., 305:49:34 pp., 2023

2023
[6]

Granville and K

A. Granville and K. Soundararajan. The spectrum of multiplicative functions.Ann. Math., 153(2):407– 470, 2001

2001
[7]

Granville and K

A. Granville and K. Soundararajan. Decay of mean values of multiplicative functions.Can. J. Math., 55(6):1191–1230, 2003

2003
[8]

Granville and K

A. Granville and K. Soundararajan. Negative values of truncations ofl(1, χ).Clay Mathematics Proceed- ings, 7:141–149, 2005

2005
[9]

Hal´ asz.¨Uber die Mittelwerte multiplikativer zahlentheoretische Funktionen.Acta Math

G. Hal´ asz.¨Uber die Mittelwerte multiplikativer zahlentheoretische Funktionen.Acta Math. Sci. Hung., 19:365–403, 1968

1968
[10]

Hal´ asz

G. Hal´ asz. On the distribution of additive and the mean values of multiplicative arithmetic functions. Studia Scient. Math. Hung., 6:211–233, 1971

1971
[11]

Halberstam and H.-E

H. Halberstam and H.-E. Richert. On a result of R.R. Hall.J. Number Theory, 11(1):76–89, 1979

1979
[12]

R.R. Hall. Proof of a conjecture of Heath-Brown concerning quadratic residues.Proc. Edin. Math. Soc., 39(3):581–588, 1996

1996
[13]

Hall and G

R.R. Hall and G. Tenenbaum.Divisors. Cambridge Tracts in Mathematics vol. 90, Cambridge University Press, 1988

1988
[14]

Haselgrove

C.B. Haselgrove. A disproof of a conjecture of p´ olya.Mathematika, 5(2):141–145, 1958

1958
[15]

Hoeffding

W. Hoeffding. Probability inequalities for sums of bounded random variables.J. Amer. Stat. Assoc., 58 (301):13––30, 1963

1963
[16]

A. Ingham. On the difference between consecutive primes.Q.J. Math., 8:255–266, 1937

1937
[17]

Kerr and O

B. Kerr and O. Klurman. How negative canP n≤x f(n)/nbe?Math. Proc. Camb. Phil. Soc., to appear:26 pp., 2026

2026
[18]

Koukoulopoulos

D. Koukoulopoulos. On multiplicative functions which are small on average.Geom. Funct. Anal., 23(5):1569–1630, 2013

2013
[19]

Kucheriaviy

P. Kucheriaviy. Positivity of partial sums of a random multiplicative function and corresponding problems for the Legendre symbol. arXiv:2510.25691v2 [math.NT]

work page arXiv
[20]

Lamzouri and A.P

Y. Lamzouri and A.P. Mangerel. Large odd order character sums and improvements of the P´ olya- Vinogradov inequality.Trans. Amer. Math. Soc., 375(6):3759–3793, 2022

2022
[21]

Matom¨ aki and X

K. Matom¨ aki and X. Shao. When the sieve works II.J. reine angew. Math. (Crelle), 763:1–24, 2020

2020
[22]

Montgomery and R.C

H.L. Montgomery and R.C. Vaughan. Mean values of multiplicative functions.Per. Math. Hung., 43:199– 214, 2002

2002
[23]

P. Shiu. A Brun-Titschmarsh inequality for multiplicative functions.J. reine angew. Math., 313:161–170, 1980

1980
[24]

Tenenbaum.Introduction to Analytic and Probabilistic Number Theory

G. Tenenbaum.Introduction to Analytic and Probabilistic Number Theory. Graduate Studies in Mathe- matics vol. 163, AMS Publications, Providence, RI, 2015

2015
[25]

Tenenbaum

G. Tenenbaum. Moyennes effectives de fonctions multiplicatives complexes.Ramanujan J., 44:641–701, 2017. School of Mathematics, University of Bristol, Woodland Road, Bristol, BS8 1UG, UK Email address:lklurman@gmail.com Department of Mathematical Sciences, Durham University, Stockton Road, Durham, DH1 3LE, UK Email address:smangerel@gmail.com

2017