Hyperreal differentiation with an idempotent ultrafilter

Bryan Dawson; Samuel Allen Alexander

REVIEW 2 major objections 5 minor 16 references

Hyperreal differentiation with an idempotent ultrafilter

T0 review · 2 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read An idempotent ultrafilter makes the hyperreal derivative operator well-defined.

desk verdict A clean, correct result: idempotent ultrafilters in 0+ make the hyperreal derivative well-defined, plus a nice strengthening of Hindman's theorem; minor blemishes don't affect the main proof. read the letter →

arxiv 2411.14689 v1 pith:5YBFFV7H submitted 2024-11-22 math.LO

Samuel Allen Alexander , Bryan Dawson This is my paper · ORCID

classification math.LO MSC 26E3526A2454D8030D20

keywords hyperrealsidempotentultrafiltersderivativeoperatorfinitecalculusHindman'stheoremwell-definednesschainruleentirenumbers

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper shows that if the ultrafilter used to build the hyperreals is idempotent and contains every interval (0,ε) for ε>0, then the derivative operation on equivalence classes of functions is well-defined whenever the represented derivative exists. This means the hyperreal [f] can be given the derivative [f'], and the operation satisfies the standard algebraic rules of calculus—linearity, the product rule, the power rule—and on a smaller class of "entire" hyperreals, composition is also well-defined so the chain rule holds. The paper further develops a finite-difference analogue on the natural numbers, proves that it tracks ordinary derivatives through an irrational remainder, and uses it to give a new proof and strengthening of Hindman's theorem.

What carries the argument

The load-bearing machinery is an idempotent ultrafilter p on R that contains all intervals (0,ε) with ε>0. Idempotency means every S∈p has a p-large collection of shifts x with S−x∈p; this shift-stability is what lets difference quotients be compared at arbitrarily small increments. The finite-calculus half uses the same idea on N with the irrational remainder rm_γ(n), whose p-a.e. smallness makes Δ/[rm_γ] behave like the ordinary derivative.

What would settle it

To test Theorem 9, look for an idempotent ultrafilter p containing all (0,ε) and an everywhere-differentiable h with h≡0 on some S∈p while h' does not vanish on any set in p; such a pair would directly contradict the theorem. If instead one can prove that no idempotent ultrafilter can contain all the intervals (0,ε), the theorem becomes vacuous.

Watch

Extended reading notes

Core claim

The central discovery is that, for an idempotent ultrafilter p with every (0,ε)∈p, the map [f] ↦ [f'] is well-defined on the set D = {[f] : [f'] exists}. In proof, whenever f and g agree on a set S∈p and both derivatives exist, idempotency supplies a shift-set {x : S−x∈p}∈p; for each x there, S−x meets every (0,ε), so the difference quotients of f and g agree on points arbitrarily close to x, forcing f'(x)=g'(x). As a result the structure (D,1,Ω,+,·,') satisfies every positive formula in the language of elementary calculus functions, and on entire numbers—hyperreals represented by functions equal to their own Taylor series at 0—composition is well-defined, yielding the chain rule. The finite derivative Δ over an idempotent ultrafilter on N is similarly well-defined and, after dividing by the remainder of an irrational multiple, recovers standard derivatives; this leads to a proof of Hindman's theorem and its strengthening, Theorem 35.

Load-bearing premise

The proof depends on assuming there is a way of choosing a 'large' set of real numbers—an ultrafilter—that is self-similar under shifts and contains every tiny interval just to the right of zero; if no such choice exists, the main theorem applies to nothing.

Editorial extensions

If this is right

The hyperreal derivative on D obeys the product rule, linearity, power rule, and nontriviality, so a numerical model of a substantial fragment of elementary-calculus-function theory exists.
Adding symbols for functions like sin and cos, the structure satisfies all positive formulas whose hyperreal representatives have derivatives.
On entire numbers, composition is well-defined and the chain rule (x∘y)' = (x'∘y)y' holds.
The finite derivative Δ over an idempotent ultrafilter on N is well-defined on hyperreals and, through D_γ, differentiates functions of the discrete remainder variable by [rm_γ] with standard derivative rules.
Hindman's theorem follows from the well-definedness of Δ, and the same proof strengthens it to allow Ω as an additional element in finite sums.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural next step the paper leaves open is whether D is all of *R; if every hyperreal has a differentiable representative, the derivative operator extends to every hyperreal, while if not, the missing hyperreals would mark a genuine boundary of this 'number calculus.'
The use of idempotency in Corollary 30 suggests that any non-idempotent ultrafilter making Δ well-defined would yield a Hindman proof without idempotency; since the paper calls that possibility surprising, testing whether non-idempotent q can satisfy the conclusion would reveal whether idempotency is truly essential.
The secant-method example raises a concrete extension: try the same one-step hyperreal secant method on nonlinear differential equations to see whether the exactness seen for polynomial equations persists, and if not, what the failure reveals about the method.
The 'entire numbers' suggest a maximality question: determine whether the set of entire functions is the largest class on which composition can be defined consistently with a well-defined derivative and the chain rule.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Worth a look. The central result (Theorem 9/Corollary 10) is correct: for an idempotent ultrafilter p in 0+, the map [f] -> [f'] is well-defined on hyperreals whose representatives have derivatives p-a.e. The proof is short and sound: from [f]=[g] you get a p-large set where f and g agree, idempotency gives a p-large set of shifts preserving that set, and because p contains every (0,epsilon), those shifts bring you arbitrarily close to any point in the domain. The finite-calculus derivative in Section 4 is genuinely new to me, and Theorem 26 connecting Delta/rrm_gamma to the ordinary derivative for gamma-periodic functions is a nice observation. The strengthened Hindman theorem (Theorem 35) is a real bonus: the usual proof actually yields that the same c works for sums involving Omega as well. The authors are honest about what they don't know (whether D is proper in *R, whether non-idempotent ultrafilters could work, and the secant-method application being far from practical). External inputs, like the existence of an idempotent ultrafilter in 0+, are standard and properly cited. Soft spots, in proportion. The claim that (D, 1, Omega, +, *, ') satisfies every positive formula in the theory of elementary calculus functions is asserted without proof. That's stronger than anything established and would need a real argument about quantifiers and additional function symbols. The 'equivalence' in Section 3 about existence of [f'] is indeed false as a converse: containing all (0,epsilon) doesn't imply that a set intersecting all those intervals belongs to p. It's a side remark, not used in the proof of Theorem 9, so it's a blemish rather than a break. A few proofs are sketched (parts (2) and (3) of Theorem 26, parts of Theorem 32), but the missing details are routine. Who this is for: people in nonstandard analysis, ultrafilter combinatorics, and the 'differentiating numbers' program. It's not a breakthrough, but it's a solid, honest contribution with verifiable mathematics and no fitted parameters. I'd send it to a serious referee; the authors can fix the equivalence and either prove or soften the positive-formula claim. I'd cite it if I worked in this area.

Referee Report

2 major / 5 minor

Summary. The paper studies the possibility of defining a derivative on hyperreals *R_p by setting [f]' = [f']_p, and shows (Theorem 9) that if p is an idempotent ultrafilter on R that belongs to 0+ (i.e., contains every interval (0,ε) with ε>0), then this definition is well-defined on the domain D = {[f] : [f'] exists}. The proof of Theorem 9 is correct and uses idempotency to produce p-large sets on which the relevant difference quotients agree. The paper then introduces a finite-calculus analogue Δf(x) = *f(x+Ω) - f(x) associated with an idempotent ultrafilter q on N, proves a connection with ordinary derivatives for γ-periodic functions, establishes well-definedness of the corresponding hyperreal operator (Theorem 29), and uses it to give an alternative proof and a strengthening of Hindman's theorem (Theorem 35). It also discusses composition on 'entire numbers', an approximately space-filling property of the derivative graph, and a counterexample to the intuition that zero derivative should force constancy.

Significance. If the main result holds, it cleanly resolves a natural question in nonstandard analysis: an idempotent ultrafilter in 0+ makes hyperreal differentiation well-defined on the maximal natural domain (functions whose derivative exists). The proof of Theorem 9 is rigorous and the paper also offers an interesting new perspective on Hindman's theorem, with a strengthened statement involving the nonstandard element Ω that appears original and verifiable. The external inputs (existence of idempotent ultrafilters in 0+ and Bergelson's theorem) are standard and appropriately cited. However, the paper contains several over-strong auxiliary claims that need correction or qualification, and the exposition could be tightened in places.

major comments (2)

[Section 3, after Corollary 10] The assertion that the structure (D,1,Ω,+,·,') satisfies every positive formula in the theory of elementary calculus functions is not proved, and it does not follow directly from Theorem 9. The map f ↦ [f] is a homomorphism, but its surjectivity from a standard function algebra onto D is not established, and positive universal formulas are not automatically preserved when the map is not known to be surjective. Since D contains elements represented by arbitrary functions that are differentiable only on a p-large set, the claim is not obviously true as stated. Please either provide a proof of the preservation claim or weaken it to the specific axioms (Leibniz rule, linearity, power rule schema) that are actually established.
[Section 3, Proposition 11] The proof that {x : f(x)=r} ∉ p for all r is not justified. From p ∈ 0+ it follows only that every set in p has 0 as a right accumulation point; a level set that is a union of intervals shrinking to 0 may still belong to p. The choice of the Cantor set C needs to ensure that no level set of the associated devil's staircase is p-large (for example, by arranging that C has no gap adjacent to 0 and that each level set is countable or otherwise p-small), or the claim should be revised.

minor comments (5)

[Section 3, before Theorem 9] The statement that 'since p∈0+, the existence of [f'] is equivalent to the statement that for all real ε>0, there exists real δ∈(0,ε) such that f'(δ) exists' is false as written. The forward direction holds, but the converse does not: the condition only ensures that dom(f') is dense near 0, which does not force dom(f') ∈ p. This equivalence is not used in the proof of Theorem 9, but it should be corrected.
[Section 3, after Corollary 10] The discussion of positive formulas would benefit from a precise definition of the language and the theory of elementary calculus functions; currently the language has both id and Ω, and the relationship between them is not formalized. This makes the claim about satisfying every positive formula hard to verify.
[Section 3.2, Definition 13] The term 'entire' is normally reserved for complex-analytic functions on C; for real functions, 'real-analytic on R' would be clearer and avoid confusion with standard usage.
[Section 4, Definition 27] The /llbracket·/rrbracket notation is introduced for functions N → *R; please ensure the notation is typeset consistently and explicitly distinguished from the [·] notation used for ordinary hyperreals.
[Section 4, Theorem 40] The notation Dγ[f∘rmγ] is potentially confusing because Dγ is defined on hyperreals represented by functions on N, while f∘rmγ is a function on the domain S. Please clarify the intended interpretation and domain restrictions.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the derivative well-definedness proof is a direct consequence of the idempotent ultrafilter hypothesis, with external combinatorial inputs cited from independent sources.

full rationale

The paper's central claim (Theorem 9) is derived, not assumed: given an idempotent p in 0+, the proof takes S = S0 ∩ dom(f') ∩ dom(g'), uses idempotency to obtain {x : S-x ∈ p} ∈ p, and then shows the difference quotients of f and g agree on a p-large set of points x with S-x containing arbitrarily small positive h. Since f' and g' exist at such x, the limits agree p-a.e., giving [f'] = [g']. No fitted parameter is introduced and no conclusion is used as a hypothesis. Lemma 7 (existence of an idempotent ultrafilter in 0+) is delegated to the standard reference Hindman and Strauss, an independent external source; it is a nonconstructive existence theorem, not a restatement of Theorem 9. The finite-calculus half similarly relies on the standard existence of idempotent ultrafilters on N (Lemma 21, cited to [9]) and on Bergelson's Theorem 7.2 (Lemma 24), both external and unrelated to the paper's conclusions. The proof of Hindman's theorem (Theorem 34) is an alternate proof in the usual Glazer style: it uses Corollary 30, which itself depends only on the existence of an idempotent ultrafilter, a theorem independent of Hindman's theorem; the paper does not assume Hindman's theorem. The strengthening (Theorem 35) is a new statement proved from the same independent machinery. There are no self-citations among the load-bearing references, no fitted inputs renamed as predictions, and no definitional equivalence between the derivative operator and the property it is claimed to satisfy. The paper is self-contained against external benchmarks in the sense that every external input is a standard published theorem not equivalent to the target results. No circularity is present.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

No free parameters or invented entities are introduced. The paper depends on standard ultrafilter theory and nonstandard analysis; the key external inputs are the existence of idempotent ultrafilters (on R and N) and Bergelson's theorem on fractional parts of irrational multiples.

assumptions (7)

standard math There exists an idempotent ultrafilter p on R that contains every interval (0, epsilon) for epsilon > 0.
Lemma 7, citing Hindman-Strauss [9] (Theorems 13.29(a), 13.31). Nonconstructive existence used to define the derivative operator.
standard math There exists an idempotent ultrafilter q on N.
Lemma 21, citing Hindman-Strauss [9]. Used for the finite calculus section.
standard math For idempotent q on N and irrational gamma > 0, the set {n : |rm_gamma(n)| < epsilon} is in q for all epsilon > 0.
Lemma 24, from Bergelson [3], Theorem 7.2. Load-bearing for Theorem 26 and Theorem 32.
standard math Los's theorem and the standard nonstandard extension construction, so [f] = *f(Omega) and field operations behave as expected.
Used throughout; standard background for ultrafilter ultrapowers.
standard math Iterated ultrapower construction (Chang-Keisler Proposition 6.5.2) allows /llbracket f /rrbracket to be viewed as a hyperreal in the iterated ultrapower.
Remark 28; needed to interpret Delta[f] as a hyperreal in Corollary 30.
standard math Every entire function is determined by its derivatives at 0.
Used in Proposition 14 to prove well-definedness of composition of entire numbers.
standard math There exist disjoint Cantor sets on R.
Theorem 1.14 of Bankston-McGovern [1]; used in Proposition 11.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Hyperreal differentiation with an idempotent ultrafilter." pith.science (2026). https://pith.science/paper/5YBFFV7H

@misc{pith2026241114689,
  author       = {Pith},
  title        = {Pith review of: Hyperreal differentiation with an idempotent ultrafilter},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5YBFFV7H}},
  note         = {Machine review of arXiv:2411.14689}
}

read the original abstract

In the hyperreals constructed using a free ultrafilter on R, where [f] is the hyperreal represented by f:R->R, it is tempting to define a derivative operator by [f]'=[f'], but unfortunately this is not generally well-defined. We show that if the ultrafilter in question is idempotent and contains (0,epsilon) for arbitrarily small real epsilon then the desired derivative operator is well-defined for all f such that [f'] exists. We also introduce a hyperreal variation of the derivative from finite calculus, and show that it has surprising relationships to the standard derivative. We give an alternate proof, and strengthened version of, Hindman's theorem.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

16 extracted references · 16 canonical work pages

[1]

P Bankston, R J McGovern, T opological partitions, General Topology and its Appli cations 10 (1979) 215–229

work page 1979
[2]

E Barbeau , Remarks on an arithmetic derivative , Canadian Mathematical Bulletin 4 (1961) 117–122

work page 1961
[3]

Ultraﬁlters across Mathematics

V Bergelson, Ultraﬁlters, IP sets, dynamics, and combinatorial number t heory, from: “Ultraﬁlters across Mathematics”, American Mathematical Society (2010) 23–47

work page 2010
[4]

A Buium, Arithmetic differential equations, American Mathematical Society (2005)

work page 2005
[5]

Arithmetic and Geometry, Lon don Mathematical Society Lecture Note Series

A Buium, Differential calculus with integers, from: “Arithmetic and Geometry, Lon don Mathematical Society Lecture Note Series”, Cambridge U niversity Press (2015) 139–187

work page 2015
[6]

A Buium , F oundations of arithmetic differential geometry, American Mathematical Society (2023)

work page 2023
[7]

C Chang, H J Keisler, Model Theory, 3rd edition, New Y ork, Elsevier (1990)

work page 1990
[8]

R L Graham , D E Knuth , O Patashnik , Concrete Mathematics: A F oundation for Computer Science, 2nd edition, AddisonWesley (1994)

work page 1994

Show all 16 references

[9]

N Hindman , D Strauss , Algebra in the Stone ˇCech compactiﬁcation: theory and applications, Walter de Gruyter (2011)

2011
[10]

J Jeffries, Differentiating by prime numbers, Notices of the AMS 70 (2023) Hyperreal differentiation with an idempotent ultraﬁlter 17

2023
[11]

J Kovic, The arithmetic derivative and antiderivative , Journal of Integer Sequences 15 (2012)

2012
[12]

H Pasten, Arithmetic derivatives through geometry of numbers , Canadian Mathemati cal Bulletin 65 (2022) 906–923

2022
[13]

J M Shelly, Una cuesti ´on de la teor ´ıa de los n ´umeros, Asociaci ´on espa ˜nola, Granada (1911) 1–12

1911
[14]

M Stay, Generalized number derivatives, arXiv preprint math/0508364 (2005)

2005 arXiv
[15]

T T ossavainen , P Haukkanen , J K Merikoski , M Mattila , W e Can Differentiate Numbers, T oo, The College Mathematics Journal 55 (2024) 100–108

2024
[16]

Securities and Exchange Commission Union University samuelallenalexander@gmail.com, bdawson@uu.edu Received: aa bb 20YY Revised: cc dd 20ZZ

V Ufnarovski , B ˚Ahlander, How to differentiate a number , Journal of Integer Se quences 6 (2003) 03–3 The U.S. Securities and Exchange Commission Union University samuelallenalexander@gmail.com, bdawson@uu.edu Received: aa bb 20YY Revised: cc dd 20ZZ

2003

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

P Bankston, R J McGovern, T opological partitions, General Topology and its Appli cations 10 (1979) 215–229

work page 1979

[2] [2]

E Barbeau , Remarks on an arithmetic derivative , Canadian Mathematical Bulletin 4 (1961) 117–122

work page 1961

[3] [3]

Ultraﬁlters across Mathematics

V Bergelson, Ultraﬁlters, IP sets, dynamics, and combinatorial number t heory, from: “Ultraﬁlters across Mathematics”, American Mathematical Society (2010) 23–47

work page 2010

[4] [4]

A Buium, Arithmetic differential equations, American Mathematical Society (2005)

work page 2005

[5] [5]

Arithmetic and Geometry, Lon don Mathematical Society Lecture Note Series

A Buium, Differential calculus with integers, from: “Arithmetic and Geometry, Lon don Mathematical Society Lecture Note Series”, Cambridge U niversity Press (2015) 139–187

work page 2015

[6] [6]

A Buium , F oundations of arithmetic differential geometry, American Mathematical Society (2023)

work page 2023

[7] [7]

C Chang, H J Keisler, Model Theory, 3rd edition, New Y ork, Elsevier (1990)

work page 1990

[8] [8]

R L Graham , D E Knuth , O Patashnik , Concrete Mathematics: A F oundation for Computer Science, 2nd edition, AddisonWesley (1994)

work page 1994