betaFeedback

WorksMathematics

Integral zero-sum incidence identities and Davenport constants in rank three

Abstract

We study the Davenport constant of finite abelian groups through signed identities among zero-sum subsequences. An integral incidence identity of degree tt for a sequence assigns integer weights to its nonempty zero-sum subsequences so that the weights add up to one, while those of the subsequences containing any given nonempty set of at most tt terms add up to zero. Over a finite abelian pp-group HH, the zero-sum subsequences of a sequence of any length have an indicator polynomial of degree at most D(H)−1D(H)-1, and this gives a transfer principle: if, for each t≥0t\ge0, every sequence of L(t)L(t) terms over QQ admits such an identity of degree tt, then D(Q⊕H)≤L(D(H)−1)D(Q\oplus H)\le L(D(H)-1) for every finite abelian pp-group HH. For m∈{2,3,4}m\in\{2,3,4\} we construct integral identities at length L(t)=m(t+3)−2L(t)=m(t+3)-2 over Cm3C_m^3, and prove that these thresholds are sharp over fields of characteristic zero or characteristic not dividing mm. For a prime p∤mp\nmid m and 0≤a≤b≤c0\le a\le b\le c, we prove that D(Cmpa⊕Cmpb⊕Cmpc)=m(pa+pb+pc)−2. D(C_{mp^a}\oplus C_{mp^b}\oplus C_{mp^c}) =m(p^a+p^b+p^c)-2. In particular D(C153)=43D(C_{15}^3)=43 and D(C203)=58D(C_{20}^3)=58, which are the two smallest cases of the conjecture D(Cn3)=3n−2D(C_n^3)=3n-2 left open in the literature. The proofs of the base cases for m∈{3,4}m \in \{3,4\} are computer-assisted. All of the results in this paper have been formalised in Lean 4.

D(Cmpa⊕Cmpb⊕Cmpc)=m(pa+pb+pc)−2.\mathrm{D}(C_{mp^a}\oplus C_{mp^b}\oplus C_{mp^c})=m(p^a+p^b+p^c)-2.

In particular D(C153)=43\mathrm{D}(C_{15}^3)=43 and D(C203)=58\mathrm{D}(C_{20}^3)=58, which are the two smallest cases of the conjecture D(Cn3)=3n−2\mathrm{D}(C_n^3)=3n-2 left open in the literature. The proofs of the base cases for m∈{3,4}m \in\{3,4\} are computer-assisted. All of the results in this paper have been formalised in Lean 4.

Introduction

Let GG be a finite abelian group, written additively. A sequence over GG is a finite list of elements of GG in which repetitions are allowed, and the Davenport constant D(G)\mathrm{D}(G) is the least integer LL such that every sequence of LL terms over GG has a nonempty subsequence with sum zero. If

G≅Cn1⊕⋯⊕Cnr,1<n1∣⋯∣nr,G \cong C_{n_1}\oplus\cdots\oplus C_{n_r},\qquad1<n_1\mid\cdots\mid n_r,

then ni−1n_i-1 copies of each basis element form a zero-sum free sequence, so

D(G)≥D∗(G):=1+∑i=1r(ni−1).\mathrm{D}(G)\ge\mathrm{D}^*(G):=1+\sum_{i=1}^{r}(n_i-1).

Olson proved that equality holds for pp-groups and for groups of rank at most two [16, 17]. Equality fails for infinitely many groups of rank four [11, 8]. In rank three, it is conjectured that

D(Cn1⊕Cn2⊕Cn3)=n1+n2+n3−2(1<n1∣n2∣n3).(1)\mathrm{D}(C_{n_1}\oplus C_{n_2}\oplus C_{n_3})=n_1+n_2+n_3-2\qquad(1<n_1\mid n_2\mid n_3). \tag*{(1)}

For background on the conjecture, see [7, 8, 19, 10]. In this paper, we develop a method for attacking particular cases of this conjecture. We write the group as Q⊕HQ\oplus H with HH a pp-group, and read a sequence over it as a sequence over the quotient QQ whose terms carry labels in the kernel HH. We first study identities among the zero-sum subsequences of sequences over a small quotient, and then use those identities for arbitrary labels in a pp-group kernel.

Our method allows us to prove the conjecture (1) for infinitely many mixed-prime groups of rank three with a quotient Cp3C_p^3 or Cp4C_p^4, and the values it produces include the two smallest cases of (1) for Cn3C_n^3 that the classical theory leaves open, n=15n = 15 and n=20n = 20.

The inductive method. The standard way to bound D(G)D(G) from above when GG is not a pp-group reduces the question to two smaller groups, a subgroup HH and the quotient Q=G/HQ = G/H. It goes back to the earliest work on the subject [17, 20], and is known in the zero-sum literature as the inductive method [19 Section 3], [3, 14]. A zero sum in a quotient need not be a zero sum in the original group. Project a sequence over GG to QQ, and call a subsequence a block if its projection has sum zero. Each block has a sum in HH; if we can find D(H)D(H) pairwise disjoint blocks, some nonempty collection of their sums adds to zero in HH, and the union of those blocks is a zero sum in GG [5 Proposition 2.6].

Short blocks make this argument efficient. Write η(Q)\eta(Q) for the least length such that every sequence over QQ of that length contains a nonempty zero-sum subsequence of length at most exp⁡(Q)\exp(Q). Extracting short blocks successively gives, in particular,

D(Q⊕H)≤η(Q)+(D(H)−1)exp⁡(Q).(2)D(Q \oplus H) \le\eta(Q) + (D(H) - 1)\exp(Q). \tag*{(2)}

The same estimate holds for GG in the extension above.

The last block need not be short, and this improves the count. With D(H)≥2D(H) \ge2, extract D(H)−1D(H) - 1 short blocks and then one arbitrary zero-sum block. This requires at most

max⁡{η(Q)+(D(H)−2)exp⁡(Q), D(Q)+(D(H)−1)exp⁡(Q)}\max\{\eta(Q) + (D(H) - 2)\exp(Q),\ D(Q) + (D(H) - 1)\exp(Q)\}

terms. The first expression ensures that the last short block can be extracted; the second ensures that a final zero sum remains. The maximum is the second expression precisely when

η(Q)≤D(Q)+exp⁡(Q).\eta(Q) \le D(Q) + \exp(Q).

For Q=Cp2Q = C_p^2, the values η(Q)=3p−2\eta(Q) = 3p - 2 and D(Q)=2p−1D(Q) = 2p - 1 satisfy this inequality [17, 8]; this is the count that makes the inductive method exact in rank two. For Q=C33Q = C_3^3 and Q=C43Q = C_4^3, however, the comparison is 17>7+317 > 7 + 3 and 22>10+422 > 10 + 4, respectively [9, 15]. For the groups Cm3⊕HC_m^3 \oplus H studied below, the second expression is the basis lower bound, so for m=3,4m = 3, 4 the count overshoots it. Nor is the loss due to the insistence on short blocks: ten terms over C33C_3^3 need not contain two disjoint blocks of any size (Appendix § 8), although D(C33⊕C2)=10D(C_3^3 \oplus C_2) = 10.

Once disjoint blocks have been chosen, this argument uses only their sums in HH. It forgets the other zero sums of the projection and the ways in which they overlap. Our approach keeps that incidence information and combines the zero sums with signed weights.

Incidence identities. We regard sequences as indexed, so equal group elements at different positions are considered different terms, since the kernel labels attached to those positions may differ. For U=(u1,…,uN)U = (u_1,\ldots,u_N) over GG, write

Z(U)={∅≠E⊆[N]:∑j∈Euj=0}\mathcal{Z}(U) = \{ \varnothing\ne E \subseteq[N] : \sum_{j \in E} u_j = 0 \}

for the family of its zero sets, and let 1E∈{0,1}N\mathbf{1}_E \in\{0,1\}^N be the incidence vector of EE.

Let tt be a non-negative integer, and let RR be a commutative ring with identity. An incidence identity of degree tt over RR, or It\mathcal{I}_t-identity, is a family (cE)E∈Z(U)(c_E)_{E \in\mathcal{Z}(U)} of elements of RR satisfying

∑EcE=1,∑E⊇JcE=0for every J⊆[N] with 1≤∣J∣≤t.(3)\sum_E c_E = 1,\qquad\sum_{E \supseteq J} c_E = 0 \quad\text{for every } J \subseteq[N] \text{ with } 1 \le|J| \le t. \tag*{(3)}

The sum indexed by JJ is the moment at JJ. Thus the weights add to one, but the weights through any prescribed nonempty set of at most tt positions cancel. We say the identity is integral when R=ZR = \mathbb{Z}.

For example, if BB and CC are disjoint nonempty zero sets, give weights 11, 11, −1-1 to BB, CC, B∪CB \cup C, respectively. The total weight is one, and the total through each individual position is zero. These are exactly the conditions for degree one.

The polynomial meaning is the reason for requiring all these cancellations. On Boolean inputs, the monomial ∏j∈Jzj\prod_{j \in J} z_j records whether a set contains every position of JJ. Consequently (3) is equivalent to

∑EcEP(1E)=P(0)for every polynomial P of degree at most t.\sum_E c_E P(\mathbf{1}_E) = P(0) \quad\text{for every polynomial } P \text{ of degree at most } t.

We prove this in Lemma 2.1. The empty subsequence receives no weight; nevertheless, these signed evaluations reproduce its value for every polynomial of the indicated degree.

For an abelian pp-group HH, the test for a selected set of kernel labels to have sum zero is such a polynomial over Fp\mathbb{F}_p: its degree is at most D(H)−1D(H) - 1, independently of the number of labels (Lemma 3.1). An identity of this degree for the sequence over the quotient can therefore be evaluated at the kernel’s zero-sum test, and this gives the transfer principle.

Theorem A (Transfer principle). Let QQ be a finite abelian group, let pp be a prime, and let L(0),L(1),L(2),…L(0), L(1), L(2), \ldots be integers such that every sequence of L(t)L(t) terms over QQ has an It\mathcal{I}_t-identity over Fp\mathbb{F}_p. Then

D(Q⊕H)≤L(D(H)−1)for every finite abelian p-group H.D(Q \oplus H) \leq L(D(H) - 1) \quad\text{for every finite abelian } p\text{-group } H.

If the identities are integral, the conclusion holds for every prime pp.

To see the mechanism, let PP be the zero-sum indicator for the kernel labels and evaluate the quotient identity at PP. The result is

1=P(0)=∑EcEP(1E).1 = P(0) = \sum_E c_E P(\mathbf{1}_E).

Some summand is nonzero, so some nonempty zero set in the quotient also has sum zero in the kernel. For a fixed degree, the weights depend only on the quotient sequence; the same identity works for every choice of kernel labels. Integer weights can be reduced modulo any prime, which makes a single certificate useful for many different kernels.

Over a field KK, the alternative to an It\mathcal{I}_t-identity is a separator of degree at most $t: a polynomial PP of degree at most tt such that P(0)=1P(0) = 1 and P(1E)=0P(\mathbf{1}_E) = 0 for every E∈Z(U)E \in\mathcal{Z}(U). Lemma 2.1 shows that exactly one of these alternatives occurs. We write κt(G;K)\kappa_t(G; K) for the least NN such that no sequence of NN terms over GG has a separator of degree at most tt. In degree zero the only possible separator is the constant polynomial $1$, which works exactly when there is no nonempty zero sum. Thus κ0(G;K)=D(G)\kappa_0(G; K) = D(G).

The quotient hierarchy. For Q=C33Q = C_3^3 and H=Cpa⊕Cpb⊕CpcH = C_{p^a} \oplus C_{p^b} \oplus C_{p^c} with p∤mp \nmid m, Olson’s theorem and the Chinese remainder theorem give

D∗(Cm3⊕H)=m(D(H)−1)+3m−2.D^*(C_m^3 \oplus H) = m(D(H) - 1) + 3m - 2.

This singles out the target length Lm(t)=mt+3m−2=m(t+3)−2L_m(t) = mt + 3m - 2 = m(t + 3) - 2: a quotient identity at this length has exactly the degree and length needed to reach the basis lower bound after transfer.

Theorem B. Let m∈{2,3,4}m \in\{2,3,4\} and let t≥0t \geq0 be an integer. Every indexed sequence of m(t+3)−2m(t+3)-2 terms over Cm3C_m^3 has an integral It\mathcal{I}_t-identity. Consequently κt(Cm3;K)≤m(t+3)−2\kappa_t(C_m^3; K) \leq m(t + 3) - 2 for every field KK, with equality whenever the characteristic of KK is zero or does not divide mm.

The m=2m=2 case gives a small geometric model of the argument: a signed identity on the Fano plane with one point removed. For m=3m=3 and m=4m=4, finite computations supply the base cases, and the three constructions of Lemma 2.2 propagate them to all degrees.

Combining Theorems A and B gives D(Cm3⊕H)≤m(D(H)−1)+3m−2D(C_m^3 \oplus H) \le m(D(H)-1)+3m-2 for every finite abelian pp-group HH. When HH has rank at most three and p∤mp \nmid m, the right-hand side is the basis lower bound, giving the following consequence.

Corollary C. Let m∈{2,3,4}m \in\{2,3,4\}, let pp be a prime with p∤mp \nmid m, and let $0 \le a \le b \le c$ be integers. Then

D(Cmpa⊕Cmpb⊕Cmpc)=m(pa+pb+pc)−2.(4)D(C_{mp^a} \oplus C_{mp^b} \oplus C_{mp^c}) = m(p^a+p^b+p^c)-2. \tag*{(4)}

In particular D(C3pa3)=9pa−2D(C_{3p^a}^3)=9p^a-2 for p≠3p \ne3 and D(C4pa3)=12pa−2D(C_{4p^a}^3)=12p^a-2 for odd pp.

For example, D(C153)=43D(C_{15}^3)=43. As C153≅C33⊕C53C_{15}^3 \cong C_3^3 \oplus C_5^3 and D(C53)=13D(C_5^3)=13, the classical estimate (2) only gives the upper bound 5353.

Trivial kernel factors are allowed, with Cp0=C1C_{p^0}=C_1. When p∣mp \mid m, the rank-three groups displayed in (4) are themselves pp-groups, and their Davenport constants are already given by Olson’s theorem.

Earlier work and our contribution. Corollary C includes classical families: the doubled pp-groups and tripled 22-groups of van Emde Boas and Kruyswijk [20, 21], as well as special cases covered by later rank-three results [4, 3, 19, 14]. For Cn3C_n^3 earlier work covers, besides the prime powers, the exponents n=2pan=2p^a and n=3⋅2an=3\cdot2^a [19, 13], but not n=3pan=3p^a or n=4pan=4p^a with p≥5p \ge5.3

The group-algebra input goes back to Olson, and polynomial methods have a substantial history in zero-sum theory [16, 1]. Finite computational input to the inductive method also occurs in the work of Bhowmik, Halupczok and Schlage-Puchta [3]. Related signed zero-block and moment methods occur in Volkmann’s work on nonabelian groups of exponent three [22, 23, 24]. Here we establish complete incidence identities over the integers, throughout the degree hierarchy for three small quotients. The transfer principle explains how those identities control arbitrary labels in an abelian pp-group kernel, while integrality supplies a common certificate in every characteristic.

Scope. The bounded-degree indicator in Theorem A uses a pp-group kernel. Remark 3.3 explains why a degree bound independent of the number of labels fails for other kernels in characteristic pp. If nn has at least three distinct prime divisors, separating off one primary component of Cn3C_n^3 leaves a quotient Cm3C_m^3 with at least two prime divisors. An incidence hierarchy at length m(t+3)−2m(t+3)-2 for such a quotient would already include the Davenport conjecture for that quotient in degree zero. Deleting short zero sets, as in Section 5, reduces a proposed hierarchy to finitely many base degrees for each fixed mm; establishing those base cases is still a mathematical task.

Organisation. § 2 develops the identities and their three constructions. § 3 proves the transfer principle, and § 4 gives the lower bound. Section 5 establishes the quotient hierarchies and § 6 is dedicated to the proof of the announced Davenport values. § 7 records the computational dependencies and the formal verification, with the coverage arguments and certificate specifications in Appendix § 8.

AI disclosure. Large language models (OpenAI’s Codex and ChatGPT, Anthropic’s Claude) were used in an essential way for this work: in the exploration that led to the incidence identities, which play a key role in the main results of this paper, in coding the enumeration programs, in writing and revising the exposition and the proofs, and in a substantial part of the Lean formalisation. The authors made a significant effort to improve the exposition of the AI-produced mathematics and to give proper credit to prior work that uses similar ideas. Every argument was check independently by at least one author and the author take full responsibility for the paper.

Incidence identities and three constructions

For a sequence U=(u1,…,uN)U=(u_1,\ldots,u_N) write σ(U)=∑j=1Nuj\sigma(U)=\sum_{j=1}^{N}u_j; the notation xkx^k denotes kk occurrences of xx. A subsequence is specified by its set of positions.

Lemma 2.1. Let (cE)(c_E) be a family of elements of RR indexed by Z(U)\mathcal{Z}(U).

  1. (cE)(c_E) is an Lt\mathcal{L}_t-identity over RR if and only if

∑E∈Z(U)cEP(1E)=P(0)(5)\sum_{E\in\mathcal{Z}(U)}c_E P(1_E)=P(0) \tag*{(5)}

for every polynomial P∈R[z1,…,zN]P\in R[z_1,\ldots,z_N] of degree at most tt.

  1. An Lt\mathcal{L}_t-identity over Z\mathbb{Z} is an Lt\mathcal{L}_t-identity over every commutative ring, by reduction of the coefficients.

  1. If R=KR=K is a field, UU has an Lt\mathcal{L}_t-identity over KK if and only if UU has no separator of degree at most tt over KK.

Proof. Both sides of (5) are linear in PP. On Boolean inputs, every polynomial agrees with a multilinear polynomial of no greater degree, so it is enough to test the monomials zJ=∏j∈Jzjz_J=\prod_{j\in J}z_j with ∣J∣≤t|J|\le t, including z∅=1z_{\varnothing}=1. For these monomials the equality is ∑E⊇JcE=[J=∅]\sum_{E\supseteq J}c_E=[J=\varnothing], exactly the moment conditions.

Part (2) is clear, because we always have a ring morphism Z→R\mathbb{Z}\to R.

For (3), consider the vectors mE=(1EJ)∣J∣≤t∈K(N≤t)m_E=(1_{E}^{J})_{|J|\le t}\in K^{\binom{N}{\le t}} of monomial values, for E∈Z(U)E\in\mathcal{Z}(U), and the vector m∅=(1,0,…,0)m_{\varnothing}=(1,0,\ldots,0) of the empty set. Observe that an identity is an expression of m∅m_{\varnothing} as a linear combination of the mEm_E, and a separator is a linear functional vanishing on all mEm_E and equal to 11 on m∅m_{\varnothing}. Over a field, either m∅m_{\varnothing} belongs to the span of the vectors mEm_E, or there is a linear functional vanishing on that span and taking the value 11 at m∅m_{\varnothing}. The first alternative gives the identity; the second gives the separator.

An identity on a subsequence extends to the full sequence by assigning zero weight to the other zero sets. Conversely, a separator for the full sequence restricts to one for a subsequence by setting the omitted coordinates to zero. Thus existence of an identity, and absence of a separator, are inherited by longer sequences. The invariant κt\kappa_t is finite: (t+1)D(G)(t+1)D(G) terms contain t+1t+1 disjoint zero sets, and Lemma 2.2(1) supplies an identity.

The identities that Section 5 obtains by computer are stored with integer weights. Reducing their coefficients modulo pp gives the identities needed for Theorem A, so one exact certificate check serves every characteristic. Solvability over each field separately would give neither the integral assertion of Theorem B nor such a common certificate.

We use three ways to build identities. Disjoint zero sets give inclusion–exclusion. Adjoining a disjoint zero set turns a finite difference into an identity of one higher degree. Taking complements inside a zero set cancels the highest-degree terms in a different way.

Lemma 2.2 (Three constructions). Let UU be an indexed sequence and t≥0t\ge0. The following produce integral Lt\mathcal{L}_t-identities for UU from integral identities of lower degree.

  1. If UU has t+1t+1 pairwise disjoint zero sets, then UU has an Lt\mathcal{L}_t-identity.

  1. Let t≥1t \ge1. If B∈Z(U)B \in\mathcal{Z}(U) and a subsequence of UU disjoint from BB has an It−1I_{t-1}-identity, then UU has an ItI_t-identity.

  1. Let t≥1t \ge1. If Z∈Z(U)Z \in\mathcal{Z}(U) and a proper subsequence T⊊ZT \subsetneq Z has an It−1I_{t-1}-identity, then UU has an ItI_t-identity.

Proof. For (1), let B1,…,Bt+1B_1,\ldots,B_{t+1} be disjoint nonempty zero sets. Give the union of every nonempty collection of kk blocks the weight (−1)k+1(-1)^{k+1}, and give every other zero set weight zero. Distinct collections have distinct unions. If a nonempty set JJ of at most tt positions is contained in such a union, it misses some block BiB_i. The collections whose union contains JJ are paired by adding or removing BiB_i, with opposite weights. Thus every required nonconstant moment vanishes. The total weight is ∑k=1t+1(−1)k+1(t+1k)=1\sum_{k=1}^{t+1}(-1)^{k+1}\binom{t+1}{k}=1.

For (2), let (cE)(c_E) be the identity of degree t−1t-1 on a subsequence U′U' disjoint from BB. Define new weights by

dE=cE,dE∪B=−cE(E∈Z(U′)),dB=1,d_E=c_E,\qquad d_{E\cup B}=-c_E\quad(E\in\mathcal{Z}(U')),\qquad d_B=1,

and set the remaining weights to zero. These sets are nonempty zero sets and are pairwise distinct. Let PP be multilinear of degree at most tt. Restricting to the coordinates outside BB, the polynomial

P(z+1B)−P(z)P(z+\mathbf{1}_B)-P(z)

has degree at most $t-1. Indeed, terms involving no coordinate of BB cancel, and each remaining term has spent at least one degree on a coordinate of BB. Applying the old identity gives

∑E∈Z(U′)cE(P(1E∪B)−P(1E))=P(1B)−P(0).\sum_{E\in\mathcal{Z}(U')}c_E\bigl(P(\mathbf{1}_{E\cup B})-P(\mathbf{1}_E)\bigr)=P(\mathbf{1}_B)-P(0).

Rearranging yields ∑FdFP(1F)=P(0)\sum_F d_FP(\mathbf{1}_F)=P(0).

For (3), let (cE)(c_E) be the identity on T⊊ZT\subsetneq Z, and put s=(−1)ts=(-1)^t. Give weight cEc_E to EE, weight −scE-sc_E to Z∖EZ\setminus E, and weight ss to ZZ, with zero weights elsewhere. The complements are nonempty zero sets. The assigned sets are distinct: each complement contains Z∖TZ\setminus T, whereas each EE lies in TT; and ZZ is neither an EE nor the complement of a nonempty EE. Restrict a multilinear polynomial PP of degree at most tt to the coordinates of ZZ. Then

RP(z)=P(z)−sP(1Z−z)R_P(z)=P(z)-sP(\mathbf{1}_Z-z)

has degree at most $t-1: complementation multiplies each degree-tt term by (−1)t(-1)^t, so the degree-tt parts cancel. Applying the identity on TT to RPR_P and rearranging gives

∑EcEP(1E)−s∑EcEP(1Z∖E)+sP(1Z)=P(0).\sum_E c_EP(\mathbf{1}_E)-s\sum_E c_EP(\mathbf{1}_{Z\setminus E})+sP(\mathbf{1}_Z)=P(0).

All three constructions use only addition and subtraction of coefficients, so they preserve integrality.

The transfer principle

For a cyclic kernel H=CpH=C_p, identified with the additive group of Fp\mathbb{F}_p, the required indicator is familiar. Fermat’s theorem gives

P(z)=1−(∑jvjzj)p−1.P(z)=1-\left(\sum_j v_jz_j\right)^{p-1}.

At a Boolean vector z=1Ez=\mathbf{1}_E, this is 1 precisely when ∑j∈Evj=0\sum_{j\in E}v_j=0, and it has degree at most $p-1, however many labels there are. Multiplying one such factor for each coordinate gives the same phenomenon for CprC_p^r. The next lemma extends it to arbitrary abelian pp-groups, including cyclic factors of order pap^a.

Lemma 3.1. Let H=⨁i=1rCpaiH = \bigoplus_{i=1}^{r} C_{p^{a_i}} and M=∑i(pai−1)M = \sum_i(p^{a_i}-1). For arbitrary v1,…,vN∈Hv_1,\ldots,v_N \in H, the Boolean function

f(z)=1{∑jzjvj=0}∈Fp,z∈{0,1}N,f(z)=\mathbf{1}_{\{\sum_j z_jv_j=0\}}\in\mathbb{F}_p,\qquad z\in\{0,1\}^{N},

is represented by a multilinear polynomial over Fp\mathbb{F}_p of degree at most MM.

Proof. Write XhX^h for the basis element of Fp[H]\mathbb{F}_p[H] indexed by h∈Hh\in H, so that XaXb=Xa+bX^aX^b=X^{a+b}. Choose generators eie_i of the cyclic factors, with eie_i of order paip^{a_i}, and put Yi=Xei−1Y_i=X^{e_i}-1. Then

Fp[H]≅Fp[Y1,…,Yr]/(Y1pa1,…,Yrpar).\mathbb{F}_p[H]\cong\mathbb{F}_p[Y_1,\ldots,Y_r]/(Y_1^{p^{a_1}},\ldots,Y_r^{p^{a_r}}).

and the augmentation ideal I=(Y1,…,Yr)I=(Y_1,\ldots,Y_r) satisfies IM+1=0I^{M+1}=0, since a monomial of total degree M+1M+1 in the YiY_i has some exponent at least paip^{a_i}. Now we consider the multilinear polynomial

P(z)=[X0]∏j=1N(1+zj(Xvj−1)).P(z)=[X^0]\prod_{j=1}^{N}(1+z_j(X^{v_j}-1)).

where [X0][X^0] extracts the identity coefficient in the group basis. Multiplication records the selected sum: setting zj=0z_j=0 omits vjv_j, and setting zj=1z_j=1 includes it. For a Boolean vector 1E\mathbf{1}_E the product collapses to X∑j∈EvjX^{\sum_{j\in E}v_j}, so P(1E)=f(1E)P(\mathbf{1}_E)=f(\mathbf{1}_E). The coefficient of zJz_J in PP is [X0]∏j∈J(Xvj−1)[X^0]\prod_{j\in J}(X^{v_j}-1). Before taking [X0][X^0], this is a product of ∣J∣|J| elements of the augmentation ideal, and therefore lies in I∣J∣I^{|J|}. Once ∣J∣>M|J|>M, that product, and hence the corresponding scalar coefficient of PP, is zero. This is why the degree is bounded by MM rather than by the number NN of labels. Thus the degree of our polynomial PP is at most MM. □

Olson’s theorem. The same nilpotence gives D(H)=M+1D(H)=M+1. If M+1M+1 terms were zero-sum free, the product ∏j(1−Xvj)\prod_j(1-X^{v_j}) would have identity coefficient 11, contributed only by the empty selection. But this product lies in IM+1=0I^{M+1}=0. Hence D(H)≤M+1D(H)\le M+1. Conversely, pai−1p^{a_i}-1 copies of each eie_i form a zero-sum free sequence of length MM, giving equality. For this basis sequence, only the empty subset has sum zero. Its unique multilinear indicator is ∏j=1M(1−zj)\prod_{j=1}^{M}(1-z_j), of degree MM. Thus the indicator degree bound is best possible in general.

Proof of Theorem A. Put M=D(H)−1M=D(H)-1 and take L(M)L(M) terms (uj,vj)(u_j,v_j) in Q⊕H\mathbb{Q}\oplus H. The quotient sequence (uj)(u_j) has an IM\mathcal{I}_M-identity (cE)(c_E) over Fp\mathbb{F}_p. Let PP be the indicator polynomial for the kernel labels (vj)(v_j) supplied by Lemma 3.1. Since deg⁡P≤M\deg P\le M,

1=P(0)=∑E∈Z((uj))cEP(1E).1=P(0)=\sum_{E\in\mathcal{Z}((u_j))}c_EP(\mathbf{1}_E).

Some summand is nonzero. For its nonempty set EE, ∑j∈Euj=0\sum_{j\in E}u_j=0 by definition, while P(1E)=1P(\mathbf{1}_E)=1 gives ∑j∈Evj=0\sum_{j\in E}v_j=0. Thus the original sequence has a nonempty zero sum, and D(Q⊕H)≤L(M)D(\mathbb{Q}\oplus H)\le L(M). If the identities are integral, reducing their coefficients modulo pp makes the same argument available for every prime. □

Remark 3.2 (The elementary block bound). A sequence of length η(Q)+texp⁡(Q)\eta(Q)+t\exp(Q) contains t+1t+1 disjoint short zero sets, obtained by repeated extraction. Their unions, with inclusion–exclusion weights, give an It\mathcal{I}_t-identity by Lemma 2.2(1). For a pp-group kernel, Theorem A therefore recovers (2), and in the same way every bound obtained from D(H)D(H) disjoint blocks, whatever their sizes. Incidence identities allow further cancellations among zero sets that do not arise from a disjoint block decomposition. The six-term Fano example below already requires such overlap.

Remark 3.3 (The kernel hypothesis). The degree bound in Lemma 3.1 cannot hold uniformly in the number of labels when HH is not a pp-group. Choose an element gg of prime order q≠pq \ne p and take all N=psN = p^s labels equal to gg. The coefficient of z1⋯zNz_1 \cdots z_N in their zero-sum indicator is

[X0](Xg−1)N=[X0](XNg−1)=−1≠0,[X^0](X^g-1)^N=[X^0](X^{Ng}-1)=-1\ne0,

where the middle equality is Frobenius and Ng≠0Ng \ne0 because q∤Nq \nmid N. A function on {0,1}N\{0,1\}^N has a unique multilinear representative over Fp\mathbb{F}_p, so the indicator has degree exactly NN. Since ss is arbitrary, there is no bound depending only on HH.

For example, when gg has order two and pp is odd, the indicator is

12(1+∏j=1N(1−2zj)).\frac{1}{2}\left(1+\prod_{j=1}^{N}(1-2z_j)\right).

the test that the number of selected labels is even. Thus the bounded-degree indicator argument cannot be applied unchanged to such kernels.

Sharpness

Proposition 4.1. Let m≥2m \ge2 and t≥0t \ge0, and let KK be a field of characteristic zero or of characteristic not dividing mm. Then

κt(Cm3;K)≥m(t+3)−2.\kappa_t(C_m^3;K)\ge m(t+3)-2.

Proof. We choose a sequence whose zero sets are especially simple:

U=e1m−1e2m−1e3m(t+1)−1,U=e_1^{m-1}e_2^{m-1}e_3^{m(t+1)-1},

of length m(t+3)−3m(t+3)-3. A zero set contains no first- or second-axis term, because there are fewer than mm of each. It therefore consists of third-axis positions, and its size is mkmk for some 1≤k≤t1\le k\le t.

We look for a separator among the linear combinations of the elementary symmetric polynomials s0,…,sts_0,\ldots,s_t in those third-axis coordinates. On a set of size ℓ\ell, their values are sj(1E)=(ℓj)s_j(1_E)=\binom{\ell}{j}. It suffices to show that the integer matrix

B=((mkj))0≤k,j≤tB=\left(\binom{mk}{j}\right)_{0\le k,j\le t}

is invertible over KK.

To compute its determinant, let aj,ia_{j,i} count the jj-element subsets of a fixed union of ii disjoint mm-element blocks that meet every block. Put a0,0=1a_{0,0}=1 and aj,0=0a_{j,0}=0 for j>0j>0. A jj-element subset of a union of kk blocks first chooses the ii blocks it meets, giving

(mkj)=∑i=0k(ki)aj,i.\binom{mk}{j}=\sum_{i=0}^{k}\binom{k}{i}a_{j,i}.

Thus B=STB=ST, where S=((ki))0≤k,i≤tS=\left(\binom{k}{i}\right)_{0\le k,i\le t} is lower triangular with diagonal entries one, and T=(aj,i)0≤i,j≤tT=(a_{j,i})_{0\le i,j\le t} is upper triangular. Moreover, aj,j=mja_{j,j}=m^j: a jj-element subset meeting jj blocks chooses one element from each. Consequently

det⁡B=∏j=0tmj=mt(t+1)/2.\det B=\prod_{j=0}^{t}m^j=m^{t(t+1)/2}.

This identity was obtained over the integers without division. The determinant has nonzero image in KK under the stated characteristic hypothesis. We can therefore choose a linear combination of s0,…,sts_0,\ldots,s_t taking the values 1,0,…,01,0,\ldots,0 at subset sizes 0,m,…,tm0,m,\ldots,tm. It is a separator for UU of degree at most tt. Hence κt(Cm3;K)≥m(t+3)−2\kappa_t(C_m^3;K)\ge m(t+3)-2.

Remark 4.2. The restriction on the characteristic cannot be dropped, and the smallest case shows the threshold moving: κ1(C23;F2)=5\kappa_1(C_2^3;\mathbb{F}_2)=5, one less than the value 6 that Theorem B gives over every field of characteristic different from two. Indeed, for five terms u1,…,u5u_1,\ldots,u_5 over C23=F23C_2^3=\mathbb{F}_2^3 the linear map F25→F23\mathbb{F}_2^5\to\mathbb{F}_2^3, z↦∑jzjujz\mapsto\sum_j z_j u_j, has a null space of dimension at least two, and its nonzero elements are the incidence vectors of the zero sets. If xx and yy are two independent vectors of this null space, an affine separator 1+ℓ1+\ell with ℓ\ell linear would have to satisfy ℓ(x)=ℓ(y)=ℓ(x+y)=1\ell(x)=\ell(y)=\ell(x+y)=1, which is impossible; so no five-term sequence has an affine separator over F2\mathbb{F}_2. On the other hand e1e2e32e_1e_2e_3^2 has the affine separator 1+z31+z_3, its only zero set being {3,4}\{3,4\}. Thus κ1(C23;F2)=5\kappa_1(C_2^3;\mathbb{F}_2)=5. The integral six-term identity supplied by Theorem B remains valid in characteristic two; the least length is simply smaller there.

The quotients C23C_2^3, C33C_3^3 and C43C_4^3

We now prove that every sequence of Lm(t)=m(t+3)−2L_m(t)=m(t+3)-2 terms over Cm3C_m^3 has an integral It\mathcal{I}_t-identity. Call a sequence short-zero-free if it has no nonempty zero set of size at most the exponent. For t≥1t\ge1, if a sequence of length Lm(t)L_m(t) has a zero set BB with ∣B∣≤m|B|\le m, deleting it leaves at least

Lm(t)−m=Lm(t−1)L_m(t)-m=L_m(t-1)

terms. An identity of degree t−1t-1 on the remainder therefore gives one of degree tt by Lemma 2.2(2). We call this step the short-zero recurrence. Once Lm(t)≥η(Cm3)L_m(t)\ge\eta(C_m^3), it is automatic. The work is concentrated in the degrees before that point.

Identities of degree one and two are tested on affine and on quadratic polynomials, and we call them affine and quadratic identities. The following table summarises where the finite work lies; the computations behind the finite certificates are described in Appendix § 8, and sharpness is Proposition 4.1.

The case C23C_2^3

The Fano configuration shows the mechanism in miniature. Degree zero is D(C23)=4D(C_2^3)=4. For degree one, consider six terms. If a term is zero or repeated, there is a zero set of at most two positions; deleting it leaves at least four terms and hence a second, disjoint zero set. Lemma 2.2(1) applies.

Otherwise the six terms are distinct nonzero vectors of F23\mathbb{F}_2^3: the Fano plane with one point xx removed. Its lines are the triples {u,v,u+v}\{u,v,u+v\}, each of which has sum zero. Give weight 1 to the four lines not through xx, and weight −1-1 to the complements, in the seven-point plane, of the three lines through xx. The latter are zero quadruples and do not contain xx. Each remaining point lies on two of the selected lines and in two of the selected quadruples. Its total weight is therefore 2−2=02-2=0, whereas the total weight of the seven sets is 4−3=14-3=1.

This identity uses overlap in an essential way. Every zero set among six distinct nonzero vectors has at least three terms. Two disjoint zero sets would therefore exhaust the six points, but their total is the omitted nonzero point xx, since all seven Fano points sum to zero.

Finally, any eight terms contain zero or a repetition, so η(C23)≤8\eta(C_2^3)\le8. The short-zero recurrence gives all degrees t≥2t\ge2 at length 2t+42t+4. These are integral identities, valid after reduction in every characteristic; Proposition 4.1 gives equality in characteristic zero and in odd characteristic.

The case C33C_3^3

We need the ten-term incidence identity and two elementary facts about supports.

Lemma 5.1 (Ternary base facts). The following assertions hold.

  1. Every ten-term sequence over C33C_3^3 has an integral L1\mathcal{L}_1-identity.

  2. The support of a short-zero-free sequence over C33C_3^3 has at most eight points. Every such support with seven points has nonzero sum, and every such support with eight points has sum zero.

We start by proving the second assertion.

Proof of Lemma 5.1(2). Let AA be the support of a short-zero-free sequence and put C=A∪{0}C=A\cup\{0\}. Then CC is a cap, that is, it contains no three distinct collinear points (this is the connection with affine caps studied in [6]). Indeed, over F3\mathbb{F}_3, three distinct points are collinear exactly when their sum is zero. A line through 00 would give an opposite pair in AA, and any other line would give a zero triple in AA.

An affine plane contains at most four cap points. Otherwise choose five. In each of the four directions, the three parallel lines would contain 2,2,12,2,1 of these points, accounting for two pairs in that direction. This accounts for at most eight pairs, rather than (52)=10\binom{5}{2}=10.

Put N=∣C∣N=|C| and let nin_i be the number of affine planes meeting CC in exactly ii points, for 0≤i≤40\leq i\leq4. There are (33−1)/2=13(3^3-1)/2=13 parallel classes of planes, each with three members, hence 39 planes. A point belongs to 13 planes, a pair to (32−1)/2=4(3^2-1)/2=4, and a noncollinear triple to one. Counting these incidences gives

∑i=04ni=39,∑i=04ini=13N,\sum_{i=0}^{4}n_i=39,\qquad\sum_{i=0}^{4}in_i=13N,
∑i=04(i2)ni=4(N2),∑i=04(i3)ni=(N3).\sum_{i=0}^{4}\binom{i}{2}n_i=4\binom{N}{2},\qquad\sum_{i=0}^{4}\binom{i}{3}n_i=\binom{N}{3}.

For N=10N=10, these equations give n3+4n0=−4n_3+4n_0=-4, a contradiction. A larger cap would contain a ten-point cap, so N≤9N\leq9 and ∣A∣≤8|A|\leq8.

Suppose ∣A∣=7|A|=7, so N=8N=8. The same equations give n2=6n0−2n_2=6n_0-2, hence some affine plane is disjoint from CC. Write it as ℓ=c\ell=c, where ℓ\ell is a nonzero linear functional. Since 0∈C0\in C, one has c≠0c\ne0. The other two parallel planes contain four points each, and therefore

ℓ(∑x∈Cx)=4∑a∈F3∖{c}a=−c≠0.\ell\left(\sum_{x\in C}x\right)=4\sum_{a\in\mathbb{F}_3\setminus\{c\}}a=-c\ne0.

Thus σ(A)=σ(C)≠0\sigma(A)=\sigma(C)\ne0.

Finally suppose ∣A∣=8|A|=8, so N=9N=9. No affine plane is empty, because the other two parallel planes could hold at most eight cap points. Thus n0=0n_0=0, and the incidence equations give n2=6n0=0n_2=6n_0=0. For every nonzero linear functional ℓ\ell, the three parallel slices have sizes 3,3,33,3,3 or a permutation of 1,4,41,4,4. In either case their sizes are congruent modulo three, so ℓ(σ(C))=0\ell(\sigma(C))=0. This holds for every ℓ\ell, hence σ(A)=σ(C)=0\sigma(A)=\sigma(C)=0. □

Proof of Lemma 5.1(1), computer-assisted. If a ten-term sequence has a zero set of at most three positions, its deletion leaves at least seven terms and hence a disjoint zero set, since D(C33)=7D(C_3^3)=7. Lemma 2.2(1) then gives the required affine identity.

It remains to consider the short-zero-free sequences. Their multiplicities are at most two, and their supports satisfy the cap conditions used above. The complete enumeration and its symmetry reductions are given in Appendix § 8. Each case is supplied either with two disjoint zero sets or with an explicit integral affine identity. The certificate check verifies the zero sums, total weight one, and zero total weight through each individual position. The transports between normalised representatives and the indexed sequences they represent are checked together with the certificates.

Call a set A⊆C33A \subseteq C_3^3 allowed if it contains neither zero nor a pair of opposite points, and no three distinct points of AA sum to zero. Write 2A2A for the sequence containing two copies of each point of AA.

Lemma 5.2. Every short-zero-free sequence of thirteen terms over C33C_3^3 has a zero set of size at least eleven.

Proof. The plan is to write σ(U)\sigma(U) as a sum of at most two terms of UU: deleting them leaves a zero set of size at least eleven. The representation will come from the maximality of an allowed support, and the only point to watch is whether it uses a term more often than UU contains it.

We first observe that if AA is an inclusion-maximal allowed support, then every x∈C33x \in C_3^3 is a sum of at most two terms of 2A2A. For x=0x=0 take no terms, for x∈Ax \in A one term, and if −x∈A-x \in A take two copies of −x-x. Otherwise −x-x is a nonzero point outside AA whose opposite xx is also outside AA, so adjoining −x-x to AA can only violate the third condition: there is a zero triple −x+a+b=0-x+a+b=0 with distinct a,b∈Aa,b \in A, which represents x=a+bx=a+b.

Now let UU be short-zero-free of length thirteen. Its multiplicities are at most two, so its support has seven or eight points, and there are three cases.

Case 1: the support is an eight-set BB. Then U=2B−b−c−dU=2B-b-c-d for three distinct points b,c,db,c,d of multiplicity one. By Lemma 5.1, σ(B)=0\sigma(B)=0, so σ(U)=−b−c−d\sigma(U)=-b-c-d, and BB is maximal by the support bound. A representation of σ(U)\sigma(U) by at most two terms of 2B2B is unavailable in UU, that is, it uses a term more often than UU contains it, only if it uses one of b,c,db,c,d twice, say 2b=−b−c−d2b=-b-c-d; this gives c+d=0c+d=0, impossible in an allowed support.

Case 2: the support is a seven-set AA which is maximal. Then U=2A−a0U=2A-a_0 for one point a0a_0 of multiplicity one, and the only unavailable representation is 2a0=σ(U)=2σ(A)−a02a_0=\sigma(U)=2\sigma(A)-a_0, which forces σ(A)=0\sigma(A)=0, contradicting Lemma 5.1.

Case 3: the support is a seven-set AA which is not maximal. Again U=2A−a0U=2A-a_0, and AA extends to an allowed eight-set B=A∪{b}B=A\cup\{b\}, which is maximal, with σ(B)=0\sigma(B)=0 and hence σ(U)=b−a0\sigma(U)=b-a_0. A representation of b−a0b-a_0 by at most two terms of 2B2B is unavailable in UU only if it uses bb, which UU does not contain, or a0a_0 twice, so it is one of

b,2b,b+c with c∈A,2a0,b,\qquad2b,\qquad b+c\ \text{with }c\in A,\qquad2a_0,

and these force a0=0a_0=0, a0=−ba_0=-b, c=−a0c=-a_0 and b=0b=0 respectively. Each contradicts the support conditions.

Proof of Theorem B for m=3m=3. The target lengths are 3t+73t+7. Degree zero is D(C33)=7\mathrm{D}(C_3^3)=7; degree one is Lemma 5.1(1).

For degree two, consider thirteen terms. A zero set of at most three positions leaves at least ten terms, so Lemma 2.2(2) applies to the affine identity on the remainder. Otherwise Lemma 5.2 gives a zero set of at least eleven positions. It properly contains a ten-term subsequence, and Lemma 2.2(3) raises that subsequence’s affine identity to degree two.

For degree three, a short zero set among sixteen terms again reduces to the preceding degree. If there is none, the support bound and the multiplicity bound force eight support points, each appearing twice. Their sum is zero by Lemma 5.1(2), so the whole sequence is a zero set. A thirteen-term subsequence properly inside it has a degree-two identity, which Lemma 2.2(3) raises to degree three.

The support and multiplicity bounds also show that every seventeen terms contain a short zero set: η(C33)≤17\eta(C_3^3)\leq17 (in fact η(C33)=17\eta(C_3^3)=17, see [9]). For t≥4t\geq4, the target length is at least nineteen, so the short-zero recurrence proves all further degrees. Proposition 4.1 gives sharpness under its characteristic hypothesis.

The case C43C_4^3

Lemma 5.3 (Quaternary finite input). Every fourteen-term sequence over C43C_4^3 has an integral L1\mathcal{L}_1-identity, and every eighteen-term sequence over C43C_4^3 has an integral L2\mathcal{L}_2-identity.

The proof is computer-assisted; its reductions are as follows. Since D(C43)=10\mathrm{D}(C_4^3)=10, a zero set of at most four positions among fourteen terms leaves a disjoint zero set. The two sets give an affine identity by Lemma 2.2(1). Each remaining fourteen-term case receives either a disjoint-pair witness or an explicit integral affine identity.

For eighteen terms, a short zero set reduces to the fourteen-term claim by Lemma 2.2(2). Three disjoint zero sets also suffice, by part (1). A zero set of at least fifteen positions gives a third route: it properly contains fourteen positions, whose affine identity lifts by part (3). The remaining cases are supplied with explicit integral quadratic identities.

Appendix § 8 gives the complete finite domains, the coverage arguments for two normalisations, and the counts of the positive witnesses. The checks verify the group sums and every required incidence moment over the integers. The enumeration includes multiplicities and accounts for sequences in proper subgroups; all transports preserve indexed occurrences.

The following example shows what the finite search can find beyond these length-based reductions. Take

U=(100)3(010)3(110)3(001)3(101)3(131)3U=(100)^3(010)^3(110)^3(001)^3(101)^3(131)^3

in (Z/4Z)3(\mathbb{Z}/4\mathbb{Z})^3, with positions numbered 1,…,181,\ldots,18 in this order.

Inspecting the first, third and second coordinates shows that a nonempty zero set of at most four terms would require four equal terms, which are unavailable. Also σ(U)=(0,3,1)\sigma(U)=(0,3,1). A zero set of at least fifteen terms would leave at most three terms with this sum. Their first coordinates force them to be 010 or 001, requiring three copies of 010 and one of 001, a contradiction. Three disjoint zero sets would have a zero union of at least fifteen terms, so they are impossible too.

There is nevertheless a useful five-term zero set

B={9,11,13,14,16}.B=\{9,11,13,14,16\}.

Deleting it leaves only thirteen terms, one fewer than the length at which an affine identity is guaranteed for every sequence. For this particular remainder, such an identity can be written down. The following five sets are zero sets disjoint from BB:

A={1,2,4,5,7,8},C={3,8,10,12,15,18},D={6,7,12,15,17,18},C′={4,5,6,7,8,10,12,15,18},D′={1,2,3,7,8,12,15,17,18}.\begin{aligned} A&=\{1,2,4,5,7,8\},\\ C&=\{3,8,10,12,15,18\},\\ D&=\{6,7,12,15,17,18\},\\ C'&=\{4,5,6,7,8,10,12,15,18\},\\ D'&=\{1,2,3,7,8,12,15,17,18\}. \end{aligned}

Position by position, they satisfy the integer identity

1A+1C+1D=1C′+1D′.\mathbf{1}_A+\mathbf{1}_C+\mathbf{1}_D=\mathbf{1}_{C'}+\mathbf{1}_{D'}.

The coefficients 1,1,1,−1,−11,1,1,-1,-1 add to one and cancel through every position, so they give an integral L1\mathcal{L}_1-identity on the remainder. Lemma 2.2(2) now produces an L2\mathcal{L}_2-identity on UU: keep the five weights, give their negatives to the unions of the five sets with BB, and give weight 1 to BB.

Thus the eleven-set quadratic certificate is a lower-degree identity lifted by one of the elementary constructions. The useful extra information is the affine identity on this particular thirteen-term remainder; its existence does not follow from the universal fourteen-term threshold. The certificate itself, in the form in which it is stored and checked, is reproduced in the supplement.

Proof of Theorem B for m=4m=4. The target lengths are 4t+104t+10. Degree zero is D(C43)=10D(C_4^3)=10, and degrees one and two are Lemma 5.3. The bound η(C43)≤22\eta(C_4^3)\le22 follows from the classical result η(C43)=22\eta(C_4^3)=22 of Harborth [15], see [9 Theorem 2.2(2)]. For t≥3t\ge3, the target length is at least 22; deleting a short zero set leaves at least 4(t−1)+104(t-1)+10 terms. Lemma 2.2(2) therefore proves every further degree. Proposition 4.1 gives sharpness under its characteristic hypothesis.

Davenport constants

Proof of Corollary C. Theorems A and B give

D(Cm3⊕H)≤m(D(H)−1)+3m−2(6)D(C_m^3\oplus H)\le m(D(H)-1)+3m-2 \tag*{(6)}

for every finite abelian pp-group HH. Take H=Cpa⊕Cpb⊕CpcH=C_{p^a}\oplus C_{p^b}\oplus C_{p^c}, allowing trivial factors. Olson’s theorem gives

D(H)−1=pa+pb+pc−3.D(H)-1=p^a+p^b+p^c-3.

Since p∤mp\nmid m, the Chinese remainder theorem identifies Cm3⊕HC_m^3\oplus H with G=Cmpa⊕Cmpb⊕CmpcG=C_{mp^a}\oplus C_{mp^b}\oplus C_{mp^c}. Its invariant factors are in divisibility order, so substitution yields

D(G)≤m(pa+pb+pc)−2=D∗(G).D(G)\le m(p^a+p^b+p^c)-2=D^*(G).

The basis lower bound gives the reverse inequality.

The general bound (6) holds for kernels of every rank and for every prime. Coprimality and the rank bound are used in the comparison with D∗D^* above.

For example, take m=3m=3 and H=C1⊕C5⊕C25H=C_1\oplus C_5\oplus C_{25}. Here D(H)−1=28D(H)-1=28, so the quotient identity is used in degree 28, at length 3(28+3)−2=913(28+3)-2=91. Transfer gives

D(C3⊕C15⊕C75)=91.D(C_3\oplus C_{15}\oplus C_{75})=91.

No enumeration at length 91 is needed: the constructions have already extended the finite base cases to degree 28. The basis sequence of length 2+14+74=902+14+74=90 shows why the resulting value is sharp.

Verification, reproducibility and dependencies

The proofs of Lemma 5.1(1) and Lemma 5.3 are computer assisted. The support statement in Lemma 5.1(2) follows from affine-plane counting, and the proof of Lemma 5.2 uses it. The later incidence degrees are deductions from the finite base identities by the three constructions. Table 1 records these dependencies. Olson’s small pp-group values for the Davenport constant D(C23)=4D(C_2^3)=4, D(C33)=7D(C_3^3)=7 and D(C43)=10D(C_4^3)=10 also supply degree zero and the short-zero reductions in the relevant rows.

QuotientTarget lengthBefore the automatic recurrenceAutomatic recurrence
C23C_2^32t+42t+4t=0t=0: D(C23)=4D(C_2^3)=4; t=1t=1: the Fano configurationt≥2t\ge2, using η(C23)≤8\eta(C_2^3)\le8
C33C_3^33t+73t+7t=0t=0: D(C33)=7D(C_3^3)=7; t=1t=1: finite affine certificates; t=2,3t=2,3: the support facts and the constructionst≥4t\ge4, using η(C33)≤17\eta(C_3^3)\le17
C43C_4^34t+104t+10t=0t=0: D(C43)=10D(C_4^3)=10; t=1,2t=1,2: finite certificatest≥3t\ge3, using η(C43)≤22\eta(C_4^3)\le22

Table 1.

For each computed case, the verification checks a positive witness: disjoint zero sets, a suitable long zero set, or an integer incidence identity. An incorrect negative classification, such as failing to detect a disjoint pair, cannot by itself certify a case; a valid positive witness must still pass the check. Appendix § 8 specifies the full finite domains and justifies the reductions used to enumerate them. The certificate verification uses exact arithmetic and does not rely on the optimisation software used to discover the weights. All certificates and the instructions for reproducing their results are collected in a separate Supplement available on our Github repository [18].

The definitions, the lemmas, Proposition 4.1, Theorems A and B and Corollary C have also been formalised in Lean 4 with Mathlib [18], with proofs that follow the ones given here. In developing the formalisation, we made use of Scott Armstrong’s LeanAutomatizationSkills collection [2]. For Lemma 5.1(1) and Lemma 5.3, the formalisation performs its own enumerations; their completeness is proved in Lean, together with the soundness of the certificate checks (see the supplement).

StatementNatureEssential inputs
Lemma 2.1 (duality)ordinaryBoolean multilinearisation; linear algebra over a field
Lemma 2.2 (three constructions)ordinaryLemma 2.1(1); inclusion–exclusion; two degree-cancellation arguments
Lemma 3.1 (indicator)ordinarytruncated group algebra; nilpotence of the augmentation ideal
Theorem Aordinaryevaluation identity; Lemma 3.1; Olson’s value of D(H)D(H)
Proposition 4.1 (sharpness)ordinaryinteger triangular factorisation of the binomial-value matrix
Theorem B, m=2m = 2ordinaryFano identity; η(C32)≤8\eta(C_3^2) \le8; Lemma 2.2; Proposition 4.1
Lemma 5.1(1)computer-assistedcomplete ten-term coverage (Appendix § 8); positive affine or disjoint-pair certificates and their transports
Lemma 5.1(2)ordinarycap–plane incidence counts; linear functionals
Lemma 5.2ordinarymaximal allowed supports; Lemma 5.1(2); multiplicity checks
Theorem B, m=3m = 3deduction using finite inputLemmas 5.1, 5.2, 2.2; D(C33)=7D(C_3^3) = 7; η(C33)≤17\eta(C_3^3) \le17; Proposition 4.1
Lemma 5.3computer-assistedtwo quaternary enumerations (Table 3); exact certificates; the proper-subgroup bound used by the support-basis enumeration
Theorem B, m=4m = 4deduction using finite inputLemmas 5.3, 2.2; D(C34)=10D(C_3^4) = 10; η(C34)≤22\eta(C_3^4) \le22; Proposition 4.1
Corollary Cdeduction using finite inputTheorems A and B; Olson’s theorem; Chinese remainder theorem; basis lower bound

Table 1 (PDF p. 14). Dependency map. “Ordinary” refers to the proof and its stated inputs; “computer-assisted” means an exhaustive enumeration with positive witnesses checked by an independent program; “deduction” means an ordinary argument some of whose inputs are computer-assisted statements.

QuantityLength 14Length 18
Generated by projection and lifts308,05118,224
Generated with a fixed support basis24,445,3771,267,272
Distinct linear orbits106,6554,941
Disjoint-zero witnesses284,99413,549
Long-zero witnesses (after disjoint cases)—4,171
Direct integer certificates23,057504
Orbits receiving direct certificates6,96364

Table 3. Quaternary finite coverage. Generated cases may repeat orbits; orbit counts refer to the full linear group.

Computational coverage and certificates

This appendix collects the parts of the proofs of Lemma 5.1(1) and Lemma 5.3 that justify the finite computations: the exact domains that were enumerated and the arguments showing that these domains cover every required sequence. Nothing is inferred from samples: each enumeration establishes a universal statement about all sequences of the stated length. The positive witnesses themselves, their statistics, the stored form of the eighteen-term example of Section 5 and the instructions for an independent replay are collected in the supplement.

Ternary coverage

A short-zero-free sequence over C33C_3^3 has all multiplicities at most two, and its support AA contains no zero, no pair of opposite points and no three distinct points with sum zero; equivalently, A∪{0}A \cup\{0\} is an affine cap in F33\mathbb{F}_3^3, a connection developed in [6]. A support with at least five points spans F33\mathbb{F}_3^3, since a plane contains only four pairs of opposite nonzero points. Mapping three independent support points to the standard basis leaves ten pairs of opposite points, each contributing none or exactly one of its two points, and the complete enumeration of these 310=59 0493^{10} = 59\,049 choices produces the support counts of Table 2; no support of size nine or more occurs. All 296 seven-point supports have nonzero sum and all 18 eight-point supports have sum zero. A separate enumeration by direct basis completion confirms these two facts. These are the three facts of Lemma 5.1(2); they are proved in Section 5 by counting affine planes, and the enumeration is a redundant check of them.

Support sizeSupportsCasesDisjoint pairOther identities
51081088127
62964,4402,7691,671
729610,3607,0523,308
8185045040
Total15,41210,4065,006

Table 2. The ternary ten-term computation, after normalising a support basis. Support counts are not orbit counts.

For Lemma 5.1(1), a zero set of size at most three leaves seven terms and hence a disjoint zero set, since D(C33)=7D(C_3^3)=7. It remains to treat the short-zero-free multiplicity vectors of total ten; their exact counts are given in Table 2. Most of them contain two disjoint zero sets. The remaining 5 006 cases form 43 orbits under the linear group, and each orbit representative receives an explicit integral identity with 5, 7 or 9 zero sets and coefficients ±1\pm1. An independent checker regenerates the complete case space and verifies every zero set and every moment of every identity, and a second one checks all transport matrices and index permutations between representatives and cases. No affine translations are used anywhere.

Quaternary coverage by reduction modulo two

Represent C43C_4^3 as (Z/4Z)3(\mathbb{Z}/4\mathbb{Z})^3 and reduce modulo two. In a short-zero-free sequence, the fibre over 0∈F230 \in\mathbb{F}_2^3 contains at most three terms, which must be distinct and independent in C43C_4^3. A nonzero fibre contains at most four distinct elements, since its eight elements form four pairs of opposite elements, and at most one of these elements can repeat, because two distinct repeated elements g,hg,h in the same fibre satisfy 2g+2h=02g+2h=0. Each multiplicity is at most three, so a fibre carries at most six terms.

We enumerate every projected multiplicity vector (μv)v∈F23(\mu_v)_{v\in\mathbb{F}_2^3} of the prescribed total with μ0≤3\mu_0\le3 and μv≤6\mu_v\le6 otherwise, up to the 168 elements of GL⁡3(F2)\operatorname{GL}_3(\mathbb{F}_2). Every such linear map lifts to an automorphism of (Z/4Z)3(\mathbb{Z}/4\mathbb{Z})^3, because a binary lift has odd determinant. Write a lifted element as v~+2w\tilde v+2w with v,w∈F23v,w\in\mathbb{F}_2^3, where v~\tilde v is the binary representative. The automorphisms that reduce to the identity modulo two, and their action on high parts, are

{I+2N:N∈M3(F2)},w↦w+Nv,\{I+2N:N\in M_3(\mathbb{F}_2)\}, \qquad w\mapsto w+Nv,

so choosing one term in each of the first independent occupied nonzero fibres, a suitable NN makes each chosen high part ww zero. Since zero is the least high part, requiring the first high part in each of these fibres to vanish, after sorting high parts within fibres, retains a representative of every orbit. This enumerator does not assume that the projection spans, so it covers sequences contained in proper subgroups as well.

Backtracking then generates all lifts in nondecreasing order within each fibre. It maintains the sets R1,R2,R3R_1,R_2,R_3 of sums of one, two and three previously placed terms with distinct indices. Appending xx is legal exactly when x≠0x\ne0 and −x∉R1∪R2∪R3-x\notin R_1\cup R_2\cup R_3, and the update is

R1′=R1∪{x},R2′=R2∪(R1+x),R3′=R3∪(R2+x),R_1'=R_1\cup\{x\},\qquad R_2'=R_2\cup(R_1+x),\qquad R_3'=R_3\cup(R_2+x),

with the old sets on the right. This is an exact test for short-zero-freeness that keeps track of repetitions.

Independent support-basis enumeration

A second enumeration normalises a support basis instead. An exhaustive computation in the subgroup C42⊕C2C_4^2 \oplus C_2 shows that short-zero-free sequences there have length at most thirteen. Every proper subgroup of C43C_4^3 lies in an index-two subgroup, all index-two subgroups are linearly equivalent, and therefore every short-zero-free sequence of length at least fourteen spans modulo two; three of its support vectors then have odd determinant and form a basis of (Z/4Z)3(\mathbb{Z}/4\mathbb{Z})^3. We generate all short-zero-free multisets containing the standard basis and canonicalise each by the least sorted image under the maps that send an ordered support basis to the standard basis; this is a complete invariant for linear equivalence. The two enumerations produce identical canonical orbit lists.

References

  1. [1]N. Alon, Combinatorial Nullstellensatz, Combin. Probab. Comput. 8 (1999), 7–29. doi:10.1017/S0963548398003411.DOI
  2. [2]S. Armstrong, LeanAutoformalizationSkills: Portable Lean 4 autoformalization skills for Codex and Claude Code, GitHub repository (2026). github.com/scott narmstrong/LeanAutoformalizationSkills.
  3. [3]G. Bhowmik, I. Halupczok, and J.-C. Schlage-Puchta, Inductive methods and zero-sum free sequences, Integers 9 (2009), A40, 515–536. doi:10.1515/INTEG.2009.042; arXiv:0711.1262.DOI
  4. [4]G. Bhowmik and J.-C. Schlage-Puchta, Davenport’s constant for groups of the form ℤ₃ ⊕ ℤ₃ ⊕ ℤ₃d, in: Additive Combinatorics, CRM Proc. Lecture Notes 43, Amer. Math. Soc., Providence, RI, 2007, pp. 307–326. arXiv:math/0610416.
  5. [5]C. Delorme, O. Ordaz, and D. Quiroz, Some remarks on Davenport constant, Discrete Math. 237 (2001), 119–128.DOI
  6. [6]Y. Edel, C. Elsholtz, A. Geroldinger, S. Kubertin, and L. Rackham, Zero-sum problems in finite abelian groups and affine caps, Q. J. Math. 58 (2007), 159–186. doi:10.1093/qmath/ham003.DOI
  7. [7]W. Gao, On Davenport’s constant of finite abelian groups with rank three, Discrete Math. 222 (2000), 111–124. doi:10.1016/S0012-365X(00)00010-8.DOI
  8. [8]W. Gao and A. Geroldinger, Zero-sum problems in finite abelian groups: a survey, Expo. Math. 24 (2006), 337–369. doi:10.1016/j.exmath.2006.07.002.DOI
  9. [9]W. D. Gao, Q. H. Hou, W. A. Schmid, and R. Thangadurai, On short zero-sum subsequences. II, Integers 7 (2007), Paper A21. doi:10.5281/zenodo.8281159.DOI
  10. [10]A. Geroldinger, D. J. Grynkiewicz, and Q. Zhong, Combinatorial Factorization Theory. Factorization Theory of Krull Monoids via Additive Combinatorics, Math. Surveys Monogr. 296, Amer. Math. Soc., Providence, RI, 2026. doi:10.1090/surv/296.
  11. [11]A. Geroldinger and R. Schneider, On Davenport’s constant, J. Combin. Theory Ser. A 61 (1992), 147–152. doi:10.1016/0097-3165(92)90061-X.DOI
  12. [12]A. Geroldinger and W. Yang, On a classical zero-sum invariant, preprint (2026). arXiv:2608.19090v1.arxiv.org/abs/2608.19090
  13. [13]B. Girard, An asymptotically tight bound for the Davenport constant, J. Éc. polytech. Math. 5 (2018), 605–611. doi:10.5802/jep.79.DOI
  14. [14]B. Girard and W. A. Schmid, Direct zero-sum problems for certain groups of rank three, J. Number Theory 197 (2019), 297–316. doi:10.1016/j.jnt.2018.08.016.DOI
  15. [15]H. Harborth, Ein Extremalproblem für Gitterpunkte, J. Reine Angew. Math. 262/263 (1973), 356–360. doi:10.1515/crll.1973.262-263.356.DOI
  16. [16]J. E. Olson, A combinatorial problem on finite Abelian groups. I, J. Number Theory 1 (1969), 8–10. doi:10.1016/0022-314X(69)90021-3.DOI
  17. [17]J. E. Olson, A combinatorial problem on finite Abelian groups. II, J. Number Theory 1 (1969), 195–199. doi:10.1016/0022-314X(69)90037-7.DOI
  18. [18]C. G. Pop, G. C. Rad, and G. C. Turcaș, Davenport_Project: a Lean 4 formalization of “Integral zero-sum incidence identities and Davenport constants in rank three”, 2026. https://github.com/CiprianRad/Davenport_Project.
  19. [19]W. A. Schmid, The inverse problem associated to the Davenport constant for C₂⊕C₂⊕C₂ₙ, and applications to the arithmetical characterization of class groups, Electron. J. Combin. 18(1) (2011), Paper P33. doi:10.37236/520.DOI
  20. [20]P. van Emde Boas, A combinatorial problem on finite abelian groups. II, Report ZW 7/69, Mathematisch Centrum, Amsterdam, 1969. CWI institutional repository.
  21. [21]P. van Emde Boas and D. Kruyswijk, A combinatorial problem on finite abelian groups. III, Report ZW 8/69, Mathematisch Centrum, Amsterdam, 1969. CWI institutional repository.
  22. [22]A. Volkmann, The small Davenport constant of H₂₇×C₃ʳ, preprint (2026). arXiv:2609.19810.
  23. [23]A. Volkmann, A twelve-term exclusion for the small Davenport constant of E₂×C₃ʳ, preprint (2026). arXiv:2609.22972.
  24. [24]A. Volkmann, The small Davenport constant of E₂ × C₃ʳ for 0 ≤ r ≤ 3, preprint (2026). arXiv:2609.37262.

Paper details

vlnc:W_gewyfvqv
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents