Lattices of partial sums

In this paper we introduce and study a class of partially ordered sets that can be interpreted as partial sums of indeterminate real numbers. An important example of these partially ordered sets, is the classical Young lattice Y of the integer partitions. In this context, the sum function associated...

Повний опис

Збережено в:
Бібліографічні деталі
Опубліковано в: :Algebra and Discrete Mathematics
Дата:2014
Автори: Chiaselotti, G., Gentile, T., Oliverio, P.A.
Формат: Стаття
Мова:English
Опубліковано: Інститут прикладної математики і механіки НАН України 2014
Онлайн доступ:https://nasplib.isofts.kiev.ua/handle/123456789/153329
Теги: Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Цитувати:Lattices of partial sums / G. Chiaselotti, T. Gentile, P.A. Oliverio // Algebra and Discrete Mathematics. — 2014. — Vol. 18, № 2. — С. 171–185. — Бібліогр.: 23 назв. — англ.

Репозитарії

Digital Library of Periodicals of National Academy of Sciences of Ukraine
id nasplib_isofts_kiev_ua-123456789-153329
record_format dspace
spelling Chiaselotti, G.
Gentile, T.
Oliverio, P.A.
2019-06-14T03:21:25Z
2019-06-14T03:21:25Z
2014
Lattices of partial sums / G. Chiaselotti, T. Gentile, P.A. Oliverio // Algebra and Discrete Mathematics. — 2014. — Vol. 18, № 2. — С. 171–185. — Бібліогр.: 23 назв. — англ.
1726-3255
2010 MSC:Primary: 06A06, Secondary: 06A07, 05A17.
https://nasplib.isofts.kiev.ua/handle/123456789/153329
In this paper we introduce and study a class of partially ordered sets that can be interpreted as partial sums of indeterminate real numbers. An important example of these partially ordered sets, is the classical Young lattice Y of the integer partitions. In this context, the sum function associated to a specific assignment of real values to the indeterminate variables becomes a valuation on a distributive lattice.
en
Інститут прикладної математики і механіки НАН України
Algebra and Discrete Mathematics
Lattices of partial sums
Article
published earlier
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
collection DSpace DC
title Lattices of partial sums
spellingShingle Lattices of partial sums
Chiaselotti, G.
Gentile, T.
Oliverio, P.A.
title_short Lattices of partial sums
title_full Lattices of partial sums
title_fullStr Lattices of partial sums
title_full_unstemmed Lattices of partial sums
title_sort lattices of partial sums
author Chiaselotti, G.
Gentile, T.
Oliverio, P.A.
author_facet Chiaselotti, G.
Gentile, T.
Oliverio, P.A.
publishDate 2014
language English
container_title Algebra and Discrete Mathematics
publisher Інститут прикладної математики і механіки НАН України
format Article
description In this paper we introduce and study a class of partially ordered sets that can be interpreted as partial sums of indeterminate real numbers. An important example of these partially ordered sets, is the classical Young lattice Y of the integer partitions. In this context, the sum function associated to a specific assignment of real values to the indeterminate variables becomes a valuation on a distributive lattice.
issn 1726-3255
url https://nasplib.isofts.kiev.ua/handle/123456789/153329
citation_txt Lattices of partial sums / G. Chiaselotti, T. Gentile, P.A. Oliverio // Algebra and Discrete Mathematics. — 2014. — Vol. 18, № 2. — С. 171–185. — Бібліогр.: 23 назв. — англ.
work_keys_str_mv AT chiaselottig latticesofpartialsums
AT gentilet latticesofpartialsums
AT oliveriopa latticesofpartialsums
first_indexed 2025-11-25T23:08:43Z
last_indexed 2025-11-25T23:08:43Z
_version_ 1850581515379933184
fulltext Algebra and Discrete Mathematics RESEARCH ARTICLE Volume 18 (2014). Number 2, pp. 171–185 © Journal “Algebra and Discrete Mathematics” Lattices of partial sums Giampiero Chiaselotti, Tommaso Gentile, Paolo Antonio Oliverio Communicated by V. V. Kirichenko Abstract. In this paper we introduce and study a class of partially ordered sets that can be interpreted as partial sums of indeterminate real numbers. An important example of these partially ordered sets, is the classical Young lattice Y of the integer partitions. In this context, the sum function associated to a specific assignment of real values to the indeterminate variables becomes a valuation on a distributive lattice. Introduction In the present paper we build a formal order context within which we examine the analogies that arise between combinatorial sum problems on multisets of real numbers and combinatorial set problems. Our con- struction provides a class of distributive lattices that we will call partial sums lattices. The elements of these lattices correspond to indeterminate partial sums of real numbers and any effective assignment of real values to such indeterminate becomes a valuation on these lattices (classical results concerning the valuations on a distributive lattice and their links with Euler characteristic and Mobius function can be found in [10–12,20]). For a wide range of extremal combinatorial problems concerning partial sums on real numbers we refer to [2, 14–18]. In particular, a beautiful example of similarity between combinatorial sum problems and combinatorial set 2010 MSC: Primary: 06A06, Secondary: 06A07, 05A17. Key words and phrases: Distributive Lattices, Real Partial Sums, Integer Par- titions. 172 Lattices of partial sums problems is a still unsolved conjecture raised in [16], which has a dual formulation (in terms of partial sums on multi-sets of real numbers) of the famous theorem of Erdös-Ko-Rado [9]. This conjecture and related problems have been studied in [2–8,14,17–19,21,23]. In [16] have been raised several combinatorial problems concerning partial sums on multi- sets of n real numbers, and it is interesting to observe as such problems have a strict analogy with corresponding problems concerning families of subsets of the set of indexes {1, . . . , n} (see [3, 4, 16, 23] for a more detailed description of some such analogies). If n is a positive integer we set [n] = {1, . . . , n}. We denote by P([n]) the power set of [n] and we set Pk([n]) = {A ∈ P([n]) : |A| = k}, for each k = 0, 1, . . . , n. In this context we call indexes the element of [n]. Let xn, . . . , x1 be n indeterminate real numbers (not necessarily distinct) such that xn > xn−1 > · · · > x2 > x1 (1) Our basic motivation is to investigate the analogies between order proper- ties on the subsets of indexes A ∈ P([n]) and on the partial sums ∑ i∈A xi, where each xi (i ∈ A) satisfies (1). For example, it is obviously false that if A, B ∈ P([n]) and A ⊆ B then ∑ i∈A xi 6 ∑ i∈B xi, since in (1) can also be negative numbers. We refine then the previous question as follows. Let us suppose that our indeterminate real numbers satisfy the condition xn > xn−1 > · · · > x2 > x1 > 0 (2) Then, under the hypothesis (2), it is true that if A, B ∈ P([n]) and A ⊆ B then ∑ i∈A xi 6 ∑ i∈B xi. But the reverse implication is not true, i.e. if for each assignment of real values in (2) we have that ∑ i∈A xi 6 ∑ i∈B xi, then not necessarily A ⊆ B. For example, for each assignment that satisfies (2) we have x2 > x1, but {2} + {1}. We ask therefore the following question: does there exist a partial order ⊑ on P([n]) such that A ⊑ B iff ∑ i∈A xi 6 ∑ i∈B xi for each assignment of n real values that satisfies (2)? More generally, if xn > · · · > xn−r+1 > 0 > xn−r > · · · > x1 > 0 (3) does there exist a partial order ⊑ on P([n]) such that A ⊑ B iff ∑ i∈A xi 6 ∑ i∈B xi for each assignment of n real values that satisfies (3)? In this paper we reformulate and study the previous question in a more general form, when i.e. the set [n] of the indexes is substituted with an arbitrary total ordered set I and with an its fixed partition (I+, I−) G. Chiaselotti, T. Gentile, P. A. Oliverio 173 in two blocks: the elements of I+ will be thought of as "positive" indexes and that of I− of as "negative" indexes. In such an abstract context we obtain a class of posets, which we call partial sums posets on I. On such posets we define a particular class of order-preserving maps which are the natural generalizations of the partial sums ∑ i∈A xi, when A is a subset of [n] and we study some basic properties of such maps. The class of the partial sums posets includes, as particular cases, the classical Young lattice Y of all the the integer partitions ordered by means of the inclusion of the corresponding Young diagrams, and other sublattices of Y. For example, the lattice M(n) of all the integer partitions having all distinct parts and maximum part at most n, or, also, the lattice L(m, n) of all the integer partitions with at most m parts and maximum part not exceeding n (both these lattices were introduced by Stanley in [22]). 1. A quasi-order for partial sums Let I a totally ordered set with order �. If λ, λ′ ∈ I, we set λ ≺ λ′ if λ � λ′ and λ 6= λ′; moreover, we write in an equivalent way λ′ ≻ λ instead of λ ≺ λ′ and λ′ � λ instead of λ � λ′. We fix a set partition (I+, I−) of I with two disjoint subsets I+ and I− such that λ ≻ γ if λ ∈ I+ and γ ∈ I− (we admit the possibility that one of such subsets is empty). The elements of I are called indexes, those of I+ positive indexes and those of I− negative indexes. Let {xλ : λ ∈ I+} ∪ {yγ : γ ∈ I−} a family of indeterminate real numbers with running indexes in I. We assume that xλ > xλ′ > 0 > yγ > yγ′ (4) if λ, λ′ ∈ I+ with λ ≻ λ′ and if γ, γ′ ∈ I− with γ ≻ γ′. We refer to (4) as to the (I+, I−)-initial conditions. We denote by M(I) the family of the finite multisets of I. Definition 1. We say that a function f : I → R realizes the (I+, I−)- initial conditions if: the assignment xλ = f(λ) when λ ∈ I+ and yγ = f(γ) when γ ∈ I−, satisfies the inequalities in (4). We denote by Φ(I+, I−) the set of all the functions that realize the (I+, I−)-initial conditions. We introduce now the concept of partial sums order on I. Definition 2. A partial sums order on I that realizes the (I+, I−)-initial conditions (briefly a (I|I+, I−)-partial sums order) is a partial order ⊑ on M(I) such that if A, B ∈ M(I) then A ⊑ B ⇐⇒ ∑ α∈A f(α) 6 ∑ β∈B f(β) for each f ∈ Φ(I+, I−) (5) 174 Lattices of partial sums A partial sums poset on I that realizes the initial conditions (I+, I−) (briefly a (I|I+, I−)-partial sums poset) is an induced sub-poset (T, ⊑) of (M(I), ⊑), for some subset T of M(I), where ⊑ is a (I|I+, I−)-partial sums order. If (T, ⊑) is a lattice, we say that (T, ⊑) is a (I|I+, I−)-partial sums lattice. If the partition (I+, I−) of I is clear from the context, we only say that (T, ⊑) is a I-partial sums poset. The aim of this section is to build explicitly a quasi-order which will induce a (I|I+, I−)-partial sums order on an arbitrary totally ordered set I. We now introduce a new symbol o and we formally assume that λ ≻ o if λ ∈ I+ and o ≻ γ if γ ∈ I−. We set Io = I ∪ {o}. Of course Io is also a totally ordered set with the obvious extension of the relation ≻ to Io. Definition 3. Let q and p be two non-negative integers. A Io-string w with balance (q, p) is a finite sequence of the form (λq, . . . , λ1|γ1, . . . , γp), where λq, . . . , λ1 ∈ I+ ∪ {o}, γp, . . . , γ1 ∈ I− ∪ {o} and λq � · · · � λ1, γ1 � · · · � γp. The elements λq, . . . , λ1, γ1, . . . , γp are called indexes of w. A Io-string w is a Io-string having balance (q, p), for some non-negative integers q and p. When I+ is a subset of the real interval (0, +∞) and I− is a sub- set of the real interval (−∞, 0), we assume the following convention. If w = (λq, . . . , λ1|γ1, . . . , γp) is a Io-string, we also write w in the form λq . . . λ1|µ1 . . . µp, where µj is the absolute value of γj , for j = 1, . . . , p. For example, with 44o|oo11 we mean (4, 4, o|o, o, −1, −1), whereas the Io-string (44, o|o, o, −11) will be write as (44)o|oo(11). We denote by M(Io) the set of all the Io-strings. We call λq, . . . , λ1 the non-negative indexes of w and γ1, . . . , γp the non-positive indexes of w; also, we call positive indexes of w the elements λi with λi ≻ o and negative indexes of w the elements γj with γj ≺ o. We assume that q = 0 [p = 0] iff there are no non-negative [non-positive] indexes of w; in particular, it results that p = 0 and q = 0 iff w is the empty string, that we denote by (|). If (w = λq, . . . , λ1|γ1, . . . , γp), we set w+ = λq . . . λ1| and w− = |µ1 . . . µp. If w is a Io-string, we denote by {w � o} [{w ≻ o}] the multi-set of all the non-negative [positive] indexes of w, by {w � o} [{w ≺ o}] the multi-set of all the non-positive [negative] indexes of w and by {w} the multi-set of all the indexes of w (i.e. {w} = {w � o} ∪ {w � o}). We denote respectively with |w|�, |w|�, |w|≻, |w|≺ the cardinality of {w � o}, {w � o}, {w ≻ o}, {w ≺ o}. We call the ordered couple (|w|≻, |w|≺) the signature of w and we note that (|w|�, |w|�) is exactly the balance of w . G. Chiaselotti, T. Gentile, P. A. Oliverio 175 For example, if I = Z\{0}, I+ = Z+, I− = Z− and w = 444221ooo|o11333, then {w � o} = {43, 22, 11, o3}, {w � o} = {o1, (−1)2, (−3)3}, {w ≻ o} = {43, 22, 11}, {w � o} = {(−1)2, (−3)3}, w has balance (|w|�, |w|�) = (9, 6) and signature (|w|≻, |w|≺) = (6, 5). In the sequel, two Io-strings w = (λq, . . . , λ1|γ1, . . . , γp) and w′ = (λ′ q′ , . . . , λ′ 1|γ′ 1, . . . , γ′ p′) are considered equals (and we shall write w = w′) if and only if q = q′, p = p′ and λi = λ′ i, µj = µ′ j for i = 1, . . . , q and j = 1, . . . , p. Therefore, for example, the two Io-strings (λ2, λ1, o, o, o|γ1, γ2) and (λ2, λ1, o, o|o, γ1, γ2) are considered different between them in our context. If w is a Io-string having signature (t, s) and balance (q, p), then w has the form w = (λq, . . . , λq−t+1, λq−t, . . . λ1|γ1, . . . , γp−s, γp−s+1, . . . , γp) (6) where λq � · · · � λq−t+1 ≻ 0 ≻ γp−s+1 � · · · � γp, λq−t = . . . = λ1 = o and γ1 = . . . = γp−s = o. We also write w in (6) in the following form: w = (λq, . . . , λq−t+1, oq−t|op−s, γp−s+1, . . . , γp) (7) If w is a Io-string as in (7), we call reduced Io-string of w the following Io-string: w∗ = (λq, . . . , λq−t+1|γp−s+1 . . . γp) (8) If W is a subset of M(Io), we set W∗ = {w∗ : w ∈ W}. Then it is obvious that we can identify the set M(I) with the set M(Io)∗. Let us note that {w ≻ o} = {w∗ ≻ o} and {w ≺ o} = {w∗ ≺ o}. If U is a subset Io-strings, we say that U is uniform if all the Io-strings in U have the same balance; in particular, if all the Io-strings in U have balance (q, p), we also say that U is (q, p)-uniform. If v1, . . . , vk are Io-strings, we say that they are uniform [(q, p)-uniform] if the subset U = {v1, . . . , vk} is uniform [(q, p)-uniform]. If F is a finite subset of Io-strings, we define a way to make uniform all the Io-strings in F : we set qF = max{|v|� : v ∈ F}, pF = max{|v|� : v ∈ F}; moreover, if v = (λq, . . . , λ1|γ1, . . . , γp) ∈ F we also set vF = (λq, . . . , λ1, oqF −q|opF −p, γ1, . . . , γp) and F = {vF : v ∈ F}. Then F is (qF , pF )-uniform and |F | 6 |F |. If F is uniform we note that vF = v for each v ∈ F , hence F = F . We call F the uniform closure of F . When F is clear from the context we simply write v instead of vF . In particular, if v and w are two Io-strings, when we write v and w without further specification, we always mean vF and wF , where F = {v, w}. We observe that if v and w are two uniform Io-strings 176 Lattices of partial sums then v = v and w = w; moreover, for each finite subset F ⊆ P such that w ∈ F we have {w ≻ 0} = {wF ≻ 0} and {w ≺ 0} = {wF ≺ 0}. If u = (λq, . . . , λ1|γ1 . . . γp) and u′ = (λ′ q, . . . , λ′ 1|γ′ 1, . . . , γ′ p) are two uniform Io-strings, we set: u 0 u′ ⇐⇒ λi � λ′ i and γj � γ′ j (9) for all i = 1, . . . , q and j = 1, . . . , p. We also set u △ u′ = (min{λq, λ′ q}, . . . , min{λ1, λ′ 1}| min{γ1, γ′ 1}, . . . , min{γp, γ′ p}) and u ▽ u′ = (max{λq, λ′ q}, . . . , max{λ1, λ′ 1}| max{γ1, γ′ 1}, . . . , max{γp, γ′ p}). Let us introduce now a quasi-order ⊑ (i.e. a reflexive and transitive binary relation) on the set M(Io) of all the Io-strings. Definition 4. If v, w ∈ M(Io), we set v ⊑ w if v 0 w. (10) In particular, if v and w are uniform, then v ⊑ w iff v 0 w. The following result is simple but useful: Proposition 1. ⊑ is a quasi-order on the set M(Io). Moreover, if v, w ∈ M(Io) the following conditions are equivalent: (i) v ⊑ w. (ii) There exists a finite subset F of M(Io) containing v and w such that vF 0 wF . (iii) For each finite subset F of M(Io) containing v and w we have vF 0 wF . (iv) v∗ ⊑ w∗. Proof. It is immediate to verify that the relation ⊑ is a quasi-order. The unique implication that requires some comment is (ii) ⇒ (iii). For this, it is enough to observe that if {v, w} ⊆ H ⊆ F , with H and F both finite, then vH 0 wH if and only if vF 0 wF . G. Chiaselotti, T. Gentile, P. A. Oliverio 177 2. Abstract partial sums posets We consider now the equivalence relation on M(Io) induced from the quasi-order ⊑, i.e. if v and w are two Io-strings in M(Io), we set v ∼ w ⇐⇒ v ⊑ w and w ⊑ v (11) The proof of the following result is immediate from the above definitions. Proposition 2. (i) If v and w are two Io-strings, then v = w if and only if v and w are uniform and v ∼ w. (ii) If F is a finite subset of M(Io) such that v, w ∈ F , then v ∼ w if and only if vF = wF . (iii) If F is a finite subset of M(Io) such that v ∈ F then v ∼ vF . (iv) If v is a Io-string then v ∼ v∗. If F is any subset of M(Io), we can consider on the quotient set F/ ∼ the usual partial order induced by ∼, which will be denoted by .. We recall that . is defined as follows: if Z, Z ′ ∈ F/ ∼ then Z . Z ′ ⇐⇒ v ⊑ v′ (12) for any/all v, w ∈ F such that v ∈ Z and v′ ∈ Z ′. If w ∈ F , in some case we set [w]F∼ = {v ∈ F : v ∼ w}, that is the equivalence class of w in F/ ∼. Remark 1. If F ⊆ H ⊆ M(Io) we can consider F/ ∼ as a subset of H/ ∼ through the identification of [v]F∼ with [v]H∼ , for each v ∈ F . Therefore, if F ⊆ H ⊆ M(Io) we always can assume that (F/ ∼ ,.) is a sub-poset of (H/ ∼ ,.). The next definition describes a type of subsets of Io-strings which shall permit us to find several partial sums lattices. Definition 5. We say that a subset F ⊆ M(Io) is lattice-inductive if for each finite subset F ⊆ F it results that: (i) F ⊆ F ; (ii) if v, w ∈ F , then vF △ wF ∈ F and vF ▽ wF ∈ F . Let us note that obviously M(Io) is lattice-inductive. The relevance of the lattice-inductive subsets of P is established in the following result. 178 Lattices of partial sums Theorem 1. Let F a lattice-inductive subset of M(Io). Then (F/ ∼ ,.) is a distributive lattice. Proof. To prove that (F/ ∼) is a lattice, we take two equivalence classes [v]F∼ and [w]F∼ in (F/ ∼) and we define the following operations: [v]F∼ ∧ [w]F∼ = [v △ w]F∼ and [v]F∼ ∨ [w]F∼ = [v ▽ w]F∼. It easy to see that the operations ∧ and ∨ are well defined because they do not depend on the choice of representatives in the respective equivalence classes and that [v]F∼ ∧ [w]F∼ . [v]F∼, [v]F∼ ∧ [w]F∼ . [w]F∼. Let now [z]∼ ∈ F/ ∼ such that [z]F∼ . [v]F∼ and [z]F∼ . [w]F∼. We set F = {v, w, z}. Then: ([z]F∼ . [v]F∼ and [z]F∼ . [w]F∼) ⇐⇒ (by definition of .) (z ⊑ v and z ⊑ w) ⇐⇒ (by Proposition 1 (iii)) zF 0 vF and zF 0 wF ⇐⇒ (by definition of 0 and of △) zF 0 vF △ wF ⇐⇒ (since the Io-strings are uniform) zF ⊑ vF △ wF ⇐⇒ (by definition of . and by Proposition 2 (iii)) [z]F∼ = [zF ]F∼ . [vF △ wF ]F∼. Now, since [vF △ wF ]F∼ = (by definition of ∧) [vF ]F ∧ [wF ]F = (by Proposition 2 (iii)) [v]F∼ ∧ [w]F∼, it follows that [z]F∼ . [v]F∼ ∧ [w]F∼. This proves that the operation ∧ defines effectively the inf in (F/ ∼,.). In the same way we can proceed for the operation ∨ in the sup-case. Finally, let us note that the distributivity holds because the operations △ and ▽ are defined on the components of the uniform Io-strings. Since M(Io) is obviously lattice-inductive, it follows that: Corollary 1. (M(Io)/ ∼ ,.) is a distributive lattice. When F is finite and uniform is simple to verify if F is lattice-inductive: Corollary 2. Let F a finite uniform subset of M(Io), then F is lattice- inductive if and only if whenever v, w ∈ F also v △ w ∈ F and v ▽ w ∈ F Let us observe that if v ∈ F , by Proposition 2 (iv) we can always choose v∗ as a representative for the equivalence class [v]F∼. In such a way we identify the quotient set F/ ∼ with the subset F∗ of M(Io), and we shall write F/ ∼≡ F∗. Therefore, if F ⊆ M(Io), Z, Z ′ are two any equivalence classes in F/ ∼ and v, v′ are two any elements in F such that v ∈ Z, v′ ∈ Z ′, the order (12) will have the following equivalent form (by Proposition 1 (iv)): Z . Z ′ ⇐⇒ v∗ ⊑ v′ ∗ (13) Hence, taking into account the conventions established in (13), we can always think that on each quotient set F/ ∼ the partial order . is identified with ⊑, therefore: G. Chiaselotti, T. Gentile, P. A. Oliverio 179 Remark 2. If F ⊆ M(Io) we shall identify the quotient poset (F/ ∼,.) with (F∗, ⊑), and this means that we shall consider (F/ ∼,.) as a sub- poset of (M(Io)∗, ⊑) ≡ (M(I), ⊑). In particular, if F coincides with M(Io) then (M(Io)/ ∼,.) ≡ (M(Io)∗, ⊑) ≡ (M(I), ⊑) (14) If f ∈ Φ(I+, I−) we define the function f̃ : Io → R setting f̃(α) = { f(α) if α ∈ I 0 if α = o Definition 6. If f ∈ Φ(I+, I−), we denote by ∑ f the function ∑ f : M(Io) → R such that ∑ f (w) = ∑ α∈{w} f̃(α) for each w ∈ M(Io), and we call ∑ f the sum function of f . Let us note that if w ∈ M(Io) then ∑ f (w) = ∑ α∈{w∗} f(α). The next is the main result of this section. Theorem 2. (i) The partial order ⊑ on M(I) in (14) is a (I|I+, I−)- partial sums order. (ii) If F is a subset of M(Io), then (F∗, ⊑) is a (I|I+, I−)-partial sums poset. (iii) If F is a lattice-inductive subset of M(Io), then (F∗, ⊑) is a dis- tributive (I|I+, I−)-partial sums lattice. Proof. (i) By Proposition 1 (iii) it is sufficient to prove that if w and w′ are two uniform Io-strings then w ⊑ w′ ⇐⇒ ∑ f (w) 6 ∑ f (w′) (15) Let w = (λq, . . . , λ1|γ1, . . . , γp) and w′ = (λ′ q, . . . , λ′ 1|γ′ 1, . . . , γ′ p) two uni- form Io-strings. If w ⊑ w′, by (9) we have λi � λ′ i and γj � γ′ j for all i = 1, . . . , q and j = 1, . . . , p. Therefore, if f ∈ Φ(I+, I−), from the hypothesis that f realizes the (I+, I−)-initial conditions and by definition of ∑ f , we obtain ∑ f (w) 6 ∑ f (w′). We assume now that ∑ f (w) 6 ∑ f (w′) for each f ∈ Φ(I+, I−) and that the condition w ⊑ w′ is false. This means that there exists some i ∈ {1, . . . , q} such that λi ≻ λ′ i or some j ∈ {1, . . . , p} such that γj ≻ γ′ j . Let us suppose first that there exists i ∈ {1, . . . , q} such that λi ≻ λ′ i and 180 Lattices of partial sums we assume that i is maximal among all the positive integers l ∈ {1, . . . , q} such that λl ≻ λ′ l, therefore λq � · · · � λi+1 � λi ≻ λ′ i � λ′ i−1 � · · · � λ′ 1 ; λ′ q � λq, . . . , λ′ i+1 � λi+1 (16) We consider now the following function : f(α) :=      −1 if α ∈ I− 0 if α ∈ I+ and α ≺ λi +1 if α ∈ I+ and α � λi Then f ∈ Φ(I+, I−) and by (16) it follows that ∑ f (w) > (q − i + 1) + ∑ 16j6p f̃(γj) > (q−i)+ ∑ 16j6p f̃(γj) = (q−i)+ ∑ 16j6p f̃(γ′ j) = ∑ f (w′), that is a contradiction. We can suppose then λi � λ′ i for all i = 1, . . . , q, so there exists j ∈ {1, . . . , p} such that γj ≻ γ′ j and we assume that j is minimal among all the positive integers l ∈ {1, . . . , p} such that γl ≻ γ′ l, therefore γ1 � · · · � γj−1 � γj ≻ γ′ j � γ′ j+1 � · · · � γ′ p ; γ′ 1 � γ1, . . . , γ′ j−1 � γj−1 (17) We must now distinguish two cases. First we suppose that γj = o. In this case we consider the following function: h(α) := { 0 if α ∈ I+ −1 if α ∈ I− Then h ∈ Φ(I+, I−) and by (17) it follows that ∑ h(w) > (−1)(p − j) > (−1)(p − j + 1) = ∑ h(w′), that is a contradiction. We assume now that γj ≺ o. In this case we consider the following function: g(α) :=      +1 if α ∈ I+ −1 if α ∈ I− and α � γj −2 if α ∈ I− and α ≺ γj Then g ∈ Φ(I+, I−) and by (17) we have: ∑ g (w) > ∑ 16l6j−1 g̃(γl) + (−1) + ∑ j+16l6p g̃(γl) > ∑ 16l6j−1 g̃(γl) + (−2) + ∑ j+16l6p g̃(γl) = ∑ 16l6j−1 g̃(γ′ l) + g̃(γ′ j) + ∑ j+16l6p g̃(γl) > ∑ 16l6j−1 g̃(γ′ l) + (−2) + (−2)(p − j) = ∑ g (w′) G. Chiaselotti, T. Gentile, P. A. Oliverio 181 that is a contradiction. This complete the proof of (i). Parts (ii) and (iii) are direct consequence of Remark 2 and Theorem 1. We establish now some direct consequences of the previous theorem. Corollary 3. (M(I), ⊑) is a distributive (I|I+, I−)-partial sums lattice. Proof. We take F = M(Io). Since F is lattice-inductive, by Theorem 2 (iii) it follows that (M(Io)∗, ⊑) is a distributive (I|I+, I−)-partial sums lattice. The result follows then by (14). Corollary 4. The classical Young lattice Y of the integer partitions is a distributive partial sums lattice. Proof. In the particular case I = N, I+ = N and I− = ∅ we have Y = M(I) and the partial order on Y is exactly ⊑. Hence the thesis follows from the previous corollary. We recall now the definition of the lattice L(m, n) introduced by Stanley in [22]: if m and n are two positive integers, L(m, n) is the sub-lattice of Y of all the integer partitions with at most m parts and maximum part not exceeding n. Corollary 5. L(m, n) is a distributive partial sums lattice. Proof. We take I = {n > n − 1 > · · · > 1}, I+ = I and I− = ∅. We consider the subset F of M(Io) whose elements are the Io-strings w such that |w|> = m. Then F is lattice-inductive and F∗ = L(m, n). Hence the thesis follows by (iii) of the previous theorem. Corollary 6. If F is a subset of M(Io) and f ∈ Φ(I+, I−), the restricted sum function ∑ f : F∗ → R is order-preserving. Proof. It follows from the definition of (I|I+, I−)-partial sums poset and by Theorem 2 (ii). In terms of partial sums on a numerable quantity of real variables, we must take I− = Z− = {−1, −2, −3, . . . } and I+ = Z+ = {1, 2, 3, . . . }. In this case, (4) becomes . . . > xr > xr−1 > . . . > x1 > 0 > y1 > y2 > . . . (18) In this case, if w = (λt, . . . , λ1|µ1, . . . , µs) ∈ M(I), we set ∑ (w) = xλt + · · · + xλ1 + yµ1 + · · · + yµs . Then, by the Theorem 2 we can to 182 Lattices of partial sums think the partial order ⊑ on M(I) as the natural order induced from the linear systems inequalities (18) on the partial sum of the real variables . . . xr, xr−1, . . . , x1, y1, . . . , yq, . . . In other terms, if we formally identify the signed partitions w and w′ respectively with the indeterminate real partial sums ∑ (w) and ∑ (w′), then the result of the Theorem 2 tell us that w ⊑ w′ if and only if the real inequality ∑ (w) 6 ∑ (w′) holds and it can be deduced by using only the inequalities in (18). We can therefore use a more suggestive terminology and to think the signed partition lattice (M(I), ⊑) as a lattice of indeterminate partial sums take over a numerable quantity of real variables . . . xr, xr−1, . . . , x1, y1, . . . , yq, . . . that satisfy (18). Using this interpretation a finite piece of the Hasse diagram of M(I) is the following: 2y2 x1 + 2y2 y1 + y2 x2 + 2y2 2x1 + 2y2 x1 + y1 + y2 y2 2y1 x2 + x1 + 2y2 x2 + y1 + y2 2x1 + y1 + y2 x1 + y2 x1 + 2y1 y1 . . . . . . A particularly relevant lattice-inductive subset of M(Io) is the subset of all the Io-strings without repeated negative indexes and without repeated positive indexes. We denote such a subset by S(Io). Theorem 3. S(Io) is lattice-inductive. Proof. Let F be a finite subset of S(Io) and let w ∈ F , then, by definition of F , there exists v ∈ F such that w = vF . Since v is also an element of S(Io), it has no repeated negative or positive indexes, therefore also vF has no repeated negative or positive indexes, since vF is built by adding only eventual further o’s to v. Hence w ∈ S(Io), and this shows that F ⊆ S(Io). Let now v, v′ ∈ F , and w = vF = (λq, . . . , λ1|γ1, . . . , γp), w′ = v′F = (λ′ q, . . . , λ′ 1|γ′ 1, . . . , γ′ p), then w, w′ ∈ F ⊆ S(Io). We must to prove that w △ w′ ∈ S(Io) and w ▽ w′ ∈ S(Io). We show only that w △ w′ ∈ S(Io) because the proof that w ▽ w′ ∈ S(Io) is similar. Let w′′ = w ▽ w′, with G. Chiaselotti, T. Gentile, P. A. Oliverio 183 w′′ = (λ′′ q , . . . , λ′′ 1|γ′′ 1 , . . . , γ′′ p ). Let us suppose by absurd that w′′ /∈ S(Io); by definition of S(Io) this means that there exist two positive indexes or two negative indexes in w′′ that must be equal. We can assume that there exist λ′′ k and λ′′ l such that k > l and λ′′ k = λ′′ l ≻ o (the case when λ′′ k = λ′′ l ≺ o is analogue). Since λ′′ k = λ′′ l and w′′ ∈ M(Io), we have λ′′ k = λ′′ k−1 = · · · = λ′′ l+1 = λ′′ l (19) We note that λl ≻ o and λ′ l ≻ o because λ′′ l ≻ o, therefore, since w, w′ ∈ S(Io), by definition of S(Io) it follows that λk ≻ λk−1 ≻ · · · ≻ λl+1 ≻ λl (20) and λ′ k ≻ λ′ k−1 ≻ · · · ≻ λ′ l+1 ≻ λ′ l (21) We assume now that λ′′ k = λk. Then, if λ′′ k−1 = λk−1, by (19) we have λk = λk−1, that contradicts (20). Therefore it must be λ′′ k−1 = λ′ k−1 . Then by (19) it follows that λk = λ′ k−1 and hence, by (20), λ′ k−1 ≻ λk−1, but this contradicts the equality λ′′ k−1 = min{λk−1, λ′ k−1 } = λ′ k−1 . On the other side, also the equality λ′′ k = λ′ k leads to an absurd if we use (19) and (21). This complete the proof. Obviously we can identify S(Io)∗ with the subset of M(I) whose elements are the finite subsets (i.e. the finite multi-sets without repeated elements) of I, that we denote by S(I). Therefore, by the previous propo- sition it follows that (S(I), ⊑) ≡ (S(Io)∗, ⊑) is a distributive (I|I+, I−)- partial sums lattice. We recall now the definition of the Stanley lattice M(n) (see [22]): M(n) is the sub-lattice of the Young lattice Y whose elements are the integer partitions having all distinct parts and whose maximum is at most n, where n is an arbitrary positive integer. We have then: Corollary 7. M(n) is a distributive partial sums lattice. Proof. If we take I = {n, n − 1, . . . , 1}, I+ = I and I− = ∅, then M(n) coincides exactly with S(Io)∗. Hence the thesis is a direct consequences of the previous theorem. If n > r > 0 are two integers, we can take I = {r > · · · > 1 > −1 > · · · > −(n − r)}, (I+ = {r, . . . , 1} and I− = {−1, . . . , −(n − r)}). In this case S(Io)∗ coincides with the lattice S(n, r) introduced in [3] and [4] in 184 Lattices of partial sums order to study some extremal combinatorial sum problems. As observed in [8], S(n, r) is isomorphic to the direct product M(r) × M(n − r)∗. We recall now the definition of valuation on an arbitrary lattice X. If X is a lattice, a map η : X → R is called a valuation on X if for all a, b ∈ X: η(a ∧ b) + η(a ∨ b) = η(a) + η(b). Proposition 3. Let F be a lattice-inductive subset of M(Io) and f ∈ Φ(I+, I−). Then the restricted sum function ∑ f : F∗ → R is a valuation on (F∗, ⊑). Proof. Since the partial order ⊑ is made on the components of the Io- strings, the result follows. In [20] was shown that a valuation on a distributive lattice is uniquely determined by the values that it takes on the join-irreducible elements of the lattice, therefore, in our case, this means that ∑ f is uniquely determined by the values that it takes on the join-irreducible elements of the distributive lattice F∗. References [1] K. Alladi, P. Erdös and J.D. Vaaler, Multiplicative functions and small divi- sors, Analytic Number Theory and Diophantine Problems, Vol. 70, Progress in Mathematics, Birkhäuser, Boston, MA, 1987, pp. 1–13. [2] T. Bier and N. Manickam, The first distribution invariant of the Johnson-scheme, SEAMS Bull. Math., 11 (1987), 61–68. [3] C. Bisi and G. Chiaselotti, A class of lattices and boolean functions related to the Manickam-Miklös-Singhi Conjecture, Adv. Geom. 13, Issue 1, (2013), 1–27. [4] C. Bisi and G. Chiaselotti, Extension results for boolean maps and a class of systems of linear inequalities e-print arXiv:1012.5486 (2010), 1–22 [5] G. Chiaselotti, On a problem concerning the weight functions, European J. Combin., 23 (2002), 15–22. [6] G. Chiaselotti, G. Infante, G. Marino, New results related to a conjecture of Manickam and Singhi. European J. Combin. 29 (2008), n. 2, 361–368. [7] G. Chiaselotti, G. Marino, C. Nardi, A minimum problem for finite sets of real numbers with nonnegative sum, J. Appl. Math, 2012 (2012), Article ID 847958, 15 pages. [8] K. Engel and C. Nardi, Solution of a problem on non-negative subset sums, European J. Combin., 33, (2012), 1253–1256. [9] P. Erdös, C. Ko and R. Rado Intersection theorems for systems of finite sets. Quart. J. Math. Oxford Ser. (2), 12 (1961), 313–320. [10] L. Geissinger, Valuations on distributive lattices.I. Arch. Mat (1973), no. 24, 230–239 G. Chiaselotti, T. Gentile, P. A. Oliverio 185 [11] L. Geissinger, Valuations on distributive lattices.II. Arch. Mat (1973), no. 24, 337–345 [12] L. Geissinger, Valuations on distributive lattices.III. Arch. Mat (1973), no. 24, 475–481 [13] G. O. H. Katona, A simple proof of the Erdös-Ko-Rado theorem. J. Combinatorial Theory Ser. B, 13 (1972), 183–184. [14] N. Manickam, First distribution invariants of association shemes, Ph. D. Thesis, The Ohio State University, (1986). [15] N. Manickam, W -complement d-spreads, Singleton systems, European J. Combin., 8 (4) (1987), 437–439. [16] N. Manickam and D. Miklos, On the number of non–negative partial sums of a non–negative sum, Colloq. Math. Soc. Janos Bolyai, 52 (1987), 385–392. [17] N.Manickam, Distribution Invariants of Association Schemes, Congress Numeran- tum, 61 (1988), 121–131. [18] N. Manickam, First distributed sets in the association scheme of bilinear forms, Colloq. Math. Soc. Janos Bolyai, 60 (1991), 465–468. [19] G. Marino and G. Chiaselotti, A method to count the positive 3-subsets in a set of real numbers with non-negative sum, European J. Combin., 23 (2002), 619–629. [20] G.C. Rota, On the combinatorics of the Euler characteristic, Studies in Pure Math., Ac.Press, London, 1971, 221–233. [21] S. Srinivasan, On an Arithmetical Inequality II, Contemporary Mathematics, American Mathematical Society, Vol. 210, Providence, RI, 1998, pp. 299–301. [22] R.P. Stanley, Weyl groups, the hard Lefschetz theorem, and the Sperner property SIAM J. Alg. Disc. Meth., 1 (1980), 168–184. [23] M. Tyomkyn, An improved bound for the Manickam-Miklós-Singhi conjecture, European J. Combin., 33 (1), (2012), 27–32. Contact information G. Chiaselotti, T. Gentile, P. A. Oliverio Dipartimento di Matematica, Universitá della Calabria, Via Pietro Bucci, Cubo 30B, 87036 Arcavacata di Rende (CS), Italy. E-Mail(s): chiaselotti@unical.it, gentile@mat.unical.it, oliverio@unical.it Received by the editors: 03.10.2012 and in final form 06.06.2013.