On the dimension of Kirichenko space

We introduce a notion of the Kirichenko space which is connected with the notion of Gorenstein matrix (see [2], ch. 14). Every element of Kirichenko space is an n × n matrix, whose elements are solutions of the equations ai,j +aj,σ(i) = ai,σ(i) ; a₁,i = 0 for i, j = 1, . . . , n determined by a...

Full description

Saved in:
Bibliographic Details
Published in:Algebra and Discrete Mathematics
Date:2006
Main Author: Plakhotnyk, M.
Format: Article
Language:English
Published: Інститут прикладної математики і механіки НАН України 2006
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/157355
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:On the dimension of Kirichenko space / M. Plakhotnyk // Algebra and Discrete Mathematics. — 2006. — Vol. 5, № 2. — С. 87–126. — Бібліогр.: 8 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
id nasplib_isofts_kiev_ua-123456789-157355
record_format dspace
spelling Plakhotnyk, M.
2019-06-20T02:46:01Z
2019-06-20T02:46:01Z
2006
On the dimension of Kirichenko space / M. Plakhotnyk // Algebra and Discrete Mathematics. — 2006. — Vol. 5, № 2. — С. 87–126. — Бібліогр.: 8 назв. — англ.
1726-3255
https://nasplib.isofts.kiev.ua/handle/123456789/157355
We introduce a notion of the Kirichenko space which is connected with the notion of Gorenstein matrix (see [2], ch. 14). Every element of Kirichenko space is an n × n matrix, whose elements are solutions of the equations ai,j +aj,σ(i) = ai,σ(i) ; a₁,i = 0 for i, j = 1, . . . , n determined by a permutation σ which has no cycles of the length 1. We give a formula for the dimension of this space in terms of the cyclic type of σ.
en
Інститут прикладної математики і механіки НАН України
Algebra and Discrete Mathematics
On the dimension of Kirichenko space
Article
published earlier
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
collection DSpace DC
title On the dimension of Kirichenko space
spellingShingle On the dimension of Kirichenko space
Plakhotnyk, M.
title_short On the dimension of Kirichenko space
title_full On the dimension of Kirichenko space
title_fullStr On the dimension of Kirichenko space
title_full_unstemmed On the dimension of Kirichenko space
title_sort on the dimension of kirichenko space
author Plakhotnyk, M.
author_facet Plakhotnyk, M.
publishDate 2006
language English
container_title Algebra and Discrete Mathematics
publisher Інститут прикладної математики і механіки НАН України
format Article
description We introduce a notion of the Kirichenko space which is connected with the notion of Gorenstein matrix (see [2], ch. 14). Every element of Kirichenko space is an n × n matrix, whose elements are solutions of the equations ai,j +aj,σ(i) = ai,σ(i) ; a₁,i = 0 for i, j = 1, . . . , n determined by a permutation σ which has no cycles of the length 1. We give a formula for the dimension of this space in terms of the cyclic type of σ.
issn 1726-3255
url https://nasplib.isofts.kiev.ua/handle/123456789/157355
citation_txt On the dimension of Kirichenko space / M. Plakhotnyk // Algebra and Discrete Mathematics. — 2006. — Vol. 5, № 2. — С. 87–126. — Бібліогр.: 8 назв. — англ.
work_keys_str_mv AT plakhotnykm onthedimensionofkirichenkospace
first_indexed 2025-11-25T12:56:41Z
last_indexed 2025-11-25T12:56:41Z
_version_ 1850512637367943168
fulltext Algebra and Discrete Mathematics RESEARCH ARTICLE Number 2. (2006). pp. 87 – 126 c© Journal “Algebra and Discrete Mathematics” On the dimension of Kirichenko space Makar Plakhotnyk Communicated by M. Ya. Komarnytskyj Dedicated to the memory of V. M. Usenko Abstract. We introduce a notion of the Kirichenko space which is connected with the notion of Gorenstein matrix (see [2], ch. 14). Every element of Kirichenko space is an n × n matrix, whose elements are solutions of the equations ai,j +aj,σ(i) = ai,σ(i); a1,i = 0 for i, j = 1, . . . , n determined by a permutation σ which has no cycles of the length 1. We give a formula for the dimension of this space in terms of the cyclic type of σ. Introduction We remind some definitions and notions from [2], ch 14. Denote by Mn(B) the ring of all n× n matrices over the ring B. Let Z be a ring of all integers. An integer matrix E = (aij) is called an exponent matrix, if aij + ajk ≥ aik for all i, j, k; an reduced exponent, if aij + aji > 0 for all i, j. A reduced exponent matrix E will be called Gorenstein, if there exists a permutation σ of {1, . . . , n} such that ai,k + ak,σ(i) = ai,σ(i). We will denote σ by σ(E). As usual, we will call σ(E) the Kirichenko permutation. Note that σ(E) of Gorenstein matrix has no cycles of length 1. We will name relations of the type ai,k + ak,σ(i) = ai,σ(i) Gorenstein relations. Exponent matrices are widely used in the theory of tiled orders over a discrete valuation ring. Many properties of tiled orders and their quivers are fully determined by their exponent matrices. Gorenstein matrices appeared at the first time in [5]. Type your thanks here.. Key words and phrases: Gorenstein matrix, Gorenstein tiled order. 88 On the dimension of Kirichenko space It is easy to see, that from the condition ai,k + ai,σ(k) = ai,σ(i) for i = k we obtain ak,k = 0 for every k, 1 ≤ k ≤ n. Theorem. (see [1], p. 17) Let permutation σ of the set {1, 2, . . . n} be arbitrary permutation without fixed elements. Then there exists Goren- stein matrix A with σ as Kirichenko permutation, such that all elements of A belong to the set {0, 1, 2}. Proof. Matrix A, which has σ as correspondent Kirichenko permutation may be constructed in direct way: ai,i = 0 and ai,σ(i) = 2 for i = 1, . . . , n; ai,j = 1 for all i, j ∈ [1, n], such that i 6= j and j 6= σ(i). It is obvious that such matrix will be Gorenstein and σ will be Kirichenko permutation. Exponent matrices A, B are called equivalent, if one may be ob- tained from another by the following transformations: 1) adding an integer to all elements of some row with simultaneous subtracting it from the elements of the column with the same number. 2) simultaneous interchanging of two rows and equally numbered columns, or by compositions of such transformations. Theorem (see [1], p. 15): Under the transformations of the first type Gorenstein matrix goes to Gorenstein one with the same Kirichenko permutation. Note that transformations of the first type define free commutative group with n generators. If one consider matrices Hn =   0 0 0 · · · 0 0 1 0 0 · · · 0 0 1 1 0 · · · 0 0 · · · · · · · · · · · · · · · · · · 1 1 1 · · · 0 0 1 1 1 · · · 1 0   and G2m = ( Hm H (1) m H (1) m Hm ) , where H (1) m = E + Hm, then it will be easy to see, that Hn is Gorenstein matrix with Kirichenko permutation σ(Hn) = (n, n−1, . . . 2, 1), and G2m is Gorenstein one with Kirichenko permutation σ(G2m) = m∏ i=1 (i, m + i). A matrix B = (bi,j) is called (0, 1) matrix, if bi,j is either zero or one. Theorem (see [1], p. 15): Gorenstein (0, 1) matrix is equivalent either Hn or G2m. M. Plakhotnyk 89 Theorem (see [1], p. 15): If A is Gorenstein matrix with Kirichenko permutation σ, and B is obtained from it by transformation of the second type, defined by transposition τ , then B is Gorenstein matrix with the permutation τστ . According to the last theorem, without bounding of generality we may suppose that for the cyclic index {l1, . . . , lq} of correspond Kirichenko permutation the conditions 2 ≤ l1 ≤ . . . ≤ lq and l1 + . . . + lq = n take place. Denote ni = i∑ k=1 lk for every 1 ≤ i ≤ q. Then σ will look as σ = (1, 2 . . . n1)(n1 + 1, n1 + 2 . . . , n2) · · · (nq−1 + 1, nq−1 + 2, . . . , nq). Further we will consider only Gorenstein matrices and under equiv- alence of matrices we will consider possibility of obtaining one from another only by transformations of the first type. One may come to a question of describing of Gorenstein matrices in inverse way to the one, presented in the definition. In the set of square matrices of order n over the field K of characteristic 0 consider linear space K(n, σ) of all matrices A = (ai,j) such that Gorenstein relations take place for A. When a problem of describing of such matrices will be solved, it will become possible to talk about restrictions, generated by inequalities from the definition of reduced exponent matrix. If A ∈ K(n, σ), B is equivalent to it, then B ∈ K(n, σ), and ev- ery equivalence class contains the unique matrix with zeros at the first line, which is canonical representor of the equivalence class. Call ma- trix is called Kirichenko matrix. More over, it’s easy to see, that for Kirichenko matrix from the condition ai,k + ak,σ(i) = ai,σ(i) for i = 1 one may obtain ak,σ(1) = 0 for all k. It is easy to prove that the set of Kirichenko matrices is linear space and we will name it Kirichenko space. We need the next notations. Let A = (ai,j) be Kirichenko matrix of order n with Kirichenko permutation σ, and σ is the product of q cycles of length l1, l2, . . . , lq, respectively 2 ≤ li ≤ li+1, 1 ≤ i ≤ q − 1 and l1 + . . . + lq = n. The main result of this work is calculating the dimension of Kirichen- ko space. To do this, some elements of Kirichenko matrix were considered as parameters, and represent all another elements of matrix in terms of these parameters. The relations which represent the elements of matrix through parameters we name element relations, and relations between parameters we name count relations. We were succeed to find the set of parameters, element and count relations in the form such that it is 90 On the dimension of Kirichenko space possible to prove the equivalence of the set of Gorenstein relations to the set of element relations and count relations. After this, we were succeed to find the formula for the defect of the system of count relations i.e. the dimension of Kirichenko space may be represented in terms of cyclic index of Kirichenko permutation σ as 2 − 2q + q∑ r=1 [ lr 2 ] + 1 2 ∑ r 6=s (ls, lr), where (a, b) denotes the greatest common divisor of numbers a and b, and [x] denotes integer part of a number x which is the greatest integer, not larger then x. Definition 1. Denote xk = ak,1 for every k, 2 ≤ k ≤ n. For arbitrary r, 0 < r < q, and for every k, nr + 2 ≤ k ≤ n, denote as well zk,r = = anr+1,k. Variables xk and zk,r we name parameters. Definition 2. Element relation is formula which is corollary of Goren- stein relations and expresses some element of Gorenstein matrix as linear function of parameters. Full set of element relations is set of relations which contains the formula for every element of matrix A. Example 1. Relations ak,σ(k) = xk, 2 ≤ k ≤ n (1) are element. Really, if we substitutes value i = 1 to the equality ak,i + ai,σ(k) = ak,σ(k) and take into account the fact that the first row of A is zero, we obtain the necessary relation. Example 2. Relations ak,2 = 0, 1 ≤ k ≤ n ak,k = 0, 1 ≤ k ≤ n a1,k = 0, 1 ≤ k ≤ n (2) and definition of parameters are element relations. Element relations from two previous examples we name trivial. M. Plakhotnyk 91 1. Element relations are corollaries of Gorenstein rela- tions Proposition 1. Equalities (1)-(11) are full set of element relations, where a2,k = { xns+1 , if k = ns + 1 for some s xk−1, if k 6= ns + 1 for any s (3) For every m, 3 ≤ m ≤ n1, every s, 1 ≤ s ≤ q − 1 and every k, ns + 1,≤ k ≤ ns+1 or m + 1 ≤ k ≤ n1 write out ak,m =    m−1∑ i=2 xi − m−3∑ i=0 xns+1−i, if k = ns + 1; m−1∑ i=2 xi − l−1∑ i=1 xk−i − m−l−2∑ i=0 xns+1−i, if k = ns + l, 2 ≤ l ≤ m −2; m−1∑ i=2 xi − m−2∑ i=1 xk−i, if ns + m − 1 ≤ k ≤ ns+1, or m < k ≤ n1 (4) and am,k =    m−2∑ i=0 xns+1−i − m−1∑ i=2 xi, if k = ns + 1, l−1∑ i=1 xk−i + m−l−1∑ i=0 xns+1−i − m−1∑ i=2 xi, if k = ns + l, 2 ≤ l < m m−1∑ i=1 xk−i − m−1∑ i=2 xi, if ns + m ≤ k ≤ ns+1, or m < k ≤ n1 (5) Writing out element relations for other elements of matrix A, we will write indices of its elements as ak,m and am,k, where k > m for nr + 1 ≤ m ≤ nr+1 and some r, 1 ≤ r ≤ q − 1. in this case for every s, r ≤ s ≤ q − 1, and k, ns + 1 ≤ k ≤ ns+1 relations (6)-(11) look like: ak,nr+1 = { xk − zns+1,r, if k = ns+1, s > r xk − zk+1,r, if ns + 1 ≤ k < ns+1, s > r (6) ak,nr+1 = { xk, if k = nr+1. xk − zk+1,r, if nr + 1 < k < nr+1 (7) ak,nr+2 = xnr+1 − zk,r, if k > nr + 2 (8) anr+2,k = { xns+1 − xnr+1 + zns+1,r, if k = ns + 1, s > r xk−1 − xnr+1 + zk−1,r, if ns + 2 ≤ k ≤ ns+1, k > nr + 2 (9) 92 On the dimension of Kirichenko space ak,nr+m =    −zns+1−m+3,r + m−1∑ i=1 xnr+i − m−3∑ i=0 xns+1−i, if k = ns + 1 > nr + m; −zns+1−m+l+2,r + m−1∑ i=1 xnr+i − m−l−2∑ i=0 xns+1−i − l−1∑ i=1 xk−i, if k = ns + l > nr + m 2 ≤ l ≤ m − 2; −zk−m+2,r + m−1∑ i=1 xnr+i − m−2∑ i=1 xk−i, if ns + m − 1 ≤ k ≤ ns+1, k > nr + m; (10) anr+m,k =    zns+1−m+2,r − m−1∑ i=1 xnr+i + m−2∑ i=0 xns+1−i, if k = ns + 1 > nr + m; zns+1−m+l+1,r − m−1∑ i=1 xnr+i + l−1∑ i=1 xk−i + m−l−1∑ i=0 xns+1−i, if k = ns + l > nr + m, 2 ≤ l ≤ m − 1; zk−m+1,r − m−1∑ i=1 xnr+i + m−1∑ i=1 xk−i, if ns + m ≤ k ≤ ns+1, k > nr + m. (11) For proving this proposition we need the following lemmas. Lemma 1. If 3 ≤ m ≤ n1, max(ns + 1, m + 1) ≤ k ≤ ns+1, then relations (4) and (5) are corollaries of Gorenstein ones. Proof. Consider equality a2,k + ak,3 = x2, whence, using (3) obtain, that ak,3 = x2−xns+1 , if k = ns+1, and ak,3 = x2−xk−1 if ns+2 ≤ k ≤ ns+1, which is equality (4) for m = 3. Consider ak,3+a3,k+1 = xk, k 6= ns+1. For use (4), consider situations k = ns +1 and k 6= ns +1. For k = ns +1, obtain (x2−xns+1 )+a3,ns+2 = xns+1, whence a3,p = xp−1+xns+1 −x2, for p = ns+2. Consider situation, when ns + 2 ≤ k < ns+1. In this case (x2 − xk−1) + a3,k+1 = xk, whence a3,k+1 = xk + xk−1 − x2, that is a3,p = xp−1 + xp−2 − x2 for ns + 3 ≤ p ≤ ns+1. Now, substituting k = ns+1 to an equality ak,3 + a3,σ(k) = xk, we obtain equality (5) for m = 3. Let us prove the validity of relations (4) and (5) for 3 ≤ m ≤ n1 by induction. The induction base is just proved. Suppose these relations to be valid for some m, 3 ≤ m < n1,. Let us proof their validity for some m + 1 too. M. Plakhotnyk 93 Consider equality am,k + ak,m+1 = xm, m < n1, whence ak,m+1 = xm − am,k. For k = ns + 1 we obtain ak,m+1 = xm − ( m−2∑ i=0 xns+1−i − m−1∑ i=2 xi ) = (m+1)−1∑ i=2 xi − (m+1)−3∑ i=0 xns+1−i, which is necessary. Fix l ∈ [2, m−1]. Then for k = ns+l we obtain ak,m+1 = xm−am,k = xm − ( l−1∑ i=1 xk−i + m−l−1∑ i=0 xns+1−i − m−1∑ i=2 xi ) = (m+1)−1∑ i=2 xi − l−1∑ i=1 xk−i− − (m+1)−l−2∑ i=0 xns+1−i for k = ns+l, 2 ≤ l ≤ (m+1)−2, which is necessary. Consider equality am,k + ak,σ(m) = xm for k, ns + m ≤ k ≤ ns+1 and obtain ak,m+1 = xm − am,k = xm − ( m−1∑ i=1 xk−i − m−1∑ i=2 xi ) = (m+1)−1∑ i=2 xi − (m+1)−2∑ i=1 xk−i, ns + (m + 1) − 1 ≤ k ≤ ns+1. Thus we have proved that ak,m+1 = =    (m+1)−1∑ i=2 xi − (m+1)−3∑ i=0 xns+1−i, ifk = ns + 1, (m+1)−1∑ i=2 xi − l−1∑ i=1 xk−i − (m+1)−2−l∑ i=0 xns+1−i, ifk = ns + l, 2 ≤ l < m (m+1)−1∑ i=2 xi − (m+1)−2∑ i=1 xk−i, ifk > ns + m − 1 . Consider equality ak,m+1 + am+1,k+1 = xk, k 6= ns. Let k = ns + 1. Then am+1,ns+2 = xns+1 − ans+1,m+1 = xns+1 −( m∑ i=2 xi − m−2∑ i=0 xns+1−i ) = xp−1 − (m+1)−1∑ i=2 xi + (m+1)−3∑ i=0 xns+1−i = am+1,p , for p = ns + l, l = 2. Let k = ns + l, 2 ≤ l ≤ m − 1. Then am+1,ns+(l+1) = xns+l − −ans+l,m+1 = xns+l− ( m∑ i=2 xi − l−1∑ i=1 xk−i − m−l−1∑ i=0 xns+1−i ) = − (m+1)−1∑ i=2 xi+ (l+1)−1∑ i=1 x(k+1)−i + (m+1)−(l+1)−1∑ i=0 xns+1−i, which is the same as am+1,ns+l = = − (m+1)−1∑ i=2 xi + l−1∑ i=1 xk−i + (m+1)−l−1∑ i=0 xns+1−i, for 3 ≤ l ≤ (m + 1) − 1. 94 On the dimension of Kirichenko space Let ns + m ≤ k ≤ ns+1. Then am+1,k+1 = xk − ak,m+1 = xk − − ( m∑ i=2 xi − m−1∑ i=1 xk−i ) = − m∑ i=2 xi + m∑ i=1 xk+1−i, for ns + m ≤ k ≤ ns+1, which is the same as am+1,k = − (m+1)−1∑ i=2 xi + (m+1)−1∑ i=1 xk−i for ns + (m + 1) ≤ k ≤ ns+1. Consider equality ak,m+1 + am+1,σ(k) = xk for k = ns+1, and ob- tain am+1,ns+1 = xns+1 − ans+1,m+1 = xns+1 − ( m∑ i=2 xi − m−1∑ i=1 xns+1−i ) = (m+1)−2∑ i=0 xns+1−i − (m+1)−1∑ i=2 xi = am+1,p , where p = ns + 1. thus we have proved that am+1,k = =    (m+1)−2∑ i=0 xns+1−i − (m+1)−1∑ i=2 xi, if k = ns + 1, l−1∑ i=1 xk−i + (m+1)−l−1∑ i=0 xns+1−i − (m+1)−1∑ i=2 xi, if k = ns + l, 2 ≤ l < m + 1 (m+1)−1∑ i=1 xk−i − (m+1)−1∑ i=2 xi, if ns + (m + 1) ≤ k ≤ ns+1 , and it finishes the proof of relations (4) and (5). Fix arbitrary r, 1 ≤ r ≤ q−1, arbitrary s, r ≤ s ≤ q−1, and consider k, ns + 1 ≤ k ≤ ns+1. For convenience instead of zk,r we will write zk. Lemma 2. For r, s and k from the intervals noted above, equalities (6)- (9) take place. Proof. Consider equality ak,nr+1 + anr+1,σ(k) = xk. For k 6= ns+1 obtain ak,nr+1 + zk+1 = xk, whence ak,nr+1 = xk − zk+1. Considering k = nr+1 obtain anr+1,nr+1 +anr+1,nr+1 = xnr+1 , whence ak,nr+1 = xk, that is ak,nr+1 = { xk, if k = nr+1. xk − zk+1, if nr + 1 < k < nr+1 , which coincides with (7). For k = ns+1, s > r obtain ans+1,nr+1 + anr+1,ns+1 = xns+1 , whence ak,nr+1 = xk − zns+1, that is ak,nr+1 = { xk − zns+1, if k = ns+1. xk − zk+1, if ns + 1 ≤ k < ns+1 , which coincides with (6). M. Plakhotnyk 95 Consider equality anr+1,k + ak,nr+2 = xnr+1, whence, according to denotations, ak,nr+2 = xnr+1 − anr+1,k = xnr+1 − zk, which is (8). If lr > 2, then consider equality ak,nr+2+anr+2,k+1 = xk for k 6= ns+1. Whence anr+2,k+1 = xk − ak,nr+2 = xk − xnr+1 + zk, that is anr+2,p = xp−1 − xnr+1 + zp−1 for p 6= ns + 1. For k = ns+1 obtain ans+1,nr+2 + anr+2,ns+1 = xns+1 , and so anr+2,k = xns+1 − ans+1,nr+2 = xns+1 − (xnr+1 − zns+1 ) for k = ns + 1, whence anr+2,p = { xns+1 − xn1+1 + zns+1 , if k = ns + 1, which is (9). xk−1 − xn1+1 + zk−1, if ns + 2 < k ≤ ns+1 Leaving in force denotations for boundaries for r, s and k, consider arbitrary m, 3 ≤ m ≤ lr. Lemma 3. For every r, s, k and m from noted intervals, relations (10) and (11) are corollaries of Gorenstein relations. Proof. Proof by induction for m. Consider the equality anr+2,k + ak,nr+3 = xnr+2, whence ak,nr+3 = xnr+2 − anr+2,k, i.e. ak,nr+3 = { xnr+1 + xnr+2 − xns+1 − zns+1 , if k = ns + 1. xnr+1 + xnr+2 − xk−1 − zk−1, if k 6= ns + 1 , and so obtain induction base for (10). Consider equality ak,nr+3 + anr+3,k+1 = xk for k 6= ns+1, whence anr+3,p = xp−1 − ap−1,nr+3, for ns + 2 ≤ p ≤ ns+1. For k 6= ns + 2 then the last equality is equivalent to anr+3,k = xk−1−(xnr+1+xnr+2−xk−2−zk−2) = xk−1+xk−2−xnr+1−xnr+2+zk−2. For k = ns+2 obtain anr+3,k = xk−1−(xnr+1+xnr+2−xns+1 −zns+1 ) = xk−1 − xnr+1 − xnr+2 + xns+1 + zns+1 Substitute k = ns+1 to equality ak,nr+3 + anr+3,σ(k) = xk, and obtain anr+3,ns+1 = xns+1 − ans+1,nr+3 = xns+1 − (xnr+1 + xnr+2 − xns+1−1 − zns+1−1) = xns+1 + xns+1−1 − xnr+1 − xnr+2 + zns+1−1. Whence obtain anr+3,k =    xns+1 + xns+1−1 − xnr+1 − xnr+2 + zns+1−1; if k = ns + 1, xk−1 − xnr+1 − xnr+2 + xns+1 + zns+1 ; if k = ns + 2, xk−1 + xk−2 − xnr+1 − xnr+2 + zk−2, if ns + 3 ≤ k ≤ ns+1, which gives induction base for (11). 96 On the dimension of Kirichenko space Assume that relations (10) and (11) are valid for some 3 ≤ m ≤ lr−1, and proof, that in this case they will be valid for m + 1 also. Consider equality anr+m,k +ak,nr+m+1 = xnr+m, whence ak,nr+m+1 = xnr+m − anr+m,k, that is ak,nr+m+1 = xnr+m − ( zns+1−m+2 − m−1∑ i=1 xnr+i + m−2∑ i=0 xns+1−i ) if k = ns + 1; For the case k = ns + l, 2 ≤ l ≤ m − 1 the formula for ak,nr+m+1 will be ak,nr+m+1 = = xnr+m − ( zns+1−m+l+1 − m−1∑ i=1 xnr+i + l−1∑ i=1 xk−i + m−l−1∑ i=0 xns+1−i ) , and if ns + m ≤ k ≤ ns+1, then ak,nr+m+1 = xnr+m − ( zk−m+1 − m−1∑ i=1 xnr+i + m−1∑ i=1 xk−i ) , which is necessary. Consider equality ak,nr+m+1+anr+m+1,σ(k) = xk, whence for k 6= ns+1 obtain anr+m+1,k+1 = xk − ak,nr+m+1. Substitute k = ns + 1 to the last equality and obtain anr+m+1,ns+2 = xns+1 − ans+1,nr+m+1 = xns+1 − ( m∑ i=1 xnr+i− − m−2∑ i=0 xns+1−i − zns+1−m+2 ) = zns+1−m+2 + xns+1 − m∑ i=1 xnr+i+ + m−2∑ i=0 xns+1−i, that is xnr+m+1,p = zns+1−(m+1)+3 + xp−1− − (m+1)−1∑ i=1 xnr+i + (m+1)−3∑ i=0 xns+1−i for p = ns + 2, which is necessary. Substitute k = ns + l, 2 ≤ l ≤ m− 1 to equality anr+m+1,k+1 = xk − ak,nr+m+1, and obtain anr+m+1,k+1 = xk− ( −zns+1−m+l+1 + m∑ i=1 xnr+i− − l−1∑ i=1 xk−i − m−1−l∑ i=0 xns+1−i ) = zns+1−m+l+1 − m∑ i=1 xnr+i + l∑ i=1 xk+1−i + M. Plakhotnyk 97 m−l−1∑ i=0 xns+1−i, that is anr+m+1,p = = zns+1−(m+1)+l+2 − (m+1)−1∑ i=1 xnr+i + l−1∑ i=1 xp−i + m−l−1∑ i=0 xns+1−i for p = ns + l, 3 ≤ l ≤ (m + 1) − 1, which is necessary. Substitute ns + m ≤ k ≤ ns+1 − 1 to an equality ak,nr+m+1 + anr+m+1,σ(k) = xk, whence, according to assumption of induction, obtain anr+m+1,k+1 = = xk − ak,nr+m+1 = xk − ( m∑ i=1 xnr+i − zk−m+1 − m−1∑ i=1 xk−i ) = = zk−m+1 − m∑ i=1 xnr+i + m−1∑ i=0 xk−i, that is anr+m+1,p = zp−(m+1)+2 − (m+1)−1∑ i=1 xnr+i + (m+1)−1∑ i=1 xp−i for ns + (m + 1) ≤ p ≤ ns+1, which is necessary. Substitute k = ns+1 to an equality anr+m+1,σ(k) = xk − ak,nr+m+1, and obtain anr+m+1,ns+1 = xns+1 − ans+1,nr+m+1 = xns+1 − − ( m∑ i=1 xnr+i − zns+1−m+1 − m−1∑ i=1 xns+1−i ) = zns+1−(m+1)+2− − (m+1)−1∑ i=1 xnr+i + (m+1)−2∑ i=0 xns+1−i, which is necessary to prove. 2. Count relations are corollaries of Gorenstein relations Definition 3. Untrivial relation between parameters is called count re- lation, if it is provided by Gorenstein relations. Proposition 2. Relations (12)-(18) are count relations. If n1 = 2, then for every s, 1 ≤ s ≤ q − 1, equalities { xk−1 + xk = x2, if ns + 2 ≤ k ≤ ns+1, xns+1 + xk = x2, if k = ns + 1 (12) 98 On the dimension of Kirichenko space take place. For every m, 1 ≤ m ≤ n1 − 1, equalities xm+1 = xn1−m+1 (13) take place. If n1 > 2, then for arbitrary s, 1 ≤ s ≤ q − 1, equalities xns+1 + n1−2∑ i=0 xns+1−i = n1∑ i=2 xi, (14) { xns+p = xns+1−n1+p, if 1 ≤ p ≤ n1 xns+p = xns+p−n1 , if n1 < p ≤ ls (15) take place. For every r, 1 ≤ r ≤ q−1, and m, 2 ≤ m ≤ lr−1, equalities (16)-(18) znr+2,r = xnr+1, (16) m∑ i=1 xnr+i − m−2∑ i=0 xnr+1−i = znr+1−m+2,r + znr+m+1,r, (17)    zns+1 − zns+1−(lr−1) = ns+1∑ i=ns+1−(lr−1) xi − nr+1∑ i=nr+1 xi =: f0 zns+p+1 − zns+1−lr+p+1 = = p−1∑ i=0 x(ns+p)−i + lr−p−1∑ i=0 xns+1−i − nr+1∑ i=nr+1 xi =: fp for 1 ≤ p ≤ lr − 1 zns+p+1 − zns+p−(lr−1) = lr−1∑ i=0 xns+p−i − nr+1∑ i=nr+1 xi =: fp for lr ≤ p ≤ ls − 1 (18) take place. The proof of this proposition is broken to some natural parts which we will formulate as lemmas. For convenience of further calculations we will write zk instead of zk,r, if it well be obvious r, 0 < r < q, which is under consideration. Lemma 4. If n1 = 2, then relation (12) takes place for every s, 1 ≤ s ≤ q − 1 . Proof. Assume that n1 = 2 and consider the equality a2,k + ak,1 = a2,1. For ns +2 ≤ k ≤ ns+1, according to (3) obtain xk−1 +xk = x2, which coincides with the first line of (12). According to (3) for k = ns + 1 we obtain xns+1 + xns = x2, which coincides with second line of (12). M. Plakhotnyk 99 Lemma 5. The equality (13) takes place for every m, 1 ≤ m ≤ n1 − 1. Proof. We will prove this lemma by induction for m. Consider equality ak,2 + a2,σ(k) = xk for k = n1, and obtain an1,2 + a2,1 = xn1 . According to definition and according to (2) we have an1,2 = 0 and a2,1 = x2, whence x2 = xn1 , that is (13) for m = 1 and gives the induction base. Let for some m, 1 < m ≤ n1 − 1, the equality (13) takes place for every i < m. Show, that in this case it well take place for i = m also. Substitute k = n1 into equality ak,m+1 + am+1,σ(k) = ak,σ(k), and obtain an1,m+1 +xm+1 = xn1 . To express an1,m+1 we use (4) substituting respectively n1 and m+1 instead of k and m, whence using the third line of (4) we obtain ( m∑ i=2 xi − m−1∑ i=1 xn1−i ) + xm+1 = xn1 , that is m+1∑ i=2 xi = m−1∑ i=0 xn1−i, whence, using induction base obtain that xm+1 = xn1−m+1, which is necessary. Lemma 6. If n1 > 2, then for arbitrary s, 1 ≤ s ≤ q−1 the equality (14) takes place. Proof. For arbitrary s, 1 ≤ s ≤ q−1 consider equality an1,ns+1+ans+1,1 = an1,1. Then according to (5) for m := n1 and k = ns + 1 (for calculating an1,ns+1, according to the first line of this relation), and notation, obtain n1−2∑ i=0 xns+1−i − n1−1∑ i=2 xi + xns+1 = xn1 , that is n1−2∑ i=0 xns+1−i + xns+1 = n1∑ i=2 xi, which is (14). Lemma 7. For every n1 < p ≤ ls the equalities, which are second line of (15), are corollaries of element relations and Gorenstein ones. Proof. Fix arbitrary value of p, n1 < p ≤ ls, and consider equality an1,k + ak,1 = an1,1 for k = ns + p − 1 and k = ns + p, whence obtain n1−1∑ i=0 xns+p−1−i = n1∑ i=2 xi, that is n1∑ i=1 xns+p−i = n1∑ i=2 xi and n1−1∑ i=0 xns+p−i = n1∑ i=2 xi, whence, subtracting the last equality from previous one, obtain xns+p = xns+p−n1 , that proves lemma. 100 On the dimension of Kirichenko space Lemma 8. For every 2 ≤ s ≤ q − 1 and p, 1 ≤ p ≤ n1 the equalities, defined by the first line of (15) are corollaries of Gorenstein relations and element relations. Proof. We prove this lemma by induction for p. According to (2,5), for k = ns + p, 1 < p < n1 equality an1,ns+p + ans+p,1 = an1,1 may be changed to p−1∑ i=1 xk−i + n1−p−1∑ i=0 xns+1−i − n1−1∑ i=2 xi + xns+p = xn1 , whence p−1∑ i=0 xns+p−i + n1−p−1∑ i=0 xns+1−i = n1∑ i=2 xi (19) Consider equality an1,ns+1 + ans+1,1 = an1,1, whence, according to (5), obtain ( n1−2∑ i=0 xns+1−i − n1−1∑ i=2 xi ) + xns+1 = xn1 whence (19) is valid for every p, 1 ≤ p < n1. Using (19) for p = 2 obtain xns+1 + xns+2 + n1−3∑ i=0 xns+1−i = n1∑ i=2 xi. Subtracting from it equality for p = 1, obtain xns+2 = xns+1−(n1−2), and so, the induction base, that is the first line of (15) for p = 2 is proved. Consider arbitrary arbitrary p, from the interval [3, n1−1]. Substitute k = ns + p − 1 and k = ns + p into equality an1,k + ak,1 = an1,1, and using (5) for m = n1 obtain   (p−1)−1∑ i=1 xns+p−1−i + n1−(p−1)−1∑ i=0 xns+1−i − n1−1∑ i=2 xi  + xns+p−1 = xn1 , that is p−1∑ i=1 xns+p−1−i + n1−(p−1)−1∑ i=0 xns+1−i = n1∑ i=2 xi and p−1∑ i=0 xns+p−i + n1−p−1∑ i=0 xns+1−i = n1∑ i=2 xi, whence, subtracting the last equality from previous one, obtain xns+p = xns+1−(n1−p), which proves the validity of lemma for 3 ≤ p ≤ n1 − 1. For k = ns + n1, equality an1,k + ak,1 = an1,1 according to (5) one may change to the form n1−1∑ i=1 xns+n1−i − n1−1∑ i=2 xi + xns+n1 = xn1 , that is M. Plakhotnyk 101 n1−1∑ i=0 xns+n1−i = n1∑ i=2 xi. Subtracting this equality from (19) for p = n1 −1 obtain (15) for p = n1. Let’s return to equality an1,k + ak,1 = xn1 . According to (5), it may be rewritten in the form ( n1−1∑ i=1 xk−i − n1−1∑ i=2 xi ) + xk = xn1 , that is n1−1∑ i=0 xk−i = n1∑ i=2 xi. From the second line of equality (15) (which is proved in previous lemma) obtain, that for arbitrary s, 2 ≤ s ≤ q − 1 equalities xns+1 = xns+n1+1, xns+2 = xns+n1+2 and so on up to xns+1−n1 = xns+1 take place, so sequence {xns+1, . . . , xns+1 } is periodic with period n1. It means that n1−1∑ i=0 xk−i is sum of n1 elements of this sequence, which are equal to it’s last elements, whence obtain equality n1−1∑ i=0 xk−i = n1∑ i=2 xi, which according to (14) one may change to the form n1−1∑ i=0 xns+1−i = n1−2∑ i=0 xns+1−i+xns+1, whence xns+1−n1+1 = xns+1, which is (15) for p = 1, and finishes the proof of equalities (15) in general. Substitute k = nr + 1 into equality ak,nr+1 + anr+1,σ(k) = xk, and obtain znr+2 = xnr+1, which is (16). Lemma 9. For every 2 ≤ m ≤ lr − 1 equalities (17) are corollaries of element relations and Gorenstein ones. Proof. Proof by induction for m. Substitute k = nr+1 into equality ak,nr+2+anr+2,σ(k) = xk and obtain anr+1,nr+2+anr+2,nr+1 = xnr+1 , whence, according to (8), and (9), obtain (xnr+1 − znr+1 ) + (xnr+2 − znr+3) = xnr+1 , that is znr+1 + znr+3 = xnr+1 + xnr+2 − xnr+1 , which is the equality (17) for m = 2 and so induction base is proved. For every m, 3 ≤ m ≤ lr − 1, substitute k = nr+1 into equal- ity ak,nr+m + anr+m,σ(k) = xk, and obtain anr+1,nr+m + anr+m,nr+1 = xnr+1, whence, according to (7) and (10) obtain ( m∑ i=1 xnr+i − m−1∑ i=1 xnr+1−i − znr+1−m+1 ) +(xnr+m+1 − znr+m+2) = xnr+1 , 102 On the dimension of Kirichenko space whence m∑ i=1 xnr+i − m−2∑ i=0 xnr+1−i = znr+1−m+2 + znr+m+1, and lemma is proved. Lemma 10. For arbitrary p, 1 ≤ p ≤ lr −1, equalities, which are second line of (18) are corollaries of element relations and Gorenstein ones. Proof. Consider equality anr+1,k + ak,nr+1 = anr+1,nr+1 = xnr+1 for k = ns + p, p > 0. For p = 1 obtain anr+1,k, using (11) for k = nr + 1 and m = lr. More over, according to (6), ak,nr+1 = xk − zk+1, k 6= ns, whence obtain( − lr−1∑ i=1 xnr+i + lr−2∑ i=0 xns+1−i + zns+1−lr+2 ) +(xns+1−zns+2) = xnr+1 , that is lr−2∑ i=0 xns+1−i + zns+1−lr+2 + xns+1 − zns+2 = nr+1∑ i=n1+1 xi For 2 ≤ p ≤ lr−1 obtain anr+1,k using (11), for k = ns+l, substituting l = p; m = lr, whence ( − lr−1∑ i=1 xnr+i + p−1∑ i=1 xns+p−i + lr−p−1∑ i=0 xns+1−i + zns+1−lr+p+1 ) + (xns+p − zns+p+1) = xnr+1 , that is p−1∑ i=0 xns+p−i + lr−p−1∑ i=0 xns+1−i + zns+1−lr+p+1 − zns+p+1 = nr+1∑ i=nr+1 xi. For p = ns+1 − ns the formula (6) gives ans+1,nr+1 = xns+1 − zns+1 whence obtain lr−1∑ i=0 xns+1−i + zns+1−lr+1 − zns+1 = nr+1∑ i=nr+1 xi, which gives the first line of (18). Lemma 11. For every p, lr ≤ p ≤ ls − 1 equalities, which are second line of (18) are corollaries of element relations and Gorenstein relations. Proof. Consider equality anr+1,k + ak,nr+1 = anr+1,nr+1. Let k = ns + p. M. Plakhotnyk 103 For lr − 1 < p < ls from the formula (11) for k, ns + m ≤ k ≤ ns+1, and m = nr+1 − nr obtain anr+1,k = − lr−1∑ i=1 xnr+i + lr−1∑ i=1 xk−i + + zk−lr+1, and as ns + k < ns+1, then ak,nr+1 = xk − zk+1. Thus obtain( − lr−1∑ i=1 xnr+i + lr−1∑ i=1 xns+p−i + zns+p−lr+1 ) + (xns+p − zns+p+1) = xnr+1 , that is lr−1∑ i=0 xns+p−i + zns+p−lr+1 − zns+p+1 = nr+1∑ i=nr+1 xi, which is necessary. 3. Gorenstein relations are corollaries of element relations and count ones For an arbitrary i, 1 ≤ i ≤ n the equality ai,k + ak,σ(i) = ai,σ(i) for k = i is valid. Later for every i we will consider equalities ai,k +ak,σ(i) = ai,σ(i) and ak,i + ai,σ(k) = ak,σ(k) for k > i, which coincides with the whole set of Gorenstein relations. Lemma 12. Equalities ai,k + ak,σ(i) = ai,σ(i) and ak,i + ai,σ(k) = ak,σ(k) for 0 < i < n1 and i < k ≤ n1 are corollaries of element relations and count ones. Proof. For i = 1 Consider equalities ak,i + ai,σ(k) = ak,σ(k). As the first line of matrix is zero, these equalities go to ak,1 = ak,σ(k) which is corollary of (1). For i = 2 consider equalities ak,i + ai,σ(k) = ak,σ(k), which are equiv- alent to equalities ak,2 + a2,σ(k) = ak,σ(k), which take place in order to (1), (2) and denotations. For 3 ≤ m ≤ n1 consider equality ak,m + am,σ(k) = ak,σ(k), which is ak,m +am,σ(k) = xk. If k < n1, then according to (3) and (4) this equality is equivalent to ( m−1∑ i=2 xi − m−2∑ i=1 xk−i ) + ( m−1∑ i=1 xk+1−i − m−1∑ i=2 xi ) = xk, which is identity. For i = 1 consider equalities ai,k + ak,σ(i) = ai,σ(i). they take place in order to (2) and because the first line of matrix is zero. For i = 2 consider equalities ai,k + ak,σ(i) = ai,σ(i). Consider case, when n1 > 2. In this case σ(2) = 3, and for finding out the formula for ak,σ(i) one may use (4), whence ak,3 = x2 −xk−1, and according to (3) we obtain a2,k = xk−1, whence these equalities are valid. For n1 = 2 one obtain equalities a2,k + ak,1 = a2,1. According to (3) for k 6= ns + 1 they are xns+1 + xns = x2, which coincides with (12) 104 On the dimension of Kirichenko space For 3 ≤ m ≤ n1 the equality am,k +ak,σ(m) = am,σ(m) according to (1) is equivalent to am,k + ak,m+1 = xm. As k ≤ n1, then according to (4) and (5) it is equivalent to identity ( m−1∑ i=1 xk−i − m−1∑ i=2 xi ) +   (m+1)−1∑ i=2 xi − (m+1)−2∑ i=1 xk−i   = xm. For k = n1 consider equality ak,2 + a2,σ(k) = ak,σ(k), that is an1,2 + a2,1 = xn1 , whence according to denotations x2 = xn1 , and it is corollary of (13). For k = n1 and m, 3 ≤ m ≤ n1−1 consider equality ak,m +am,σ(k) = ak,σ(k) for m, 3 ≤ m ≤ n1, that is an1,m + xm = xn1 . And according to (4) one may rewrite it in the form ( m−1∑ i=2 xi − m−2∑ i=1 xn1−i ) + xm = xn1 , that is m∑ i=2 xi = m−2∑ i=0 xn1−i, which is the same as m−1∑ i=1 (x1+i −xn1−i+1) = 0, which is corollary of (13). Lemma 13. For every i, 1 ≤ i ≤ n1 and s, 1 ≤ s ≤ q − 1 equalities ai,k + ak,σ(i) = ai,σ(i) and ak,i + ai,σ(k) = ak,σ(k) for k, ns < k ≤ ns+1 are corollaries of element relations and count ones. Proof. For i = 1 consider equalities ai,k + ak,σ(i) = ai,σ(i) and ak,i + ai,σ(k) = ak,σ(k). They are equivalent to equalities a1,k + ak,2 = a1,2 and ak,1 + a1,σ(k) = ak,σ(k). The former of these equalities are valid in order to (2) and because the first line of matrix is zero. Last is valid in order to definitions, (1), and because the first line of matrix is zero. Let n1 > 2. Then σ(2) = 3. For i = 2 consider equalities ai,k + ak,σ(i) = ai,σ(i) and ak,i + ai,σ(k) = ak,σ(k). They are equal to equalities a2,k +ak,3 = a2,3 and ak,2 +a2,σ(k) = ak,σ(k). According to (4), one obtain ak,3 = { x2 − xns+1 , if k = ns + 1 x2 − xk−1, if k 6= ns + 1 , and according to (3) a2,k = { xns+1 , if k = ns + 1 xk−1, if k 6= ns + 1 , whence the first equality takes place. Second equality takes place in order to (1), (2) and denotations. In the case n1 = 2, equality ai,k+ak,σ(i) = ai,σ(i) for i = 2 is equivalent to a2,k+ak,1 = x2 one. From (3) and denotations, for k = ns+1 we obtain xns+1 + xk = x2 , which is corollary of (12). If ns + 2 ≤ k ≤ ns+1, then the former equality is equivalent to xk−1 + xk = x2, which is corollary of (12). For every m, 3 ≤ m ≤ n1 − 1 equality am,k + ak,σ(m) = am,σ(m) according to (1) is equivalent to am,k + ak,m+1 = xm one. Fix arbitrary 0 < s < q − 1. For using (4) and (5) consider some different cases for k. M. Plakhotnyk 105 For k = ns + 1 obtain the identity ( m−2∑ i=0 xns+1−i − m−1∑ i=2 xi ) + ( m∑ i=2 xi − m−2∑ i=0 xns+1−i ) = xm. For k = ns + l, 2 ≤ l ≤ m − 1 equality am,k + ak,m+1 = xm is equivalent to ( l−1∑ i=1 xk−i + m−l−1∑ i=0 xns+1−i − m−1∑ i=2 xi ) + ( m∑ i=2 xi − l−1∑ i=1 xk−i− − m−1−l∑ i=0 xns+1−i ) = xm one, which is identity. For k, ns+m ≤ k ≤ ns+1 equality am,k+ak,m+1 = xm is equivalent to( m−1∑ i=1 xk−i − m−1∑ i=2 xi ) + ( m∑ i=2 xi − m−1∑ i=1 xk−i ) = xm one, which is identity also. For m = n1 the equality am,k + ak,σ(m) = am,σ(m) according to (1) is equivalent to an1,k + xk = xn1 . Consider cases. Let k = ns + 1. Then according to (5) this equality is equivalent to( n1−2∑ i=0 xns+1−i − n1−1∑ i=2 xi ) +xns+1 = xn1 one, that is n1−2∑ i=0 xns+1−i+xns+1 = n1∑ i=2 xi, and coincides with (14). For k = ns+l and arbitrary l, 2 ≤ l ≤ n1−1 the equality an1,k +xk = xn1 is equivalent to ( l−1∑ i=1 xns+l−i + n1−l−1∑ i=0 xns+1−i − n1−1∑ i=2 xi ) + xns+l = xn1 , that is l∑ i=1 xns+i+ n1−l−1∑ i=0 xns+1−i = n1∑ i=2 xi, whence xns+1+ l∑ i=2 xns+i+ n1−2∑ i=0 xns+1−i− n1−2∑ i=n1−l xns+1−i = n1∑ i=2 xi. Subtracting (14) from this equality one may obtain equivalent equality l∑ i=2 xns+i − n1−2∑ i=n1−l xns+1−i = 0, that is l∑ i=2 xns+i − l∑ i=2 xns+1−n1+i = 0, whence l∑ i=2 ( xns+i − xns+1−n1+i ) = 0, which is corollary of (15). For ns+n1 ≤ k ≤ ns+1 the equality an1,k+xk = xn1 according to (5) is equivalent to ( n1−1∑ i=1 xk−i − n1−1∑ i=2 xi ) + xk = xn1 one, that is n1−1∑ i=0 xk−i = n1∑ i=2 xi. From the second line of equality (15) it follows that equalities xns+1 = xns+n1+1, xns+2 = xns+n1+2 and so on up to xns+1−n1 = xns+1 take place, whence the sequence is periodic with period n1. It means that n1−1∑ i=0 xk−i−1 is the sum of n1 elements of this sequence, and this sum equals 106 On the dimension of Kirichenko space the sum of n1 last elements, whence the equality n1−1∑ i=0 xk−i = n1−1∑ i=1 x1+i according to (14) one may transform to n1−1∑ i=0 xns+1−i = n1−2∑ i=0 xns+1−i + xns+1, whence xns+1−n1+1 = xns+1, which is (15) for p = 1. Now for every m, 3 ≤ m ≤ n1 − 1 consider equality ak,m + am,σ(k) = ak,σ(k). For k = ns + 1, using (4) and (5) one obtains( m−1∑ i=2 xi− m−3∑ i=0 xns+1−i ) + ( 2−1∑ i=1 x(k+1)−i + m−2−1∑ i=0 xns+1−i − m−1∑ i=2 xi ) = xns+1, which is identity. For k = ns + l, and every l, 2 ≤ l ≤ m − 1 according to (4) and (5), the equality ak,m + am,k+1 = ak,k+1 may be transformed to( m−1∑ i=2 xi − l−1∑ i=1 xk−i − m−l−2∑ i=0 xns+1−i ) + ( l∑ i=1 xk+1−i + m−l−2∑ i=0 xns+1−i− − m−1∑ i=2 xi ) = xns+l one, which is identity. For ns + m ≤ k ≤ ns+1 equality ak,m + am,k+1 = ak,k+1 may be transformed to ( m−1∑ i=2 xi − m−2∑ i=1 xk−i ) + ( m−1∑ i=1 xk+1−i − m−1∑ i=2 xi ) = xk one, which is identity. Consider equality ak,n1 + an1,σ(k) = xk, ns + 1 ≤ k ≤ ns+1. For k = ns + 1 obtain ans+1,n1 + an1,ns+2 = xns+1 and using (4) and (5), this is equivalent to( n1−1∑ i=2 xi − n1−3∑ i=0 xns+1−i ) + ( 1∑ i=1 xns+2−i + n1−3∑ i=0 xns+1−i − n1−1∑ i=2 xi ) = xns+1, which is identity. For k = ns + p, 1 < p < n1 − 1 one may obtain ans+p,n1 = n1−1∑ i=2 xi − p−1∑ i=1 xns+p−i − n1−p−2∑ i=0 xns+1−i, and an1,ns+p+1 = p∑ i=1 xns+p−i+1 + + n1−p−2∑ i=0 xns+1−i − n1−1∑ i=2 xi from the relations (4) and (5). Substituting these values to the equality ans+p,n1 + an1,ns+p+1 = xns+p, obtain equiv- alent equality ( n1−1∑ i=2 xi − p−1∑ i=1 xns+p−i − n1−p−2∑ i=0 xns+1−i ) + ( p∑ i=1 xns+p−i+1 + n1−p−2∑ i=0 xns+1−i − n1−1∑ i=2 xi ) = xns+p, which is identity also. For k = ns + p, n1 ≤ p one may obtain ( n1−1∑ i=2 xi − n1−2∑ i=1 xns+p−i ) + + ( n1−1∑ i=1 xns+p+1−i − n1−1∑ i=2 xi ) = xns+p, which is identity, in order to the M. Plakhotnyk 107 equality ak,n1 + an1,σ(k) = xk according to (4) and (5). Lemma 14. For arbitrary r, 0 < r < q − 1, equalities anr+m,k + ak,σ(nr+m) = anr+m,σ(nr+m) and ak,nr+m + anr+m,σ(k) = ak,σ(k) for m, 1 ≤ m ≤ lr and k, nr + 1 < k ≤ nr+1 are corollaries of element relations and count ones. Proof. Consider equality anr+1,k + ak,nr+2 = xnr+1. Using denotations and (8) one may obtain, that it is equivalent to zk +(xnr+1−zk) = xnr+1, which is identity. Consider equality anr+2,k + ak,σ(nr+2) = xnr+2. If lr > 2, then using (9) and (10) one may obtain that it is equivalent to (xk−1 −xnr+1 + zk−1)+ (xnr+1 +xnr+2 −xk−1 − zk−1) = xnr+2, which is identity. If lr = 2, then equality anr+2,k + ak,σ(nr+2) = xnr+2 is equivalent to anr+2,k + ak,nr+1 = xnr+2. As nr + 1 < k ≤ nr+1 then according to conditions of the lemma this equivalence is possible to consider only in the case when k = nr+1 = nr + 1, but in this case it is trivial. For arbitrary m, 3 ≤ m ≤ lr − 1 consider equality anr+m,k + ak,σ(nr+m) = xnr+m. As nr + m < k < nr+1, this equality is anr+m,k + ak,nr+m+1 = xnr+m and according to (10) and (11) is equivalent to ( − m−1∑ i=1 xnr+i + m−1∑ i=1 xk−i+ +zk−m+1 ) + ( m∑ i=1 xnr+i − m−1∑ i=1 xk−i − zk−m+1 ) = xnr+m one, which is identity. Consider equality ak,nr+1 + anr+1,σ(k) = xk for k 6= nr+1. Using (6) and denotations one may obtain that it is equivalent to xk−zk+1+zσ(k) = xk, which is identity. Consider equality ak,nr+2 + anr+2,σ(k) = xk for k 6= nr+1. Using (8) and (9) one may obtain that it is equivalent to (xnr+1 − zk) + (xk − xnr+1 + zk) = xk one, which is identity. For arbitrary m, 3 ≤ m ≤ lr consider equality ak,nr+m +anr+m,σ(k) = ak,σ(k) for nr +m < k < nr+1. According to (11) and (10) it is equivalent to ( m−1∑ i=1 xnr+i − m−2∑ i=1 xk−i − zk−m+2 ) + ( − m−1∑ i=1 xnr+i + m−1∑ i=1 xk+1−i+ +zk−m+2) = xk one, which is identity. For k = nr+1 consider equalities ak,nr+m + anr+m,σ(k) = ak,σ(k), that is anr+1,nr+m + anr+m,nr+1 = xnr+1 for 1 ≤ m < lr. For m = 1 one obtains anr+1,nr+1 + anr+1,nr+1 = anr+1,nr+1, which is valid. 108 On the dimension of Kirichenko space For m = 2 the equality anr+1,nr+m + anr+m,nr+1 = xnr+1 is anr+1,nr+2 + anr+2,nr+1 = xnr+1 and according to (6) and (8) it is equiv- alent to (xnr+1 − znr+1 ) + (xnr+2 − znr+3) = xnr+1 one, which follows from (17) for m = 2. For 3 ≤ m ≤ lr−1 consider equality anr+1,nr+m+anr+m,nr+1 = xnr+1 , and, according to (10) and (6) this equality is equivalent to ( m−1∑ i=1 xnr+i− − m−2∑ i=1 xnr+1−i − znr+1−m+2 ) + (xnr+m − znr+m+1) = xnr+1 , whence m∑ i=1 xnr+i − m−2∑ i=0 xnr+1−i = znr+1−m+2 + znr+m+1, which follows from (17) for 3 ≤ m ≤ lr − 1. For m = lr equality anr+1,nr+1 + anr+1,nr+1 = xnr+1 is corollary of (7). Lemma 15. For an arbitrary r, 0 < r < q − 1, equalities anr+m,k + ak,σ(nr+m) = anr+m,σ(nr+m) and ak,nr+m + anr+m,σ(k) = ak,σ(k) for every s, r < s < q; m, 1 ≤ m ≤ lr and k, ns + 1 < k ≤ ns+1 are corollaries of element relations and count ones. Consider equality anr+1,k+ak,nr+2 = xnr+1. According to denotations and (8) it is equivalent to zk + (xnr+1 − zk) = xnr+1, which is identity. Consider equality ak,nr+1 + anr+1,σ(k) = xk. According to (6) and denotations it is equivalent to { xk − zns+1, if k = ns+1 xk − zk+1, if k 6= ns + zσ(k) = xk, which is identity. Consider equality anr+2,k + ak,σ(nr+2) = xnr+2. If lr > 2, then σ(nr + 2) = nr + 3, and according to (9) and (10) it is equivalent to { xns+1 − xnr+1 + zns+1 , if k = ns + 1 xk−1 − xnr+1 + zk−1, if k 6= ns + 1 + + { xnr+1 + xnr+2 − xns+1 − zns+1 , if k = ns + 1. xnr+1 + xnr+2 − xk−1 − zk−1, if k 6= ns + 1 = xnr+2, which is identity. If lr = 2, then equality anr+2,k + ak,σ(nr+2) = xnr+2 is equivalent to anr+2,k + ak,nr+1 = xnr+2. Let ns+1 − ns > 2. Then according to (6) and (9) for k = ns+1 it is the same as (xk − zns+1) + (xk−1 − xnr+1 + zk−1) = xnr+2. For k = ns + 1 it is the same as (xns+1 − xnr+1 + + zns+1) + (xk − zk+1) = xnr+2. And for ns + 1 < k < ns+1 the equality anr+2,k + ak,nr+1 = xnr+2 is equivalent to (xk−1 − xnr+1 + zk−1) + (xk − zk+1) = xnr+2. Note, that all of these equalities are corollaries of (18). Consider equality ak,nr+2 + anr+2,σ(k) = xk. According to (9) it is equivalent to anr+2,σ(k) = { xns − xnr+1 + zns , if k = ns xk − xnr+1 + zk, if k 6= ns , whence, ac- M. Plakhotnyk 109 cording to (8) it is equivalent to (xnr+1 − zk) + { xns − xnr+1 + zns , if k = ns x(k+1)−1 − xnr+1 + zk, if k 6= ns = xk which is identity. For arbitrary 3 ≤ m ≤ lr consider equality ak,nr+m + anr+m,σ(k) = ak,σ(k). Consider cases for k. For k = ns+1 according to (1) it is ans+1,nr+m+anr+m,ns+2 = xns+1, whence, according to (10) and (11) it is equivalent to ( m−1∑ i=1 xnr+i − − m−3∑ i=0 xns+1−i − zns+1−m+3 ) + ( − m−1∑ i=1 xnr+i + 2−1∑ i=1 x(ns+2)−i+ m−2−1∑ i=0 xns+1−i + zns+1−m+2+1 ) = xns+1, which is identity. For k = ns + l, and l, 2 ≤ l ≤ m − 1, according to (10) and (11) the equality ans+l,nr+m+anr+m,ns+l+1 = xns+l is equivalent to ( m−1∑ i=1 xnr+i − − m−l−2∑ i=0 xns+1−i − l−1∑ i=1 xk−i − zns+1−m+l+2 ) + ( l∑ i=1 x(k+1)−i − m−1∑ i=1 xnr+i+ m−l−2∑ i=0 xns+1−i + zns+1−m+l+2 ) = xns+l, which is identity. For every l, ns +m ≤ l ≤ ns+1−1 equality ak,nr+m +anr+m,k+1 = xk according to (10) and (11) is equivalent to( m−1∑ i=1 xnr+i − m−2∑ i=1 xk−i − zk−m+2 ) + ( − m−1∑ i=1 xnr+i + m−1∑ i=1 x(k+1)−i+ +zk−m+2) = xk, which is identity. Consider equality anr+m,k+ak,σ(nr+m) = xnr+m for m = lr and obtain anr+1,k + ak,nr+1 = xnr+1 . For k = ns + 1 obtain anr+1,ns+1 + ans+1,nr+1 = xnr+1 , whence, ac- cording to (11) and (6), it is equivalent to ( − lr−1∑ i=1 xnr+i + lr−2∑ i=0 xns+1−i+ zns+1−lr+2 ) +(xns+1 − zns+2) = xnr+1 , whence lr−2∑ i=0 xns+1−i+zns+1−lr+2+ xns+1 − zns+2 = lr∑ i=1 xnr+i which is (18) for p = 1. For k = ns+l and l, 2 ≤ l ≤ lr−1, the equivalence anr+1,k+ak,nr+1 = xnr+1 is equivalent to ( − lr−1∑ i=1 xnr+i + l−1∑ i=1 xk−i + lr−l−1∑ i=0 xns+1−i+ +zns+1−lr+l+1 ) + (xns+l − zns+l+1) = xnr+1 , whence l−1∑ i=0 xns+l−i+ 110 On the dimension of Kirichenko space + lr−l−1∑ i=0 xns+1−i + zns+1−lr+l+1 − zns+l+1 = lr∑ i=1 xnr+i = nr+1∑ i=nr+1 xi, which is (18) for 2 ≤ p ≤ lr − 1. For k = ns+l and l, lr ≤ l ≤ ls−1, the equivalence anr+1,k+ak,nr+1 = xnr+1 according to (11) and (6) is equivalent to ( − lr−1∑ i=1 xnr+i+ + lr−1∑ i=1 xns+l−i + zns+l−lr+1 ) + (xns+l − zns+l+1) = xnr+1 , whence lr−1∑ i=0 xns+l−i + zns+l−lr+1 − zns+l+1 = nr+1∑ i=nr+1 xi, which is (18) for lr ≤ p ≤ ls − 1. For k = ns+1 the equality anr+1,k + ak,nr+1 = xnr+1 according to (11) and (6) is equivalent to ( − lr−1∑ i=1 xnr+i + lr−1∑ i=1 xns+1−i + zns+1−lr+1 ) + + ( xns+1 − zns+1 ) = xnr+1 , whence lr−1∑ i=0 xns+1−i + zns+1−lr+1 − zns+1 = nr+1∑ i=nr+1 xi, which coincides with the first line of (18). 4. Lemma about permutation Let’s prove lemma for counting the number of cycles the permutation, whose bottom line of standard representation is (l + 1, . . . , n, 1, . . . , l) is decomposed to. This lemma will be used for counting the dimension of Kirichenko space. Lemma 16. The permutation π = ( 1 l + 1 · · · n − l n − l + 1 n 1 · · · n l ) is decomposed to (n, l) cycles, where (n, l) is greatest common divisor of numbers n and l. In this case each cycle consists of numbers which give the same remainder in division by (n, l). Proof. Remark, that if n is divisible by l, then lemma is obvious. Let n be indivisible by l. Denote by ξ(π) = ξ(n, l) the quantity of cycles in the decomposition of the permutation of the type, specified above. Remark, that this quantity is equal to one for permutation π−1 = ( 1 n − l + 1 · · · l l + 1 n 1 · · · n n − l ) , M. Plakhotnyk 111 whence obtain relation ξ(n, l) = ξ(n, n − l). Consider the cases n > 2l and n < 2l. Let n > 2l. For n > 3l we will show that ξ(n, l) = ξ(n − l, l) which will give us a reason to reduce this case to 2l < n < 3l one. Write out the permutation π, decomposing it into blocks, as π = ( 1 l + 1 · · · l 2l l + 1 2l + 1 · · · 2l 3l 2l + 1 3l + 1 · · · n − l n n − l + 1 1 · · · n l ) . Pay attention to that cycles of π which contain elements of the first block, which is interval [1, l]. Consider some element xi ∈ [1, l] of some cyclic trajectory, and it’s obvious that for the previous element xi−1 of this trajectory we have enclosure xi−1 ∈ [n − l + 1, n], which means, that xi−1 belongs to the last block of π. It is also obvious that for the next element of trajectory xi+1 we have inclosure xi+1 ∈ [l+1, 2l] which means, that is belongs to the second block of π. More over, as lengthes of the first, second and the last cycles are equal to each other, then for arbitrary element xi−1 ∈ [n − l + 1], which belongs to some cyclic trajectory, an element xi+1 of the same trajectory will belong to interval [l + 1, 2l]. That is why the quantity of cycles, which the permutation π is de- composed to, is equal to one for permutation π1, where π1 = ( l + 1 2l + 1 · · · 2l 3l 2l + 1 3l + 1 · · · n − l n n − l + 1 l + 1 · · · n 2l ) . Reducing all numbers, which figure in the record of the permutation π1 on l we will not change the permutation itself and so we will not change the quantity of cycles it is decomposed to, and obtain π2 = ( 1 l + 1 · · · l 2l l + 1 2l + 1 · · · n − 2l n − l n − 2l + 1 1 · · · n − l l ) . So, ξ(π) = ξ(π2), whence ξ(n, l) = ξ(n − l, l). Repeating some times, if necessary, these reasonings, obtain ξ(n, l) = ξ(n − pl, l), where 2l < n − pl < 3l. Then ξ(π2) = ξ(π̃2), where π̃2 = ( 1 l + 1 · · · l 2l l + 1 2l + 1 · · · n1 − l n1 n1 − l + 1 1 · · · n1 l ) , and n1 = n − pl. Let us consider the permutation π3 = π̃−1 2 , and denote k = n1 − l. As 2l < n1 < 3l, then 3k < 2n1 < 4k. So the permutation π3 112 On the dimension of Kirichenko space looks like π3 = ( 1 k + 1 · · · n1 − k n1 n1 − k + 1 1 · · · k 2k − n1 k + 1 2k − n1 + 1 · · · n1 k ) . Notice that number k does not belong to the first line of the first block, i.e. k > n1 − k (because 4k > 2n1), which gives possibility to decompose the permutation π3 to three blocks, taking k as a border between second and third block. We may note also, that number 2n1 − 3k belongs to the first block, because 2n1 − 3k < n1 − k (as 2n1 < 4k), and the number 2n1 − 2k belongs to third block, because k +1 < 2n1 − 2k +1 (as 2n1 > 3k). That is why, the permutation π3 may be represented as π3 = ( 1 k + 1 · · · 2n1 − 3k 2n1 − 2k 2n1 − 3k + 1 2n1 − 2k + 1 · · · n1 − k n1 n1 − k + 1 1 · · · k 2k − n1 k + 1 2k − n1 + 1 · · · 2n1 − 2k n1 − k 2n1 − 2k + 1 n1 − k + 1 · · · n1 k ) , that is both the first and third blocks are decomposed to two blocks. Pay attention on elements of the forth block. Like the way we have considered the permutation π, it is easy to see that each cycle which contains some element xi of the forth block contains some element xi−1 of the first block, more over, each cyclic trajectory which contains some element of the first block contains some element of the forth block on the next place. Pay attention on elements of the fifth block. Each cycle, which contains some element xi ∈ [2n1 − 2k + 1, n1], contains also xi−1 ∈ [2n1 − 3k + 1, n1 − k] (from the second block) and xi+1 ∈ [1, 2k − 1] (from the third block), that is why, like in considering the permutation π, it is possible to regard that elements of interval [2n1 − 3k + 1, n1 − k] (second block) go accordingly to elements of interval [1, 2k − n1] (third block) which gives us possibility to state that the number of cycles, the permutation π3 is decomposed to, is equal to one for permutation π4 = ( 1 2k − n1 + 1 · · · 2n1 − 3k n1 − k 2n1 − 3k + 1 1 · · · n1 − k 2k − n1 ) . M. Plakhotnyk 113 So one may see that ξ(n1, n1 − k) = ξ(n1 − k, 2k − n1) = ξ(l, n1 − 2l) = ξ(l, n1 mod l) = = ξ(l, n mod l). Let n < 2l. then the permutation π looks like π = ( 1 l + 1 · · · n − l n n − l + 1 1 · · · l 2l − n l + 1 2l − n + 1 · · · n l ) . Reasoning like in consideration of the permutation in the case n > 3l it is possible to state that permutation π is decomposed to the same quantity of cycles as permutation π5 π5 = ( 1 2n − l + 1 · · · n − l l n − l + 1 1 · · · l 2l − n ) , whence ξ(n, l) = ξ(l, 2l−n) = ξ(l, l−(2l−n)) = ξ(l, n−l) = ξ(l, n mod l). So we have obtained that for every n and l the equality ξ(n, l) = ξ(l, n mod l) takes place, which is the equality used in the algorithm of finding the greatest common divisor which, as known, may be obtained by this step, after using it some number of times. Show, that all the numbers which belong to the same cycle give the same remainders in division by (n, l). Let x be an element of some cycle. If x+ l ≤ n, then the next element of this cycle is x+ l, which, obviously, gives the same remainder in dividing by l as x does and so the same one in dividing by (n, l). If x+ l > n then the next element of the cycle is x+ l−n, which also gives the same remainder in dividing by (n, l) as x does. 5. The dimension of Kirichenko space. The purpose of this section is to count the defect of the matrix of count relations which is the dimension of Kirichenko space. Lemma 17. Consider the relation l1 ls∑ i=1 xns+i = ls l1∑ i=2 xi. (14a) Systems of relations (14; 15) and (14a; 15) are equivalent. Proof. Consider the system of relations n1−2∑ i=0 xns+1−i + xns+1 = n1∑ i=2 xi, (14) 114 On the dimension of Kirichenko space { xns+p = xns+1−(n1−p) 1 ≤ p ≤ n1 xns+p = xns+p−n1 n1 < p ≤ ns+1 − ns (15) According to the first line of the last relation, the relation (14) is equivalent to n1∑ i=1 xns+i = n1∑ i=2 xi (14b) one. From second line of equalities (15) we may obtain xns+n1+1 = xns+1, xns+n1+2 = xns+2 and so on up to xns+1 = xns+1−n1 , that is the set of parameters {xns+1, . . . , xns+1 } is periodic with period n1. Here consider the periodicity in the sense, that if i − j is divisible by n1, then xi = xj . Without bounding of generality one may consider this set as consisted of elements {xns+1, . . . , xns+n1 }. It goes from the first line of equalities (15) that the last n1 elements of a set {xns+1, . . . , xns+1 } coincide with {xns+1, xns+2, . . . , xns+n1 } cor- respondingly. Taking into account the fact that the set {xns+1, . . . , xns+1 }, is pe- riodic, it is easy to see that the problem of calculating the quantity of different parameters of the set {xns+1, . . . , xns+n1 } is equivalent to one of finding the quantity of different cycles, the permutation π, π = ( 1 l + 1 · · · n1 − l n1 − l + 1 n1 1 · · · n1 − 1 n1 l − 1 l ) , which the first line corresponds to the last n1 elements of a set {xns+1, . . . , xns+1 } (which are equal to former ones), and second line cor- responds to the last elements of a set {xns+1, . . . , xns+1 } is decomposed to. According to lemma 16 (about permutation), the set of parameters {xns+1, . . . , xns+1 } is decomposed to (ls, l1) sets of pairwise equal ones, and indices of elements of each cycle give the same remainders in division by (ls, l1). Consider ones more the equality (14b) which is n1∑ i=1 xns+i = n1∑ i=2 xi. Taking into account facts proved above, sum of first n1 parameters is equal to sum of all (ls, l1) pairwise unequal ones taken l1 (ls, l1) times. So, it is obvious, that (ls, l1) l1 l1∑ i=1 xns+i is sum of one gang of pairwise M. Plakhotnyk 115 unequaled parameters, whence l1∑ i=1 xns+i = l1 ls ls∑ i=1 xns+i, which is l1 ls∑ i=1 xns+i = ls l1∑ i=2 xi. (14a) Using the same steps (with revers order) one may prove that the system of equations (14; 15) is a corollary of (14a; 15). Lemma 18. If n1 = 2, then the system of equations { xk−1 + xk = x2 ns + 2 ≤ k ≤ ns+1, xns+1 + xk = x2 k = ns + 1 (12) is a partial case of relations system (14a; 15). Proof. Write out the matrix of relation (12) and obtain   1 1 0 0 · · · 0 x2 0 1 1 0 · · · 0 x2 ... . . . . . . . . . . . . ... ... 0 · · · 0 1 1 0 x2 0 0 · · · 0 1 1 x2 1 0 · · · · · · 0 1 x2   If one subtract from each line (except the first) the previous one, and add all the lines to the last one then the matrix which determine the system of equations it will appear. This matrix is equivalent to former one, where (just for convenience) it is possible to add the last line of a former matrix to penultimate place, and obtain matrix   1 0 −1 0 · · · 0 0 0 1 0 −1 · · · 0 0 ... . . . . . . . . . . . . ... ... 0 · · · 0 1 0 −1 0 −1 0 · · · 0 1 0 0 0 –1 · · · 0 0 1 0 2 2 · · · · · · 2 2 lsx2   where bold type is used for added line. It’s obvious, that this system coincides with the system of relations (14a, 15) for n1 = 2. 116 On the dimension of Kirichenko space Thus, the problem of calculating the dimension of Kirichenko space has reduced to finding the defect of matrix K whose columns correspond to parameters {x2, . . . xn, zn1+2,1, zn1+3,1 . . . zn,1, zn2+2,2, . . . zn,2 . . . . . . , znq−1+2,q−1, . . . zn,q−1} and lines correspond to relations (13; 14a; 15- 18) ordered in a natural way. It’s obvious, that this matrix looks like K = ( X 0 X ′ Z ) =   X 0 0 · · · 0 X1 Z1 0 · · · 0 X2 0 Z2 · · · 0 ... ... . . . . . . ... Xq−1 0 · · · 0 Zq−1   , where matrices X, X1, . . . , Xq−1, Z1, . . . Zq−1 are block matrices, matrix X corresponds to relations (13), and each block ( Xr 0 Zr 0 ) cor- responds to count relations (14a; 15-18) for every r, 1 ≤ r ≤ q − 1. Blocks which correspond to different values of r have similar form and in some sense are differ only by dimension. Consider their structure in more details. Like in proving relations, we may fix arbitrary r, 1 ≤ r ≤ n − 1 and so, for admissible k instead of zk,r we will write zk. Consider the system of relations m∑ i=1 xnr+i − m−2∑ i=0 xnr+1−i = znr+m+1 + znr+1−m+2, 2 ≤ m ≤ lr − 1. (17) Fix arbitrary m0, 2 ≤ m0 ≤ lr − 1. It’s obvious that parameter znr+m0+1 will be met in the equality for m = m0 and in the equality for such m = m1, that nr+1 −m1 + 2 = nr + m0 + 1, because znr+m0+1 may be met not only as first, but also as second item in relations (17). Let’s find indices for z, which will appear in relations for m = m1. Sure, one of them will be nr + m0 + 1. As nr+1 − m1 + 2 = nr + m0 + 1, then m1 = nr+1 − nr − m0 + 1, whence the index, we need, is nr + m1 + 1 = nr+1−m0+2, which coincides with index of second element in the relation for m = m0. It means, for any relation in (17), there is one more, which has the same right side. Show, that these two relations coincide, that is they are one relation, written out two times. This statement may be formulated as such lemma. Lemma 19. For every r, 1 ≤ r ≤ q − 1 and every m, 2 ≤ m ≤ lr − 1 denote f(m) = m∑ i=1 xnr+i − m−2∑ i=0 xnr+1−i. Then f(m) = f(nr+1 − nr − m + 1). M. Plakhotnyk 117 Proof. Let’s prove, that f(m) − f(nr+1 − nr − m + 1) = 0. f(m) − f(nr+1 − nr − m + 1) = ( m∑ i=1 xnr+i − m−2∑ i=0 xnr+1−i ) − − ( nr+1−nr−m+1∑ i=1 xnr+i −− nr+1−nr−m−1∑ i=0 xnr+1−i ) = ( nr+m∑ i=nr+1 xi− − nr+1∑ i=nr+1−m+2 xi  − ( nr+1−m+1∑ i=nr+1 xi − nr+1∑ i=nr+m+1 xi ) = 0 Thus, lines of matrix, which correspond to fixed r and relation (17) looks like   Xr,2 0 1 0 0 · · · 0 0 1 0 Xr,3 0 0 1 0 · · · 0 1 0 0 Xr,4 0 0 0 1 · · · 1 0 0 0 ... ... 0 0 0 . . . 0 0 0 0 Xr,lr−3 0 0 0 1 . . . 1 0 0 0 Xr,lr−2 0 0 1 0 · · · 0 1 0 0 Xr,lr−1 0 1 0 0 · · · 0 0 1 0   where by the bold font we type block-lines, and the first half of lines of this part of K are pairwise different, but others are copies of some first lines. The problem of calculating the defect of this part of K is reduced to calculating the quantity of independent parameters in the set (17). This set is decomposed into independent pairs. Each such pair is relation, which looks like sum zi + zj equals to some expression, depends only on the set x2, . . . , xn. It is obvious, that if nr+m0+1 6= nr+1+m0+2 for any m0, then one of parameters znr+m0+1, znr+1+m0+2 we may consider as independent, and the other as expressed through independent ones, and so in this case, the relation (17) for m = m0 gives us one independent parameter. If for some m0 we have equality nr +m0 +1 = nr+1 +m0 +2, then znr+m0+1 is expressed through parameters of the set x2, . . . , xn and is dependent. That’s why the quantity of independent parameters equals lr − 2 2 , if lr is even, and equals lr − 3 2 , if lr is odd, which is [ lr − 2 2 ] in general case and is equal to the defect of matrix, determined by (17). Consider the part of matrix K which corresponds to the system of relations (18), and looks like ( X(r) 0 Z(r) 0 ) (indices are bracketed 118 On the dimension of Kirichenko space to make difference with denotations for blocks, which appeared in some previous propositions) which may be written in the matrix form as Z(r) =   1 0 · · · 0 −1 0 · · · 0 f0 0 1 · · · 0 0 −1 · · · 0 f1 ... ... . . . . . . ... ... 0 0 1 0 0 · · · −1 fls−lr−1 −1 0 · · · 0 1 0 · · · 0 fls−lr 0 −1 · · · 0 0 1 · · · 0 fls−lr+1 ... ... . . . . . . ... ... 0 0 · · · −1 0 0 · · · 1 fls−1   . Let’s show, that by means of interchanging lines and columns, this matrix may be reduced to block-diagonal form. The possibility of such reduc- ing means that the set of parameters, which correspond to columns of this matrix may by decomposed into such disjunct join, that there is no relation, which connects parameters from different sets. Write out the connections between parameters zns+1, . . . , zns+ls in the form of permu- tation π, where equality π(i) = j will mean that there is an equation zns+j − zns+i = fj+1 in the system (18). From the fact that for each i, 1,≤ i ≤ ls, parameter zns+i is met only two times i.e. once with positive sign and once with negative one, we obtain that the quantity of blocks the left part of Z(r) may be decomposed to is the quantity of cycles the permutation π is decomposed to. The first two lines of system (18) give that π will be like π = ( · · · ls − lr + 1 ls − lr + 2 · · · ls · · · 1 2 · · · lr ) . And in the same time, the last line of (18) gives, that π looks like ( 1 · · · ls − lr · · · lr + 1 · · · ls · · · ) , and in general π is π = ( 1 2 · · · ls − lr ls − lr + 1 · · · ls lr + 1 lr + 2 · · · ls 1 · · · lr ) . According to lemma 16 it is decomposed into composition of (ls, lr) = t cycles, and each cycle in this decomposition contains exactly w = ls t elements. Note, that all elements from each cycle in this decomposition give the same remainders in division by t. M. Plakhotnyk 119 Make the simultaneous interchanging of lines and columns of matrix Zr determined by permutation τ which looks as( 1 π(1) π2(1) . . . πw−1(1) 2 π(2) · · · πw−1(2) · · · 1 2 3 . . . w w + 1 w + 2 · · · 2w · · · · · · t · · · πw−1(t) · · · ls − w + 1 · · · ls ) . After this matrix Z(r) will be transformed into   Zr,1 · · · 0 Fr,1 ... . . . ... ... 0 · · · Zr,t Fr,t   where Fr,i is column of functions f̃(i−1)t+1, f̃(i−1)t+2, . . . f̃(i−1)t+w, ob- tained during transformations, and matrices Zr,i are equal and look like Zr,i =   1 −1 0 · · · 0 0 1 −1 · · · 0 ... ... . . . . . . ... 0 0 · · · 1 −1 −1 0 · · · 0 1   . Let’s do transformations with lines of each block of obtained matrix. In each block of obtained matrix add all previous lines to each one except the first line (because it has no previous), and obtain matrix Z̃r,i =   1 −1 0 · · · 0 f̃(i−1)t+1 1 0 −1 · · · 0 f̃(i−1)t+1 + f̃(i−1)t+2 ... ... . . . . . . ... ... 1 0 · · · 0 −1 w−1∑ j=1 f̃(i−1)t+j 0 0 · · · 0 0 w∑ j=1 f̃(i−1)t+j   . Now, my means of addition of all columns of matrix Z̃r,i except the first one to other columns of matrix K with necessary coefficients we may make zero all elements of Z̃r,i except ones, belongs to it’s first sub diagonal, which will stay be equal to −1. Show that the sum w∑ j=1 f̃(i−1)t+j doesn’t depend on i. The last propo- sition is a corollary of next lemma. 120 On the dimension of Kirichenko space Lemma 20. For every common divisor t of numbers lr and ls and for every b, 1 ≤ b ≤ t the equality (ls/t)−1∑ i=0 f̃b−1+it = lr t ls∑ i=1 xns+i− ls t lr∑ i=1 xnr+i takes place. Proof. As functions f̃0, . . . f̃ls−1 are obtained from the set of ones {f0, . . . fls−1} after interchanging of lines and columns of matrix, then for every p, 0 ≤ p ≤ lq − 1 the equality f̃p = τ∗(fτ(p+1)−1), where τ∗ is automorphism of the set {xns+1, . . . xns+1 } such that τ∗(xns+i) = xns+τ(i) for every i, 1 ≤ i ≤ ls takes place. As for every b, 1 ≤ b ≤ t the set {b − 1 + it, 1 ≤ b ≤ t} is invariant for τ , then the proposition of lemma is equivalent to (ls/t)−1∑ i=0 fb−1+it = lr t ls∑ i=1 xns+i − ls t lr∑ i=1 xnr+i, which we will prove. Denote gk = fk − nr+1∑ i=nr+1 xi. It is obvious that it is necessary and enough for proving lemma to prove that (ls/t)−1∑ i=0 gb−1+it = lr t ls∑ i1 xns+i, because items − nr+1∑ i=nr+1 xi are met exactly once in each fk, and that is why alter taking the sum of fk we will obtain that quantity of appearing of − nr+1∑ i=nr+1 xi is ls t . Prove, that (ls/t)−1∑ i=0 gb−1+it = lr t ls∑ i1 xns+i. Let us write out the matrix G, which columns correspond to parameters xns+1, . . . , xns+1 and ls lines correspond to relations g0, . . . gls−1. When k runs through the interval [0, lr − 1], then lines of matrix G which correspond to relations gk look like G∗∗ =   0 0 0 · · · · · · 1 1 1 · · · 1 1 1 1 0 0 · · · · · · 0 1 1 · · · 1 1 1 1 1 . . . . . . · · · 0 0 . . . . . . 1 1 1 1 1 . . . 0 · · · 1 0 0 . . . . . . 1 1 1 1 . . . 1 · · · 1 . . . 0 . . . 0 . . . 1 1 1 . . . 1 1 1 · · · 1 0 · · · 0 1   , or G∗∗ =   0 0 0 · · · · · · 0 1 1 1 · · · 1 1 1 1 0 0 · · · · · · 0 0 1 1 · · · 1 1 1 1 1 0 · · · · · · 0 0 0 1 · · · 1 1 1 1 1 . . . . . . · · · 0 0 0 . . . . . . 1 1 1 1 1 . . . 1 0 0 0 0 0 · · · 0 1 1 1 1 . . . 1 1 0 0 0 0 · · · 0 0 1   M. Plakhotnyk 121 which depends on whether 2lr ≥ ls, or not, where the first line of G∗∗ contains lr identities and ls − lr zeros. This matrix is decomposed in the natural way to blocks Gl and Gr, the former of which corresponds to parameters xns+1, . . . , xns+1−lr and the last - to xns+1−lr+1, . . . , ns+1 ones. When k runs through interval [lr, ls − 1], then the part of matrix G, whose lines correspond to relations gk for these k looks like G∗ =   1 1 1 · · · · · · 1 1 0 0 0 · · · 0 0 1 1 · · · · · · 1 1 1 0 0 · · · 0 0 0 1 · · · · · · 1 1 1 1 0 · · · 0 . . . . . . · · · . . . ... 0 0 · · · 0 1 1 1 · · · · · · 1 1 0   where the former lr elements of the first line are equal to identity, and the last ones are zero. The statement of this lemma in terms of matrix G is the following. If we write out the lines of g with the step t, and arbitrary beginning b, 1 ≤ b ≤ t then the sum of ones in each column will be the same and will be equal to lr t . Note that as the quantity of lines of matrix G∗∗ is equal to lr which is divisible by t, then such writing out the lines means that we obtain a matrix Gb, which consists of writing out one above another matrices G∗ b and G∗∗ b , each of them is obtained by writing out the lines of matrices G∗∗ and G∗ with the step t starting from b. Denote by Glb and Grb those parts of matrix G∗∗ b which are obtained during writing out the lines with step t. Let’s show that matrix Gb will be column-block matrix, i.e. it will be consisted of such vertical blocks with ls t lines that each of them will have the same element in each line. We will say about such matrix, that it satisfies column-block condition. Let’s show, that for G∗∗ b the correspond widthes of blocks will be b− 1, t, t, . . . , t, t− b + 1, b− 1, t, t, . . . , t, t− b + 1, and the first blocks of matrices Glb and Grb have width b − 1. For proving this proposition, consider matrices Glb, Grb and G∗ b sep- arately. Consider matrix Glb. The first line of this matrix have b− 1 ones and zeros at the other places, and satisfies column-blocks condition. Each next line of this matrix will have t ones more at the beginning and t zeros less at the end and so, satisfies the column-block condition also. Consider matrix Grb. The first line of this matrix have b − 1 zeros and ones at the other places, and satisfies the column-block condition. Each next line of this matrix will have t zeros more at the beginning and t ones less at the end and so, satisfies the column-block condition also. 122 On the dimension of Kirichenko space Consider matrix G∗ b . The first line of this matrix have b − 1 zeros at the beginning, then lr ones, and at last (ls − lr − t) + (t− b + 1) zeros at the end, and satisfies the column-block condition. Each next line of this matrix has t zeros more at the beginning, and t zeros less at the end, and so, satisfies the column-block condition also. Let us count the quantity of ones in the first column of matrix Gb that is equal to this quantity for the first block-column. If b = 1 then each of lr t lines except the first one of matrix Glb will have some number of ones at its beginning, and the first one will have zero there, so the quantity of ones in the first block of matrix Glb is lr t − 1. At the same time, the first line of the matrix G∗ b has one at its beginning and other ones have zero at their beginnings, so the quantity of ones of the first block of matrix Gb is lr t . If b > 1, then the first element of each line of matrix Glb will be equal to one, but the first element of each line of matrix G∗ b will be equal to zero, whence the quantity of ones at first places of all lines of Gb is equal to lr t . It’s easy to see that every next block of Glb has exactly 1 one less, but every next block of G∗ b has 1 one more, so general quantity of ones is not changed and is equal to lr t for all columns from the first up to one number ls − lr. Count the quantity of ones in the last column of matrix Gb. The last element of each of lr t lines of matrix Grb is equal to one, and in the came time the last element of each line of G∗ b is equal to zero. That is why the quantity of ones in the last column of Gb is equal to lr t . It’s easy to see that moving left from last block of Grb each block we meet, will have 1 one less then previous in matrix Grb and 1 one more in the matrix G∗ b , saving the general quantity of ones in columns with numbers from ls − lr +1 up to ls. So the quantity of ones in each column of Gb is equal to lr t . Corollary 1 of lemma: The part ( X(k) 0 Z(k) 0 ) of matrix K which corresponds to count relations (18) for fixed r by transforming lines and columns may be reduced to being like Ns =   lr ls∑ i1 xns+i − ls lr∑ i1 xnr+i 0 0 0 0 0 E 0   , M. Plakhotnyk 123 where E is identity matrix of order ls − (ls, lr). Corollary 2 of lemma: Using transformations of lines and columns, matrix K = ( X 0 X ′ Z ) may be reduced to being like   X 0 0 X ′′ 0 0 0 E 0   , where E in such identity matrix that ( E 0 ) has defect q−1∑ s=2 q∑ r=s+1 (ls, lr) + q−2∑ s=2 [ ls − 2 2 ] , and matrix X ′′ looks like X ′′ =   0 L3 −L2 0 0 L4 0 −L2 ... . . . 0 Lq−1 0 0 · · · −L2 0 0 Lq 0 0 · · · 0 −L2 0 0 Ls+1 −Ls 0 0 0 Ls+2 0 −Ls ... . . . 0 0 Lq 0 0 −Ls ... 0 0 Lq−1 −Lq−2 0 0 0 Lq 0 −Lq−2 0 0 0 Lq −Lq−1   where all elements at full places are zero, and by Li we denote a matrix which has one line all whose elements are equal to li. In this case the quantity of blocks-columns of matrix is equal to q, the width of first of them is equal to l1 − 1, and the widthes of others are equal to li, i > 1. Let’s go to describing a matrix X. According to relations (13; 14a; 15), it is like X = ( A B ) , where A =   1 0 · · · 0 0 −1 0 1 · · · 0 −1 0 . . . 0 −1 · · · 0 1 0 −1 0 · · · 0 0 1 X3 X4 . . . Xq   124 On the dimension of Kirichenko space where the first block has l1 − 1 columns, and each next one has li, i > 1, columns, and Xi is like Xi =   1 0 · · · 0 −1 0 · · · 0 0 1 · · · 0 0 −1 · · · 0 ... ... . . . . . . ... 0 0 1 0 0 · · · −1 −1 0 · · · 0 1 0 · · · 0 0 −1 · · · 0 0 1 · · · 0 ... ... . . . . . . ... 0 0 · · · −1 0 0 · · · 1   , where the sub diagonal, whose elements are equal to −1 has number n1, i.e. the element of first line of Xi, which is equal to −1, is in the column number n1 + 1. Matrix B looks like B =   L2 −L1 0 · · · 0 0 L3 0 −L1 · · · 0 0 ... . . . Lq−1 0 0 · · · −L1 0 Lq 0 0 · · · 0 −L1   where Li means the same as in matrix X ′′. Consider block matrix ( X X ′′ ) . Note that widthes of blocks of matrices A, B, and X ′′ are equal, and its easy to see, that if we will subtract the first column of each block of the width l1 − 1, l2, . . . lq, from all other it’s columns, then each block of matrices B and X ′′ will have number li at the first place and zeros at the others. Denote them by L̃i, and matrices which appear after transformation of B and X ′′ by B̃ and X̃ ′′. Let’s find the rang of matrix ( B̃ X̃ ′′ ) . It’s obvious that it is not less then q − 1, because its first q − 1 lines are linear independent. Let’s consider the first q − 1 lines and arbitrary other line, which contains elements li and lj , obtaining det   l2 −l1 0 · · · 0 0 l3 0 −l1 · · · 0 0 ... . . . lq−1 0 0 · · · −l1 0 lq 0 0 · · · 0 −l1 0 li · · · −lj · · · 0   = M. Plakhotnyk 125 = (−l1) q−3 · det   lj −l1 0 li 0 −l1 0 li −lj   = (−l1) q−3 · (l1lilj − l1lilj) = 0. And, the rang of matrix ( B̃ X̃ ′′ ) is equal to q − 1. That is why the rang of matrix ( X X ′′ ) is equal to one of matrix A minus q − 1, And we may generalize our calculations of dimension of Kirichenko space. Thus, if q > 2, then the dimension of Kirichenko space may be written out as [ l1 2 ] + q∑ s=2 (ls, l1)+ q∑ r=2 [ lr − 2 2 ] + q∑ r=2 q∑ s=r+1 (ls, lr)−q+1 = q∑ r=1 [ lr 2 ] + q∑ r=1 q∑ s=r+1 (ls, lr) − 2q + 2. If q = 2, then matrix X ′′ will be absent, whence the dimension of Kirichenko space will be [ l1 2 ] + [ l2 − 2 2 ] + (l2, l1) − 1 = [ l1 2 ] + [ l2 2 ] + (l2, l1) − 2 If q = 1, then all count relations will be represented by equality (13), and the dimension of Kirichenko space will be equal to [n1 2 ] . Thus, we have proved the main theorem Main Theorem. Let σ be a permutation, which is composition of cycles of lengthes l1, . . . , lq. Then the dimension of Kirichenko space is equal to 2−2q+ q∑ r=1 [ lr 2 ] + 1 2 ∑ r 6=s (ls, lr), where (a, b) denotes the greatest common divisor of numbers a and b. References [1] M. A. Dokuchaev, V. V. Kirichenko, A. V. Zelensky, V. N. Zhuravlev Gorenstein Matrices, Algebra and discrete mathematics N1, 2005 pp. 8-29. [2] M. Hazewinkel, N. Gubareni and V. V. Kirichenko, Algebras, Rings and Modules, V.I, Mathematics and Its Applications, V.575, Kluwer Academic Publishers, 2004. [3] H. Fujita, Full matrix algebras with structure systems, Colloq. Math. 2003. 98(2) pp. 249-258. [4] H. Fujita, and Y. Sakai, Frobenius full matrix algebras and Gorenstein tiled orders, Communications in algebra, 2006, 34: pp. 1181-1203. [5] V. V. Kirichenko, Quasi-Frobenius rings and Gorenstein orders Trudy Mat. Steklov Inst. 148, (1978) 168-174 (in Russian), English translation in Proceedings of the Steklov Institute of Mathematics (1980), issue 4, pp. 171-177. 126 On the dimension of Kirichenko space [6] V. V. Kirichenko, A. V. Zelensky, V. N. Zhuravlev, Exponent matrices and tiled orders over discrete valuation rings, Intern. Journ. of Algebra and Computation, Vol 15, N. 5 and 6, 2005, pp. 997-1012. [7] M. Plakhotnyk, On the dimension of the space of Gorenstein matrices for some types of correspond permutations, 5th Iinternational Algebraic Conference in Ukraine, 2005, p. 157. [8] K. W. Roggenkamp, V. V. Kirichenko, M. A. Khibina, V. N. Zhuravlev, Gorenstein tiled orders, Comm. Algebra, 29(9), 2001, pp. 4231-4247. Contact information M. Plakhotnyk Department of Mechanics and Mathemat- ics, Kyiv National Taras Shevchenko Univ., Volodymyrska str., 64, 01033 Kyiv, Ukraine E-Mail: Makar_Plakhotnyk@ukr.net Received by the editors: 31.10.2005 and in final form 06.10.2006.