The Covering Excess Method in the Theory of Covering Codes

Size: px
Start display at page:

Download "The Covering Excess Method in the Theory of Covering Codes"

Transcription

1 The Covering Excess Method in the Theory of Covering Codes Wolfgang Haas Dissertation zur Erlangung des Doktorgrades der Fakultät für Mathematik und Physik der Albert-Ludwigs-Universität Freiburg im Breisgau Oktober 008

2 Dekan: Prof. Dr. Ludger Rüschendorf (Freiburg) 1. Referent: Prof. Dr. Jan-Christoph Schlage-Puchta (Gent). Referent: Prof. Dr. Laurent Habsieger (Lyon) Betzreuung: Prof. Dr. Dieter Wolke (Freiburg) Datum der Promotion:

3 Für Lara

4

5 Acknowledgements First of all I wish like to thank Dieter Wolke (Freiburg) for his friendly support over many, many years. Without his permanent encouragement this thesis would not have been possible. I thank Laurent Habsieger (Lyon), whose fine papers on the subject are an important source of inspiration to me. Moreover I wish like to thank Jörn Quistorff (Berlin), who brought me nearer to the world of mathematics. Finally I thank Michael Botzet (Hamburg), who brought me to coding theory.

6

7 Contents 1 Introduction A sphere-covering problem Overview of the thesis Preliminaries 5.1 Codes The covering excess A bound for K(n, 1) with even n Bounds for K(n, 1) with odd n The case n 5 (mod 6) The bound of van Wee The Theorem of Blass and Litsyn The new bound The case n 1, 3 (mod 6) Steiner systems Perfect coverings of spheres The general excess bound Habsiegers improvement and the excess matrix Congruence properties for the covering excess function Some lemmas Proof of Theorem Proof of Theorem The main property of the excess matrix Congruence set systems and their resolvability Simple properties of congruence set systems The classification of congruence set systems on small sets

8 3.6.3 Open questions Proof of Theorem The general excess bound improved Bounds for K(n, R) The bounds of van Wee and Honkala An improvement of Honkala s bound The case ɛ = R A Further results 71 A.1 Notes to Chapter and A. Notes to Chapter B The new bounds 73 Bibliography 75

9 Chapter 1 Introduction Let K(n, R) denote the minimal cardinality of a binary code of length n and covering radius R. This thesis is concerned with lower bounds on K(n, R) by the method of covering excess, which was first proposed by Johnson [18] in 197 and later rediscovered by van Wee [9] in A sphere-covering problem Let F = {0, 1} denote the finite field with two elements. To measure the similarity between two words of length n over F, we introduce their Hamming distance, which simply counts the number of places, where the words differ. The space F n of the binary words of length n over F, equipped with the Hamming distance, is one of the most important finite metric spaces. Like in every (finite) metric space there are two fundamental questions, the sphere-packing and the sphere-covering problem. (i) Determine the maximal number A(n, e+1) of non-overlapping spheres of given radius e, that can be placed in the Hamming space F n. (ii) Determine the minimal number K(n, R) of spheres of given radius R needed to cover the whole space F n. The quantity A(n, d) was widely studied during the last six decades in the framework of error-correcting codes. The sphere-covering problem (ii) however was not studied equally extensive, but in the last two or three decades, it attracted much attention to the mathematicians. In generally the determination of K(n, R) is very difficult and we have to be content with lower 3

10 and upper bounds. A trivial lower bound for K(n, R) certainly is the cardinality of the whole space F n divided by the cardinality of a Hamming sphere with radius R. Many methods are known to improve on this lower bound, the so-called sphere-covering bound. One of the most efficient method is the covering excess counting method, which is the object of the present thesis. It relies on the fact, that for arithmetical reasons under certain conditions a sphere-covering is forced to cover words in a small Hamming sphere more than once, the so-called covering excess. 1. Overview of the thesis In chapter we give the necessary background on codes as well as the notations and terminology on covering excess, which are used throughout this work. As a first instance of the covering excess counting method we give Johnson s and van Wee s bound on K(n, 1) for even n, which makes use of a forced covering excess in a sphere of radius one. In chapter 3 we study bounds for K(n, 1) when n is odd. We first use a method of Blass and Litsyn [1] to give the presently best known lower bound for K(n, 1) if n 5 (mod 6). Then we discuss the general excess bound, which follows from the fact, that if n 1 (mod p) for an odd prime p, there exists a covering excess on a sphere with radius p 1. We give a new lower bound for the covering excess, leading to a general improvement of the general excess bound for binary codes with covering radius one. In chapter 4 we study bounds for K(n, R), which rely on a covering excess in spheres with radius one. We describe a classical bound of van Wee for this case from 1988 and a slight improvement of this bound by Honkala. We present a short proof of a further improvement of Honkala s bound. Moreover we give an improvement of the van Wee bound in case that R + 1 divides n, when both bounds coincide and no general further progress of the van Wee bound is known up to now. 4

11 Chapter Preliminaries.1 Codes Let F = {0, 1} denote the finite field with two elements. The Hamming distance d(x, y) between x = (x 1,..., x n ) F n and y = (y 1,..., y n ) F n is defined by d(x, y) = {i {1,..., n} x i y i }. One easily checks, that F n, equipped with the Hamming distance, is a finite metric space, the so-called Hamming space. Any nonempty subset C F n is called a code. For x F n the distance d(x, C) is defined by d(x, C) = min d(x, c). c C We say that x R-covers y, if d(x, y) R holds. C R-covers x means d(x, C) R. In both cases we simply say cover instead of R-cover, if R is clear from the context. If C, then d(c) denotes the minimum distance of C: d(c) = min d(c 1, c ). c 1,c C,c 1 c Let B(x, e) denote the Hamming sphere with radius e centered on x F n, B(x, e) = {y F n : d(x, y) e}. One easily sees, that the cardinality of a sphere B(x, e) does not depend on 5

12 x F n. So we may set V (n, e) = B(x, e) = 0 i e ( ) n i (.1) for any x F n. Like in every (finite) metric space there exists the fundamental spherepacking problem: (i) Determine the maximal number of non-overlapping spheres of given radius e, that can be placed in the Hamming space F n. One easily sees, that the centers of the spheres in such a packing form a code C F n with minimum distance at least e + 1. Thus the number requested in (i) is A(n, e+1), where A(n, d) denotes the maximal cardinality of a code C F n with minimum distance at least d. Assume that a set of messages is encoded by the words of a code C F n with minimum distance at least e + 1. If during the transmission of such a word there occur at most e errors, then these errors can be corrected by the receiver. Thus A(n, d) is the most basic quantity in coding theory, since A(n, e + 1) is the maximal cardinality of a code correcting up to e errors. For a monograph on error-correcting codes see [6]. Updated internet tables for A(n, d) are given by Brouwer []. In this thesis however we are concerned with the dual sphere-covering problem: (ii) Determine the minimal number of spheres of given radius R needed to cover the whole space F n. The covering radius R = R(C) of a code C F n denotes the least integer R, such that the spheres with radius R centered on the codewords cover the whole space. It is clear, that the centers of the spheres of a covering described in (ii) form a code with covering radius at most R. Equivalently, a code C F n has covering radius (at most) R, if x F n c C with d(x, c) R (.) holds or, simply, if each x F n is R-covered by C. K(n, R) denotes the minimal cardinality of a code C F n with covering radius (at most) R. 6

13 In the last two or three decades the theory of covering codes attracted a great amount of attention. A monograph [4] appeared in Updated internet tables for K(n, R) are given by Kéri [19]. An internet bibliography on covering codes is compiled by Lobstein [5]. Let 0 = (0,..., 0) F n denote the all-zero word of length n. Replacing the ith coordinate of the all-zero word by one yields the unit vector e i F n, 1 i n. By the weight w(x) of a word x F n we denote the number of its nonzero coordinates. The support of x F n = (x 1,..., x n ), supp(x), is defined by supp(x) = {1 i n x i 0}. (.3) Binary vectors may be identified with their supports and if x, y F n, then x y, x y and x y refer to the supports of x and y. We further set F n i = {x F n the ith coordinate of x equals 1}. Let C F n be a code with covering radius R and cardinality K(n, R). Then by F n c C B(c, R) we get Fn c C B(c, R) = C V (n, R) and thus K(n, R) n V (n, R). (.4) This is the most basic lower bound on K(n, R), called the sphere-covering bound. In generally this bound cannot be improved, since there are codes attaining equality in (.4). These codes are called perfect. One especially famous family of perfect codes are the so-called Hamming codes. Theorem 1 (Hamming [11]). If n = m 1, then there exists a perfect code C F n with covering radius one. Especially K(n, 1) = n m. In the majority of cases however the sphere-covering bound can be improved. Many different methods are known to do this, see [4]. One of the most efficient method is the covering excess counting method, which is the object of the present thesis. It relies on the fact, that sometimes a covering code is forced to cover words in a small Hamming sphere more than once, the 7

14 so-called covering excess. As a simple example it could be possible, that a codeword covers always an even number of points in a Hamming sphere, but the cardinality of the sphere is odd. Then a covering excess in that sphere is unavoidable. The first who introduced the method of covering excess was Johnson [18] in 197, but his remarkable paper remained unnoticed until the appearance of a monograph on covering codes [4] in The method was rediscovered by van Wee [9] in 1988, so it should be quite fair, to attribute the method both to Johnson and van Wee. After some prerequisites in section. we recall their most prominent example in section.3, a lower bound for K(n, 1) for even n.. The covering excess In this section we give the definition of covering excess, some notations and a few basic equations, which will be used throughout this work. Let C F n be a code with covering radius R. For x F n we define the covering excess E(x) by E(x) = E C (x) = B(x, R) C 1. (.5) By (.) E(x) 0 for every x F n, and E(x) is strictly positive, if x is covered by at least two codewords from C. For V F n, i 0 and x F n we set E(V ) = y V E(y), E i (x) = E(y), y F n,d(x,y)=i Z = {z F n E(z) > 0}, Z i = {z F n E(z) = i}. 8

15 For V F n we have E(V ) = ( B(x, R) C 1) (.6) x V = 1 V x V c B(x,R) C = 1 V c C x B(c,R) V = B(c, R) V V. c C In the special case V = F n we get E(F n ) = C V (n, R) n, (.7) which in case of R = 1 by (.1) further simplifies to E(F n ) = C (n + 1) n. (.8).3 A bound for K(n, 1) with even n The case of binary codes with covering radius one is of special interest. Here the sphere-covering bound (.4) reads K(n, 1) n n + 1. (.9) We now present the argument of Johnson and van Wee to improve on this bound for even n, which is probably the most famous application of the covering excess method. Theorem (Johnson [18], van Wee [9]). Let C F n be a binary code with covering radius one. If x F n, then { 1 (mod ) if n is even and x C, E(B(x, 1)) 0 (mod ) if n is odd. 9

16 Proof. Let δ(x) = 1 if x C, otherwise δ(x) = 0. If c C, then B(c, 1) B(x, 1) equals the number of words in the sphere B(x, 1), which were 1- covered by c. One easily sees, that B(x, 1) = n + 1 if d(c, x) = 0, B(c, 1) B(x, 1) = if d(c, x) = 1 or, 0 if d(c, x) 3. Now (.6) applied to V = B(x, 1) yields E(B(x, 1)) = c C and Theorem follows. B(c, 1) B(x, 1) B(x, 1) = δ(x)(n + 1) + c C, 1 d(c,x) (δ(x) 1)(n + 1) (mod ) 1 (n + 1) Now let n be even and C F n a binary code with covering radius one. Since E(B(x, 1)) is always nonnegative, from Theorem follows E(B(x, 1)) 1 if x C. (.10) Thus there exists a covering excess on every sphere of radius one not centered on a codeword. This was used by Johnson and van Wee to derive the following improvement of the sphere-covering bound (.9), the first excess bound for binary codes with covering radius one (see [4]). Theorem 3 (Johnson [18], van Wee [9]). When n is even we have K(n, 1) n n. Proof. Let C F n be a binary code with covering radius one. Note, that z Z means E(z) 1 and thus B(z, 1) C, implying B(z, 1) \ C B(z, 1) = n 1 if z Z. (.11) 10

17 We now get n C = 1 x F n \C x F n \C = E(B(x, 1)) by (.10) E(z) x F n \C z B(x,1) = z Z E(z) x B(z,1)\C (n 1)E(F n ) by (.11) = (n 1)( C (n + 1) n ) by (.8) 1 implying C n /n and Theorem 3 follows. In general, Theorem 3 cannot be improved, as we see in the next theorem. Theorem 4 (Johnson [18], van Wee [9]). For m 1 we have K( m, 1) = m m. Proof. By Theorem 1 there exists a perfect code C F m 1 of length m 1, covering radius one and cardinality C = m m 1. One easily sees, that the code C F m of length m defined by C = {(0, x) x C} {(1, x) x C} has covering radius one and satisfies C = m m. Therefore K( m, 1) m m. The corresponding lower bound follows from Theorem 3. If n is odd, there is no forced covering excess in a sphere of radius one. For instance the Hamming sphere B(0, 1) centered on the all zero word 0 of length n could be perfectly covered (i.e. without covering excess) by the words e 1, e + e 3, e 4 + e 5,..., e n 1 + e n. What can we do to improve on the sphere-covering bound? We have to look for a covering excess in a sphere with radius greater than one! 11

18 1

19 Chapter 3 Bounds for K(n, 1) with odd n In this chapter we study bounds for K(n, 1), where n is odd. We start with the case n 5 (mod 6), where a covering excess in a sphere of radius two can be found. We describe a classical bound of van Wee [9] for this case as well as an improvement due to Blass and Litsyn [1]. We further improve on this bound, giving the presently best known lower bound for K(n, 1) if n 5 (mod 6). Next we discuss the general excess bound for K(n, 1) (due to Habsieger [9] and Honkala [16]), which follows from the fact, that if n 1 (mod p) for an odd prime p, there exists a covering excess p 1 on a sphere with radius p 1. Answering a question of Habsieger, we prove that this covering excess indeed grows quadratic in p. This yields a general improvement of the general excess bound for binary codes with covering radius one. The proof uses a classification theorem for certain subset systems as well as new congruence properties for the excess function, which were conjectured by Habsieger. 3.1 The case n 5 (mod 6) The bound of van Wee Let C F n be a binary code with covering radius one. As we have already seen, there is no need for a covering excess in a sphere of radius one in case of odd n, so lets hope n 1 (mod 3), when the following theorem of van Wee applies. 13

20 Theorem 5 (van Wee [9]). Let C F n be a binary code with covering radius one and n 1 (mod 3). If x F n, then E(B(x, )) 1 (mod 3). (3.1) Proof. If c C, then B(c, 1) B(x, ) equals the number of words in the sphere B(x, ), which were 1-covered by c. One easily sees, that B(c, 1) = n + 1 if d(c, x) 1, B(c, 1) B(x, ) = 3 if d(c, x) = or 3, 0 if d(c, x) 4. Now (.6) applied to V = B(x, ) yields E(B(x, )) = c C = (n + 1) B(c, 1) B(x, ) B(x, ) c C, d(c,x) 1 ( ) n (mod 3) 1 (mod 3) by (.1) and n 1 (mod 3) c C, d(c,x) 3 1 ( 1 + n + ( )) n If n 1 (mod 3) with even n, then the better bound in Theorem 3 applies, so we may restrict us to the case n 5 (mod 6). Theorem 6 (van Wee [9]). When n 5 (mod 6), we have ( ) 4 n K(n, 1) 1 + n + n + n + 1. Proof. Let C F n be a binary code with covering radius one and C = K(n, 1). Since E(B(x, )) is always nonnegative for x F n, from (3.1) follows E(B(x, )) for each x F n. (3.) 14

21 We now get n = x F n x F n E(B(x, )) by (3.) = x F n z B(x,) = z F n E(z) E(z) x B(z,) = V (n, )E(F n ) = V (n, )( C (n + 1) n ) by (.8) and Theorem 6 follows by V (n, ) = 1 + n + ( n ) (see (.1)) The Theorem of Blass and Litsyn Theorem 6 was further improved by Blass and Litsyn [1], Habsieger [9], Honkala [15], [16], [17] and van Wee [31]. The best of these bounds is due to Blass and Litsyn [1]: Theorem 7 (Blass and Litsyn [1]). If n 5 (mod 6), then ( ) 40 n K(n, 1) n 44n + 77 n + 1. (3.3) In section we give a simplification of the proof of Theorem 7, which at the same time leads to a better result. We give the complete proof (for n 17), since it is divided on the three papers [1], [7], [16] and it saves the efforts of the reader to find out, which of the material of [1] is needed and which not. For the rest of this section we give the necessary lemmas, which already can be found in [1] and [16], in a slightly modified manner. All improvements on Theorem 6 mentioned above rely on the fact, that not always equality can hold in (3.) and so one has to upper-bound B, where B is defined by 1 B = {x F n E(B(x, )) = }. (3.4) Crucial is a structural property of B described in the next lemma. Assume for the rest of this section, that C F n is a fixed binary code with covering radius one and n 5 (mod 6), n

22 Lemma 8 (van Wee [31]). For every x B there exists a unique z B(x, ) with E(z) > 0. Consequently E(z) = (i.e. z Z ). Proof. Assume to the contrary, there exist distinct z 1, z B(x, ) with E(z 1 ) > 0 and E(z ) > 0. One easily sees, that in all cases there exists y B(x, 1) such that, say z 1 B(y, 1), but z B(y, 1). Now E(B(y, 1)) E(z 1 ) 1. On the other hand by Theorem we have E(B(y, 1)) 0 (mod ), implying E(B(y, 1)). This however gives the contradiction = E(B(x, )) E(B(y, 1)) + E(z ) 3. For x F n we set L = {y F n : E(B(y, )) = 5}, B(x) = B(x, ) B. One easily checks, that if z Z, then E(B(z, )) > and thus E(B(z, )) 5 by Theorem 5. We partition Z into the following subsets (different from [1]): Z (1) = L Z, { ( )} Z () = z Z \ Z (1) n 3 B(z) >, { ( )} Z (3) = z Z \ Z (1) n 3 B(z). (3.5) The next lemmas are from [1]. Lemma 9 (Blass, Litsyn [1]). If z Z, then ( ) n B(z). (3.6) If additionally z Z (1), then there are exactly three distinct words y i B(z, ), i = 1,, 3, such that E(y i ) = 1. For every c C with d(y i, c) 1 for at least one y i we also have d(z, c) 1. Proof. By z Z, z is covered exactly three times. Case I. z C. Without loss z = 0 and C = {0, e 1, e } C. Clearly E(e 1 ) 1, E(e ) 1, E(e 1 + e ) 1. (3.7) 16

23 To show (3.6), simply note that for x B(0) we have d(x, e 1 ) >, since otherwise E(B(x, )) E(0) + E(e 1 ) 3 contradicting E(B(x, )) = for x B(0) B. Similar d(x, e ) >. Thus every word in B(0) has weight two and starts with two zeros, i.e. B(0) {e i + e j 3 i < j n}, (3.8) and (3.6) follows in case I. If z = 0 Z (1), then E(B(0, )) = 5 and by (3.7) in fact we have E(e 1 ) = E(e ) = E(e 1 + e ) = 1. (3.9) Thus we may set y 1 = e 1, y = e and y 3 = e 1 + e. Assume c C with d(y i, c) 1 for at least one y i. By (3.9), y i is covered exactly twice. Since each y i, i = 1,, 3, is covered by two words from C, we find c C and thus d(z, c) = d(0, c) 1. Case II. z C. Without loss z = 0 and {e 1, e, e 3 } C. Clearly E(e 1 + e ) 1, E(e 1 +e 3 ) 1 and like in case I one shows d(x, e 1 +e ) > and d(x, e 1 +e 3 ) > whenever x B(0). Thus every word in B(0) has weight one or two and starts with three zeros, i.e. B(0) {e i 4 i n} {e i + e j 4 i < j n}, (3.10) implying B(0) n 3 + ( ) ( n 3 = n ) and (3.6) follows in case II too. Setting y 1 = e 1 + e, y = e + e 3 and y 3 = e 1 + e 3, the proof of the second half of the Lemma is very similar like in case I. Lemma 10 (Blass, Litsyn [1]). We have Z 1 3 Z (1). (3.11) Proof. By Lemma 9, Z 1 B(z, ) = 3 whenever z Z (1). Thus Lemma 10 follows, if we can show, that no y Z 1 satisfies y B(z 1, ) B(z, ) whenever z 1, z Z (1) with z 1 z. Assume to the contrary, that y Z 1 B(z 1, ) B(z, ). Since y is covered by C, there exists c C with d(c, y) 1. By Lemma 9 we get d(c, z 1 ) 1 as well as d(c, z ) 1, implying d(z 1, z ). But then B(z 1, ) contains two words from Z (z 1 and z ) as well as three words from Z 1 by Lemma 9 and thus E(B(z 1, )) 7 contradicting z 1 Z (1). 17

24 Lemma 11 (Blass, Litsyn [1]). If z Z (), there are exactly three words y i B(z, ) \ {z} (i = 1,, 3) satisfying E(y i ) =. Proof. Case I. z C. Without loss z = 0 and {0, e 1, e } C. We set y 1 = e 1, y = e and y 3 = e 1 + e. By (3.7) we have y i B(z, ) \ {z} and E(y i ) > 0 (i = 1,, 3). We next prove, that there is no other word y B(0, ) \ {0} satisfying E(y) > 0. Without loss it suffices to consider the cases y = e 3 or y = e i + e 3 with i 3. By E(e 3 ) > 0 (resp. E(e i + e 3 ) > 0) and E(0) = we have E(B(e 3 + e j, )) > for j 3, implying e 3 + e j B(0). Therefore by (3.8) each word of B(0) must be of the form e i + e j, 4 i < j n. The number ) (), contradicting the definition of Z of such vectors is at most ( n 3. It remains to show E(y 1 ) = E(y ) = E(y 3 ) =. The case E(y 1 ) = E(y ) = E(y 3 ) = 1 is impossible, since it yields E(B(z, )) = 5 and thus z Z (1), contradicting the definition of Z (). If E(y 1 ) or E(y 3 ), then e 1 + e (with e = 0 or e = e ) must be covered by a codeword c C, C 0, e 1, e. Now c = e 1 + e + e j with j 3 is impossible, since then E(y) > 0 with y = e 1 + e j and y B(z, ) \ {z, y 1, y, y 3 }. There remains the possibility c = e 1 + e C and E(y 1 ) = E(y 3 ) = follows. For reasons of symmetry we also have E(y ) =. Case II. z C. Without loss z = 0 and {e 1, e, e 3 } C. We set y 1 = e 1 +e, y = e +e 3 and y 3 = e 1 + e 3. Clearly E(y i ) > 0 (i = 1,, 3). We show, that there is no other word y B(0, ) \ {0} with E(y) > 0. It suffices to consider the cases y = e i0 or y = e i0 + e 4, 1 i 0 n. Similar like in case I we see, that we then have B(0) B 1 := {e i + e j 4 i < j n} in the first case and B(0) B := {e i 5 i n} {e i + e j 5 i < j n} in the second case (use (3.10)). Since B 1 = B = ( ) n 3 we get a contradiction to the definition of Z (). It remains to show E(y 1 ) = E(y ) = E(y 3 ) =. Like in case I E(y 1 ) = E(y ) = E(y 3 ) = 1 is impossible. For reasons of symmetry it suffices to assume E(y 1 ). Then y 1 = e 1 +e must be covered by a codeword c C, c = e 1 + e + e with e = 0 or e = e i, i 3. Now e = 0 (resp. e = e i, i 4) is impossible, since then E(e 1 ) > 0 (resp. E(e 1 + e i ) > 0). Hence y 1 C, c = e 1 + e + e 3 C and E(y 1 ) = follows. Since c also covers y and y 3, we get E(y ) and E(y 3 ). Now like in the case of y 1 we have 18

25 E(y ) = E(y 3 ) = also. For z Z () we set H(z) = {z, y 1, y, y 3 } Z () Z (3), (3.1) where y 1, y and y 3 are defined in Lemma 11. Clearly H(z) = 4 if z Z (). (3.13) Following the proof of Lemma 11, we notice, that if z C, then for some i, j {1,..., n}, i j If z C, then H(z) = {z, z + e i, z + e j, z + e i + e j } C. (3.14) H(z) = {z, z + e k + e l, z + e k + e m, z + e l + e m } with H(z) C = (3.15) for some k, l, m {1,..., n}, k l m k. {z + e k, z + e l, z + e m, z + e k + e l + e m } C. Moreover in that case Lemma 1 (Blass, Litsyn [1]). If z 1, z Z (), then either H(z 1 ) H(z ) = or H(z 1 ) = H(z ). Proof. Assume y H(z 1 ) H(z ). By definition of H(z) we have E(y) =. Case I. y C. Then z 1, z C (otherwise H(z 1 ) C = ). From (3.14) follows H(z 1 ) = {z 1, z 1 + e i, z 1 + e j, z 1 + e i + e j } = {y, y + e i, y + e j, y + e i + e j } C for some i, j {1,..., n}, i j. Similar by y H(z ) we have H(z ) = {y, y + e k, y + e l, y + e k + e l } C for some k, l {1,..., n}, k l. We now get {i, j} = {k, l} (otherwise E(y) > ), i.e. H(z 1 ) = H(z ). Case II. y C. Then z 1, z C (otherwise H(z 1 ), H(z ) C). From (3.15) follows H(z 1 ) = {z 1, z 1 + e k + e l, z 1 + e k + e m, z 1 + e l + e m } = {y, y + e k + e l, y + e k + e m, y + e l + e m } for some k, l, m {1,..., n}, k l m k and {z 1 + e k, z 1 + e l, z 1 + e m, z 1 + e k + e l + e m } = {y + e k, y + e l, y + e m, y + e k + e l + e m } C. Similar by y H(z ) we have H(z ) = {y, y + e k + e l, y + e k + e m, y + e l + e m } for some k, l, m {1,..., n}, k l m k and {y + e k, y + e l, y + e m, y + e k + e l + e m } C. Like in case I we now get {k, l, m} = {k, l.m }, i.e. H(z 1 ) = H(z ). 19

26 The proof of the case z C of the following Lemma already appeared in Honkala [16]. Lemma 13 (Blass, Litsyn [1], Honkala [16]). If z Z () we have ( ) n 3 B(y) 4. (3.16) y H(z) Proof. By the definition of H(z) we have y Z whenever y H = H(z). Indeed we will show ( ) n 3 B(y) 3 (3.17) y H\{x} whenever x H, since by (3.13) this implies 3 B(y) = ( ) n 3 B(y) 1 y H x H and (3.16) follows. y H\{x} Case I. z C. Without loss z = 0 and H = {0, e 1, e, e 1 + e }, see (3.14). Moreover, it suffices to show (3.17) for x = e 1 + e, i.e. H := H \ {x} = {0, e 1, e }. For y H we have B(y) B (y) := {y + e i + e j 3 i < j n}, (3.18) (compare (3.8)). Note, that the sets B (y), y H are pairwise disjoint. We now consider the covering excess in a sphere of radius two centered on e i, 3 i n. Clearly B(e i, ) H = H and thus E(B(e i, ) H) = 6 by E(y) = whenever y H. By Theorem 5 however we have E(B(e i, )) 1 (mod 3). This implies 3 i n u i B(e i, ) \ H with E(u i ) > 0. (3.19) We set B = y H B (y) and consider two subcases. Subcase A. Each word u i has weight at least two, 3 i n. We first show B(u i, ) B F n i 6 (3 i n). (3.0) 0

27 To see this, fix i with 3 i n. Note that in subcase A by u i B(e i, ) we have u i F n i. If u i F n 1, then B(u i, ) B (e1) F n i = B (e1) F n i implying B(u i, ) B (e1) F n i = B (e1) F n i = n 3 6 and (3.0) follows. If u i F n or w(u i ) =, then the argument is very similar. There remains the case, that u i starts with two zeros and satisfies w(u i ) = 3. Then for each of the three elements y H we have B(u i, ) B (y) F n i = and (3.0) follows again. Each word in B occurs in exactly two of the sets B F n i, 3 i n. Therefore (B(u i, ) B ) (3.1) 3 i n (B(u i, ) B F n i ) 1 B(u i, ) B F n i 3(n 3). 3 i n 3 i n by (3.0). Now, if x B(u i, ) B, (3 i n), then there exists y H with x B (y ) and thus y B(x, ). Therefore E(B(x, )) E(y ) + E(u i ) >, which implies x y H B(y) whenever x B(u i, ) B, 3 i n. Using (3.18) and (3.1) we now get B(y) B (B(u i, ) B ) y H 3 i n ( ) ( ) n n 3 3 3(n 3) = 3 and (3.17) follows, completing the proof of Lemma 13 in subcase A. Subcase B. There is a word u i with w(u i ) = 1. If e j0 with a suitable j 0, 3 j 0 n occurs among the words u i, then (3.1) follows again, since then each of the 3(n 3) words of the form y + e i + e j0, y H, 3 i n, i j 0 lies in B(e j0, ) B. Since we have already seen, that (3.1) implies (3.17), the proof of Lemma 13 is complete in case I. Case II. z C. In this case by (3.15) we may assume z = 0 and H = {0, e 1 + e, e 1 + e 3, e + e 3 }. It suffices to show (3.17) for x = e + e 3, i.e. H = H \ {x} = {0, e 1 + e, e 1 + e 3 }. For y H we have B(y) B (y) := {y + e i 4 i n} {y + e i + e j 4 i < j n}, 1

28 see (3.10). The excess argument from case I leading to (3.19) applies equally well to the spheres B(e 1 + e i, ), 4 i n and yields 4 i n v i B(e 1 + e i, ) \ H with E(v i ) > 0. Again we set B = y H B (y) and consider two subcases. Subcase A. Neither e 1 nor one of the words e 1 + e j occurs among the words v i, 4 i, j n. We show B(v i, ) B F n i 6 (4 i n). (3.) To see this, fix i with 4 i n. Again in subcase A we have v i F n i. If v i does not start with (100), then B(v i, ) B (y) F n i = n 3 6 for a suitable y H and (3.) follows. If v i starts with (100) then B(v i, ) B (y) F n i = for every y H and (3.) follows again. Now (3.1) (with u i replaced by v i ) follows like in case I. The rest of the proof of Lemma 13 in subcase A is the same. Subcase B. Either e 1 or one of the words e 1 + e j, 4 j n occurs among the words v i, 4 i n. If e 1 occurs among the words v i, then each of the 3(n 3) words of the form y+e i, y H, 4 i n lies in B(e 1, ) B. If e 1 +e j0 with a suitable j 0, 4 j 0 n occurs among the words v i, then each of the 3(n 3) words of the form either y + e j0 or y + e i + e j0, y H, 4 i n, i j 0 lies in B(e 1 + e j0, ) B. In both cases the proof proceeds exactly like in case I The new bound In this section we further improve on (3.3). The bound given in the next theorem appears to be the presently best lower bound on K(n, 1) when n 5 (mod 6). Theorem 14 (Haas [7]). If n 11 and n 5 (mod 6), then ( ) 8 n K(n, 1) 1 + n 10n + 19 n + 1. As a corollary we get the following new bounds on K(n, 1). records in parenthesis follow from (3.3). The old

29 Corollary 15. K(17, 1) 7419 (7414 [1]), K(3, 1) 3587 (35736 [1]), K(9, 1) ( [1]). Proof of Theorem 14. We use the notations of the previous section. Let n be an integer with n 11 and n 5 (mod 6). By (3.3) Theorem 14 does hold for n = 11, so we may assume n 17. Let C F n be a binary code of length n with covering radius one. By M Z () we denote a set of maximal cardinality satisfying H(z 1 ) H(z ) = whenever z 1, z M with z 1 z. Moreover we set H = H(z). Then by Lemma 1, (3.1), (3.13) we have H = 4 M and z M Lemma 16. If z L, then B(z, ) Z 1. Z () Z (3) = H (Z (3) \ H). (3.3) Proof. By E(B(z, )) = 5 there exists u B(z, ) with E(u) odd. Fix any x B(u, 1) B(z, 1). By Theorem we have E(B(x, 1)) 0 (mod ). Thus there exists v B(x, 1) B(z, ) with v u and E(v) odd. Now, since E(B(z, )) is odd, there must exist w B(z, ) with E(w) odd, w u and w v. Since E(u), E(v) and E(w) are odd, nonnegative integers adding to at most 5, at least two of them must equal one and Lemma 16 follows. We now may complete the proof of Theorem 14. From Lemma 8 follows B z Z B(z) (remember, that B is defined in (3.4)). Thus by (3.3) we get B z Z B(z) = z Z (1) B(z) z M y H(z) B(y) z Z (3) \H B(z), 3

30 which by (3.5), (3.3), Lemma 9 and Lemma 13 implies B B(z) + B(y) + B(z) (3.4) z Z (1) z M y H(z) z Z (3) ( ) ( ) ( \H ) n Z (1) n 3 n M + (Z () Z (3) ) \ H ( ) ( ) n = Z (1) n 3 + Z () Z (3) ( ) n 3 = Z + (n 3) Z (1). Moreover E(F n ) = i 0 i Z i Z + Z 1. Thus By Theorem 5 we have ( ( )) n 1 + n + E(F n ) = V (n, )E(F n ) Z E(F n ) Z 1. (3.5) = x F n E(B(x, )) B + 5 L + 8( n B L ) = 8 n 6 B 3 L. Substituting (3.11), (3.4) and (3.5) on the right-hand side yields ( ( ) ( )) n n n E(F n ) (3.6) ( ( ) ) n 3 8 n + 3 (n 3) Z 1 3 L. By Lemma 16 for n 17 we have ( ( ) ) n 3 3 (n 3) Z 1 3 = ( 1 + n + z Z 1 x B(z,) ( )) n Z x L z B(x,) Z 1 3 L.

31 This inserted in (3.6) yields E(F n ) and Theorem 14 follows by (.8). 8 ( ) n ( ) n n n The case n 1, 3 (mod 6) We have seen, that in the case n 5 (mod 6) a binary code of length n and covering radius one forces a covering excess in every sphere of radius two (Theorem 5). In this section we have a look on the remaining odd cases n 1, 3 (mod 6) and show, that then there are coverings of spheres with radius three indeed without covering excess. For the proof we have to review shortly some basic facts on Steiner systems Steiner systems We shortly give the definition and a few basic facts for Steiner systems, which are basic objects in design theory. They are related to the construction of perfect coverings of Hamming spheres, i.e. covering codes without covering excess on certain spheres. Definition 17. A Steiner system S(t, k, n) is a collection of k-subsets (called blocks) of {1,..., n}, such that every t-subset of {1,..., n} is contained in exactly one of the blocks. The connection to coding theory is given by the quantity supp defined in (.3), which yields a natural one to one correspondence between the words x F n and the subsets of {1,..., n}. The history of Steiner systems goes back until 1847, when Kirkman considered the special case t = and k = 3. A Steiner system S(, 3, n) is called a Steiner triple system. Theorem 18 (Kirkman [0] 1847). A Steiner triple system S(, 3, n) exists if and only if n 1, 3 (mod 6). A Steiner system S(3, 4, n) is called a Steiner quadrupel system. Theorem 19 (Hanani [1] 1960). A Steiner quadrupel system S(3, 4, n) exists if and only if n, 4 (mod 6). 5

32 3.. Perfect coverings of spheres Definition 0. A perfect R-covering of a Hamming sphere B(x, r) with x F n is a code C F n, such that each y B(x, r) is R-covered exactly once by C. Clearly a perfect code C F n with covering radius R is a perfect R- covering of any sphere B(x, r), x F n, r 0. Perfect coverings of spheres are related to Steiner systems as we will show in the next theorem. Theorem 1. A perfect 1-covering of a Hamming sphere B F n with radius three exists, if and only if there exists a Steiner quadruple system S(3, 4, n + 1), i.e. if n 1, 3 (mod 6). Proof. It suffices to consider the case B = B(0, 3) with the all-zero word 0 F n. Assume D 1,..., D r {1,..., n+1} are the blocks of a Steiner quadruple system S(3, 4, n + 1) on the set {1,..., n + 1}. Then one easily sees, that the code C = {0, supp 1 (D 1 \ {n + 1}),..., supp 1 (D r \ {n + 1})} F n is a perfect 1-covering of B(0, 3). Conversely assume that C F n is a perfect 1-covering of B(0, 3). Let C i C be the words of weight i in C, i 0. If 0 C, then C 1 = C =. One easily sees, that in this case {supp(u) {n+1} u C 3 } {supp(u) u C 4 } is a Steiner quadruple system S(3, 4, n + 1) on the set {1,..., n + 1}. Consider the case 0 C. Then C 1 = 1, say C 1 = {e n }. Then supp(u) {1,..., n 1} and u v = 4 whenever u, v C, u v. One checks, that {u v u, v C, u v} {supp(u) {n, n + 1} u C } {supp(u) {n + 1} u C 3 } {supp(u) u C 4 } is a Steiner quadruple system S(3, 4, n + 1) on the set {1,..., n + 1}. The last part of the claim now follows from Theorem 19. Similar (somewhat easier) one can show, that a perfect 1-covering of a Hamming sphere B F n with radius two exists if and only if a Steiner triple system S(, 3, n) exists, which by Theorem 18 is the case, if n 1, 3 (mod 6). One might try to extend the construction from Theorem 1 to spheres of radius four, thereby obtaining the statement, that a perfect 1-covering of 6

33 a Hamming sphere B F n with radius four exists if and only if a Steiner quintuple system S(4, 5, n + ) exists. This however in generally is incorrect, since for instance by Theorem 1 there exists a perfect 1-covering of a Hamming sphere B F 15 with radius four, but there is no Steiner system S(4, 5, 17) (Östergård and Pottonen [7]). At the present nothing seems to be known concerning perfect 1-coverings of spheres with radius greater than three. Question. For which values of n there exists a perfect 1-covering of a Hamming sphere B F n with radius four? By Theorem 1 there is no forced covering excess on a sphere with radius three if n 1, 3 (mod 6). In order to improve on the sphere-covering bound (.9) we will have to look on a covering excess in a sphere with radius greater than three! 3.3 The general excess bound Again let C F n be a code with covering radius one. Theorem 5 was independently generalized by Habsieger [9] and Honkala [16], who showed that there is a covering excess in a sphere of radius p 1 whenever n 1 (mod p) for an odd prime p. For the proof we need a simple lemma, which was first stated in Cohen et al. [5]. For x F n and an integer i we define A i (x) by Note that A i (x) = 0, if i is negative. A i (x) = {c C d(c, x) = i}. (3.7) Lemma 3 (Cohen et al. [5]). For x F n and i 0 we have ( ) n E i (x) = (n i + 1)A i 1 (x) + A i (x) + (i + 1)A i+1 (x). (3.8) i Proof. It suffices to consider the case x = 0. Then A j (0) is the number of codewords of weight j, j 0. Let S(0, i) denote the set of words with weight i. One easily sees, that for c C B(c, 1) S(0, i) = n i + 1 if w(c) = i 1, 1 if w(c) = i, i + 1 if w(c) = i + 1, 0 otherwise. 7

34 Therefore by (.6) applied to V = S(0, i) and R = 1 we have E i (0) = E(S(0, i)) = c C = j B(c, 1) S(0, i) S(0, i) c C w(c)=j B(c, 1) S(0, i) S(0, i) = (n i + 1)A i 1 (0) + A i (0) + (i + 1)A i+1 (0) ( ) n. i Theorem 4 (Habsieger [9], Honkala [16]). Let C F n be a binary code with covering radius one and n 1 (mod p) for an odd prime p. If x F n, then E(B(x, p 1)) = E 0 (x) + E 1 (x) E p 1 (x) 1 (mod p). (3.9) Proof. Summing (3.8) with 0 i p 1 gives E i (x) = 0 i p 1 0 i p 1 0 i p 1 = (n + 1) 0 i p 1 0 i p 1 (n i + 1)A i 1 (x) + ( ) n i 0 i p ( ) n i 1 (mod p) by n 1 (mod p). 0 i p 1 A i (x) + A i (x) + p(a p 1 (x) + A p (x)) (mod p) ( 1) i (mod p) 0 i p 1 0 i p 1 (i + 1)A i+1 (x) ( ) n i 8

35 As a consequence we get the following bound on K(n, 1), the general excess bound for binary codes with covering radius one. Theorem 5 (Habsieger [9], Honkala [16]). When n 1 (mod p) for an odd prime p, then ( ) p 1 n K(n, 1) 1 + V (n, p 1) n + 1. (3.30) Proof. Let C F n be a binary code with covering radius one and C = K(n, 1). From (3.9) follows E(B(x, p 1)) p 1, since E(B(x, p 1)) is always nonnegative for x F n. We now get (p 1) n = x F n (p 1) and Theorem 5 follows. x F n E(B(x, p 1)) = x F n z B(x,p 1) = z F n E(z) E(z) x B(z,p 1) = V (n, p 1)E(F n ) = V (n, p 1)( C (n + 1) n ) by (.8) The condition n 1 (mod p) is always satisfied for a suitable odd prime p, except if n = m 1, but then the value of K(n, 1) is known by Theorem 1. If we define the residue classes R p N for an odd prime p by we have R p = {n N n 1 (mod p)}, N \ { m 1 m N} = R 3 R 5... R p.... Then, as a major statement, the general excess bound says K(n, 1) n n + 1 if n, n R p for a fixed odd prime p. (3.31) 1 9

36 Note, that (3.31) (and (3.30)) does not imply K(n, 1) n n + 1 if n, n N \ {m 1 m N}, (3.3) since (3.30) cannot improve on the sphere covering bound, if p is too large compared to n. The quality of the general excess bound depends on the least odd prime divisor of n + 1. A proof of (3.3) could be found in [6]. The rest of this chapter is devoted to the possibility for further improvements of the general excess bound. 3.4 Habsiegers improvement and the excess matrix Let C F n be a code with covering radius one, n odd. Theorem 4 states, that whenever n 1 (mod p) for an odd prime p, then there exists a covering excess in a sphere with radius p 1. This gives rise to the natural question, under which circumstances there exists a perfect 1-covering of a sphere with radius p. A necessary condition is, that p is the least odd prime satisfying n 1 (mod p), since if n 1 (mod p ) with an odd prime p < p, then by Theorem 4 there exists a covering excess in a sphere of radius p 1 p. This condition indeed is sufficient in the cases p = 3 (see the last paragraph of section.3) and p = 5 (see Theorem 1). At the present state of knowledge the existence of a perfect 1-covering of a sphere with radius p, where n is odd and p is the least odd prime divisor of n + 1 cannot proved or refuted when p 7, see section 3... It is perhaps worth noting, that another necessary condition for E i (x) = 0 whenever 0 i p arising from (3.8) is always satisfied. Theorem 6. Let n m 1 be odd and p the least odd prime satisfying n 1 (mod p). Then there exist nonnegative integers A 0,..., A p 1 satisfying ( ) n E i := (n i + 1)A i 1 + A i + (i + 1)A i+1 = 0 (3.33) i for 0 i p (assume A 1 = 0). 30

37 The proof of Theorem 6 is postponed to section 3.5, where the necessary tools are developed. If it cannot be excluded, that there is no covering excess in a sphere of radius p, what can be done to further improve on the general excess bound? Let us note the following immediate consequence of Theorem 4. Theorem 7 (Habsieger [9], Honkala [16]). Assume C F n is a binary code of length n and covering radius one with n 1 (mod p) for an odd prime p. Then for every x F n E 0 (x) = E 1 (x) =... = E p (x) = 0 (3.34) implies E p 1 (x) p 1. (3.35) Habsiegers idea was, that if (3.34) cannot excluded, it could be possible to improve on (3.35). His result is conditional and uses another set of congruence properties of the excess function. Definition 8. Assume C F n is a binary code of length n and covering radius one with odd n. We say, that C satisfies condition (q), if there exist integers β 0,..., β q, such that 0 i q holds whenever x F n. β i E i (x) + E q 1 (x) 0 (mod q) (3.36) Theorem 9 (Habsieger [9]). If p 5, then in Theorem 7, (3.35) may be replaced by E p 1 (x) p 1, (3.37) if additionally C satisfies condition (q) for q = p 1 and q = p. We give the proof of Theorem 9 in a framework different from the original, since we develop material needed later. We first introduce the notion of the excess matrix. For the study of covering excess it suffices to consider spheres centered on 0 and we use the abbreviation E i = E i (0) for i 0. 31

38 Definition 30. Let C F n be a binary code of length n and covering radius one with n 1 (mod p) for an odd prime p. Assume E 0 = E 1 =... = E p = 0. (3.38) Let d j (j A = {1,..., E p 1 }) run over the set Z B(0, p 1), such that every d Z B(0, p 1) occurs exactly E(d) times. Then the E p 1 n- matrix D, whose rows consists of the words d j (j A) is called the excess matrix of C. The column vectors of D are the incident vectors of a subset system B 1,..., B n A, i.e. for 1 i n Note that B i = {j A : the ith coordinate of d j is nonzero} A. w(d j ) = p 1 for j A (3.39) by (3.38). Moreover E p 1 > 0 by (3.35). Thus the excess matrix is nonempty. We next show, that condition (q) implies a strong structure on the column sets B 1,..., B n A of the excess matrix. Theorem 31. Assume the propositions of Definition 30 are satisfied. Let s, r 1,..., r s be any integers satisfying 1 s p 1 and 1 r 1 <... < r s n. We set q = p s. If C satisfies condition (q), then the column sets B i, 1 i n of the excess matrix satisfy B ri 0 (mod q). (3.40) 1 i s Proof. Let y F n denote the unique vector, whose ith coordinate is nonzero if and only if i {r 1,..., r s } (1 i n). Now if z B(y, q ), then d(z, 0) d(z, y) + d(y, 0) p and thus z B(0, p ) implying E(z) = 0 by (3.38). Therefore E i (y) = 0 for 0 i q, which yields E q 1 (y) 0 (mod q) (3.41) by (3.36) (applied to x = y). If z Z B(y, q 1), then d(z, 0) d(z, y) + 3

39 d(y, 0) p 1 and thus z Z B(0, p 1). Therefore E q 1 (y) = E(z) (3.4) We next show = z Z d(z,y)=q 1 z Z B(0,p 1) d(z,y)=q 1 E(z) = {j A d(d j, y) = q 1}. {j A : d(d j, y) = q 1} = 1 i s B ri. (3.43) Assume first d(d j, y) = q 1 for an integer j with 1 j E p 1. To show j 1 i s B r i it suffices to show, that the ith coordinates of d j are nonzero whenever i {r 1,..., r s }. Assume to the contrary, that, say the r 1 th coordinate of d j is zero. This means that there are at most s 1 nonzero coordinates among the set {r 1,..., r s }. Since d j has exactly p 1 nonzero coordinates, d j must have at least (p 1) (s 1) = q nonzero coordinates outside of {r 1,..., r s }. In these coordinates y is zero however, implying d(d j, y) q, contradicting d(d j, y) = q 1. Conversely assume j 1 i s B r i. This means that the ith coordinates of d j are nonzero whenever i {r 1,..., r s }. Moreover d j has exactly q 1 nonzero coordinates outside the set {r 1,..., r s }. Using the definition of y we now see, that y and d j exactly differ in these q 1 coordinates, i.e. d(d j, y) = q 1. This completes the proof of (3.43). Now (3.40) follows by (3.41), (3.4) and (3.43). Proof of Theorem 9. It suffices to consider the case x = 0. Assume equality holds in (3.35), i.e. E p 1 = A = p 1. Since any row of the excess matrix D has weight p 1, D has at least two nonempty column sets, say B 1, B A. Since C satisfies condition (p 1) we may apply Theorem 31 with s = 1 and r 1 = 1 (resp. r 1 = ) and get B 1 B 0 (mod p 1), implying B 1 = B = A by A = p 1. Again by Theorem 31 (with s = and r 1 = 1, r = ) however we find B 1 B 0 (mod p ), since C satisfies condition (p ). This gives the contradiction p 1 = A = B 1 = B 1 B 0 (mod p ). Therefore E p 1 > p 1 and (3.37) follows by (3.34) and Theorem 4. 33

40 In [9] Habsieger explicitly gave congruence properties of type (3.36) for q 5 and raised the question, if the conditions (p 1) and (p ) are always satisfied for C under the propositions of Theorem 9. He wrote: It is probably true, that a necessary and sufficient condition for the existence of (3.36) (for q = p 1 and q = p ) is that p divides n + 1. Since the applications of... require explicit congruence properties, we shall not try to prove this characterization. In the next section we prove the sufficient part of Habsiegers conjecture. Motivated by important structural properties of the excess matrix (see Theorem 31), we show more generally, that condition (q) is satisfied for any binary code of length n and covering radius one, if q and n + 1 do not have a common odd prime divisor. More precisely we prove the following theorem. Theorem 3. Let n, q be integers with 1 q n and odd n. Assume q and n + 1 do not have a common odd prime divisor. Then there are integers β 0,..., β q depending only on q and the residue class of n (mod q), such that for every binary code C F n with covering radius one and each x F n we have β i E i (x) + E q 1 (x) 0 (mod q). (3.44) 0 i q Note that Habsiegers conjecture follows from Theorem 3 applied to q = p resp. q = p 1, since by (3.9) and (3.34) p is the least odd prime satisfying n 1 (mod p) and therefore q and n + 1 do not have a common odd prime divisor. As the second major question concerning Theorem 9, Habsieger [9] asked whether the estimation E p 1 p 1 could be improved further. Indeed Honkala [17] showed E p 1 3p 1. In Habsiegers words:... the inequality E p 1 p 1 can probably sharpened. One might even get a bound which is quadratic in p. As a main goal for the rest of this chapter we will answer this question in the affirmative. Under essential use of the structural properties of the excess matrix stated in Theorem 31 together with the congruence properties described in Theorem 3 we will give a proof of the following unconditional result. 34

41 Theorem 33. Assume C F n is a binary code of length n and covering radius one with n 1 (mod p) for an odd prime p. Then for every x F n E 0 (x) = E 1 (x) =... = E p (x) = 0 (3.45) implies E p 1 (x) (p )p 1. (3.46) In section 3.7 we will use Theorem 33 to give a general improvement of the general excess bound for binary codes with covering radius one. 3.5 Congruence properties for the covering excess function In this section we prove Theorem 3 and discuss some of its consequences including the outstanding proof of Theorem 6. In this section we always assume, that variables with negative index equal zero Some lemmas For the proof of Theorem 3 we need some lemmas. Lemma 34. Let m, n be nonnegative integers with n odd. Then there are integers γ i (0 i n+1) satisfying ( ) n + 1 (n i + 1)γ i + (i + )γ i+1 (mod m ) (3.47) i + 1 for 0 i n 1. Proof. Recursively choose integers u i (0 i n+1 ) satisfying (n i)u i+1 1 (i + 1)u i (mod m ) for 0 i n 1 (here u 0 may be chosen arbitrarily). This is possible, since n i is odd (0 i n 1) and therefore n i and the modul m do not have a common prime divisor. We now set ( ) ( n + 1 γ i = u i 0 i n + 1 ) i 35

42 and find (0 i n 1 ) (n i + 1)γ i + (i + )γ ( ) i+1 ( ) n + 1 n + 1 = (n i + 1)u i + (i + )u i+1 i i + ( ) ( ) n + 1 n + 1 = (n i + 1)u i + (n i)u i+1 i i + 1 ( ) ( ) n + 1 n + 1 (n i + 1)u i + (1 (i + 1)u i ) i i + 1 ( ) n + 1 =. i + 1 (mod m ) Lemma 35. Let m, n be nonnegative integers with n odd. Then there are integers α i (0 i n) satisfying ( ) n (n i + 1)α i 1 + α i + (i + 1)α i+1 (mod m ) (3.48) i for 0 i n 1. Proof. We use Lemma 34 to get integers γ i (0 i n+1 ) satisfying (3.47). The integers α i (0 i n) are defined recursively. When i is even with 0 i n we set α i = γ i α i 1 (0 i n, i even). (3.49) When i is odd with 0 i n we choose any integer α i satisfying ( ) n iα i (n i + )α i α i 1 i 1 (mod m ) (0 i n, i odd). (3.50) This is possible, since i and m do not have a common prime divisor if i is odd. When i is even with 0 i n 1, then (3.48) follows from (3.50) 36

43 (with i + 1 instead of i). When i is odd with 0 i n 1 we have (n i + 1)α i 1 + α i + (i + 1)α i+1 = (n i + )(α i + α i 1) + (i + 1)(α i + α i+1) = = and (3.48) holds again. ((n i + )α i + α i 1 + iα i) ( n i 1 ) ( + 1 γ i 1 + i 1 ) + γ i+1 by (3.49) ((n i + )α i + α i 1 + iα i) ( ) ( ) n + 1 n (mod m ) by (3.47), (3.50) i i 1 ( ) n i Lemma 36. Assume n, u are nonnegative odd integers, such that n+1 and u do not have a common prime divisor. Then there are integers α i (0 i n) satisfying ( ) n (n i + 1)α i 1 + α i + (i + 1)α i+1 (mod u) (3.51) i for 0 i n 1. Proof. Choose an integer α satisfying α(n + 1) 1 (mod u). (3.5) This is possible, since u and n + 1 are assumed to be without common prime divisor. We set ( ) n α i = α (0 i n) i 37

44 and have (0 i n 1) (n i + 1)α i 1 + α i + (i + 1)α i+1 ( ( ) ( ) ( )) n n n = α (n i + 1) + + (i + 1) i 1 i i + 1 ( ( ) ( ) ( )) n n n = α i + + (n i) i i i ( ) n = α(n + 1) i ( ) n (mod u) by (3.5). i Lemma 37. Let n, q be positive integers with n odd. Moreover assume, that n+1 and q do not have a common odd prime divisor. Then there are integers α i (0 i n) satisfying ( ) n (n i + 1)α i 1 + α i + (i + 1)α i+1 (mod q) (3.53) i for 0 i n 1. Proof. Write q = m u (3.54) with m 0 and odd u. Then the propositions of Lemma 35 and Lemma 36 are satisfied and we get integers α i, α i (0 i n) satisfying (3.48) and (3.51). Now by the well-known chinese remainder theorem there are integers α i (0 i n) satisfying α i α i (mod m ), α i α i (mod u) for 0 i n, since m and u do not have a common prime divisor. By (3.48) and (3.51) this implies ( ) n (n i + 1)α i 1 + α i + (i + 1)α i+1 (mod m ), i ( ) n (n i + 1)α i 1 + α i + (i + 1)α i+1 (mod u) i for 0 i n 1 and (3.53) follows by (3.54). 38

Lower Bounds for q-ary Codes with Large Covering Radius

Lower Bounds for q-ary Codes with Large Covering Radius Lower Bounds for q-ary Codes with Large Covering Radius Wolfgang Haas Immanuel Halupczok Jan-Christoph Schlage-Puchta Albert-Ludwigs-Universität Mathematisches Institut Eckerstr. 1 79104 Freiburg, Germany

More information

On Locating-Dominating Codes in Binary Hamming Spaces

On Locating-Dominating Codes in Binary Hamming Spaces Discrete Mathematics and Theoretical Computer Science 6, 2004, 265 282 On Locating-Dominating Codes in Binary Hamming Spaces Iiro Honkala and Tero Laihonen and Sanna Ranto Department of Mathematics and

More information

On non-antipodal binary completely regular codes

On non-antipodal binary completely regular codes On non-antipodal binary completely regular codes J. Borges, J. Rifà Department of Information and Communications Engineering, Universitat Autònoma de Barcelona, 08193-Bellaterra, Spain. V.A. Zinoviev Institute

More information

Support weight enumerators and coset weight distributions of isodual codes

Support weight enumerators and coset weight distributions of isodual codes Support weight enumerators and coset weight distributions of isodual codes Olgica Milenkovic Department of Electrical and Computer Engineering University of Colorado, Boulder March 31, 2003 Abstract In

More information

32 Divisibility Theory in Integral Domains

32 Divisibility Theory in Integral Domains 3 Divisibility Theory in Integral Domains As we have already mentioned, the ring of integers is the prototype of integral domains. There is a divisibility relation on * : an integer b is said to be divisible

More information

Spanning and Independence Properties of Finite Frames

Spanning and Independence Properties of Finite Frames Chapter 1 Spanning and Independence Properties of Finite Frames Peter G. Casazza and Darrin Speegle Abstract The fundamental notion of frame theory is redundancy. It is this property which makes frames

More information

We saw in the last chapter that the linear Hamming codes are nontrivial perfect codes.

We saw in the last chapter that the linear Hamming codes are nontrivial perfect codes. Chapter 5 Golay Codes Lecture 16, March 10, 2011 We saw in the last chapter that the linear Hamming codes are nontrivial perfect codes. Question. Are there any other nontrivial perfect codes? Answer. Yes,

More information

NUMBERS WITH INTEGER COMPLEXITY CLOSE TO THE LOWER BOUND

NUMBERS WITH INTEGER COMPLEXITY CLOSE TO THE LOWER BOUND #A1 INTEGERS 12A (2012): John Selfridge Memorial Issue NUMBERS WITH INTEGER COMPLEXITY CLOSE TO THE LOWER BOUND Harry Altman Department of Mathematics, University of Michigan, Ann Arbor, Michigan haltman@umich.edu

More information

Improved Upper Bounds on Sizes of Codes

Improved Upper Bounds on Sizes of Codes 880 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 48, NO. 4, APRIL 2002 Improved Upper Bounds on Sizes of Codes Beniamin Mounits, Tuvi Etzion, Senior Member, IEEE, and Simon Litsyn, Senior Member, IEEE

More information

Notes on Systems of Linear Congruences

Notes on Systems of Linear Congruences MATH 324 Summer 2012 Elementary Number Theory Notes on Systems of Linear Congruences In this note we will discuss systems of linear congruences where the moduli are all different. Definition. Given the

More information

Equidivisible consecutive integers

Equidivisible consecutive integers & Equidivisible consecutive integers Ivo Düntsch Department of Computer Science Brock University St Catherines, Ontario, L2S 3A1, Canada duentsch@cosc.brocku.ca Roger B. Eggleton Department of Mathematics

More information

Properties of the Integers

Properties of the Integers Properties of the Integers The set of all integers is the set and the subset of Z given by Z = {, 5, 4, 3, 2, 1, 0, 1, 2, 3, 4, 5, }, N = {0, 1, 2, 3, 4, }, is the set of nonnegative integers (also called

More information

a + b = b + a and a b = b a. (a + b) + c = a + (b + c) and (a b) c = a (b c). a (b + c) = a b + a c and (a + b) c = a c + b c.

a + b = b + a and a b = b a. (a + b) + c = a + (b + c) and (a b) c = a (b c). a (b + c) = a b + a c and (a + b) c = a c + b c. Properties of the Integers The set of all integers is the set and the subset of Z given by Z = {, 5, 4, 3, 2, 1, 0, 1, 2, 3, 4, 5, }, N = {0, 1, 2, 3, 4, }, is the set of nonnegative integers (also called

More information

COUNTING NUMERICAL SEMIGROUPS BY GENUS AND SOME CASES OF A QUESTION OF WILF

COUNTING NUMERICAL SEMIGROUPS BY GENUS AND SOME CASES OF A QUESTION OF WILF COUNTING NUMERICAL SEMIGROUPS BY GENUS AND SOME CASES OF A QUESTION OF WILF NATHAN KAPLAN Abstract. The genus of a numerical semigroup is the size of its complement. In this paper we will prove some results

More information

Packing and Covering Dense Graphs

Packing and Covering Dense Graphs Packing and Covering Dense Graphs Noga Alon Yair Caro Raphael Yuster Abstract Let d be a positive integer. A graph G is called d-divisible if d divides the degree of each vertex of G. G is called nowhere

More information

Covering Subsets of the Integers and a Result on Digits of Fibonacci Numbers

Covering Subsets of the Integers and a Result on Digits of Fibonacci Numbers University of South Carolina Scholar Commons Theses and Dissertations 2017 Covering Subsets of the Integers and a Result on Digits of Fibonacci Numbers Wilson Andrew Harvey University of South Carolina

More information

Math 324 Summer 2012 Elementary Number Theory Notes on Mathematical Induction

Math 324 Summer 2012 Elementary Number Theory Notes on Mathematical Induction Math 4 Summer 01 Elementary Number Theory Notes on Mathematical Induction Principle of Mathematical Induction Recall the following axiom for the set of integers. Well-Ordering Axiom for the Integers If

More information

= 1 2x. x 2 a ) 0 (mod p n ), (x 2 + 2a + a2. x a ) 2

= 1 2x. x 2 a ) 0 (mod p n ), (x 2 + 2a + a2. x a ) 2 8. p-adic numbers 8.1. Motivation: Solving x 2 a (mod p n ). Take an odd prime p, and ( an) integer a coprime to p. Then, as we know, x 2 a (mod p) has a solution x Z iff = 1. In this case we can suppose

More information

PUTNAM TRAINING NUMBER THEORY. Exercises 1. Show that the sum of two consecutive primes is never twice a prime.

PUTNAM TRAINING NUMBER THEORY. Exercises 1. Show that the sum of two consecutive primes is never twice a prime. PUTNAM TRAINING NUMBER THEORY (Last updated: December 11, 2017) Remark. This is a list of exercises on Number Theory. Miguel A. Lerma Exercises 1. Show that the sum of two consecutive primes is never twice

More information

Instructor: Bobby Kleinberg Lecture Notes, 25 April The Miller-Rabin Randomized Primality Test

Instructor: Bobby Kleinberg Lecture Notes, 25 April The Miller-Rabin Randomized Primality Test Introduction to Algorithms (CS 482) Cornell University Instructor: Bobby Kleinberg Lecture Notes, 25 April 2008 The Miller-Rabin Randomized Primality Test 1 Introduction Primality testing is an important

More information

Week 15-16: Combinatorial Design

Week 15-16: Combinatorial Design Week 15-16: Combinatorial Design May 8, 2017 A combinatorial design, or simply a design, is an arrangement of the objects of a set into subsets satisfying certain prescribed properties. The area of combinatorial

More information

11 Minimal Distance and the Parity Check Matrix

11 Minimal Distance and the Parity Check Matrix MATH32031: Coding Theory Part 12: Hamming Codes 11 Minimal Distance and the Parity Check Matrix Theorem 23 (Distance Theorem for Linear Codes) Let C be an [n, k] F q -code with parity check matrix H. Then

More information

This is a recursive algorithm. The procedure is guaranteed to terminate, since the second argument decreases each time.

This is a recursive algorithm. The procedure is guaranteed to terminate, since the second argument decreases each time. 8 Modular Arithmetic We introduce an operator mod. Let d be a positive integer. For c a nonnegative integer, the value c mod d is the remainder when c is divided by d. For example, c mod d = 0 if and only

More information

FRACTIONAL FACTORIAL DESIGNS OF STRENGTH 3 AND SMALL RUN SIZES

FRACTIONAL FACTORIAL DESIGNS OF STRENGTH 3 AND SMALL RUN SIZES FRACTIONAL FACTORIAL DESIGNS OF STRENGTH 3 AND SMALL RUN SIZES ANDRIES E. BROUWER, ARJEH M. COHEN, MAN V.M. NGUYEN Abstract. All mixed (or asymmetric) orthogonal arrays of strength 3 with run size at most

More information

Modular Arithmetic Instructor: Marizza Bailey Name:

Modular Arithmetic Instructor: Marizza Bailey Name: Modular Arithmetic Instructor: Marizza Bailey Name: 1. Introduction to Modular Arithmetic If someone asks you what day it is 145 days from now, what would you answer? Would you count 145 days, or find

More information

Ahlswede Khachatrian Theorems: Weighted, Infinite, and Hamming

Ahlswede Khachatrian Theorems: Weighted, Infinite, and Hamming Ahlswede Khachatrian Theorems: Weighted, Infinite, and Hamming Yuval Filmus April 4, 2017 Abstract The seminal complete intersection theorem of Ahlswede and Khachatrian gives the maximum cardinality of

More information

PRACTICE PROBLEMS: SET 1

PRACTICE PROBLEMS: SET 1 PRACTICE PROBLEMS: SET MATH 437/537: PROF. DRAGOS GHIOCA. Problems Problem. Let a, b N. Show that if gcd(a, b) = lcm[a, b], then a = b. Problem. Let n, k N with n. Prove that (n ) (n k ) if and only if

More information

Triangle-free graphs with no six-vertex induced path

Triangle-free graphs with no six-vertex induced path Triangle-free graphs with no six-vertex induced path Maria Chudnovsky 1, Paul Seymour 2, Sophie Spirkl Princeton University, Princeton, NJ 08544 Mingxian Zhong Columbia University, New York, NY 10027 June

More information

ORTHOGONAL ARRAYS OF STRENGTH 3 AND SMALL RUN SIZES

ORTHOGONAL ARRAYS OF STRENGTH 3 AND SMALL RUN SIZES ORTHOGONAL ARRAYS OF STRENGTH 3 AND SMALL RUN SIZES ANDRIES E. BROUWER, ARJEH M. COHEN, MAN V.M. NGUYEN Abstract. All mixed (or asymmetric) orthogonal arrays of strength 3 with run size at most 64 are

More information

Packing cycles with modularity constraints

Packing cycles with modularity constraints Packing cycles with modularity constraints Paul Wollan Mathematisches Seminar der Universität Hamburg Bundesstr. 55 20146 Hamburg, Germany Abstract We prove that for all positive integers k, there exists

More information

Irredundant Families of Subcubes

Irredundant Families of Subcubes Irredundant Families of Subcubes David Ellis January 2010 Abstract We consider the problem of finding the maximum possible size of a family of -dimensional subcubes of the n-cube {0, 1} n, none of which

More information

Transposition as a permutation: a tale of group actions and modular arithmetic

Transposition as a permutation: a tale of group actions and modular arithmetic Transposition as a permutation: a tale of group actions and modular arithmetic Jeff Hooper Franklin Mendivil Department of Mathematics and Statistics Acadia University Abstract Converting a matrix from

More information

Lecture 6: Finite Fields

Lecture 6: Finite Fields CCS Discrete Math I Professor: Padraic Bartlett Lecture 6: Finite Fields Week 6 UCSB 2014 It ain t what they call you, it s what you answer to. W. C. Fields 1 Fields In the next two weeks, we re going

More information

3. Coding theory 3.1. Basic concepts

3. Coding theory 3.1. Basic concepts 3. CODING THEORY 1 3. Coding theory 3.1. Basic concepts In this chapter we will discuss briefly some aspects of error correcting codes. The main problem is that if information is sent via a noisy channel,

More information

Decomposing Bent Functions

Decomposing Bent Functions 2004 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 49, NO. 8, AUGUST 2003 Decomposing Bent Functions Anne Canteaut and Pascale Charpin Abstract In a recent paper [1], it is shown that the restrictions

More information

The Witt designs, Golay codes and Mathieu groups

The Witt designs, Golay codes and Mathieu groups The Witt designs, Golay codes and Mathieu groups 1 The Golay codes Let V be a vector space over F q with fixed basis e 1,..., e n. A code C is a subset of V. A linear code is a subspace of V. The vector

More information

MATH 291T CODING THEORY

MATH 291T CODING THEORY California State University, Fresno MATH 291T CODING THEORY Spring 2009 Instructor : Stefaan Delcroix Chapter 1 Introduction to Error-Correcting Codes It happens quite often that a message becomes corrupt

More information

On some inequalities between prime numbers

On some inequalities between prime numbers On some inequalities between prime numbers Martin Maulhardt July 204 ABSTRACT. In 948 Erdős and Turán proved that in the set of prime numbers the inequality p n+2 p n+ < p n+ p n is satisfied infinitely

More information

WORKSHEET ON NUMBERS, MATH 215 FALL. We start our study of numbers with the integers: N = {1, 2, 3,...}

WORKSHEET ON NUMBERS, MATH 215 FALL. We start our study of numbers with the integers: N = {1, 2, 3,...} WORKSHEET ON NUMBERS, MATH 215 FALL 18(WHYTE) We start our study of numbers with the integers: Z = {..., 2, 1, 0, 1, 2, 3,... } and their subset of natural numbers: N = {1, 2, 3,...} For now we will not

More information

On Anti-Elite Prime Numbers

On Anti-Elite Prime Numbers 1 2 3 47 6 23 11 Journal of Integer Sequences, Vol. 10 (2007), Article 07.9.4 On Anti-Elite Prime Numbers Tom Müller Institut für Cusanus-Forschung an der Universität und der Theologischen Fakultät Trier

More information

8 Primes and Modular Arithmetic

8 Primes and Modular Arithmetic 8 Primes and Modular Arithmetic 8.1 Primes and Factors Over two millennia ago already, people all over the world were considering the properties of numbers. One of the simplest concepts is prime numbers.

More information

Chapter 2. Error Correcting Codes. 2.1 Basic Notions

Chapter 2. Error Correcting Codes. 2.1 Basic Notions Chapter 2 Error Correcting Codes The identification number schemes we discussed in the previous chapter give us the ability to determine if an error has been made in recording or transmitting information.

More information

Facets for Node-Capacitated Multicut Polytopes from Path-Block Cycles with Two Common Nodes

Facets for Node-Capacitated Multicut Polytopes from Path-Block Cycles with Two Common Nodes Facets for Node-Capacitated Multicut Polytopes from Path-Block Cycles with Two Common Nodes Michael M. Sørensen July 2016 Abstract Path-block-cycle inequalities are valid, and sometimes facet-defining,

More information

COMPLETION OF PARTIAL LATIN SQUARES

COMPLETION OF PARTIAL LATIN SQUARES COMPLETION OF PARTIAL LATIN SQUARES Benjamin Andrew Burton Honours Thesis Department of Mathematics The University of Queensland Supervisor: Dr Diane Donovan Submitted in 1996 Author s archive version

More information

Solutions to the 74th William Lowell Putnam Mathematical Competition Saturday, December 7, 2013

Solutions to the 74th William Lowell Putnam Mathematical Competition Saturday, December 7, 2013 Solutions to the 74th William Lowell Putnam Mathematical Competition Saturday, December 7, 213 Kiran Kedlaya and Lenny Ng A1 Suppose otherwise. Then each vertex v is a vertex for five faces, all of which

More information

arxiv: v1 [math.nt] 19 Dec 2018

arxiv: v1 [math.nt] 19 Dec 2018 ON SECOND ORDER LINEAR SEQUENCES OF COMPOSITE NUMBERS DAN ISMAILESCU 2, ADRIENNE KO 1, CELINE LEE 3, AND JAE YONG PARK 4 arxiv:1812.08041v1 [math.nt] 19 Dec 2018 Abstract. In this paper we present a new

More information

Roth s Theorem on 3-term Arithmetic Progressions

Roth s Theorem on 3-term Arithmetic Progressions Roth s Theorem on 3-term Arithmetic Progressions Mustazee Rahman 1 Introduction This article is a discussion about the proof of a classical theorem of Roth s regarding the existence of three term arithmetic

More information

QUADRATIC RINGS PETE L. CLARK

QUADRATIC RINGS PETE L. CLARK QUADRATIC RINGS PETE L. CLARK 1. Quadratic fields and quadratic rings Let D be a squarefree integer not equal to 0 or 1. Then D is irrational, and Q[ D], the subring of C obtained by adjoining D to Q,

More information

2x 1 7. A linear congruence in modular arithmetic is an equation of the form. Why is the solution a set of integers rather than a unique integer?

2x 1 7. A linear congruence in modular arithmetic is an equation of the form. Why is the solution a set of integers rather than a unique integer? Chapter 3: Theory of Modular Arithmetic 25 SECTION C Solving Linear Congruences By the end of this section you will be able to solve congruence equations determine the number of solutions find the multiplicative

More information

THESIS. Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School of The Ohio State University

THESIS. Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School of The Ohio State University The Hasse-Minkowski Theorem in Two and Three Variables THESIS Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School of The Ohio State University By

More information

The Hamming Codes and Delsarte s Linear Programming Bound

The Hamming Codes and Delsarte s Linear Programming Bound The Hamming Codes and Delsarte s Linear Programming Bound by Sky McKinley Under the Astute Tutelage of Professor John S. Caughman, IV A thesis submitted in partial fulfillment of the requirements for the

More information

D-bounded Distance-Regular Graphs

D-bounded Distance-Regular Graphs D-bounded Distance-Regular Graphs CHIH-WEN WENG 53706 Abstract Let Γ = (X, R) denote a distance-regular graph with diameter D 3 and distance function δ. A (vertex) subgraph X is said to be weak-geodetically

More information

SUMS PROBLEM COMPETITION, 2000

SUMS PROBLEM COMPETITION, 2000 SUMS ROBLEM COMETITION, 2000 SOLUTIONS 1 The result is well known, and called Morley s Theorem Many proofs are known See for example HSM Coxeter, Introduction to Geometry, page 23 2 If the number of vertices,

More information

Number Theory. Final Exam from Spring Solutions

Number Theory. Final Exam from Spring Solutions Number Theory. Final Exam from Spring 2013. Solutions 1. (a) (5 pts) Let d be a positive integer which is not a perfect square. Prove that Pell s equation x 2 dy 2 = 1 has a solution (x, y) with x > 0,

More information

Orthogonal Arrays & Codes

Orthogonal Arrays & Codes Orthogonal Arrays & Codes Orthogonal Arrays - Redux An orthogonal array of strength t, a t-(v,k,λ)-oa, is a λv t x k array of v symbols, such that in any t columns of the array every one of the possible

More information

Groups that Distribute over Stars

Groups that Distribute over Stars Groups that Distribute over Stars Arthur Holshouser 3600 Bullard St Charlotte, NC, USA, 808 Harold Reiter Department of Mathematics UNC Charlotte Charlotte, NC 83 hbreiter@emailunccedu 1 Abstract Suppose

More information

1 Basic Combinatorics

1 Basic Combinatorics 1 Basic Combinatorics 1.1 Sets and sequences Sets. A set is an unordered collection of distinct objects. The objects are called elements of the set. We use braces to denote a set, for example, the set

More information

Measures and Measure Spaces

Measures and Measure Spaces Chapter 2 Measures and Measure Spaces In summarizing the flaws of the Riemann integral we can focus on two main points: 1) Many nice functions are not Riemann integrable. 2) The Riemann integral does not

More information

Hypercube Coloring and the Structure of Binary Codes

Hypercube Coloring and the Structure of Binary Codes Hypercube Coloring and the Structure of Binary Codes by James Gregory Rix A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in The College of Graduate Studies

More information

THE TRIANGULAR THEOREM OF THE PRIMES : BINARY QUADRATIC FORMS AND PRIMITIVE PYTHAGOREAN TRIPLES

THE TRIANGULAR THEOREM OF THE PRIMES : BINARY QUADRATIC FORMS AND PRIMITIVE PYTHAGOREAN TRIPLES THE TRIANGULAR THEOREM OF THE PRIMES : BINARY QUADRATIC FORMS AND PRIMITIVE PYTHAGOREAN TRIPLES Abstract. This article reports the occurrence of binary quadratic forms in primitive Pythagorean triangles

More information

Finite Mathematics. Nik Ruškuc and Colva M. Roney-Dougal

Finite Mathematics. Nik Ruškuc and Colva M. Roney-Dougal Finite Mathematics Nik Ruškuc and Colva M. Roney-Dougal September 19, 2011 Contents 1 Introduction 3 1 About the course............................. 3 2 A review of some algebraic structures.................

More information

Constructions with ruler and compass.

Constructions with ruler and compass. Constructions with ruler and compass. Semyon Alesker. 1 Introduction. Let us assume that we have a ruler and a compass. Let us also assume that we have a segment of length one. Using these tools we can

More information

Rigid Divisibility Sequences Generated by Polynomial Iteration

Rigid Divisibility Sequences Generated by Polynomial Iteration Rigid Divisibility Sequences Generated by Polynomial Iteration Brian Rice Nicholas Pippenger, Advisor Christopher Towse, Reader May, 2008 Department of Mathematics Copyright c 2008 Brian Rice. The author

More information

CS264: Beyond Worst-Case Analysis Lecture #11: LP Decoding

CS264: Beyond Worst-Case Analysis Lecture #11: LP Decoding CS264: Beyond Worst-Case Analysis Lecture #11: LP Decoding Tim Roughgarden October 29, 2014 1 Preamble This lecture covers our final subtopic within the exact and approximate recovery part of the course.

More information

Cross-Intersecting Sets of Vectors

Cross-Intersecting Sets of Vectors Cross-Intersecting Sets of Vectors János Pach Gábor Tardos Abstract Given a sequence of positive integers p = (p 1,..., p n ), let S p denote the set of all sequences of positive integers x = (x 1,...,

More information

Packing and decomposition of graphs with trees

Packing and decomposition of graphs with trees Packing and decomposition of graphs with trees Raphael Yuster Department of Mathematics University of Haifa-ORANIM Tivon 36006, Israel. e-mail: raphy@math.tau.ac.il Abstract Let H be a tree on h 2 vertices.

More information

Incidence Structures Related to Difference Sets and Their Applications

Incidence Structures Related to Difference Sets and Their Applications aòµ 05B30 ü èµ Æ Òµ 113350 Æ Æ Ø Ø K8: 'u8'é(9ùa^ = Ø K8: Incidence Structures Related to Difference Sets and Their Applications úôœææ Æ Ø ž

More information

SUBGROUPS OF CYCLIC GROUPS. 1. Introduction In a group G, we denote the (cyclic) group of powers of some g G by

SUBGROUPS OF CYCLIC GROUPS. 1. Introduction In a group G, we denote the (cyclic) group of powers of some g G by SUBGROUPS OF CYCLIC GROUPS KEITH CONRAD 1. Introduction In a group G, we denote the (cyclic) group of powers of some g G by g = {g k : k Z}. If G = g, then G itself is cyclic, with g as a generator. Examples

More information

Course MA2C02, Hilary Term 2013 Section 9: Introduction to Number Theory and Cryptography

Course MA2C02, Hilary Term 2013 Section 9: Introduction to Number Theory and Cryptography Course MA2C02, Hilary Term 2013 Section 9: Introduction to Number Theory and Cryptography David R. Wilkins Copyright c David R. Wilkins 2000 2013 Contents 9 Introduction to Number Theory 63 9.1 Subgroups

More information

2x 1 7. A linear congruence in modular arithmetic is an equation of the form. Why is the solution a set of integers rather than a unique integer?

2x 1 7. A linear congruence in modular arithmetic is an equation of the form. Why is the solution a set of integers rather than a unique integer? Chapter 3: Theory of Modular Arithmetic 25 SECTION C Solving Linear Congruences By the end of this section you will be able to solve congruence equations determine the number of solutions find the multiplicative

More information

Baltic Way 2003 Riga, November 2, 2003

Baltic Way 2003 Riga, November 2, 2003 altic Way 2003 Riga, November 2, 2003 Problems and solutions. Let Q + be the set of positive rational numbers. Find all functions f : Q + Q + which for all x Q + fulfil () f ( x ) = f (x) (2) ( + x ) f

More information

6 Cosets & Factor Groups

6 Cosets & Factor Groups 6 Cosets & Factor Groups The course becomes markedly more abstract at this point. Our primary goal is to break apart a group into subsets such that the set of subsets inherits a natural group structure.

More information

Every SOMA(n 2, n) is Trojan

Every SOMA(n 2, n) is Trojan Every SOMA(n 2, n) is Trojan John Arhin 1 Marlboro College, PO Box A, 2582 South Road, Marlboro, Vermont, 05344, USA. Abstract A SOMA(k, n) is an n n array A each of whose entries is a k-subset of a knset

More information

Functions and cardinality (solutions) sections A and F TA: Clive Newstead 6 th May 2014

Functions and cardinality (solutions) sections A and F TA: Clive Newstead 6 th May 2014 Functions and cardinality (solutions) 21-127 sections A and F TA: Clive Newstead 6 th May 2014 What follows is a somewhat hastily written collection of solutions for my review sheet. I have omitted some

More information

On the order bound of one-point algebraic geometry codes.

On the order bound of one-point algebraic geometry codes. On the order bound of one-point algebraic geometry codes. Anna Oneto and Grazia Tamone 1 Abstract. Let S ={s i} i IN IN be a numerical semigroup. For each i IN, let ν(s i) denote the number of pairs (s

More information

Chapter 5. Number Theory. 5.1 Base b representations

Chapter 5. Number Theory. 5.1 Base b representations Chapter 5 Number Theory The material in this chapter offers a small glimpse of why a lot of facts that you ve probably nown and used for a long time are true. It also offers some exposure to generalization,

More information

ABOUT THE CLASS AND NOTES ON SET THEORY

ABOUT THE CLASS AND NOTES ON SET THEORY ABOUT THE CLASS AND NOTES ON SET THEORY About the Class Evaluation. Final grade will be based 25%, 25%, 25%, 25%, on homework, midterm 1, midterm 2, final exam. Exam dates. Midterm 1: Oct 4. Midterm 2:

More information

arxiv: v1 [math.co] 27 Aug 2015

arxiv: v1 [math.co] 27 Aug 2015 P-positions in Modular Extensions to Nim arxiv:1508.07054v1 [math.co] 7 Aug 015 Tanya Khovanova August 31, 015 Abstract Karan Sarkar In this paper, we consider a modular extension to the game of Nim, which

More information

Tilings of Binary Spaces

Tilings of Binary Spaces Tilings of Binary Spaces Gérard Cohen Département Informatique ENST, 46 rue Barrault 75634 Paris, France Simon Litsyn Department of Electrical Engineering Tel-Aviv University Ramat-Aviv 69978, Israel Alexander

More information

Multiple coverings of the farthest-off points and multiple saturating sets in projective spaces

Multiple coverings of the farthest-off points and multiple saturating sets in projective spaces Thirteenth International Workshop on Algebraic and Combinatorial Coding Theory June 15-21, 2012, Pomorie, Bulgaria pp. 53 59 Multiple coverings of the farthest-off points and multiple saturating sets in

More information

Sydney University Mathematical Society Problems Competition Solutions.

Sydney University Mathematical Society Problems Competition Solutions. Sydney University Mathematical Society Problems Competition 005 Solutions 1 Suppose that we look at the set X n of strings of 0 s and 1 s of length n Given a string ɛ = (ɛ 1,, ɛ n ) X n, we are allowed

More information

Number Theory and Graph Theory. Prime numbers and congruences.

Number Theory and Graph Theory. Prime numbers and congruences. 1 Number Theory and Graph Theory Chapter 2 Prime numbers and congruences. By A. Satyanarayana Reddy Department of Mathematics Shiv Nadar University Uttar Pradesh, India E-mail: satya8118@gmail.com 2 Module-1:Primes

More information

Perfect matchings in highly cyclically connected regular graphs

Perfect matchings in highly cyclically connected regular graphs Perfect matchings in highly cyclically connected regular graphs arxiv:1709.08891v1 [math.co] 6 Sep 017 Robert Lukot ka Comenius University, Bratislava lukotka@dcs.fmph.uniba.sk Edita Rollová University

More information

ON A PROBLEM OF PILLAI AND ITS GENERALIZATIONS

ON A PROBLEM OF PILLAI AND ITS GENERALIZATIONS ON A PROBLEM OF PILLAI AND ITS GENERALIZATIONS L. HAJDU 1 AND N. SARADHA Abstract. We study some generalizations of a problem of Pillai. We investigate the existence of an integer M such that for m M,

More information

A LOWER BOUND FOR THE SIZE OF A MINKOWSKI SUM OF DILATES. 1. Introduction

A LOWER BOUND FOR THE SIZE OF A MINKOWSKI SUM OF DILATES. 1. Introduction A LOWER BOUND FOR THE SIZE OF A MINKOWSKI SUM OF DILATES Y. O. HAMIDOUNE AND J. RUÉ Abstract. Let A be a finite nonempty set of integers. An asymptotic estimate of several dilates sum size was obtained

More information

Discrete Mathematics

Discrete Mathematics Discrete Mathematics 309 (009) 3811 380 Contents lists available at ScienceDirect Discrete Mathematics journal homepage: www.elsevier.com/locate/disc Disjoint hamiltonian cycles in bipartite graphs Michael

More information

Math 117: Topology of the Real Numbers

Math 117: Topology of the Real Numbers Math 117: Topology of the Real Numbers John Douglas Moore November 10, 2008 The goal of these notes is to highlight the most important topics presented in Chapter 3 of the text [1] and to provide a few

More information

Rank and Kernel of binary Hadamard codes.

Rank and Kernel of binary Hadamard codes. 1 Rank and Kernel of binary Hadamard codes. K.T. Phelps, J. Rifà Senior Member IEEE, M. Villanueva Abstract In this paper the rank and the dimension of the kernel for (binary) Hadamard codes of length

More information

1. A Little Set Theory

1. A Little Set Theory . A Little Set Theory I see it, but I don t believe it. Cantor to Dedekind 29 June 877 Functions are the single most important idea pervading modern mathematics. We will assume the informal definition

More information

Standard forms for writing numbers

Standard forms for writing numbers Standard forms for writing numbers In order to relate the abstract mathematical descriptions of familiar number systems to the everyday descriptions of numbers by decimal expansions and similar means,

More information

Quadratic reciprocity and the Jacobi symbol Stephen McAdam Department of Mathematics University of Texas at Austin

Quadratic reciprocity and the Jacobi symbol Stephen McAdam Department of Mathematics University of Texas at Austin Quadratic reciprocity and the Jacobi symbol Stephen McAdam Department of Mathematics University of Texas at Austin mcadam@math.utexas.edu Abstract: We offer a proof of quadratic reciprocity that arises

More information

Lecture 4: Proof of Shannon s theorem and an explicit code

Lecture 4: Proof of Shannon s theorem and an explicit code CSE 533: Error-Correcting Codes (Autumn 006 Lecture 4: Proof of Shannon s theorem and an explicit code October 11, 006 Lecturer: Venkatesan Guruswami Scribe: Atri Rudra 1 Overview Last lecture we stated

More information

TOPOLOGICAL COMPLEXITY OF 2-TORSION LENS SPACES AND ku-(co)homology

TOPOLOGICAL COMPLEXITY OF 2-TORSION LENS SPACES AND ku-(co)homology TOPOLOGICAL COMPLEXITY OF 2-TORSION LENS SPACES AND ku-(co)homology DONALD M. DAVIS Abstract. We use ku-cohomology to determine lower bounds for the topological complexity of mod-2 e lens spaces. In the

More information

Codewords of small weight in the (dual) code of points and k-spaces of P G(n, q)

Codewords of small weight in the (dual) code of points and k-spaces of P G(n, q) Codewords of small weight in the (dual) code of points and k-spaces of P G(n, q) M. Lavrauw L. Storme G. Van de Voorde October 4, 2007 Abstract In this paper, we study the p-ary linear code C k (n, q),

More information

Optimal XOR based (2,n)-Visual Cryptography Schemes

Optimal XOR based (2,n)-Visual Cryptography Schemes Optimal XOR based (2,n)-Visual Cryptography Schemes Feng Liu and ChuanKun Wu State Key Laboratory Of Information Security, Institute of Software Chinese Academy of Sciences, Beijing 0090, China Email:

More information

Lecture 2: Vector Spaces, Metric Spaces

Lecture 2: Vector Spaces, Metric Spaces CCS Discrete II Professor: Padraic Bartlett Lecture 2: Vector Spaces, Metric Spaces Week 2 UCSB 2015 1 Vector Spaces, Informally The two vector spaces 1 you re probably the most used to working with, from

More information

Introduction to Abstract Mathematics

Introduction to Abstract Mathematics Introduction to Abstract Mathematics Notation: Z + or Z >0 denotes the set {1, 2, 3,...} of positive integers, Z 0 is the set {0, 1, 2,...} of nonnegative integers, Z is the set {..., 1, 0, 1, 2,...} of

More information

New infinite families of Candelabra Systems with block size 6 and stem size 2

New infinite families of Candelabra Systems with block size 6 and stem size 2 New infinite families of Candelabra Systems with block size 6 and stem size 2 Niranjan Balachandran Department of Mathematics The Ohio State University Columbus OH USA 4210 email:niranj@math.ohio-state.edu

More information

CHAOTIC BEHAVIOR IN A FORECAST MODEL

CHAOTIC BEHAVIOR IN A FORECAST MODEL CHAOTIC BEHAVIOR IN A FORECAST MODEL MICHAEL BOYLE AND MARK TOMFORDE Abstract. We examine a certain interval map, called the weather map, that has been used by previous authors as a toy model for weather

More information

Number Theory Solutions Packet

Number Theory Solutions Packet Number Theory Solutions Pacet 1 There exist two distinct positive integers, both of which are divisors of 10 10, with sum equal to 157 What are they? Solution Suppose 157 = x + y for x and y divisors of

More information