Morita Equivalence for Unary Varieties

Morita Equivalence for Unary Varieties Tobias Rieck Dissertation zur Erlangung des Grades eines Doktors der Naturwissenschaften Dr. rer. nat. Vorgelegt im Fachbereich 3 (Mathematik & Informatik) der Universität Bremen im Januar 2003

Datum des Promotionskolloquiums: 3. März 2003 Gutachter: Prof. Dr. Hans-Eberhard Porst (Universität Bremen) Prof. Dr. Horst Herrlich (Universität Bremen)

Contents Preface v 1 Introduction 1 1.1 Universal Algebra and Varieties................... 2 1.2 Lawvere Theories........................... 4 1.3 Varietal Generators.......................... 5 1.4 Matrix Power and u-modification.................. 7 2 The Matrix Power 11 2.1 The Variety V (n)........................... 12 2.1.1 V (n) is the Matrix Power of V................ 13 2.2 An Alternative Description of the Matrix Power: V {n}...... 15 2.2.1 V {n} is the Matrix Power of V................ 16 3 The Matrix Powers of Set 19 3.1 n-set 1................................. 20 3.1.1 The Equivalence of n-set 1 to Set............. 22 3.2 n-set 2................................. 24 3.2.1 The Equivalence of n-set 2 to Set............. 25 3.3 n-set 3................................. 27 3.3.1 The Equivalence of n-set 3 to Set............. 28 3.4 n-set 4................................. 30 3.4.1 The Equivalence of n-set 4 to Set............. 31 iii

iv CONTENTS 4 u-modification for Unary Varieties 33 4.1 Varietal Generators and Invertible Unary Operations....... 34 4.2 Clifford Monoids and V(M, u).................... 35 4.2.1 V(M, u) is the u-modification of M Act [n]......... 38 4.3 Equivalence for M-Acts....................... 44 5 Morita Equivalence for Boolean Algebras 47 5.1 The Matrix Powers of BOOL.................... 47 5.2 The u-modification.......................... 48 5.3 The Equivalence of Post Algebras and Boolean Algebras..... 49 5.4 Hu s Primal Algebra Theorem.................... 54 References 57 Index of Symbols 61 Index 63

Preface The problem of determining all varieties W (categorically) equivalent to a given variety V has been addressed in numerous ways. Isbell first posed the problem in the early 1970 s in [15]. There are several solutions for special cases. The most prominent example is classical Morita theory which describes for a given ring R all rings S such that the varieties of R-modules and S-modules are equivalent (see e.g. in [16]). At first sight, this seems to be a more restricted problem. But since any variety equivalent to a category of modules is itself a variety of S-modules for a suitable ring S (cf. e.g. [12]), classical Morita theory characterizes all varieties equivalent to the variety of R-modules for a given ring R. Classical Morita theory is based on certain generators in the category of R- modules. This has been generalized. General Morita theory, as described in [27], is based on the characterization of varietal generators. This characterization gives rise to two constructions: the n-th matrix power V [n] of a variety V for natural numbers n 2 and the u-modification V(u) of V for an idempotent and invertible term u in V. The varieties equivalent to a given variety V are, up to concrete isomorphism, precisely the varieties V [n] (u) for some n N, n 1, and some idempotent and invertible term u for V [n]. These constructions are already described in R. N. McKenzie s paper [25], which is based on results from J. J. Dukarm [9]. The construction of the matrix power was already known to F. E. J. Linton during the 1960 s according to R. N. McKenzie (cf. [25]). The role played by the varietal generators and how the concepts of the matrix power and u-modification arise from them only becomes fully clear in H.-E. Porst s paper [27]. Using basic results from categorical algebra, as developed mainly by Lawvere, Isbell and Linton, he characterizes all Lawvere theories S Morita equivalent to a given Lawvere theory T. Lawvere theories T and S are called Morita equivalent provided that the categories of their models ModT and ModS are equivalent categories. Thus the Lawvere theories correspond to the rings in classical Morita theory. In this thesis we give new descriptions for the n-th matrix power V [n] of a (finitary) variety V for natural numbers n 2 and the u-modification V(u) of V for an idempotent and invertible term u in V. The aim is to simplify the syntax. Especially in the case of the matrix power we have succeeded in giving a very simple characterization by adding just one binary operation to the original v

vi Preface operations of a given variety V as well as adding equations to the original ones. The u-modification is more elusive and we have to restrict ourselves to the unary case. Again we gain a characterization by adding operations and equations to the originally given ones. Furthermore we treat some special cases like the variety of Boolean algebras as illustrating examples. The outline of this thesis is the following: The first chapter provides a brief summary of the fundamental concepts needed in the following chapters. After sketching both Birkhoff s as well as Lawvere s approach to varieties, we give a characterization of varietal generators. Finally, the notions of n-th matrix power V [n] of a variety V for natural numbers n 2 and the u-modification V(u) of V for an idempotent and invertible term u in V are introduced. After the first chapter we leave it to the reader to decide in which order to read Chapter 2 and Chapter 3. Both chapters can be read on their own. Whereas Chapter 2 describes the matrix power in the general case, Chapter 3 contains the matrix powers of the category Set of sets and maps. In the second chapter we characterize the n-th matrix power V [n] of a variety V for all natural numbers n 2. We give two different descriptions, both times by adding operations and equations to the existing syntax. After explaining how these operations work for a given algebra, we show that they indeed completely characterize the matrix powers. The first description V (n) is more elegant whereas the second characterization V {n} is easier to work with in the case of unary varieties in Chapter 4. The results in the second chapter are a generalization of the results in Chapter 3. Chapter 3 contains four different descriptions of the matrix powers Set [n] of Set for all natural numbers n 2. As in the second chapter we add operations and equations to the existing ones. The characterizations are generalizations of special instances of the matrix powers of Set. The first two, n-set 1 and n-set 2, are direct generalizations of R. Börger s example in [5], the third, n-set 3, uses R. N. MacKenzie s operations taken from Example 2 in [25]. The fourth and most refined solution n-set 4, is based on a variety constructed by Saade [31]. Börger s and MacKenzies s examples just give the n-th matrix power Set [n] of Set for n = 2. Chapter 4 is dedicated to finding all varieties equivalent to a given unary variety. For this it is sufficient to deal with M-acts, since all unary varieties can be described as varieties of M-acts. First we determine the varietal generators in M Act and thereby find the idempotent invertible unary operations u needed for the u-modification. We do not fully achieve our goal of describing Morita equivalence for all unary varieties since we have to restrict ourselves to the varieties M Act of M-acts where M is a Clifford monoid. But for those cases we construct the u-modification of the matrix powers M Act {n} of M Act for all suitable idempotent and invertible terms u in M Act {n}. We define a variety V(M, u) via explicitly given operations and equations and show that it is indeed the u-modification of the n-th matrix powers of M Act if M is a Clifford Monoid. We finish the fourth chapter by answering a similar question

vii to the one Morita treated: Given a monoid M, for which monoids N is the variety N Act equivalent to M Act? Knauer [19] and Banachschewsky [4] both independently found an answer to this question in 1972. We give a short proof of their result using categorical algebra. In the fifth and last chapter we prove that the varieties equivalent to the variety BOOL of Boolean algebras are (up to concrete isomorphism) precisely the varieties P n of Post algebras of order n for n N, n 2. This is done by writing down the constructions of the matrix power and the u-modification explicitly, thus giving a complete characterization through operations and equations. It turns out that we get exactly the description of Post algebras by axioms which were given by T. Traczyk in [32]. Of course, we also show the reverse direction, that every Post algebra of order n+1 (n N) is isomorphic to the u-modification of an n-th matrix power of a Boolean algebra. The result is already contained in [27] but we make the construction of the matrix power and u-modification of the variety BOOL of Boolean algebras and some other parts of the proof more explicit than they are given by Porst. We finish the chapter by shortly pointing out the connection to Hu s primal algebra theorem. Notation Some symbols describe different operations in different sections, but only when the operations have the same underlying principle. We refrained from using indices in these cases since we had to use more than enough already. The categorical algebra notations are mostly from [27, 29, 28]. Other categorical notations have been taken from [1]. Acknowledgements In particular I would like to express my gratitude to my supervisor Prof. Dr. Hans-Eberhard Porst for his support, guidance and encouragement. General thanks are due to all other members of the research group KatMAT. I owe special thanks to Ina Bergen and Christoph Schubert for proofreading. I would also like to acknowledge a generous bursary from the University of Bremen.

viii Preface

Chapter 1 Introduction In this first chapter we introduce the fundamental concepts which are used throughout this thesis. We shortly sketch traditional universal algebra and finitary varieties in Birkhoff s sense. It is important to have at least an intuitive understanding of these concepts because they will be needed constantly. We continue by describing the more modern approach to varieties by Lawvere s theories [20] and the concept of varietal generators. These concepts are important for the description of equivalence between varieties. Knowing the varietal generators of a given variety V provides at least theoretical knowledge of all varieties equivalent to V. These results mostly go back to Lawvere [20], Isbell [15] and Linton [21]. Most important of all we introduce the n-th matrix power V [n] of a variety V for n N and the u-modification V(u) of V for an idempotent and invertible term u in V. They are the fundamental concepts of this thesis. The varieties equivalent to a given variety V are, up to concrete isomorphism, precisely the varieties V [n] (u) for some n N, n 1, and some idempotent and invertible term u for V [n]. We have taken most definitions and theorems in this chapter from [27, 28, 29]. Porst s paper [27], in which he makes the n-th matrix power construction of a variety and McKenzie s u-modification of a variety [25] more transparent, is the basis for this thesis. The proofs omitted here can be found in [27] (unless mentioned otherwise). We assume that the reader has an understanding of basic categorical notions. Especially equivalence is of course a fundamental concept for this thesis. Be aware that we use the term in the categorical sense, i.e. what is called categorical equivalence in the terminology of universal algebra. Equivalence in the sense of algebraical texts like [25, 24] is concrete equivalence in category theory. For more information on concrete equivalence we refer to [26]. In general we use categorical language as found in [27]. For more information on category theory we refer to [1, 12] or [22]. 1

2 Chapter 1. Introduction 1.1 Universal Algebra and Varieties The concept of varieties goes back to Birkhoff who wrote the first papers on universal algebra in the 1930s. Here we just recall the basic notions. Readers who are not familiar with universal algebra may refer to [24]. A very concise introduction can be found in [17]. Let A be a set and n a natural number. An n-ary operation on A is a map from A n to A. A signature is a pair (F, E) where F is a set of operations and E is a set of equations. An F -algebra is a pair (A, F ) where A is a set and F is the set which contains the A-interpretations f A : A n A of the operations in F. It is called an (F, E)-algebra if all equations from E are satisfied. An F -homomorphism ϕ : A B between two F -algebras A and B is a map such that for each n N and each f F the following diagram commutes: A n ϕ n B n f A f B A ϕ B Now we can define a variety: Definition 1.1 Let (F, E) be a signature. By Alg(F, E) we denote the category of all (F, E)-algebras and all F -homomorphisms. A category of the form Alg(F, E) is called a variety. Please note that a variety V is always a concrete category via the canonical functor : V Set which sends each algebra to its underlying set and each homomorphism to its underlying map. As an example we define M-acts and the variety Set, which we will need in Chapter 4. Definition 1.2 Let M be a monoid. An M-act is an (F, E)-algebra where the set of operations F contains one unary operation m for every m M and the set of equations E consists of the following equations: ex = x where e is the unary operation which corresponds to the neutral element of M. m(nx) = (mn)x for all unary operations m, n and mn where m, n and mn are the unary operations which correspond to the elements m, n and mn M By M Act we denote the variety of all M-acts Alg(F, E). M-acts can be defined without making the set F of operations explicit. In fact the following is the more common definition which is equivalent to the first:

1.1 Universal Algebra and Varieties 3 Definition 1.3 Let M be a monoid. An M-act is a set X with a map such that φ : M X X, φ(e, x) = x and φ(m, φ(n, x)) = φ(mn, x) for e, n, m M, e the neutral element, x X. We write ax instead of φ(a, x). By M Act we denote the variety of all M-acts. M-acts describe all unary varieties: Lemma 1.4 Each unary variety is concretely isomorphic to M Act for a suitable monoid M. Proof: Let V be a unary variety and let F be the set of operations and E the set of equations defining V. Let M be the free monoid generated by the elements of F subject to the equations of E, where the binary monoid-operation is the composition of the unary operations. Then each term in V corresponds to an element of M and vice versa. Now let A be any V-algebra. We can fully determine how M acts on a set by defining how the generating elements act on the set. If we define that the generating elements act on the underlying set of A like the unary operations which they are, it is obvious that the such defined M-act and A have the same underlying set and the same clone. Also each M-act is a unary algebra defined by the operations in F and satisfying the equations of E and is thus a V-algebra. Unary varieties can also be characterized by a categorical property: Lemma 1.5 A variety is unary if and only if its underlying functor preserves coproducts. Proof: Coproducts in Set are disjoint unions. Thus the underlying functor of a variety preserves coproducts if and only if the coproduct of a family (A i ) i I in the variety has the disjoint union of the family (A i ) i I as underlying set. It is easy to see that this is the case for unary varieties: Let V be a unary variety and let (A i ) i I be a family of V-algebras. The disjoint union (A i ) i I is a V- algebra if we define the unary operations to act as in the A i, i.e. for an element a ( (A i ) i I ) there exists a k I such that a A k and each operation f acts as f A k on a. This is indeed the coproduct of the family (A i ) i I. Let C be any V-algebra and (h i : A i C) i I a family of homomorphisms. Then there exists exactly one map h such that the following diagram commutes: µ j A i (Ai ) i I h i h C

4 Chapter 1. Introduction For each element a ( (A i ) i I ) there exists a k I such that a A k, thus we must define h(a) = h k (a). The map h is obviously a homomorphism, since for any operation f the image f(a) lies again in A k. Now let V be a variety which has an r-ary operation g for an r 2 which cannot be reduced to a unary one by equations. Let (A i ) i I be a family of V- algebras. Assume that the coproduct (A i ) i I has the disjoint union (A i ) i I as underlying set. Now let C be the term algebra over (A i ) i I modulo the congruence generated by all equations in V, i.e. the free algebra generated by (Ai ) i I. Let (h i : A i C) i I be the family of homomorphisms which inserts the generators into C. As above there exists exactly one map h such that the following diagram commutes: µ j A i (Ai ) i I h i h C There exist elements a k1,..., a kr of the coproduct (A i ) i I, where a kj A kj, such that g C (a k1,..., a kr ) is a term in C and not one of the generating elements from (A i ) i I. Otherwise there would be equations reducing g to a unary operation contrary to the assumption. But g (A i ) i I (a k1,..., a kr ) must be an element of (A i ) i I. Thus h(g (A i ) i I (a k1,..., a kr )) g C (h(a k1 ),..., h(a kr )). Hence h is not a homomorphism and therefore the coproduct (A i ) i I cannot have the disjoint union (A i ) i I as underlying set. While Birkhoff s definition of a variety is certainly the most intuitive one there is a different more modern approach via categories and functors. 1.2 Lawvere Theories Lawvere s approach to varieties was first described in his Ph.D. thesis [20]. It is fundamental for Porst s description of Morita equivalence. The concept of a Lawvere theory arises from extending the clone of a variety into a category such that clone composition becomes composition in this category. For further information about theories see e.g. [6]. An introduction can also be found in [27]. Definitions 1.6 A (Lawvere) theory T is a category of countably many objects T 0, T 1, T 2,... together with a distinguished family of morphisms (π n i : T n T 1 ) 1 i n

1.3 Varietal Generators 5 for each n N such that T n is an n-fold power of T 1. A theory morphism is a functor Φ : S T between theories which preserves the distinguished product families (π n i : T n T 1 ) 1 i n. A T-model (or T-algebra) is a functor from T into the category Set of sets and mappings which preserves all finite products. For a given Lawvere theory T, the T-models and natural transformations between them form the category ModT. Hence the category of all T-models is a full subcategory of the category of all functors from T to Set. The category ModT is a concrete category. It has a canonical underlying functor U T : ModT Set which is evaluation at T 1 : U T (H µ K) = H(T 1 ) µ T 1 K(T1 ). Definition 1.7 Lawvere theories T and S are called Morita equivalent provided that ModT and ModS are equivalent categories. A paradigmatic way to construct a theory is the following: Let G be an object in a category V which admits all its finite copowers. If we take the dual of the full subcategory of V spanned by the chosen n-fold copowers ng of G for each n N we have a Lawvere theory. We call this theory the Lawvere theory generated by G and we denote it by Th V (G). Let V be a variety and G = F 1 the freely generated algebra on one generator. Then the theory Th V (F 1) is called the theory of V and is denoted by ThV. The connection between Birkhoff s and Lawvere s approaches to varieties is given by the fact that every variety V is concretely equivalent to ModThV and conversely. 1.3 Varietal Generators Varietal generators are crucial for equivalence between varieties. Knowing the varietal generators of a variety V means at least theoretical knowledge of all varieties equivalent to V as Theorem 1.10 shows. However it seems that there is no one-to-one correspondence between varietal generators in V and and varieties W equivalent to V. The results in this section mostly go back to Lawvere [20], Isbell [15] and Linton [21] except for Theorem 1.12 which is due to Porst [27]. Definition 1.8 An object G in a cocomplete category is called a varietal generator if G is (i) a regular generator,

6 Chapter 1. Introduction (ii) regularly projective, (iii) finitely presentable. Remark 1.9 In this thesis we just need varietal generators in varieties. An algebra G in a variety is a varietal generator provided (i) G is a generator, i.e. each object is a homomorphic image of some copower of G, (ii) G is projective (with respect to surjective homomorphisms), (iii) G is presentable by finitely many generators and equations. In each variety the finitely generated free algebras F n for n N are varietal generators. They may serve as the paradigmatic example here. Combining the concepts of theories and varietal generators allows the following characterization of equivalent varieties. Theorem 1.10 For varieties V and W the following are equivalent: (i) W is equivalent to V. (ii) Th W (F 1) = Th V (G) for a varietal generator G in V. For a proof see [6]. With a suitable underlying functor the equivalence even becomes concrete. This will be of importance for finding the right underlying set when trying to construct equivalent varieties. Theorem 1.11 Let Φ : W V be an equivalence between varieties. Then the following diagram commutes Φ W V Set hom(g, ) where G is the varietal generator from Theorem 1.10, i.e. equivalence between (W, ) and (V, hom(g, )). Φ is a concrete For a proof see [6]. We have seen that the knowledge of all varietal generators is fundamental. The following theorem (cf. Porst [27]) characterizes the varietal generators as retracts of the finitely generated free algebras F n. The morphism u in this theorem delivers the operation u needed for the u-modification in the next section. We use the same notation for morphisms and operations.

1.4 Matrix Power and u-modification 7 Theorem 1.12 For an object G in V the following are equivalent: (i) G is a varietal generator. (ii) There are retractions G s F n r G = id G and F 1 p mg d F 1 = id F 1 with n, m natural numbers. (iii) For some natural number n there exists a morphism r : F n G such that (r, G) is a coequalizer of (u, id F n ), where u is an idempotent endomorphism of F n for which there exists for some natural number m a retraction F 1 p d F (nm) F 1 = id F 1 such that d (mu) p = id F 1. 1.4 Matrix Power and u-modification In this section we present two constructions: the n-th matrix power A [n] of an algebra A for natural numbers n 2 and the u-modification A(u) of an algebra A for an idempotent and invertible term u. They are the fundamental concepts of this thesis. In this section we just give the definitions as they can be found in [27] and the main theorem from [25] which characterizes all varieties equivalent to a given one. Definition 1.13 Let A be an algebra in a given variety V and n N, n 2. The n-th matrix power A [n] of A is the following algebra: 1. The underlying set of A [n] is A n. 2. The r-ary operations A [n] r = A nr A n = A [n] are those maps f whose composition with the projections yield nr-ary V-operations f i A nr f=(f 1,...,f n) A n f i π i A By V [n] we denote the category of all algebras isomorphic to some A [n] for A V and all homomorphisms between them.

8 Chapter 1. Introduction V [n] is in fact a variety (cf. [27]). In Chapter 2 we will present two new ways to describe the n-th matrix power V [n] of a variety V for n N. Definition 1.14 An idempotent morphism u : A A in any category V admitting finite products of A is called invertible if there exists some m N and morphisms p : A m A and d : A A m such that A d A m u m A m p A = 1 A The idempotent endomorphisms u : F n F n obtained in Theorem 1.12 (iii) correspond exactly to the idempotent and invertible terms u needed for the u-modification defined in the following definition. For more details and proofs please refer to [27]. Definition 1.15 Let A be an algebra in a given variety V and u be an idempotent and invertible term in V. Then the u-modification A(u) of A is the following algebra: 1. The underlying set of A(u) is u A [ A ]. 2. The r-ary operations (u A [ A ]) r u A [ A ] are those maps fu A which arise as restrictions to (u A [ A ]) r of maps A r f A u A A u[ A ] where f A is the A-interpretation of a operation f in V. By V(u) we denote the category of all algebras isomorphic to some A(u) for A V and all homomorphisms between them. V(u) is in fact a variety (cf. [27]). In Chapter 4 we examine the u-modification of unary varieties. We show a new way to construct the u-modification for a special class of unary varieties. Remark 1.16 As we have seen in Theorem 1.12 the varietal generators G in a variety V correspond to retractions G s F n r G = id G. In the case of r = id F n the equivalent variety determined by G is V [n]. The nontrivial retractions G s F 1 r G = id G correspond to V(u) where u = sr(1) F 1. transparent in [27]. These correspondences become

1.4 Matrix Power and u-modification 9 The following theorem is one of the main results of McKenzie s paper [25]. It characterizes all varieties equivalent to a given one via the construction of the matrix power and the u-modification. We present it as Porst in [27] where you can also find a proof which makes use of categorical algebra and clarifies the necessary constructions much better than McKenzie s original proof. Theorem 1.17 The varieties equivalent to a given variety V are, up to concrete isomorphism, precisely the varieties V [n] (u) for some n N, n 1, and some idempotent and invertible term u for V [n].

10 Chapter 1. Introduction

Chapter 2 The Matrix Power As we have seen in the introduction we are already able to construct the matrix power for arbitrary varieties. But this construction is very abstract and difficult to apply in concrete cases since we need the whole clone for its description. Here we provide a different approach, a characterization of the matrix power via a few operations and equations. We give two similar descriptions. The first one, V (n), is more elegant, since it consists of adding only one binary operation and three basically different equations to the already existing operations and equations. The real number of equations added depends on the number of operations the variety we start with has: one of the equations is in fact a family of equations, one equation for every original operation. The second, V {n}, needs 2n 1 operations for the n-th matrix power (n N), which are basically just two families of operations, and we have more equations as well. However it turns out to be a lot easier to work with when we construct the u-modification for unary varieties in Chapter 4, since the binary operations are less complex. The matrix power is sufficient to describe all Morita equivalent varieties for those varieties in which each finitely generated projective algebra is free, i.e. where the finitely generated free algebras are all varietal generators. This is for example the case in the category of all (Abelian) groups or in the category of all sets (cf. [27]). The results in this chapter are a generalization of the description of varieties equivalent to the variety Set of sets and maps which we present in the next chapter. Thus the next chapter can be read first, as already mentioned in the preface, depending on whether one prefers the theory or the example first. Both chapters can be read on their own. In the next chapter we develop several different descriptions of the matrix powers of Set. They are generalizations of examples of n-th matrix powers which were published earlier for special values of n (Example 1.2 in [5] and Example 2 in [25]). Each description is a refinement of the previous and might help to understand how the operations work. The last description in the next chapter, n-set 4, is due to results by Marshall Saade [31]. He did not work on the same problem as far as I can discern from his paper 11

12 Chapter 2. The Matrix Power but his operations work very well for our problem. All the motivation given in his paper is the following sentence: We describe in this paper, for each integer n > 2, a curious variety V of groupoids originally introduced in [10] by Evans for the purpose of describing the spectrum.. The binary operation we use for our variety V (n) is from his curious variety. The variety V {n} which is our second description of the n-th matrix power of a variety V is a generalization of the the first construction n-set 1 in the next chapter. Some ideas for these generalizations are due to Fajtlowicz. In his paper on n- dimensional dice [11] he finds necessary and sufficient conditions for an algebra to have an n-th Cartesian power as underlying set. Especially the idea for equation (i), which is needed in both descriptions, comes from his paper. 2.1 The Variety V (n) Definition 2.1 Let V be a variety which is defined by a set F of operations and a set E of equations. For each natural number n 2 let V (n) be the variety defined by adding one binary operation, which we do not denote by a symbol, to F and adding the following equations to E: (i) (ii) (iii) f(x 1,..., x k )f(y 1,..., y k ) = f(x 1 y 1,..., x k y k ) for all f F x 1... x n y = x 1... x n z (x 1... x n 1 x)(x n... x 2n 3 x)... (x p 2 x p 1 x)(x p x)(xy) = x n(n 1) where p = 2 Here x 1 x 2... x m denotes the product (x 1 (x 2 (... (x m 2 (x m 1 x m ))... ))); we leave out the brackets wherever possible. By x k we denote the product xx }{{... x}. k times Remark 2.2 We can turn any V-algebra A into a V (n) -algebra A (n) by taking its Cartesian power A n with operations acting coordinatewise and defining the binary operation by (x 1,..., x n )(y 1,..., y n ) (x n, y 1, y 2,..., y n 1 ). It is straightforward to check that this is indeed a V (n) -algebra. Definition 2.3 Through repeated application of the binary operation on a single variable, we define new unary operations D i by D i (x) := (x n+1 i ) n+1, i = 1,..., n. Remark 2.4 In A (n) it is straightforward to see that the D i operate as (x 1,..., x n ) (x i,..., x i ).

2.1 The Variety V (n) 13 x n+1 i has the i-th coordinate of x as its last coordinate. Applying the binary operation on x n+1 i n + 1 times, copies the i-th coordinate of x into every coordinate. Lemma 2.5 The following equations can be derived from equations (ii) and (iii) in V (n) : (iv) (x 1... x n y)z = x n z (v) (x 1... x n 1 yx n ) n+1 = y n+1 (vi) x n+1 = ((xy) n ) n+1 (vii) (y n k ) n+1 = ((xy n k 1 ) n+1 for k = 0, 1,..., n 2 (viii) (x n ) n+1 (x n 1 ) n+1... (x 2 ) n+1 x n+1 = x (ix) (x n+1 ) 2 = x n+1 (x) ((x k+1 ) n k ) n+1 = ((x n 1 )(x n 1 2 )... (x n t t+1 )... (x2 n 1)(x 2 n)) n k ) n+1 for k = 0, 1,..., n 1 Proof: The proof consists of straightforward calculations and can be found in [31]. 2.1.1 V (n) is the Matrix Power of V Definition 2.6 For A V (n) we define the set D A := {x A xx = x} A. A simple calculation shows that D A is closed under all operations f F due to equation (i). Thus, it is even a V-algebra. It follows from equation (ix) that the D i give maps from A to D A. Lemma 2.7 Let A be a V (n) -algebra. Then A is isomorphic to (D A ) (n). Proof: Define maps ϕ : A (D A ) n by: x (D 1 x,..., D n x) = ((x n ) n+1, (x n 1 ) n+1,..., x n+1 ) and ψ : (D A ) n A by: (x 1,..., x n ) x 1 x 2... x n Then ϕ ψ = id (DA ) n : ϕ ψ(x 1,..., x n ) = (((x 1 x 2... x n ) n ) n+1, ((x 1 x 2... x n ) n 1 ) n+1,..., ((x 1 x 2... x n ) 1 ) n+1 ) (x) = ((x 1 ) n+1,..., (x n ) n+1 ) (since all x i D A we have x t+1 = (x t+1 ) n t ) = (x 1,..., x n ) since all x i D A

14 Chapter 2. The Matrix Power as well as ψ ϕ = id A : ψ ϕ(x) = (x n ) n+1 (x n 1 ) n+1... (x 2 ) n+1 n+1 (viii) x = x Hence A = (D A ) n in Set. If we turn (D A ) n into a V (n) -Algebra D (n) A according to Remark 2.2 then ϕ is a homomorphism: ϕ(xy) = (((xy) n ) n+1, ((xy) n 1 ) n+1,..., (xy) n+1 ) (vi)+(vii) = ((x) n+1, ((y) n ) n+1,..., ((y) 2 ) n+1 ) = ϕ(x)ϕ(y) Let f F ϕ(f(x 1,..., x k )) = (D 1 f(x 1,..., x k ),..., D n f(x 1,..., x k )) (i) = (f(d 1 x 1,..., D 1 x k ),..., f(d n x 1,..., D n x k )) = f n ((D 1 x 1,..., D n x 1 ),..., (D 1 x k,..., D n x k )) = f n (ϕ(x 1 ),..., ϕ(x n )) Thus ϕ is a V (n) -isomorphism. Theorem 2.8 For a given variety V and any natural number n 2 the variety V (n) and the matrix power V [n] are concretely isomorphic. Proof: The lemma above shows that every algebra in V (n) is isomorphic to an algebra of the form A (n) for an algebra A in V. Every algebra in V [n] is isomorphic to an algebra of the form A [n] for an algebra A in V by definition. Since A (n) and A [n] have the same underlying set it remains to show that they have the same clone. The operations of the Cartesian power A n which are the operations from A just acting coordinatewise are obviously contained in the clone of A [n] and the binary operation as defined in Remark 2.2 can be easily derived from the appropriate projections and thus also lies in the clone of A [n]. Now let f be an r-ary A [n] operation. By definition of the matrix power, composing f with the projections gives nr-ary A operations f i = π i f for 1 i n. A nr f=(f 1,...,f n) A n f i π i A

2.2 An Alternative Description of the Matrix Power: V {n} 15 The operations f (n) i : A nnr A n which operate like the f i on the single coordinates are contained in the clone of A (n). For a given nr-tuple x = (x 1,..., x r ) with x i A n we introduce the following notation: Obviously thus x = (x 11,..., x }{{ 11, x } 12,..., x 12,..., x }{{} rn,..., x rn ) A nnr }{{} n n n f (n) i ( x) = (f i (x),..., f i (x)), f(x) = f (n) 1 ( x)f (n) 2 ( x)... f n (n) ( x). Therefore f is an element of the clone of A (n) 2.2 An Alternative Description of the Matrix Power: V {n} Now we introduce a second way to construct the matrix power. This approach needs more operations and equations but is better suited when we construct the u-modification of M Act [n] in Chapter 4. Definition 2.9 Let V be a variety which is defined by a set F of operations and a set E of equations. For each natural number n 2 let V {n} be the variety defined by adding unary operations D i for i = 1,..., n and binary operations j for j = 1,..., n 1 to F and adding the following equations to E: (i) D i (f(x 1,..., x k )) = f(d i x 1,..., D i x k ) f F, i {1,..., n} (ii) D j D i x = D i x i, j {1,..., n} (iii) D 1 x 1 D 2 x 2 D 3 x 3 n 1 D n x = x (iv) D i (x 1 1 n 1 x n ) = D i x i i {1,..., n} { D i x for i = j (v) D i (x j y) = i {1,..., n}, j {1,..., n 1} D i y for i j x 1 x 2 x m denotes the product (x 1 (x 2 ( (x m 2 (x m 1 x m ))... ))), again we leave out the brackets wherever possible. Remark 2.10 The D i in this section are the same operations as the derived operations D i in the previous section. Thus we use the same notation. The binary operations i are very similar to the binary operation in V (n). They basically just lack the shifting of coordinates which makes them easier to handle.

16 Chapter 2. The Matrix Power Remark 2.11 We can turn any V-algebra A into a V {n} -algebra A {n} by taking its Cartesian power A n with operations acting coordinatewise and defining the unary operations D i by D i (x 1,..., x n ) = (x i..., x i ) and the binary operations i by (x 1,..., x n ) i (y 1,..., y n ) = (y 1,..., y i 1, x i, y i+1,..., y n ). It is straightforward to check that this is indeed a V {n} -algebra. 2.2.1 V {n} is the Matrix Power of V Definition 2.12 For A V {n} we define the sets D A i := {x A D i x = x} A for i = {1,..., n}. From equation (ii) we get D A := D A 1 = DA 2 = = DA n. D A is closed under all operations f F due to equation (i), and thus it is even a V-algebra. The D i give maps from A to D A. Lemma 2.13 Let A be a V {n} -algebra. Then A is isomorphic to (D A ) {n}. Proof: Define maps ϕ : A (D A ) n by x (D 1 x,..., D n x) and ψ : (D A ) n A by (x 1,..., x n ) x 1 1 x 2 2 n 1 x n. Then ψ ϕ = id A : ψ ϕ(x) = D 1 x 1 D 2 x 2 D 3 x 3 n 1 D n x (iii) = x as well as ϕ ψ = id (D A ) n : ϕ ψ(x 1,..., x n ) = (D 1 (x 1 1 n 1 x n ),..., D n (x 1 1 n 1 x n )) (iv) = (D 1 x 1,..., D n x n ) x i D A = (x 1,..., x n ). Hence A = D A in Set.

2.2 An Alternative Description of the Matrix Power: V {n} 17 If we turn (D A ) n into a V {n} -algebra D {n} A even a homomorphism: according to Remark 2.11 then ϕ is ϕ(d i x) = (D 1 D i x,..., D n D i x) (ii) = (D i x,..., D i x) = D i (D 1 x,..., D n x) = D i ϕ(x) ϕ(x i y) = (D 1 (x i y),..., D n (x i y)) (v) = (D 1 y,..., D i 1 y, D i x, D i+1 y,..., D n y) = (D 1 x,..., D n x) i (D 1 y,..., D n y) = ϕ(x) i ϕ(y) Let f F ϕ(f(x 1,..., x k )) = (D 1 f(x 1,..., x k ),..., D n f(x 1,..., x k )) (i) = (f(d 1 x 1,..., D 1 x k ),..., f(d n x 1,..., D n x k )) = f n ((D 1 x 1,..., D n x 1 ),..., (D 1 x k,..., D n x k )) = f n (ϕ(x 1 ),..., ϕ(x n )) Thus ϕ is a V {n} -isomorphism. Theorem 2.14 For a given variety V and any natural number n 2 the variety V {n} and the matrix power V [n] are concretely isomorphic. Proof: The lemma above shows that every algebra in V {n} is isomorphic to an algebra of the form A {n} for an algebra A in V. Every algebra in V [n] is isomorphic to an algebra of the form A [n] for an algebra A in V by definition. Since A (n) and A [n] have the same underlying set it remains to show that they have the same clone. This can be done in the same way as in the case of V (n). The operations of the Cartesian power A n which are the operations from A just acting coordinatewise are obviously contained in the clone of A [n] and the unary operation D i as well as the binary operations i as defined in Remark 2.11 can be easily derived from the appropriate projections and thus also lie in the clone of A [n]. Now let f be an r-ary A [n] operation. By definition of the matrix power composing f with the projections gives nr-ary A operations f i = π i f for 1 i n. A nr f=(f 1,...,f n) A n f i π i A

18 Chapter 2. The Matrix Power The operations f {n} i : A nnr A n which operate like the f i on the single coordinates are contained in the clone of A {n}. For a given nr-tuple x = (x 1,..., x r ) with x i A n we introduce the following notation x = (x 11,..., x }{{ 11, x } 12,..., x 12,..., x }{{} rn,..., x rn ) A nnr }{{} n n n Obviously f {n} i ( x) = (f i (x),..., f i (x)) thus f(x) = f {n} 1 ( x) 1 f {n} 2 ( x) n 1 f n {n} ( x). Therefore f is an element of the clone of A {n}.

Chapter 3 The Matrix Powers of Set This chapter provides four different descriptions of the matrix powers of Set, the variety of sets and maps. Each one is described independently, but their order shows how I initially got to understand and refine the operations needed to describe the matrix powers. This chapter can be read on its own to get a better understanding of the operations before reading Chapter 2 or it can be treated as an example for the theory in Chapter 2. The varieties equivalent to Set are known: the nonempty finite sets are the varietal generators of Set, i.e. up to isomorphism the natural numbers without zero are all varietal generators. The corresponding Lawvere theories are therefore easy to describe. Thus every variety V equivalent to Set is concretely equivalent to (Set, hom(n, )) for some natural number n 0 (see Theorem 1.11). But this only gives us a very abstract description. The aim of this chapter is to find a small number of operations and equations which define the varieties in question. When first starting to work on this problem our interest was sparked by two existing solutions for the special case of the second matrix power. One from Börger (Example 1.2 in [5]) and one from McKenzie (Example 2 in [25]). For the general case we present four solutions. The first two, n-set 1 and n-set 2, are direct generalizations of Börger s example whereas the third, n-set 3, uses McKenzie s operations. The fourth description, n-set 4, is based on a variety constructed by Saade [31]. It seems that Saade did not realize the significance of his construction in this context because he only talks about describing a curious variety. The description n-set 4 is an example of the first characterization V (n) in the previous chapter. Many of the proofs are analogous in the different sections of this chapter. Nevertheless, they are all given explicitly, so that it is not necessary to read every section. In this chapter, we use the symbols of some operations repeatedly. They can represent different operations in different sections. But they are always based on the same principle, especially the operations D i. We refrain from distinguishing between them via indices since the differences should be clear and we have to use more than enough indices already. 19

20 Chapter 3. The Matrix Powers of Set Remark 3.1 Due to their concrete equivalence to (Set, hom(n, )) for some natural number n 0, the varieties equivalent to Set contain empty sets and thus they cannot contain nullary operations (constants). It is impossible to describe those varieties only by unary operations since otherwise unions of subalgebras would always be subalgebras. But that is not true in general because of the cardinality restriction induced by hom(n, ) as forgetful functor. Only sets that are an n-th Cartesian power for n 2 appear as underlying sets. 3.1 n-set 1 Definition 3.2 For each natural number n 2 let n-set 1 be the variety defined by the following operations and equations: Unary operations: D i for i = 1,..., n Binary operations: j for j = 1,..., n 1 Equations: α ij : D j D i = D i i, j {1,..., n} with i j { D i x for i j β ij : D i (x j y) = i {1,..., n}, j {1,..., n 1} D i y for i > j γ : D 1 x 1 D 2 x 2 D 3 x 3 n 1 D n x = x Remark 3.3 We use the following conventions for greater clearness of representation: the D i have higher priority than the j and x 1 1 x 2 2 n 1 x n := (... ((x 1 1 x 2 ) 2 x 3 ) 3... ) n 1 x n. Note that this implies that in this section the terms have to be read from the left to the right opposite to Chapter 2 where equations had to be read from the right to the left. Remark 3.4 We can turn a set of the form X n into an n-set 1 -algebra by defining D Xn i : X n X n by (x 1,..., x n ) (x i,..., x i ) for all i = 1,..., n and Xn j : (X n ) 2 X n by (x, y) (x 1,..., x j, y j+1,..., y n ) for all j = 1,..., n 1. required equations. The such defined operations obviously satisfy the Lemma 3.5 If D i x = D i y for all i = 1,..., n then x = y.

3.1 n-set 1 21 Proof: Let D i x = D i y for all i = 1,..., n. Then x = γ D 1 x 1 n 1 D n x = D 1 y 1 n 1 D n y = γ y. Lemma 3.6 In n-set 1 the following equations are satisfied: α ii : D i D i = D i i {1,..., n} ˆβ i : D i (x 1 1 x 2 2 n 1 x n ) = D i x i i {1,..., n} δ j : (x 1 1 n 1 x n ) j (y 1 1 n 1 y n ) = x 1 1 j 1 x j j y j+1 j+1 n 1 y n j {1,..., n 1} ε j : x j x = x j {1,..., n 1} Proof: Repeated application of β ij gives ˆβ i. For all i = 1,..., n: D i D i x ˆβ i = Di (D 1 x 1 D 2 x 2 D 3 x 3 n 1 D n x) = γ D i x which proves α ii. Ad δ j : Let j = 1,..., n 1 Case 1: For all i j Case 2: For all i > j D i ((x 1 1 n 1 x n ) j (y 1 1 n 1 y n )) β ij = D i (x 1 1 n 1 x n ) ˆβ i = Di x i ˆβ i = Di (x 1 1 j 1 x j j y j+1 j+1 n 1 y n ) D i ((x 1 1 n 1 x n ) j (y 1 1 n 1 y n )) β ij = D i (y 1 1 n 1 y n ) ˆβ i = Di y i ˆβ i = Di (x 1 1 j 1 x j j y j+1 j+1 n 1 y n ) Cases 1 and 2 give δ j by Lemma 3.5. For all i = 1,..., n we have D i (x j x) = D i x according to β ij. Hence Lemma 3.5 gives ε j. Remark 3.7 For n = 2 we get Example 1.2 from Börger in [5] where p = D 1, q = D 2 and = 1. The equations are exactly the same.

22 Chapter 3. The Matrix Powers of Set Remark 3.8 For n = 2 we can derive an associative law for the binary operation by applying δ: (x y) z ε = ((x y) (z z)) δ = x z δ = ((x x) (y z)) ε = x (y z). Hence for n = 2 we get semigroups with two additional unary operations (cf. Example 2 in [25]). 3.1.1 The Equivalence of n-set 1 to Set Definition 3.9 For A n-set 1 we define the set D A := {x A D 1 x = x} A. Obviously D i x = x for all x D A and all i = 1,..., n due to equation α ij. Lemma 3.10 Let A be an n-set 1 -algebra. Then A is isomorphic to (D A ) n. Proof: Define maps ϕ : A (D A ) n by x (D 1 x,..., D n x) and ψ : (D A ) n A by (x 1,..., x n ) x 1 1 n 1 x n. Then ϕ ψ = id (DA ) n : ϕ ψ(x 1,..., x n ) = (D 1 (x 1 1 n 1 x n ),..., D n (x 1 1 n 1 x n )) as well as ψ ϕ = id A : ˆβ i = (D1 x 1,..., D n x n ) = (x 1,..., x n ) since x i D A ψ ϕ(x) = D 1 x 1 D 2 x 2 D 3 x 3 n 1 D n x γ = x. Hence A = (D A ) n in Set. If we turn (D A ) n into an n-set 1 -algebra according to Remark 3.4 then ϕ is a homomorphism since for all i = 1,..., n ϕ(d i x) = (D 1 D i x,..., D n D i x) α ij = (D i x,..., D i x) = D i (D 1 x,..., D n x) = D i ϕ(x)

3.1 n-set 1 23 and for all j = 1,..., n 1 ϕ(x j y) = (D 1 (x j y),..., D n (x j y)) β ij = (D 1 x,..., D j x, D j+1 y,..., D n y) δ j = (D 1 x,..., D n x) j (D 1 y,..., D n y) = ϕ(x) j ϕ(y). Thus ϕ is an n-set 1 -isomorphism. Proposition 3.11 (n-set 1, ) is concretely equivalent to (Set, hom(n, )). Proof: We define a functor F : Set n-set 1 which maps each map f : X Y to the n-set 1 -homomorphism f n : X n Y n, where X n and Y n are n-set 1 - algebras as defined in Remark 3.4. It is clear that F is faithful. Let g : X n Y n be an n-set 1 -homomorphism. Then for (x,..., x) D X n all coordinates of g(x,..., x) are the same since g(x,..., x) = gd i (x,..., x) = D i g(x,..., x) for all i = 1,..., n. Thus we can define a map f : X Y which maps each x X to the first coordinate of g(x,..., x). Obviously f n equals g on D X n. But this implies f n = g on the whole of X n by applying equation γ. Therefore F is full. F is isomorphism dense by Lemma 3.10. Thus F is an equivalence. Concreteness is obvious. Remark 3.12 Here is an alternative proof of the equivalence between n-set 1 and Set. We construct a functor G which is an equivalence inverse to the functor F in the previous proof. Proof: We construct a functor G : n-set 1 Set which assigns to each n-set 1 -morphism f : A B the map f D B D A : D A D B (Imf DA D B, since for each x D A we have f(x) = f(d 1 x) = D 1 f(x) D B ). Each map g : D A D B yields a unique homomorphism ḡ : A B in n-set 1 defined by x g(d 1 x) 1... n 1 g(d n x). (i) ḡ : A B is a homomorphism: (a) For all i = 1,..., n and all x A we get ḡ(d i x) = D i ḡ(x), since ḡ(d i x) = g(d 1 D i x) 1 n 1 g(d n D i x) α ij = g(d i x) 1 n 1 g(d i x) ε j = g(d i x) = D i g(d i x) since g(d i x) D B ˆβ i = Di (g(d 1 x) 1 n 1 g(d n x)) = D i ḡ(x)

24 Chapter 3. The Matrix Powers of Set (b) For j = 1,..., n 1 we have ḡ(x j y) = ḡ(x) j ḡ(y). Define (x/y) ij by (x/y) ij = x for i j and (x/y) ij = y for i > j. Then D i ḡ(x j y) = ḡ(d i (x j y)) = ḡ(d i ((x/y) ij )) = D i ḡ(x/y) ij ) = D i (ḡ(x) j ḡ(y)) for all i = 1,..., n. Hence ḡ(x j y) = ḡ(x) j ḡ(y) by Lemma 3.5. (ii) ḡ D B D A = g since for each x D A we have ḡ(x) = ḡ(d 1 x) (i)(a) = g(d 1 x) = g(x). (iii) Let f : A B be an n-set 1 -homomorphism and let g := G(f) = f D B D A. Then ḡ = f since f(x) γ = f(d 1 x 1 n 1 D n x) = f(d 1 x) 1 n 1 f(d n x) = g(d 1 x) 1 n 1 g(d n x) = ḡ(x) Thus G is fully faithful. G is isomorphism dense. Let X be a set and let A = X n an n-set 1 -algebra with operations defined according to Remark 3.4. Then D A = {((x,..., x) A x X} = X. Therefore G is an equivalence. 3.2 n-set 2 Instead of n 1 binary equations we can alternatively use just one n-ary operation. It is left to the reader to decide whether it is preferable to have a small (and fixed) arity or a smaller number of operations. This case is analogous to the previous n-set 1 case except for minor details. Definition 3.13 For each natural number n 2 let n-set 2 be the variety defined by the following operations and equations: Unary operations: D i for i = 1,..., n n-ary operation: Equations: α ij : D j D i = D i i, j {1,..., n} with i j β i : γ : D i (x 1,..., x n ) = D i x i (D 1 x,..., D n x) = x

3.2 n-set 2 25 Remark 3.14 We can turn a set of the form X n into an n-set 2 -algebra by defining D Xn i : X n X n by (x 1,..., x n ) (x i,..., x i ) for all i = 1,..., n and Xn : (X n ) n X n by ( x 1,..., x n ) (x 11,..., x nn ). The such defined operations obviously satisfy the required equations. Lemma 3.15 If D i x = D i y for all i = 1,..., n then x = y. Proof: Let D i x = D i y for all i = 1,..., n. Then x = γ (D 1 x,..., D n x) = (D 1 y,..., D n y) = γ y. Lemma 3.16 In n-set 2 the following equations are satisfied for i = 1,..., n: α ii : D i D i = D i Proof: For all i = 1,..., n: D i D i x β i = D i (D 1 x,..., D n x) = γ D i x. Remark 3.17 For n = 2 we get Example 1.2 from Börger in [5] again, where p = D 1, q = D 2 and =. The equations are the same. 3.2.1 The Equivalence of n-set 2 to Set Definition 3.18 For A n-set 2 we define the set D A := {x A D 1 x = x} A. Obviously we have D i x = x for all x D A and all i = 1,..., n. Lemma 3.19 Let A be an n-set 2 -algebra. Then A is isomorphic to (D A ) n. Proof: Define maps ϕ : A (D A ) n by x (D 1 x,..., D n x) and ψ : (D A ) n A by (x 1,..., x n ) (x 1,..., x n ).

26 Chapter 3. The Matrix Powers of Set Then ϕ ψ = id (DA ) n : ϕ ψ(x 1,..., x n ) = (D 1 (x 1,..., x n ),..., D n (x 1,..., x n )) as well as ψ ϕ = id A : β i = (D1 x 1,..., D n x n ) ψ ϕ(x) = (D 1 x,..., D n x) γ = x = (x 1,..., x n ) since x i D A Hence A = (D A ) n in Set. If we turn (D A ) n into an n-set 2 -algebra according to Remark 3.14 then ϕ is a homomorphism. Because for all i = 1,..., n ϕ(d i x) = (D 1 D i x,..., D n D i x) α ij = (D i x,..., D i x) = D i (D 1 x,..., D n x) = D i ϕ(x) and for all j = 1,..., n 1 ϕ( (x 1,..., x n )) = (D 1 (x 1,..., x n ),..., D n (x 1,..., x n )) β i = (D1 x 1,..., D n x n ) = ((D 1 x 1,..., D n x 1 ),..., (D 1 x n,..., D n x n )) = (ϕ(x 1 ),..., ϕ(x n )) Thus ϕ is even an n-set 2 -isomorphism. Proposition 3.20 (n-set 2, ) is concretely equivalent to (Set, hom(n, )). Proof: We define a functor F : Set n-set 2 which maps each map f : X Y to the n-set 2 -homomorphism f n : X n Y n, where X n and Y n are defined as in Remark 3.14. It is clear that F is faithful. Let g : X n Y n be an n-set 2 -homomorphism. Then for (x,..., x) D X n coordinates of g(x,..., x) are the same since we have all g(x,..., x) = gd i (x,..., x) = D i g(x,..., x) for all i = 1,..., n. Thus we can define a map f : X Y which maps each x X to the first coordinate of g(x,..., x). Obviously f n equals g on D X n. But this implies f n = g on the whole of X n by applying equation γ. Therefore F is full. F is isomorphism dense by Lemma 3.19. Thus F is an equivalence. Concreteness is obvious.