A Dissertation entitled
The Steenrod Algebra is a Prime Ring and the Krull Dimensions of the Steenrod Algebra
by Robert P. Stephens
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Mathematics
Dr. Charles J. Odenthal, Committee Chair
Dr. Paul R. Hewitt, Committee Member
Dr. John H. Palmieri, Committee Member
Dr. Martin R. Pettet, Committee Member
Dr. Patricia R. Komuniecki, Dean College of Graduate Studies
The University of Toledo August 2011 Copyright 2011, Robert P. Stephens
This document is copyrighted material. Under copyright law, no parts of this document may be reproduced without the expressed permission of the author. An Abstract of The Steenrod Algebra is a Prime Ring and the Krull Dimensions of the Steenrod Algebra by Robert P. Stephens
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Mathematics The University of Toledo August 2011
Kashkarev has shown that the mod 2 Steenrod algebra is a prime ring. For any odd prime p, we prove that the mod p Steenrod algebra is also a prime ring. In sequel, for any prime p, we show that the mod p Steenrod algebra (a local ring with nil maximal ideal) has infinite little Krull dimension. This contrasts sharply with the case of a commutative (or noetherian) local ring with nil maximal ideal which must have little Krull dimension equal to 0. Also, we show that the Steenrod algebra has no Krull dimension, classical Krull dimension, or Gabriel dimension.
iii Acknowledgments
I would like to thank my advisor Charles Odenthal for his guidance and encour- agement. His advice helped me focus my thoughts and ideas into sentences that were more precise and easier to understand.
I also thank Rose Kneen for her friendship and help throughout my time here at the University of Toledo.
I thank the anonymous reviewer for all his or her suggestions and comments.
Lots of appreciation goes to my Mom and Dad, Mary Ann and Joe Stephens, and my sister Katie and her family, Dylan and Kenneth. My family helped kept me afloat through tough times.
Finally, I’d like to think Jeffery Davis. His sense of humor could always cut through any bad mood and make me smile, which in turn helped keep me sane.
iv Contents
Abstract iii
Acknowledgments iv
Contents v
1 History and Conventions 2
1.1 Background ...... 2
1.1.1 Cohomology ...... 2
1.2 The Steenrod Squares and Powers ...... 4
1.2.1 Axiomatic Description of the Steenrod Squares and Powers . .5
1.2.2 Algebraic Description of the Steenrod Algebras ...... 8
1.2.3 Early Uses of the Steenrod Algebra ...... 10
1.3 Prime Ideals and Krull Dimensions ...... 11
1.4 Definitions and Conventions ...... 17
2 The Steenrod algebra is a Prime Ring 22
2.1 Preliminary Lemmas ...... 22
2.2 Main Primality Theorem ...... 26
2.3 Miscellaneous ...... 33
3 The Krull Dimensions of the Steenrod Algebra 36
v 3.1 Preliminary Results ...... 37
3.2 The Little Krull Dimension ...... 38
3.2.1 The Ideals Dj ...... 38 3.2.2 The map ψ ...... 51
3.2.3 The Hopf Algebra Approach ...... 54
3.3 Other Dimensions ...... 55
3.4 Future Work ...... 57
3.5 The case p =2 ...... 58
References 59
A Construction of the Steenrod Squares and Powers 64
B Hopf Algebras 69
vi Introduction
The main topic of this work is the Steenrod algebra, which plays an important role in the study of algebraic topology, but this treatise mostly concentrates on some of the algebraic properties of the Steenrod algebra. Namely, we show that, for an odd prime p, then Steenrod algebra is a prime ring. This work is an extension of the work by Kashkarev in [25], who showed that the mod 2 Steenrod algebra was a prime ring. In sequel, we show, for any prime p, that the mod p Steenrod algebra has infinite little Krull dimension and that the Steenrod algebra has no Krull dimension, classical Krull dimension, or Gabriel dimension. We also show that a semiprime ring with a nonzero nil ideal has no Gabriel dimension.
The current work is broken into 3 chapters. The first chapter presents some of the history and general knowledge of the Steenrod algebra, followed by some of the notational conventions we will use. Chapter 2 concentrates on showing the primality of the Steenrod algebra for odd primes, and Chapter 3 discusses the various Krull dimensions of the Steenrod algebra.
1 Chapter 1
History and Conventions
1.1 Background
Before we can introduce the Steenrod algebra, we will very briefly explain what is cohomology. This will allow us to better describe the Steenrod squares and Steenrod powers and how they are used.1 Some axioms and properties of these squares and powers will then be presented, followed by the algebraic description of the Steenrod algebra. We will also discuss some of the early uses of the Steenrod algebra.
We will then give an introduction to prime ideals and prime rings, followed by the various Krull dimensions and Gabriel dimension. Then we will set the notation that we will use for the rest of paper.
1.1.1 Cohomology
Our exposition won’t be the most general, but only what we need. An interested reader who is unfamiliar with cohomology can refer to [21], [36], and [41], for a topological viewpoint of cohomology, or [6], [7], [8], and [13], for group cohomology.
Details about constructing the chains below can be found in those references.
1For another general overview of the Steenrod algebra, see [49].
2 A chain complex over a commutative ring R is a sequence of R-modules and
R-maps C = (Cn, ∂n):
∂n ∂n+1 · · · ←− Cn−1 ←− Cn ←−−− Cn+1 ←−· · ·
with ∂n∂n+1 = 0 for all n. Since img ∂n+1 ⊆ ker ∂n, we can form the quotient module
n th H (C) = ker ∂n/img ∂n+1, which we call the n homology of C.
∗ Taking duals and writing C = HomR (Cn,R), we have the cochain complex over R ∗ ∂∗ ∗ ∂n ∗ n+1 ∗ · · · −→ Cn−1 −→ Cn −−−→ Cn+1 −→· · ·
∗ ∗ ∗ ∗ with ∂n+1∂n = 0 for all n. This time, img ∂n ⊆ ker ∂n+1, and we form the quotient
n ∗ ∗ th module H (C) = ker ∂n+1/img ∂n, the n cohomology of C. In topology, starting with a space X and a commutative ring R, there are several
ways to construct associated chain complexes of R-modules. Conveniently, these all
produce the same homology and cohomology, which are denoted
M ∗ M n H∗ (X; R) = Hn (X; R) and H (X; R) = H (X; R) , n n
∗ respectively, both H∗(X; R) and H (X; R) are graded R-modules. Given a map of
spaces f : X → Y , there are induced homomorphisms f∗ : Hn(X; R) → Hn(Y ; R) and f ∗ : Hn(Y ; R) → Hn(X; R). As an example, the diagonal map d : X → X × X, defined by d(x) = (x, x), induces a homomorphism of graded R-modules
∗ H∗ (X × X; R) −→d H∗ (X; R) .
We also have the cross product (see [21, Page 210])
Hk (X; R) × H` (Y ; R) −→c Hk+` (X × Y ; R) .
3 When Y = X, the composition d∗ ◦ c helps define a multiplicative structure on
H∗ (X; R) called the cup product (see [21] and [41] for details): for α ∈ Hk (X; R)
and β ∈ H` (X; R), the cup product is written α ^ β ∈ Hk+` (X; R). We note
here that the cup product is anticommutative; that is, α ^ β = (−1)k`β ^ α for
α ∈ Hk (X; R) and β ∈ H` (X; R). With this multiplicative structure, H∗ (X; R) is a
graded R-algebra
1.2 The Steenrod Squares and Powers
Before introducing the Steenrod squares and powers, we give definitions to the
properties that the Steenrod squares and powers have.
m A cohomology operation (see [21, 44]) is a transformation Θ = ΘX : H (X; R) → Hn (X; T ) defined for all spaces X, with fixed choice of m, n, R, and T , and satisfying
the naturality property f ∗Θ = Θf ∗ for all maps f : X → Y ; that is, Θ satisfies the
commuting diagram
ΘY Hm (Y ; R) / Hn (Y ; T )
f ∗ f ∗
ΘX Hm (X; R) / Hn (X; T ) for all maps f : X → Y . These operations “are used to decide questions about the existence of continuous mappings which cannot be settled by examining cohomology groups alone” [44]. We define the suspension of a space X, denoted SX, as the quotient of X × [0, 1] obtained by collapsing X × {0} to a point and X × {1} to
another point. A cohomology operation Θ is said to be stable if it commutes with the suspension operator; notationally, SΘ = ΘS. Though we will not make use of this property, it is useful in the study of algebraic topology. For instance, Hatcher
[21, Theorem 4.L2] uses stability to show that the stable homotopy groups of certain spheres are nontrivial. See [21, 44] for more information about suspension.
4 Two continuous maps f, g : X → Y are said to be homotopic, written f ∼ g, if
f can be continuously deformed into g; that is, f ∼ g if there exists a continuous family of maps ft : X → Y (0 ≤ t ≤ 1), such that f0 = f and f1 = g. Two spaces X and Y are homotopy-equivalent (or of the same homotopy type), written X ∼ Y ,
if there exists maps f : X → Y and g : Y → X, such that gf ∼ 1X and fg ∼ 1Y ,
where 1X and 1Y are the identity maps of X and Y respectively. The relation ∼ on spaces is an equivalence relation, and one can try to enumerate the homotopy classes
of continuous maps f : X → Y . This is called the homotopy classification problem, and the enumeration should be in terms of algebraic invariants of X and Y (see [12]).
We introduce the Bockstein homomorphism (see [21, 44]), which is the coboundary
n n+1 operation δ : H (X; Zp) → H (X; Zp) associated with the coefficient sequence
p 0 → Zp → Zp2 → Zp → 0.
The Bockstein homomorphism has the properties
δδ = 0 and δ (α ^ β) = δα ^ β + (−1)|α| α ^ δβ,
as well as naturality and stability.
1.2.1 Axiomatic Description of the Steenrod Squares and
Powers
In 1947, Steenrod [42] introduced a set of stable cohomology operators, later
named the Steenrod squares, as a generalization of the cup product, which he used
to solve the homotopy classification problem for maps of an (n + 1)-complex K in Sn.
5 These cohomology operators are denoted
i n n+i Sq : H (X; Z2) → H (X; Z2) , for i ≥ 0, (1.1)
∗ where the cohomology groups H (X; Z2) of some complex X have Z2 coefficients. These are the unique operations that satisfy the following axioms (see [44]):
1. Naturality: Sqi f ∗ = f ∗ Sqi for f : X → Y , a map of complexes.
2. Linearity: Sqi (α + β) = Sqi (α) + Sqi (β).
3. Sq0 = 1.
4. If i = |α|, then Sqi (α) = α2.
5. If i > |α|, then Sqi (α) = 0.
i P j k 6. The Cartan Formula [9]: Sq (α ^ β) = j+k=i Sq (α) ^ Sq (β).
These 6 axioms imply the following properties:
1. The operations Sqi are stable operations.
2. Sq1 = δ.
3. The Adem Relations [5]: If a < 2b, then
[a/2] X b − 1 − j SqaSqb = Sqa+b−jSqj, a − 2j j=1
with the binomial coefficients taken modulo 2.
Later on in 1953, Steenrod [43] introduced what were later called the Steenrod powers, which are an extension of the Steenrod squares to any odd prime p. These are denoted
i n n+2i(p−1) P : H (X; Zp) → H (X; Zp) , for i ≥ 0, (1.2)
6 ∗ where the cohomology groups H (X; Zp) of some complex X have Zp coefficients with p an odd prime. These are the unique operations that satisfy the following axioms
(see [44]):
1. Naturality: Pif ∗ = f ∗Pi for f : X → Y , a map of complexes.
2. Linearity: Pi (α + β) = Pi (α) + Pi (β).
3. P0 = 1.
4. If 2i = |α|, then Pi (α) = αp.
5. If 2i > |α|, then Pi (α) = 0.
i P j k 6. The Cartan Formula: P (α ^ β) = j+k=i P (α) ^ P (β).
These 6 axioms imply the following properties:
1. The operations Pi are stable operations.
2. The Adem Relations [5]: If a < pb, then
[a/p] X a+t (p − 1) (b − t) − 1 PaPb = (−1) Pa+b−tPt, a − pt t=0
and if a ≤ pb, then
[a/p] X a+t (p − 1) (b − t) PaδPb = (−1) δPa+b−tPt a − pt t=0 [(a−1)/p] X a+t−1 (p − 1) (b − t) − 1 + (−1) Pa+b−tδPt, a − pt − 1 t=0
with the binomial coefficients taken modulo p.
For the interested reader, the construction of the Steenrod squares and powers can be found in Appendix A.
7 1.2.2 Algebraic Description of the Steenrod Algebras
An iterated p-power is a composition of two or more of the operations Sqi, when
p = 2, or δ and Pi, when p > 2. As examples of iterated powers, we have PiδPk and
Sqi Sqj, for i, j, k ≥ 0. It is clear that any iterated power is also a cohomology operator
∗ i on H (X; Zp). So, one could form an algebra, over Zp, involving the operations Sq from (1.1), when p = 2, and δ and Pi from (1.2), when p > 2. This is the so-called
Steenrod algebra.
Following Adem [5] in 1957, the Steenrod algebra Ap can be defined as follows:
0 1 Let R denote the free associative algebra over Zp generated by Sq , Sq ,..., if p = 2 and δ, P0, P1,..., if p > 2. Let I be the ideal of R consisting of elements f with
∗ the property that fα = 0 for all complexes K and all α ∈ H (K; Zp). Elements
of I are called relations. The Steenrod Algebra is then the quotient Ap = R/I. Adem showed in [5] that the relations, now called the Adem relations, given on Pages
6 and 7 generate the ideal I. Thus, the Steenrod algebra can be defined purely
algebraically by modding out these Adem relations from the algebra R. Adem also
n 2i o showed that the set Sq : i = 0, 1, 2,... generates the algebra A2, and the set
n pi o P : i = 0, 1, 2,... ∪ {δ} generates the algebra Ap for p > 2. We note here that the Steenrod algebra consists of all the stable cohomology operations.
∗ For any space X, the cohomology ring H (X; Zp) is an Ap-module. Given prop- erty 5 above for both p = 2 and p > 2, we have a certain property called instability:
∗ i • if x ∈ H (X; Zp) and i > |x|, then Sq x = 0, for p = 2;
∗ e i • if x ∈ H (X; Zp) and e + 2i > |x|, e = 0, 1, then δ P x = 0, for p > 2.
2 We say an Ap-module is unstable if it satisfies the preceding property. See [39] for details about unstable modules.
2The use of the words “unstable” and “instability” is not related to the use of the word “stable.”
8 The Steenrod algebra is a graded ring of finite type (that is the grading consists of finite dimensional vector spaces) with grading as follows: |Sqr| = r, for p = 2, and
|Pr| = 2r (p − 1) and |δ| = 1, for p > 2. Here, for example, the notation |δ| denotes the degree of the element δ.
In 1958, Milnor [33] showed that the Steenrod algebra is a cocommutative Hopf
3 algebra. The coalgebra structure of the Steenrod algebra Ap is given by
α if |α| = 0 n P i j ε (α) = ∆ (Sq ) = i+j=n Sq ⊗ Sq 0 if |α| > 0. for p = 2 and
∆ (Pn) = P Pi ⊗ Pj α if |α| = 0 i+j=n ε (α) = ∆ (δ) = δ ⊗ 1 + 1 ⊗ δ 0 if |α| > 0 for p > 2. The formulas for ∆ come from the Cartan formulas presented on Pages
6 and 7. We see that Ap is cocommutative. Since Ap is graded, it has an antipode, which Milnor discusses in [33, Section 7].
∗ Since Ap is of finite type, the dual of Ap, written Ap, is a commutative Hopf
∗ algebra. Milnor (see [33, Theorem 2] for details) showed that Ap is the tensor product
k of an exterior algebra generated by τ0, τ1,..., where |τk| = 2p − 1, and a polynomial
k algebra generated by ξ1, ξ2,..., where |ξk| = 2p − 2, when p is odd. When p = 2,
∗ k A2 is just a polynomial algebra generated by ξ1, ξ2,..., where |ξk| = 2 − 1. The
∗ coalgebra structure of Ap (see [33, Theorem 3]) for p > 2 is given by
∗ Pk pi m (ξk) = ξ ⊗ ξi α if |α| = 0 i=0 k−i u∗ (α) = ∗ Pk pi m (τk) = τk ⊗ 1 + i=0 ξk−i ⊗ τi 0 if |α| > 0.
3See Appendix B for an introduction to Hopf algebras.
9 ∗ ∗ Pk 2i The comultiplication for A2 is simply m (ζk) = i=0 ζk−i ⊗ ζi. We see, as expected,
∗ that Ap is not cocommutative. Milnor used the simpler commutative multiplicative structure of the dual to form a new basis (defined in §1.4 below) for the Steenrod algebra, now called the Milnor basis, that allows one to calculate the composition, or multiplication, of the Steenrod operations in terms of his basis. Furthermore, Milnor showed that the Steenrod algebra has a filtration by finite dimensional subalgebras (see §1.4 below), which showed that every element of positive degree in the Steenrod algebra is nilpotent. We also note that the Steenrod algebra is a local ring where the maximal ideal consists of all the elements of positive degree.
1.2.3 Early Uses of the Steenrod Algebra
The Steenrod algebra has become an important tool in the subject of algebraic topology. Some of the early applications of the Steenrod algebra applied to unsolved problems were the Hopf invariant 1 problem and vector fields on spheres. We will briefly describe these below, and the reader may refer to [21] for more details.
For each n, let Sn−1 denote the unit sphere in Euclidian n-space Rn. When given a
m n 4 m+1 n map f : S → S , one can form a complex Cf by attaching a cell e to S by using
∗ f. When m = 2n − 1, cup products in H (Cf ) have a chance of being nontrivial.
n 2n Given generators α ∈ H (Cf ) and β ∈ H (Cf ), the multiplicative structure of
∗ 2 H (Cf ) is determined by the equation α = H(f)β, where H(f) is an integer called the Hopf invariant of f. The Hopf invariant 1 problem [23] asks for which values of n does H(f) = 1. Using the Steenrod algebra, Adams [2] showed that the map f : S2n−1 → Sn has Hopf invariant 1 only when n = 2, 4, 8.
On the subject of vector fields on spheres, there was a question about the maxi- mum number of orthogonal tangent vector fields on a sphere. Write n = (2a + 1) 2b
4The reader may refer to [21, Page 5] for more information about cells.
10 with b = c + 4d, where a, b, c, d ∈ Z and 0 ≤ c ≤ 3. Define ρ(n) = 2c + 8b. Following Eckmann [11], there exists ρ(n)−1 linearly independent vector fields on Sn−1. Adams
[3] showed that there does not exists ρ(n) linearly independent vector fields on Sn−1.
As Adams noted, Steenrod and Whitehead [45] showed this to be true for b ≤ 3; their proof uses the Steenrod squares. Instead of using the Steenrod squares, Adams refor- mulates the problem into K-theory, and chooses an appropriate cohomology operator to suit his needs. An interested reader may refer to Hatcher [21] for a proof for all n 6≡ −1 mod 16; his proof rephrases the stated problem in terms of Stiefel manifolds and uses the Steenrod squares.
1.3 Prime Ideals and Krull Dimensions
Since this paper discusses the topics of prime rings and Krull dimensions, we will briefly discuss the history of these topics. The first attempt to study prime ideals in a general ring – not necessarily a commutative ring – was by McCoy [32] in 1949. In a commutative ring R, an ideal P is called a prime ideal if ab ∈ P implies a ∈ P or b ∈ P , were a and b are elements of R. Fitting [15] used this definition for an arbitrary ring, calling the ideal a completely prime ideal. McCoy defines a prime ideal in weaker terms: In a ring R, an ideal P is called a prime ideal if AB ⊆ P implies A ⊆ P or
B ⊆ P , where A and B are ideals of R. McCoy derives several equivalent statements in [32] for his definition of prime ideal, including the following definition: In a ring R, an ideal P is called a prime ideal if aRb ⊆ P implies a ∈ P or b ∈ P , where a and b are elements of R. McCoy shows that there is a characterization for prime ideals: an ideal P is prime if and only if the complement of P is an m-system. An m-system M is a subset M of R such that for all a, b ∈ M, there exists x ∈ R such that axb ∈ M.
A prime ring R is a ring where the zero ideal is prime; an equivalent definition
is, for a, b ∈ R, it is true that aRb = 0 implies a = 0 or b = 0. This allows us to
11 present another description of prime ideals, which is what we use in this paper: an ideal P of a ring R is a prime ideal if R/P is a prime ring. The idea of prime rings in noncommutative ring theory is a generalization of the idea of integral domains in commutative ring theory. We note here a generalization of prime ideals (see [30] for more information): An ideal C of a ring R is called a semiprime ideal if for any ideal
A of R, it is true that A2 ⊆ C implies A ⊆ C. Equivalently, and ideal C is semiprime if aRa ⊆ C implies a ∈ C for all a ∈ R. Likewise, a semiprime ring is a ring where the zero ideal is semiprime. A prime ring is also semiprime.
There is a relationship between algebraic varieties and chains of prime ideals, studied by Emmy Noether. For a field F , recall that an affine algebraic set is a subset V of F n which is the set of common zeros of some collection I of polynomials in F [x1, x2, . . . , xn]. The set V is called an algebraic variety if V is irreducible; that is, V cannot be written as V1 ∪ V2 for proper algebraic sets in V . Letting
F [V ] = F [x1, x2, . . . , xn] /I denote the coordinate ring of the variety, the dimension of V is defined to be the transcendence degree of the field of fractions of F [V ] over
F (see [10]). If P is a prime ideal in F [x1, x2, . . . , xn] consisting of those polynomials that vanish on V , then this dimension of V equals the maximum length of a chain of prime ideals ascending from P (see [17]).
In 1928, Krull’s work (see [28]) on the principle ideal theorem used chains of prime ideals and the height of ascending chains of prime ideals. We define the height of a prime ideal P in a commutative ring R, denoted ht(P ), to be the largest integer n such that there exists a chain P = P0 > P1 > ··· > Pn. If the chain is unbounded, then we write ht(P ) = ∞. This definition led to the definition of the dimension of a commutative ring, later called the Krull dimension: the dimension of a commutative ring R is
sup {ht(P ): P ∈ Spec R} , where Spec R is the set of all proper prime ideals of R.
12 Not all noncommutative rings have enough prime ideals to make the definition above interesting or useful. We define here various types of Krull dimensions and the
Gabriel dimension. The rings mentioned below in these definitions are not necessarily commutative. These definitions can all be found in [18].
Little Krull Dimension
One attempt to adapt the above definition to arbitrary rings and ordinal numbers is the following definition from [18, Page 47].
A limit ordinal is an ordinal number that is neither zero nor a successor of an ordinal. Let λ (W ) denote the length of a well-ordered set W defined by
ord W if the order type of W is a limit ordinal, λ (W ) = ord W − 1 otherwise.
The little classical Krull dimension or, for short, little Krull dimension, which we denote by k dim R, for an arbitrary ring R is defined by
k dim R = sup {λ (W ): W ⊂ Spec R, well-ordered under reverse inclusion} .
For example, given any ordinal α, there is a principle right ideal domain with little Krull dimension α, and the dimension is given by the length of a single chain, see [18]. Also, for a countable field F , the polynomial ring F [x1, x2,...] has little
Krull dimension of ω1, the first uncountable ordinal (see Proposition 3.4.1 below). As another example, consider the nth Weyl algebra over a ring R, which is defined as
An(R) = R [x1, θ1, x2, θ2, . . . , xn, θn] ,
where the xi commute with themselves, the θi commute with themselves, and θxj −
13 5 xjθi = δij, the Kronecker delta. If F has characteristic zero, then An(F ) is a simple algebra, from which it follows that k dim An(F ) = 0. In fact, all simple algebras have little Krull dimension equal to zero. In this case, the little Krull dimension is does not supply useful information.
It is worth noting that in some cases, the little Krull dimension is the most in- teresting case of all the Krull dimensions. The Steenrod algebra is an example; the little Krull dimension of the Steenrod algebra is infinite while the other two Krull dimensions, defined below, do not exist.
Classical Krull Dimension
There is a more general definition for a Krull dimension that also uses ordinal numbers, which first was introduced by Krause in [26, Definition 11]. The following definition is from [18].
Let P0 denote the set of all maximal ideals of an arbitrary ring R. Then, if α > 0 is an ordinal, denote by Pα the set of prime ideals P of R such that each prime
Q properly containing P belongs to Pβ for some β < α. Then the classical Krull dimension of R, denoted by cl K dim R, is defined to be the first ordinal α such that
Pα = Spec R.
As simple examples, we have cl K dim Z = 1, where Z denotes the ring of integers, and cl K dim F [x1, x2, . . . , xn] = n, where F is a field. As with the little Krull dimension, all simple algebras have a classical Krull dimension equal to zero. For example, when F is a field with characteristic zero, we have cl K dim An(F ) = 0 since the nth Weyl algebra is simple. In this case, the classical Krull dimension does not yield much information about such algebras.
5 We use the notation An(R) for the Weyl algebra and hope it is not confused with the notation Ap for the Steenrod algebra or the notation for Ap(n), defined on Page 20.
14 Krull Dimension
The Krull dimension for a module was defined in [16] for finite ordinals. This
definition was extended to arbitrary ordinals in [18], and we present this definition
below.
Let R be a ring and M a right R-module. The Krull dimension of M, denoted
K dim M, is defined by transfinite recursion as follows: if M = 0, then K dim M =
−1. If α is an ordinal and K dim M 6< α, then K dim M = α provided there is no infinite descending chain M = M0 ⊃ M1 ⊃ · · · of submodules Mi such that, for i = 1, 2,..., K dim (Mi−1/Mi) 6< α. It is possible that there is no ordinal α such that K dim M = α, in which case we say M has no Krull dimension.
The Krull dimension of a ring R, denoted by K dim R, is defined to be the Krull dimension of the right R-module R.
The Krull dimension is a measure of how close a module or ring is to being ar- tinian. Notherian modules always have Krull dimension, but there are non-noetherian rings and modules that have Krull dimension, which have many noetherian-like con- sequences. The modules of Krull dimension 0 are precisely nonzero artinian modules, and K dim Z = 1. We note here that when F is a field with characteristic zero, then K dim An(F ) = n (see, for example, [17]), which differs from the other Krull dimensions mentioned above.
From [18, Proposition 7.9], if R is a ring with Krull dimension then cl K dim R
exists and
k dim R ≤ cl K dim R ≤ K dim R.
Gabriel Dimension
When studying the Krull dimension of modules, two special types of modules arise
(see [18]): critical modules and monoform modules. A nonzero module M is called
α-critical if K dim M = α and K dim M 0 < α for each proper homomorphic image
15 M 0 of M.A critical module is one which is α-critical for some ordinal α. Any nonzero module with Krull dimension has a critical module [18, Theorem 2.1].
If a nonzero module M has the property that, for each submodule N every homo- morphism ϕ : N → M is either zero or a monomorphism, then we call M a monoform module. Every critical module is monoform [18, Corollary 2.5]. Using this definition, one can describe, as in [18], monoform modules in terms of Serre subcategories. With this description, monoform modules show up in different places and in different forms: for instance, as simple objects in a quotient category or as supporting modules for kernel functions. Eventually, this leads to the definition of the Gabriel dimension given below (from [18]).
The Gabriel dimension is defined via a certain ascending chain, indexed by or- dinals, of localizing subcategories Mα of M = ModR, the category of R-modules.
Briefly, M0 = {0}, Mα+1 for α ≥ 0 is the smallest localizing subcategory containing all R-modules which have finite length considered as objects of the quotient category
M/Mα and, at limit ordinals, Mα is the smallest localizing subcategory containing S β<α Mβ.
If it happens that Mα = ModR for some ordinal α then the smallest such α is the
Gabriel dimension of R, which we denote G dim R. Similarly, if M ∈ Mα for some α, the least such α is the Gabriel dimension of M, denoted G dim M.
From [18, Page 16], every module with Krull dimension (K dim R) automatically has Gabriel dimension. Following [18], if a module contains no infinite direct sum of nonzero submodules, then there is a fixed bound on the number of summands. The module is said to have finite uniform dimension. As stated in [18] (proved in [19]), a module has Krull dimension if and only if it has Gabriel dimension and all its factors have finite uniform dimension. Furthermore, the dimensions differ by at most one.
We mention here that it is possible for a ring to have Gabriel dimension, but no
Krull dimension. We follow Hansen in [20]. Let K be the field of rational functions
16 over Q and let ’ denote the usual derivation on K. Also, let S = K [x;0 ] be the ring of differential polynomials where the coefficients are written on the right. For example, kx = xk + k0 for k ∈ K. It is known that S is a simple right and left ideal domain.
Let L be a proper, nonzero right ideal of S. The ideal L is a simple ring. Each nonzero ideal of Z + L = I contain L. Also, I has a Gabriel dimension of 3 and no Krull dimension.
We note here that Theorem 3.3.3 below shows that a semiprime ring with a nonzero nil ideal has no Gabriel dimension.
1.4 Definitions and Conventions
In this section we collect some standard facts from the literature and in the process
fix our notation for the rest of the paper. Nothing in this section is new.
∗ As mentioned in 1.1, the dual of the Steenrod algebra Ap is a commutative algebra.
Let R = (r1, r2,...) be a sequence of nonnegative integers which are almost all are zero; we will call such sequences Milnor sequences. Also, let E = (ε0, ε1,...) be a sequence of ones and zeros which are almost all zero; we call such sequences binary
ε0 ε1 r1 r2 sequences. Define τ(E) = τ0 τ1 ··· and ξ(R) = ξ1 ξ2 ··· . The elements {τ(E)ξ(R)}
∗ form an additive basis for Ap and are dual to some basis of Ap. Milnor [33, Theorem
4a] showed that this basis for Ap consists of elements of the form
Sq{R} for p = 2 and
ε0 ε1 Q0 Q1 ···P{R} for p > 2,
where R is a Milnor sequence and almost all of the εi ∈ {0, 1} are zero. (We use the notation P{R} in place of to Milnor’s notation PR in the interest of greater
ε0 ε1 legibility.) If we define Q(E) = Q0 Q1 ··· for a binary sequence E = (ε0, ε1,...),
17 then basis elements can be written in the form Q(E)P{R}, and we have
1 if E = E0 and R = R0 hτ(E)ξ(R),Q(E0)P{R0}i = 0 otherwise.
For any Milnor sequence R, we can think of R as a finite sequence and write R =
(r1, . . . , rk) if ri = 0 for all i > k. Also, for R = (r1, . . . , rk), we’ll write P{r1, . . . , rk}
r1 r2 for P{R}. Note that P{r1, r2,...} 6= P P ··· in general, but we do have P{r} =
r P . The notation Pk(b) will be used to denote the element P{0,..., 0, b}, where b
th r is in the k position. For example, P1(r) = P . The same conventions are used for
Sq{R}. The elements Sq{R} = Sq{r1, . . . , rk} have degree