1: \section{Overview} In recent years there has been a lot of interest in the interactions between homotopy theory and its allied fields. Many fruitful interactions have been
2: revealed by the study of triangulated structures naturally arising in homotopy theory, homological algebra, algebraic geometry, and
3: modular representation theory. The far-reaching applications and generalisations of Brown representability \cite{ne5} \cite{ChKeNe} \cite{kr1},
4: Bousfield localisation \cite{Ri1} \cite{AlLeJe}, and nilpotence detection \cite{Ho} \cite{car}, to other fields are some striking examples of such interactions.
5: Several mathematicians brought to light many amazing a priori different theories
6: by classifying various subcategories of triangulated categories arising in these various fields. Following the seminal work of Devinatz,
7: Hopkins, and Smith \cite{dhs} in stable homotopy theory, this line of research was initiated
8: by Hopkins in the 80s. In his famous 1987 paper \cite{Ho}, Hopkins classified the thick subcategories (triangulated categories that are closed under retractions)
9: of the finite $p$-local spectra and those of perfect complexes over a noetherian ring. He showed that thick subcategories of the finite spectra are determined by the
10: Morava $K$-theories and those of perfect complexes by the prime spectrum of the ring. These results have had tremendous impacts in their respective fields.
11: The thick subcategory theorem for finite spectra played a vital role in the study of nilpotence and periodicity. For example, using this theorem
12: Hopkins and Smith \cite{hs} were able to settle the class-invariance conjecture of Ravenel \cite{rav} which classified the Bousfield classes of finite spectra.
13: Similarly the thick subcategory theorem for the derived category establishes a surprising connection between stable homotopy theory and algebraic
14: geometry; using this theorem one is able to recover the spectrum of a ring from the homotopy structure of its derived category! These ideas were later pushed further
15: into the world of derived categories of rings and schemes by Neeman \cite{Ne} and Thomason \cite{Th}, and into modular representation theory by Benson, Carlson and
16: Rickard \cite{bcr}. Motivated by the work of Hopkins \cite{Ho}, Neeman \cite{Ne} classified the Bousfield classes and localising subcategories in the derived category of a
17: noetherian ring. In modular representation theory, the Benson-Carlson-Rickard classification of the thick subcategories of stable modules over group algebras has led to
18: some deep structural information on the representation theory of finite groups. Finally the birth of axiomatic stable homotopy theory \cite{mps} in the mid 90s encompassed
19: all these various theories and ideas and studied them all in a more general framework. With all these developments over the last 30 years, the importance of triangulated
20: categories in modern mathematics is by now abundantly clear.
21:
22: Motivated by these results, in this thesis we focus on the study of triangulated subcategories in stable homotopy categories. (Loosely speaking, a stable
23: homotopy category is a sufficiently well-behaved triangulated category that is formally similar to the homotopy category of spectra, e.g., the derived category of
24: a ring; for more details, see section \ref{sec:tc}) We focus on classifications of triangulated subcategories and decompositions of thick subcategories of compact
25: objects in stable homotopy categories which include the homotopy category of spectra, derived categories of rings, and the stable module categories of some finite
26: dimensional co-commutative Hopf algebras. Our main tools will be the thick subcategory theorems and $K$-theory.
27:
28:
29:
30: \section{Main Results}
31:
32: One of our goals is to classify the triangulated subcategories of finite spectra and those of perfect complexes over a noetherian ring. To this end,
33: we adopt a $K$-theoretic approach of Thomason. To set the stage, let $\T$ denote a triangulated category (we have the compact objects of a stable homotopy category
34: in mind) and let $K_0(\T)$ denote the \emph{Grothendieck group} of $\T$. Thomason
35: established in \cite{Th} a bijection between the dense (see definition \ref{defn:dense}) triangulated subcategories of $\T$ and the subgroups of $K_0(\T)$.
36: This theorem gives a key to the problem of classifying triangulated subcategories of $\T$:
37: \emph{First classify all thick subcategories of $\T$ and then compute the Grothendieck groups of these thick subcategories.
38: For every subgroup of each of these Grothendieck groups we get a triangulated subcategory of $\T$, and every triangulated subcategory
39: arises this way.} Thus we can try to refine all the aforementioned thick subcategory theorems (\cite{hs}, \cite{Ho,Ne}, \cite{bcr}) to obtain a classification of
40: triangulated subcategories. By ``chromatic tower for a stable homotopy category'', we mean the lattice of all the thick subcategories of compact objects. Since a
41: thick subcategory is also a triangulated subcategory, this $K$-theoretic approach gives a refinement of the chromatic towers, which explains the first part of the
42: title of this thesis.
43:
44: In chapter 3 we apply this $K$-theoretic technique to the category $\F_p$ of finite $p$-local spectra.
45: This gives rise to the following triangulated subcategories of $\F_p$. Let $\C_n$ denote the thick subcategory of $\F_p$ consisting of $K(n-1)$-acyclics, and
46: for non-negative integers $k$ and $n \ge 1$, define a full subcategory of $\F_p$ as
47: \[ \C_n^k = \{ X \in \C_n \; | \; \chi_n(X) := \sum_i (-1)^i \log_p |\bp{n-1}_i X| \equiv 0 \ \text{mod} \, l_n k \},\]
48: where $l_n$ is the smallest positive value of $\chi_n(X)$ on $\C_n$.
49: Define $\C_0^k$ to be the full subcategory of spectra in $\F_p$ whose (rational) Euler characteristic is divisible by $k$.
50: Our understanding of the lattice of triangulated subcategories of $\F_p$ can be summarised in the following theorem.
51:
52:
53: \begin{named}{\textbf{Theorem A}} \emph{For each $n \ge 0$ and $k \ne 1$, the subcategory $\C_n^k$ is a dense triangulated subcategory of $\C_n$.
54: These triangulated subcategories satisfy the inclusions
55: \[ \C_{n+1} \subsetneq \C_n^k \subsetneq \C_n .\]
56: Moreover, when $n=0$ or $1$, every dense triangulated subcategory of $\C_n$ is equal to $\C_n^k$ for some $k$. }
57: \end{named}
58:
59: One of the interesting parts of this theorem is to establish the inclusion $\C_{n+1} \subseteq \C_n^k$. This is shown using a spectral sequence calculation.
60: For a fixed $X \in \C_{n+1}$, we construct a strongly convergent Bockstein spectral sequence
61: \[ E_r^{*,*} \Rightarrow E(n)_*(X)\]
62: whose $E_1$ term is build out of $\bp{n-1}_* X$, and which abuts to
63: \[ \holim{v_n} \bp{n}_* X = E(n)_*(X) \quad (= 0 \; \mbox{by assumption}).\]
64: Now working backward with this spectral sequence we are able to show that $X \in \C_n^k$ for all $k$.
65: This theorem gives some new evidence to the following conjecture of Adams.
66:
67:
68: \begin{conj} \cite[Page 529]{MayTho} The Smith-Toda complex $V(1)$ generates the thick subcategory $\C_2$ by iterated cofiberings.
69: \end{conj}
70: It is not hard to see that this conjecture implies that $\C_2 \subseteq \C_1^k$, which is part of the above theorem.
71:
72: Among other things, the above theorem is saying that the thick subcategory $\C_1$ can be generated by iterated cofiberings of $M(p)$, the mod-$p$ Moore
73: spectrum; if we also allow retractions, then this is a well-known consequence of the thick subcategory theorem of Hopkins and Smith \cite{hs}.
74: This gives the following interesting corollary. For a $p$-torsion spectrum $X$, define its Euler characteristic by
75: \[\chi_1(X): = \sum_i (-1)^i \log_p |H\mathbb{Z}_i (X)|. \]
76:
77: \begin{cor} Let $X$ be a type-$1$ spectrum and $Y$ an arbitrary finite $p$-torsion spectrum. Then $Y$ can be generated by $X$ via cofibrations
78: if and only if $\chi_1(X)$ divides $\chi_1(Y).$
79: \end{cor}
80:
81: This corollary can be compared with the following result of Hopkins \cite{Ho}: Let $X$ and $Y$ be finite $p$-local spectra. Then $Y$ can be generated by
82: $X$ via cofibrations and retractions if and only $\Supp(Y) \subseteq \Supp(X)$. ($\Supp(X)$ is the chromatic support of $X$; the set of all
83: non-negative integers $n$ such that $K(n)_* X \ne 0$.)
84:
85: %If $X$ and $Y$ are perfect complexes over a noetherian ring, then $Y$ can be
86: %generated by $X$ via cofibrations and retractions if and only if $\Supp(Y) \subseteq \Supp(X).$
87:
88: In chapter 4, we do similar $K$-theoretic analysis in the derived category and obtain complete classifications of perfect complexes over some noetherian rings
89: (PIDs and Artin rings). See section \ref{se:classifications} for these results.\vskip 3mm \noindent
90:
91: We now move on to the second part of this thesis -- Krull-Schmidt decompositions for thick subcategories; see definition \ref{defn:KS}. Our motivation for studying such
92: decompositions comes partly from the aforementioned classifications of triangulated subcategories. We have noticed several situations where the Grothendieck group
93: under study is a direct sum of infinite cyclic groups. A careful investigation has led us to the observation that this behaviour is in fact induced by a decomposition
94: of the underlying category. More precisely, if $\A = \coprod \A_i$ is a Krull-Schmidt decomposition of a thick subcategory $\A$, then under certain conditions (which are
95: always satisfied in the examples we consider,
96: e.g., $\Hom(\A_i, \, \A_j) = 0$ for all $i \ne j$), $K_0(\A) \cong \bigoplus_i K_0(\A_i)$. In short, Krull-Schmidt decompositions respect $K$-theory.
97:
98: We now summarise our Krull-Schmidt results in the next theorem.
99:
100: \begin{named} {\textbf{Theorem B}} \emph{Let $R$ denote a noetherian ring and let $B$ denote a finite dimensional graded co-commutative Hopf algebra satisfying the
101: tensor product property (e.g., finite dimensional sub-Hopf algebras of the mod-$2$ Steenrod algebra). Then we have the following.
102: \begin{enumerate}
103: \item Every thick subcategory of perfect complexes over $R$ admits a Krull-Schmidt decomposition.
104: \item Every thick ideal of small objects in the chain homotopy category of projective $B$-modules is indecomposable.
105: \item Every thick ideal of finite dimensional stable $B$-modules admits a Krull-Schmidt decomposition.
106: \end{enumerate}
107: Further all these Krull-Schmidt decompositions are unique and they respect $K$-theory. }
108: \end{named}
109:
110: The essential idea behind the proofs of the decompositions in the above theorem is quite simple. We use thick subcategory theorems which establish a bijection
111: between thick subcategories of small objects and some geometric spaces. We obtain the above decompositions by decomposing the corresponding geometric spaces in an
112: appropriate way. These decompositions give some interesting splitting results. For example,
113:
114: \begin{cor} Let $X$ be a perfect complex over a noetherian ring $R$. Then $X$ admits a unique decomposition into perfect complexes,
115: \[ X \cong \underset{i \in I}{\bigoplus}\; X_i,\]
116: such that the supports of the $X_i$ are pairwise disjoint and indecomposable.
117: \end{cor}
118:
119: We now move on to the last chapter. Since computing the Grothendieck groups of thick subcategories is the key to the problem of classifying triangulated subcategories,
120: it is natural to investigate different ways to compute these groups.
121: So the motivating question now is the following. Can one compute the Grothendieck groups of thick subcategories of perfect complexes in the
122: $R$-module category? More precisely, if $\C$ is a thick subcategory of perfect complexes over $R$, is there a subcategory of
123: $R$-modules whose Grothendieck group is naturally isomorphic to that of $\C$?
124:
125: Towards this, it is natural to investigate the right analogue of a thick subcategory of complexes
126: in the world of modules. This has been investigated by Mark Hovey \cite{wide}. It turns out that the right
127: notion of ``thickness for modules'' is ``wideness'': A subcategory $\C$ of an abelian category
128: is said to be \emph{wide} if it is an abelian subcategory that is closed under extensions.
129:
130: Hovey \cite{wide} showed that when $R$ is regular and coherent, the wide subcategories of $\Wide (R)$ (the wide subcategory generated by $R$)
131: are in bijection with the thick subcategories of $\Thick (R)$ (the thick subcategory generated by $R$).
132: More precisely, he showed that the map
133: \[f: L_{\Wide} (R) \rightarrow L_{\Thick}(R),\]
134: from the lattice of wide subcategories to that of thick subcategories defined by
135: \[f(\C) = \{ X \in \Thick (R) : H_n(X) \in \C \;\; \text{for all}\; \; n \},\]
136: is an isomorphism when $R$ is regular and coherent. We use this classification result of Hovey to prove the following theorem.
137:
138: \begin{named} {\textbf{Theorem C}} \emph{Let $R$ be a regular coherent ring. If $\C$ is any wide subcategory of $\Wide(R)$, then
139: \[ K_0(\C) \cong K_0(f(\C)). \]
140: In addition if $R$ is noetherian, then every wide subcategory $\mathcal{W}$ of finitely generated modules
141: admits a Krull-Schmidt decomposition: $\mathcal{W} = \coprod_{i \in I} \mathcal{W}_i$. Moreover,
142: $K_0(\mathcal{W}) \cong \underset{i \in I}{\bigoplus} K_0(\mathcal{W}_i).$
143: }
144: \end{named}
145:
146: As a corollary to this theorem we recover a known splitting result for finitely generated modules over a noetherian ring; see corollary \ref{cor:splittingformodules}.
147:
148: \section{Organisation}
149:
150: This thesis is organised as follows. In chapter 2 we set up the categorical stage for our work. Here we recall some basic
151: facts about triangulated categories and Grothendieck groups, and state the main $K$-theory recipe of Thomason. In chapters 3 and 4, we apply these tools to the study of
152: classifications of triangulated subcategories of finite spectra and perfect complexes respectively. Chapter 5 deals with Krull-Schmidt decompositions in these various
153: stable homotopy categories. Finally in the last chapter we study wide subcategories where we mimic some of these ideas (from the previous chapters) in the
154: $R$-module category.
155:
156: Chapters 3 and 4 can be read independent of each other. Although most of Chapter 5 can be read separately, it would be well motivated if read after chapters 3 and 4.
157: Chapter 6 takes some input from chapter 4 and very little from chapter 5. The following picture shows the logical interdependencies of the chapters.
158: (A dotted arrow indicates a mild dependency.)
159:
160: \begin{figure}[!h]
161: \[
162: \xymatrix{
163: & *++[o][F-]{1} \ar[d] & \\
164: & *++[o][F-]{2} \ar[dl] \ar[dr] \ar[dd] & \\
165: *++[o][F-]{3} \ar@{..>}[dr] & & *++[o][F-]{4} \ar@{..>}[dl] \ar[dd] \\
166: & *++[o][F-] {5} \ar@{..>}[dr] & \\
167: & & *++[o][F-]{6}
168: }
169: \]
170: \label{logic}
171: \caption{Interdependencies of chapters}
172: \end{figure}
173:
174: Every chapter begins with an agenda and ends (except chapter 2) with a discussion of questions and further directions.
175:
176:
177:
178: %%% Local Variables:
179: %%% mode: latex
180: %%% tex-main-file: "uwthesis.tex"
181: %%% End:
182: