1: %%%%%%%%%%%%%%%%%%%%%%%%%%% Start dr.tex %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2:
3: %% This document created by Scientific Word (R)
4: %% Version 2.0
5:
6:
7: \documentstyle[amssymb,thmsd,a4,11pt]{article}
8: %%%%%%%%%%%%%%%%%%%%%%%%%%%%
9: \input tcilatex
10: \input diagram
11: \setlength{\oddsidemargin}{-0.1 cm}
12: \setlength{\evensidemargin}{-0.1 cm}
13: \setlength{\textwidth}{16 cm}
14: \setlength{\textheight}{22 cm}
15: \setlength{\topmargin}{-1.5 cm}
16: \newtheorem{examples}[theorem]{Examples}
17: \newtheorem{PropDef}[theorem]{Proposition and Definition}
18: \newtheorem{Observation}[theorem]{Observation}
19: \QQQ{Language}{
20: American English
21: }
22:
23: \begin{document}
24:
25: \author{Gabriele Vezzosi \\
26: %EndAName
27: {\em Dipartimento di Matematica}\\
28: {\em Universit\`a di Bologna}\\
29: {\em Piazza di Porta S. Donato 5, 40127}\\
30: {\em Bologna - Italy}\\
31: e-mail: {\tt vezzosi@dm.unibo.it}\\
32: \smallskip\ \\
33: Alexandre M. Vinogradov \\
34: %EndAName
35: {\em Dipartimento di Matematica e Informatica}\\
36: {\em Universit\`a di Salerno}\\
37: {\em Via S. Allende} {\em 84081-Baronissi (SA) Italy }\\
38: {\em and}\\
39: {\em the Diffiety Institute 45/6, Polevaya st.,}\\
40: {\em \ Pereslavl-Zalessky, Russia.}\\
41: {\em \ }e-mail: {\em \ }{\tt vinograd@ponza.dia.unisa.it}}
42: \title{{\bf On higher order analogues of de Rham cohomology }}
43: \date{}
44: \maketitle
45:
46: \begin{abstract}
47: If $K$ is a commutative ring and $A$ is a $K$-algebra, for any sequence $%
48: \sigma $ of positive integers there exists \cite{Vi2} an higher order
49: analogue ${\bf dR}_\sigma $ of the standard de Rham complex ${\bf dR}\equiv
50: {\bf dR}_{\left( 1,...,1,..\right) }$, which can also be defined starting
51: from suitable (=differentially closed) subcategories of $A-{\bf Mod}$. The
52: main result of this paper is that the cohomology of ${\bf dR}_\sigma $ does
53: not depend on $\sigma $, under some smoothness assumptions on the ambient
54: category. In \cite{VV}, a weaker result was proved by completely different
55: methods.
56:
57: Before proving the main theorem we give a rather detailed exposition of all
58: relevant (for our present purposes) functors of differential calculus on
59: commutative algebras. This part can be also of an independent interest.
60:
61: \smallskip
62: \ \
63: \end{abstract}
64:
65: \tableofcontents
66:
67: \section{Introduction}
68:
69: Higher analogues of the standard de Rham complex were found by one of the
70: authors about 20 years ago in course of his study of basic functors of
71: differential calculus (see \cite{Vi1}). More exactly, such an analogue $%
72: dR_\sigma $ can be associated with any sequence $\sigma $ of positive
73: integers. The standard de Rham complex turns out to be associated with the
74: simplest sequence of this kind, namely, with $\sigma =(1,1,...)$.
75: Differentials of these complexes are natural differential operators of,
76: generally, higher orders. Complexes $dR_\sigma $`s as well as related
77: functors they represent play an important role inside differential calculus
78: in the sense of \cite{Vi1} and as such are worth to be better investigated.
79: It is very plausible that they may have important applications to the formal
80: theory of partial differential equations as well as to Secondary Calculus
81: (see \cite{Vi3}), to mention the most direct ones. Of course, the first
82: natural question concerning these complexes is : what are their
83: cohomologies? At the end of eighties the second author jointly with Yu. N.
84: Torkhov tested some simple cases (unpublished) in order to check a natural
85: feeling that all complexes $dR_\sigma $ have the same cohomology. After that
86: in 1994 the authors proved (see \cite{VV-esi} and \cite{VV}) an analogue of
87: the ''infinitesimal Stokes formula'' for all $dR_\sigma $`s with
88: non-decreasing $\sigma $, which allows to develop a homotopy techniques
89: sufficient to show that these complexes are quasi-isomorphic. One year later
90: we found a somewhat involved proof of this fact in full generality. Its
91: essentially simplified version is the main result of this paper.
92:
93: In this paper we present also for the first time a rather detailed
94: exposition of all relevant for our purposes functors of differential
95: calculus (sections 1-3) and as such it could be of an independent interest.
96:
97: A preliminary version of this paper appeared as a Preprint \cite{VV-sns} of
98: the Scuola Normale Superiore, Pisa.
99:
100: Finally we note that recently A.M.Verbovetsky sketched in [Ve] an
101: alternative proof of the main theorem of this paper based on the theory of
102: compatibility complexes.
103:
104: \bigskip\
105:
106: \begin{center}
107: {\bf Notations and Conventions}
108:
109: \smallskip\ \
110: \end{center}
111:
112: $K:$ \qquad a commutative ring with unit ;
113:
114: $A:\qquad $ a commutative $K$-algebra with unit;
115:
116: $R-{\bf Mod:\qquad }$ the category of $R$-modules for a commutative ring $R$;
117:
118: ${\bf {\rm Diff}}_A:\qquad $the category whose objects are the $A$-modules
119: and the morphisms are differential operators (Section 1) between them;
120:
121: ${\bf Ens}:\qquad $the category of sets;
122:
123: $\left[ {\bf C},{\bf C}\right] :\qquad $the category of functors ${\bf %
124: C\longrightarrow C}$, ${\bf C}$ being a category;
125:
126: $Ob\left( {\bf C}\right) :\qquad $the objects of ${\bf C}$, ${\bf C}$ being
127: a category; $C\in Ob\left( {\bf C}\right) $ means that $C$ is an object of $%
128: {\bf C}$;
129:
130: $A-{\bf BiMod:\qquad }$ the category of $A$-bimodules, whose objects are
131: understood as {\it ordered couples} $(P,P^{+})$ of $A$-modules and whose
132: morphisms are the usual morphisms of bimodules. Note that $P$ and $P^{+}$
133: coincide as $K$-modules, hence as sets;
134:
135: $*A-{\bf BiMod:\qquad }$the category of ''biads'' (see Section 1);
136:
137: {\bf K}$\left( A-{\bf Mod}\right) $ (resp. {\bf K}$\left( K-{\bf Mod}\right)
138: $, resp. {\bf K}$\left( {\bf Diff}_A\right) $)$:$\quad \quad the category of
139: complexes in $A-{\bf Mod}$ (resp. $K-{\bf Mod}$, resp. ${\bf {\rm Diff}}_A$);
140:
141: ${\frak Ens}:\qquad $ the forgetful functor from a concrete category to $%
142: {\bf Ens}$;
143:
144: If ${\frak D}$ is a full subcategory of $A-{\bf Mod}$, a functor $T:{\frak D}%
145: {\bf \longrightarrow }{\frak D}$ will be said {\em strictly representable}
146: in ${\frak D}$ if it exists $\tau \in Ob\left( {\frak D}\right) $ and a
147: functorial isomorphism $T\simeq Hom_A\left( \tau ,\cdot \right) $ in $\left[
148: {\frak D},{\frak D}\right] $.
149:
150: $A-{\bf BiMod}_{{\frak D}}$ (resp. $*A-{\bf BiMod}_{{\frak D}}$, resp. {\bf K%
151: }$\left( {\bf Diff}_{A,{\frak D}}\right) $) will be the subcategory of $A-%
152: {\bf BiMod}$ whose objects are couples of objects in ${\frak D}$ (resp. the
153: subcategory of $*A-{\bf BiMod}$ whose objects are triples of objects in $%
154: {\frak D}$, resp. the subcategory of {\bf K}$\left( {\bf Diff}_A\right) $
155: whose objects are complexes of objects in ${\frak D}$).
156:
157: A sequence $T_1\rightarrow T_2\rightarrow T_3$ of functors $T_i:{\frak D}%
158: {\bf \longrightarrow }{\frak D}$, $i=1,2,3$, (and functorial morphisms) $%
159: {\frak D}$ being an abelian subcategory of $A-{\bf Mod}$, is said {\em exact
160: in }$\left[ {\frak D},{\frak D}\right] $ if it is exact in ${\frak D}$ when
161: applied to any object of ${\frak D}$.
162:
163: Let ${\bf N}_{+}^\infty =\underleftarrow{\lim }{\bf N}_{+}^k$ be the set of
164: infinite sequences of positive integers. If $\sigma \in {\bf N}_{+}^n$ (or $%
165: \sigma \in {\bf N}_{+}^\infty $) then $\sigma \left( r\right) \doteq \left(
166: \sigma _1,...,\sigma _r\right) $ for $r\leq n$ (or any $r\in {\bf N}_{+}$).
167: We denote by ${\bf 1}$ the element $(1,...,1,1,...1,...)\in {\bf N}%
168: _{+}^\infty $.
169:
170: \section{Absolute and relative functors}
171:
172: We recall here in a slightly different way the necessary definitions from
173: \cite{Vi1} ,\cite{Vi2},\cite{Kr} (see also \cite{KLV}).
174:
175: If $P$ and $Q$ are $A$-modules and $a\in A$ we define:
176: \[
177: \begin{array}{c}
178: \delta _a:Hom_K(P,Q)\longrightarrow Hom_K(P,Q) \\
179: \Phi \longmapsto \left\{ \delta _a\Phi :p\longmapsto \Phi (ap)-a\Phi
180: (p)\right\} \quad ,\text{ }p\in P
181: \end{array}
182: \]
183: (where we use juxtaposition to indicate both $A$-module multiplications in $%
184: P $ and $Q$). For each $a\in A$ $\delta _a$ is a homomorphism of $K$%
185: -modules, and commutativity of $A$ implies that
186: \[
187: \delta _{a_1}\circ \delta _{a_2}=\delta _{a_2}\circ \delta _{a_1}\quad
188: ,\forall a_1,a_2\in A\text{.}
189: \]
190: \
191:
192: \begin{definition}
193: A $K-${\em differential operator }({\em DO}){\em \ of order} $\leq s$ from
194: an $A$-module $P$ to another $A$-module $Q$ is an element $\Delta \in
195: Hom_K(P,Q)$ such that:
196: \[
197: \left[ \delta _{a_0}\circ \delta _{a_1}\circ ...\circ \delta _{a_s}\right]
198: (\Delta )=0,\text{ }\forall \left\{ a_0,a_1,...,a_s\right\} \subset A.
199: \]
200: \end{definition}
201:
202: We will write synthetically $\delta _{a_0,\ldots ,a_s}$ for $\delta
203: _{a_0}\circ \delta _{a_1}\circ ...\circ \delta _{a_s}$.
204:
205: The {\em set }{\it Diff}$_k(P,Q)$ of differential operators of order $\leq k$
206: from $P$ to $Q$ is endowed naturally with two different $A$-module
207: structures:
208:
209: (i) ({\it Diff}$_k(P,Q),\tau $) $\doteq {\rm Diff}_k(P,Q)$ (left)$,$
210:
211: $\tau :A\times ${\it Diff}$_k(P,Q)\longrightarrow ${\it Diff}$%
212: _k(P,Q):(a,\Delta )\longmapsto \tau (a,\Delta ):p\longmapsto a\Delta (p)$
213:
214: (ii) ({\it Diff}$_k(P,Q),\tau ^{+}$) $\doteq {\rm Diff}_k^{+}(P,Q)$ (right)$%
215: , $
216:
217: $\tau ^{+}:A\times ${\it Diff}$_k(P,Q)\longrightarrow ${\it Diff}$%
218: _k(P,Q):(a,\Delta )\longmapsto \tau ^{+}(a,\Delta ):p\longmapsto \Delta
219: (ap). $
220:
221: We will often write, to be concise, $\tau (a,\Delta )\equiv a\Delta $ and $%
222: \tau ^{+}(a,\Delta )\equiv a^{+}\Delta .$ It is easy to see that $(${\it Diff%
223: }$_k(P,Q),(\tau ,\tau ^{+}))\equiv \left( {\rm Diff}_k^{}(P,Q),{\rm Diff}%
224: _k^{+}(P,Q)\right) \doteq {\rm Diff}_k^{(+)}(P,Q)$ is an $A$-bimodule.\
225:
226: \begin{remark}
227: \label{r1}Since
228: \[
229: \delta _{a_0}(\Delta )\equiv 0\Leftrightarrow \Delta (a_0p)=a_0\Delta \left(
230: p\right) \text{ ,\quad }\forall \text{ }a_0\in A,\;\forall p\in P
231: \]
232: ${\rm Diff}_0(P,Q)$ and $Hom_A(P,Q)$ are identified as $A$-(bi)modules:
233: \[
234: Hom_A(P,Q)\simeq {\rm Diff}_0(P,Q)\simeq {\rm Diff}_0^{+}(P,Q).
235: \]
236: \end{remark}
237:
238: The obvious inclusion (of sets):
239: \[
240: {\rm Diff}_k(P,Q)\hookrightarrow {\rm Diff}_l(P,Q),\text{ }k\leq l
241: \]
242: induces a monomorphism of $A$-bimodules:
243: \[
244: {\rm Diff}_k^{(+)}(P,Q)\hookrightarrow {\rm Diff}_l^{(+)}(P,Q),\text{ }k\leq
245: l\text{;}
246: \]
247: the direct limit of the system in $A-{\bf BiMo{\rm D}}$ :
248:
249: \[
250: {\rm Diff}_0^{(+)}(P,Q)\hookrightarrow {\rm Diff}_1^{(+)}(P,Q)%
251: \hookrightarrow ...\hookrightarrow {\rm Diff}_n^{(+)}(P,Q)\hookrightarrow
252: ...
253: \]
254: is denoted by ${\rm Diff}^{(+)}(P,Q)=({\rm Diff}(P,Q),{\rm Diff}^{+}(P,Q))$.
255:
256: With a given $A$-module $P$ the following three functors are associated
257:
258: \[
259: \begin{array}{c}
260: {\rm Diff}_k:Q\longmapsto {\rm Diff}_k\left( P,Q\right) \\
261: {\rm Diff}_k^{+}:Q\longmapsto {\rm Diff}_k^{+}\left( P,Q\right) \\
262: {\rm Diff}_k^{\left( +\right) }:Q\longmapsto {\rm Diff}_k^{\left( +\right)
263: }\left( P,Q\right) .
264: \end{array}
265: \]
266: Let us put ${\rm Diff}_k^{\left( +\right) }\left( A,Q\right) \equiv {\rm Diff%
267: }_k^{\left( +\right) }Q$. Remark \ref{r1} implies ${\rm Diff}_0^{+}={\rm Diff%
268: }_0={\rm Id}_{A-{\bf MoD}}$.
269:
270: To simplify notations we will write ${\rm Diff}_{\sigma _1,\ldots ,\sigma
271: _n}^{+}$ instead of ${\rm Diff}_{\sigma _1}^{+}\circ \cdots \circ {\rm Diff}%
272: _{\sigma _n}^{+}$.
273:
274: \begin{definition}
275: For any $s,t\geq 0$,\label{alucce} define an $A$-module homomorphism
276: \[
277: {\em C}_{s,t}(P):{\rm Diff}_s^{+}({\rm Diff}_t^{+}P)\longrightarrow {\rm Diff%
278: }_{s+t}^{+}P
279: \]
280: by
281:
282: \[
283: {\em C}_{s,t}(P)(\Delta ):a\longmapsto \Delta (a)(1)\text{,\quad }\Delta \in
284: {\rm Diff}_s^{+}({\rm Diff}_t^{+}P).
285: \]
286: Then $P\longmapsto {\em C}_{s,t}(P)$ defines a morphism ${\rm Diff}%
287: _{s,t}^{+}\longrightarrow {\rm Diff}_{s+t\text{ }}^{+}$ of functors called
288: the {\em composition} or {\em ''gluing'' morphism}.
289: \end{definition}
290:
291: Note that ${\rm D}_{\left( k\right) }\left( Q\right) \doteq \left\{ \Delta
292: \in {\rm Diff}_kQ\mid \Delta \left( 1\right) =0\right\} $ is an $A$%
293: -submodule of ${\rm Diff}_kQ$ (but not of ${\rm Diff}_k^{+}Q$ !). The
294: functor ${\rm D}_{\left( k\right) }:Q\mapsto D_{\left( k\right) }\left(
295: Q\right) $ allows to form the following short exact sequence:
296:
297: \begin{equation}
298: 0\rightarrow {\rm D}_{\left( k\right) }\stackrel{i_k}{\longrightarrow }{\rm %
299: Diff}_k\stackrel{p_k}{\longrightarrow }{\rm Id}_{A-{\bf Mod}}\rightarrow 0
300: \label{du}
301: \end{equation}
302: in $\left[ A-{\bf Mod,}A-{\bf Mod}\right] $, where $i_k$ is the canonical
303: inclusion and $p_k$ is defined by:
304:
305: \[
306: p_k\left( Q\right) :{\rm Diff}_kQ\rightarrow Q:\Delta \mapsto \Delta \left(
307: 1\right) \text{\quad , }\Delta \in {\rm Diff}_kQ
308: \]
309: for any $A$-module $Q$. The functor monomorphism ${\rm Id}_{A-{\bf Mod}%
310: }\equiv {\rm Diff}_0\hookrightarrow {\rm Diff}_k$ splits (1), so that ${\rm %
311: Diff}_k={\rm D}_{\left( k\right) }\bigoplus {\rm Id}_{A-{\bf Mod}}$. Note
312: that ${\rm D}_{\left( 1\right) }\left( Q\right) $ is nothing but the $A$%
313: -module of all $Q$-valued $K$-linear {\it derivations} on $A$, denoted in
314: the literature usually by ${\rm Der}_{A/K}\left( Q\right) $ (see for example
315: \cite{Bou X}).
316:
317: Let $Q$ be an $A$-module and $P$, $P^{+}$ be the left and right $A$-modules
318: corresponding to an $A$-bimodule $P^{\left( +\right) }\equiv \left(
319: P,P^{+}\right) $. Let's denote by ${\rm Diff}_k^{\bullet }\left(
320: Q,P^{+}\right) $ (resp. ${\rm D}_{\left( k\right) }^{\bullet }\left(
321: P^{+}\right) $) the $A$-module which coincides with ${\rm Diff}_k\left(
322: Q,P^{+}\right) $ (resp. ${\rm D}_{\left( k\right) }\left( P^{+}\right) $) as
323: $K$-modules and whose $A$-module structures are inherited by that of $P$
324: \footnote{%
325: These $A$-module structures are well defined due to the fact that $\left(
326: P,P^{+}\right) \equiv P^{\left( +\right) }$ is a bimodule. Moreover one can
327: give similar definitions with $P^{+}$ replaced by $P$ i.e. using the
328: canonical involution $A-{\bf BiMod\longrightarrow }A-{\bf BiMod:}\left(
329: P,P^{+}\right) \rightarrow \left( P^{+},P\right) $.}:
330:
331: \begin{eqnarray*}
332: \text{(mult. by }a\text{ in }{\rm Diff}_k^{\bullet }\left( Q,P^{+}\right)
333: \text{)}\qquad \left( a^{\bullet }\Delta \right) \left( q\right) &\doteq
334: &a\Delta \left( q\right) \\
335: \text{(mult. by }a\text{ in }D_k^{\bullet }\left( P^{+}\right) \text{)}%
336: \qquad \left( a^{\bullet }\delta \right) \left( q\right) &\doteq &a\delta
337: \left( q\right)
338: \end{eqnarray*}
339: where both $a\Delta \left( q\right) $ and $a\delta \left( q\right) $ denote
340: the multiplication by $a$ in $P$. The correspondence
341:
342: \[
343: {\rm D}_{(k)}^{(\bullet )}:P^{(+)}\longmapsto (D_{(k)}(P^{+}),\text{ }%
344: D_{(k)}^{\bullet }(P^{+}))
345: \]
346: defines a functor $A-{\bf BiMod\longrightarrow }A-{\bf BiMod}$ in an obvious
347: way. If $Q=A$ we write ${\rm Diff}_k^{\bullet }\left( P^{+}\right) $ for $%
348: {\rm Diff}_k^{\bullet }\left( A,P^{+}\right) $. Obviously, ${\rm D}_{\left(
349: k\right) }^{\bullet }\left( P^{+}\right) $ is an $A$-submodule of ${\rm Diff}%
350: _k^{\bullet }(P^{+})$.
351:
352: Let us introduce now the {\it category of biads}, $*A-{\bf BiMod}$, whose
353: objects are ordered triples of $A$-modules:
354:
355: \[
356: (P,P^{+};Q)
357: \]
358: with $P^{\left( +\right) }\doteq (P,P^{+})$ being an $A$-bimodule and $Q$ an
359: $A$-submodule of $P$. The corresponding morphisms are those of underlying $A$%
360: -bimodules ''respecting'' the selected submodules, i.e.:
361:
362: \[
363: f:(P,P^{+})\longrightarrow (\overline{P},\overline{P}^{+})\text{ such that }%
364: f(Q)\subset \overline{Q}.
365: \]
366:
367: \begin{examples}
368: \label{e1}(i) If $P$ is an $A$-module and $s\leq k$, then $({\rm Diff}_kP,%
369: {\rm Diff}_k^{+}P;{\rm D}_{(s)}(P))$ is a biad for each $s\leq k$.
370:
371: (ii) If $^{}P^{\left( +\right) }\doteq (P,P^{+})$ is an $A$- bimodule, then
372: the following are biads:
373: \[
374: \begin{array}{c}
375: ({\rm Diff}_k^{\bullet }P^{+},{\rm Diff}_k^{+}P^{+};{\rm D}_{(k)}^{\bullet
376: }(P^{+})) \\
377: ({\rm Diff}_kP,{\rm Diff}_k^{+}P;{\rm D}_{(k)}(P)) \\
378: ({\rm Diff}_kP^{+},{\rm Diff}_k^{+}P^{+};{\rm D}_{(k)}(P^{+}))\text{.}
379: \end{array}
380: \]
381: \end{examples}
382:
383: We associate to a biad $(P^{(+)};Q)$ the following $K$-modules:
384:
385: \[
386: \begin{array}{c}
387: {\rm Diff}_k(Q\subset P^{+})\doteq \left\{ \Delta \in {\rm Diff}_kP^{+}\mid
388: im(\Delta )\subset Q\right\} \\
389: {\rm D}_{(k)}(Q\subset P^{+})\doteq \left\{ \Delta \in {\rm D}%
390: _{(k)}(P^{+})\mid im(\Delta )\subset Q\right\} .
391: \end{array}
392: \]
393: The $A$-module structure of ${\rm Diff}_k^{\bullet }(P^{+})$ (resp., of $%
394: {\rm Diff}_k^{+}(P^{+})$ or of ${\rm D}_{(k)}^{\bullet }(P^{+})$) induces an
395: $A$-module structure on ${\rm Diff}_k(Q\subset P^{+})$ (resp., on ${\rm Diff}%
396: _k^{+}(Q\subset P^{+})$, resp. on ${\rm D}_{(k)}(Q\subset P^{+})$), which is
397: denoted by ${\rm Diff}_k^{\bullet }(Q\subset P^{+})$ (resp. ${\rm Diff}%
398: _k^{+}(Q\subset P^{+})$ or ${\rm D}_{(k)}(Q\subset P^{+})$). So, the
399: following inclusions take place in $A-{\bf Mod}$
400:
401: \[
402: \begin{array}{c}
403: {\rm Diff}_k^{\bullet }(Q\subset P^{+})\subset {\rm Diff}_k^{\bullet }(P^{+})
404: \\
405: {\rm Diff}_k^{+}(Q\subset P^{+})\subset {\rm Diff}_k^{+}(P^{+}) \\
406: {\rm D}_{(k)}(Q\subset P^{+})\subset {\rm D}_{(k)}^{\bullet }(P^{+})\text{.}
407: \end{array}
408: \]
409: We have then a biad:
410:
411: \[
412: ({\rm Diff}_k^{\bullet }(Q\subset P^{+}),\text{ }{\rm Diff}_k^{+}(Q\subset
413: P^{+}),\text{ }{\rm D}_{(k)}(Q\subset P^{+}))
414: \]
415: which is a sub-biad of $({\rm Diff}_k^{\bullet }(P^{+}),$ ${\rm Diff}%
416: _k^{+}(P^{+}),$ ${\rm D}_{(k)}^{\bullet }(P^{+}))$. The following result is
417: straightforward\
418:
419: \begin{lemma}
420: \TeXButton{sissa}{\label{sissa}}\label{l2}\label{l2}\label{l2}\label{l2} If $%
421: \overline{P}^{(+)}$ is a sub-bimodule of $P^{(+)}$ and $Q\subset \overline{P}
422: $, then we have:
423:
424: \[
425: \begin{array}{c}
426: {\rm D}_{(k)}(Q\subset \overline{P}^{+})={\rm D}_{(k)}(Q\subset P^{+}) \\
427: {\rm Diff}_k^{+}(Q\subset \overline{P}^{+})={\rm Diff}_k^{+}(Q\subset P^{+})
428: \\
429: {\rm Diff}_k^{\bullet }(Q\subset \overline{P}^{+})={\rm Diff}_k^{\bullet
430: }(Q\subset P^{+}).
431: \end{array}
432: \]
433: \end{lemma}
434:
435: Canonical functors
436:
437: \[
438: A-{\bf Mod}\stackrel{i}{\bf \hookrightarrow }A-{\bf BiMod}
439: \begin{tabular}{l}
440: $\stackrel{j}{\bf \hookrightarrow }$ \\
441: $\stackunder{h}{\leftarrow }$%
442: \end{tabular}
443: {\bf *}A-{\bf BiMod}
444: \]
445: are defined as follows
446: \[
447: i:P\longmapsto \left( P,P\right) ,j:\left( P,P^{+}\right) \longmapsto \left(
448: P,P^{+};P\right) ,
449: \]
450: \[
451: h:\left( P,P^{+};Q\right) \longmapsto \left( P,P^{+}\right) \qquad \text{%
452: (forgetful).}
453: \]
454: Obviously, $i$ is fully faithful while $h$ is left inverse and adjoint to $j$%
455: . So, $j$ is fully faithful. Moreover there are three obvious forgetful
456: functors $p_1,p_2,p_3:$ ${\bf *}A-{\bf BiMod\rightarrow }A-{\bf Mod}$ with $%
457: p_1\left( P,P^{+};Q\right) =P,$ $p_2\left( P,P^{+};Q\right) =P^{+},$ $%
458: p_3\left( P,P^{+};Q\right) =Q$.
459:
460: We define now some {\em absolute functors} we need (see Definition \ref
461: {def
462: 1.18} for their ''relative'' version).
463:
464: \begin{definition}
465: \TeXButton{garp}{\label{garp}} For $k\geq 0$, $\underline{{\cal P}}_{(k)}:*A-%
466: {\bf BiMod}\longrightarrow *A-{\bf BiMod}$ is the functor
467:
468: \[
469: \underline{{\cal P}}_{(k)}:(P^{(+)};Q)\mapsto \left( {\rm Diff}_k^{\bullet
470: }(Q\subset P^{+}),\text{ }{\rm Diff}_k^{+}(Q\subset P^{+});\text{ }{\rm D}%
471: _{(k)}(Q\subset P^{+})\right)
472: \]
473: \ For $\sigma =\left( \sigma _1,...,\sigma _n\right) \in {\bf N}_{+}^n$,
474: define \underline{${\cal P}$}$_\sigma \equiv $\underline{${\cal P}$}$%
475: _{\left( \sigma _1,...,\sigma _n\right) }:*A-{\bf BiMod}\longrightarrow *A-%
476: {\bf BiMod}$ as the composition:
477:
478: \[
479: \underline{{\cal P}}_{\left( \sigma _1,...,\sigma _n\right) }\doteq
480: \underline{{\cal P}}_{\left( \sigma _1\right) }\circ \cdots \circ \underline{%
481: {\cal P}}_{\left( \sigma _n\right) }
482: \]
483: We put
484: \begin{eqnarray*}
485: {\cal P}_{\left( \sigma _1,...,\sigma _n\right) }^{\bullet }=p_1\circ
486: \underline{{\cal P}}_{\left( \sigma _1,...,\sigma _n\right) } \\
487: {\cal P}_{\left( \sigma _1,...,\sigma _n\right) }^{+}=p_2\circ \underline{%
488: {\cal P}}_{\left( \sigma _1,...,\sigma _n\right) } \\
489: {\rm D}_{\left( \sigma _1,...,\sigma _n\right) }^{\bullet }=p_3\circ
490: \underline{{\cal P}}_{\left( \sigma _1,...,\sigma _n\right) }.
491: \end{eqnarray*}
492:
493: Thanks to the full faithfulness of $i$ and $j$ mentioned above, we can
494: simply write $\underline{{\cal P}}_{\left( \sigma _1,...,\sigma _n\right) }$
495: for both $\underline{{\cal P}}_{\left( \sigma _1,...,\sigma _n\right) }\circ
496: j$ and $\underline{{\cal P}}_{\left( \sigma _1,...,\sigma _n\right) }\circ
497: j\circ i$ by specifying the source category only in case of a possible
498: confusion.
499: \end{definition}
500:
501: \begin{remark}
502: \TeXButton{rem 1.14}{\label{rem 1.14}} {\it (}{\em Words }{\rm \cite{Vi1}}%
503: {\it )} If $\tau \in {\bf N}_{+}^k$ and $\sigma \in {\bf N}_{+}^n$, the
504: functor
505:
506: \[
507: \underline{{\cal P}}_\tau ^{}({\rm D}_\sigma ^{\bullet }\subset {\cal P}%
508: _\sigma ^{+})\equiv \left( {\cal P}_\tau ^{\bullet }({\rm D}_\sigma
509: ^{\bullet }\subset {\cal P}_\sigma ^{+}),{\cal P}_\tau ^{+}({\rm D}_\sigma
510: ^{\bullet }\subset {\cal P}_\sigma ^{+});{\rm D}_\tau ^{\bullet }({\rm D}%
511: _\sigma ^{\bullet }\subset {\cal P}_\sigma ^{+})\right)
512: \]
513:
514: is obviously identified with $\underline{{\cal P}}_{(\tau _1)}\circ \ldots
515: \circ \underline{{\cal P}}_{(\tau _k)}\circ \underline{{\cal P}}_\sigma $ ,
516: i.e.
517:
518: \[
519: \underline{{\cal P}}_\tau ({\rm D}_\sigma \subset {\cal P}_\sigma ^{+})=%
520: \underline{{\cal P}}_{(\tau ,\sigma )}
521: \]
522: where $(\tau ,\sigma )\doteq (\tau _1,\ldots ,\tau _k,\sigma _1,\ldots
523: ,\sigma _n)\in {\bf N}_{+}^{n+k}$. This gives a ''closed'' functorial
524: meaning to ''words'' composed of functors defined above.
525: \end{remark}
526:
527: If $P$ is an $A$-module we write ${\rm D}_{\left( \sigma _1,\ldots ,\sigma
528: _n\right) }(P)$ for ${\rm D}_{\left( \sigma _1,\ldots ,\sigma _n\right)
529: }^{\bullet }(j\circ i\left( P\right) )$). It is not difficult to see that
530: \[
531: {\rm D}_{\left( \sigma _1,\ldots ,\sigma _n\right) }(P)={\rm D}_{\left(
532: \sigma _1\right) }\left( {\rm D}_{\left( \sigma _2,\ldots ,\sigma _n\right)
533: }(P)\subset {\rm Diff}_{\sigma _2,\ldots ,\sigma _n}^{+}(P)\right) .
534: \]
535: Furthermore,
536:
537: \[
538: {\rm D}_{\left( \sigma _1,\ldots ,\sigma _n\right) }(P)\hookrightarrow {\rm D%
539: }_{\left( \sigma _1\right) }^{\bullet }\left( {\rm Diff}_{\sigma _2,\ldots
540: ,\sigma _n}^{+}(P)\right) \hookrightarrow {\rm Diff}_{\sigma _1}^{\bullet
541: }\left( {\rm Diff}_{\sigma _2,\ldots ,\sigma _n}^{+}(P)\right)
542: \]
543: are inclusions in $A-{\bf Mod}$ , while the inclusion ${\rm D}_{\left(
544: \sigma _1,\ldots ,\sigma _n\right) }(P)\hookrightarrow {\rm Diff}_{\sigma
545: _1,\ldots ,\sigma _n}^{+}(P)$ is a DO of order $\leq \sigma _1$. So, the
546: functors ${\rm D}_\sigma $'s can be defined also inductively:
547:
548: \begin{definition}
549: Let $\sigma =\left( \sigma _1,\sigma _2,\ldots \sigma _n,\ldots \right) \in
550: {\bf N}_{+}^\infty $ and $\sigma \left( n\right) =\left( \sigma _1,\ldots
551: ,\sigma _n\right) $. Then functors ${\rm D}_{\sigma \left( n\right) }:A-{\bf %
552: Mod\rightarrow }A-{\bf Mod}$ are defined by induction:
553: \[
554: \begin{array}{c}
555: {\rm D}_{\sigma \left( 1\right) }\doteq {\rm D}_{\left( \sigma _1\right) }
556: \\
557: {\rm D}_{\sigma \left( n\right) }:P\longmapsto {\rm D}_{\left( \sigma
558: _1\right) }\left( {\rm D}_{\left( \sigma _2,\ldots ,\sigma _n\right) }\left(
559: P\right) \subset {\rm Diff}_{\sigma _2,\ldots ,\sigma _n}^{+}\left( P\right)
560: \right) .
561: \end{array}
562: \]
563: \end{definition}
564:
565: If $\sigma =(1,..1,1,...)$ we also write ${\rm D}_n$ for ${\rm D}_{\sigma
566: \left( n\right) }$.\
567:
568: For any $\sigma \in {\bf N}_{+}^\infty $ and $n\in {\bf N}_{+}$, the
569: sequence in $\left[ A-{\bf Mod},A-{\bf Mod}\right] $:
570:
571: \begin{equation}
572: 0\rightarrow {\rm D}_{\sigma \left( n\right) }\stackrel{{\rm I}_{\sigma
573: \left( n\right) }}{\longrightarrow }{\rm D}_{\sigma \left( n-1\right)
574: }^{\bullet }\circ {\rm Diff}_{\sigma _n}^{\left( +\right) }\stackrel{\pi
575: _{\sigma \left( n\right) }}{\longrightarrow }{\rm D}_{\left( \sigma
576: _1,\ldots ,\sigma _{n-2},\sigma _{n-1}+\sigma _n\right) } \label{(2)}
577: \end{equation}
578: where ${\rm I}_{\sigma \left( n\right) }$ is the natural inclusion and $\pi
579: _{\sigma \left( n\right) \text{ }}$ is the composition (see Definition \ref
580: {alucce})
581:
582: \[
583: {\rm D}_{\sigma \left( n-1\right) }^{\bullet }\circ {\rm Diff}_{\sigma
584: _n}^{\left( +\right) }\hookrightarrow {\rm D}_{\sigma \left( n-2\right)
585: }^{\bullet }\circ {\rm Diff}_{\sigma _{n-1}}^{\left( +\right) }\circ {\rm %
586: Diff}_{\sigma _n}^{+}\stackrel{{\rm D}_{\sigma \left( n-2\right) }^{\bullet
587: }\left( {\em C}_{\sigma _{n-1},\sigma _n}\right) }{\longrightarrow }{\rm D}%
588: _{\sigma \left( n-2\right) }^{\bullet }\circ {\rm Diff}_{\sigma
589: _{n-1}+\sigma _n}^{\left( +\right) }\text{ ,}
590: \]
591: is exact.
592:
593: \begin{remark}
594: \label{new}Let $\sigma =\left( \sigma _1,\sigma _2,\ldots \sigma _n\right)
595: \in {\bf N}_{+}^n$. We have a canonical split exact short exact sequence in $%
596: \left[ A-{\bf Mod},A-{\bf Mod}\right] $%
597: \[
598: 0\rightarrow {\rm D}_\sigma \stackrel{l_\sigma }{\stackunder{\rho _\sigma }{%
599: \rightleftarrows }}{\cal P}_\sigma ^{\bullet }\stackrel{\psi _\sigma }{%
600: \longrightarrow }{\rm D}_{\left( \sigma _2,\ldots ,\sigma _n\right)
601: }\rightarrow 0
602: \]
603: where $l_\sigma $ is the canonical inclusion, $\psi _\sigma $ is given by ($%
604: P $ is an $A$-module)
605: \[
606: \psi _\sigma (P)(\Delta )=\Delta \left( 1\right)
607: \]
608: and $\rho _\sigma $ by
609:
610: \[
611: \rho _\sigma \left( P\right) \left( \Delta \right) =\Delta -\Delta \left(
612: 1\right) .
613: \]
614: Hence ${\cal P}_\sigma ^{\bullet }\simeq {\rm D}_{\left( \sigma _2,\ldots
615: ,\sigma _n\right) }\oplus {\rm D}_\sigma $.
616: \end{remark}
617:
618: \smallskip\ \
619:
620: Now we define the {\em relative} (i.e. relative to an arbitrary $A$-module $%
621: P $) {\em functors}.
622:
623: \begin{definition}
624: \TeXButton{def 1.18}{\label{def 1.18}}If $k\geq 0$ and $P$ is an $A$-module,
625: define
626: \[
627: \underline{{\cal P}}_{\left( k\right) }\left[ P\right] :*A-{\bf %
628: BiMod\longrightarrow }A-{\bf BiMod}
629: \]
630: to be the functor
631:
632: \[
633: \left( Q,Q^{+};S\right) \longmapsto \left( {\rm Diff}_k^{\bullet }\left(
634: P,S\subset Q^{+}\right) ,{\rm Diff}_k^{+}\left( P,S\subset Q^{+}\right)
635: \right) .
636: \]
637: If $\sigma \left( n\right) =\left( \sigma _1,\ldots ,\sigma _n\right) \in
638: {\bf N}_{+}^n$, $n>1$, define the functor
639: \[
640: \underline{{\cal P}}_{\sigma \left( n\right) }\left[ P\right] :*A-{\bf %
641: BiMod\longrightarrow }\left( A,A\right) -{\bf BiMod}
642: \]
643: as the composition
644: \[
645: \underline{{\cal P}}_{\sigma \left( n\right) }\left[ P\right] \doteq
646: \underline{{\cal P}}_{\left( \sigma _1\right) }\left[ P\right] \circ
647: \underline{{\cal P}}_{\left( \sigma _2,...,\sigma _n\right) }.
648: \]
649: As in the absolute case (Def. \ref{garp}), we set
650: \begin{eqnarray*}
651: \underline{{\cal P}}_{\sigma \left( n\right) }^{\bullet }\left[ P\right]
652: \doteq p_1\circ \underline{{\cal P}}_{\sigma \left( n\right) }\left[ P\right]
653: \\
654: \underline{{\cal P}}_{\sigma \left( n\right) }^{+}\left[ P\right] \doteq
655: p_2\circ \underline{{\cal P}}_{\sigma \left( n\right) }\left[ P\right]
656: \end{eqnarray*}
657: and still denote by $\underline{{\cal P}}_{\sigma \left( n\right) }\left[
658: P\right] $ both $\underline{{\cal P}}_{\sigma \left( n\right) }\left[
659: P\right] \circ j$ and $\underline{{\cal P}}_{\sigma \left( n\right) }\left[
660: P\right] \circ j\circ i$.
661: \end{definition}
662:
663: By Lemma \ref{sissa}, we have $\left( {\cal P}_{\sigma \left( n\right)
664: }^{\bullet },{\cal P}_{\sigma \left( n\right) }^{+}\right) \equiv \left(
665: {\cal P}_\sigma ^{\bullet }\left[ A\right] ,{\cal P}_\sigma ^{+}\left[
666: A\right] \right) $. Moreover, $\underline{{\cal P}}_{\sigma \left( n\right)
667: }\left[ P\right] $ is (controvariantly) functorial in $P$ and we have
668:
669: \begin{lemma}
670: \TeXButton{split}{\label{split}}If $0\rightarrow P_1\stackrel{f}{%
671: \longrightarrow }P_2\stackrel{g}{\longrightarrow }P_3\rightarrow 0$ is exact
672: (resp. split exact) in $A-{\bf Mod}$, then
673: \begin{eqnarray*}
674: 0\rightarrow \underline{{\cal P}}_{\sigma \left( n\right) }^{\bullet }\left[
675: P_3\right] \stackrel{g^{\vee }}{\longrightarrow }\underline{{\cal P}}%
676: _{\sigma \left( n\right) }^{\bullet }\left[ P_2\right] \stackrel{f^{\vee }}{%
677: \longrightarrow }\underline{{\cal P}}_{\sigma \left( n\right) }^{\bullet
678: }\left[ P_1\right] \\
679: 0\rightarrow \underline{{\cal P}}_{\sigma \left( n\right) }^{+}\left[
680: P_3\right] \stackrel{g^{\vee }}{\longrightarrow }\underline{{\cal P}}%
681: _{\sigma \left( n\right) }^{+}\left[ P_2\right] \stackrel{f^{\vee }}{%
682: \longrightarrow }\underline{{\cal P}}_{\sigma \left( n\right) }^{+}\left[
683: P_1\right]
684: \end{eqnarray*}
685: are exact (resp.
686:
687: \begin{eqnarray*}
688: 0\rightarrow \underline{{\cal P}}_{\sigma \left( n\right) }^{\bullet }\left[
689: P_3\right] \stackrel{g^{\vee }}{\longrightarrow }\underline{{\cal P}}%
690: _{\sigma \left( n\right) }^{\bullet }\left[ P_2\right] \stackrel{f^{\vee }}{%
691: \longrightarrow }\underline{{\cal P}}_{\sigma \left( n\right) }^{\bullet
692: }\left[ P_1\right] \rightarrow 0 \\
693: 0\rightarrow \underline{{\cal P}}_{\sigma \left( n\right) }^{+}\left[
694: P_3\right] \stackrel{g^{\vee }}{\longrightarrow }\underline{{\cal P}}%
695: _{\sigma \left( n\right) }^{+}\left[ P_2\right] \stackrel{f^{\vee }}{%
696: \longrightarrow }\underline{{\cal P}}_{\sigma \left( n\right) }^{+}\left[
697: P_1\right] \rightarrow 0
698: \end{eqnarray*}
699: are exact).
700: \end{lemma}
701:
702: \TeXButton{Proof}{\proof} Straightforward, by induction on $n$.
703: \TeXButton{End Proof}{\endproof}
704:
705: If $P$ is an $A$-module, there are exact sequences in $[A-{\bf Mod},A-{\bf %
706: Mod}]$
707:
708: \begin{equation}
709: 0\rightarrow {\cal P}_{\sigma \left( n\right) }^{\bullet }\left[ P\right]
710: \hookrightarrow \stackunder{}{\cal P}_{\sigma \left( n-1\right) }^{\bullet
711: }\left[ P\right] \circ {\rm Diff}_{\sigma _n}^{\left( +\right) }\stackrel{%
712: q_{\sigma \left( n\right) }}{\longrightarrow }{\cal P}_{\left( \sigma
713: _1,\ldots \sigma _{n-2},\sigma _{n-1}+\sigma _n\right) }^{\bullet }\left[
714: P\right] \label{relHol}
715: \end{equation}
716: where the monomorphism is the natural inclusion while $q_{\sigma (n)}$ is
717: induced by the ''gluing'' morphism with respect to the pair of indexes $%
718: \left( \sigma _{n-1},\sigma _n\right) $, i.e.:
719:
720: \begin{eqnarray*}
721: \stackunder{}{\cal P}_{\left( \sigma _1,\ldots ,\sigma _{n-1}\right)
722: }^{\bullet }\left[ P\right] \circ {\rm Diff}_{\sigma _n}^{\left( +\right)
723: }\left( Q\right) &\ni &\Delta \longmapsto q_{\sigma (n)}(\Delta )=\overline{%
724: \Delta }\in {\cal P}_{\left( \sigma _1,\ldots \sigma _{n-2},\sigma
725: _{n-1}+\sigma _n\right) }^{\bullet }\left[ P\right] \left( Q\right) \\
726: \left( \ldots \left( \left( \overline{\Delta }\left( p\right) \right) \left(
727: a_1\right) \right) \cdots \right) \left( a_{n-2}\right) &\doteq &\left(
728: \left( \ldots \left( \left( \Delta \left( p\right) \right) \left( a_1\right)
729: \right) \cdots \right) \left( a_{n-2}\right) \right) \left( 1\right) ,
730: \end{eqnarray*}
731: where $p\in P$ and $a_1,...,a_{n-2}\in A$. We have analogous exact sequences
732: in $[A-{\bf Mod},A-{\bf Mod}]$:
733:
734: \begin{equation}
735: 0\rightarrow {\cal P}_{\sigma \left( n\right) }^{\bullet }\left[ P\right]
736: \hookrightarrow {\rm Diff}_{\sigma _1}^{\bullet }\left( P,\cdot \right)
737: \circ {\cal P}_{\left( \sigma _2,\ldots ,\sigma n\right) }^{+}\left[
738: P\right] \stackrel{g_{\sigma (n)}}{\longrightarrow }{\cal P}_{\left( \sigma
739: _1+\sigma _2,\sigma _3,\ldots ,\sigma _n\right) }^{\bullet }\left[ P\right]
740: \label{relantiHol}
741: \end{equation}
742: where $g_{\sigma (n)}:\Delta \mapsto \widehat{\Delta }$ with $\widehat{%
743: \Delta }\left( p\right) \doteq \Delta \left( p\right) \left( 1\right) $, $%
744: p\in P$ (i.e. we ''glue'' with respect to the first two indexes); the upper
745: boldface dot in ${\rm Diff}_{\sigma _1}^{\bullet }\left( P,\cdot \right) $
746: denotes the $A$-module structure induced by ${\cal P}_{\left( \sigma
747: _2,\ldots ,\sigma _n\right) }^{\bullet }\left[ P\right] $.
748:
749: The following definition will allow us to be concise in the next Section:
750:
751: \begin{definition}
752: For any $n>0$ and any $\sigma \in {\bf N}_{+}^n$, the functors (in $[A-{\bf %
753: Mod},A-{\bf Mod}]$) ${\cal P}_\sigma ^{\bullet }$, $D_\sigma ^{\bullet }$
754: are called the {\em relevant absolute functors} while, if $P$ is an $A$%
755: -module, the functors ${\cal P}_\sigma ^{\bullet }\left[ P\right] $ , are
756: called the {\em relevant functors relative} to the{\it \ }$A$-module $P$.
757: \end{definition}
758:
759: \section{Absolute and relative representative objects}
760:
761: In this Section we consider (strict) representative objects of the functors
762: introduced in the previous Section. We obtain, as particular cases, the
763: standard modules of K\"ahler differential forms of Algebraic Geometry and
764: the de Rham forms of Differential Geometry. We emphasize that in our
765: approach all these (and not only those of degree one) are obtained as
766: representative objects of suitable functors. One of the major advantages of
767: this approach is to allow natural generalizations.
768:
769: Let ${\frak D}$ be a full subcategory of $A-{\bf Mod}$. We denote by $A-{\bf %
770: BiMod}_{{\frak D}}$ the subcategory of $A-{\bf BiMod}$ whose objects are
771: couples of objects of ${\frak D}$ and by $*A-{\bf BiMod}_{{\frak D}}$ the
772: subcategory of $*A-{\bf BiMod}$ consisting of triples whose elements are
773: objects of ${\frak D}$ (Section 1).
774:
775: \begin{definition}
776: \TeXButton{marker def 2.1}{\label{def 2.1}}A full {\em abelian} subcategory $%
777: {\frak D}$ of $A-{\bf Mod}$ is said to be {\em differentially closed} if the
778: following properties are satisfied:
779:
780: {\it (a)} each functor defined in the previous Section, when restricted to $%
781: {\frak D}$ (resp. $A-{\bf BiMod}_{{\frak D}}$, resp. $*A-{\bf BiMod}_{{\frak %
782: D}}$) has values in ${\frak D}$ (resp. $A-{\bf BiMod}_{{\frak D}}$, resp. $%
783: *A-{\bf BiMod}_{{\frak D}}$);
784:
785: {\it (b)} if $T:A-{\bf Mod\rightarrow }$ $A-{\bf Mod}$ is a {\em relevant
786: absolute} functor or a {\em relevant relative} functor, {\em relative to an
787: object of} ${\frak D}$, then $T_{\mid {\frak D}}:{\frak D}\longrightarrow
788: {\frak D}$ is strictly representable in ${\frak D}$;
789:
790: {\it (c)} $A\in Ob\left( {\frak D}\right) $;
791:
792: {\it (d)} ${\frak D}$ is closed under tensor product (over $A$);
793:
794: (e) ${\frak D}$ is closed under taking subobjects (i.e. if $P\subseteq Q$ in
795: $A-{\bf Mod}$ and $Q$ is in ${\frak D}$ then $P$ is in ${\frak D}$).
796: \end{definition}
797:
798: Condition {\it (a)} is needed to have an ambient category which is
799: ''closed'' with respect to functorial differential calculus; as it will be
800: clear in the following, since among the functors of Section 1 there are also
801: compositions of relevant ''elementary'' ones, we would like that
802: representative objects of these nonelementary functors\footnote{%
803: Relevant or not.} (for example ${\rm D}_{(s)}^{\bullet }\circ {\rm Diff}%
804: _t^{\left( +\right) }$), if existing, {\it could be expressed in terms} of
805: representative objects of the relevant ''elementary'' ones (${\rm D}%
806: _{(s)}^{} $ and ${\rm Diff}_t$ in the example). Condition ({\it d}) makes it
807: possible.
808:
809: ${\frak D}$ being abelian and satisfying ({\it b}), exactness of sequences
810: of strictly representable functors yields exactness of the ''dual''
811: sequences of representative objects in ${\frak D}$. Condition (e) is related
812: to the existence of canonical generators for some representative objects and
813: will become clear in the sequel. Note also that (e) implies that if $f$ is a
814: morphism in ${\frak D}$, $im\left( f\right) $ (resp. $\ker \left( f\right) $%
815: ) is the same when considered in ${\frak D}$ or in $A-{\bf Mod}$.
816:
817: Let us recall some elementary facts about bimodules, mainly to fix our
818: notations.
819:
820: If $P^{(+)}=(P,P^{+})$ is an $A$-bimodule and $a\in A$, we write $a$ for the
821: multiplication in $P$ and $a^{+}$ for the multiplication in $P^{+}$. If $Q$
822: is an $A$-module we denote by:
823:
824: ({\bf I}) $P^{+}\otimes _A^{\bullet }Q$ the $A$-module obtained by the
825: abelian group $P^{+}\otimes _AQ$ with multiplication by elements of $A$
826: defined as
827:
828: \[
829: a^{\bullet }(p\otimes q)\doteq (ap)\otimes q\text{, \quad }a\in A,\text{ }%
830: p\in P^{+},\text{ }q\in Q
831: \]
832: (note that $a^{\bullet }(p\otimes q)\neq p\otimes aq$). Then $P^{+}\otimes
833: _A^{(\bullet )}Q$ $\doteq \left( P^{+}\otimes _A^{\bullet }Q,P^{+}\otimes
834: _AQ\right) $ is an $A$-bimodule;
835:
836: ({\bf II}) $Hom_A^{\bullet }(Q,P^{+})$ the $A$-module obtained by the
837: abelian group $Hom_A(Q,P^{+})$ with multiplication by elements of $A$
838: defined as :
839:
840: \[
841: \left[ a^{\bullet }f\right] (p)\doteq a_0\cdot (f(p))\text{, }\quad a\in A,%
842: \text{ }p\in P,\text{ }f\in Hom_A(Q,P^{+}).
843: \]
844: Denote by $Hom_A^{(\bullet )}(Q,P^{+})\doteq \left( Hom_A^{\bullet
845: }(Q,P^{+}),Hom_A(Q,P^{+})\right) $ the corresponding $A$-bimodule;
846:
847: ({\bf III}) $Hom_A^{+}(P,Q)$ the $A$-module obtained by the abelian group $%
848: Hom_A(P,Q)$ with multiplication by elements of $A$:
849: \[
850: \left[ a^{+}f\right] (p)\doteq f(a^{+}p).
851: \]
852: Then
853: \[
854: Hom_A^{(+)}(P,Q)\doteq \left( Hom_A(P,Q),Hom_A^{+}(P,Q)\right)
855: \]
856: is an $A$-bimodule.
857:
858: In the same way we can define the $A$-modules $P\otimes _A^{+}Q$ , $%
859: Hom_A^{+}(P,Q)$ and $Hom_A^{\bullet }(P^{+},Q)$.
860:
861: \begin{example}
862: \TeXButton{marker es 2.2}{\label{es 2.2}}\label{c6}If $P$ and $Q$ are $A$%
863: -modules, we have an isomorphism in \ $\left[ A-{\bf Mod,}A-{\bf Mod}\right]
864: $
865:
866: \[
867: {\rm Diff}_s(\cdot ,Q)\simeq Hom_A^{\bullet }\left( \cdot ,{\rm Diff}%
868: _s^{+}Q\right) \text{.}
869: \]
870: \end{example}
871:
872: It is not diificult to prove the following\
873:
874: \begin{lemma}
875: \label{l3}\label{l3}\TeXButton{marker lem 2.3}{\label{lem 2.3}}Let $R$ and $%
876: P $ be $A$-modules and $(Q,Q^{+})$ an $A$-bimodule. Then we have a canonical
877: isomorphism in $A-{\bf BiMod}$:
878: \[
879: Hom_A^{\left( \bullet \right) }\left( R,Hom_A^{+}(Q,P)\right) \stackrel{%
880: \varphi }{\simeq }Hom_A^{(+)}(Q^{+}\otimes _A^{\bullet }R,P)
881: \]
882: \end{lemma}
883:
884: \begin{proposition}
885: \label{jet}Let $P$ be an $A$-module and $k\in {\bf N}_{+}$. Then ${\rm Diff}%
886: _k\left( P,\cdot \right) :A-{\bf Mod\rightarrow }$ $A-{\bf Mod}$ is strictly
887: representable by the so-called $k${\em -jet module} ${\bf J}^k\left(
888: P\right) $.
889: \end{proposition}
890:
891: \TeXButton{Proof}{\proof} See \cite{KLV}, p. 12. \TeXButton{End Proof}
892: {\endproof}\
893:
894: In other words, there exists a universal DO $j_k(P):P\longrightarrow {\bf J}%
895: ^k(P)$, of order $\leq k$ (often denoted simply by $j_k$), such that for
896: each DO $\Delta :P\rightarrow Q$ of order $\leq k$, there is a unique $A$%
897: -homomorphism $f^\Delta $ and a commutative diagram
898:
899: \[
900: \begin{array}{ccccc}
901: P & \stackrel{j_k(P)}{\longrightarrow } & {\bf J}^k(P) & & \\
902: & \stackunder{\Delta }{\searrow } & \downarrow & f^\Delta & \\
903: & & Q & &
904: \end{array}
905: \text{.}
906: \]
907: The $A$-module ${\bf J}^k(P)$ is generated by $\left\{ j_k(p)\mid p\in
908: P\right\} $. Moreover, ${\bf J}^k(P)$ has a bimodule structure ${\bf J}%
909: _{\left( +\right) }^k\left( P\right) =\left( {\bf J}^k\left( P\right) ,{\bf J%
910: }_{+}^k\left( P\right) \right) $ which can be described in the following,
911: purely functorial, way. Suppose ${\frak D}\subseteq A-{\bf Mod}$ is a
912: subcategory such that $\forall P\in Ob\left( {\frak D}\right) $ the functor $%
913: {\rm Diff}_k\left( P,\cdot \right) $ when restricted to ${\frak D}$ has
914: values in ${\frak D}$ and is strictly representable in ${\frak D}$ (e.g. $%
915: {\frak D}=A-{\bf Mod}$ by proposition \ref{lem 2.3}). Let ${\bf J}_{{\frak D}%
916: }^k\left( P\right) $ be the corresponding representative object i.e. $%
917: Hom_A\left( {\bf J}_{{\frak D}}^k\left( P\right) ,\cdot \right) \simeq {\rm %
918: Diff}_k\left( P,\cdot \right) $ in $\left[ {\frak D},{\frak D}\right] $. If $%
919: j_k^{{\frak D}}:P\rightarrow {\bf J}_{{\frak D}}^k\left( P\right) $
920: corresponds to the identity morphism of ${\bf J}_{{\frak D}}^k\left(
921: P\right) $, we can define, for each $a\in A$, the DO
922: \[
923: a^{+}:P\rightarrow {\bf J}_{{\frak D}}^k\left( P\right) :p\longmapsto j_k^{%
924: {\frak D}}\left( ap\right) \text{.\quad }
925: \]
926: The corresponding $A$-endomorphism of ${\bf J}_{{\frak D}}^k$ is still
927: denoted by $a^{+}$ and gives the required second $A$-module structure ${\bf J%
928: }_{{\frak D}+}^k$ on the abelian group ${\bf J}_{{\frak D}}^k$.
929:
930: Using the bimodule ${\bf J}_{(+)}^k$ and Proposition \ref{jet}, we get an
931: isomorphism in
932:
933: $\left[ A-{\bf Mod,}A-{\bf Mod}\right] $
934:
935: \begin{equation}
936: {\rm Diff}_k^{+}\simeq Hom_A^{+}({\bf J}^k,\cdot )\text{.} \label{plus}
937: \end{equation}
938:
939: If $P,Q$ are $A$-modules, then:
940:
941: \[
942: Hom_A({\bf J}_{+}^k\otimes _A^{\bullet }P,Q)\simeq Hom_A^{\bullet
943: }(P,Hom_A^{+}({\bf J}^k,Q))
944: \]
945: by Lemma \ref{l3}; therefore
946:
947: \[
948: Hom_A^{\bullet }(P,Hom_A^{+}({\bf J}^s,Q))\simeq Hom_A^{\bullet }(P,{\rm Diff%
949: }_s^{+}Q)\text{. }
950: \]
951: and, by (\ref{plus}) and Example \ref{es 2.2}, we finally get:
952:
953: \[
954: Hom_A^{\bullet }(P,{\rm Diff}_s^{+}Q)\simeq {\rm Diff}_s(P,Q)\text{. }
955: \]
956: Since representative objects of the same functor are canonically isomorphic,
957: we have proved:
958:
959: \begin{lemma}
960: \TeXButton{lem 2.5}{\label{lem 2.5}}\label{ozu}There are canonical
961: isomorphisms in $\left[ A-{\bf Mod,}A-{\bf Mod}\right] $:
962: \[
963: \begin{array}{c}
964: {\bf J}^k(\cdot )\simeq {\bf J}_{+}^k\otimes _A^{\bullet }\left( \cdot
965: \right) \\
966: {\bf J}_{+}^k(\cdot )\simeq {\bf J}_{+}^k\otimes _A\left( \cdot \right)
967: \text{ .}
968: \end{array}
969: \]
970: \end{lemma}
971:
972: We are now able to prove a basic result
973:
974: \begin{lemma}
975: \TeXButton{lem 2.6}{\label{lem 2.6}}({\bf a}) If $\tau \in {\bf N}%
976: _{+}^\infty $, $n>0$, $t\geq 0$ and ${\rm D}_{\tau \left( n\right) }$ is
977: strictly representable in $A-{\bf Mod}$ by ${\bf \Lambda }^{\tau \left(
978: n\right) }$, then ${\rm D}_{\tau \left( n\right) }^{\bullet }\circ {\rm Diff}%
979: _t^{\left( +\right) }$ is strictly representable in $A-{\bf Mod}$ by ${\bf J}%
980: ^t\left( {\bf \Lambda }^{\tau \left( n\right) }\right) $;
981:
982: \label{l5}({\bf b}) If $s,t\geq 0$, then
983: \[
984: {\cal P}_{(s)}^{\bullet }\circ {\rm Diff}_t^{\left( +\right) }\equiv {\rm %
985: Diff}_s^{\bullet }\circ {\rm Diff}_t^{\left( +\right) }:A-{\bf %
986: Mod\rightarrow }A-{\bf Mod}
987: \]
988: is strictly representable by ${\bf J}^t({\bf J}^s).$
989:
990: ({\bf c}) If $s,t\geq 0$ and $P$ is an $A$-module, then
991: \[
992: {\cal P}_{(s)}^{\bullet }\left[ P\right] \circ {\rm Diff}_t^{\left( +\right)
993: }\equiv {\rm Diff}_s^{\bullet }\left( P,{\rm Diff}_t^{\left( +\right)
994: }\left( \cdot \right) \right) :A-{\bf Mod\rightarrow }A-{\bf Mod}
995: \]
996: is strictly representable by ${\bf J}^t({\bf J}^s\left( P\right) )$.
997:
998: ({\bf d}) If ${\cal P}_{\sigma (n)}^{\bullet }$ is strictly representable by
999: ${\rm Hol}^{\sigma \left( n\right) }$, then
1000: \[
1001: {\cal P}_{\sigma (n)}^{\bullet }\circ {\rm Diff}_k^{(+)}:A-{\bf %
1002: Mod\rightarrow }A-{\bf Mod}
1003: \]
1004: is strictly representable by ${\bf J}^k\left( {\rm Hol}^{\sigma \left(
1005: n\right) }\right) $;
1006:
1007: ({\bf e}) If $P$ is an $A$-module and $\stackunder{}{\cal P}_{\sigma
1008: (n)}^{\bullet }\left[ P\right] $ is strictly representable by ${\rm Hol}%
1009: ^{\sigma \left( n\right) }\left[ P\right] $, then
1010: \[
1011: {\cal P}_{\sigma (n)}^{\bullet }\left[ P\right] \circ {\rm Diff}_k^{(+)}:A-%
1012: {\bf Mod\rightarrow }A-{\bf Mod}
1013: \]
1014: is strictly representable by ${\bf J}^k\left( {\rm Hol}^{\sigma \left(
1015: n\right) }\left[ P\right] \right) $;
1016: \end{lemma}
1017:
1018: \TeXButton{Proof}{\proof} The proofs are very similar. We prove only \label
1019: {l5}({\bf b}) and ({\bf d}).
1020:
1021: ({\bf b})
1022: \[
1023: {\rm Diff}_s^{\bullet }({\rm Diff}_t^{+}P)\simeq Hom_A^{\bullet }({\bf J}^s,%
1024: {\rm Diff}_t^{+}P)\simeq Hom_A^{\bullet }({\bf J}^s,Hom_A^{+}({\bf J}^t,P))%
1025: \text{ ;}
1026: \]
1027: by (\ref{plus}) this is isomorphic to $Hom_A({\bf J}_{+}^t\otimes ^{\bullet }%
1028: {\bf J}^s,P)$ and, finally by Proposition \ref{ozu}, to $Hom_A({\bf J}^t(%
1029: {\bf J}^s),P)$.
1030:
1031: ({\bf d})
1032: \begin{eqnarray*}
1033: {\cal P}_{\sigma (n)}^{\bullet }\left( {\rm Diff}_k^{(+)}\left( P\right)
1034: \right) &\simeq &Hom_A^{\bullet }({\rm Hol}^{\sigma \left( n\right) },{\rm %
1035: Diff}_t^{+}P)\simeq Hom_A({\rm Hol}^{\sigma \left( n\right) },Hom_A^{+}({\bf %
1036: J}^k,P))\simeq \\
1037: \ &\simeq &Hom_A({\bf J}_{+}^k\otimes ^{\bullet }{\rm Hol}^{\sigma \left(
1038: n\right) },P)\simeq Hom_A\left( {\bf J}^k\left( {\rm Hol}^{\sigma \left(
1039: n\right) }\right) ,P\right) \text{.}
1040: \end{eqnarray*}
1041:
1042: \qquad \TeXButton{End Proof}{\endproof}
1043:
1044: Note that if ${\rm D}_\sigma $ is representable for any $\sigma \in {\bf N}%
1045: _{+}^n$, $n>0$ then ${\cal P}_\sigma ^{\bullet }$ is representable for any $%
1046: \sigma \in {\bf N}_{+}^n$, $n>0$ by Remark \ref{new}.
1047:
1048: \begin{remark}
1049: Note that, for example, ${\rm Diff}_k^{\bullet }\circ {\rm Diff}_l^{\left(
1050: +\right) }\circ {\rm Diff}_m^{+}$ is representable but not strictly
1051: representable in $A-{\bf Mod}$:
1052:
1053: \begin{eqnarray*}
1054: {\rm Diff}_k^{\bullet }({\rm Diff}_l^{+}\left( {\rm Diff}_m^{+}P\right)
1055: )\simeq Hom_A({\bf J}^l({\bf J}^k),{\rm Diff}_m^{+}P)\simeq \\
1056: \ \simeq Hom_A\left( {\bf J}^l({\bf J}^k),Hom_A^{+}\left( {\bf J}^m,P\right)
1057: \right) \simeq Hom_A^{+}\left( {\bf J}_{+}^m\otimes _A^{\bullet }{\bf J}^l(%
1058: {\bf J}^k),P\right) \simeq \\
1059: \ \simeq Hom_A^{+}\left( {\bf J}^m\left( {\bf J}^l({\bf J}^k)\right)
1060: ,P\right) \text{.}
1061: \end{eqnarray*}
1062: \end{remark}
1063:
1064: \smallskip\
1065:
1066: We conclude this preliminaries with the following elementary result
1067:
1068: \begin{lemma}
1069: {\rm (''Third-representable'' lemma)} \label{terzo}Let $0\rightarrow T_1%
1070: \stackrel{i}{\rightarrow }T_2\stackrel{\varphi }{\rightarrow }T_3$ be an
1071: exact sequence in $\left[ A-{\bf Mod,}A-{\bf Mod}\right] $, with $T_2$ and $%
1072: T_3$ strictly representable by $\tau _2$ and $\tau _3$, respectively; then $%
1073: T_1$ is also strictly representable by:
1074: \[
1075: \frac{\tau _2}{\varphi ^{\vee }(\tau _3)}\text{,}
1076: \]
1077: where $\varphi ^{\vee }:\tau _3\rightarrow \tau _2$ is the
1078: dual-representative of $\varphi $.
1079: \end{lemma}
1080:
1081: \TeXButton{Proof}{\proof} If $P$ is an $A$-module, the morphism
1082:
1083: \[
1084: \begin{array}{c}
1085: \chi _P:T_1(P)\rightarrow Hom_A(\frac{\tau _2}{\varphi ^{\vee }(\tau _3)},P)
1086: \\
1087: q\longmapsto \chi _p(q):\left[ t_2\right] _{{\bf mod}\varphi ^{\vee }(\tau
1088: _3)}\longmapsto \stackrel{\wedge }{q}(t_2)
1089: \end{array}
1090: \]
1091: where $\stackrel{\wedge }{q}\doteq i(P)(q)$, is well defined (since $%
1092: \stackrel{\wedge }{q}\circ \varphi ^{\vee }=\varphi (P)(\stackrel{\wedge }{q}%
1093: )=0$) and is an isomorphism, natural in $P$.\TeXButton{End Proof}{\endproof}
1094:
1095: The next Theorem, collecting some of the results above, asserts that $A-{\bf %
1096: Mod}${\it \ is itself differentially closed }(see Definition{\it \ }\ref
1097: {def
1098: 2.1}).
1099:
1100: \begin{theorem}
1101: \TeXButton{thm 2.7}{\label{thm 2.7}}Let $P$ be an $A$-module, $\sigma \in $ $%
1102: {\bf N}_{+}^\infty $ and $k\in {\bf N}$. Then:
1103:
1104: ({\bf i}) ${\rm Diff}_k\left( P,\cdot \right) $ is strictly representable in
1105: $A-{\bf Mod}$ by the $k${\em -jet module} ${\bf J}^k\left( P\right) $;
1106:
1107: ({\bf ii}) for each $n>0$, the functor ${\rm D}_{\sigma \left( n\right) }$
1108: is strictly representable in $A-{\bf Mod}$ by the so-called {\em higher de
1109: Rham forms}{\it ' module of type} $\sigma \left( n\right) $, ${\bf \Lambda }%
1110: ^{\sigma \left( n\right) }$;
1111:
1112: ({\bf iii}) ${\cal P}_{\sigma (n)}^{\bullet }$ and ${\cal P}_{\sigma
1113: (n)}^{\bullet }\left[ P\right] $ $:A-{\bf Mod\rightarrow }$ $A-{\bf Mod}$
1114: are strictly representable by the so-called {\em absolute holonomy module}
1115: of type $\sigma \left( n\right) $, ${\rm Hol}^{\sigma \left( n\right) }$ and
1116: {\em relative holonomy module} of type $\sigma \left( n\right) $, ${\rm Hol}%
1117: ^{\sigma \left( n\right) }\left[ P\right] $.
1118: \end{theorem}
1119:
1120: \TeXButton{Proof}{\proof} ({\bf i}) is Proposition \ref{jet}.
1121:
1122: ({\bf ii})\ The strict representability of ${\rm D}_{\sigma \left( n\right)
1123: } $ in $A-{\bf Mod}$ may be proved by induction on $n$. The case $n=1$
1124: follows from the exact sequence (\ref{du}), the ''third-representable''
1125: lemma and ({\bf i}). Now, suppose we have proved strict representability of $%
1126: {\rm D}_{\tau (k)}$ for each $\tau \in {\bf N}_{+}^\infty $ and each $k\leq
1127: n-1$. From the exact sequence in $\left[ A-{\bf Mod,}A-{\bf Mod}\right] $
1128:
1129: \begin{equation}
1130: 0\rightarrow {\rm D}_{\sigma \left( n\right) }\hookrightarrow {\rm D}%
1131: _{\sigma \left( n-1\right) }^{\bullet }\circ {\rm Diff}_{\sigma
1132: _n}^{+}\longrightarrow {\rm D}_{(\sigma \left( n-2\right) ,\sigma
1133: _{n-1}+\sigma _n)}\text{,} \label{ghazzali}
1134: \end{equation}
1135: the last morphism being
1136:
1137: \[
1138: \Delta \longmapsto \widehat{\Delta }
1139: \]
1140:
1141: \[
1142: \left( \left( ...\left( \left( \widehat{\Delta }(a_1)\right) \left(
1143: a_2\right) \right) ...\right) \left( a_{n-2}\right) \right) \left(
1144: a_{n-1}\right) \doteq \left( \left( \left( ...\left( \left( \Delta
1145: (a_1)\right) \left( a_2\right) \right) ...\right) \left( a_{n-2}\right)
1146: \right) \left( a_{n-1}\right) \right) \left( 1\right)
1147: \]
1148: (i.e. we use the ''gluing'' morphism of definition \ref{alucce} with respect
1149: to the last two indexes), Lemma \ref{lem 2.6} ({\bf a}) and the
1150: ''third-representable'' lemma, we obtain strict representability for ${\rm D}%
1151: _{\sigma \left( n\right) }$.
1152:
1153: ({\bf iii}) the case of ${\cal P}_{\sigma (n)}^{\bullet }$ follows, as for (%
1154: {\bf ii}), by induction via Lemma \ref{lem 2.6} , Lemma \ref{terzo} and by
1155: any of the following two exact sequences in $\left[ A-{\bf Mod,}A-{\bf Mod}%
1156: \right] $:
1157:
1158: \begin{equation}
1159: 0\rightarrow {\cal P}_{\sigma (n)}^{\bullet }\hookrightarrow {\cal P}%
1160: _{\sigma (n-1)}^{\bullet }\circ {\rm Diff}_{\sigma _n}^{(+)}\stackrel{}{%
1161: \longrightarrow }{\cal P}_{\left( \sigma (n-2),\sigma _{n-1}+\sigma
1162: _n\right) }^{\bullet } \label{Hol}
1163: \end{equation}
1164:
1165: \begin{equation}
1166: 0\rightarrow {\cal P}_{\sigma (n)}^{\bullet }\hookrightarrow {\rm Diff}%
1167: _{\sigma _1}^{\bullet }\circ {\cal P}_{\left( \sigma _2,...,\sigma _n\right)
1168: }^{+}\stackrel{{\em C}_{\sigma _1,\sigma _2}\left( {\rm Diff}_{\sigma
1169: _3,...,\sigma _n}^{+}\right) }{\longrightarrow }{\cal P}_{\left( \sigma
1170: _1+\sigma _2,\sigma _3,...,\sigma _n\right) }^{\bullet } \label{antiHol}
1171: \end{equation}
1172:
1173: where:
1174:
1175: (a) the upper boldface dot in ${\rm Diff}_{\sigma _1}^{\bullet }$ in (\ref
1176: {antiHol}) refers to the $A$-bimodule structure $\left( {\cal P}_{\left(
1177: \sigma _2,...,\sigma _n\right) }^{\bullet },{\cal P}_{\left( \sigma
1178: _2,...,\sigma _n\right) }^{+}\right) $;
1179:
1180: (b) the morphisms on the right are defined in the only natural way by using
1181: the ''gluing'' morphism of Definition \ref{alucce} : for (\ref{Hol}) we
1182: ''glue'' with respect to the last two indexes while in (\ref{antiHol}) we
1183: ''glue'' with respect to the first two. The case of ${\cal P}_{\sigma
1184: (n)}^{\bullet }\left[ P\right] $ is proved analogously, using (\ref{relHol})
1185: instead of (\ref{Hol}) or (\ref{relantiHol}) in place of (\ref{antiHol}).
1186: \qquad \TeXButton{End Proof}{\endproof}\
1187:
1188: \begin{remark}
1189: \TeXButton{rem 2.12}{\label{rem 2.12}} For any $k>0$, we have ${\bf \Lambda }%
1190: ^{\left( k\right) }\simeq \dfrac{{\rm I}}{{\rm I}^{k+1}}$ where ${\rm I}$ is
1191: the kernel of the ring multiplication $A\otimes _KA\rightarrow A$; hence $%
1192: {\bf \Lambda }^{\left( 1\right) }\simeq \Omega _{A/K}^1$ is just the $A$%
1193: -module of {\em K\"ahler differentials} (relative to $K$). Moreover it is
1194: not difficult to show (\cite{KLV} p. 17) that ${\bf \Lambda }^{\left(
1195: 1,...,1\right) }\simeq {\bf \Lambda }^n\doteq {\bf \Lambda }^1\wedge \ldots
1196: \wedge {\bf \Lambda }^1$ ($n$ times) $\simeq \Omega _{A/K}^n$ and that for
1197: each $k,l\in {\bf N}_{+}$ the map $\Delta \longmapsto \Delta $ induces a
1198: monomorphism $\left[ A-{\bf Mod,}A-{\bf Mod}\right] $:
1199:
1200: \[
1201: {\rm D}_{{\bf 1}\left( k+l\right) }\hookrightarrow {\rm D}_{{\bf 1}\left(
1202: k\right) }\circ {\rm D}_{{\bf 1}\left( l\right) }\text{, }
1203: \]
1204: whose dual representative $A$-homomorphism is just the wedge product $\wedge
1205: :{\bf \Lambda }^k(A)\otimes _A{\bf \Lambda }^l(A)\rightarrow {\bf \Lambda }%
1206: ^{k+l}(A)$.\
1207: \end{remark}
1208:
1209: If \ ${\frak D}$ is a differentially closed subcategory, we will denote the
1210: strict representatives in ${\frak D}$ of the relevant functors by adding $%
1211: {\frak D}$ as a subscript to the symbol used to denote the corresponding
1212: representative object in $A-{\bf Mod}$; for example, we write ${\bf \Lambda }%
1213: _{{\frak D}}^{\sigma \left( n\right) }$ for the representative object in $%
1214: {\frak D}$ of the functor ${\rm D}_{\sigma (n)}:{\frak D}\rightarrow {\frak D%
1215: }$.
1216:
1217: \begin{remark}
1218: \label{grate} As we did for ${\bf J}_{{\frak D}}^{\sigma _1}={\rm Hol}_{%
1219: {\frak D}}^{\left( \sigma _1\right) }$, we can exhibit another compatible $A$%
1220: -module structure on ${\rm Hol}_{{\frak D}}^{\left( \sigma _1,...,\sigma
1221: _n\right) }$, $\forall n>0$. Let $a\in A$ and $\widehat{{\rm id}}\in {\cal P}%
1222: _{\sigma (n)}^{\bullet }\left( {\rm Hol}_{{\frak D}}^{\left( \sigma
1223: _1,...,\sigma _n\right) }\right) $ correspond to the identity of ${\rm Hol}_{%
1224: {\frak D}}^{\left( \sigma _1,...,\sigma _n\right) }$ under the
1225: representability isomorphism. Since ${\cal P}_{\sigma (n)}^{\bullet }\left(
1226: {\rm Hol}_{{\frak D}}^{\left( \sigma _1,...,\sigma _n\right) }\right) $ and $%
1227: {\cal P}_{\sigma (n)}^{+}\left( {\rm Hol}_{{\frak D}}^{\left( \sigma
1228: _1,...,\sigma _n\right) }\right) $ coincide as sets, we can consider $a^{+}%
1229: \widehat{{\rm id}}$ (multiplication in ${\cal P}_{\sigma (n)}^{+}\left( {\rm %
1230: Hol}_{{\frak D}}^{\left( \sigma _1,...,\sigma _n\right) }\right) $) as an $A$%
1231: -endomorphism of ${\rm Hol}_{{\frak D}}^{\left( \sigma _1,...,\sigma
1232: _n\right) }$. It is easy to verify that this choice defines another $A$%
1233: -module structure on ${\rm Hol}_{{\frak D}}^{\left( \sigma _1,...,\sigma
1234: _n\right) }$, denoted by ${\rm Hol}_{{\frak D},+}^{\left( \sigma
1235: _1,...,\sigma _n\right) }$ and that $\left( {\rm Hol}_{{\frak D}}^{\left(
1236: \sigma _1,...,\sigma _n\right) },{\rm Hol}_{{\frak D},+}^{\left( \sigma
1237: _1,...,\sigma _n\right) }\right) $ is an $A$-bimodule.
1238: \end{remark}
1239:
1240: \begin{lemma}
1241: \label{pimpa}If $P\in Ob\left( {\frak D}\right) $ we have a canonical
1242: isomorphism ${\rm Hol}_{{\frak D}}^{\left( \sigma _1,...,\sigma _n\right)
1243: }\left[ P\right] \simeq {\rm Hol}_{{\frak D},+}^{\left( \sigma _1,...,\sigma
1244: _n\right) }\otimes ^{\bullet }P$ in ${\frak D}$.
1245: \end{lemma}
1246:
1247: \TeXButton{Proof}{\proof}Let $Q$ be an object in ${\frak D}$. By lemma \ref
1248: {lem 2.3}, we have
1249: \[
1250: Hom\left( {\rm Hol}_{{\frak D},+}^{\left( \sigma _1,...,\sigma _n\right)
1251: }\otimes ^{\bullet }P,Q\right) \simeq Hom^{\bullet }\left( P,Hom^{+}\left(
1252: {\rm Hol}_{{\frak D}}^{\left( \sigma _1,...,\sigma _n\right) },Q\right)
1253: \right) \simeq
1254: \]
1255: \[
1256: \ \simeq Hom^{\bullet }\left( P,{\cal P}_{\sigma (n)}^{+}\left( Q\right)
1257: \right)
1258: \]
1259: and by definition of ${\cal P}_{\sigma (n)}^{+}$ and example \ref{es 2.2}
1260: \[
1261: Hom^{\bullet }\left( P,{\cal P}_{\sigma (n)}^{+}\left( Q\right) \right)
1262: \simeq {\cal P}_{\left( \sigma _1\right) }^{\bullet }\left[ P\right] \left(
1263: {\cal P}_{\left( \sigma _2,...,\sigma _n\right) }^{}\left( Q\right) \right) =%
1264: {\cal P}_{\left( \sigma _1,...,\sigma _n\right) }^{\bullet }\left[ P\right]
1265: \left( Q\right) \text{.}
1266: \]
1267: \TeXButton{End Proof}{\endproof}
1268:
1269: \begin{remark}
1270: \TeXButton{jam}{\label{jam}} For any differentiable closed subcategory $%
1271: {\frak D}\subseteq A-{\bf Mod}$ it is still true, as in the case ${\frak D}%
1272: =A-{\bf Mod}$, that ${\bf J}_{{\frak D}}^k\left( P\right) $ is generated as
1273: an $A$-module by
1274: \[
1275: \ \ \left\{ j_k\left( p\right) \mid p\in P\right\} .
1276: \]
1277: In fact, let ${\bf J}_{{\frak D}}^k\left( P\right) ^{\sim }$ denote the $A$%
1278: -submodule of ${\bf J}_{{\frak D}}^k\left( P\right) $ generated by $\left\{
1279: j_k\left( p\right) \mid p\in P\right\} $; this is still an object of ${\frak %
1280: D}$ by Definition \ref{def 2.1} (e). Now, the composition
1281: \[
1282: P\stackrel{j_k^{{\frak D}}}{\longrightarrow }{\bf J}_{{\frak D}}^k\left(
1283: P\right) \stackrel{\pi }{\twoheadrightarrow }\frac{{\bf J}_{{\frak D}%
1284: }^k\left( P\right) }{{\bf J}_{{\frak D}}^k\left( P\right) ^{\sim }}
1285: \]
1286: is the DO of order $\leq k$ corresponding to $\pi $ under the isomorphism
1287: \[
1288: Hom_A\left( {\bf J}_{{\frak D}}^k\left( P\right) ,{\bf J}_{{\frak D}%
1289: }^k\left( P\right) /{\bf J}_{{\frak D}}^k\left( P\right) ^{\sim }\right)
1290: \simeq {\rm Diff}_k\left( P,{\bf J}_{{\frak D}}^k\left( P\right) /{\bf J}_{%
1291: {\frak D}}^k\left( P\right) ^{\sim }\right) .
1292: \]
1293: But $\pi \circ j_k^{{\frak D}}$ is zero hence $\pi =0$ and we conclude. Note
1294: that this also shows that the canonical morphism
1295: \[
1296: {\bf J}^k\left( P\right) \rightarrow {\bf J}_{{\frak D}}^k\left( P\right)
1297: \text{ }
1298: \]
1299: is an $A$-epimorphism.
1300: \end{remark}
1301:
1302: If $P\in Ob({\frak D})$, the monomorphism in $\left[ {\frak D}{\bf ,}{\frak D%
1303: }\right] $:
1304:
1305: \[
1306: {\rm Diff}_s(P,\cdot )\subset {\rm Diff}_t(P,\cdot )\text{,\qquad }t\geq
1307: s,\quad
1308: \]
1309: gives rise to a ${\frak D}$-epimorphism (also an $A$-epimorphism by Remark
1310: \ref{jam}) between representative objects:
1311:
1312: \[
1313: \pi _{t,s}(P):{\bf J}_{{\frak D}}^t(P)\rightarrow {\bf J}_{{\frak D}}^s(P)
1314: \]
1315: which fits in the commutative diagram
1316:
1317: \[
1318: \begin{array}{ccccc}
1319: P & \stackrel{j_s(P)}{\longrightarrow } & {\bf J}_{{\frak D}}^s(P) & & \\
1320: & \stackunder{j_t(P)}{\searrow } & \uparrow & \pi _{t,s}(P) & \\
1321: & & {\bf J}_{{\frak D}}^t(P) & &
1322: \end{array}
1323: \]
1324: The rule $P\longmapsto {\bf J}_{{\frak D}}^t(P)$ defines in the obvious way
1325: a (covariant) functor ${\frak D}\rightarrow {\frak D}$ (\cite{Kr} or \cite
1326: {KLV}).\
1327:
1328: The following example shows the importance\ of the appropriate choice of the
1329: differentially closed subcategory of $A-{\bf Mod}$ in determining the ''
1330: geometrical effectiveness'' and size of the representative objects of the
1331: relevant functors.
1332:
1333: \begin{example}
1334: Let $M$ be a smooth real manifold (which we assume Hausdorff and with a
1335: countable basis), $K={\Bbb R}$ and $A=C^\infty \left( M;{\Bbb R}\right) $.
1336: Then (\cite{MVi}) ${\bf \Lambda }^{\sigma \left( n\right) }$ is in general
1337: neither projective nor of finite type over $A$: in particular, when $\sigma
1338: =\left( 1,...,1,...\right) $, {\it it does not} coincide with the $A$-module
1339: of differential $n$-forms on the manifold $M$. To obtain these
1340: ''geometrical'' objects we must choose an appropriate subcategory of $A-{\bf %
1341: Mod}$ : in our approach, choosing a ''geometry'' is equivalent to select a
1342: differentially closed subcategory ${\frak D}$. For finite dimensional (real)
1343: differential geometry we may choose ${\frak D}\doteq A-{\bf Mod}_{geom}$,
1344: the full subcategory of geometric $A$-modules, i.e. of $A$-modules $P$ such
1345: that $\bigcap\limits_{x\in M}I_xP=(0)$, $I_x$ being the maximal ideal of
1346: smooth functions on $M$ vanishing at $x\in M$.
1347:
1348: Note that $A-{\bf Mod}_{geom}\supseteq A-{\bf Mod}_{pr,\text{ }f.t.}$, the
1349: full subcategory of projective $A$-modules of finite type, since $A$ itself
1350: is a geometric $A$-module; however, $A-{\bf Mod}_{pr,\text{ }f.t.}$ is not
1351: differentially closed because it is not abelian (and does not satisfy (e)).
1352: Another reason that makes us prefer working with the bigger $A-{\bf Mod}%
1353: _{geom}$ is its better functoriality with respect to change of algebras
1354: induced by pull backs of smooth mappings of manifolds\footnote{%
1355: If $f:M\rightarrow N$ is a smooth map and $P$ is a geometric $C^\infty (M)$%
1356: -module then $P$ is still geometric when viewed as a $C^\infty (N)$-module
1357: via the pull back $f^{*}:C^\infty (N)\longrightarrow C^\infty (M)$.
1358: Projectivity is not preserved, instead.}. $A-{\bf Mod}_{geom}$ is
1359: differentially closed due to the fact that the ''geometrization'' functor
1360: \begin{eqnarray*}
1361: \left( \text{ }\cdot \text{ }\right) _{geom}:A-{\bf Mod}\longrightarrow A-%
1362: {\bf Mod}_{geom} \\
1363: P\longmapsto \text{ }P_{geom}\doteq \frac P{\bigcap\limits_{x\in M}I_xP}
1364: \end{eqnarray*}
1365: sends representative objects in $A-{\bf Mod}_{geom}$ to representative
1366: objects in $A-{\bf Mod}_{geom}$ for all the relevant functors (\cite{KLV}).
1367: ''Geometrical'' objects are obtained as representative objects; for example $%
1368: {\bf \Lambda }_{geom.}^{\left( 1,...,1\right) }$, with $(1,...,1)\in {\bf N}%
1369: _{}^k$, is isomorphic to $\Gamma \left( \bigwedge\limits^kT^{*}M\right) $,
1370: the module of sections of the $k$-th exterior power of the cotangent bundle
1371: of $M$, i.e. the module of $k$-differential forms of $M$.
1372: \end{example}
1373:
1374: It is possible to encode the ''{\em smoothness}'' of the geometry we want to
1375: describe, completely in the choice of the differentially closed subcategory:
1376:
1377: \begin{definition}
1378: A differentially closed subcategory ${\frak D}$ of $A-{\bf Mod}$ is called
1379: {\em smooth} if $\Lambda _{{\frak D}}^{\left( 1\right) }$ is a projective $A$%
1380: -module of finite type.
1381: \end{definition}
1382:
1383: \begin{examples}
1384: (i) If $K$ is an algebraically closed field of zero characteristic and $A$
1385: is the coordinate ring of a {\it regular} affine $K$-variety, then ${\frak D}%
1386: =A-{\bf Mod}$ is smooth (\cite{Ha}, II.8).
1387:
1388: (ii) If $M$ is a smooth manifold and $A=C^\infty \left( M;{\bf R}\right) $
1389: then $A-{\bf Mod}_{geom}$ is smooth while $A-{\bf Mod}$ is not.\
1390: \end{examples}
1391:
1392: It can be proved (as in the proof of Theorem \ref{thm 2.7}) that if ${\frak D%
1393: }$ is smooth then {\em all the representative objects of relevant functors
1394: are indeed projective and of finite type as }$A${\em -modules}. However, we
1395: want to stress that since representative objects may be constructed also in
1396: non-smooth cases, our approach works also in describing singular and even
1397: infinite dimensional geometrical situations. However, to resort with useful
1398: objects one has to make an adequate choice of ${\frak D}$.\
1399:
1400: The following proposition will be useful in the next sections:
1401:
1402: \begin{proposition}
1403: \label{stop} Let ${\frak D}$ be smooth. Then
1404:
1405: (i) ${\bf \Lambda }_{{\frak D}}^{{\bf 1}_n}=(0)$ for $n>>0$;
1406:
1407: (ii) If $P\in Ob\left( {\frak D}\right) $, ${\rm Hol}_{{\frak D}}^{{\bf 1}%
1408: _n}\left[ P\right] =(0)$ for $n>>0$.
1409: \end{proposition}
1410:
1411: \TeXButton{Proof}{\proof} By lemma \ref{pimpa} and remark \ref{new}, (ii)
1412: follows from (i). Remark \ref{rem 2.12} together with the fact that ${\bf %
1413: \Lambda }_{{\frak D}}^{(1)}$ is of finite type proves (i).
1414: \TeXButton{End Proof}{\endproof}
1415:
1416: \section{Higher de Rham complexes}
1417:
1418: In this Section we use the functors introduced in Section 1 and their
1419: representative objects (Section 2) to build higher order analogs of the de
1420: Rham complex. Their cohomology will be studied in Section 4.
1421:
1422: Let ${\frak D}$ be a differentially closed subcategory of $A-{\bf Mod}$. The
1423: dual representative of the monomorphism in $\left[ {\frak D},{\frak D}%
1424: \right] $:
1425:
1426: \[
1427: {\rm D}_{(\sigma \left( n\right) ,k)}\hookrightarrow {\rm D}_{\sigma \left(
1428: n\right) }^{\bullet }\circ {\rm Diff}_k^{(+)}\text{, \qquad }\sigma \left(
1429: n\right) \in {\bf N}_{+}^n,\text{ }k\in {\bf N}_{+}\text{,}
1430: \]
1431: is a ${\frak D}$-epimorphism:
1432:
1433: \[
1434: {\bf J}_{{\frak D}}^k({\bf \Lambda }_{{\frak D}}^{\sigma \left( n\right)
1435: })\rightarrow {\bf \Lambda }_{{\frak D}}^{(\sigma \left( n\right) ,k)}\text{;%
1436: }
1437: \]
1438: define $d_{(\sigma \left( n\right) ,k)}^{{\frak D}}$ to be the composition
1439:
1440: \begin{equation}
1441: {\bf \Lambda }_{{\frak D}}^{\sigma \left( n\right) }\stackrel{j_k}{%
1442: \longrightarrow }{\bf J}_{{\frak D}}^k({\bf \Lambda }_{{\frak D}}^{\sigma
1443: \left( n\right) })\rightarrow {\bf \Lambda }_{{\frak D}}^{(\sigma \left(
1444: n\right) ,k)}\text{.} \label{scioglimento}
1445: \end{equation}
1446: Obviously, $d_{(\sigma \left( n\right) ,k)}^{{\frak D}}$ is a DO of order $%
1447: \leq k$.
1448:
1449: \smallskip\
1450:
1451: \begin{definition}
1452: If $\sigma \in {\bf N}_{+}^\infty $, the sequence in ${\bf DIFF}_A$%
1453: \begin{equation}
1454: 0\rightarrow A\stackrel{d_{(\sigma _1)}^{{\frak D}}}{\rightarrow }\Lambda _{%
1455: {\frak D}}^{(\sigma _1)}\stackrel{d_{\left( \sigma _1,\sigma _2\right) }^{%
1456: {\frak D}}}{\longrightarrow }\Lambda _{{\frak D}}^{\sigma (2)}\rightarrow
1457: \ldots \stackrel{d_{\sigma \left( k\right) }^{{\frak D}}}{\longrightarrow }%
1458: \Lambda _{{\frak D}}^{\sigma \left( k\right) }\rightarrow \ldots
1459: \label{dRsup}
1460: \end{equation}
1461: is called {\em higher de Rham sequence}{\it \ }of type $\sigma $ of the $K$%
1462: -algebra $A$ and is denoted by ${\bf dR}_\sigma ^{{\frak D}}(A)$ or simply
1463: by ${\bf dR}_\sigma ^{{\frak D}}$; each $d_{\sigma \left( k\right) }^{{\frak %
1464: D}}$, $k>0$, is called {\em higher de Rham differential} and is a DO of
1465: order $\leq \sigma _k$.
1466: \end{definition}
1467:
1468: \begin{remark}
1469: When $\sigma ={\bf 1}\in {\bf N}_{+}^\infty $, the corresponding de Rham
1470: sequence is called {\em ordinary}. In this case we write ${\bf \Lambda }_{%
1471: {\frak D}}^k$ for ${\bf \Lambda }_{{\frak D}}^{\left( 1,...,1\right) }$ , $%
1472: \left( 1,...,1\right) \in {\bf N}_{+}^k$, $\forall k>0$, so that:
1473:
1474: \begin{equation}
1475: {\bf dR}_{(\underline{1})}^{{\frak D}}\equiv {\bf dR}^{{\frak D}}{\bf :\quad
1476: }0\rightarrow A\stackrel{d}{\rightarrow }{\bf \Lambda }_{{\frak D}}^1%
1477: \stackrel{d}{\rightarrow }{\bf \Lambda }_{{\frak D}}^2\rightarrow \ldots
1478: \stackrel{d}{\rightarrow }{\bf \Lambda }_{{\frak D}}^k\rightarrow \ldots
1479: \label{dR}
1480: \end{equation}
1481: and each differential is a DO of order $\leq 1$.
1482: \end{remark}
1483:
1484: Each $d^{{\frak D}}$ in (\ref{dRsup}) is in fact a differential according to
1485: the following:\
1486:
1487: \begin{proposition}
1488: $\forall \sigma \in {\bf N}_{+}^\infty $ the higher de Rham sequence ${\bf dR%
1489: }_\sigma ^{{\frak D}}$ is a complex.\
1490: \end{proposition}
1491:
1492: \TeXButton{Proof}{\proof} Let $n\geq 0$ and consider the diagram defining
1493: two consecutive higher de Rham differentials:
1494:
1495: \[
1496: \begin{array}{cccccc}
1497: {\bf \Lambda }_{{\frak D}}^{\sigma \left( n\right) } & \stackrel{\doteq
1498: d_{\sigma \left( n+1\right) }^{{\frak D}}}{\longrightarrow } & {\bf \Lambda }%
1499: _{{\frak D}}^{\sigma \left( n+1\right) } & \stackrel{\doteq d_{\sigma \left(
1500: n+2\right) }^{{\frak D}}}{\longrightarrow } & {\bf \Lambda }_{{\frak D}%
1501: }^{\sigma \left( n+2\right) } & \\
1502: & \searrow ^{j_{\sigma _{n+1}}} & \nearrow _{\pi _1} & \searrow ^{j_{\sigma
1503: _{n+2}}} & \nearrow _{\pi _2} & \\
1504: & {\bf J}_{{\frak D}}^{\sigma _{n+1}}({\bf \Lambda }_{{\frak D}}^{\sigma
1505: \left( n\right) }) & & {\bf J}_{{\frak D}}^{\sigma _{n+2}}({\bf \Lambda }_{%
1506: {\frak D}}^{\sigma \left( n+1\right) }) & &
1507: \end{array}
1508: \text{.}
1509: \]
1510: Since $d_{\sigma \left( n+2\right) }^{{\frak D}}\circ d_{\sigma \left(
1511: n+1\right) }^{{\frak D}}\equiv $ $\pi _2\circ j_{\sigma _{n+2}}\circ \pi
1512: _1\circ j_{\sigma _{n+1}}$ is a DO of order $\leq k+l$ , there exists a
1513: unique $A$-homomorphism
1514:
1515: \[
1516: \varphi _{d_{\sigma \left( n+2\right) }^{{\frak D}}\circ d_{\sigma \left(
1517: n+1\right) }^{{\frak D}}}:{\bf J}_{{\frak D}}^{\sigma _{n+2}+\sigma _{n+1}}(%
1518: {\bf \Lambda }_{{\frak D}}^{\sigma \left( n\right) })\rightarrow {\bf %
1519: \Lambda }_{{\frak D}}^{\sigma \left( n+2\right) }
1520: \]
1521: which makes the following diagram commutative:
1522:
1523: \[
1524: \text{\TeXButton{Vtriangle1}
1525: {\settriparms[1`1`-1;500]
1526: \Vtriangle[\QTR{bf}{\Lambda }_{\QTR{frak}{D}}^{\sigma \left( n\right) }`\QTR{bf}{\Lambda }_{\QTR{frak}{D}}^{\sigma \left( n+2\right) }`\QTR{bf}{J}_{\QTR{frak}{D}}^{\sigma _{n+2}+\sigma _{n+1}}(\QTR{bf}{\Lambda }_{\QTR{frak}{D}}^{\sigma \left( n\right) });d_{\sigma \left( n+2\right) }^{\QTR{frak}{D}}\circ d_{\sigma \left( n+1\right) }^{\QTR{frak}{D}}`j_{\sigma _{n+2}+\sigma _{n+1}}`\varphi _{d_{\sigma \left( n+2\right) }^{\QTR{frak}{D}}\circ d_{\sigma \left( n+1\right) }^{\QTR{frak}{D}}}]}%
1527: .}
1528: \]
1529:
1530: It is not difficult to check that $\varphi _{d_{\sigma \left( n+2\right) }^{%
1531: {\frak D}}\circ d_{\sigma \left( n+1\right) }^{{\frak D}}}$ is just the dual
1532: representative of the composition:
1533:
1534: \[
1535: {\rm D}_{\sigma \left( n+2\right) }\hookrightarrow {\rm D}_{\sigma \left(
1536: n+1\right) }^{\bullet }\circ {\rm Diff}_{\sigma _{n+2}}^{(+)}\longrightarrow
1537: {\rm D}_{\sigma \left( n\right) }^{\bullet }\circ {\rm Diff}_{\sigma
1538: _{n+1}+\sigma _{n+2}}^{(+)}
1539: \]
1540: which is immediately checked to be zero; therefore $\varphi _{d_{\sigma
1541: \left( n+2\right) }^{{\frak D}}\circ d_{\sigma \left( n+1\right) }^{{\frak D}%
1542: }}=0$ and $d_{\sigma \left( n+2\right) }^{{\frak D}}\circ d_{\sigma \left(
1543: n+1\right) }^{{\frak D}}=0$ as well.\qquad \TeXButton{End Proof}{\endproof}\
1544:
1545: \begin{remark}
1546: \TeXButton{rem 3.6}{\label{rem 3.6}}({\bf i}) Let ${\frak D}=A-{\bf Mod}$.
1547: By induction on $n$ we can prove (using for $n=1$ the explicit description
1548: of ${\bf \Lambda }^{\left( \sigma _1\right) }$ given in Section 2) that
1549: formula (\ref{scioglimento}) implies that ${\bf \Lambda }^{\sigma \left(
1550: n\right) }$ is generated by the set
1551: \[
1552: \left\{ d_{\sigma \left( n\right) }\left( a_1d_{\sigma \left( n-1\right)
1553: }\left( a_2\ldots d_{\sigma \left( 1\right) }^{}\left( a_n\right) \ldots
1554: \right) \right) \mid a_1,a_2,\ldots ,a_n\in A\right\} .
1555: \]
1556: In fact, ${\bf J}^{\sigma _n}({\bf \Lambda }^{\sigma \left( n-1\right) })$
1557: is known to be generated over $A$ by the elements $j_{\sigma _n}\left(
1558: \omega \right) $, $\omega \in {\bf \Lambda }^{\sigma \left( n-1\right) }$
1559: and ${\bf J}^{\sigma _n}({\bf \Lambda }^{\sigma \left( n-1\right)
1560: })\rightarrow {\bf \Lambda }^{\sigma \left( n\right) }$ is an epimorphism.
1561: This result still holds for ${\bf \Lambda }_{{\frak D}}^{\sigma \left(
1562: n\right) }$ with $d$ replaced by $d^{{\frak D}}$, ${\frak D}\subseteq A-{\bf %
1563: Mod}$ being any differentiable closed subcategory: the proof is analogous to
1564: the argument used in Remark \ref{jam} (a). This also shows that the
1565: canonical morphisms
1566: \[
1567: \text{ \qquad }{\bf \Lambda }^{\sigma \left( n\right) }\rightarrow {\bf %
1568: \Lambda }_{{\frak D}}^{\sigma \left( n\right) }
1569: \]
1570: are $A$-epimorphisms.
1571:
1572: ({\bf ii}) In the ''ordinary'' case $\left( \sigma _1,...,\sigma _n\right)
1573: =\left( 1,...,1\right) \equiv {\bf 1}\left( n\right) $, the $A$-module
1574: structure of ${\rm Hol}_{{\frak D},+}^{{\bf 1}_n}$ (remark \ref{grate}) can
1575: be expressed via the isomorphism (remark \ref{new})
1576: \[
1577: {\rm Hol}_{{\frak D}}^{\left( \sigma _1,...,\sigma _n\right) }\simeq {\bf %
1578: \Lambda }_{{\frak D}}^{\left( \sigma _2,...,\sigma _n\right) }\oplus {\bf %
1579: \Lambda }_{{\frak D}}^{\left( \sigma _1,...,\sigma _n\right) }
1580: \]
1581: as $a^{+}\left( \rho ,\omega \right) =\left( a\rho ,a\omega +\left(
1582: d_{\left( 1\right) }a\right) \wedge \rho \right) $.
1583:
1584: ({\bf iii}) If ${\frak D}=A-{\bf Mod}$, ${\bf dR}_{{\bf 1}}$ coincides with
1585: the usual algebraic de Rham complex of the $K$-algebra $A$ (\cite{Bou III}
1586: and \cite{Bou X}).
1587:
1588: ({\bf iv}) If $K={\Bbb R}$, $M$ is a smooth manifold, $A=C^\infty \left( M;%
1589: {\Bbb R}\right) $ and ${\frak D}$ is the category of geometric $A$-modules
1590: (see Section 2) then ${\bf dR}_{{\bf 1}}^{{\frak D}}$ is the geometric de
1591: Rham complex on $M$. It turns out that any natural differential operator
1592: occurring in differential geometry can be recovered functorially using our
1593: approach: see \cite{VV} for the case of the Lie derivative and the
1594: corresponding homotopy formula.
1595: \end{remark}
1596:
1597: If $\sigma ,\tau \in {\bf N}_{+}^\infty $ with $\sigma \geq \tau $ (i.e. $%
1598: \sigma _i\geq \tau _i$, $\forall i\geq 1$), then for each $n>0$ we have a
1599: monomorphism ${\rm D}_{\tau \left( n\right) }\hookrightarrow {\rm D}_{\sigma
1600: \left( n\right) }$ in $\left[ {\frak D},{\frak D}\right] $; this induces a $%
1601: {\frak D}$-epimorphism on representatives ${\bf \Lambda }_{{\frak D}%
1602: }^{\sigma \left( n\right) }\rightarrow {\bf \Lambda }_{{\frak D}}^{\tau
1603: \left( n\right) }$, $\forall n>0$. By \ref{rem 3.6}, this is also an $A$%
1604: -epimorphism. All these epimorphisms commute with higher de Rham
1605: differentials and therefore define a morphism of complexes
1606:
1607: \begin{equation}
1608: {\bf dR}_\sigma ^{{\frak D}}\rightarrow {\bf dR}_\tau ^{{\frak D}}
1609: \label{luppolo}
1610: \end{equation}
1611: (if $\sigma \geq \tau $). So we can consider the ($A$-epimorphic) inverse
1612: system $\left\{ {\bf dR}_\sigma ^{{\frak D}}\right\} _{\sigma \in {\bf N}%
1613: _{+}^\infty }$ and give the following:
1614:
1615: \begin{definition}
1616: \TeXButton{infty}{\label{infty}}The {\it infinitely prolonged} (or, simply,
1617: {\it infinite}) {\it de Rham complex} of the $K$-algebra $A$, is the complex
1618: in {\bf K}$\left( K-{\bf Mod}\right) $%
1619: \begin{equation}
1620: \begin{array}{c}
1621: {\bf dR}_\infty ^{{\frak D}}(A)\doteq \underleftarrow{\lim }_{\sigma \in
1622: {\bf N}_{+}^\infty }{\bf dR}_\sigma ^{{\frak D}}(A), \\
1623: {\bf dR}_\infty ^{{\frak D}}(A):\text{ }0\rightarrow A\stackrel{d_{\left(
1624: \infty \right) }}{\longrightarrow }{\bf \Lambda }_{{\frak D}}^{\left( \infty
1625: \right) }\stackrel{d_{\left( \infty ,\infty \right) }}{\longrightarrow }{\bf %
1626: \Lambda }_{{\frak D}}^{\left( \infty ,\infty \right) _2}\rightarrow \cdots
1627: \rightarrow {\bf \Lambda }_{{\frak D}}^{\left( \infty ,\ldots ,\infty
1628: \right) _n}\rightarrow \cdots
1629: \end{array}
1630: \label{infinito}
1631: \end{equation}
1632: where ${\bf \Lambda }_{{\frak D}}^{\left( \infty ,\ldots ,\infty \right)
1633: _n}\doteq \underleftarrow{\lim }_{\sigma \left( n\right) \in {\bf N}_{+}^n}%
1634: {\bf \Lambda }_{{\frak D}}^{\sigma \left( n\right) }$, $\forall n>0$.
1635: \end{definition}
1636:
1637: \begin{remark}
1638: \label{photos}\TeXButton{photos}{\label{photos}}{\em Three descriptions of
1639: DO's between strict representative objects.}
1640:
1641: We work in a fixed differentially closed subcategory ${\frak D}$ of $A-{\bf %
1642: Mod}$ and all representative objects will be in ${\frak D}$.
1643:
1644: Let $F_1$ and $F_2$ be representative objects of differential functors $%
1645: {\cal F}_1$ and ${\cal F}_2$, respectively. Suppose that ${\cal F}_1$ has an
1646: associated functor ${\cal F}_1^{\bullet }$ (having as domain $A-{\bf BiMod}_{%
1647: {\frak D}}$) such that ${\cal F}_1^{\bullet }\left( {\rm Diff}_k^{\left(
1648: +\right) }\right) $ is strictly representable by ${\bf J}^k\left( F_1\right)
1649: $: this is the case, for example, of ${\cal F}_1={\rm D}_{\sigma \left(
1650: n\right) }$ or ${\rm Diff}_l$. Let
1651:
1652: \begin{equation}
1653: \Delta :F_1\longrightarrow F_2 \label{are}
1654: \end{equation}
1655: be a DO of order $\leq k$. Then, there exists a unique $A$-homomorphism (%
1656: \cite{Kr}: jet-associated to $\Delta $)
1657:
1658: \begin{equation}
1659: f_\Delta :{\bf J}^k\left( F_1\right) \longrightarrow F_2 \label{ere}
1660: \end{equation}
1661: which represents $\Delta $ by duality: $\Delta =f_\Delta \circ j_k\left(
1662: F_1\right) $. Since ${\bf J}^k\left( F_1\right) $ is the representative
1663: object of ${\cal F}_1^{\bullet }\left( {\rm Diff}_k^{\left( +\right)
1664: }\right) $, $f_\Delta $ defines a unique morphism in $\left[ {\frak D},%
1665: {\frak D}\right] $:
1666: \begin{equation}
1667: f^\Delta :{\cal F}_2\longrightarrow {\cal F}_1^{\bullet }\left( {\rm Diff}%
1668: _k^{\left( +\right) }\right) \text{,} \label{ire}
1669: \end{equation}
1670: called {\em generator} {\em morphism} of $\Delta $.
1671:
1672: Formulas (\ref{ere}) and (\ref{ire}) give two different descriptions of a DO
1673: between representative objects. Formula (\ref{ire}) allows one to identify
1674: it with a functorial morphism which, as a rule, may be established in a
1675: straightforward way and can then be used to define the corresponding natural
1676: DO (\ref{are}). The following examples show this procedure at work in two
1677: canonical cases; we assume for simplicity ${\frak D}=A-{\bf Mod}$.
1678:
1679: (i) {\em Higher de Rham differential} $d_{\sigma \left( n\right) }$.
1680:
1681: If ${\cal F}_2$ $={\rm D}_{\sigma \left( n\right) }$ , ${\cal F}_1={\rm D}%
1682: _{\sigma \left( n-1\right) }$ , $k=\sigma _n$ and we take for (\ref{ire})
1683: the natural inclusion
1684: \[
1685: {\rm D}_{\sigma \left( n\right) }\hookrightarrow {\rm D}_{\sigma \left(
1686: n-1\right) }^{\bullet }\left( {\rm Diff}_{\sigma _n}^{\left( +\right)
1687: }\right) ,
1688: \]
1689: then $d_{\sigma \left( n\right) }:{\bf \Lambda }^{\sigma \left( n-1\right)
1690: }\rightarrow {\bf \Lambda }^{\sigma \left( n\right) }$ is the corresponding
1691: DO (\ref{are}).
1692:
1693: (ii) {\em ''Absolute'' jet-operator} $j_k$.
1694:
1695: In this almost tautological case, ${\cal F}_1=Hom_A\left( A,\cdot \right)
1696: \equiv {\rm Diff}_0$ and ${\cal F}_2$ $={\rm Diff}_k\equiv $
1697:
1698: $Hom_A^{\bullet }\left( A,\cdot \right) $ $\left( {\rm Diff}_k^{\left(
1699: +\right) }\right) $ ; if we take (\ref{ire}) to be the identity
1700: \[
1701: {\rm Id}:{\rm Diff}_k\rightarrow Hom_A^{\bullet }\left( A,\cdot \right)
1702: \left( {\rm Diff}_k^{\left( +\right) }\right) \equiv {\rm Diff}_k,
1703: \]
1704: then (\ref{are}) is just $j_k:A\rightarrow $ ${\bf J}^k$.
1705: \end{remark}
1706:
1707: \medskip
1708:
1709: \ Rigidity of higher de Rham cohomology
1710:
1711: In this Section we prove the main result of this paper i.e. that in the
1712: smooth case the higher-order de Rham cohomologies coincide with the ordinary
1713: ( i.e. lowest order) one. Essentially this amounts to a fairly intuitive
1714: assert: raising the order of the natural DOs involved in the ${\bf dR}$%
1715: -complexes does not change the cohomological information, provided the
1716: situation in which we are working is smooth.
1717:
1718: In this Section (and in the Appendix), $A$ is a $K$-algebra of zero
1719: characteristic, containing $K$ as a subring and ${\frak D}$ a differentially
1720: closed {\it smooth} subcategory of $A-{\bf Mod}$. As in the previous
1721: Section, all representative objects, unless otherwise stated, will be
1722: considered in ${\frak D}$.
1723:
1724: Smoothness of ${\frak D}$ implies that for any $k,l\geq 0$, the gluing
1725: morphism in $\left[ {\frak D},{\frak D}\right] $
1726: \[
1727: {\rm Diff}_k^{\bullet }\circ {\rm Diff}_l^{\left( +\right) }\stackrel{C_{k,l}%
1728: }{\longrightarrow }{\rm Diff}_{k+l}^{}
1729: \]
1730: is surjective i.e. that any DO can be expressed as composition of lower
1731: order ones. This can be seen as follows. Let us fix $k$ and proceed by
1732: induction on $l$. The case $l=0$ is trivial since ${\rm Diff}_0={\rm id}_{%
1733: {\frak D}}$. To prove the inductive step let us consider the commutative
1734: diagram
1735: \[
1736: \begin{tabular}{lll}
1737: ${\rm Diff}_k^{\bullet }\circ {\rm Diff}_{l+1}^{\left( +\right) }$ & $%
1738: \stackrel{C_{k,l+1}}{\longrightarrow }$ & ${\rm Diff}_{k+l+1}$ \\
1739: $\qquad \uparrow $ & & $\quad \uparrow $ \\
1740: ${\rm Diff}_k^{\bullet }\circ {\rm Diff}_l^{\left( +\right) }$ & $%
1741: \stackunder{C_{k,l}}{\longrightarrow }$ & ${\rm Diff}_{k+l}$%
1742: \end{tabular}
1743: \]
1744: (where the vertical arrows are natural inclusions) and suppose $C_{k,l}$ is
1745: epic. Passing to the corresponding diagram of reperesentative objects
1746: completing it with kernels, we get a commutative diagram (\cite{KLV} p. 52)
1747: with exact columns
1748: \[
1749: \begin{tabular}{lll}
1750: $0$ & & $0$ \\
1751: $\downarrow $ & & $\downarrow $ \\
1752: ${\rm S}^{k+l+1}\left( {\bf \Lambda }^1\right) $ & $\stackrel{\rho }{%
1753: \longrightarrow }$ & ${\rm S}^{l+1}\left( {\bf \Lambda }^1\right) \otimes
1754: {\bf J}^k$ \\
1755: $\downarrow $ & & $\downarrow $ \\
1756: ${\bf J}^{k+l+1}$ & $\stackrel{C^{k,l+1}}{\longrightarrow }$ & ${\bf J}%
1757: ^{l+1}\left( {\bf J}^k\right) $ \\
1758: $\downarrow $ & & $\downarrow $ \\
1759: ${\bf J}^{k+l}$ & $\stackunder{C^{k,l}}{\longrightarrow }$ & ${\bf J}%
1760: ^l\left( {\bf J}^k\right) $ \\
1761: $\downarrow $ & & $\downarrow $ \\
1762: $0$ & & $0$%
1763: \end{tabular}
1764: \]
1765: where ${\rm S}^r$ denotes the $r$-th symmetric power and $C^{s,t}$ is the
1766: dual-representative of $C_{s,t}$. By duality it is enough to prove that $%
1767: C^{k,l+1}$ is monic. By induction hypothesis $C^{k,l}$ is monic so we are
1768: reduced to showing that $\rho $ is monic. It is not difficult to prove (e.g.
1769: again by induction on $l$) that $\rho $ is just the composition
1770: \[
1771: \begin{tabular}{lll}
1772: ${\rm S}^{k+l+1}\left( {\bf \Lambda }^1\right) $ & $\stackrel{\alpha }{%
1773: \longrightarrow }\stackrel{}{{\rm S}^{l+1}\left( {\bf \Lambda }^1\right)
1774: \otimes {\rm S}^k\left( {\bf \Lambda }^1\right) \stackrel{\beta }{%
1775: \longrightarrow }}$ & ${\rm S}^{l+1}\left( {\bf \Lambda }^1\right) \otimes
1776: {\bf J}^k$%
1777: \end{tabular}
1778: \]
1779: where $\alpha :\omega _1\cdots \omega _{k+l+1}\longmapsto \sum \left( \omega
1780: _{i_1}\cdots \omega _{i_{l+1}}\right) \otimes \left( \omega _{j_1}\cdots
1781: \omega _{j_k}\right) $ where the sum is extended to all partitions $\left(
1782: \left( i_1,...,i_{l+1}\right) ,\left( j_1,...,j_k\right) \right) $ of $%
1783: \left\{ 1,...,k+l+1\right\} $ of (ordered) length $(l+1,k)$ and $\beta ={\rm %
1784: id}_{{\rm S}^{l+1}\left( {\bf \Lambda }^1\right) }\otimes i$ with
1785: \[
1786: i:{\rm S}^k\left( {\bf \Lambda }^1\right) \hookrightarrow {\bf J}^k
1787: \]
1788: the inclusion of the kernel of ${\bf J}^k\rightarrow {\bf J}^{k-1}$ (\cite
1789: {KLV}, p. 52). $A$ is of zero characteristic hence $\alpha $ is well defined
1790: and monic; ${\bf \Lambda }^1$ is projective hence $\beta $ is monic too.
1791: Thus $\rho $ is monic and we conclude.
1792:
1793: As a consequence $\forall n\geq 1$, we have the following short exact
1794: sequence in $\left[ {\frak D},{\frak D}\right] $:
1795:
1796: \begin{equation}
1797: 0\rightarrow {\rm D}_{\sigma \left( n\right) }\hookrightarrow {\rm D}%
1798: _{\sigma \left( n-1\right) }^{\bullet }\circ {\rm Diff}_{\sigma _n}^{\left(
1799: +\right) }\rightarrow {\rm D}_{\left( \sigma _1,...,\sigma _{n-2},\sigma
1800: _{n-1}+\sigma _n\right) }\rightarrow 0 \label{(13)}
1801: \end{equation}
1802: (the new fact is that the last arrow of the sequence is epic since it is
1803: induced by the gluing morphism).
1804:
1805: The $n$-th cohomology $K$-module of the complex
1806: \[
1807: {\bf dR}_\sigma :\quad 0\rightarrow A\stackrel{d_{\sigma \left( 1\right) }}{%
1808: \longrightarrow }{\bf \Lambda }^{\sigma \left( 1\right) }\rightarrow \ldots
1809: \rightarrow {\bf \Lambda }^{\sigma \left( n\right) }\stackrel{d_{\sigma
1810: \left( n+1\right) }}{\longrightarrow }{\bf \Lambda }^{\sigma \left(
1811: n+1\right) }\rightarrow \ldots
1812: \]
1813: is denoted by:
1814:
1815: \[
1816: H_\sigma ^n\doteq \frac{\ker \left( d_{\sigma \left( n+1\right) }\right) }{%
1817: im\left( d_{\sigma \left( n\right) }\right) }\equiv H^n\left( {\bf dR}%
1818: _\sigma \right) \text{.}
1819: \]
1820: Since $H_\sigma ^n$ only depends on $\sigma \left( n+1\right) $, we will
1821: write also $H_{\sigma \left( n+1\right) }^n$ in place of $H_\sigma ^n$.
1822:
1823: Note that in the situation of Remark \ref{rem 3.6} ({\bf iv}), $H_{\sigma
1824: \left( n+1\right) }^n$ is the $n$-th de Rham cohomology ${\Bbb R}$-vector
1825: space of the smooth manifold $M$.
1826:
1827: \smallskip\
1828:
1829: The rest of this Section will be devoted to proving the following result:
1830:
1831: \begin{theorem}
1832: \TeXButton{main}{\label{main}}\label{t1}{\rm (''Smooth'' rigidity of higher
1833: de Rham cohomologies )}
1834:
1835: If ${\frak D}$ is a smooth subcategory of $A-{\bf Mod}$, then, for each $%
1836: \tau ,\sigma \in {\bf N}_{+}^\infty $ with $\tau \geq \sigma $, the
1837: canonical ${\frak D}$-epimorphism (\ref{luppolo}):
1838:
1839: \[
1840: {\bf dR}_\tau \rightarrow {\bf dR}_\sigma
1841: \]
1842: is a quasi-isomorphism; so:
1843:
1844: \begin{equation}
1845: H_\sigma ^n\simeq H_\tau ^n\text{, }\forall n\geq 0\text{.} \label{(14)}
1846: \end{equation}
1847: \
1848: \end{theorem}
1849:
1850: \begin{corollary}
1851: ({\bf i}) If $M$ is a {\it smooth} manifold, $A=C^\infty \left( M;{\Bbb R}%
1852: \right) $ and ${\frak D}=C^\infty \left( M;{\Bbb R}\right) -{\bf Mod}_{geom}$%
1853: , then the higher de Rham cohomologies coincide with the standard de Rham
1854: cohomology of $M$.
1855:
1856: ({\bf ii}) If $K$ is an algebraically closed field of zero characteristic
1857: and $A$ is the coordinate ring of a {\it regular} affine variety over $K$,
1858: then the higher de Rham cohomologies coincide with the standard algebraic
1859: one.
1860: \end{corollary}
1861:
1862: Note that the last Corollary is false, in general, for a singular manifold
1863: or a non-regular affine variety.
1864:
1865: The strategy of the proof of Theorem \ref{t1} is the following.
1866:
1867: Keeping $n\geq 0$ fixed, we prove the thesis by reducing, step by step, each
1868: entry of $\sigma \left( n+1\right) $ to $1$, starting from $\sigma _{n+1}$,
1869: i.e. we prove the chain of isomorphisms
1870:
1871: \begin{equation}
1872: H_{\sigma \left( n+1\right) }^n\simeq H_{\left( \sigma \left( n\right)
1873: ,1\right) }^n\simeq H_{\left( \sigma \left( n-1\right) ,1,1\right) }^n\simeq
1874: \cdots \simeq H_{\left( \sigma _1,1,\ldots ,1\right) }^n\simeq H_{dR}^n
1875: \label{sediz}
1876: \end{equation}
1877: where $H_{dR}^n$ stands for $H_{\left( 1,...,1\right) }^n$, $\left(
1878: 1,...,1\right) \in {\bf N}_{+}^{n+1}$ (the $n$-th ordinary de Rham
1879: cohomology).
1880:
1881: The first step in the chain (\ref{sediz}) is obtained via the following%
1882: \footnote{%
1883: This Lemma has been proved, independently, also by Yu. Torkhov.}:
1884:
1885: \smallskip
1886:
1887: \begin{lemma}
1888: Let $n\in {\bf N}_{+}$. If $\sigma ,\tau \in {\bf N}_{+}^\infty $ are such
1889: that $\sigma \left( n\right) =\tau \left( n\right) $, then:
1890:
1891: (i) $\ker d_{\sigma \left( n+1\right) }=\ker d_{\tau \left( n+1\right) }$ ;
1892:
1893: (ii) $im\left( d_{\sigma \left( n+1\right) }\right) \simeq im\left( d_{\tau
1894: \left( n+1\right) }\right) $
1895:
1896: (where $\simeq $ means $K-{\bf Mod}$-isomorphism).\
1897: \end{lemma}
1898:
1899: \TeXButton{Proof}{\proof} (ii) follows trivially from (i). Let $\sigma \in
1900: {\bf N}_{+}^n$ and $k>1$. Consider the short exact sequence:
1901:
1902: \[
1903: 0\rightarrow {\rm D}_{\left( \sigma ,k-1,1\right) }\hookrightarrow {\rm D}%
1904: _{\left( \sigma ,k-1\right) }^{\bullet }\circ {\rm Diff}_1^{\left( +\right) }%
1905: \stackrel{i}{\longrightarrow }{\rm D}_{\left( \sigma ,k\right) }\rightarrow
1906: 0
1907: \]
1908: whose dual-representative:
1909:
1910: \begin{equation}
1911: 0\rightarrow {\bf \Lambda }^{\left( \sigma ,k\right) }\stackrel{i^{\vee }}{%
1912: \longrightarrow }{\bf J}^1\left( {\bf \Lambda }^{\left( \sigma ,k-1\right)
1913: }\right) \longrightarrow {\bf \Lambda }^{\left( \sigma ,k-1,1\right)
1914: }\rightarrow 0 \label{garrafugio}
1915: \end{equation}
1916: is likewise exact (in ${\frak D}$). We embed the latter in the commutative
1917: diagram:
1918:
1919: \[
1920: \begin{array}{ccccccccc}
1921: 0 & \rightarrow &
1922: \begin{array}{cc}
1923: &
1924: \end{array}
1925: {\bf \Lambda }^{\left( \sigma ,k\right) } & \stackrel{i^{\vee }}{%
1926: \longrightarrow } & {\bf J}^1\left( {\bf \Lambda }^{\left( \sigma
1927: ,k-1\right) }\right) & \longrightarrow & {\bf \Lambda }^{\left( \sigma
1928: ,k-1,1\right) } & \rightarrow & 0 \\
1929: & & d_{\left( \sigma ,k\right) }\uparrow & & \uparrow j_1 & & & & \\
1930: & &
1931: \begin{array}{cc}
1932: &
1933: \end{array}
1934: {\bf \Lambda }^\sigma & \stackunder{d_{\left( \sigma ,k-1\right) }}{%
1935: \longrightarrow } & {\bf \Lambda }^{\left( \sigma ,k-1\right) } & & & &
1936: \end{array}
1937: \]
1938:
1939: Now, if $\omega \in {\bf \Lambda }^\sigma $ is such that $d_{\left( \sigma
1940: ,k-1\right) }\left( \omega \right) =0$, then $j_1\left( d_{\left( \sigma
1941: ,k-1\right) }\left( \omega \right) \right) =0$ and, by commutativity, $%
1942: \left( i^{\vee }\circ d_{\left( \sigma ,k\right) }\right) \left( \omega
1943: \right) =0$. But $i^{\vee }$ is a monomorphism, so:
1944:
1945: \[
1946: \ker d_{\left( \sigma ,k-1\right) }\subseteq \ker d_{\left( \sigma ,k\right)
1947: }\text{, }\forall k>1\text{.}
1948: \]
1949: Since the inverse inclusion is obvious, (i) is proved.
1950:
1951: \TeXButton{End Proof}{\endproof}\
1952:
1953: To prove the ''$k$-th step'' of chain (\ref{sediz}), it is enough to show
1954: that:
1955: \begin{equation}
1956: H_{\left( \sigma \left( k-1\right) ,\sigma _k+1,1,\ldots ,1\right)
1957: _{n+1}}^n\simeq H_{\left( \sigma \left( k-1\right) ,\sigma _k,1,\ldots
1958: ,1\right) _{n+1}}^n \label{sizdiz}
1959: \end{equation}
1960: where we write $\left( \rho \right) _r$ if $\rho \in {\bf N}_{+}^r$.
1961:
1962: To prove (\ref{sizdiz}) we construct an auxiliary complex.
1963:
1964: \medskip\
1965:
1966: Let $P$ be an object in${\frak D}$ and $\tau \in {\bf N}_{+}^\infty $. As
1967: shown at the beginning of this section, smoothness of ${\frak D}$ implies
1968: that $\forall n>0$ the ''relative'' sequence (\ref{relHol})
1969:
1970: \begin{equation}
1971: 0\rightarrow {\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right]
1972: \longrightarrow \stackunder{}{\cal P}_{\tau \left( n-1\right) }^{\bullet
1973: }\left[ P\right] \circ {\rm Diff}_{\tau _n}^{\left( +\right)
1974: }\longrightarrow {\cal P}_{\left( \tau \left( n-2\right) ,\tau _{n-1}+\tau
1975: _n\right) }^{\bullet }\left[ P\right] \rightarrow 0 \label{(18)}
1976: \end{equation}
1977: is exact also in the last term; hence, its dual representative:
1978:
1979: \begin{equation}
1980: 0\rightarrow {\rm Hol}^{\left( \tau \left( n-2\right) ,\tau _{n-1}+\tau
1981: _n\right) }\left[ P\right] \longrightarrow {\bf J}^{\tau _n}\left( {\rm Hol}%
1982: ^{\tau \left( n-1\right) }\left[ P\right] \right) \longrightarrow {\rm Hol}%
1983: ^{\tau \left( n\right) }\left[ P\right] \rightarrow 0 \label{dizet}
1984: \end{equation}
1985: is exact also in the first term. Furthermore, when $P$ varies in $Ob\left(
1986: {\frak D}\right) $, (\ref{dizet}) gives rise to a short exact sequence in $%
1987: \left[ {\frak D},{\frak D}\right] $. We will refer to ${\rm Hol}^{\tau
1988: \left( n\right) }\left[ P\right] $ as the ${\rm Hol}${\em -object of type }$%
1989: \tau \left( n\right) $ of the $A$-module $P$; we have
1990:
1991: \begin{equation}
1992: {\rm Hol}^{\tau \left( n\right) }\left[ P\right] \simeq \frac{{\bf J}^{\tau
1993: _n}\left( {\rm Hol}^{\tau \left( n-1\right) }\left[ P\right] \right) }{{\rm %
1994: Hol}^{\left( \tau \left( n-2\right) ,\tau _{n-1}+\tau _n\right) }\left[
1995: P\right] } \label{dinoef}
1996: \end{equation}
1997: as $A$-modules. This allows us to give the following:
1998:
1999: \begin{PropDef}
2000: \TeXButton{plotigno}{\label{plotigno}}\label{p12} Let $P\in Ob\left( {\frak D%
2001: }\right) $ and $\tau \in {\bf N}_{+}^\infty $. We define the sequence in $%
2002: {\bf DIFF}_{A,{\frak D}}$:
2003: \[
2004: \begin{array}{c}
2005: {\bf Hol}^\tau \left( P\right) :\text{ }0\rightarrow {\rm Hol}^\emptyset
2006: \left[ P\right] \doteq P\stackrel{\delta _{\tau \left( 1\right) }\left(
2007: P\right) }{\longrightarrow }{\rm Hol}^{\tau \left( 1\right) }\left[ P\right]
2008: \equiv {\bf J}^{\tau _1}\left( P\right) \rightarrow ... \\
2009: \ldots \rightarrow {\rm Hol}^{\tau \left( n\right) }\left[ P\right]
2010: \stackrel{\delta _{\tau \left( n+1\right) }\left( P\right) }{\longrightarrow
2011: }{\rm Hol}^{\tau \left( n+1\right) }\left[ P\right] \rightarrow ...
2012: \end{array}
2013: \]
2014: where, for each $n\geq 0$,
2015: \begin{equation}
2016: \delta _{\tau \left( n+1\right) }\left[ P\right] :{\rm Hol}^{\tau \left(
2017: n\right) }\left[ P\right] \longrightarrow {\rm Hol}^{\tau \left( n+1\right)
2018: }\left[ P\right] \label{(20)}
2019: \end{equation}
2020: is defined to be the DO whose description (\ref{ire}) of Remark \ref{photos}
2021: is the canonical inclusion
2022: \[
2023: {\cal P}_{\tau \left( n+1\right) }^{\bullet }\left[ P\right] \hookrightarrow
2024: \stackunder{}{\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right]
2025: \circ {\rm Diff}_{\tau _{n+1}}^{\left( +\right) };
2026: \]
2027: equivalently, $\delta _{\tau \left( n+1\right) }\left[ P\right] $ is the
2028: composition:
2029:
2030: \[
2031: {\rm Hol}^{\tau \left( n\right) }\left[ P\right] \stackrel{j_{\tau
2032: _{n+1}}\left( {\rm Hol}^{\tau \left( n\right) }\left[ P\right] \right) }{%
2033: \longrightarrow }{\bf J}^{\tau _{n+1}}\left( {\rm Hol}^{\tau \left( n\right)
2034: }\left[ P\right] \right) \stackrel{p_{\tau \left( n+1\right) }(P)}{%
2035: \longrightarrow }\text{ }\frac{{\bf J}^{\tau _{n+1}}\left( {\rm Hol}^{\tau
2036: \left( n\right) }\left[ P\right] \right) }{{\rm Hol}^{\left( \tau \left(
2037: n-1\right) ,\tau _n+\tau _{n+1}\right) }\left[ P\right] }\simeq {\rm Hol}%
2038: ^{\tau \left( n+1\right) }\left[ P\right]
2039: \]
2040: where $p_{\tau \left( n+1\right) }(P)$ is the canonical quotient projection.
2041: ${\bf Hol}^\tau \left[ P\right] $ is a {\em complex} in ${\bf DIFF}_{A,%
2042: {\frak D}}$, called ${\rm Hol}^\tau ${\em -complex of }$P$; moreover, ${\bf %
2043: Hol}^\tau \left[ P\right] $ is natural in $P$ and defines a functor ${\bf Hol%
2044: }^\tau :{\frak D}\rightarrow ${\bf K}$\left( {\bf DIFF}_{A,{\frak D}}\right)
2045: $\footnote{%
2046: We recall that {\bf K}$\left( DIFF_{A,{\frak D}}\right) $ denotes the
2047: category of complexes of differential operators formed by objects of ${\frak %
2048: D}$.}.
2049: \end{PropDef}
2050:
2051: \TeXButton{Proof}{\proof} As always, it is better to work with functors
2052: (i.e. differential operators) than with representative objects. In the
2053: notations of Remark \ref{photos}, we have that $\varphi ^{\delta _{\tau
2054: \left( n+1\right) }\circ \delta _{\tau \left( n\right) }}$\footnote{%
2055: We write shortly $\delta _{\tau \left( k\right) }$ instead of $\delta _{\tau
2056: \left( k\right) }\left[ P\right] $, for any $k\geq 0$.} coincides with the
2057: composition:
2058:
2059: \[
2060: \begin{array}{cc}
2061: & \\
2062: {\cal P}_{\tau \left( n+1\right) }^{\bullet }\left[ P\right] & \\
2063: \downarrow & \varphi ^{\delta _{\tau \left( n+1\right) }{}_{}} \\
2064: {\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right] \circ {\rm Diff}%
2065: _{\tau _{n+1}}^{\left( +\right) } & \\
2066: \downarrow & \varphi ^{\delta _{\tau \left( n\right) }}\left( {\rm Diff}%
2067: _{\tau _{n+1}}^{+}\right) \\
2068: {\cal P}_{\tau \left( n-1\right) }^{\bullet }\left[ P\right] \circ \left(
2069: {\rm D}_{\left( \tau _n,\tau _{n+1}\right) }^{\bullet },{\rm D}_{\left( \tau
2070: _n,\tau _{n+1}\right) }^{+}\right) & \\
2071: \downarrow & {\cal P}_{\tau \left( n-1\right) }^{\bullet }\left[ P\right]
2072: \left( {\em C}_{\tau _n,\tau _{n+1}}\right) \\
2073: {\cal P}_{\tau \left( n-1\right) }^{\bullet }\left[ P\right] \circ {\rm Diff}%
2074: _{\tau _{n+1}+\tau _n}^{\left( +\right) } &
2075: \end{array}
2076: \]
2077: where the first two arrows are monomorphisms and the last is the ''gluing''
2078: morphism with respect to the indexes $\left( \tau _n,\tau _{n+1}\right) $.
2079: This composition is zero. In fact, if $Q\in Ob\left( {\frak D}\right) $ and $%
2080: \Delta \in \left[ {\cal P}_{\tau \left( n+1\right) }^{\bullet }\left[
2081: P\right] \right] \left( Q\right) $, then the image $\overline{\Delta }$ of $%
2082: \Delta $ via this composition, is defined by:
2083:
2084: \[
2085: \left( \overline{\Delta }\left( p\right) \right) \left( a_1\right) \cdots
2086: \left( a_{n-1}\right) \doteq \left( \left( \Delta \left( p\right) \right)
2087: \left( a_1\right) \cdots \left( a_{n-1}\right) \right) \left( 1\right) \text{%
2088: ,}
2089: \]
2090: and is zero because $\left( \Delta \left( p\right) \right) \left( a_1\right)
2091: \cdots \left( a_{n-1}\right) \in {\rm D}_{\tau _{n+1}}\left( Q\right) $, for
2092: each $p\in P$, $a_1,\ldots ,a_{n-1}\in A$.\qquad \TeXButton{End Proof}
2093: {\endproof}
2094:
2095: Now we show that if $\tau \in {\bf N}_{+}^\infty $ is {\it regular}, then,
2096: for any object $P$ in ${\frak D}$, ${\bf Hol}^\tau \left[ P\right] $ is {\it %
2097: acyclic}. In order to do this, we will exhibit (functorially in $P$) a {\it %
2098: trivializing homotopy}.
2099:
2100: Define:
2101: \begin{eqnarray*}
2102: \varphi _\emptyset \left( P\right) &:&{\cal P}_{\tau \left( -1\right)
2103: }^{\bullet }\left[ P\right] \doteq 0\longrightarrow {\cal P}_{\tau \left(
2104: 0\right) }^{\bullet }\left[ P\right] \doteq Hom_A\left( P,\cdot \right) \\
2105: \varphi _{\tau \left( 1\right) }\left( P\right) &:&{\cal P}_{\tau \left(
2106: 0\right) }^{\bullet }\left[ P\right] \doteq Hom_A\left( P,\cdot \right)
2107: \hookrightarrow {\cal P}_{\tau \left( 1\right) }^{\bullet }\left[ P\right]
2108: \equiv {\rm Diff}_{\tau _1}\left( P,\cdot \right)
2109: \end{eqnarray*}
2110: which are morphisms in $\left[ {\frak D},{\frak D}\right] $; then define, by
2111: induction on $n$,
2112:
2113: \[
2114: \varphi ^{\tau \left( n+1\right) }\left( P\right) \doteq i_{\tau \left(
2115: n+1\right) }\left( P\right) -\widehat{\varphi }_{\tau \left( n\right)
2116: }\left( P\right) \text{,}
2117: \]
2118: where
2119:
2120: \[
2121: i_{\tau \left( n+1\right) }\left( P\right) :{\cal P}_{\tau \left( n\right)
2122: }^{\bullet }\left[ P\right] \simeq {\cal P}_{\tau \left( n\right) }^{\bullet
2123: }\left[ P\right] \circ {\rm Diff}_0^{\left( +\right) }\hookrightarrow {\cal P%
2124: }_{\tau \left( n\right) }^{\bullet }\left[ P\right] \circ {\rm Diff}_{\tau
2125: _{n+1}}^{\left( +\right) }
2126: \]
2127:
2128: \begin{eqnarray*}
2129: \widehat{\varphi }_{\tau \left( n\right) }\left( P\right) &:&{\cal P}_{\tau
2130: \left( n\right) }^{\bullet }\left[ P\right] \hookrightarrow {\cal P}_{\tau
2131: \left( n-1\right) }^{\bullet }\left[ P\right] \circ {\rm Diff}_{\tau
2132: _n}^{\left( +\right) }\stackrel{\varphi _{\tau \left( n+1\right) }\left(
2133: P\right) \left( {\rm Diff}_{\tau _n}^{+}\right) }{\longrightarrow } \\
2134: \ &\rightarrow &{\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right]
2135: \circ {\rm Diff}_{\tau _n}^{\left( +\right) }\hookrightarrow {\cal P}_{\tau
2136: \left( n\right) }^{\bullet }\left[ P\right] \circ {\rm Diff}_{\tau
2137: _{n+1}}^{\left( +\right) }.
2138: \end{eqnarray*}
2139: With this definition, $\varphi _{\tau \left( n+1\right) }\left( P\right) :%
2140: {\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right] \rightarrow {\cal %
2141: P}_{\tau \left( n\right) }^{\bullet }\left[ P\right] \circ {\rm Diff}_{\tau
2142: _{n+1}}^{\left( +\right) }$, but it is easy to resolve the inductive
2143: definition in the following one:
2144:
2145: \begin{eqnarray}
2146: \left\{ \left[ \varphi _{\tau \left( n+1\right) }\left( P\right) \left(
2147: Q\right) \right] \left( \Delta \right) \right\} \left( p\right) \left(
2148: a_1\right) \cdots \left( a_n\right) &=&a_n\Delta \left( p\right) \left(
2149: a_1\right) \cdots \left( a_{n-1}\right) + \label{lilla} \\
2150: &&\ \sum\limits_{k=1}^{n-1}\left( -1\right) ^{n-k}\Delta \left( p\right)
2151: \left( a_1\right) \cdots \left( a_ka_{k+1}\right) \cdots \left( a_n\right) +
2152: \nonumber \\
2153: &&\ +\left( -1\right) ^n\Delta \left( a_1p\right) \left( a_2\right) \cdots
2154: \left( a_n\right) \nonumber
2155: \end{eqnarray}
2156: ($Q\in Ob\left( {\frak D}\right) $, $p\in P$, $a_1,\ldots ,a_n\in A$ and $%
2157: \Delta \in {\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right] \left(
2158: Q\right) $). This shows that actually
2159: \[
2160: \varphi _{\tau \left( n+1\right) }\left( P\right) :{\cal P}_{\tau \left(
2161: n\right) }^{\bullet }\left[ P\right] \rightarrow {\cal P}_{\tau \left(
2162: n+1\right) }^{\bullet }\left[ P\right] .
2163: \]
2164: Therefore, we get a family
2165:
2166: \[
2167: \left\{ \varphi _{\tau \left( n\right) }\left( P\right) :{\cal P}_{\tau
2168: \left( n-1\right) }^{\bullet }\left[ P\right] \rightarrow {\cal P}_{\tau
2169: \left( n\right) }^{\bullet }\left[ P\right] \right\} _{n>0}
2170: \]
2171: of morphisms in $\left[ {\frak D},{\frak D}\right] $. Of course, formula (%
2172: \ref{lilla}) can equally be taken as the definition of the family $\left\{
2173: \varphi _{\tau \left( n\right) }\left( P\right) \right\} _{n>0}$ but the
2174: inductive definition can be ''dualized'', to representative objects, to give
2175: the following (keeping the notations of Proposition \ref{plotigno}):
2176:
2177: \begin{eqnarray*}
2178: \varphi ^\emptyset \left( P\right) &:&{\rm Hol}^{\tau \left( 0\right)
2179: }\left[ P\right] \doteq P\longrightarrow {\rm Hol}^{\tau \left( -1\right)
2180: }\left[ P\right] \doteq 0 \\
2181: \varphi ^{\tau \left( 1\right) }\left( P\right) &:&{\rm Hol}^{\tau \left(
2182: 1\right) }\left( P\right) \doteq {\bf J}^{\tau _1}\left( P\right)
2183: \longrightarrow {\rm Hol}^{\tau \left( 0\right) }\left[ P\right] \doteq P%
2184: \text{ (natural projection)} \\
2185: \varphi ^{\tau \left( n+1\right) }\left( P\right) &:&{\rm Hol}^{\tau \left(
2186: n+1\right) }\left[ P\right] \longrightarrow {\rm Hol}^{\tau \left( n\right)
2187: }\left[ P\right]
2188: \end{eqnarray*}
2189: where $\varphi ^{\tau \left( n+1\right) }\left( P\right) $ is the only $%
2190: {\frak D}$-morphism corresponding to the DO of order $\leq \tau _{n+1}$
2191:
2192: \[
2193: \Delta _{\tau \left( n\right) }\doteq {\rm id}_{{\rm Hol}^{\tau \left(
2194: n\right) }\left[ P\right] }-\delta _{\tau \left( n\right) }\left( P\right)
2195: \circ \varphi ^{\tau \left( n\right) }\left( P\right) :{\rm Hol}^{\tau
2196: \left( n\right) }\left[ P\right] \longrightarrow {\rm Hol}^{\tau \left(
2197: n\right) }\left[ P\right] \text{.}
2198: \]
2199: As before, but dually\footnote{%
2200: Subfunctors of strictly representable functors correspond to quotient
2201: objects of the representatives.}, this definition gives apparently a ${\frak %
2202: D}$-morphism:
2203:
2204: \[
2205: \widehat{\varphi }^{\tau \left( n+1\right) }\left( P\right) :{\bf J}^{\tau
2206: _{n+1}}\left( {\rm Hol}^{\tau \left( n\right) }\left[ P\right] \right)
2207: \longrightarrow {\rm Hol}^{\tau \left( n\right) }\left[ P\right]
2208: \]
2209: ($\tau $ being regular) but formula (\ref{lilla}) shows that actually
2210:
2211: \[
2212: \ker \left( \widehat{\varphi }^{\tau \left( n+1\right) }\left( P\right)
2213: \right) \supseteq {\rm Hol}^{\left( \tau \left( n-1\right) ,\tau _n+\tau
2214: _{n+1}\right) }\left[ P\right]
2215: \]
2216: so that, by formula (\ref{dinoef}), $\widehat{\varphi }^{\tau \left(
2217: n+1\right) }\left( P\right) $ induces, by passing to the quotient, the
2218: morphism $\varphi ^{\tau \left( n+1\right) }\left( P\right) $ we wanted.
2219:
2220: Now we have a family of ${\frak D}$-morphisms $\left\{ \varphi ^{\tau \left(
2221: n\right) }\left( P\right) :{\rm Hol}^{\tau \left( n\right) }\left[ P\right]
2222: \rightarrow {\rm Hol}^{\tau \left( n-1\right) }\left[ P\right] \right\}
2223: _{n\geq 0}$ dual to $\left\{ \varphi _{\tau \left( n\right) }\left( P\right)
2224: :{\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right] \rightarrow
2225: {\cal P}_{\tau \left( n-1\right) }^{\bullet }\left[ P\right] \right\}
2226: _{n\geq 0}$.
2227:
2228: \begin{proposition}
2229: \TeXButton{dan}{\label{dan}}For each object $P$ in ${\frak D}$ and for each
2230: regular $\tau \in {\bf N}_{+}^\infty $, $\left\{ \varphi ^{\tau \left(
2231: n\right) }\left( P\right) \right\} _{n\geq 0}$ is a trivializing homotopy
2232: for ${\bf Hol}^\tau \left( P\right) $. Furthermore, $\left\{ \varphi ^{\tau
2233: \left( n\right) }\left( P\right) \right\} _{n\geq 0}$ is natural in $P$.
2234: \end{proposition}
2235:
2236: \TeXButton{Proof}{\proof} We must show that the {\it sum} $L+R$ of the two
2237: compositions:
2238:
2239: \begin{eqnarray*}
2240: L &:&{\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right]
2241: \hookrightarrow \stackunder{}{\cal P}_{\tau \left( n-1\right) }^{\bullet
2242: }\left[ P\right] \circ {\rm Diff}_{\tau _n}^{\left( +\right) }\stackrel{%
2243: \varphi _{\tau \left( n\right) }\left( P\right) \left( {\rm Diff}_{\tau
2244: _n}^{+}\right) }{\longrightarrow } \\
2245: &\rightarrow &\stackunder{}{\cal P}_{\tau \left( n\right) }^{\bullet }\left[
2246: P\right] \circ {\rm Diff}_{\tau _n}^{\left( +\right) }\hookrightarrow
2247: \stackunder{}{\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right]
2248: \circ {\rm Diff}_{\tau _{n+1}}^{\left( +\right) }
2249: \end{eqnarray*}
2250:
2251: \[
2252: R:{\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right] \stackrel{%
2253: \varphi _{\tau \left( n+1\right) }\left( P\right) }{\longrightarrow }{\cal P}%
2254: _{\tau \left( n+1\right) }^{\bullet }\left[ P\right] \hookrightarrow
2255: \stackunder{}{\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right]
2256: \circ {\rm Diff}_{\tau _{n+1}}^{\left( +\right) }
2257: \]
2258: equals ${\rm id}_{{\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right]
2259: }$ (which is then homotopic to the zero map) or, equivalently, that the
2260: diagram:
2261:
2262: \begin{equation}
2263: \begin{tabular}{lll}
2264: $\quad {\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right] $ & $%
2265: \stackrel{L+R}{\longrightarrow }$ & $\stackunder{}{\cal P}_{\tau \left(
2266: n\right) }^{\bullet }\left[ P\right] \circ {\rm Diff}_{\tau _{n+1}}^{\left(
2267: +\right) }$ \\
2268: $\qquad \quad \quad {\rm id}$ & $\searrow $ & $\quad \quad \cup $ \\
2269: & & $\stackunder{}{\cal P}_{\tau \left( n\right) }^{\bullet }\left[
2270: P\right] \circ {\rm Diff}_0^{\left( +\right) }\quad $%
2271: \end{tabular}
2272: \label{tomaso}
2273: \end{equation}
2274: is commutative. For $Q\in Ob\left( {\frak D}\right) $ and $\Delta \in {\cal P%
2275: }_{\tau \left( n\right) }^{\bullet }\left[ P\right] \left( Q\right) $, we
2276: have by (\ref{lilla}):
2277:
2278: \[
2279: L\left( \Delta \right) \left( p\right) \left( a_1\right) \cdots \left(
2280: a_n\right) =
2281: \]
2282: \[
2283: =\left[ a_{n-1}^{+}\Delta \left( p\right) \left( a_1\right) \cdots \left(
2284: a_{n-2}\right) +\sum\limits_{s=1}^{n-2}\left( -1\right) ^{n-1-s}\Delta
2285: \left( p\right) \left( a_1\right) \cdots \left( a_sa_{s+1}\right) \cdots
2286: \left( a_{n-1}\right) +\right.
2287: \]
2288: \[
2289: \left. +\left( -1\right) ^{n-1}\Delta \left( a_1p\right) \left( a_2\right)
2290: \cdots \left( a_{n-1}\right) \right] \left( a_n\right)
2291: =-\sum\limits_{s=1}^{n-1}\left( -1\right) ^{n-s}\Delta \left( p\right)
2292: \left( a_1\right) \cdots \left( a_sa_{s+1}\right) \cdots \left( a_n\right) +
2293: \]
2294:
2295: \[
2296: +\left( -1\right) _{}^{n-1}\Delta \left( a_1p\right) \left( a_2\right)
2297: \cdots \left( a_n\right)
2298: \]
2299: while
2300:
2301: \begin{eqnarray*}
2302: R\left( \Delta \right) \left( p\right) \left( a_1\right) \cdots \left(
2303: a_n\right) &=&a_n\Delta \left( p\right) \left( a_1\right) \cdots \left(
2304: a_{n-1}\right) + \label{lilla} \\
2305: &&\ \sum\limits_{k=1}^{n-1}\left( -1\right) ^{n-k}\Delta \left( p\right)
2306: \left( a_1\right) \cdots \left( a_ka_{k+1}\right) \cdots \left( a_n\right) +
2307: \\
2308: &&\ +\left( -1\right) ^n\Delta \left( a_1p\right) \left( a_2\right) \cdots
2309: \left( a_n\right)
2310: \end{eqnarray*}
2311: so that
2312:
2313: \[
2314: \left( L+R\right) \left( \Delta \right) \left( p\right) \left( a_1\right)
2315: \cdots \left( a_n\right) =a_n\Delta \left( p\right) \left( a_1\right) \cdots
2316: \left( a_{n-1}\right)
2317: \]
2318: i.e. $\left( L+R\right) \left( \Delta \right) $ coincides with the image of $%
2319: \Delta $ via the inclusion
2320:
2321: \[
2322: {\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right] \left( Q\right)
2323: \simeq \stackunder{}{\cal P}_{\tau \left( n\right) }^{\bullet }\left[
2324: P\right] \circ {\rm Diff}_0^{(+)}\left( Q\right) \hookrightarrow \stackunder{%
2325: }{\cal P}_{\tau \left( n\right) }^{\bullet }\left[ P\right] \circ {\rm Diff}%
2326: _{\tau _{n+1}}^{(+)}\left( Q\right) \text{.}
2327: \]
2328: \TeXButton{End Proof}{\endproof}
2329:
2330: \medskip\
2331:
2332: We now use acyclicity of the ${\bf Hol}^{{\bf 1}}$-complex,${\bf 1}%
2333: =(1,...,1,1,...,1,...)\in {\bf N}_{+}^\infty $, to prove the ''$k$-th step''
2334: i.e. formula (\ref{sizdiz}). Let $\sigma =(\sigma _1,...,\sigma _k+1)\in
2335: {\bf N}_{+}^k$, $(\sigma ,{\bf 1})=(\sigma _1,...,\sigma
2336: _k+1,1,1,...,1,...)\in {\bf N}_{+}^\infty $ and
2337: \[
2338: {\bf K}_{(\sigma ,{\bf 1})}^{\left( k\right) }\doteq \ker \left( {\bf dR}%
2339: _{(\sigma ,{\bf 1})}\rightarrow {\bf dR}_{(\sigma _1,...,\sigma
2340: _k,1,1,...,1,...)}\right) .
2341: \]
2342: For each $\left( \mu \right) _s=$ $\left( \mu _1,...,\mu _s\right) \in {\bf N%
2343: }_{+}^s$, $1\leq r\leq s$, $r,s\in {\bf N}_{+}$, we put:
2344:
2345: \[
2346: K_{(\mu )_s}^{\left( r\right) }\doteq \ker \left( {\bf \Lambda }^{\left( \mu
2347: _1,...,\mu _r,...\mu _s\right) }\rightarrow {\bf \Lambda }^{\left( \mu
2348: _1,...,\mu _r-1,...\mu _s\right) }\right) \text{.}
2349: \]
2350: To prove the ''$k$-th step'' it is enough to show acyclicity of ${\bf K}%
2351: _{(\sigma ,{\bf 1})}^{\left( k\right) }$. We claim that there exists a
2352: resolution of ${\bf K}_{(\sigma ,{\bf 1})}^{\left( k\right) }$ of the form:
2353: \[
2354: \cdots \rightarrow {\bf Hol}^{{\bf 1}}\left[ K_{\left( \sigma _1,...,\sigma
2355: _k+l+1\right) }^{(k)}\right] \left[ -k-l\right] \stackrel{\psi _l\left[
2356: -k+l\right] }{\longrightarrow }{\bf Hol}^{{\bf 1}}\left[ K_{\left( \sigma
2357: _1,...,\sigma _k+l\right) }^{(k)}\right] \left[ -k-l+1\right] \rightarrow
2358: \cdots
2359: \]
2360: \begin{equation}
2361: \cdots \stackunder{}{\stackunder{\psi _2\left[ -\left( k+2\right) \right] }{%
2362: \rightarrow }}{\bf Hol}^{{\bf 1}}\left[ K_{\left( \sigma _1,...,\sigma
2363: _k+2\right) }^{(k)}\right] \left[ -k-1\right] \stackunder{\psi _1\left[
2364: -k+1\right] }{\longrightarrow }{\bf Hol}^{{\bf 1}}\left[ K_{\left( \sigma
2365: _1,...,\sigma _k+1\right) }^{(k)}\right] \left[ -k\right] \stackunder{\rho }{%
2366: \longrightarrow }{\bf K}_{(\sigma ,{\bf 1})}^{\left( k\right) }\rightarrow 0
2367: \label{res}
2368: \end{equation}
2369: where if $r\in {\bf N}_{+}$, $\left( \cdot \right) \left[ r\right] $
2370: denotes, as usual, the $r$-shift both for complexes and morphisms of
2371: complexes. We postpone in the Appendix the definition of the maps of
2372: complexes
2373: \begin{eqnarray*}
2374: \psi _l &:&{\bf Hol}^{{\bf 1}}\left[ K_{\left( \sigma _1,...,\sigma
2375: _k+l+1\right) }^{(k)}\right] \stackrel{}{\longrightarrow }{\bf Hol}^{{\bf 1}%
2376: }\left[ K_{\left( \sigma _1,...,\sigma _k+l\right) }^{(k)}\right] \left[
2377: 1\right] ,\text{ \quad }l\in {\bf N}_{+}, \\
2378: \rho &:&{\bf Hol}^{{\bf 1}}\left[ K_{\left( \sigma _1,...,\sigma _k+1\right)
2379: }^{(k)}\right] \left[ -k\right] \stackunder{}{\longrightarrow }{\bf K}%
2380: _{(\sigma ,{\bf 1})}^{\left( k\right) }
2381: \end{eqnarray*}
2382: and the proof that (\ref{res}) is actually a resolution.
2383:
2384: Assuming the existence of resolution (\ref{res}), the acyclicity of ${\bf K}%
2385: _{(\sigma ,{\bf 1})}^{\left( k\right) }$ is then an immediate consequence of
2386: acyclicity of ${\bf Hol}^{{\bf 1}}$-complexes together with the following
2387: elementary fact
2388:
2389: \begin{lemma}
2390: Let $C^{\cdot },$ $P_i^{\cdot },$ $i>0,$ be cochain complexes in $A-{\bf Mod}
2391: $ and
2392: \[
2393: \cdots \longrightarrow P_n^{\cdot }\longrightarrow P_{n-1}^{\cdot
2394: }\longrightarrow \cdots \longrightarrow P_1^{\cdot }\longrightarrow C^{\cdot
2395: }\rightarrow 0
2396: \]
2397: be a resolution of $C^{\cdot }.$ Suppose that $\forall i\geq 1$, $P_i^k=(0)$
2398: $\forall k<0$ (so that $C^k=(0)$ $\forall k<0$ too). If each $P_i^{\cdot }$
2399: is acyclic then so is $C^{\cdot }$.
2400: \end{lemma}
2401:
2402: \TeXButton{Proof}{\proof} It follows from the hypotheses that $C^{\cdot }$
2403: is isomorphic in the derived category $D^{+}\left( A-{\bf Mod}\right) $ to
2404: the total complex associated to the double complex induced by
2405: \[
2406: \cdots \longrightarrow P_n^{\cdot }\longrightarrow P_{n-1}^{\cdot
2407: }\longrightarrow \cdots \longrightarrow P_1^{\cdot }\text{ }
2408: \]
2409: which is acyclic.
2410:
2411: For those readers who feel uncomfortable with derived categories, here is a
2412: more ''step-by-step'' proof. By the usual sign-trick we can associate to the
2413: given resolution a $1^{st}$- quadrant double complex $R_{\cdot }^{\cdot
2414: }=\left( R_q^p\right) $ which is mixed: homological in the vertical (i.e.
2415: with $p$ fixed) direction and cohomological in the horizontal (i.e. with $p$
2416: fixed) direction. We turn it into a 2$^{nd}$-quadrant homological double
2417: complex $\widehat{R}_{\cdot \text{ }\cdot }=\left( \widehat{R}_{pq}\right) $
2418: with $\widehat{R}_{pq}\doteq R_q^{-p}$.
2419:
2420: Consider the spectral sequence induced by the ''filtration by rows'' on $%
2421: \widehat{R}_{\cdot \text{ }\cdot }$ (e.g. \cite{We} p. 142):
2422:
2423: \[
2424: ^{II}E_{pq}^0=\widehat{R}_{qp}
2425: \]
2426: with $^{II}d_{pq}^0$ given by the horizontal differential in $\widehat{R}%
2427: _{\cdot \text{ }\cdot }$ . Then $\left\{ ^{II}E_{pq}^1,^{II}d_{pq}^1\right\}
2428: $ is just:
2429:
2430: \begin{eqnarray*}
2431: &&
2432: \begin{tabular}{l}
2433: $q=n+1$ \\
2434: $q=n$ \\
2435: $\quad \vdots $ \\
2436: $q=2$ \\
2437: $q=1$ \\
2438: $q=0$%
2439: \end{tabular}
2440: \begin{tabular}{|l|l|l|l|l|l|}
2441: \hline
2442: $\leftarrow 0$ & $\leftarrow H^{n+1}\left( C^{\cdot }\right) $ & $\leftarrow
2443: 0$ & $\leftarrow 0$ & & $\leftarrow 0$ \\ \hline
2444: $\leftarrow 0$ & $\leftarrow H^n\left( C^{\cdot }\right) $ & $\leftarrow 0$
2445: & $\leftarrow 0$ & & $\leftarrow 0$ \\ \hline
2446: $\quad \vdots $ & $\qquad \vdots $ & \quad $\vdots $ & \quad $\vdots $ & &
2447: \quad $\vdots $ \\ \hline
2448: $\leftarrow 0$ & $\leftarrow H^2\left( C^{\cdot }\right) $ & $\leftarrow 0$
2449: & $\leftarrow 0$ & & $\leftarrow 0$ \\ \hline
2450: $\leftarrow 0$ & $\leftarrow H^1\left( C^{\cdot }\right) $ & $\leftarrow 0$
2451: & $\leftarrow 0$ & & $\leftarrow 0$ \\ \hline
2452: $\leftarrow 0$ & $\leftarrow H^0(C^{\cdot })$ & $\leftarrow 0$ & $\leftarrow
2453: 0$ & & $\leftarrow 0$ \\ \hline
2454: \end{tabular}
2455: \\
2456: &&
2457: \begin{tabular}{lllllll}
2458: $\quad \quad \quad p=-1\quad $ & $\quad 0\quad $ & $\quad \qquad 1\quad $ & $%
2459: \quad 2$ & $\cdots $ & $\quad n$ &
2460: \end{tabular}
2461: \end{eqnarray*}
2462: (with differential induced by vertical differential in $\widehat{R}_{\cdot
2463: \text{ }\cdot }$). So $^{II}E$ degenerates at $^{II}E^1$. But $\widehat{R}%
2464: _{\cdot \text{ }\cdot }$ is a 2$^{nd}$-quadrant double complex hence $^{II}E$
2465: converges to $H^{*}\left( Tot\left( \widehat{R}_{\cdot \text{ }\cdot
2466: }\right) \right) $, $Tot\left( \widehat{R}_{\cdot \text{ }\cdot }\right) $
2467: being the total complex associated to $\widehat{R}_{\cdot \text{ }\cdot }$,
2468: which is zero since by hypothesis $\widehat{R}_{\cdot \text{ }\cdot }$ has
2469: exact columns. So $^{II}E_{pq}^1=(0)$ $\forall p,q$ and $C^{\cdot }$ is
2470: acyclic. \TeXButton{End Proof}{\endproof}
2471:
2472: \begin{corollary}
2473: \TeXButton{ty}{\label{ty}}Let ${\frak D}\subseteq A-{\bf Mod}$ be a
2474: differentially closed smooth subcategory. If we define the {\em stable}
2475: infinite de Rham complex to be
2476: \[
2477: {\bf dR}_\infty ^{\text{st}}\doteq \underleftarrow{\lim }_{k>0}{\bf dR}_{%
2478: {\bf k}}
2479: \]
2480: (where ${\bf k}\doteq \left( k,...,k,k,...k,...\right) $ ) then the
2481: canonical morphism ${\bf dR}_\infty ^{\text{st}}\longrightarrow {\bf dR}_{%
2482: {\bf 1}}$ is a quasi-isomorphism.
2483: \end{corollary}
2484:
2485: \TeXButton{Proof}{\proof} We use the following facts:
2486:
2487: (i) the index category of the inverse system which defines ${\bf dR}_\infty
2488: ^{\text{st}}$ is countable;
2489:
2490: (ii) the canonical ${\frak D}$-morphisms ${\bf dR}_{\underline{k^{\prime }}%
2491: }\rightarrow {\bf dR}_{\underline{k}}$, $k^{\prime }\geq k$, in the inverse
2492: system are epimorphisms.
2493:
2494: If we denote by $\underleftarrow{\lim }_k^1$ the first right derived functor
2495: of $\underleftarrow{\lim }_k$, (i) and (ii) imply, via standard spectral
2496: sequence's arguments (e.g. \cite{Lu} Cor. 1.1, p. 535), that there is a
2497: short exact sequence
2498:
2499: \[
2500: 0\rightarrow \underleftarrow{\lim }_{k>0}^1\left[ H^{n-1}\left( {\bf dR}_{%
2501: {\bf k}}\right) \right] \rightarrow H^n\left( {\bf dR}_\infty ^{\text{st}%
2502: }\right) \rightarrow \underleftarrow{\lim }_{k>0}\left[ H^n\left( {\bf dR}_{%
2503: {\bf k}}\right) \right] \rightarrow 0.
2504: \]
2505: By theorem \ref{main}, the term on the right is isomorphic to $H_{dR}^n$, so
2506: we are left to prove that $\underleftarrow{\lim }_{k>0}\left[ H^{n-1}\left(
2507: {\bf dR}_{\underline{k}}\right) \right] =\left( 0\right) $. But $%
2508: \underleftarrow{\lim }_{k>0}^1$is right exact and
2509:
2510: \[
2511: H^{n-1}\left( {\bf dR}_{{\bf k}}\right) \equiv H_{{\bf k}}^{n-1}\cong
2512: H_{dR}^{n-1}\doteq H_{{\bf 1}}^{n-1}\text{, }\forall n\geq 1\text{, }
2513: \]
2514: by theorem \ref{main}, therefore it will be enough to prove the vanishing of
2515: $\underleftarrow{\lim }_{k>0}^1$ for the constant inverse system
2516: \[
2517: \cdots \rightarrow H_{dR}^{n-1}\stackrel{id}{\longrightarrow }H_{dR}^{n-1}%
2518: \stackrel{id}{\longrightarrow }H_{dR}^{n-1}\rightarrow \cdots
2519: \]
2520:
2521: But this is an easy consequence of the following description (due to
2522: Eilenberg) of $\underleftarrow{\lim }_{k>0}^1$for constant systems.
2523:
2524: If we define
2525:
2526: \[
2527: D_0:\prod\limits_{k\in {\bf N}_{+}}H_{dR}^{n-1}\longrightarrow
2528: \prod\limits_{k\in {\bf N}_{+}}H_{dR}^{n-1}:\left( \alpha _k\right) _{k\in
2529: {\bf N}_{+}}\longmapsto \left( \alpha _{k+1}-\alpha _k\right) _{k\in {\bf N}%
2530: _{+}}\text{ ;}
2531: \]
2532: then
2533:
2534: \[
2535: {\rm co}\ker \left( D_0\right) =\underleftarrow{\lim }_{k>0}^1\left[
2536: H_{dR}^{n-1}\right] \text{.}\label{4.30}
2537: \]
2538: Let $\left( \omega _k\right) _{k>0}\in \prod\limits_{k\in {\bf N}%
2539: _{+}}H_{dR}^{n-1}$ and define $\left( \overline{\omega }_k\right) _{k>0}$ as
2540: $\overline{\omega }_k\doteq \sum\limits_{l=1}^{k-1}\omega _l$; then
2541: \[
2542: D_0\left( \left( \overline{\omega }_k\right) _{k>0}\right) =\left( \overline{%
2543: \omega }_{k+1}-\overline{\omega }_k\right) _{k>0}=\left( \omega _k\right)
2544: _{k>0}.
2545: \]
2546: Therefore $D_0$ is surjective and we conclude. \TeXButton{End Proof}
2547: {\endproof}
2548:
2549: \begin{corollary}
2550: \label{markandeja} Let ${\frak D}\subseteq A-{\bf Mod}$ be a differentially
2551: closed smooth subcategory such that $\forall \sigma \in {\bf N}_{+}^\infty $%
2552: , $\exists n_\sigma \in {\bf N}_{+}$ :
2553:
2554: \[
2555: {\bf \Lambda }^{\sigma \left( r\right) }=(0)\text{, }\forall r>n_\sigma
2556: \text{.}
2557: \]
2558: Then the canonical morphism ( Def. \ref{infty})
2559:
2560: \[
2561: {\bf dR}_\infty \rightarrow {\bf dR}_\sigma
2562: \]
2563: is a quasi-isomorphism $\forall \sigma \in {\bf N}_{+}^\infty $.
2564: \end{corollary}
2565:
2566: \TeXButton{Proof}{\proof} Under our hypotheses
2567:
2568: \[
2569: \widehat{{\bf N}}_{+}\doteq \bigsqcup\limits_{k\in {\bf N}_{+}}\left\{
2570: \underline{k}\in {\bf N}_{+}^\infty \mid \underline{k}\left( n\right) \equiv
2571: \left( k,...,k\right) \in {\bf N}_{+}^n\text{, }\forall n\in {\bf N}%
2572: _{+}\right\}
2573: \]
2574: is cofinal in the index category of the system $\left\{ {\bf dR}_\sigma
2575: \right\} $; hence ${\bf dR}_\infty \simeq \underleftarrow{\lim }_{k>0}{\bf dR%
2576: }_{{\bf k}}$ and the thesis follows from corollary \ref{ty}.
2577: \TeXButton{End Proof}{\endproof}
2578:
2579: \section{Appendix}
2580:
2581: This appendix is devoted to defining the maps in the sequence (\ref{res})
2582: and to showing that (\ref{res}) is exact.
2583:
2584: There are two kinds of maps of complexes to be defined:
2585: \[
2586: \rho \equiv \left( \rho ^n:{\bf Hol}^{{\bf 1}_{n-k}}\left[ K_{\left( \sigma
2587: _1,...,\sigma _k+1\right) }^{\left( k\right) }\right] \rightarrow K_{\left(
2588: \sigma _1,...,\sigma _k+1,1,...,1\right) _n}^{\left( k\right) }\right)
2589: _{n\geq 0}
2590: \]
2591: \[
2592: \psi _l\equiv \left( \psi _l^n\right) _{n\geq 0}:{\bf Hol}^{{\bf 1}}\left[
2593: K_{\left( \sigma _1,...,\sigma _k+l+1\right) }^{\left( k\right) }\right]
2594: \rightarrow {\bf Hol}^{{\bf 1}}\left[ K_{\left( \sigma _1,...,\sigma
2595: _k+l\right) }^{\left( k\right) }\right] \left[ 1\right] .
2596: \]
2597:
2598: Let us first define $\rho $. We will define a functorial morphism
2599: \[
2600: \Theta :{\rm D}_{\left( \sigma _1,...,\sigma _k+1,1,...,1\right)
2601: _n}\longrightarrow {\cal P}_{{\bf 1}_{n-k}}^{\bullet }\left[ K_{\left(
2602: \sigma _1,...,\sigma _k+1\right) }^{\left( k\right) }\right]
2603: \]
2604: and show that the sequence
2605: \[
2606: 0\rightarrow {\rm D}_{\left( \sigma _1,...,\sigma _k,1,...,1\right) _n}%
2607: \stackrel{}{\stackrel{\Phi }{\hookrightarrow }}{\rm D}_{\left( \sigma
2608: _1,...,\sigma _k+1,1,...,1\right) _n}\stackrel{\Theta }{\longrightarrow }%
2609: {\cal P}_{{\bf 1}_{n-k}}^{\bullet }\left[ K_{\left( \sigma _1,...,\sigma
2610: _k+1\right) }^{\left( k\right) }\right]
2611: \]
2612: is exact so that the dual representative of $\Theta $ will pass to the
2613: quotient defining our surjective $\rho $.
2614:
2615: Define $\Theta $ to be the following composition:
2616: \begin{eqnarray*}
2617: {\rm D}_{\left( \sigma _1,...,\sigma _k+1,1,...,1\right) _n}\simeq {\rm D}%
2618: _{\left( \sigma _1,...,\sigma _k+1,1\right) _{k+1}}^{\bullet }\left( {\rm D}%
2619: _{{\bf 1}_{n-k-1}}\subset {\rm Diff}_{{\bf 1}_{n-k-1}}^{+}\right) \simeq
2620: \end{eqnarray*}
2621: \begin{eqnarray*}
2622: \ &\simeq &Hom_A^{\bullet }\left( {\bf \Lambda }^{\left( \sigma
2623: _1,...,\sigma _k+1,1\right) },{\rm D}_{{\bf 1}_{n-k-1}}\subset {\rm Diff}_{%
2624: {\bf 1}_{n-k-1}}^{+}\right) \stackrel{\circ \text{ }d_{\left( \sigma
2625: _1,...,\sigma _k+1,1\right) }}{\longrightarrow } \\
2626: \ &\longrightarrow &{\rm Diff}_1^{\bullet }\left( {\bf \Lambda }^{\left(
2627: \sigma _1,...,\sigma _k+1\right) },{\rm D}_{{\bf 1}_{n-k-1}}\subset {\rm Diff%
2628: }_{{\bf 1}_{n-k-1}}^{+}\right) \simeq
2629: \end{eqnarray*}
2630: \[
2631: \simeq {\cal P}_{{\bf 1}_{n-k}}^{\bullet }\left[ {\bf \Lambda }^{\left(
2632: \sigma _1,...,\sigma _k+1\right) }\right] \stackrel{{\cal P}_{{\bf 1}%
2633: _{n-k}}^{\bullet }\left[ j\right] }{\longrightarrow }{\cal P}_{{\bf 1}%
2634: _{n-k}}^{\bullet }\left[ K_{\left( \sigma _1,...,\sigma _k+1\right)
2635: }^{\left( k\right) }\right]
2636: \]
2637: where
2638: \[
2639: 0\rightarrow K_{\left( \sigma _1,...,\sigma _k+1\right) }^{\left( k\right) }%
2640: \stackrel{j}{\longrightarrow }{\bf \Lambda }^{\left( \sigma _1,...,\sigma
2641: _k+1\right) }\stackrel{\pi }{\longrightarrow }{\bf \Lambda }^{\left( \sigma
2642: _1,...,\sigma _k\right) }\rightarrow 0.
2643: \]
2644: Then, $\Phi \circ \Theta $ coincides with the following
2645: \begin{eqnarray*}
2646: {\rm D}_{\left( \sigma _1,...,\sigma _k,1,...,1\right) _n}\simeq {\rm D}%
2647: _{\left( \sigma _1,...,\sigma _k,1\right) _{k+1}}^{\bullet }\left( {\rm D}_{%
2648: {\bf 1}_{n-k-1}}\subset {\rm Diff}_{{\bf 1}_{n-k-1}}^{+}\right) \simeq
2649: \end{eqnarray*}
2650: \begin{eqnarray*}
2651: \ &\simeq &Hom_A^{\bullet }\left( {\bf \Lambda }^{\left( \sigma
2652: _1,...,\sigma _k,1\right) },{\rm D}_{{\bf 1}_{n-k-1}}\subset {\rm Diff}_{%
2653: {\bf 1}_{n-k-1}}^{+}\right) \stackrel{\circ \text{ }d_{\left( \sigma
2654: _1,...,\sigma _k,1\right) }}{\longrightarrow } \\
2655: \ &\longrightarrow &{\rm Diff}_1^{\bullet }\left( {\bf \Lambda }^{\left(
2656: \sigma _1,...,\sigma _k\right) },{\rm D}_{{\bf 1}_{n-k-1}}\subset {\rm Diff}%
2657: _{{\bf 1}_{n-k-1}}^{+}\right) \simeq
2658: \end{eqnarray*}
2659: \[
2660: \simeq {\cal P}_{{\bf 1}_{n-k}}^{\bullet }\left[ {\bf \Lambda }^{\left(
2661: \sigma _1,...,\sigma _k\right) }\right] \stackrel{{\cal P}_{{\bf 1}%
2662: _{n-k}}^{\bullet }\left[ \pi \right] }{\longrightarrow }{\cal P}_{{\bf 1}%
2663: _{n-k}}^{\bullet }\left[ {\bf \Lambda }^{\left( \sigma _1,...,\sigma
2664: _k+1\right) }\right] \stackrel{{\cal P}_{{\bf 1}_{n-k}}^{\bullet }\left[
2665: j\right] }{\longrightarrow }{\cal P}_{{\bf 1}_{n-k}}^{\bullet }\left[
2666: K_{\left( \sigma _1,...,\sigma _k+1\right) }^{\left( k\right) }\right] ;
2667: \]
2668: but $\pi \circ j=0$ hence $im\left( \Phi \right) \subseteq \ker \left(
2669: \Theta \right) $. We prove the reverse inclusion.
2670:
2671: Let $P$ be an object in ${\frak D}$ and $h\in Hom_A^{\bullet }\left( {\bf %
2672: \Lambda }^{\left( \sigma _1,...,\sigma _k+1,1\right) },{\rm D}_{{\bf 1}%
2673: _{n-k-1}}\left( P\right) \subset {\rm Diff}_{{\bf 1}_{n-k-1}}^{+}\left(
2674: P\right) \right) \simeq {\rm D}_{\left( \sigma _1,...,\sigma
2675: _k+1,1,...,1\right) _n}\left( P\right) $ be such that
2676: \[
2677: \Theta \left( h\right) =h\circ d_{\left( \sigma _1,...,\sigma _k+1,1\right)
2678: }\circ j=0;
2679: \]
2680: we claim that $h\in im\left( \Phi \right) $. Now
2681: \[
2682: h\circ d_{\left( \sigma _1,...,\sigma _k+1,1\right) }\circ j=h\circ
2683: j^{\prime }\circ d_{\left( \sigma _1,...,\sigma _k+1,1\right) }\mid
2684: _{K_{\left( \sigma _1,...,\sigma _k+1\right) }^{\left( k\right) }}\text{ }
2685: \]
2686: where
2687: \[
2688: 0\rightarrow K_{\left( \sigma _1,...,\sigma _k+1,1\right) }^{\left( k\right)
2689: }\stackrel{j^{\prime }}{\longrightarrow }{\bf \Lambda }^{\left( \sigma
2690: _1,...,\sigma _k+1,1\right) }\stackrel{\pi ^{\prime }}{\longrightarrow }{\bf %
2691: \Lambda }^{\left( \sigma _1,...,\sigma _k,1\right) }\rightarrow 0.
2692: \]
2693: But $h\in im\left( \Phi \right) $ iff $h\circ j^{\prime }=0$ so it is enough
2694: to show that $im\left( d_{\left( \sigma _1,...,\sigma _k+1,1\right) }\mid
2695: _{K_{\left( \sigma _1,...,\sigma _k+1\right) }^{\left( k\right) }}\right) $
2696: generates $K_{\left( \sigma _1,...,\sigma _k+1,1\right) }^{\left( k\right) }$
2697: over $A$ (since both $h$ and $j^{\prime }$ are $A$-homomorphisms). We know
2698: that $im\left( j_1:Q\rightarrow {\bf J}^1\left( Q\right) \right) $ generates
2699: ${\bf J}^1\left( Q\right) $ over $A$ for any object $Q$ in ${\frak D}$
2700: (Section 2). Moreover, the 3$\times $3 lemma\footnote{%
2701: Smoothness of ${\frak D}$ enters here.} gives us an exact commutative
2702: diagram
2703: \[
2704: \begin{tabular}{llllllllll}
2705: & & $0$ & & $0$ & & $0$ & & & \\
2706: & & $\downarrow $ & & $\downarrow $ & & $\downarrow $ & & & \\
2707: $0$ & $\rightarrow $ & $K_{\left( \sigma _1,...,\sigma _k+2\right) }^{\left(
2708: k\right) }$ & $\rightarrow $ & ${\bf J}^1\left( K_{\left( \sigma
2709: _1,...,\sigma _k+1\right) }^{\left( k\right) }\right) $ & $\stackrel{t}{%
2710: \rightarrow }$ & $K_{\left( \sigma _1,...,\sigma _k+1,1\right) }^{\left(
2711: k\right) }$ & $\rightarrow $ & $0$ & \\
2712: & & $\downarrow $ & & $\downarrow $ & & $\downarrow $ & & & \\
2713: $0$ & $\rightarrow $ & ${\bf \Lambda }^{\left( \sigma _1,...,\sigma
2714: _k+2\right) }$ & $\rightarrow $ & ${\bf J}^1\left( {\bf \Lambda }^{\left(
2715: \sigma _1,...,\sigma _k+1\right) }\right) $ & $\rightarrow $ & ${\bf \Lambda
2716: }^{\left( \sigma _1,...,\sigma _k+1,1\right) }$ & $\rightarrow $ & $0$ & \\
2717: & & $\downarrow $ & & $\downarrow $ & & $\downarrow $ & & & \\
2718: $0$ & $\rightarrow $ & ${\bf \Lambda }^{\left( \sigma _1,...,\sigma
2719: _k+1\right) }$ & $\rightarrow $ & ${\bf J}^1\left( {\bf \Lambda }^{\left(
2720: \sigma _1,...,\sigma _k\right) }\right) $ & $\rightarrow $ & ${\bf \Lambda }%
2721: ^{\left( \sigma _1,...,\sigma _k,1\right) }$ & $\rightarrow $ & $0$ & \\
2722: & & $\downarrow $ & & $\downarrow $ & & $\downarrow $ & & & \\
2723: & & $0$ & & $0$ & & $0$ & & & \\
2724: & & & & & & & & &
2725: \end{tabular}
2726: \]
2727: (where we used the fact that the functor ${\bf J}^k\left( \cdot \right) $ is
2728: exact if ${\frak D}$ is smooth: this follows from lemma \ref{lem 2.5} since $%
2729: {\bf J}^k$ is projective); but $d_{\left( \sigma _1,...,\sigma _k+1,1\right)
2730: }\mid _{K_{\left( \sigma _1,...,\sigma _k+1\right) }^{\left( k\right)
2731: }}=t\circ j_1$ (by definition of $d$), hence $im\left( d_{\left( \sigma
2732: _1,...,\sigma _k+1,1\right) }\mid _{K_{\left( \sigma _1,...,\sigma
2733: _k+1\right) }^{\left( k\right) }}\right) $ generates $K_{\left( \sigma
2734: _1,...,\sigma _k+1,1\right) }^{\left( k\right) }$ over $A$ and we have
2735: finished.
2736:
2737: \smallskip\
2738:
2739: Now let's turn ourselves to the definition of
2740: \[
2741: \psi _l\equiv \left( \psi _l^n:{\bf Hol}^{{\bf 1}_n}\left[ K_{\left( \sigma
2742: _1,...,\sigma _k+l+1\right) }^{\left( k\right) }\right] \rightarrow {\bf Hol}%
2743: ^{{\bf 1}_{n+1}}\left[ K_{\left( \sigma _1,...,\sigma _k+l\right) }^{\left(
2744: k\right) }\right] \right) _{n\geq 0}\text{ .}
2745: \]
2746: First of all
2747: \[
2748: \psi _l^0=0:\left( {\bf Hol}^{{\bf 1}}\left[ K_{\left( \sigma _1,...,\sigma
2749: _k+l+1\right) }^{\left( k\right) }\right] \right) ^0=\left( 0\right)
2750: \rightarrow \left( {\bf Hol}^{{\bf 1}}\left[ K_{\left( \sigma _1,...,\sigma
2751: _k+l\right) }^{\left( k\right) }\right] \right) ^1={\bf J}^1\left( K_{\left(
2752: \sigma _1,...,\sigma _k+l\right) }^{\left( k\right) }\right) .
2753: \]
2754: For $n>0$%
2755: \[
2756: \psi _l^n:{\rm Hol}^{{\bf 1}_n}\left[ K_{\left( \sigma _1,...,\sigma
2757: _k+l+1\right) }^{\left( k\right) }\right] \rightarrow {\rm Hol}^{{\bf 1}%
2758: _{n+1}}\left[ K_{\left( \sigma _1,...,\sigma _k+l\right) }^{\left( k\right)
2759: }\right]
2760: \]
2761: will be defined as the dual representative of a functorial morphism
2762: \[
2763: \psi _n^l:{\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ K_{\left( \sigma
2764: _1,...,\sigma _k+l\right) }^{\left( k\right) }\right] \rightarrow {\cal P}_{%
2765: {\bf 1}_n}^{\bullet }\left[ K_{\left( \sigma _1,...,\sigma _k+l+1\right)
2766: }^{\left( k\right) }\right] .
2767: \]
2768: From the exact sequence
2769: \begin{eqnarray*}
2770: 0 &\rightarrow &K_{\left( \sigma _1,...,\sigma _k+l\right) }^{\left(
2771: k\right) }\stackrel{i}{\longrightarrow }{\bf \Lambda }^{\left( \sigma
2772: _1,...,\sigma _k+l\right) }\stackrel{p}{\longrightarrow }{\bf \Lambda }%
2773: ^{\left( \sigma _1,...,\sigma _k+l-1\right) }\rightarrow 0 \\
2774: \text{(resp. }0 &\rightarrow &K_{\left( \sigma _1,...,\sigma _k+l+1\right)
2775: }^{\left( k\right) }\stackrel{i^{\prime }}{\longrightarrow }{\bf \Lambda }%
2776: ^{\left( \sigma _1,...,\sigma _k+l+1\right) }\stackrel{p^{\prime }}{%
2777: \longrightarrow }{\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right)
2778: }\rightarrow 0\text{ )}
2779: \end{eqnarray*}
2780: and the fact that ${\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l-1\right)
2781: }$ (resp. ${\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }$) is
2782: projective we get (Prop. \ref{split}) an exact sequence of functors ${\frak D%
2783: }\rightarrow {\frak D}$%
2784: \begin{eqnarray*}
2785: 0 &\rightarrow &{\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ {\bf \Lambda }%
2786: ^{\left( \sigma _1,...,\sigma _k+l-1\right) }\right] \stackrel{\epsilon }{%
2787: \rightarrow }{\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ {\bf \Lambda }%
2788: ^{\left( \sigma _1,...,\sigma _k+l\right) }\right] \stackrel{\eta }{%
2789: \rightarrow }{\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ K_{\left( \sigma
2790: _1,...,\sigma _k+l\right) }^{\left( k\right) }\right] \rightarrow 0 \\
2791: \text{(resp. }0 &\rightarrow &{\cal P}_{{\bf 1}_n}^{\bullet }\left[ {\bf %
2792: \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right] \stackrel{%
2793: \epsilon ^{\prime }}{\rightarrow }{\cal P}_{{\bf 1}_n}^{\bullet }\left[ {\bf %
2794: \Lambda }^{\left( \sigma _1,...,\sigma _k+l+1\right) }\right] \stackrel{\eta
2795: ^{\prime }}{\rightarrow }{\cal P}_{{\bf 1}_n}^{\bullet }\left[ K_{\left(
2796: \sigma _1,...,\sigma _k+l+1\right) }^{\left( k\right) }\right] \rightarrow 0%
2797: \text{ )}
2798: \end{eqnarray*}
2799: with $\epsilon ={\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ p\right] $ and $%
2800: \eta ={\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ i\right] $ (resp. $\epsilon
2801: ^{\prime }={\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ p^{\prime }\right] $
2802: and $\eta ^{\prime }={\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ i^{\prime
2803: }\right] $). To define $\psi _n^l$ it will be then enough to define
2804: \[
2805: \overline{\psi }_n^l:{\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ {\bf \Lambda }%
2806: ^{\left( \sigma _1,...,\sigma _k+l\right) }\right] \rightarrow {\cal P}_{%
2807: {\bf 1}_n}^{\bullet }\left[ K_{\left( \sigma _1,...,\sigma _k+l+1\right)
2808: }^{\left( k\right) }\right]
2809: \]
2810: and show that $\overline{\psi }_n^l\circ \epsilon =0$. We know (Section 2 )
2811: that there is an exact sequence
2812: \[
2813: 0\rightarrow {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l+1\right) }%
2814: \stackrel{s}{\longrightarrow }{\bf J}^1\left( {\bf \Lambda }^{\left( \sigma
2815: _1,...,\sigma _k+l\right) }\right) \stackrel{q}{\longrightarrow }{\bf %
2816: \Lambda }^{\left( \sigma _1,...,\sigma _k+l,1\right) }\rightarrow 0
2817: \]
2818: and this (Prop. \ref{split}) gives us the exact sequence
2819: \[
2820: 0\rightarrow {\cal P}_{{\bf 1}_n}^{\bullet }\left[ {\bf \Lambda }^{\left(
2821: \sigma _1,...,\sigma _k+l,1\right) }\right] \stackrel{\alpha }{\rightarrow }%
2822: {\cal P}_{{\bf 1}_n}^{\bullet }\left[ {\bf J}^1\left( {\bf \Lambda }^{\left(
2823: \sigma _1,...,\sigma _k+l\right) }\right) \right] \stackrel{\beta }{%
2824: \rightarrow }{\cal P}_{{\bf 1}_n}^{\bullet }\left[ {\bf \Lambda }^{\left(
2825: \sigma _1,...,\sigma _k+l+1\right) }\right] \rightarrow 0
2826: \]
2827: (with $\alpha ={\cal P}_{{\bf 1}_n}^{\bullet }\left[ q\right] $ and $\beta =%
2828: {\cal P}_{{\bf 1}_n}^{\bullet }\left[ s\right] $). Then we take $\overline{%
2829: \psi }_n^l$ to be the composition
2830: \begin{eqnarray*}
2831: {\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ {\bf \Lambda }^{\left( \sigma
2832: _1,...,\sigma _k+l\right) }\right] \simeq {\cal P}_{\left( 1,1\right)
2833: }^{\bullet }\left[ {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right)
2834: }\right] \left( {\rm D}_{{\bf 1}_{n-1}}\subset {\rm Diff}_{{\bf 1}%
2835: _{n-1}}^{+}\right) \simeq
2836: \end{eqnarray*}
2837: \[
2838: \simeq Hom_A^{\bullet }\left( {\rm Hol}^{(1,1)}\left[ {\bf \Lambda }^{\left(
2839: \sigma _1,...,\sigma _k+l\right) }\right] ,{\rm D}_{{\bf 1}_{n-1}}\subset
2840: {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right) \stackrel{\circ \text{ }\delta
2841: _{\left( 1,1\right) }\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma
2842: _k+l\right) }\right) }{\longrightarrow }
2843: \]
2844: \begin{eqnarray*}
2845: \ &\rightarrow &{\rm Diff}_1^{\bullet }\left( {\rm Hol}^{(1)}\left[ {\bf %
2846: \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right] ,{\rm D}_{{\bf 1}%
2847: _{n-1}}\subset {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right) \simeq \\
2848: \ &\simeq &{\cal P}_{\left( 1\right) }^{\bullet }\left[ {\bf J}^1\left( {\bf %
2849: \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right) \right] \left(
2850: {\rm D}_{{\bf 1}_{n-1}}\subset {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right) \simeq
2851: \end{eqnarray*}
2852: \[
2853: \simeq {\cal P}_{{\bf 1}_n}^{\bullet }\left[ {\bf J}^1\left( {\bf \Lambda }%
2854: ^{\left( \sigma _1,...,\sigma _k+l\right) }\right) \right] \stackrel{\beta }{%
2855: \longrightarrow }{\cal P}_{{\bf 1}_n}^{\bullet }\left[ {\bf \Lambda }%
2856: ^{\left( \sigma _1,...,\sigma _k+l+1\right) }\right] \stackrel{\eta ^{\prime
2857: }}{\longrightarrow }{\cal P}_{{\bf 1}_n}^{\bullet }\left[ K_{\left( \sigma
2858: _1,...,\sigma _k+l+1\right) }^{\left( k\right) }\right] .
2859: \]
2860: Now we show that $\overline{\psi }_n^l\circ \epsilon =0$.
2861:
2862: Note that using the identifications
2863: \begin{eqnarray*}
2864: {\cal P}_{{\bf 1}_n}^{\bullet }\left[ {\bf J}^1\left( {\bf \Lambda }^{\left(
2865: \sigma _1,...,\sigma _k+l\right) }\right) \right] &\simeq &{\rm Diff}%
2866: _1^{\bullet }\left( {\bf J}^1\left( {\bf \Lambda }^{\left( \sigma
2867: _1,...,\sigma _k+l\right) }\right) ,{\rm D}_{{\bf 1}_{n-1}}\subset {\rm Diff}%
2868: _{{\bf 1}_{n-1}}^{+}\right) \simeq \\
2869: \ &\simeq &Hom_A^{\bullet }\left( {\bf J}^1\left( {\bf J}^1\left( {\bf %
2870: \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right) \right) ,{\rm D}%
2871: _{{\bf 1}_{n-1}}\subset {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right)
2872: \end{eqnarray*}
2873: and
2874: \[
2875: Hom_A^{\bullet }\left( {\bf J}^1\left( {\bf \Lambda }^{\left( \sigma
2876: _1,...,\sigma _k+l+1\right) }\right) ,{\rm D}_{{\bf 1}_{n-1}}\subset {\rm %
2877: Diff}_{{\bf 1}_{n-1}}^{+}\right) \simeq {\cal P}_{{\bf 1}_n}^{\bullet
2878: }\left[ {\bf J}^1\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma
2879: _k+l+1\right) }\right) \right]
2880: \]
2881: (resp. the identifications
2882: \[
2883: {\cal P}_{{\bf 1}_n}^{\bullet }\left[ {\bf \Lambda }^{\left( \sigma
2884: _1,...,\sigma _k+l+1\right) }\right] \simeq Hom_A^{\bullet }\left( {\bf J}%
2885: ^1\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l+1\right) }\right) ,%
2886: {\rm D}_{{\bf 1}_{n-1}}\subset {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right)
2887: \]
2888: and
2889: \[
2890: Hom_A^{\bullet }\left( {\bf J}^1\left( K_{\left( \sigma _1,...,\sigma
2891: _k+l+1\right) }^{\left( k\right) }\right) ,{\rm D}_{{\bf 1}_{n-1}}\subset
2892: {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right) \simeq {\cal P}_{{\bf 1}_n}^{\bullet
2893: }\left[ K_{\left( \sigma _1,...,\sigma _k+l+1\right) }^{\left( k\right)
2894: }\right] \text{ )}
2895: \]
2896: $\beta $ (resp. $\eta ^{\prime }$ ) is given by
2897: \begin{eqnarray*}
2898: &&\ \ \ \ Hom_A^{\bullet }\left( {\bf J}^1\left( {\bf J}^1\left( {\bf %
2899: \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right) \right) ,{\rm D}%
2900: _{{\bf 1}_{n-1}}\subset {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right) \stackrel{%
2901: \circ \text{ }{\bf J}^1\left( s\right) }{\longrightarrow } \\
2902: \ &\longrightarrow &Hom_A^{\bullet }\left( {\bf J}^1\left( {\bf \Lambda }%
2903: ^{\left( \sigma _1,...,\sigma _k+l+1\right) }\right) ,{\rm D}_{{\bf 1}%
2904: _{n-1}}\subset {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right)
2905: \end{eqnarray*}
2906: (resp. by
2907: \begin{eqnarray*}
2908: &&\ \ \ \ \ \ Hom_A^{\bullet }\left( {\bf J}^1\left( {\bf \Lambda }^{\left(
2909: \sigma _1,...,\sigma _k+l+1\right) }\right) ,{\rm D}_{{\bf 1}_{n-1}}\subset
2910: {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right) \stackrel{\circ \text{ }{\bf J}%
2911: ^1\left( i^{\prime }\right) }{\longrightarrow } \\
2912: \ &\rightarrow &Hom_A^{\bullet }\left( {\bf J}^1\left( K_{\left( \sigma
2913: _1,...,\sigma _k+l+1\right) }^{\left( k\right) }\right) ,{\rm D}_{{\bf 1}%
2914: _{n-1}}\subset {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right) \text{ ).}
2915: \end{eqnarray*}
2916: With a similar analysis we see that $\epsilon $, viewed as a morphism
2917: \begin{eqnarray*}
2918: &&\ \ \ \ \ \ Hom_A^{\bullet }\left( {\rm Hol}^{\left( 1,1\right) }\left(
2919: {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l-1\right) }\right) ,{\rm D}_{%
2920: {\bf 1}_{n-1}}\subset {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right) \stackrel{%
2921: \epsilon }{\rightarrow } \\
2922: \ &\rightarrow &Hom_A^{\bullet }\left( {\rm Hol}^{\left( 1,1\right) }\left(
2923: {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right) ,{\rm D}_{%
2924: {\bf 1}_n}\subset {\rm Diff}_{{\bf 1}_n}^{+}\right)
2925: \end{eqnarray*}
2926: is given by taking the composition with $\left[ {\bf J}^1\left( {\bf J}%
2927: ^1\left( p\right) \right) \right] $ where
2928: \begin{eqnarray*}
2929: \left[ {\bf J}^1\left( {\bf J}^1\left( p\right) \right) \right] &:&\frac{%
2930: {\bf J}^1\left( {\bf J}^1\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma
2931: _k+l-1\right) }\right) \right) }{{\bf J}^2\left( {\bf \Lambda }^{\left(
2932: \sigma _1,...,\sigma _k+l-1\right) }\right) }\simeq {\rm Hol}^{\left(
2933: 1,1\right) }\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l-1\right)
2934: }\right) \rightarrow \\
2935: \ &\rightarrow &{\rm Hol}^{\left( 1,1\right) }\left( {\bf \Lambda }^{\left(
2936: \sigma _1,...,\sigma _k+l\right) }\right) \simeq \frac{{\bf J}^1\left( {\bf J%
2937: }^1\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right)
2938: \right) }{{\bf J}^2\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma
2939: _k+l\right) }\right) }
2940: \end{eqnarray*}
2941: is the quotient map of ${\bf J}^1\left( {\bf J}^1\left( p\right) \right) :%
2942: {\bf J}^1\left( {\bf J}^1\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma
2943: _k+l-1\right) }\right) \right) \rightarrow {\bf J}^1\left( {\bf J}^1\left(
2944: {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right) \right) $.
2945:
2946: Therefore $\overline{\psi }_n^l\circ \epsilon $, viewed as a morphism
2947: \[
2948: Hom_A^{\bullet }\left( {\rm Hol}^{\left( 1,1\right) }\left( {\bf \Lambda }%
2949: ^{\left( \sigma _1,...,\sigma _k+l-1\right) }\right) ,{\rm D}_{{\bf 1}%
2950: _{n-1}}\subset {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right) \longrightarrow
2951: \]
2952: \[
2953: \longrightarrow \ \ \ Hom_A^{\bullet }\left( {\bf J}^1\left( K_{\left(
2954: \sigma _1,...,\sigma _k+l+1\right) }^{\left( k\right) }\right) ,{\rm D}_{%
2955: {\bf 1}_{n-1}}\subset {\rm Diff}_{{\bf 1}_{n-1}}^{+}\right)
2956: \]
2957: is given by
2958: \[
2959: f\longmapsto f\circ \left[ {\bf J}^1\left( {\bf J}^1\left( p\right) \right)
2960: \right] \circ \xi _{\left( \sigma _1,...,\sigma _k+l\right) }\circ {\bf J}%
2961: ^1\left( s\right) \circ {\bf J}^1\left( i^{\prime }\right)
2962: \]
2963: where
2964: \[
2965: \xi _{\left( \sigma _1,...,\sigma _k+l\right) }:{\bf J}^1\left( {\bf J}%
2966: ^1\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right)
2967: \right) \rightarrow \frac{{\bf J}^1\left( {\bf J}^1\left( {\bf \Lambda }%
2968: ^{\left( \sigma _1,...,\sigma _k+l\right) }\right) \right) }{{\bf J}^2\left(
2969: {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right) }
2970: \]
2971: is the natural projection. (Recall from Section 4 that $\delta _{\left(
2972: 1,1\right) }\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right)
2973: }\right) $ is given by the composition
2974: \[
2975: {\bf J}^1\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right)
2976: }\right) \stackrel{j_1\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma
2977: _k+l\right) }\right) }{\longrightarrow }{\bf J}^1\left( {\bf J}^1\left( {\bf %
2978: \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right) \right)
2979: \stackrel{\xi _{\left( \sigma _1,...,\sigma _k+l\right) }}{\longrightarrow }%
2980: \frac{{\bf J}^1\left( {\bf J}^1\left( {\bf \Lambda }^{\left( \sigma
2981: _1,...,\sigma _k+l\right) }\right) \right) }{{\bf J}^2\left( {\bf \Lambda }%
2982: ^{\left( \sigma _1,...,\sigma _k+l\right) }\right) }\text{ ).}
2983: \]
2984: But $\left[ {\bf J}^1\left( {\bf J}^1\left( p\right) \right) \right] \circ
2985: \xi $ coincides with
2986: \[
2987: {\bf J}^1\left( {\bf J}^1\left( {\bf \Lambda }^{\left( \sigma _1,...,\sigma
2988: _k+l\right) }\right) \right) \stackrel{{\bf J}^1\left( {\bf J}^1\left(
2989: p\right) \right) }{\longrightarrow }{\bf J}^1\left( {\bf J}^1\left( {\bf %
2990: \Lambda }^{\left( \sigma _1,...,\sigma _k+l-1\right) }\right) \right)
2991: \stackrel{\xi _{\left( \sigma _1,...,\sigma _k+l-1\right) }}{\longrightarrow
2992: }\frac{{\bf J}^1\left( {\bf J}^1\left( {\bf \Lambda }^{\left( \sigma
2993: _1,...,\sigma _k+l\right) }\right) \right) }{{\bf J}^2\left( {\bf \Lambda }%
2994: ^{\left( \sigma _1,...,\sigma _k+l\right) }\right) }
2995: \]
2996: ($\xi _{\left( \sigma _1,...,\sigma _k+l-1\right) }$ being again the natural
2997: projection), so that
2998: \[
2999: \begin{tabular}{lll}
3000: $\overline{\psi }_n^l\circ \epsilon :f$ & $\longmapsto $ & $f\circ \xi
3001: _{\left( \sigma _1,...,\sigma _k+l\right) }\circ {\bf J}^1\left( {\bf J}%
3002: ^1\left( p\right) \right) \circ {\bf J}^1\left( s\right) \circ {\bf J}%
3003: ^1\left( i^{\prime }\right) =$ \\
3004: & & $=f\circ \xi _{\left( \sigma _1,...,\sigma _k+l\right) }\circ {\bf J}%
3005: ^1\left( {\bf J}^1\left( p\right) \circ s\circ i^{\prime }\right) .$ \\
3006: & &
3007: \end{tabular}
3008: \]
3009: Again as above, the 3$\times $3 lemma gives us an exact commutative diagram:
3010: \[
3011: \begin{tabular}{ccccccccc}
3012: & & $0$ & & $0$ & & $0$ & & \\
3013: & & $\downarrow $ & & $\downarrow $ & & $\downarrow $ & & \\
3014: $0$ & $\rightarrow $ & $K_{\left( \sigma _1,...,\sigma _k+l+1\right)
3015: }^{\left( k\right) }$ & $\rightarrow $ & ${\bf J}^1\left( K_{\left( \sigma
3016: _1,...,\sigma _k+l\right) }^{\left( k\right) }\right) $ & $\stackrel{}{%
3017: \rightarrow }$ & $K_{\left( \sigma _1,...,\sigma _k+l,1\right) }^{\left(
3018: k\right) }$ & $\rightarrow $ & $0$ \\
3019: & & $i^{\prime }\downarrow \quad $ & & $\downarrow $ & & $\downarrow $ &
3020: & \\
3021: $0$ & $\rightarrow $ & ${\bf \Lambda }^{\left( \sigma _1,...,\sigma
3022: _k+l+1\right) }$ & $\stackrel{s}{\rightarrow }$ & ${\bf J}^1\left( {\bf %
3023: \Lambda }^{\left( \sigma _1,...,\sigma _k+l\right) }\right) $ & $\rightarrow
3024: $ & ${\bf \Lambda }^{\left( \sigma _1,...,\sigma _k+l,1\right) }$ & $%
3025: \rightarrow $ & $0$ \\
3026: & & $\downarrow $ & & $_{{\bf J}^1\left( p\right) }\downarrow $ $\qquad $
3027: & & $\downarrow $ & & \\
3028: $0$ & $\rightarrow $ & ${\bf \Lambda }^{\left( \sigma _1,...,\sigma
3029: _k+l\right) }$ & $\rightarrow $ & ${\bf J}^1\left( {\bf \Lambda }^{\left(
3030: \sigma _1,...,\sigma _k+l-1\right) }\right) $ & $\rightarrow $ & ${\bf %
3031: \Lambda }^{\left( \sigma _1,...,\sigma _k+l-1,1\right) }$ & $\rightarrow $ &
3032: $0$ \\
3033: & & $\downarrow $ & & $\downarrow $ & & $\downarrow $ & & \\
3034: & & $0$ & & $0$ & & $0$ & &
3035: \end{tabular}
3036: \]
3037: which finally shows that ${\bf J}^1\left( p\right) \circ s\circ i^{\prime }$
3038: $=0$ and hence $\overline{\psi }_n^l\circ \epsilon =0$.
3039:
3040: Therefore
3041: \[
3042: \psi _n^l:{\cal P}_{{\bf 1}_{n+1}}^{\bullet }\left[ K_{\left( \sigma
3043: _1,...,\sigma _k+l\right) }^{\left( k\right) }\right] \rightarrow {\cal P}_{%
3044: {\bf 1}_n}^{\bullet }\left[ K_{\left( \sigma _1,...,\sigma _k+l+1\right)
3045: }^{\left( k\right) }\right]
3046: \]
3047: is well defined as well as its dual representative
3048: \[
3049: \psi _l^n:{\rm Hol}^{{\bf 1}_n}\left[ K_{\left( \sigma _1,...,\sigma
3050: _k+l+1\right) }^{\left( k\right) }\right] \rightarrow {\rm Hol}^{{\bf 1}%
3051: _{n+1}}\left[ K_{\left( \sigma _1,...,\sigma _k+l\right) }^{\left( k\right)
3052: }\right]
3053: \]
3054: as we wanted.
3055:
3056: Just as in the case of $\rho $, an easy application of the 3$\times $3 lemma
3057: proves that $im\left( \psi _{n+1}^{l-1}\right) =\ker \left( \psi _n^l\right)
3058: $.
3059:
3060: It is easy to verify that $\rho $ and $\psi _l$ so defined are maps of
3061: complexes; therefore (\ref{res}) is a resolution of ${\bf K}_{\left( \sigma
3062: _1,...,\sigma _k+1,{\bf 1}\right) }^{(k)}$ as desired.
3063:
3064: \begin{center}
3065: \smallskip
3066: \
3067: \end{center}
3068:
3069: \begin{thebibliography}{Bou III}
3070: \bibitem[Bou III]{Bou III} N. Bourbaki, {\em El\'ements de Math\'ematique:
3071: Alg\`ebre Ch. III}, Hermann, Paris, 1971, pp. 133-138.
3072:
3073: \bibitem[Bou X]{Bou X} N. Bourbaki, {\em El\'ements de Math\'ematique:
3074: Alg\`ebre Ch. X (Alg\`ebre Homologique)}, Masson, Paris, 1980, pp. 43-46.
3075:
3076: \bibitem[Ha]{Ha} R. Hartshorne, {\em Algebraic Geometry}, Springer Verlag,
3077: Berlin, 1977.
3078:
3079: \bibitem[Kr]{Kr} I. S. Krasil'shchik, {\em Calculus over commutative
3080: algebras: a concise user guide}, Acta Appl. Math. 49 (3), 1997, 235-248.
3081:
3082: \bibitem[KLV]{KLV} I. S. Krasil'schik, V. V. Lychagin, A. M. Vinogradov,
3083: {\em Geometry of jet spaces and nonlinear partial differential equations},
3084: Gordon and Breach, 1986.
3085:
3086: \bibitem[Lu]{Lu} S. Lubkin, {\em Cohomology of completions}, North-Holland,
3087: 1980.
3088:
3089: \bibitem[Ver]{Ver} A. Verbovetsky, {\em On the cohomology of the
3090: compatibility complex}, Uspekhi Mat. Nauk, 53 (1998), no. 1, 213-214
3091: (Russian); English translation in Russian Math. Surveys 53 (1998), no. 1,
3092: 225-226.
3093:
3094: \bibitem[VV]{VV} G. Vezzosi, A. M. Vinogradov, {\em Infinitesimal Stokes'
3095: formula for higher-order de Rham complexes}, Acta Appl. Math. 49 (3), 1997,
3096: 311-329.
3097:
3098: \bibitem[VV-esi]{VV-esi} G. Vezzosi. A. M. Vinogradov, {\em On Higher
3099: Analogs of the de Rham Complex}, Preprint ESI 202, Wien, 1995.
3100:
3101: \bibitem[VV-sns]{VV-sns} G. Vezzosi, A. M. Vinogradov, {\em On higher order
3102: de Rham complexes}, Preprints di Matematica n.19, Scuola Normale Superiore,
3103: Pisa, June 1998.
3104:
3105: \bibitem[Vi1]{Vi1} A. M. Vinogradov, {\em The logic algebra for the theory
3106: of linear differential operators}, Sov. Math. Dokl. {\bf 13} (1972),
3107: 1058-1062.
3108:
3109: \bibitem[Vi2]{Vi2} A. M. Vinogradov, {\em Some homological systems
3110: associated with the differential calculus in commutative algebras}, Russian
3111: Math. Surveys {\bf 34}:6 (1979), 250-255.
3112:
3113: \bibitem[Vi3]{Vi3} A. M. Vinogradov, {\em Introduction to Secondary Calculus%
3114: }, Cont. Math., 1999, vol. 219, pp. 241-272.
3115:
3116: \bibitem[MVi]{MVi} M. M. Vinogradov, {\em On Spencer and de Rham algebraic
3117: cohomology}, Sov. Math. Dokl. 19 (1978).
3118:
3119: \bibitem[We]{We} C. A. Weibel, {\em An introduction to homological algebra}%
3120: , Cambridge University Press, 1995.
3121: \end{thebibliography}
3122:
3123: \end{document}
3124:
3125: %%%%%%%%%%%%%%%%%%%%%%%%%%%% End dr.tex %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3126:
3127: