1: \section{The Hochschild-Kostant-Rosenberg theorem and the Chern character}
2: \label{sec:hkr}
3:
4: In this section we study the relationship between the Hochschild and
5: harmonic structures. We provide a discussion of the connection
6: between the usual Chern character and the one introduced
7: in~\cite{CalHH1}.
8:
9: \subsection{}
10: The starting point of our analysis is the following theorem:
11: \begin{theorem}[Hochschild-Kostant-Rosenberg~\cite{HKR},
12: Kontsevich~\cite{Kon}, Swan~\cite{Swa}, Yekutieli~\cite{Yek}]
13: \label{thm:hkr}
14: Let $X$ be a smooth, quasi-projec\-tive variety of dimension $n$,
15: and let $\Delta:X\ra X\times X$ be the diagonal embedding. Then there
16: exists a quasi-isomorphism
17: \[ I:\Delta^*\cO_\Delta \stackrel{\sim}{\lra}
18: \bigoplus_i \Omega_X^i[i], \]
19: where the right hand side denotes the complex whose $-i$-th term is
20: $\Omega_X^i$, and all differentials are zero.
21: \end{theorem}
22:
23: \begin{proof}
24: (This is nothing but a brief recounting of the results in~\cite{Yek},
25: and the reader should consult [loc.cit.] for more details.) Recall
26: that if $R$ is a commutative $\C$-algebra there exists a standard
27: resolution of $R$ as an $R^e = R\otimes_\C R$-module. For $i\geq 0$
28: let
29: \[ \cB_i(R) = R^{\otimes (i+2)}, \]
30: where the tensor product is taken over $\C$. It is an $R^e$-module by
31: multiplication in the first and last factor. The bar resolution is
32: defined to be the complex of $R^e$-modules
33: \[ \cdots\ra \cB_i(R) \ra \cdots \ra \cB_1(R) \ra \cB_0(R) \ra 0, \]
34: with differential
35: \begin{eqnarray*}
36: \lefteqn{d(a_0\otimes a_1\otimes\cdots\otimes a_i) = } \\
37: & & a_0a_1 \otimes a_2 \otimes\cdots \otimes a_i \,-\,
38: a_0\otimes a_1a_2 \otimes \cdots \otimes a_i \,+\, \cdots \,+\, \\
39: & & (-1)^{i-1}a_0\otimes a_1\otimes\cdots \otimes
40: a_{i-1}a_i.
41: \end{eqnarray*}
42: It is an exact complex, except at the last step where the cohomology
43: is $R$. Thus it is a resolution of $R$ in
44: $R^e$-$\gMod$~\cite[1.1.12]{Lod}.
45:
46: If $X$ were affine, $X=\Spec R$, we could use the above resolution to
47: compute $\Delta^* \cO_\Delta$: indeed, $\cO_\Delta$ is nothing but $R$
48: viewed as an $R^e = \cO_{X\times X}$-module, and the modules
49: $\cB_i$ are $R^e$-flat. The complex obtained by tensoring the
50: bar resolution over $R^e$ with $R$ is called the bar complex:
51: \[ \cdots\ra \cC_i(R) \ra \cdots \ra \cC_1(R) \ra \cC_0(R) \ra 0, \]
52: where
53: \[ \cC_i(R) = \cB_i(R) \otimes_{R^e} R, \]
54: and the differential is obtained from the differential of $\cB_\cdot(R)$.
55:
56: Problems arise when one tries to sheafify the bar resolution to
57: obtain a complex of sheaves on a scheme: the resulting sheaves are
58: ill-behaved (in particular, not quasi-coherent).
59: As a replacement, Yekutieli proposed to used the complete bar
60: resolution, which he defined in~\cite{Yek}. For $i\geq 0$, let
61: $\sX^i$ be the formal completion of the scheme $X^i = X\times
62: \cdots\times X$ along the small diagonal. Define
63: \[ \hcB_i(X) = \cO_{\sX^{i+2}}, \]
64: which is a sheaf of abelian groups on the topological space $X$.
65: Yekutieli argued that one can formally complete and sheafify the
66: original bar resolution to get the complete bar resolution
67: \[ \cdots\ra \hcB_i(X) \ra \cdots \ra \hcB_1(X) \ra \hcB_0(X) \ra 0, \]
68: where the maps are locally obtained from the maps of the original bar
69: complex, by noting that these are continuous for the topologies with
70: respect to which we are completing. The complete bar resolution is an
71: exact resolution of $\cO_\Delta$ by sheaves of flat $\cO_{X\times
72: X}$-modules (see remark following Proposition 1.4 and proof of
73: Proposition 1.5 in~\cite{Yek}). Over an affine open set $U = \Spec R$
74: of $X$, $\Gamma(U, \hcB_i(X))$ is the completion $\hcB_i(R)$ of
75: $\cB_i(R)$ at the ideal $I_i$ which is the kernel of the
76: multiplication map $\cB_i(R) = R^{\otimes i} \ra R$.
77:
78: One can take the complete bar resolution as a flat resolution of
79: $\cO_\Delta$ on $X\times X$, and use it to compute
80: $\Delta^*\cO_\Delta$. This is the same as tensoring the complete bar
81: resolution over $\cO_{X\times X}$ with $\cO_\Delta$. The resulting
82: complex is called the complex of complete Hochschild chains of $X$
83: (see~\cite[Definition 1.3]{Yek} for details),
84: \[ \cdots\ra \hcC_i(X) \ra \cdots \ra \hcC_1(X) \ra \hcC_0(X) \ra 0, \]
85: where
86: \[ \hcC_i(X) = \hcB_i(X) \otimes_{\cO_{X\times X}} \cO_\Delta. \]
87: Over an affine open set $U=\Spec R$, $\Gamma(U, \hcC_i(X))$ is the
88: completion $\hcC_i(R)$ of $\cC_i(R)$ at $I_i$ (as a $\cB_i(R)$-module).
89:
90: Over any affine open $U= \Spec R$ define
91: \[ I_i:\cC_i(R) \ra \Omega_{R/k}^i \]
92: by setting
93: \[ I_i((1\otimes a_1\otimes\cdots\otimes a_i \otimes 1) \otimes_{R^e}
94: 1) = d a_1\wedge da_2 \wedge\cdots \wedge da_i. \]
95: These maps are continuous with respect to the topology that is used
96: for completing~(\cite[Lemma 4.1]{Yek}), so they can be completed and
97: sheafified to maps
98: \[ I_i:\hcC_i(X) \ra \Omega_X^i. \]
99: They also commute with the zero differentials of the complex $\oplus_i
100: \Omega_X^i$, so they assemble to a morphism of complexes
101: \[ I:\Delta^*\cO_\Delta \ra \bigoplus_i \Omega_X^i[i] \]
102: which can be seen to be a quasi-isomorphism in characteristic
103: 0~(\cite[Theorem 4.6.1.1]{Kon},~\cite[Proposition 4.4]{Yek}). In the
104: affine case this is essentially the Hochschild-Kostant-Rosenberg
105: theorem~\cite{HKR}.
106: \end{proof}
107:
108: \begin{corollary}
109: \label{cor:ihkr}
110: The Hochschild-Kostant-Rosenberg isomorphism $I$ induces isomorphisms
111: of graded vector spaces
112: \begin{align*}
113: I^\HKR & : HH^*(X) \stackrel{\sim}{\lra} HT^*(X), \\
114: I_\HKR & : HH_*(X) \stackrel{\sim}{\lra} H\Omega_*(X).
115: \end{align*}
116: \end{corollary}
117:
118: \begin{proof}
119: \begin{align*}
120: HH^k(X) & = \Hom_{X\times X}(\cO_\Delta, \cO_\Delta[k]) \iso
121: \Hom_X(\Delta^*\cO_\Delta, \cO_X[k]) \\
122: & \iso \Hom_X(\bigoplus_i \Omega_X^i[i], \cO_X[k]) = \bigoplus_i
123: H^{k-i}(X, \bigwedge^i T_X) = HT^k(X),
124: \intertext{and}
125: HH_k(X) & = \Hom_{X\times X}(\Delta_!\cO_X[k],\cO_\Delta) \iso
126: \Hom_X(\cO_X[k], \Delta^*\cO_\Delta) \\
127: & \iso \Hom_X(\cO_X[k], \bigoplus_i \Omega_X^i[i]) = \bigoplus_i
128: H^{i-k}(X, \Omega_X^i) = H\Omega_k(X).
129: \end{align*}
130: \end{proof}
131:
132: \subsection{}
133: We are now interested in understanding how the above isomorphisms
134: relate the Chern character $K_0(X) \ra HH_0(X)$ defined in the
135: introduction to the usual Chern character.
136:
137: Let $\Omega_\Delta^{\otimes i}$ and $\Omega_\Delta^i$ denote the
138: push-forwards by $\Delta$ of $\Omega_X^{\otimes i}$ and $\Omega_X^i$,
139: respectively. (Here the tensor product is taken over $\cO_X$.) Let
140: \[ \epsilon:\Omega_X^{\otimes i} \ra \Omega_X^i \]
141: be the antisymmetrization map which acts as
142: \[ v_1\otimes v_2\otimes\cdots v_i \mapsto \frac{1}{i!}
143: \sum_{\sigma\in\Sigma_i} (-1)^{\epsilon(\sigma)} v_{\sigma_1}\otimes
144: v_{\sigma_2}\otimes\cdots v_{\sigma_i}. \]
145: By an abuse of notation, we shall also denote by $\epsilon$ the
146: push-forward
147: \[ \Delta_*\epsilon:\Omega_\Delta^{\otimes i} \ra \Omega_\Delta^i. \]
148:
149: \begin{definition}
150: Define the universal Atiyah class to be the class
151: \[ \alpha_1\in\Ext^1_{X\times X}(\cO_\Delta, \Omega_\Delta^1), \]
152: of the extension
153: \[ 0\ra \Omega_\Delta^1 \ra \cO_{\Delta^{(2)}} \ra \cO_\Delta \ra 0,
154: \]
155: where $\cO_{\Delta^{(2)}}$ is the second infinitesimal neighborhood of
156: the diagonal in $X\times X$. Furthermore, define $\alpha_i$ for
157: $i\geq 0$ by the formula
158: \[ \alpha_i = \epsilon \circ (\pi_2^*\Omega_X^{\otimes (i-1)}
159: \otimes \alpha_1) \circ (\pi_2^*\Omega_X^{\otimes (i-2)} \otimes
160: \alpha_1) \circ \cdots\circ \alpha_1:\cO_\Delta\ra \Omega_\Delta^i[i]. \]
161: The exponential Atiyah class $\exp(\alpha)$ is defined by the formula
162: \[ \exp(\alpha) =
163: 1+\alpha_1+\alpha_2+\cdots+\alpha_n:\cO_\Delta\ra\bigoplus_i
164: \Delta_* \Omega_X^i[i]. \]
165: \end{definition}
166:
167: This definition requires a short explanation. Recall that given an
168: object $\cE\in\D(X)$, the Atiyah class of $\cE$ is the class
169: \[ \alpha_1(\cE)\in\Ext^1_X(\cE, \cE\otimes \Omega_X^1) \]
170: of the extension on $X$
171: \[ 0\ra \cE\otimes \Omega_X^1\ra J^1(\cE)\ra \cE \ra 0 \]
172: where $J^1(\cE)$ is the first jet bundle of $\cE$~\cite[1.1]{KapAt}.
173: A natural way to construct this extension is to consider the natural
174: transformation $\FMXX^{\alpha_1}$ associated to the universal Atiyah
175: class
176: \[ \alpha_1:\cO_\Delta\ra \Omega_\Delta^1[1] \]
177: between the identity functor and the ``tensor by $\Omega_X^1[1]$''
178: functor. The value $\FMXX^{\alpha_1}(\cE)$ of this natural
179: transformation on $\cE$ is precisely the Atiyah class $\alpha_1(\cE)$
180: of $\cE$ (see, for example,~\cite[10.1.5]{BlueBook}). The $i$-th
181: component of the Chern character of $\cE$ is then obtained as
182: \[ \ch_i(\cE) = \Tr_\cE(\alpha_i(\cE)) \]
183: where
184: \[ \alpha_i(\cE) = \epsilon\circ
185: (\Omega_X^{\otimes(i-1)}\otimes\alpha_1(\cE)) \circ
186: (\Omega_X^{\otimes(i-2)}\otimes\alpha_1(\cE)) \circ \cdots \circ
187: \alpha_1(\cE):\cE\ra \cE\otimes \Omega_X^i[i]. \]
188: (See~\cite[10.1.6]{BlueBook} for details.) Our definition of
189: $\alpha_i:\cO_\Delta\ra \Delta_*\Omega_X^i[i]$ has been tailored to
190: mimic this definition: $\alpha_i(\cE)$ will be precisely the value on
191: $\cE$ of the natural transformation associated to the morphism
192: $\alpha_i$. Therefore, if we consider the natural transformation
193: $\FMXX^{\exp(\alpha)}$ associated to $\exp(\alpha)$, its value
194: \[ \Phi^{\exp(\alpha)}(\cE):\cE\ra\bigoplus_i \cE\otimes
195: \Omega_X^i[i] \]
196: on $\cE$ will satisfy
197: \[ \ch_{\mathrm{orig}}(\cE) = \Tr_\cE(\Phi^{\exp(\alpha)}(\cE)), \]
198: where $\ch_{\mathrm{orig}}(\cE)$ is the usual Chern character of
199: $\cE$.
200:
201: \begin{proposition}
202: \label{prop:etaatyiah}
203: The exponential $\exp(\alpha)$ of the universal Atiyah class is
204: precisely the map
205: \[ \cO_\Delta\stackrel{\eta}{\lra}\Delta_*\Delta^*\cO_\Delta
206: \stackrel{\Delta_*I}{\lra} \bigoplus_i \Delta_*\Omega_X^i[i], \]
207: where $\eta$ is the unit of the adjunction $\Delta^*\adjoint \Delta_*$.
208: \end{proposition}
209:
210: \begin{proof}
211: We divide the proof of this proposition into several steps, to make it
212: more manageable. We will use the notations used in the proof of
213: Theorem~\ref{thm:hkr}.
214: \vspace{2mm}
215:
216: \noindent {\em Step 1.}
217: Consider the exact sequence
218: \[ 0\ra \Omega_\Delta^1 \ra \cO_{\Delta^{(2)}} \ra \cO_\Delta \ra 0 \]
219: which defines the universal Atiyah class $\alpha_1$. Tensoring it by
220: the locally free sheaf $\pi_2^* \Omega_X^{\otimes i}$ yields the exact
221: sequence
222: \[ 0 \ra \Omega_\Delta^{\otimes (i+1)} \ra \cO_{\Delta^{(2)}} \otimes
223: \pi_2^* \Omega_X^{\otimes i}\ra \Omega_\Delta^{\otimes i} \ra 0. \]
224: Stringing together these exact sequences for successive values of $i$ we
225: construct the exact sequence
226: \[ 0\ra \Omega_\Delta^{\otimes i} \ra \cO_{\Delta^{(2)}} \otimes
227: \pi_2^*\Omega_X^{\otimes (i-1)} \ra \cO_{\Delta^{(2)}} \otimes
228: \pi_2^*\Omega_X^{\otimes (i-2)} \ra \cdots \ra \cO_{\Delta^{(2)}}\ra
229: \cO_\Delta \ra 0, \]
230: whose extension class is precisely
231: \[ (\pi_2^*\Omega_X^{\otimes (i-1)}
232: \otimes \alpha_1) \circ (\pi_2^*\Omega_X^{\otimes (i-2)} \otimes
233: \alpha_1) \circ \cdots\circ \alpha_1:\cO_\Delta\ra
234: \Omega_\Delta^{\otimes i}[i]. \]
235:
236: \noindent {\em Step 2.}
237: We claim that there exists a map $\phi_\cdot$ of exact sequences
238: \begin{diagram}[height=2em,width=2em,labelstyle=\scriptstyle]
239: \cdots & \rTo & \hcB_i(X) & \rTo & \hcB_{i-1}(X) &
240: \rTo & \cdots & \rTo & \hcB_0(X) & \rTo & \cO_\Delta & \rTo & 0 \\
241: & & \dTo_{\phi_i'} & & \dTo_{\phi_{i-1}} & & & & \dTo_{\phi_0} & & \dEqual
242: & & \\
243: 0 & \rTo & \Omega_\Delta^{\otimes i} & \rTo & \cO_{\Delta^{(2)}}
244: \otimes \pi_2^*\Omega_X^{\otimes (i-1)} & \rTo & \cdots
245: & \rTo & \cO_{\Delta^{(2)}} & \rTo & \cO_\Delta & \rTo & 0,
246: \end{diagram}
247: where the top row is the (augmented) completed bar resolution defined
248: in the proof of Theorem~\ref{thm:hkr}, and the bottom row is the one
249: defined in Step 1. It is sufficient to define the maps in a local
250: patch $U=\Spec R$. Let $I = I_2 = \ker(R\otimes R \ra R)$ be the
251: ideal defining the diagonal in $U\times U$. Consider the maps
252: \[ \phi_i:\cB_i(R) = R^{\otimes (i+2)} \ra (R\otimes R)/I^2 \otimes_R
253: \Omega_R^{\otimes_R i} \]
254: defined by
255: \[ \phi_i(a_0\otimes a_1\otimes \cdots \otimes a_{i+1}) = (a_0\otimes
256: a_{i+1} + I^2) \otimes_R da_1 \otimes_R da_2
257: \otimes_R \cdots \otimes_R da_i \]
258: (we write $\Omega_R$ on the right because we use $\pi_2^*$). The
259: same argument as the one in the proof of~\cite[Lemma 4.1]{Yek} shows
260: that these maps are continuous with respect to the adic topology used
261: to complete $\cB_i(R) = R^{\otimes (i+2)}$, thus the maps $\phi_i$
262: descend to maps
263: \[ \phi_i:\hcB_i(R) \ra (R\otimes R)/I^2 \otimes_R
264: \Omega_R^{\otimes_R i} \]
265: which sheafify to give the desired maps
266: \[ \phi_i: \hcB_i(X) \ra \cO_{\Delta^{(2)}} \otimes \pi_2^*
267: \Omega_X^{\otimes i}. \]
268: The map $\phi_i'$ is the composition
269: \[ \hcB_i(X) \stackrel{\phi_i}{\lra} \cO_{\Delta^{(2)}}
270: \otimes \pi_2^* \Omega_X^{\otimes i} \ra \cO_\Delta \otimes \pi_2^*
271: \Omega_X^{\otimes i} = \Omega_\Delta^{\otimes i}. \]
272: \vspace{2mm}
273:
274: \noindent {\em Step 3.}
275: We now need to check the commutativity of the squares in the above
276: diagram. Note that since everything is local, we can assume we are in
277: an open patch $U=\Spec R$, $U\times U = \Spec R\otimes R$. The ideal
278: $I$ in $R\otimes R$ is generated by expressions of the form $r\otimes 1
279: -1\otimes r$ for $r\in R$. Then a relevant square in the above
280: diagram (before completing) is
281: \[
282: \begin{diagram}[height=2em,width=2em,labelstyle=\scriptstyle]
283: R\otimes R\otimes R\otimes R & \rTo^{d_1} & R\otimes R\otimes R \\
284: \dTo_{\phi_2} & & \dTo_{\phi_1} \\
285: (R\otimes R)/I^2 \otimes_R I/I^2 \otimes_R I/I^2 & \rTo^{d_1'} &
286: (R\otimes R)/I^2 \otimes_R I/I^2,
287: \end{diagram}
288: \]
289: where $(R\otimes R)/I^2$ is considered a right $R$-module by
290: multiplication in the second factor, and $I/I^2$ is considered an
291: $R$-module by multiplication in either factor (the two module
292: structures are the same). The maps in this diagram are:
293: \begin{align*}
294: & d_1(1\otimes b\otimes c\otimes 1) = b\otimes c\otimes 1 - 1\otimes
295: bc\otimes 1 + 1\otimes b\otimes c,\quad\mbox{the Hochschild
296: differential} \\
297: & d_1'((r+I^2)\otimes_R (i+I^2)\otimes_R (i'+I^2)) = (ri+I^2)\otimes_R
298: (i'+I^2), \\
299: & \phi_1(a\otimes b\otimes c) = (a\otimes c + I^2)\otimes_R (b\otimes
300: 1 - 1\otimes b + I^2), \\
301: & \phi_2(1\otimes b\otimes c\otimes 1) = (1\otimes 1 + I^2)\otimes_R
302: (b\otimes 1 - 1\otimes b + I^2) \otimes_R (c\otimes 1 - 1\otimes c +
303: I^2).
304: \end{align*}
305: Omitting the $+I^2$ terms, we have
306: \begin{align*}
307: \phi_1(d_1 & (1\otimes b\otimes c\otimes 1)) =
308: \phi_1(b\otimes c\otimes
309: 1 - 1\otimes bc\otimes 1 + 1\otimes b\otimes c) \\
310: & = (b\otimes 1)\otimes_R (c\otimes 1 - 1\otimes c) - (1\otimes
311: 1)\otimes_R (bc\otimes 1-1\otimes bc) + (1\otimes c)\otimes_R (b\otimes
312: 1 - 1\otimes b) \\
313: \intertext{which, using the right module structure on $R\otimes R$, equals}
314: & = (b\otimes 1)\otimes_R (c\otimes 1 - 1\otimes c) - (1\otimes
315: 1)\otimes_R (bc\otimes 1-1\otimes bc) + (1\otimes
316: 1)\otimes_R (b\otimes c - 1\otimes bc) \\
317: & = (b\otimes 1)\otimes_R (c\otimes 1 - 1\otimes c) - (1\otimes
318: 1)\otimes_R (bc\otimes 1 - b\otimes c) \\
319: \intertext{and note that, since the last term is equal to
320: $1\otimes bc - c\otimes b$ modulo $I^2$,}
321: & = (b\otimes 1)\otimes_R (c\otimes 1 - 1\otimes c) - (1\otimes
322: 1)\otimes_R (1\otimes bc - c\otimes b) \\
323: & = (b\otimes 1)\otimes_R (c\otimes 1 - 1\otimes c) - (1\otimes
324: b)\otimes_R (1\otimes c - c\otimes 1) \\
325: & = (b\otimes 1 - 1\otimes b) \otimes_R (c\otimes 1 - 1\otimes
326: c) \\
327: & = d_1'((1\otimes 1)\otimes_R (b\otimes 1 - 1\otimes b)
328: \otimes_R (c\otimes 1 - 1\otimes c)) \\
329: & = d_1'(\phi_2(1\otimes b\otimes c\otimes 1)).
330: \end{align*}
331: Similar computations ensure the commutativity of the other squares.
332: \vspace{2mm}
333:
334: \noindent {\em Step 4.}
335: Observe that there exists a natural map $\eta$ from the bar resolution
336: $\hcB_\cdot(X)$ to the bar complex $\hcC_\cdot(X) = \hcB_\cdot(X)
337: \otimes_{X\times X} \cO_\Delta$, simply given by $1\otimes \mu$ where
338: $\mu:\cO_{X\times X} \ra \cO_\Delta$ is the natural projection. This
339: map is immediately seen to be precisely the unit $\eta$ of the
340: adjunction $\Delta^*\adjoint \Delta_*$.
341:
342: It is now obvious that the composite
343: \[ \hcB_i(X) \stackrel{\phi_i'}{\lra}
344: \Omega_\Delta^{\otimes i} \stackrel{\epsilon}{\lra} \Omega_\Delta^i \]
345: is precisely the same as the map
346: \[ \hcB_i(X) \stackrel{\eta_i}{\lra} \hcC_i(X)
347: \stackrel{\Delta_*I_i}{\lra} \Omega_\Delta^i, \]
348: where $\eta_i$ is the $i$-th component of $\eta$, locally (before
349: completion) given by
350: \[ a_0 \otimes a_1 \otimes \cdots \otimes a_{i+1} \mapsto a_0 a_{i+1}
351: \otimes a_1 \otimes \cdots \otimes a_i, \]
352: and $\Delta_*I_i$ is the $i$-th component of the HKR isomorphism.
353:
354: Now chopping off at the last step the two exact sequences we have
355: studied above we get the diagram
356: \[
357: \begin{diagram}[height=2em,width=2em,labelstyle=\scriptstyle]
358: \cdots & \rTo & \hcB_i(X) &
359: \rTo & \hcB_{i-1}(X) & \rTo & \cdots & \rTo & \hcB_0(X) & \rTo & 0\\
360: & & \dTo_{\phi_i'} & & \dTo_{\phi_{i-1}} & & & &
361: \dTo_{\phi_0} & & \\
362: 0 & \rTo & \Omega_\Delta^{\otimes i}
363: & \rTo & \cO_{\Delta^{(2)}} \otimes \pi_2^*\Omega_X^{\otimes (i-1)} & \rTo & \cdots
364: & \rTo & \cO_{\Delta^{(2)}} & \rTo & 0 \\
365: & & \dTo_{p_i} & & & & & & & & & &\\
366: & & \Omega_\Delta^{\otimes i} & & & & & & & & & & \\
367: & & \dTo_{\epsilon} & & & & & & & & & & \\
368: & & \Omega_\Delta^i, & & & & & & & & & &
369: \end{diagram}
370: \]
371: which can be thought of as a map from the top complex (which
372: represents $\cO_\Delta$) to $\Omega_\Delta^i[i]$. In fact what we
373: have is a factoring
374: \[ \cO_\Delta \stackrel{p_i\circ\phi_\cdot}{\lra} \Omega_\Delta^{\otimes i}
375: \stackrel{\epsilon}{\lra} \Omega_\Delta^i \]
376: of the map
377: \[ \epsilon \circ p_i \circ \phi_\cdot = \Delta_* I_i \circ \eta, \]
378: where $\phi_\cdot$ is the map of complexes appearing at the top of the
379: above diagram. However, note that both the source and the target of
380: $\phi_\cdot$ are naturally isomorphic (in $\D(X\times X)$) to
381: $\cO_\Delta$, and then $\phi_\cdot$ can be viewed as the identity map
382: $\cO_\Delta \ra \cO_\Delta$. Under these identifications we conclude
383: \[ \epsilon \circ p_i = \Delta_* I_i \circ \eta. \]
384:
385: But the construction of $p_i$ is such that it is represented by the
386: $i$-step extension
387: \[
388: \begin{diagram}[height=2em,width=2em,labelstyle=\scriptstyle]
389: 0 & \rTo & \Omega_\Delta^{\otimes i}
390: & \rTo & \cO_{\Delta^{(2)}} \otimes \pi_2^*\Omega_X^{\otimes (i-1)} & \rTo & \cdots
391: & \rTo & \cO_{\Delta^{(2)}} & \rTo &\cO_\Delta & \rTo & 0,
392: \end{diagram}
393: \]
394: whose class we argued is
395: \[ (\pi_2^*\Omega_X^{\otimes (i-1)}
396: \otimes \alpha_1) \circ (\pi_2^*\Omega_X^{\otimes (i-2)} \otimes
397: \alpha_1) \circ \cdots\circ \alpha_1:\cO_\Delta\ra
398: \Omega_\Delta^{\otimes i}[i]. \]
399: Therefore
400: \[ p_i = (\pi_2^*\Omega_X^{\otimes (i-1)}
401: \otimes \alpha_1) \circ (\pi_2^*\Omega_X^{\otimes (i-2)} \otimes
402: \alpha_1) \circ \cdots\circ \alpha_1:\cO_\Delta\ra
403: \Omega_\Delta^{\otimes i}[i], \]
404: and hence
405: \[ \alpha_i = \epsilon \circ p_i = \Delta_* I_i \circ \eta. \]
406: We conclude that
407: \[ \exp(\alpha) = \bigoplus_i \alpha_i = \bigoplus \Delta_* I_i \circ
408: \eta = \Delta_*I \circ \eta. \]
409: \end{proof}
410:
411: \begin{theorem}
412: \label{thm:chernhar}
413: The composition
414: \[
415: \begin{diagram}[height=2em,width=2em,labelstyle=\scriptstyle]
416: K_0(X) & \rTo^{\ch}& HH_0(X)&\rTo^{I_\HKR} \bigoplus_i
417: H^i(X, \Omega^i_X)
418: \end{diagram} \]
419: is the usual Chern character map.
420: \end{theorem}
421:
422: \begin{proof}
423: Let $\cF\in K_0(X)$, and let
424: \[ \ch(\cF)\in HH_0(X) = \Hom_{X\times X}(\Delta_!\cO_{\Delta}, \cO_\Delta)
425: \]
426: be the Chern character defined in~(\ref{subsec:chernchar}). Let
427: \[ \ch'(\cF)\in \Hom_X(\cO_X, \Delta^*\cO_\Delta) \]
428: be the element that corresponds to $\ch(\cF)$ under the adjunction
429: $\Delta_!\adjoint \Delta^*$. If $\mu'$ is any element of
430: $\Hom_X(\Delta^*\cO_\Delta, S_X)$ and
431: \[ \mu = \Delta_*\mu'\circ \eta \]
432: is the corresponding element of $\Hom_{X\times X}(\cO_\Delta,
433: S_\Delta)$ under the adjunction $\Delta^*\adjoint\Delta_*$, the
434: construction of $\Delta_!$ is such that
435: \[ \Tr_X(\mu'\circ \ch'(\cF)) = \Tr_{X\times X}(\mu\circ \ch(\cF)). \]
436: (Here $\eta:\cO_\Delta \ra \Delta_*\Delta^*\cO_\Delta$ is the unit of
437: the adjunction.)
438:
439: On the other hand, the definition of $\ch(\cF)$ is such that for any
440: $\mu$,
441: \[ \Tr_{X\times X}(\mu\circ \ch(\cF)) =
442: \Tr_X(\pi_{2,*}(\pi_1^*\cF\otimes \mu)), \]
443: and $\ch(\cF)$ is the unique element in $HH_0(X)$ with this property.
444: We then have
445: \begin{align*}
446: \Tr_X(\mu'\circ \ch'(\cF)) & = \Tr_{X\times X}(\mu \circ \ch(\cF)) =
447: \Tr_X(\pi_{2,*}(\pi_1^*\cF\otimes \mu)) \\
448: & = \Tr_X(\pi_{2,*}(\pi_1^*\cF\otimes(\Delta_*\mu' \circ \eta))) \\
449: & = \Tr_X(\pi_{2,*}(\pi_1^*\cF\otimes \Delta_*\mu') \circ
450: \pi_{2,*}(\pi_1^*\cF\otimes \eta)) \\
451: & = \Tr_X(\cF\otimes \mu' \circ \Phi^\eta(\cF))\\
452: & = \Tr_X(\mu' \circ \Tr_\cF(\Phi^{\eta}(\cF))),
453: \end{align*}
454: where the last equality is~\cite[Lemma~2.4]{CalHH1}.
455: Since the trace induces a non-degenerate pairing and the above
456: equalities hold for any $\mu'$, it follows that
457: \[ \ch'(\cF) = \Tr_\cF(\Phi^{\eta}(\cF)). \]
458: Applying the isomorphism $I$ to both sides we conclude that
459: \[ I_\HKR(\ch(\cF)) = I\circ\ch'(\cF)) = I\circ\Tr_\cF(\Phi^{\eta}(\cF)) =
460: \Tr_\cF(\Phi^{\exp(\alpha)}(\cF)) = \ch_{\mathrm{orig}}(\cF), \]
461: where the third equality is Proposition~\ref{prop:etaatyiah}.
462: \end{proof}
463:
464:
465: