1: % \input{bheader}
2: % \input{bmacros}
3: % \input{dmacros}
4:
5: % new
6: % ---
7:
8: % \newcommand {\hsmooth}{\H^{\infty}}
9: % \newcommand {\hi}{\H_{i}}
10: % \newcommand {\hj}{\H_{j}}
11: % \newcommand {\pl}{\mathcal P_{\ell}}
12: % \newcommand {\fin}{^{\scriptscriptstyle{\operatorname{fin}}}}
13: % \newcommand {\hfin}{\H\fin}
14:
15: % \newcommand {\Vbox}{V_{\square}}
16: \newcommand {\sbar}{\overline{s}}
17: % \newcommand {\vertex}[3]{\left(\begin{array}{c}#1\\#2\thickspace #3\end{array}\right)}
18: % \renewcommand {\SS}{\mathcal S}
19:
20: %\begin{document}
21:
22: \chapter{Analytic properties of primary fields}
23: \label{ch:sobolev fields}
24: %//////////////////////////////////////////////
25:
26: This chapter is devoted to the study of the continuity properties of primary fields
27: of $L\Spin_{2n}$. These are required to construct (at first unbounded) explicit
28: intertwiners for the local loop groups by smearing the fields on functions supported
29: in complementary intervals. We show that any primary field
30: $\phi:\hfin_{i}\otimes V_{k}[z,z^{-1}]\rightarrow\hfin_{j}$ such that one of the
31: $\Spin_{2n}$--modules $\H_{i}(0),V_{k},\H_{j}(0)$ is minimal extends to a jointly
32: continuous operator--valued distribution
33: $\hsmooth_{i}\otimes C^{\infty}(S^{1},V_{k})\rightarrow\hsmooth_{j}$ satisfying as
34: expected
35: \begin{equation}\label{eq:smeared eq}
36: \pi_{j}(\gamma)\phi(f)\pi_{i}(\gamma)^{*}=\phi(\gamma f)
37: \end{equation}
38: for any $\gamma\in L\Spin_{2n}$.
39: When the charge $V_{k}$ is the vector representation, $\phi$ satisfies stronger
40: continuity properties and extends to a bounded map
41: $L^{2}(S^{1},V_{k})\rightarrow\B(\H_{i},\H_{j})$.\\
42:
43: The level 1 result is obtained in section \ref{se:sobolev 1} from the bosonic
44: construction of the primary fields given in chapter \ref{ch:vertex operator}.
45: The level $\ell$ result is proved in section \ref{se:sobolev l} and follows
46: because the primary fields belonging to the above class may be obtained as
47: $\ell$--fold tensor products of level 1 primary fields.
48: Some care is required in checking this and the corresponding finite--dimensional
49: analysis is carried out in section \ref{se:level l for intertwiners}. The identity
50: \eqref{eq:smeared eq} is proved in section \ref{se:smeared inter}.
51:
52: \ssection{Continuity of level 1 primary fields}
53: \label{se:sobolev 1}
54: %==============================================
55:
56: As observed by Wassermann \cite{Wa5}, the continuity of the level 1 primary
57: fields depends upon the fact that they may be written as the product of
58: generating functions whose modes are bounded operators. We begin by studying
59: these. The notation follows chapter \ref{ch:vertex operator}.
60:
61: \ssubsection{Norm boundedness of the vertex operators for $T\times\check T$}
62: %---------------------------------------------------------------------------
63:
64: Recall that the action of $\tc$ on $\IC[\weight]$ is given by
65: \begin{equation}\label{eq:reminder}
66: h \delta_{\mu}=\<h,\mu\>\delta_{\mu}
67: \end{equation}
68:
69: \begin{lemma}\label{zero modes boundedness}
70: Let $\eta$ be a $\T$--valued function on $\weight$ and
71: \begin{equation}
72: X_{\lambda}(z)=
73: V_{\lambda}z^{\lambda+\half{\sqnm{\lambda}}}\eta(\cdot)=
74: \sum_{m}X_{\lambda}(m)z^{-m}
75: \end{equation}
76: Then, $X_{\lambda}(m)$ is bounded in norm by one.
77: \end{lemma}
78: \proof
79: By \eqref{eq:reminder}, $X_{\lambda}(m)=V_{\lambda}\eta(\cdot)P_{m}$ where $P_{m}$
80: is the orthogonal projection onto
81: \begin{equation}
82: \bigoplus_{\substack{\mu\in\weight\\ \<\mu,\lambda\>+\half{\sqnm{\lambda}}=-m}}
83: \IC\cdot\delta_{\mu}
84: \end{equation}
85: The claimed boundedness follows since $V_{\lambda}\eta(\cdot)$ is unitary \halmos
86:
87: \ssubsection{Norm boundedness of pre-vertex operators of small conformal dimension}
88: %---------------------------------------------------------------------------------
89:
90: This subsection follows \cite{Wa5}. Recall from \S \ref{ss:Stone for nonzero} of
91: chapter \ref{ch:vertex operator} the definition of the exponential operators
92: \begin{align}
93: E^{-}(\alpha,z)&=
94: \exp\Bigl(-\sum_{n<0}\frac{\alpha(n)}{n}z^{-n}\Bigr)&
95: E^{+}(\alpha,z)&=
96: \exp\Bigl(-\sum_{n>0}\frac{\alpha(n)}{n}z^{-n}\Bigr)
97: \end{align}
98: which are formal power series with coefficents in $\End(\SS)$ and the fact that
99: \begin{equation}
100: E^{+}(\alpha,z)E^{-}(\beta,\zeta)=
101: \Bigl(1-\frac{\zeta}{z}\Bigr)^{\<\alpha,\beta\>}
102: E^{-}(\beta,\zeta)E^{+}(\alpha,z)
103: \end{equation}
104: We have now
105:
106: \begin{proposition}\label{non zero norm boundedness}
107: For any $\alpha\in i\t\cong\IR^{n}$ such that $\sqnm{\alpha}\leq 1$, the
108: modes $Y(n)$ of the pre--vertex operator
109: \begin{equation}
110: Y_{\alpha}(z)=
111: E^{-}(\alpha,z)E^{+}(\alpha,z)=
112: \exp\Bigl(-\sum_{n<0}\frac{\alpha(n)}{n}z^{-n}\Bigr)
113: \exp\Bigl(-\sum_{n>0}\frac{\alpha(n)}{n}z^{-n}\Bigr)
114: \end{equation}
115: satisfy
116: \begin{equation}\label{eq:bounded modes}
117: \|Y(n)\psi\|\leq\|\psi\|
118: \end{equation}
119: for any $\n\in\IZ$ and $\psi\in\SS$.
120: \end{proposition}
121: \proof
122: Consider first the case $\sqnm{\alpha}=1$. Then,
123: \begin{equation}
124: Y_{\alpha}(z)Y_{-\alpha}(\zeta)+
125: \Bigl(\frac{\zeta}{z}\Bigr)^{-1}Y_{-\alpha}(\zeta)Y_{\alpha}(z)=
126: \biggl[\Bigl(1-\frac{\zeta}{z}\Bigr)^{-1}+
127: \Bigl(1-\frac{z}{\zeta}\Bigr)^{-1}\Bigl(\frac{\zeta}{z}\Bigr)^{-1}\biggr]
128: \reg(\alpha,z,\zeta)
129: \end{equation}
130: where
131: \begin{equation}
132: \reg(\alpha,z,\zeta)=
133: E^{-}(\alpha,z)E^{-}(-\alpha,\zeta)E^{+}(\alpha,z)E^{+}(-\alpha,\zeta)=
134: \reg(-\alpha,\zeta,z)
135: \end{equation}
136: By (\ref{ch:vertex operator}.\ref{eq:delta}), the bracketed term is equal to
137: $\delta\Bigl(\frac{\zeta}{z}\Bigr)$ and since $\reg(\alpha,\zeta,\zeta)=1$,
138: we find by (\ref{ch:vertex operator}.\ref{eq:delta identity})
139: \begin{equation}
140: Y_{\alpha}(z)Y_{-\alpha}(\zeta)+
141: \Bigl(\frac{\zeta}{z}\Bigr)^{-1}Y_{-\alpha}(\zeta)Y_{\alpha}(z)=
142: \delta\Bigl(\frac{\zeta}{z}\Bigr)
143: \end{equation}
144: Writing $Y_{\alpha}(z)=\sum_{n}a_{n}z^{-n}$,
145: $Y_{-\alpha}(\zeta)=\sum_{n}b_{n}z^{-n}$ and recalling that the formal adjunction
146: property $Y_{\alpha}(z)^{*}=Y_{-\alpha}(z)$ implies that $b_{n}^{*}=a_{-n}$, we get
147: by taking modes on both sides
148: \begin{equation}
149: a_{n}a_{m}^{*}+a_{m-1}^{*}a_{n-1}=\delta_{n,m}
150: \end{equation}
151: and in particular
152: \begin{equation}
153: \|a_{n}\psi\|\leq\|\psi\|
154: \end{equation}
155:
156: The general case may be settled by the following factorisation trick. The
157: space $V_{+}=z^{-1}\tc[z^{-1}]$ splits as $V_{+}^{1}\bigoplus V_{+}^{2}$ for
158: any orthogonal decomposition $\tc=\tc^{1}\oplus\tc^{2}$. Correspondingly,
159: \begin{equation}
160: \SS=\bigoplus_{k}S^{k}V_{+}=\SS^{1}\bigotimes\SS^{2}
161: \end{equation}
162: and
163: \begin{equation}
164: Y_{\lambda\oplus\lambda^\prime}(z)=Y_{\lambda}(z)\otimes Y_{\lambda^\prime}(z)
165: \end{equation}
166: for any $\lambda\in\tc^{1}$, $\lambda'\in\tc^{2}$. If $\sqnm{\alpha}\leq 1$,
167: we may find, by possibly enlarging $\tc$ if it is one--dimensional,
168: $i\t\ni\alpha'\perp\alpha$ such that $\sqnm{\alpha\oplus\alpha'}=1$.
169: The modes $Y_{\alpha\oplus\alpha'}(n)$ satisfy \eqref{eq:bounded modes} and are
170: equal to
171: \begin{equation}
172: \sum_{p+q=n} Y_{\alpha}(p)\otimes Y_{\alpha'}(q)
173: \end{equation}
174: Let $d=d_{1}+d_{2}$ be the infinitesimal generators of rotations on $\SS^{1},\SS^{2}$.
175: Let $\eta\in\SS^{2}$ be the lowest energy vector so that
176: $Y_{\alpha'}(z)\eta=E^{-}(\alpha',z)\eta$ and therefore
177: $Y_{\alpha'}(0)\eta=\eta$. If $\xi\in\SS^{1}$ is any eigenvector of $d_{1}$, then
178: \begin{equation}
179: \|\xi\|^{2}\|\eta\|^{2}=\|\xi\otimes\eta\|^{2}\geq
180: \|Y_{\alpha\oplus\alpha'}(n)\xi\otimes\eta\|^{2}=
181: \sum_{p'+q'=n}\|Y_{\alpha}(p')\xi\|^{2}\|Y_{\alpha'}(q')\eta\|^{2}\geq
182: \|Y_{\alpha}(n)\xi\|^{2}\|\eta\|^{2}
183: \end{equation}
184: since the vectors $Y_{\alpha}(p)\xi$, $Y_{\alpha}(p')\xi$ have different energies
185: and are therefore orthogonal for $p\neq p'$ and the same holds for
186: $Y_{\alpha}(q)\eta$ and $Y_{\alpha}(q')\xi$ whenever $q\neq q'$. Thus
187: \begin{equation}
188: \|Y_{\alpha}(n)\xi\|\leq \|\xi\|
189: \end{equation}
190: Lastly, if $\xi=\sum \xi_{n}$ is a sum of eigenvectors of $d_{1}$ with
191: distinct eigenvalues, then
192: \begin{equation}
193: \|Y_{\alpha}(m)\xi\|^{2}=
194: \sum\|Y_{\alpha}(m)\xi_{n}\|^{2}\leq
195: \sum\|\xi_{n}\|^{2}=\|\xi\|^{2}
196: \end{equation}
197: \halmos
198:
199: \ssubsection{Sobolev estimates for products of norm bounded homogeneous fields}
200: %------------------------------------------------------------------------------
201:
202: This subsection follows \cite{Wa5}.
203: Let $\F_{i}$, $i=1\ldots k$ be inner product spaces supporting positive energy
204: representations $U_{\theta}^{i}=e^{i\theta d_{i}}$ of (a cover of) $\rot$. Call
205: a field $Y_{i}(z)=\sum_{n}Y_{i}(n)z^{-n}\in\End(\F_{i})[[z,z^{-1}]]$
206: {\it homogeneous} if $[d_{i},Y_{i}(n)]=-nY_{i}(n)$.
207:
208: \begin{proposition}\label{factorisation}
209: Let $Y_{1}(z)\cdots Y_{k}(z)$ be homogeneous fields with uniformly bounded modes
210: acting on $\F_{1}\cdots\F_{k}$. Then, the modes of
211: $Y(z)=Y_{1}(z)\otimes\cdots\otimes Y_{k}(z)$ satisfy
212: \begin{equation}
213: \|Y(m)\xi\|\leq C(1+|m|)^{k-1}\|(1+d)^{k-1}\xi\|
214: \end{equation}
215: for any $\xi\in\F=\F_{1}\otimes\cdots\otimes\F_{k}$ where
216: $d=d_{1}\otimes1\otimes\cdots\otimes1+
217: \cdots+1\otimes\cdots\otimes 1\otimes d_{k}$
218: \end{proposition}
219: \proof We have
220: \begin{equation}
221: Y(m)=\sum_{p_{1}+\cdots+p_{k}=m}Y_{1}(p_{1})\otimes\cdots\otimes Y_{k}(p_{k})
222: \end{equation}
223: Assume the $\F_{i}$ support positive energy representations of an $s$--sheeted cover
224: of $\rot$. If the right hand--side is applied to $\xi\in\F$ of energy $n\geq 0$, the
225: sum reduces to one involving only terms for which $p_{i}\leq n$ for any $i$. Their
226: number is bounded by that of solutions of $\sum p_{i}=m$, $p_{i}\leq n$,
227: $p_{i}\in s^{-1}\IZ$. However, $p_{i}=m-\sum_{j\neq i}p_{j}\geq m-n(k-1)$ and hence
228: for any $i$,
229: \begin{equation}
230: m-n(k-1)\leq p_{i}\leq n
231: \end{equation}
232: so that there are at most $s^{k-1}(1+nk-m)^{k-1}\leq s^{k-1}(1+nk+|m|)^{k-1}$
233: solutions since $p_{k}$ is determined once $p_{1},\ldots p_{k-1}$ are fixed.
234: It follows from $\|A\otimes B\|\leq\|A\|\|B\|$ for operators $A$, $B$
235: (Cauchy--Schwarz) that
236: \begin{equation}
237: \begin{split}
238: \|Y(m)\xi\|
239: &\leq Cs^{k-1}(1+nk+m)^{k-1}\|\xi\|\\
240: &\leq C'(1+|m|)^{k-1}(1+n)^{k-1}\|\xi\|\\
241: &=C'(1+|m|)^{k-1}\|(1+d)^{k-1}\xi\|
242: \end{split}
243: \end{equation}
244: where $C,C'$ are constants independent of $m$ and $\xi$ and we have used
245: $(1+d)\xi=(1+n)\xi$. The claimed inequality therefore holds for eigenvectors
246: of $d$ and hence for any $\xi\in\F$ since it is stable under taking orthogonal
247: sums of eigenvectors \halmos
248:
249: \ssubsection{Continuity of the level 1 spin primary fields}
250: %----------------------------------------------------------
251:
252: \begin{theorem}\label{th:spin sobolev}
253: Let $\phi_{s}:\hfin_{i}\otimes V_{s}[z,z^{-1}]\rightarrow\hfin_{j}$ be a
254: level 1 primary field of $L\Spin_{2n}$ whose charge is one of the spin
255: modules. Then $\phi_{s}$ extend to a jointly continuous map
256: $\hsmooth_{i}\otimes C^{\infty}(S^{1},V_{s})\rightarrow\hsmooth_{j}$.
257: \end{theorem}
258: \proof
259: Denote as customary by $\H_{i}^{t},\H_{j}^{t}$ the completion of $\hfin_{i},
260: \hfin_{j}$ with respect to the norm $\|\xi\|_{t}=\|(1+d)^{t}\xi\|$ where
261: $d$ is the infinitesimal generator of rotations.
262: The weights of the spin representations are of the form
263: $\mu=\half{1}(\epsilon_{1}\theta_{1}+\cdots+\epsilon_{n}\theta_{n})$
264: where $\epsilon_{i}\in\{\pm 1\}$ and may therefore be decomposed as orthogonal
265: sums of $\Delta=\left\lceil\frac{n}{4}\right\rceil$ vectors $\lambda_{i}$ with
266: $\sqnm{\lambda_{i}}\leq 1$. By theorem
267: \ref{ch:vertex operator}.\ref{th:level 1 fields}, the corresponding component
268: of $\phi_{s}$ factorises as
269: \begin{equation}
270: \Phi_{\mu}(z)=
271: Y_{\lambda_{1}}(z)\otimes\cdots\otimes Y_{\lambda_{\Delta}}(z)
272: \otimes V_{\mu}z^{\mu+\half{\sqnm{\mu}}}\eta(\cdot)
273: \end{equation}
274: for some $\T$--valued function $\eta$ on $\weight$. By lemma
275: \ref{zero modes boundedness} and proposition \ref{non zero norm boundedness},
276: all factors have modes bounded in norm by 1. Thus, if $\xi\in\hfin_{i}$ is of
277: energy $n\geq m$, proposition \ref{factorisation} yields
278: \begin{align}
279: \|\Phi_{\mu}(m)\xi\|_{t}
280: &=(1+n-m)^{t}\|\Phi_{\mu}(m)\xi\|\notag\\
281: &\leq C(1+|m|)^{\Delta}(1+n-m)^{t}\|\xi\|_{\Delta}\notag\\
282: &=C(1+|m|)^{\Delta}\frac{(1+n-m)^{t}}{(1+n)^{t}}\|\xi\|_{\Delta+t}\\
283: &\leq C(1+|m|)^{\Delta+\half{|t|}}\|\xi\|_{\Delta+t}\notag\\
284: \intertext{whence, for any $\xi\in\hfin_{i}$}
285: \|\Phi_{\mu}(m)\xi\|_{t}
286: &\leq C(1+|m|)^{\Delta+|t|}\|\xi\|_{{\Delta+t}}
287: \end{align}
288: since $\xi$ may be written as an orthogonal sum of eigenvectors of $d$.
289: Next, if
290: \begin{equation}
291: f=\sum_{\substack{m\in\IZ\\ \mu\in\Pi(s)}}
292: a_{m,\mu}v_{\mu}(m)\in V_{s}[z,z^{-1}]
293: \end{equation}
294: where $\Pi(s)$ is the set of weights of $V_{s}$, we have, by definition
295: $\phi_{s}(f)=\sum_{m,\mu}a_{m,\mu}\Phi_{\mu}(m)$. Then,
296: \begin{equation}\label{eq:spin estimate}
297: \|\phi_{s}(f)\xi\|_{t}\leq
298: C\sum_{m,\mu}|a_{m,\mu}|(1+|m|)^{\Delta+|t|}\|\xi\|_{\Delta+t}\leq
299: C' |f|_{\Delta+|t|}\|\xi\|_{\Delta+t}
300: \end{equation}
301: and $\phi_{s}$ extends to a continuous map
302: $C^{\infty}(S^{1},V_{s})\rightarrow\B(\H_{i}^{\Delta+t},\H_{j}^{t})$
303: \halmos\\
304:
305: \remark Notice that the estimates \eqref{eq:spin estimate} are not quite
306: optimal. For example, for $L\Spin_{8}$, the spin primary fields are Fermi
307: fields by proposition \ref{ch:vertex operator}.\ref{pr:vector in vertex}
308: and therefore extend to bounded maps
309: $L^{2}(S^{1},V_{s_{\pm}})\rightarrow\B(\H_{i},\H_{j})$.
310:
311: \ssection{Finite--dimensional $\Spin_{2n}$--intertwiners}
312: \label{se:level l for intertwiners}
313: %=======================================================
314:
315: By lemma \ref{ch:classification}.\ref{le:l lemma}, any $\Spin_{2n}$--module
316: admissible at level $\ell$ is contained in an $\ell$--fold tensor product
317: $V_{p_{1}}\otimes\cdots\otimes V_{p_{\ell}}$ where the $V_{p}$ are admissible
318: at level 1 and therefore minimal. In this section, we prove an analogous
319: factorisation for intertwiners $\phi:V_{i}\otimes V_{k}\rightarrow V_{j}$
320: when $V_{i},V_{k}, _{j}$ are admissible at level $\ell$ and one of them is
321: minimal. This will be used in the next section to show that level $\ell$
322: primary fields corresponding to the vertex $\vertex{V_{k}}{V_{j}}{V_{i}}$
323: can be written as the tensor product of level 1 primary fields.
324:
325: \ssubsection{Minimal representations}
326: %------------------------------------
327:
328: Recall from proposition \ref{ch:classification}.\ref{pr:weights of minimal} that
329: the weights of a minimal $G$--module $V$ lie on a single orbit of the Weyl group
330: and satisfy % They are the set of elements $\mu$ of minimal
331: % length in a given $\weight/\root$--coset and therefore satisfy
332: $\<\mu,\alpha^{\vee}\>\in\{1,0,-1\}$ for any root $\alpha$ and corresponding coroot
333: $\alpha^{\vee}=h_{\alpha}$.
334: If $\mu,\wt\mu$ are weights of $V$,
335: %distinct and of minimal length in the same $\weight/\root$-coset
336: then $\wt\mu=\mu+\alpha_{1}+\cdots+\alpha_{k}$ where the $\alpha_{i}$ are possibly
337: repeated roots. Since $\<\mu,\alpha_{i}\>\geq 0$ for all $i$ would imply
338: $\|\wt\mu\|>\|\mu\|$
339: there exists an $\alpha_{i}$ such that $\<\mu,\alpha_{i}^{\vee}\>=-1$ and therefore
340: $\|\mu+\alpha_{i}\|=\|\mu\|$. An iteration of this argument produces a permutation
341: $\tau$ of $\{1,\ldots,k\}$ such that
342: $\|\mu+\alpha_{\tau(1)}+\cdots+\alpha_{\tau(j)}\|=\|\mu\|$ for any $j=1\ldots k$.
343: In representation theoretic terms, the weight spaces of a minimal representation
344: are one--dimensional and if $v_{\wt\mu}$, $v_{\mu}$ are eigenvectors corresponding
345: to the weights $\wt\mu$, $\mu$ then, up to a non--zero multiplicative constant,
346: $v_{\wt\mu}=e_{\alpha_{\tau(k)}}\cdots e_{\alpha_{\tau(1)}}v_{\mu}$. Indeed, by
347: elementary $\mathfrak{sl}_{2}(\IC)$ theory, for any $j\in\{1\ldots k\}$,
348: $e_{\alpha_{\tau(j)}}e_{\alpha_{\tau(j-1)}}\cdots e_{\alpha_{\tau(1)}}v_{\mu}\neq 0$
349: since $\<\alpha_{j}^{\vee},\mu+\alpha_{1}+\cdots+\alpha_{j-1}\>=-1$ and therefore
350: $v_{\wt\mu}$ and $e_{\alpha_{\tau(k)}}\cdots e_{\alpha_{\tau(1)}}v_{\mu}$ are proportional
351: since they lie in the same weight space.\\
352:
353: If $V_{\delta},V_{\nu}$ are irreducible representations of highest weights $\delta,
354: \nu$ and $V_{\delta}$ is minimal, the tensor product $V_{\nu}\otimes V_{\delta}$
355: decomposes according to proposition \ref{ch:classification}.\ref{pr:tensor with minimal}
356: as
357: \begin{equation} \label{decompose}
358: V_{\nu}\otimes V_{\delta}=\bigoplus_{\wt\delta} V_{\nu+\wt\delta}
359: \end{equation}
360: where $\wt\delta$ varies among the weights of $V_{\delta}$ such that $\nu+\wt\delta$
361: is dominant.
362:
363: \begin{lemma}\label{approx}
364: Let $V_{\nu}$, $V_{\delta}$ be irreducible representations with highest weights $\nu$,
365: $\delta$ and corresponding eigenvectors $v_{\nu}$, $v_{\delta}$. Assume $V_{\delta}$
366: is minimal so that $V_{\nu}\otimes V_{\delta}$ decomposes according to \eqref{decompose}.
367: Then, if $\wt\delta$ is a weight of $V_{\delta}$ such that $\nu+\wt\delta$ is dominant
368: and $v_{\wt\delta}$ is a corresponding non--zero weight vector, the orthogonal projection
369: of $v_{\nu}\otimes v_{\wt\delta}$ on $V_{\nu+\wt\delta}\subset V_{\nu}\otimes V_{\wt\delta}$
370: is non--zero.
371: \end{lemma}
372: \proof
373: Let $\Omega_{\nu+\wt\delta}\in V_{\nu+\wt\delta}\subset V_{\nu}\otimes V_{\wt\delta}$ be a
374: non--zero highest weight vector. We claim that
375: $(\Omega_{\nu+\wt\delta},v_{\nu}\otimes v_{\wt\delta})\neq 0$. Indeed, if the contrary
376: holds, we shall prove inductively that
377: \begin{equation}\label{vanish}
378: (\Omega_{\nu+\wt\delta},
379: f_{\alpha_{\sigma(k)}}\cdots f_{\alpha_{\sigma(1)}}v_{\nu}\otimes
380: e_{\alpha_{k}}\cdots e_{\alpha_{1}} v_{\wt\delta})=0
381: \end{equation}
382: where $\alpha_{1}\ldots\alpha_{k}$ are (possibly repeated) simple roots and $\sigma$ is any
383: permutation of $\{1,\ldots, k\}$. Since such vectors form a spanning set of the eigenspace
384: of $V_{\nu}\otimes V_{\delta}$ corresponding to the weight $\nu+\wt\delta$, it follows that
385: $\Omega_{\nu+\wt\delta}=0$, a contradiction. To prove the inductive claim, notice that
386: \begin{equation}\label{expand}
387: \begin{split}
388: &
389: f_{\alpha_{\sigma(k)}}(
390: f_{\alpha_{\sigma(k-1)}}\cdots f_{\alpha_{\sigma(1)}}v_{\nu}\otimes
391: e_{\alpha_{k}}\cdots e_{\alpha_{1}}v_{\wt\delta})\\
392: =&
393: f_{\alpha_{\sigma(k)}}\cdots f_{\alpha_{\sigma(1)}}v_{\nu}\otimes
394: e_{\alpha_{k}}\cdots e_{\alpha_{1}} v_{\wt\delta}\\
395: -&
396: \<\alpha_{k}^{\vee},\wt\delta+(\alpha_{1}+\cdots+\alpha_{j-1})\>
397: f_{\alpha_{\sigma(k-1)}}\cdots f_{\alpha_{\sigma(1)}}v_{\nu}\otimes
398: e_{\alpha_{k}}\cdots \widehat{e_{\alpha_{j}}}\cdots e_{\alpha_{1}}v_{\wt\delta}\\
399: +&
400: f_{\alpha_{\sigma(k-1)}}\cdots f_{\alpha_{\sigma(1)}}v_{\nu}\otimes
401: e_{\alpha_{k}}\cdots e_{\alpha_{j}}f_{\alpha_{\sigma_{k}}}
402: e_{\alpha_{j-1}}\cdots e_{\alpha_{1}}v_{\wt\delta}
403: \end{split}
404: \end{equation}
405: where $j$ is the largest element of $\{1,\ldots,k\}$ such that $\alpha_{j}=\alpha_{\sigma(k)}$.
406: Now, if $f_{\alpha_{\sigma(k)}}\cdots f_{\alpha_{\sigma(1)}}v_{\nu}\otimes
407: e_{\alpha_{k}}\cdots e_{\alpha_{1}} v_{\wt\delta}$ is non--zero, then
408: $e_{\alpha_{j}}e_{\alpha_{j-1}}\cdots e_{\alpha_{1}}v_{\wt\delta}\neq 0$ and therefore by
409: elementary $\mathfrak{sl}_{2}(\IC)$ theory,
410: $f_{\alpha_{\sigma_{k}}}e_{\alpha_{j-1}}\cdots e_{\alpha_{1}}v_{\wt\delta}=0$
411: since $\wt\delta+(\alpha_{1}+\cdots+\alpha_{j-1})$ is a minimal weight and therefore
412: $\<\alpha_{j}^{\vee},\wt\delta+(\alpha_{1}+\cdots+\alpha_{j-1})\>\in\{1,0,-1\}$.
413: Therefore, if $f_{\alpha_{\sigma(k)}}\cdots f_{\alpha_{\sigma(1)}}v_{\nu}\otimes
414: e_{\alpha_{k}}\cdots e_{\alpha_{1}} v_{\wt\delta}\neq 0$, the last term on the right hand-side
415: of \eqref{expand} vanishes and by the inductive hypothesis and the fact that
416: $\Omega_{\nu+\wt\delta}$ is a highest weight vector, \eqref{vanish} holds \halmos\\
417:
418: \ssubsection{Admissible intertwiners for $\Spin_{2n}$}
419: %-----------------------------------------------------
420:
421: We begin by collecting some elementary facts about the spin representations of
422: $\Spin_{2n}$ which may be found in section \ref{ch:fermionic}.\ref{se:fd spinors}. The
423: complexified definining representation $V=\Vbox=\IC^{2n}$ has a natural bilinear,
424: non--degenerate symmetric form $B(\cdot,\cdot)$ which yields an isomorphism
425: $\so_{2n,\IC}\cong V\wedge V$ where the latter space acts on $V$ by
426: $u\wedge v\medspace w=uB(v,w)-v(u,w)$. If $f_{j}$, $0\neq j=-n\ldots n$ is an orthonormal
427: basis of $V$ satisfying $B(f_{j},f_{k})=\delta_{j+k,0}$, a basis for $\so_{2n,\IC}$
428: is given by the elementary matrices $F_{ij}=f_{i}\wedge f_{j}$. The Cartan subalgebra
429: corresponding to the block diagonal matrices in $\so_{2n,\IC}$ is then spanned by $F_{i,-i}$,
430: $i=1,\ldots n$. The roots of $\Spin_{2n}$ are $\theta_{k}+\theta_{l}$, $-n\leq k\neq\pm l
431: \leq n$ where the $\theta_{i}$, $i=1\ldots n$ are the dual basis to $F_{i,-i}$ and, by
432: definition $\theta_{-i}=-\theta_{i}$. The $\mathfrak{sl}_{2}(\IC)$--subalgebra
433: $\{e_{\alpha},f_{\alpha},h_{\alpha}\}$ of $\so_{2n,\IC}$ corresponding
434: to $\alpha=\theta_{k}+\theta_{l}$ is given by $\{F_{k,l},F_{k,-k}+F_{l,-l},F_{-l,-k}\}$.
435: The weight vectors in $V$ are exactly the $f_{j}$ since
436: $F_{i,-i}f_{j}=(\delta_{ij}-\delta_{i,-j})f_{i}=\theta_{j}(F_{i,-i})f_{j}$.\\
437:
438: The spin representations are obtained via the representation of the Clifford algebra of $V$
439: generated by the $\IC$-linear symbols $\psi(u)$, $u\in V$ subject to the relations
440: $\psi(u)\psi(v)+\psi(v)\psi(u)=2B(u,v)$, on the exterior algebra $\Lambda V^{1,0}$ where
441: $V^{1,0}$ is the subspace spanned by the $f_{j}$, $j>0$. The action is given explicitly by
442: \begin{equation}\label{eq:action of}
443: \psi(u) v_{1}\wedge\cdots\wedge v_{k}=
444: \left\{\begin{array}{rl}
445: \sqrt{2}
446: u\wedge v_{1}\wedge\cdots\wedge v_{k}&\text{if $u\in V^{1,0}$}\\[1.2 ex]
447: \sqrt{2} \sum_{j} (-1)^{j+1}B(u,v_{j})
448: v_{1}\wedge\cdots\wedge\widehat{v_{j}}\wedge\cdots\wedge v_{k}&\text{if $u\in V^{0,1}$}
449: \end{array}\right.
450: \end{equation}
451:
452: The representation is obtained by letting the Lie algebra element $u\wedge v$ act as
453: $\frac{1}{4}(\psi(u)\psi(v)-\psi(v)\psi(u))=\half{1}(\psi(u)\psi(v)-B(u,v))$. The even
454: and odd parts of the exterior algebra are clearly invariant under this action and are
455: irreducible. Their weights are easily read from
456: \begin{equation}
457: F_{i,-i} f_{j_{1}}\wedge\cdots\wedge f_{j_{k}}=
458: (\psi(f_{i})\psi(f_{-i})-\half{1})f_{j_{1}}\wedge\cdots\wedge f_{j_{k}}=
459: \left\{\begin{array}{rl}
460: \half{1}f_{j_{1}}\wedge\cdots\wedge f_{j_{k}}&
461: \text{if $i\in\{j_{1},\ldots,j_{k}\}$}\\[1.2 ex]
462: -\half{1}f_{j_{1}}\wedge\cdots\wedge f_{j_{k}}&
463: \text{if $i\notin\{j_{1},\ldots,j_{k}\}$}
464: \end{array}\right.
465: \end{equation}
466: so that the vector $f_{J}=f_{j_{1}}\wedge\cdots\wedge f_{j_{k}}$ corresponds to the weight
467: $-\half{1}\sum\theta_{i}+\sum_{p}\theta_{j_{p}}$. The highest weight of the half of the
468: exterior algebra containing the top exterior power $\Lambda^{n} V^{1,0}$ has therefore
469: highest weight $s_{+}=\half{1}(\theta_{1}+\cdots+\theta_{n})$ and the other has highest
470: weight $s_{-}=\half{1}(\theta_{1}+\cdots+\theta_{n-1}-\theta_{n})$.\\
471:
472: Finally, the Clifford multiplication map $V\otimes\Lambda V^{1,0}\rightarrow\Lambda V^{1,0}$
473: given by $v\otimes f_{J}\rightarrow\psi(v)f_{J}$ commutes with the action of $\Spin_{2n}$
474: and gives therefore rise to two intertwiners $V\otimes V_{s_{\pm}}\rightarrow V_{s_{\mp}}$.
475:
476: \begin{lemma}\label{fd intertwiners}
477: Let $V_{i},V_{j},V_{k}$ be irreducible representations of $G=\Spin_{2n}$, one of which
478: is minimal so that $\Hom_{G}(V_{i}\otimes V_{k},V_{j})$ is at most one--dimensional. If
479: all are admissible at level $\ell$ and $\Hom_{G}(V_{i}\otimes V_{k},V_{j})=\IC$, there
480: exist minimal $G$--modules $V_{i_{p}},V_{j_{p}},V_{k_{p}}$, $p=1\ldots\ell$ and
481: intertwiners $\phi_{p}\in\Hom_{G}(V_{i_{p}}\otimes V_{k_{p}},V_{j_{p}})$ such that
482: \begin{xalignat}{3}
483: V_{i}&\subset\bigotimes_{p=1}^{\ell}V_{i_{p}}&
484: V_{j}&\subset\bigotimes_{p=1}^{\ell}V_{j_{p}}&
485: V_{k}&\subset\bigotimes_{p=1}^{\ell}V_{k_{p}}
486: \end{xalignat}
487: and the corresponding restriction of $\otimes_{p}\phi_{p}$ to an intertwiner
488: $V_{i}\otimes V_{k}\rightarrow V_{j}$ is non--zero.
489: Moreover, if $V_{k}$ is minimal then one may choose $V_{i_{p}}=V_{j_{p}}$ and $V_{k_{p}}=\IC$
490: for $p=1\ldots\ell-1$ and $V_{k_{\ell}}=V_{k}$ so that $\otimes_{p}\phi_{p}$ is of the form
491: $1\otimes\cdots\otimes 1\otimes\phi_{p}$.
492: \end{lemma}
493: \proof
494: Up to a permutation of the modules, we may assume that $V_{k}$ is minimal and
495: that $\<\mu,\theta\>\leq\<\lambda,\theta\>$ where $\mu$, $\lambda$ are the
496: highest weights of $V_{i}$ and $V_{j}$ respectively and $\theta$ is the highest
497: root. If $\<\mu,\theta\>\leq\ell-1$, $V_{i}$ is contained, by lemma
498: \ref{ch:classification}.\ref{le:l lemma} in some tensor product
499: $V_{i_{1}}\otimes\cdots\otimes V_{i_{\ell-1}}$ with minimal factors. Then,
500: \begin{align}
501: V_{i}&\subset V_{i_{1}}\otimes\cdots\otimes V_{i_{\ell-1}}\otimes \IC\\[1.5 ex]
502: V_{k}&\subset \IC \otimes\cdots\otimes \IC \otimes V_{k}\\[1.5 ex]
503: V_{j}&\subset V_{i_{1}}\otimes\cdots\otimes V_{i_{\ell-1}}\otimes V_{k}
504: \end{align}
505: and $\phi=1\otimes\cdots\otimes 1\otimes 1$ clearly restricts to a non--zero intertwiner.
506: Assume now that $\<\mu,\theta\>=\<\lambda,\theta\>=\ell\geq 2$. If $V_{k}$ is the trivial
507: representation then $V_{i}=V_{j}$ and, by lemma \ref{ch:classification}.\ref{le:l lemma},
508: $V_{i}\subset V_{i_{1}}\otimes\cdots\otimes V_{i_{\ell}}$ for some minimal $V_{i_{p}}$
509: and the lemma holds. We shall treat the cases when $V_{k}$ is the vector or one of the
510: spin representations separately.\\
511:
512: $\mathbf{V_{k}=V_{s_{\pm}}}$.\\
513: Let $s$ and $\sbar$ be the highest weights of $V_{k}$ and of the other spin representation
514: so that $s=s_{\pm}=\half{1}(\theta_{1}+\cdots+\theta_{n-1}\pm\theta_{n})$ and
515: $\sbar=s_{\mp}$. We have $\lambda=\mu+\sigma$ where
516: $\sigma=\half{1}(\epsilon_{1}\theta_{1}+\cdots+\epsilon_{n}\theta_{n})$ for some
517: $\epsilon_{i}\in\{\pm 1\}$ is a weight of $V_{k}=V_{s}$. Up to a permutation of $V_{i}$,
518: $V_{j}$, we may assume that $\mu_{1}=\lambda_{1}+\half{1}$. Since
519: $\lambda_{1}+\lambda_{2}=\<\lambda,\theta\>=\<\mu,\theta\>=\mu_{1}+\mu_{2}$, we have
520: $\mu_{2}=\lambda_{2}-\half{1}$ and therefore
521: $\mu_{1}-\mu_{2}=\lambda_{1}-\lambda_{2}+1\geq 1$ so that $\rho=\mu-\theta_{1}$ is a
522: dominant weight. Let $V_{\rho}$ be the corresponding irreducible representation. Since
523: $\<\rho,\theta\>=\ell-1$, $V_{\rho}$ is contained in some tensor product
524: $V_{i_{1}}\otimes\cdots\otimes V_{i_{\ell-1}}$ with minimal factors. Thus, by
525: \eqref{decompose}
526: \begin{align}
527: V_{i} \subset V_{\rho}\otimes\Vbox
528: &\subset V_{i_{1}}\otimes\cdots\otimes V_{i_{\ell-1}}\otimes\Vbox\\[1.5 ex]
529: V_{k}&\subset\IC\otimes\cdots\otimes\IC\otimes V_{s}\\[1.5 ex]
530: V_{j} \subset V_{\rho}\otimes V_{\sbar}
531: &\subset V_{i_{1}}\otimes\cdots\otimes V_{i_{\ell-1}}\otimes V_{\sbar}
532: \end{align}
533: since $\mu=\rho+\theta_{1}$ and $\lambda=\mu+\sigma=\rho+\sigma'$ where
534: $\sigma'=\theta_{1}+\sigma$ is a weight of $V_{\sbar}$.
535: If $\psi:\Vbox\otimes V_{s}\rightarrow V_{\sbar}$ is Clifford multiplication, we claim
536: that the intertwiner $\phi=1\otimes\cdots\otimes 1\otimes\psi$ has a non--zero restriction
537: to $V_{i}\otimes V_{k}\rightarrow V_{j}$.
538: To see this, notice that the highest weight vector $v_{\mu}$ in
539: $V_{i}\subset V_{\rho}\otimes \Vbox$
540: is the product $v_{\rho}\otimes v_{\theta_{1}}=v_{\rho}\otimes f_{1}$ of the corresponding
541: highest weight vectors.
542: If $v_{\sigma}\in V_{s}$ is of weight $\sigma$ so that
543: $v_{\sigma}=\wedge_{j:\sigma_{j}=\half{1}}f_{j}$, then
544: $\phi(v_{\mu}\otimes v_{\sigma})=
545: v_{\rho}\otimes\psi(f_{1}\otimes v_{\sigma})=
546: v_{\rho}\otimes f_{1}\wedge v_{\sigma}$. Since $f_{1}\wedge v_{\sigma}\in V_{\sbar}$
547: is of weight $\sigma+\theta_{1}$ and $\lambda=\rho+(\sigma+\theta_{1})$, lemma
548: \ref{approx} implies that $\phi(v_{\mu}\otimes v_{\sigma})$ has a non--zero projection
549: on $V_{j}\subset V_{\rho}\otimes V_{\sbar}$ whence the conclusion.\\
550:
551: $\mathbf{V_{k}=\Vbox}$.\\
552: Up to a permutation of $V_{i}$ and $V_{j}$, we may assume that $\lambda$ is obtained
553: from $\mu$ by adding a box to the corresponding Young diagram, {\it i.e.}~ $\lambda=\mu+\theta_{j}$
554: where $j\geq 3$ since $\<\lambda,\theta\>=\<\mu,\theta\>$.
555: Let
556: \begin{equation}
557: \sigma=\half{1}(\theta_{1}+\cdots+\theta_{j-1}-\theta_{j}-\cdots-\theta_{n})
558: \end{equation}
559: so that it is a weight of $V_{s}$ with $s=s_{\pm}$ according to the parity of $n-j+1$.
560: As is readily verified, $\rho=\mu-\sigma$ is dominant and satisfies $\<\rho,\theta\>=\ell-1$.
561: Thus, if $V_{\rho}$ is the corresponding highest weight representation, then
562: $V_{\rho}\subset V_{i_{1}}\otimes\cdots\otimes V_{i_{\ell-1}}$ where the $V_{i_{p}}$
563: are minimal. It follows by \eqref{decompose} that
564: \begin{align}
565: V_{i}\subset V_{\rho} \otimes V_{s}
566: &\subset V_{i_{1}}\otimes\cdots\otimes V_{i_{\ell-1}}\otimes V_{s}\\[1.5 ex]
567: V_{k}&\subset \IC \otimes\cdots\otimes \IC \otimes \Vbox\\[1.5 ex]
568: V_{j}\subset V_{\rho} \otimes V_{\sbar}
569: &\subset V_{i_{1}}\otimes\cdots\otimes V_{i_{\ell-1}}\otimes V_{\sbar}
570: \end{align}
571: since $\mu=\rho+\sigma$ and $\lambda=\mu+\theta_{j}=\rho+\sigma+\theta_{j}$ where
572: $\sigma$, $\sigma+\theta_{j}$ are weights of $V_{s}$ and $V_{\sbar}$ respectively.
573: Let $\psi:V_{s}\otimes\Vbox\rightarrow V_{\sbar}$ be Clifford multiplication and
574: $\phi=1\otimes\cdots\otimes 1\otimes\psi$.
575: We claim that if $v_{\mu}\in V_{i}\subset V_{\rho}\otimes V_{s}$ is the highest
576: weight vector then $\phi(v_{\mu}\otimes f_{j})$ is a non--zero highest weight
577: vector in $V_{j}\subset V_{\rho}\otimes V_{\sbar}$ so that $\phi$ restricts to
578: a non--zero intertwiner $V_{i}\otimes V_{k}\rightarrow V_{j}$.
579: To prove our claim, assume that $\phi(v_{\mu}\otimes f_{j})=0$.
580: Let $v_{\rho}\in V_{\rho}$ be the highest weight vector and
581: $v_{\sigma}=\wedge_{i=1\ldots j-1}f_{i}\in V_{s}$ a vector
582: of weight $\sigma$ so that $\psi(f_{-j})v_{\sigma}=0$. Then,
583: \begin{equation}
584: \begin{split}
585: 0
586: &=(\phi(v_{\mu}\otimes f_{j}),v_{\rho}\otimes\psi(f_{j})v_{\sigma})\\[1.4 ex]
587: &=(1\otimes\psi(f_{j})v_{\mu},v_{\rho}\otimes\psi(f_{j})v_{\sigma})\\[1.4 ex]
588: &=(v_{\mu},v_{\rho}\otimes\psi(f_{-j})\psi(f_{j})v_{\sigma}) \\[1.4 ex]
589: &=2(v_{\mu},v_{\rho}\otimes v_{\sigma})
590: \end{split}
591: \end{equation}
592: in contradiction with lemma \ref{approx}.
593: Let now $e_{\alpha}\in\so_{2n,\IC}$ be a root vector corresponding to the
594: simple root $\alpha$. Then,
595: $e_{\alpha}\phi(v_{\mu}\otimes f_{j})=\phi(v_{\mu}\otimes e_{\alpha}f_{j})$.
596: Since $e_{\alpha}f_{j}$ has weight $\alpha+\theta_{j}$, this does not vanish
597: only if $\alpha=\theta_{j-1}-\theta_{j}$ and in that case is
598: proportional to $\phi(v_{\mu}\otimes f_{j-1})$. To conclude, it is therefore
599: sufficient to prove that $\phi(v_{\mu}\otimes f_{j-1})=0$.
600: To this end, notice that the weight spaces of $V_{s}$ are one--dimensional and
601: $v_{\mu}$ may therefore be written as
602: \begin{equation}
603: v_{\mu}=\sum_{} v_{\rho-(\sigma'-\sigma)}\otimes v_{\sigma'}
604: \end{equation}
605: where $\sigma'$ ranges over all weights of $V_{s}$ differing from $\sigma$ by
606: a sum of positive roots, $v_{\sigma'}=\wedge_{i:\sigma'_{i}=\half{1}}f_{i}$
607: and the $v_{\rho-(\sigma'-\sigma)}\in V_{\rho}$ have weight $\rho-(\sigma'-\sigma)$ .
608: Since such $\sigma'$ are of the form
609: $\sigma'=
610: \half{1}(\theta_{1}+\cdots+\theta_{j-1}-\epsilon_{j}
611: \theta_{j}-\cdots-\epsilon_{n}\theta_{n})$
612: where $\epsilon$ ranges over all even sign changes of the variables
613: $\theta_{j},\ldots,\theta_{n}$, we find that
614: $f_{j-1}\wedge v_{\sigma'}=0$ and therefore
615: \begin{equation}
616: \phi(v_{\mu}\otimes f_{j-1})=1\otimes\psi(f_{j-1})v_{\mu}=0
617: \end{equation}
618: \halmos
619:
620: \ssection{Continuity of higher level primary fields}
621: \label{se:sobolev l}
622: %===================================================
623:
624: \begin{theorem}\label{th:minimal is sobolev}
625: Let $\H_{i},\H_{j}$ be irreducible positive energy representations at level $\ell$ with
626: lowest energy subspaces $V_{i},V_{j}$ and $V_{k}$ an irreducible $\Spin_{2n}$--module
627: admissible at level $\ell$. If one of $V_{i},V_{j},V_{k}$ is minimal, the (projectively
628: unique) primary field $\psi:\hfin_{i}\otimes V_{k}[z,z^{-1}]\rightarrow \hfin_{j}$
629: extends to a jointly continuous operator--valued distribution
630: $\H_{i}^{\infty}\otimes C^{\infty}(S^{1},V_{k})\rightarrow\H_{j}^{\infty}$.
631: If, in addition, $V_{k}\cong\IC^{2n}$ is the
632: vector representation, $\psi$ extends to a bounded map
633: $L^{2}(S^{1},V_{k})\rightarrow\B(\H_{i},\H_{j})$.
634: \end{theorem}
635: \proof
636: By corollary \ref{ch:classification}.\ref{co:one dim},
637: $D=\dim(\Hom_{\Spin_{2n}}(V_{i}\otimes V_{k},V_{j}))\leq 1$. If $D=0$, then
638: $\psi=0$ and there is nothing to prove. If, on the other hand $D=1$, then by
639: lemma \ref{fd intertwiners}, we have
640: \begin{xalignat}{3}
641: V_{i}&\subset\bigotimes_{p=1}^{\ell} V_{i_{p}}&
642: V_{j}&\subset\bigotimes_{p=1}^{\ell} V_{j_{p}}&
643: V_{k}&\subset\bigotimes_{p=1}^{\ell} V_{k_{p}}
644: \end{xalignat}
645: where the $V_{i_{p}},V_{j_{p}},V_{k_{p}}$ are minimal and therefore admissible at
646: level 1. Moreover, there exist intertwiners
647: $\varphi_{p}\in\Hom_{\Spin_{2n}}(V_{i_{p}}\otimes V_{k_{p}},V_{j_{p}})$
648: such that $\otimes_{p}\varphi_{p}$ restricts to a non--zero intertwiner
649: $\varphi:V_{i}\otimes V_{k}\rightarrow V_{j}$.
650: Let $\H_{i_{p}},\H_{j_{p}}$ be the irreducible level 1 positive energy representations
651: with lowest energy subspaces $V_{i_{p}}$ and $V_{j_{p}}$ respectively.
652: The lowest energy suspaces of the level $\ell$ positive energy representations
653: \begin{xalignat}{2}
654: \H_{i_{1},\ldots,i_{\ell}}&=\bigotimes_{p=1}^{\ell}\H_{i_{p}}&
655: \H_{j_{1},\ldots,j_{\ell}}&=\bigotimes_{p=1}^{\ell}\H_{j_{p}}
656: \end{xalignat}
657: contain $V_{i}$ and $V_{j}$ respectively and therefore, by lemma
658: \ref{ch:classification}.\ref{le:generation},
659: $\H_{i_{1},\ldots,i_{\ell}}$ and $\H_{j_{1},\ldots,j_{\ell}}$ contain as submodules
660: $\H_{i}$ and $\H_{j}$ respectively. Denote by $P_{i}$ and $P_{j}$ the corresponding
661: orthogonal projections.
662: Let $\phi_{p}(z):\H_{i_{p}}\otimes V_{k_{p}}[z,z^{-1}]\rightarrow\H_{j_{p}}$ be the
663: level 1 primary field with initial term $\varphi_{p}$. Then
664: \begin{equation}
665: \phi(z)=P_{j}\phi_{1}(z)\otimes\cdots\otimes\phi_{\ell}(z)P_{i}
666: \end{equation}
667: is easily seen to be
668: a primary field of type $\H_{i}\otimes V_{k}[z,z^{-1}]\rightarrow\H_{j}$ when
669: $V_{k}[z,z^{-1}]$ is embedded in $V_{k_{1}}[z,z^{-1}]\otimes\cdots\otimes V_{k_{\ell}}
670: [z,z^{-1}]$ in the obvious way. Since the initial term of $\phi$ is $\varphi$,
671: $\phi\neq 0$ and therefore, up to a non--zero multiplicative constant, $\psi=\phi$.
672: We proceed now as in the proof of theorem \ref{th:spin sobolev}. By theorem
673: \ref{ch:vertex operator}.\ref{th:level 1 fields}, each $\phi_{p}$ is a product of vertex
674: operators with uniformly bounded modes and therefore so is $\phi$. It follows, by
675: proposition \ref{factorisation} that $\phi$ extends to a jointly continuous
676: bilinear map $C^{\infty}(S^{1},V_{k})\otimes\H_{i}^{\infty}\rightarrow\H_{j}^{\infty}$.
677: Finally, if $V_{k}$ is the vector representation we may, by lemma \ref{fd intertwiners},
678: choose $\phi(z)$ of the form
679: \begin{equation}
680: P_{j}1\otimes\cdots\otimes1\otimes\Psi(z)P_{i}
681: \end{equation}
682: where $\Psi$ is a level 1 vector primary field and is therefore continuous for
683: the $L^{2}$--norm by proposition \ref{ch:fermionic}.\ref{pr:L2 vector} \halmos\\
684:
685: \remark Let $\phi,\phi^{*}$ be a primary field and its adjoint with corresponding vertices
686: $\vertex{V_{k}}{V_{j}}{V_{i}}$ and $\vertex{V_{k}^{*}}{V_{i}}{V_{j}}$. If one of $V_{i},
687: V_{j},V_{k}$ is minimal then, by theorem \ref{th:minimal is sobolev}, the defining identity
688: for $\phi^{*}$, namely
689: \begin{equation}
690: (\phi(f)\xi,\eta)=(\xi,\phi^{*}(\overline{f})\eta)
691: \end{equation}
692: where $\xi\in\hfin_{i}$, $\eta\in\hfin_{j}$ and $f\in V_{k}[z,z^{-1}]$, extends to
693: $\xi\in\hsmooth_{i},\eta\in\hsmooth_{j}$ and $f\in C^{\infty}(S^{1},V_{k})$. In particular,
694: \begin{xalignat}{3}
695: \phi^{*}(\overline{f})&\subseteq \phi(f)^{*}&
696: &\text{and}&
697: \phi(f)&\subseteq{\phi^{*}(\overline{f})}^{*}
698: \end{xalignat}
699: so that, for any $f\in C^{\infty}(S^{1},V_{k})$, $\phi(f)$ and $\phi^{*}(\overline{f})$ are
700: densely defined, closeable operators.
701:
702: \ssection{Intertwining properties of smeared primary fields}
703: \label{se:smeared inter}
704: %===========================================================
705:
706: \begin{proposition}\label{pr:LG equivariance}
707: Let $\phi:\hfin_{i}\otimes V_{k}[z,z^{-1}]\rightarrow\hfin_{j}$ be a primary field
708: extending to an operator--valued distribution. Then, the following holds projectively
709: on $\hsmooth_{i}$, for any $\gamma\in LG\rtimes\rot$ and $f\in C^{\infty}(S^{1},V_{k})$.
710: \begin{equation}\label{eq:primary covariance}
711: \pi_{j}(\gamma)\phi(f)\pi_{i}(\gamma)^{*}=\phi(\gamma f)
712: \end{equation}
713: Moreover, if $d$ is the integrally--moded infinitesimal generator of rotations, then
714: \begin{equation}
715: e^{i\theta d}\phi(f)e^{-i\theta d}=\phi(f_{\theta})
716: \end{equation}
717: where $f_{\theta}(\phi)=f(\phi-\theta)$.
718: \end{proposition}
719: \proof
720: We use a standard ODE argument. Notice first that by continuity, $\phi$ intertwines the action
721: of $L\g$ on $\H_{i}^{\infty}$ and
722: $\H_{j}^{\infty}$, {\it i.e.}~
723: \begin{equation}
724: \pi_{i}(X)\phi(f)\xi-\phi(f)\pi_{j}(X)\xi=\phi(Xf)\xi
725: \end{equation}
726: for any $X\in L\g$, $f\in C^{\infty}(S^{1},V_{k})$, and $\xi\in\H_{i}^{\infty}$. Define now
727: $F(t)=e^{-t\pi_{j}(X)}\phi(e^{tX}f)e^{t\pi_{i}(X)}\xi$ where $X,f,\xi$ are as above. Then,
728: using the invariance of $\H_{i}^{\infty}$ under $LG$
729: (proposition \ref{ch:analytic}.\ref{invariance}) and the $C^{\infty}$--continuity of $\phi$,
730: \begin{equation}
731: \begin{split}
732: F(t+s)
733: &=
734: e^{-(t+s)\pi_{j}(X)}\phi(e^{tX}f+sXe^{tX}f+o(s))
735: \Bigl(e^{t\pi_{i}(X)}\xi+s\pi_{i}(X)e^{t\pi_{i}(X)}\xi+o(s)\Bigr)\\
736: &=
737: e^{-(t+s)\pi_{j}(X)}\Bigl(
738: \phi(e^{tX}f)e^{t\pi_{i}(X)}\xi+
739: s\phi(Xe^{tX}f)e^{t\pi_{i}(X)}\xi+s\phi(e^{tX}f)\pi_{i}(X)e^{t\pi_{i}(X)}\xi+o(s)\Bigr)\\
740: &=
741: F(t)+
742: s\Bigl(-\pi_{j}(X)\phi(e^{tX}f)e^{t\pi_{i}(X)}\xi+
743: \phi(Xe^{tX}f)e^{t\pi_{i}(X)}\xi+s\phi(e^{tX}f)\pi_{i}(X)e^{t\pi_{i}(X)}\xi\Bigr)+o(s)\\
744: &=
745: F(t)+o(s)
746: \end{split}
747: \end{equation}
748: where $o(s)s^{-1}\rightarrow 0$ as $s\rightarrow 0$ in $\H^{\infty}$ and therefore in $\H$.
749: Thus, $F\in C^{1}(\IR,\H)$ and $\dot F\equiv 0$ so that $F\equiv F(0)=\xi$. It follows
750: that $\phi$ intertwines the one--parameter subgroups of $LG$ and therefore $LG$ itself.
751: The commutation properties with $\rot$ follow in a similar way \halmos\\
752:
753: % \remark If $\phi:\hfin_{i}\otimes V_{k}[z,z^{-1}]\rightarrow\hfin_{j}$ is a primary field, and
754: % $U_{\theta}=e^{id\theta}$ is the integrally moded action of $\rot$, the previous proposition
755: % shows that $U_{\theta}\phi(f)U_{\theta}^{*}=\phi(f_{\theta})$ where
756: % $f_{\theta}(\phi)=f(\phi-\theta)$. Since on $\H_{i}$, $L_{0}=d+\Delta_{i}$ it follows that
757: % $e^{i\theta L_{0}}\phi(f)e^{-i\theta L_{0}}=e^{i\theta(\Delta_{i}-\Delta_{j})}\phi(f_{\theta})$.\\
758:
759: \remark\hfill
760: \begin{enumerate}
761: \item Recall that the central extensions of $L\Spin_{2n}$ corresponding to $\pi_{i},\pi_{j}$
762: are canonically isomorphic by proposition \ref{ch:analytic}.\ref{classify extension}
763: and denote either of them by $\L\Spin_{2n}$. It is not difficult to show that a more
764: astringent relations than \eqref{eq:primary covariance} holds, namely the (non--projective)
765: identity
766: \begin{equation}\label{eq:primary covariance 2}
767: \pi_{j}(\wt\gamma)\phi(f)\phi_{i}(\wt\gamma)^{*}=\phi(\wt\gamma f)
768: \end{equation}
769: for any $\wt\gamma\in\L\Spin_{2n}$.
770: \item Somewhat conversely, Wassermann has pointed out that it might be possible to
771: use the continuity of the level 1 spin primary fields and \eqref{eq:primary covariance}
772: to give an alternative proof of
773: proposition \ref{ch:analytic}.\ref{classify extension} for $G=\Spin_{2n}$.
774: Notice first that it is only necessary to prove this at level 1 since the level
775: $\ell$ representations are contained in the $\ell$--fold tensor product of the
776: level 1 representations. Moreover, we already know that the central extensions
777: corresponding to the level 1 vacuum and vector representations are isomorphic
778: since these arise as summands of the Neveu--Scharz Fock space
779: (proposition \ref{ch:fermionic}.\ref{pr:level 1 reps}). The same holds for the
780: level 1 spin representations since they are summands of the Ramond sector. Thus,
781: it is sufficient to prove the isomorphism of the central extensions corresponding
782: to the the vacuum representation $(\pi_{0},\H_{0})$ and one of the spin
783: representations $(\pi_{s},\H_{s})$ at level 1. Let
784: $\phi_{s}:\hfin_{0}\otimes V_{s}[z,z^{-1}]\rightarrow\hfin_{s}$ be the corresponding
785: primary field. By theorem \ref{th:spin sobolev} and proposition
786: \ref{pr:LG equivariance}, the following holds projectively for any
787: $f\in C^{\infty}(S^{1},V_{s})$ and $\gamma\in LG$
788: \begin{equation}\label{eq:correcting phase}
789: \pi_{s}(\gamma)\phi_{s}(f)\pi_{0}(\gamma)^{*}=\phi_{s}(\gamma f)
790: \end{equation}
791: This identity may be used to define a continuous map between the two central extensions
792: by associating to any lift $\wt\pi_{0}(\gamma)\in U(\H_{0})$ of $\pi_{0}(\gamma)$
793: the unique unitary lift $\wt\pi_{s}(\gamma)$ of $\pi_{s}(\gamma)$ such that
794: \eqref{eq:correcting phase} holds without any additional phase corrections. However,
795: the proof that it is a homomorphism requires, to the best of our knowledge, a proof
796: of the fact that the definition of the map is independent of the particular $f$ chosen.
797: \end{enumerate}
798:
799: %\input{thesisbiblio}
800: %\end{document}