1: \magnification=\magstep1 \voffset0.4in
2: \vsize=8.9truein
3: \hsize=138truemm
4: \overfullrule=0pt \input epsf
5: %-----------example from page 252 of texbook----
6: %this gives pages numbers on top, running title, no title on p.1
7: \nopagenumbers %this supresses pagenos at foot
8: \headline={\ifnum\pageno=1 \hfil\else\ifodd\pageno\rightheadline \else\leftheadline\fi\fi}
9: \def\rightheadline{\tenrm\hfil CFT and mapping class groups\hfil\folio}
10: \def\leftheadline{\tenrm\folio\hfil CFT and mapping class groups\hfil}
11: \voffset=2\baselineskip
12: %-------------
13: %\def\eqde{\,{\buildrel \rm def \over =}\,}
14: \def\eqde{\,{:=}} \def\leaderfill{\leaders\hbox to 1em{\hss.\hss}\hfill}
15: \def\al{\alpha} \def\la{\lambda} \def\be{\beta} \def\ga{\gamma}
16: \def\J{{\cal J}} \def\c{{\cal C}} \def\i{{\rm i}} \def\B{{\cal B}}
17: \def\si{\sigma} \def\eps{\epsilon} \def\om{\omega} \def\ka{{\kappa}}
18: \def\s{{\cal S}} \def\k{\bar{k}} \def\I{{\cal I}}
19: \def\r{\bar{r}} \def\L{{L}} \def\M{{\cal M}}
20: \def\E{{\cal E}} \def\p{{\cal P}} \def\D{{\cal D}} \def\F{{\cal F}}
21: \def\N{{\cal N}} \def\W{{\cal W}} \def\H{{\Bbb H}}
22: \font\huge=cmr10 scaled \magstep2
23: \def\QED{\vrule height6pt width6pt depth0pt}
24: \font\smcap=cmcsc10 \font\smit=cmmi7 \font\smal=cmr7 %\font\smbf=cmb7
25: \def\z{{\cal Z}} \def\G{\Gamma}
26: \def\va{\varphi}
27: %%%%% Blackboard bold characters %%%%%
28: \input amssym.def
29: \def\Z{{\Bbb Z}} \def\R{{\Bbb R}} \def\Q{{\Bbb Q}} \def\Qq{{\Bbb Q}}
30: \def\C{{\Bbb C}} \def\Fb{{\Bbb F}} \def\Nn{{\Bbb N}} \def\M{{\Bbb M}}
31: \font\smit=cmmi7
32: \def\sdprod{{\times\!\vrule height5pt depth0pt width0.4pt\,}}
33: \def\boxit#1{\vbox{\hrule\hbox{\vrule{#1}\vrule}\hrule}}
34: \def\splus{\,\,{\boxit{$+$}}\,\,}
35: \def\stimes{\,\,{\boxit{$\times$}}\,\,}
36: \def\sdot{\,{\boxit{$\cdot$}}\,}
37: \def\buildrel#1\over#2{{\mathrel{\mathop{\kern0pt #2}\limits^{#1}}}}
38:
39: \centerline{{\bf \huge Conformal field theory and mapping class groups}}
40: \bigskip
41:
42: \centerline{{Terry Gannon}}\smallskip
43:
44: \centerline{{Mathematics Dept, U of Alberta, Canada}}\bigskip
45:
46: {\narrower\noindent{\bf Abstract.} Rational conformal field theories
47: produce a tower of finite-dimensional representations of surface
48: mapping class groups, acting on the conformal blocks of the theory.
49: We review this formalism. We show that many recent mathematical developments
50: can be fit into the first 2 floors of this tower. We also review what is known
51: in higher genus.
52:
53: }\bigskip
54:
55: \noindent{{\bf 1. Introduction.}}
56:
57: Conformal field theory (CFT) is an exceptionally symmetric
58: quantum field theory. It takes classical mathematical structures, and
59: `loops' or `complexifies' them, to produce
60: infinite-dimensional structures such as affine algebras or vertex operator
61: algebras. Its importance to math -- which is considerable -- is that the
62: resulting structures tend to straddle several areas, such as geometry,
63: algebra, number theory, functional analysis, ....
64:
65: One of its most beautiful but least appreciated accomplishments is the
66: reorganisation of several recent mathematical developments, such as
67: Monstrous Moonshine, Jones' knot invariant,
68: the modularity of affine Kac-Moody characters, and
69: braid group monodromy of the KZ equation,
70: into the first two floors of an infinite tower. This paper describes the
71: resulting picture. For more details see the book [18] and references therein.
72:
73:
74: \bigskip\noindent{{\bf 2. Mapping class groups.}}
75:
76: Up to homeomorphism, a (connected oriented real) surface is completely
77: characterised by its genus $g$ and number $n$ of boundary components (punctures). In fact,
78: there's only one way (up to equivalence) to give it a
79: real-differential structure.
80: But a real surface can also be a complex curve -- it can usually be
81: given a complex-differential (equivalently, a conformal) structure in
82: infinitely many different ways.
83:
84: For example, the torus $\R^2/\Z^2$ can be given a complex
85: structure by replacing $\R^2$ with $\C$ and $\Z^2$ by
86: $\Z+\tau\Z$ for any $\tau\in\C$
87: with nonzero imaginary part. In fact, any torus is
88: conformally equivalent to one of the form $\C/(\Z+\tau\Z)=:T_\tau$, where
89: $\tau$ lies in the {upper-half plane}
90: $\H:=\{x+\i y\in\C\,|\,y>0\}$.
91: Moreover, the tori $T_\tau$ and $T_{\tau'}$ are themselves conformally equivalent, iff
92: $\tau'={a\tau+b\over c\tau+d}$
93: for a matrix $\left(\matrix{a&b\cr c&d}\right)\in {\rm SL}_2(\Z)$.
94:
95: The set of possible complex structures on the torus forms the
96: {\it moduli space} ${\frak M}_{1,0}$, so labelled because the torus has
97: genus 1 and 0 punctures. This moduli space can be identified with
98: the orbifold $\H/{\rm SL}_2(\Z)$:
99: we call SL$_2(\Z)=:\Gamma_{1,0}$ its {\it mapping class
100: group}, and $\H=:{\frak T}_{1,0}$ its {\it Teichm\"uller
101: space}. Similarly, $\Gamma_{g,n}$,
102: ${\frak T}_{g,n}$, and ${\frak M}_{g,n}={\frak T}_{g,n}/\Gamma_{g,n}$ can be
103: defined for any other $g,n\ge 0$. In particular,
104: the Teichm\"uller space ${\frak T}_{g,n}$ (a simply connected
105: complex manifold) accounts for `continuous'
106: conformal equivalences, while the mapping class group $\Gamma_{g,n}$
107: (almost always an infinite discrete group) contains the left-over `discontinuous'
108: ones. The moduli spaces usually have conical singularities, corresponding
109: to surfaces with extra symmetries; taking into account these stabilisers,
110: $\Gamma_{g,n}$ will be the (orbifold) fundamental group of ${\cal M}_{g,n}$.
111:
112:
113: For example,
114: ${\frak M}_{1,0}$ is a sphere with a puncture (corresponding to the cusp
115: $\Bbb{Q}\cup \i\infty$), and
116: conical singularities at $\tau=\i$ and $e^{2\pi\i/3}$.
117: Because $\C/(\Z+\tau\Z)$ can also be interpreted as a torus with a special point, namely the
118: additive identity 0, we also have ${\frak T}_{1,1}=\H$ and
119: $\Gamma_{1,1}={\rm SL}_2(\Z)$.
120:
121:
122:
123:
124: The surfaces relevant to our story possess additional structure.
125: Let $\Sigma$ be a compact genus-$g$ surface
126: with $n$ marked points $p_i\in\Sigma$. About
127: each point $p_i$ choose a local coordinate $z_i$, vanishing at $p_i$ --
128: this identifies a neighbourhood
129: of $p_i$ with a neighbourhood of $0\in\C$. We call $(\Sigma,\{p_i\},\{z_i\})$ an {\it enhanced
130: surface}{} of type $(g,n)$. The resulting moduli space $\widehat{{\frak M}}_{g,n}$
131: is infinite-dimensional, but its mapping class group $\widehat{\Gamma}_{g,n}$ is an extension
132: of $\Gamma_{g,n}$ by $\Z^n$.
133: For example, $\widehat{\Gamma}_{1,1}$ is the braid group ${\cal B}_3$.
134:
135: As we will see below, a
136: rational conformal field theory gives finite-dimensional representations
137: of each $\widehat{\Gamma}_{g,n}$ -- merely projective for $n=0$, but it seems
138: a true one for $n\ge 1$ (though a proof of trueness is to my knowledge only available
139: for $g\le 1$). Enhanced surfaces are important because they have canonical sewings.
140: Nevertheless it is common to restrict instead to the projective
141: representations of $\Gamma_{g,n}$, and pay at most lipservice to the coordinates
142: $z_i$.
143:
144: \bigskip\noindent{{\bf 3. Conformal field theory.}}
145:
146: This section introduces the correlation functions and chiral blocks of conformal
147: field theory.
148:
149: A {conformal field theory} (CFT) is a quantum field theory, usually on a two-dimensional space-time
150: $\Sigma$, whose symmetries include the conformal transformations
151: (so conformally equivalent space-times are identified).
152: We restrict to compact orientable $\Sigma$. The same
153: CFT lives simultaneously on all such $\Sigma$. See e.g. [9,16,19,27],
154: and Chapter 4 of [18] for reviews.
155:
156:
157: Two dimensions are special for CFT because the local
158: conformal maps, which form the Lie algebra ${\frak{so}}_{n+1,1}
159: (\R)$ in $\R^n$ for $n>2$, becomes infinite-dimensional in $\R^2$
160: (thanks to their identification with (anti-)holomorphic maps).
161: The conformal algebra in two dimensions consists of two commuting
162: copies of the Witt algebra $\frak{Witt}$ (one for the
163: holomorphic maps, and the other for anti-holomorphic ones).
164: $\frak{Witt}$ is the infinite-dimensional Lie algebra of vector fields
165: on $S^1$, and has a basis $\ell_{n}$, $n\in\Z$, satisfying
166: $$[\ell_m,\ell_n]=(m-n)\ell_{m+n}\ .\eqno(3.1)$$
167: Its unique nontrivial central extension is the Virasoro algebra $\frak{Vir}$,
168: with basis $L_n,C$ satisfying $[L_n,C]=0$ and
169: $$[L_m,L_n]=(m-n)L_{m+n}+\delta_{n,-m}{m\,(m^2-1)\over 12}\,C\ .
170: \eqno(3.2)$$
171:
172:
173: Basic data in the CFT are the quantum fields $\varphi(z)$,
174: called {\it vertex operators}, living on space-time $\Sigma=S^2=\C\cup\{\infty
175: \}$ and centred at $z=0$. Being quantum
176: fields, these $\varphi$ are
177: `operator-valued distributions' on $\Sigma$, acting on the space ${\cal H}$ of
178: states for $\Sigma$. The most important vertex operators
179: are the {\it stress-energy tensors}
180: $T(z),\overline{T}(z)$,
181: which are the conserved currents of the conformal symmetry,
182: as promised by Noether's Theorem; the corresponding conserved charges are
183: operators $L_n,\overline{L}_m$ defining a $\frak{Vir}$-representation, with central term $C$
184: given by scalars $cI,\overline{c}I$ called the {\it central charges}.
185:
186:
187: In a typical quantum field theory, a theoretical physicist
188: makes contact with experiment by computing transition amplitudes
189: $\langle{\rm out}|{\rm in}\rangle$ between in-coming and out-going states,
190: given mathematically by a Hermitian product $(|{\rm
191: out}\rangle,S|{\rm in}\rangle)$ in the Hilbert space ${\cal H}$ of states,
192: for some operator $S$ called the scattering matrix. In practise these can
193: only be calculated in infinite time ($t\rightarrow\pm \infty$)
194: limits. The typical way (`LSZ reduction formulae') to express these
195: asymptotic amplitudes is via artifacts sometimes called {\it correlation
196: functions}. The theory is regarded as solved if all correlation
197: functions can be computed. We are interested in the correlation functions
198: $$\left\langle \varphi_{1}(z_1)\,\varphi_{2}(z_2)\cdots\varphi_{n}(z_n)
199: \right\rangle_{\Sigma;p_1,\ldots,p_n}\eqno(3.3)$$
200: of CFT, for any vertex operators $\varphi_i$ and
201: any enhanced surface $(\Sigma,\{p_i\},z_i)$ (so
202: $\varphi_{i}(z_i)$ is `centred' at $p_i\in\Sigma)$.
203: The remainder of this section explains how physicists think of these
204: CFT correlation functions. Their intuition is provided by string theory.
205:
206: In a typical quantum theory, correlation functions
207: are calculated perturbatively by Taylor-expanding in some coupling constant.
208: For this purpose,
209: Feynman's path integral formulation -- the quantisation of Hamilton's
210: action principle
211: in classical mechanics -- is convenient.
212: Each term in this perturbation series is computed
213: separately using Feynman diagrams and rules.
214: The Feynman diagrams of quantum field theory
215: are graphs, with a different kind of edge for each species of particle, and
216: a different kind of vertex for every term in the interaction part of the Lagrangian;
217: Feynman's rules describe how to go from these diagrams
218: to certain integral expressions and hence to the individual terms
219: in the Taylor series expansion of the given correlation function.
220: Feynman diagrams are combinatorial artifacts describing `virtual' (non-real)
221: processes; topologically equivalent ones are identified,
222: and in practise only the simplest are ever considered.
223:
224: Applying this perturbation formalism to string theory
225: recovers CFT. Consider for convenience closed strings. Then
226: CFT lives on the world-sheet
227: $\Sigma$ (string theory's Feynman diagrams) traced by the strings as they
228: virtually evolve, colliding and separating, through time:
229: string amplitudes (in e.g.\ 26-dimensional space-time)
230: can be expressed as correlation functions of a (point-particle) CFT
231: in two dimensions. The boundaries
232: of these world-sheets are the in-coming and out-going strings;
233: the world-sheets for asymptotic amplitudes have semi-infinite end-tubes and
234: can be conformally mapped to compact surfaces with punctures $p_i$
235: (one for every external string). The corresponding Feynman integral
236: is over moduli space $\widehat{{\cal M}}_{g,n}$. The data
237: of those external strings are stored in the appropriate vertex operator
238: attached to that point $p_i$.
239: The Witt algebra arises here as infinitesimal reparametrisations
240: of the string (a circle).
241:
242: Everything in CFT comes in
243: a combination of strictly holomorphic, and strictly anti-holomorphic,
244: quantities. Here,
245: `holomorphic' is in terms of space-time $\Sigma$ (which locally looks
246: like $\C$), or the appropriate
247: moduli space (which usually locally looks like $\C^\infty$). These holomorphic and
248: anti-holomorphic building blocks are called {\it chiral}. A CFT is studied by
249: first analysing its chiral parts, and then determining explicitly how they
250: piece together to form the physical(=bi-chiral) quantities.
251: Almost all attention by
252: mathematicians has focused on the chiral (as opposed to physical) data.
253: In string theory, this holomorphic/anti-holomorphic alternative corresponds to
254: classical ripples travelling clockwise/anti-clockwise around the string.
255:
256:
257: A typical vertex operator $\varphi(z)$ depends neither holomorphically nor
258: anti-holomorphically on $z$.
259: Let ${\cal V}$ consist of all the holomorphic, and
260: $\overline{\cal V}$ the anti-holomorphic, ones.
261: For example, the {stress-energy tensor} $T(z)$ and all of its derivatives
262: lie in ${\cal V}$.
263: In the very simplest CFTs, called the minimal models, ${\cal V}$
264: consists only of $T(z)$ and its derivatives.
265:
266: These {\it chiral algebras} ${\cal V},\overline{\cal V}$ have a
267: rich mathematical structure, with a `multiplication' coming from the so-called
268: operator product expansion, and are examples of {\it vertex operator algebras}
269: (see [28], or Chapter 5 of [18]).
270: ${\cal V}$ and $\overline{\cal V}$ mutually commute and
271: the full symmetry `algebra' of the CFT can be identified with
272: ${\cal V}\oplus\overline{\cal V}$.
273: Since quantum fields act on state-space ${\cal H}$, it
274: carries a representation of ${\cal V}\oplus\overline{
275: {\cal V}}$ and decomposes into a direct integral of
276: irreducible ${\cal V}\oplus\overline{{\cal V}}$-modules.
277: A {\it rational conformal field theory} (RCFT) is one whose
278: state-space ${\cal H}$ decomposes in fact into a {\it finite direct sum}
279: $${\cal H}=\oplus_{M\in\Phi,\overline{N}\in\overline{\Phi}}\,
280: {\cal Z}_{M,\overline{N}}\,M\otimes \overline{N}\ ,\eqno(3.4)$$
281: where $\Phi$ and $\overline{\Phi}$ denote the (finite) sets of
282: irreducible ${\cal V}$- and $\overline{\cal V}$-modules,
283: and the ${\cal Z}_{M,\overline{N}}\ge 0$ are multiplicities.
284: The RCFT are especially symmetric and well-defined quantum field theories
285: and are the CFTs we're interested in.
286: The name `rational' arises
287: because their central charges $c,\overline{c}$ lie in $\Q$.
288:
289: The correlation functions (3.3) can be expressed in terms of purely
290: chiral
291: quantities called {\it conformal} or {\it chiral blocks}, denoted
292: $$\left\langle {\cal I}_1(v_1,z_1)\,{\cal I}_2(v_2,z_2)\cdots
293: {\cal I}_n(v_n,z_n)\right\rangle_{(\Sigma;p_1,\ldots,p_n;M^1,\ldots,M^n)}\ .
294: \eqno(3.5)$$
295: As usual, $(\Sigma,\{p_i\},\{z_i\})$ is an enhanced surface, and
296: to each $p_i$ we assign a ${\cal V}$-module $M^i\in \Phi$ and a
297: state $v_i\in M^i$. The holomorphic field ${\cal I}_i(v_i,z_i)$ centred at
298: $p_i$ is an operator-valued distribution called an {\it intertwiner}
299: sending (`intertwining') one ${\cal V}$-module (say $M\in
300: \Phi$) to another (say $N\in\Phi$). For fixed $M,N,M^i$,
301: the dimension of the space of intertwiners
302: is called the {\it fusion coefficient} ${\cal N}_{M^i,M}^N$, and is given by
303: Verlinde's formula
304: $${\cal N}_{M^i,M}^N=\sum_{P\in \Phi}{S_{M^iP}S_{MP}\overline{S_{NP}}
305: \over S_{{\cal V}P}}\ ,\eqno(3.6)$$
306: where the matrix $S$ (no relation to the scattering matrix) is defined in
307: (5.1a) below. For example, ${\cal V}$ is always a module
308: for itself and ${\cal N}_{{\cal V},{\cal V}}^{{\cal V}}=1$; the unique
309: (up to scaling) intertwiner ${\cal I}(v,z)$ bijectively associates states
310: $v$ with vertex operators $\varphi(z)={\cal I}(v,z)$ (the so-called
311: `state-field correspondence'). Thus intertwiners generalise vertex operators
312: $\varphi\in{\cal V}$.
313:
314: To solve a given RCFT, the strategy then is to:
315:
316: \smallskip \item{(a)} construct all possible chiral
317: blocks (3.5); and
318:
319: \smallskip\item{(b)} construct the correlation
320: functions (3.3) from those chiral blocks.
321:
322:
323:
324:
325: \bigskip\noindent{\bf 4. The chiral blocks of RCFT.}
326:
327: For a fixed $(g,n;M^1,\ldots,M^n)$, an RCFT assigns
328: a finite-dimensional space ${\frak F}^{(g,n)}_{(M^i)}$
329: of chiral blocks (3.5). Chiral blocks are important to
330: RCFT because finite combinations of them are the correlation functions, and
331: knowing the latter is equivalent to solving the theory.
332:
333:
334:
335: Each chiral block depends multi-linearly
336: on the states $v_i\in M^i$, and holomorphically on the $z_i$, provided branch-cuts
337: in $\Sigma$ between $p_i$ are made; locally, it can be regarded as a
338: holomorphic function on $\widehat{{\cal M}}_{g,n}$. The dimension of this
339: space is given by Verlinde's formula
340: $${\rm dim}\,{\frak F}^{(g,n)}_{(M^i)}=\sum_{P\in\Phi}
341: {S_{M^1 P}\over S_{{\cal V} P}}\cdots {S_{M^n P}\over S_{{\cal V} P}}
342: S_{{\cal V}P}^{2(1-g)}\ ,\eqno(4.1)$$
343: a generalisation of (3.6).
344:
345: Moore and Seiberg [30] -- see also [2] -- isolated
346: the data (finite-dimensional vector spaces and linear transformations)
347: defining each chiral half of RCFT, and provided a complete set of relations they satisfy.
348: Huang is pursuing the explicit construction
349: for all sufficiently nice chiral algebras ${\cal V}$ (see e.g.\ [21]).
350:
351: A basis for ${\frak F}^{(g,n)}_{(M^i)}$ is found by performing the following
352: Feynman rules (called `conformal bootstrap'). Fix a surface $\Sigma$
353: of type $(g,n)$. The space ${\frak F}^{(0,3)}_{(M,N,P)}$ consists of
354: intertwiners; arbitrarily fix bases for all those spaces. Now,
355: dissect $\Sigma$ into pairs-of-pants, as in Figure 1;
356: assign a dummy label $N_j\in\Phi$ to each internal cut in the dissection;
357: to each vertex in your dissection,
358: choose an intertwining operator from the
359: basis of the appropriate space ${\frak F}^{(0,3)}$;
360: `evaluate' the corresponding chiral block -- e.g.\ for
361: each cut, a trace is taken of the product of intertwiners.
362: Repeating, by running through all possible values of the dummy labels,
363: the result is a basis of chiral blocks.
364:
365:
366:
367: \medskip\epsfysize=1in\centerline{ \epsffile{figy1.eps}}\medskip
368: \centerline{{\smcap Figure 1.} Dissecting a surface into pairs-of-pants}\medskip
369:
370:
371: For an important example, let $\Sigma$ be the torus $\C/(\Z+\tau\Z)$ with one
372: puncture $p$ (say at 0),
373: assigned the module $M^1={\cal V}$
374: and state $v_1=|0\rangle$ (the vacuum, the state of lowest energy). One cut
375: suffices to unfold it into a sphere with 3 punctures, assigned ${\cal V}$-modules
376: ${\cal V},M,M\in\Phi$ ($M$ is the dummy label). The fusion coefficient
377: ${\cal N}_{{\cal V},M}^M$ always equals 1, and so for each
378: $M\in\Phi$ there is a unique intertwiner, say ${\cal I}_1^{(M)}$.
379: Hence dim$\,{\frak F}^{(1,1)}_{{\cal V}}=\|\Phi\|$.
380: These Feynman rules yield the {chiral block}
381: $$\chi_M(\tau):={\rm tr}_M e^{2\pi\i\tau\,(L_0-c/24)}\ ,\eqno(4.2)$$
382: where $c$ is the central charge and
383: the Virasoro element $L_0$ corresponds to energy (the trace comes from the
384: dissection). These span ${\frak F}^{(1,0)}$.
385:
386: One of the simplest RCFT is the Ising model, a minimal model. It has central
387: charge $c=\overline{c}=0.5$, and its
388: chiral algebra has 3 irreducible modules, which we'll label $\Phi=
389: \overline{\Phi}=\{{\cal V},\epsilon,\sigma\}$. Its toroidal chiral blocks
390: (4.2) are
391: $$\eqalignno{\chi_{{\cal V}}(\tau)=&\,q^{-1/48}\,(1+q^2+q^3+2q^4+2q^5+3q^6+3q^7+\cdots)
392: \ ,&\cr
393: \chi_\epsilon(\tau)=&\,q^{23/48}\,(1+q+q^2+q^3+2q^4+2q^5+3q^6+3q^7+\cdots)
394: \ ,&(4.3)\cr
395: \chi_\sigma(\tau)=&\,q^{1/24}\,(1+q+q^2+2q^3+2q^4+3q^5+4q^6+5q^7+\cdots)\ ,
396: &}$$
397: where $q=e^{2\pi\i\tau}$.
398:
399: Other important RCFT are the {\it Wess--Zumino--Witten}{} (WZW{})
400: models. These correspond to strings living on a compact Lie group $G$.
401: The chiral algebra ${\cal V}$ is closely
402: related to the affine Kac--Moody algebra ${\frak{g}}^{(1)}$ (see
403: [23]), where
404: ${\frak{g}}$ is the Lie algebra of $G$ (${\frak{g}}^{(1)}$
405: is the nontrivial central extension of the loop algebra ${\frak{g}}
406: \otimes\C[z^{\pm 1}]$).
407: Its modules $M\in\Phi$ can be identified with the integrable highest-weight
408: modules $L(\lambda)$ at a level $k$ determined by the central charge $c$.
409: The chiral blocks $\chi_M(\tau)$ for the WZW models are a specialisation of
410: the corresponding affine algebra ${\frak{g}}^{(1)}$-character $\chi_\lambda(h)$,
411: and for this reason $\chi_M(\tau)$ in any RCFT is called the character of the
412: ${\cal V}$-module $M$. We'll return to the WZW and Ising models shortly.
413:
414:
415: Each dissection produces a basis for the space ${\frak F}^{(g,n)}_{(M^i)}$.
416: However, any $\Sigma$ can be dissected in different
417: ways. The over-used term {\it duality} means here for the
418: invertible matrices
419: relating the chiral blocks of different dissections. For example,
420: the left dissection in Figure 2 of the $(g,n)=(0,4)$ surface
421: corresponds to a matrix
422: $F\left[\matrix{M&N\cr L&P}\right]$ of size $n\times n$ for $n=
423: {\rm dim}\,\frak{F}^{(0,4)}_{(L,M,N,P)}$ (for an appropriate
424: orientation of surfaces and punctures -- a minor technicality we've been
425: ignoring), called the fusing matrix.
426: Likewise, the right dissection defines the braiding
427: matrix $B$.
428:
429:
430: \medskip\epsfysize=.8in \centerline{\epsffile{figy2.eps}}\medskip
431:
432: \centerline{{\smcap Figure 2.} The fusing and braiding matrices $F,B$}\medskip
433:
434:
435: All duality transformations are built up from a few elementary
436: ones, like $B$ and $F$. By decomposing surfaces
437: in different ways, we get relations between these elementary dualities.
438: For example, the $B$-matrices obey an equation of the form $BBB=BBB$ called
439: the Yang--Baxter equation, and this is the source of its name `braiding'.
440:
441: Consider four marked points $w_i$ on the sphere $\C\cup\{\infty\}$.
442: Using the M\"obius(=conformal) symmetry of the sphere,
443: move $w_i$ to $0,w,1,\infty$, respectively, where $w$ is the cross-ratio
444: ${(w_1-w_2)(w_3-w_4)\over (w_1-w_3)(w_2-w_4)}$.
445: If we
446: label all four marked points
447: with the Ising module $\sigma\in\Phi$,
448: then the space ${\frak F}^{(0,4)}_{(\sigma\sigma
449: \sigma\sigma)}$ of chiral blocks is two-dimensional;
450: choosing the smallest energy state in $\sigma$, the chiral blocks are spanned by
451: $$\eqalignno{{\cal F}_1(w)=&\,{\sqrt{1+\sqrt{1-w}}\over (w(1-w))^{1/8}}
452: \ ,&(4.4)\cr
453: {\cal F}_2(w)=&\,{\sqrt{1-\sqrt{1-w}}\over (w(1-w))^{1/8}}
454: \ .&}$$
455: Fusing interchanges $w_1=0$ and $w_3=1$, hence involves the M\"obius transformation
456: $w\mapsto (1-w)/(1-0)=1-w$. Likewise, braiding interchanges $w_2$ with $w_3$,
457: and so involves $w\mapsto (0-1)/(0-w)=1/w$.
458: The braiding and fusing matrices here become
459: $$\eqalignno{B\left[\matrix{\sigma&\sigma\cr\sigma&\sigma}\right]&\,
460: ={e^{-\pi \i/8}\over \sqrt{2}}\left(\matrix{1&\i\cr \i&1}\right)\ ,&(4.5)\cr
461: F\left[\matrix{\sigma&\sigma\cr\sigma&\sigma}\right]&\,
462: ={1\over \sqrt{2}}\left(\matrix{1&1\cr 1&-1}\right)\ .&}$$
463:
464:
465:
466: \noindent{{\bf 5. Monodromy in RCFT.}}
467:
468: The fractional powers in the Ising blocks (4.4) tell us they
469: have branch-point singularities -- we must make cuts in the $w$-plane
470: to get holomorphic functions there. If instead
471: we analytically continue these functions along a closed curve, the value
472: of the block need not return to the same value. For example,
473: consider the circle
474: $w(t)=r\,e^{2\pi\i t}$ for $r$ small: the value of
475: ${\cal F}_i(w)$ at $t=1$ is $e^{-\pi\i/4}$ times its value at $t=0$.
476: This factor $e^{-\pi\i/4}I$ is the {monodromy} about $w=0$. Likewise,
477: their monodromy about $w=1$ is
478: $$\left(\matrix{{\cal F}_1(w)\cr{\cal F}_2(w)}\right)\mapsto\left(\matrix{
479: 0&e^{-2\pi\i/8}\cr e^{-2\pi\i/8}&0}\right)\left(\matrix{{\cal F}_1(w)\cr
480: {\cal F}_2(w)}\right)\ .$$
481:
482: Reintroducing the four
483: coordinates $w_i$, the chiral blocks ${\cal F}_i$ will be
484: holomorphic on the universal cover
485: of the configuration space $\frak{C}_4(S^2)=\{(w_1,w_2,w_3,w_4)\in(S^2)^4\,|\,
486: w_i\ne w_j\}$ of the Riemann sphere. Analytically
487: continuing along any closed path $\gamma$ in $\frak{C}_4(S^2)$
488: defines an action of the fundamental group $\pi_1(\frak{C}_4(S^2))$
489: -- the pure braid group of the sphere with four strands --
490: on the space ${\frak F}^{(0,4)}_{(\sigma\sigma\sigma\sigma)}$ of chiral blocks.
491: For example, the monodromy about $w=1$ found above
492: corresponds to the pure braid $\sigma_{1w}^2$, where $\sigma_{1w}$ is the
493: twist of the 1-strand with the $w$-strand (each strand corresponds to one
494: of the points $0,1,w,\infty$). Actually, the full spherical braid group
495: ${\frak B}_4(S^2)$ acts: $\beta\in{\frak B}_4(S^2)$ maps the space
496: ${\frak F}_{(M^1,M^2,M^3,M^4)}^{(0,4)}$ to ${\frak F}^{(0,4)}_{(M^{\beta 1},
497: M^{\beta 2},M^{\beta 3},M^{\beta 4})}$, where $\beta i$
498: is the associated permutation. For example, the twist $\sigma_{1w}\in{\frak B}_4(S^2)$
499: is the braiding matrix $B\left[\matrix{\sigma&\sigma\cr\sigma&\sigma}\right]$.
500:
501: Equivalently, as a
502: `function' on the configuration space, the chiral blocks form
503: holomorphic horizontal sections of a projectively flat vector bundle. What this means is
504: that each chiral block satisfies a system of partial differential equations
505: (the {\it Knizhnik--Zamolodchikov} or
506: KZ equations) describing how to parallel-transport it around
507: configuration space, and flatness says it {\it locally} depends only
508: on the moduli space parameters (and not on the path chosen). {\it Globally},
509: however, there will be monodromy.
510:
511:
512: More generally, a chiral block ${\cal F}$ for an enhanced surface $\Sigma$
513: is a multi-valued function on the appropriate moduli space.
514: To make it well-defined,
515: ${\cal F}$ can be lifted to the corresponding Teichm\"uller space. There
516: will be an action of the corresponding mapping class group $\widehat{\Gamma}_{g,n}$, coming from
517: monodromy. In other words, the space ${\frak F}_{(M^i)}^{(g,n)}$ of chiral
518: blocks carries a representation $\rho^{(g,n)}_{(M^i)}$ of $\widehat{\Gamma}_{g,n}$.
519: This $\widehat{\Gamma}_{g,n}$-representation is built up from the duality
520: matrices, such as the braiding and fusing matrices. As we shall see below,
521: this picture unifies the Jones knot polynomial, the modularity of Monstrous
522: Moonshine, and many other phenomena.
523:
524: For example, we can dissect the torus using a single vertical cross-sectional
525: cut, or using a
526: horizontal equatorial cut; one basis is $\chi_M(\tau)$ and the other is
527: $\chi_M(-1/\tau)$. Duality says that they both span the same space:
528: $$\chi_M(-1/\tau)=\sum_{N\in\Phi} S_{MN}\,\chi_N(\tau)\ .\eqno(5.1a)$$
529: Likewise, performing a Dehn twist about the vertical cut, we obtain
530: $$\chi_M(\tau+1)=\sum_{N\in\Phi} T_{MN}\,\chi_N(\tau)\ .\eqno(5.1a)$$
531: Here, $S,T$ are complex matrices. Together, $\tau\mapsto-1/\tau$ and
532: $\tau\mapsto\tau+1$ generate the modular group PSL$_2(\Z)$, and
533: $S,T$ generate a true representation $\rho^{(1,0)}$ of the central extension
534: SL$_2(\Z)=\Gamma_{1,0}$.
535: Hence RCFT characters $\chi_M$ form a weight-0 vector-valued modular form for
536: SL$_2(\Z)$, with multiplier $\rho^{(1,0)}$.
537:
538:
539:
540: For example, the matrix $T$ for WZW models
541: involves the quadratic Casimir of $G$, while the matrix $S$
542: involves characters of $G$ evaluated at elements of finite order.
543: For the Ising model, these matrices are
544: $$S={1\over 2}\left(\matrix{1&1&\sqrt{2}\cr 1&1&-\sqrt{2}\cr
545: \sqrt{2}&-\sqrt{2}&0}\right)\, ,\
546: T=\left(\matrix{e^{-\pi\i/24}&0&0\cr 0&-e^{-\pi\i/24}&0\cr 0&0&e^{\pi\i/12}}
547: \right)\eqno(5.2) $$
548:
549:
550: Perhaps the most elegant treatment of the finite-dimensional representations
551: of a compact Lie group $G$ is Borel--Weil theory, which constructs them
552: via the $G$-action on line bundles over the flag manifold
553: $G_\C/B$. Something similar
554: happens to the Virasoro algebra $\frak{Vir}$, with now the moduli spaces of
555: curves playing the role of the flag manifold and mapping class groups taking
556: the place of the Weyl group. A copy
557: of $\frak{Witt}=Vect(S^1)$ attached to the $i$th puncture on an enhanced
558: surface of type $(g,n)$ acts naturally on the moduli
559: space $\widehat{{\frak M}}_{g,n}$:
560: the vector field $z_i^\ell\partial/\partial z_i$ for $\ell\ge 1$ changes
561: the local coordinate $z_i$; $\partial/\partial z_i$ moves the puncture; and
562: $z_i^\ell\partial/\partial z_i$ for $\ell\le -1$ can change the
563: conformal structure of the surface. This infinitesimal action fills out the
564: tangent space to any point on $\widehat{{\frak M}}_{g,n}$. In this picture,
565: the central extension of $\frak{Witt}$ to $\frak{Vir}$ arises geometrically
566: as a curvature effect. The KZ equations say roughly that the desired
567: sections of $\widehat{{\frak M}}_{g,n}$-vector bundles should respect this
568: $\frak{Vir}$ action.
569:
570:
571:
572:
573: \bigskip\noindent{\bf 6. Correlation functions.}
574:
575: Our primary interest is the ${{\Gamma}}_{g,n}$-action $\rho^{(g,n)}_{(M^i)}$
576: on the spaces
577: ${\frak F}_{(M^i)}^{(g,n)}$. Correlation functions -- the quantities of
578: physical interest -- are sesquilinear combinations of chiral blocks which have
579: trivial monodromy. In a typical quantum field theory the correlation functions
580: are computed perturbatively, but in RCFT they can be found exactly.
581:
582: For example, the toroidal {correlation function} is
583: $${\cal Z}(\tau):=\sum_{M\in\Phi,
584: \overline{N}\in\overline{\Phi}} {\cal Z}_{M,\overline{N}}\,
585: \chi_M(\tau)\,\chi_{\overline{N}}(\overline{\tau})\ \eqno(6.1)$$
586: (recall (3.4)). It is required to be invariant under the SL$_2(\Z)$ action
587: (5.1) on the chiral blocks $\chi_M(\tau)$ of the torus. For the special
588: case of the Ising model (recall (5.2)), the unique solution to the various
589: constraints is
590: $${\cal Z}(\tau)=\chi_{{\cal V}}(\tau)\,{\chi_{{\cal V}}(\overline{\tau})}+
591: \chi_\epsilon(\tau)\,{\chi_\epsilon(\overline{\tau})}+
592: \chi_\sigma(\tau)\,{\chi_\sigma(\overline{\tau})}\ .\eqno(6.2)$$
593: Its SL$_2(\Z)$-invariance follows from the unitarity of
594: the matrices (5.2). The classification of possible toroidal correlation
595: functions for WZW models involves quite interesting Lie theory and
596: number theory -- see e.g. [17] for a short proof of
597: Cappelli--Itzykson--Zuber's A-D-E classification of SU$_2(\C)$.
598:
599: The most elegant and general construction of correlation functions from
600: chiral blocks uses topological
601: field theories and the language of category theory [14].
602:
603: \bigskip\noindent{\bf 7. Genus 0: braids and knots.}
604:
605: In the 1980s, Jones studied the combinatorial characterisation of embedding
606: one von Neumann algebra (a factor) in another, and as an unexpected by-product
607: obtained new representations of braid groups ${\cal B}_n$. We can obtain a
608: knotted link from a braid by gluing the $n$ top endpoints of the braid to
609: the corresponding bottom ones. It is possible to characterise completely
610: (using the `Markov moves') the different braids which yield the same knot,
611: and remarkably Jones' ${\cal B}_n$-representations respect this redundancy,
612: in the sense that Jones could obtain from his representations (using the
613: trace in the underlying von Neumann algebra), a polynomial knot invariant [22].
614:
615:
616: Witten [32] reinterpreted Jones' braid group representations as due to the
617: (projective) representation of the genus-0 mapping class groups $\Gamma_{0,n}$,
618: coming from the SU$_2(\C)$ WZW model (as we shall see, $\Gamma_{0,n}$ is
619: essentially a braid group). Witten showed how the $\Gamma_{0,n}$-representation
620: of any other RCFT similarly gives rise
621: to other knot invariants, thus generalising Jones' invariant considerably by embedding it
622: naturally in a much broader context. von Neumann algebras arise in quantum field theory
623: (hence RCFT) through the assignment to each region of space-time
624: of the observables measurable in that region;
625: when one region is a subset of another, then its algebra of
626: observables is embedded in the other.
627:
628: $\Gamma_{0,n}$ has presentation
629: $$\eqalignno{\Gamma_{0,n}=\langle \sigma_1,\ldots,\sigma_{n-1}\,|&\,
630: \sigma_i\sigma_j=\sigma_j\sigma_i\
631: (|i-j|>1),\ \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},&\cr
632: &\sigma_1\cdots
633: \sigma_{n-1}\sigma_{n-1}\cdots\sigma_1=1=(\sigma_1\cdots\sigma_{n-1})^4\rangle
634: &\cr&\cong{\cal B}_n
635: (S^2)/\Z_2\ ,&(7.1)}$$
636: the quotient of the spherical braid group by its centre. RCFT obtains the
637: $\Gamma_{0,n}$-representation $\rho^{(0,n)}_{(M^i)}$
638: by assigning each generator $\sigma_i$ to a matrix in block-diagonal form, whose
639: blocks are braiding matrices. We get a different representation, of dimension given by
640: (4.1), for every choice $M^1,\ldots,M^n$ of ${\cal V}$-modules. An element $\beta\in\Gamma_{0,n}$
641: sends the space ${\frak F}^{(0,n)}_{(M^i)}$ to ${\frak F}^{(0,n)}_{(M^{\beta i})}$,
642: so to get a representation of the full $\Gamma_{0,n}$ we should sum
643: ${\frak F}^{(0,n)}_{(M^i)}$ over all reorderings of $(M^i)$.
644: It is common to lift this $\Gamma_{0,n}$-action to the
645: braid group ${\cal B}_n$ in the manner clear from (7.1). In all known examples
646: it seems, these representations are always defined over some cyclotomic field.
647:
648: The groups $\Gamma_{0,0}\cong\Gamma_{0,1}\cong 1$, $\Gamma_{0,2}\cong\Z_2$, $\Gamma_{0,3}
649: \cong S_3$ are all finite and so aren't very interesting. This is because the M\"obius
650: symmetry on $S^2$ is triply transitive. However, the image of $\Gamma_{0,n}$ for $n\ge 4$
651: will usually be infinite -- e.g. in the special case of two-dimensions,
652: the question of $\Gamma_{0,4}$ having finite image reduces to Schwarz'
653: classical analysis of the finite monodromy of the hypergeometric equation,
654: and as such is very rare.
655:
656: $\Gamma_{0,4}$ is an extension of PSL$_2(\Z)$ by $\Z_2\times\Z_2$, and the part of
657: $\Gamma_{0,4}$ corresponding to a trivial permutation $\beta i=i$ is isomorphic
658: to the principal congruence subgroup $\Gamma(2)/
659: \pm 1$. Using the $\Gamma(2)$-Hauptmodul $\theta_2(\tau)^4/\theta_3(\tau)^4$, we can lift
660: the chiral blocks ${\cal F}(w)$ to the upper-half plane $\H$, and in this way interpret these chiral
661: blocks as vector-valued modular forms for SL$_2(\Z)$. For example, the
662: Ising blocks (4.4) would now become
663: $$\eqalignno{{\cal F}_1(\tau)=&\,q^{-{1\over 16}}(1+q^{{1\over 2}}+3q+4q^{{3\over 2}}+5q^2
664: +8q^{{5\over 2}}+11q^3+\cdots)\ ,&\cr
665: {\cal F}_2(\tau)=&\,q^{{3\over 16}}(2+2q^{{1\over 2}}+2q+4q^{{3\over 2}}+8q^2+10q^{{5\over 2}}+12q^3+\cdots)\ .&(7.2)}$$
666: These lifts ${\cal F}(\tau)$ will always be holomorphic in $\H$, but can have poles
667: at the cusps. The weight can be rational because the $\Gamma(2)$-representation will
668: typically be projective; in fact arbitrary rational weight is possible. The theory of these
669: vector-valued modular forms of arbitrary rational weight for PSL$_2(\Z)$, and no restriction
670: on the kernel of the multiplier, has been developed recently [4], [26] and is quite rich.
671: It is interesting that RCFT produces plenty of examples of these.
672:
673: The $\Gamma_{0,n}$-action coming from WZW models is especially interesting.
674: Consider SU$_2(\C)$ for concreteness.
675: Choose $n$ distinct points $z_1,\ldots,z_n\in\C$ and $n$ $\frak{sl}_2(\C)$-modules
676: $V^i$; write $\widehat{V}^i$ for the corresponding $\frak{sl}_2^{(1)}$-modules. Then
677: the conformal blocks ${\cal F}\in {\frak{F}}^{(0,n)}_{(\hat{V}^i)}$ are
678: precisely the functions ${\cal F}:\frak{C}_n(S^2)\rightarrow V_1\otimes\cdots\otimes
679: V_n$ satisfying the {\it KZ equations} [27]
680: $${\partial {\cal F}\over\partial z_i}={1\over k+2}\sum_{j\ne i}{\Omega_{ij}\over
681: z_i-z_j}\,{\cal F}\ ,\eqno(7.3)$$
682: where
683: $\Omega_{ij}/(z_i-z_j)$ is the classical Yang--Baxter $r$-matrix for SU$_2(\C)$.
684: As mentioned earlier, any solution to (7.3) can be parallel-transported
685: through $\frak{C}_n(S^2)$; projective flatness means that this
686: parallel-transport along a closed loop depends (up to a projective
687: factor) only on the homotopy-class of the loop. In other words,
688: the space ${\frak{F}}^{(0,n)}_{(\hat{V}^i)}$ of solutions to (7.3) carries a
689: projective reprsentation of the pure spherical braid group
690: $\pi_1(\frak{C}_n(S^2))$. The Drinfel'd--Kohno monodromy theorem expresses
691: this monodromy in terms of the $6j$-symbols of the quantum group $U_q(\frak{sl}_2(\C))$,
692: for $q=e^{\pi \i/(k+2)}$, which are straightforward to compute [25]. Something
693: similar happens for any $G$.
694:
695: The infinitely many irreducible finite-dimensional modules of a simple Lie
696: algebra $\frak{g}$ naturally span a symmetric monoidal category
697: (see [31] for definitions); its representation ring is isomorphic to
698: a polynomial ring in $r$ variables, where $r$ is the rank of the algebra. On
699: the other hand, the finitely many level $k$ irreducible integrable
700: modules of the affine algebra $\frak{g}^{(1)}$ span (among other things) a
701: braided monoidal category; the corresponding representation
702: ring is called a {\it fusion ring}{} and has structure constants equal to
703: the fusion coefficients (3.6).
704: The key ingredient in this category -- the braiding -- comes from the
705: braid group monodromy of (7.3). Something similar happens for any RCFT.
706:
707:
708:
709:
710: \bigskip\noindent{\bf 8. Genus 1: modularity.}
711:
712:
713:
714: The characters $\chi_\la$ of the affine algebra $\frak{g}^{(1)}$ are defined
715: exactly as for semi-simple $\frak{g}$, as
716: a sum of exponentials of the Cartan
717: subalgebra, though the sum will now be infinite.
718: In fact a miracle happens: the character $\chi_\la$ will be a
719: modular function for some subgroup of SL$_2(\Z)$! One of the coordinates
720: of the Cartan subalgebra of ${\frak{g}}^{(1)}$ plays the role of $\tau\in\H$,
721: and the others come along for the ride.
722: The algebraic proof of this modularity makes it
723: look accidental: the character $\chi_\la$ is expressed as a fraction;
724: the denominator is automatically a modular form for SL$_2(\Z)$, by the simple
725: combinatorics of affine algebras; the numerator is a modular form (in fact a
726: lattice theta function) for some congruence
727: subgroup of SL$_2(\Z)$, because the Weyl group of ${\frak{g}}^{(1)}$
728: contains translations in a lattice; their quotient yields a modular function.
729:
730: RCFT provides a much more satisfying explanation for this
731: unexpected modularity. The characters $\chi_\la(\tau)$ of these
732: affine algebra modules equal the chiral blocks (4.2) of the corresponding
733: WZW model, and the action (5.1) of $\Gamma_{1,0}={\rm SL}_2(\Z)$ coming from
734: RCFT explains their unexpected
735: modularity. This relation with RCFT also tells us the
736: ${{\frak g}}^{(1)}$-modules are simultaneously ${\frak{Vir}}$-modules.
737: All of this was known to algebraists before the
738: relation of affine algebras to RCFT was developed, but this relation emphasises
739: that these properties of affine algebra modules are not accidental
740: but naturally fit into a much broader perspective.
741:
742:
743: There is more to being a modular form or function
744: than transforming nicely with respect to SL$_2(\Z)$. Good behaviour at the
745: cusps of $\H$ is also crucial, as they compactify the domain. These cusps
746: correspond to a pinched torus; their analogue
747: for the other moduli spaces are surfaces with nodes (this is the Deligne--Mumford
748: compactification). RCFT requires nice behaviour (`factorisation')
749: of chiral blocks as we move in moduli space toward these degenerate
750: surfaces. This connects the moduli spaces of different topologies, and
751: tells us RCFT is naturally defined on a `universal tower' of moduli spaces.
752:
753:
754: Interesting modularity certainly isn't restricted to affine algebras. Indeed,
755: the Monstrous Moonshine conjectures (see e.g. [18]) relate character
756: values of the monster finite simple group to various Hauptmoduls. For example,
757: the first nontrivial coefficient (196884) in the $j$-function nearly
758: equals the dimension 196883 of the first nontrivial representation of the
759: monster. Conjecturally, Hauptmoduls are associated to pairs of commuting
760: elements in the Monster -- e.g. the $j$-function is assigned to $(e,e)$.
761: The starting point to our (still
762: incomplete) understanding of these conjectures is the construction [13]
763: of an RCFT (with central charge $c=24$,
764: $\|\Phi\|=1$, and anti-holomorphic
765: chiral algebra $\overline{{\cal V}}=\C$) whose symmetry group equals the
766: Monster and
767: whose single character (4.2) equals the $j$-function.
768:
769: To some crude extent, Moonshine can then be interpreted as the conjunction of
770: two different pictures of quantum field theory, applied to that very special
771: RCFT: the Hamiltonian picture, which provides us a
772: Hilbert space (state-space) carrying an action of the Monster, and an energy
773: operator $L_0$ such that (4.2) is defined; and the Feynman picture, which
774: lives in moduli space and which makes modularity manifest.
775: As explained at the end of Section 5, the Virasoro algebra, through its
776: action on the moduli spaces $\widehat{\frak{M}}_{g,n}$,
777: lies at the heart of Moonshine.
778:
779:
780:
781: Can we see more directly why the
782: RCFT characters $\chi_M(\tau)$ of (4.2) should have anything to do with
783: modularity? The chiral blocks on the torus
784: can be obtained from those of the plane $\C$, by first
785: considering the map $z\mapsto w:=e^{2\pi\i z}$. Though holomorphic,
786: it changes the global topology,
787: sending the plane $\C$ to the annulus $\C\backslash\{0\}$, and this topology
788: change is responsible for the $-c/24$ in (4.2). To
789: obtain our torus, we now identify $w$ and $qw$, where as always $q=e^{2\pi\i\tau}$.
790: This is equivalent to taking the finite annulus $\{w\in\C\,|\,|q|<|w|<1\}$
791: and sewing together its two boundary circles appropriately.
792: The resulting torus is conformally equivalent to $\C/(\Z+\Z\tau)$.
793: Applying this construction to chiral blocks, we find that those
794: for the torus are indeed given by (4.2) (e.g. the trace comes from sewing).
795: The proof [33] of modularity of vertex operator algebra characters follows
796: this outline.
797:
798: The SL$_2(\Z)$-representation (5.1) is defined over a cyclotomic field for any
799: RCFT [8], and its kernel contains a congruence subgroup [3]. Perhaps the
800: latter isn't so surprising, considering that ${\frak{F}}^{(1,0)}$ has a basis
801: (4.2) with integer $q$-expansions. Intimately connected with this
802: congruence subgroup property, the matrices $S,T$ have nice properties with
803: respect to the cyclotomic Galois group [8,3].
804:
805: $\Gamma_{1,1}$ is also SL$_2(\Z)$, and so the chiral blocks in
806: ${\frak{F}}^{(1,1)}_M$ also form vector-valued modular forms for SL$_2(\Z)$.
807: The weight though can be arbitrary rational numbers and the kernel need
808: not be of finite index. So together with the ${\cal F}\in\frak{F}^{(0,4)}_{(
809: M^i)}$ (as explained last section), RCFT is a rich source of vector-valued
810: modular forms.
811: These $\Gamma_{1,1}$-representations are explicitly known in terms of the
812: duality matrices (see e.g. [30]), and so the machinery of [4] allows these
813: chiral blocks to be explicitly found [5].
814:
815: But specialising to the WZW models, we should expect a nice Lie theoretic
816: answer, and indeed the complete answer is known for SU$_2(\C)$ [24,11].
817: The matrix representing $\tau\mapsto\tau+1$ will again be given by quadratic
818: Casimirs, but the matrix representing $\tau\mapsto -1/\tau$ involves the
819: continuous $q$-ultraspherical polynomials, also
820: known as the Macdonald polynomials for $\frak{sl}_2(\C)$ (and as such are a
821: natural generalisation of SU$_2(\C)$-characters, which describe $\tau\mapsto
822: -1/\tau$ for $\rho^{(1,0)}$). Similarly, the chiral blocks in
823: ${\frak{F}}^{(1,1)}_\lambda$ can be expressed using Macdonald `polynomials'
824: for $\frak{sl}_2^{(1)}$ (a natural generalisation of the
825: $\frak{sl}_2^{(1)}$-characters which are the chiral blocks in
826: $\frak{F}^{(1,0)}$). Something similar can be expected for the other
827: WZW models on the punctured torus.
828: Likewise, evaluating these chiral blocks for the Moonshine RCFT should
829: extend Moonshine to noncommuting pairs $g,h$ in the monster [5].
830:
831: The space $\frak{F}_{{\cal V}}^{(1,1)}$, with the ${\cal V}$-module ${\cal V}$,
832: can be identified with $\frak{F}^{(1,0)}$
833: except that in the former we have the freedom to evaluate the block at
834: any state $v\in{\cal V}$. In particular, taking $v$ to have the minimum energy
835: -- the vacuum -- recovers the characters (4.2), but taking other $v$ will
836: give a vector-modular form of even weight, with the same SL$_2(\Z)$-multiplier
837: $\rho^{(1,0)}$ as in (5.1). [10] evaluated these for the Moonshine RCFT and
838: found that all modular forms (of the right shape) arise.
839: What is interesting is that the coefficients of these modular forms will
840: have Moonshine-like interpretations involving characters of the stabiliser
841: of the state $v$.
842:
843: \bigskip\noindent{\bf 9. Higher genus.}
844:
845: For a fixed RCFT, chiral blocks ${\cal F}\in\frak{F}^{(g,n)}_{(M^1,\ldots,M^n)}$
846: yield vector-valued automorphic functions for the infinite
847: discrete groups $\Gamma_{g,n}$, each realising a finite-dimensional
848: representation $\rho^{(g,n)}_{(M^i)}$ of $\Gamma_{g,n}$. This tower of automorphic functions
849: is coherent in the sense that it respects basic operations like sewing or
850: pinching the surfaces. As mentioned earlier, $(g,n)=(1,0),(1,1),(0,4)$ all
851: give a vector-valued modular form for SL$_2(\Z)$; for $(1,0)$ this is
852: a classical object, being weight-0 and invariant under some
853: congruence subgroup, but for $(1,1)$ and $(0,4)$ the weight is rational
854: and the image of the multiplier will usually be infinite.
855: Relatively little is known in
856: higher genus $g$, and surely it is a direction for important future
857: research. The main open challenge is to identify the special features and
858: structures occurring here. In this sense most of the work done has been negative.
859: In this section we sample a few of the highlights.
860:
861:
862: Most of the work has focused on the kernel and image of these representations
863: $\rho^{(g,n)}_{(M^i)}$. The groups $\Gamma_{0,n}$, $n\le 3$, are finite;
864: all other ker$\,\rho^{(g,n)}_{(M^i)}$ will be infinite, since
865: $\Gamma_{g,n}$ is generated by
866: infinite-order Dehn twists but $\rho^{(g,n)}_{(M^i)}$ maps each of these
867: to a finite-order matrix. However, for fixed $n$, the intersection over
868: all $k$ of the kernel of $\rho^{(g,0)}$ for the SU$_n(\C)$ WZW model
869: at level $k$, is trivial in any genus $g>2$ [1].
870:
871: The image of $\rho^{(1,0)}$ is always finite [3], but this is atypical:
872: it is expected that a generic RCFT will have
873: all other images infinite. For example, Funar [15] found that all
874: im$\,\rho^{(g,0)}$ will be infinite for SU$_2(\C)$ WZW models at all levels
875: $k>8$, and all genus $g>1$. Moreover, Masbaum [29] found an infinite-order matrix in
876: im$\,\rho^{(0,4)}$ for those RCFT.
877:
878: In the RCFT associated to even self-dual lattices $L$ (where the strings live
879: on the torus $\R^n/L$ for $n={\rm dim}\,L$), the conformal blocks in
880: ${\frak{F}}^{(g,0)}$ can be expressed in terms of Siegel theta functions,
881: and the Torelli subgroup of $\Gamma_{g,0}$ is in the kernel of $\rho^{(g,0)}$.
882: This is very atypical for RCFT, e.g. it is known to fail for SU$_2(\C)$ WZW at
883: most levels.
884:
885: On the other hand, these representations for all known RCFT seem to be
886: always definable over a cyclotomic field. A notion of integrality for
887: these representations is being developed [20].
888:
889: A class of RCFT very conducive to study are the so-called holomorphic orbifolds,
890: associated to the Drinfel'd double of a finite group $G$. In this case,
891: the chiral blocks in $\frak{F}^{(g,0)}$ are parametrised by Hom$(\pi_1(\Sigma_g),
892: G)/G$, i.e. group
893: homomorphisms $\varphi$ from $\pi_1$ of a genus-$g$ surface $\Sigma_g$ into $G$,
894: where we identify $\varphi(\sigma)$ and $g^{-1}\varphi(\sigma)g$.
895: $\Gamma_{g,0}$ acts naturally
896: on $\pi_1(\Sigma_g)$ and hence $\rho^{(g,0)}$ here becomes a permutation
897: representation. This means im$\,\rho^{(g,0)}$ here is manifestly finite.
898: However, we can see from this explicitly that ker$\,\rho^{(g,0)}$ won't
899: contain the Torelli generators listed by [6], at least for generic groups
900: $G$, and so even in this extremely well-behaved theory, $\rho^{(g,0)}$
901: doesn't factor through to a representation of Siegel's modular group
902: Sp$_{2g}(\Z)$. [12] show
903: that im$\,\rho^{(0,n)}$ will always be finite here, and it is tempting
904: to guess that all im$\,\rho^{(g,n)}$ is finite here.
905:
906:
907:
908: \bigskip\noindent{{\bf Acknowledgements.}} I've benefitted greatly from
909: communications with Peter B\'antay, Gregor Masbaum, and Mark Walton. This
910: research is supported in part by NSERC.
911:
912: \bigskip\noindent{\bf Bibliography.}\medskip
913:
914: \item{[1]} J. E. Andersen, `Asymptotic faithfulness of the quantum $SU(n)$
915: representations of th mapping class groups', {\it Ann.\ Math.} {\bf 163}
916: (2006) 347--368.
917:
918: \item{[2]} {B.\ Bakalov} and {A.\ Kirillov, Jr.}, {\it Lectures on Tensor
919: Categories and Modular Functors} (American Mathematical Society, Providence 2001).
920:
921:
922: \item{[3]} P. B\'{a}ntay, `The kernel of the modular representation
923: and the Galois action in RCFT', {\it Commun.\ Math.\ Phys.} {\bf 233} (2003)
924: 423--438.
925:
926: \item{[4]} P. B\'{a}ntay and T. Gannon, `Vector-valued modular functions
927: for the modular group and the hypergeometric equation', arXiv:math/0705.2467.
928:
929: \item{[5]} P. B\'{a}ntay and T. Gannon, in preparation.
930:
931:
932: \item{[6]} J. S. Birman, `On Siegel's modular group', {\it Math. Ann.}
933: {\bf 191} (1971) 59--68.
934:
935: \item{[7]} {J.\ S.\ Birman}, {\it Braids, Links, and Mapping Class Groups}
936: (Princeton University Press, Princeton 1974).
937:
938: \item{[8]} {A.\ Coste} and {T.\ Gannon}, `Remarks on
939: Galois in rational conformal field theories', {\it Phys.\ Lett.}\ {\bf B323}
940: (1994) 316--321.
941:
942:
943: \item{[9]} {P.\ Di Francesco, P.\ Mathieu}, and {D.\ S\'en\'echal}, {\it
944: Conformal Field Theory} (Springer, New York 1996).
945:
946:
947: \item{[10]} C. Dong and G. Mason, `Monstrous moonshine at higher weight',
948: {\it Acta Math.} {\bf 185} (2000) 101--121.
949:
950: \item{[11]} P. I. Etingof and A. A. Kirillov Jr, `On the affine analogue of
951: Jack and Macdonald polynomials', {\it Duke Math. J.} {\bf 78} (1995) 229--256.
952:
953: \item{[12]} P. Etingof, E. Rowell and S. Witherspoon, `Braid group reprsentations
954: from twisted quantum doubles of finite groups', math/070327.
955:
956:
957: \item{[13]} {I.\ Frenkel}, {J.\ Lepowsky}, and {A.\ Meurman}, {\it
958: Vertex Operator Algebras and the Monster} (Academic Press, San Diego
959: 1988).
960:
961: \item{[14]} {J.\ Fuchs, I.\ Runkel,} and {C.\ Schweigert},
962: `Boundaries, defects and Frobenius algebras', {\it Fortsch.\ Phys.} {\bf 51}
963: (2003) 850--855.
964:
965:
966: \item{[15]} L. Funar, `On the TQFT representations of the mapping class groups',
967: {\it Pacif. J. Math.} {\bf 188} (1999) 251--274.
968:
969: \item{[16]} {M.\ R.\ Gaberdiel}, `Introduction to conformal field
970: theory', {\it Rep.\ Prog.\ Phys.} {\bf 63} (2000) 607--667.
971:
972: \item{[17]} {T.\ Gannon}, `The Cappelli--Itzykson--Zuber A-D-E classification',
973: {\it Rev.\ Math.\ Phys.} {\bf 12} (2000) 739--748.
974:
975:
976: \item{[18]} T.\ Gannon, {\it Moonshine Beyond the Monster} (Cambridge University
977: Press, 2006).
978:
979:
980: \item{[19]} {K.\ Gawedzki}, `Conformal field theory', S\'eminaire
981: Bourbaki, {\it Ast\'erisque} {\bf 177-178} (1989) 95--126.
982:
983:
984: \item{[20]} P. M. Gilmer and G. Masbaum, `Integral lattices in TQFT', arXiv:
985: math.QA/0411029.
986:
987: \item{[21]} {Y.-Z.\ Huang}, {\it Two-Dimensional Conformal Geometry and
988: Vertex Operator Algebras} (Birkh\"auser, Boston 1997).
989:
990: \item{[22]} V. Jones and V.\ S.\ Sunder, {\it Introduction to Subfactors}
991: (Cambridge University Press, 1997).
992:
993: \item{[23]} {V.\ G.\ Kac}, {\it Infinite Dimensional Lie Algebras},
994: 3rd edn (Cambridge University Press, Cambridge 1990).
995:
996:
997: \item{[24]} A. A. Kirillov Jr, `On an inner product in modular tensor categories',
998: {\it J. Amer. Math. Soc.} {\bf 9} (1996) 1135--1169.
999:
1000: \item{[25]} A.\ N.\ Kirillov and N.\ Yu.\ Reshetikhin, `Representations of the
1001: algebra $U_q(sl(2))$, $q$-orthogonal polynomials and invariants of links',
1002: in: {\it Infinite-dimensional Lie algebras and groups (Luminy-Marseille, 1988)}
1003: (World Scientific, Teaneck NJ 1989) 285--339.
1004:
1005:
1006: \item{[26]} M. Knopp and G. Mason, `Vector-valued modular
1007: forms and Poincar\'{e} series', {\it Illinois J. Math.} {\bf 48} (2004) 1345-1366.
1008:
1009: \item{[27]} T. Kohno, {\it Conformal Field Theory and Topology} (American Math.
1010: Soc., 2002).
1011:
1012: \item{[28]} {J.\ Lepowsky} and {H.\ Li}, {\it Introduction to
1013: Vertex Operator Algebras and Their Representations} (Birkh\"auser, Boston
1014: 2004).
1015:
1016: \item{[29]} G. Masbaum, `An element of infinite order in TQFT-representations
1017: of mapping class groups', {\it Contemp. Math.} {\bf 233} (1999) 137--139.
1018:
1019: \item{[30]} {G.\ Moore} and {N.\ Seiberg}, `Classical and
1020: quantum conformal field theory', {\it Commun.\ Math.\ Phys.} {\bf 123}
1021: (1989) 177--254.
1022:
1023:
1024: \item{[31]} V. G. Turaev, {\it Quantum Invariants of Knots and 3-manifolds}
1025: (de Gruyter, Berlin 1994).
1026:
1027:
1028: \item{[32]} E.\ Witten, `Quantum field theory and the Jones polynomial',
1029: {\it Commun.\ Math.\ Phys.} {\bf 121} (1989) 351--399.
1030:
1031: \item{[33]} {Y.\ Zhu}, `Modular invariance of characters
1032: of vertex operator algebras', {\it J.\ Amer.\ Math.\ Soc.} {\bf 9}
1033: (1996) 237--302.
1034:
1035:
1036: \end
1037: