1: \magnification=\magstep1
2: \overfullrule=0pt \input epsf
3: %\def\eqde{\,{\buildrel \rm def \over =}\,}
4: \def\eqde{\,{:=}} \def\leaderfill{\leaders\hbox to 1em{\hss.\hss}\hfill}
5: \def\al{\alpha} \def\la{\lambda} \def\be{\beta} \def\ga{\gamma}
6: \def\i{{\rm i}} \def\B{{\cal B}}
7: \def\si{\sigma} \def\eps{\epsilon} \def\om{\omega} \def\ka{{\kappa}}
8: \def\N{{\cal N}}
9: \font\huge=cmr10 scaled \magstep2
10: \def\QED{\vrule height6pt width6pt depth0pt}
11: \font\smcap=cmcsc10 \font\smit=cmmi7 \font\smal=cmr7
12: \def\z{{\cal Z}} \def\G{\Gamma} \def\va{\varphi}
13: %%%%% Blackboard bold characters %%%%%
14: \input amssym.def
15: \def\Z{{\Bbb Z}} \def\R{{\Bbb R}} \def\Q{{\Bbb Q}} \def\Qq{{\Bbb Q}}
16: \def\C{{\Bbb C}} \def\Fb{{\Bbb F}} \def\Nn{{\Bbb N}} \def\M{{\Bbb M}}
17: \font\smit=cmmi7 \def\H{{\Bbb H}} \def\g{{\frak g}}
18: \def\sdprod{{\times\!\vrule height5pt depth0pt width0.4pt\,}}
19: \def\boxit#1{\vbox{\hrule\hbox{\vrule{#1}\vrule}\hrule}}
20: \def\splus{\,\,{\boxit{$+$}}\,\,}
21: \def\stimes{\,\,{\boxit{$\times$}}\,\,}
22: \def\sdot{\,{\boxit{$\cdot$}}\,}
23:
24: \centerline{{\huge Monstrous Moonshine: The first twenty-five years}}
25: \bigskip
26:
27: \centerline{{\smcap Terry Gannon}}\medskip
28:
29: \centerline{{Abstract}}
30:
31: \medskip\noindent{{\smal Twenty-five years ago, Conway and Norton
32: published in this journal\footnote{$^\dagger$}{{\smal
33: This is an invited paper for the {\smit Bull}.\ {\smit London}\
34: {\smit Math}.\ {\smit Soc}. Comments are welcome.}}
35: their remarkable paper `Monstrous Moonshine',
36: proposing a completely unexpected relationship between finite simple groups
37: and modular functions. This paper reviews the progress made in broadening and
38: understanding that relationship.}}\footnote{}{{\smal 2000 {\smit Mathematics}\
39: {\smit Subject}\ {\smit Classification}\ 11F22, 17B69, 17B67, 81T40}}
40:
41: \bigskip\centerline{\it 1. Introduction}
42: \medskip
43:
44: \noindent It has been approximately twenty-five years since John McKay remarked that
45: $$196\,884=196\,883+1\ .\eqno(1.1)$$
46: That time has seen the discovery of important structures, the establishment of
47: another deep connection between number theory and algebra, and a reinforcement
48: of a new era of cooperation between pure mathematics and mathematical
49: physics. It is a beautiful and accessible example of how mathematics can be driven by strictly conceptual
50: concerns, and of how the particular and the general can feed off each other.
51: Now, six years after Borcherds' Fields Medal,
52: the original flurry of activity is over; the new period should be one
53: of consolidation and generalisation and should witness the gradual movement of
54: this still rather esoteric corner of mathematics toward the mainstream.
55:
56: The central question McKay's equation (1.1) raises, is: What does the $j$-function
57: (the left side) have to do with the Monster finite group (the right side)?
58: Many would argue that we still don't have our finger on the essence of the matter.
59: But what is clear is that we understand far more about this central question
60: today than we did in 1978. Today we say that there is a vertex operator
61: algebra, called the {\it Moonshine module} $V^\natural$, which interpolates
62: between the left and right sides of (1.1): its automorphism
63: group is the Monster and its graded dimension is the $j$-function ($-744$).
64:
65: This paper tries to summarise this work of the past twenty-five
66: years in about as many pages. The original article [{\bf 24}] is
67: still very readable and contains a wealth of information not found in other
68: sources. Other reviews are [{\bf 21}], [{\bf 87}], [{\bf 12}],
69: [{\bf 39}], [{\bf 90}], [{\bf 48}], [{\bf 15}], [{\bf 78}], [{\bf 102}], [{\bf 16}],
70: [{\bf 46}] and the introductory
71: chapter in [{\bf 44}], and each has its own emphasis. Our own bias here has been to
72: breadth at the expense of depth, which probably limits this review to be a mere annotated
73: sampling of representative literature.
74:
75:
76:
77:
78: \bigskip\centerline{{\it 2. Background}}\medskip
79:
80: In \S2.1 we describe the finite simple groups and in
81: particular the Monster. In \S2.2 we focus on the modular groups and
82: functions which arise in Monstrous Moonshine.
83:
84: \medskip{{\it 2.1. The Monster.}}
85: By definition, a simple group is one whose only normal subgroups are the
86: trivial ones: $\{1\}$ and the group itself. The importance of the
87: finite simple groups lies in their role as building blocks,
88: in the sense that any finite group $G$ can be constructed from $\{1\}$ by
89: extending successively by a (unique up to order) sequence of finite simple
90: groups. For example the symmetric group $S_4$ arises in this way from the
91: cyclic groups $C_2,C_2,C_3,C_2$.
92:
93: A formidable accomplishment last century was determining the explicit list
94: of all finite simple groups. See e.g.\ [{\bf 49}] for more details and references.
95: These groups are:\smallskip
96:
97: \item{(i)} the cyclic groups $C_p$, $p$ prime;
98:
99: \item{(ii)} the alternating groups $A_n$, $n\ge 5$;
100:
101: \item{(iii)} 16 infinite families of groups of Lie type;
102:
103: \item{(iv)} 26 sporadic groups.\smallskip
104:
105: An example of a family of finite simple group of Lie type is PSL$_n({\Bbb F}_q)$, i.e.\
106: the group of $n\times n$ matrices with determinant 1, and entries from the
107: finite field ${\Bbb F}_q$, quotiented out by its centre
108: (the scalar matrices $aI$, where $a^n=1$).
109:
110: The mysteriousness of the sporadics is due to their falling outside those infinite
111: families. They range in size from the Mathieu group $M_{11}$, with order 7920
112: and discovered in 1861, to the Monster ${\Bbb M}$, with order
113: $$|\M|=2^{45}\cdot 3^{20}\cdot 5^9\cdot 7^6\cdot 11^2\cdot 13^3\cdot 17\cdot
114: 19\cdot 23\cdot 29\cdot 31\cdot 41 \cdot 47 \cdot 59 \cdot 71
115: \approx 8\times 10^{53}\ .\eqno(2.1)$$
116: The existence of $\M$ was conjectured in 1973 by Fischer and Griess, and
117: finally constructed in 1980 by Griess [{\bf 50}]. Most sporadics arise in $\M$
118: (e.g.\ as quotients of subgroups). We'll encounter many of these sporadics in the
119: coming pages, but most of our attention will be directed at $\M$.
120:
121: Griess showed in fact that $\M$ was the automorphism group of a 196883-dimensional
122: commutative nonassociative algebra, now called the {\it Griess algebra},
123: but the construction was somewhat artificial. We now understand [{\bf 44}] the
124: Griess algebra
125: as the first nontrivial tier of an infinite-dimensional graded algebra, the {\it Moonshine module}
126: $V^\natural$, which lies at the heart of Monstrous Moonshine. We'll discuss
127: $V^\natural$ in \S4.2; we will find that it
128: has a very rich algebraic structure, is conjectured to obey a strong
129: uniqueness property, and has automorphism group $\M$.
130:
131: The Monster $\M$ has a remarkably simple presentation. As with any noncyclic
132: finite simple group, it is generated by its involutions (i.e.\ elements of
133: order 2) and so will be a homomorphic image of a Coxeter group. Let ${\cal G}_{pqr}$,
134: $p\ge q\ge r\ge 2$, be the graph consisting of three strands of lengths
135: $p+1,q+1,r+1$, sharing a common endpoint. Label the $p+q+r+1$ nodes as in
136: Figure 1. %The Coxeter group ${\cal W}({\cal G})$ associated to any graph
137: %${\cal G}$ is defined to have a generator $g_i$ for each node, obeying the
138: %relations $g_i^2=1$ and $(g_ig_j)^{a_{ij}+1}=1$, where $a_{ij}$ equals the
139: %number of edges between nodes $i$ and $j$.
140: Given any graph ${\cal G}_{pqr}$,
141: define $Y_{pqr}$ to be the group consisting of a generator for each node,
142: obeying the usual Coxeter group relations (i.e.\ all generators are involutions,
143: and the product $gg'$ of two generators has order 3 or 2, depending on whether
144: or not the two nodes are adjacent), together with one more relation:
145: $$(ab_1b_2ac_1c_2ad_1d_2)^{10}=1\ .\eqno(2.2)$$
146: The groups $Y_{pqr}$, for $p\le 5$, have now all been identified (see e.g.\
147: [{\bf 61}]).
148: Conway conjectured and, building on work by Ivanov [{\bf 60}], Norton proved [{\bf 98}]
149: that $Y_{555}\cong Y_{444}$ is the `Bimonster', the wreathed-square ${\Bbb M}
150: \wr C_2$ of the Monster (so has order $2|{\Bbb M}|^2$). A closely related presentation
151: of the Bimonster has 26 involutions as generators and has relations given by
152: the incidence graph of the
153: projective plane of order 3; the Monster itself arises from 21 involutions and
154: the affine plane of order 3 [{\bf 25}]. Likewise, $Y_{553}\cong
155: Y_{443}\cong \M\times C_2$. Other sporadics
156: arise in e.g.\ $Y_{533}\cong Y_{433}$ (the Baby Monster ${\Bbb B}$), $Y_{552}\cong Y_{442}$
157: (the Fischer group $Fi_{24}'$), and $Y_{532}\cong Y_{432}$ (the Fischer group
158: $Fi_{23}$). The Coxeter groups of ${\cal G}_{555}$, ${\cal G}_{553}$, ${\cal G}_{
159: 533}$, ${\cal G}_{ 552}$, and ${\cal G}_{ 532},$ are all infinite groups of
160: hyperbolic reflections in e.g.\ $\R^{17,1}$, and contain
161: copies of groups such as the affine $E_8$ Weyl group, so the geometry
162: here should be quite pretty. What role, if any,
163: these remarkable presentations have in Moonshine hasn't been established yet.
164: As a first step though, [{\bf 97}] identifies in Aut$(V^\natural)$ the 21
165: involutions generating $\M$.
166:
167:
168:
169: \medskip\epsfysize=1.5in\centerline{ \epsffile{lms1.eps}}\medskip
170: \centerline{{\bf Figure 1. The graph ${\cal G}_{555}$ presenting the Bimonster}}\medskip
171:
172: The Monster has 194 conjugacy classes, and so that number of irreducible
173: representations. Its character table (and much other useful information)
174: is given in the Atlas [{\bf 22}], where we also find analogous data for
175: the other simple groups of `small' order. For example,
176: we find that $\M$ has exactly 2, 3, 4 conjugacy classes of elements of
177: order 2, 3, 4, respectively --- these classes are named 2A, 2B, 3A, etc.
178: We also find that the dimensions of the smallest irreducible representations
179: of $\M$ are 1, 196883, 21296876, and 842609326. This is the same 196883 as
180: on the right side of (1.1), and as the dimension of the Griess algebra.
181:
182:
183:
184:
185:
186:
187:
188:
189: \medskip{{\it 2.2. The $j$-function.}}
190: The group SL$_2(\R)$, consisting of $2\times 2$ matrices of determinant 1 with
191: real entries, acts on the upper half-plane $\H:=\{\tau\in\C\,|\,{\rm Im}(\tau)>0\}$
192: by fractional linear transformations
193: $$\left(\matrix{a&b\cr c&d}\right).\tau={a\tau+b\over c\tau+d}\ .\eqno(2.3)$$
194: Of course this is really an action of PSL$_2(\R):={\rm SL}_2(\R)/\{\pm I\}$
195: on $\H$, but it is more convenient to work with SL$_2(\R)$. $\H$ is the
196: hyperbolic plane, one of the three possible geometries in two dimensions (the others
197: are the sphere and the Euclidean plane), and PSL$_2(\R)$ is its group of
198: orientation-preserving isometries.
199:
200: Let $G$ be a discrete subgroup of SL$_2(\R)$. Then the space $G\backslash\H$
201: has a natural structure of an orientable surface, and inherits a complex
202: structure from $\H$ (so can be regarded as a complex curve). By the {\it genus}
203: of the group $G$, we mean the genus of the resulting real surface $G\backslash\H$.
204: For example, the choice $G={\rm SL}_2(\Z)$ yields
205: the sphere with one puncture, so SL$_2(\Z)$ has genus 0. Moreover, any curve
206: $\Sigma$ with genus $g$ and $n$ punctures, for $3g+n>3$, is equivalent as a
207: complex curve to the space $G\backslash\H$, for some subgroup $G$ of SL$_2(\R)$
208: isomorphic to the fundamental group $\pi_1(\Sigma)$.
209:
210: The most important choice for $G$ is SL$_2(\Z)$, thanks
211: to its interpretation as the modular group of the torus. Most groups $G$ of
212: interest are commensurable with SL$_2(\Z)$, i.e.\ $G\cap{\rm SL}_2(\Z)$
213: has finite index in both $G$ and SL$_2(\Z)$. Examples of these
214: are the congruence subgroups
215: $$\eqalignno{\Gamma(N):=&\,\{\left(\matrix{a&b\cr c&d}\right)\in{\rm SL}_2(\Z)\,|\,
216: \left(\matrix{a&b\cr c&d}\right)\equiv \left(\matrix{1&0\cr 0&1}\right)\
217: ({\rm mod}\ N)\}\ ,&(2.4a)\cr
218: \Gamma_0(N):=&\,\{\left(\matrix{a&b\cr c&d}\right)\in{\rm SL}_2(\Z)\,|\,
219: N\ {\rm divides}\ c\}\ .&(2.4b)}$$
220: For example $\Gamma_0(N)$ has genus 0 for $N=2,13,25$, while $N=50$ has genus 2 and
221: $N=24$ has genus 3.
222: The following definition includes all groups arising in Monstrous Moonshine.
223:
224: \medskip{{\smcap Definition 1.}} {\it Call a discrete subgroup $G$ of
225: SL$_2(\R)$ a} moonshine-type modular group, {\it if it contains some $\Gamma_0(N)$, and
226: also obeys the condition that}
227: $$\left(\matrix{1&t\cr 0&1}\right)\in G\ {\rm iff}\ t\in\Z\ .$$
228:
229: Such a modular group is necessarily commensurable with SL$_2(\Z)$. Note that
230: for such a $G$, any meromorphic function $f:G\backslash\H\rightarrow\C$ will
231: have a Fourier expansion of the form $f(\tau)=\sum_{n=-\infty}^\infty a_nq^n$,
232: where $q=e^{2\pi\i\tau}$.
233:
234: \medskip{{\smcap Definition 2.}} {\it Let $G$ be any subgroup of
235: SL$_2(\R)$ commensurable with SL$_2(\Z)$. By a} modular function $f$ for $G$
236: {\it we mean any meromorphic function $f:\H\rightarrow\C$, such that}
237: $$f({a\tau+b\over c\tau+d})=f(\tau)\qquad \forall\left(\matrix{a&b\cr c&d}
238: \right)\in G\eqno(2.5)$$
239: {\it and such that, for any $A\in{\rm SL}_2(\Z)$, the function $f(A.\tau)$ has
240: Fourier expansion of the form $\sum_{n=-\infty}^\infty b_n q^{n/N}$ for some
241: $N$ and $b_n$ (both depending on $A$), and where $b_n=0$ for all but finitely
242: many negative $n$.}\medskip
243:
244: This definition simply states that $f$ is a meromorphic function
245: on the {\it compact} surface $\Sigma_G:=
246: G\backslash\overline{\H}$, where $\overline{\H}:=\H\cup\Q\cup\{\i\infty\}$.
247: The $G$-orbits of $\Q\cup\{\i\infty\}$ are called {\it cusps}; their role
248: is to fill in the punctures of $G\backslash\H$, compactifying the surface, as
249: there are much fewer meromorphic functions on compact surfaces than
250: on noncompact ones (compare the Riemann sphere to the complex plane!).
251:
252: We are especially interested in genus 0 groups $G$ of moonshine-type. Their modular
253: functions are particularly easy to characterise: there will be a unique modular
254: function $J_G$ for $G$, with $q$-expansion of the form
255: $$J_G(\tau)=q^{-1}+\sum_{n=1}^\infty a_n q^n\ ;\eqno(2.6)$$
256: the modular functions for $G$ are precisely the rational functions
257: $f(\tau)={{\rm poly}(J_G(\tau))\over {\rm poly}(J_G(\tau))}$ in $J_G$.
258: This function $J_G$ is called the (normalised) {\it Hauptmodul} for the genus 0 group $G$.
259: For example, the modular group SL$_2(\Z)$ has Hauptmodul
260: $$J_{{\rm SL}_2(\Z)}(\tau)=J(\tau)=q^{-1}+196884\,q+214\,93760\,q^2+
261: 8642\,909970\,q^3+\cdots\ .\eqno(2.7)$$
262: This 196884 is the same as that on the left-side of (1.1). Historically, in place
263: of this Hauptmodul was the equivalent
264: $$j(\tau)={(\theta_2(\tau)^8+\theta_3(\tau)^8+\theta_4(\tau)^8)^3\over 8\eta(\tau)^{24}}
265: =J(\tau)+744\ .$$
266:
267: As we know, there are other genus 0 modular groups. For example the Hauptmoduls for
268: $\Gamma_0(2)$, $\Gamma_0(13)$, and $\Gamma_0(25)$, are respectively
269: $$\eqalignno{J_2(\tau)=&\,q^{-1}+276q-2048q^2+ 11202q^3 -49152q^4+ 184024q^5+\cdots\ ,&(2.8a)\cr
270: J_{13}(\tau)=&\,q^{-1}-q+ 2q^2+ q^3 +2q^4 -2q^5 -2q^7 -2q^8 +q^9+\cdots\ ,&(2.8b)\cr
271: J_{25}(\tau)=&\,q^{-1}-q+q^4 +q^6 -q^{11} -q^{14} +q^{21}+q^{24}-q^{26}+\cdots\ .&(2.8c)}$$
272:
273: Thompson [{\bf 115}] proved there are only finitely many modular groups of moonshine-type
274: in each genus. Cummins [{\bf 28}] has found all of these of genus 0 and 1. In particular
275: there are precisely 6486 genus 0 moonshine-type groups. Exactly 616 of these
276: have Hauptmoduls with rational (in fact integral) coefficients, the remainder have
277: cyclotomic integer coefficients. There are some
278: natural equivalences (e.g.\ a Galois action) which collapse this number to
279: 371, 310 of which have integral Hauptmoduls.
280:
281: In genus $>0$, two functions are needed to generate the function field. A
282: complication facing the development of a higher-genus Moonshine is that,
283: unlike the situation in genus 0 considered here, there is no canonical choice
284: for these generators.
285:
286: See e.g.\ [{\bf 92}] for a very readable account of some of the
287: circle of ideas meandering through this subsection. Modular functions are
288: discussed in e.g.\ [{\bf 75}].
289:
290: \bigskip\centerline{{\it 3. The Monstrous Moonshine conjectures}}\medskip
291:
292:
293: The number on
294: the left of (1.1) is the first nontrivial coefficient of the $j$-function, and
295: the numbers on the right are the dimensions of the smallest irreducible
296: representations of the Fischer--Griess Monster ${\Bbb M}$. On the one
297: side we have a modular function; on the other, a sporadic finite
298: simple group. Moonshine is the explanation and generalisation of this
299: unlikely connection.
300:
301: But first, why can't (1.1) merely be a coincidence? This is soon dispelled
302: by comparing the next few coefficients of $J$ with the dimensions of
303: irreducible representations of $\M$:
304: $$\eqalignno{214\,93760=&\,212\,96876+196883+1\ ,&(3.1a)\cr
305: 8642\,99970=&\,8426\,09326+212\,96876+2\cdot 196883+2\cdot 1\ .&(3.1b)}$$
306:
307:
308:
309: \medskip{{\it 3.1. The fundamental conjecture of Conway and Norton.}}
310: The central structure in the attempt to understand equations (1.1) and (3.1)
311: is an infinite-dimensional graded module for the Monster:
312: $$V=V_0\oplus V_1\oplus V_2\oplus V_3\oplus\cdots\ .\eqno(3.2a)$$
313: If we let $\rho_d$ denote the $d$-dimensional
314: irreducible representation of $\M$, then the first few subspaces will be
315: $V_{0}=\rho_1$, $V_1=\{0\}$,
316: $V_2=\rho_1\oplus \rho_{196883}$, and
317: $V_3=\rho_1\oplus\rho_{196883}\oplus\rho_{21296876}$. This module is to have
318: graded dimension
319: $${\rm dim}_V(\tau)=\sum_{n=0}q^n{\rm dim}(V_n)=1+196884q^2+214\,93760q^3+\cdots
320: =q J(\tau)\ .\eqno(3.2b)$$
321: Of course, (3.2b) alone certainly doesn't uniquely determine $V$, but assume
322: for now this $V$ has been found. Thompson [{\bf 114}] suggested studying
323: in addition the graded traces
324: $$T_g(\tau):=q^{-1}\sum_{n=0}^\infty {\rm ch}_{V_n}(g)\,q^n\eqno(3.2c)$$
325: for all $g\in\M$, where the ch$_{V_n}$ are characters. As taking $g=1$ recovers
326: $J$, (3.2c) is
327: a natural twist of (3.2b). The functions $T_g$ are now called the {\it McKay--Thompson
328: series}.
329:
330:
331: Conway and Norton conjectured [{\bf 24}] that for each element $g$ of the Monster $\M$,
332: $T_g$ is the Hauptmodul
333: $$J_{G_g}(\tau)=q^{-1}+\sum_{n=1}^\infty a_n(g)\,q^n\eqno(3.3)$$
334: for a genus 0 subgroup $G_g$ of SL$_2(\R)$.
335: So for each $n$ the coefficient $g\mapsto a_n(g)$ defines a character ch$_{V_n}(g)$ of $\M$.
336: They explicitly identify
337: each of the groups $G_g$; these groups each contain $\G_0(N)$ as a normal subgroup, for
338: some $N$ dividing $o(g)\,{\rm
339: gcd}(24,o(g))$ ($o(g)$ is the order of $g$), and the quotient group $G_g/\Gamma_0(N)$
340: has exponent 2 (or 1).
341:
342: Since $T_g=T_{hgh^{-1}}$ by definition, there are at most 194 distinct McKay--Thompson
343: series.
344: All coefficients $a_n(g)$ are integers (as are in fact most
345: entries of the character table of $\M$).
346: This implies that $T_{g}=T_{h}$ whenever
347: the cyclic subgroups $\langle g\rangle$ and $\langle h\rangle$ are equal.
348: In fact, the total number of distinct McKay--Thompson series
349: $T_g$ arising in Monstrous Moonshine turns out to be only 171.
350: The first 50 coefficients $a_n(g)$ of each $T_g$ are given in [{\bf 91}].
351: Together with the recursions given in
352: \S3.3 below, this allows one to effectively compute arbitrarily many coefficients $a_n(g)$
353: of the Hauptmoduls. It is also this which uniquely defines $V$, up to equivalence,
354: as a graded $\M$-module.
355:
356: For example, there are two different conjugacy classes of
357: order 2 elements. One of these gives the Hauptmodul $J_2$ in (2.8a), while
358: the other corresponds to (3.4) below.
359: Similarly, (2.8b) corresponds to an order 13 element, but $J_{25}$ in (2.8c)
360: doesn't equal any $T_g$. Recall that there are exactly 616 Hauptmoduls of
361: moonshine-type with integer coefficients, so most of these don't arise as
362: $T_g$. Recently [{\bf 23}], a fairly simple characterisation has been found of the
363: groups arising as $G_g$ in Monstrous Moonshine. Their proof of this
364: characterisation is by exhaustion.
365:
366: Conway coined this conjecture {\it Monstrous Moonshine}. The word `moonshine'
367: here is English slang for `insubstantial or unreal', `idle talk or speculation',
368: `an illusive shadow'. It is meant to give the impression that matters here
369: are dimly lit, and that [{\bf 24}] is `distilling information illegally' from
370: the character table of $\M$.
371:
372: Monstrous Moonshine began, unofficially, in 1975 when Andrew Ogg remarked
373: that the list of primes $p$ for which the group
374: $$\Gamma_0(p)+:=\langle \Gamma_0(p),{1\over \sqrt{p}}\left(\matrix{0&-1\cr p&0}
375: \right)\rangle\eqno(3.4)$$
376: has genus 0, is precisely equal to the list of primes $p$ dividing the order
377: of $\M$. Indeed, in the tables of [{\bf 24}] we find that, for each prime $p$
378: dividing $|\M|$,
379: an element $g$ of $\M$ of order $p$ is assigned the group $G_g=\Gamma_0(p)+$.
380:
381: \medskip{{\it 3.2. Lie theory and Moonshine.}}
382: McKay not only noticed (1.1), but also observed that
383: $$j(\tau)^{{1\over 3}}=q^{-{1\over 3}}\,(1+248q+4124q^2+34\,752q^3+\cdots)\ .\eqno(3.5)$$
384: The point is that 248 is the dimension of the defining representation of the $E_8$ Lie
385: group, while $4124=3875+248+1$ and $34\,752=30\,380+3875+2\cdot 248+1$.
386: Incidentally, $j^{{1\over 3}}$ is a generating modular function for the genus-0
387: group $\Gamma(3)$. Thus Moonshine is related somehow to Lie theory.
388:
389: McKay later found independent relationships with Lie theory [{\bf 89}],
390: [{\bf 15}], [{\bf 47}],
391: reminiscent of his famous A-D-E correspondence with finite subgroups of
392: SU$_2(\C)$. As mentioned earlier, $\M$ has two conjugacy classes of involutions.
393: Let $K$ be the smaller one, called `2A' in [{\bf 22}] (the alternative, class
394: `2B', has almost 100 million times more elements). The product of any two
395: elements of $K$ will lie in one of nine conjugacy classes: namely, 1A, 2A, 2B,
396: 3A, 3C, 4A, 4B, 5A, 6A, corresponding respectively to elements of orders
397: 1, 2, 2, 3, 3, 4, 4, 5, 6. It is surprising that,
398: for such a complicated group as $\M$, that list stops at only 6 --- we call
399: $\M$ a {\it 6-transposition group} for this reason (more on this in \S5.2).
400: The punchline: McKay noticed that those nine numbers are precisely the labels
401: of the affine $E_8$ diagram (see Figure 2). Thus we can attach a conjugacy
402: class of $\M$ to each vertex of the $\widehat{E}_8$ diagram.
403: An interpretation of the {\it
404: edges} in the $\widehat{E_8}$ diagram, in terms of $\M$, is unfortunately not
405: known.
406:
407:
408: \medskip\epsfysize=1.5in\centerline{ \epsffile{lms2.eps}}\medskip
409: \centerline{{\bf Figure 2. The affine $E_8$, $F_4$, and $G_2$ diagrams with labels}}\medskip
410:
411:
412: We can't get the affine $E_7$ labels in a similar way, but McKay noticed that
413: an order {two} folding of affine $E_7$ gives the affine $F_4$ diagram, and we
414: can obtain its labels using the Baby Monster ${\Bbb B}$ (the second largest sporadic).
415: In particular, let $K$ now be the smallest conjugacy class of involutions in
416: ${\Bbb B}$ (also labelled `2A' in [{\bf 22}]); the conjugacy classes in $KK$ have
417: orders 1, 2, 2, 3, 4 (${\Bbb B}$ is a 4-transposition group), and these are
418: the labels of $\widehat{F_4}$. Of course we'd prefer $\widehat{E_7}$ to
419: $\widehat{F_4}$, but perhaps that {\it two}-folding has something to do with
420: the fact that an order-{\it two} central extension of ${\Bbb B}$ is the
421: centraliser of an element $g\in\M$ of order {\it two}.
422:
423: Now, the {\it triple}-folding of affine $E_6$ is affine $G_2$. The Monster has three
424: conjugacy classes of order {\it three}. The smallest of these (`3A') has a centraliser
425: which is a {\it triple} cover of the Fischer group $Fi_{24}'.2$. Taking the smallest
426: conjugacy class of involutions in $Fi_{24}'.2$, and multiplying it by itself,
427: gives conjugacy classes with orders 1, 2, 3 (hence $Fi_{24}'.2$ is a 3-transposition
428: group) --- and those not surprisingly are the labels of $\widehat{G_2}$!
429:
430: Although we now understand (3.5) (see \S4.1) and have proven the fundamental
431: Conway--Norton conjecture (see \S\S4.2--4.4), McKay's $\widehat{E_8},\widehat{F_4},
432: \widehat{G_2}$ observations still have no explanation. In [{\bf 47}] these patterns
433: are extended, by relating various simple groups to the $\widehat{E_8}$ diagram
434: with deleted nodes.
435:
436:
437: \medskip{{\it 3.3. Replicable functions.}}
438: There are several other less important conjectures. One which played an important
439: role in ultimately proving the main conjecture involves the {\it replication
440: formulae}. Conway--Norton want to think of the Hauptmoduls $T_g$ as
441: being intimately connected with $\M$; if so, then the group structure
442: of $\M$
443: should somehow directly relate different $T_g$. Considering the
444: power map $g\mapsto g^n$ leads to the following.
445:
446: It was well-known classically that $j(\tau)$ has the property that
447: $j(p\tau)+j({\tau\over p})+j({\tau+1\over p})+\cdots+j({\tau+p-1\over p})$
448: is a polynomial in $j(\tau)$, for any prime $p$ ({\it proof:}
449: it's a modular
450: function for SL$_2(\Z)$, and hence equals a rational function of $j(\tau)$; since its
451: only poles will be at the cusps, the denominator polynomial must be trivial).
452: Hence the same will hold for $J$. More generally, we get
453: $$\sum_{ad=n,0\le b<d}J({a\tau+b\over d})=Q_{n}(J(\tau))\ ,\eqno(3.6a)$$
454: where $Q_{n}$ is the unique polynomial for which
455: $Q_{n}(J(\tau))-q^{-n}$ has a $q$-expansion with only strictly positive
456: powers of $q$. For example, $Q_2(x)=x^2-2a_1$ and $Q_3(x)=x^3-3a_1x-3a_2$,
457: where we write $J(\tau)=\sum_n a_nq^n$. The left side of (3.6a) is really a
458: Hecke operator applied to $J$. These equations (3.6a) can be rewritten into
459: recursions such as
460: $a_4=a_3+(a_1^2-a_1)/2$, or collected together into the remarkable expression
461: (originally due to Zagier)
462: $$p^{-1}\prod_{{m>0\atop n\in\Z}}(1-p^mq^n)^{a_{mn}}=J(z)-J(\tau)\ ,\eqno(3.6b)$$
463: where $p=e^{2\pi \i z}$.
464:
465:
466:
467: Conway and Norton conjectured [{\bf 24}] that these formulas
468: have an analogue for any McKay--Thompson series $T_g$. In particular,
469: (3.6a) becomes
470: $$\sum_{ad=n,0\le b<d}T_{g^a}({a\tau+b\over d})=Q_{n,g}(T_g(\tau))\ ,\eqno(3.7a)$$
471: where $Q_{n,g}$ plays the same role for $T_g$ that $Q_n$ played for $J$.
472: These are called the replication formulae. Again, these yield recursions like
473: $a_4(g)=a_2(g)+(a_1(g)^2-a_1(g^2))/2$, or can be collected into the expression
474: $$p^{-1}\exp\bigl[-\sum_{k>0}\sum_{{m>0\atop
475: n\in\Z}}a_{mn}(g^k){p^{mk}q^{nk}\over k}\bigr]=T_g(z)-T_g(\tau)\ .\eqno(3.7b)$$
476: This looks a lot more complicated than (3.6b), but you can glimpse the Taylor
477: expansion of ln$(1-p^mq^n)$ there and in fact for $g=1$ (3.7b) reduces
478: to (3.6b).
479:
480: Axiomatising (3.7a) leads to Norton's notion of {\it replicable function}
481: [{\bf 96}], [{\bf 1}]. Write $f^{(1)}(\tau)=q^{-1}+\sum_{k=1}^\infty
482: b_k^{(1)}q^k$, and replacing each
483: $T_{g^a}$ in (3.7a) with $f^{(a)}$, use (3.7a) to recursively define each
484: $f^{(n)}$. If each $f^{(n)}$ has a $q$-expansion of the form $f^{(n)}(\tau)=
485: q^{-1}+\sum_{k=1}^\infty b_k^{(n)}q^k$ --- i.e.\ no fractional powers of $q$
486: arise --- then we call $f=f^{(1)}$ replicable. Equation (3.7a) says the McKay--Thompson
487: series are replicable, and [{\bf 30}] proved that the Hauptmodul of any genus 0 modular
488: group of moonshine-type is replicable, provided its coefficients are rational.
489: Conversely, Norton conjectured that any replicable function with rational
490: coefficients is either such a Hauptmodul, or one of the `modular fictions'
491: $f(\tau)=q^{-1}$, $f(\tau)=q^{-1}\pm q$. This conjecture seems difficult and
492: is still open.
493: Incidentally, if the coefficients $b_k^{(1)}$ are irrational, then the definition (3.7a) of
494: replicability should be modified to include Galois automorphisms (see \S8 of
495: [{\bf 29}]). Replication in positive genus is discussed in [{\bf 109}].
496:
497: Replication (3.7a) concerns the power map $g\mapsto g^n$ in $\M$. Can
498: Moonshine see more of the group structure of $\M$? We explored one step in
499: this direction in \S3.2, where McKay modeled products of conjugacy
500: classes using Dynkin diagrams. A different idea is given in \S5.1.
501: It would be very desirable to find other direct connections between the
502: group operation in $\M$ and e.g.\ the McKay--Thompson series.
503:
504:
505:
506: \medskip{{\it 3.4. The Leech lattice and Moonshine.}}
507: The Leech lattice $\Lambda=\Lambda_{24}$ is a 24-dimensional even self-dual lattice [{\bf 26}] which is to
508: lattices much as the $\M$-module $V$ of (3.2a) turns out to be for vertex operator
509: algebras (see \S4.2 below). $\Lambda$ has no vectors of odd norm, no norm-2
510: vectors, and precisely 196560 norm-4
511: vectors --- a number remarkably close to the monstrous 196883. In fact its
512: theta series $\Theta_{\Lambda}(\tau)
513: =\sum_{v\in\Lambda}q^{v\cdot v/2}$, when divided by $\eta(\tau)^{24}$, equals
514: $J(\tau)+24$. Is this another example of Moonshine?
515:
516: Indeed it is. However we have:
517:
518: \medskip{{\smcap Theorem 3.}} {\it Let $L\subset\R^n$ be any $n$-dimensional
519: positive-definite lattice whose norms ${\bf v}\cdot{\bf v}$ are all rational. Let
520: ${\bf t}\in \R^n$ be any vector with finite order in $L$: i.e.\ $m{\bf t}
521: \in L$ for some nonzero $m\in\Z$. Then the theta series}
522: $$\Theta_{L+{\bf t}}(\tau):=\sum_{{\bf v}\in L}e^{\pi\i\tau\,({\bf v}
523: +{\bf t})^2}\ ,$$
524: {\it divided by $\eta(\tau)^n$, is a modular function for some
525: $\Gamma(N)$.}\medskip
526:
527: See e.g.\ Theorem 20 of [{\bf 100}] for a proof of this classical result. If $L$ is in fact
528: an {\it even} lattice (i.e.\ all norms ${\bf v}\cdot{\bf v}$ lie in $2\Z$),
529: we can say more. Let $L^*:=\{{\bf x}\in\R^n\,|\,{\bf x}\cdot L\subseteq
530: \Z\}$ be the dual lattice. It contains $L$ with finite index; write
531: ${\bf t}_i+L$, $i=1,\ldots,M$, for the finitely many cosets in $L^*/L$.
532: Define a column vector $\vec{\chi}_L(\tau)$ with $i$th component
533: $\Theta_{{\bf t}_i+L}(\tau)/\eta(\tau)^n$. Then $\vec{\chi}_L$ forms a {\it vector-valued
534: modular function} for SL$_2(\Z)$: for any $A=\left(\matrix{a&b\cr c&d}\right)
535: \in{\rm SL}_2(\Z)$,
536: $$\vec{\chi}_L({a\tau+b\over c\tau+d})=\rho(A)\,\vec{\chi}_L(\tau)\eqno(3.8)$$
537: for some $M$-dimensional
538: unitary matrix representation $\rho$ of SL$_2(\Z)$. In particular, for the
539: Leech lattice $L=\Lambda$, $M=1$ and we can quickly identify
540: $\Theta_{\Lambda}(\tau)$ in terms of $J(\tau)$. Although the $196560\approx 196884$
541: coincidence is thus trivial
542: to explain, it will turn out to be a very instructive example of Moonshine.
543:
544: The lattices are related to groups through their automorphism groups, which will
545: always be finite for positive-definite lattices. The automorphism group $Co_0:=
546: {\rm Aut}(\Lambda)$
547: of the Leech lattice has order about $8\times 10^{18}$, and
548: is the direct product of $C_2$ with Conway's sporadic group $Co_1$.
549: Several other sporadics are also involved in Aut($\Lambda)$.
550: To each automorphism $\alpha \in {\rm Aut}(\Lambda)$, let $\theta_\alpha$
551: denote the theta series of the sublattice of $\Lambda$ fixed by $\alpha$.
552: [{\bf 24}] also associate to each automorphism $\alpha$ a certain function $\eta_\alpha(\tau)$
553: of the form $\prod_i \eta(a_i\tau)/\prod_j\eta(b_j\tau)$ built out of the
554: Dedekind eta. Both $\theta_\alpha$ and $\eta_\alpha$ are constant on each
555: conjugacy class in Aut($\Lambda)$, of which there are 202. [{\bf 24}] remarks that
556: the ratio $\theta_\alpha/\eta_\alpha$ always seems to equal some McKay--Thompson
557: series $T_{g(\alpha)}$. See also [{\bf 87}].
558:
559: It turns out that this observation isn't quite correct [{\bf 74}]. For each $\alpha\in{\rm
560: Aut}(\Lambda)$, the subgroup of SL$_2(\R)$ which fixes $\theta_\alpha/\eta_\alpha$
561: is indeed genus 0, but for exactly 15 conjugacy classes in Aut$(\Lambda)$,
562: $\theta_\alpha/\eta_\alpha$ is not the Hauptmodul.
563:
564: Similarly, one can ask this for the $E_8$ lattice, whose automorphism group
565: is the Weyl group (of order $\approx 7\times 10^8$) of the $E_8$ Lie group.
566: The automorphisms $\alpha$ of the
567: $E_8$ lattice which yield a Hauptmodul were classified in [{\bf 19}].
568:
569: \bigskip\centerline{{\it 4. Proof of the Monstrous Moonshine conjectures}}\medskip
570:
571:
572: At first glance, the significance of the Moonshine conjectures seems very
573: unlikely: they constitute after all a finite set of very specialised coincidences.
574: The whole point though
575: is to try to understand {\it why} such seemingly incomparable objects as the
576: Monster and the Hauptmoduls can be so related,
577: and to try to extend and apply this understanding to other contexts.
578: Establishing the truth (or falsity) of the conjectures was merely meant
579: as an aid to uncovering the why of Moonshine. Indeed, in achieving this
580: understanding,
581: important new algebraic structures were formulated. We will sketch this
582: theory below.
583:
584: The main Conway--Norton conjecture was attacked almost immediately. Thompson
585: showed [{\bf 113}] (see also [{\bf 103}]) that if $g\mapsto a_n(g)$ is a character for all sufficiently small
586: $n$ (apparently $n\le 1300$ is sufficient), then
587: it will be for all $n$. He also showed that if certain congruence conditions
588: hold for a certain number of $a_n(g)$ (all with $n\le 100$), then all
589: $g\mapsto a_n(g)$ will be {\it virtual} characters (i.e.\ differences of
590: true characters of $\M$).
591: Atkin, Fong, and Smith (see [{\bf 110}] for details) used that to prove on a computer
592: that indeed all $a_n(g)$ were virtual characters (they didn't quite reach
593: $n=1300$ though). But their work doesn't say
594: anything more about the underlying (possibly virtual) representation $V$,
595: other than its existence. Their work plays no role in the following.
596:
597: We want to show that the McKay--Thompson series $T_g(\tau)$
598: of (3.2c) equals the Hauptmodul $J_{G_g}(\tau)$ in (3.3).
599: First, we need to construct the infinite-dimensional module $V$ of $\M$.
600: We discuss this, and the underlying theory of vertex operator algebras, in
601: \S4.2. Borcherds' strategy [{\bf 11}] was to
602: bring in Lie theory, by associating to the module $V$ a `Monster Lie algebra'.
603: This algebra, and the underlying theory of generalised Kac--Moody algebras,
604: is described in \S4.3. In the final subsection we go from the Monster Lie algebra
605: to the replication formulae, and conclude the proof.
606: We begin this section though by explaining the much simpler connection of
607: $E_8$ with $j^{{1\over 3}}$.
608:
609: \medskip{{\it 4.1. $E_8$ and $j^{{1\over 3}}$.}}
610: An explanation for the relation between $E_8$ and $j^{{1\over 3}}$ was found
611: almost immediately, by Kac and Lepowsky [{\bf 65}], [{\bf 76}]: $j^{{1\over 3}}$ is the (normalised)
612: character of a representation of the affine Kac--Moody algebra $E^{(1)}_8$.
613: Given a finite-dimensional simple Lie algebra $\g$, the affine algebra
614: $\g^{(1)}$ is the infinite-dimensional Lie algebra consisting of all Laurent
615: polynomials $\sum_{n=-\infty}^\infty a_nt^n$ where $a_n\in\g$ and $t$ is an
616: indeterminate, together with
617: a central term and derivation $D=-L_0$ (see [{\bf 66}], [{\bf 70}]). Highest weight representations are defined in
618: the usual way. Thanks largely to the fact that the affine Weyl group is a semi-direct
619: product of the additive group $\Z^r$ ($r={\rm rank}(\g)$) with the finite
620: Weyl group, the characters of these representations
621: (especially the `integrable' highest weight ones, which are the direct analogue
622: of the finite-dimensional representations of $\g$) transform nicely with respect
623: to SL$_2(\Z)$. See e.g.\ Chapter 13 of [{\bf 70}] for details. This is probably the
624: single biggest reason Kac--Moody algebras are so well-known.
625:
626: \medskip{{\smcap Theorem 4.}} [{\bf 69}] {\it Let $\g$ be any finite-dimensional Lie algebra,
627: and $\g^{(1)}$ denote the corresponding affine algebra. Let $P_+^k$ denote the finitely
628: many `level $k$ integrable highest weight modules' $L_\lambda$ of $\g^{(1)}$.
629: The $\g^{(1)}$-module $L_\la$ has a natural $\Z$-grading $L_\la=\oplus_{n=0}^\infty(L_\la)_n$
630: into finite-dimensional $\g$-modules $(L_\la)_n$. Let $\chi_\la(\tau)=
631: q^{h_\la-c/24}\sum_{n=0}^\infty {\rm dim}((L_\la)_n) \,q^n$ be the corresponding normalised
632: character (for some appropriate choice of $h_\la-c/24\in\Q$). Then each
633: $\chi_\la$ is a holomorphic function in $\H$, and the vector $\vec{\chi}_k(\tau)$
634: with entry $\chi_\la(\tau)$ for each $\la\in P_+^k$ defines a vector-valued
635: modular function for SL$_2(\Z)$, as in (3.8),
636: for some finite-dimensional unitary representation $\rho$ of SL$_2(\Z)$.}\medskip
637:
638: In fact each character $\chi_\la$ will be a rational function in lattice theta
639: series, and so will be a modular function for some
640: $\Gamma(N)$. It turns out that there is only
641: one level 1 integrable highest-weight representation of $E_8^{(1)}$, and
642: its character equals $j^{{1\over 3}}$. The modularity of $j^{{1\over 3}}$ is
643: thus predicted
644: by Kac--Moody theory, and the fact that the coefficients are dimensions of
645: $E_8$ representations is automatic. We will see a simultaneous generalisation
646: of Theorems 3 and 4 next section.
647:
648: We've already encountered the mysterious normalisation $q^{-1}$ of the McKay--Thompson
649: series, and the $q^{-{1\over 3}}$ of the $E_8^{(1)}$ character, and more
650: generally the $q^{h_\la-c/24}$ of $\chi_\la$. Many explanations have been
651: provided for this pervasive factor. For example, to [{\bf 2}] it is topological
652: in origin, and related to the Atiyah--Singer Index Theorem; a geometric
653: interpretation using determinant line bundles is due to Segal [{\bf 107}]. In
654: quantum
655: physics it's called the {\it conformal anomaly} (a breakdown of manifest
656: conformal symmetry when the classical system is quantised), and is introduced
657: in regularisation as a vacuum energy. Probably the simplest instance of it is the prefactor in the
658: familiar definition $\eta(\tau)=q^{{1\over 24}}\prod_{n=1}^\infty(1-q^n)$
659: for the Dedekind eta: reading through classical proofs for its modularity
660: we find that `${1\over 24}$' here arises through the combination $\zeta(2)/
661: (2\pi)^2$; $\eta$ appears also in physics in the partition function of the bosonic string,
662: and that same `${1\over 24}$' arises there via regularisation as $-\zeta(-1)/2$.
663: The equivalence of these two expressions for ${1\over 24}$ comes from the
664: functional equation of the Riemann zeta.
665: This same zeta value appears famously in the central term of the Virasoro
666: algebra (4.3), and Bloch [{\bf 8}] found other zeta values appearing in other algebras
667: of differential operators, many of which have now been interpreted and
668: generalised (starting with [{\bf 77}])
669: within the vertex operator algebra framework.
670:
671: Although a direct explanation for Monstrous Moonshine using affine algebras
672: has never been found (and certainly isn't expected), the theory of Kac--Moody algebras
673: influenced every stage
674: of the ultimate proof. For this reason we'll briefly sketch their theory.
675: A simple finite-dimensional Lie algebra $\g$ is built out of the 3-dimensional
676: algebra sl$_2$, in a simple way; the Dynkin diagram of $\g$ encodes the
677: exact presentation. In the identical way, Kac--Moody algebras are also built out
678: of copies of sl$_2$ --- the only difference is that the finite-dimensionality
679: constraint (a positive-definiteness condition on the Cartan matrix) is lifted.
680: Their structure is completely analogous to that of the simple Lie algebras:
681: e.g.\ it has a grading by roots into finite-dimensional spaces; it has a
682: triangular decomposition (making Verma modules possible); it has an invariant
683: symmetric bilinear form. See e.g.\ [{\bf 66}], [{\bf 70}] for details. The
684: affine algebras are the class of Kac--Moody algebras
685: especially analogous to the finite-dimensional ones.
686:
687:
688:
689: \medskip{{\it 4.2. The Moonshine module $V^\natural$.}}
690: A vital component of the Monstrous Moonshine conjectures came a few years
691: after [{\bf 24}]. In a deep work, Frenkel--Lepowsky--Meurman [{\bf 43}],
692: [{\bf 44}] constructed a graded infinite-dimensional
693: representation $V^{\natural}$ of $\M$ and conjectured (correctly) that it
694: is the representation $V$ in (3.2a). $V^\natural$ has a very rich algebraic
695: structure: it is in fact a {\it vertex operator algebra!}
696:
697: A vertex operator algebra [{\bf 9}], [{\bf 44}], [{\bf 67}], [{\bf 39}],
698: [{\bf 79}] is an infinite-dimensional vector space
699: $V$ with infinitely many heavily constrained bilinear products $u*_nv$.
700: The name means `algebra of (generalised) vertex operators'; vertex operators
701: are formal differential operators which originally appeared in physics as quantum
702: fields describing the creation and propagation of physical strings (see
703: \S6 below), and were
704: constructed later but independently by Lie theorists (starting with Lepowsky
705: and Wilson)
706: to realise affine Kac--Moody algebras as algebras of differential operators.
707: Because there were vertex operator constructions associated to lattices,
708: affine algebra modules, and string theory, and all of these have connections
709: to modular functions, it was natural to use vertex operators
710: to try to construct the $\M$-module $V$ of (3.2a).
711:
712: The definition of vertex operator algebra (VOA) is too complicated to give in
713: detail here. A VOA is a graded infinite-dimensional vector
714: space ${\cal V}=\oplus_{n=0}^\infty {\cal V}_n$, where each ${\cal V}_n$ is finite-dimensional.
715: To simplify the discussion, we will limit ourselves in this paper to VOAs
716: with one-dimensional
717: ${\cal V}_0$, which is typical of the examples relevant to Moonshine (and conformal
718: field theory).
719: To each vector $v\in {\cal V}$ we assign a {\it vertex operator} $Y(v,z)$, which is a
720: formal power series $Y(v,z)=\sum_{m\in\Z}v_{(m)} z^{-m-1}$, with coefficients
721: $v_{(m)}\in{\rm End}({\cal V})$. The vertex operator is just the generating function
722: for the products: $u*_nv=u_{(n)}(v)$. These products respect the grading ---
723: in particular,
724: $${\cal V}_k*_n{\cal V}_\ell\subseteq {\cal V}_{k+\ell-n-1}\ .\eqno(4.1)$$
725: A key axiom, which collects together all the identities obeyed by the
726: products $u*_nv$, can be written as
727: $$(z-w)^M[Y(u,z),Y(v,w)]=0\qquad\forall u,v\in {\cal V}\ ,\eqno(4.2a)$$
728: for some integer $M$ (depending on $u,v$), where the bracket in (4.2a) means
729: the commutator $Y(u,z)\,Y(v,w)-Y(v,w)\,Y(u,z)$. This strange-looking formula
730: really says that each such commutator is a linear combination of Dirac deltas
731: and their derivatives, all centred at $z=w$ (see e.g.\ Corollary 2.2 in
732: [{\bf 67}]). Equation (4.2a) implies more down-to-earth identities, such as
733: $$(u_\ell v)_n=\sum_{i\ge 0}(-1)^i\left({\ell\atop i}\right)(u_{\ell-i}\circ
734: v_{n+i}-(-1)^\ell v_{\ell+n-i}\circ u_i)\ .\eqno(4.2b)$$
735:
736: There are two distinguished elements in ${\cal V}$: the identity ${\bf 1}\in {\cal V}_0$
737: (so ${\cal V}_0=\C{\bf 1}$) and the {\it conformal vector} $\omega\in {\cal V}_2$.
738: The identity obeys $Y({\bf 1},z)=id$,
739: i.e.\ ${\bf 1}_{(n)}v=\delta_{n,-1}v$. More interesting is the conformal vector:
740: writing $L_n=\omega_{(n+1)}$, the operators $L_n$ are required to form a representation
741: on ${\cal V}$ of the {\it Virasoro algebra}:
742: $$[L_m,L_n]=(m-n)L_{m+n}+\delta_{m,-n}{m^3-m\over 12}c \,id_{{\cal V}}\ ,\eqno(4.3)$$
743: for some number $c\in\R$ (an important numerical invariant of ${\cal V}$) called the
744: {\it rank} or {\it central charge} of the VOA. In addition we
745: require $L_0u=nu$ whenever $u\in {\cal V}_n$, and $L_{-1}$ acts on ${\cal V}$ as a derivation.
746:
747: The appearance here of the Virasoro algebra is fundamental. It is the
748: unique nontrivial central extension of the (polynomial) vector fields on $S^1$,
749: which in turn is the Lie algebra associated to the group Diff($S^1$) of
750: diffeomorphisms of the circle.
751:
752: The notion of a VOA may seem very arbitrary, but as we'll mention in \S6 it
753: is the `chiral algebra' of a conformal field theory.
754: The simplest (and least interesting) special case of a VOA occurs when $M=0$
755: in (4.2a) --- i.e.\ when all vertex operators commute. Then it is not hard to see
756: that, for each choice of $z\ne 0$, ${\cal V}$ would be a commutative associative
757: algebra with unit, whose product is given by $u*_zv:=Y(u,z)v$. A more honest
758: way to motivate VOAs has been suggested by Huang: binary trees can be used
759: to keep track of the brackets in nested products, e.g.\ $a((bc)d)$, and
760: e.g.\ Lie algebras can be easily formulated using this language [{\bf 58}]; in a monumental
761: work [{\bf 59}], Huang `two-dimensionalised' this Lie algebra formulation by replacing
762: binary trees with spheres with tubes, and showed that the result is equivalent
763: to a VOA.
764:
765: A relation between VOAs and Lie algebras also exists at a more elementary level.
766: Placing $\ell=n=0$ in (4.2b), and evaluating it on the right by $w\in {\cal V}$,
767: gives
768: $$(u_0v)_0w=u_0(v_0w)-v_0(u_0w)\ .\eqno(4.4)$$
769: Writing $[xy]$ for $x_0y$, this becomes
770: the Lie algebra Jacobi identity, at least when $[xy]=-[yx]$. There
771: are different ways to obtain from this a true Lie algebra. The simplest is
772: that $V_1$ will be a Lie algebra for that bracket;
773: moreover, the number $\langle u,v\rangle$ defined by
774: $u_1v=\langle u,v\rangle{\bf 1}$ is an invariant bilinear form for this
775: Lie algebra (by (4.2b) with $\ell=0,n=1$). Equation (4.4) tells us each
776: ${\cal V}_n$ is a ${\cal V}_1$-module, and in fact $e^u$ will be an automorphism of ${\cal V}$
777: for any $u\in {\cal V}_1$. In the most common examples, ${\cal V}_1$ will be
778: reductive (i.e.\ a direct sum of simple and abelian Lie algebras).
779:
780: Modules $M$ of ${\cal V}$ can be defined in the obvious way [{\bf 42}],
781: [{\bf 79}] --- e.g.\ for each
782: $u\in {\cal V}$, each $u_{(n)}$ will be in End$(M)$.
783: All (irreducible) modules $M$ come with a $\Z$-grading $M=\oplus_{n=0}^\infty
784: M_n$, where $u_kM_n\subseteq M_{n+\ell-k-1}$ for all $u\in {\cal V}_\ell$,
785: and $L_0x=(n+h)x$ for any $x\in M_n$, where $h$ is some number (the {\it conformal
786: weight}) depending only on $M$. The (normalised) character of $M$ is
787: $$\chi_M(\tau)=q^{-c/24}{\rm Tr}_M q^{L_0}=q^{h-c/24}\sum_{n=0}^\infty {\rm dim}
788: (M_n)\,q^{n}\ .$$
789:
790: It takes a little effort to construct even the simplest examples of VOAs.
791: The best-behaved ones are called {\it rational VOAs} [{\bf 124}] (borrowing on terminology
792: from physics) and have only finitely many irreducible modules. A rational VOA
793: is associated to any even positive-definite lattice $L$, and their modules are
794: in one-to-one correspondence with the cosets $L^*/L$. Another
795: important example: to any affine nontwisted
796: Kac--Moody algebra and choice of positive integer $k$ (the `level'), the highest
797: weight module $L_{k\Lambda_0}$ has a natural VOA structure, and its modules are
798: precisely the affine algebra modules $L_\la$ for each highest weight
799: $\la\in P_+^k$.
800:
801: One of the deepest results in the theory of VOAs is due to Zhu:
802:
803: \medskip{{\smcap Theorem 5.}} [{\bf 124}] {\it Let ${\cal V}$ be a rational VOA. Its
804: characters $\chi_M(\tau)$ are holomorphic in ${\Bbb H}$, and the subspaces $M_n$
805: carry representations for Aut$({\cal V}$). Write $\vec{\chi}_{{\cal V}}(\tau)$
806: for the vector whose components are the characters $\chi_M(\tau)$ of irreducible
807: modules $M$. Then
808: $\vec{\chi}_{{\cal V}}$ is a vector-valued modular function for SL$_2(\Z)$.}\medskip
809:
810: It is believed that the characters $\chi_M(\tau)$ themselves will be modular
811: functions for some $\Gamma(N)$; significant progress towards this
812: was made in [{\bf 5}] (see also [{\bf 71}]).
813: The proof of Zhu's Theorem is much more difficult than that of Theorem 4, which
814: it generalises.
815:
816: The automorphism group Aut$({\cal V})$ is by definition required to fix $\omega$,
817: which is why it respects the grading of ${\cal V}$. Aut$({\cal V})$ is how group theory
818: impinges on VOA theory. Since the automorphism group Aut$({\cal V})$ of a VOA contains
819: $e^{{\cal V}_1}$ as a (normal) subgroup, Aut(${\cal V}$) can be finite only when ${\cal V}_1=0$.
820: Zhu's Theorem tells us that Moonshine (without the genus-0 aspect) will hold
821: between the group Aut$({\cal V})$ and the functions $\chi_M(\tau)$, for
822: any rational VOA.
823:
824: The most famous example of a VOA is the Moonshine module $V^\natural$ of
825: [{\bf 44}]. It is the orbifold of the Leech lattice VOA ${\cal V}_{\Lambda}$
826: by the $\pm 1$-symmetry of $\Lambda$,
827: which means it's the direct sum of two parts: an invariant part $V^\natural_+$
828: and a twisted part $V^\natural_-$ (more on this in \S5.1). The orbifold serves
829: two purposes: it removes the constant
830: term `24' from the graded dimension $J+24$ (hence the subspace $({\cal V}_{\Lambda})_1$)
831: of ${\cal V}_{\Lambda}$; and it enhances the symmetry from the discrete part
832: of Aut$({\cal V}_{\Lambda})$,
833: which is an extension of $Co_0$ by $(C_2)^{24}$, to all of $\M$.
834:
835:
836:
837: A major claim of [{\bf 44}]
838: was that $V^\natural$ is a `natural' structure (hence their notation).
839: Even so, this bipartite structure to $V^\natural$ complicates its study.
840: We have $V^\natural_0=\C{\bf 1}$, as usual, but the Lie algebra ${\cal V}_1=\{0\}$
841: is trivial. For any such VOA, the space ${\cal V}_2$ will be a commutative nonassociative
842: algebra with product $u\times v:=u_1v$ and identity ${1\over 2}\omega$.
843: For the Moonshine VOA $V^\natural$, this can be shown (with effort!) to
844: be the 196883-dimensional {Griess algebra} extended by an identity
845: element. From this, we find the automorphism group of $V^\natural$ to be
846: the Monster $\M$. The only irreducible module for $V^\natural$ is itself
847: --- such a VOA is called {\it holomorphic}.
848: Together with Zhu's Theorem, this implies that its character, namely $J(\tau)$,
849: must be a modular function for SL$_2(\Z)$ (strictly speaking, we only get
850: invariance up to a 1-dimensional character of SL$_2(\Z)$, but it is easy to
851: show that character must be identically 1). We'll see in \S5.1 how to obtain the other
852: McKay--Thompson series from $V^\natural$.
853:
854: Conjecturally, there are 71 holomorphic VOAs with rank $c=24$ [{\bf 106}].
855: Much as the Leech lattice is the unique even self-dual positive-definite lattice of
856: dimension 24 containing no norm-2 vectors [{\bf 26}], the Moonshine module
857: $V^\natural$ is {\it conjecturally} [{\bf 44}] the unique holomorphic VOA with
858: $c=24$ and with trivial ${\cal V}_1$.
859: Thus, just as the Leech lattice is the unique lattice with
860: theta series $\Theta_{\Lambda}$, so (conjecturally) is the Moonshine
861: module the unique holomorphic VOA with (normalised) graded dimension $J$. Proving this is one of the
862: most important (and difficult) challenges in the subject.
863:
864:
865: \medskip{{\it 4.3. The Monster Lie algebra ${\frak m}$}}.
866: To show that all of the McKay--Thompson series $T_g$ are indeed Hauptmoduls,
867: Borcherds needed identities satisfied by their $q$-expansions. He obtained
868: these through a Lie algebra he associated to $V^\natural$. Before discussing
869: it, let's briefly describe Borcherds' generalisation of Kac--Moody algebras
870: [{\bf 10}].
871:
872: A Borcherds--Kac--Moody algebra differs from a Kac--Moody algebra
873: in that it is built up from Heisenberg algebras as well as sl$_2$, and
874: these subalgebras intertwine in more complicated ways. Nevertheless
875: much of
876: the theory for finite-dimensional simple Lie algebras continues to find
877: an analogue in this much more general setting (e.g.\ root-space decomposition,
878: Weyl group, character formula,...). This unexpected fact is the point
879: of Borcherds--Kac--Moody algebras. For reasons of space we avoid giving here the
880: fairly simple definition, but for this and much more see the review articles
881: [{\bf 53}], [{\bf 63}], [{\bf 102}].
882:
883: Their basic structure theorem is that of Kac--Moody algebras. In particular,
884: there is a grading by roots into finite-dimensional spaces (except that the
885: 0-graded piece, corresponding to the Cartan subalgebra, may be infinite-dimensional).
886: They also have a triangularisable decomposition and an invariant symmetric
887: bilinear form. Indeed, these structural properties {\it characterise}
888: Borcherds--Kac--Moody algebras. In this sense
889: Borcherds--Kac--Moody algebras are the ultimate
890: generalisation
891: of simple Lie algebras, in that any further generalisation would lose
892: some basic structural ingredient.
893:
894:
895: In short, Borcherds' algebras strongly resemble the Kac--Moody
896: ones and constitute a natural and nontrivial generalization. The
897: main differences are that they can be generated by copies of
898: the 3-dimensional Heisenberg algebra as well as sl$_2$, and that there can be
899: imaginary simple roots. Borcherds introduced these algebras and developed their
900: theory in order to understand the Monster Lie algebra ${\frak m}$.
901:
902: We want to construct ${\frak m}$ from the Moonshine module $V^{\natural}
903: =V^\natural_0\oplus V^\natural_1\oplus\cdots$. For later convenience,
904: relabel its subspaces $V^i:=V^\natural_{i+1}$.
905: Of course the obvious choice $V^{\natural}_1=V^0$ is 0-dimensional, so we must
906: modify $V^\natural$ first. Let $II_{1,1}$ denote the even self-dual indefinite
907: lattice consisting of all pairs $(m,n)\in\Z^2$ with inner product
908: $(m,n)\cdot(m',n')=mn'+nm'$. Because it is indefinite, the usual construction
909: of a VOA from a lattice will fail here to produce a true VOA, but most properties
910: will be obtained. Call this near-VOA, ${\cal V}_{{1,1}}$.
911:
912:
913: The Monster Lie algebra ${\frak m}$ is a Lie algebra associated to the near-VOA $V^\natural
914: \otimes {\cal V}_{{1,1}}$ --- see [{\bf 11}] for the details.
915: ${\frak m}$ inherits a $II_{1,1}$-grading
916: from ${\cal V}_{{1,1}}$, and this is its root space decomposition: the
917: $(m,n)$ root space
918: is isomorphic (as a vector space) to $V^{mn}$, if $(m,n)\ne (0,0)$;
919: the $(0,0)$ piece is isomorphic to $\R^2$. Structurally,
920: the Monster Lie algebra has a decomposition ${\frak m}=u^+\oplus{\rm gl}_2\oplus
921: u^-$ into a sum of Lie subalgebras, where $u^{\pm}$ are free Lie algebras
922: (see e.g.\ [{\bf 64}]). It inherits the action of $\M$ from $V^\natural$.
923:
924: This construction of ${\frak m}$ may seem indirect; an alternate approach,
925: anticipated in [{\bf 11}] and [{\bf 12}], uses
926: {\it Moonshine cohomology} [{\bf 81}] --- a functor, inspired by BRST
927: cohomology in conformal field theory, assigning
928: to certain $c=2$ near-VOAs some Lie algebra
929: carrying an action of $\M$. To ${\cal V}_{1,1}$ this functor associates ${\frak m}$.
930:
931:
932: \medskip{{\it 4.4. Denominator identities and modular equations.}}
933: It was discovered early on that the
934: Hauptmoduls all obey the replication formulae, and that anything obeying
935: those formulae will be determined by their first few coefficients. The idea
936: then is to show that the McKay--Thompson series $T_g$ of (3.2c) also are
937: replicable. Borcherds did this using Lie algebra denominator identities [{\bf 11}].
938:
939: Finite-dimensional simple Lie algebras ${\frak g}$ possess a very useful formula for their characters,
940: due to Weyl: the (formal) character $\chi_\la$ of a module $L_\la$ equals
941: $$\chi_\la:=\sum_{\mu}{\rm dim}(L_\la(\mu))\,e^{\mu
942: }=e^{-\rho}\,{\sum_{w\in W}\eps(w)\, e^{w(\la+\rho)}\over
943: \prod_{\alpha\in\Delta_+} (1-e^{-\alpha})}\ ,\eqno(4.5)$$
944: where $W$ is the Weyl group, $\Delta_+$ the positive roots, $\eps(w)={\rm det}(w)$
945: is a sign, and where
946: $\oplus_\mu L_\la(\mu)$ is the weight-space decomposition of $L_\la$.
947: As the weights $\mu$ by definition lie in the dual ${\frak h}^*$ of the Cartan
948: subalgebra of ${\frak g}$, the character $\chi_\la$ can be regarded as a
949: complex-valued function on the space ${\frak h}\cong{\Bbb C}^r$ ($r={\rm rank}
950: ({\frak g})$).
951:
952:
953: Consider
954: the trivial representation: i.e.\ $x\mapsto 0$ for all $x\in \g$. Its
955: character $\chi_0$ will be identically 1. Thus the
956: character formula (4.5) tells us that a certain alternating sum over a Weyl group, equals
957: a certain product over positive roots. These formulas, called {\it
958: denominator identities}, are nontrivial even in this finite-dimensional case.
959:
960: In a famous paper [{\bf 83}], Macdonald generalised the denominator
961: identity for (4.5), to infinite sum/product identities,
962: corresponding to the extended Dynkin diagrams. The simplest one was known
963: classically as the Jacobi triple product identity:
964: $$\sum_{n=-\infty}^\infty(-1)^nx^{n^2}y^n=
965: \prod_{m=1}^\infty(1-x^{2m})(1-x^{2m-1}y)(1-x^{2m-1}y^{-1})\ .\eqno(4.6)$$
966: Macdonald's identities were later reinterpreted, by Kac and Moody, as denominator
967: identities for the affine algebras. For example, we
968: now know (4.6) to be the denominator identity for the algebra $A_1^{(1)}$.
969:
970:
971: In particular,
972: the same formula (4.5) holds for Kac--Moody algebras, except that the
973: sum and product are now infinite, the positive roots now come with
974: multiplicities, and the characters are usually normalised by a prefactor
975: $q^{h_\la-c/24}$.
976: The variable $\tau$ in Theorem 4 is one of the coordinates in the
977: Cartan subalgebra ${\Bbb C}^{r+2}$ of the affine algebra (see e.g.\ equation
978: (13.2.4) of [{\bf 66}]). In that theorem we dropped the
979: remaining variable dependence of the $\chi_\lambda$ for readability, although
980: those additional coordinates serve the important role of guaranteeing linear
981: independence of the characters, and of giving us an action of SL$_2(\Z)$
982: rather than merely PSL$_2(\Z)$.
983:
984:
985: Because a Borcherds--Kac--Moody algebra $\g$ is triangularisable, highest weight
986: $\g$-modules can be defined in the usual way from Verma modules.
987: The character formula becomes
988: $$\chi_{\la}=e^{-\rho}{\sum_{w\in W}\epsilon(w)\,w(e^{\la+\rho}S_\lambda)\over
989: \prod_{\alpha\in\Delta_+}(1-e^{-\alpha})^{{\rm mult}\,\alpha}}\ ,\eqno(4.7)$$
990: where $S_\lambda$ is a correction factor due to imaginary simple roots.
991:
992:
993:
994:
995:
996: The corresponding denominator identity of the Monster Lie algebra ${\frak m}$ can be
997: computed, and is given in (3.6b). Its Weyl group is $C_2$ and sends the
998: $(m,n$)-root space to $(n,m$); the $(m,n)$ root has multiplicity given by
999: coefficient $a_{mn}$ of $J$; for each $n>0$ we have an imaginary simple root
1000: $(1,n)$ with multiplicity $a_{n}$. Because of a cohomological interpretation
1001: of all denominator identities, (3.6b) can be `twisted' by each $g\in\M$,
1002: and this gives (3.7b).
1003: These formulas are equivalent to
1004: the replication formula (3.7a) conjectured in \S 3.3.
1005:
1006: Identities equivalent to (3.7b) were obtained by more elementary means --- i.e.\
1007: methods requiring less of the theory of Borcherds--Kac--Moody algebras --- in
1008: [{\bf 64}] and [{\bf 68}], permitting a simplification of Borcherds' proof
1009: at this stage.
1010:
1011: Now, it turns out that if we verify for each conjugacy class $K_g$ of $\M$
1012: that the first, second, third, fourth and sixth coefficients of the
1013: McKay--Thompson series $T_g$ and the corresponding Hauptmodul $J_{G_g}$ agree,
1014: then indeed $T_g=J_{G_g}$.
1015: That is precisely what Borcherds then did: he compared finitely many
1016: coefficients, and as they all equalled what they should, this concluded
1017: the proof [{\bf 11}] of Monstrous Moonshine!\medskip
1018:
1019: However, this case-by-case verification
1020: occurred at the critical point where the McKay--Thompson series
1021: were being compared directly to the Hauptmoduls, and so provides little insight
1022: into why the $T_g$ are genus 0. Fortunately a more conceptual
1023: explanation of their equality has since been found.
1024:
1025: A function $f$ obeying the replication formulae (3.7a) will also obey
1026: {\it modular equations} --- i.e.\ a 2-variable polynomial
1027: identity satisfied by $f(x)$ and $f(nx)$. The simplest examples come from
1028: the exponential and cosine functions: note that for any $n>0$,
1029: $\exp(nx)=(\exp(x))^n$ and $\cos(nx)=T_n(\cos(x))$ where $T_n$ is a
1030: Tchebychev polynomial. It was known classically that $j$ (hence $J$) satisfied
1031: a modular equation for any $n$: e.g.\ put $X=J(\tau)$ and
1032: $Y=J(2\tau)$, then
1033: $$\eqalignno{(X^2-Y)(Y^2-X)=&\,393768\,(X^2+Y^2)+42987520\,XY+40491318744\,(X+Y)
1034: &\cr&\,-120981708338256\ .&\cr}$$
1035:
1036: The only functions $f(\tau)=q^{-1}+a_1q+\cdots$
1037: which obey modular equations for all $n$, are $J(\tau)$ and the `modular
1038: fictions' $q^{-1}$ and $q^{-1}\pm q$ (which are essentially exp, cos, and
1039: sin) [{\bf 72}]. More generally, we have:
1040:
1041: \medskip{\smcap Theorem 6.} [{\bf 29}] {\it A function $B(\tau)
1042: =q^{-1}+\sum_{n=1}^\infty b_nq^n$ which obeys a modular equation for all $n\equiv 1$
1043: (mod $N$), will either be of the form $B(\tau)=q^{-1}+b_1q$, or will be a
1044: Hauptmodul for a modular group of moonshine-type.} \medskip
1045:
1046:
1047: The converse is also true [{\bf 29}]. The denominator identity argument tells us each
1048: $T_g$ obeys a modular equation for each $n\equiv 1$ modulo the order
1049: of $g$, so Theorem 6 then concludes the proof of Monstrous Moonshine.
1050:
1051: The computer searches in [{\bf 20}] suggest that the hypothesis of
1052: Theorem 6 may be considerably weakened, perhaps all the way down to the
1053: existence of modular equations for any two distinct primes.
1054:
1055: \bigskip\centerline{{\it 5. Further developments}}%\medskip
1056:
1057:
1058: \medskip{{\it 5.1. Orbifolds.}} About a third of the McKay--Thompson series $T_g$ will
1059: have some negative coefficients. In \S5.4 we'll see Borcherds interpret them
1060: as dimensions of superspaces (which come with signs).
1061: In an important announcement [{\bf 97}], on par with [{\bf 24}], Norton proposed
1062: that, although $T_g(-1/\tau)$ will not usually be another McKay--Thompson series, it will
1063: always have nonnegative integer $q$-coefficients, and these can be interpreted as
1064: ordinary dimensions. In the process, he extended the $g\mapsto T_g$ assignment
1065: to commuting pairs
1066: $(g,h)\in \M\times\M$.
1067:
1068: In particular, to each such pair we have a
1069: function $N(g,h;\tau)$, which we will call a {\it Norton series}, such that
1070: $$N(g^ah^c,g^bh^d;\tau)=\alpha\,N(g,h;{a\tau+b\over c\tau+d})\qquad
1071: \forall\left(\matrix{a&b\cr c&d}\right)\in {\rm SL}_2(\Z)\ ,\eqno(5.1)$$
1072: for some root of unity
1073: $\alpha$ (of order dividing 24, and depending on $g,h,a,b,c,d$).
1074: The Norton series $N(g,h;\tau)$ is either constant, or generates the modular
1075: functions for a genus-0 subgroup of SL$_2(\Z)$ containing some $\Gamma(N)$
1076: (but otherwise not necessarily of moonshine-type). Constant $N(g,h;\tau)$
1077: arise when all elements of the form $g^ah^b$ (for gcd($a,b)=1$) are
1078: `non-Fricke' (an element
1079: $g\in\M$ is called {\it Fricke} if the group $G_g$ contains an element
1080: sending 0 to $\i\infty$ --- the identity 1 is Fricke, as are 120 of the 171
1081: $G_g$). Each $N(g,h;\tau)$ has a $q^{{1\over N}}$-expansion for that $N$; the
1082: coefficients of this expansion
1083: are characters evaluated at $h$ of some central extension of the centralizer
1084: $C_{\M}(g)$. Simultaneous conjugation of $g,h$ leaves the Norton series
1085: unchanged: $N(aga^{-1},aha^{-1};\tau)=N(g,h;\tau)$.
1086:
1087: For example, when $\langle g,h\rangle\cong C_2\times C_2$ and $g,h,gh$ are
1088: all in class 2A, then $N(g,h;\tau)=\sqrt{J(\tau)-984}$.
1089: The McKay--Thompson series are
1090: recovered by the $g=1$ specialisation: $N(1,h;\tau)=T_h(\tau)$.
1091: This action (5.1) of SL$_2(\Z)$ is related to its natural action
1092: on the fundamental group $\Z^2$ of the torus, as we'll see in \S6, as well as
1093: a natural action of the braid group, as we'll see next subsection.
1094: Norton arrived at his conjecture empirically, by studying the data of Queen
1095: (see \S5.3).
1096:
1097: The basic tool we have for approaching Moonshine conjectures is the theory
1098: of VOAs, so we need to understand Norton's suggestion from that point of
1099: view. For reasons of space, we'll limit this discussion to $V^\natural$, but
1100: it generalises. Given any automorphism $g\in{\rm Aut}(V^\natural)$, we can
1101: define $g$-twisted modules in the obvious way [{\bf 36}]. Then for each $g\in\M$,
1102: there is a unique $g$-twisted module, call it $V^\natural(g)$, for $V^\natural$
1103: --- this statement generalises
1104: the holomorphicity of $V^\natural$ mentioned in \S4.2.
1105: More generally, given any automorphism $h\in{\rm Aut}(V^\natural)$ commuting
1106: with $g$, $h$ will yield an automorphism of $V^\natural(g)$, so we can
1107: perform Thompson's twist (3.2c) and write
1108: $$q^{-c/24}{\rm Tr}_{V^\natural(g)}h\, q^{L_0}=:{\cal Z}(g,h;\tau)\ .\eqno(5.2)$$
1109: These ${\cal Z}(g,h)$'s can be thought of as the building blocks of the graded
1110: dimensions of various eigenspaces in $V^\natural(g)$: e.g.\ if $h$ has
1111: order $m$, then the subspace of $V^\natural(g)$ fixed by automorphism $h$
1112: will have graded dimension $m^{-1}\sum_{i=1}^m{\cal Z}(g,h^i)$.
1113: In the case of the Monster considered here, we have ${\cal Z}(g,h)=N(g,h)$.
1114:
1115: The important paper [{\bf 36}] proves that, whenever the subgroup $\langle g,h\rangle$
1116: generated by $g$ and $h$ is cyclic, then $N(g,h)$ will be a
1117: Hauptmodul satisfying (5.1). One way this will happen of course is
1118: whenever the orders of
1119: $g$ and $h$ are coprime. Extending [{\bf 36}] to all commuting
1120: pairs $g,h$ is one of the most pressing tasks in Moonshine.
1121:
1122: This orbifold construction is the same as was used to construct $V^\natural$
1123: from ${\cal V}_{\Lambda}$: $V^\natural$ is the sum of the `$\iota$'-invariant
1124: subspace $V^\natural_+$ of ${\cal V}_{\Lambda}$ with the `$\iota$'-invariant subspace
1125: $V^\natural_-$ of the unique `$-1$'-twisted module for ${\cal V}_{\Lambda}$,
1126: where $\iota\in{\rm Aut}(\Lambda)$ is some involution. The
1127: graded dimensions of $V^\natural_{\pm}$ are $2^{-1}({\cal Z}(\pm 1,1)+{\cal Z}
1128: (\pm 1,\iota))$, respectively, and these sum to $J$.
1129:
1130: The orbifold construction is also involved in an interesting reformulation
1131: of the Hauptmodul property, due to Tuite [{\bf 116}]. Assume the uniqueness
1132: conjecture: $V^\natural$ is the only VOA with graded dimension $J$. He argues
1133: from this that, for each $g\in\M$, $T_g$ will be a Hauptmodul iff the only
1134: orbifolds of $V^\natural$ are ${\cal V}_{\Lambda}$ and $V^\natural$ itself. In e.g.\ [{\bf
1135: 62}], this analysis is extended to some of Norton's $N(g,h)$'s,
1136: where the subgroup $\langle g,h\rangle$ is not cyclic (thus going beyond [{\bf 36}]),
1137: although again assuming the uniqueness conjecture.
1138:
1139: \medskip{{\it 5.2. Why the Monster?}}
1140: That $\M$ is associated with {\it modular functions} can be explained
1141: by it being the automorphism group of the Moonshine VOA $V^\natural$.
1142: But what is so special about this group $\M$ that
1143: these modular functions $T_g$ and $N(g,h)$ should be Hauptmoduls?
1144: This is still open. One approach is due to
1145: Norton, and was first (rather cryptically) stated in [{\bf 97}]: the Monster is probably the
1146: largest (in a sense) group with the 6-transposition property. Recall from
1147: \S3.2 that a $k$-transposition group $G$ is one generated by a conjugacy class
1148: $K$ of involutions, where the product $gh$ of any two elements of $K$ has order
1149: $\le k$. For example, taking $K$ to be the transpositions in the symmetric
1150: group $G=S_n$, we find that $S_n$ is 3-transposition.
1151:
1152: A transitive
1153: action of $\Gamma:={\rm PSL}_2(\Z)$ on a finite set $X$ with one distinguished
1154: point $x_0\in X$,
1155: is equivalent to specifying a finite index subgroup $\Gamma_0$ of $\Gamma$.
1156: In particular, $\Gamma_0$ is the stabiliser $\{g\in\Gamma\,|\,g.x_0=x_0\}$
1157: of $x_0$, $X$ can be identified with the cosets $\Gamma_0\backslash \Gamma$,
1158: and $x_0$ with the coset $\Gamma_0$. (If we avoid specifying $x_0$, then
1159: $\Gamma_0$ will be identified only up to conjugation.)
1160:
1161: To such an action, we can associate an interesting triangulation of the closed surface
1162: $\Gamma_0\backslash\overline{\H}$, called a (modular) {\it quilt}. The definition,
1163: originally due to Norton and further developed by Parker, Conway, and Hsu,
1164: is somewhat involved and will be avoided here (but see especially Chapter 3
1165: of [{\bf 57}]). It is so-named because there is a polygonal `patch' covering every
1166: cusp of $\Gamma_0\backslash\H$, and the closed surface is formed by sewing together
1167: the patches along their edges (`seams'). There are a total of $2n$ triangles
1168: and $n$ seams in the triangulation, where $n$ is the index $\|\Gamma_0\backslash
1169: \Gamma\|=\|X\|$. The boundary of each patch has an even number of edges,
1170: namely the double of the corresponding cusp width. The familiar formula
1171: $$\gamma={n\over 12}-{n_2\over 4}-{n_3\over 3}-{n_\infty\over 2}+1$$
1172: for the genus $\gamma$ of $\Gamma_0\backslash\H$ in terms of the index $n$ and the
1173: numbers $n_i$ of $\Gamma_0$-orbits of fixed-points of order $i$, can be
1174: interpreted in terms of the data of the quilt (see (6.2.3) of [{\bf 57}]),
1175: and we find in particular that if every patch of the quilt has at most 6
1176: sides, then the genus will be 0 or 1, and genus 1 only exceptionally.
1177:
1178: In particular, we're interested in one class of these $\Gamma$-actions (actually
1179: an SL$_2(\Z)$-action, but this doesn't matter). Recall that the braid group
1180: $B_3$ has presentation
1181: $$\langle \sigma_1,\sigma_2\,|\,\sigma_1\sigma_2\sigma_1=\sigma_2\sigma_1\sigma_2
1182: \rangle\ ,\eqno(5.3a)$$
1183: and centre ${Z}=\langle (\sigma_1\si_2\si_1)^2\rangle$ [{\bf 7}]. It
1184: is related to the modular group by
1185: $$B_3/{Z}\cong {\rm PSL}_2(\Z)\ ,\quad B_3/\langle (\sigma_1\si_2\si_1)^4
1186: \rangle\cong {\rm SL}_2(\Z)\ .\eqno(5.3b)$$
1187:
1188: Fix a finite group $G$ (we're most interested in the choice $G=\M$).
1189: We can define a right action of $B_3$ on triples $(g_1,g_2,g_3)\in G^3$ by
1190: $$(g_1,g_2,g_3)\si_1=(g_1g_2g_1^{-1},g_1,g_3)\ ,\qquad(g_1,g_2,g_3)\si_2=
1191: (g_1,g_2g_3 g_2^{-1},g_2)\ .\eqno(5.4a)$$
1192: We will be interested in this action on the subset of $G^3$ where all
1193: $g_i\in G$ are involutions. The action (5.4a) is equivalent to a reduced version, where
1194: we replace $(g_1,g_2,g_3)$ with $(g_1g_2,g_2g_3)\in G^2$. Then (5.4a) becomes
1195: $$(g,h)\si_1=(g,gh)\ ,\qquad(g,h)\si_2=(gh^{-1},h)\ .\eqno(5.4b)$$
1196: These $B_3$ actions come from specialisations of the Burau and reduced Burau
1197: representations [{\bf 7}], respectively, and generalise to actions of
1198: $B_n$ on $G^n$ and $G^{n-1}$. We can get an action of SL$_2(\Z)$ from the
1199: $B_3$ action (5.4b) in two ways: either
1200:
1201: \smallskip\item{(i)} by restricting to commuting pairs $g,h$; or
1202:
1203: \smallskip\item{(ii)} by identifying
1204: each pair $(g,h)$ with all its conjugates $(aga^{-1}, aha^{-1})$.
1205:
1206: \smallskip\noindent Norton's
1207: SL$_2(\Z)$ action of \S5.1 arises from the $B_3$ action (5.4b), when we
1208: perform both (i) and (ii).
1209:
1210: The quilt picture was designed for this SL$_2(\Z)$ action.
1211: The point of this construction is that the number of sides in each patch
1212: is determined by the orders of the corresponding elements $g,h$.
1213: If $G$ is say a
1214: 6-transposition group (such as the Monster), and we take the involutions
1215: $g_i$ from 2A, then each patch will have $\le 6$ sides, and the corresponding
1216: genus will be 0 (usually) or 1 (exceptionally if at all). In this way we can relate
1217: the Monster with a genus-0 property.
1218:
1219: Based on the actions (5.4), Norton anticipates some analogue of Moonshine
1220: valid for noncommuting pairs. CFT considerations (`higher genus orbifolds')
1221: alluded to in \S6 suggest that more natural should be
1222: e.g.\ quadruples $(g,g',h,h')\in\M^4$ obeying $ghg^{-1}h^{-1}=h'g'h'{}^{-1}
1223: g'{}^{-1}$.
1224:
1225: \smallskip
1226: An interesting question is, how much does Monstrous Moonshine determine the
1227: Monster? How much of $\M$'s structure can be deduced from e.g.\ McKay's
1228: $\widehat{E}_8$ Dynkin diagram observation, and/or the (complete) replicability
1229: of the $T_g$, and/or Norton's conjectures in \S5.1, and/or Modular Moonshine
1230: in \S5.4 below? A small start toward this is taken in
1231: [{\bf 99}], where some control on the subgroups of $\M$ isomorphic to $C_p\times
1232: C_p$ ($p$ prime) was obtained, using only the properties of the series $N(g,h)$.
1233: For related work, see Chapter 8 of [{\bf 57}].
1234:
1235:
1236: \medskip{{\it 5.3. Other finite groups.}}
1237: It is natural to ask about Moonshine for other groups.
1238: Indeed, the Hauptmodul for $\Gamma_0(2)+$ looks like
1239: $$q^{-1}+4372q+96256q^2+12\,40002q^3+\cdots\eqno(5.5a)$$
1240: and we find the relations
1241: $$4372=4371+1\ ,\qquad 96256=96255+1\ ,\qquad 12\,40002=11\,39374+4371+2\cdot 1
1242: \ ,\eqno(5.5b)$$
1243: where 1, 4371, 96255, and $11\,39374$ are all dimensions of irreducible
1244: representations of the Baby Monster ${\Bbb B}$. Thus we find Moonshine
1245: for ${\Bbb B}$! We will return to this example shortly.
1246:
1247:
1248: Of course any subgroup of $\M$ automatically inherits Moonshine by
1249: restriction, but obviously this isn't interesting.
1250: Most constructions of the Leech lattice start with Mathieu's sporadic
1251: $M_{24}$ (see e.g.\ Chapters 10 and 11 of [{\bf 26}]), and most
1252: constructions of the Monster involve the Leech lattice. Thus we are led
1253: to the following natural hierarchy of (most) sporadics:
1254: \smallskip \item{(1)} $M_{24}$ (from which we can get $M_{11}$, $M_{12}$,
1255: $M_{22}$, $M_{23}$); which leads to
1256:
1257: \item{(2)} $Co_0=C_2\times Co_1$ (from which we get $HJ$, $HS$,
1258: $McL$, $Suz$, $Co_3$, $Co_2$); which leads to
1259:
1260: \item{(3)} $\M$ (from which we get $He$, $Fi_{22}$, $Fi_{23}$, $Fi_{24}'$,
1261: $HN$, $Th$, ${\Bbb B}$).\smallskip
1262:
1263: It can thus be argued that we could approach problems in Monstrous Moonshine,
1264: by first addressing in order $M_{24}$ and $Co_1$, which should be much simpler. Indeed,
1265: the full VOA orbifold theory --- i.e.\ the complete analogue of \S5.1 ---
1266: for $M_{24}$ has been established in [{\bf 38}] (the relevant series
1267: ${\cal Z}(g,h)$ had already been constructed in [{\bf 88}]).
1268:
1269: Largely by trial and error, Queen [{\bf 101}] established Moonshine for the
1270: following groups (all essentially centralisers of elements of $\M$):
1271: $Co_0$, $Th$, $3.2.Suz$, $2.HJ$, $HN$, $2.A_7$, $He$, $M_{12}$ (by e.g.\
1272: `$2.HJ$' we mean $C_2$ is normal and $HJ$ is the quotient $2.HJ/C_2$).
1273: In particular,
1274: to each element $g$ of these groups, there corresponds a series $Q_g(\tau)=
1275: q^{-1}+\sum_{n=0}^\infty a_n(g)q^n$, which is a Hauptmodul for some
1276: modular group of moonshine-type, and where each $g\mapsto a_n(g)$ is a
1277: virtual character. For $Th$,
1278: $HN$, $He$ and $M_{12}$ it is a proper character. Other differences with
1279: Monstrous Moonshine are that there can be a preferred nonzero value for the
1280: constant term $a_0$, and that although $\Gamma_0(N)$ will be a subgroup of
1281: the fixing group, it won't necessarily be normal. We will return to these
1282: results next section, where we will see that many seem to come out of
1283: the Moonshine for $\M$. About half of Queen's Hauptmoduls $Q_g$ for $Co_0$ do
1284: not arise as a McKay--Thompson series for $\M$. Norton's conjectures in \S5.1
1285: are a reinterpretation and extension of Queen's work.
1286:
1287: Queen never reached ${\Bbb B}$ because of its size. However, the Moonshine
1288: (5.5) for ${\Bbb B}$ falls into her and Norton's scheme because (5.5a)
1289: is the McKay--Thompson series associated to class 2A of $\M$, and the centraliser
1290: of an element in 2A is a double cover of ${\Bbb B}$.
1291:
1292: There can't be a VOA $V=\oplus_nV_n$ with graded dimension (5.5a) and automorphisms
1293: in ${\Bbb B}$, because e.g.\ the ${\Bbb B}$-module $V_3$ doesn't contain $V_2$ as
1294: a submodule. However, H\"ohn deepened the analogy between $\M$ and ${\Bbb B}$
1295: by constructing a vertex operator superalgebra $V{\Bbb B}^\natural$ of rank
1296: $c=23.5$, called
1297: the {\it shorter Moonshine module}, closely related to $V^\natural$ (see e.g.\
1298: [{\bf 56}]). Its automorphism group is
1299: $C_2\times {\Bbb B}$. Just as $\M$ is the automorphism group of the Griess
1300: algebra $V^\natural_2$, so is ${\Bbb B}$ the automorphism group of the algebra
1301: $(V{\Bbb B}^\natural)_2$. Just as $V^\natural$ is associated to the Leech lattice
1302: $\Lambda$, so is $V{\Bbb B}^\natural$ associated to the shorter Leech lattice $O_{23}$,
1303: the unique 23-dimensional positive-definite self-dual lattice with no vectors
1304: of length 2 or 1 (see e.g.\ Chapter 6 of [{\bf 26}]). The automorphism group of
1305: $O_{23}$ is $C_2\times Co_2$.
1306:
1307: There has been no interesting Moonshine rumoured for the remaining six
1308: sporadics (the {\it pariahs} $J_1$, $J_3$, $Ru$, $ON$, $Ly$, $J_4$).
1309: There will be some sort of Moonshine for any group which is an
1310: automorphism group of a vertex operator algebra (so this means
1311: any finite group [{\bf 37}]!). Many finite groups of
1312: Lie type should arise as automorphism groups of VOAs associated to
1313: affine algebras except defined over finite fields.
1314: But apparently all known examples of genus-0 Moonshine are limited to the
1315: groups involved with $\M$.
1316:
1317: \medskip{{\it 5.4. Modular Moonshine}}.
1318: Consider an element $g\in\M$. We expect from [{\bf 101}], [{\bf 97}], [{\bf 36}]
1319: that there is a Moonshine for the centraliser $C_{\M}(g)$ of $g$ in $\M$,
1320: governed by the $g$-twisted module $V^\natural(g)$. Unfortunately, $V^\natural
1321: (g)$ is not usually itself a VOA, so the analogy with $\M$ is not perfect.
1322: Ryba found it interesting that, for $g\in\M$ of prime order $p$, Norton's
1323: series $N(g,h)$ is a McKay--Thompson series (and has all the associated nice
1324: properties) whenever $h$ is $p$-{\it regular}
1325: (i.e.\ $h$ has order coprime to $p$). This special behaviour of $p$-regular
1326: elements suggested to him to look at modular representations.
1327:
1328: The basics of modular representations and {Brauer characters} are discussed in
1329: sufficient detail in Chapter 2 of [{\bf 31}]. A {\it modular representation} $\rho$
1330: of a group $G$
1331: is a representation defined over a field of positive characteristic $p$ dividing
1332: the order $|G|$ of $G$. Such representations possess many special (i.e.\
1333: unpleasant) features. For one
1334: thing, they are no longer completely reducible (so the role of irreducible
1335: modules as direct summands will be replaced with their role as composition
1336: factors). For another, the usual notion
1337: of character (the trace of representation matrices) loses its usefulness and is
1338: replaced by the more subtle {\it Brauer character} $\beta(\rho)$: a complex-valued class function
1339: on $\M$ which is only well-defined on the $p$-regular elements of $G$.
1340:
1341: \medskip{\smcap Theorem 7.} [{\bf 105}], [{\bf 17}], [{\bf 13}] {\it Let
1342: $g\in \M$ be any element of prime order $p$, for any $p$ dividing $|\M|$.
1343: Then there is a vertex operator
1344: superalgebra ${}^g{\cal V}=\oplus_{n\in\Z}{}^g{\cal V}_n$ defined over the
1345: finite field ${\Bbb F}_p$ and acted on by the centraliser $C_{\M}(g)$. If $h\in C_\M(g)$
1346: is $p$-regular, then the graded Brauer character
1347: $$R(g,h;\tau):=q^{-1}\sum_{n\in\Z}\beta({}^g{\cal V}_n)(h)\, q^n$$
1348: equals the McKay--Thompson series $T_{gh}(\tau)$. Moreover, for $g$ belonging to
1349: any conjugacy class in $\M$ except 2B, 3B, 5B, 7B, or 13B, this is in fact
1350: an ordinary VOA (i.e.\ the `odd' part vanishes),
1351: while in the remaining cases the graded Brauer characters of both the
1352: odd and even parts can separately be expressed using McKay--Thompson series.}\medskip
1353:
1354: By a vertex operator superalgebra, we mean there is a $\Z_2$-grading into
1355: even and odd subspaces, and for $u,v$ both odd the commutator in (4.2a) is
1356: replaced by an anticommutator. In the proof, the superspaces arise as
1357: cohomology groups, which naturally form an alternating sum.
1358: The centralisers $C_\M(g)$ in the Theorem are quite nice: e.g.\ for $g$ in
1359: classes 2A, 2B, 3A, 3B, 3C, 5A, 5B, 7A, 11A, respectively, these involve
1360: the sporadic
1361: groups ${\Bbb B}$, $Co_1$, $Fi_{24}'$, $Suz$, $Th$, $HN$, $HJ$, $He$, and $M_{12}$.
1362: The proof for $p=2$ is not complete at the present time.
1363: The conjectures in [{\bf 105}] concerning modular
1364: analogues of the Griess algebra for several sporadics follow from Theorem 7.
1365:
1366: Can these modular ${}^g{\cal V}$s be interpreted as a reduction mod $p$ of
1367: (super)algebras in characteristic 0? Also, what about elements $g$ of composite
1368: order?
1369:
1370: \medskip{{\smcap Conjecture 8.}} [{\bf 13}] {\it Choose any $g\in\M$ and let
1371: $n$ denote its order. Then there is a ${1\over n}\Z$-graded superspace
1372: ${}^g\widehat{{\cal V}}=\oplus_{i\in {1\over n}\Z}{}^g\widehat{{\cal V}}_i$
1373: over the ring of cyclotomic integers $\Z[e^{2\pi\i/n}]$. It is often (but
1374: probably not always) a vertex operator superalgebra --- in particular
1375: ${}^1\widehat{{\cal V}}$ is an integral form of the Moonshine module
1376: $V^\natural$. Each ${}^g\widehat{{\cal V}}$ carries a representation of a
1377: central extension of $C_{\M}(g)$ by $C_n$. Define the graded trace
1378: $$B(g,h;\tau)=q^{-1}\sum_{i\in {1\over n}\Z}{\rm ch}_{{}^g\widehat{{\cal V}}_i}
1379: (h)\,q^i\ .$$
1380: If $g,h\in\M$ commute and have
1381: coprime orders, then $B(g,h;\tau)=T_{gh}(\tau)$. If all $q$-coefficients of
1382: $T_g$ are nonnegative, then the `odd' part of ${}^g\widehat{{\cal V}}$
1383: vanishes, and ${}^g\widehat{{\cal V}}$ is the $g$-twisted
1384: module $V^\natural(g)$ of} [{\bf 36}]. {\it If $g$ has prime order $p$, then the
1385: reduction mod $p$ of ${}^g\widehat{{\cal V}}$ is the modular vertex operator
1386: superalgebra ${}^g{\cal V}$ of Theorem 7.}\medskip
1387:
1388: When we say ${}^1\widehat{{\cal V}}$ is an integral form for $V^\natural$,
1389: we mean that ${}^1\widehat{{\cal V}}$ has the same structure as a VOA, with
1390: everything defined over $\Z$, and tensoring it with $\C$ recovers $V^\natural$.
1391: This remarkable conjecture, which tries to explain Theorem 7, is completely
1392: open.
1393:
1394:
1395:
1396:
1397: \medskip{{\it 5.5. The geometry of Moonshine.}}
1398: Algebra is the mathematics of structure, and so of course it has a profound
1399: relationship with every area of mathematics. Therefore the trick for finding
1400: possible fingerprints of Moonshine in say geometry is to look there for modular
1401: functions. And that search quickly leads to the elliptic genus.
1402:
1403: For details see e.g.\ [{\bf 55}], [{\bf 108}], [{\bf 112}]. All manifolds here
1404: are compact, oriented and differentiable. In Thom's cobordism
1405: ring $\Omega$, elements are equivalence classes of cobordant manifolds, addition
1406: is connected sum, and multiplication is Cartesian product.
1407: The {\it universal elliptic genus} $\phi(M)$ is a ring homomorphism
1408: from $\Q\otimes\Omega$ to the ring of power series in $q$, which sends $n$-dimensional
1409: manifolds with spin connections to a weight $n/2$ modular form of $\Gamma_0(2)$ with
1410: integer coefficients. Several variations and generalisations have been
1411: introduced, e.g.\ the Witten genus assigns spin manifolds with vanishing first
1412: Pontryagin class a weight $n/2$ modular form of SL$_2(\Z)$ with integer
1413: coefficients.
1414:
1415: Several deep relationships between elliptic genera and the general material reviewed
1416: elsewhere in this paper, have been uncovered. For instance, the important
1417: rigidity property of the Witten genus with respect to any compact Lie group
1418: action on the manifold, is a consequence of the modularity of the characters
1419: of affine algebras (our Theorem 4) [{\bf 81}]. The elliptic genus of a
1420: manifold $M$ has been interpreted as the graded dimension of a vertex operator
1421: superalgebra constructed from $M$ [{\bf 111}]. Seemingly related to this,
1422: [{\bf 18}] recovered the elliptic genus of a Calabi--Yau manifold $X$ from the
1423: sheaf of vertex algebras in the chiral de Rham complex [{\bf 85}] attached to
1424: $X$. Unexpectedly, the elliptic genus of even-dimensional projective spaces
1425: $P^{2n}$ has nonnegative coefficients and in fact equals the graded
1426: dimension of a certain vertex algebra [{\bf 86}]; this suggests interesting
1427: representation-theoretic questions in the spirit of Monstrous Moonshine.
1428: In physics, elliptic
1429: genera arise as partition functions of $N=2$ superconformal field theories
1430: [{\bf 120}]. Mason's constructions [{\bf 88}] associated to Moonshine for the Mathieu
1431: group $M_{24}$ have been interpreted as providing a geometric model (`elliptic
1432: system') for elliptic
1433: cohomology Ell${}^*(BM_{24})$ of the classifying space of $M_{24}$ [{\bf 112}],
1434: [{\bf 39}].
1435: The Witten genus (normalised by $\eta^8$) of the Milnor--Kervaire manifold
1436: $M_0^8$, an 8-dimensional manifold built from the $E_8$ diagram, equals
1437: $j^{{1\over 3}}$ [{\bf 55}] (recall (3.5)).
1438:
1439: Hirzebruch's `prize question'
1440: (p.86 of [{\bf 55}]) asks for the construction of a 24-dimensional manifold
1441: $M$ with Witten genus $J$ (after being normalised by $\eta^{24}$). We would
1442: like $\M$ to act on $M$ by diffeomorphisms, and the twisted Witten genera to be
1443: the McKay--Thompson series $T_g$. It would also be nice to associate Norton's
1444: series $N(g,h)$ to this Moonshine manifold. Constructing such a manifold
1445: is perhaps the remaining Holy Grail of Monstrous Moonshine.
1446:
1447: Hirzebruch's question was partially answered by Mahowald and Hopkins [{\bf 84}],
1448: who constructed a manifold with Witten genus $J$, but couldn't show it would
1449: support an effective action of $\M$. Related work is [{\bf 3}], who constructed
1450: several actions of $\M$ on e.g.\ 24-dimensional manifolds (but none of which
1451: could have genus $J$), and [{\bf 73}], who showed the graded dimensions of the
1452: subspaces $V^\natural_\pm$ of the Moonshine module are twisted $\widehat{A}$-genera
1453: of Milnor--Kervaire's manifold $M^8_0$ (the $\widehat{A}$-genus is the specialisation
1454: of elliptic genus to the cusp $\i\infty$).
1455:
1456: \smallskip There has been a second conjectured relationship between geometry and
1457: Monstrous Moonshine. {\it Mirror symmetry} says that most Calabi--Yau manifolds come
1458: in closely related pairs. Consider a 1-parameter family $X_z$ of
1459: Calabi--Yau manifolds, with mirror $X^*$ given by the resolution of an orbifold
1460: $X/G$ for $G$ finite and abelian. Then the Hodge numbers $h^{1,1}(X)$ and
1461: $h^{2,1}(X^*)$ will be equal, and more precisely the moduli space of (complexified) K\"ahler
1462: structures on $X$ will be locally isometric to the moduli space of complex
1463: structures on $X^*$. The `mirror map' $z(q)$, which can be defined using
1464: the Picard--Fuchs equation [{\bf 95}], gives a canonical map between those moduli
1465: spaces. For example, $x_1^4+x_2^4+x_3^4+x_4^4+z^{-1/4}x_1x_2x_3x_4=0$
1466: is such a family of K3 surfaces, where $G=C_4\times C_4$. Its mirror map
1467: is given by
1468: $$z(q)=q-104q^2+6444q^3-311744q^4+13018830q^5-493025760q^6+\cdots\ .\eqno(5.6)$$
1469:
1470: Lian--Yau [{\bf 80}] noticed that the reciprocal $1/z(q)$ of the mirror map
1471: in (5.6) equals the McKay--Thompson series $T_g(\tau)+104$ for $g$ in class
1472: 2A of $\M$. After looking
1473: at several other examples with similar conclusions, they proposed their
1474: {\it Mirror-Moonshine Conjecture}: The reciprocal $1/z$ of the mirror
1475: map of a 1-parameter family of K3 surfaces with an orbifold mirror, will be
1476: a McKay--Thompson series (up to an additive constant).
1477:
1478: A counterexample (and more examples) are given in \S7 of [{\bf 118}]. In
1479: particular, although there are relations between mirror symmetry and
1480: modular functions (see e.g.\ [{\bf 51}] and [{\bf 54}]),
1481: there doesn't seem to be any special relation with the Monster. Doran [{\bf
1482: 40}] `demystifies the Mirror-Moonshine phenomenon' by finding necessary and
1483: sufficient conditions for $1/z$ to be a modular function for a modular
1484: group commensurable with SL$_2(\Z)$.
1485:
1486: \bigskip\centerline{{\it 6. The physics of Moonshine}}\medskip
1487:
1488: The physical side (perturbative string theory, or equivalently conformal
1489: field theory) of Moonshine was noticed early on, and has profoundly influenced
1490: the development of Moonshine and VOAs. This is a very rich subject, which
1491: we can only superficially touch on. The book [{\bf 32}], with its extensive
1492: bibliography, provides an introduction but will be difficult reading for many
1493: mathematicians (as will this section!) The treatment in [{\bf 45}] is more
1494: accessible and shows how naturally VOAs arise from the physics.
1495: This effectiveness of physical interpretations isn't magic --- it merely
1496: tells us that many of our finite-dimensional objects are seen much more
1497: clearly when studied through infinite-dimensional structures (often by being
1498: `looped'). Of course Moonshine, which teaches us to study the finite group
1499: $\M$ via its infinite-dimensional module $V^\natural$, fits perfectly into
1500: this picture.
1501:
1502: A conformal field theory (CFT) is a quantum field theory on 2-dimensional
1503: space-time, whose symmetries include the conformal transformations. In string
1504: theory the basic objects are finite curves (`strings') rather than points
1505: (`particles'), and the CFT lives on the surface traced by the strings as they
1506: evolve (colliding and separating) through time. Each CFT is associated with
1507: a pair ${\cal V}_L,{\cal V}_R$ of mutually commuting VOAs, called its {\it
1508: chiral algebras} [{\bf 6}]. For example, strings living on a compact Lie group manifold
1509: (the so-called Wess--Zumino--Witten model) will have chiral algebras given by
1510: affine algebra VOAs. The space ${\cal H}$ of states for the
1511: CFT carries a representation of ${\cal V}_L\otimes\overline{{\cal V}}_R$,
1512: and many authors have (somewhat optimisticly) concluded that the study of
1513: CFTs reduces to that of VOA representation theory. Rational VOAs correspond
1514: to the important class of {\it rational} CFTs, where ${\cal H}$ decomposes into a
1515: finite sum $\oplus M_L\otimes M_R$ of irreducible modules. The Virasoro
1516: algebra (4.3) arises naturally in CFT through infinitesimal conformal transformations.
1517: The vertex operator $Y(\phi,z)$, for the space-time parameter $z=e^{t+\i x}$,
1518: is the quantum field which creates from the vacuum $|0\rangle\in{\cal H}$ the
1519: state $|\phi\rangle\in{\cal H}$ at time $t=-\infty$: $|\phi\rangle={\rm lim}_{z
1520: \rightarrow 0}Y(\phi,z)\,|0\rangle$. In particular, Borcherds' definition
1521: [{\bf 9}] of VOAs can be
1522: interpreted as an axiomatisation of the notion of chiral algebra in CFT, and
1523: for this reason alone is important.
1524:
1525:
1526: In CFT, the Hauptmodul property of Moonshine is hard to interpret, and a less
1527: direct formulation like that in [{\bf 116}] is needed. However, both the statement
1528: and proof of Theorem 5 are natural from the CFT framework (see [{\bf 45}])
1529: --- e.g.\ the modularity of the series $T_g$ and $N(g,h)$ are automatic in
1530: CFT. This modularity arises in CFT through the equivalence of the Hamiltonian
1531: formulation, which describes concretely the graded spaces we take traces on
1532: (and hence the coefficients of our $q$-expansions), and the Feynman path
1533: formalism, which interprets these graded traces as sections over moduli spaces
1534: (and hence makes modularity manifest). Beautiful reviews are sketched in
1535: [{\bf 119}], [{\bf 120}].
1536:
1537: Because $V^\natural$ is so mathematically special, it may be expected that it
1538: corresponds to interesting physics. Certainly it has been the subject of
1539: some speculation. There will be a $c=24$ rational CFT whose chiral algebra
1540: ${\cal V}_L$ and state space ${\cal H}$ are both $V^\natural$, while
1541: ${\cal V}_R$ is trivial (this is possible
1542: because $V^\natural$ is holomorphic). This CFT is nicely described in [{\bf 34}];
1543: see also [{\bf 35}].
1544: The Monster is the symmetry of that CFT, but the Bimonster $\M\wr C_2$ will
1545: be the symmetry of a rational CFT with ${\cal H}=V^\natural\otimes \overline{V^\natural}$.
1546: The paper [{\bf 27}] finds a family of D-branes for the latter theory which
1547: are in one-to-one correspondence with the elements of $\M$, and their `overlaps'
1548: $\langle\!\langle g\|q^{{1\over 2}(L_0+\bar{L}_0-{c\over 24})}\|h\rangle\!\rangle$
1549: equal the McKay--Thompson series $T_{g^{-1}h}$. However, we still lack any explanation
1550: as to why a CFT involving $V^\natural$ should yield interesting {\it physics}.
1551:
1552: Almost every facet of Moonshine finds a natural formulation in CFT, where it
1553: often was discovered first. For example, the `No-Ghost' Theorem of
1554: Brower--Goddard--Thorn
1555: was used to great effect in [{\bf 11}] to understand the structure of the
1556: Monster Lie algebra ${\frak m}$. On a finite-dimensional manifold $M$, the
1557: index of the Dirac operator $D$ in the heat kernel interpretation is a path
1558: integral in supersymmetric quantum mechanics, i.e\ an integral over the
1559: free loop space ${\cal L}M=\{\gamma:S^1\rightarrow M\}$; the string theory
1560: version of this is that the index of the Dirac operator on ${\cal L}M$ should
1561: be an integral over ${\cal L}({\cal L}M)$, i.e.\ over smooth maps of tori into
1562: $M$, and this is just the elliptic genus, and explains why it should be
1563: modular. The orbifold construction of [{\bf 36}]
1564: comes straight from CFT (although [{\bf 43}]'s construction of $V^\natural$
1565: predates CFT orbifolds by a year and in fact influenced their development
1566: in physics). That said, the translation process from
1567: physics to mathematics of course is never easy --- Borcherds' definition
1568: [{\bf 9}] is a prime example!
1569:
1570: But from this standpoint,
1571: what is most exciting is what hasn't yet been fully exploited. String theory
1572: tells us that CFT can live on any surface $\Sigma$. The VOAs, including the
1573: geometric VOAs of [{\bf 59}], capture CFT in genus 0. The graded dimensions and
1574: traces considered above concern CFT quantities (`conformal blocks') at genus
1575: 1: $\tau\mapsto e^{2\pi\i\tau}$ maps $\H$ onto a cylinder, and the trace
1576: identifies the two ends. But there are analogues of all this at higher genus
1577: [{\bf 123}] (though the formulas can rapidly
1578: become awkward). For example, the graded dimension of e.g.\ the $V^\natural$
1579: CFT in genus 2 is computed in [{\bf 117}], and involves e.g.\ Siegel theta
1580: functions. The orbifold theory in \S5.1 is genus 1: each `sector' $(g,h)$
1581: corresponds to a homomorphism from the fundamental group $\Z^2$ of the torus
1582: into the orbifold group $G$ (e.g\ $G=\M$) --- $g$ and $h$ are the targets of
1583: the two generators of $\Z^2$ and hence must commute. More generally, the
1584: sectors will correspond to each homomorphism $\varphi:\pi_1(\Sigma)\rightarrow
1585: G$, and to each we will get a higher genus trace ${\cal Z}(\varphi)$, which
1586: will be a function on the Teichm\"uller space $T_g$ (generalising the upper half-plane
1587: ${\Bbb H}$ for genus 1). The action of SL$_2(\Z)$ on the $N(g,h)$
1588: generalises to the action of the mapping class group on $\pi_1(\Sigma)$ and
1589: $T_g$. See e.g.\ [{\bf 4}] for some thoughts in this direction.
1590:
1591:
1592:
1593: \bigskip\centerline{{\it 7. Conclusion}}\medskip
1594:
1595: There are different basic aspects to Monstrous Moonshine: (i) why modularity enters
1596: at all; (ii) why in particular we have genus 0; and (iii) what does it have to
1597: do with the Monster. We understand (i) best. There will be a Moonshine-like
1598: relation between any (subgroup of the) automorphism
1599: group of any rational VOA, and the characters $\chi_M$, and the
1600: same can be expected to hold of the orbifold characters ${\cal Z}$ in \S5.1.
1601:
1602: To prove the genus 0 property of the $T_g$, we needed recursions obtained one way
1603: or another from the Monster Lie algebra ${\frak m}$, and from these we apply
1604: Theorem 6. These recursions are very special, but so presumably is the
1605: genus 0 property. The suggestion of [{\bf 20}] though is that we may be able
1606: to considerably simplify this part of the argument.
1607:
1608: Every group known to have rich Moonshine properties is contained in the Monster.
1609: Our understanding of this seemingly central role of $\M$ is the poorest of those three
1610: aspects.
1611:
1612: It should be clear
1613: from this review, of the central role VOAs play in our current understanding
1614: of Moonshine. The excellent review [{\bf 39}] makes this point even more
1615: forcefully. It can be (and has been) questioned
1616: though whether the full and difficult machinery of VOAs is really needed to
1617: understand
1618: this, i.e.\ whether we really have isolated the key conjunction of properties
1619: needed for Moonshine to arise. CFT has been an invaluable guide thus far, but
1620: perhaps we are a little too steeped in its lore.
1621:
1622: Moonshine (in its more general sense) is a relation between algebra and number theory, and its impact on
1623: algebra has been dramatic (e.g.\ VOAs, $V^\natural$, Borcherds--Kac--Moody algebras).
1624: Its impact on number theory has been far less so. This may merely be a
1625: temporary accident due to the backgrounds of most researchers (including the
1626: mathematical physicists) working to date in the area. But
1627: the most exciting prospects for the future of Moonshine (in this writer's
1628: opinion) are in the direction of
1629: number theory.
1630: Hints of this future can be found in e.g.\ [{\bf 121}], [{\bf 41}], [{\bf 14}],
1631: [{\bf 33}], [{\bf 52}], [{\bf 94}].
1632:
1633: \bigskip{{\it Acknowledgements.}} This paper was written at IHES
1634: and at U of Wales Swansea, and I thank both for their hospitality. Much of
1635: what I've learned of Moonshine came from conversations over the years with
1636: Chris Cummins and John McKay, and I warmly thank both. I appreciate the many
1637: comments and suggestions provided by the readers of the first draft;
1638: in this vein I would
1639: particularly like to mention James Lepowsky and Fyodor Malikov.
1640: My research is supported in part by NSERC.
1641:
1642: \bigskip\centerline{{\it References}\footnote{$^\dagger$}{{\smal
1643: For a more comprehensive Moonshine bibliography, especially for older
1644: papers, see}}\footnote{}{{\smal
1645: http:$/\!/$cicma.mathstat.concordia.ca/faculty/cummins/moonshine.html}}}\medskip
1646:
1647:
1648:
1649: \item{{\bf 1.}} {\smcap D.\ Alexander, C.\ Cummins, J.\ McKay,} and {\smcap
1650: C.\ Simons}, `Completely replicable functions', {\it Groups, Combinatorics
1651: and Geometry} (ed.\ M.W.\ Liebeck and J.\ Saxl, Cambridge Univ.\ Press, 1992)
1652: 87--98.
1653:
1654: \item{{\bf 2.}} {\smcap O.\ Alvarez}, `Conformal anomalies and the index theorem',
1655: {\it Nucl.\ Phys.} B286 (1987) 175--188.
1656:
1657: \item{{\bf 3.}} {\smcap M.G.\ Aschbacher}, `Finite groups acting on homology
1658: manifolds',
1659: {\it Olga Taussky-Todd: in Memoriam, Pacific J.\ Math.\ Special Issue} (ed.\
1660: M.\ Aschbacher et al, Pacific Journal of Mathematics, Berkeley, 1997) 3--36.
1661:
1662: \item{{\bf 4.}} {\smcap P.\ B\'antay}, `Higher genus moonshine', {\it Moonshine,
1663: the Monster, and Related Topics, Contemp.\
1664: Math.} 193 (Amer.\ Math.\ Soc., Providence, 1996) 1--8.
1665:
1666: \item{{\bf 5.}} {\smcap P.\ B\'antay}, `The kernel of the modular representation
1667: and the Galois action in RCFT', {\it Commun.\ Math.\ Phys.} 233 (2003) 423--438.
1668:
1669: \item{{\bf 6.}} {\smcap A.\ Belavin, A.M.\ Polyakov,} and {\smcap A.B.\
1670: Zamolodchikov}, `Infinite conformal symmetries in two-dimensional quantum
1671: field theory', {\it Nucl.\ Phys.} {\bf B241} (1984) 333--380.
1672:
1673: \item{{\bf 7.}} {\smcap J.S.\ Birman}, {\it Braids, Links and Mapping Class Groups}
1674: (Princeton Univ.\ Press, 1974).
1675:
1676: \item{{\bf 8.}} {\smcap S.\ Bloch}, `Zeta values and differential operators
1677: on the circle', {\it J.\ Alg.} 182 (1996) 476--500.
1678:
1679: \item{{\bf 9.}} {\smcap R.E.\ Borcherds}, `Vertex algebras, Kac--Moody algebras, and the
1680: Monster', {\it Proc.\ Natl.\ Acad.\ Sci.\ (USA)} 83 (1986) 3068--3071.
1681:
1682: \item{{\bf 10.}} {\smcap R.E.\ Borcherds}, `Generalized Kac--Moody Lie algebras',
1683: {\it J.\ Alg.} 115 (1988) 501--512.
1684:
1685: \item{{\bf 11.}} {\smcap R.E\ Borcherds}, `Monstrous moonshine and monstrous Lie superalgebras',
1686: {\it Invent.\ Math.} 109 (1992) 405--444.
1687:
1688: \item{{\bf 12.}} {\smcap R.E\ Borcherds}, `Sporadic groups and string theory',
1689: {\it First European Congress of Math.\ Paris (1992)}, Vol.I (Birkh\"auser,
1690: Basel, 1994) 411--421.
1691:
1692: \item{{\bf 13.}} {\smcap R.E.\ Borcherds}, `Modular moonshine III', {\it Duke Math.\ J.}\ 93 (1998)
1693: 129--154.
1694:
1695: \item{{\bf 14.}} {\smcap R.E.\ Borcherds}, `Automorphic forms with
1696: singularities on Grassmannians', {\it Invent.\ Math.} 132 (1998) 491--562.
1697:
1698: \item{{\bf 15.}} {\smcap R.E.\ Borcherds}, `What is Moonshine?', {\it Proc. Intern.\
1699: Congr.\ Math.\ 1998 (Berlin)}, Vol. 1 (Documenta Mathematica, Bielefeld, 1998)
1700: 607--615.
1701:
1702: \item{{\bf 16.}} {\smcap R.E.\ Borcherds}, `Problems in Moonshine',
1703: {\it First Intern.\ Congr.\ Chinese Math.}
1704: (Amer. Math. Soc., Providence, 2001) 3--10.
1705:
1706: \item{{\bf 17.}} {\smcap R.E.\ Borcherds} and {\smcap A.J.E.\ Ryba},
1707: `Modular moonshine II', {\it Duke Math.\ J.} 83 (1996) 435--459.
1708:
1709: \item{{\bf 18.}} {\smcap L.A.\ Borisov} and {\smcap A.\ Libgober}, `Elliptic
1710: genera of toric varieties and applications to mirror symmetry', {\it Invent.\
1711: Math.} 140 (2000) 453--485.
1712:
1713: \item{{\bf 19.}} {\smcap S.-P. Chan}, {\smcap M.-L. Lang}, and {\smcap C.-H.\
1714: Lim}, `Some modular functions
1715: associated to Lie algebra $E_8$', {\it Math.\ Z.} 211 (1992) 223--246.
1716:
1717: \item{{\bf 20.}} {\smcap H.\ Cohn} and {\smcap J.\ McKay}, `Spontaneous
1718: generation of modular invariants', {\it Math.\ of Comput.} 65 (1996) 1295--1309.
1719:
1720: \item{{\bf 21.}} {\smcap J.H.\ Conway}, `Monsters and Moonshine', {\it Math.\ Intell.}\ 2 (1980) 165--171.
1721:
1722:
1723: \item{{\bf 22.}} {\smcap J.H.\ Conway}, {\smcap R.T.\ Curtis}, {\smcap S.P.\ Norton},
1724: {\smcap R.A.\ Parker} and {\smcap R.A.\
1725: Wilson}, {\it An Atlas of Finite Groups} (Clarendon Press, Oxford, 1985).
1726:
1727: \item{{\bf 23.}} {\smcap J.H.\ Conway}, {\smcap J.\ McKay} and {\smcap A.\
1728: Sebbar}, `On the discrete groups of moonshine', {\it Proc.\ Amer.\ Math.\ Soc.}
1729: (to appear).
1730:
1731: \item{{\bf 24.}} {\smcap J.H.\ Conway} and {\smcap S.P.\ Norton}, `Monstrous Moonshine', {\it Bull.\ London
1732: Math.\ Soc.} 11 (1979) 308--339.
1733:
1734: \item{{\bf 25.}} {\smcap J.H.\ Conway}, {\smcap S.P.\ Norton}, and {\smcap
1735: L.H.\ Soicher}, `The Bimonster, the group $Y_{555}$, and the projective plane
1736: of order $3$', {\it Computers in Algebra} (Dekker, New York, 1988) 27--50.
1737:
1738:
1739: \item{{\bf 26.}} {\smcap J.H.\ Conway} and {\smcap N.J.A.\ Sloane}, {\it Sphere Packings,
1740: Lattices and Groups}, 3rd edn (Springer, Berlin, 1999).
1741:
1742: \item{{\bf 27.}} {\smcap B.\ Craps}, {\smcap M.R.\ Gaberdiel}, {\smcap
1743: J.A.\ Harvey}, `Monstrous branes', {\it Commun.\ Math.\ Phys.} 234 (2003)
1744: 229--251.
1745:
1746: \item{{\bf 28.}} {\smcap C.J.\ Cummins}, `Congruence subgroups of groups
1747: commensurable with PSL$(2,\Z)$ of genus 0 and 1', Preprint.
1748:
1749: \item{{\bf 29.}} {\smcap C.J. Cummins} and {\smcap T.\ Gannon}, `Modular equations and the genus
1750: zero property of moonshine functions', {\it Invent.\ Math.} 129 (1997) 413--443.
1751:
1752: \item{{\bf 30.}} {\smcap C.J.\ Cummins} and {\smcap S.P.\ Norton}, `Rational
1753: Hauptmodul are replicable', {\it Canad.\ J.\ Math.} 47 (1995) 1201--1218.
1754:
1755: \item{{\bf 31.}} {\smcap C.W.\ Curtis} and {\smcap I.\ Reiner}, {\it Methods of
1756: Representation Theory with Applications to Finite Groups and Orders}, Vol.I
1757: (Wiley, New York, 1981).
1758:
1759: \item{{\bf 32.}} {\smcap P.\ Di Francesco}, {\smcap P.\ Mathieu} and {\smcap
1760: D.\ S\'en\'echal}, {\it Conformal Field Theory} (Springer, New York, 1997).
1761:
1762: \item{{\bf 33.}} {\smcap R.\ Dijkgraaf}, `The mathematics of fivebranes',
1763: {\it Proc.\ Intern.\ Congr.\ Math.\ 1998 (Berlin)}, Vol.III (Documenta Mathematica,
1764: Bielefeld, 1998) 133--142.
1765:
1766: \item{{\bf 34.}} {\smcap L.\ Dixon}, {\smcap P.\ Ginsparg} and {\smcap J.A.\ Harvey},
1767: `Beauty and the beast:
1768: superconformal symmetry in a monster module', {\it Commun.\ Math.\ Phys.} 119 (1988)
1769: 221--241.
1770:
1771: \item{{\bf 35.}} {\smcap L.\ Dolan}, {\smcap P.\ Goddard} and {\smcap P.\
1772: Montague}, `Conformal field theory of twisted vertex operators', {\it Nucl.\
1773: Phys.} B338 (1990) 529--601.
1774:
1775: \item{{\bf 36.}} {\smcap C.\ Dong}, {\smcap H.\ Li} and {\smcap G.\ Mason},
1776: `Modular invariance of trace functions in orbifold theory and generalized
1777: moonshine', {\it Commun.\ Math.\ Phys.} 214 (2000) 1--56.
1778:
1779: %\item{{\bf DLM2.}} {\smcap C.\ Dong}, {\smcap K.\ Liu} and {\smcap X.\ Ma},
1780: %`Elliptic genus and vertex operator algebras', Preprint (arXiv: math.DG/0201135).
1781:
1782: \item{{\bf 37.}} {\smcap C.\ Dong} and {\smcap G.\ Mason}, `Nonabelian
1783: orbifolds and the boson-fermion correspondence', {\it Commun.\ Math.\ Phys.}
1784: 163 (1994) 523--559.
1785:
1786: \item{{\bf 38.}} {\smcap C.\ Dong} and {\smcap G.\ Mason}, `An orbifold
1787: theory of genus zero associated to the sporadic group $M_{24}$', {\it
1788: Commun.\ Math.\ Phys.} 164 (1994) 87--104.
1789:
1790: \item{{\bf 39.}} {\smcap C.\ Dong} and {\smcap G.\ Mason}, `Vertex operator algebras
1791: and moonshine: A survey',
1792: {\it Progress in Algebraic Combinatorics, Adv.\ Stud.\ Pure Math.}\ 24 (Math.\
1793: Soc.\ Japan, Tokyo, 1996) 101--136.
1794:
1795: \item{{\bf 40.}} {\smcap C.F.\ Doran}, `Picard--Fuchs uniformization
1796: and modularity of the mirror map', {\it Commun.\ Math.\ Phys.} 212 (2000) 625--647.
1797:
1798: \item{{\bf 41.}} {\smcap V.G.\ Drinfeld}, `On quasitriangular quasi-Hopf algebras
1799: on a group that is closely related with Gal$\,\overline{\Q}/\Q$', {\it Leningrad.\
1800: Math.\ J.} 2 (1991) 829--860.
1801:
1802: \item{{\bf 42.}} {\smcap I.\ Frenkel}, {\smcap Y.-Z.\ Huang} and {\smcap J.\
1803: Lepowsky}, `On Axiomatic Approaches
1804: to Vertex Operator Algebras and Modules', {\it Mem.\ Amer.\ Math.\ Soc.} 104
1805: (1993) 1--63.
1806:
1807: \item{{\bf 43.}} {\smcap I.\ Frenkel}, {\smcap J.\ Lepowsky} and {\smcap A.\ Meurman}, `A natural representation
1808: of the Fischer--Griess monster with the modular function $J$ as character',
1809: {\it Proc.\ Natl.\ Acad.\ Sci.\ USA} 81 (1984) 3256--3260.
1810:
1811: \item{{\bf 44.}} {\smcap I.\ Frenkel}, {\smcap J.\ Lepowsky} and {\smcap A.\ Meurman},
1812: {\it Vertex Operator Algebras and the Monster} (Academic Press, New York, 1988).
1813:
1814: \item{{\bf 45.}} {\smcap M.R.\ Gaberdiel} and {\smcap P.\ Goddard}, `Axiomatic
1815: conformal field theory', {\it Commun.\ Math.\ Phys.} 209 (2000) 549--594.
1816:
1817: \item{{\bf 46.}} {\smcap T.\ Gannon}, {\it Moonshine Beyond the Monster}
1818: (Cambridge Univ.\ Press, to appear).
1819:
1820: \item{{\bf 47.}} {\smcap G.\ Glauberman} and {\smcap S.P.\ Norton}, `On McKay's connection
1821: between the affine $E_8$ diagram and the Monster', {\it Proc.\ on Moonshine and
1822: Related Topics} (Amer.\ Math.\ Soc., Providence, 2001) 37--42.
1823:
1824: \item{{\bf 48.}} {\smcap P.\ Goddard}, `The work of Richard Ewen Borcherds',
1825: {\it Proc.\ Intern.\ Congr.\ Math.\ 1998 (Berlin)}, Vol.\ I
1826: (Documenta Mathematica, Bielefeld, 1998) 99--108.
1827:
1828: \item{{\bf 49.}} {\smcap D.\ Gorenstein}, {\it Finite Simple Groups: An Introduction to their
1829: Classification} (Plenum, New York, 1982).
1830:
1831: \item{{\bf 50.}} {\smcap R.\ Griess}, `The friendly giant', {\it Invent.\ Math.}
1832: 68 (1982) 1--102.
1833:
1834: \item{{\bf 51.}} {\smcap V.A.\ Gritsenko} and {\smcap V.V.\ Nikulin}, `The
1835: arithmetic mirror symmetry and Calabi--Yau manifolds', {\it Commun.\ Math.\
1836: Phys.} 210 (2000) 1--11.
1837:
1838: \item{{\bf 52.}} {\smcap S.\ Gukov} and {\smcap C.\ Vafa}, `Rational conformal
1839: field theories and complex multiplication', Preprint (arXiv: hep-th/0203213).
1840:
1841: \item{{\bf 53.}} {\smcap K.\ Harada}, {\smcap M.\ Miyamoto}, and {\smcap H.\
1842: Yamada}, `A generalization of Kac--Moody algebras', {\it Groups, Difference Sets,
1843: and the Monster} (de Gruyter, Berlin, 1996) 377--408.
1844:
1845: \item{{\bf 54.}} {\smcap J.A.\ Harvey} and {\smcap G.\ Moore}, `Algebras, BPS
1846: states, and strings', {\it Nucl.\ Phys.} B463 (1996) 315--368.
1847:
1848: \item{{\bf 55.}} {\smcap F.\ Hirzebruch}, {\smcap T.\ Berger}, and {\smcap R.\ Jung}, {\it Manifolds and Modular
1849: Forms} (Friedr.\ Vieweg \& Sohn, 1991).
1850:
1851: %\item{{\bf Ho.}} {\smcap G.\ H\"ohn}, `Selbst duale Vertexoperatoralgebren und das
1852: %Babymonster', PhD thesis, Universit\"at Bonn, 1995.
1853:
1854: \item{{\bf 56.}} {\smcap G.\ H\"ohn}, `The group of symmetries of the shorter
1855: Moonshine module', Preprint (arXiv:math.QA/0210076).
1856:
1857: \item{{\bf 57.}} {\smcap T.\ Hsu}, {\it Quilts: Central Extensions, Braid Actions, and Finite
1858: Groups}, {\it Lecture Notes in Mathematics 1731} (Springer, Berlin, 2000).
1859:
1860: \item{{\bf 58.}} {\smcap Y.-Z.\ Huang}, `Binary trees and finite-dimensional
1861: Lie algebras', {\it Algebraic Groups and their Generalizations: Quantum and
1862: Infinite-Dimensional Methods}, Proc.\ Symp.\ Pure Math.\ 56, pt.\ 2 (Amer.\
1863: Math.\ Soc., Providence, 1994) 337--348.
1864:
1865: \item{{\bf 59.}} {\smcap Y.-Z.\ Huang}, {\it Two-dimensional Conformal Geometry
1866: and Vertex Operator Algebras} (Birkh\"auser, Boston, 1997).
1867:
1868: \item{{\bf 60.}} {\smcap A.A.\ Ivanov}, `Geometric presentations of groups with an application
1869: to the Monster', {\it Proc.\ Intern.\ Congr.\ Math.\ 1990 (Kyoto)}, Vol.\
1870: II (Springer, Hong Kong, 1991) 1443--1453.
1871:
1872: \item{{\bf 61.}} {\smcap A.A.\ Ivanov}, `$Y$-groups via transitive extension',
1873: {\it J.\ Alg.} 218 (1999) 412--435.
1874:
1875: \item{{\bf 62.}} {\smcap R.\ Ivanov} and {\smcap M.P.\ Tuite}, `Some irrational
1876: generalized Moonshine from orbifolds', {\it Nucl.\ Phys.} B635 (2002) 473--491.
1877:
1878: \item{{\bf 63.}} {\smcap E.\ Jurisich}, `An exposition of generalized Kac--Moody
1879: algebras', {\it Contemp.\ Math.} 194 (Amer.\ Math.\ Soc., Providence, 1996)
1880: 121--159.
1881:
1882: \item{{\bf 64.}} {\smcap E.\ Jurisich}, {\smcap J.\ Lepowsky} and {\smcap
1883: R.L.\ Wilson}, `Realizations of the Monster Lie algebra', {\it Selecta Math.\
1884: (NS)} 1 (1995) 129--161.
1885:
1886: \item{{\bf 65.}} {\smcap V.G.\ Kac}, `An elucidation of: Infinite-dimensional algebras,
1887: Dedekind's $\eta$-function, classical M\"obius function and the very strange
1888: formula. $E_8^{(1)}$ and the cube root of the modular invariant $j$',
1889: {\it Adv.\ in Math.} 35 (1980) 264--273.
1890:
1891:
1892: \item{{\bf 66.}} {\smcap V.G.\ Kac}, {\it Infinite Dimensional Lie Algebras}, 3rd edn,
1893: (Cambridge Univ.\ Press, 1990).
1894:
1895: \item{{\bf 67.}} {\smcap V.G.\ Kac}, {\it Vertex Operators for Beginners}, Univ.\
1896: Lecture Series, Vol.\ 10 (Amer.\ Math.\ Soc., Providence, 1997).
1897:
1898: \item{{\bf 68.}} {\smcap V.G.\ Kac} and {\smcap S.-J.\ Kang}, `Trace formula for
1899: graded Lie algebras and Monstrous Moonshine', {\it Representations of Groups, Banff
1900: (1994)} (Amer.\ Math.\ Soc., Providence, 1995) 141--154.
1901:
1902: \item{{\bf 69.}} {\smcap V.G.\ Kac} and {\smcap D.H.\ Peterson}, `Infinite dimensional
1903: Lie algebras, theta functions, and modular forms', {\it Adv.\ in
1904: Math.} 53 (1984) 125--264.
1905:
1906: \item{{\bf 70.}} {\smcap S.\ Kass}, {\smcap R.V.\ Moody}, {\smcap J.\ Patera} and
1907: {\smcap R.\ Slansky}, {\it Affine Lie Algebras,
1908: Weight Multiplicities, and Branching Rules}, Vol.\ 1 (Univ.\ California
1909: Press, Berkeley, 1990).
1910:
1911: \item{{\bf 71.}} {\smcap M.\ Knopp} and {\smcap G.\ Mason}, `Generalized
1912: modular forms', {\it J.\ Number Theory} 99 (2003) 1--28.
1913:
1914: \item{{\bf 72.}} {\smcap D.N.\ Kozlov}, `On completely replicable functions and extremal
1915: poset theory', MSc thesis, Univ.\ of Lund, Sweden, 1994.
1916:
1917: \item{{\bf 73.}} {\smcap R.\ Kultze}, `Elliptic genera and the moonshine module',
1918: {\it Math.\ Z.} 223 (1996) 463--471.
1919:
1920: \item{{\bf 74.}} {\smcap M.-L.\ Lang}, `On a question raised by Conway--Norton',
1921: {\it J. Math.\ Soc.\ Japan} 41 (1989) 263--284.
1922:
1923: \item{{\bf 75.}} {\smcap S.\ Lang}, {\it Elliptic Functions}, 2nd edn
1924: (Springer, New York, 1997).
1925:
1926: \item{{\bf 76.}} {\smcap J.\ Lepowsky}, `Euclidean Lie algebras and the modular
1927: function $j$', {\it The Santa Cruz Conference on Finite Groups},
1928: {\it Proc.\ Sympos.\ Pure Math.} 37 (Amer.\ Math.\ Soc., Providence, 1980)
1929: 567--570.
1930:
1931: \item{{\bf 77.}} {\smcap J.\ Lepowsky}, `Vertex operator algebras and the zeta
1932: function', {\it Recent Developments in Quantum Affine Algebras and Related Topics,
1933: Contemp.\ Math.} 248 (Amer.\ Math.\ Soc., Providence, 1999) 327--340.
1934:
1935: \item{{\bf 78.}} {\smcap J.\ Lepowsky}, `The work of Richard E.\ Borcherds',
1936: {\it Notices Amer.\ Math.\ Soc.} 46 (1999) 17--19.
1937:
1938: \item{{\bf 79.}} {\smcap J.\ Lepowsky} and {\smcap H.\ Li}, {\it Introduction
1939: to Vertex Operator Algebras and their Representations} (Birkh\"auser, Boston,
1940: 2004).
1941:
1942: \item{{\bf 80.}} {\smcap B.H.\ Lian} and {\smcap S.-T.\ Yau}, `Arithmetic
1943: properties of mirror map and quantum coupling', {\it Commun.\ Math.\ Phys.}
1944: 176 (1996) 163--191.
1945:
1946: \item{{\bf 81.}} {\smcap B.H.\ Lian} and {\smcap G.J.\ Zuckerman}, `Moonshine
1947: cohomology',
1948: {\it Moonshine and Vertex Operator Algebras}
1949: (Surikaisekikenkyusho Kokyuroku No.\ 904, Kyoto, 1995) 87--115.
1950:
1951: \item{{\bf 82.}} {\smcap K.\ Liu}, `On modular invariance and rigidity theorems',
1952: {\it J.\ Diff.\ Geom.} 41 (1995) 343--396.
1953:
1954: \item{{\bf 83.}} {\smcap I.G.\ Macdonald}, `Affine root systems and Dedekind's
1955: $\eta$-function', {\it Invent.\ Math.} 15 (1972) 91--143.
1956:
1957: \item{{\bf 84.}} {\smcap M.\ Mahowald} and {\smcap M.\ Hopkins},
1958: `The structure of 24 dimensional manifolds having normal bundles which lift to $B{\rm O}[8]$',
1959: {\it Recent Progress in Homotopy Theory, Contemp.\ Math.} 293
1960: (Amer.\ Math.\ Soc., Providence, 2002) 89--110.
1961:
1962: \item{{\bf 85.}} {\smcap F.\ Malikov, V.\ Schechtman} and {\smcap A.\ Vaintrob},
1963: `Chiral de Rham complex', {\it Commun.\ Math.\ Phys.} {\bf 204} (1999) 439--473.
1964:
1965: \item{{\bf 86.}} {\smcap F.\ Malikov} and {\smcap V.\ Schechtman}, `Deformations
1966: of vertex algebras, quantum cohomology of toric varieties, and elliptic
1967: genus', {\it Commun.\ Math.\ Phys.} 234 (2003) 77--100.
1968:
1969: \item{{\bf 87.}} {\smcap G.\ Mason}, `Finite groups and modular functions', {\it
1970: The Arcata Conference on Representations of Finite Groups, Proc.\ Sympos.\
1971: Pure Math.} 47 (Amer.\ Math.\ Soc., Providence, 1987) 181--209.
1972:
1973: \item{{\bf 88.}} {\smcap G.\ Mason},
1974: `On a system of elliptic modular forms attached to the large Mathieu group',
1975: {\it Nagoya Math. J.} 118 (1990) 177--193.
1976:
1977: \item{{\bf 89.}} {\smcap J.\ McKay}, `Graphs, singularities, and finite groups',
1978: {\it The Santa Cruz Conference on Finite Groups,
1979: Proc.\ Sympos.\ Pure Math.} 37 (Amer.\ Math.\ Soc., Providence, 1980) 183--186.
1980:
1981: \item{{\bf 90.}} {\smcap J.\ McKay}, `The essentials of Monstrous Moonshine', {\it Groups and
1982: Combinatorics -- in Memory of M.\ Suzuki, Adv.\ Studies
1983: in Pure Math.} 32 (Math.\ Soc.\ Japan, Tokyo, 2001) 347--353.
1984:
1985: \item{{\bf 91.}} {\smcap J.\ McKay} and {\smcap H.\ Strauss}, `The q-series
1986: of monstrous moonshine and the decomposition of the head characters', {\it
1987: Commun.\ Alg.} 18 (1990) 253--278.
1988:
1989: \item{{\bf 92.}} {\smcap H.\ McKean} and {\smcap V.\ Moll}, {\it Elliptic Curves:
1990: Function Theory, Geometry, Arithmetic} (Cambridge Univ.\ Press,
1991: 1999).
1992:
1993: \item{{\bf 93.}} {\smcap M.\ Miyamoto}, `21 involutions acting on the Moonshine
1994: module', {\it J.\ Alg.} 175 (1995) 941--965.
1995:
1996: \item{{\bf 94.}} {\smcap G.W.\ Moore}, `Les Houches lectures on strings and
1997: arithmetic', Preprint (arXiv: hep-th/0401049).
1998:
1999: \item{{\bf 95.}} {\smcap D.R.\ Morrison}, `Picard--Fuchs equations and mirror
2000: maps for hypersurfaces', {\it Essays on Mirror Manifolds} (ed.\ S.-T.\ Yau,
2001: Intern.\ Press, Hong Kong, 1992) 241--264.
2002:
2003: \item{{\bf 96.}} {\smcap S.P.\ Norton}, `More on Moonshine', In: {\it Computational
2004: Group Theory} (ed.\ M.D.\ Atkinson, Academic Press, New York, 1984) 185--193.
2005:
2006: \item{{\bf 97.}} {\smcap S.P.\ Norton}, `Generalized moonshine', {\it The Arcata Conference on
2007: Representations of Finite Groups, Proc.\ Sympos.\ Pure Math.}
2008: 47 (Amer.\ Math.\ Soc., Providence, 1987) 208--209.
2009:
2010: \item{{\bf 98.}} {\smcap S.P.\ Norton}, `Constructing the Monster', {\it
2011: Groups, Combinatorics, and Geometry} (ed.\ M.W.\ Liebeck and J.\ Saxl,
2012: Cambridge Univ.\ Press, 1992) 63--76.
2013:
2014: \item{{\bf 99.}} {\smcap S.P.\ Norton}, `From moonshine to the Monster', {\it Proc.\
2015: on Moonshine and Related Topics} (Amer.\ Math.\ Soc, Providence, 2001)
2016: 163--171.
2017:
2018: \item{{\bf 100.}} {\smcap A.\ Ogg}, {\it Modular forms and Dirichlet series}
2019: (W. A. Benjamin, New York, 1969).
2020:
2021:
2022: \item{{\bf 101.}} {\smcap L.\ Queen}, `Modular functions arising from some finite groups',
2023: {\it Math.\ of Comput.} 37 (1981) 547--580.
2024:
2025: \item{{\bf 102.}} {\smcap U.\ Ray}, `Generalized Kac--Moody algebras
2026: and some related topics', {\it Bull.\ Amer.\ Math.\ Soc.} 38 (2000) 1--42.
2027:
2028: \item{{\bf 103.}} {\smcap W.F.\ Reynolds}, `Thompson's characterization of
2029: characters and sets of primes', {\it J.\ Alg.} 156 (1993) 237--243.
2030:
2031: \item{{\bf 104.}} {\smcap S.-S.\ Roan}, `Mirror symmetry of elliptic curves
2032: and Ising model', {\it J.\ Geom.\ Phys.} 20 (1996) 273--296.
2033:
2034: \item{{\bf 105.}} {\smcap A.J.E.\ Ryba}, `Modular moonshine?', {\it Moonshine, the Monster,
2035: and Related Topics, Contemp.\ Math.} 193 (Amer.\ Math.\ Soc., Providence, 1996)
2036: 307--336.
2037:
2038: \item{{\bf 106.}} {\smcap A.N.\ Schellekens}, `Meromorphic $c=24$ conformal
2039: field theories', {\it Commun.\ Math.\ Phys.} 153 (1993) 159--185.
2040:
2041: \item{{\bf 107.}} {\smcap G.\ Segal}, `The definition of conformal field theory',
2042: {\it Differential Geometric Methods in Theoretical Physics} (ed.\ K.\ Bleuler and
2043: M.\ Werner, Academic Press, Boston, 1988) 165--171.
2044:
2045: \item{{\bf 108.}} {\smcap G.\ Segal}, `Elliptic cohomology', {\it S\'eminaire
2046: Bourbaki 1987-88, no.\ 695, Ast\'erisque} 161-162 (1988) 187--201.
2047:
2048: \item{{\bf 109.}} {\smcap G.W.\ Smith}, `Replicant powers for higher genera',
2049: {\it Moonshine, the Monster, and Related Topics, Contemp.\ Math.}
2050: 193 (Amer.\ Math.\ Soc., Providence, 1996) 337--352.
2051:
2052: \item{{\bf 110.}} {\smcap S.D.\ Smith}, `On the head characters of the monster
2053: simple group',
2054: {\it Finite Groups -- Coming of Age} (ed.\ J.\ McKay, Amer.\ Math.\ Soc.,
2055: Providence, 1985) 303--313.
2056:
2057: \item{{\bf 111.}} {\smcap H.\ Tamanoi}, {\it Elliptic Genera and Vertex Operator
2058: Superalgebras, Lecture Notes in Mathematics 1704} (Springer, Berlin, 1999).
2059:
2060: \item{{\bf 112.}} {\smcap C.B.\ Thomas}, {\it Elliptic Cohomology} (Kluwer, New York, 1999).
2061:
2062: \item{{\bf 113.}} {\smcap J.G.\ Thompson}, `Finite groups and modular functions',
2063: {\it Bull.\ London Math.\ Soc.} 11 (1979) 347--351.
2064:
2065: \item{{\bf 114.}} {\smcap J.G.\ Thompson}, `Some numerology between the Fischer--Griess
2066: Monster and the elliptic modular function', {\it Bull.\ Lond.\ Math.\ Soc.}
2067: 11 (1979) 352--353.
2068:
2069: \item{{\bf 115.}} {\smcap J.G.\ Thompson}, `A finiteness theorem for subgroups
2070: of PSL(2,$\R$) which are commensurable with PSL($2,\Z)$', {\it Santa Cruz
2071: Conference on Finite Groups, Proc.\ Symp.\ Pure Math.} 37 (Amer.\ Math.\
2072: Soc., Providence, 1980) 533--555.
2073:
2074: \item{{\bf 116.}} {\smcap M.P.\ Tuite}, `On the relationship between Monstrous Moonshine
2075: and the uniqueness of the Moonshine module', {\it Commun.\ Math.\ Phys.} 166 (1995)
2076: 495--532.
2077:
2078: \item{{\bf 117.}} {\smcap M.P.\ Tuite}, `Genus two meromorphic conformal field
2079: theories', {\it Proc.\ on Moonshine and Related Topics} (Amer.\ Math.\ Soc.,
2080: Providence, 2001) 231--251.
2081:
2082: \item{{\bf 118.}} {\smcap H.\ Verrill} and {\smcap N.\ Yui}, `Thompson series
2083: and the mirror maps of pencils of K3 surfaces', {\it The Arithmetic and Geometry
2084: of Algebraic Cycles} (ed.\ B.\ Gordon et al, Centre Res.\ Math.\ Proc.\ and
2085: Lecture notes 24, 2000) 399--432.
2086:
2087: \item{{\bf 119.}} {\smcap E.\ Witten}, `Physics and geometry', {\it Proc.\
2088: Intern.\ Congr.\ Math.\ 1986 (Berkeley)} (Amer.\ Math.\ Soc., Providence, 1987).
2089:
2090: \item{{\bf 120.}} {\smcap E.\ Witten}, `Elliptic genera and quantum field theory',
2091: {\it Commun.\ Math.\ Phys.} 109 (1987) 525--536.
2092:
2093: \item{{\bf 121.}} {\smcap E.\ Witten}, `Quantum field theory, Grassmannians,
2094: and algebraic curves', {\it Commun.\ Math.\ Phys.} 113 (1988) 529--600.
2095:
2096: \item{{\bf 122.}} {\smcap E.\ Witten}, `Geometry and quantum field theory',
2097: {\it Mathematics Into the Twenty-first Century}, Vol.II (Amer.\ Math.\ Soc.,
2098: Providence, 1992) 479--491.
2099:
2100: \item{{\bf 123.}} {\smcap Y.\ Zhu}, `Global vertex operators on Riemann
2101: surfaces', {\it Commun.\ Math.\ Phys.} 165 (1994) 485--531.
2102:
2103: \item{{\bf 124.}} {\smcap Y.\ Zhu}, `Modular invariance of characters of vertex operator
2104: algebras', {\it J.\ Amer.\ Math.\ Soc.} 9 (1996) 237--302.
2105:
2106: \bigskip\bigskip{\it Terry Gannon}
2107:
2108: {\it Department of Mathematics}
2109:
2110: {\it University of Alberta}
2111:
2112: {\it Edmonton, CANADA T6G 2G1}
2113:
2114: \medskip tgannon@math.ualberta.ca
2115:
2116: \end
2117:
2118: