1:
2: % Final draft -- last modified 2-2-05, 7 am
3:
4: % Revised 6-6-05, 8 pm
5:
6: \documentclass[12pt]{article}
7: \oddsidemargin 0in
8: \textwidth 6.5in
9: \topmargin 0in
10: \headheight 0in
11: \textheight 8.5in
12: \parskip 2ex
13:
14: \usepackage{amsmath,amssymb,latexsym,psfig}
15: \usepackage{color}
16: \usepackage[all]{xy}
17:
18:
19: % The next two definitions allow me to use squares in
20: % discussions of twisted sectors.
21: % They were copied verbatim from Blumenhagen-Sethi.
22:
23:
24: \def\sqr#1#2{{\vcenter{\vbox{\hrule height.#2pt
25: \hbox{\vrule width.#2pt height#1pt \kern#1pt
26: \vrule width.#2pt}\hrule height.#2pt}}}}
27:
28: \def\square
29: {\mathop{\mathchoice{\sqr{12}{15}}{\sqr{9}{12}}
30: {\sqr{6.3}{9}}{\sqr{4.5}{9}}}}
31: %%%%%%%%%%%%%%
32:
33: \begin{document}
34:
35: \hfill hep-th/0502027
36:
37: \vspace{0.5in}
38:
39: \begin{center}
40:
41:
42: {\large\bf Notes on gauging noneffective group actions}
43:
44: \vspace{0.25in}
45:
46: Tony Pantev$^1$ and Eric Sharpe$^{2}$ \\
47: $^1$ Department of Mathematics \\
48: University of Pennsylvania \\
49: David Rittenhouse Lab.\\
50: 209 South 33rd Street \\
51: Philadelphia, PA 19104-6395\\
52: $^2$ Departments of Physics, Mathematics \\
53: University of Utah \\
54: Salt Lake City, UT 84112 \\
55: {\tt tpantev@math.upenn.edu},
56: {\tt ersharpe@math.utah.edu} \\
57:
58:
59: $\,$
60:
61:
62: \end{center}
63:
64:
65: In this paper we study sigma models in which a
66: noneffective group
67: action has been gauged. Such gauged sigma
68: models turn out to be different from gauged sigma models in which an
69: effectively-acting group is gauged, because of nonperturbative effects
70: on the worldsheet. We concentrate on finite noneffectively-acting
71: groups, though we also outline how analogous phenomena also happen in
72: nonfinite noneffectively-acting groups. We find that understanding
73: deformations along twisted sector moduli in these theories leads one
74: to new presentations of CFT's, defined by fields valued in roots of unity.
75:
76: \begin{flushleft}
77: January 2005
78: \end{flushleft}
79:
80: \newpage
81:
82: \tableofcontents
83:
84: \newpage
85:
86:
87: \section{Introduction}
88:
89: In this paper we shall collect some results on the physics of gauged
90: sigma models in which a noneffectively-acting group has been gauged.
91: By ``noneffectively-acting,'' we mean that some nontrivial elements of
92: the group act trivially, {\it i.e.} $g \cdot x = x$ for all $x$, for
93: some $g$ other than the identity. Such trivially-acting elements form
94: a normal subgroup, call it $K$, of the gauge group $G$, and so as
95: $G/K$ is a group, the reader might suspect that a $G$-gauging would be
96: physically equivalent to a $G/K$ gauging. However, that is not the
97: case, as we shall see in numerous examples. Gauging $G$ is a distinct
98: physical operation from gauging $G/K$, because of
99: nonperturbative effects.
100:
101: In this paper, we shall concentrate on understanding finite
102: noneffectively-acting groups. We will briefly outline how analogous
103: phenomena also happen when gauging nonfinite noneffectively-acting
104: groups, but a more extensive discussion will appear in \cite{tonyme,glsm}.
105: We will discuss massless spectrum computations in such orbifolds,
106: which have features that make them a bit more subtle than ordinary
107: orbifolds by effectively-acting groups. We shall also discuss various
108: other technical issues in noneffective orbifolds, such as possible
109: D-branes, and, to a limited extent, mirror symmetry. (Mirror symmetry
110: in noneffective gaugings will be discussed much more extensively in
111: \cite{glsm}.) Curiously, we shall see that twist fields associated to
112: trivially-acting group elements are often equivalent to fields valued
113: in roots of unity, a fact which will play an important part in the
114: sequels \cite{tonyme,glsm}, where we will rederive the same
115: description from completely independent lines of reasoning.
116:
117: Part of the purpose of this paper is to lay part of the physical
118: groundwork for the upcoming papers \cite{tonyme,glsm}, which will
119: describe what it means to compactify a string on a general Calabi-Yau
120: stack. In a nutshell, under some mild conditions,
121: every stack\footnote{When we speak of stacks,
122: we will always assume the stacks are smooth,
123: complex, algebraic, Deligne-Mumford stacks. This means, for example,
124: that all $G$-actions will be assumed to have finite stabilizers, throughout
125: this paper as well as \cite{tonyme,glsm}. We will explain the mild conditions
126: alluded to above in \cite{tonyme}.} has a
127: presentation of the form $[X/G]$ for some manifold $X$ and some group
128: $G$ with an action on $X$, where $G$ need not be finite and need not
129: act effectively. To such a presentation $[X/G]$, one associates a
130: $G$-gauged sigma model on $X$. Thus, studying string
131: compactifications on stacks boils down to studying gauged sigma
132: models. The first important point is that a stack can have many
133: presentations of the form $[X/G]$, which can define very different
134: gauged sigma models. For example, if $G$ is finite in one
135: presentation and nonfinite in another, then in the first presentation,
136: the gauged sigma model is a CFT, whereas typically in the second
137: presentation, the gauged sigma model will not be a CFT. Thus, stacks
138: cannot classify gauged sigma models; rather, the most one can hope
139: for is that universality classes\footnote{These are
140: equivalence classes of gauged sigma models, where two such models are
141: declared equivalent if they are related by worldsheet RG flow.} of
142: gauged sigma models are classified by stacks. Such a claim cannot be
143: checked directly, but numerous indirect tests are possible, as we
144: shall describe in \cite{tonyme,glsm}. In those papers we also resolve
145: various obstacles to the consistency of this claim,
146: perhaps most importantly the mismatch between physical CFT
147: deformations and mathematical deformations of stacks. Finally, in
148: order to understand both the resolution of the puzzles posed by
149: deformation theory, as well as mirror symmetry, one is led to study
150: new presentations of abstract CFT's defined by fields valued in roots of
151: unity.
152:
153: In this paper, we shall concentrate on the physics of gauging
154: noneffective (and primarily finite) group actions, though we shall
155: occasionally use the language of stacks to help make contact with the
156: sequels \cite{tonyme,glsm}. Among other things, in this paper we
157: will see that twist fields associated to trivially-acting group elements
158: are equivalent to fields valued in roots of unity; we will also recover
159: such fields from completely independent lines of reasoning in
160: \cite{tonyme,glsm}.
161:
162:
163:
164: We begin in section~\ref{basicexs} with a discussion of several
165: examples of gauged finite noneffectively-acting groups. We explicitly
166: compute that gauging a noneffectively-acting group is distinct from
167: gauging an effectively-acting one, and also check
168: that these noneffective gaugings are consistent -- the theories are
169: modular-invariant, for example. Mathematically, the types of stacks
170: associated with noneffective gaugings are known as `gerbes,' and since
171: we will be using the language of stacks in \cite{tonyme,glsm}, we
172: relate our examples to that mathematical language.
173:
174: In section~\ref{nonmincharge} we briefly outline analogous phenomena
175: in gauging noneffective nondiscrete groups. A much more extensive
176: discussion of such gaugings, together with numerous examples and
177: computations, will appear in \cite{tonyme,glsm}; for the purposes of
178: this paper, we merely point out the existence of analogous phenomena
179: there.
180:
181: In section~\ref{spectra} we discuss massless spectrum computations in
182: orbifolds by noneffectively-acting finite groups. We find that the
183: massless spectrum has the same general form as for finite
184: effectively-acting groups, {\it i.e.} one twisted sector for each
185: conjugacy class, even if the elements of that class act trivially.
186: However, the reasoning behind this result is a bit subtle, and since
187: we have not seen a detailed explanation of massless spectra in
188: noneffective orbifolds in the physics literature previously, we spend
189: a great deal of time discussing potentially confusing issues. One of
190: the more important issues is understanding the physical infinitesimal
191: deformations dictated by the results of the massless spectrum
192: computation. In typical cases, the noneffective orbifold has more
193: (unobstructed) moduli fields than its effectively-acting
194: counterpart, but the only ones that have a clear geometric
195: meaning are that subset in the effectively-acting theory. This
196: issue can be restated more formally as a mismatch between the number
197: of physical moduli of gauged sigma models and the number of
198: mathematical moduli of the stack. As the physical moduli are, by
199: definition, part of the massless spectrum, it is very important to
200: understand this issue to properly understand the massless spectrum.
201:
202: In section~\ref{defthy} we discuss this issue and outline how resolving it
203: leads to new presentations of CFT's. In \cite{tonyme} such deformation theory
204: issues will play a much more important role, and will be discussed much
205: more extensively.
206:
207: One prerequisite for the deformation theory discussion is to rewrite
208: twist fields for trivially-acting group elements in a
209: different-looking fashion. In section~\ref{trivgerbesection} we
210: discuss how such twist fields are equivalent to fields valued in roots
211: of unity, which gives us a very algebraic description of many of the
212: twisted sector fields. Such a description is impossible for a twist
213: field associated to a nontrivially-acting group element, but
214: trivially-acting group elements are special in this regard. We also
215: take the oppotunity in this section to compare an orbifold of a space
216: $X$ by a trivially-acting ${\bf Z}_k$ to the CFT of a sigma model on
217: $k$ disjoint copies of $X$. The two theories are distinct, but do
218: share certain features.
219:
220: In section~\ref{Dbranenoneff} we discuss D-branes in noneffective orbifolds.
221: Even if a group element acts trivially on the space,
222: it can still act nontrivially on the Chan-Paton factors, as this is
223: consistent with the Cardy condition.
224:
225: In section~\ref{mirrors} we briefly discuss mirror symmetry in the special
226: case of noneffective orbifolds in which the entire orbifold group acts
227: trivially. We will discuss mirror symmetry for gauged sigma models much more
228: extensively in \cite{glsm}.
229:
230:
231: The noneffective gaugings we discuss in this paper all have
232: multiple dimension zero operators in their spectru, which signals
233: a failure of cluster decomposition. Such a failure is not fatal
234: in two-dimensional conformal field theories, however,
235: as for example a sigma model on a disjoint union of spaces also
236: has multiple dimension zero operators, and so also fails cluster
237: decomposition. These issues are not unrelated;
238: in the followup work \cite{clusterdecomp} we shall argue that
239: the conformal field theories obtained by these noneffective gaugings
240: are equivalent to conformal field theories describing disjoint unions
241: of spaces.
242:
243:
244:
245:
246:
247:
248:
249:
250: \section{Examples of global quotients by finite
251: noneffectively-acting groups} \label{basicexs}
252:
253:
254: Consider an orbifold of a space $X$ by a group $G$,
255: containing elements that act trivially. As
256: mentioned in the introduction, those elements form a normal subgroup
257: of $G$, call it $K$. Now, the reader might at first glance suspect
258: that gauging $G$ would be physically equivalent to gauging $G/K$, but
259: we shall see in examples in this section that this
260: is not the case, by computing one-loop partition functions (and,
261: incidentally, checking modular invariance). Gauging a
262: noneffectively-acting group is not the same as gauging an
263: effectively-acting group.
264:
265:
266: In \cite{tonyme} we will give such orbifolds an alternative
267: interpretation, as examples of sigma models\footnote{To be precise,
268: a sigma model is defined on a presentation of a stack, not precisely on
269: the stack itself. In the discussion in the paragraph above, each orbifold
270: is canonically associated with a particular presentation, and sigma models
271: are defined on those presentations.} on Calabi-Yau gerbes.
272: A gerbe is, in essence, a local orbifold by a trivially-acting group,
273: and can be presented as global quotients by larger
274: noneffectively-acting groups.
275: A ``Calabi-Yau gerbe'' is a gerbe that can be presented as
276: a quotient of a Calabi-Yau by a $G$-action that preserves
277: the holomorphic top-form.
278: If $K$ lies in the center of $G$, we say the gerbe is `banded.'
279: If $K$ does not lie in the center of $G$, we say the gerbe is
280: `non-banded.' We shall see that this banded versus non-banded distinction
281: is reflected in the one-loop partition functions.
282:
283:
284:
285:
286:
287:
288:
289:
290:
291: \subsubsection{First example: trivially-acting ${\bf Z}_2$ center}
292: \label{ex:d4elliptic}
293:
294: We shall begin by considering a family of examples in which
295: the trivially-acting subgroup is $K = {\bf Z}_2$,
296: and with the
297: property that $K$ lies in the center of $G$.
298: To be specific, take $G = D_4$, the eight-element dihedral group,
299: which is
300: a nonabelian group that can be described
301: as a\footnote{Nontrivial central extensions of
302: ${\bf Z}_2 \times {\bf Z}_2$ by ${\bf Z}_2$
303: are not unique. For example, the quaternions define another
304: eight-element nonabelian group, which is such
305: a central extension. The quaternions and $D_4$ are not isomorphic,
306: as can be checked by {\it e.g.} comparing the orders of their elements.}
307: nontrivial central extension of ${\bf Z}_2 \times {\bf Z}_2$
308: by ${\bf Z}_2$:
309: \begin{equation} \label{Hext}
310: 1 \: \longrightarrow \: {\bf Z}_2 \: \longrightarrow \: D_4 \:
311: \longrightarrow \:
312: {\bf Z}_2 \times {\bf Z}_2 \: \longrightarrow \: 1
313: \end{equation}
314: We can alternately
315: describe $D_4$ as upper-triangular
316: $3 \times 3$ matrices with $1$'s on the diagonal and the strictly
317: upper-triangular elements in ${\bf F}_2$.
318: Define
319: \begin{displaymath}
320: z \: = \: \left[ \begin{array}{ccc}
321: 1 & 0 & 1 \\
322: 0 & 1 & 0 \\
323: 0 & 0 & 1 \end{array} \right], \: \:
324: a \: = \: \left[ \begin{array}{ccc}
325: 1 & 1 & 0 \\
326: 0 & 1 & 0 \\
327: 0 & 0 & 1 \end{array} \right], \: \:
328: b \: = \: \left[ \begin{array}{ccc}
329: 1 & 1 & 1 \\
330: 0 & 1 & 1 \\
331: 0 & 0 & 1 \end{array} \right]
332: \end{displaymath}
333: It is straightforward to check that everything commutes with $z$
334: (in fact, the image of ${\bf Z}_2$ in $D_4$ is $\{ I, z \}$ where
335: $I$ denotes the $3 \times 3$ identity matrix), and that
336: \begin{displaymath}
337: z^2 \: = \: a^2 \: = I, \: \: b^2 \: = \: z, \: \:
338: ba \: = \: abz
339: \end{displaymath}
340: The eight group elements are
341: \begin{displaymath}
342: I, \: z, \: a, \: b, \: ab, \: az, \: bz, \: ba
343: \end{displaymath}
344: and $H$ has five conjugacy classes, given by
345: \begin{displaymath}
346: \{ I \}, \: \{ z \}, \: \{ a, az \}, \: \{ b, bz \}, \:
347: \{ab, ba \}
348: \end{displaymath}
349:
350: Let $D_4$ act on a manifold $X$ by first projecting to ${\bf Z}_2 \times
351: {\bf Z}_2$, and then letting the ${\bf Z}_2 \times {\bf Z}_2$ act,
352: so that the ${\bf Z}_2$ center acts trivially. Gauging the $D_4$ action
353: means that we must sum over principal $D_4$ bundles on the worldsheet,
354: so that, for example, the one-loop partition function of a $D_4$ gauged
355: sigma model has the same form as if the $D_4$ were acting effectively:
356: \begin{displaymath}
357: Z(D_4) \: = \: \frac{1}{|D_4|} \sum_{g,h \in D_4, gh=hg} Z_{g,h}
358: \end{displaymath}
359:
360:
361:
362: How does this string orbifold $[ X/D_4 ]$ compare to the string orbifold
363: defined by the stack $[X/ ({\bf Z}_2 \times {\bf Z}_2)]$?
364: It is straightforward to check that, for each twisted sector
365: of $[X/ ({\bf Z}_2 \times {\bf Z}_2)]$ there are $| {\bf Z}_2 |^2 = 4$
366: twisted sectors
367: in $[X/D_4]$ with the same boundary conditions on the fields,
368: {\it except} for the\footnote{Each one-loop twisted sector is defined by
369: an equivalence class of principal $G$-bundles over an elliptic curve,
370: or, equivalently, a pair of commuting group elements. We denote such
371: one-loop twisted sectors by a square with the commuting group elements on the
372: sides, corresponding to a representation of the elliptic curve as a square
373: with sides identified, and marking how the commuting group elements appear.
374: }
375: \begin{displaymath}
376: {\scriptstyle a} \square_b \, , \: \: \:
377: {\scriptstyle a} \square_{ab} \, , \: \: \:
378: {\scriptstyle b} \square_{ab}
379: \end{displaymath}
380: twisted sectors
381: of $[X/ ({\bf Z}_2 \times {\bf Z}_2)]$.
382: Since there is no way to lift those pairs of group elements to
383: commuting pairs of group elements in $D_4$, there are no corresponding
384: one-loop twisted sectors. The one-loop partition functions of the two theories
385: are related as
386: \begin{displaymath}
387: Z(D_4) \: = \: \frac{ | {\bf Z}_2 \times {\bf Z}_2 | }{ |D_4|} |{\bf Z}_2|^2
388: \left[ Z({\bf Z}_2 \times {\bf Z}_2 ) \: - \:
389: \left( \mbox{some twisted sectors} \right) \right]
390: \end{displaymath}
391: Moreover, it is easy to check in examples that the omitted one-loop
392: twisted sectors are nonzero in general.
393: Thus, this physical theory is
394: distinct from the orbifold $[X/({\bf Z}_2 \times {\bf Z}_2)]$,
395: with a manifestly different partition function.
396: Gauging the noneffectively-acting
397: $D_4$ is {\it not} the same as gauging the effectively-acting
398: $D_4/{\bf Z}_2 = {\bf Z}_2 \times {\bf Z}_2$.
399:
400: Omitting twisted sectors from a string orbifold partition function
401: runs the risk of destroying modular invariance. After all,
402: in a one-loop partition function,
403: \begin{displaymath}
404: \left[ \begin{array}{cc}
405: m & n \\
406: p & q \end{array} \right] \: \in \: SL(2, {\bf Z})
407: \end{displaymath}
408: sends the
409: \begin{displaymath}
410: {\scriptstyle g} \square_h
411: \end{displaymath}
412: twisted sector to the
413: \begin{displaymath}
414: {\scriptstyle g^m h^n} \square_{g^p h^q}
415: \end{displaymath}
416: twisted
417: sector.
418:
419: In the case of the $[X / ({\bf Z}_2 \times {\bf Z}_2)]$ orbifold, however,
420: it is nonetheless possible to omit some of the twisted sectors without
421: destroying modular invariance, and such a truncation is precisely what we
422: have obtained from our noneffective orbifold. The omitted twisted sectors
423: precisely fill an $SL(2,{\bf Z})$ orbit.
424: None of the remaining
425: twisted sectors can be mapped into the omitted twisted sectors
426: by the $SL(2,{\bf Z})$ action, so our $[X/D_4]$ orbifold is modular-invariant.
427: In hindsight, the fact that modular invariance is not broken should
428: not surprise us. If we think of the orbifold as a $[X/D_4]$ quotient,
429: and keep track of the non-effectively-acting part of the group,
430: then the partition function is manifestly modular invariant -- we sum
431: over all commuting pairs of elements of $D_4$. It is only when we try
432: to think of the orbifold in terms of some operation on a ${\bf Z}_2 \times
433: {\bf Z}_2$ orbifold that modular invariance becomes more obscure.
434: More generally, whenever one has an $[X/G]$ quotient for $G$ finite
435: but not necessarily effectively-acting, the one-loop partition function
436: will contain copies of some of the twisted sectors in an $[X/H]$
437: orbifold for some effectively-acting $H$,
438: and the multiplicities between different one-loop
439: twisted sectors might vary, but the resulting partition function will
440: always be modular invariant.
441:
442:
443:
444:
445:
446: So far we have not specified the manifold $X$,
447: but examples are easy to construct. One well-known example of a space
448: with a ${\bf Z}_2 \times {\bf Z}_2$ action is $T^6$.
449: One can then define a $D_4$ action on $T^6$ by first projecting to
450: ${\bf Z}_2 \times {\bf Z}_2$, and then letting the
451: ${\bf Z}_2 \times {\bf Z}_2$ act.
452:
453: We can also construct examples in which the ${\bf Z}_2 \times {\bf Z}_2$
454: acts freely. For example, ${\bf Z}_2 \times {\bf Z}_2$ has a free
455: action on $T^2$, defined by translations by the
456: 2-torsion points on $T^2$, \cite[section II.1]{silvtate}
457: ${\bf Z}_2 \times {\bf Z}_2$. These points are sketched in
458: figure~\ref{2tr}.
459:
460:
461: \begin{figure}
462: \centerline{\psfig{file=2tor.eps,width=2in}}
463: \caption{\label{2tr} The four two-torsion points on an elliptic curve.}
464: \end{figure}
465:
466: We have previously mentioned that such quotients by noneffectively-acting
467: groups are examples of gerbes. Since we will be using that language
468: more extensively in the follow-up works \cite{tonyme,glsm},
469: let us explore what that means in the
470: present example.
471: Since the ${\bf Z}_2$ subgroup acts trivially, the quotient is a
472: ${\bf Z}_2$-gerbe over $T^2/ ({\bf Z}_2 \times {\bf Z}_2)$,
473: {\it i.e.} a local orbifold by a trivially-acting ${\bf Z}_2$,
474: and it can be shown that it is a nontrivial ${\bf Z}_2$ gerbe.
475: Also, since the ${\bf Z}_2$ lies in the center of $D_4$, this is
476: a banded ${\bf Z}_2$ gerbe, and since the group action preserves
477: the holomorphic top-form, it is a Calabi-Yau banded ${\bf Z}_2$ gerbe.
478:
479: It is instructive to note that this is a nontrivial
480: gerbe.
481: Let $X$ be any Calabi-Yau, with an action of ${\bf Z}_2 \times
482: {\bf Z}_2$. Then, in particular, $X$ is a principal
483: ${\bf Z}_2 \times {\bf Z}_2$ bundle over the stack
484: $[ X / ( {\bf Z}_2 \times {\bf Z}_2 ) ]$, and as such,
485: is classified by an element
486: \begin{displaymath}
487: \xi \: \in \: H^1\left( [ X/ ( {\bf Z}_2 \times {\bf Z}_2 ) ],
488: {\bf Z}_2 \times {\bf Z}_2 \right)
489: \end{displaymath}
490: The short exact sequence~(\ref{Hext}) induces a long exact sequence
491: containing a map
492: \begin{displaymath}
493: H^1\left( [ X/ ( {\bf Z}_2 \times {\bf Z}_2 ) ],
494: {\bf Z}_2 \times {\bf Z}_2 \right)
495: \: \longrightarrow \:
496: H^2\left( [ X/ ( {\bf Z}_2 \times {\bf Z}_2 ) ],
497: {\bf Z}_2 \right)
498: \end{displaymath}
499: and the image of $\xi$ under this map is the characteristic
500: class of the gerbe we are currently interested in.
501: This characteristic class of the gerbe will be trivial
502: if and only if $\xi$ is in the image of the map
503: \begin{equation} \label{eq:mapD4}
504: H^1\left( [ X/ ( {\bf Z}_2 \times {\bf Z}_2 ) ],
505: D_4 \right)
506: \: \longrightarrow \:
507: H^1\left( [ X/ ( {\bf Z}_2 \times {\bf Z}_2 ) ],
508: {\bf Z}_2 \times {\bf Z}_2 \right)
509: \end{equation}
510: {\it i.e.} if and only if the principal ${\bf Z}_2 \times {\bf Z}_2$
511: bundle $X$ lifts to a principal $D_4$ bundle over $[X/\left( {\bf Z}_2 \times
512: {\bf Z}_2\right)]$. In the present case,
513: ${\bf Z}_2 \times {\bf Z}_2$ acts freely and
514: $Y:= [X/\left({\bf Z}_2 \times {\bf Z}_2\right)]$ is
515: a smooth elliptic
516: curve.
517: But it is easy to see that there are no such
518: principal $D_4$ bundles. Every such $D_{4}$ bundle
519: corresponds to a
520: representation $\rho : \pi_{1}(Y) \to D_{4}$, such that
521: $\operatorname{im}(\rho) \subset D_{4}$ surjects onto ${\bf Z}_2
522: \times {\bf Z}_2$. However $D_{4}$ does not contain any abelian
523: subgroups that surject onto $D_{4}$. Thus
524: $\xi$ cannot be in the image of the map \eqref{eq:mapD4}, and so the
525: characteristic
526: class must be nontrivial.
527:
528:
529:
530:
531: \subsubsection{Another set of ${\bf Z}_2$ gerbes}
532:
533: Another example of a ${\bf Z}_2$ gerbe can be built
534: from an orbifold $[X/D_8]$ where the ${\bf Z}_2$ center of
535: $D_8$ acts trivially.
536: In general, the group $D_{n}$ is generated by two elements,
537: call them $a$, $b$, subject to the relations
538: \begin{displaymath}
539: a^2=1, \: \: b^{n} = 1, \: \: aba \: = \: b^{-1}
540: \end{displaymath}
541: The center of $D_{2n}$ is ${\bf Z}_2$, generated by $b^n$,
542: and $D_{2n}/{\bf Z}_2 = D_n$, {\it i.e.},
543: \begin{displaymath}
544: 1 \: \longrightarrow \: {\bf Z}_2 \: \longrightarrow \:
545: D_{2n} \: \longrightarrow \: D_n \: \longrightarrow \: 1
546: \end{displaymath}
547: is exact, and describes $D_{2n}$ as a (nontrivial) central extension of $D_n$.
548: In an orbifold $[X/D_8]$ where the center of $D_8$ acts trivially,
549: so that only $D_4$ acts effectively on $X$,
550: we find that the resulting theory looks much like a $D_4$ orbifold,
551: except that some one-loop twisted sectors are omitted,
552: but the remaining theory is still modular invariant.
553: The remaining one-loop twisted sectors follow the pattern that
554: for any $g \in D_4$ along one side, the allowed group elements on
555: the other side are
556: generated by $g$. Since the ${\bf Z}_2$ lies in the center,
557: and so each remaining one-loop twisted sector appears in the $D_8$ orbifold
558: with multiplicity $| {\bf Z}_2 |^2$, the one-loop partition function has
559: the form
560: \begin{displaymath}
561: Z(D_8) \: = \: \frac{| D_4 | }{| D_8 |} | {\bf Z}_2 |^2 \left[
562: Z(D_4) \: - \: \left( \mbox{some one-loop sectors} \right) \right]
563: \end{displaymath}
564: the same general form as in the previous example.
565:
566:
567:
568: \subsubsection{${\bf Z}_2$ gerbe over a dihedral orbifold}
569:
570: Another example of a gerbe over a space can be obtained as follows.
571: Let $DD_n$ denote the binary dihedral group, generated (as a subgroup
572: of $SL(2,{\bf C})$) by the matrices
573: \begin{displaymath}
574: a \: = \: \left[ \begin{array}{cc}
575: \xi & 0 \\
576: 0 & \xi^{-1} \end{array} \right], \: \:
577: b \: = \: \left[ \begin{array}{cc}
578: 0 & 1 \\
579: -1 & 0 \end{array} \right]
580: \end{displaymath}
581: for $\xi$ an $n$th root of unity, obeying relations including
582: $b^2 = -I$ and $aba = b$. For simplicity,
583: we shall assume $n=3$.
584: This group projects onto a representation of the dihedral
585: group in $SL(3,{\bf C})$, generated by the matrices
586: \begin{displaymath}
587: a' \: = \: \left[ \begin{array}{ccc}
588: \xi & 0 & 0 \\
589: 0 & 1 & 0 \\
590: 0 & 0 & \xi^{-1} \end{array} \right], \: \:
591: b' \: = \: \left[ \begin{array}{ccc}
592: 0 & 0 & 1 \\
593: 0 & -1 & 0 \\
594: 1 & 0 & 0 \end{array} \right]
595: \end{displaymath}
596: obeying relations including $b'^2 = + I$ and $a' b' a' = b'$.
597: If we denote the first group by $DD$ and the second by $D$,
598: then as the first projects onto the second with kernel ${\bf Z}_2$,
599: we have a short exact sequence
600: \begin{displaymath}
601: 1 \: \longrightarrow \: {\bf Z}_2 \: \longrightarrow \:
602: DD \: \longrightarrow \: D \: \longrightarrow \: 1
603: \end{displaymath}
604: that we can use to define a ${\bf Z}_2$ gerbe over an orbifold
605: $[X/D]$. For example, consider $X = E \times A \times E$,
606: where $E$ and $A$ are both elliptic curves, and $E$ has a complex
607: multiplication by a cube root of unity.
608: Let $D$ act on $E \times A \times E$ in the obvious way, via its
609: description in terms of $SL(3,{\bf C})$ matrices.
610:
611: It is not hard to check that this gerbe is not
612: trivial as well. Indeed, a trivialization of this gerbe is the same
613: thing as a principal ${\bf Z}_{2}$-bundle on $E \times A \times E$
614: which is equipped with an action of $DD$ that lifts the natural action
615: of $D$ on $E \times A \times E$. Every such ${\bf Z}_{2}$-bundle
616: corresponds to a character $\pi_{1}(E\times A\times E) \to {\bf
617: Z}_{2}$ which is invariant under the natural action of $D$ on
618: $\pi_{1}(E\times A\times E)$. Using the identifications $\pi_{1}(E)
619: \cong \pi_{1}(A) \cong {\bf Z}^{2}$ one checks immediately that $a'$
620: and $b'$ act on $\pi_{1}(E\times A\times E) = {\bf Z}^{2}\oplus {\bf
621: Z}^{2}\oplus {\bf Z}^{2}$ via the block matrices
622: \[
623: \begin{pmatrix} {\mathbb X} & & \\ & {\mathbb I} & \\ & & {\mathbb X}
624: \end{pmatrix} \qquad \text{and} \qquad
625: \begin{pmatrix}
626: & & {\mathbb I} \\ & - {\mathbb I} & \\ {\mathbb I} & & ,
627: \end{pmatrix}
628: \]
629: respectively. Here ${\mathbb I}$ is the $2\times 2$ identity matrix
630: and ${\mathbb X}$ is the $2\times 2$ matrix
631: \[
632: {\mathbb X} = \begin{pmatrix} 0 & -1 \\ 1 & -1 \end{pmatrix}.
633: \]
634: Using this explicit form for the $D$ action on $\pi_{1}(E\times
635: A\times E)$, one checks directly that the only $D$-invariant ${\bf
636: Z}_{2}$-valued character of $\pi_{1}(E\times A\times E)$ is the
637: trivial one. Thus $[X/DD]$ will be trivial only if $D\times {\bf
638: Z}_{2}$ admits a surjective homomorphism onto $DD$, which is clearly
639: impossible. One can also check that in $[X/DD]$,
640: at one-loop one gets exactly $| {\bf Z}_2 |^2$ copies of each
641: one-loop twisted sector of $[X/D]$, so the resulting one-loop
642: partition function is manifestly modular-invariant, and the closed
643: string massless spectrum of $[X/DD]$ is the same as that of $[X/D]$.
644:
645:
646:
647: \subsubsection{A non-banded gerbe}
648:
649:
650: So far we have only discussed banded gerbes, {\it i.e.} the
651: trivially-acting part of the group has been central.
652: Let us next consider a more general example.
653: Let ${\bf H}$ denote the eight-element group of
654: quaternions, {\it i.e.}
655: \begin{displaymath}
656: {\bf H} \: = \: \{ \pm 1, \pm i, \pm j, \pm k \}
657: \end{displaymath}
658: Consider a nontrivial ${\bf Z}_4$ gerbe over
659: the orbifold $[X/{\bf Z}_2]$ constructed by using the fact
660: that ${\bf H}$ can be expressed as
661: \begin{displaymath}
662: 1 \: \longrightarrow \: \langle i \rangle \: \longrightarrow \: {\bf H}
663: \: \longrightarrow \: {\bf Z}_2 \: \longrightarrow \: 1
664: \end{displaymath}
665: where $\langle i \rangle$ denotes the cyclic subgroup of order four generated
666: by $i \in {\bf H}$.
667: The subgroup $\langle i \rangle$ is not in the center of ${\bf H}$,
668: hence this extension is not central, and so the gerbe
669: $[X/ {\bf H}]$ (in which ${\bf H}$ acts by first projecting to
670: ${\bf Z}_2$ and then using the given ${\bf Z}_2$ action)
671: is not a ${\bf Z}_4$-banded gerbe, but merely a ${\bf Z}_4$ gerbe.
672: Nontrivial Calabi-Yau gerbes of this form can be constructed by {\it e.g.}
673: taking $X$ to be a Calabi-Yau with $\pi_1$ containing a
674: ${\bf Z}_2$ whose generator preserves the holomorphic volume form.
675: When we apply the same analysis as above to this particular
676: gerbe, we find that the resulting theory has all the same one-loop
677: twisted sectors as $[X/{\bf Z}_2]$.
678:
679:
680: However, there is a new complication arising in this non-banded
681: gerbe. Although all the same one-loop twisted sectors as in a $[X/{\bf Z}_2]$
682: orbifold arise, none are omitted;
683: they arise with different multiplicities.
684: If we let $\xi$ denote the generator of ${\bf Z}_2$,
685: then the
686: \begin{displaymath}
687: {\scriptstyle 1} \square_{1}
688: \end{displaymath}
689: one-loop twisted sector of $[X/{\bf Z}_2]$ arises from
690: \begin{displaymath}
691: {\scriptstyle \pm 1, \pm i} \square_{ \pm 1, \pm i}
692: \end{displaymath}
693: twisted sectors in $[X/{\bf H}]$
694: {\it i.e.} has multiplicity sixteen,
695: whereas the
696: \begin{displaymath}
697: {\scriptstyle 1} \square_{ \xi}
698: \end{displaymath}
699: one-loop twisted sector of $[X/{\bf Z}_2]$ arises from
700: the $[X/{\bf H}]$ twisted sectors
701: \begin{displaymath}
702: {\scriptstyle \pm 1} \square_{ \pm j, \pm k},
703: \end{displaymath}
704: and so only has multiplicity eight.
705: Because $\langle i \rangle$ is not central in ${\bf H}$,
706: there are no
707: \begin{displaymath}
708: {\scriptstyle \pm i} \square_{ \pm j, \pm k}
709: \end{displaymath}
710: twisted sectors
711: in the $[X/{\bf H}]$ orbifold, and so the multiplicity is reduced.
712: Similarly, the
713: \begin{displaymath}
714: {\scriptstyle \xi } \square_{ \xi}
715: \end{displaymath}
716: twisted sector of the $[X/{\bf Z}_2]$
717: orbifold arises from the $[X/{\bf H}]$ twisted sectors
718: \begin{displaymath}
719: {\scriptstyle \pm j} \square_{\pm j} \, , \: \: \:
720: {\scriptstyle \pm k} \square_{\pm k} \, ,
721: \end{displaymath}
722: and so again has
723: multiplicity eight.
724:
725: Since the $[X/{\bf H}]$ orbifold is explicitly modular-invariant when described
726: as an ${\bf H}$ orbifold, it must also be modular-invariant
727: when described in terms of twisted sectors of the $[X/{\bf Z}_2]$
728: orbifold, and indeed it is straightforward to check that this is
729: the case. The $SL(2,{\bf Z})$ action on $[X/{\bf Z}_2]$ one-loop
730: twisted sectors has two orbits, given by
731: \begin{displaymath}
732: \left\{ {\scriptstyle 1} \square_1 \: \right\}
733: \end{displaymath}
734: and
735: \begin{displaymath}
736: \left\{ {\scriptstyle 1} \square_{\xi} \, , \: {\scriptstyle \xi}
737: \square_1 \, , \: {\scriptstyle \xi} \square_{\xi} \: \right\} \\
738: \end{displaymath}
739: so as multiplicities are constant within each individual
740: $SL(2,{\bf Z})$ orbit, again we see the theory is modular-invariant.
741:
742:
743:
744:
745:
746: \subsubsection{Another non-banded gerbe}
747:
748: Another example of a non-banded gerbe can be obtained as follows.
749: Consider the nonabelian group $A_4$ \cite[chapter I.5]{lang}
750: of alternating permutations of
751: four elements. One uses the notation
752: $(a b c \cdots d)$ to indicate a permutation mapping $a$ to $b$,
753: $b$ to $c$, and so forth, eventually wrapping around to map
754: $d$ to $a$. This group has a ${\bf Z}_2 \times {\bf Z}_2$
755: normal subgroup described by the nontrivial elements
756: \begin{eqnarray*}
757: \alpha & \equiv & (1 4) (2 3) \\
758: \beta & \equiv & (1 3) (2 4) \\
759: \gamma & \equiv & (1 2) (3 4)
760: \end{eqnarray*}
761: The group $A_4$ has a total of twelve elements,
762: and the three elements of the quotient
763: \begin{displaymath}
764: A_4 / {\bf Z}_2 \times {\bf Z}_2 \cong {\bf Z}_{3}
765: \end{displaymath}
766: have representatives
767: \begin{displaymath}
768: \begin{array}{c}
769: \left\{ 1, \alpha, \beta, \gamma \right\} \\
770: \left\{ (1 2 3), (1 4 2), (2 4 3), (1 3 4) \right\} \\
771: \left\{ (1 3 2), (1 4 3), (1 2 4), (2 3 4) \right\}
772: \end{array}
773: \end{displaymath}
774: One can form a non-banded ${\bf Z}_2 \times {\bf Z}_2$ gerbe
775: over an orbifold $[X/{\bf Z}_3]$ as,
776: $[X/A_4]$, where $A_4$ acts on $X$ by first projecting to ${\bf Z}_3$
777: and then using the ${\bf Z}_3$ action.
778:
779: The analysis of strings on this non-banded gerbe proceeds much as before.
780: We find that every possible one-loop twisted sector of $[X/{\bf Z}_3]$
781: reappears in $[X/A_4]$ -- no twisted sectors are omitted.
782: However, the one-loop twisted sectors appear with different multiplicities.
783: Let us work through a few examples.
784: Let $\xi$ denote the generator of ${\bf Z}_3$.
785: The
786: \begin{displaymath}
787: {\scriptstyle 1} \square_1
788: \end{displaymath}
789: one-loop twisted sectors of the $[X/{\bf Z}_3]$ orbifold
790: arise from
791: \begin{displaymath}
792: {\scriptstyle 1, \alpha, \beta, \gamma} \square_{1, \alpha, \beta, \gamma}
793: \end{displaymath}
794: one-loop twisted sectors of the $[X/A_4]$ orbifold, and so have
795: multiplicity $4^2=16$. The
796: \begin{displaymath}
797: {\scriptstyle 1} \square_{\xi}
798: \end{displaymath}
799: one-loop twisted sector of
800: the $[X/{\bf Z}_3]$ orbifold arises from the
801: \begin{displaymath}
802: {\scriptstyle 1} \square_{ (123) } \, , \: \: \:
803: {\scriptstyle 1} \square_{ (142) } \, , \: \: \:
804: {\scriptstyle 1} \square_{ (243) } \, , \: \: \:
805: {\scriptstyle 1} \square_{ (134) }
806: \end{displaymath}
807: one-loop twisted sectors of the $[X/A_4]$ orbifold, and so have
808: multiplicity $4$. The
809: \begin{displaymath}
810: {\scriptstyle \xi} \square_{\xi^2}
811: \end{displaymath}
812: one-loop twisted sectors
813: of the $[X/{\bf Z}_3]$ orbifold arise from the
814: \begin{displaymath}
815: {\scriptstyle (123)}\square_{(132)} \, , \: \: \:
816: {\scriptstyle (142)} \square_{(124)} \, , \: \: \:
817: {\scriptstyle (243)} \square_{(234)} \, , \: \: \:
818: {\scriptstyle (134)} \square_{(143)}
819: \end{displaymath}
820: one-loop twisted sectors of the $[X/A_4]$ orbifold, and so have
821: multiplicity $4$.
822: Proceeding in this fashion, one can show that the
823: \begin{displaymath}
824: {\scriptstyle 1} \square_1
825: \end{displaymath}
826: twisted
827: sector of the $[X/{\bf Z}_3]$ orbifold appears with multiplicity
828: sixteen, but all other one-loop twisted sectors appear with
829: multiplicity four.
830:
831: Modular invariance of the theory is guaranteed by its presentation
832: as an $[X/A_4]$ orbifold, but is also straightforward to check
833: in terms of $[X/{\bf Z}_3]$ twisted sectors.
834: The $SL(2,{\bf Z})$ orbits of one-loop twisted sectors of
835: $[X/{\bf Z}_3]$ are given by
836: \begin{displaymath}
837: \left\{ {\scriptstyle 1} \square_1 \: \right\}
838: \end{displaymath}
839: \begin{displaymath}
840: \left\{ {\scriptstyle 1} \square_{\xi} \, , \:
841: {\scriptstyle \xi} \square_1 \, , \:
842: {\scriptstyle \xi} \square_{\xi} \, , \:
843: {\scriptstyle \xi} \square_{\xi^2} \, , \:
844: {\scriptstyle \xi^2} \square_{\xi} \, , \:
845: {\scriptstyle \xi^2} \square_{\xi^2} \: \right\}
846: \end{displaymath}
847: so again we see that multiplicities are constant on elements of
848: any given $SL(2,{\bf Z})$ orbit, and so the theory is modular-invariant.
849:
850:
851:
852:
853:
854:
855:
856:
857:
858:
859:
860:
861:
862:
863:
864:
865:
866:
867:
868:
869: \section{Gauging nonfinite noneffective groups}
870: \label{nonmincharge}
871:
872: In the previous section we discussed several examples of gauged
873: finite noneffectively-acting groups. Now, it is also certainly possible
874: to gauge a nonfinite group acting noneffectively but
875: with finite stabilizers.
876: For example, consider a sigma model on the total space of a principal
877: $U(1)$ bundle, in which a $U(1)$ is gauged which rotates the fibers
878: $k$ times instead of once. By comparison to a gauging which rotates
879: the fibers only once, rotating the fibers $k$ times means giving
880: the fields in the worldsheet theory nonminimal $U(1)$ charges.
881:
882: Since we have seen in examples that gauging a noneffectively-acting
883: finite group is not equivalent to gauging an effectively-acting group,
884: we would expect the same to be true of nonfinite groups.
885: After all, one can describe both as local orbifolds by trivially-acting
886: groups, so one would expect qualitatively similar behavior.
887: Thus, in order to be consistent with the observations
888: of the last section, one expects that a two-dimensional
889: gauge theory with nonminimal charges must be different from a two-dimensional
890: gauge theory with minimal charges.
891:
892: Indeed, that is the case. Although such two-dimensional gauge
893: theories are the same perturbatively, they are very different
894: nonperturbatively.
895:
896: This fact will play a crucial role in \cite{glsm}, where we will
897: study gauged linear sigma models for toric stacks, which
898: look like ordinary gauged linear sigma models, but with nonminimal charges.
899: There, we will explicitly calculate some of the many ways in which
900: the theories differ -- from different correlation functions to
901: different R-symmetry anomalies.
902:
903: Since this physical effect is obscure, let us take a moment to
904: describe
905: more carefully the
906: general reasons why these theories are distinct.
907: (We
908: would like to thank J.~Distler and R.~Plesser for providing the
909: detailed argument that we review in this section.)
910: For a different
911: discussion of two-dimensional gauge theories with fermions of
912: nonminimal charges, see \cite[section 4]{edold}. (The discussion
913: there is most applicable to the present situation when $m \ll M$, in
914: the notation of that reference.)
915:
916:
917: To be specific, consider a gauged linear sigma model with a single
918: $U(1)$ gauge field, and with chiral superfields, all of charge $k$,
919: with $k>1$. (Mathematically, this corresponds to a ${\bf Z}_k$ gerbe
920: on a projective space, as we shall review in \cite{tonyme,glsm}.)
921: One might argue that this theory should be the same as a theory
922: with chiral superfields of charge $1$, as follows.
923: Since instanton number is essentially monopole number,
924: from Dirac quantization since the electrons have charges a multiple
925: of $k$, the instantons must have charge a multiple of $1/k$,
926: and so zero modes of the Higgs fields in a minimal nonzero instanton
927: background would be sections of ${\cal O}(k/k) = {\cal O}(1)$,
928: just as in a minimal charge GLSM. Making the charges nonminimal
929: has not changed the physics.
930: In order to recover the physics we have described,
931: we require the Higgs fields to have charge $k$ while the
932: instanton numbers are integral, not fractional.
933:
934: Closer analysis reveals subtleties.
935: Let us break up the analysis into two separate cases: first,
936: the case that the worldsheet is noncompact, second,
937: that the worldsheet is compact. For both cases, it will be important
938: that the worldsheet theory is two-dimensional.
939:
940: First, the noncompact case.
941: Since the $\theta$ angle couples to $\mbox{Tr }F$,
942: we can determine the instanton numbers through the periodicity of
943: $\theta$. Suppose we have the physical theory described above,
944: namely a GLSM with Higgs fields of charge $k$,
945: plus two more massive fields, of charges $+1$ and $-1$.
946: In a two-dimensional theory, the $\theta$ angle acts as an electric
947: field, which can be screened by pair production, and that screening
948: determines the periodicity of $\theta$.
949: If the only objects we could pair produce were the Higgs fields
950: of charge $k$, then the theta angle would have periodicity
951: $2 \pi k$, and so the instanton numbers would be multiples
952: of $1/k$. However, since the space is noncompact, and the
953: electric field fills the entire space, we can also pair produce
954: arbitrary numbers of the massive fields, which have charges
955: $\pm 1$, and so the $\theta$ angle has periodicity $2 \pi$,
956: so the instantons have integral charges.
957:
958:
959: We can phrase this more simply as follows.
960: In a theory with only Higgs fields of charge $k$,
961: the instanton numbers are multiples of $1/k$, and so the resulting
962: physics is equivalent to that of a GLSM with minimal charges.
963: However, if we add other fields of charge $\pm 1$,
964: then the instanton numbers are integral,
965: and if those fields become massive, and we work at an energy scale
966: below that of the masses of the fields, then we have a theory
967: with Higgs fields of charge $k$, and integral instanton numbers,
968: giving us the physics that corresponds to a gerbe target.
969:
970: Thus, we see in the noncompact case that there are two
971: possible physical theories described by Higgs fields of charge $k$:
972: one is equivalent to the GLSM with minimal charges,
973: and the other describes the gerbe.
974:
975:
976: The analysis for the compact worldsheet case is much shorter.
977: Strictly speaking, to define the theory nonperturbatively on a
978: compact space, we must specify, by hand, the bundles that the
979: Higgs fields couple to. If the gauge field is described by
980: a line bundle $L$, then coupling all of the Higgs fields to
981: $L^{\otimes k}$ is a different prescription from coupling all
982: of the Higgs fields to $L$. As a result, the spectrum of zero modes
983: differs between the two theories, hence correlation functions and
984: anomalies differ between the two theories,
985: and so the two physical theories are very different,
986: as we shall see in examples later.
987:
988: We shall assume throughout this paper that the worldsheet is
989: compact, though as we have argued the same subtlety shows up
990: for noncompact worldsheets.
991:
992: Again, we shall discuss this matter in much greater detail
993: in \cite{tonyme,glsm}, but to help whet the reader's appetite, let us
994: review how this works in a simple example. Consider the ${\bf C}
995: {\bf P}^{N-1}$ model, realized as $N$ chiral superfields each of
996: charge $1$ with respect to a gauged $U(1)$.
997: Let us construct a model which we shall denote the $G^k_{-1} {\bf P}^{N-1}$
998: model (notation to be explained in \cite{glsm}),
999: or $G {\bf P}^{N-1}$ for brevity,
1000: consisting of $N$ chiral superfields each of charge $k$ with respect
1001: to a single gauged $U(1)$. Although perturbatively these two two-dimensional
1002: gauge theories are equivalent, nonperturbatively they are distinct.
1003: For example, in the ordinary ${\bf C} {\bf P}^{N-1}$ model,
1004: anomalies break the $U(1)_A$ to a ${\bf Z}_{2N}$ subgroup,
1005: whereas in the $G {\bf P}^{N-1}$ model, anomalies break the
1006: $U(1)_A$ to a ${\bf Z}_{2kN}$ subgroup. The quantum cohomology
1007: ring of the ordinary ${\bf C} {\bf P}^{N-1}$ model is given by
1008: \begin{displaymath}
1009: {\bf C}[x]/(x^N \: - \: q)
1010: \end{displaymath}
1011: whereas the quantum cohomology ring of the $G {\bf P}^{N-1}$ model
1012: is given by
1013: \begin{displaymath}
1014: {\bf C}[x]/(x^{kN} \: - \: q)
1015: \end{displaymath}
1016: reflecting the fact that A model correlation functions in the two
1017: theories are different.
1018: We shall explore this in much more detail in \cite{glsm}.
1019:
1020:
1021:
1022:
1023:
1024: \section{Closed string spectra} \label{spectra}
1025:
1026: \subsection{Quotients by finite noneffectively-acting groups}
1027: \label{spectra:noneff}
1028:
1029:
1030: To compute the massless spectrum of a sigma model on $X$ with a gauged
1031: noneffectively-acting finite group $G$,
1032: one way to proceed is to do the computation formally the same way
1033: as for an effectively-acting finite group: for each principal $G$-bundle
1034: on $S^1$, we have a branch of the semiclassical moduli space,
1035: and so quantizing that branch we get a sector of the Hilbert space.
1036: In this fashion we are led to
1037: a massless spectrum given by
1038: \begin{displaymath}
1039: \oplus_{[g]} H^*(X^g; {\bf C})^{Z(g)}
1040: \end{displaymath}
1041: where the sum is over conjugacy classes in $G$, and $Z(g)$ is the centralizer
1042: of a given element $g$ representing some conjugacy class.
1043: The inertia stack of $[X/G]$ for $G$ finite and noneffectively-acting
1044: has the same form as for $G$ finite and effectively-acting, namely
1045: \begin{displaymath}
1046: I_{[X/G]} \: = \: \prod_{[g]} [ X^g/Z(g) ]
1047: \end{displaymath}
1048: and so proceeding as before, the massless spectrum is the same as the
1049: de Rham cohomology of the inertia stack.
1050:
1051: A skeptic might well argue that this calculation is somewhat naive.
1052: Let us work through a simple example, and examine the details of
1053: the calculation.
1054:
1055: Consider for example $[X/{\bf Z}_k]$, where the ${\bf Z}_k$ acts
1056: completely trivially on $X$. According to the proposed
1057: massless spectrum calculation
1058: above, since ${\bf Z}_k$ is abelian, the Hilbert space should contain
1059: $k$ sectors, and since the ${\bf Z}_k$ acts trivially, $X^g = X$ for all
1060: $g$, so each twisted sector contains a copy of $H^*(X; {\bf C})$.
1061: In other words, according to the calculation above, the massless spectrum
1062: of this orbifold should be $k$ copies of the massless spectrum of a
1063: sigma model on $X$.
1064:
1065: The one-loop partition function of this gauged sigma model is given by
1066: \begin{eqnarray*}
1067: Z_{[X/G]} & = & \frac{1}{| {\bf Z}_k | } \sum_{g,h} Z_{g,h} \\
1068: & = & \frac{1}{ | {\bf Z}_k | } | {\bf Z}_k |^2 Z_{1,1} \\
1069: & = & k Z_X
1070: \end{eqnarray*}
1071: just a factor of $k$ times the one-loop partition function for $X$.
1072:
1073: Now, ordinarily in quantum field theory, multiplying a partition function
1074: by a constant has no effect on the physics, so a skeptic might argue that
1075: in this case, gauging the trivially-acting ${\bf Z}_k$ should have no effect,
1076: and the massless spectrum should be given by one copy of $H^*(X;{\bf C})$,
1077: not the $k$ copies we obtained above.
1078: However, because this sigma model is ultimately coupled to worldsheet gravity,
1079: we must be more careful. In a theory coupled to gravity, factors in front
1080: of partition functions cannot be ignored, for the same reasons that
1081: one cannot ignore contributions to a cosmological constant
1082: (see \cite[section 7.3]{pol1} for more details on this).
1083: Thus, the multiplicative factor of $k$ in the one-loop partition function
1084: cannot be consistently ignored.
1085:
1086: We can see the effect of such multiplicative factors by closer examination
1087: of the one-loop partition function. For example, if $X = {\bf R}^d$, then
1088: the one-loop partition function of a sigma model on $X$ can be written
1089: in the form \cite[equ'n~(7.3.8b)]{pol1}:
1090: \begin{displaymath}
1091: Z \: = \: i V_d \int_{F_0} \frac{ d \tau d \overline{\tau} }{ 4 \tau_2 }
1092: \left( 4 \pi^2 \alpha' \tau_2 \right)^{-d/2}
1093: \sum_{i \in {\cal H}^{\perp}} q^{h_i-1} \overline{q}^{ \overline{h}_i-1}
1094: \end{displaymath}
1095: (see the reference for notation) where ${\cal H}^{\perp}$ is
1096: (most of) the closed string Hilbert space. Multiplying this partition
1097: function by a factor of $k$ looks formally equivalent to increasing
1098: the multiplicity of closed string states by a factor of $k$,
1099: and that is precisely the result we obtained originally for the massless
1100: spectrum.
1101:
1102: Another check can be performed by interpreting the one-loop partition function
1103: as a string propagator and counting poles.
1104: For a bosonic string on flat space, the full one-loop partition function
1105: in the regime where $\tau_2 \rightarrow \infty$ has the expansion
1106: \cite[equ'n~(7.3.15)]{pol1}:
1107: \begin{displaymath}
1108: 2 \pi i V_{26} \int^{\infty} \frac{d \tau_2}{2 \tau_2}
1109: \left( 4 \pi^2 \alpha' \tau_2\right)^{-13}
1110: \left[ \exp(4 \pi \tau_2) \: + \: 24^2 \: + \: \cdots \right]
1111: \end{displaymath}
1112: The exponential term corresponds to the tachyon in the closed bosonic
1113: string spectrum, the $24^2$ term corresponds to the $24^2$ massless states
1114: of the closed bosonic string
1115: \begin{displaymath}
1116: \alpha_0^{\mu} \overline{\alpha}_0^{\nu} | 0 \rangle
1117: \end{displaymath}
1118: and so forth.
1119: If we were to quotient ${\bf R}^{24}$ by a trivially-acting ${\bf Z}_k$,
1120: the effect would be to multiply this partition function by a factor of $k$.
1121: Then, in this pole expansion, instead of a $24^2$ term, we would have
1122: a $24^2 k$ term, which would indicate $24^2 k$ massless states,
1123: consistent with our calculation of the massless spectrum in the
1124: noneffective orbifold.
1125:
1126:
1127:
1128: A skeptic might nevertheless still want to try to argue that the spectrum of a
1129: trivially-acting
1130: orbifold should only be one copy of the massless spectrum of the cover.
1131: In special cases, namely when the full orbifold group is a central
1132: extension by a trivially-acting group, it is possible to find an
1133: alternative spectrum computation. If the full orbifold group is nonabelian,
1134: then the one-loop partition function will be proportional to
1135: the partition function of an effectively-acting orbifold (with group
1136: given by the quotient of the full group by the trivially-acting part),
1137: with an $SL(2,{\bf Z})$-orbit of one-loop twisted sectors omitted,
1138: which could be interpreted as modifying the projection operator.
1139: To satisfy such skeptics, we pursue this spectrum calculation program
1140: in section~\ref{falselead}. Although this direction might sound promising,
1141: ultimately it fails, because the resulting physical theory is
1142: non-unitary. This is essentially because you cannot consistently
1143: multiply even $SL(2,{\bf Z})$-orbits of one-loop twisted sectors by
1144: zero and get a unitary theory -- although modular invariance is preserved,
1145: multiloop factorization is not. Unitarity is, in fact, the origin
1146: of the cocycle condition in discrete torsion. Thus, since this alternative
1147: spectrum calculation leads to nonunitary results, we do not believe
1148: this alternative spectrum calculation is correct.
1149: To help convince skeptics, we work out the details of this false
1150: lead extensively
1151: in section~\ref{falselead}.
1152:
1153: Another potential interpretation of the massless spectrum
1154: requires a homomorphism from the trivially-acting group to $U(1)$.
1155: After all, given a banded $G$-gerbe, classified by an element of
1156: $H^2(X,G)$ for $G$ finite, and a homomorphism from $G$ to $U(1)$,
1157: we can construct an element of $H^2(X,U(1))$, which defines a flat
1158: $B$ field. If such a map arose naturally in these constructions,
1159: then perhaps the correct massless spectrum calculation would be
1160: in terms of an effectively-acting orbifold with a flat $B$ field
1161: background. However, no such homomorphism arises physically, so far
1162: as we have been able to determine, so this potential massless spectrum
1163: calculation is not well-defined, much less tenable.
1164:
1165:
1166:
1167: A more subtle difficulty, that we have not discussed so far,
1168: involves deformation theory. The mathematical notion of deformation
1169: theory of a stack encodes only the untwisted sector moduli;
1170: there is no mathematics corresponding to twist field moduli.
1171: Thus, for example, in the orbifold $[X/{\bf Z}_k]$ where
1172: the ${\bf Z}_k$ acts trivially on $X$, the mathematical deformations
1173: are those of $X$. This would appear to be a problem for the massless
1174: spectrum calculation presented here, as ordinarily the physical
1175: moduli have a geometric understanding as the moduli of the target.
1176:
1177: We will briefly discuss this issue later in section~\ref{defthy},
1178: and will discuss the issue much more extensively in \cite{tonyme,glsm}.
1179: Twist fields for trivially-acting group elements can be understood
1180: algebraically, and giving such twist field moduli a vev takes us
1181: into new presentations of abstract CFT's of a form not previously discussed.
1182: We simply have more physical deformations than can be understood
1183: mathematically, and we will be able to see their effects explicitly.
1184:
1185: More to the point, we will see explicitly in \cite{glsm} that these
1186: twist field moduli play a critical role in understanding mirror symmetry.
1187: Usual mirror constructions, when applied to quotients by noneffectively-acting
1188: groups, naturally produce the abstract CFT's alluded to above.
1189: Furthermore,
1190: the structure of these abstract CFT's plays a crucial role in understanding
1191: how to generalize Batyrev's mirror construction to stacks.
1192:
1193: Since we see these nonmathematical twist field moduli explicitly
1194: giving rise to abstract CFT's, and since we see the same moduli playing
1195: a crucial role in understanding mirror symmetry, we are led to believe
1196: that the proposed calculation of massless spectra in noneffective orbifolds
1197: is correct, and that we have not overcounted states.
1198:
1199:
1200:
1201:
1202:
1203:
1204:
1205:
1206:
1207:
1208:
1209:
1210:
1211:
1212:
1213:
1214: \subsection{An instructive false lead on
1215: finite noneffectively-acting
1216: groups} \label{falselead}
1217:
1218: We have just argued that the correct massless spectrum
1219: of an orbifold by a finite non\-effec\-ti\-ve\-ly-acting group should
1220: be computed in formally the same way as for a finite
1221: effectively-acting group: the Hilbert space has as many sectors
1222: as conjugacy classes of the group, and in each sector,
1223: one takes the part of the cohomology of the fixed-point locus
1224: that is invariant under centralizers. We have seen how this is
1225: consistent with spectrum calculations based on one-loop partition
1226: function calculations, discussed some alternatives, and also outlined
1227: how this is consistent with deformation theory and mirror symmetry,
1228: topics we shall discuss more extensively later.
1229:
1230:
1231: We argued in the previous section that multiplicative factors in
1232: orbifold partition functions play a crucial role in checking
1233: state degeneracies, and give a solid test of our massless spectrum
1234: calculation. To help convince remaining skeptics,
1235: in this subsection we shall see what happens when one ignores those
1236: multplicative factors, and assume that one gets only one copy of untwisted
1237: sector states in noneffective orbifolds, not multiple copies.
1238: This leads to an alternative spectrum calculation, in which omission of
1239: one-loop twisted sectors implies a modified projection operation on the spectrum
1240: of an effectively-acting orbifold.
1241:
1242: We shall see in this subsection that this alternative spectrum calculation
1243: is not consistent, because the resulting physical theories are not
1244: unitary, and moreover this approach does not work in all cases.
1245: To help convince readers that this approach is not fruitful,
1246: and to add support for our proposal, let us work through the details
1247: of this alternative approach, to see in greater detail why it is wrong.
1248:
1249:
1250:
1251:
1252:
1253:
1254:
1255: \subsubsection{Basic calculations}
1256:
1257: Let us take the attitude that in an orbifold by a noneffectively-acting
1258: finite group, {\it i.e.} a string compactification on a gerbe,
1259: the result should be closely related to the massless spectrum
1260: of an orbifold by an effectively-acting group, given by quotienting out
1261: the noneffectively-acting normal subgroup.
1262: In particular, there should be only one dimension-zero operator,
1263: and omission of some of the one-loop twisted sectors should be interpreted
1264: as modifying the projection operator. We shall refer to the specific
1265: example of an $[X/D_4]$ orbifold~\ref{ex:d4elliptic},
1266: where the $D_4$ acts by first
1267: projecting to a ${\bf Z}_2 \times {\bf Z}_2$, which acts effectively:
1268: \begin{displaymath}
1269: 1 \: \longrightarrow \: {\bf Z}_2 \: \longrightarrow \: D_4 \:
1270: \longrightarrow \: {\bf Z}_2 \times {\bf Z}_2 \:
1271: \longrightarrow \: 1
1272: \end{displaymath}
1273:
1274: To understand why the projection operation is modified,
1275: recall that one of the functions of
1276: the one-loop twisted sector sum is to enforce a projection onto
1277: $G$-invariant states in a $G$-orbifold. Mechanically, summing over
1278: twisted sectors is equivalent to inserting a projection operator
1279: \begin{displaymath}
1280: \frac{1}{|G|} \sum g
1281: \end{displaymath}
1282: in the string propagator that only allows $G$-invariant states to
1283: propagate. By omitting some of the one-loop twisted sectors,
1284: we no longer have the complete projection operator, so only
1285: a partial projection is enforced.
1286:
1287: To see what the projection operator becomes on each $S^1$ twisted
1288: sector, we need to look at the surviving $T^2$ twisted sectors.
1289: Since all twisted sectors of the form $(1 | g)$ for any element
1290: $g \in {\bf Z}_2 \times {\bf Z}_2$ survive, the projection operator
1291: on the untwisted states is the usual one. Thus, for untwisted states,
1292: we take ${\bf Z}_2 \times {\bf Z}_2$ invariants.
1293: The other $S^1$ twisted sectors are more interesting.
1294: For each $g \in {\bf Z}_2 \times {\bf Z}_2$,
1295: the only surviving $T^2$ twisted sectors involving $g$
1296: are $(1 | g)$ and $(g | g)$. Thus, in a $g$ twisted sector,
1297: for $g \neq 1$, the projection operator reduces to a projection
1298: onto states invariant under the cyclic subgroup of ${\bf Z}_2 \times
1299: {\bf Z}_2$ generated by $g$.
1300:
1301:
1302: More generally, given any banded $K$-gerbe $[X/G]$ over an orbifold $[X/H]$
1303: where
1304: \begin{displaymath}
1305: 1 \: \longrightarrow \: K \: \longrightarrow \: G \:
1306: \stackrel{\alpha}{\longrightarrow} \: H \: \longrightarrow \: 1
1307: \end{displaymath}
1308: is a central extension involving finite groups,
1309: it is straightforward to see that
1310: all the $[X/H]$ twisted sectors that appear, appear with the same
1311: multiplicity, so that the massless spectrum is given by
1312: \begin{displaymath}
1313: \bigoplus_{ (h) \subset H } H^*\left( [ X^h / Z'(h) ]; {\bf C} \right)
1314: \end{displaymath}
1315: where the sum is over conjugacy classes in $H$, and
1316: $Z'(h) = \alpha( Z( \alpha^{-1}(h) ))$.
1317:
1318: In the present case, since ${\bf Z}_2 \times {\bf Z}_2$ acts
1319: freely, the massless spectrum would be just the ${\bf Z}_2 \times {\bf Z}_2$
1320: invariant part of the cohomology of the elliptic curve.
1321: In the language of the paragraph above, whenever $H$ acts freely on $X$,
1322: the massless spectrum would just be the $H$-invariant part of the
1323: massless spectrum of $X$.
1324:
1325:
1326:
1327:
1328:
1329:
1330:
1331:
1332:
1333:
1334:
1335:
1336:
1337: \subsubsection{An example} \label{ex:falselead}
1338:
1339:
1340: Next, let us consider a specific example.
1341: Consider the ${\bf Z}_2 \times {\bf Z}_2$ action on
1342: $T^6$ in which each ${\bf Z}_2$ flips the signs of
1343: two of the three complex coordinates, as in \cite{vafaed}.
1344: We can define an action of the group $D_4$ on $T^6$,
1345: where $D_4$ is the nontrivial ${\bf Z}_2$ extension of ${\bf Z}_2 \times
1346: {\bf Z}_2$ discussed above, in which $D_4$ acts on $T^6$ by
1347: first projecting to ${\bf Z}_2 \times {\bf Z}_2$ and then
1348: ${\bf Z}_2 \times {\bf Z}_2$ acts on $T^6$ as just discussed.
1349: The ${\bf Z}_2$ subgroup of $D_4$ acts trivially, so this corresponds
1350: to a sigma model on a ${\bf Z}_2$ gerbe over the stack
1351: $[ T^6 / {\bf Z}_2 \times {\bf Z}_2 ]$.
1352: This gerbe is Calabi-Yau, and can be shown to be nontrivial.
1353:
1354:
1355:
1356: The physical analysis of closed strings on this gerbe proceeds just
1357: as before. Recall from \cite{vafaed} that the Hodge diamond of massless
1358: closed string states of the
1359: original $[T^6/ {\bf Z}_2 \times {\bf Z}_2 ]$ is given by
1360: \begin{displaymath}
1361: \begin{array}{ccccccc}
1362: & & & 1 & & & \\
1363: & & 0 & & 0 & & \\
1364: & 0 & & 51 & & 0 & \\
1365: 1 & & 3 & & 3 & & 1 \\
1366: & 0 & & 51 & & 0 & \\
1367: & & 0 & & 0 & & \\
1368: & & & 1 & & & \end{array}
1369: \end{displaymath}
1370: where
1371: \begin{displaymath}
1372: \begin{array}{ccccccc}
1373: & & & 1 & & & \\
1374: & & 0 & & 0 & & \\
1375: & 0 & & 3 & & 0 & \\
1376: 1 & & 3 & & 3 & & 1 \\
1377: & 0 & & 3 & & 0 & \\
1378: & & 0 & & 0 & & \\
1379: & & & 1 & & & \end{array}
1380: \end{displaymath}
1381: states are ${\bf Z}_2 \times {\bf Z}_2$-invariant untwisted sector
1382: states, and the remaining states are $3 \cdot 16$ copies
1383: (one for each nontrivial element of ${\bf Z}_2 \times {\bf Z}_2$,
1384: and one for each fixed point locus under a fixed element) of
1385: the ${\bf Z}_2 \times {\bf Z}_2$-invariant elements of
1386: \begin{displaymath}
1387: \begin{array}{ccccccc}
1388: & & & 0 & & & \\
1389: & & 0 & & 0 & & \\
1390: & 0 & & 1 & & 0 & \\
1391: 0 & & 1 & & 1 & & 0 \\
1392: & 0 & & 1 & & 0 & \\
1393: & & 0 & & 0 & & \\
1394: & & & 0 & & & \end{array}
1395: \end{displaymath}
1396: which is to say, $3 \cdot 16$ copies of
1397: \begin{displaymath}
1398: \begin{array}{ccccccc}
1399: & & & 0 & & & \\
1400: & & 0 & & 0 & & \\
1401: & 0 & & 1 & & 0 & \\
1402: 0 & & 0 & & 0 & & 0 \\
1403: & 0 & & 1 & & 0 & \\
1404: & & 0 & & 0 & & \\
1405: & & & 0 & & & \end{array}
1406: \end{displaymath}
1407: The spectrum calculation for the ${\bf Z}_2$ gerbe
1408: $[ T^6 / D_4 ]$ over $[ T^6 / {\bf Z}_2 \times {\bf Z}_2 ]$
1409: is almost identical, except that now in the twisted sectors, we only
1410: project onto states invariant under the subgroup generated by the
1411: group element associated with the twisted sector, not onto states
1412: invariant under all of ${\bf Z}_2 \times {\bf Z}_2$.
1413: Thus, the Hodge diamond of massless states on the gerbe is given by
1414: a sum of
1415: \begin{displaymath}
1416: \begin{array}{ccccccc}
1417: & & & 1 & & & \\
1418: & & 0 & & 0 & & \\
1419: & 0 & & 3 & & 0 & \\
1420: 1 & & 3 & & 3 & & 1 \\
1421: & 0 & & 3 & & 0 & \\
1422: & & 0 & & 0 & & \\
1423: & & & 1 & & & \end{array}
1424: \end{displaymath}
1425: states from the untwisted sector, invariant under the entire
1426: ${\bf Z}_2 \times {\bf Z}_2$, plus $3 \cdot 16$ copies of
1427: \begin{displaymath}
1428: \begin{array}{ccccccc}
1429: & & & 0 & & & \\
1430: & & 0 & & 0 & & \\
1431: & 0 & & 1 & & 0 & \\
1432: 0 & & 1 & & 1 & & 0 \\
1433: & 0 & & 1 & & 0 & \\
1434: & & 0 & & 0 & & \\
1435: & & & 0 & & & \end{array}
1436: \end{displaymath}
1437: which are twisted sector states,
1438: invariant under the relevant subgroup of ${\bf Z}_2 \times {\bf Z}_2$.
1439: Thus, the Hodge diamond of massless states on the gerbe
1440: $[T^6/D_4]$ is given by
1441: \begin{displaymath}
1442: \begin{array}{ccccccc}
1443: & & & 1 & & & \\
1444: & & 0 & & 0 & & \\
1445: & 0 & & 51 & & 0 & \\
1446: 1 & & 51 & & 51 & & 1 \\
1447: & 0 & & 51 & & 0 & \\
1448: & & 0 & & 0 & & \\
1449: & & & 1 & & & \end{array}
1450: \end{displaymath}
1451: We see that $H^{1,1}$ and $H^{2,2}$ of the gerbe $[T^6/D_4]$
1452: are identical to $H^{1,1}$ and $H^{2,2}$ of the underlying orbifold
1453: $[T^6/{\bf Z}_2 \times {\bf Z}_2]$, but $H^{1,2}$ and $H^{2,1}$ of
1454: the gerbe are significantly larger -- the gerbe has $48$ extra
1455: complex structure deformations beyond those possessed by the
1456: underlying orbifold, according to this proposed massless spectrum calculation.
1457:
1458:
1459: We shall see in the next subsection that this proposed alternative
1460: massless spectrum
1461: calculation fails the test of unitarity.
1462:
1463:
1464:
1465:
1466:
1467:
1468:
1469:
1470:
1471:
1472: \subsubsection{Unitarity fails in the alternate interpretation}
1473:
1474: Now that we have examined one-loop twisted sectors,
1475: let us take a moment to consider higher-loop
1476: twisted sectors.
1477: In particular, we will show
1478: that the one-loop $[T^2/{\bf Z}_2 \times {\bf Z}_2]$
1479: twisted sectors that are
1480: `forbidden' in the $[T^2/D_4]$ orbifold can reappear at higher string
1481: loop order, which is a sign of nonunitarity in this
1482: alternate interpretation of noneffective orbifolds.
1483: We take this result as another indication that our original interpretation
1484: of noneffective orbifolds and their spectra is correct.
1485:
1486: For example, consider a two-loop twisted sector in
1487: $[T^2/D_4]$. It is defined by four group elements $g_1$, $h_1$,
1488: $g_2$, $h_2$ which must obey the relation
1489: \begin{displaymath}
1490: h_1 g_1^{-1} h_1^{-1} g_1 \: = \: g_2^{-1} h_2 g_2 h_2^{-1}
1491: \end{displaymath}
1492: in the conventions of \cite[section 4.3.2]{dt3},
1493: just as the two group elements defining a one-loop twisted sector
1494: must obey the constraint that they commute.
1495: When both sides of the equation above are separately equal to the
1496: identity, the two-loop diagram factors through the identity operator,
1497: and can degenerate into
1498: a pair of one-loop twisted sectors joined by a long thin handle.
1499: In the present case of a $[T^2/D_4]$ orbifold,
1500: consider the case that
1501: \begin{eqnarray*}
1502: g_1 & = & a \\
1503: h_1 & = & ab \\
1504: g_2 & = & a \\
1505: h_2 & = & b
1506: \end{eqnarray*}
1507: These four group elements satisfy the condition above,
1508: and so define a two-loop twisted sector.
1509: Moreover, these four group elements obey the condition
1510: \begin{displaymath}
1511: h_1 g_1^{-1} h_1^{-1} g_1 \: = \: z \: = \:
1512: g_2^{-1} h_2 g_2 h_2^{-1}
1513: \end{displaymath}
1514: Since $z$ acts trivially on $T^2$, in the target space this two-loop
1515: diagram appears factorizable -- it looks like a product of two
1516: one-loop diagrams. The one-loop factors, however, are
1517: `forbidden' diagrams -- since $a$ and $b$ do not commute
1518: as elements of $D_4$,
1519: there is no $(a|b)$ one-loop twisted sector,
1520: and similarly there is no $(a|ab)$ one-loop twisted sector,
1521: despite the fact that both reappear inside this two-loop diagram.
1522:
1523: This lack of factorization is a signal of failure of unitarity of the
1524: target space theory. Recall that
1525: the optical theorem \cite[section 3.6]{weinberg1},
1526: an immediate consequence of unitarity of the
1527: S-matrix, says that the imaginary parts of a scattering amplitude
1528: can be obtained by cutting the diagram in half,
1529: multiplying the amplitudes for the two separate halves,
1530: and integrating over intermediate momenta.
1531: In a little more detail, following \cite[section 3.6]{weinberg1},
1532: if we write S-matrix elements as
1533: \begin{displaymath}
1534: S_{\beta \alpha} \: = \: \delta(\beta - \alpha) \: - \:
1535: 2 \pi i \delta(p_{\beta} - p_{\alpha}) M_{\beta \alpha}
1536: \end{displaymath}
1537: (see the reference for notation)
1538: then the relation $\sum_{\beta} S_{\beta \gamma}^* S_{\beta \alpha} =
1539: \delta(\gamma - \alpha)$ implies that
1540: \begin{displaymath}
1541: \mbox{Im } M_{\alpha \alpha} \: = \: - \pi \sum_{\beta}
1542: \delta(p_{\beta} - p_{\alpha}) | M_{\beta \alpha} |^2
1543: \end{displaymath}
1544: In the present case, since the one-loop diagrams vanish,
1545: in a unitary theory we would expect that the two-loop diagram has to vanish,
1546: but that is not what we found -- the two-loop diagram is nonvanishing,
1547: whereas the one-loop diagram vanishes. Thus, we appear to violate
1548: the optical theorem, and hence violate unitarity
1549: (unless all relevant scattering amplitudes have no imaginary part,
1550: which seems extremely unlikely).
1551: More generally, the optical theorem in the target-space theory
1552: is the reason why factorization of higher-loop amplitudes is necessary
1553: for unitarity.
1554:
1555: In passing, note that this same argument does {\it not} apply when we
1556: calculate the massless spectrum using the methods we support.
1557: If we do not omit noneffectively-acting twist fields, if the
1558: massless spectrum contains multiple dimension zero operators
1559: in different (noneffective) twisted sectors, then the two-loop diagram
1560: above does not factorize on the identity, but rather on a dimension-zero
1561: twist field. In this case, the two one-loop diagrams appearing on either
1562: side of the cut are not one-loop vacuum diagrams, but rather contain a
1563: (noneffective) twist field insertion, and so need not vanish, thereby
1564: preventing a contradiction with unitarity.
1565:
1566: There is an another way to see that unitarity is broken in this alternative
1567: interpretation of the noneffective orbifold, based on a difficulty with the
1568: fusion rules.
1569: Consider a gerbe over an orbifold, with extra twisted sector
1570: states. In particular, consider a gerbe over a ${\bf Z}_2 \times {\bf Z}_2$
1571: orbifold, such as our noneffective $D_4$ orbifold example.
1572: If we have a state in the $a$ twisted sector that is not invariant
1573: under $b$, where $a$ and $b$ generate ${\bf Z}_2 \times {\bf Z}_2$,
1574: then consider the product of that state with another $a$ sector
1575: state that is invariant under $b$. Since $a^2=1$, the result is an untwisted
1576: sector state that's not invariant under $b$ -- which cannot be allowed!
1577: (Unless it is the zero state.)
1578: The algebra does not close, so the theory does not make sense,
1579: as the operator products are not well-defined.
1580:
1581:
1582: Although lack of unitarity is not necessarily
1583: completely fatal (for example, noncommutative field theories are
1584: often nonunitary \cite{gomis}), in the present case we find this explicit
1585: failure of unitarity to be suggestive, and in light of other
1586: arguments presented earlier, we do not believe this alternative calculation
1587: of massless spectra to be correct.
1588: Thus, we are led to believe
1589: that the correct massless spectrum of a noneffective orbifold
1590: has as many sectors in the Hilbert space as conjugacy classes in
1591: the group, even if some of the group elements act trivially.
1592:
1593:
1594:
1595:
1596:
1597:
1598: \subsection{Quantum symmetries in noneffective orbifolds} \label{quantumsymm:noneff}
1599:
1600: We have argued that the Hilbert space of a noneffective orbifold
1601: should be computed in the same form as that of an effective orbifold.
1602: Recall that an effective orbifold has a `quantum symmetry'
1603: \cite{vafaqs}.
1604: In an abelian effective orbifold $[X/G]$, for $G$ finite,
1605: the quantum symmetry is $G$, and gives phases to twisted sectors.
1606:
1607: Because of the form of our result,
1608: noneffective orbifolds also trivially possess the same quantum
1609: symmetry. For example, in an abelian orbifold $[X/G]$ where $G$ is finite
1610: and acts trivially, there is a quantum symmetry $G$ which multiplies
1611: the twisted sectors by phases.
1612:
1613: In effective abelian orbifolds, orbifolding the orbifold by the
1614: quantum symmetry restores the original theory. The same arguments
1615: used to establish
1616: this fact (see {\it e.g.} \cite[section 8.5]{ginsparg}) can now
1617: be trivially extended to noneffective orbifolds, where one can
1618: easily see the same result is obtained.
1619:
1620: Now, suppose ${\cal C}$ is a CFT with a ${\bf Z}_n$ action,
1621: so that the orbifold CFT ${\cal C}' \equiv [ {\cal C} / {\bf Z}_n]$
1622: has a ${\bf Z}_n$ quantum symmetry. Suppose we now orbifold
1623: ${\cal C}'$ by ${\bf Z}_{kn}$ where ${\bf Z}_{kn}$ acts
1624: (noneffectively) on ${\cal C}'$ by first projecting to ${\bf Z}_n$,
1625: \begin{displaymath}
1626: 1 \: \longrightarrow \: {\bf Z}_k \: \longrightarrow \:
1627: {\bf Z}_{kn} \: \longrightarrow \: {\bf Z}_n \: \longrightarrow \: 1
1628: \end{displaymath}
1629: and then letting the ${\bf Z}_n$ act on the quantum symmetry.
1630: A natural guess is that the orbifold $[{\cal C}'/{\bf Z}_{kn}]$
1631: should give the same physical theory as the orbifold
1632: of the original CFT ${\cal C}$ by a trivially-acting ${\bf Z}_k$.
1633: Let us take a moment to see that explicitly,
1634: following \cite[section 8.5]{ginsparg}.
1635:
1636: First, let us recall why the ${\bf Z}_n$ orbifold of a $[ {\cal C}/{\bf Z}_n]$
1637: orbifold is again the original CFT ${\cal C}$.
1638: Let the generator of (either) ${\bf Z}_n$ be denoted $g$,
1639: let the one-loop twisted sector with boundaries $g^a$, $g^b$ in
1640: the $[ {\cal C}/{\bf Z}_n]$ orbifold be denoted
1641: \begin{displaymath}
1642: {\scriptstyle a} \square_b,
1643: \end{displaymath}
1644: and the one-loop twisted sector with analogous boundary conditions
1645: in the ${\bf Z}_n$ orbifold of the $[ {\cal C}/{\bf Z}_n]$ orbifold be
1646: denoted \begin{displaymath}
1647: {\scriptstyle a} \square_b\,{'}.
1648: \end{displaymath}
1649: Let $\xi$ be the generator of the $n$th roots of
1650: unity. Then, it is straightforward to show that
1651: \begin{displaymath}
1652: {\scriptstyle a} \square_{b}\,{'} \: = \: \frac{1}{n} \sum_{c,d} \xi^{ac} \xi^{bd}
1653: \left( {\scriptstyle c} \square_d \right)
1654: \end{displaymath}
1655: so that the complete one-loop partition function is
1656: \begin{eqnarray*}
1657: Z'' & = & \frac{1}{n} \sum_{a,b} \left( {\scriptstyle a} \square_b\,{'} \right) \\
1658: & = & \frac{1}{n^2} \sum_{a,b} \sum_{c,d} \xi^{ac} \xi^{bd}
1659: \left( {\scriptstyle c} \square_d \right) \\
1660: & = & \frac{1}{n^2} \sum_{c,d} n^2 \delta_{c,0} \delta_{d,0}
1661: \left( {\scriptstyle c} \square_d \right) \\
1662: & = & {\scriptstyle 0} \square_0
1663: \end{eqnarray*}
1664: which is the
1665: one-loop partition function for the original CFT ${\cal C}$.
1666:
1667: Now, let us repeat this analysis for a ${\bf Z}_{kn}$ orbifold of
1668: $[ {\cal C}/ {\bf Z}_n]$, where the ${\bf Z}_k$ kernel acts trivially
1669: and the ${\bf Z}_n$ projection acts as the quantum symmetry.
1670: Using indices $i,j \in \{ 0, 1, \cdots, kn-1\}$,
1671: and the fact that each twisted sector in this orbifold will be the
1672: same as a twisted sector in the ${\bf Z}_n$ orbifold of
1673: $[ {\cal C}/{\bf Z}_n]$, we have that
1674: \begin{displaymath}
1675: {\scriptstyle i} \square_j\,{'}
1676: \: = \: \frac{1}{n} \sum_{a,b=0}^{n-1} \xi^{a [i/k]}
1677: \xi^{b [j/k]} \left( {\scriptstyle a} \square_b \right)
1678: \end{displaymath}
1679: Thus, the full one-loop partition function of the final orbifold is
1680: given by
1681: \begin{eqnarray*}
1682: Z'' & = & \frac{1}{kn} \sum_{i,j=0}^{kn-1} \left( {\scriptstyle i} \square_j\,{'}
1683: \right) \\
1684: & = & \frac{1}{kn^2} \sum_{i,j=0}^{kn-1} \sum_{a,b=0}^{n-1}
1685: \xi^{a [i/k]}
1686: \xi^{b [j/k]} \left( {\scriptstyle a} \square_b \right) \\
1687: & = & \frac{1}{kn^2} \sum_{a,b=0}^{n-1} (kn)^2 \delta_{a,0} \delta_{b,0}
1688: \left( {\scriptstyle a} \square_b \right) \\
1689: & = & k \left( {\scriptstyle 0} \square_0 \right)
1690: \end{eqnarray*}
1691: the same as the one-loop partition function of
1692: the orbifold $[ {\cal C}/{\bf Z}_k]$ where the ${\bf Z}_k$ acts trivially.
1693: Thus, we have confirmation of our conjecture.
1694:
1695: We have only described one-loop partition functions, but the calculation
1696: can be repeated at arbitrary genus.
1697: It is straightforward to compute
1698: that the $g$-loop partition function of the ${\bf Z}_{kn}$ orbifold
1699: of $[ {\cal C}/{\bf Z}_n]$ is given by
1700: \begin{displaymath}
1701: \frac{1}{(kn)^g} \frac{1}{n^g} (kn)^{2g} \: = \: k^g
1702: \end{displaymath}
1703: times the $g$-loop partition function
1704: of ${\cal C}$, which is the same as the $g$-loop partition function
1705: of the orbifold $[ {\cal C}/{\bf Z}_k]$ for a trivially-acting
1706: ${\bf Z}_k$.
1707:
1708:
1709:
1710:
1711:
1712:
1713:
1714:
1715: \section{CFT and trivial group actions} \label{trivgerbesection}
1716:
1717: In this section we collect some remarks on
1718: gauging a $G$-action on $X$ where all of $G$ acts trivially.
1719: We will assume that $G$ is finite.
1720:
1721: In order to make contact with our upcoming work
1722: \cite{tonyme,glsm}, let us note that mathematically,
1723: a quotient by a $G$-action in which all of $G$ acts trivially
1724: is the same as the trivial $G$-gerbe.
1725:
1726:
1727:
1728:
1729: \subsection{Trivial group actions and product CFT's} \label{trivgcft}
1730:
1731: We have argued that the massless
1732: spectrum of a global quotient by a
1733: finite noneffectively-acting group should be computed in exactly
1734: the same fashion as an effectively-acting group, with one sector
1735: of the Hilbert space for each conjugacy class in $G$, and so forth.
1736: In the case of a trivial gerbe presented as above,
1737: this means that there are as many twisted sectors as conjugacy classes
1738: of $G$, and each twisted sector is additively a copy of the untwisted
1739: sector, with a dimension zero operator in each twisted sector
1740: corresponding to the identity of the untwisted sector.
1741: Furthermore, on the basis of quantum numbers it is clear that
1742: a state in any given twisted sector can be obtained from its counterpart
1743: in the untwisted sector by acting on the untwisted state with the
1744: dimension zero twist field.
1745:
1746: Put more simply, the massless spectrum, both additively and in its
1747: product structures, looks like the tensor product of the CFT for
1748: the underlying Calabi-Yau $X$ and the CFT for the orbifold $[\mbox{point}/G]$.
1749: The spectrum of the latter orbifold contains only dimension zero operators,
1750: one for each conjugacy class of $G$, and multiplying them by the
1751: identity operator in the CFT for $X$ generates the dimension zero twist
1752: fields in the CFT of $[X/G]$.
1753:
1754: This similarity with the tensor product extends to one-loop
1755: partition functions.
1756: Recall the one-loop partition function for the trivial gerbe on $X$
1757: is given by
1758: \begin{displaymath}
1759: |G| Z(X)
1760: \end{displaymath}
1761: By comparison, in conventions in which the partition function for
1762: a sigma model on a point is $1$, the one-loop partition function for the
1763: orbifold $[\mbox{point}/G]$ is given by $|G|$. The one-loop partition
1764: function of the tensor product is the product of the partition functions
1765: for the separate theories, so we see that the partition function
1766: for the tensor product of a sigma model on $X$ ($Z(X)$)
1767: and the orbifold $[\mbox{point}/G]$ is given by $|G|Z(X)$, matching
1768: the one-loop partition function for the trivial gerbe.
1769:
1770:
1771: On the basis of the massless spectrum, correlation functions,
1772: and the one-loop partition functions,
1773: we claim that physically the CFT corresponding to an orbifold
1774: $[X/G]$ by a global trivial $G$ action
1775: on a Calabi-Yau $X$ is the same as the tensor product of the CFT
1776: ${\cal C}_X$ corresponding to $X$ and
1777: the CFT ${\cal C}_G$ corresponding to the orbifold $[\mbox{point}/G]$,
1778: {\it i.e.}
1779: \begin{displaymath}
1780: {\cal C}_{ [X/G] } \: \cong \: {\cal C}_X \otimes {\cal C}_G.
1781: \end{displaymath}
1782:
1783:
1784:
1785: This claim about physics has a mathematical counterpart.
1786: Mathematically, a trivial $G$-gerbe $[X/G]$ over a manifold $X$ can be
1787: expressed as the product $X \times BG$ of stacks,
1788: {\it i.e.}
1789: \begin{displaymath}
1790: [X/G] \: \cong \: X \times BG
1791: \end{displaymath}
1792: where $BG = [\mbox{point}/G]$.
1793: It is straightforward to check this statement at the level of incoming
1794: maps. A map from a manifold $Y$ into the trivial $G$-gerbe $[X/G]$ is a pair
1795: consisting of a principal $G$-bundle $E$ over $Y$, together
1796: with a $G$-equivariant map $f: E \rightarrow X$.
1797: Since $G$ acts trivially on $X$, $f$ is equivalent to a map
1798: $f': Y \rightarrow X$. Thus, our map from $Y$ into $[X/G]$ is the
1799: same as a principal $G$-bundle $E$ over $Y$ together with a map
1800: $f': Y \rightarrow X$. However, that pair also specifies a map
1801: from $Y$ into $X \times BG$. The relevance of the map $f': Y \rightarrow X$
1802: is clear, and since $BG = [\mbox{point}/G]$,
1803: a map $Y \rightarrow BG$ is just\footnote{This entertaining fact, an immediate
1804: consequence of the definition, makes the stack $[\mbox{point}/G]$ behave
1805: analogously to the classifying space for $G$, and is a reason for the similar
1806: notation and the similar name (classifying stack). } a principal $G$-bundle $E$ over $Y$.
1807:
1808: Intuitively, if we try to compare gerbes to fiber bundles,
1809: then $BG$ is the analogue of the fiber of a $G$-gerbe.
1810: A trivial $G$-gerbe over $X$ is the product $X \times BG$.
1811: Also,
1812: all $G$-gerbes over $X$ look locally like $X \times BG$,
1813: though only the trivial gerbe has that form globally.
1814:
1815: For $X$ and $Y$ Calabi-Yau spaces, the CFT of $X \times Y$ is the same
1816: as the tensor product of the CFT's corresponding to $X$ and $Y$,
1817: so it is very natural for the CFT of $X \times BG$ to be the tensor product
1818: of the CFT's for $X$ and $BG$.
1819:
1820:
1821:
1822:
1823:
1824:
1825:
1826: \subsection{$[\mbox{point}/{\bf Z}_k]$ and finite-group physics}
1827: \label{finitegroupfieldintro}
1828:
1829: Consider the CFT defined by the ${\bf Z}_k$-orbifold of a point,
1830: $[\mbox{point}/{\bf Z}_k]$. The corresponding massless spectrum is
1831: generated by a single twist field $\xi$, as seen in
1832: section~\ref{spectra:noneff},
1833: and because of selection
1834: rules for noneffective orbifolds discussed in
1835: section~\ref{quantumsymm:noneff},
1836: correlation functions $\langle \xi^n \rangle$ vanish unless $n$ is a
1837: multiple of $k$.
1838:
1839: The same result can be obtained from a slightly different-looking setup.
1840: Consider a physical theory defined by a ${\bf Z}_k$-valued field
1841: $\phi$, ${\bf Z}_k$-valued in the sense that it takes values in
1842: the $k$th roots of unity. One can build a very trivial QFT of this
1843: field: the path integral measure is just a sum over the $k$ possible
1844: values of $\phi$, the action vanishes identically,
1845: and correlation functions are just simple
1846: statistical measures:
1847: \begin{displaymath}
1848: \langle \phi^n \rangle \: = \: \sum_k \phi^k
1849: \end{displaymath}
1850: Since the path integral measure is just a sum over $k$th roots of unity,
1851: the correlation functions in this trivial theory vanish unless $n$ is
1852: a multiple of $k$.
1853:
1854: Since this trivial QFT has the same fields as the CFT
1855: $[\mbox{point}/{\bf Z}_k]$, and those fields have the same correlation
1856: functions, we claim that the trivially-acting orbifold of a point is
1857: isomorphic to this trivial theory of a ${\bf Z}_k$-valued field.
1858:
1859: This observation will play an important role later in \cite{tonyme}
1860: when we study deformation theory of stacks -- we will use fields
1861: valued in roots of unity to make some physical deformations without
1862: mathematical counterparts explicit, and so verify their existence.
1863: We shall also, independently, find such fields valued in roots of unity
1864: occurring in \cite{glsm} when we study mirrors to stacks.
1865:
1866:
1867:
1868:
1869:
1870:
1871:
1872:
1873:
1874:
1875:
1876: \subsection{Trivial group actions versus disconnected targets}
1877:
1878: While discussing orbifolds by trivially-acting
1879: ${\bf Z}_k$'s, {\it i.e.} trivial ${\bf Z}_k$ gerbes, let us take a moment
1880: to compare their physics to that of sigma models with target space
1881: $k$ disjoint copies of a manifold $X$.
1882:
1883: The partition functions of these two theories match.
1884: As already discussed, the $g$-loop partition function of the
1885: orbifold of $X$ by a trivially-acting ${\bf Z}_k$ is
1886: $k^{2g}/k^g = k^g$ times the $g$-loop partition
1887: function of a sigma model on $X$.
1888: This is also true for the $g$-loop partition function of a sigma
1889: model on $k$ disjoint copies of $X$. To see this, note that such a sigma
1890: model has $k$ times as many states as a sigma model on $X$, given by
1891: the $k$-fold tensor product of the states of a sigma model on $X$.
1892: Since in a genus $g$ partition function one has states propagating on
1893: each of $g$ loops, the result is that a genus $g$ partition function
1894: when the target is $k$ copies of $X$ should be $k^g$ times the partition
1895: function for $X$, matching the $g$-loop partition function of the trivial
1896: gerbe.
1897:
1898: Similarly, the massless spectra are also the same.
1899: The massless spectrum of the $[X/{\bf Z}_k]$ orbifold is the sum of
1900: $k$ copies of the cohomology of $X$. Thus, for example, it contains
1901: $k$ dimension zero operators, in each of $k$ twisted sectors.
1902:
1903: The massless spectrum of the disjoint union of $k$ copies of $X$ is the
1904: direct sum of $k$ copies of the cohomology of $X$. In other words,
1905: a state in this sigma model is a $k$-tuple of states in a sigma model on $X$.
1906: Just as in the orbifold $[X/{\bf Z}_k]$, in this theory there are $k$
1907: dimension zero operators, corresponding to the fact that the cohomology
1908: of the disjoint union of $k$ copies of $X$ is dimension $k$ in degree zero.
1909:
1910: In fact, we believe these conformal field theories are the same,
1911: and more generally, we believe that noneffective gaugings are
1912: at least often described by the same conformal field theories as
1913: disjoint unions of spaces.
1914: Furthermore, this identification solves an important technical problem
1915: involving cluster decomposition in these theories: having multiple
1916: dimension zero operators violates cluster decomposition, but conformal
1917: field theories describing disjoint unions of spaces violate
1918: cluster decomposition in the mildest possible way, causing no other
1919: physical inconsistencies.
1920: We shall discuss these issues in much greater detail in
1921: \cite{clusterdecomp}, and defer further discussion to that work.
1922:
1923:
1924:
1925:
1926:
1927:
1928:
1929: \section{Deformation theory issues} \label{defthy}
1930:
1931: One important issue we have not addressed so far concerns the interpretation
1932: of the marginal operators in the CFT's we have described.
1933: In typical examples, there are more marginal operators than there are
1934: geometric moduli, so naturally one must ask, what does it mean to deform
1935: along those directions?
1936:
1937: For example, for a trivial ${\bf Z}_k$ orbifold of a space $X$,
1938: {\it i.e.} the trivial gerbe $[X/{\bf Z}_k] = X \times B {\bf Z}_k$,
1939: we have argued in section~\ref{spectra:noneff} that the massless spectrum
1940: is $k$ copies of the cohomology of $X$. However, the only obvious geometric
1941: deformations are just deformations of $X$, the original untwisted sector
1942: moduli. What does it mean to deform along the other $k-1$
1943: marginal operators?
1944:
1945: This physical puzzle has a mathematical analogue.
1946: The mathematical infinitesimal
1947: moduli of the algebraic stack corresponding to this gerbe
1948: contain only one copy of the moduli of $X$, not $k$ copies.
1949: Again, we have a mismatch.
1950: Moreover, in a sigma
1951: model on a smooth manifold $X$, the physical moduli match
1952: mathematical moduli, so the present mismatch is a potential problem,
1953: just as for quotients by effectively-acting finite groups.
1954:
1955: One conceivable answer is that the `extra' $k-1$ marginal operators
1956: are obstructed. However, it is easy to check that since they differ
1957: from the untwisted sector operators merely by the addition of a twist
1958: field associated to a trivially-acting group element, there is no way
1959: to get nonvanishing correlation functions involving these operators
1960: unless there are some nonvanishing correlation functions among the
1961: original untwisted sector operators. If the untwisted sector operators
1962: are truly marginal, describing unobstructed moduli, then the twisted
1963: sector operators must also be unobstructed.
1964:
1965: Another conceivable answer is that there is some subtle
1966: problem with our massless
1967: spectrum calculation. After all, the marginal operators are, by definition,
1968: part of the massless spectrum, so if we have miscomputed the massless
1969: spectrum, then we may also have miscomputed the number of marginal
1970: operators. However, we have performed extensive independent tests
1971: of the massless spectrum calculation, and we do not believe that it is
1972: in error.
1973:
1974: This mismatch of deformations is part of a larger issue involving
1975: how physical deformations and mathematical moduli of stacks are related.
1976: We will discuss this matter extensively in \cite{tonyme}.
1977: As we shall discuss in that reference,
1978: in general terms our resolution of such mismatches is that
1979: the mathematical moduli correspond to deformations of the stack
1980: which result in (weakly-coupled) physical theories with well-behaved
1981: mathematical
1982: interpretations. The `extra' physical moduli result in physical
1983: theories which do not appear to have clean mathematical interpretations.
1984: We believe this claim because we are able to very explicitly describe
1985: and manipulate the theories that result from such deformations,
1986: as we shall outline below.
1987:
1988: In the case of the trivial ${\bf Z}_k$ orbifold of $X$, as described
1989: above, the untwisted sector moduli merely deform the covering space $X$,
1990: an operation which has a clean mathematical understanding.
1991: Giving a vev to twisted sector moduli has a different effect in
1992: conformal perturbation theory: formally, if we try to insert an
1993: exponential of a second descendant of a twisted sector marginal operator
1994: in correlation functions, defined by its Taylor expansion,
1995: then by the usual selection rules many of the terms in the Taylor expansion
1996: drop out. This formal operation is no longer anything as clean or simple
1997: as merely an ordinary geometric deformation of $X$.
1998: Rather, one appears to get a new and different family of conformal field theories.
1999:
2000: Another example should make the analysis clearer.
2001: Begin with a Landau-Ginzburg model corresponding to a Calabi-Yau hypersurface,
2002: so that the superpotential is the hypersurface polynomial.
2003: As is well-known,
2004: a marginal deformation of the theory corresponds to a deformation
2005: of that superpotential by terms which do not change the degree of
2006: homogeneity of the polynomial.
2007:
2008: Now, construct a trivial ${\bf Z}_k$ orbifold of that Landau-Ginzburg
2009: model. As discussed previously in section~\ref{finitegroupfieldintro},
2010: this is equivalent to adding a field $\Upsilon$ that takes values
2011: in $k$th roots of unity. According to our massless spectrum calculation,
2012: we now have $k$ times as many moduli in the physical theory as before,
2013: given by multiplying any vertex operator corresponding to a modulus
2014: of the original theory by a power of $\Upsilon$.
2015: Giving a vev to such a twisted sector modulus is equivalent to
2016: adding a term to the Landau-Ginzburg superpotential which has a factor
2017: of $\Upsilon$ to some power, since such terms are just supersymmetry
2018: transformations of the relevant vertex operator.
2019: For example, in \cite{glsm} we shall see Landau-Ginzburg superpotentials
2020: of the form
2021: \begin{displaymath}
2022: W \: = \: x_1^5 \: + \: \cdots \: + \: x_5^5 \: + \:
2023: \Upsilon \psi x_1 x_2 x_3 x_4 x_5
2024: \end{displaymath}
2025: where the $x_i$ are chiral superfields, $\psi$ is a complex number,
2026: and $\Upsilon$ takes values in roots of unity, which are summed over
2027: in the path integral measure.
2028: Thus, we can see these new physical deformations very explicitly,
2029: as {\it e.g.} Landau-Ginzburg superpotential terms with factors
2030: of $\Upsilon$, the field valued in roots of unity.
2031:
2032: In principle, we can interpret this in the same way as in the
2033: previous example, adding a formal
2034: exponential of a second descendant of the twisted sector modulus,
2035: and because of the usual selection rules, many of the terms will drop out.
2036: This is an equivalent description. By working with fields valued in roots
2037: of unity, however, we have a more algebraic description of the
2038: new conformal field theories, something much easier to work with than
2039: the description provided directly by conformal perturbation theory.
2040:
2041: The reader might ask why one cannot do the same for twist fields
2042: in effective orbifolds.
2043: Ordinarily, giving a vev to a twist field is somewhat messy,
2044: as the twist field introduces a branch cut, and as this changes the moding
2045: of worldsheet fields, the resulting marginal operators are no longer
2046: quite so simple to express. In the present case, however,
2047: since the twist field in question corresponds to a group element that
2048: acts trivially, the moding of worldsheet fields does not change,
2049: and so an algebraic description of the process of giving a vev to a
2050: twist field, as we have outlined above, becomes possible.
2051:
2052: We will return to these abstract CFT's and study them more extensively
2053: in \cite{glsm}, where they will be derived from a completely different
2054: direction. Here, we have derived Landau-Ginzburg models with fields
2055: taking values in roots of unity from considering physical deformations
2056: of noneffective orbifold theories. In \cite{glsm} we will find that the
2057: same sort of Landau-Ginzburg theories appear when one builds mirrors
2058: to stacks. The fact that we are seeing these same physical theories
2059: appear in a different context is an excellent check that our
2060: analysis is consistent.
2061:
2062:
2063:
2064:
2065:
2066:
2067:
2068:
2069:
2070:
2071:
2072:
2073:
2074:
2075:
2076:
2077:
2078:
2079:
2080:
2081: \section{D-branes in noneffective orbifolds} \label{Dbranenoneff}
2082:
2083: Earlier in section~\ref{spectra} we
2084: argued that the
2085: closed string massless spectrum in a noneffective orbifold
2086: should have exactly the same general form
2087: as that for effectively-acting finite groups.
2088: For example,
2089: even for an orbifold $[X/{\bf Z}_n]$ where the ${\bf Z}_n$ acts
2090: completely trivially, there should still be $n$ distinct twisted
2091: sectors in the massless spectrum, although additively each sector
2092: is identical to the untwisted sector.
2093:
2094: For open strings in noneffective orbifolds,
2095: although the group acts trivially on the base space,
2096: it can still act nontrivially on the Chan-Paton factors.
2097: Checking this statement requires verifying the Cardy condition,
2098: as discussed for orbifolds in {\it e.g.} \cite{bcr}.
2099: Recall that Cardy's condition \cite{cardy,cardylew,lew}
2100: amounts to the statement that
2101: the physics of an
2102: annulus diagram (see figure~\ref{c1}) should be independent of whether we
2103: interpret it as an open string propagating
2104: at one-loop or a closed string propagating at tree level
2105: between boundary states.
2106:
2107:
2108: \begin{figure}
2109: \centerline{\psfig{file=cardy2.eps,width=2.5in}}
2110: \caption{\label{c1} An annulus diagram, interpreted in two ways}
2111: \end{figure}
2112:
2113: In the case of an orbifold, the relevant annulus diagram is as
2114: shown in figure~\ref{c2}, and has a branch cut running between
2115: the boundary states. We can interpret this as either an open
2116: string propagating in a loop, coming back to itself up to the
2117: action of some element $g$, or alternately as a closed string
2118: in the $g$ twisted sector, propagating between two boundary states.
2119:
2120:
2121: \begin{figure}
2122: \centerline{\psfig{file=cardy1.eps,width=2.5in}}
2123: \caption{\label{c2} An annulus diagram in an orbifold }
2124: \end{figure}
2125:
2126: The analysis of {\it e.g.} \cite{bcr} also applies to the case of
2127: noneffective orbifolds, and allows us to give nontrivial $G$-actions
2128: to Chan-Paton factors even if $G$ acts trivially on the base,
2129: so long as we are careful to count all boundary states.
2130: For example, in $[X/{\bf Z}_n]$ where the ${\bf Z}_n$ acts trivially
2131: on $X$, then even though $G$ acts trivially on $X$,
2132: we still must distinguish boundary states in each
2133: of $| {\bf Z}_n | = n$ twisted
2134: sectors, even though many of these boundary states appear otherwise
2135: identical, just as the closed string massless spectrum is $n$ copies
2136: of the untwisted sector.
2137:
2138:
2139: Mathematically, the combination of a trivial action on the underlying space
2140: and a nontrivial action on the Chan-Paton factors means that B-branes
2141: in such orbifolds are twisted sheaves on the underlying space.
2142: In fact, there is a general statement that sheaves on gerbes are the
2143: same as twisted sheaves on the underlying space. These matters will
2144: be discussed in detail in \cite{tonyme}.
2145:
2146:
2147:
2148:
2149:
2150:
2151: \section{Mirror symmetry for completely trivial group actions} \label{mirrors}
2152:
2153:
2154: In section~\ref{trivgcft}, we argued that the CFT of a an
2155: orbifold of a sigma model on $X$
2156: by a completely trivially-acting group $G$,
2157: {\it i.e.} a trivial
2158: $G$-gerbe over a Calabi-Yau $X$, decomposes as a tensor product
2159: \begin{displaymath}
2160: {\cal C}_X \otimes {\cal C}_G
2161: \end{displaymath}
2162: where ${\cal C}_X$ is the CFT associated to a sigma model on $X$,
2163: and ${\cal C}_G$ is the $G$-orbifold of a single point, $[\mbox{point}/G]$.
2164:
2165: Given that result, we can immediately read off how mirror symmetry
2166: must work for trivial gerbes over spaces.
2167: If $X$ and $Y$ are a mirror pair of Calabi-Yau manifolds,
2168: then by definition of mirror symmetry, ${\cal C}_X \cong {\cal C}_Y$,
2169: hence
2170: \begin{displaymath}
2171: {\cal C}_X \otimes {\cal C}_G \: \cong \: {\cal C}_Y \otimes {\cal C}_G
2172: \end{displaymath}
2173: so we have that the trivial $G$-gerbe on $X$ is mirror to the trivial
2174: $G$-gerbe on $Y$.
2175:
2176:
2177: It is very easy to check that this prediction is compatible with
2178: massless spectra. Recall that if $X$ and $Y$ have
2179: complex dimension $n$, then their Hodge numbers satisfy
2180: \begin{displaymath}
2181: h^{i,j}(X) \: = \: h^{n-i,j}(Y)
2182: \end{displaymath}
2183: Now, the massless spectrum of
2184: the trivial gerbe $[X/G]$ ({\it i.e.} $G$ acts trivially on $X$)
2185: is just copies of the massless spectrum of $X$, one copy for each
2186: conjugacy class of $G$, with $U(1)_R$ charges and conformal weights
2187: unchanged, hence we immediately have the trivial result
2188: that
2189: \begin{displaymath}
2190: h^{i,j}([X/G]) \: = \: h^{n-i,j}([Y/G])
2191: \end{displaymath}
2192: confirming our claim above.
2193:
2194: We shall discuss mirror symmetry for noneffective quotients and stacks
2195: much more extensively in \cite{glsm}.
2196:
2197:
2198:
2199:
2200:
2201:
2202: \section{Conclusions}
2203:
2204: In this paper we have discussed some basic features of noneffective
2205: orbifolds, {\it i.e.} orbifolds in which nontrivial elements of the
2206: orbifold group act trivially. We have seen that the resulting physical
2207: theories are very different from orbifolds by effectively-acting groups.
2208: We have discussed their consistency in a variety of examples,
2209: studied closed string massless spectrum computations in detail,
2210: discussed D-branes in such orbifolds, and looked at some of the special
2211: properties of orbifolds in which all elements of the group act trivially.
2212:
2213: An important issue in understanding such gauged sigma models is
2214: the interpretation of the moduli fields. There are typically
2215: more (unobstructed) moduli fields than there are geometric moduli;
2216: how are the rest
2217: interpreted?
2218: Understanding the resolution of this puzzle has led us to a class of CFT's
2219: with a novel description, in terms of fields valued in roots of unity,
2220: an algebraic description of twist fields associated to trivially-acting
2221: group elements.
2222:
2223: We will return to these issues in \cite{tonyme,glsm}, where we will
2224: describe a more complete classification of universality classes of
2225: worldsheet RG flow of gauged sigma models. There, we will find general
2226: results for things ranging from massless spectra of IR fixed points
2227: to mirror symmetry, and will describe some of the new physics that
2228: arises in such considerations, such as examples of multiple
2229: distinct nonperturbative completions
2230: of perturbative two-dimensional gauge theories, and more independent
2231: derivations
2232: of fields valued in roots of unity.
2233:
2234: All of this has a mathematical interpretation in terms of stacks,
2235: as we have begun to outline in this paper.
2236:
2237:
2238:
2239:
2240:
2241: \section{Acknowledgements}
2242:
2243: We would like to thank A.~Adams, J.~Distler, S.~Katz, A.~Knutson,
2244: J.~McGreevy, and
2245: R.~Plesser for useful conversations. We would also like to thank the
2246: Aspen Center for Physics for hospitality while this work was being
2247: done and the UPenn Math-Physics group for the excellent
2248: conditions for collaboration it provided during several stages of this
2249: work. T.P. was partially supported by NSF grants DMS
2250: 0403884 and FRG 0139799.
2251:
2252:
2253:
2254:
2255:
2256: %\bibliographystyle{my-h-elsevier}
2257: %\bibliography{eric}
2258:
2259:
2260:
2261: \begin{thebibliography}{10}
2262:
2263: \addcontentsline{toc}{section}{References}
2264:
2265: \bibitem{tonyme}
2266: T. Pantev and E. Sharpe,
2267: \newblock ``String compactifications on {C}alabi-{Y}au stacks,'' 2005,
2268: \newblock {\tt hep-th/0502044}.
2269:
2270: \bibitem{glsm}
2271: T. Pantev and E. Sharpe,
2272: \newblock ``{G}{L}{S}{M}'s for gerbes (and other toric stacks),'' 2005,
2273: \newblock {\tt hep-th/0502053}.
2274:
2275: \bibitem{clusterdecomp} S. Hellerman, A. Henriques, T. Pantev, E. Sharpe,
2276: \newblock to appear.
2277:
2278: \bibitem{silvtate}
2279: J. Silverman and J. Tate,
2280: \newblock {\it Rational points on elliptic curves}, Undergraduate Texts in Mathematics
2281: (Springer-Verlag, New York, 1992).
2282:
2283: \bibitem{lang}
2284: S. Lang,
2285: \newblock {\it Algebra}, Graduate Texts in Mathematics Vol. 211, third ed.
2286: (Springer-Verlag, New York, 2002).
2287:
2288: \bibitem{edold}
2289: E. Witten,
2290: \newblock ``Instantons, the quark model, and the $1/{N}$ expansion,''
2291: \newblock Nucl. Phys. {\bf B149} (1979) 285--320.
2292:
2293: \bibitem{pol1}
2294: J. Polchinski,
2295: \newblock {\it String theory}. {V}ol. {I} Cambridge Monographs on Mathematical Physics
2296: (Cambridge University Press, Cambridge, 1998).
2297:
2298: \bibitem{vafaed}
2299: C. Vafa and E. Witten,
2300: \newblock ``On orbifolds with discrete torsion,''
2301: \newblock J. Geom. Phys. {\bf 15} (1995) 189, {\tt hep-th/9409188}.
2302:
2303: \bibitem{dt3}
2304: E. Sharpe,
2305: \newblock ``Discrete torsion,''
2306: \newblock Phys. Rev. D (3) 68 (2003) 126003, 20,
2307: \newblock {\tt hep-th/0008154}.
2308:
2309: \bibitem{weinberg1}
2310: S. Weinberg,
2311: \newblock {\it The quantum theory of fields}. {V}ol. {I} (Cambridge University Press,
2312: Cambridge, 1996).
2313:
2314: \bibitem{gomis}
2315: J. Gomis and T. Mehen,
2316: \newblock ``Space-time noncommutative field theories and unitarity,''
2317: \newblock Nuclear Phys. {\bf B591} (2000) 265--276,
2318: \newblock {\tt hep-th/0005129}.
2319:
2320: \bibitem{vafaqs}
2321: C. Vafa,
2322: \newblock ``Quantum symmetries of string vacua,''
2323: \newblock Modern Phys. Lett. {\bf A4} (1989) 1615.
2324:
2325: \bibitem{ginsparg}
2326: P. Ginsparg,
2327: \newblock ``Applied conformal field theory,''
2328: \newblock {\it Champs, cordes et ph\'enom\`enes critiques} (Les Houches, 1988), pp.
2329: 1--168, North-Holland, Amsterdam, 1990.
2330:
2331: \bibitem{bcr}
2332: M. Bill{\'o}, B. Craps and F. Roose,
2333: \newblock ``Orbifold boundary states from {C}ardy's condition,''
2334: \newblock J. High Energy Phys. (2001) Paper 38, 51,
2335: \newblock {\tt hep-th/0011060}.
2336:
2337: \bibitem{cardy}
2338: J. Cardy,
2339: \newblock ``Boundary conditions, fusion rules and the {V}erlinde formula,''
2340: \newblock Nucl. Phys. {\bf B324} (1989) 581--596.
2341:
2342: \bibitem{cardylew}
2343: J. Cardy and D. Lewellen,
2344: \newblock ``Bulk and boundary operators in conformal field theory,''
2345: \newblock Phys. Lett. {\bf B259} (1991) 274--278.
2346:
2347: \bibitem{lew}
2348: D. Lewellen,
2349: \newblock ``Sewing constraints for conformal field theories on surfaces with
2350: boundaries,''
2351: \newblock Nucl. Phys. {\bf B372} (1992) 654--682.
2352:
2353: \end{thebibliography}
2354:
2355:
2356: \end{document}
2357:
2358: