1: \NeedsTeXFormat{LaTeX2e}
2: \documentclass[12pt]{article}
3: \usepackage{amsmath,amssymb}
4: \usepackage{graphicx} % including PostScript
5: \usepackage{bbm}
6: \usepackage{longtable}
7: \usepackage[small,loose]{subfigure}
8: \usepackage{fleqn} % equations are not centered
9: \usepackage{mathrsfs} % e.g. nice Lagrange-L with \mathscr{L}
10: \usepackage[small,sf]{caption2}
11:
12:
13: % ============== Some page parameters ==========================================
14: \addtolength\textheight{115pt}
15: \addtolength\textwidth{60pt}
16: \addtolength\oddsidemargin{-37pt}
17: \setlength{\parindent}{20pt}
18: \setlength{\parskip}{4pt}
19: \frenchspacing
20: \sloppy
21: \headheight 12pt
22: \headsep 30pt
23: \footskip 24pt
24: \renewcommand{\baselinestretch}{1.2}
25: \addtolength{\topmargin}{-1.5cm}
26: \advance \headheight by 3.0truept % for 12pt mandatory...
27:
28: % ==============================================================================
29: % ============== User defined commands =========================================
30: \DeclareMathOperator{\re}{Re}
31: \DeclareMathOperator{\im}{Im}
32: \DeclareMathOperator{\tr}{tr}
33: \DeclareMathOperator{\diag}{diag}
34: \DeclareMathOperator{\quabla}{\boldsymbol{\square}}
35: \newcommand{\CenterEps}[2][1]{\ensuremath{\vcenter{\hbox{\includegraphics[scale=#1]{#2.eps}}}}} % Input eps files - Usage: \CenterEps[ScaleFactor]{FileName}
36: \newcommand{\CenterObject}[1]{\ensuremath{\vcenter{\hbox{#1}}}}
37: \newcommand{\D}{\mathrm{d}}
38: \newcommand{\I}{\mathrm{i}}
39: \newcommand{\BmL}{\ensuremath{B\!-\!L} }
40: \newcommand{\SimpleRoot}[1]{\alpha_{(#1)}}
41: \newcommand{\FundamentalWeight}[1]{\mu^{(#1)}}
42: \newcommand{\E}[1]{\ensuremath{\mathrm{E}_{#1}}} % e.g. \E{8}
43: \newcommand{\G}[1]{\ensuremath{\mathrm{G}_{#1}}}
44: \newcommand{\SO}[1]{\ensuremath{\mathrm{SO}(#1)}}
45: \newcommand{\SU}[1]{\ensuremath{\mathrm{SU}(#1)}}
46: \newcommand{\U}[1]{\ensuremath{\mathrm{U}(#1)}}
47: \newcommand{\Z}[1]{\ensuremath{\mathbbm{Z}_{#1}}} % Z_N ->\Z{N}
48: \newcommand{\SL}[1]{\ensuremath{\mathrm{SL}(#1)}}
49: \newcommand{\Spin}[1]{\ensuremath{\mathrm{Spin}(#1)}}
50:
51: \allowdisplaybreaks[0]
52:
53:
54: % ==============================================================================
55: % ============== Here we go! ===================================================
56: \begin{document}
57: \title{
58: \hfill {\normalsize TUM-HEP-659/07}\\[2cm]
59: {\bf\Huge Mirage torsion}\\[0.8cm]}
60:
61: \setcounter{footnote}{0}
62: \renewcommand{\thefootnote}{\alph{footnote}}
63:
64: \author{{\bf\normalsize
65: Felix~Pl\"oger$^{\boldsymbol{1},}$\footnote{\texttt{ploeger@th.physik.uni-bonn.de}},~
66: Sa\'ul~Ramos-S\'anchez$^{\boldsymbol{1},}$\footnote{\texttt{ramos@th.physik.uni-bonn.de}}},\\
67: {\bf\normalsize
68: Michael Ratz$^{\boldsymbol{2},}$\footnote{\texttt{mratz@ph.tum.de}}~~and
69: Patrick~K.~S.~Vaudrevange$^{\boldsymbol{1},}$\footnote{\texttt{patrick@th.physik.uni-bonn.de}}}\\[1cm]
70: {\it\normalsize
71: ${}^1$ Physikalisches Institut der Universit\"at Bonn, Nussallee 12, 53115 Bonn,
72: Germany.}\\[0.1cm]
73: {\it\normalsize
74: ${}^2$ Physik Department T30, Technische Universit\"at M\"unchen, 85748 Garching,
75: Germany.}
76: }
77:
78: \date{}
79:
80: \maketitle \thispagestyle{empty}
81:
82: \abstract{%
83: $\Z{N}\times\Z{M}$ orbifold models admit the introduction of a discrete torsion
84: phase. We find that models with discrete torsion have an alternative
85: description in terms of torsionless models. More specifically, discrete
86: torsion can be `gauged away' by changing the shifts by lattice vectors.
87: Similarly, a large class of the so-called generalized discrete torsion phases
88: can be traded for changing the background fields (Wilson lines) by lattice
89: vectors. We further observe that certain models with generalized discrete
90: torsion are equivalent to torsionless models with the same gauge embedding but
91: based on different compactification lattices. We also present a method of
92: classifying heterotic $\Z{N}\times\Z{M}$ orbifolds.
93: }
94:
95:
96: \setcounter{footnote}{0}
97: \renewcommand{\thefootnote}{\arabic{footnote}}
98:
99: \clearpage
100:
101:
102:
103: \section{Introduction}
104:
105:
106: Among the currently available frameworks, superstring theory appears to have the
107: greatest prospects for yielding a unified description of nature. Optimistically
108: one may hope to identify a string compactification that reproduces all
109: observations.
110: The perhaps simplest way to obtain a chiral spectrum in four dimensions, as
111: required by observation, is to compactify on an
112: orbifold~\cite{Dixon:1985jw,Dixon:1986jc}.
113: Although it is straightforward to compute orbifold spectra, a deep understanding
114: of these constructions, including an interpretation of the zero-modes, is harder
115: to obtain. Obstructions arise from the large number of possible gauge embeddings
116: and geometries, as well as other degrees of freedom. The classification of gauge
117: embeddings has been accomplished only in prime orbifolds. The generalization to
118: $\Z{N}\times\Z{M}$ orbifolds with or without Wilson lines has not been discussed
119: in the literature so far. $\Z{N}\times\Z{M}$ orbifolds are particularly rich as
120: they can be generalized by turning on certain phases which are known as discrete
121: torsion~\cite{Vafa:1986wx,Font:1988mk,Vafa:1994rv,Sharpe:2000ki,Gaberdiel:2004vx}.
122:
123: Aiming at a systematic understanding of heterotic $\Z{N}\times\Z{M}$ orbifolds, we set out
124: to survey the possibilities arising in these constructions. In the course of our
125: investigations we obtain rather surprising results. First of all, discrete
126: torsion can be `gauged away' in the sense that models with discrete torsion have
127: an alternative description in terms of torsionless models. Moreover, we shall
128: see that the so-called `non-factorizable' orbifolds are equivalent to
129: factorizable orbifolds with (generalized) discrete torsion or different gauge embedding.
130:
131: This paper is organized as follows. In section~\ref{sec:ZNxZM} we collect some
132: basic facts on the construction of $\Z{N}\times\Z{M}$ orbifold models. We
133: encourage readers who are familiar with the construction of orbifolds to
134: skip this section. In section~\ref{sec:DiscreteTorsion}, we establish the
135: equivalence between switching on a discrete torsion phase and changing the gauge
136: embedding by elements of the weight lattice.
137: Section~\ref{sec:GeneralizationDTAndBrothers} is devoted to the generalization
138: to orbifolds with Wilson lines. In section~\ref{sec:ClassifZ3xZ3} we outline a
139: prescription for a classification of $\Z{N}\times\Z{M}$ orbifold models.
140: Finally, section~\ref{sec:Conclusions} contains a discussion of our results.
141: Some issues concerning the transformation phases are discussed in the appendix.
142:
143:
144: \section{$\boldsymbol{\Z{N}\times\Z{M}}$ orbifold compactifications}
145: \label{sec:ZNxZM}
146:
147: \subsection{Setup}
148:
149: Let us start by reviewing some basic facts on orbifold compactifications
150: \cite{Dixon:1986jc}. To construct an orbifold, one first considers a
151: $d$-dimensional torus $\mathbbm{T}^d$, which can be understood as
152: $\mathbbm{R}^d/\Gamma$, i.e.\ as the $d$-dimensional space with points
153: differing by lattice vectors $e_\alpha \in \Gamma$ identified. In this study we will take
154: $d=6$ in order to arrive at an effective four-dimensional theory at low
155: energies. If the torus lattice enjoys one or more discrete rotational symmetries
156: comprising the point group $P$, one can define an orbifold as the quotient
157: $\mathbbm{O}=\mathbbm{T}^6/P$. Equivalently one can describe the orbifold by
158: \begin{equation}
159: \mathbbm{O}~=~\frac{\mathbbm{R}^6}{S}\;,
160: \end{equation}
161: where $S$ is the space group. Space group elements consist of discrete
162: rotations and translations by lattice vectors $e_\alpha$. We will be mostly
163: interested in $\Z{N}\times\Z{M}$ orbifolds, in which the torus lattice has two
164: discrete rotational symmetries described by the independent twists $\theta$ and
165: $\omega$, whereby $\theta^N=\omega^M=\mathbbm{1}$ and $N$ is a multiple of $M$.
166: Space group elements $g\in S$ are then given by $g = (\theta^{k_1}\,\omega^{k_2},
167: n_\alpha\, e_\alpha)$ where $0\le k_1 \le N-1$, $0\le k_2\le M-1$ and
168: $n_\alpha\in\mathbbm{Z}$.
169: Further, we restrict our analysis to models with \SU3 holonomy, where the
170: rotations can be diagonalized,
171: \begin{equation}
172: \theta\,z^i~=~\exp(2\pi\I\,v_1^i)\,z^i
173: \quad\text{and}\quad
174: \omega\,z^i~=~\exp(2\pi\I\,v_2^i)\,z^i\;,
175: \end{equation}
176: with $z^{1,2,3}$ being the complex coordinates of the compact space, and $\sum_i
177: v_{1,2}^i=0$. Unless stated otherwise, we use
178: \begin{equation}
179: v_1~=~\frac{1}{N}(1,0,-1;0)
180: \quad\text{and}\quad
181: v_2~=~\frac{1}{M}(0,1,-1;0)\;.
182: \end{equation}
183:
184: The space group action is to be embedded in the gauge degrees of freedom
185: according to
186: \begin{equation}
187: g = (\theta^{k_1}\,\omega^{k_2},n_\alpha\,e_\alpha)~\hookrightarrow~
188: (\mathbbm{1},V_g)\qquad V_g = k_1\,V_1+k_2\,V_2+n_\alpha\,A_\alpha\;,
189: \end{equation}
190: where $V_1$, $V_2$ are the shifts, $A_\alpha$ are the Wilson lines, and $V_g$ denotes the local shift corresponding to the twist $v_g = k_1 v_1 + k_2 v_2$. Due to the embedding, they have to be of appropriate orders:
191: \begin{equation}\label{eq:latticeconditions}
192: N\, V_{1} ~\in~ \Lambda\;,\quad
193: M\, V_{2} ~\in~ \Lambda\;,\quad
194: N_{\alpha}\, A_{\alpha} ~\in~ \Lambda\;.
195: \end{equation}
196: Here $\Lambda$ is the $\E8\times\E8$ or $\mathrm{Spin}(32)/\Z2$ weight
197: lattice,\footnote{Since these lattices are self-dual, we denote the root and
198: weight lattice by the same symbol.}
199: and $N_\alpha$ denotes the order of the Wilson line $A_\alpha$, which is
200: constrained by geometry.
201:
202: Modular invariance of one--loop amplitudes imposes strong conditions on the shifts and Wilson lines. In $\Z{N}$ orbifolds, the shift $V$ and the twist $v$ must fulfill~\cite{Dixon:1986jc,Vafa:1986wx}:
203: \begin{equation}\label{eq:znmodularinv}
204: N \,\left(V^2 - v^2\right)~=~0 \mod 2 \,.
205: \end{equation}
206: In $\Z{N}\times\Z{M}$ orbifolds with Wilson lines, modular invariance, together with consistency
207: requirements (see appendix \ref{app:ModularInvariance}), requires
208: \begin{subequations}\label{eq:newmodularinv}
209: \begin{eqnarray}
210: N\,\left(V_{1}^{2} - v_{1}^{2} \right) & = & 0 \mod 2\;, \\
211: \label{eq:fsmi1}
212: M\,\left(V_{2}^{2} - v_{2}^{2} \right) & = & 0 \mod 2\;, \\
213: \label{eq:fsmi2}
214: M\,\left(V_{1}\cdot V_{2} - v_{1}\cdot v_{2} \right) & = & 0 \mod 2\;, \\
215: \label{eq:fsmi3}
216: N_\alpha\,\left(A_{\alpha}\cdot V_{i}\right) & = & 0 \mod 2\;, \\
217: \label{eq:fsmi4}
218: N_\alpha\,\left(A_{\alpha}^2\right) & = & 0 \mod 2\;, \\
219: \label{eq:fsmi5}
220: Q_{\alpha\beta}\,\left(A_{\alpha}\cdot A_{\beta}\right) & = & 0 \mod 2 \quad (\alpha \neq \beta)\;,
221: \label{eq:fsmi6}
222: \end{eqnarray}
223: \end{subequations}
224: where $Q_{\alpha\beta}\equiv\text{gcd}(N_{\alpha},N_{\beta})$ denotes the greatest common divisor of
225: $N_{\alpha}$ and $N_{\beta}$.
226:
227:
228:
229: \subsection{Spectrum}
230:
231: Given a compactification lattice, the discrete rotations described by $v_{1,2}$,
232: shifts and Wilson lines, there exists a standard procedure to calculate the
233: massless spectrum (cf.\
234: \cite{Ibanez:1987pj,Font:1989aj,Forste:2004ie,Kobayashi:2004ya,Buchmuller:2004hv,Buchmuller:2006ik}). The Hilbert space
235: decomposes in untwisted and various twisted sectors, denoted by $U$ and
236: $T_{(k_1,k_2,n_\alpha)}$, respectively. The gauge group after compactification
237: is generated by the 16 Cartan generators plus roots $p\in\Lambda$ ($p^2=2$)
238: fulfilling
239: \begin{equation}
240: p\cdot V_i~=~0\mod1 \quad\forall i\;,\quad
241: p\cdot A_\alpha~=~0\mod1\quad\forall\alpha\;.
242: \end{equation}
243: Chiral untwisted sector states are described by $p\in\Lambda$ and
244: $q\in\Lambda_{\SO8}$ ($q^2=1$) satisfying
245: \begin{subequations}
246: \begin{eqnarray}
247: p\cdot V_i-q\cdot v_i & = &0\mod1\;,\quad i \in 1,2\;,\\
248: q\cdot v_i & \ne & 0\;,\quad i=1\:\text{and/or}\:2\;, \\
249: p\cdot A_\alpha & = & 0\mod 1\quad\forall\alpha\;.
250: \end{eqnarray}
251: \end{subequations}
252: Twisted sector zero modes are associated to the inequivalent `constructing
253: elements' $g=(\theta^{k_1}\omega^{k_2},n_\alpha e_\alpha)\in S$, corresponding
254: to the inequivalent fixed points and fixed planes. For each such $g$ one solves
255: the mass equations
256: \begin{subequations}\label{eq:MassEquation}
257: \begin{eqnarray}
258: \frac{1}{8}m_\mathrm{L}^2
259: & = &
260: \frac{1}{2}p_\mathrm{sh}^2-1+
261: \omega_i\,\widetilde{N}_{g,i}
262: +\overline{\omega}_i\,\widetilde{N}_{g,i}^*
263: +\delta c ~\stackrel{!}{=}~0\;,\\
264: \frac{1}{8}m_\mathrm{R}^2
265: & = &
266: \frac{1}{2}q_\mathrm{sh}^2-\frac{1}{2}+\delta c
267: ~\stackrel{!}{=}~0\;,
268: \end{eqnarray}
269: \end{subequations}
270: with the shifted momenta ($p\in\Lambda,q\in\Lambda_{\SO8}$)
271: \begin{subequations}
272: \begin{eqnarray}
273: p_\mathrm{sh} & = & p+V_g\;,\\
274: q_\mathrm{sh} & = & q+v_g\;.
275: \end{eqnarray}
276: \end{subequations} Here $\omega_i=(v_g)_i\mod1$ and
277: $\overline{\omega}_i=-(v_g)_i\mod1$, such that
278: $0<\omega_i,\,\overline{\omega}_i\le1$. Moreover, $\widetilde{N}_{g,i}$ and
279: $\widetilde{N}_{g,i}^*$ are integer oscillator numbers. Finally, $\delta
280: c=\frac{1}{2}\sum_i\omega_i\left(1-\omega_i\right)$.
281:
282: The states $|q_\mathrm{sh}\rangle_\mathrm{R} \otimes |p_\mathrm{sh}\rangle_\mathrm{L}$, where
283: $q_\mathrm{sh}$ and $p_\mathrm{sh}$ are solutions of the mass
284: equations~(\ref{eq:MassEquation}), are subject to certain invariance conditions: commuting elements
285: $h=(\theta^{t_1}\omega^{t_2},m_\alpha e_\alpha)$, with $[g,h]=0$, have to
286: act as the identity on physical states. This leads to the projection condition
287: \begin{equation}\label{eq:transformation}
288: |q_\mathrm{sh}\rangle_\mathrm{R} \otimes |p_\mathrm{sh}\rangle_\mathrm{L}
289: ~\stackrel{h}{\longmapsto}~
290: \Phi\, |q_\mathrm{sh}\rangle_\mathrm{R} \otimes |p_\mathrm{sh}\rangle_\mathrm{L}
291: ~\stackrel{!}{=}~
292: |q_\mathrm{sh}\rangle_\mathrm{R} \otimes |p_\mathrm{sh}\rangle_\mathrm{L}\;.
293: \end{equation}
294: Here the transformation phase $\Phi$ is given by
295: \begin{equation}\label{eq:transformationphase}
296: \Phi
297: ~\equiv~
298: e^{2\pi\I\,[p_\mathrm{sh}\cdot V_h- q_\mathrm{sh} \cdot v_h
299: + (\widetilde{N}_g - \widetilde{N}_g^*)\cdot v_h]}\,
300: \Phi_\mathrm{vac}\;,
301: \end{equation}
302: where (cf.\ appendix \ref{app:ModularInvariance})
303: \begin{equation}
304: \Phi_\mathrm{vac}
305: ~=~
306: e^{2\pi\I\,[-\frac{1}{2}(V_g\cdot V_h - v_g\cdot v_h)]}\;.
307: \end{equation}
308: Equation \eqref{eq:transformation} states that the transformation phase $\Phi$
309: has to vanish, which will be important for the following discussion.
310:
311:
312:
313: \section{Brother models and discrete torsion}
314: \label{sec:DiscreteTorsion}
315:
316: In this section we start by examining a new possibility to find inequivalent
317: models. We discuss under what circumstances models with shifts differing by
318: lattice vectors have different spectra and are thus inequivalent. Then
319: we review the concept of \textit{discrete torsion}, and clarify its relation to
320: models in which shifts differ by lattice vectors.
321:
322: \subsection[Brother models]{Brother models}
323: \label{sec:brothermodel}
324:
325: Let us start by clarifying under which conditions two models M and M$'$ are
326: equivalent. First, we restrict to the case without Wilson lines, where the
327: models M and M$'$ are described by the set of shifts $(V_1,V_2)$ and
328: $(V_1',V_2')$, respectively. Clearly, if the shifts are related by Weyl
329: reflections, i.e.\
330: \begin{equation}\label{eq:Weyl}
331: (V_1',V_2')~=~(W\,V_1,W\,V_2)\;,
332: \end{equation}
333: where $W$ represents a series of Weyl reflections, one does obtain equivalent
334: models. Let us now turn to comparing the spectra of two models M and M$'$, where
335: \begin{equation}\label{eq:BrotherDef}
336: (V_1',V_2')~=~(V_1+\Delta V_1,V_2+\Delta V_2)\;,
337: \end{equation}
338: with $\Delta V_1,\Delta V_2 \in \Lambda$. For future reference, we call models
339: related by equation~\eqref{eq:BrotherDef} `\textit{brother models}'.
340:
341: Brother models are also subject to modular invariance constraints. For the sake
342: of keeping the expressions simple, we restrict here to models fulfilling the
343: following (stronger) conditions:
344: \begin{subequations}\label{eq:strongmodularinv2}
345: \begin{eqnarray}
346: V_{i}^{2} - v_{i}^{2} & = & 0 \mod 2 \quad (i = 1, 2)\;,\label{eq:strong1} \\
347: V_{1} \cdot V_{2} - v_{1} \cdot v_{2} & = & 0 \mod 2\;. \label{eq:strong2}
348: \end{eqnarray}
349: \end{subequations}
350: (In section~\ref{sec:GeneralizationDTAndBrothers} we will relax these
351: conditions.)
352: Equations \eqref{eq:strongmodularinv2} imply that $\Phi_\mathrm{vac}~=~1 $ in
353: the transformation phase \eqref{eq:transformationphase}. The condition that
354: $(V'_1,\,V'_2)$ fulfill \eqref{eq:strongmodularinv2} leads to the following
355: constraints on $(\Delta V_1,\Delta V_2)$:
356: \begin{subequations}\label{eq:modinvconditions}
357: \begin{eqnarray}
358: V_{i} \cdot \Delta V_{i} & = & 0 \mod 1 \quad \quad i = 1, 2 \;,\\
359: V_{1} \cdot \Delta V_{2} + \Delta V_{1} \cdot V_{2} + \Delta V_1\cdot\Delta V_2& = & 0 \mod 2\;.
360: \end{eqnarray}
361: \end{subequations}
362: Consider now the massless spectrum corresponding to the constructing
363: element
364: \begin{equation}
365: g~=~(\theta^{k_1} \omega^{k_2}, n_\alpha e_\alpha)~\in~S
366: \end{equation}
367: of the models M and M$'$. For simplicity, we restrict our attention to
368: non-oscillator states. Physical states arise from tensoring together left- and
369: right-moving solutions of the masslessness condition
370: equation~\eqref{eq:MassEquation},
371: \begin{eqnarray}
372: |q + k_1 v_1 + k_2 v_2\rangle_\mathrm{R} \, \otimes\, |p + k_1 V_1 + k_2 V_2\rangle_\mathrm{L}
373: &&\text{for M}\;,\\
374: |q + k_1 v_1 + k_2 v_2\rangle_\mathrm{R} \, \otimes\, |p' + k_1 V'_1 + k_2 V'_2\rangle_\mathrm{L} &&\text{for M}' \;,
375: \end{eqnarray}
376: where $p' = p - k_1\Delta V_1 - k_2 \Delta V_2$ and the shifted momenta of the
377: left-movers are identical for M and M$'$. According to equation~(\ref{eq:transformationphase}) with
378: $\Phi_\mathrm{vac}=1$, these massless states transform under the action of a commuting element
379: \begin{equation}
380: h~=~(\theta^{t_1} \omega^{t_2}, m_\alpha e_\alpha) \in S \quad\text{with}\quad
381: [h,g]~=~0
382: \end{equation}
383: with the phases
384: \begin{eqnarray}
385: \Phi & = &
386: e^{2\pi \I\, \left[(p + k_1 V_1 + k_2 V_2)\cdot (t_1 V_1 + t_2 V_2) - (q + k_1 v_1 + k_2 v_2)\cdot (t_1
387: v_1 + t_2 v_2)\right]}
388: \qquad\text{for M}\;,\nonumber\\
389: \Phi' & = & e^{2\pi \I\, \left[(p' + k_1 V'_1 + k_2 V'_2)\cdot (t_1 V'_1 + t_2 V'_2) - (q + k_1 v_1 +
390: k_2 v_2 )\cdot (t_1 v_1 + t_2 v_2)\right]}
391: \ \quad\text{for M}'\;.\nonumber
392: \end{eqnarray}
393: By using the constraints~\eqref{eq:modinvconditions} and the properties of an integral lattice,
394: $p\cdot \Delta V_i \in {\mathbb Z}$ for $p, \Delta V_i \in \Lambda$, the
395: mismatch between the phases can be simplified to
396: \begin{equation}
397: \Phi'~=~\Phi \, e^{-2\pi \I\, (k_1 t_2 - k_2 t_1)V_2\cdot\Delta V_1}\;.
398: \end{equation}
399: That is, the transformation phase of states in model M$'$ differs from the
400: transformation phase of states in model M by a relative phase
401: \begin{equation}\label{eq:brotherphase}
402: \widetilde{\varepsilon}~=~e^{-2\pi \I (k_1 t_2 - k_2 t_1)V_2\cdot\Delta V_1}\;.
403: \end{equation}
404: According to the nomenclature `brother models', the relative phase
405: $\widetilde{\varepsilon}$ will be referred to as `brother phase'. It is
406: straightforward to see that the same relative phase occurs for oscillator
407: states, and the derivation can be repeated for shifts satisfying
408: \eqref{eq:newmodularinv} rather than \eqref{eq:strongmodularinv2}, yielding the
409: same qualitative result.
410:
411: The (brother) phase $\widetilde{\varepsilon}$ has certain properties and the
412: fact that it can be non-trivial has important consequences. First of all,
413: $\widetilde{\varepsilon}$ depends on the definition of the model M$'$, i.e.\ on
414: the lattice vectors $(\Delta V_1,\Delta V_2)$. Furthermore, it clearly depends
415: on the constructing element $g$ and on the commuting element $h$,
416: \begin{equation}
417: \widetilde{\varepsilon}~=~ \widetilde{\varepsilon}(g,h)\;.
418: \end{equation}
419: It follows from the construction that the brother phase vanishes for $g =
420: (\mathbbm{1}, 0)$, i.e.\ for the untwisted sector. Thus the gauge group and the
421: untwisted sector coincide for brother models. On the other hand,
422: since the brother phase does not vanish in general, the brother models M and
423: M$'$ may have different twisted sectors, and therefore be inequivalent. This result extends also to the
424: case where we subject the shifts only to the weaker constraints \eqref{eq:newmodularinv}.
425:
426: \subsubsection*{A $\boldsymbol{\mathbbm{Z}_3 \times \mathbbm{Z}_3}$ example}
427: \label{ex1}
428:
429: Let us now study an example to illustrate the results obtained so far. Consider
430: a $\mathbbm{Z}_3 \times \mathbbm{Z}_3$ orbifold of $\text{E}_8 \times
431: \text{E}_8$ with standard embedding~\cite{Font:1989aj}, i.e.\ model M is defined by
432: \begin{equation}
433: V_1~=~\frac{1}{3}\left(1,0,-1,0^{5} \right) \left( 0^{8} \right)
434: \quad\text{and}\quad
435: V_2~=~\frac{1}{3}\left(0,1,-1,0^{5} \right) \left( 0^{8} \right)\;.
436: \end{equation}
437: The resulting model has an $\text{E}_6 \times \text{U}(1)^2 \times \text{E}_8$
438: gauge group, 84 $(\boldsymbol{\overline{27}},\boldsymbol{1})$ and 243 non-abelian
439: singlets with non-zero U(1) charges.\footnote{There are three additional singlets
440: $|q\rangle_\mathrm{R}\otimes\widetilde{\alpha}^i_{-1}|0\rangle_\mathrm{L}$ from the
441: 10d SUGRA multiplet. All orbifold spectra are computed using~\cite{Orbifolder}.} Now define the brother model M$'$ by
442: \begin{equation}
443: \Delta V_1~=~\left(0,-1,0,1,0^{4}\right) \left( 0^{8} \right)
444: \quad\text{and}\quad
445: \Delta V_2~=~\left(1,0,0,0,1,0^{3}\right) \left( 0^{8} \right)\;,
446: \end{equation}
447: which fulfill the conditions~\eqref{eq:modinvconditions}.
448: From equation~\eqref{eq:brotherphase} we find the following non-trivial brother
449: phase
450: \begin{equation}\label{eq:brotherphasez3xz3}
451: \widetilde{\varepsilon}(g,h)
452: ~=~
453: \widetilde{\varepsilon}(\theta^{k_1}\omega^{k_2},\theta^{t_1}\omega^{t_2})
454: ~=~e^{\frac{2\pi\I}{3}\, (k_1\, t_2 - k_2\, t_1)}.
455: \end{equation}
456: As expected, the gauge group and the untwisted matter of model M$'$ remain the
457: same as in model M.
458: However, the twisted sectors get modified. The total number of generations is
459: reduced to 3 $(\boldsymbol{\overline{27}},\boldsymbol{1})$ and 27
460: $(\boldsymbol{27},\boldsymbol{1})$. The number of singlets remains the same as
461: before, but their localization properties change.
462:
463: Model M$'$ is not an unknown construction, but has been studied in the
464: literature in the context of $\mathbbm{Z}_3 \times \mathbbm{Z}_3$ orbifolds with
465: discrete torsion \cite{Font:1988mk}. As we shall see, the brother phase,
466: equation~\eqref{eq:brotherphasez3xz3}, is nothing but the discrete torsion phase (equation~(4) in
467: Ref.~\cite{Font:1988mk}). To make this statement more precise, we
468: briefly review discrete torsion in section \ref{sec:discretetorsionphase}, and analyze
469: its relation to the brother phase in section \ref{sec:comparison}.
470:
471:
472: \subsection[Discrete torsion phase for $\mathbbm{Z}_{N} \times \mathbbm{Z}_{M}$ orbifolds]{Discrete
473: torsion phase for $\boldsymbol{\mathbbm{Z}_{N} \times \mathbbm{Z}_{M}}$ orbifolds}
474: \label{sec:discretetorsionphase}
475:
476: Let us start with a brief review of discrete torsion in
477: orbifolds, following Vafa~\cite{Vafa:1986wx}.
478: %
479: The one-loop partition function $Z$ for a $\mathbbm{Z}_{N}\times \mathbbm{Z}_{M}$
480: orbifold has the overall structure
481: \begin{equation}
482: Z~=~\sum_{\substack{g,h \\ [g,h]=0}} \varepsilon (g,h)\, Z(g,h)\;,
483: \end{equation}
484: where the sum runs over pairs of commuting space group elements $g,h\in
485: \mbox{S}$ and the $\varepsilon (g,h)$ are relative phases between the different
486: terms in the partition function and thus between the different sectors.
487: Different assignments of phases lead, in general, to different orbifold models.
488:
489: Modular invariance strongly constrains the torsion phases
490: \cite{Vafa:1986wx}:
491: \begin{subequations}\label{eq:phaseconstraints}
492: \begin{eqnarray}
493: \varepsilon(g_{1} g_{2},g_{3}) &=&
494: \varepsilon(g_{1},g_{3})\, \varepsilon(g_{2},g_{3})\;,
495: \label{eq:phaseconstraints1} \\
496: \varepsilon(g_{1},g_{2}) &=&
497: \varepsilon(g_{2},g_{1})^{-1}\;. \label{eq:phaseconstraints2}
498: \end{eqnarray}
499: Further, we use the convention
500: \begin{equation}
501: \label{eq:phaseconstraints3} \varepsilon(g,g)
502: ~=~ 1 \; .
503: \end{equation}
504:
505: At two--loop, the partition function allows to switch on analogous phases,
506: $\varepsilon(g_{1},h_{1};g_{2},h_{2})$. From the requirement of factorizability
507: of the two--loop partition function one infers \cite{Vafa:1986wx}
508: \begin{equation}
509: \varepsilon(g_{1},h_{1};g_{2},h_{2})
510: ~=~\varepsilon(g_{1},h_{1})\,\varepsilon(g_{2},h_{2})\;.
511: \end{equation}
512: \end{subequations}
513:
514: Following the discussion of Ref.~\cite{Font:1988mk}, in orbifolds without Wilson lines $g,h$ are chosen to be
515: elements of the point group $P$. In $\mathbbm{Z}_N$ orbifolds, due to this choice and
516: equations~\eqref{eq:phaseconstraints} the phases have to be trivial,
517: \begin{equation}
518: \varepsilon(g,h)~=~1 \qquad \forall g,h \in P\;.
519: \end{equation}
520: Therefore, in the case of $\mathbbm{Z}_{N}$ orbifolds without Wilson lines,
521: non-trivial discrete torsion cannot be introduced.
522:
523: In $\mathbbm{Z}_{N}\times \mathbbm{Z}_{M}$ orbifolds, still without Wilson
524: lines, the situation is different because there are independent pairs of
525: elements (such that the first element is not a power of the second) which
526: commute with each other. If we take two point group elements
527: $g=\theta^{k_1}\omega^{k_2}$ and
528: $h=\theta^{t_1}\omega^{t_2}$, the equations~\eqref{eq:phaseconstraints}
529: determine the shape of the corresponding phase,
530: \begin{equation}\label{eq:torsionphase2}
531: \varepsilon(g,h)
532: ~=~
533: \varepsilon(\theta^{k_1}\omega^{k_2},\theta^{t_1}\omega^{t_2})
534: ~=~
535: e^{\frac{2\pi\I\, m}{M}(k_1 t_2-k_2 t_1)}\,\,,
536: \end{equation}
537: where $m\in \mathbbm{Z}$ \cite{Font:1988mk}. In particular, there are only $M$
538: inequivalent assignments of $\varepsilon$.
539:
540: The most important consequence of non-trivial $\varepsilon$-phases for our
541: discussion is that they modify the boundary conditions for twisted states and
542: thus change the twisted spectrum. This can be seen from the transformation phase
543: of equation~(\ref{eq:transformationphase}), which is modified in the presence of
544: discrete torsion according to
545: \begin{equation}\label{eq:modifiedtransformationphase}
546: \Phi \longmapsto \varepsilon(g,h) \,\Phi\;.
547: \end{equation}
548:
549: \subsection{Brother models versus discrete torsion}
550: \label{sec:comparison}
551:
552: Let us now come back to the task of establishing the relation between the
553: discrete torsion phase and the brother phase as introduced in section
554: \ref{sec:brothermodel}. From equations~\eqref{eq:brotherphase} and~\eqref{eq:torsionphase2} it is clear
555: that both phases can be made to coincide.
556: More precisely, since $V_2$ can be written as $V_{2}=\frac{\lambda_{2}}{M}$ with
557: $\lambda_{2} \in \Lambda$ (cf.\ equation~\eqref{eq:latticeconditions}), one can
558: achieve
559: \begin{equation}\label{eq:mdef}
560: -V_{2} \cdot \Delta V_{1}
561: ~=~
562: \frac{m}{M}
563: \end{equation}
564: for an appropriate choice of $\Delta V_1\in\Lambda$. Since the solutions to the
565: mass equations and the projection conditions are the same in a model with
566: discrete torsion and a brother model, whose associated phases fulfill equation~\eqref{eq:mdef}, the spectra of
567: both models coincide. We will therefore regard both models as \emph{equivalent}.
568: This means that introducing a discrete torsion phase, equation~\eqref{eq:torsionphase2}, is equivalent to
569: changing the gauge embedding according to
570: \begin{equation}
571: (V_1,V_2) ~ \to ~ (V_1+\Delta V_1,V_2+\Delta V_2)
572: \end{equation}
573: with $\Delta V_i\in\Lambda$ and $-V_{2} \cdot \Delta V_{1}=m/M$. In particular,
574: the assignment of discrete torsion to a given $\mathbbm{Z}_N\times\mathbbm{Z}_M$
575: model is a `gauge-dependent' statement in the sense that torsion can be traded
576: for changing the gauge embedding (cf.\ Fig.~\ref{fig:MirageTorsion}).
577:
578:
579: \begin{figure}[!t]
580: \centerline{\includegraphics{MirageTorsion.eps}}
581: \caption{Models with non-trivial discrete torsion have an equivalent
582: description as models with trivial discrete torsion but a different gauge
583: embedding.}
584: \label{fig:MirageTorsion}
585: \end{figure}
586:
587:
588: To illustrate our result, we construct the standard embedding models for
589: $\mathbbm{Z}_N \times \mathbbm{Z}_M$ orbifolds with an $\mathrm{E}_8 \times
590: \mathrm{E}_8$ lattice of Ref.~\cite{Font:1988mk} with discrete torsion in terms of
591: non-standard embedding shifts without discrete torsion (brother models). We use
592: the following recipe to construct brother models, i.e.\ mimic models with
593: discrete torsion:\\
594: For a given set of shifts $V_1$ and $V_2$ fulfilling the modular invariance conditions,
595: find a new set of shifts $V_1'=V_1+\Delta V_1$ and $V_2'=V_2+\Delta V_2$
596: with the following properties:
597: \renewcommand{\labelenumi}{(\roman{enumi})}
598: \begin{enumerate}
599: \item the new shifts differ from the original set only by lattice vectors,
600: i.e.\ $\Delta V_1,\Delta V_2\in\Lambda$
601: \item the new shifts also fulfill the modular invariance conditions,
602: and
603: \item the `interference term' $V_2\cdot \Delta V_1$ is not an integer.
604: \end{enumerate}
605: In practice (and for any $N,M$), the above properties can be expressed in
606: terms of linear Diophantine equations for which we always find solutions.
607:
608: Possible choices for the shifts $(V_1+\Delta V_1,V_2+\Delta V_2)$ are shown in
609: Tab.~\ref{tab:DiscreteTorsionBrothers}, where we list the shifts of torsionless
610: models equivalent to the discrete torsion model of Ref.~\cite{Font:1988mk}.
611:
612:
613: \begin{table}[!h]
614: \begin{center}
615: \begin{tabular}{|l|r|l|l|}
616: \hline
617: orbifold & torsion $\varepsilon$ & shift $V_1$ & shift $V_2$\\
618: \hline\hline
619: $\Z2\times\Z2$ &
620: $1$ &
621: $\displaystyle \left(\tfrac{1}{2},0,-\tfrac{1}{2},0,0,0,0,0\right)$ &
622: $\displaystyle \left(0,\tfrac{1}{2},-\tfrac{1}{2},0 ,0, 0, 0, 0\right)$\\
623: & $-1$ &
624: $\displaystyle \left(\tfrac{1}{2}, -1,-\tfrac{1}{2},1,0,0,0,0\right)$ &
625: $\displaystyle \left(1,\tfrac{1}{2},-\tfrac{1}{2},0,1,0,0,0\right)$\\
626: \hline
627: $\Z4\times\Z2$ &
628: $1$ &
629: $\displaystyle \left(\tfrac{1}{4},0,-\tfrac{1}{4},0,0,0,0,0\right)$ &
630: $\displaystyle \left(0,\tfrac{1}{2},-\tfrac{1}{2},0,0,0,0,0\right)$ \\
631: & $-1$ &
632: $\displaystyle \left(\tfrac{1}{4},-1,-\tfrac{1}{4},1,0,0,0,0\right)$ &
633: $\displaystyle \left(2,\tfrac{1}{2},-\tfrac{1}{2},0,0,0,0,0\right)$ \\
634: \hline
635: $\Z6\times\Z2$ &
636: $1$ &
637: $\displaystyle \left(\tfrac{1}{6},0,-\tfrac{1}{6},0,0,0,0,0\right)$ &
638: $\displaystyle \left(0,\tfrac{1}{2},-\tfrac{1}{2},0 ,0, 0, 0, 0\right)$\\
639: & $-1$ &
640: $\displaystyle \left(\tfrac{1}{6},-1,-\tfrac{1}{6},1,0,0,0,0\right)$ &
641: $\displaystyle \left(3,\tfrac{1}{2},-\tfrac{1}{2},0,1,0,0,0\right)$ \\
642: \hline
643: $\Z6'\times\Z2$ &
644: $1$ &
645: $\displaystyle \left(\tfrac{1}{6},\tfrac{1}{6},-\tfrac{1}{3},0,0,0,0,0\right)$ &
646: $\displaystyle \left(\tfrac{1}{2},0,-\tfrac{1}{2},0 ,0, 0, 0, 0\right)$\\
647: & $-1$ &
648: $\displaystyle \left(-\tfrac{5}{6},\tfrac{7}{6},-\tfrac{1}{3},1,1,0,0,0\right)$ &
649: $\displaystyle \left(\tfrac{1}{2},3,-\tfrac{1}{2},1,0,0,0,0\right)$\\
650: \hline
651: $\Z3\times\Z3$ &
652: $1$ &
653: $\displaystyle \left(\tfrac{1}{3},0,-\tfrac{1}{3},0,0,0,0,0\right)$ &
654: $\displaystyle \left(0,\tfrac{1}{3},-\tfrac{1}{3},0 ,0, 0, 0, 0\right)$\\
655: & $\mathrm{e}^{2\pi\I\tfrac{1}{3}}$ &
656: $\displaystyle \left(\tfrac{1}{3},-1,-\tfrac{1}{3},1,0,0,0,0\right)$ &
657: $\displaystyle \left(1,\tfrac{1}{3},-\tfrac{1}{3},0,1,0,0,0\right)$ \\
658: & $\mathrm{e}^{2\pi\I\tfrac{2}{3}}$ &
659: $\displaystyle \left(\tfrac{1}{3},-2,-\tfrac{1}{3},0,0,0,0,0\right)$ &
660: $\displaystyle \left(2,\tfrac{1}{3},-\tfrac{1}{3},0,0,0,0,0\right)$ \\
661: \hline
662: $\Z6\times\Z3$ &
663: $1$ &
664: $\displaystyle \left(\tfrac{1}{6},0,-\tfrac{1}{6},0,0,0,0,0\right)$ &
665: $\displaystyle \left(0,\tfrac{1}{3},-\tfrac{1}{3},0 ,0, 0, 0, 0\right)$\\
666: & $\mathrm{e}^{2\pi\I\tfrac{1}{3}}$ &
667: $\displaystyle \left(\tfrac{1}{6},-1,-\tfrac{1}{6},1,0,0,0,0\right)$ &
668: $\displaystyle \left(2,\tfrac{1}{3},-\tfrac{1}{3},0,0, 0, 0, 0\right)$\\
669: & $\mathrm{e}^{2\pi\I\tfrac{2}{3}}$ &
670: $\displaystyle \left(\tfrac{1}{6},-2,-\tfrac{1}{6},0,0,0,0,0\right)$ &
671: $\displaystyle \left(4,\tfrac{1}{3},-\tfrac{1}{3},0,0, 0, 0, 0\right)$\\
672: \hline
673: $\Z4\times\Z4$ &
674: $1$ &
675: $\displaystyle \left(\tfrac{1}{4},0,-\tfrac{1}{4},0,0,0,0,0\right)$ &
676: $\displaystyle \left(0,\tfrac{1}{4},-\tfrac{1}{4},0,0,0,0,0\right)$\\
677: & $\I$ &
678: $\displaystyle \left(\tfrac{1}{4},-1,-\tfrac{1}{4},1,0,0,0,0\right)$ &
679: $\displaystyle \left(1,\tfrac{1}{4},-\tfrac{1}{4},0,1,0,0,0\right)$\\
680: & $-1$ &
681: $\displaystyle \left(\tfrac{1}{4},-2,-\tfrac{1}{4},0,0,0,0,0\right)$ &
682: $\displaystyle \left(2,\tfrac{1}{4},-\tfrac{1}{4},0,0,0,0,0\right)$\\
683: & $-\I$ &
684: $\displaystyle \left(\tfrac{1}{4},-3,-\tfrac{1}{4},1,0,0,0,0\right)$ &
685: $\displaystyle \left(3,\tfrac{1}{4},-\tfrac{1}{4},0,1,0,0,0\right)$\\
686: \hline
687: $\Z6\times\Z6$ &
688: $1$ &
689: $\displaystyle \left(\tfrac{1}{6},0,-\tfrac{1}{6},0,0,0,0,0\right)$ &
690: $\displaystyle \left(0,\tfrac{1}{6},-\tfrac{1}{6},0,0,0,0,0\right)$\\
691: & $\mathrm{e}^{2\pi\I\tfrac{1}{6}}$ &
692: $\displaystyle \left(\tfrac{1}{6},-1,-\tfrac{1}{6},1,0,0,0,0\right)$ &
693: $\displaystyle \left(1,\tfrac{1}{6},-\tfrac{1}{6},0,1,0,0,0\right)$\\
694: & $\mathrm{e}^{2\pi\I\tfrac{1}{3}}$ &
695: $\displaystyle \left(\tfrac{1}{6},-2,-\tfrac{1}{6},0,0,0,0,0\right)$ &
696: $\displaystyle \left(2,\tfrac{1}{6},-\tfrac{1}{6},0,0,0,0,0\right)$\\
697: & $-1$ &
698: $\displaystyle \left(\tfrac{1}{6},-3,-\tfrac{1}{6},1,0,0,0,0\right)$ &
699: $\displaystyle \left(3,\tfrac{1}{6},-\tfrac{1}{6},0,1,0,0,0\right)$\\
700: & $\mathrm{e}^{2\pi\I\tfrac{2}{3}}$ &
701: $\displaystyle \left(\tfrac{1}{6},-4,-\tfrac{1}{6},0,0,0,0,0\right)$ &
702: $\displaystyle \left(4,\tfrac{1}{6},-\tfrac{1}{6},0,0,0,0,0\right)$\\
703: & $\mathrm{e}^{2\pi\I\tfrac{5}{6}}$ &
704: $\displaystyle \left(\tfrac{1}{6},-5,-\tfrac{1}{6},1,0,0,0,0\right)$ &
705: $\displaystyle \left(5,\tfrac{1}{6},-\tfrac{1}{6},0,1,0,0,0\right)$\\
706: \hline
707: \end{tabular}
708: \caption{$\Z{N}\times\Z{M}$ models with discrete torsion and standard embedding are equivalent to
709: models without discrete torsion and non-standard embedding. We write the torsion phase factor as
710: $\varepsilon=\mathrm{e}^{-2\pi\I\, V_2\cdot\Delta V_1}$. The components of the shifts within the second
711: $\E8$ all vanish. This result also applies to orbifold models in SO(32).}
712: \label{tab:DiscreteTorsionBrothers}
713: \end{center}
714: \end{table}
715:
716: Our result has important consequences for the classification of
717: $\mathbbm{Z}_N\times\mathbbm{Z}_M$ orbifolds. Introducing a discrete torsion
718: phase in the sense of Ref.~\cite{Font:1988mk} does not lead to new models. That
719: is, all models with this discrete torsion can be equivalently obtained by
720: scanning over torsionless models only. This will be important for our
721: classification in section~\ref{sec:ClassifZ3xZ3}.
722:
723: It is also instructive to interpret the equivalence between discrete torsion and
724: changing the gauge embedding in terms of geometry. Discrete torsion can be
725: regarded as a property of the 6D compact space while changing the gauge
726: embedding affects the (left-moving) coordinates of the gauge lattice only. Hence
727: one might argue that discrete torsion and choosing a different gauge embedding
728: are two different features of orthogonal dimensions. However, by embedding the
729: `spatial' twist in the gauge degrees of freedom, these features get combined in
730: such a way that it is no longer possible to make a clear separation. Using a
731: more technical language one might rephrase this statement by saying that, since
732: physical states arise from tensoring left- and right-movers together, the phases
733: $\varepsilon$ and $\widetilde{\varepsilon}$ cannot be distinguished.
734: Consequently, properties of the zero-modes cannot be ascribed neither to the gauge
735: embedding alone nor to the presence of discrete torsion, but only to both.
736:
737:
738: \section{Generalized discrete torsion}
739: \label{sec:GeneralizationDTAndBrothers}
740:
741: The results of the previous section can be generalized. To see this, we first
742: generalize the brother phase of section~\ref{sec:brothermodel} for orbifolds
743: with Wilson lines. In a second step, we compare the emerging phases to what is
744: known as generalized discrete torsion \cite{Gaberdiel:2004vx}. As before, we can
745: relate both phases.
746:
747:
748: \subsection{Generalized brother models}
749: \label{sec:GeneralizedBrothers}
750:
751: Let us turn to the discussion of orbifolds with Wilson lines~\cite{Ibanez:1986tp}. A (torsionless) model
752: M is defined by $(V_1,V_2,A_\alpha)$. A brother model M$'$ appears by adding lattice vectors to
753: the shifts and Wilson lines, i.e.\ M$'$ is defined by
754: \begin{equation}
755: (V'_1, V'_2, A'_\alpha)~=~(V_1+\Delta V_1, V_2+\Delta V_2, A_\alpha+\Delta A_\alpha)\;,
756: \end{equation}
757: with $\Delta V_i,\Delta A_\alpha \in \Lambda$. From the
758: conditions~\eqref{eq:newmodularinv}, the choice of lattice vectors $(\Delta
759: V_i,\Delta A_\alpha)$ is constrained by
760: \begin{subequations}\label{eq:newModularInvforDelta}
761: \begin{eqnarray}
762: M \left(V_1\cdot\Delta V_2 + V_2\cdot\Delta V_1 + \Delta V_1 \cdot\Delta V_2 \right) & = & 0\text{ mod }
763: 2~\equiv~ 2\,x \;,\label{eq:newModularInvforDeltaVV}\\
764: N_\alpha \left(V_i\cdot\Delta A_\alpha + A_\alpha\cdot\Delta V_i + \Delta V_i \cdot\Delta A_\alpha
765: \right) & = & 0\text{ mod }
766: 2~\equiv~ 2\,y_{i\alpha} \;,\label{eq:newModularInvforDeltaVA}\\
767: Q_{\alpha\beta} \left(A_\alpha\cdot\Delta A_\beta + A_\beta\cdot\Delta A_\alpha + \Delta A_\alpha\cdot\Delta A_\beta
768: \right) & = & 0\text{ mod }
769: 2~\equiv~ 2\,z_{\alpha\beta}\;,\label{eq:newModularInvforDeltaAA}
770: \end{eqnarray}
771: \end{subequations}
772: where $x,\,y_{i\alpha},\,z_{\alpha\beta}\,\in{\mathbb Z}$.
773:
774: Repeating the steps of section~\ref{sec:brothermodel} one arrives at a
775: `generalized brother phase'
776: \begin{eqnarray}
777: \widetilde{\varepsilon}&=&\exp\left\{-2\pi \I\,\left[
778: (k_1\, t_2-k_2\, t_1)\left(V_{2} \cdot \Delta V_{1} - \frac{x}{M}\right) \phantom{\frac{z_{\alpha\beta}}{Q_{\alpha\beta}}}\right.\right.\nonumber\\*
779: & & \hphantom{\exp\left\{\right.}{}\left.\left.
780: + (k_1\, m_{\alpha} - t_1\, n_{\alpha})\left(A_{\alpha} \cdot \Delta V_{1}-\frac{y_{1\alpha}}{N_\alpha}\right)\right.\right.
781: \nonumber\\*
782: & & \hphantom{\exp\left\{\right.}{}\left.\left.
783: + (k_2\, m_{\alpha}-t_2\, n_{\alpha})\left(A_{\alpha} \cdot \Delta
784: V_{2}-\frac{y_{2\alpha}}{N_\alpha}\right) \right.\right.
785: \nonumber\\*
786: & & \hphantom{\exp\left\{\right.}{}\left.\left.
787: + n_{\alpha}\, m_{\beta}\left( A_{\beta} \cdot \Delta
788: A_{\alpha}-\frac{z_{\alpha\beta}}{Q_{\alpha\beta}}\right)\right]
789: \right\}\;,
790: \label{eq:generalbrotherphase}
791: \end{eqnarray}
792: corresponding to the constructing element
793: $g=(\theta^{k_1}\omega^{k_2},n_{\alpha}e_{\alpha})$ and the commuting element
794: $h=(\theta^{t_1}\omega^{t_2},m_{\alpha}e_{\alpha})$. One can see that $D_{\alpha\beta}\equiv A_\beta\cdot\Delta
795: A_\alpha - z_{\alpha\beta}/Q_{\alpha\beta}$ is (almost) antisymmetric in $\alpha,\,\beta$,
796: \begin{equation}
797: D_{\alpha\beta} = -D_{\beta\alpha}\text{ mod }1\,.
798: \end{equation}
799:
800: Notice that also in the case of orbifolds with lattice-valued Wilson lines,
801: $A_\alpha\in\Lambda$, the last three terms of
802: equation~\eqref{eq:generalbrotherphase} can be non-trivial, giving rise to new
803: brother models.
804:
805: \subsubsection*{Brother models in $\boldsymbol{\mathbbm{Z}_N}$ orbifolds}
806:
807: From equation~\eqref{eq:generalbrotherphase}, it is clear that the generalized brother phase is also
808: important for $\mathbbm{Z}_N$ orbifolds. More precisely, in $\mathbbm{Z}_N$ orbifolds with Wilson lines,
809: the second and fourth lines of equation~\eqref{eq:generalbrotherphase} are not always trivial and thus
810: also lead to brother models.
811:
812: Let us illustrate this with an example in $\mathbbm{Z}_4$ with twist
813: $v=\frac{1}{4}(-2,\,1,\,1;\,0)$ acting on the compactification lattice $\Gamma=$
814: SO(4)$^3$, and standard embedding~\cite{Katsuki:1990bf}.
815: \label{ex2}
816: The gauge group is \E{6}$\times$SU(2)$\times$\E{8}. By turning on the
817: lattice-valued Wilson lines
818: \begin{equation}
819: A_1=\left(0^8 \right)\left(1^2,0^6\right), \quad\quad\quad A_5=A_6=\left(0^8 \right)\left(0,1^2,0^5\right),
820: \end{equation}
821: a non-trivial generalized brother phase with $D_{15}=D_{16}=-\frac{1}{2}$ is introduced. The untwisted
822: and first twisted sectors remain unchanged, but the number of (anti-) families in
823: the second twisted sector is reduced from 10 $(\boldsymbol{\overline{27}},\boldsymbol{1},\boldsymbol{1})$
824: + 6 $(\boldsymbol{27},\boldsymbol{1},\boldsymbol{1})$ to 6 $(\boldsymbol{\overline{27}},\boldsymbol{1},\boldsymbol{1})$
825: + 2 $(\boldsymbol{27},\boldsymbol{1},\boldsymbol{1})$.
826:
827: \subsection{Generalized discrete torsion}
828: \label{sec:GeneralizedDiscreteTorsion}
829:
830: In section~\ref{sec:discretetorsionphase} we have discussed the discrete torsion
831: phase as introduced in Ref.~\cite{Font:1988mk}. More recently, this concept has
832: been extended by introducing a generalized discrete torsion phase in the context
833: of type IIA/B string theory~\cite{Gaberdiel:2004vx}. This generalized torsion
834: phase depends on the fixed points of the orbifold. It weights differently terms
835: in the partition function corresponding to the same twisted sector but different
836: fixed points, and is constrained by modular invariance.
837:
838: Following the steps of section~\ref{sec:discretetorsionphase} and considering
839: $g,h\in S$, we write down the general solution of
840: equations~\eqref{eq:phaseconstraints} for the discrete torsion phase
841: as\footnote{Note that we employ the stronger
842: constraints~\eqref{eq:phaseconstraints} rather than the conditions presented
843: in~\cite{Gaberdiel:2004vx}. It might be possible to relax
844: condition~\eqref{eq:phaseconstraints2}, in which case additional possibilities
845: could arise. We ignore this possibility in the present study.}
846: \begin{equation}\label{eq:generalizedtorsionphase}
847: \varepsilon(g,h)~=~e^{2 \pi \I\,[a\, (k_1\, t_2 - k_2\, t_1)
848: + b_{\alpha}\, (k_1\, m_{\alpha} - t_1\, n_{\alpha})
849: + c_{\alpha}\, (k_2\, m_{\alpha} - t_2\, n_{\alpha})
850: + d_{\alpha \beta}\, n_{\alpha}\, m_{\beta}]}\;.
851: \end{equation}
852: Modular invariance constrains the values of $a,b_{\alpha},c_{\alpha},d_{\alpha
853: \beta}$. Therefore $a=\widetilde{a}/M,\,b_\alpha =
854: \widetilde{b}_\alpha/N_\alpha$, $c_\alpha = \widetilde{c}_\alpha/N_\alpha$,
855: $d_{\alpha\beta}=\widetilde{d}_{\alpha\beta}/N_{\alpha\beta}$ with
856: $\widetilde{a},\,\widetilde{b}_\alpha$, $\widetilde{c}_\alpha$,
857: $\widetilde{d}_{\alpha\beta} \in \mathbbm{Z}$, $N_{\alpha\beta}$ being the
858: greatest common divisor of $N_\alpha$ and $N_\beta$. In addition, $d_{\alpha
859: \beta}$ must be antisymmetric in $\alpha,\,\beta$.
860:
861: The parameters $b_\alpha$, $c_\alpha$, $d_{\alpha\beta}$ are additionally
862: constrained by the geometry of the orbifold. It is not hard to see that if
863: $e_\alpha \simeq e_\beta$ on the orbifold, then $b_\alpha=b_\beta,\,
864: c_\alpha=c_\beta$ and $d_{\alpha\beta}=0$ must hold (cf.\ the examples below).
865:
866: The generalized discrete torsion is not restricted only to $\mathbbm{Z}_N\times\mathbbm{Z}_M$
867: orbifolds, as the usual discrete torsion was, but will likewise appear in the $\mathbbm{Z}_N$
868: case. Clearly, since in $\mathbbm{Z}_N$ orbifolds there is only one shift, the parameters $a$ and
869: $c_\alpha$ vanish.
870:
871: \subsubsection*{Examples}
872: \label{ex3}
873:
874: Let us consider the $\mathbbm{Z}_3\times \mathbbm{Z}_3$ orbifold compactified on
875: an $\SU3^3$ lattice. In this case we have $e_1 \simeq e_2,\, e_3 \simeq e_4$
876: and $e_5 \simeq e_6$ on the orbifold. This implies that there are only three
877: independent $b_\alpha$, namely $b_1,\,b_3,\,b_5$, while $b_2 = b_1,\,b_4 =
878: b_3,\,b_6 = b_5$. Analogously, only $c_1,\,c_3,\,c_5$ are independent. Further,
879: the antisymmetric matrix $d_{\alpha\beta}$ takes the form
880: \begin{equation}
881: d_{\alpha\beta}~=~ \left(
882: \begin{array}{cccccc}
883: 0&0&d_1&d_1&d_2&d_2\\
884: 0&0&d_1&d_1&d_2&d_2\\
885: -d_1&-d_1&0&0&d_3&d_3\\
886: -d_1&-d_1&0&0&d_3&d_3\\
887: -d_2&-d_2&-d_3&-d_3&0&0\\
888: -d_2&-d_2&-d_3&-d_3&0&0
889: \end{array}
890: \right)\;.
891: \label{eq:dmatrixZ3xZ3}
892: \end{equation}
893: Including the parameter $a$, there are 10 independent discrete torsion
894: parameters, which can take values $0$, $\tfrac{1}{3}$ or $\tfrac{2}{3}$.
895:
896: For the $\mathbbm{Z}_2 \times \mathbbm{Z}_2$ orbifold on an $\SU2^6$ lattice an
897: analogous consideration shows that there are no restrictions for the discrete
898: torsion parameters. Therefore, there are $1+6+6+15=28$ independent parameters
899: $a,\,b_\alpha,\,c_\alpha,\,d_{\alpha\beta}$, with values either 0 or
900: $\tfrac{1}{2}$. However, since the coefficients $n_\alpha m_\beta$ of
901: $d_{\alpha\beta}$ for $(\alpha,\,\beta)\in\{(1,2),\,(3,4),\,(5,6)\}$ vanish, the
902: corresponding $d_{\alpha\beta}$ are not physical, leading to 25 effective
903: parameters.
904:
905: \subsubsection*{Generalized discrete torsion and local spectra}
906:
907: In order to understand the action of the generalized discrete torsion, let us
908: consider the following example.
909: \label{ex4}
910: We start with the $\mathbbm{Z}_3\times\mathbbm{Z}_3$ standard embedding without
911: Wilson lines, $A_\alpha=0$, and switch on the discrete torsion phase,
912: equation~\eqref{eq:generalizedtorsionphase}, with $b_3=b_4=\frac{1}{3}$. The
913: total number of families is reduced from 84
914: $(\boldsymbol{\overline{27}},\boldsymbol{1})$ to 24
915: $(\boldsymbol{\overline{27}},\boldsymbol{1})$ and 12
916: $(\boldsymbol{27},\boldsymbol{1})$.
917:
918: Due to its form, the discrete torsion phase
919: $\varepsilon=e^{2\pi\I\,b_{\alpha}\,(k_1\, m_{\alpha}-t_1\,n_\alpha)}$
920: distinguishes between different fixed points of a particular twisted sector.
921: That is, generalized discrete torsion can be thought of as a local feature. In
922: general, the additional phase at a given fixed point coincides with a brother
923: phase of the torsionless model (cf.\ first term of
924: equation~\eqref{eq:generalbrotherphase}), i.e.\ locally one can find $\Delta
925: V_i$ such that
926: \begin{equation}\label{eq:epsilonloc}
927: \varepsilon~=~e^{2\pi\I\,b_{\alpha}\,(k_1\, m_{\alpha}-t_1\,n_\alpha)}
928: ~=~e^{-2\pi\I\,(k_1\, t_2-k_2\, t_1)
929: \left(V_2\cdot\Delta V_1 - \frac{x}{3}\right)}
930: \end{equation}
931: with appropriate $x$.
932: Then, each local spectrum coincides with the local spectrum of some brother
933: model. The interpretation of generalized discrete torsion in terms of
934: `localized discrete torsion' parallels the concept of local shifts (cf.\
935: \cite{Buchmuller:2004hv,Buchmuller:2006ik}) in orbifolds with Wilson lines.
936:
937: Note that $\Delta V_i$ as in \eqref{eq:epsilonloc} cannot be found for twisted
938: sectors where $b_\alpha$ corresponds to a direction $e_\alpha$ of a fixed torus,
939: where $b_\alpha$ projects out all states of the sector.
940:
941: \begin{figure}[!t!]
942: \centerline{\subfigure[{}]{\includegraphics{TorsionPillow1.eps}}
943: \qquad
944: \subfigure[{}]{\includegraphics{TorsionPillow2.eps}}}
945: \caption{Sketch of a (2D) \SU3 plane of a $\mathbbm{Z}_3\times\mathbbm{Z}_3$
946: orbifold (the second plane in the example). Parts (`corners') from different
947: brother models (a) can be `sewed together' to a model in which the torsion phase
948: differs for different fixed points. This is equivalent to switching on the
949: generalized discrete torsion phase $b_\alpha$ (b).}
950: \label{fig:puzzle}
951: \end{figure}
952:
953: For concreteness, we first focus on the three fixed points in the second torus of the $T_{(0,1)}$
954: twisted sector. As depicted in Fig.~\ref{fig:puzzle}, the local spectra of the three brother
955: models, $a\equiv -\left(V_2\cdot\Delta V_1 - \frac{x}{3}\right)=0,\,\frac{1}{3},\,\frac{2}{3}$, can be combined consistently into one model with
956: $b_3=b_4=\frac{1}{3}$. On the other hand, in the $T_{(1,0)}$
957: twisted sector there is a fixed torus in the directions $e_3,\,e_4$; thus the sector is empty.
958:
959: This procedure can also be applied to the terms $c_\alpha$ and $d_{\alpha\beta}$ of the generalized
960: discrete torsion phase, equation~\eqref{eq:generalizedtorsionphase}.
961:
962: \subsubsection*{Generalized brother models versus generalized discrete torsion}
963:
964: As in our previous discussion in chapter 3, also the generalized versions of the
965: discrete torsion phase and the brother phase have a very similar form. Indeed,
966: whenever there are non-trivial solutions to
967: equations~\eqref{eq:newModularInvforDelta}, one can equivalently describe models
968: with generalized discrete torsion phase in terms of generalized brother models.
969: This is the generic case.
970:
971: However, there are exceptions. Namely, as we will explain below, models with
972: $d_{\alpha\beta}\neq 0$ in $\mathbbm{Z}_3\times\mathbbm{Z}_3$ orbifolds without Wilson lines cannot be
973: interpreted in terms of brother models.
974:
975: Consider the fourth part of the generalized discrete torsion phase of
976: equation~\eqref{eq:generalizedtorsionphase},
977: \begin{equation}\label{eq:dtermdiscretetorsion}
978: \varepsilon~=~e^{2\pi\I\,d_{\alpha\beta}\,n_{\alpha} m_{\beta}}\;,
979: \end{equation}
980: with $d_{\alpha\beta}\in \left\lbrace 0,\tfrac{1}{3},\tfrac{2}{3}\right\rbrace$. An analogous
981: term appears in the generalized brother phase as
982: \begin{equation}\label{eq:dtermbrotherphase}
983: \widetilde{\varepsilon}~=~\exp\left[-2 \pi \I\, n_{\alpha}m_{\beta}\left(A_\beta \cdot \Delta
984: A_{\alpha}-\frac{z_{\alpha\beta}}{Q_{\alpha\beta}} \right)\right]\;,
985: \end{equation}
986: where $Q_{\alpha\beta}~=~3$, since the Wilson lines have order 3. In general, both phases can be made
987: coincide by choosing $\Delta A_{\alpha}\in \Lambda$ such that
988: \begin{equation}
989: -\left(A_\beta \cdot \Delta A_{\alpha}-\frac{z_{\alpha\beta}}{3} \right)=d_{\alpha\beta}\;.
990: \end{equation}
991: On the other hand, in the case when $A_\alpha=0$ and $\Delta A_\alpha\neq 0$,
992: equation~\eqref{eq:dtermbrotherphase} simplifies to
993: \begin{equation}\label{eq:dtermbrotherphaseZ3xZ3}
994: \widetilde{\varepsilon}
995: ~=~e^{2\pi \I\, n_{\alpha}m_{\beta}\left(\frac{z_{\alpha\beta}}{3}\right)}
996: ~=~e^{2\pi \I\, n_{\alpha}m_{\beta}\left(\frac{\Delta A_\alpha\cdot\Delta A_\beta}{2} \right)}\;,
997: \end{equation}
998: where the second equality follows from the definition of $z_{\alpha\beta}$,
999: equation~\eqref{eq:newModularInvforDeltaAA}. As $\Delta A_\alpha$
1000: are lattice vectors, this equality can only hold if $z_{\alpha\beta}=0\text{ mod }3$, which implies
1001: that the brother phase equation~\eqref{eq:dtermbrotherphaseZ3xZ3} is trivial. Thus, in this case, the
1002: generalized discrete torsion phase leads to models which cannot arise by adding lattice vectors to shifts
1003: and Wilson lines.
1004:
1005:
1006: In summary, the generalized discrete torsion phases admit more possible
1007: assignments than the generalized brother phases. Nevertheless, a large class of
1008: the models with generalized discrete torsion has an equivalent description in
1009: terms of models with a modified gauge embedding.
1010:
1011: Our results have important implications. By introducing generalized discrete
1012: torsion, or lattice-valued Wilson lines, one can control the local spectra. We
1013: therefore expect that introducing generalized discrete torsion, or alternatively
1014: shifting the Wilson lines by lattice vectors, will gain a similar importance as
1015: discrete Wilson lines~\cite{Ibanez:1986tp} for orbifold model building.
1016:
1017: As stated above, switching on generalized discrete torsion can lead to the
1018: disappearance of complete local spectra. This raises the question of how to
1019: interpret this fact in terms of geometry. Some of the localized zero-modes can
1020: be viewed as blow-up modes which allow to resolve the orbifold singularity
1021: associated to a given fixed point
1022: \cite{Walton:1988bu,Aspinwall:1994ev,Vafa:1994rv} (see
1023: \cite{Lust:2006zh,Honecker:2006qz,Nibbelink:2007rd} for recent developments). If
1024: at a given fixed point there are no zero modes, one might argue that, therefore,
1025: the associated singularity cannot be `blown up'. In what follows, we shall
1026: advertise an alternative interpretation.
1027:
1028: \subsection{Connection to non-factorizable orbifolds}
1029: \label{sec:Z2xZ2}
1030:
1031: We find that in many cases orbifold models M with certain geometry, i.e.\
1032: compactification lattice $\Gamma$, and generalized discrete torsion switched on
1033: are equivalent to torsionless models M$'$ based on a different lattice
1034: $\Gamma'$. Model M$'$ has less fixed points than M, and the mismatch turns out
1035: to constitute precisely the `empty' fixed points of model M.
1036:
1037:
1038: The simplest examples are based on $\mathbbm{Z}_2\times\mathbbm{Z}_2$ orbifolds
1039: with standard embedding and without Wilson lines.\label{ex5}
1040: %
1041: As compactification lattice $\Gamma$, we choose an $\SU2^6$
1042: lattice~\cite{Forste:2004ie}. As we have seen in
1043: section~\ref{sec:GeneralizedDiscreteTorsion}, in this case there are 25 physical
1044: parameters for generalized discrete torsion, with values either 0 or
1045: $\frac{1}{2}$. For concreteness, we restrict to the 12 $d_{\alpha\beta}$
1046: parameters and scan over all $2^{12}$ models.
1047:
1048: \begin{table}[b!]
1049: \def\oldwidth{\captionwidth}
1050: \setlength{\captionwidth}{0.85\textwidth}
1051: {\small{
1052: \begin{center}
1053: \begin{tabular}{|c||c|c|c|c|r||c||l|}
1054: \hline
1055: & $T_{(1,0)}$ & $T_{(0,1)}$ & $T_{(1,1)}$ & total & $\#_S$ & $d_{\alpha\beta} = \tfrac{1}{2}$ & $A_\alpha \neq 0$\\
1056: \hline
1057: \hline
1058: A.1 & $(16,0)$ & $(16,0)$ & $(16,0)$ & $(51,3)$ & 246 & $-$ & $-$ \\
1059: \hline
1060: A.2 & $(12,4)$ & $(8,0)$ & $(8,0)$ & $(31,7)$ & 166 & $d_{24}$ & $A_2=(S)(0^8)$, $A_4 =(V)(0^8)$ \\
1061: \hline
1062: A.3 & $(10,6)$ & $(4,0)$ & $(4,0)$ & $(21,9)$ & 126 & $d_{14}, d_{23}$ & $A_1=(S)(0^8)$, $A_2 =(0^8)(S)$,\\
1063: & & & & & & & $A_3=(0^8)(V)$, $A_4 =(V)(0^8)$ \\
1064: \hline
1065: A.4 & $(8,0)$ & $(8,0)$ & $(8,0)$ & $(27,3)$ & 126 & $d_{26}, d_{46}$ & $A_2 =(V')(0^8)$, $A_4 =(V)(0^8)$,\\
1066: & & & & & & & $A_6 =(S)(0^8)$ \\
1067: \hline
1068: A.5 & $(6,2)$ & $(6,2)$ & $(4,0)$ & $(19,7)$ & 106 & $d_{24}, d_{36}$ & $A_2 =(S)(0^8)$, $A_3 =(0^8)(S)$,\\
1069: & & & & & & & $A_4 =(V)(0^8)$, $A_6 =(0^8)(V)$\\
1070: \hline
1071: A.6 & $(6,2)$ & $(4,0)$ & $(4,0)$ & $(17,5)$ & 86 & $d_{16}, d_{24},$ & $A_1 =(V)(0^8)$, $A_2 =(0^8)(V)$,\\
1072: & & & & & & $d_{36}$ & $A_3 =(V')(0^8)$, $A_4 =(0^8)(S)$,\\
1073: & & & & & & & $A_6 =(S)(0^8)$ \\
1074: \hline
1075: A.7 & $(4,0)$ & $(4,0)$ & $(4,0)$ & $(15,3)$ & 66 & $d_{16}, d_{25},$ & $A_1=(V)(0^8)$, $A_2 =(0^8)(V)$,\\
1076: & & & & & & $d_{36}, d_{45}$ & $A_3 =(V')(0^8)$, $A_4 =(0^8)(V')$,\\
1077: & & & & & & & $A_5 =(0^8)(S)$, $A_6 =(S)(0^8)$\\
1078: \hline
1079: A.8 & $(3,1)$ & $(3,1)$ & $(3,1)$ & $(12,6)$ & 66 & $d_{16}, d_{24},$ & $A_1 =(W_1)(0^8)$, $A_2 =(0^8)(W_1)$,\\
1080: & & & & & & $d_{35}$ & $A_3 =(0^8)(W_1')$, $A_4 =(0^8)(W_2)$,\\
1081: & & & & & & & $A_5 =(0^8)(W_2')$, $A_6 =(W_2)(0^8)$\\
1082: \hline
1083: \end{tabular}
1084: \end{center}
1085: \caption{Survey of $\Z2\times\Z2$ orbifolds with generalized discrete torsion.
1086: The $2^\mathrm{nd}$--$4^\mathrm{th}$ columns list the number of anti-families
1087: and families, respectively, for the various twisted sectors $T_{(k_1,k_2)}$.
1088: In all models, the untwisted sector gives a contribution of $(3,3)$
1089: (anti-)families. $\#_S$ denotes the total number of singlets. These spectra
1090: can either be obtained by turning on generalized discrete torsion
1091: $d_{\alpha\beta}$ as specified in the next-to-last column, or by using
1092: lattice-valued Wilson lines $A_\alpha$ as listed in the last column. The
1093: building blocks are defined in equation \eqref{eq:buildingblocks}.}
1094: \label{tab:Z2xZ2}
1095: }}
1096: \setlength{\captionwidth}{\oldwidth}
1097: \end{table}
1098:
1099: Beside other models with a net number of zero families, we find eight models
1100: (and their mirrors, i.e.\ models where families and anti-families are exchanged).
1101: They are listed in Tab.~\ref{tab:Z2xZ2}, where we present the number of
1102: (anti-)families for each twisted sector and the total number of singlets. As
1103: discussed in section~\ref{sec:GeneralizedDiscreteTorsion}, models with
1104: non-trivial $d_{\alpha\beta}$ are equivalent to torsionless models with
1105: lattice-valued Wilson lines. Possible representatives of these Wilson lines can
1106: be composed out of the building blocks
1107: \begin{align}
1108: W_1& =~(0^6,1,1)\;,
1109: & W_2 &=~(0^5,1,1,0)\;,
1110: & W_1'& =~(1,1,0^6) \;,
1111: & W_2' & =~(0,1,1,0^5)\;, \nonumber \\
1112: S&=~(\tfrac{1}{2}^8)\;,
1113: & V & =~(0^7,2)\;,
1114: & V' & =~(0^6,2,0)\;,\label{eq:buildingblocks}
1115: & & \phantom{I^I}
1116: \end{align}
1117: and are listed in the last column of Tab.~\ref{tab:Z2xZ2}.
1118:
1119:
1120: Models leading to spectra coinciding with what we got in Tab.~\ref{tab:Z2xZ2}
1121: have already been discussed in the literature. They appeared first in
1122: Ref.~\cite{Donagi:2004ht} in the context of free fermionic string models related
1123: to the $\mathbbm{Z}_2\times\mathbbm{Z}_2$ orbifold with an additional freely
1124: acting shift. More recently, new $\mathbbm{Z}_2\times\mathbbm{Z}_2$ orbifold
1125: constructions have been found in studying orbifolds of non-factorizable
1126: six-tori~\cite{Faraggi:2006bs,Forste:2006wq}. We find that for each model M of
1127: Tab.~\ref{tab:Z2xZ2} there is a corresponding `non-factorizable' model M$'$
1128: with the following properties:
1129: \begin{enumerate}
1130: \item
1131: Each `non-empty' fixed point, i.e.\ each fixed point with local zero-modes, in the model M can be mapped
1132: to a fixed point with the same spectrum in model M$'$.
1133: \item
1134: The number of `non-empty' fixed points in M coincides with the total number of
1135: fixed points in M$'$.
1136: \end{enumerate}
1137:
1138:
1139: These relations are not limited to $\mathbbm{Z}_2\times\mathbbm{Z}_2$ orbifolds,
1140: rather we find an analogous connection also in other
1141: $\mathbbm{Z}_N\times\mathbbm{Z}_M$ cases ($\mathbbm{Z}_N\times\mathbbm{Z}_M$
1142: orbifolds based on non-factorizable compactification lattices have recently
1143: been discussed in~\cite{Takahashi:2007qc}). This result hints towards an
1144: intriguing impact of generalized discrete torsion on the interpretation of
1145: orbifold geometry. What the (zero-mode) spectra concerns, introducing
1146: generalized discrete torsion (or considering generalized brother models) is
1147: equivalent to changing the geometry of the underlying compact space,
1148: $\Gamma\to\Gamma'$. To establish complete equivalence between these models would
1149: require to prove that the couplings of the corresponding states are the same,
1150: which is beyond the scope of the present study. It is, however, tempting to
1151: speculate that non-resolvable singularities, as discussed above, do not `really'
1152: exist as one can always choose (for a given spectrum) the compactification
1153: lattice $\Gamma$ in such a way that there are no `empty' fixed points.
1154:
1155:
1156: \section{How to classify $\boldsymbol{\mathbbm{Z}_{N}\times\mathbbm{Z}_{M}}$\ orbifolds}
1157: \label{sec:ClassifZ3xZ3}
1158:
1159: Let us now turn to describing a method of classifying heterotic
1160: $\mathbbm{Z}_{N}\times\mathbbm{Z}_{M}$ orbifolds, taking into account
1161: generalized discrete torsion. To illustrate our methods, we focus on
1162: $\mathbbm{Z}_{3}\times\mathbbm{Z}_{3}$ orbifold compactifications of the
1163: $\E8\times\E8$ heterotic string. It is straightforward to generalize the
1164: discussion to other $\Z{N}\times\Z{M}$ orbifolds and to the \SO{32} case.
1165:
1166: %\subsection{Classification without Wilson lines}
1167:
1168: To classify an orbifold requires an efficient prescription of how to obtain all
1169: inequivalent models. The first step in a classification is to get all admissible
1170: choices for the shift vector $V_1$. For this purpose, we make use of Dynkin
1171: diagram techniques (see e.g.\ \cite{Katsuki:1990bf}). These techniques are
1172: advantageous since, when writing down $V_1$, one has the freedom of choosing the
1173: basis of the weight lattice $\Lambda$ in such a way that the shift has a very
1174: simple form. Clearly, this freedom is lost when one introduces the second shift
1175: (and Wilson lines). This complicates the construction of all inequivalent shifts
1176: $V_2$.
1177:
1178: To obtain all inequivalent $V_2$ we utilize a method introduced by
1179: Giedt~\cite{Giedt:2000bi} (see also~\cite{Nilles:2006np}), i.e.\ use an adequate
1180: minimal ansatz which avoids redundancies due to lattice translations and some
1181: Weyl reflections. This ansatz restricts the shifts to be only in a certain cell
1182: $\widetilde{\Lambda}_N$ of the lattice $\Lambda$ in such a way that any possible
1183: $\mathbbm{Z}_N$ shift can be written as an element of this cell plus a lattice
1184: vector. That is, an arbitrary $\Z{N}$ shift has a unique decomposition
1185: \begin{equation}\label{eq:shiftdecomposition}
1186: V~=~\widetilde{V}+\Delta V\;,\quad \text{where}\:
1187: \widetilde{V}\in\widetilde{\Lambda}_N\:\text{and}\:\Delta V\in\Lambda\;.
1188: \end{equation}
1189: Consider now a consistent gauge embedding $(V_1,V_2)$. According to
1190: equation~\eqref{eq:shiftdecomposition} the shifts can be decomposed into
1191: $(\widetilde{V}_1+\Delta V_1,\widetilde{V}_2+\Delta V_2)$ with
1192: $\widetilde{V}_1\in\widetilde{\Lambda}_{N}$,
1193: $\widetilde{V}_2\in\widetilde{\Lambda}_{M}$ and $\Delta V_i\in\Lambda$. It is
1194: not hard to see that the conditions \eqref{eq:newmodularinv} imply
1195: \begin{subequations}
1196: \label{eq:necessaryconditions}
1197: \begin{eqnarray}
1198: N\,\left(\widetilde{V}_1^2 - v_1^2\right)&~=~ 0\text{ mod }2 \;,\label{eq:necessarycondition1}\\
1199: M\,\left(\widetilde{V}_2^2 - v_2^2\right)&~=~0\text{ mod }2 \;,\label{eq:necessarycondition2}\\
1200: M\,\left(\widetilde{V}_1\cdot \widetilde{V}_2 - v_1\cdot v_2\right) &~=~0\text{ mod }1 \;.\label{eq:necessarycondition3}
1201: \end{eqnarray}
1202: \end{subequations}
1203: To scan over all possible shift embeddings is therefore reduced to the task of
1204: \begin{enumerate}
1205: \item specifying $\widetilde{V}_1$ (satisfying \eqref{eq:necessarycondition1})
1206: by Dynkin techniques,
1207: \item scanning $\widetilde{\Lambda}_{M}$ for $\widetilde{V}_2$ fulfilling equations
1208: \eqref{eq:necessarycondition2} and \eqref{eq:necessarycondition3}, and
1209: \item examining all possible $(V_1,V_2)$ related to
1210: $(\widetilde{V}_1,\widetilde{V}_2)$ by lattice translations and satisfying
1211: \eqref{eq:newmodularinv}.
1212: \end{enumerate}
1213: At first sight, the last step seems to require a scan over an infinite number of
1214: shifts. However, as we have seen in section~\ref{sec:DiscreteTorsion}, given one
1215: representative $(V_1,V_2)$ satisfying \eqref{eq:newmodularinv}, this scan can be
1216: replaced by switching on all inequivalent discrete torsion phases. Since there
1217: is only a finite number of torsion phase assignments, we have found a
1218: prescription to obtain all inequivalent models by scanning only over a finite set
1219: of inputs.
1220:
1221: As already stated, in a complete classification it is necessary to take into
1222: account generalized discrete torsion as well. Thus, to get all admissible
1223: shifts, one has to scan over all appropriate values for the parameters
1224: $b_\alpha,\,c_\alpha$ and $d_{\alpha\beta}$.
1225:
1226: All statements made for $V_2$ apply also to the Wilson lines. That is, in order
1227: to obtain all inequivalent Wilson line embeddings, one can also scan the finite
1228: cell $\widetilde{\Lambda}_{N_\alpha}$ (fulfilling consistency conditions
1229: analogous to \eqref{eq:necessaryconditions}), and then switch on generalized
1230: discrete torsion.
1231:
1232: So far, we have described how to obtain all inequivalent models. However, some
1233: of the inputs, specified by $(V_1,V_2)$, Wilson lines and generalized torsion
1234: phases, turn out to be equivalent. Apart from the ambiguities related to Weyl
1235: reflections (cf.\ \cite{Giedt:2000bi}), in the framework of $\Z{N}\times\Z{M}$
1236: orbifolds further complications arise. For instance, we note that models with
1237: shifts related by discrete rotations, i.e.\
1238: $(V_1,V_2)\rightarrow(a_1\,V_1+a_2\,V_2,b_1\,V_1+b_2\,V_2)$ with proper values of
1239: $a_i,b_i\in\mathbbm{Z}$, can be equivalent. For example, in $\mathbbm{Z}_3\times\mathbbm{Z}_3$ a model
1240: with shifts $(V_1,V_2)$ is equivalent to a model with shifts $(V_1, V_1+2\,V_2)$.
1241:
1242: In our classification below, we consider two models as inequivalent if and only
1243: if their massless spectra are different.\footnote{In practice, we compare the
1244: non-Abelian massless spectra and the number of singlets. This underestimates the
1245: true number of models somewhat.}
1246:
1247:
1248: \subsubsection*{Sample classification of $\boldsymbol{\mathbbm{Z}_{3}\times\mathbbm{Z}_{3}}$ without Wilson lines}
1249:
1250: As a concrete application, let us describe the classification of $\Z3\times\Z3$
1251: orbifolds without Wilson lines.\footnote{As we allow for generalized discrete
1252: torsion, some of the models can be interpreted as being endowed with
1253: lattice-valued Wilson lines, see section \ref{sec:GeneralizationDTAndBrothers}.}
1254: By using Dynkin diagram techniques, one finds that there are only five
1255: consistent shift vectors $V_1$, which can be written in the generic form
1256: \begin{equation}\label{eq:MinForm1}
1257: \widetilde{V}_1~=~\frac{1}{3}(0^{n_0},1^{n_1},2^\alpha)\,(0^{m_0},1^{m_1},2^\beta),
1258: \end{equation}
1259: where $\alpha$ and $\beta$ can be either 0 or 1, and $n_0,n_1,m_0,m_1\in\mathbbm{Z}$, such that
1260: $n_0+n_1+\alpha=m_0+m_1+\beta=8$.
1261:
1262: A general ansatz describing the second shift $V_2$ of order three is given by~\cite{Giedt:2000bi}:
1263: \begin{eqnarray}\label{eq:MinForm2}
1264: \widetilde{V}_2~=~ &\dfrac{1}{3}\left( \left(
1265: \begin{array}{c}
1266: 3\\
1267: \vdots \\
1268: -2
1269: \end{array}
1270: \right), \left(
1271: \begin{array}{c}
1272: 1\\
1273: 0
1274: \end{array}
1275: \right)^{n_0-1}, \left(
1276: \begin{array}{c}
1277: 1\\
1278: 0\\
1279: -1
1280: \end{array}
1281: \right)^{n_1+\alpha}\; \right) \nonumber \\
1282: &\quad
1283: \left( \left(
1284: \begin{array}{c}
1285: 3\\
1286: \vdots\\
1287: -2
1288: \end{array}
1289: \right), \left(
1290: \begin{array}{c}
1291: 1\\
1292: 0
1293: \end{array}
1294: \right)^{m_0-1}, \left(
1295: \begin{array}{c}
1296: 1\\
1297: 0\\
1298: -1
1299: \end{array}
1300: \right)^{m_1+\beta}\; \right),
1301: \end{eqnarray}
1302: subject to lattice conditions, equations~\eqref{eq:latticeconditions}, and to
1303: the necessary conditions~\eqref{eq:necessaryconditions}. Some of these models
1304: do not fulfill the consistency requirements~\eqref{eq:newmodularinv}. As
1305: explained above, in these cases we proceed by identifying lattice vectors
1306: $\Delta V_i$ with the property that $(\widetilde{V}_1+\Delta
1307: V_1,\widetilde{V}_2+\Delta V_2)$ satisfy \eqref{eq:newmodularinv}. The problem
1308: of finding those lattice vectors can be reduced to a set of linear
1309: Diophantine equations.
1310:
1311:
1312: To generate all shift embeddings, we compute the spectra of models with
1313: different values of $a$ in the discrete torsion phase. In
1314: $\mathbbm{Z}_3\times\mathbbm{Z}_3$, the parameter $a$ can have values
1315: $0,\,\frac{1}{3}$ or $\frac{2}{3}$. This gives a factor of three to the total
1316: number of models. However, it turns out that not all of them are inequivalent.
1317: Counting only inequivalent spectra, we find that there are
1318: 120 inequivalent shift embeddings.
1319:
1320: We use now the set of shift embeddings to generate all admissible models. As
1321: discussed in the examples of section~\ref{sec:GeneralizedDiscreteTorsion},
1322: excluding $a$, there are 9 independent generalized discrete torsion parameters,
1323: whose values can be again $0,\,\frac{1}{3}$ or $\frac{2}{3}$. Although the
1324: number of models is multiplied by a factor $3^9$, the number of all inequivalent
1325: models (spectra) is 1082. These models comprise the complete set of admissible
1326: models without Wilson lines, or, more precisely, the complete set of models
1327: which can be described by vanishing Wilson lines. The model definitions and the
1328: resulting spectra are given in \cite{WebTables:2007mt}.
1329:
1330: The above procedure can straightforwardly be carried over to the SO(32) case.
1331: The results turn out to be similar. Repeating the steps of our $\E8\times\E8$
1332: discussion, we find that there are 131 shift embeddings. The total number of
1333: inequivalent models is very similar to the $\E8\times\E8$ case.
1334:
1335: \clearpage
1336: \section{Discussion}
1337: \label{sec:Conclusions}
1338:
1339: Aiming at a systematic understanding of heterotic $\Z{N}\times\Z{M}$ orbifolds,
1340: we have investigated the possibilities arising in these constructions. We find
1341: that, unlike in the case of prime orbifolds, adding elements of the weight
1342: lattice to the shifts, $(V_1,V_2)\to(V_1+\Delta V_1,V_2+\Delta V_2)$, changes in
1343: general the spectrum. Interestingly, the same spectra are obtained by switching
1344: on discrete torsion. Stated differently, one can trade discrete torsion for a
1345: change of the gauge embedding by lattice vectors.
1346: %
1347: We have extended our analysis such as to include generalized discrete torsion.
1348: We find that a large class of generalized discrete torsion phases can be
1349: mimicked by lattice-valued Wilson lines. Interestingly, an analog of a
1350: generalized discrete torsion phase appears in certain \Z{N} orbifolds which
1351: admit at least two discrete Wilson lines of coinciding orders. Another
1352: remarkable result is that switching on certain types of generalized discrete
1353: torsion is equivalent to changing the 6D compactification lattice. This implies
1354: that the field-theoretic interpretation of the model input can be somewhat
1355: subtle. We provided various explicit examples, illustrating all main statements.
1356:
1357: Our results have important consequences. At the more practical side, we were
1358: able to formulate a straightforward method of classifying $\Z{N}\times\Z{M}$
1359: orbifolds. We have also seen that switching on generalized discrete torsion
1360: allows to change the local spectra, i.e.\ one obtains different twisted states,
1361: which correspond to brane fields in the field-theoretic description. We expect
1362: this to become important for orbifold model building, where one can use this
1363: observation, for instance, for reducing the number of generations without
1364: modifying the gauge group.
1365:
1366: At a more conceptual level, our findings imply that a given spectrum cannot be
1367: ascribed neither to properties of the 6D internal space alone, i.e.\ whether
1368: discrete torsion is switched on or not, nor to the gauge embedding, but only to
1369: both. This implies that the same models, leading to the same spectra, can be
1370: regarded as resulting from what one might consider as different geometries.
1371: Although we cannot claim to have identified the deeper reasons for these
1372: relations, we feel that our observations constitute some progress in the task of
1373: understanding stringy geometry.
1374:
1375: \paragraph{Acknowledgments.}
1376: We would like to thank S.~F\"{o}rste, A.~Micu and H.~P.~Nilles for valuable
1377: discussions, and T.~Kobayashi and K.-J.~Takahashi for correspondence.
1378:
1379: This work was partially supported by the cluster of excellence ``Origin and
1380: Structure of the Universe'', the EU 6th Framework Program MRTN-CT-2004-503369 ``Quest
1381: for Unification'', MRTN-CT-2004-005104 ``ForcesUniverse'', MRTN-CT-2006-035863
1382: ``UniverseNet'' and the Transregios 27 '`Neutrinos and Beyond'' and 33 ``The Dark
1383: Universe'' by Deutsche Forschungsgemeinschaft (DFG).
1384:
1385:
1386: \appendix
1387: \renewcommand{\theequation}{\Alph{section}.\arabic{equation}}
1388: \setcounter{section}{0}
1389: \setcounter{equation}{0}
1390: \renewcommand{\thetable}{\Alph{section}.\arabic{table}}
1391: \clearpage
1392:
1393:
1394: \section{Transformation phases}
1395: \label{app:ModularInvariance}
1396:
1397:
1398: The aim of this appendix is to clarify the transformation law of physical
1399: states. Let us start with the simplest example, a $\Z{N}$ orbifold without
1400: Wilson lines. Modular invariance requires \cite{Vafa:1986wx}
1401: \begin{equation}\label{eq:weakmodularinv}
1402: N\,(V^2-v^2)~=~0\mod 2\;.
1403: \end{equation}
1404: To see that there are some subtleties, consider a `constructing element'
1405: $g=(\theta^k,n_\alpha e_\alpha)$. Zero-modes in the $g$-twisted sector have to
1406: fulfill the masslessness condition \eqref{eq:MassEquation}. Focus, for
1407: simplicity, on non-oscillator states, for which the masslessness conditions read
1408: \begin{subequations}
1409: \begin{eqnarray}
1410: \frac{1}{2}p_\mathrm{sh}^2-1+\delta c & = & 0\;,\\
1411: \frac{1}{2}q_\mathrm{sh}^2-\frac{1}{2}+\delta c & = & 0\;,
1412: \end{eqnarray}
1413: \end{subequations}
1414: where $p_\mathrm{sh}=(p+V_g)$, $q_\mathrm{sh}=(q+v_g)$, $V_g=k\,
1415: V+n_\alpha\,A_\alpha$ and $v_g=k\,v$. Using the properties $p^2=\text{even}$ and
1416: $q^2=\text{odd}$ one derives
1417: \begin{subequations}\label{eq:p.V}
1418: \begin{eqnarray}
1419: p\cdot V_g & = & 1-\delta c-\frac{V_g^2}{2}+\text{integer}\;,\\
1420: q\cdot v_g & = & -\delta c-\frac{v_g^2}{2}+\text{integer}\;.
1421: \end{eqnarray}
1422: \end{subequations}
1423: Now let us study the transformation of a solution of the mass equation
1424: $|\psi\rangle$. The shifted momenta $p_\mathrm{sh}$ and $q_\mathrm{sh}$ represent the gauge
1425: and Lorentz (or $\SO8$) quantum numbers. This fixes the transformation phase of
1426: the $g$-twisted string $|\psi\rangle$ to be
1427: \begin{equation}\label{eq:ProjectionCondition2}
1428: 2\pi\,[p_\mathrm{sh}\cdot V_g-q_\mathrm{sh}\cdot v_g]~=~
1429: 2\pi\,\left[\frac{1}{2}\left(V_g^2-v_g^2\right)\mod 1\right]
1430: \end{equation}
1431: under the action of $g$. From \eqref{eq:weakmodularinv} we infer that this phase
1432: does not vanish in general, rather it is of the form $2\pi\,m/N$ with
1433: $m\in\mathbbm{Z}$. This raises the question of how a state associated to $g$ can
1434: be invariant under the action of $g$. In the literature, the transformation
1435: behavior of states associated to twisted strings has often been `repaired',
1436: i.e.\ an additional transformation phase has been introduced by hand. In what
1437: follows, we present a geometric explanation of how such additional phases arise.
1438:
1439:
1440: Consider a string twisted by the space group element $g$ in the `upstairs'
1441: picture, i.e.\ on the torus. Twisted strings end at the borders of the
1442: fundamental domain of the orbifold. The fundamental domain of the orbifold
1443: together with its non-trivial images under $g$, $g^2$ etc.\ comprise the
1444: fundamental domain of the torus. Therefore, on the torus a twisted string
1445: appears in $n$ copies where $n$ is the order of $g$, i.e.\ the minimal positive
1446: integer with $g^n=\mathbbm{1}$. A 2D illustration is shown in
1447: figure~\ref{fig:Transformation}.
1448: %
1449: \begin{figure}[!h]
1450: \centerline{\CenterObject{\includegraphics{SU3TorusLoc2.eps}}}
1451: \caption{$\mathbbm{T}^2_{\SU3}/\Z3$ orbifold. The fundamental domain of the
1452: torus can be taken to be the darker area. Transformation of
1453: $g=(\theta,e_1+e_2)$-twisted strings under (the constructing element) $g$: the
1454: (red) right string is mapped to the (green) upper, and the (green) upper to the
1455: (blue) left, and the (blue) left back to the (red) upper one.}
1456: \label{fig:Transformation}
1457: \end{figure}
1458: %
1459: That is, a $g$-twisted state appears as a linear combination of the state
1460: $|\psi\rangle$ corresponding to the $g$-twisted string in the fundamental domain
1461: and its images under $g^m$. This linear combination can involve phase factors
1462: with the constraint being that $g^n$ acts as identity. There are $n$
1463: different linear combinations labeled by $m\in\mathbbm{Z}$,
1464: \begin{equation}\label{eq:LinComb}
1465: |\psi^{(m)}\rangle~=~
1466: \frac{1}{\sqrt{n}}
1467: \left(|\psi\rangle+e^{-2\pi\I\, m/n}\,|g\,\psi\rangle
1468: + \dots + e^{-2\pi\I\, m\,(n-1)/n}\,|g^{n-1}\,\psi\rangle\right)\;,
1469: \end{equation}
1470: on the torus. In figure~\ref{fig:Transformation} this would correspond to a
1471: superposition of the red, green and blue string, weighted by phase factors. It
1472: is clear that, under the action of $g$, such a linear combination picks up a
1473: phase $e^{2\pi\I\, m/n}$. In other words, the transformation phase
1474: \eqref{eq:ProjectionCondition2} is to be amended by $2\pi\, m/n$.
1475:
1476: It is straightforward to apply these observations to the above-mentioned problem
1477: of $g$-invariance of states associated to the constructing element $g$. Clearly,
1478: only the linear combination \eqref{eq:LinComb} with $m=-\frac{n}{2}(V_g^2-v_g^2)$
1479: is $g$-invariant. So we conclude that for every solution of the mass equation
1480: one finds precisely one $g$-invariant state. In addition to the phase arising
1481: from the gauge and Lorentz quantum numbers \eqref{eq:ProjectionCondition2}, this
1482: state picks up a compensating phase
1483: \begin{equation}\label{eq:PhiVac}
1484: \Phi_\mathrm{vac}~=~ \exp\left\{2\pi\I\,\left[-\frac{1}{2}(V_g\cdot V_g-v_g\cdot v_g)\right]\right\}
1485: \end{equation}
1486: under the action of $g$.
1487:
1488: Before turning to the discussion of the transformation of such states under
1489: commuting elements $h$, let us briefly comment on a technical simplification
1490: that is possible in many models.
1491: It is also clear that, in \Z{N} orbifolds without Wilson lines, one can
1492: `transform' a model M, with a `weak' shift $V$, to the model M$'$, with shift
1493: $V'= V+\Delta V$ where $\Delta V\in\Lambda$. The solutions of the mass equation
1494: coincide in the models M and M$'$. It is, up to some exceptional cases which we
1495: will discuss below, always possible to find a $\Delta V$ with
1496: $\frac{1}{2}[(V+\Delta V)^2-v^2]\in\mathbbm{Z}$ so that
1497: \eqref{eq:ProjectionCondition2} is automatically fulfilled for any
1498: $|\psi\rangle$ (and therefore also for trivial linear combinations).
1499: For a large class of constructions one can therefore adopt the following logic.
1500: Models with an input fulfilling only the (`weak') modular invariance constraints
1501: \eqref{eq:weakmodularinv} might not be considered as they have an alternative
1502: description in terms of a model with input fulfilling the stronger constraints
1503: \begin{equation}\label{eq:strongmodularinv}
1504: V^2-v^2~=~0\mod 2\;.
1505: \end{equation}
1506: That is, in order to avoid double-counting, one can restrict to the stronger
1507: constraints \eqref{eq:strongmodularinv} in many cases. Similar statements
1508: apply to the case with non-trivial Wilson lines (an example has been given in
1509: \cite{Buchmuller:2006ik}).
1510: %
1511: However, there is a caveat, namely the above-mentioned exceptional cases in
1512: which a `weak' modular invariant input cannot be transformed to the `strong'
1513: form. The simplest example for such a case is a \Z3 orbifold with $V=0$. Further
1514: examples arise in non-prime \Z{N\cdot M} orbifolds ($N,M>1$) where $N\cdot
1515: V\in\Lambda$.
1516:
1517: Let us now return to the question of how a ($g$-invariant) state associated to
1518: the constructing element $g$ transforms under the action of a commuting element
1519: $h$. In the following, we denote the corresponding transformation phase of
1520: equation~\eqref{eq:transformationphase} by $\Phi(g,h)$. Clearly, this phase has to comply with the
1521: space group multiplication law, thus
1522: \begin{equation}
1523: \Phi_\mathrm{vac}(g,\,g^p\,h^q)
1524: ~=~
1525: [\Phi_\mathrm{vac}(g,\,g)]^p\,[\Phi_\mathrm{vac}(g,\,h)]^q\;.
1526: \end{equation}
1527: Since $\Phi_\mathrm{vac}(g,\,g)$ is already fixed by \eqref{eq:PhiVac}, this leads
1528: to the conclusion
1529: \begin{equation}
1530: \Phi_\mathrm{vac}(g,\,h)
1531: ~=~
1532: \exp\left\{2\pi\I\,\left[-\frac{1}{2}\left(V_g\cdot V_h-v_g\cdot v_h\right)\right]\right\}
1533: \end{equation}
1534: for any commuting $h\in S$. Therefore, the full transformation phase of the physical states has to be defined as in
1535: equation~\eqref{eq:transformationphase}. But there are still some constraints which have to be fulfilled
1536: for the sake of consistency. To illustrate them, let us consider $g,h \in S$ with $g^n = \mathbbm{1} =
1537: h^s$. From the definition of the full transformation phase $\Phi$, it is clear that one has to demand
1538: \begin{equation} \label{eq:TransfModified}
1539: \Phi(g,h) ~\stackrel{!}{=}~ \Phi(g^{n+1},h)
1540: ~=~\Phi(g,h)\,\Phi_\mathrm{vac}(g,h)^{-n} \,,
1541: \end{equation}
1542: where the second equality is obtained by replacing, according to the
1543: usual embedding, $V_g\to (n+1)\,V_g$ and $v_g\to(n+1)\,v_g$ in equation~\eqref{eq:transformationphase}.
1544: Thus, $\Phi_\mathrm{vac}(g,h)^{n}\stackrel{!}{=}1$. An analogous
1545: reasoning starting with $\Phi(g,h^{s+1})$ leads to
1546: $\Phi_\mathrm{vac}(g,h)^{s}~\stackrel{!}{=}~1$ and thus finally to
1547: \begin{equation}\label{eq:consistency}
1548: \Phi_\mathrm{vac}(g,h)^{\mathrm{gcd}(n,s)}~\stackrel{!}{=}~1\;.
1549: \end{equation}
1550: Formulating equation~\eqref{eq:consistency} in terms of the gauge embedding
1551: shifts leads to the consistency conditions~\eqref{eq:newmodularinv} on shifts
1552: and Wilson lines.\footnote{In the case of two different \Z2 Wilson lines we find
1553: that \eqref{eq:fsmi5} can be relaxed, i.e.\ $\text{gcd}(N_{\alpha},N_{\beta})$
1554: can be replaced by $N_{\alpha}\,N_{\beta}=4$, provided there exists no $g \in
1555: P$ with the property $g\,e_\alpha~\neq~e_\alpha$ but $g\,e_\beta~=~e_\beta$.
1556: Imposing the weaker condition leads, as we find, to anomaly-free spectra.}
1557:
1558:
1559: \providecommand{\bysame}{\leavevmode\hbox to3em{\hrulefill}\thinspace}
1560: \begin{thebibliography}{10}
1561:
1562: \bibitem{Dixon:1985jw}
1563: L.~J. Dixon, J.~A. Harvey, C.~Vafa, and E.~Witten, Nucl. Phys. \textbf{B261}
1564: (1985), 678--686.
1565: %%CITATION = NUPHA,B261,678;%%
1566:
1567: \bibitem{Dixon:1986jc}
1568: L.~J. Dixon, J.~A. Harvey, C.~Vafa, and E.~Witten, Nucl. Phys. \textbf{B274}
1569: (1986), 285--314.
1570: %%CITATION = NUPHA,B274,285;%%
1571:
1572: \bibitem{Vafa:1986wx}
1573: C.~Vafa, Nucl. Phys. \textbf{B273} (1986), 592.
1574: %%CITATION = NUPHA,B273,592;%%
1575:
1576: \bibitem{Font:1988mk}
1577: A.~Font, L.~E. Ib{\`a}{\~n}ez, and F.~Quevedo, Phys. Lett. \textbf{B217}
1578: (1989), 272.
1579: %%CITATION = PHLTA,B217,272;%%
1580:
1581: \bibitem{Vafa:1994rv}
1582: C.~Vafa and E.~Witten, J. Geom. Phys. \textbf{15} (1995), 189--214,
1583: [hep-th/9409188].
1584: %%CITATION = HEP-TH 9409188;%%
1585:
1586: \bibitem{Sharpe:2000ki}
1587: E.~R. Sharpe, Phys. Rev. \textbf{D68} (2003), 126003, [hep-th/0008154].
1588: %%CITATION = HEP-TH 0008154;%%
1589:
1590: \bibitem{Gaberdiel:2004vx}
1591: M.~R. Gaberdiel and P.~Kaste, JHEP \textbf{08} (2004), 001, [hep-th/0401125].
1592: %%CITATION = HEP-TH 0401125;%%
1593:
1594: \bibitem{Ibanez:1987pj}
1595: L.~E. Ib{\'a}{\~n}ez, J.~Mas, H.-P. Nilles, and F.~Quevedo, Nucl. Phys.
1596: \textbf{B301} (1988), 157.
1597: %%CITATION = NUPHA,B301,157;%%
1598:
1599: \bibitem{Font:1989aj}
1600: A.~Font, L.~E. Ib{\'a}{\~n}ez, F.~Quevedo, and A.~Sierra, Nucl. Phys.
1601: \textbf{B331} (1990), 421--474.
1602: %%CITATION = NUPHA,B331,421;%%
1603:
1604: \bibitem{Forste:2004ie}
1605: S.~F{\"o}rste, H.~P. Nilles, P.~K.~S. Vaudrevange, and A.~Wingerter, Phys. Rev.
1606: \textbf{D70} (2004), 106008, [hep-th/0406208].
1607: %%CITATION = HEP-TH 0406208;%%
1608:
1609: \bibitem{Kobayashi:2004ya}
1610: T.~Kobayashi, S.~Raby, and R.-J. Zhang, Nucl. Phys. \textbf{B704} (2005),
1611: 3--55, [hep-ph/0409098].
1612: %%CITATION = HEP-PH 0409098;%%
1613:
1614: \bibitem{Buchmuller:2004hv}
1615: W.~Buchm{\"u}ller, K.~Hamaguchi, O.~Lebedev and M.~Ratz,
1616: Nucl.\ Phys.\ \textbf{B712} (2005), 139,
1617: [hep-ph/0412318].
1618: %%CITATION = NUPHA,B712,139;%%
1619:
1620: \bibitem{Buchmuller:2006ik}
1621: W.~Buchm{\"u}ller, K.~Hamaguchi, O.~Lebedev, and M.~Ratz, (2006),
1622: hep-th/0606187.
1623: %%CITATION = HEP-TH 0606187;%%
1624:
1625: \bibitem{Orbifolder}
1626: S.~Ramos-S\'anchez, P.~K.~S.~Vaudrevange, and A.~Wingerter,
1627: \emph{C++ Orbifolder}
1628:
1629: \bibitem{Ibanez:1986tp}
1630: L.~E. Ib{\'a}{\~n}ez, H.~P. Nilles, and F.~Quevedo, Phys. Lett. \textbf{B187}
1631: (1987), 25--32.
1632: %%CITATION = PHLTA,B187,25;%%
1633:
1634: \bibitem{Katsuki:1990bf}
1635: Y.~Katsuki, Y.~Kawamura, T.~Kobayashi, N.~Ohtsubo, Y.~Ono, and K.~Tanioka,
1636: Nucl. Phys. \textbf{B341} (1990), 611--640.
1637: %%CITATION = NUPHA,B341,611;%%
1638:
1639: \bibitem{Walton:1988bu}
1640: M.~A. Walton, Phys. Rev. \textbf{D37} (1988), 377.
1641: %%CITATION = PHRVA,D37,377;%%
1642:
1643: \bibitem{Aspinwall:1994ev}
1644: P.~S. Aspinwall, (1994), hep-th/9403123.
1645: %%CITATION = HEP-TH 9403123;%%
1646:
1647: \bibitem{Lust:2006zh}
1648: D.~L{\"u}st, S.~Reffert, E.~Scheidegger, and S.~Stieberger, (2006),
1649: hep-th/0609014.
1650: %%CITATION = HEP-TH 0609014;%%
1651:
1652: \bibitem{Honecker:2006qz}
1653: G.~Honecker and M.~Trapletti, JHEP \textbf{01} (2007), 051, [hep-th/0612030].
1654: %%CITATION = HEP-TH 0612030;%%
1655:
1656: \bibitem{Nibbelink:2007rd}
1657: S.~G. Nibbelink, M.~Trapletti, and M.~Walter, (2007), hep-th/0701227.
1658: %%CITATION = HEP-TH 0701227;%%
1659:
1660: \bibitem{Donagi:2004ht}
1661: R.~Donagi and A.~E. Faraggi, Nucl. Phys. \textbf{B694} (2004), 187--205,
1662: [hep-th/0403272].
1663: %%CITATION = HEP-TH 0403272;%%
1664:
1665: \bibitem{Faraggi:2006bs}
1666: A.~E. Faraggi, S.~F{\"o}rste, and C.~Timirgaziu, JHEP \textbf{08} (2006), 057,
1667: [hep-th/0605117].
1668: %%CITATION = HEP-TH 0605117;%%
1669:
1670: \bibitem{Forste:2006wq}
1671: S.~F{\"o}rste, T.~Kobayashi, H.~Ohki, and K.-j. Takahashi, (2006),
1672: hep-th/0612044.
1673: %%CITATION = HEP-TH 0612044;%%
1674:
1675: \bibitem{Takahashi:2007qc}
1676: K.-j. Takahashi, (2007), hep-th/0702025.
1677: %%CITATION = HEP-TH 0702025;%%
1678:
1679: \bibitem{Giedt:2000bi}
1680: J.~Giedt, Ann. Phys. \textbf{289} (2001), 251, [hep-th/0009104].
1681: %%CITATION = HEP-TH 0009104;%%
1682:
1683: \bibitem{Nilles:2006np}
1684: H.~P. Nilles, S.~Ramos-S{\'a}nchez, P.~K.~S. Vaudrevange, and A.~Wingerter,
1685: JHEP \textbf{04} (2006), 050, [hep-th/0603086].
1686: %%CITATION = HEP-TH 0603086;%%
1687:
1688: \bibitem{WebTables:2007mt}
1689: F.~Pl{\"o}ger, S.~Ramos-S{\'a}nchez, M.~Ratz, and P.~K.~S. Vaudrevange,
1690: \emph{$\mathbbm{Z}_3\times\mathbbm{Z}_3$ orbifold tables}, 2007,
1691: {\texttt{http://www.th.physik.uni-bonn.de/nilles/Z3xZ3orbifold/}}.
1692:
1693: \end{thebibliography}
1694: \end{document}
1695: \endinput
1696: