hep-th0507016/kdkn.tex
1: \documentclass{article}
2: \usepackage{graphicx}
3: \usepackage{latexsym}
4: \usepackage{amssymb,amsfonts,amsmath}
5: %\usepackage{epsfig,epic}
6: \usepackage{epsf}
7: \usepackage[dvips]{color}
8: \title{The decay of unstable k-strings in SU(N) gauge theories at zero and 
9: finite temperature}
10: \author{Ferdinando Gliozzi}
11: %\date{October, 2004}
12: \newcommand{\eq}{\begin{equation}}
13: \newcommand{\en}{\end{equation}}
14: \newcommand{\ear}{\begin{eqnarray}}
15: \newcommand{\rae}{\end{eqnarray}}
16: \newcommand{\Z}{\mathbb{Z}}
17: \newcommand{\C}{\mathbb{C}}
18: \newcommand{\uu}{\mathbb{I}}
19: \newcommand{\R}{{\cal R}}
20: \newcommand{\s}{{\cal S}}
21: \newcommand{\bra}{\langle}
22: \newcommand{\ket}{\rangle}
23: \newcommand{\um}{\frac12}
24: \newcommand{\tr}{{\rm tr}\,}
25: \newcommand{\Tr}{{\rm tr}^{~}}
26: \definecolor{M_Beige}         {rgb}{0.96 , 0.96 , 0.86}
27: \newcommand{\CBei}[1]{\textcolor{M_Beige}{#1}}
28: 
29: \definecolor{M_Brown}         {rgb}{0.65 , 0.16 , 0.16}
30: \newcommand{\CBro}[1]{\textcolor{M_Brown}{#1}}
31: 
32: \definecolor{M_Gold}          {rgb}{1.00 , 0.84 , 0.00} 
33: \newcommand{\CGol}[1]{\textcolor{M_Gold}{#1}}
34: 
35: \definecolor{M_LemonChiffon}  {rgb}{1.00 , 0.98 , 0.80}
36: \newcommand{\CLem}[1]{\textcolor{M_LemonChiffon}{#1}}
37: \definecolor{M_Orange}        {rgb}{1.00 , 0.60 , 0.00}
38: \newcommand{\COra}[1]{\textcolor{M_Orange}{#1}}
39: 
40: \definecolor{M_Pink}          {rgb}{1.00 , 0.75 , 0.80}
41: \newcommand{\CPin}[1]{\textcolor{M_Pink}{#1}}
42: 
43: \definecolor{M_Violet}        {rgb}{0.93 , 0.51 , 0.93}
44: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
45: \newcommand{\Cbla}[1]{\textcolor{black}{#1}}
46: \newcommand{\Cblu}[1]{\textcolor{blue}{#1}}
47: \newcommand{\Ccya}[1]{\textcolor{cyan}{#1}}
48: \newcommand{\Cgre}[1]{\textcolor{green}{#1}}
49: \newcommand{\Cred}[1]{\textcolor{red}{#1}}
50: \newcommand{\Cyel}[1]{\textcolor{yellow}{#1}}
51: \newcommand{\Cwhi}[1]{\textcolor{white}{#1}}
52: \newcommand{\CVio}[1]{\textcolor{M_Violet}{#1}}
53: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
54:  \newcommand{\resection}[1]{\setcounter{equation}{0}\section{#1}}
55: \newcommand{\appsection}[1]{\addtocounter{section}{1}
56: \setcounter{equation}{0} \section*{Appendix \Alph{section}~~#1}}
57: \renewcommand{\theequation}{\thesection.\arabic{equation}}
58: 
59: \begin{document}
60: %\begin{flushright}
61: %IFUM-793-FT
62: %\end{flushright}
63: \maketitle
64: \noindent
65: Dipartimento di Fisica Teorica, Universit\`a di Torino and\\ INFN,
66: sezione di Torino, via P. Giuria, 1, I-10125 Torino, Italy.\\
67: \vskip0.5cm\noindent
68: e-mail: gliozzi@to.infn.it\\
69: %\begin{minipage}[0cm]{\textwidth}
70: %\vspace{-16cm}
71: %\hfill IFUM-793-FT
72: %\end{minipage}
73: \begin{abstract}
74: Sources in higher representations of $SU(N)$ gauge theory at $T=0$ 
75: couple with apparently stable strings with tensions depending on 
76: the specific representation rather than on its $N-$ality. Similarly
77: at the deconfining temperature these sources carry their own 
78: representation-dependent critical exponents. It is pointed out that in 
79: some instances one can evaluate exactly these exponents by fully 
80: exploiting the correspondence between the 2+1 dimensional critical 
81: gauge theory and the $2d$ conformal field theory in the same universality 
82: class. The emerging functional form of the Polyakov-line correlators 
83: suggests a similar form 
84: for Wilson loops in higher representations which helps in understanding 
85: the behaviour of unstable strings at $T=0$. A generalised Wilson loop in 
86: which along part of its trajectory a source is converted in a gauge 
87: invariant way into higher representations with  same $N-$ality could 
88: be used as a tool to estimate the decay scale of the unstable strings.   
89: 
90: \end{abstract}
91: 
92: \section{Introduction}
93: 
94: One of the most fascinating aspects of numerical experiments in lattice 
95: gauge theory is the possibility to make many controlled changes to 
96: explore the response of the strong interaction dynamics. In particular, we 
97: can vary the quark masses, the number $N$ of colours and even remove
98: the sea quarks. In this way one is led to study pure $SU(N)$ gauge theory
99: to probe the main properties of the confining vacuum.   
100: 
101: In this context, the linear rising of the static potential between a pair 
102: of quarks in the fundamental representation is well described  
103: by a thin flux tube, or string, joining the two quarks.
104: Excited colour sources, {\it i.e.} sources in a higher representation $\R$ of 
105: the gauge group, behave at intermediate distances in a similar way, 
106: giving rise to the formation of a confining string with a string tension 
107: $\sigma_\R$. However most strings of this kind are expected to be unstable: the
108: long distance properties of the string attached  to a source in a 
109: representation $\R$ built up of $j$ copies of the fundamental representation 
110: should depend only on its $N-$ality $k^{~}_\R\equiv j~({\rm mod} N)$ because 
111: all representations with same $k$ can be converted into each other by the 
112: emission of a proper number of soft gluons. As a consequence, the heavier 
113: $\R-$strings are expected to decay into the string with smallest string 
114: tension within the same $N$-ality class, called $k-$string. 
115: For a recent discussion on this subject see \cite{as}.
116: 
117: General heuristic 
118: arguments suggest that stable $k-$strings belong to the anti-symmetric 
119: representation with $k$ quarks, as it has been supported by exact results 
120: in various approaches to ${\cal N}=1$ supersymmetric  $SU(N)$ gauge theories 
121: \cite{ds,hsz,hk} as well as by  lattice calculations in some $SU(N)$ pure 
122: gauge models \cite{dprv,lt,ltw,dprv2}.  
123: 
124: In this paper we address the question of stability of strings 
125: attached to sources in different representations with the same $N-$ality.
126: Although most numerical experiments based on large Wilson loops 
127: \cite{cj,pt,sd,ba,cp} seem to defy the above theoretical arguments, 
128: yielding  apparently stable string tensions which depend on $\R$ rather 
129: than on its $N-$ality, this kind of behaviour has been fully understood, 
130: at least in the case of the adjoint string. Since the $N-$ality of the adjoint 
131: representation is zero, the static potential of two adjoint sources is 
132: expected to level off at large 
133: separations. Correspondingly, the associated string should decay into 
134: a pair of bound sates of a static adjoint colour source and a gluon field, 
135: sometimes called glue-lump in the literature. Knowing the mass of the glue-lumps
136: one can even evaluate the scale $r_{adj}$ at which the adjoint string 
137: breaks \cite{pt}. 
138: 
139: The lack of any sign of adjoint string breaking 
140: in the above-mentioned studies, while measuring the static potential from  
141: Wilson loops only, indicates that such an  operator has  a  
142: poor overlap with the true ground state. This fact has been directly 
143: demonstrated  in $2+1$ $SU(2)$ gauge theory  \cite{krde} where,
144: using a variance reduction algorithm allowing to detect signals down to 
145: $10^{-40}$, it has been clearly observed a rectangular Wilson loop 
146:  $W(r>r_{adj},t)$
147:  changing sharply its slope as a function of $t$ from that associated to 
148: the unbroken string (area-law decay) to that of the broken-string state 
149: (perimeter-law decay) at a distance much longer than the adjoint string 
150: breaking scale $r_{adj}$. 
151: 
152: Alternatively, one can enlarge the basis of the operators used to extract 
153: the adjoint potential 
154: in order to find a better overlap to the true ground state, following a 
155: multichannel method originally advocated in \cite{cm}. 
156: Indeed adjoint sources, contrarily to what happens in the case of fundamental 
157: representation, can form gauge invariant \Cblu{open} Wilson lines, like those 
158: depicted in Fig.\ref{Figure:1} $a$ and $b$, having a good overlap with 
159: the two glue-lump state. In this way a rather abrupt crossover between 
160: string-like and broken string states has been clearly seen at the expected 
161: distance $r_{adj}$ both in $2+1$ \cite{st} and $3+1$ \cite{fp} $SU(2)$ 
162: gauge models.
163: The same multichannel method allowed to observe breaking of
164: fundamental string in various gauge theories coupled to dynamical matter
165: fields in 2+1 \cite{pw} and 3+1 dimensions \cite{ks}, including QCD with two 
166: flavours \cite{bdl}. Also in these cases ordinary Wilson loops do not show 
167: clear signs of string breaking even at sizes much larger than the expected
168: breaking scale, except in a particularly simple case, the 2+1 
169: $\Z_2$ gauge-Higgs model \cite{Gliozzi:2004cs}, where, using  a 
170: variance reduction algorithm like in \cite{krde}, fundamental string 
171: breaking has been convincingly demonstrated.     
172: \par
173: One of the results of the present work is the generalisation of the multichannel method to the decay of unstable strings of non-vanishing $N-$ality.
174: To this aim we define in Sect.\ref{mixed} a new kind of gauge-invariant 
175: operators, the \Cblu{mixed Wilson loops}, (see Fig.\ref{Figure:1} $c$ and $d$)
176: where  along one or more segments of the closed path $\gamma$ 
177: (the dashed lines in Fig.\ref{Figure:1}) the static source carries the 
178: quantum numbers of an excited representation $\R$ of $N-$ality $k$, 
179: while in the remaining path 
180: (solid lines) the source lies in the stable, fully anti-symmetric 
181: representation. These operators have a good overlap with the stable 
182: $k-$string state and constitute the most obvious generalisation of the 
183: glue-lump operators drawn in Fig.\ref{Figure:1} $a$ and $b$. 
184: \par The above construction follows a general discussion  of the expected 
185: overlap properties of ordinary Wilson loops in excited representation. 
186: This leads to a rough estimate of the decay scale of the unstable 
187: strings and offers a simple explanation of the apparent stability of 
188: these strings at $T=0$.
189: \vskip .3 cm   
190: \par
191: The $T>0$ situation is less problematic. According to  a simple diagrammatic 
192: argument \cite{gp}
193: the correlator of  Polyakov lines is expected to have a good 
194: overlap with the true ground state. These correlators  
195:  in non-fundamental representations have been studied in 
196: various instances \cite{cb,ot,Muller:1991xj,DelDebbio:2004zc}
197: and it has easily observed the screening of the adjoint representation, 
198: and the decay of excited representations of non vanishing triality in $SU(3)$
199: and even the decay of the symmetric 2-index representation into the 
200: anti-symmetric one in $SU(4)$ \cite{DelDebbio:2003tk}. A difficulty 
201: of the $T>0$ approach is to extract reliable estimates of the breaking 
202: or decay scales of unstable strings at $T=0$.
203: \begin{figure}
204: \centering
205: \includegraphics[width=0.9\textwidth]{newop.eps} 
206: %\vspace{5cm}  % amount of vertical space needed
207: \caption{Glue-lump operators and mixed Wilson lines. The operators depicted in 
208: $a$ and $b$ describe adjoint Wilson lines decaying in the vacuum. They 
209: are used to extract the adjoint static potential in the multichannel method. 
210: The operators $c$ and $d$, described in the text, generalise the glue-lump operators to the case of representations of non zero $N-$ality .} 
211: \label{Figure:1}
212: \end{figure}
213: \vskip .3 cm
214: \par
215: A much more challenging problem emerges at the deconfining point: in gauge 
216: theories with continuous  phase transition one 
217: obvious question concerns the critical behaviour of the Polyakov lines 
218: in arbitrary representations. Over the years,many studies have been 
219: dedicated to this subject  \cite{gw,dam,jk,ctd,dh,pi}. 
220: \par The surprising result is 
221: that sources of higher representations, according to various numerical 
222: experiments \cite{dam,ctd,dh}, correspond to \Cblu{different} magnetisation 
223: exponents, one exponent for each representation. Universality arguments would,
224: for such continuous phase transitions, place the finite-temperature $SU(N)$ 
225: gauge theory in the universality class of $\Z_N$ invariant spin model in one 
226: dimension less \cite{sy}. 
227: The spin operator is mapped into the Polyakov line in the fundamental
228: representation. What about the Polyakov lines in higher representations of 
229: $SU(N)$? There appears to be no room for independent exponents for these 
230: higher representations from the point of view of the abelian spin system, 
231: since there is simply no obvious analogue of the non-abelian process of 
232: soft gluons emission, creating higher representations in the same $N-$ality 
233: class. 
234: 
235: Actually a mean field approximation of the the effective $SU(2)$ 
236: Polyakov-line action at criticality in the $d\to \infty$ limit shows that the
237: leading amplitudes of higher representations vanish at strong coupling, 
238: and the sub-leading exponents become dominant, thus each higher representation 
239: source carries its own critical exponent\cite{dh}.      
240: 
241: Here we reach a similar conclusion for $SU(3)$ and $SU(4)$  critical 
242: theories in $2+1$ dimensions starting from a different point of view:
243: we find a map between the operator product expansion (OPE) of the Polyakov 
244: operators in the gauge theory and the corresponding spin operators in the 
245: two-dimensional conformal field theory which describes the associated 
246: spin system at criticality. As a result, we are able to conjecture, 
247: for instance, an exact value for the $\eta$ exponent (or anomalous dimension) 
248: of the sources in the \Cblu{\sl symmetric}  representation made with 
249: two quarks (Sect.\ref{polyakov}). The resulting functional form of the 
250: Polyakov-Polyakov correlator at criticality of this excited representation 
251: is the starting point of an Ansatz for the vacuum expectation value 
252: of the  Wilson loop associated to sources in higher representations at $T=0$
253: (Sect.\ref{decay}). Section \ref{mixed} is dedicated to 
254: the construction of mixed Wilson loops and finally in Sect.
255: \ref{conclusion} we draw some conclusions.       
256: \section{Polyakov loops at criticality}
257: \label{polyakov}
258: Consider a $d+1$ dimensional pure gauge theory undergoing a continuous 
259: deconfinement transition at the critical temperature $T_c$. The effective model describing the behaviour of Polyakov lines at finite $T$ will be  a
260:  $d-$dimensional spin model with a global symmetry group coinciding with 
261: the center of the gauge group. Svetitsky and Yaffe \cite{sy} (SY) were 
262: able to show that this effective model has only short-range 
263: interactions, then it follows from universality arguments that the 
264: spin model belongs in the same universality class of the original 
265: gauge model and the high temperature phase of the gauge theory is like the 
266: low temperature phase of the spin system.
267: 
268: It is clear that this SY conjecture, which has passed several numerical 
269: tests, becomes very predictive for $d=2$, where, using the methods of 
270: conformal field theory (CFT), the critical behaviour can be determined 
271: exactly. For example, the critical properties of $2+1$ dimensional $SU(3)$ 
272: gauge theory at deconfinement coincide with those of the 3-state Potts model,
273: as it has been checked in numerical simulations \cite{ctd}.
274: 
275: What is needed to fully exploit the predictive power of the SY conjecture 
276: is a mapping relating the physical observables of the gauge theory to the 
277: operators of the dimensionally reduced model, as first advocated in
278: \cite{Gliozzi:1997yc}.
279: 
280: The correspondence between the  Polyakov line 
281: in the fundamental representation
282: $f$ and the order parameter $\sigma$ of the spin model is the first entry  
283: in this mapping:
284: \eq
285: \Tr_f(U_{\vec{x}}) ~\sim~ \sigma(\vec{x})~, 
286: \en
287: $U_{\vec{x}}$ is the gauge group element  associated to the closed path 
288: winding once around the periodic imaginary time direction intersecting 
289: the spatial plane at the point $\vec{x}$. The above equivalence is only 
290: valid in a weak sense, that is, when the left-hand-side of the equation  
291: is inserted in a correlation function of the gauge theory and the right-hand 
292: side in the corresponding correlator of the spin model.
293:  \par It is now natural to ask what operators in the CFT correspond to
294: Polyakov lines in  higher representations. \Cblu{On the gauge side} 
295: these can be obtained by a proper combination of products of Polyakov 
296: lines in the fundamental representation, using repeatedly  the  general 
297: property
298: \eq
299: \Tr_\R(U)\,\Tr_\s(U)=\sum_{\R'\in\R\otimes\s}\Tr_{\R'}(U)~,
300: \label{decompo}
301: \en  
302: valid for any pair of irreducible representations of an arbitrary group.
303: In order to control the  singularities in the correlator functions 
304: due to evaluation of local operators at the same point 
305: we may resort to the operator product expansion (OPE). The OPE of Polyakov operator in the fundamental representation can be written in the form
306: \eq
307:   \Tr_f(U_{\vec{x}})\; \tr_f(U_{\vec{y}})=
308: \sum_{\R\in f\otimes f}C_\R(\vert\vec{x}-\vec{y}\vert)\,\Tr_{\R}(U_{(\vec{x}+\vec{y})/2})
309: +\dots
310: \en
311: where the coefficients $C_\R(r)$ are suitable functions (they become  powers
312: of $r$ at the critical point) and the dots represent the contribution of 
313: higher dimensional local operators. The important property of  this OPE
314: is that the local operators are classified according to the irreducible 
315: representations of $G$ obtained by the decomposition of the direct product 
316: of the representations of the two local operators in the left-hand side.
317: 
318: \Cblu{On the CFT side} we have a similar structure.  The order parameter
319: $\sigma$ belongs to an irreducible representation $[\sigma]$ of the Virasoro 
320: algebra \cite{bpz} and the local operators contributing to an OPE are 
321: classified according to the decomposition of the direct product of the 
322: Virasoro representations of the left-hand-side operators. This decomposition 
323: is known as fusion algebra and can be written generically as
324: \eq
325: [\chi_i]\star[\chi_j]= c_{ij}^k[\chi_k]
326: \label{fusion}
327: \en
328: where the integers $ c_{ij}^k$  are the fusion coefficients.
329: In the case of three-state Potts model there is  a finite number of 
330: representations that we list along with their scaling dimensions
331: \footnote{Actually the critical three-state Potts model is invariant under
332: a larger algebra than that of Virasoro, the so-called ${\cal W}_3$ algebra,
333: and the representations listed in (\ref{potts}) are irreducible 
334: representations of such a larger algebra. A complete list of 
335: these fusion rules can be found for instance in \cite{Fuchs:1998qn}.}
336:   
337: 
338: \eq
339: \begin{matrix}
340: {\uu}~({\rm identity});& \sigma,\,\sigma^+~({\rm spin~fields});&
341: \epsilon~({\rm energy});&\psi,\psi^+\\
342: ~&~&~&\\
343: x_{\uu}=0&x_\sigma=\frac2{15}& x_\epsilon=
344: \frac4{15}& x_\psi=\frac43\\
345: \end{matrix}
346: \label{potts}
347: \en 
348: 
349: The fusion rules we need are
350: \eq
351: [\sigma]\star[\sigma]=[\sigma^+]+[\psi^+]~;~\,[\sigma]\star[\sigma^+]=
352: [{\uu}]+[\epsilon]~;~\,[\psi]\star[\epsilon]=[\sigma]~.
353: \label{fusion3}
354: \en 
355: 
356: Comparison of the first equation with the analogous one of the gauge side
357: \eq
358: \{3\}\otimes\{3\}=\{\bar{3}\}+\{6\}~,
359: \en
360: owing to the correspondence $\Tr_{\bar{f}}(U_{\vec{x}})\sim 
361: \sigma^+(\vec{x})$, yields a new entry of the gauge/CFT correspondence
362:  \eq
363: \Tr_{\{6\}}(U_{\vec{x}}) ~\sim~ \psi^+(\vec{x})+
364: c\,\sigma^+(\vec{x})~,
365: \label{c6}
366: \en
367: where there are no a priori reasons for the vanishing of the 
368: coefficient $c$. Hence the Polyakov-Polyakov critical correlator
369: of the symmetric representation $\{6\}$ is expected to have the 
370: following general form in the thermodynamic limit
371: \eq
372: \bra\Tr_{\{6\}}(U_{\vec{x}})\;\Tr_{\{\bar{6}\}}(U_{\vec{y}})
373: \ket_{SU(3)}^{~}=
374: \frac{c_s}{r^{2x_\sigma}}+\frac{c_u}{r^{2x_\psi}}~,
375: \label{c66}
376: \en
377:  with $r=\vert\vec{x}-\vec{y}\vert$ and $c_s,c_u$ suitable coefficients.
378: Since $x_\sigma<x_\psi$, the second term drops off more rapidly than the 
379: first, thus at large distance this correlator behaves like that of 
380: the anti-symmetric representation $\{\bar{3}\}$ as expected also at zero 
381: temperature. 
382: 
383: A similar reasoning  can be applied to the second  fusion rule 
384: (\ref{fusion3}), which is the general form relating the order parameter to 
385: the energy operator in spin systems. The gauge side of this equation is
386: \eq
387: \{3\}\otimes\{\bar{3}\}=\{1\}+\{8\}
388: \label{b33}
389: \en
390: which leads to the new entry
391: \eq
392: \Tr_{adj}(U_{\vec{x}})~\sim~ a+\epsilon(\vec{x})~,
393: \label{cadj}
394: \en
395: which is expected to be valid for any $SU(N)$ gauge theory undergoing 
396: a continuous phase transition.
397: The constant $a$ can be numerically evaluated\cite{jk} using the 
398: expected finite-size behaviour  
399: \eq
400: \bra\Tr_{adj}(U)\ket=a+\frac b{L^{2-1/\nu}}~, 
401: \en
402: where $L$ is the spatial size of the system and we used the general relation 
403: $x_\epsilon=d-1/\nu$ relating the scaling 
404: dimension of the energy operator to the thermal exponent  $\nu$. 
405: 
406: Finally the third fusion rule ( \ref{fusion3}), combined with (\ref{cadj}), 
407: can be interpreted as the CFT counterpart of the soft gluon emission which 
408: converts into each other stable $(\sim\sigma)$ and unstable $(\sim\psi)$
409: strings.
410: \vskip .3 cm
411: 
412: Another example of this gauge/CFT correspondence can be worked out in the 
413: 2+1 dimensional $SU(4)$ gauge model, where now the $2d$ CFT is the symmetric
414: Ashkin-Teller model \cite{Kadanoff:1978pv}. It describes two  Ising 
415: models with spin fields $\sigma(\vec{x})$ and $\tau(\vec{x})$ with scaling 
416: dimensions $x_\sigma=x_\tau=\frac18$ coupled through a local four spin 
417: interaction which depends on a coupling constant $g$. 
418: 
419: The model is invariant under  three different $\Z_2$ transformations:
420: \eq
421: \sigma\to-\sigma~;~~\tau\to-\tau~;~~\sigma\leftrightarrow\tau~.
422: \label{sym}
423: \en
424: The first two correspond to symmetries that can be spontaneously broken 
425: by  a thermal variation, leading to an order-disorder transition. Hence 
426: they should be associated to the center $\Z_4$ of the gauge group. The  
427: $\sigma\leftrightarrow\tau $ symmetry, on the contrary, cannot 
428: be spontaneously broken by  a thermal variation  and corresponds 
429: to charge conjugation. 
430: The first entry of the gauge/CFT correspondence can therefore be written
431: in the form
432: \eq
433: \Tr_{\{4\}}(U_{\vec{x}}) ~\sim~ \theta(\vec{x})=
434: e^{-i\pi/4}\,\left(\sigma(\vec{x})+i\tau(\vec{x})\right)~. 
435: \label{f4}
436: \en    
437: The phase $\pi/4$ comes from the observation that  
438: charge conjugation $ \sigma\leftrightarrow\tau $ should correspond, on the 
439: gauge side, to  complex conjugation; this gives the constraint
440: \eq
441: \begin{matrix} \theta(\vec{x})& \longrightarrow &\theta^*(\vec{x})\\
442: ~& \sigma\leftrightarrow\tau &~\\
443: \end{matrix}~,
444: \en
445: which fixes such a phase factor.
446: 
447: The local operators which occur in the OPE of 
448: $\theta(\vec{x})\;\theta(\vec{y})$ or 
449: $\theta(\vec{x})\;\theta^*(\vec{y})$
450: are the so-called ``ploarization'' $\pi(\vec{x})\in[\pi]=[\sigma]\star[\tau]$
451: and the ``cross-over'' operator   $\rho(\vec{x})\in[\rho]=
452: [\sigma]\star[\sigma]-[\tau]*[\tau]$ as well as the energy $\epsilon$ 
453: and the identity.
454: 
455: The Ashkin-Teller model is not an isolated 
456: critical system like the three-state Potts model, but describes a 
457: line of fixed points depending on the coupling parameter $g$.
458: While the spin fields retain along the fixed line their scaling dimensions,
459: $\epsilon$, $\pi$ and $\rho$ have $g-$dependent scaling 
460: dimensions. They obey, however, the simple relations
461: \eq
462: \frac{x_\pi}{x_\epsilon}=\frac14~,~~x_\rho\,x_\epsilon=1~.
463: \en
464: Numerical experiments \cite{deForcrand:2003wa} indicate that the critical 
465: $SU(4)$ gauge theory is located near the four-state Potts model, 
466: corresponding to $x_\epsilon=\um$  \cite{Kadanoff:1978pv}.
467: 
468: Let us consider the $SU(4)$ representations made with two quarks. We have
469: \eq
470: \{4\}\otimes\{4\}=\{6\}+\{10\}~,
471: \en
472: where the anti-symmetric representation  $\{6\}$ is real, 
473: $i.e.$ $[\Tr_{\{6\}}(U)]^*=  \Tr_{\{6\}}(U)$, being the vector 
474: representation of $SO(6)\sim SU(4)$. On the CFT side, using (\ref{f4})
475: in the OPE of   $\theta(\vec{x})\;\theta(\vec{y})$   yields at once
476: \eq
477: [\theta]\star[\theta]=[\pi]+[\rho]~,
478: \en 
479: where $[\pi]=[\sigma]\star[\tau]$  is real and even under charge 
480: conjugation  $\sigma\leftrightarrow\tau$, while $[\rho]$ is purely imaginary and odd. We have therefore 
481: two new entries for the $SU(4)$/CFT correspondence
482: \eq
483: \begin{split}
484: \Tr_{\{6\}}(U_{\vec{x}})~\sim ~\pi(\vec{x})~,\\
485: \Tr_{\{10\}}(U_{\vec{x}})~\sim ~\rho(\vec{x})+c\,\pi(\vec{x})~,
486: \end{split}
487: \en 
488: where $c$ must be different from zero, because $\Tr_{\{10\}}(U)$ 
489: is not purely imaginary.
490: 
491: In analogy with Eq.(\ref{c66}) we can write
492: \eq
493: \bra\Tr_{\{10\}}(U_{\vec{x}})\,\,\Tr_{\{\bar{10}\}}(U_{\vec{y}})
494: \ket^{~}_{SU(4)}=
495: \frac{c_s}{r^{x_\epsilon/2}}+\frac{c_u}{r^{2/x_\epsilon} }~,
496: \label{c10}
497: \en
498: with both $c_s$ and $c_u$ are different from zero. This
499:  shows that  the Polyakov-Polyakov correlator  in the symmetric 
500: representation with $N-$ality 2, even if at short distance is controlled 
501: by the irrelevant operator $\rho$, at large 
502: distance  behaves exactly like the  anti-symmetric rep. with the 
503: same $N-$ality, as expected under physical grounds.
504: \section{Decay of unstable strings at zero temperature} 
505: \label{decay}
506: 
507:  As mentioned in the introduction, the difficulty in observing  string 
508: breaking  or string decay with the Wilson loop seems to indicate nothing 
509: more than that it has a very small overlap with the broken-string or 
510: stable string state.
511: Why? being a general phenomenon which occurs for any gauge group, 
512: including $\Z_2$, in pure gauge models as well as in models coupled 
513: to whatever kind of matter, it requires a general explanation which should 
514: not depend on detailed dynamical properties of the model. 
515: A simple explanation in the case of gauge models coupled with matter was 
516: proposed in  \cite{Gliozzi:2004cs}, which now we enforce in the 
517: present case.
518: 
519: The general form of the Polyakov correlator in higher representations of 
520: $SU(N)$ found in Eq.s(\ref{c66}) and (\ref{c10}) suggests a simple  
521: Ansatz describing the asymptotic functional form
522: of the vacuum expectation value of a large, rectangular, Wilson loop in 
523: a higher representation $\R$ coupled to an unstable string which should 
524: decay into a stable  $k-$string
525: \footnote{For sake of simplicity we neglect the $1/r$ term in the potential 
526: which accounts for the quantum fluctuations of the flux tube.}
527: \eq 
528: \bra W_\R(r,t)\ket\simeq c_u\exp[-2\mu_{\R}^{~}(r+t)-\sigma_{\R}^{~} rt]+c_s
529: \exp[-2\mu_\R' (r+t)-\sigma_k^{~} rt]
530: \label{Ansatz} 
531: \en
532: Similar proposals are described in \cite{krde,as}.
533: The first term describes the typical area-law decay produced at intermediate 
534: distances by the unstable string with tension $\sigma_{\R}^{~}$. 
535: The second term is  instead the contribution expected 
536: by the stable $k-$string in which the $\R-$string decays. In the case 
537: of adjoint representation (zero $N-$ality) 
538: one has $\sigma_0=0$ and the perimeter term $\mu'_{adj}$ denotes
539: the mass of the lowest glue-lump. Eq.(\ref{Ansatz}) has to be understood as an 
540: asymptotic expansion  which approximates  $\bra W_\R(r,t)\ket$ when
541: $r,t>r_o$, where $r_o$ may be interpreted as the  scale where the confining 
542: string forms.
543: \begin{figure}
544: \centering
545: \includegraphics[width=0.9\textwidth]{potr.eps} 
546: %\vspace{5cm}  % amount of vertical space needed
547: \caption{A schematic view of the static potential between sources belonging to 
548: an excited representation $\R$ (dashed line). 
549: $V_u$ is the potential experienced at 
550: intermediate distances generated by the unstable string. $V_s$ is the 
551: asymptotic behaviour, controlled by the string tension $\sigma^{~}_k$ 
552: of the stable string in which the unstable string decays. $r^{~}_\R$ is 
553: the decay scale.} 
554: \label{Figure:2}
555: \end{figure}
556: 
557: When $t$ and $r$ are sufficiently large, no 
558: matter how small $c_s$ is, the above Ansatz 
559: implies that at  long distances the stable string  
560: eventually prevails, since 
561: $\Delta\sigma\equiv\sigma_\R-\sigma_k>0$, hence the first
562:  term drops off more rapidly than the second. With the emergence of this 
563: long-distance effect a closely related question comes in: at what distance 
564: the stable string shows up? it depends on the  
565: difference $\Delta\mu\equiv\mu'_\R-\mu^{~}_\R$. Since
566: \eq
567: -\frac1t\log\bra W_\R(r,t)\ket=2\mu'_\R+\sigma_k^{~}r-
568: \frac1t\log\left[c_s+{c_u}e^{r2\Delta\mu}
569: e^{-t(r\Delta\sigma -2\Delta\mu)}\right]~,
570: \label{logw}
571: \en
572: we have  (see Fig.\ref{Figure:2})
573: \eq
574: V_\R(r)=V_s(r)=2\mu'_\R+\sigma_k^{~}r~,~~~ r>r_\R^{~}~,
575: \label{pots}
576: \en
577: where 
578: \eq
579: V_\R(r)=-\lim_{t \to\infty}
580: \frac1t\log\bra W_\R(r,t)\ket
581: \label{poto} 
582: \en
583:  is the static potential 
584: and $r_\R^{~}$ is the  value of $r$ which annihilates the exponent in
585: Eq.(\ref{logw}) 
586: \eq
587: r_\R^{~}=\frac{2\Delta \mu}{\Delta \sigma}\equiv2\frac{\mu'_\R-\mu^{~}_\R}
588: {\sigma^{~}_\R-\sigma^{~}_k}~.
589: \label{scale}
590: \en
591: Creation of unstable strings requires  $\mu'_\R>\mu^{~}_\R$
592: (see Fig.\ref{Figure:2}). In the case of 
593: zero $N-$ality the above equation yields the usual estimate of the adjoint 
594: string breaking scale. The inclusion of the $\frac1r$ terms in this analysis 
595: does not modify substantially the numerical estimates.  
596:  Notice that the mass $\mu_\R^{~}$ and 
597: $\mu'_\R$ are not UV finite because of the additive  self-energy divergences
598: (linear in 3+1 dimensions, logarithmic in 2+1 dimensions), which cannot be 
599: absorbed in a parameter of the theory. However, these divergences should 
600: cancel in their difference, hence $r^{~}_\R$ is a purely dynamical scale, 
601: defined  for any non fully antisymmetric representation of $SU(N)$, which 
602: cannot be tuned by any bare parameter of the theory. 
603: 
604: When $r$ is less than the decay scale $r^{~}_\R$ Eq.(\ref{pots}) is no longer 
605: valid and is replaced by
606: \eq
607: V_\R(r)=V_u(r)=2\mu^{~}_\R+\sigma_\R^{~}r~,~~~ r_o<r<r_\R^{~}~.
608: \label{potu}
609: \en
610: Thus the Ansatz (\ref{Ansatz}) describes the unstable string decay as a level
611: crossing phenomenon, as observed in the adjoint string.
612: \par
613: It is worth noting that, though $c_u$ and $c_s$ must be non-vanishing 
614: quantities, they  do not contribute to $V_\R(r)$, whereas  play  a 
615: fundamental role in the 
616: possibility to observe string decay when both $r$ \Cblu{and} $t$ are finite. 
617: It is easy to see that  for  $r<r^{~}_\R$ and $t$ large enough the unstable 
618: string cannot decay, the reason being that $V_u(r)<V_s(r)$ in this range.
619: In order to avoid the unphysical behaviour in which the decay is visible 
620: only in a  finite interval of  $t$,  it is  obvious that $c_s$ cannot be 
621: too big. More precisely, we must assume
622: \eq 
623: c_u\, e^{-2\mu^{~}_\R(r+t)-\sigma_\R^{~} rt}\geq c_s\, 
624: e^{-2\mu'_\R(r+t)-\sigma_k^{~} rt}~,~  
625: r_o\leq r\leq r^{~}_\R~,~
626: t\ge r_o~.
627: \label{Ineq}
628: \en
629: With the help of Eq.(\ref{scale}), this inequality can  be recast into the form
630: \eq
631: \log\frac{c_s}{c_u}\leq\Delta\sigma\,
632: \left[r^{~}_\R(r+t)-rt\right]~~,
633: ~r_o\leq r\leq r^{~}_\R,~t\ge r_o~.
634: \label{ineq}
635: \en
636: To minimise the right-hand-side we put $r=t=r_o$ and get
637: 
638: \eq
639: \log\frac{c_s}{c_u}\leq\Delta\sigma\,
640: r_o\,(2r^{~}_\R-r_o)~~.
641: \label{bound}
642: \en
643: Such an upper bound constitutes the main obstruction to observe string decay
644: in the ordinary Wilson loop $W_\R$. Indeed, when $r>r^{~}_\R$ the 
645: distance $t^{~}_\R$
646: where the second term of the Ansatz (\ref{Ansatz}) equals the first, then 
647: it makes possible to see the decay, is given by
648: \eq
649: t^{~}_\R\,(r-r^{~}_\R)=r\,r^{~}_\R-\frac1{\Delta\sigma}
650: \log\frac{c_s}{c_u}\ge\, r^{~}_\R \,(r- 2r_o)+r_o^2~~,
651: \en
652:  which shows that $ t^{~}_\R\gg r^{~}_\R$  unless   $ r\gg r^{~}_\R$,
653: and we shall argue shortly that $r^{~}_\R> r_{adj}$.
654: From a 
655: computational point of view it is very challenging  to reach such 
656: length scales \footnote {In 3+1 dimensional $SU(2)$ 
657: gauge model, for instance, $r_{adj}\sim 1.25$ fm \cite{fp}.}
658: in the measure of $ \bra W_\R(r,t)\ket $. 
659: This explains why unstable string decay has not yet been observed at zero 
660: temperature. 
661: \par The evaluation of $r^{~}_\R$
662: is  problematic, because  Eq.(\ref{scale})   implies an estimate of
663: the self-energy $\mu'_\R$, which contributes to the sub-dominant term of the 
664: Ansatz (\ref{Ansatz}), hence from a numerical point of view it is almost 
665: hopeless. We shall see that the mixed Wilson loop can be used  as a 
666: simple tool to extract this quantity.
667: 
668: Waiting for a computational work which will provide us with this information, 
669: we now try to combine together some known facts in order to get a rough 
670: estimate of $r^{~}_\R$.  Numerical data on \Cblu{unstable} 
671: strings in 3+1 dimensional $SU(3)$ \cite{sd,ba} and $SU(2)$ \cite{cp} 
672: gauge theories seem to support Casimir scaling \cite{aop}, which tells us that
673: the static potential between sources in the representation $\R$ is proportional 
674: to that of fundamental sources according to
675: \eq
676: V_\R(r)\simeq \frac{{\cal C}_\R}{{\cal C}_f}\,V_f(r)~~,
677: \label{casimir}
678: \en
679: where ${\cal C}_\R$ is the quadratic Casimir operator of the representation 
680: $\R$.  This implies 
681: \eq
682: \sigma_\R^{~} \simeq \frac{{\cal C}_\R}{{\cal C}_f}\,\sigma~,  
683: ~~\mu_\R^{~} \simeq\frac{{\cal C}_\R}{{\cal C}_f} \, \mu~,
684: \label{cas}
685: \en
686: where $\sigma$ and $\mu$ refer to the fundamental representation.
687: Since the additive UV divergence of $\mu^{~}_\R$ is cancelled by that of  
688:  $\mu'_\R$ in the difference $\Delta\mu$, it is quite natural to expect 
689:  $\mu'_\R$ also  proportional to ${\cal C}_\R$. However, only the 
690: unstable $\R-$strings have a non-vanishing  $\mu'$, hence we assume
691: \eq
692: \mu'_\R\simeq\frac{{\cal C}_\R}{{\cal C}_{adj}}\,\,\mu'_{adj}~,
693: \label{mumu}
694: \en
695: where, as noted above, $\mu'_{adj}$ is the mass of the lowest glue-lump.
696: Thus, combining Eq.s(\ref{scale}),(\ref{cas}) and (\ref{mumu}) leads to
697: \eq
698: r_\R^{~}\simeq \,\frac{{\cal C}_\R}
699: {{\cal C}_\R-\sigma^{~}_k/{\sigma}}\;r_{adj}^{~}~,
700: \label{esca}
701: \en
702: hence  $r^{~}_\R> r_{adj}$ for any \Cblu{unstable} string coupled to 
703: a colour source with non-vanishing $N-$ality. For instance, in the $SU(2)$ 
704: case, the $j=\frac32$ is unstable against  decaying into the fundamental 
705: string with $j=\frac12$, thus $r^{~}_{\frac32}\simeq\frac54 r_{adj}$. 
706:  
707: Notice that, unlike the unstable strings, the  
708: \Cblu{stable} $k-$strings tension 
709: $\sigma_k/\sigma$ does not obey Eq.(\ref{cas}).  
710: Numerical experiments 
711: \cite{dprv,lt,ltw,dprv2} as well as  consistency of $N\to\infty$  limit
712: \cite{as} point to a sizable
713: violation of the Casimir scaling and some works \cite{dprv,dprv2} find 
714: good agreement with the so called sine law discovered in some supersymmetric 
715: gauge models\cite{ds,hsz,hk}.
716: \par 
717: Our rough estimate of the decay lengths of unstable strings were derived under
718: a number of (strong) assumptions. A better definition, based on the 
719: multichannel method, will be described in the next section. 
720: \section{Mixed Wilson loops}
721: \label{mixed}
722: To warm up, let us consider the construction of a $SU(N)$ glue-lump operator 
723: which creates a glue-lump at the point $\vec{x}$ and annihilates it at the 
724: point $\vec{y}$.
725: \begin{figure}
726: \centering
727: \includegraphics[width=0.95\textwidth]{gluelump.eps} 
728: %\vspace{5cm}  % amount of vertical space needed
729: \caption{Generating operator of the gluelump, as described in 
730: Eq.(\ref{glu}).} 
731: \label{Figure:3}
732: \end{figure}
733: 
734: 
735: The starting point is the gauge-invariant operator depicted in Fig.
736: \ref{Figure:3}
737: \eq
738: P(\vec{x})^i_j\;\;U^\dagger(\vec{x},\vec{y})^j_k\;\; 
739: Q(\vec{y})^k_l\;\; U(\vec{x},\vec{y})^l_i~~,
740: \label{glu}
741: \en
742: where $P(\vec{x})$ and  $Q(\vec{y})$ are the source and the sink of the 
743: gluelump. For simplicity, the points $\vec{x}$ and $\vec{y}$ 
744: are chosen on the same coordinate axis. In a 3D cubic lattice, for instance, 
745: a good overlap with the 
746: lowest $0^+$ state is obtained by 
747: choosing $P$ and $Q$ as the `clover leaves' \cite{mc} formed by the four 
748: plaquettes orthogonal to  $\vec{y}-\vec{x}$.  $U(\vec{x},\vec{y})$ 
749: is a shorthand  notation for the parallel transporter formed by the  product 
750: of the link variables in the fundamental 
751: representation along the straight path connecting the sites 
752: $\vec{x}$ and $\vec{y}$. Let us focus on the indices $i$ and $j$ of the 
753: operators $U$ and $U^\dagger$. They belong to the reducible representation
754: $\{N\}\otimes\{\bar{N}\}=\{N^2-1\}+\{1\}$. In order to project out the singlet 
755: and allow propagating the adjoint representation only, we 
756: perform the substitution
757: \eq
758:  {U^\dagger}^j_k \,U^l_i\to   {U^\dagger}^j_k\, U^l_i-\frac1N\,\delta^j_i \,
759:  {U^\dagger}^m_k\, U^l_m= {U^\dagger}^j_k\, U^l_i-\frac1N\,\delta^j_i\;\delta^l_k~.
760: \en
761: Inserting this projection in (\ref{glu}) yields
762: \eq
763: G(\vec{x},\vec{y})=\tr(PUQU^\dagger)-\frac1N\,\tr P\;\tr Q~.
764: \label{gluelump}
765: \en
766: In the case of $SU(2)$ we can use the trace identity
767: \eq
768: \tr A\,\tr B=\tr(A\,B)+\tr(A\,B^{-1})~,
769: \label{ti}
770: \en
771: valid for any pair of $2\times2$ unimodular matrices
772: \footnote{Actually it can be shown that the most general solution of 
773: the functional equation  $\phi(A)\,\phi(B)=\phi(A\,B)+\phi(A\,B^{-1})$, 
774: where $A$ and $B$ are arbitrary elements of an unspecified group $G$ and 
775: $\phi$ is a class function, is  $G\equiv SL(2,\C)$ and $\phi$ is the character 
776: of its fundamental representation. A study of this kind of relationships for various groups can be found in\cite{gv}. }, 
777: with $A=P$ and $B=UQU^\dagger$ and recast 
778: Eq.(\ref{gluelump}) in the form used by the lattice community
779: \eq
780: G(\vec{x},\vec{y})=\um\tr[PU(Q-Q^\dagger)U^\dagger]\equiv
781: \um\tr(P\sigma^a)\Gamma_{ab}\tr(Q\sigma^b)~,
782: \en
783: where the $\sigma^a\equiv\sigma_a$ are the Pauli matrices , $a=1,2,3$, and
784: \eq
785: \Gamma_{ab}=\um\tr(\sigma_aU\sigma_bU^\dagger)~.
786: \en 
787: 
788: 
789: \begin{figure}
790: \centering
791: \includegraphics[width=0.95\textwidth]{wilmix.eps} 
792: %\vspace{5cm}  % amount of vertical space needed
793: \caption{Mixed  Wilson loop. The loop $(PUVUQU^\dagger)$ denotes a static source in the fundamental representation $f$. The  three coincident 
794: lines $U,U,U^\dagger$ at  the top are drawn separately for clarity. They 
795: belong to the reducible representation $f\otimes f\otimes \bar{f}$. 
796: Projection on  a irreducible component is described in the text. } 
797: \label{Figure:4}
798: \end{figure}
799: 
800: To construct a \Cblu{mixed Wilson loop} let us start by considering an 
801: arbitrary closed path $\gamma=u\,v$ made by the composition of two open paths
802: $u$ and $v$. Let $U$ and $V$ be the group elements associated with 
803: these two paths respectively. The associated standard Wilson loop is
804: \eq
805: W(\gamma)=\tr(UV)~.
806: \label{wo}
807: \en  
808: where the trace is taken , here and in the following, in the fundamental 
809: representation $f$  of the \Cblu{\sl non-abelian} gauge group $G$. We want to 
810: transform $W$ in a mixed Wilson loop in which the source along the path $u$ 
811: carries the quantum numbers of an
812: higher representation $\R$ belonging to the same $N-$ ality class of $f$.
813: In analogy with the construction of the gluelump operator, we start 
814: by considering  the gauge-invariant 
815: generating operator (see Fig.\ref{Figure:4})
816: \eq
817: P^m_i\,U^i_l\,V^l_j\,U^j_k\,Q^k_n\,{U^\dagger}^n_m
818: \label{genera}
819: \en  
820: where $P$ and $Q$ can be taken as the source operators  associated to the 
821: ``clovers'' orthogonal to the path $\gamma$ at the junctions separating $u$
822: and $v$.    Along the path $u$ now propagates a  source belonging to the 
823: reducible $f\otimes f\otimes\bar{f} $. We have then to project on 
824: some irreducible component.
825: 
826: 
827:  To make a specific, illustrative example, let us consider the case of 
828: $SU(3)$, where we have
829: \eq
830: \{3\}\otimes \{3\}\otimes\{\bar{3}\}=2\{3\}+\{\bar{6}\}+\{15\}~.
831: \en
832: We want to project on the $\{\bar{6}\}$ representation which has the same
833: triality of $\{3\}$. It may be selected, for instance, by anti-symmetrizing 
834: the indices $i$ and $j$ and eliminating the traces 
835: with the index $m$ in the matrix elements $P^m_i\,V^l_j$ . 
836: Namely, we take the combination
837: 
838: \eq
839: \frac32(P^m_j\,V^l_i-P^m_i\,V^l_j)+ 
840: \frac34[(\delta^m_i\,V^l_j -\delta^m_j\,V^l_i) \tr P-(VP)^l_j\,\delta^m_i+
841: \delta^m_j\,(VP)^l_i]
842: \label{prj6}
843: \en
844: and saturate with $ U^i_l\,U^j_k\,Q^k_n\,{U^\dagger}^n_m$, getting finally
845: \eq
846: \begin{split}
847: W^{~}_{\{3\}\to\{\bar{6}\}}(\gamma)=\frac32\left[\tr(PUQU^\dagger)\,\tr(VU)-
848: \tr(PUVUQU^\dagger)\right]+\\
849: \frac34\left[\tr P\, \tr(VUQ)-\tr(VPUQ)+\tr (VPU)\,\tr Q-\tr P\tr(VU)\,\tr Q
850: \,\right]~.
851: \label{w36}
852: \end{split}
853: \en
854: The  fraction $u$ of $\gamma$ belonging to $\{\bar{6}\}$ is of course arbitrary. Moving the two junctions along $\gamma$ we can manage as to shrink the 
855: length $\vert v\vert$ of $v$ to zero, hence $V$  is the identity matrix
856: $V=\uu$ and the whole loop carries the quantum numbers of $\{\bar{6}\}$.
857: In this limit the two clovers associated to $P$ and $Q$, which have opposite 
858: orientations, overlap, hence $P=Q^\dagger$. 
859: As a check of Eq.(\ref{w36}) we can now 
860: integrate over $Q$, using the standard orthogonality relations of the 
861: irreducible characters
862: \eq
863: \begin{split}
864: \int_{Q\in G}dQ\,\Tr_\R(Q\,A)\Tr_{\R'}(Q^\dagger B)=\delta_{\R,\R'}
865: \,\frac 1{d_\R}\Tr_\R(A^\dagger B)\\
866:   \int_{Q\in G}dQ\,\Tr_\R(Q\,A\,Q^\dagger B)=
867: \frac 1{d_\R}\Tr_\R(A^\dagger B)~,
868: \end{split}
869: \en
870:  written for any arbitrary compact group $G$. $\Tr_\R$ is the character,
871: {\sl i.e.} the trace calculated in the irreducible representation $\R$ and 
872: $d_\R=\Tr_\R(\uu)$ its dimension.
873: 
874: Putting in  Eq.(\ref{w36}) $V=\uu$, $P=Q^\dagger$ and integrating on $Q$ yields
875: \eq
876: \um\left[ (\tr U)^2-\tr(U^2)\right]\tr U^\dagger-\tr U=
877: \um\left[(\tr U^\dagger)^2+\tr({U^\dagger}^2)\right]=
878: \Tr_{\{\bar{6}\}}(U)~,
879: \en
880: where the $SU(3)$ identity $\um[ (\tr U)^2-\tr(U^2)]=\tr U^\dagger$ and 
881: its conjugate have been used.
882: 
883: Similarly, in the $SU(2)$ case, the mixed Wilson loop  associated to the 
884: pair  of representations $j=\um$ and $j=\frac32$ turns out to be
885: \eq
886: \begin{split}
887: W^{~}_{\um\to\frac32}(\gamma)=\tr(PUVUQU^\dagger)+\tr(PUQU^\dagger)\tr(UV)-\\
888: \frac13\left[\tr(PUQV)+\tr(PUV)\,\tr Q+\tr P\left(\tr(VUQ)+\tr(VU)\,\tr Q
889: \right)\right]~,
890: \end{split}
891: \en
892: where the normalisation is chosen, like in the $SU(3)$ case, in such a way 
893: that integration on $Q$ after putting $V=\uu$ and $P=Q^\dagger$ gives
894: \eq
895: \int_{Q\in SU(2)}dQ\,W^{~}_{\um\to\frac32}=
896: \Tr_{\frac32}(U)\equiv\tr(U^2)\,\tr U~.
897: \en
898:  \par
899: It is clear that the above construction can be generalised to any non-abelian 
900: group and, in particular, to any fully anti-symmetric representation 
901: of $N-$ality $k$ of $SU(N)$, which can be converted through the emission and 
902: the reabsorption of a glue-lump to an excited representation $\R$.
903: It is also clear that one can build up mixed Wilson loops of the 
904: type drawn in Fig.\ref{Figure:1} $c$ and $d$ that we denote respectively as 
905: $W^{~}_{k\to\R}(r,t)$ and $W^{~}_{\R\to k\to\R}(r,t)$\footnote{The ``time'' 
906: variable $t$ is the vertical line in Fig.\ref{Figure:1}.}.
907: \par The static potential $V_\R(r)$ between the $\R$ sources and the 
908: decay of the 
909: associated unstable string into the stable $k-$string can then be extracted 
910: from measurements of the matrix correlator
911: \eq
912: C(r,t)=\left(
913: \begin{matrix}
914: \bra W^{~}_{\R}(r,t)\ket &\bra W^{~}_{k\to\R}(r,t)\ket\\
915: \bra W^{~}_{k\to\R}(r,t)\ket&\bra W^{~}_{\R\to k\to\R}(r,t)\ket
916: \end{matrix}
917: \right)~,
918: \label{cor}
919: \en   
920: where $W^{~}_{\R}(r,t)$ is the ordinary Wilson loop  when the whole 
921: rectangle $(r,t)$ is in the $\R$ representation. This is the 
922: generalisation of the multichannel method we alluded in the Introduction.
923: Denoting by $\lambda(r,t)$  the highest eigenvalue of Eq.(\ref{cor})
924: we have
925: \eq
926: V_\R(r)=-\lim_{t\to\infty}\frac1t\log\lambda(r,t)~,
927: \en
928: which from a computational point of view is much better than the definition
929: (\ref{poto}), because the mixed Wilson loops have, by construction, a much 
930: better overlap with the stable $k-$string state. 
931: In this way it will possible to evaluate the decay scale $r_\R$. 
932: One could also try to get a rough estimate of this quantity through 
933: Eq.(\ref{scale}). Indeed the instability of the $\R-$string leads to conjecture that the vacuum expectation value of 
934: $W^{~}_{k\to\R}(\gamma)$ should behave asymptotically as 
935: \eq
936: \bra W^{~}_{k\to\R}(\gamma)\ket\propto e^{-\mu_\R^{~}\vert v\vert-
937: \mu_\R'\vert u\vert-\sigma^{~}_k A}~,
938: \en
939: where $A$ is the area of the minimal surface encircled by $\gamma$ and 
940: $\vert v\vert$ and $\vert u\vert$  are the lengths of the paths which carry 
941:  the quantum numbers of the representations $k$ and $\R$, respectively.
942: This seems the most effective way to estimate the quantity $\mu'_\R$
943: and therefore $r^{~}_\R$.
944:  \section{Conclusion}
945: \label{conclusion}
946: In this paper we gained some insight into the physics of $SU(N)$ pure
947: gauge theory by using in two different ways the standard decomposition 
948: of the direct product of irreducible representations of the gauge group.
949: 
950: First, we mapped this decomposition into the fusion rules of the effective CFT 
951: describing the critical behaviour of the finite-temperature deconfinement 
952:  of those 2+1 dimensional gauge theories which undergo a continuous phase 
953: transition. 
954: 
955: We worked out two specific examples, $SU(3)$ and $SU(4)$, which 
956: led us to conjecture the exact critical exponents to be attributed to Polyakov 
957: lines in some higher representations of the gauge group. 
958: The resulting functional form of these Polyakov-Polyakov correlators at 
959: critically suggests generalising it to $T=0$ Wilson loops in higher 
960: representations for any $SU(N)$. The proposed Ansatz offers a simple, 
961: general explanation of the difficulty to observe decaying unstable strings 
962: while measuring Wilson loops only.
963: 
964: The other way we exploited the mentioned decomposition in irreducible 
965: representations has been used to define a new gauge-invariant operator,
966: the mixed Wilson loop $W^{~}_{\R\to\s} (\gamma)$ , which describes a 
967: source belonging to an irreducible representation $\R$ which is converted into 
968: another representation $\s$ 
969: along part of its trajectory $\gamma$. Gauge invariance requires $\R$ and 
970: $\s$ having the same $N-$ality. This new operator could be used as a tool 
971: to study the decay of the unstable strings, a recurrent theme of this work. 
972:     
973:   
974: \begin{thebibliography}{9}
975: \bibitem{as} A. Armoni and M. Shifman, Nucl. Phys. B {\bf 671}, 67 (2003)
976:  [arXiv:hep-th/0307020].
977: \bibitem{ds} M.R. Douglas and S.H. Shenker, Nucl. Phys. B {\bf 481}, 513 (1995)
978: [arXiv:hep-th/9503163].
979: \bibitem{hsz}  A. Hanany, M.J. Strassler and A. Zaffaroni, Nucl Phys. B 
980: {\bf 513}, 87 (1998)[arXiv:hep-th/9707244].
981: \bibitem{hk} C.P. Herzog and I. R. Klebanov, Phys. Lett. B {\bf 526}, 388 
982: (2002) [arXiv:hep-th/0111078].
983:  \bibitem{dprv} L. Del Debbio, H. Panagopoulos, P. Rossi and E. Vicari, Phys.
984: Rev. D {\bf 65}, 021501 (2002) [arXiv:hep-th/0106185].
985: \bibitem{lt}B. Lucini and M. Teper, Phys. Rev. D {\bf 64}, 105019 (2001)
986: [arXiv:hep-lat/0107007].
987:  \bibitem{ltw}B. Lucini, M. Teper and U. Wengler, JHEP {\bf 0406}, 012 
988: (2004 ) [arXiv:hep-lat/0404008].
989: \bibitem{dprv2} L. Del Debbio, H. Panagopoulos, P. Rossi and E. Vicari, 
990:  JHEP {\bf 01} , 009 (2002) [arXiv:hep-th/0111090].
991: \bibitem{cj} N.A. Campbell, I.H. Jorysz, C. Michael, Phys. Lett.B {\bf 167}
992: (1986) 91. 
993: \bibitem{pt} G. Poulis and H.D. Trottier, Phys. Lett. B {\bf 400}(1997)358
994: [arXiv:hep-lat/9504015];  C.  Michael, ArXiv:hep-lat/9809211.
995: \bibitem{sd} S. Deldar, Phys. Rev. D {\bf 62}(2000) 034509 
996: [arXiv:hep-lat/9911008]
997: \bibitem{ba}G. Bali, Phys. Rev. D {\bf 62} (2000) 114503 
998: [arXiv:hep-lat/0006022]. 
999: \bibitem{cp} C. Piccioni, arXiv:hep-lat/0503021.
1000: \bibitem{krde} S.Kratochvila and P.de Forcrand, Nucl Phys. B {\bf 671}
1001:  (2003) 103 [arXiv:hep-lat/0306011].
1002: \bibitem{cm}  C. Michael, Nucl Phys. B (Proc. Suppl.){\bf 26} (1992) 417.
1003: \bibitem{st}P.W. Stephenson, Nucl. Phys. B {\bf 550} (1999) 427 
1004: [arXiv:hep-lat/9902002]; O. Philipsen and H. Wittig, Phys. Lett. 
1005: B {\bf 451} (1999) 146 [arXiv:hep-lat/9902003].
1006: \bibitem{fp}P. de Forcrand and O.Philipsen, Phys. Lett. B {\bf 475} (2000) 280
1007: [arXiv:hep-lat/9912050]; K.Kallio and H.D. Trottier, 
1008: Phys. Rev. D {\bf 66} (2002) 
1009: 034503 [arXiv:hep-lat/0001020].
1010: \bibitem{pw} O. Philipsen and H.Wittig, Phys. Rev. Lett. {\bf 81}
1011: (1998) 4056 [Erratum-ibid.{\bf 83} (1999) 2684][hep-lat/9807020];
1012:  F.Gliozzi and A.Rago, Phys. Rev. D {\bf 66} (2002) 074511
1013: [arXiv:hep-lat/0206017]. 
1014:  \bibitem{ks}F. Knechtli and R. Sommer [ALPHA Collaboration], Phys. Lett. B 
1015: {\bf 440} (1998) 345 [arXiv:hep-lat/9807022] and Nucl. Phys. 
1016: B {\bf 590} (2000) 309
1017: [hep-lat/0005021].
1018: \bibitem{bdl}G.S. Bali, H. Neff, Th. D\"ussel, Th Lippert, 
1019: K.Schilling, Phys. Rev. D {\bf 71} (2005) 114513 [hep-lat/0505012].
1020: 
1021: \bibitem{mc} C. Michael, Nucl. Phys. B {\bf 259} (1985) 58
1022: \bibitem{Gliozzi:2004cs}
1023:   F.~Gliozzi and A.~Rago,
1024:   %``Overlap of the Wilson loop with the broken-string state,''
1025:   Nucl.\ Phys.\ B {\bf 714} (2005) 91
1026:   [arXiv:hep-lat/0411004].
1027:   %%CITATION = HEP-LAT 0411004;%%
1028: 
1029: 
1030: \bibitem{gp}F.Gliozzi and P. Provero, Nucl. Phys. B {\bf 556} (1999) 76
1031: [arXiv:hep-lat/9903013]; Nucl. Phys. B (Proc. Suppl.) {\bf 83} (2000) 46 
1032: [arXiv:hep-lat/99070231]. 
1033: 
1034: 
1035: 
1036: \bibitem{cb} C. Bernard, Phys. Lett. B {\bf 108} (1982) 431; Nucl Phys. B 
1037: {\bf 219} (1983) 341.
1038: \bibitem{ot} S.Ohta, M. Fukugita and A. Ukawa, Phys. Lett B {\bf 173}(1986)
1039: 15.
1040: \bibitem{mf} H. Markum and M.E. Faber, Phys. Lett. B {\bf 200} (1988) 343.
1041: 
1042: %\cite{Muller:1991xj}
1043: \bibitem{Muller:1991xj}
1044:   M.~M\"uller, W.~Beirl, M.~Faber and H.~Markum,
1045:   %``Universality of the confinement string in multiplet potentials,''
1046:   Nucl.\ Phys.\ Proc.\ Suppl.\  {\bf 26} (1992) 423.
1047:   %%CITATION = NUPHZ,26,423;%%
1048: 
1049: %\cite{DelDebbio:2004zc}
1050: \bibitem{DelDebbio:2004zc}
1051:   L.~Del Debbio, H.~Panagopoulos and E.~Vicari,
1052:   %``N-ality and topology at finite temperature,''
1053:   arXiv:hep-th/0409203.
1054:   %%CITATION = HEP-TH 0409203;%%
1055: %\cite{DelDebbio:2003tk}
1056: \bibitem{DelDebbio:2003tk}
1057:   L.~Del Debbio, H.~Panagopoulos and E.~Vicari,
1058:   %``Confining strings in representations with common n-ality,''
1059:   JHEP {\bf 0309} (2003) 034
1060:   [arXiv:hep-lat/0308012].
1061:   %%CITATION = HEP-LAT 0308012;%%
1062: \bibitem{gw} M. Gross and J. F. Wheater, Phys. Rev. Lett. {\bf 54} (1985) 389;
1063: P. H. Damgaard , Phys. Lett. B {\bf 183}(1987) 81;
1064: M.E.Faber, H. Markum and M. Meinahart, Phys. Rev. D {\bf 36} (1987)
1065: 632.
1066: \bibitem{dam} P.H. Damgaard, Phys. Lett. B {\bf 194} (1987) 107;
1067: K. Redlich and H. Satz, Phys. Lett. B {\bf 213} (1988) 191;
1068: J. Christensen and P.H.Damagaard, Nucl. Phys. B {\bf 348}
1069: (1991) 226; Nucl. Phys. B {\bf 354} (1991) 339.
1070: \bibitem{jk} J. Kiskis, Phys. Rev. D {\bf 41} (1990) 3204; J.Kiskis and 
1071: P. Vranas, Phys. Rev. D {\bf 49} (1994) 528.
1072: \bibitem{ctd} Christensen, G. Thorleifsson, P.H.Damagaard and 
1073: J.F. Wheater, Nucl. Phys. B{\bf 374} (1992) 225.
1074: \bibitem{dh} P.H. Damgaard and M. Hasenbusch, Phys. Lett. B 
1075: {\bf 331} (1994) 400.
1076: \bibitem{pi} R. D. Pisarski, Phys. Rev. D {\bf 62} (2000) 111501
1077: \bibitem{sy}B. Svetitsky and L.G. Yaffe, Nucl. Phys. B {\bf 210} [FS6]
1078: (1982) 423.
1079: 
1080: %\cite{Gliozzi:1997yc}
1081: \bibitem{Gliozzi:1997yc}
1082:   F.~Gliozzi and P.~Provero,
1083:   %``The Svetitsky-Yaffe conjecture for the plaquette operator,''
1084:   Phys.\ Rev.\ D {\bf 56} (1997) 1131
1085:   [arXiv:hep-lat/9701014].
1086:   %%CITATION = HEP-LAT 9701014;%%
1087: 
1088: 
1089: \bibitem{bpz} A.A. Belavin, A.M. Polyakov and  A. B. Zamolodchikov, 
1090: Nucl. Phys. B {\bf 241} (1984) 333.
1091: 
1092: \bibitem{Fuchs:1998qn}
1093:   J.~Fuchs and C.~Schweigert,
1094:   %``Completeness of boundary conditions for the critical three-state Potts
1095:   %model,''
1096:   Phys.\ Lett.\ B {\bf 441} (1998) 141
1097:   [arXiv:hep-th/9806121].
1098:   %%CITATION = HEP-TH 9806121;%%
1099: %\cite{Kadanoff:1978pv}
1100: \bibitem{Kadanoff:1978pv}
1101:   L.~P.~Kadanoff and A.~C.~Brown,
1102:   %``Correlation Functions On The Critical Lines Of The Baxter And Ashkin-Teller
1103:   %Models,''
1104:   Annals Phys.\  {\bf 121} (1979) 318.
1105:   %%CITATION = APNYA,121,318;%%
1106: %\cite{deForcrand:2003wa}
1107: \bibitem{deForcrand:2003wa}
1108:   P.~de Forcrand and O.~Jahn,
1109:   %``Deconfinement transition in 2+1-dimensional SU(4) lattice gauge theory,''
1110:   Nucl.\ Phys.\ Proc.\ Suppl.\  {\bf 129} (2004) 709
1111:   [arXiv:hep-lat/0309153].
1112:   %%CITATION = HEP-LAT 0309153;%%
1113: 
1114: 
1115: 
1116: 
1117: \bibitem{aop} J. Ambj\o rn, P.Olesen and C. Peterson, Nucl Phys. B 
1118: {\bf 240},186 (1984; Nucl. Phys. B {\bf 240}, 533 (1984).
1119: 
1120: \bibitem{gv} F.Gliozzi and M.A. Virasoro, Nucl Phys. B {\bf 164}(1980) 141. 
1121: 
1122: \end{thebibliography}
1123: 
1124: \end{document}  
1125: 
1126: