hep-lat0510076/gn.tex
1: \documentclass[12pt]{article}
2: %\documentclass[11pt]{article}
3: \usepackage{amsmath,epsfig,amssymb,eufrak}
4: \usepackage{amscd,latexsym}
5: \usepackage{color}
6: 
7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
8: 
9: \headheight=12pt
10: %\headsep=36pt
11: \footskip=48pt
12: \hoffset=-.3truecm
13: \voffset=-1.2truecm
14: 
15: \textwidth=15truecm
16: \textheight=22truecm
17: 
18: \date{}
19: 
20: \def\ppall{\mathaccent23p}
21: \def\ppall{\mathaccent23p}
22: \def\kpall{\mathaccent23k}
23: \def\qpall{\mathaccent23q}
24: \def\xlf{\raisebox{+0.2em}{\color{red}\boldmath{$\chi$}}\hspace{-0.2ex}\raisebox{-0.2em}{\color{green}L}
25: \hspace{-1.5ex}\raisebox{+0.14em}{\color{blue}F}\hspace{2mm}}
26: 
27: \def\lsi{\raise0.3ex\hbox{$<$\kern-0.75em\raise-1.1ex\hbox{$\sim$}}}
28: \def\gsi{\raise0.3ex\hbox{$>$\kern-0.75em\raise-1.1ex\hbox{$\sim$}}}
29: 
30: \newcommand{\lsim}{\mathop{\lsi}}
31: \newcommand{\gsim}{\mathop{\gsi}}
32: 
33: \newcommand{\e}{{\rm{e}}}
34: \newcommand{\msb}{\overline{\mathrm{MS}}}
35: \newcommand{\dds}{\stackrel{\leftrightarrow}{D}}
36: \renewcommand{\thefootnote}{\fnsymbol{footnote}}
37: 
38: \newcommand{\beq}{\begin{equation}}
39: \newcommand{\eeq}{\end{equation}}
40: \newcommand{\beqn}{\begin{eqnarray}}
41: \newcommand{\eeqn}{\end{eqnarray}}
42: \newcommand{\bpsi}{\bar\psi}
43: \newcommand{\bchi}{\bar\chi}
44: 
45: 
46: \input macros.sty      % \eqalign, \meqalign
47: \input eqalign.sty     % \eqalign, \meqalign
48: 
49: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
50: 
51: \begin{document}
52: 
53: \begin{titlepage}
54: 
55: \title{
56:   {\vspace{-0cm} \normalsize
57:   \hfill \parbox{40mm}{DESY/05-206\\
58:                        October 2005}}\\[10mm]
59: Two--dimensional lattice Gross--Neveu model\\
60: with Wilson twisted mass fermions}
61: \author{Kei-ichi Nagai and Karl Jansen          
62: %
63: \\
64: %\\
65: %   {\bf \xlf Collaboration}\\
66: \\
67: %
68: {\small   John von Neumann-Institut f\"ur Computing NIC,} \\
69: %{\small         Deutsches Elektronen-Synchrotron DESY,}\\ 
70: {\small         Platanenallee 6, D-15738 Zeuthen, Germany} \\ \ \\
71: %
72: }
73: 
74: \maketitle
75: 
76: \begin{abstract}
77: We study the two-dimensional lattice Gross--Neveu model
78: with Wilson twisted mass fermions 
79: in order to explore the phase structure in this setup. 
80: In particular, we investigate the behaviour of the phase transitions
81: found earlier with standard Wilson fermions as a function of the 
82: twisted mass parameter $\mu$. We find that qualitatively the dependence 
83: of the phase transitions on $\mu$ is very similar to the 
84: case of lattice QCD.
85:   \vspace{0.75cm}
86: \noindent
87: %%%{\it PACS:}  ; ; \\
88: %%%{\it Keywords:}  ; .
89: \end{abstract}
90: 
91: \end{titlepage}
92: 
93: 
94: %%%%%%%%%%%%%%%%%%%%%%%%%%%%% SECTION %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
95: \section{Introduction}
96: \label{sec:intro}
97: 
98: Recent results from simulations of lattice QCD with Wilson twisted 
99: mass fermions \cite{Frezzotti:2000nk,Frezzotti:2003ni} 
100: revealed an 
101: intriguing phase structure 
102: with an Aoki phase at strong coupling \cite{Ilgenfritz:2003gw,Sternbeck:2003gy} 
103: and pronounced first order phase transitions at intermediate 
104: values of the coupling which 
105: weakens when the continuum limit is approached 
106: \cite{Farchioni:2004us,Farchioni:2004ma,Farchioni:2004fs,Farchioni:2005tu}.  
107: These findings are supported by calculations in chiral perturbation 
108: theory for standard Wilson fermions \cite{Sharpe:1998xm} which were 
109: later extended to the case of Wilson twisted mass fermions
110: \cite{Munster:2004am,Scorzato:2004da,Sharpe:2004ny,Sharpe:2004ps,Aoki:2004ta,Sharpe:2005rq}.
111: See refs.~\cite{Frezzotti:2004pc,Shindler} for reviews.
112: The appearance of the first order phase transition turns out to be a 
113: generic phenomenon of Wilson fermions and takes place independent of 
114: whether the twisted 
115: mass parameter is switched on or not. 
116: 
117: Of course, one may argue that on a finite lattice no phase transition 
118: can occur and that the metastabilities observed in 
119: refs.~\cite{Farchioni:2004us,Farchioni:2004ma,Farchioni:2004fs,Farchioni:2005tu} 
120: are purely an artefact of the algorithm employed. 
121: From the analytical side, 
122: chiral perturbation theory ($\chi$PT) is a powerful tool
123: to determine the phase structure. 
124: However, the assumption of $\chi$PT is that $a \Lambda \ll 1$,
125: where $a$ is the lattice spacing and $\Lambda$ is the scale parameter.
126: Moreover, the analysis by $\chi$PT is done up to $O(a^2)$ only.
127: Thus, it is unclear whether at large values of $a$ and by inclusion 
128: of higher order terms the picture can change substantially.
129: 
130: Motivated by these considerations and 
131: in order to have another angle on the problem, we decided
132: to study the phase structure of the Gross-Neveu model \cite{Gross:1974jv} 
133: on the lattice including a twisted mass term, 
134: taking this 
135: as a toy model of twisted mass lattice QCD.
136: In the large-$N$ limit, where $N$ is the number of flavours,
137: the Gross-Neveu (GN) model is asymptotically free and has bound states 
138: \cite{Gross:1974jv}.  
139: The GN-model is one-loop exact in the large-$N$ limit
140: and hence, in this limit, higher order contributions of the lattice spacing 
141: are taken into account.
142: In this sense, the large-$N$ investigation of the 
143: GN-model goes beyond $\chi$PT and can provide therefore a complementary 
144: picture of the phase diagram, although one should always keep in mind
145: that we are considering a two-dimensional model only.
146: 
147: The GN-model has been studied 
148: for pure Wilson fermions 
149: in the large-$N$ limit 
150: in refs.~\cite{Izubuchi:1998hy,Aoki:1985jj}.
151: In ref.~\cite{Aoki:1983qi}, the phase structure of the lattice 
152: GN-model with Wilson fermions  
153: is first predicted and a 
154: so-called {\it Parity-Flavour broken phase} (or {\it Aoki phase}) was 
155: found. A refined analysis \cite{Izubuchi:1998hy} showed that even the 
156: zero temperature phase diagram is rather complex and reveals 
157: besides the second order phase transition to the Aoki phase 
158: strong first order phase transitions. 
159: In this work we want to extend the investigations of 
160: refs.~\cite{Aoki:1983qi,Aoki:1985jj,Izubuchi:1998hy} to the case when 
161: a twisted mass parameter $\mu$ is switched on. In particular,
162: we want to address the question
163: of the $\mu$-dependence
164: of the first order phase transition.
165: We hope that our findings can give some, at least qualitative insight also 
166: into the generic phase diagram of lattice QCD with Wilson fermions.
167: 
168: 
169: 
170: 
171: %%%%%%%%%%%%%%%%%%%%%%%%%%%%% SECTION %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
172: \section{Lattice Gross--Neveu model with twisted mass fermion}
173: \label{sec:lat}
174: 
175: %%%%%%%%%%%%%%%%%%%%%%%%%%%%% SUBSECTION %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
176: 
177: A formulation of the two-dimensional Gross--Neveu model on the lattice 
178: with Wilson fermions
179: is given in refs.~\cite{Aoki:1983qi,Aoki:1985jj,Izubuchi:1998hy}.
180: Our study here follows  
181: refs.~\cite{Aoki:1985jj,Izubuchi:1998hy} where the 
182: the lattice couplings of the model are treated separately.
183: We choose the so-called  twisted basis
184: \cite{Frezzotti:2003ni} for our investigation in which the Lagrangian 
185: of the model reads
186: \beq
187: {\cal L}_{GN}=\bchi \left[\gamma \cdot \nabla^{sym} -
188: r \nabla_\mu \nabla_\mu^\dagger + M_{cr}(r) 
189: + m_q e^{i \gamma_5 \tau_3 \omega} \right] \chi
190: -\frac{g_\sigma^2}{2 N}  (\bchi \chi)^2 
191: -\frac{g_\pi^2}{2 N}  (\bchi i \gamma_5 \tau_3\chi)^2 ,
192: \label{eq:lat-gn-chi}
193: \eeq
194: where $(\bchi,\chi)$ denote the fermion fields in the ``twisted basis''
195: or $\chi$-basis,
196: and $N$ is the number of the ``flavours''. 
197: The parameters of the model are the critical fermion mass 
198: $M_{cr}$, the off-set fermion mass $m_q$, the twist angle 
199: $\omega$ and the couplings $g_\sigma$ and $g_\pi$.
200: In this version of twisted mass fermions, the Wilson term is not twisted.
201: An important remark is that in general the coupling constant 
202: $g_\sigma \ne g_\pi$ 
203: on the lattice\cite{Aoki:1985jj,Izubuchi:1998hy}.
204: 
205: In momentum space, the Lagrangian in the $\chi$-basis is given as
206: \beq
207: {\cal L}_{GN}=\bchi \left[K(p) + W(p) + m_0 
208: + i \gamma_5 \tau_3 \mu  \right] \chi
209: -\frac{g_\sigma^2}{2 N}  (\bchi \chi)^2 
210: -\frac{g_\pi^2}{2 N}  (\bchi i \gamma_5 \tau_3  \chi)^2 ,
211: \label{eq:gn-chi-mom}
212: \eeq
213: where $K(p)=i \sum_\nu \gamma_\nu \sin (p_\nu a)$,
214: $W(p)=r \sum_\nu (1-\cos(p_\nu a))$,
215: $m_0 = M_{cr} + m_q \cos \omega$,
216: $\mu = m_q \sin \omega$
217: and
218: $r$ the Wilson parameter.
219: The situation of ``full twist'' is realized when $\omega=\pm \pi/2$ 
220: or, equivalently, $m_0=M_{cr}$. 
221: 
222: In the  
223: large-$N$ limit we obtain the Lagrangian
224: \be
225: {\cal L}_{GN}= 
226: \bchi \left[K(p) + W(p) +
227: \sigma_L
228: + i \gamma_5 \tau_3 \pi_L  \right] \chi 
229: +\frac{N}{2 g_\sigma^2}  (\sigma_L- m_0)^2
230: +\frac{N}{2 g_\pi^2}  (\pi_L - \mu)^2 ,
231: \label{eq:gn-chi-1}
232: \eeq
233: where, by the equations of motion, the bosonic fields are defined as
234: \beq
235: \sigma_L = - \frac{g_\sigma^2}{N} ( \bchi \chi) + m_0 \quad, \quad 
236: \pi_L =-\frac{g_\pi^2}{N} (\bchi i \gamma_5 \tau_3  \chi )+ \mu \quad.
237: \eeq
238: 
239: To solve the model, we integrate out the fermion fields, treating 
240: the $\pi$- and $\sigma$-fields as constant since 
241: the quantum fluctuations are suppressed in the large-$N$ limit.
242: We then obtain the effective potential
243: $V_{eff}$,
244: \beq
245: Z=\int [d \chi][d \bchi] \exp \{- \int d^2 x {\cal L}_{GN}\} 
246: \equiv \exp \{ -N \Omega V_{eff} \} \, ,
247: \eeq
248: where $\Omega$ is the space-time volume.
249: 
250: \noindent $V_{eff}$ in large-$N$ limit is given by       
251: (setting here and in the following $r=1$),
252: \beqn
253: \hspace{-1cm}
254: &&V_{eff}=\frac{1}{2 g_\sigma^2}(\sigma_L - m_0)^2
255: +\frac{1}{2 g_\pi^2}(\pi_L - \mu)^2 
256: \label{eq:eff-pot-chi} \\
257: \hspace{-1cm}
258: && - \int_p \ln [ \sum_\nu \sin^2 (p_\nu a) 
259: +\{\sum_\nu (1 -\cos (p_\nu a)) \}^2 \nonumber 
260:  +(\sigma_L^2+\pi_L^2) 
261: + 2 \sigma_L
262:  \sum_\nu (1 -\cos (p_\nu a))] ,
263: \nonumber 
264: \eeqn
265: where $\int_p=\int_{-\pi}^\pi \frac{d^2(p a)}{(2 \pi)^2}$.
266: Note that eq.~(\ref{eq:eff-pot-chi}) has the same form
267: as the effective potential  
268: in the Wilson fermion formulation of 
269: the GN-model on the 
270: lattice \cite{Aoki:1985jj,Izubuchi:1998hy}.
271: The difference shows up only in the definition of the 
272: field $\pi_L$, where there is an explicit appearance 
273: of the twisted mass parameter $\mu$.  
274: 
275: %%%%%%%%%%%%%%%%%%%%%%%%%%%%% SECTION %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
276: \subsection{Phase structure and Pion mass in twisted basis}
277: \label{sec:phase}
278: 
279: At $\mu=0$, i.e. for pure Wilson fermions, the phase structure 
280: of the Gross-Neveu model is 
281: already quite complicated showing
282: first order and 
283: second order phase transitions. 
284: It is unclear, whether the presence of the twisted mass term 
285: can change the phase structure.
286: In order to explore the phase structure of the lattice GN-model, 
287: it is sufficient to solve 
288: the saddle point equations as they can be derived
289: from the effective potential $V_{eff}$.
290: 
291: In the following we will
292: concentrate on the twisted basis,
293: because then the effective potential 
294: has the same form as the 
295: one for
296: ordinary Wilson fermion.
297: The effect of the twisted mass parameter is, to 1-loop order, 
298: only included 
299: via the field $\pi_L$ by a shift. 
300: 
301: The saddle point equations are 
302: \beqn
303: &&\frac{\partial V_{eff}}{\partial \sigma_L}
304: = \frac{\sigma_L - m_0}{g_\sigma^2} - \sigma_L F(\sigma_L,\pi_L)
305:  - G(\sigma_L,\pi_L) = 0, 
306: \label{eq:saddle-s} \\
307: &&\frac{\partial V_{eff}}{\partial \pi_L }
308: = \frac{\pi_L - \mu}{g_\pi^2} - \pi_L F(\sigma_L,\pi_L)=0,
309: \label{eq:saddle-p} 
310: \eeqn
311: where
312: \beqn
313: F(\sigma_L,\pi_L) &=& \int_p \frac{2}{A(\sigma_L,\pi_L)} , \quad
314: G(\sigma_L,\pi_L) = \int_p \frac{2 \sum_\nu (1-\cos (p_\nu a))}{A(\sigma_L,\pi_L)} , \nonumber \\
315:  A(\sigma_L,\pi_L) &=& (\sigma_L^2+\pi_L^2) 
316: + \{\sum_\nu (1-\cos (p_\nu a)) \}^2 + \sum_\nu \sin^2 (p_\nu a) \nonumber \\
317: & & + 2 \sigma_L \sum_\nu (1-\cos (p_\nu a)) \, . \nonumber
318: \label{eq:func}
319: \eeqn
320: 
321: In the saddle point equations 
322: the 1-loop tadpole contributions are canceled 
323: by the potential terms of $\sigma_L$ and $\pi_L$.
324: The solution of the saddle point equations
325: is non-trivial and will differ from the one obtained with 
326: pure Wilson fermions opening thus the possibility of new
327: phenomena.
328: 
329: The neutral pion mass and the sigma particle mass are computed 
330: from the second derivative of the effective potential at the 
331: solution of the saddle point equations,
332: \beqn
333: \hspace{-1cm}
334: &&M_\pi^2 = \frac{1}{2} \frac{\partial^2 V_{eff}}{\partial \pi_L^2}
335: = \frac{1}{2} \left[ \frac{1}{g_\pi^2} - F(\sigma_L,\pi_L)
336: - \pi_L \frac{\partial}{\partial \pi_L} F(\sigma_L,\pi_L)  \right] \, ,
337: \label{eq:pionmass1}  \\
338: \hspace{-1cm}
339: &&M_\sigma^2 = \frac{1}{2} \frac{\partial^2 V_{eff}}{\partial \sigma_L^2}
340: = \frac{1}{2} \left[ \frac{1}{g_\sigma^2} - F(\sigma_L,\pi_L)
341: - \sigma_L \frac{\partial}{\partial \sigma_L} F(\sigma_L,\pi_L)
342: - \frac{\partial}{\partial \sigma_L} G(\sigma_L,\pi_L) \right] .
343: \label{eq:sigmamass1} 
344: \eeqn
345: 
346: 
347: Our strategy is the following. We first fix the values 
348: the couplings $g_\sigma^2$ and $g_\pi^2$ such  
349: that the perturbative tuning 
350: relation between 
351: $g_\pi^2$ 
352: and  $g_\sigma^2$ as computed for the ordinary Wilson fermion formulation 
353: \cite{Izubuchi:1998hy} holds, 
354: \beq
355: \frac{1}{g_\pi^2} = \frac{1}{g_\sigma^2} + 4 C_2 + O(a) \, , 
356: \quad \quad C_2=\frac{2 \sqrt{3}}{27}+\frac{1}{12 \pi} \, .
357: %\frac{m_0}{g_\sigma^2} = -2 r C_1 + O(a^2)\; ,\;\; C_2=  \, .  \\
358: \label{tuning}
359: \eeq
360: In the following table,
361: the (perturbative) tuning values for various $g_\sigma^2$ are given.
362: 
363: \begin{center}
364: \begin{tabular}[t]{|c|c|c|c|c|}
365: \hline
366: $g_\sigma^2$ & 0.1   & 0.5   & 1.0  & 2.0 \\
367: \hline
368: $g_\pi^2$    & 0.094 & 0.381 & 0.617& 0.893 \\
369: \hline 
370: % 3.0  &  1.049  \\
371: %\hline
372: \end{tabular}
373: \end{center}
374: 
375: \vspace{0.5cm}
376: 
377: We then also fix the value of 
378: the twisted mass $\mu$ and 
379: determine 
380: the values of $\sigma_L$ and $\pi_L$  
381: from the solutions of the saddle point equations 
382: eqs.~(\ref{eq:saddle-s}, \ref{eq:saddle-p}) as a function of $m_0$.
383: In this way, we can study whether $\sigma_L$ or $\pi_L$ undergo    
384: a phase transition. Once we have the value for $\sigma_L$ and $\pi_L$ 
385: we can insert them into the gap equations, 
386: eqs.~(\ref{eq:pionmass1}, \ref{eq:sigmamass1}) to compute also the 
387: corresponding masses.
388: 
389: Let us remark that
390: in principle there is also an additional        
391: tuning relation between the coupling 
392: $g_\sigma^2$ and the fermion mass $m_0$. To explore the phase structure, 
393: we will, however, vary $m_0$ and hence will not make use of 
394: this tuning relation.
395: We will also restrict ourselves to values of $g_\sigma^2 \le 2$, since for
396: larger values of the coupling the tuning relation of 
397: eq.~(\ref{tuning}) will not hold anymore. 
398: 
399: 
400: %%%%%%%%%%%%%%%%%%%%%%%%%%%%% SUBSECTION %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
401: \section{Results}
402: \label{sec:results}
403: 
404: In this section we will give our results obtained 
405: from the solution of the saddle point 
406: equations. We used a lattice of 
407: size $128^2$ and solved the equations numerically. We always checked that
408: we indeed found the correct solutions by inspecting the extrema of 
409: the effective 
410: potential itself. 
411: 
412: \begin{figure}[h!tb] 
413: \centering
414: \vspace*{-0.8cm}
415: \centerline{
416: \resizebox{9.0cm}{!}{\rotatebox{-90}{\includegraphics{plots/fig-pi-gs2.ps}}}
417: }
418: \vspace*{-0.6cm}
419: \centerline{
420: \resizebox{9.0cm}{!}{\rotatebox{-90}{\includegraphics{plots/fig-pi-gs1.ps}}}
421: }
422: \vspace*{-0.6cm}
423: \centerline{
424: \resizebox{9.0cm}{!}{\rotatebox{-90}{\includegraphics{plots/fig-pi-gs0.5.ps}}}
425: }
426: \caption{$\langle \pi_L \rangle$ as a function of 
427: the fermion mass $m_0$. The coupling $g_\pi$ is the perturbatively 
428: tuned one obtained from eq.~(\ref{tuning}) and 
429: assumes values of $g_\pi^2= 0.893$, $g_\pi^2= 0.617$ and $g_\pi^2= 0.381$ 
430: from top to bottom. 
431: We represent some values of the twisted mass parameter $\mu$ by open 
432: symbols. Other, intermediate values of $\mu$ are represented
433: by solid lines in the graph.
434: }
435: \label{fig:exppi}
436: \end{figure}
437: 
438: In fig.~\ref{fig:exppi} we show the expectation value 
439: $\langle\pi_L\rangle$ of the $\pi$-field as a function of the mass parameter 
440: $m_0$ for various values of the twisted mass parameter $\mu$. 
441: Let us summarize the results of our large-$N$ computations.
442: At $\mu=0$ and
443: increasing the values of $m_0$, we find 
444: in accordance with ref.~\cite{Izubuchi:1998hy} first   
445: and second order phase transitions, where the 
446: $\pi$-field assumes a non-zero value. For larger values of 
447: $m_0$ there is a first order 
448: transition where $\langle\pi_L\rangle$ jumps back to a zero value. 
449: This structure changes in two aspects when the twisted mass parameter 
450: $\mu$ is switched on. First, the second order phase transition vanishes 
451: and $\langle \pi_L \rangle$ shows an analytical behaviour. 
452: Second, for some range 
453: of $\mu < \mu_\mathrm{crit}$ the first order phase transition remains. 
454: However, it becomes weaker when $\mu$ is increased and turns into a second
455: order phase transition when a critical value $\mu_\mathrm{crit}$ is reached. 
456: For $\mu > \mu_\mathrm{crit}$ the phase transition disappears altogether
457: and turns into a completely smooth and analytical behaviour. 
458: %Note also the value of $m_0$ at which the first order phase transition happens
459: %becomes larger when $\mu$ is increased.
460: 
461: When the coupling $g_\sigma^2$ is made smaller, with the associated
462: tuning of $g_\pi^2$, the region in $m_0$ where the 
463: phase transitions occur shrinks and also the jumps at the first order 
464: phase transitions become smaller.  
465: Thus the qualitative picture does not change when we approach the 
466: continuum limit. 
467: Note that also the value of $\mu_\mathrm{crit}$ gets smaller with 
468: decreasing coupling $g_\sigma^2$.
469: 
470: The fate of the first order phase transition 
471: is very much reminiscent of the situation of four-dimensional lattice QCD.
472: Also there it has been found from numerical simulations 
473: \cite{Farchioni:2004fs,Farchioni:2004ma,Farchioni:2004us}
474: and from calculations in chiral perturbation theory
475: \cite{Sharpe:1998xm,Sharpe:2004ny,Munster:2004am,Scorzato:2004da,Sharpe:2005rq}
476: that the first order phase transition terminates at a critical 
477: values of $\mu$. 
478: Note that the values of $m_0$ where the first order phase transition takes 
479: place moves to larger values when $\mu$ is increased. 
480: A similar phenomenon can be expected also in lattice QCD.
481: What is very interesting from our analysis in the GN-model 
482: --and apparently different from current findings
483: of lattice QCD and $O(a^2)$-$\chi$PT-- is that 
484: this first order phase transition is accompanied by a second 
485: order phase transition at $\mu=0$ or its remnant at $\mu>0$. 
486: 
487: \begin{figure}[h!tb] 
488: \centering
489: \centerline{
490: \resizebox{9.5cm}{!}{\rotatebox{-90}{\includegraphics{plots/fig-mpi-gs2.ps}}}
491: }
492: \centerline{
493: \resizebox{9.5cm}{!}{\rotatebox{-90}{\includegraphics{plots/fig-sg-gs2.ps}}}
494: }
495: \vspace*{-0.5cm}
496: \centerline{
497: \resizebox{9.5cm}{!}{\rotatebox{-90}{\includegraphics{plots/fig-msg-gs2.ps}}}
498: }
499: \caption{
500: We show (from top to bottom) 
501: the pion mass, the $\sigma$-field expectation value and 
502: the $\sigma$-mass as a function 
503: of $m_0$ at $g_\sigma^2=2$, $g_\pi^2=0.893$. 
504: }
505: \label{fig:masspi}
506: \end{figure}
507: 
508: In fig.~\ref{fig:masspi} we show at the example of the coupling values
509: $g_\sigma^2=2$, $g_\pi^2=0.893$ the behaviour of the pion mass (top), the 
510: $\sigma$-field expectation value (middle) and the $\sigma$-mass (bottom). 
511: At $\mu=0$ and 
512: increasing $m_0$, the pion mass first goes to zero 
513: and vanishes at the second order phase transition. 
514: It then assumes non-zero values again until it jumps when the first order 
515: phase transition is hit. Switching on $\mu$, the second order phase transition 
516: vanishes and the behaviour of the pion mass
517: becomes continuous. The first order phase transition remains, however, 
518: until $\mu=\mu_\mathrm{crit}$. For larger values of $\mu$ 
519: also the first order phase transition vanishes and the pion mass shows
520: an overall analytical behaviour. 
521: 
522: The $\sigma$-field expectation value and mass do not show a 
523: large sensitivity on the second order phase transition. However, this 
524: field feels the first order phase transition strongly and shows jumps 
525: in the field expectation value as well as in the mass. The picture shown 
526: here for the couplings $g_\sigma^2=2$, $g_\pi^2=0.893$ stays qualitatively 
527: unchanged when the coupling $g_\sigma^2$ is reduced with the associated
528: tuning of $g_\pi^2$. 
529: 
530: \begin{figure}[h!tb]
531: \centering
532: \centerline{
533: \resizebox{9.5cm}{!}{\rotatebox{-90}{\includegraphics{plots/fig-pi-gs2-gp0.8.ps}}}
534: }
535: \vspace*{-0.5cm}
536: \centerline{
537: \resizebox{9.5cm}{!}{\rotatebox{-90}{\includegraphics{plots/fig-sg-gs2-gp0.8.ps}}}
538: }
539: \caption{
540: The $\pi$- (top) and $\sigma$-field (bottom) expectation values 
541: as a function of $m_0$ at the so-called $\sigma$-phase transition
542: \cite{Izubuchi:1998hy}.
543: }
544: \label{fig:sigmapt}
545: \end{figure}
546: 
547: Finally and for completeness, we show in fig.~\ref{fig:sigmapt} the fate 
548: of the so-called $\sigma$-phase transition \cite{Izubuchi:1998hy}
549: for a value of the coupling $g_\sigma^2=2$.
550: Note that the value of the coupling is now $g_\pi^2=0.8$ and does 
551: not correspond to the ones of the perturbative tuning.
552: This phase transition is characterized by the fact that at $\mu=0$ the 
553: $\sigma$-field expectation value $\langle \sigma_L \rangle $ shows a discontinuous jump, 
554: whereas the 
555: $\pi$-field expectation value stays essentially zero. 
556: This phase transition 
557: survives when $\mu$ is switched on and  
558: can clearly be seen in the jump of $\langle \sigma_L \rangle$. 
559: It is interesting to observe that the point in $m_0$ where this jumps
560: happens is moved to larger values of $m_0$ when $\mu$ is increased. 
561: However, also for this case the phase transition vanishes when 
562: $\mu$ is increased above a certain critical value.  
563: Furthermore, the $\pi$-field shows now at $\mu>0$ also a jump and 
564: exhibits the first order phase transition. 
565: 
566: 
567: 
568: %%%%%%%%%%%%%%%%%%%%%%%%%%%%% SECTION %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
569: \section{Summary}
570: \label{sec:summary}
571: 
572: Following the earlier work of 
573: refs.~\cite{Aoki:1983qi,Aoki:1985jj,Izubuchi:1998hy}  
574: we studied in this letter, how the presence of the twisted mass 
575: term influences the known phase diagram of the 
576: two-dimensional  Gross--Neveu model which we consider as a 
577: toy model of lattice QCD. 
578: We found that the coexistence of the Aoki phase and the 
579: first order phase transition as it is obtained 
580: at vanishing twisted mass parameter $\mu$ \cite{Izubuchi:1998hy} 
581: is {\em not} kept 
582: at $\mu>0$. 
583: The first change is that at $\mu>0$ the second order phase transition 
584: turns into a smooth and analytical behaviour, although e.g. the 
585: $\pi$-field expectation $\langle \pi_L \rangle$ still shows a rapid change. 
586: The second change is that the gap in $\langle \pi_L \rangle$ of 
587: the first order phase transition shrinks with increasing values of $\mu$ 
588: and finally vanishes at a critical value of $\mu=\mu_\mathrm{crit}$.  
589: In addition, the values of $m_0$, where the first order phase transition 
590: takes place, is shifted to larger values when $\mu$ is increased. 
591: 
592: The motivation of our investigation has been to provide a complementary
593: analysis of a ``QCD-like'' model to the findings of numerical simulations
594: \cite{Farchioni:2004us,Farchioni:2004ma,Farchioni:2004fs,Farchioni:2005tu} 
595: and from (Wilson) chiral perturbation theory
596: \cite{Sharpe:1998xm,Munster:2004am,Scorzato:2004da,Sharpe:2004ny,Sharpe:2004ps,Aoki:2004ta,Sharpe:2005rq}.
597: Both of these approaches have some shortcomings: in the numerical simulations
598: it is not clear whether the metastability effects are due to a bad 
599: behaviour of the algorithms employed. 
600: $\chi$PT on the other hand can, in principle, only predict the 
601: phase structure reliably 
602: in a region close to the continuum limit where the
603: $O(a^2)$ expansion is valid and the quark masses are small. 
604: 
605: Our findings from the large-$N$ analysis of the GN-model 
606: as described above show a striking similarity to the results from the 
607: four-dimensional simulations and from $\chi$PT. 
608: This includes the shift in the critical mass, the weakening of the 
609: first order phase transition as function of increasing $\mu$ 
610: and the vanishing of the first order phase transition altogether when 
611: $\mu>\mu_\mathrm{crit}$.
612: However, 
613: there is also a very noticeable and important difference to the picture we have
614: presently in four-dimensions: this concerns the coexistence of the first 
615: order phase transition with a second order transition to the Aoki phase. 
616: It is unclear to us whether this difference is an artefact of the 
617: GN-model in the large-$N$ analysis. Another, very interesting possibility is
618: of course that so far both, the numerical simulations and the analysis 
619: in $\chi$PT have missed this possibility. 
620: Clearly, if the Aoki phase transition is very close to the first order 
621: phase transition it will be very difficult for the numerical simulations
622: to resolve both. It could be speculated that the coexistence of both phases 
623: will show up in $\chi$PT only when higher orders are taken 
624: into account\footnote{Indeed, in some recent calculations in $\chi$PT for 
625: four-dimensional lattice-QCD, taking higher order effects 
626: into account such a coexistence of the Aoki phase and the first order 
627: phase transition --as we found here-- seems to take 
628: place \cite{Aokiprivate}.}. 
629: From our analysis here we would just like to give a warning that maybe the
630: phase structure of Wilson lattice QCD is even more complicated than 
631: our present picture suggests. 
632: If the picture of the phase diagram from our analysis in the GN-model 
633: would be correct also for lattice QCD, 
634: it would provide an argument for using 
635: a non-vanishing twisted mass parameter, since then one would avoid the
636: complication of the existence 
637: of two phase transitions.
638: 
639: 
640: We also want to mention that  
641: the $\sigma$-phase transition, 
642: which is present in the GN-model with Wilson fermion, 
643: changes to a normal first order phase transition,
644: when $\mu$ is switched on.
645: 
646: In this letter we did not address a number of additional interesting 
647: points that would deserve a further investigation:
648: in order to determine the phase transition, 
649: in particular in order to study the second order phase transition,
650: the second moment, for instance the susceptibility, may be better.
651: %In the physical basis with {\it normal} 4-fermi interactions,
652: %the cross term of $\sigma_L$ and $\pi_L$ appears.
653: %Thus the choice of the interaction in the Gross-Neveu model
654: %is not equivalent for both, the twisted and the physical bases and it 
655: %would be nice to repeat our calculations in the physical basis, too. 
656: In principle one could also study the scaling behaviour of physical
657: quantities analytically toward the continuum limit.
658: In addition, an analysis within perturbation theory could provide
659: important insight into the renormalization properties of the GN-model 
660: with the twisted mass term.
661: 
662: 
663: Finally, it would be straightforward to change the discretization of the
664: fermions in the GN-model. For example, one could study a chiral invariant 
665: GN-model. In this way, it would be possible to learn how the change
666: of the fermion part of the action changes the phase structure. This is clearly
667: important for four-dimensional lattice QCD simulations.
668: 
669: 
670: \section*{Acknowledgments}
671: We thank Taku Izubuchi, Andrea Shindler and Giancarlo Rossi for 
672: very useful discussions and helpful communications.
673: 
674: 
675: 
676: \bibliographystyle{h-physrev4}
677: \bibliography{gn}
678: 
679: \end{document}
680: