1:
2: \section{The nature of the QCD phase transition \label{sec:pd}}
3:
4: %\subsection{Phase transitions and phase diagrams}
5: %\label{sec:pdgen}
6:
7: In the previous sections we discussed the properties of QCD matter on either side
8: of the quark hadron
9: transition, but haven't yet addressed
10: what the nature of that transition is. To obtain an unambiguous answer to
11: this question is in fact a
12: very difficult task. In statistical mechanics, phase transitions are defined
13: as singularities, or non-analyticities, in the free energy as a function of its thermodynamic parameters.
14: However, on finite volumes, free energies are always analytic functions and it can be proved that singularities only develop in the thermodynamic limit of infinitely many particles, or $V\rightarrow \infty$ \cite{ly}.
15: This is particularly obvious in the case of lattice QCD, whose partition
16: function is a functional integral
17: over a compact group with a bounded exponential as an integrand.
18: It is thus a perfectly analytic function of
19: $T,\mu,V$ for any finite $V$.
20: Hence, a theoretical establishment of a true phase transition requires
21: finite size scaling (FSS) studies on a series of increasing and
22: sufficiently large volumes in order to extrapolate to the thermodynamic limit.
23: Three different situations can emerge: a first order phase transition
24: is characterized by coexistence of two phases, and hence a discontinuous jump
25: of the order parameter (and other quantities)
26: when going through such a transition. A second order transition
27: shows a continuous transition of the order parameter accompanied by a divergence of the correlation length
28: and some other quantities, like the heat capacity.
29: Finally, a marked change in the physical properties of a system may also occur
30: without any non-analyticity of the free energy, in which case it is called
31: an analytic crossover.
32: A familiar system featuring all these possibilities is water,
33: with a weakening first oder liquid-gas phase transition terminating in a
34: critical endpoint with Z(2) universality, as well as a triple point
35: where the first order liquid-gas and
36: solid-liquid transitions meet. Similar structures are also conjectured to
37: be present in the QCD phase diagram \cite{wilc,tric}, fig.~\ref{fig:1schem} (left).
38: In this section we focus on the order of the transition for $\mu=0$
39: as a function of $N_f$ and quark masses, before tackling $\mu\neq 0$
40: in sec.~\ref{sec:mu}.
41:
42: \begin{figure}[t]
43: \vspace*{-5cm}
44: %\includegraphics*[width=0.5\textwidth]{water.ps}
45: \includegraphics*[width=0.5\textwidth]{qm99fig2.eps}\hspace*{1cm}
46: \includegraphics*[width=0.5\textwidth]{phase_diagram_trunc.ps}
47: \caption{
48: %Left: Phase diagram of water, featuring a weakening first oder liquid gas phase transition terminating in a critical endpoint with Z(2) universality, as well as a triple point.
49: Left: Conjectured phase diagram for $N_f=2$ QCD with finite light quark massees. The physical
50: case $N_f=2+1$ is believed to qualitatively look the same,
51: according to universality and continuity arguments as well as results from QCD-like
52: models \cite{wilc}.
53: Right: Schematic phase transition behaviour of $N_f=2+1$ flavor QCD at $\mu=0$ for
54: different choices of quark masses $(m_{u,d},m_s)$, at $\mu=0$.}
55: \label{fig:1schem}
56: \end{figure}
57:
58: \subsection{Universality and finite size scaling}
59:
60: A most fascinating phenomenon in physics is the ``universality'' exhibited by physical systems
61: near a critical point of second order phase transition. It is due to the divergence of the
62: correlation length, which implies that the entire system acts as a coherent collective.
63: Hence, microscopic physics becomes unimportant, the collective behaviour is determined by global symmetries.
64: With the divergence of the correlation length the characteristic length scale
65: disappears from the problem, and thermodynamic observables in the critical
66: region obey scale invariant power laws. For example, at a second order
67: ferromagnetic phase transition, the order parameter magnetization vanishes as
68: $M\sim |T-T_c|^{\beta}$, while the specific heat and correlation length diverge,
69: $C\sim |T-T_c|^{-\alpha}$ and $\xi \sim |T-T_c|^{-\nu}$, respectively.
70: The critical exponents $\alpha,\beta,\nu$ and similar ones for other quantities
71: are determined by the effective
72: global symmetries at the critical point, and thus are the same
73: for all systems within a universality class.
74: The latter are labeled by spin models, since those can be
75: readily solved numerically.
76: The power law behaviour for thermodynamic functions
77: follows from the scaling form
78: of the singular part of the free energy,
79: \be
80: f_s(t, h) = L^{-3} f_s(L^{y_t}\tau, L^{y_h}h),
81: \ee
82: where $L$ denotes the box size, and $\tau$ and $h$
83: are the relevant scaling fields. In the case of the Ising model these are the reduced
84: temperature and magnetic field.
85: Unfortunately, for most systems, and particularly QCD, it is not obvious which
86: global symmetries the system will exhibit at a critical point,
87: those are a dynamically
88: determined subset of the total symmetry of the theory.
89: Moreover, since there is no true order parameter in the case
90: of finite quark masses, fields and parameters of QCD map
91: into those of the effective model in a non-trivial way.
92:
93: \subsection{The nature of the QCD phase transition for $N_f=2+1$ at $\mu=0$: qualitative picture}
94:
95: Before discussing numerical results, let us briefly outline the qualitative picture for
96: the order of the phase transition,
97: fig.~\ref{fig:1schem} (right).
98: As mentioned in sec.~\ref{sec:limits}, for $N_f=3$
99: in the limits of zero and infinite quark masses
100: (lower left and upper right corners),
101: order parameters corresponding to the breaking of a symmetry can be defined,
102: and one finds numerically at small and large quark masses
103: that a first-order transition takes place at a finite
104: temperature $T_c$. On the other hand, one observes an analytic crossover at
105: intermediate quark masses. Hence, each corner must be surrounded by a region of
106: first-order transition, bounded by a second-order line. The situation is less clear for the chiral
107: limit of the two flavour theory in the upper left corner.
108: If the transition is second order, then chiral symmetry
109: $SU_L(2)\times SU(2)_R \sim O(4)$ puts it in the universality
110: class of 3d $O(4)$ spin models.
111: In this case there must be a tricritical strange quark mass $m_s^{tric}$,
112: where the second order chiral
113: transition ends and the first order region begins. The exponents at such a
114: tricritical point would
115: correspond to 3d mean field \cite{tric}. On the other hand,
116: a first order scenario for the chiral limit
117: of $N_f=2$ is a logical possibility as well.
118:
119: \subsection{The chiral transition for $N_f=2$: is it O(4) or first order?}
120:
121: \begin{figure}[t]
122: \centerline{
123: \includegraphics*[width=0.5\textwidth]{Cv_max_Run1.eps}\hspace*{1cm}
124: \includegraphics*[width=0.5\textwidth]{Chi_max_Run1.eps}
125: }
126: \caption{Finite volume scaling behaviour of specific heat and chiral susceptibility
127: on $N_t=4$. For O(4) viz.~O(2) behaviour, the data should fall on a horizontal line \cite{dig}.}
128: \label{fig:dig}
129: \end{figure}
130:
131: On the lattice, O(4) will effectively look like O(2) as long as there are
132: discretization effects \cite{o2}.
133: %A lot of numerical work has been done over the years, but no definite conclusion
134: %has been reached.
135: Wilson fermions appear to see O(4) scaling \cite{wil},
136: while staggered actions are inconsistent with both O(4) and O(2) \cite{s2}.
137: (The staggered strong coupling limit, however, does display O(2) scaling
138: \cite{c2}).
139: An attempt to tackle this question by means of a finite size scaling analysis
140: with unprecedented lattice sizes was made in \cite{dig}. The work simulates $L^3\times 4$ lattices with
141: $L=16-32$, using standard staggered fermions with several quark masses,
142: the smallest being $m/T\lsim 0.055$.
143: In a critical region, the specific heat or the chiral susceptibility
144: scale as
145: \ba
146: C_V - C_0 &\simeq & L^{\alpha/\nu} f_c \left(\tau L^{1/\nu}, am\, L^{y_h} \right),
147: \quad \tau=1-T/T_c\nonumber\\
148: \chi& \simeq &L^{\gamma/\nu} f_\chi \left(\tau L^{1/\nu}, am\,L^{y_h} \right).
149: \label{scale}
150: \ea
151: Here the non-singular part of the specific heat $C_0$ has been subtracted.
152: The authors of \cite{dig} thus fix $y_h$ to its critical value, and then choose $L$ and
153: $m$ for a series of simulations such as to keep $(am \,L^{y_h})$ constant, reducing the scaling problem on one variable only.
154: The infinite volume limit in this procedure thus corresponds to the chiral limit and allows to check whether the data are consistent with the predicted scaling behaviour.
155:
156: Fig.~\ref{fig:dig} shows simulation results from \cite{dig}.
157: Scaling as in eq.~(\ref{scale}) would imply the data points to fall on a
158: horizontal line, which is clearly not the case. After many variations on
159: the analysis, the authors conclude that their data are closer to first order
160: behaviour than to O(4)/O(2).
161:
162: A different conclusion was reached in \cite{ksnf2}, in which $\chi$QCD was investigated numerically.
163: This is a theory modified by an irrelevant term ({\it i.e.}~one going to zero in the continuum limit) such as
164: to allow simulations in the chiral limit. The authors compare their data with those obtained from an
165: $O(2)$ spin model on moderate to small volumes and find them to be compatible.
166: This might point at finite volume effects of current $N_f=2$ QCD simulations.
167: The scaling region of the chiral
168: limit might be excessively small, which would require extraordinarily large volumes in order to see the correct scaling.
169: Another possibility for the failure of simulations
170: to unambiguously pin down the order of this transition is of course the
171: failure of the fermion discretizations to reproduce chiral symmetry.
172: Future studies on either finer lattices or with chiral fermion actions should help to
173: clarify this issue.
174:
175: Another possibility is to study the strength of the $U_A(1)$ anomaly
176: discussed in sec.~\ref{sec:limits}, which also enters the effective
177: sigma model \cite{chiral}
178: for the critical region.
179: It has recently been non-perturbatively demonstrated that a strong anomaly is
180: required for the chiral phase transition to be second order \cite{chi}.
181:
182: \subsection{$N_f=2+1$: 3d Ising universality and the critical line $m_s^c(m_{u,d})$}
183: \label{sec:m1m2c}
184:
185: \begin{figure}[t]
186: \centerline{
187: \includegraphics*[width=0.5\textwidth]{nf3_L.eps}
188: \includegraphics*[width=0.5\textwidth]{m1m2c_mu0.0.eps}
189: }
190: \caption{Left: The Binder cumulant as a function or quark mass for $N_f=3, N_t=4$ and different
191: lattice sizes \cite{fp3}.
192: Right: The chiral critical line in the bare quark mass plane
193: at $\mu=0$, $N_t=4$.
194: $N_f=3$ is indicated by the solid line.
195: Also shown are the physical point according to
196: \cite{fk2}, and an extrapolation using mean field exponents to the chiral limit,
197: giving $m_s^{tric}\sim 2.8 T$.
198: The point marked by the arrow has $m_\pi/m_\rho=0.148(2)$, compared to
199: the physical value $0.18$. From \cite{fp3}.
200: }
201: \label{fig:B4}
202: \end{figure}
203:
204: Next we move our attention to the second order boundary
205: line separating the first order region from the crossover in
206: fig.~\ref{fig:1schem} (right), starting with degenerate quark masses, $N_f=3$.
207: In this case on $N_t=4$ lattices
208: the critical quark mass marking the boundary
209: between the crossover and the first order region is large enough for simulations to be carried out, and
210: it was possible to determine it quite accurately \cite{kls,clm,fp2,fp3},
211: together with the Z(2) universality class of the 3d Ising model
212: to which it belongs \cite{kls}. The most practical observable to determine the
213: latter is the Binder cumulant, constructed from moments of fluctuations of the order parameter,
214: \be
215: B_4(m,\mu)=\frac{\langle(\delta\bar{\psi}\psi)^4\rangle}
216: {\langle(\delta\bar{\psi}\psi)^2\rangle^2},\qquad
217: \delta \bar{\psi}\psi=\bar{\psi}\psi - \langle \bar{\psi}\psi\rangle,
218: \ee
219: evaluated on the phase boundary $\beta=\beta_c(m,\mu)$.
220: In the infinite volume limit, $B_4\rightarrow 1$ or $3$ for a first order transition or crossover, respectively, whereas
221: it approaches a value characteristic of the universality class at a critical point. For 3d Ising
222: universality one has $B_4\rightarrow 1.604$ \cite{ising}.
223: Hence $B_4$ is a non-analytic step function, which gets smoothed out to an
224: analytic curve on finite volumes,
225: with a slope increasing with volume to gradually approach the step function.
226:
227: A numerical example is shown in fig.~\ref{fig:B4} (left).
228: The slope
229: increases with lattice size, and the data fit the
230: finite volume scaling
231: formula predicted by universality,
232: \be
233: B_4(m,L)=b_0+b L^{1/\nu}(am-am^c).
234: \ee
235: One observes that all curves intersect at the critical $B_4$-value, and moreover $\nu$ is consistent with its 3d Ising value $\nu=0.63$. Hence we can read off the critical quark mass in lattice units,
236: $am^c=0.0263(3)$ or $m^c/T_c=1.052(1)$ \cite{fp3}.
237:
238: While the nature of the transition and its universality class are determined
239: by long range fluctuations and thus should be insensitive to the cut-off effects
240: on a coarse lattice, the critical quark mass is a
241: quantity requiring renormalization and is tremendously sensitive to such effects.
242: Indeed, the critical
243: quark mass and the corresponding pion mass have been computed on $N_t=4$ lattices
244: with a standard staggered and a p4-improved staggered action. The critical pion
245: mass in the improved
246: case is almost a factor of four smaller than the non-improved
247: result \cite{kls,milc}. Hence, the location
248: of the boundary line in the phase diagram is still far from the continuum result.
249:
250: Starting from the critical quark mass for $N_f=3$, the boundary line
251: has also been mapped for the non-degenerate $N_f=2+1$ theory,
252: again with $N_t=4$ \cite{fp3}.
253: The question now arises which flavours
254: to couple to the quark chemical potential. Within QCD alone there are no
255: flavour changing interactions,
256: while the initial state in heavy ion collisions has net $u,d$-quark
257: numbers only. Thus it is sensible
258: to assign a chemical potential to the two degenerate light quarks only.
259: In nature, where the weak interactions are switched on as well, the situation
260: is somewhat more complicated than this \cite{exp}.
261: The methodology then is the same as in the three flavour case:
262: fix a strange quark mass $am_s$ and
263: scan the Binder cumulant in the light quark mass $am_{u,d}$ for the
264: corresponding critical point.
265: Repeating for several strange quark masses produces the critical
266: line $am_s^c(am_{u,d})$ shown
267: in fig.~\ref{fig:B4} (right). The results are in qualitative agreement with
268: expectations from fig.~\ref{fig:1schem} (right).
269: They are also consistent with the possible existence of a
270: tricritical point
271: $(m_{u,d}=0,m_s=m_s^{tric})$. Using its known, mean field exponents, the data
272: favour a heavy $m_s^{tric} \sim 2.8 T_c$. Again large corrections,
273: presumably towards a smaller value, are expected in the continuum limit.
274: Also, $T_c$ is found to vary
275: little along the critical line, in accordance with model calculations
276: employing effective chiral lagrangians~\cite{Szep}.
277:
278:
279: \subsection{The nature of the transition at the physical point}
280: \label{sec:phys}
281:
282: \begin{figure}[t]
283: \centerline{
284: \includegraphics*[bb=20 430 580 710,width=9cm]{susc_46.eps}
285: }
286: \caption[]{Finite size scaling of the chiral susceptibility at the physical point for two lattice spacings$N_t=4$ (left) and $N_t=6$ (right). The peaks saturate at a finite height, consistent with an analytic crossover.}
287: \label{fig:fknat}
288: \end{figure}
289:
290:
291: A more immediate issue is whether the QCD physical point indeed lies on
292: the crossover side of the critical line as expected. For that purpose spectrum
293: calculations at $T\sim 0$ have been performed \cite{fp3}
294: at the parameters indicated by the arrows
295: in fig.~ \ref{fig:B4} (right). They show that
296: $m_s$ at the upper arrow is approximately tuned to its physical value
297: ($\frac{m_K}{m_\rho} \sim \frac{m_K}{m_\rho}|_{\rm phys}$=0.65),
298: while the pion is lighter than in physical QCD
299: ($\frac{m_\pi}{m_\rho}=0.148(2) < \frac{m_\pi}{m_\rho}|_{\rm phys}=0.18$).
300: This confirms that the physical point lies to the right of the critical line,
301: {\it i.e.} in the crossover region.
302:
303: An important question is whether this finding, calculated on coarse lattices $N_t=4$,
304: is stable under continuum extrapolations. This has been recently addressed and affirmed
305: in \cite{fknat}, using a Symanzik improved gauge and a stout-link improved staggered fermion action
306: for four different lattice spacings, $N_t=4,6,8,10$. The simulations were carried out along the
307: lines of constant physics, {\it i.e.}~the bare parameters were tuned such that the
308: hadron spectrum in physical units is constant for different lattice spacings.
309: The result of a finite size scaling analysis
310: of the susceptibility of the chiral condensate is shown in fig.~\ref{fig:fknat}. The peak height
311: appears to clearly saturate at some finite value, and the transition thus corresponds to a crossover. The finite peak height was calculated for all four
312: lattice spacings and extrapolates to a finite value also in the continuum limit.
313: Hence, one concludes that
314: the transition between hadronic
315: and quark gluon plasma physics at zero density indeed is an analytic crossover for
316: physical QCD. In view of the concerns about the validity of the staggered fermion
317: actions \cite{sharpe}, it would of course be desirable to reprodce this result with
318: alternative fermion discretizations.
319:
320:
321:
322: