hep-ph0006129/v2.tex
1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: % transverse.tex
3: %
4: % paper by P. Kolb, J. Sollfrank, U. Heinz on ''Anisotropic transverse flow 
5: % and the quark-hadron phase transition''
6: %
7: % first version by PK on 8 March 2000
8: % last changes  by UH on 10 June 2000
9: % submitted to PRC on 13 June 2000
10: % referee report arrived on 3 Aug. 2000
11: % revisions by UH on 9 Aug. 2000 to take care of remarks by referee
12: %
13: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
14: 
15: %documentstyle[aps,prl,epsfig]{revtex}
16: \documentstyle[prc,twocolumn,aps,epsf,eqsecnum,ifthen,epsfig]{revtex}
17: 
18: %\hoffset -0.5cm %prints out symmetrically on A4; remove for US Letter format.
19: %\hoffset 0.5cm 
20: 
21: % Special definitions
22: 
23: \renewcommand{\P}{\partial}
24: \renewcommand{\phi}{\varphi}
25: 
26: \newcommand{\nc}{\newcommand}
27: \nc{\pt}{p_{\rm T}}
28: \nc{\se}{\section}
29: \nc{\suse}{\subsection}
30: \nc{\beq}[1]{\begin{equation}\label{#1}}
31: \nc{\eeq}{\end{equation}}
32: \nc{\bea}[1]{\begin{eqnarray}\label{#1}}
33: \nc{\eea}{\end{eqnarray}}
34: \nc{\bce}{\begin{center}}
35: \nc{\ece}{\end{center}}
36: \nc{\bit}{\begin{itemize}}
37: \nc{\eit}{\end{itemize}}
38: \nc{\bmp}{\begin{minipage}}
39: \nc{\emp}{\end{minipage}}
40: 
41: \nc{\la}{\langle}       
42: \nc{\lla}{\left \langle}
43: \nc{\ra}{\rangle}       
44: \nc{\rra}{\right \rangle}
45: 
46: \newcommand{\lda}{\langle\!\langle}       
47: \newcommand{\llda}{\left\langle\!\left\langle}
48: \newcommand{\rda}{\rangle\!\rangle}       
49: \newcommand{\rrda}{\right\rangle\!\right\rangle}
50: 
51:  
52: \newcommand{\tr}{{\rm tr}}
53: \newcommand{\bp}{{\bbox{p}}}
54: \newcommand{\bb}{{\bbox{b}}}
55: \newcommand{\bv}{{\bbox{v}}}
56: \newcommand{\bq}{{\bbox{q}}}
57: \newcommand{\bk}{{\bbox{k}}}
58: \newcommand{\bK}{{\bbox{K}}}
59: \newcommand{\bx}{{\bbox{x}}}
60: \newcommand{\tp}{{\tilde p}}
61: \newcommand{\tk}{{\tilde k}}
62: 
63: \newcommand{\half}{{\textstyle {1\over 2}}}
64: 
65: %%%%%%%     Some abbreviations for writing references   %%%%%%%%%%%%%%%%%%
66: 
67: \newcommand{\ibid}[3]{{\it ibid.} {\bf #1}, #3 (#2)}
68: \newcommand{\ptp}[3]{Prog. Theor. Phys. {\bf #1}, #3 (#2)}
69: \newcommand{\ijmpa}[3]{Int. J. Mod. Phys. A {\bf #1}, #3 (#2)}
70: \newcommand{\annrev}[3]{Ann. Rev. Nucl. Part. Sci. {\bf #1}, #3 (#2)}
71: \newcommand{\prep}[3]{Phys. Rep. {\bf #1}, #3 (#2)}
72: \newcommand{\prevc}[3]{Phys. Rev. C {\bf #1}, #3 (#2)}
73: \newcommand{\prevd}[3]{Phys. Rev. D {\bf #1}, #3 (#2)}
74: \newcommand{\prevl}[3]{Phys. Rev. Lett.\ {\bf #1}, #3 (#2)}
75: \newcommand{\zpa}[3]{Z. Phys. A {\bf #1}, #3 (#2)}
76: \newcommand{\zpc}[3]{Z. Phys. C {\bf #1}, #3 (#2)}
77: \newcommand{\jpg}[3]{J. Phys. G {\bf #1}, #3 (#2)}
78: \newcommand{\epjc}[3]{Eur. Phys. J. C {\bf #1}, #3 (#2)}
79: \newcommand{\plb}[3]{Phys. Lett. B {\bf #1}, #3 (#2)}
80: \newcommand{\npa}[3]{Nucl. Phys. {\bf A#1}, #3 (#2)}
81: \newcommand{\npb}[3]{Nucl. Phys. {\bf B#1}, #3 (#2)}
82: \newcommand{\aps}[3]{Acta Phys. Slov. {\bf #1}, #3 (#2)}
83: \newcommand{\appb}[3]{Acta Phys. Polon. B {\bf #1}, #3 (#2)}
84: \newcommand{\hip}[3]{Heavy Ion Physics {\bf #1}, #3 (#2)}
85: \newcommand{\yf}[3]{Yad. Fiz. {\bf #1}, #3} 
86: \newcommand{\sjnp}[3]{Sov. J. Nucl. Phys. {\bf #1}, #3 (#2)}
87: \newcommand{\Nat}[3]{Nature {\bf #1}, #3 (#2)}
88: \newcommand{\phm}[3]{Phil. Mag. {\bf #1}, #3 (#2)}
89: \newcommand{\jcomp}[3]{J. Comp. Phys. {\bf #1}, #3 (#2)}
90: 
91: %\renewcommand{\baselinestretch}{1.5} %For draft version
92: 
93: \preprint{\\CERN-TH/2000-161\\hep-ph/0006129}
94: 
95: \title{Anisotropic transverse flow and the quark-hadron phase transition}   
96: \author{Peter F.~Kolb$^{a,b}$ \and Josef Sollfrank$^b$, 
97:         and Ulrich~Heinz$^{a,}$\thanks{On 
98:         leave of absence from Institut f\"ur Theoretische Physik, 
99:         Universit\"at Regensburg. Email: Ulrich.Heinz@cern.ch}}
100: \address{$^a$Theoretical Physics Division, CERN, CH-1211 Geneva 23, 
101:          Switzerland\\
102:          $^b$Institut f\"ur Theoretische Physik, Universit\"at
103:          Regensburg, D-93040 Regensburg, Germany
104:          }
105: \date{\today}
106: 
107: \begin{document}
108: 
109: \maketitle
110: 
111: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
112: % Abstract
113: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
114: 
115: \begin{abstract}
116: 
117: We use (3+1)-dimensional hydrodynamics with exact longitudinal 
118: boost-invariance to study the influence of collision centrality 
119: and initial energy density on the transverse flow pattern and 
120: the angular distributions of particles emitted near midrapidity in 
121: ultrarelativistic heavy-ion collisions. We concentrate on radial flow 
122: and the elliptic flow coefficient $v_2$ as functions of the impact 
123: parameter and of the collision energy. We demonstrate that the finally
124: observed elliptic flow is established earlier in the collision than the
125: observed radial flow and thus probes the equation of state at higher 
126: energy densities. We point out that a phase transition from hadronic 
127: matter to a color-deconfined quark-gluon plasma leads to non-monotonic 
128: behaviour in both beam energy and impact parameter dependences which, 
129: if observed, can be used to identify such a phase transition. Our 
130: calculations span collision energies from the Brookhaven AGS (Alternating 
131: Gradient Synchrotron) to beyond the LHC (Large Hadron Collider); the 
132: QGP phase transition signature is predicted between the lowest available 
133: SPS (CERN Super Proton Synchrotron) and the highest RHIC (Brookhaven 
134: Relativistic Heavy Ion Collider) energies. To optimize the chances for 
135: applicability of hydrodynamics we suggest to study the excitation 
136: function of flow anisotropies in central uranium-uranium collisions
137: in the side-on-side collision geometry.
138: 
139: \end{abstract}
140: 
141: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
142: \se{Introduction}
143: \label{sec1}
144: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
145: 
146: At a given beam energy, the highest energy densities can be reached
147: in central collisions (impact para\-me\-ter $b{\,=\,}0$) between the 
148: largest available nuclei. Hence for many years the exper\-i\-mental 
149: and theoretical attention has focussed on such collisions. Non-central
150: ($b\ne 0$) collisions are, however, interesting in their own right
151: since they exhibit new phenomena which are forbidden by azi\-muthal 
152: symmetry in central collisions between spherical nuclei. For 
153: non-central collisions the directions of the beam axis and the 
154: impact parameter $\bb$ define the collision plane, and many interesting
155: physical phenomena are now non-trivial functions of the azimuthal angle
156: $\phi$ relative to the collision plane. These include in particular the 
157: transverse geo\-me\-try of the collision fireball as measured with two-particle
158: Bose-Einstein correlations (see e.g. \cite{LHW00} and references therein)
159: and momentum-space anisotropies in the transverse plane due to anisotropic
160: transverse flow of the fireball matter \cite{flowrev}.
161: 
162: Aside from changing the collision energy, limited variations of the 
163: energy density of the reaction zone are also possible by varying
164: the collision centrality. Variation of the initial energy density 
165: provides the handle for studying phase transitions in nuclear matter, 
166: in particular the quark-hadron transition at a critical energy density 
167: $e_{\rm c}{\,\lesssim\,}1$\,GeV/fm$^3$ \cite{lattice}. Non-central 
168: collisions between spherical nuclei and/or central collisions between 
169: deformed nuclei provide new opportunities to correlate phenomena related 
170: to azimuthal anisotropies with the initial energy density. This may 
171: yield novel phase transition signatures. In \cite{Rischke95} this idea 
172: was exploited for the so-called directed flow at forward and backward 
173: rapidities: the softening of the equation of state (EOS) in the phase 
174: transition region was predicted to lead to a reduction of the directed 
175: flow, making the phase transition visible as a minimum in its excitation 
176: function. Sorge \cite{S97,S99} suggested analogous features for the 
177: elliptic flow \cite{St82,O92,O98} which were further studied in 
178: \cite{HL99,KSH99,TS99}. The effects of a phase transition on the 
179: excitation function of radial flow in central collisions between 
180: spherical nuclei had been discussed earlier in \cite{SZ78,vH82,KRLG86}.
181: 
182: An important difference between the radial flow observed in azimuthally 
183: symmetric central collisions and the anisotropic directed and elliptic 
184: flows in non-central collisions and/or central collisions between 
185: deformed nuclei was pointed out by Sorge in \cite{S97}: 
186: 
187: 1. {\em Directed flow} affects mostly particles at forward and backward 
188: rapidities which (at energies above a few hundred MeV/nucleon) are 
189: deflected away from the beam direction by the pressure built up between 
190: the colliding nuclei during the time of their mutual overlap. Since the 
191: thus affected particles quickly leave the central region where this 
192: transverse pressure force acts, the finally observed directed transverse 
193: flow pattern is established very early in the collision. Its natural 
194: time scale is given by the transition time of the two colliding nuclei 
195: which decreases with increasing beam energy; this causes a decrease at 
196: high collision energies (after an initial rise at low beam energies) of 
197: the directed flow \cite{flowrev}. This decrease is amplified by a lack of
198: thermalization during the very earliest stages of the collision which
199: prohibits fast enough buildup of transverse pressure and thus eventually 
200: invalidates the applicability of hydrodynamic concepts for calculating
201: the directed flow. Such pre-equilibrium features may even cover up 
202: \cite{S97} the phase transition signal \cite{Rischke95} in the 
203: excitation function of directed flow.
204: 
205: 2. The {\em elliptic flow} is strongest near midrapidity \cite{NA49flow}.
206: Its driving force is the azimuthal anisotropy of the transverse pressure
207: gradient, caused by the geometric deformation of the reaction region
208: in the transverse plane. As pointed out in \cite{S99,KSH99}, elliptic
209: flow acts against its own cause by eliminating the geometric deformation
210: which generates it, thereby shutting itself off after some time. This time 
211: is, at least at high energies, longer than the nuclear transition time.
212: Elliptic flow is thus generated {\em later} than directed flow, and 
213: hydrodynamic concepts for its description may have a larger chance of 
214: being valid, even if the spatial deformation which causes elliptic 
215: flow exists only for a fraction of the total fireball lifetime. An 
216: important focus of this work will be a quantitative determination 
217: of the time scale over which elliptic flow is generated, as a 
218: function of the collision energy. We will see that this time scale 
219: grows with the overall size of the (initially deformed) collision region 
220: \cite{O92,O98}. Studying central collisions between large deformed nuclei 
221: like $^{238}$U \cite{Sh00,Li00} therefore improves the chances that 
222: thermalization happens sufficiently early for a hydrodynamic description 
223: of elliptic flow evolution to be valid. Such collisions are the 
224: preferred proving ground for hydrodynamic predictions for the 
225: excitation function of elliptic flow.
226: 
227: 3. {\em Radial flow} is generated by the pressure gradient between the
228: interior of the collision fireball and the external vacuum; this force
229: persists throughout the fireball expansion until freeze-out. Of all
230: three transverse flow patterns it thus has the strongest weight at
231: late times. Comparing the excitation functions of elliptic and radial
232: flow with their intrinsically different weights for the EOS at early 
233: and late times (i.e. at high and low energy density) may help with 
234: the identification of phase transition signatures and their 
235: discrimination against possible non-equilibrium effects from incomplete 
236: local thermalization. Of course, the final proof for the phase transition
237: to quark matter will require an additional correlation of the here
238: predicted structures in the anisotropic flow pattern with other
239: ``quark-gluon plasma signatures'' (see \cite{BGHG99,HJ00}).
240: 
241: As already indicated we here study the evolution of transverse flow
242: in a macroscopic hydrodynamic framework (to be contrasted with
243: microscopic kinetic approaches \cite{S97,S99,S97a}). This approach, 
244: which is based on the assumption of rapid local thermalization,
245: allows the most direct connection of observables to the EOS of the hot 
246: matter in the collision fireball, including possible
247: phase transitions. Its validity can be tested both experimentally and
248: by comparison with kinetic approaches. We will not do so here (see, 
249: for example, Refs. \cite{HL99,VP00}) but rather concentrate on 
250: qualitative predictions resulting from the hydrodynamic approach. 
251: 
252: Hydrodynamics cannot describe the earliest collision stage of nuclear 
253: energy loss and entropy production by thermalization of the energy 
254: deposited in the reaction zone by the stopping process; this must be 
255: replaced by appropriate initial conditions for the hydrodynamic expansion.
256: The evolution of azimuthally asymmetric reaction zones requires a 
257: (3+1)-dimensional hydrodynamic approach. This is very time consuming 
258: and makes a tuning of initial conditions to data difficult 
259: \cite{Hirano,Netal99}. However, 
260: near midrapidity and especially for high collision energies the 
261: longitudinal expansion dynamics is expected to be given by the 
262: Bjorken scaling solution \cite{Bj83} which can be implemented
263: analytically. The remaining hydrodynamic equations for the transverse
264: dynamics live in 2 space and 1 time dimension and are much easier to 
265: solve \cite{O92,KSH99,TS99}. The hydrodynamic evolution is terminated 
266: by a freeze-out criterium (in our case a fixed decoupling energy 
267: density). At this point the energy and baryon densities are converted 
268: into temperature and chemical potentials for baryon number and 
269: strangeness, using the EOS, and the particle spectra are calculated 
270: using the Cooper-Frye prescription \cite{CF74}. With these spectra 
271: and the hydrodynamic flow pattern on the freeze-out surface the 
272: average radial flow velocity $\lda v_\perp \rda$ and the elliptic 
273: flow coefficient $v_2$ are evaluated.
274: 
275: The present paper gives technical details for our previous two short
276: reports in \cite{KSH99} and significantly extends the results presented 
277: there. The excitation function for $v_2$ is complemented by a similar 
278: one for the average radial flow and calculated up to very much higher 
279: energies. We also compute the impact parameter dependence at fixed 
280: beam energy of the elliptic flow scaled by the initial spatial 
281: anisotropy. The time evolutions of radial and elliptic flow and
282: their dependence on the collision energy are discussed in detail, in 
283: order to establish to which extent elliptic flow is really a signature
284: for {\em early pressure} in the system \cite{S97,S99,KSH99}.
285: Finally, we advertize central U+U collisions in the side-on-side 
286: configuration as an optimum system for studying the hydrodynamic 
287: evolution of elliptic flow and the quark-hadron phase transition
288: signature in its beam energy dependence \cite{KSH99}. We give predictions
289: for the time evolution of radial and elliptic flow, for their excitation
290: function and for the $p_{_{\rm T}}$-dependence of the elliptic flow 
291: coefficient at SPS energies for this particular collision system.
292: 
293: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
294: \se{The hydrodynamic model}
295: \label{sec2}
296: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
297: 
298: The equations of relativistic ideal hydrodynamics follow from the local
299: conservation laws for energy, momentum, and other conserved currents
300: (e.g. baryon number),
301:  \beq{econs}
302:    {\partial_{\mu} T^{\mu \nu}}(x)=0
303:    \mbox{\hspace{.6cm} and \hspace{.6cm}}
304:    \partial_\mu j^\mu(x) =0\,, 
305:  \eeq
306: by inserting the ideal fluid decompositions
307:  \bea{decomp1}
308:    &&T^{\mu\nu}(x)=\Bigl(e(x)+p(x)\Bigr) u^\mu(x) u^\nu(x) 
309:      - g^{\mu\nu} p(x)\,,
310:  \\
311:  \label{decomp2}
312:    &&j^\mu(x) = n(x)\, u^\mu(x)\,.
313:  \eea
314: $e(x)$ is the energy density, $p(x)$ the pressure, and $n(x)$ the 
315: conserved number density at point $x^\mu{=}(t,x,y,z)$; 
316: $u^\mu(x){\,=\,}\gamma (1,v_x,v_y,v_z)$ with 
317: $\gamma{\,=\,}1/{\textstyle\sqrt{1{-}v_x^2{-}v_y^2{-}v_z^2}}$
318: is the local four velocity of the fluid. Ideal hydrodynamics assumes
319: that local thermalization by the strong interactions among the matter 
320: constituents happens fast on the scale defined by the space-time 
321: gradients of these quantities and therefore neglects such 
322: gradient terms \cite{Cs94}.
323: 
324: We always use $x$ for the transverse coordinate inside the reaction 
325: plane, with positive values in the direction of the impact parameter
326: $\bbox{b}$, and $y$ for the transverse coordinate perpendicular to
327: $\bbox{b}$. (In momentum space $y$ denotes the rapidity; which meaning
328: is implied should follow from the context.) $z$ points in beam direction.
329: 
330: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
331: \suse{The equation of state}
332: \label{sec2a}
333: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
334: 
335: The set (\ref{econs}) contains 5 equations for 6 unknown fields 
336: $e, n, p, v_x, v_y, v_z$. To close the system one needs an equation
337: of state (EOS) which relates pressure, energy and baryon density.
338: The EOS for strongly interacting matter involves a phase transition 
339: from a hadron resonance gas (HG) phase to a color-deconfined 
340: quark-gluon plasma (QGP) phase. Like many others before (see, e.g., 
341: \cite{LRH86,Setal97}) we accomplish this by separately constructing 
342: an EOS for a resonance gas (EOS~H) and for the QGP phase (EOS~I)
343: %
344: %%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
345:  \begin{figure}[h,t,b,p]
346:  \epsfig{file=fig1.ps,width=8cm}
347:  \caption{The three equations of state discussed in the text, at 
348:           vanishing net baryon density.}
349:  \label{F1}
350:  \end{figure}
351: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
352: %
353: \noindent
354: and matching the two via the Maxwell construction, invoking a bag constant
355: $B$ to describe the different vacuum energy in the two phases. EOS~H is 
356: constructed from the contributions of all known hadron resonances of 
357: masses up to 2 GeV; their repulsive short-range interactions are 
358: parametrized via a mean-field potential ${\mathcal V}(n)=\half K n^2$ 
359: with $K=0.45$ GeV\,fm$^3$ \cite{Setal97}. The QGP is described as an 
360: ideal gas of massless quarks and gluons (EOS~I) inside a large bag with 
361: bag constant $B$. The latter is tuned to the desired phase transition 
362: temperature: $B^{1/4}=230$\,MeV gives $T_{\rm c}(n=0) = 164$\,MeV at 
363: vanishing net baryon density. EOS~I is given by the simple equation 
364: $p(e,n) = {1\over 3} e$ or $\P p/\P e = {1\over 3}$, independent of 
365: $n$.
366: 
367: In order to investigate the influence of the phase transition on 
368: the anisotropic transverse flow pattern, we stu\-died separately the 
369: equations of state EOS~H and EOS~I as well as the combined equation 
370: of state EOS~Q which includes the phase transition between the first 
371: two as obtained from the Maxwell construction. Comparisons to data 
372: are only performed for EOS~Q. Figure \ref{F1} shows all three 
373: equations of state for vanishing net baryon density $n=0$ while 
374: Fig.~\ref{F2} gives for EOS~Q the pressure as a function of both 
375: $e$ and $n$.
376: 
377: %%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
378:  \begin{figure}[h,t,b,p]
379:  \epsfig{file=fig2.ps,width=8cm}
380:  \caption{The equation of state EOS~Q with a quark-hadron phase 
381:     transition. The pressure is shown as a function of energy and
382:     net baryon density, $e$ and $n$. For each value of $n$ there exists 
383:     a minimum energy density $e_{\rm min}(n)$ with corresponding pressure
384:     $p_{\rm min}(e_{\rm min},n)$; below $e_{\rm min}$ the pressure is set 
385:     to zero by hand.}
386:  \label{F2}
387:  \end{figure}
388: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
389: 
390: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
391: \suse{Reduction to 2+1 dimensions}
392: \label{sec2b}
393: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
394: 
395: At high collision energies, relativistic kinematics and its influence 
396: on the particle production process implies longitudinal boost-invariance 
397: of the collision fireball near midrapidity \cite{Bj83}. (Of course, 
398: near the target and projectile rapidities longitudinal boost-invariance 
399: is broken by the finite amount of total available energy.) As a result,
400: the longitudinal velocity field scales as $v_z=z/t$, and it is convenient 
401: to use a coordinate system spanned by longitudinal proper time 
402: $\tau=t\sqrt{1-v_z^2}$ and space-time rapidity $\eta = \half
403: \ln[(t{+}z)/(t{-}z)]$ instead of $t$ and $z$ (see Appendix~\ref{appa}).
404: Longitudinal boost-invariance is then equivalent with 
405: $\eta$-independence.
406: 
407: Assuming the validity of this scaling ansatz near midrapidity, the
408: longitudinal expansion of the fireball can be dealt with analytically,
409: thereby reducing the numerical problem to the two transverse dimensions
410: and time \cite{O92}. This greatly reduces the numerical effort. However,
411: by doing so one gives up the possibility to study the rapidity dependence
412: of the (anisotropic) transverse flow pattern \cite{Hirano,Netal99}
413: as well as other interesting effects which occur at AGS and SPS 
414: energies, like the tilt of the longitudinal axis of the collision 
415: fireball away from the beam direction \cite{Netal99,Betal00,LHW00}. For 
416: such studies a complete solution of the (3+1)-dimensional hydrodynamics 
417: \cite{Hirano,Netal99,Betal00,RBM95} is required. We will here concentrate
418: entirely on the midrapidity region where the (2+1)-dimensional approach 
419: with exact longitudinal boost-invariance is expected to yield reasonable
420: results even at SPS energies. At higher energies the model should become
421: better and better.
422: 
423: The implementation of longitudinal boost-invariance and transformation
424: from $(t,z)$ to $(\tau,\eta)$ is described in Appendix~\ref{appa}. The
425: rewritten hydrodynamic equations read
426:  \bea{DGL}
427:    \begin{array}{c@{\,+\;}c@{\,+\;}c@{\,=\;}l}
428:    \partial_\tau\tilde{T}^{\tau\tau}                           &   
429:    \partial_x\left(\tilde{v}_{x} \tilde{T}^{\tau\tau}\right)   &   
430:    \partial_y\left(\tilde{v}_{y} \tilde{T}^{\tau\tau}\right)   &    
431:    -p\,,
432:  \\
433:    \partial_\tau\tilde{T}^{\tau x}                             &
434:    \partial_x\left(\bar{v}_x \tilde{T}^{\tau x}\right)         &
435:    \partial_y\left(\bar{v}_y \tilde{T}^{\tau x}\right)         & 
436:   -\partial_x\tilde{p}\,,
437:  \\
438:    \partial_\tau\tilde{T}^{\tau y}                             &
439:    \partial_x\left(\bar{v}_x \tilde{T}^{\tau y}\right)         &
440:    \partial_y\left(\bar{v}_y \tilde{T}^{\tau y}\right)         &
441:   -\partial_y\tilde{p}\,,
442:  \\
443:    \partial_\tau\tilde{\jmath}^\tau                            &
444:    \partial_x(\bar{v}_x \tilde{\jmath}^\tau)                   &
445:    \partial_y(\bar{v}_y \tilde{\jmath}^\tau)                   &
446:    0\,,
447:  \end{array}
448:  \eea
449: where 
450:  \bea{defns}
451:    &&\tilde T^{\mu \nu} = \tau T^{\mu \nu}\,,\quad
452:      \tilde p = \tau\, p\,,
453:  \\
454:    &&\bar v_i = v_i \cosh\eta\,, \quad
455:      \tilde v_i = {T^{\tau i}\over T^{\tau \tau}}
456:                 = {(e+p)\bar\gamma^2\bar{v}_i \over
457:                    (e+p)\bar\gamma^2{-}p}
458:      \,,\quad(i=x,y).
459:  \nonumber
460:  \eea   
461: We call $\bar v_i$ the transport velocities and $\tilde v_i$ the energy
462: flow velocities in the transverse directions. Since we work at midrapidity, 
463: $\eta=0$, the transverse transport velocities agree with the corresponding
464: fluid velocities in the c.m. frame.
465: 
466: In hydrodynamic problems phase transitions generically lead to the 
467: formation of shock waves which complicate the numerical solution.
468: To integrate the differential equations (\ref{DGL}) we use the 
469: ``Sharp and Smooth Transport Algorithm'' (SHASTA \cite{BB73}) 
470: which was shown to perform excellently even under difficult 
471: conditions \cite{RBM95}.
472: 
473: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
474: \suse{Initialization of the fields}
475: \label{sec2c}
476: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
477: 
478: In this subsection we discuss the initial conditions for the solution
479: of Eqs.~(\ref{DGL}). Strong interactions between the partons of the 
480: colliding nuclei lead to the deposition of a large fraction of the 
481: beam energy and the creation of many secondary particles in the 
482: reaction zone. The newly produced partons interact strongly with each 
483: other and, after only a few scatterings during a time interval 
484: $\tau_0 = {\cal O}$(1\,fm/$c$), the system is expected to reach a 
485: state of approximate local thermal equilibrium. Following 
486: \cite{O92,BO90} (to which we refer for details) we take the energy 
487: deposition in the transverse plane to be proportional (by a factor $K$)
488: to the number of collisions producing wounded nucleons:
489:  \bea{init}
490:    \FL
491:    && e(x,y;\tau_0) =
492:  \\
493:    && K \left\{ T_A\bigl(x{+}{\textstyle{b\over 2}},y\bigr)
494:       \Bigl[1-\Bigl(1-{\sigma T_B\bigl(x{-}{\textstyle{b\over 2}},y\bigr)
495:                        \over B}\Bigr)^B \Bigr] \right. 
496:  \nonumber\\
497:    &&\ \ + \left. T_B\bigl(x{-}{\textstyle{b\over 2}},y\bigr)
498:      \Bigl[1-\Bigl(1-{\sigma T_A\bigl(x{+}{\textstyle{b\over 2}},y)
499:                       \over A}\Bigr)^A \Bigr]\right\}\,.
500:  \nonumber
501:  \eea
502: %
503: Here $T_A$ is the nuclear thickness function of the incoming 
504: nucleus $A$,
505:  \beq{TA}
506:    T_A(x,y)=\int_{-\infty}^{+\infty} dz\, \rho_A(x,y,z) \,,
507:  \eeq
508: where the nuclear density $\rho_A$ is given by a Woods-Saxon profile,
509:  \beq{rhoA}
510:    \rho_A(\bbox{r}) = \frac{\rho_0}{1+\exp[(r-R_0)/\xi]}\,,
511:  \eeq
512: %
513: and similarly for nucleus $B$. 
514: 
515: We further assume that the initial transverse density profile of net 
516: baryon number is proportional to the initial transverse energy 
517: density profile:
518:  \beq{n}
519:    n(x,y;\tau_0)=L\,e(x,y;\tau_0)\,.
520:  \eeq
521: For Pb-Pb collisions we use in (\ref{rhoA}) a nuclear radius 
522: $R_0=6.5$\,fm and a surface thickness $\xi=0.54$\,fm \cite{BM69}. 
523: For U-U collisions we take $R_0=6.8$\,fm, with a deformation 
524: $\delta=0.27$ (\cite{BM69}, Vol.~2, p.~133). This leads to a 
525: ratio $R_l/R_s=1.29$ between the long and short axes of this 
526: nucleus; in absolute terms $R_l=8.0$\,fm and $R_s=6.2$\,fm \cite{fn1}. 
527: For the ground-state nuclear density we take $\rho_0=0.17$\,fm$^{-3}$ 
528: \cite{BM69}. 
529: 
530: %%%%%%%%%%%%%%%%%%%%%%%%% Fig. 3 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
531: \begin{figure}[h,t,b,p]
532: \hspace*{-0.4cm}\epsfig{file=fig3.ps,width=\linewidth}
533: \caption{Left: initial transverse energy density distribution for a 
534:    typical 158\,$A$\,GeV/$c$ Pb+Pb collision at impact parameter 
535:    $b{\,=\,}7$\,fm. Indicated are contours of constant ener\-gy density 
536:    between $e{\,=\,}7.0$\,GeV/fm$^3$ (innermost contour) and 
537:    $e{\,=\,}0.5$\,GeV/fm$^3$ (outermost contour) in steps of 
538:    $\Delta e$\,=\,0.5 GeV/fm$^3$. The dashed lines represent the 
539:    colliding nuclei before impact. Right: The same for a central 
540:    155\,$A$\,GeV/$c$ side-on-side U+U collision -- the innermost 
541:    (outermost) contour corresponds to $e{\,=\,}8.0$\,GeV/fm$^3$ 
542:    (0.5\,GeV/fm$^3$). 
543: \label{F3}}
544: \end{figure}
545: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
546: 
547: Three parameters thus describe the initial conditions:
548: \bit
549: \item{the maximum energy density $e_0$ in a central collision 
550:       ($b=0$) -- this fixes the parameter $K$ in (\ref{init}) at 
551:       the given beam energy;}
552: \item{the ratio $L$ in (\ref{n}) between energy and baryon density;}
553: \item{the equilibration time $\tau_0$.}
554: \eit
555: In Sec.~\ref{sec2e} we adjust the parameters by tuning the output of 
556: our calculations with EOS~Q for central ($b=0$) Pb+Pb collisions to 
557: experimental data (transverse mass spectra of negative hadrons and 
558: net protons at midrapidity \cite{NA49spec}) at 158 $A$\,GeV/$c$ beam 
559: momentum. We use the same parameters $K,L$ and $\tau_0$ for U+U 
560: collisions at 155 $A$\,GeV/$c$.
561: 
562: In Fig.~\ref{F3} we illustrate the initial conditions resulting from 
563: this tuning procedure. It shows contour plots of the energy density 
564: in the transverse plane at $z=0$ for Pb+Pb collisions with $b=7$\,fm 
565: and central U+U collisions in the side-on-side configuration at the 
566: highest SPS beam momentum of $400\times(Z/A)$ GeV/$c$. Note that at 
567: fixed collision energy the central energy density for $b{\,=\,}0$ 
568: side-on-side U+U collisions is 8\% lower than for $b{\,=\,}0$ Pb+Pb 
569: collisions, but about 14\% higher than in Pb+Pb collisions at 
570: $b{\,=\,}7$\,fm which correspond to about the same initial spatial 
571: deformation. At similar deformation, the initial volume of the 
572: elliptic fireball formed in central side-on-side U+U collisions is 
573: almost twice that of the corresponding semi-central Pb+Pb collisions.
574: 
575: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
576: \suse{Freeze-out and particle spectra}
577: \label{sec2d}
578: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
579: 
580: As the matter expands and cools, the mean free path of the matter
581: constituents grows, and the hydrodynamical description eventually
582: breaks down. The system reaches the point of ``kinetic freeze-out''
583: after which the momentum spectra are no longer significantly affected 
584: by scattering among the particles. One should stop the hydrodynamic 
585: solution when the average time between scatterings $\tau_{\rm scatt}=
586:  1/\la v\sigma \ra n$ becomes comparable to the expansion time scale 
587: $\tau_{\rm exp} = 1/\partial\cdot u$ (inverse ``Hubble constant'') 
588: \cite{BGZ78,HLR87,LHS89}. [It was shown in \cite{SH92} that in 
589: relativistic heavy ion collisions freeze-out happens {\em dynamically} 
590: rather than geometrically, i.e. it is driven by the expansion of the 
591: fireball and not by its finite size.] Numerical calculations 
592: \cite{SH92,MH97} have shown that, since the particle density in 
593: the denominator of $\tau_{\rm scatt}$ is a very steep function of 
594: $T$, this leads to freeze-out at nearly constant temperature. For 
595: low net baryon freeze-out densities, as they arise in heavy ion 
596: collisions at and above SPS energies near midrapidity, this corresponds 
597: to almost constant energy density. We here therefore impose freeze-out
598: at a constant energy density $e_{\rm dec}$ which is the most easily 
599: implemented condition in hydrodynamics. The value of $e_{\rm dec}$ (or,
600: nearly equivalently, $T_{\rm dec}$) is another model parameter to be 
601: tuned to the data. 
602: 
603: After the freeze-out hypersurface $\Sigma$ of constant energy density
604: $e_{\rm dec}$ has been determined, the $T_{\rm dec}(x)$, chemical 
605: potentials $\mu_i(x)$ and flow velocity field $u_\mu(x)$ are evaluated 
606: on this surface. To this end a tabulated version of EOS~H is used
607: for interpolation which (in addition to the pressure $p$) gives the 
608: intensive thermodynamical variables as functions of $e$ and $n$.
609: Each cell $x$ on this freeze-out hypersurface contributes particles
610: of species $i$ (where $i$ runs over all resonances included in EOS~H)
611: with a local equilibrium distribution
612:  \beq{distr}
613:    f_i(x,p) = {g_i\over (2\pi)^3} \, 
614:    {1\over e^{[p\cdot u(x) - \mu_i(x)]/T_{\rm dec}(x)}\pm 1} \,.
615:  \eeq
616: $g_i$ is the spin-isospin degeneracy factor for particle species $i$.
617: The complete momentum spectrum is obtained by summing the corresponding
618: particle flux currents across the 3-dimensional freeze-out hypersurface 
619: $\Sigma$ in space-time over all cells in $\Sigma$ (Cooper-Frye 
620: prescription \cite{CF74}):
621:  \beq{CF} 
622:    E {dN_i\over d^3p} = 
623:    {dN_i \over dy\, p_{_{\rm T}} dp_{_{\rm T}} \, d\phi} =
624:    \int_\Sigma p\cdot d^3\sigma(x)\, f_i(x,p)\,. 
625:  \eeq
626: This prescription is strictly correct only for freeze-out surfaces
627: whose normal vector $d^3\sigma(x)$ is everywhere timelike because 
628: otherwise some particles flow back into the 4-volume inside 
629: $\Sigma$. A discussion of this issue which still awaits 
630: a fully consistent solution can be found in \cite{BMGR96,Bu96}.
631: 
632: In the present paper we concentrate on flow patterns reflected in pion 
633: spectra. (Flow anisotropies for pions and protons at SPS energies were 
634: compared in \cite{KSH99}.) A significant fraction of the measured pions 
635: arises from the decays of unstable resonances after freeze-out. These
636: decays usually happen isotropically in the rest frame of the resonance
637: and tend to smear out flow anisotropies, thereby reducing the anisotropic 
638: flow signals \cite{KSH99,Hirano}. The fraction of pions from resonance 
639: decays depends strongly on the freeze-out temperature: their diluting
640: effect on the elliptic flow $v_2$, for example, is much stronger at 
641: $T_{\rm dec}{\,=\,}140$\,MeV \cite{Hirano} than at 
642: $T_{\rm dec}{\,=\,}120$\,MeV \cite{KSH99}. All our calculations fully
643: account for resonance decay contributions including the complete
644: relativistic decay kinematics \cite{SKH91}. 
645: 
646: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
647: \suse{Tuning the model}
648: \label{sec2e}
649: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
650:  
651: Since the hydrodynamic approach cannot describe the initial thermalization
652: stage directly after nuclear impact, the initial conditions for the
653: hydrodynamic expansion stage cannot be predicted but must be obtained
654: by fitting experimental data. However, once the initial conditions 
655: (in our case the parameters $K,L$, and $\tau_0$) have been fixed in 
656: central collisions, the Glauber model (\ref{init}) uniquely predicts 
657: their dependence on the impact parameter. The validity of the 
658: hydrodynamic model can thus be tested by checking the impact parameter
659: dependence of its predictions. In \cite{KSH99,Kolb} we showed that, 
660: after being tuned to central Pb+Pb collisions at 158 $A$\,GeV/$c$, the 
661: model successfully reproduces the measured pion and proton spectra near 
662: midrapidity up to impact parameters of 8-10\,fm. This was better than 
663: expected.
664: 
665: We here provide some details of the tuning procedure which were not
666: previously reported in \cite{KSH99} due to space limitations. In 
667: particular we show in Fig.~\ref{F4} our fit to the midrapidity 
668: $m_{_{\rm T}}$-spectra of negative hadrons ($h^-$) and net protons 
669: measured by the NA49 collaboration \cite{NA49spec}. The theoretical 
670: spectra are absolutely normalized. The corresponding fit parameters for 
671: the initial state are $e_0{\,=\,}9.0$ GeV/fm$^3$ for the initial energy 
672: density in the center of the fireball (corresponding to $K=2.04$\,GeV/fm 
673: in (\ref{init}) and to an initial central temperature $T_0{\,=\,}258$ 
674: MeV \cite{fn1a}), $n_0{\,=\,}1.1$ fm$^{-3}$ for the initial baryon density 
675: in the fireball center (corresponding to $L=0.122$\,GeV$^{-1}$ in (\ref{n})), 
676: and a starting time $\tau_0=0.8$\,fm/$c$ for the hydrodynamic expansion
677: (corresponding to $T_0\cdot\tau_0/\hbar{\,=\,}1.05$). $\tau_0$ controls 
678: the dilution of the matter via boost-invariant longitudinal expansion 
679: and thus the length of time available for the buildup of transverse 
680: flow before freeze-out; the latter affects the slope of the 
681: $m_{_{\rm T}}$-spectra. The total time until freeze-out and the amount 
682: of transverse flow generated can also be changed by varying the initial 
683: energy density, but this also affects the normalization of the
684: midrapidity spectra. $e_0$ and $\tau_0$ result from a suitable 
685: balance between these two effects. $n_0$ is then essentially fixed 
686: by the measured ratio between the proton and $h^-$ spectra.
687: 
688: %%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 4 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
689: \begin{figure}[htbp]
690: \epsfig{file=fig4.ps,width=8cm}
691: \caption{Particle spectra from central Pb+Pb collisions at 
692:   158\,$A$\,GeV/$c$ at midrapidity \protect\cite{NA49spec} together 
693:   with the hydrodynamical model predictions after tuning of the model 
694:   parameters (solid lines).
695: \label{F4}}
696: \end{figure}
697: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
698: 
699: The different shapes of the proton and $h^-$ spectra provide a handle
700: to separate collective transverse flow ($\lda v_\perp \rda$) from 
701: thermal motion ($T_{\rm dec}$) at freeze-out. However, it is known that
702: a thermal model analysis of particle spectra in general results in 
703: strong correlations between these two parameters \cite{LHS89,SH92}.
704: Our best fit gives $T_{\rm dec}{\,\approx\,}120$\,MeV (corresponding 
705: to $e_{\rm dec}{\,=\,}0.06$ GeV/fm$^3$) and 
706: $\lda v_\perp\rda{\,=\,}0.45\,c$, albeit with a significant 
707: uncertainty (somewhat lower $T_{\rm dec}$ with higher $\lda v_\perp \rda$ 
708: and vice versa cannot be excluded). This is in good agreement with 
709: other analyses of particle spectra \cite{Ketal97} and hydrodynamic
710: simulations \cite{HS98}; a combined analysis of spectral slopes and 
711: two-particle Bose-Einstein correlations \cite{NA49HBT,TWH99}
712: tends to give somewhat larger transverse flow velocities coupled to 
713: lower freeze-out temperatures, but still inside the region of 
714: uncertainty from the analysis of the single-particle spectra.
715: 
716: This set of fit parameters, adjusted to SPS data, is our starting 
717: point for extrapolations towards non-central collisions and into 
718: different collision energy regimes. When studying the impact parameter 
719: dependence at fixed collision energy we leave all parameters unchanged. 
720: This may be unrealistic for very peripheral collisions where the 
721: midrapidity fireball is smaller and geometric freeze-out can cut the 
722: expansion short, leading to higher decoupling temperatures. For the
723: spectral slopes this is a second order effect since earlier freeze-out
724: at higher $T_{\rm dec}$ is partially compensated for by a smaller 
725: transverse flow velocity $\lda v_\perp \rda$. As we will see below
726: (see Fig.~\ref{F7}), the elliptic flow anisotropy $v_2$ builds up early 
727: in the collision and, even at SPS energies, has almost reached its 
728: final value already several fm/$c$ before decoupling; a possible 
729: earlier decoupling in very peripheral collisions thus will not 
730: strongly affect $v_2$ either. We thus feel justified in leaving the 
731: model parameters (in particular the decoupling temperature) unchanged 
732: when studying the impact parameter dependence.
733: 
734: When investigating the excitation function of radial and elliptic
735: flow we change $K$ and $\tau_0$. This is rationalized as follows:
736: At higher energies we expect higher particle production per 
737: wounded nucleon; we cannot predict the beam energy dependence of 
738: secondary particle production, but we can parametrize it by changing 
739: $K$ and plotting our results as a function of the finally observed 
740: multiplicity density $dN/dy$. The beam energy dependence of $dN/dy$
741: will eventually be provided by experiment, then allowing to present
742: our results directly against $\sqrt{s}$. -- Higher initial particle
743: production leads to higher particle and energy densities and thereby
744: to accelerated thermalization. From relativistic kinematics and the 
745: uncertainty relation it follows that the production time of a secondary 
746: particle is inversely related to its energy \cite{KLS92}; by dimensional 
747: analysis this suggests that the thermalization time $\tau_0$ scales in 
748: inverse proportion to the initial temperature $T_0$: 
749: $T_0 \tau_0$\,=\,const. or, equivalently, $\tau_0\,K^{1/4}$\,=\,const. 
750: This is what we use in the present paper; in \cite{KSH99} we instead 
751: left $\tau_0$ constant. Within the range of collision energies studied 
752: in \cite{KSH99} the difference is negligible, but for the higher 
753: energies investigated here a reduction of $\tau_0\sim 1/T_0$ causes
754: a significant shrinkage of the horizontal axis on the excitation 
755: function in Fig.~\ref{F14} below.
756: 
757: For energies above the SPS we leave the initial baryon density
758: $n(x,y;\tau_0)$ unchanged. As a result, the ratio $L$ of baryon to
759: energy density drops, qualitatively consistent with the expectation of
760: decreasing baryon stopping. Since already at the SPS the influence of 
761: the baryons on the EOS is minor, it doesn't really matter in which way
762: $L$ approaches zero as the collision energy goes to infinity. Note
763: that we don't predict the normalization of the baryon spectra at other
764: than SPS energies. Below SPS energies we leave $L$ constant, lacking
765: motivation for a different choice. Once a better understanding of the
766: beam energy dependence of the initial conditions becomes available,
767: this can be easily improved.  
768:   
769: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
770: \se{Transverse flow phenomenology}
771: \label{sec3}
772: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
773: 
774: In this section we study generally the space-time evolution of the 
775: transverse flow pattern and how it is influenced by a phase transition
776: in the EOS. Since the finally observed particle spectra and their azimuthal
777: anisotropies reflect the full space-time history of the fireball expansion,
778: their proper interpretation requires an accurate understanding of the 
779: transverse fireball evolution. In \cite{KSH99} we showed that the 
780: softening of the EOS in the phase transition region leads at collision 
781: energies above the SPS to a reduction of the elliptic flow coefficient 
782: $v_2$ below the value expected from a hadron resonance gas. At even higher
783: energies, however, one expects to enter a regime where the initial 
784: energy density is so far above the phase transition that nearly all of
785: the expansion history happens inside the QGP phase. Since far above 
786: $T_{\rm c}$ the EOS of a QGP ($p{\,=\,}{1\over 3}e-B$) is much harder
787: than EOS~H (which in the region relevant for us can be parametrized
788: by $p{\,\approx\,}0.15\,e$), $v_2$ should eventually rise again and 
789: approach the value characteristic of EOS~I which is 30-40\% higher. 
790: In order to see whether this is true we have now studied collisions at
791: very much higher energies, even far beyond the LHC.
792:  
793: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
794: \suse{Semiperipheral Pb+Pb collisions}
795: \label{sec3a}
796: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
797: 
798: In this subsection we investigate Pb+Pb collisions at an impact
799: parameter of 7\,fm (left panel in Fig.~\ref{F3}). We begin by showing
800: the evolution of the energy distribution and flow field in the
801: transverse $(x,y)$ plane for the cases with and without a phase
802: transition. We do so for an initial central energy density in
803: $b{\,=\,}0$ Pb+Pb collisions of $e_0{\,=\,}175$\,GeV/fm$^3$
804: ($T_0{\,=\,}510$\,MeV) at $\tau_0{\,=\,}0.38$\,fm/$c$. The resulting
805: total pion multiplicity density with EOS~Q of 
806: ${dN_\pi\over dy}\big\vert_{y=0}{\,=\,}1070$ at $b{\,=\,}7$\,fm is at
807: the upper end of the range of predictions for RHIC energies \cite{RHIC}. 
808: This study was motivated by the work of Teaney and Shuryak who
809: predicted under similar conditions an interesting phenomenon which
810: they called ``nutcracker flow'' \cite{TS99} and which shows up only in
811: the presence of a phase transition. In Fig.~\ref{F5} we show the
812: evolution for EOS~I, i.e. a hard EOS without phase transition. One 
813: sees smooth expansion and a continuous transition from an initial
814: state of positive elliptic deformation (longer axis perpendicular to
815: the collision plane) to one with negative deformation, caused by the 
816: developing in-plane elliptic flow. The thicker contours correspond 
817: (from the inside outward) to $e{\,=\,}1.6$, 0.45, and 0.06\,GeV/fm$^3$; 
818: for the more realistic equation of state EOS~Q the first two values 
819: limit the mixed phase while the latter indicates freeze-out.
820: 
821: Figure \ref{F6} shows the analogous situation for EOS~Q (which includes
822: a phase transition) for identical initial conditions. Compared to 
823: Fig.~\ref{F5} one sees clear differences: the lack of a pressure
824: gradient in the mixed phase inhibits its transverse expansion; the 
825: hadronic phase ouside the mixed phase expands quickly and freezes out,
826: leaving a shell of mixed phase matter behind which inertially confines
827: the QGP matter in the center. The matter with the softest EOS (smallest
828: $p/e$) is concentrated around the QGP/mixed interface (thick contour at 
829: 1.6 GeV/fm$^3$). When the QGP matter finally pushes the mixed phase 
830: shell apart (the ``nutcracker phenomenon'' discovered in \cite{TS99}),
831: the energy density contours develop an interesting structure vaguely
832: reminiscent of two separated half shells. Compared to Fig.~\ref{F5}, 
833: the elliptic flow clearly needs more time to push the matter from 
834: a state of positive to one of negative elliptic deformation. This is 
835: due to the inertia of the mixed phase shell which does not participate 
836: in the pushing.  
837: 
838: %%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 5 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
839: \begin{figure}[htbp]
840: \bce
841: \epsfig{file=fig5.ps,width=8.5cm}
842: \caption{Time evolution for EOS~I of the transverse energy density 
843:    profile (indicated by constant energy density contours spaced by 
844:    $\Delta e{\,=\,}150$\,MeV/fm$^3$) and of the flow velocity 
845:    field (indicated by arrows) for Pb+Pb collisions at impact 
846:    parameter $b{\,=\,}7.0$\,fm. The four panels show snapshots 
847:    at times $\tau{-}\tau_0{\,=\,}3.2$, 4.0, 5.6, and 8.0\,fm/$c$. At
848:    these times the maximal energy densities in the center are 5.63,
849:    3.62, 1.31 and 0.21 GeV/fm$^3$, respectively. For further details
850:    see text.  
851: \label{F5}}
852: \ece
853: \end{figure}
854: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
855: 
856: Figures \ref{F5} and \ref{F6} emphasize the spatial structure of
857: the fireball at fixed time steps. Let us now study the time evolution
858: in more detail. To this end we condense the information contained
859: in the density and flow patterns into three time-dependent scalar 
860: quantities: 
861: 
862: {\em (i)} The ``spatial ellipticity'' 
863:  \beq{eps_x}
864:    \epsilon_x = {\lda y^2{-}x^2 \rda \over \lda y^2{+}x^2 \rda}
865:  \eeq
866: characterizes the spatial deformation of the fireball in the transverse
867: plane. The angular brackets denote energy density weighted spatial 
868: averages at a fixed time. $\epsilon_x$ causes azimuthal anisotropies 
869: in the transverse pressure gradients which would eventually drive it 
870: to zero if the hydrodynamic evolution were not cut short by the 
871: freeze-out process. 
872: 
873: 
874: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 6 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
875: \begin{figure}[htbp]
876: \bce
877: \epsfig{file=fig6.ps,width=8.5cm}
878: \caption{Same as Fig.~\ref{F5}, but for EOS~Q which features a phase
879:    transition. The spacing between energy density contours is again 
880:    150 MeV/fm$^3$, and the snapshots are taken at the same times. The 
881:    corresponding maximum energy densities are 5.97, 3.97, 1.67, and
882:    0.55 GeV/fm$^3$, respectively. See text for discussion.
883: \label{F6}}
884: \ece
885: \end{figure}
886: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
887: 
888: {\em (ii)} The momentum anisotropy 
889:  \beq{eps_p}
890:    \epsilon_p = {\lda T^{xx}{-}T^{yy} \rda \over  
891:                  \lda T^{xx}{+}T^{yy} \rda} 
892:  \eeq   
893: measures in an analogous way the anisotropy of the transverse 
894: momentum-space density. It is directly calculated from the spatial 
895: components of the energy momentum tensor but, as shown in \cite{KSH99},
896: at freeze-out it is nearly equal to the $p_{_{\rm T}}^2$-weighted 
897: elliptic flow $v_{2,p_{\rm T}^2}$ for pions as calculated from their
898: final momentum spectra \cite{fn2}. Its time-dependence thus provides a 
899: picture of the dynamical buildup of the elliptic flow even at early 
900: times when the elliptic flow coefficient $v_2$ (which is calculated
901: from hadronic momentum spectra, see Sec.~\ref{sec4}) is not yet
902: defined. For pions at freeze-out $v_2$ is given by $2\,v_2 \approx$
903: $v_{2,p_{\rm T}^2} \approx \epsilon_p$ \cite{KSH99}. 
904: 
905: {\em (iii)} The time-dependence of the average radial flow velocity
906:  \beq{vperp}
907:    \lda v_\perp \rda = {\llda \gamma {\textstyle\sqrt{v_x^2{+}v_y^2}}\rrda 
908:                         \over \lda \gamma \rda} \,.
909:  \eeq
910: characterizes the buildup of the overall transverse expansion which is
911: modulated by the elliptic flow. Comparing the time-dependencies of
912: $\lda v_\perp \rda$ and $\epsilon_p$ allows to answer the question
913: to which stages of the expansion (i.e. to which domains of the EOS) 
914: each one is most sensitive.
915: 
916: %%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 7 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
917: \begin{figure}[htbp]
918: \bce
919: \hspace*{0.3cm}\epsfig{file=fig7a.ps,height=4.9cm,width=7.1cm}\\
920: \hspace*{0.1cm}\epsfig{file=fig7b.ps,height=4.9cm,width=7.34cm}\\
921: \epsfig{file=fig7c.ps,height=4.9cm,width=7.5cm}\\
922: \caption{Time evolution of the spatial ellipticity $\epsilon_x$, the 
923:   momentum anisotropy $\epsilon_p$, and the radial flow
924:   $\lda v_\perp \rda$. The labels {\tt a, b, c} and {\tt d} denote 
925:   systems with initial energy densities of 9, 25, 175 and 25000 
926:   GeV/fm$^3$, respectively, expanding under the influence of EOS~Q. 
927:   Curves {\tt e} show the limiting behaviour for EOS~I as $e_0\to\infty$ 
928:   (see text). In the lower two panels the two vertical lines below each 
929:   of the curves {\tt a-d} limit the time interval during which the 
930:   fireball center is in the mixed phase. In the upper panel the dots
931:   (crosses) indicate the time at which the center of the reaction zone 
932:   passes from the QGP to the mixed phase (from the mixed to the HG
933:   phase). For curves {\tt a} and {\tt b} the stars indicate the
934:   freeze-out point; for curves {\tt c-e} freeze-out happens outside
935:   the diagram. 
936:  \label{F7}}
937: \ece
938: \end{figure}
939: \vspace*{-0.4cm}
940: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
941: 
942: We now give a detailed discussion of Figs.~\ref{F7}a-c which show 
943: (using EOS~Q) the time evolution for the above three quantities for 
944: a sequence of collision energies, parametrized by the initial central 
945: energy density in $b{\,=\,}0$ Pb+Pb collisions, $e_0$: $e_0$=\,9, 25, 
946: 175, and 25000 GeV/fm$^3$ (curves {\tt a} through {\tt d} in 
947: Figs.~\ref{F7}). With increasing $e_0$ the initial time $\tau_0$ 
948: was scaled down as described at the end of Sec.~\ref{sec2e}. The 
949: lowest of these $e_0$-values corresponds to 158\,$A$\,GeV Pb+Pb 
950: collisions at the SPS, while the highest value is far beyond the 
951: reach of even the LHC.
952: 
953: A calculation with EOS~I is shown for comparison as curve {\tt e}. 
954: Since EOS~I ($e=3p$) is completely scale invariant, the time 
955: evolution of the dimensionless ratios (\ref{eps_x}), (\ref{eps_p}), 
956: and (\ref{vperp}) is invariant under a rescaling of $e_0$ as long 
957: as $\tau_0$ is held fixed (see Eqs.~(\ref{DGL})). Changing 
958: $\tau_0\sim e_0^{-1/4}$ breaks this scaling, but only weakly
959: as we have checked. Curves {\tt e} in Fig.~\ref{F7} show the time
960: evolution for EOS~I in the limit $e_0\to\infty,\ \tau_0\to 0$. Not 
961: shown is a calculation with EOS~Q which was initialized with an 
962: extraordinarily high initial temperature of $T_0\approx 20$\,GeV 
963: ($e_0{\,=\,}25\times10^6$\,GeV/fm$^3$); during the first 16 fm/$c$ 
964: covered by Fig.~\ref{F7} it fully coincides with curve {\tt e}. In 
965: this case almost all of the matter stays in the QGP phase during this 
966: time period whose EOS coincides with EOS~I up to the (here negligible) 
967: bag constant. Therefore, as expected, the hydrodynamic evolution with 
968: EOS~Q approaches at asymptotically high energies that with EOS~I.
969: 
970: Inspection of Fig.~\ref{F7} shows that the elliptic flow $\epsilon_p$
971: saturates at large times while the radial flow $\lda v_\perp \rda$
972: keeps rising forever, albeit at a decreasing rate. The driving force 
973: for radial flow, the radial pressure gradient between the matter in
974: the fireball and the surrounding vacu\-um, never vanishes completely. 
975: The spatial ellipticity $\epsilon_x$, on the other hand, which is 
976: responsible for azimuthal anisotropies in the transverse pressure 
977: gradients and thus drives the evolution of $\epsilon_p$, passes 
978: through zero after some time. Afterwards the longer axis of the 
979: transverse fireball cross section no longer points perpendicular to 
980: the reaction plane, but {\em into} the reaction plane. 
981: A vanishing $\epsilon_x$ implies a vanishing growth rate for 
982: $\epsilon_p$; as $\epsilon_x$ turns negative, smaller oppositely 
983: directed anisotropies of the pressure gradients develop which can 
984: actually cause $\epsilon_p$ to decrease again. This can be seen in 
985: Fig.~\ref{F7}b for large values of $e_0$ where the sign of 
986: $\epsilon_x$ changes sufficiently early in the collision that 
987: pressures are still high enough to generate this effect. 
988: 
989: Qualitatively one hence can say that the final value of $\epsilon_p$ is 
990: established roughly at the point when $\epsilon_x$ passes through 
991: zero. For SPS energies this happens just before decoupling (implying
992: that the fireball freezes out in a nearly circular configuration), but
993: at high energies this occurs well before freeze-out. Generically the
994: freeze-out value of $\epsilon_p$ (and thus $v_2$) is sensitive to the 
995: EOS at significantly higher energy densities than the radial flow 
996: $\lda v_\perp \rda$. {\em The elliptic flow indeed measures the early
997: pressure} \cite{S97,S99}.
998:  
999: On a more detailed level, the time evolution shows an interesting 
1000: additional feature: In curves {\tt b} and {\tt c} the elliptic flow 
1001: $\epsilon_p$ is seen to peak even {\em before} $\epsilon_x$ passes 
1002: through zero. The origin of this phenomenon, which is related to the
1003: phase transition, will be discussed in Sec.~\ref{sec3c} below.
1004: 
1005: Comparison of the lower two panels in Fig.~\ref{F7} shows that the
1006: softening effect on the EOS of the phase transition affects the
1007: buildup of $\lda v_\perp \rda$ and $\epsilon_p$ at similar times.
1008: However, the influence on $\epsilon_p$ is stronger since elliptic 
1009: flow is a smaller effect (which feels only the anisotropies in
1010: the transverse pressure gradient, not its overall magnitude) and thus
1011: more fragile than radial flow. This results in a relatively larger 
1012: sensitivity of elliptic flow to the phase transition.
1013: 
1014: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1015: \suse{Central U+U collisions in the side-on-side configuration}
1016: \label{sec3b}
1017: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1018: 
1019: As discussed in Sec.~\ref{sec2c}, central U+U collisions in the
1020: side-on-side configuration provide 14\% higher energy density over
1021: nearly twice the volume at the same initial spatial deformation as
1022: Pb+Pb collisions at $b{\,=\,}7$\,fm. This leads to a longer lifetime
1023: for non-zero spatial ellipticity $\epsilon_x$, the driving force for 
1024: elliptic flow, and also for the whole fireball until freeze-out. 
1025: Hence the system has more time for thermalization, favoring the 
1026: applicability of our hydrodynamic method. For this reason we decided 
1027: to perform quantitative calculations for this system and make
1028: predictions for experiments with uranium beams at RHIC and LHC.  
1029: 
1030: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 8 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1031: \begin{figure}[htbp]
1032: \bce
1033: \epsfig{file=fig8.ps,width=8.5cm}
1034: \caption{Same as Figs.~\ref{F5} and \ref{F6} ($e_0{\,=\,}175$\,GeV/fm$^3$ 
1035:   at $\tau_0{\,=\,}0.38$\,fm/$c$, EOS~Q), but for central side-on-side 
1036:   U+U collisions. The spacing between energy density contours is again 
1037:   150 MeV/fm$^3$, and the snapshots are taken at the same times. The 
1038:   corresponding maximum energy densities are 8.71, 6.06, 3.27, and
1039:   1.47\,GeV/fm$^3$, respectively. See text for discussion.
1040: \label{F8}}
1041: \ece
1042: \end{figure}
1043: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1044: \vspace*{-0.5cm}
1045: 
1046: We first look once more at the space-time evolution of the transverse 
1047: energy density and flow profiles, shown in Fig.~\ref{F8}. The
1048: initialization corresponds to the same collision energy as in 
1049: Fig.~\ref{F6} ($e_0{\,=\,}175$ GeV/fm$^3$) but, since we now consider
1050: central ($b{\,=\,}0$) collisions, the initial energy density in the
1051: center of the deformed collision\break
1052: %
1053: %%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 9 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1054: \begin{figure}[ht]
1055: \bce
1056: \hspace*{0.2cm}\epsfig{file=fig9a.ps,height=5cm,width=7.3cm}\\
1057: \epsfig{file=fig9b.ps,height=5cm,width=7.5cm}\\
1058: \hspace*{-0.2cm}\epsfig{file=fig9c.ps,height=5cm,width=7.6cm}\\
1059: \caption{Same as Fig.~\ref{F7}, but now comparing central U+U (solid) 
1060:   to semiperipheral ($b{\,=\,}7$\,fm) Pb+Pb collisions (dashed) at
1061:   two selected beam energies. The curves labelled ``SPS'' correspond 
1062:   to $e_0$\,=\,9\,GeV/fm$^3$ (8.3\,GeV/fm$^3$) for central Pb+Pb
1063:   (side-on-side U+U) collisions, those labelled ``RHIC'' have 
1064:   $e_0$\,=\,175\,GeV/fm$^3$ in both cases.
1065:  \label{F9}}
1066: \ece
1067: \end{figure}
1068: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1069: %
1070: \noindent
1071: region is higher than in the
1072: semiperipheral Pb+Pb collisions of Fig.~\ref{F6}. As seen in 
1073: Fig.~\ref{F9}, the whole time evolution is slower for central U+U 
1074: than for semiperi\-phe\-ral Pb+Pb collisions, due to the larger system 
1075: size: At $\tau-\tau_0{\,=\,}3.2$\,fm/$c$ (the first shown snapshot)
1076: the central energy density is 50\% higher, and at
1077: $\tau-\tau_0{\,=\,}8$\,fm/$c$ (the last snapshot) it is even by a
1078: factor 3 larger than in $b{\,=\,}7$\,fm Pb+Pb collisions at the same
1079: beam energy. Freeze-out occurs nearly 30\% later in central U+U than
1080: in semiperipheral Pb+Pb collisions (see Fig.~\ref{F9}).
1081: 
1082: We note with surprise that the ``nutcracker'' pheno\-me\-non \cite{TS99} 
1083: is conspicuously missing in the U+U col\-li\-sions. We could not find
1084: it at lower and higher collision energies either. The origin of this
1085: difference between central U+U and peripheral Pb+Pb collisions will 
1086: be discussed in the following subsection.
1087: 
1088: In Fig.~\ref{F9} we compare the time evolutions of the three 
1089: characteristic quantities $\epsilon_x,\,\epsilon_p$, and 
1090: $\lda v_\perp\rda$ in central U+U and semiperipheral Pb+Pb 
1091: collisions, at SPS ($e_0$\,=\,9\,GeV/fm$^3$) and RHIC 
1092: ($e_0{\,=\,}175$\,GeV/fm$^3$) energies. We note that at freeze-out
1093: ($T_{\rm dec}{\,=\,}120$\, MeV) both systems give nearly the same 
1094: radial and elliptic flow, in spite of the different time 
1095: evolution: in the large system both flow types develop more slowly, 
1096: but over a longer time. This does not take into account that the
1097: flow gradients are smaller in the larger system, leading to later
1098: freeze-out at a lower temperature \cite{HS98}. This would not
1099: change the elliptic flow since $\epsilon_p$ has already saturated 
1100: (actually, it would lead to a very slight decrease of $\epsilon_p$, 
1101: see Fig.~\ref{F9}b). The radial flow $\lda v_\perp \rda$ would, 
1102: however, be somewhat larger. Since we enforced freeze-out at the 
1103: same value $T_{\rm dec}$, we don't see this.
1104: 
1105: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1106: \suse{What makes the nut crack?}
1107: \label{sec3c}
1108: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1109: 
1110: In this subsection we analyze two questions which so far remained 
1111: open: (1) Why does the ``nutcracker'' phenomenon arise in 
1112: semiperipheral Pb+Pb collisions, but not in central U+U collisions, 
1113: in spite of their identical initial deformation? (2) What is the 
1114: origin of the decrease of $\epsilon_p(\tau)$ before $\epsilon_x$ 
1115: passes through zero which is observed in Fig.~\ref{F9}b and curves 
1116: {\tt b} and {\tt c} of Fig.~\ref{F7}b?
1117: 
1118: To answer them requires a more detailed look at the time evolution
1119: of the transverse pressure gradients (cause) and transverse flow 
1120: profiles (effect). In Figs.~\ref{F10} and \ref{F11} we show a series
1121: of six snapshots each for semiperipheral Pb+Pb and central U+U 
1122: collisions, plotting the pressure and flow velocity profiles along 
1123: the $x$ and $y$ axis, respectively. The crucial difference between the 
1124: two collision systems is that in the semiperipheral Pb+Pb collisions
1125: the initial fireball contains a roughly 0.5\,fm thick {\em layer of
1126:   mixed phase matter with vanishing transverse flow velocity}; for
1127: central U+U collisions the initial ener\-gy density drops to zero so
1128: steeply that the mixed phase layer is initially practically absent.
1129: 
1130: As the matter begins to expand and dilute, a mixed phase layer begins
1131: to develop also in the U+U collisions; however, due to the buildup of
1132: transverse flow in the expanding matter, it is automatically created
1133: with a {\em nonvanishing} transverse flow velocity. Thus, even without
1134: pressure gradients inside the mixed phase which could accelerate it,
1135: the mixed phase matter flows in the transverse directions, with
1136: velocities exceeding those of the enclosed QGP matter (see
1137: Fig.\,\ref{F11}). The resulting transverse flow profiles are 
1138: monotonous functions of $x$ and $y$, with a selfsimilar (linear
1139: ``scaling'') pattern inside the mixed phase exactly as given by the
1140: analytic solution recently found by Bir\'o \cite{Biro}. The monotony
1141: of the transverse flow profiles is related one-to-one to the absence
1142: of the nutcracker phenomenon.
1143: 
1144: %%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 10 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1145: \begin{figure}[ht]
1146: \bce
1147: \epsfig{file=fig10a.ps,width=4.25cm}
1148: \epsfig{file=fig10b.ps,width=4.25cm}
1149: \epsfig{file=fig10c.ps,width=4.25cm}
1150: \epsfig{file=fig10d.ps,width=4.25cm}
1151: \epsfig{file=fig10e.ps,width=4.25cm}
1152: \epsfig{file=fig10f.ps,width=4.25cm}\\
1153: \vspace*{0.2cm}
1154: \caption{Transverse pressure (solid) and velocity (dashed) profile,
1155:   in $x$ (thick) and $y$ (thin) directions, for Pb+Pb collisions
1156:   at $b{\,=\,}7$\,fm. The 6 panels show snapshots at the indicated times.
1157:   The region of nearly constant pressure is in the mixed phase. The 
1158:   velocity profiles (dashed) are cut off at the freeze-out point. Initial
1159:   conditions as in Fig.~\ref{F6}.
1160:  \label{F10}}
1161: \ece
1162: \end{figure}
1163: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1164: 
1165: In the semiperipheral Pb+Pb collisions, on the other hand, the
1166: initially present mixed phase layer is at rest and, due to the lack of 
1167: pressure gradients, cannot accelerate itself in the transverse
1168: direction. As the transverse pressure gradients in the enclosed QGP
1169: matter begin to accelerate the QGP matter, the latter ``slams'' into
1170: the motionless mixed phase. This is clearly seen in the first four
1171: panels of Fig.\,\ref{F10} which show a strong radial increase of
1172: the transverse flow velocities inside the QGP phase, followed by a
1173: dramatic drop inside the mixed phase and a second rise in the HG
1174: matter near the edge. Inside the mixed phase the radial velocity
1175: profile is thus completely different from the selfsimilar scaling
1176: pattern seen in Fig.\,\ref{F11}. As time proceeds, this anomalous
1177: structure in the Pb+Pb collisions weakens, and the velocity profile
1178: begins to approach a scaling form inside the mixed phase; scaling 
1179: violations survive longest near the outer edge of the mixed phase
1180: layer. In the $y$ direction they disappear slightly earlier than
1181: in the shorter $x$ direction; this is the origin of the ``nutcracker
1182: phenomenon''. 
1183: 
1184: %%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 11 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1185: \begin{figure}[ht]
1186: \bce
1187: \epsfig{file=fig11a.ps,width=4.25cm}
1188: \epsfig{file=fig11b.ps,width=4.25cm}
1189: \epsfig{file=fig11c.ps,width=4.25cm}
1190: \epsfig{file=fig11d.ps,width=4.25cm}
1191: \epsfig{file=fig11e.ps,width=4.25cm}
1192: \epsfig{file=fig11f.ps,width=4.25cm}\\
1193: \vspace*{0.2cm}
1194: \caption{Same as Fig.~\ref{F10}, but for central U+U collisions
1195:   in the side-on-side configuration. Initial conditions as in 
1196:   Fig.~\ref{F8}.
1197:  \label{F11}}
1198: \ece
1199: \end{figure}
1200: \vspace{-0.7cm}
1201: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1202: 
1203: Now we can also understand the decrease of $\epsilon_p$ even before
1204: $\epsilon_x$ passes through zero: Figs.\,\ref{F7}, \ref{F10} and
1205: \ref{F11} show that this happens while most of the fireball is in the
1206: mixed phase. (Actually, $\epsilon_p$ begins to decrease while there is
1207: still a small QGP core in the center.) During this stage the matter
1208: expands essentially without transverse acceleration, featuring a 
1209: nearly selfsimilar transverse flow pattern. While it lasts, the
1210: selfsimilar flow dilutes the earlier developed momentum anisotropy
1211: $\epsilon_p$. This feature is therefore also directly related to the
1212: phase transition.   
1213: 
1214: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1215: \se{Experimental predictions}
1216: \label{sec4}
1217: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1218: 
1219: While the time-evolution of $\epsilon_x$, $\epsilon_p$ and 
1220: $\lda v_\perp \rda$ is interesting and helpful for an understanding
1221: of the relevant physical mechanisms, only the final values
1222: at freeze-out are observable (through the momentum spectra and, in 
1223: the case of $\epsilon_x$, possibly indirectly via two-particle momentum 
1224: correlations). The flow observables thus represent time-integrals over 
1225: the expansion history and EOS, and their measurement in a single 
1226: collision system at fixed beam energy provides very little information. 
1227: Using flow signatures as indicators for properties of the equation of 
1228: state for strongly interacting matter requires their measurement over a
1229: wide range of external control parameters, such as impact parameter, size 
1230: of the colliding nuclei, and beam energy. As discussed in the preceding
1231: section, a time-differential measurement is to some extent possible by
1232: comparing the radial and elliptic flow as functions of these parameters.
1233: 
1234: Flow anisotropies reflect themselves as non-vanishing higher order 
1235: Fourier coefficients in a Fourier expansion of the azimuthal dependence
1236: of the measured single-particle spectra around the beam direction 
1237: \cite{VZ96}:
1238:  \beq{vn}
1239:    v_n(y) = {\int_{-\pi}^\pi d\phi\,\cos(n\phi)\, {dN \over dy\, d\phi}
1240:              \over
1241:              \int_{-\pi}^\pi d\phi\, {dN \over dy\, d\phi}} \,,\quad
1242:    n=1,2,\dots  \,.
1243:  \eeq
1244: Since most experiments have limited $p_{_{\rm T}}$ acceptance, one 
1245: studies these coefficients also as functions of the transverse momentum:
1246:  \beq{vnpt}
1247:    v_n(y,p_{_{\rm T}}) = 
1248:    {\int_{-\pi}^\pi d\phi\,\cos(n\phi)\,
1249:    {dN \over dy\,p_{\rm T}\,dp_{\rm T}\, d\phi}
1250:    \over
1251:    \int_{-\pi}^\pi d\phi\,
1252:    {dN \over dy\,p_{\rm T}\,dp_{\rm T}\, d\phi}}
1253:    \,.
1254:  \eeq
1255: The $p_{_{\rm T}}^2$-weighted anisotropic flow coefficients are
1256: defined by
1257:  \beq{vnpt2}
1258:    v_{n,p_{\rm T}^2}(y) = 
1259:    {\int_{-\pi}^\pi d\phi\,\cos(n\phi)\, \int p_{\rm T}^2\, dp_{\rm T}^2
1260:     \, {dN \over dy\,dp_{\rm T}^2\, d\phi}
1261:     \over
1262:     \int_{-\pi}^\pi d\phi\,\int p_{\rm T}^2\, dp_{\rm T}^2
1263:    {dN \over dy\,dp_{\rm T}^2\, d\phi}}
1264:    \,.
1265:  \eeq
1266: In symmetric collision systems (which are the only ones we consider
1267: here) the odd order coefficients $v_1,v_3,\dots$ vanish at midrapidity 
1268: $y{\,=\,}0$ by symmetry. We here concentrate on the second harmonic 
1269: coefficient which is conventionally called ``elliptic flow''. The 
1270: $v_i$ are only defined at freeze-out but we already discussed how 
1271: $v_2$ and $v_{2,p_{\rm T}^2}$ can be related to $\epsilon_p$ which 
1272: is known also before freeze-out. 
1273: 
1274: $\epsilon_x$ and $\epsilon_p$ are functions of time; in the present 
1275: section, however, we only need the {\em initial} spatial deformation 
1276: $\epsilon_x(\tau_0)$ and the {\em final} momentum-space deformation 
1277: $\epsilon_p(\tau_{\rm f})$. For simplicity we will quote them as 
1278: $\epsilon_x$ and $\epsilon_p$, respectively, without the time 
1279: arguments.
1280: 
1281: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1282: \suse{$p_{\rm T}$-dependence of elliptic flow}
1283: \label{sec4a}
1284: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1285: 
1286: Since most experiments have a limited acceptance in transverse 
1287: momentum, the measured elliptic flow signal must be corrected for 
1288: the $p_{_{\rm T}}$-acceptance. In Fig.~\ref{F12} we show the 
1289: $p_{_{\rm T}}$-dependence of $v_2$ for pions and protons for 
1290: semiperipheral Pb+Pb and central U+U collisions. In spite of 
1291: their different masses, the predicted $v_2(p_{_{\rm T}})$ is 
1292: rather similar for the two particle species \cite{HL99}. At low
1293: $p_{_{\rm T}}$, the heavier protons show even a little less 
1294: elliptic flow than the pions. To the extent that hydrodynamics 
1295: is applicable, the larger $\la v_2\ra$ for protons than pions 
1296: measured by NA49 \cite{NA49flow} is thus predominantly due to 
1297: the different $p_{_{\rm T}}$-windows for the two particle 
1298: species (the proton elliptic flow was measured at higher 
1299: $p_{_{\rm T}}$ \cite{NA49flow}).
1300: 
1301: According to general arguments \cite{D95}, $v_2$ must vanish with 
1302: zero slope as $p_{_{\rm T}}\to 0$. We checked that this is true. 
1303: Fig.~\ref{F12} shows, however, that for pions the turnover from a 
1304: roughly linear behaviour at large $p_{_{\rm T}}$ to zero slope as 
1305: $p_{_{\rm T}}\to 0$ occurs at very small $p_{_{\rm T}}$-values,
1306: $p_{_{\rm T}}<0.1$\,GeV/$c$; for protons the corresponding scale is 
1307: somewhat larger. We have no quantitative analytic understanding of 
1308: this momentum scale but note that qualitatively similar behaviour was 
1309: found in \cite{BS00} using the kinetic UrQMD model.
1310: 
1311: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 12 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1312:  \begin{figure}[htbp]
1313:  \hspace*{-0.3cm}\epsfig{file=fig12.ps,width=8.4cm}
1314:  \caption{$p_{_{\rm T}}$-dependence of the elliptic flow coefficient
1315:      $v_2$ for pions (solid) and protons (dashed), for 158 $A$\,GeV/$c$
1316:      Pb+Pb collisions at $b{\,=\,}7$\,fm (left panel) and 
1317:      155 $A$\,GeV/$c$ U+U collisions at $b{\,=\,}0$ in the side-on-side 
1318:      configuration (right panel).
1319:  \label{F12}}
1320:  \end{figure}
1321: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1322: 
1323: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1324: \suse{Impact parameter dependence of elliptic flow}
1325: \label{sec4b}
1326: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1327: 
1328: As one changes the impact parameter, the initial spatial deformation
1329: $\epsilon_x$ of the transverse cross section through the reaction 
1330: zone varies as shown in Fig.~3 of Ref.~\cite{O92}. The stronger the 
1331: initial ellipticity, the stronger is the hydrodynamic response to it, 
1332: i.e. the larger are $v_2$ or $\epsilon_p$ at freeze-out. Ollitrault
1333: \cite{O92} showed that for an EOS with a constant velocity of sound, 
1334: $\P e/\P p$\,=\,const., the ratio $\epsilon_p/\epsilon_x$ or, 
1335: equivalently, $v_2/\epsilon_x$ is independent of the impact parameter 
1336: \cite{fn3}. (Ollitrault \cite{O92} used the variable $v_{2,p_{\rm T}^2}$ 
1337: which is closely related to $\epsilon_p$ \cite{fn2}. For pions $v_2$ 
1338: and $\epsilon_p$ are related by a factor 2 \cite{KSH99}.) This scaling 
1339: is broken only for very peripheral collisions which freeze out before 
1340: the elliptic flow builds up and saturates; thus in hydrodynamics 
1341: $v_2/\epsilon_x$ is constant over most of the impact parameter range.
1342: 
1343: A phase transition is characterized by a strong drop of the sound 
1344: velocity in the critical region (for a first order phase transition 
1345: the sound velocity vanishes in the mixed phase). It is therefore 
1346: interesting to reinvestigate the impact parameter dependence of 
1347: $v_2/\epsilon_x$ in the presence of a phase transition. The impact 
1348: parameter not only controls the initial spatial ellipticity of the 
1349: fireball, but also (with less variation) its initial energy density.
1350: At a given beam energy, it is therefore possible to probe the EOS 
1351: over a range of energy densities by varying the impact parameter. For
1352: a beam energy, at which in central collisions the initial energy density
1353: is not too far above the phase transition, it may thus be possible to
1354: study the effect of the reduced sound velocity near the phase transition 
1355: on the elliptic flow by changing the impact parameter. Weak structures
1356: in Fig.~9 of Ref.~\cite{O92} first indicated that the quark-hadron
1357: phase transition might thus become visible. Our analysis improves on
1358: that analysis by including resonance decays which tend to dilute the
1359: elliptic flow signature \cite{KSH99}.
1360: 
1361: In Fig.~\ref{F13} we study the impact parameter dependence of 
1362: $v_2/\epsilon_x$ in Pb+Pb collisions for three different initial 
1363: central energy densities: $e_0{\,=\,}25$ GeV/fm$^3$ (corresponding
1364: to a low energy run at RHIC), $e_0{\,=\,}9$ GeV/fm$^3$ (corresponding
1365: to collisions at the highest SPS energy of 158\,$A$\,GeV), and
1366: $e_0{\,=\,}4.5$ GeV/fm$^3$ (corresponding to lower SPS energies 
1367: around 40\,$A$\,GeV). The calculated total pion multiplicity densities
1368: at $b{\,=\,}0$ and midrapidity are 
1369: ${dN_\pi\over dy}\big\vert_{y=0}(b{\,=\,}0)$\,=\,859, 460, and 317, 
1370: respectively. For later comparison with Fig.~\ref{F14} we also 
1371: quote the corresponding rapidity densities for semiperipheral Pb+Pb
1372: collisions: ${dN_\pi\over dy}\big\vert_{y=0}(b{\,=\,}7\,{\rm fm})$ 
1373: =\,415, 220, and 148, respectively. 
1374: 
1375: %%%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 13 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1376:  \begin{figure}[htbp]
1377:  \epsfig{file=fig13.ps,width=8cm}
1378:  \caption{The ratio of the elliptic flow coefficient $v_2$ and
1379:     the initial spatial ellipticity $\epsilon_x$ as a function
1380:     of impact parameter $b$ for Pb+Pb collisions. Results for 
1381:     three values of the initial central energy density at 
1382:     $b{\,=\,}0$ ($e_0$\,=\,4.5, 9.0 and 25 GeV/fm$^3$) are shown.
1383:     Note the suppressed zero on the vertical axis. 
1384:  \label{F13}}
1385:  \end{figure}
1386: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1387: 
1388: Fig.~\ref{F13} shows that, at moderate impact parameters, the 
1389: largest elliptic flow is generated at the lowest of these three
1390: beam energies. At very large impact parameters (where hydrodynamics
1391: becomes doubtful) the generated elliptic flow naturally drops to 
1392: zero, since the overlap region and its initial energy density are 
1393: then too small and the matter freezes out before flow can develop. 
1394: What is interesting, however, is that at higher beam energies the 
1395: elliptic flow starts out lower than at $e_0{\,=\,}4.5$\,GeV/fm$^3$, 
1396: but then $v_2/\epsilon_x$ {\em rises} with increasing $b$. In fact, 
1397: for $e_0{\,=\,}9$\,GeV/fm$^3$ this ratio reaches at $b{\,=\,}11$\,fm 
1398: nearly the same value as for central collisions at 
1399: $e_0{\,=\,}4.5$\,GeV/fm$^3$.
1400: 
1401: The decrease with rising beam energy of $v_2/\epsilon_x$ at moderate 
1402: impact parameters was found \cite{KSH99} to result from the softening 
1403: of the EOS in the phase transition region. The soft matter near the 
1404: transition point inhibits the buildup of elliptic flow. Going at 
1405: fixed beam energy to larger impact parameters is like going at 
1406: fixed impact parameter to lower beam energies: in both cases the
1407: initial energy density in the collision zone is reduced, and 
1408: eventually the matter is dominated again by the relatively hard
1409: hadron gas. When read from right to left, the curves in Fig.~\ref{F13} 
1410: can thus be viewed as different projections of the excitation 
1411: function of elliptic flow which will be discussed below. We emphasize 
1412: in particular the rise of $v_2/\epsilon_x$ towards larger impact 
1413: parameters at the high SPS and the low RHIC energy: without a phase 
1414: transition this feature would be absent. Unfortunately, these 
1415: variations are small (at the level of a few percent), and very 
1416: accurate measurements are required to identify them.
1417: 
1418: Preliminary data from 158\,$A$\,GeV Pb+Pb collisions \cite{NA49flow}
1419: show a monotonous decrease of $v_2/\epsilon_x$ with increasing impact 
1420: parameter, instead of the nearly constant behaviour predicted by 
1421: hydrodynamics (see Fig.~\ref{F13}). For $b\to 0$, however, the data 
1422: seem to approach the hydrodynamic prediction. It is possible that 
1423: semiperipheral Pb+Pb collisions do not equilibrate quickly enough to 
1424: permit the elliptic flow to fully reach the hydrodynamic limit. Indeed, 
1425: kinetic simulations with the RQMD code \cite{S99,NA49flow,VP00},
1426: where the collision centrality is coupled to the degree of local
1427: thermalization, are able to qualitatively explain the observed
1428: decrease of $v_2/\epsilon_x$ with increasing impact parameter: more
1429: peripheral collisions lead to less equilibration and hence to a weaker
1430: elliptic flow response to the initial spatial ellipticity. When RQMD
1431: is modified to simulate an EOS with a quark-hadron phase transition
1432: \cite{S99}, the same generic decrease is superimposed on the rise of
1433: $v_2/\epsilon_x$ at large $b$ shown here (middle curve in
1434: Fig.~\ref{F13}); this results in a decrease of $(v_2/\epsilon_x)(b)$
1435: which is first steep, then flattens, then finally steepens again
1436: \cite{S99}. 
1437: 
1438: It is evident that a proper understanding of the interesting features
1439: in the impact parameter dependence of $v_2/\epsilon_x$ predicted
1440: in \cite{S99} for Pb+Pb collisions require the separation of 
1441: pre-equilibrium effects from those induced by the softening of the 
1442: EOS near the phase-transition. A collision system which is large 
1443: enough to ensure sufficiently rapid thermalization for hydrodynamics
1444: to apply would make life much easier. We therefore suggest to study 
1445: elliptic flow in side-on-side U+U collisions at zero impact parameter
1446: and search for the hydrodynamically predicted phase transition
1447: signatures in the beam energy dependence of elliptic flow.  
1448: 
1449: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1450: \suse{Beam energy dependence of elliptic flow}
1451: \label{sec4c}
1452: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1453: 
1454: The time-dependence of the flow patterns discussed in Sec.\,\ref{sec3} 
1455: reflects itself also in the beam energy dependence of elliptic
1456: flow. We already noted in \cite{KSH99} that the phase transition
1457: causes an non-monotonic excitation function for the elliptic flow
1458: coefficient $v_2$: as the collision energy is increased,
1459: $v_2$ first rises (at low energies the fireball freezes out before
1460: the elliptic flow can saturate) but then decreases again as the 
1461: initial energy density rises above the QGP threshold. We now
1462: understand that this decrease is intimately connected to the diluting
1463: effects of the selfsimilar fireball expansion in the mixed phase,
1464: even before the spatial deformation $\epsilon_x$ passes through zero
1465: (see the discussion in Sec.\,\ref{sec3c}.) Without a phase transition
1466: (EOS~H) this does not happen (see dash-dotted lines in
1467: Fig.\,\ref{F14}); the slight decrease of $v_2$ with EOS~H at
1468: asymptotically high energies has a different origin, namely a
1469: reduction of $\epsilon_p$ by the opposite sign of the spatial
1470: fireball anisotropy after $\epsilon_x$ has passed through zero.
1471: 
1472: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Fig. 14 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1473:  \begin{figure}[htbp]
1474:  \hspace*{-0.2cm}\epsfig{file=fig14a.ps,width=8.5cm}\\
1475:  \hspace*{0.15cm}\epsfig{file=fig14b.ps,width=8.5cm}\\
1476:  \caption{Excitation function of the elliptic flow coefficient $v_2$
1477:    (left vertical axis) and the radial flow $\lda v_\perp \rda/c$ 
1478:    (right vertical axis), for Pb+Pb collisions at $b{\,=\,}7$\,fm
1479:    (upper panel) and side-on-side U+U collisions at $b{\,=\,}0$
1480:    (lower panel). The horizontal axis gives the total pion
1481:    multiplicity density at midrapidity, ${dN_\pi\over
1482:      dy}\big\vert_{y=0}$, as a measure for the collision energy. 
1483:    Horizontal arrows indicate the regions covered by SPS, RHIC, 
1484:    and LHC. In the lower panel LHC would start around 5000. 
1485:  \label{F14}}
1486:  \end{figure}
1487: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1488: 
1489: The comparison of semiperipheral Pb+Pb collisions with central U+U
1490: collisions in the upper and lower panels of Fig.\,\ref{F14} shows that 
1491: this non-monotonic behaviour of the excitation function for $v_2$ is
1492: not sensitive to the exi\-stence of the ``nutcracker phenomenon'': the
1493: decrease of $v_2$ below its maximum in the SPS regime is only
1494: slightly weaker in the U+U case than for Pb+Pb, although only the
1495: latter features a ``cracking nut''. Since elliptic flow is a fragile
1496: phenomenon which is quite sensitive to incomplete thermalization, we
1497: believe that the most promising route towards experimental verification 
1498: of the phase transition signature suggested here is to study the 
1499: excitation function of $v_2$ in largest available deformed
1500: collision system, namely central side-on-side U+U collisions. 
1501:  
1502: In Ref.\,\cite{KSH99} we missed the fact that at asymptotically high
1503: energies the elliptic flow coefficient $v_2$ must approach the larger
1504: value corresponding to the stiffer QGP equation of state EOS~I. We 
1505: calculated in \cite{KSH99} the excitation function for $b{\,=\,}7$\,fm 
1506: Pb+Pb collisions only up to multiplicity densities 
1507: ${dN_\pi\over dy}\big\vert_{y=0}{\,=\,}500$ and concluded prematurely 
1508: that $v_2$ saturates at high collision energies at a value below the 
1509: value corresponding to EOS~H. Fig.\,\ref{F14} extends the excitation 
1510: functions for both Pb+Pb and U+U collisions to LHC energies and 
1511: demonstrates that $v_2$ begins to rise again, eventually approaching 
1512: the EOS~I limit. The dip, which indicates the presence of the phase 
1513: transition, thus only covers the energy range between SPS and RHIC. Note 
1514: that in the same energy region also the radial flow $\lda v_\perp\rda$ 
1515: (dashed lines in Fig.\,\ref{F14}) is predicted to grow more slowly 
1516: with $\sqrt{s}$ than at lower and higher beam energies where the 
1517: expansion is predominantly driven by pure HG or pure QGP matter. 
1518: 
1519: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1520: \suse{Elliptic flow as an estimator for the thermalization time scale}
1521: \label{sec4d}
1522: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1523: 
1524: Throughout this paper we have assumed early thermalization followed 
1525: by hydrodynamic expansion. For a given initial deformation of 
1526: the collision zone in the transverse plane (which can be calculated 
1527: from geometry once the impact parameter is known, for example by a 
1528: measurement of the number of spectator nucleons), this guarantees the 
1529: maximum possible momentum-space response in the form of elliptic 
1530: flow. Any delay in the thermalization process will lead to a reduction 
1531: of the elliptic flow: even without secondary collisions the spatial
1532: deformation of the region occupied by the produced particles decreases
1533: by free-streaming, and if thermalization effectively sets in later,
1534: the resulting anisotropies in the pressure gradients will be smaller,
1535: leading to less elliptic flow.
1536: 
1537: We can use the above demonstrated fact that, up to variations of the 
1538: order of 20\%, the hydrodynamic response $v_2$ to the elliptic spatial 
1539: deformation at thermalization is essentially constant: 
1540: $v_2^{\rm hydro}/\epsilon_x\approx$\,const.\,$\approx 0.25$. This allows 
1541: to interpret the measured $v_2$ in terms of an {\em effective} initial 
1542: spatial deformation at the point of thermalization, i.e. at the beginning of 
1543: the hydrodynamic evolution. It is clearly not a good approximation to 
1544: idealize the initial kinetic equilibration stage of the collision by 
1545: a stage of collisionless free-streaming followed by hydrodynamic 
1546: expansion, thereby assuming a sudden, but delayed transition from 
1547: a non-equilibrium initial state to a fully thermalized fluid. Still,
1548: this simple-minded picture can be used to obtain a rough first 
1549: order-of-magnitude guess of the thermalization time scale, based on 
1550: a measurement of $v_2$.
1551: 
1552: To this end we note that under free-streaming the phase-space 
1553: distribution evolves as
1554:  \beq{freestr}
1555:    f(\bbox{r},\bbox{p},t) = 
1556:    f\left(\bbox{r}-{\bbox{p}\over E}(t-t_0),\bbox{p},t_0\right)\,.
1557:  \eeq
1558: Using a Gaussian parametrization for the initial phase-space distribution
1559: of produced secondary particles,
1560:  \beq{Gauss}
1561:    f(\bbox{r},\bbox{p},t_0) = 
1562:    \exp\left[ -{x^2\over 2R_x^2}-{y^2\over 2R_y^2}
1563:               -{p_x^2+p_y^2\over2\Delta^2}\right]\,,
1564:  \eeq
1565: one easily finds
1566:  \bea{eps}
1567:   \epsilon_x(t) &=& 
1568:   {\int d^2r\,r^2\cos(2\phi_r) \int d^3p f(\bbox{r},\bbox{p},t)
1569:    \over
1570:    \int d^2r\,r^2 \int d^3p f(\bbox{r},\bbox{p},t)}
1571:  \nonumber\\
1572:   &\approx& \epsilon_x(t_0)\, {R_x^2+R_y^2
1573:                                \over R_x^2+R_y^2 + 2 (c\Delta t)^2}\,,
1574:  \eea
1575: where $\Delta t = t{-}t_0$ is the time delay between particle formation 
1576: and thermalization. Assuming that $\epsilon_x(t_0{+}\Delta t)$ can be 
1577: obtained from the measured $v_2$ by dividing by $\approx\,0.25$, we can 
1578: extract $\Delta t$ by rewriting 
1579: (\ref{eps}) as
1580:  \beq{ratioeps}
1581:    {\epsilon_x(t_0+\Delta t)\over \epsilon_x(t_0)} =
1582:    \left[ 1 + {(c\Delta t)^2 \over R^2 (1+\delta^2)}\right]^{-1}\,,
1583:  \eeq
1584: where $\delta$ parametrizes the initial deformation via $R_x=R(1{-}\delta)$,
1585: $R_y=R(1{+}\delta)$ such that $\epsilon_x(t_0) = 2\delta/(1{+}\delta^2)$.
1586: 
1587: Inserting appropriate values for $R$ and $\delta$ one finds that for
1588: Pb+Pb collisions at $b{\,=\,}7$\,fm a dilution by 50\% of the elliptic flow 
1589: signal by initial free-streaming requires a time-delay of order 3.5 fm/$c$
1590: until thermalization sets in; for central U+U collisions in the 
1591: side-on-side configuration $\Delta t\,\approx\,5$\,fm/$c$ of approximate
1592: free-streaming would be required to dilute the elliptic flow signal by 50\%.
1593: This (admittedly rough) exercise demonstrates two points: (i) U+U 
1594: collisions provide the better chance to observe the full hydrodynamic 
1595: elliptic flow signal, and (ii) the observation of less elliptic flow than
1596: hydrodynamically expected can be used to obtain a rough estimate of 
1597: the thermalization time scale in the initial collision stage.
1598: 
1599: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1600: \se{Summary}
1601: \label{sec5}
1602: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1603: 
1604: On the basis of hydrodynamic simulations we analyzed the sensitivity of
1605: radial and elliptic transverse flow at midrapidity to the quark-hadron 
1606: phase transition. We modelled this phase transition as a strongly first 
1607: order phase transition with a latent heat of about 1.15 GeV/fm$^3$. 
1608: It manifests itself dynamically as an expan\-ding shell of mixed phase 
1609: matter inside which all pressure gradients and thus all hydrodynamic
1610: acceleration forces vanish. Compared to the situation of a pure HG or
1611: a pure QGP phase this leads to a reduction of both radial and elliptic 
1612: flow. Elliptic flow, as the more fragile phenomenon which is generated
1613: only by azimuthal anisotropies in the pressure gradients, shows a larger
1614: sensitivity to the phase transition than radial flow. Also, since we 
1615: showed that it saturates well before freeze-out, it more directly 
1616: reflects the EOS during the early and dense stages of the expansion. 
1617: 
1618: As a tell-tale signature for the phase transition we predict a 
1619: non-monotonic excitation function for the elliptic flow coefficient 
1620: $v_2$ as shown in Fig.~\ref{F14}. In the present paper we explored
1621: in great detail the origin of the dip in $v_2$, which we predict 
1622: to occur in the energy region between the SPS and RHIC, by performing 
1623: a careful ana\-ly\-sis of the space-time evolution of the anisotropic 
1624: transverse flow pattern for a variety of collision energies. As the 
1625: dynamical origin of the phase transition signature in $v_2$ we 
1626: identified the existence of a large subvolume of mixed phase matter 
1627: which undergoes nearly selfsimilar, acceleration-free expansion while 
1628: it lasts. In addition to the $v_2$ excitation function it leaves 
1629: traces in the impact parameter dependence of the response 
1630: $v_2/\epsilon_x$ of the elliptic flow to the initial spatial 
1631: deformation of the collision zone, and in the (not directly 
1632: measurable) time evolution of the flow anisotropy $\epsilon_p$.
1633: 
1634: When colliding spherical nuclei with each other, the measurement of
1635: elliptic flow requires selecting collisions at rather large impact
1636: parameters ($b\gtrsim5$\,fm) in order to achieve a sufficiently large
1637: initial spatial deformation of the nuclear overlap region (reaction 
1638: zone). Correspondingly the overall size of the elliptically deformed, 
1639: expanding fireball is small, and one may doubt the applicability of 
1640: our hydrodynamic approach. We here point out that central U+U 
1641: collisions in the side-on-side configuration provide nearly twice 
1642: larger collision volumes at similar deformation as Pb+Pb collisions 
1643: at $b{\,=\,}7$\,fm and should thus exhibit hydrodynamic behaviour 
1644: much more clearly.
1645: 
1646: We therefore carefully compared central side-on-side U+U collisions 
1647: with semipe\-ri\-phe\-ral Pb+Pb collisions at all collision 
1648: energies. We showed that the phase transition signature in the $v_2$
1649: excitation function manifests itself similarly in both
1650: collision systems. The U+U system should thus be preferred for 
1651: its presumed better hydrodynamical behaviour and for the larger
1652: particle multiplicities which improve the statistics of elliptic flow
1653: measurements. The phase transition signal appears to be slightly 
1654: stronger in the smaller Pb+Pb system; we were able to trace this to
1655: the ``nutcracker phenomenon'' of Shuryak and Teaney \cite{TS99} which,
1656: unfortunately, only occurs in the Pb+Pb system. In trying to understand 
1657: the fragility of ``nutcracker flow'' we found that it crucially relies
1658: on the existence of a rather thick shell of mixed phase matter
1659: {\em at rest} in the initial state of fireball expansion, which 
1660: surrounds a significant core of QGP. In response to internal 
1661: pressure gradients the QGP core starts to expand and ``slams'' into 
1662: the surrounding shell of mixed phase at rest. This cannot happen in
1663: central U+U collisions since there the initial transverse energy 
1664: density profile drops to zero so steeply that no visible mixed phase 
1665: shell forms.
1666: 
1667: We thus conclude that the interesting ``nutcracker flow'' phenomenon 
1668: constitutes a very fragile variant of anisotropic flow which is not 
1669: generated in central U+U collisions. If the fireballs formed in 
1670: semiperipheral Pb+Pb collision should turn out to be too small to 
1671: achieve sufficient local thermalization for hydrodynamics to work, it
1672: may be unmeasurable. Fortunately, the elliptic flow signature for
1673: the phase transition is more robust and does not require the actual
1674: ``cracking of the nut''; it should be clearly visible in central U+U
1675: collisions.
1676: 
1677: This raises the question how to experimentally select the side-on-side 
1678: collision geometry. By requiring zero spectators one can trigger on 
1679: configurations in which the colliding nuclei overlap completely in the
1680: transverse plane. This still allows for arbitrary, but (up to a sign) 
1681: equal angles ($\theta_1=\pm \theta_2$) between the beam direction and 
1682: the long axes of the two deformed nuclei. The interesting side-on-side 
1683: configuration corresponds to $\theta_1{\,=\,}\theta_2{\,=\,}90^\circ$. 
1684: Since this configuration has the largest initial spatial deformation 
1685: in the transverse plane, it generates the largest elliptic flow $v_2$; 
1686: therefore, Shuryak \cite{Sh00} suggested a cut on large $v_2$ to select 
1687: the side-on-side collision geometry. Unfortunately, the event-by-event 
1688: fluctuations of $v_2$ are so large that this off-line trigger is not 
1689: expected to be very efficient \cite{Posk}; furthermore, it would 
1690: introduce an inconvenient trigger bias into our suggested investigation
1691: of the dependence of $v_2$ on various control parameters.    
1692: 
1693: We have not been able to come up with a more efficient selection 
1694: criterium. We checked that with initial conditions calculated according 
1695: to (\ref{init}), the produced charged particle multiplicity densities at
1696: midrapidity vary by less than 5\% between tip-on-tip and side-on-side
1697: collisions (with side-on-side collisions producing more particles, with 
1698: slightly smaller $\langle p_T\rangle$ at freeze-out). Again this 
1699: difference is well below the expected level of event-by-event 
1700: fluctuations. Its smallness is explained by the fact that with the 
1701: {\em Ansatz} (\ref{init}) the amount of entropy $dS/dy$ stopped at 
1702: midrapidity is essentially independent of the orientation 
1703: $\theta_1=\pm\theta_2$ (for 0$^\circ$ it is 1.3\% larger than for 
1704: 90$^\circ$), and boost-invariant longitudinal expansion conserves 
1705: $dS/dy$. At higher collision energies minijet production may overtake
1706: the soft particle production processes implicitly assumed in (\ref{init});
1707: instead of scaling with the number of wounded nucleons as in (\ref{init}),
1708: minijet production scales with the number of nucleon-nucleon collisions,
1709: involving the product rather than the sum of the nuclear thickness 
1710: functions appearing in (\ref{init}). In this case tip-on-tip collisions
1711: are expected to generate considerably more entropy in the transverse
1712: plane at midrapidity than side-on-side collisions, and one could trigger
1713: on the latter by selecting for zero spectators combined with low 
1714: $dN/dy(y{=}0)$. --- In the absence of an efficient trigger for side-on-side
1715: U+U collisions at present-day collision energies one will be forced 
1716: to compare with data which are averaged over all orientations 
1717: $\theta_1=\pm\theta_2$. The computation of an orientation-averaged 
1718: excitation function for $v_2$ is, however, numerically expensive;
1719: we therefore postpone it until experiments involving U+U are approved. 
1720: 
1721: Our prediction of a dip in the excitation function of $v_2$ at 
1722: midrapidity is directly related to the one by Rischke {\it et al.} 
1723: \cite{Rischke95} of a dip in the excitation function for directed 
1724: flow at forward and backward rapidities: both rely on the softening 
1725: of the EOS near the phase transition which results in reduced 
1726: hydrodynamic pressure gradients. We point out, however, that, as 
1727: the collision energy increases, the time interval during which 
1728: directed flow is generated (the nuclear transition time) becomes 
1729: shorter and shorter, and the prospects for sufficiently fast local 
1730: thermalization to validate hydrodynamic concepts thus become
1731: {\em worse and worse}. The opposite is true for elliptic flow: 
1732: Figs.~\ref{F7}  and \ref{F9} show that the time interval over 
1733: which elliptic flow builds up approaches at high collision energies 
1734: a finite limit of about 7\,fm/$c$ for semiperipharal Pb+Pb and about
1735: 12\,fm/$c$ for central U+U collisions. The density of produced 
1736: particles, on the other hand, continues to increase, leading to 
1737: shorter and shorter thermalization times. The hydrodynamic 
1738: description of elliptic flow buildup should thus become {\em better} 
1739: with increasing collision energy. 
1740:  
1741: We finally comment on the sensitivity of the proposed phase
1742: transition signature to our simple modelling of the phase transition: 
1743: we used a Maxwell construction between the HG and QGP equations of state, 
1744: leading to a strong first order phase transition with large latent 
1745: heat. We don't believe that smoothing the phase transition to a rapid 
1746: crossover will qualitatively alter our results: the only major change 
1747: will be a replacement of the acceleration-free mixed phase by a 
1748: transition region with non-zero, but nevertheless small pressure 
1749: gradients. However, since elliptic flow signals are generically weak
1750: and the predicted effects from the phase transition are at a level
1751: of only about 10\% of this signal, further hydrodynamic simulations 
1752: using a more realistic modelling of the EOS may be required for a 
1753: reliable quantitative assessment of the expected experimental signal.
1754:    
1755: \acknowledgments
1756: 
1757: We gratefully acknowledge fruitful discussions with Tamas Bir\'o, 
1758: Pasi Huovinen, Art Poskanzer, and Ser\-gei Voloshin. Our hydrodynamic 
1759: code is a (2+1)-dimensional generalization of a (1+1)-dimensional 
1760: algorithm for central collisions which was originally developed by 
1761: M.~Kataja, P.V.~Ruuskanen, R.~Venugopalan, and P.~Huovinen. We thank 
1762: these colleagues for allowing us to modify their code for non-central 
1763: collisions. This work was supported in part by BMBF, DFG and GSI.
1764: 
1765: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% APPENDIX %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1766: \appendix
1767: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1768: \se{Implementation of boost invariance}
1769: \label{appa}
1770: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1771: 
1772: An elegant method of introducing longitudinal boost invariance with 
1773: the longitudinal velocity field $v_z=z/t$ makes use of the notation 
1774: of general covariant derivatives. 
1775: 
1776: In an arbitrary coordinate system the equations of motion can be 
1777: written as 
1778:  \beq{A1}
1779:    {T^{mn}}_{;m} = 0\,, \qquad 
1780:    {j^m}_{;m} = 0\,,
1781:  \eeq
1782: where the semicolon indicates a covariant derivative. For tensors of 
1783: rank 1 and 2 it reads explicitly
1784:  \bea{A2}
1785:    j^i_{;p} &=& j^i_{,p} + \Gamma^i_{pk}\, j^k \,,
1786:  \\
1787:  \label{A3}
1788:    {T^{ik}}_{;p} &=& {T^{ik}}_{,p} + \Gamma^i_{pm} T^{mk} 
1789:                                    + \Gamma^k_{pm} T^{im}\,,
1790:  \eea
1791: where the komma denotes a simple partial derivative and the Christoffel 
1792: symbols $\Gamma^s_{ij}$ are given by derivatives of the metric tensor
1793: $g^{ab}(x)$:
1794:  \beq{A4}
1795:     \Gamma^s_{ij} =
1796:     \half g^{ks} \bigl( g_{ik,j} + g_{jk,i} - g_{ij,k}\bigr)\,.
1797:  \eeq
1798: We use this with the following transformation from Cartesian to light 
1799: cone coordinates:
1800:  \bea{A5}
1801:    x^\mu=(t,x,y,z) & \longrightarrow  & \bar{x}^m=(\tau,x,y,\eta)
1802:  \nonumber\\
1803:    t = \tau \cosh\eta &     &  \tau = \sqrt{t^2-z^2}
1804:  \\
1805:    z = \tau \sinh\eta &     &  \eta = \half \ln{t{+}z\over t{-}z} \,.
1806:  \eea
1807: In the new coordinate system the velocity field (after inserting $v_z=z/t$)
1808: is given by
1809:  \beq{A6}
1810:    \bar{u}^m = \bar{\gamma}(1,\bar{v}_x,\bar{v}_y,0)
1811:  \eeq
1812: with $\bar{v}_i \equiv v_i\cosh\eta$, $i=x,y$, and
1813: $\bar{\gamma} \equiv 1/{\textstyle{\sqrt{1{-}\bar{v}_x^2{-}\bar{v}_y^2}}}$.
1814: 
1815: Now we turn to the metric of the new system. We have 
1816:  \bea{A7}
1817:    ds^2 = g_{\mu \nu} dx^\mu dx^\nu 
1818:        &=& dt^2 - dx^2 - dy^2 - dz^2
1819:   \nonumber \\
1820:        &=& d\tau ^2 - dx^2 - dy^2 - \tau^2 d\eta^2
1821:  \eea
1822: and therefore
1823:  \beq{A8}
1824:    g_{mn}=\left( \begin{array}{*{4}{c}} 
1825:         1 & 0 & 0 & 0 \\ 
1826:         0 & -1 & 0 & 0 \\
1827:         0 & 0 & -1 & 0 \\ 
1828:         0 & 0 & 0 & -\tau^2 \\
1829:         \end{array} \right)\,,
1830:  \eeq
1831: The only non-vanishing Christoffel symbols are
1832:  \beq{A9}
1833:    \Gamma^{\eta}_{\eta \tau} = 
1834:    \Gamma^{\eta}_{\tau \eta} = {1\over\tau}\,,\qquad
1835:    \Gamma^\tau_{\eta \eta} = \tau\,.
1836:  \eeq
1837: Finally, by making use of the relations
1838: $T^{\tau i}= \bar{v}_i T^{\tau\tau}+\bar{v}_i p$ and 
1839: $T^{\eta \eta}= p /\tau^2$ the energy-momentum conservation equations
1840: (\ref{A1}) turn for $n=\tau,x,y,\eta$ into
1841:  \begin{mathletters}
1842:  \label{A10}
1843:  \bea{A10a}
1844:  \FL
1845:    &&{T^{\tau\tau}}_{,\tau} + (\bar{v}_x T^{\tau\tau})_{,x}
1846:      + (\bar{v}_y T^{\tau\tau})_{,y} =
1847:  \\
1848:    &&\qquad = -{p{+}T^{\tau\tau}\over \tau} - (p\, \bar{v}_x)_{,x}
1849:               - (p\, \bar{v}_y)_{,y} \,,
1850:  \nonumber\\
1851:  \label{A10b}
1852:    &&{T^{\tau x}}_{,\tau} + (\bar{v}_x T^{\tau x})_{,x}  
1853:      + (\bar{v}_y T^{\tau x})_{,y} =
1854:  \\
1855:    &&\qquad = - p_{,x} - {T^{\tau x}\over \tau}\,,
1856:  \nonumber\\
1857:  \label{A10c}
1858:    &&{T^{\tau y}}_{,\tau} + (\bar{v}_x T^{\tau y})_{,x}
1859:      + (\bar{v}_y T^{\tau y})_{,y} =
1860:  \\
1861:    &&\qquad = - p_{,y} - {T^{\tau y}\over \tau}\,,
1862:  \nonumber\\
1863:  \label{A10d}
1864:    &&{1\over \tau^2}\, p_{,\eta} = 0\,,
1865:  \eea
1866:  \end{mathletters}
1867: while the current conservation (\ref{A1}) becomes
1868:  \beq{A10e}
1869:    {j^\tau}_{,\tau} + (\bar{v}_x j^\tau)_{,x} + (\bar{v}_y j^\tau)_{,y}
1870:    = - {j^\tau\over \tau}\,.
1871:  \eeq
1872: We note the explicit appearance of $\tau$ on the r.h.s. of the 
1873: differential equations, reflecting the dilution of the matter
1874: due to the boost-variant longitudinal expansion. Connected with 
1875: this is the initial equilibration time $\tau_0$ as one of the 
1876: model parameters. Equation (\ref{A10d}) expresses the fact that, 
1877: due to longitudinal boost-invariance, the evolution is 
1878: $\eta$-independent.
1879: 
1880: Multiplying these equations by $\tau$ and introducing the scaled quantities
1881: $\tilde{\jmath}^\mu=\tau j^\mu$, $\tilde{T}^{\mu\nu}=\tau T^{\mu \nu}$, and
1882: $\tilde{p}=\tau\,p$ leads to the simple form (\ref{DGL}).
1883: 
1884: %%%%%%%%%%%%%%%%%%%%% Literature %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1885: 
1886: \begin{thebibliography}{99}
1887: 
1888: \bibitem{LHW00}
1889:   M.A. Lisa, U. Heinz, and U.A. Wiedemann, {\tt nucl-th/ 0003022}.
1890: \bibitem{flowrev}
1891:   A comprehensive list of references can be found in the recent review 
1892:   by N. Herrmann, J.P. Wessels, and T. Wienold, \annrev{49}{1999}{581}.
1893: \bibitem{lattice}
1894:   E. Laermann, \npa{610}{1996}{1c}; F. Karsch, {\tt hep-lat/9909006}.
1895: \bibitem{Rischke95}
1896:   D.H. Rischke, Y. P\"urs\"un, J.A. Maruhn, H. St\"ocker, and 
1897:   W. Greiner, \hip{1}{1995}{309}.
1898: \bibitem{S97}
1899:   H.~Sorge, \prevl{78}{1997}{2309}.
1900: \bibitem{S99}
1901:   H.~Sorge, \prevl{82}{1999}{2048}.
1902: \bibitem{St82}
1903:   H. St\"ocker {\it et al.}, \prevc{25}{1982}{1873}.
1904: \bibitem{O92}
1905:   J.Y.~Ollitrault, \prevd{46}{1992}{229}.
1906: \bibitem{O98}
1907:   J.Y.~Ollitrault, \npa{638}{1998}{195c}.
1908: \bibitem{HL99}
1909:   H. Heiselberg and A.-M. Levy, \prevc{59}{1999}{2716}.
1910: \bibitem{KSH99}
1911:   P.F.~Kolb, J.~Sollfrank, and U.~Heinz, \plb{459}{1999}{667};
1912:   P.F.~Kolb, J.~Sollfrank, P.V.~Ruuskanen and U.~Heinz,
1913:   \npa{661}{1999}{349c}.
1914: \bibitem{TS99}
1915:   E.V. Shuryak, \npa{661}{1999}{119c};
1916:   D.~Teaney and E.V.~Shuryak, \prevl{83}{1999}{4951}.
1917: \bibitem{SZ78}
1918:   E.V. Shuryak and O.V. Zhirov, \sjnp{28}{1978}{247}; 
1919:   and \plb{89}{1980}{253}. 
1920: \bibitem{vH82}
1921:   L. van Hove, \plb{118}{1982}{138}.
1922: \bibitem{KRLG86}
1923:   M. Kataja, P.V. Ruuskanen, L.D. McLerran, and H. von Gersdorff,
1924:   \prevd{34}{1986}{2755}.
1925: \bibitem{S97a}
1926:   H. Sorge, \plb{402}{1997}{251}.
1927: \bibitem{NA49flow}
1928:   NA49 Collaboration: H. Appelsh\"auser {\it et al.}, 
1929:   \prevl{80}{1998}{4136}; A.M. Poskanzer, S.A. Voloshin {\it et al.}, 
1930:   \npa{661}{1999}{341c}. 
1931: \bibitem{Sh00}
1932:   E.V. Shuryak, \prevc{61}{2000}{034905}.
1933: \bibitem{Li00}
1934:   Bao-An Li, \prevc{61}{2000}{021903(R)}.
1935: \bibitem{BGHG99}
1936:   S.A. Bass, M. Gyulassy, H. St\"ocker, and W. Greiner, \jpg{25}{1999}{R1}.
1937: \bibitem{HJ00}
1938:   CERN Press Release Feb.\,10, 2000: 
1939:   {\tt http://cern.web. cern.ch/CERN/Announcements/2000/NewStateMatter/};
1940:   U.~Heinz and M.~Jacob, {\tt nucl-th/0002042}.
1941: \bibitem{Bj83}
1942:   J.D.~Bjorken, \prevd{27}{1983}{140}.
1943: \bibitem{CF74}
1944:   F.~Cooper and G.~Frye, \prevd{10}{1974}{186}.
1945: \bibitem{VP00}
1946:   S.A. Voloshin and A.M. Poskanzer, \plb{474}{2000}{27}.
1947: \bibitem{Hirano}
1948:   T.~Hirano, {\tt nucl-th/9904082} and {\tt nucl-th/0004029}. 
1949: \bibitem{Netal99}
1950:   C.~Nonaka, N.~Sasaki, S.~Muroya, and O.~Miyamu\-ra, \npa{661}{1999}{353c};
1951:   C.~Nonaka, S.~Mu\-ro\-ya, and O.~Miyamura,
1952:   {\tt nucl-th/9907045}; 
1953:   K.~Mo\-ri\-ta, S.~Muroya, H.~Nakamura, and C.~Nonaka, 
1954:   \prevc{61}{2000}{034904}.
1955: \bibitem{Cs94}
1956:   L.P.~Csernai, {\it Introduction to relativistic heavy ion 
1957:   collisions} (Wiley{\&}Sons, Chichester (UK), 1994).
1958: \bibitem{LRH86}
1959:   K.S. Lee, M.J. Rhoades-Brown, and U. Heinz, \plb{174}{1986}{123}; and 
1960:   \prevc{37}{1988}{1452}.
1961: \bibitem{Setal97}
1962:   J.~Sollfrank {\it et al.}, \prevc{55}{1997}{392}
1963: \bibitem{Betal00}
1964:   J.~Brachmann {\it et al.}, \prevc{61}{2000}{024909}.
1965: \bibitem{BB73} 
1966:   J.P.~Boris, D.L.~Book, \jcomp{11}{1973}{38};
1967:   J.P.~Boris, D.L.~Book, and K.~Hain, \ibid{18}{1995}{248}.
1968: \bibitem{BO90}
1969:   J.-P.~Blaizot and J.-Y.~Ollitrault, in {\it Quark-Gluon Plasma} 
1970:   (R.C.~Hwa, ed.), Adv. Series on Directions in High Energy Physics 
1971:   {\bf 6}, p.393 (World Scientific, Singapore, 1990).
1972: \bibitem{RBM95}
1973:   D.H.~Rischke, S.~Bernard, and J.A.~Maruhn, \npa{595}{1995}{346};
1974:   D.H.~Rischke, Y.~P\"urs\"un, and J.A.~Ma\-ruhn, {\it ibid.}, p.~383.
1975: \bibitem{BM69} 
1976:   A.~Bohr and B.R.~Mottelson, {\it Nuclear Structure}
1977:   (Benjamin, New York, 1969).
1978: \bibitem{fn1}
1979:   Note that the box distribution radii used in \cite{Sh00} are 4.6\% 
1980:   larger than the Woods-Saxon radius parameters used by us. The values 
1981:   for $R_{l,s}$ given in \cite{Sh00} were scaled down by this factor.
1982: \bibitem{NA49spec}
1983:   NA49 Collaboration, H.~Appelsh\"auser {\it et al.}, \prevl{82}{1999}{2471}.
1984: \bibitem{fn1a}
1985:   For an approximately Gaussian initial transverse energy density profile
1986:   $e_0{\,=\,}9$\,GeV/fm$^3$ at $\tau_0{\,=\,}0.8$\,fm/$c$ corresponds to 
1987:   an average energy density $\bar e(0.8$\,fm$/c)=$ 4.5\,GeV/fm$^3$.
1988:   Due to longitudinal Bjorken expansion this decreases with $\tau^{-4/3}$
1989:   to $\bar e(1$\,fm$/c){\,=\,}3.34$\,GeV/fm$^3$ at $\tau{\,=\,}1$\,fm/$c$.
1990:   This value is consistent with $\bar e(1$\,fm/$c)=$ $(3.2\pm0.3)$\,GeV/fm$^3$ 
1991:   as estimated from the measured transverse energy $dE_T/dy$ using 
1992:   Bjorken's formula \cite{e0}.
1993: \bibitem{e0}
1994:   NA49 Collaboration, T. Alber {\it et al.}, \prevl{75}{1995}{3814};
1995:   WA98 Collaboration, T. Peitzmann {\it et al.}, \npa{610}{1996}{200c}.
1996: \bibitem{BGZ78}
1997:   J. Bondorf, S. Garpman, and J. Zim\'anyi, \npa{296}{1978}{320}.
1998: \bibitem{HLR87}
1999:   U. Heinz, K.S. Lee, and M.J. Rhoades-Brown, \prevl{58}{1987}{2292};
2000:   K.S. Lee, M.J. Rhoades-Brown, and U. Heinz, \prevc{37}{1988}{1463}.
2001: \bibitem{LHS89} 
2002:   K.S. Lee, U. Heinz, and E. Schnedermann, \zpc{48}{1989}{525}.
2003: \bibitem{SH92}
2004:   E.~Schnedermann and U.~Heinz, \prevl{69}{1992}{2908} and
2005:   \prevc{50}{1994}{1675};
2006:   E.~Schnedermann, J.~Sollfrank, and U.~Heinz, in {\it Particle 
2007:   Production in Highly Excited Matter} (H.H.~Gutbrod and J.~Rafelski,
2008:   eds.), NATO Asi Series {\bf B 303} (1993) 175 (Plenum, New York).
2009: \bibitem{MH97}
2010:   U. Mayer and U. Heinz, \prevc{56}{1997}{439}.  
2011: \bibitem{BMGR96}
2012:   S.~Bernard, J.A.~Maruhn, W.~Greiner, and D.H.~Rischke, \npa{605}{1996}{566};
2013:   D.H.~Rischke, in {\it Hadrons in Dense Matter and Hadrosynthesis} 
2014:   (J.~Cleymans, H.B. Geyer, and F.G. Scholz, eds.), Springer Lecture 
2015:   Notes in Physics {\bf 516}, 21 (1999).
2016: \bibitem{Bu96}
2017:   K.A. Bugaev, \npa{606}{1996}{559};
2018:   K.A. Bugaev and M.I. Gorenstein, {\tt nucl-th/9903072};
2019:   K.A. Bugaev, M.I. Gorenstein, and W. Greiner, \jpg{25}{1999}{2147}.
2020: \bibitem{SKH91}
2021:   J.~Sollfrank, P.~Koch, and U.~Heinz, \zpc{52}{1991}{593};
2022:   U.A.~Wiedemann and U. Heinz, \prevc{56}{1997}{3265}.
2023: \bibitem{Kolb}
2024:   P.F.~Kolb, Diploma thesis, University of Regensburg, 1999, unpublished.
2025: \bibitem{Ketal97}
2026:   B.~K\"ampfer, A.~Peshier, O.P.~Pavlenko, M.~Hentschel, and G.~Soff,
2027:   \jpg{23}{1997}{2001}.
2028: \bibitem{HS98}
2029:   C.M.~Hung and E.V. Shuryak, \prevc{57}{1998}{1891}.
2030: \bibitem{NA49HBT}
2031:   NA49 Collaboration, H. Appelsh\"auser {\it et al.}, \epjc{2}{1998}{661}.
2032: \bibitem{TWH99}
2033:   B.~Tom\'a\v{s}ik, U.A.~Wiedemann, and U.~Heinz, {\tt nucl-th/ 9907096}. 
2034: \bibitem{KLS92}
2035:   J.~Kapusta, L.~McLerran, and D.K.~Srivastava, \plb{283}{1992}{145}.
2036: \bibitem{RHIC}
2037:   S.A.~Bass {\it et al.}, \npa{661}{1999}{205c}.
2038: \bibitem{fn2}
2039:   According to (\ref{CF}) the hadronic spectra from which $v_2$ and
2040:   $v_{2,p_{\rm T}^2}$ are calculated involve an integration over the
2041:   freeze-out surface $\Sigma$ whereas $\epsilon_p$ is given by a 
2042:   spatial integral at fixed time (e.g. at $\tau_{\rm f}$). We found 
2043:   that the difference is small since most particles freeze out during 
2044:   a very short time interval. This is in particular true at very high 
2045:   energies (see Fig.~4 in the first paper of Ref.~\cite{KSH99}). Another 
2046:   difference between $\epsilon_p$ and $v_{2,p_{\rm T}^2}$ is that the 
2047:   first implicitly sums the contributions from all particle species while
2048:   the latter is defined in terms of the spectrum of a single particle
2049:   species (in our case pions). The difference is small for 
2050:   pion-dominated systems like those studied here. For the same 
2051:   reason, however, the $v_{2,p_{\rm T}^2}$ for protons or kaons 
2052:   may well differ from $\epsilon_p$.
2053: \bibitem{VZ96}
2054:   S.A. Voloshin and Y. Zhang, \zpc{70}{1996}{665}.
2055: \bibitem{D95}
2056:   P. Danielewicz, \prevc{51}{1995}{716}.
2057: \bibitem{BS00}
2058:   M. Bleicher and H. St\"ocker, {\tt hep-ph/0006147}.
2059: \bibitem{fn3}
2060:   The differing statement in \cite{HL99} that this ratio should 
2061:   increase with decreasing impact parameter (see Fig.~5 in \cite{HL99})
2062:   is incorrect. It is derived from schematic considerations which 
2063:   assume that elliptic flow scales like radial flow, not taking into
2064:   account the earlier saturation of the former.
2065: \bibitem{Biro}
2066:   T. Bir\'o, \plb{474}{2000}{21}, and {\tt nucl-th/ 0003027}.
2067: \bibitem{Posk}
2068:   A.M. Poskanzer, private communication.
2069: 
2070: \end{thebibliography}
2071: 
2072: \end{document}
2073: 
2074: 
2075: 
2076: 
2077: 
2078: 
2079: 
2080: 
2081: 
2082: 
2083: