1: \section{The properties of the gas}
2:
3: In this section, we examine both the evolution of the gas in the simulation
4: as a whole and the properties of gas within virialised dark matter halos.
5:
6: \subsection{Global properties of the gas}
7:
8: Following Kay \etal (2000), we define three different gas phases: (i)
9: cold galactic gas (temperature below $12000\K$ and overdensity greater
10: than 10); (ii) hot halo gas (temperature above $10^5\K$); and (iii)
11: cold, uncollapsed gas (overdensity less than 10; this is essentially
12: diffuse gas in ``voids.'') We ignore the effects of photoionization
13: because at this mass resolution a photoionising field has a negligible
14: effect upon galactic or halo properties.
15: In practice, little gas is both cold and at overdensities between 5
16: and 1000, so that effectively the uncollapsed and galaxy phases are
17: completely disjoint. The evolution of the fraction of gas in each of
18: these phases is illustrated in Fig.~3 for the two cosmological models
19: that we have calculated, SCDM (bold lines) and \LCDM (thin
20: lines). These gas fractions are very similar to those derived using a
21: semi-analytic model of cooling gas implemented in these simulations by
22: Benson \etal (2000c).
23:
24: \begin{figure}
25: \psfig{file=figs/madau.ps,height=8.5cm,width=8.5cm}
26: \caption{The rate at which gas cools, per unit volume, in the SCDM
27: simulation as a function of time since the Big Bang.
28: The data points are inferred star formation rates in the real
29: universe, from Steidel \etal (1999). The match between the simulation
30: results and the data indicates that, in spite of resolution effects and
31: other limitations, the simulations provide a reasonably realistic model.
32: \label{madau}}
33: \end{figure}
34:
35: The properties of cold, uncollapsed material in voids are only followed in an
36: approximate manner in our simulations. The SPH method generally has
37: poor spatial resolution in regions in which the gas is diffuse because
38: it must smooth over a sparse particle distribution. This limitation
39: is exacerbated in our specific implementation which imposes a
40: maximum search length, resulting in a minimum resolvable density of a
41: few times the mean density in the computational volume. Test
42: simulations without such a minimum resolvable density demonstrate that
43: these inaccuracies do not affect the properties of the galaxies because
44: the gas forces are small in void regions and have no effect on the
45: dense regions that ultimately form galaxies (Kay \etal 2000). However,
46: outwardly propagating shocks generated by the formation of large
47: clusters are poorly followed in underdense regions and so we can say
48: little about the detailed properties of voids (beyond their mere
49: presence). This is a regime in which Eulerian techniques are more
50: accurate than SPH (\eg Cen \& Ostriker 1996).
51:
52: As the evolution proceeds the cold, diffuse phase is depleted as gas
53: falls into dark halos where it is shocked into the hot phase,
54: thereafter cooling into galaxies. The time derivative of the dashed
55: curve that tracks the galaxy phase in Fig.~3 effectively defines the
56: global star formation rate in the computational volume. This is
57: plotted in Fig.~4. The star formation rate rises rapidly at early times, has a
58: broad maximum around $z\sim 1-2$ and then gently declines to the
59: present. This behaviour is broadly similar to that seen in real data
60: (Steidel \etal 1999), indicating that the limited resolution of the
61: simulations and other approximations do not result in too unrealistic
62: a model. At late times, Fig.~3 shows that there is a substantial
63: transfer of gas from the uncollapsed phase into the hot phase, with
64: around $50\%$ of the gas ending up at temperatures above $10^5\K$ in
65: both cosmological models. At $z\simeq 1$, when only around $30\%$ of
66: the gas is above $10^5\K$, the gas fractions in the different phases
67: in our simulations agree well with those in the simulations of Evrard
68: \etal (1994) (which were stopped at this point.)
69:
70: \begin{figure}
71: \psfig{file=figs/pscol.ps,height=8.5cm,width=8.5cm}
72: \caption{A comparison of the cumulative mass function of dark
73: halos in the \LCDM gas + dark matter simulations and an otherwise
74: identical pure dark matter simulation. The halo masses of the gas +
75: dark matter simulation have been scaled by
76: $\Omega_0/(\Omega_0-\Omega_b)$. For this plot the halos were
77: identified using a friends-of-friends algorithm with a linking length
78: of 0.2. The shaded region shows the expected number of halos, allowing
79: for $1\sigma$ Poisson errors because of the finite simulation volume, as
80: predicted by the formula given in Jenkins \etal (2000).
81: \label{ps}}
82: \end{figure}
83:
84: As expected, more gas cools into the galaxy phase in the \LCDM than in
85: the SCDM model. This is due to our adoption of a single gas particle
86: mass in the two simulations which results in the same effective
87: resolution in both cases. More gas cools at early times in the \LCDM
88: simulation because structure forms earlier in this case. If all the
89: gas that cools into galaxies were assumed to turn into stars, then the
90: SCDM model would have roughly the observed mean mass density in
91: stars. In the \LCDM model, on the other hand, about twice as much gas
92: cools and, if all this gas were assumed to turn into stars, an
93: uncomfortably large value of the mass-to-light ratio would then be
94: required to match the observed mean stellar mass density. The amount
95: of gas that cools in our simulations is determined by the mass
96: resolution and a small decrease in the resolution of the
97: \LCDM model would reduce the final amount of cold gas to the same
98: level as in the SCDM model.
99:
100: \subsection{Properties of gas in halos}
101:
102: To construct a catalogue of dark matter halos from the simulations, we
103: first identified cluster centres using the friends-of-friends grouping
104: algorithm (Davis \etal 1985) with a small value of the linking length,
105: $b=0.05$. We then grew a sphere around the centre of mass of each of these
106: halos until the mean overdensity within it reached a value of 178 in the
107: case of SCDM or 324 in the case of \LCDM. These are the overdensities
108: corresponding to the virial radius according to the spherical top-hat model
109: for clusters (Eke, Cole \& Frenk 1996.) The resulting set of halos was
110: cleaned by removing the smaller of any two overlapping halos. This
111: procedure produced catalogues of 1320 halos in SCDM and 1795 halos
112: in \LCDM, with more than 28 particles per halo in each case.
113:
114: To check whether the presence of baryons affects the halo catalogue, we ran
115: a pure N-body simulation of the \LCDM model and extracted a halo catalogue
116: from both it and the baryonic simulation with gas cooling. The mass
117: function of halos in the test simulation is almost indistinguishable from
118: that in the original simulation, as illustrated in Fig.~5. This figure also
119: compares the mass functions in the simulations with the mass function given
120: in Jenkins \etal (2000). This gives a better fit than the predictions of
121: the Press-Schechter model (Press \& Schechter 1974). Although for the
122: values of $\Omega_b$ that we have assumed here, the presence of gas has
123: little effect on the halo mass function, it does affect the central
124: structure of the halo into which it cools (Pearce \etal 2000).
125:
126: \begin{figure*}
127: \psfig{file=figs/gasfrac_comp.ps,height=8.5cm,width=17.5cm}
128: \caption{The ratio of gas to dark matter mass ($\times
129: \Omega_{cdm}/\Omega_b$, where $\Omega_{cdm}$ denotes the dark matter
130: contribution to $\Omega_0$) within the virial radius of $\sim 1500$
131: dark matter halos found in the simulations. The circles give the total gas
132: fraction, the triangles the fraction of gas below $12000\K$ and the crosses
133: the fraction of gas above $12000\K$. The solid line indicates the
134: effective resolution limit of 32 gas particles within a halo. Panel (a)
135: corresponds to the \LCDM simulation and panel (b) to the SCDM simulation.
136: \label{frac}}
137: \end{figure*}
138:
139: \begin{figure*}
140: \psfig{file=figs/massvst.ps,height=8.5cm,width=17.5cm}
141: \caption{The relation between the mean, mass-weighted temperature of
142: hot ($>10^5\K$) gas and the mass of its host dark matter halo. The
143: dashed line shows the theoretical expectation, $M \propto T^{3 \over
144: 2}$, of a simple equilibrium model for the halos. The tight
145: correlation between $T$ and $M$ indicates that the gas is close to
146: equilibrium in all but the smallest halos in the simulations (where
147: the temperatures can be overestimated due to the proximity of a larger
148: halo, see text).
149: \label{hgvstemp}}
150: \end{figure*}
151:
152: In Fig.~6 we plot the mass fraction of gas within the virial radius of the
153: halos normalized to the mass fraction in the simulation as a whole,
154: together with the proportions in the hot and cold phases. The total gas
155: mass fraction varies little with halo mass over the range $10^{12}\Mo\lsim
156: M\lsim 10^{15}\Mo$, although there is some indication of an upturn at low
157: masses in the \LCDM case and at high masses in the SCDM case. The mean
158: value of the gas fraction is $\sim 0.85$ for \LCDM and $\sim 0.8$ for SCDM
159: with a scatter of $\sim 0.05$, indicating that the gas is slightly more
160: extended than the dark matter in both models. A few small mass halos have a
161: gas mass fraction greater than unity. These halos, however, are near the
162: resolution limit of the simulations. The crosses show the mass fraction of
163: hot ($T> 12000\K$) gas in the halos. This increases rapidly with halo mass,
164: from $\sim 10-20\%$ for the smallest halos in the simulations to $\sim
165: 60-80\%$ for the largest. The fraction of cold ($T<12000\K$) gas shows the
166: opposite trend and varies from $\sim 70-80\%$ at the small mass end to
167: $\sim 10-20\%$ at the high mass end.
168:
169: \begin{figure*}
170: \psfig{file=figs/mult.ps,height=8.5cm,width=15.0cm}
171: \caption{Evolution of the abundance of ``galaxies" in the simulations as a
172: function of their cold gas mass. The mass functions are plotted at
173: the redshifts given in the legend. The left panel shows results for \LCDM
174: and the right panel for SCDM.
175: \label{mult}}
176: \end{figure*}
177:
178: The fraction of baryonic material that cools within dark matter halos can
179: also be calculated using semi-analytical techniques (White \& Frenk
180: 1991). As we discussed in the preceeding section, in semi-analytic models
181: the cooling of gas is regulated by feedback processes due to energy
182: liberated by stars. In these simulations we ignore feedback altogether and
183: cooling in small mass objects is then limited by resolution effects. This
184: leads to some differences in the gas fractions derived with these two
185: techniques. For example, in the semi-analytic model of Cole \etal (2000),
186: the gas fraction of cold gas has a maximum of $\sim 40$\% at a halo mass of
187: $10^{12}\Msun$. This is close to the resolution limit of our
188: simulations. At small masses, the cold fraction is suppressed by feedback
189: and at larger masses, it is suppressed by the long cooling time of the
190: diffuse gas in halos. The first of these effects is not present in our
191: simulations (which do not resolve the regime where the feedback is
192: important in the semi-analytic models), but the second one is. On these
193: large mass scales, the simulations agree well with the semi-analytic
194: results and even over the entire range of mass, the difference is only of
195: the order of 50\% (Benson \etal 2000c). Overall, somewhat more mass cools
196: in the simulations than in the semi-analytic model, suggesting that
197: feedback effects are not negligible even above the resolution limit of our
198: simulations.
199:
200: The hot gas that falls into a halo is quickly thermalized by shocks and
201: heated to the virial temperature. This is illustrated in Fig.~7 which shows
202: the relation between the mean, mass-weighted gas temperature (for gas with
203: $T>10^5\K$) and the virial mass of the host halo. For masses above a few
204: times $10^{12}\Msun$, there is a very tight correlation between gas
205: temperature and halo mass, indicating that the gas is relaxed and close to
206: equilibrium. At the small mass end, there are some halos with anomalously
207: large temperatures. These tend to be in the vicinity of larger halos whose
208: own hot halos contaminate the smaller ones, boosting their temperature.
209: The dashed line is the relation $M \propto T^{3
210: \over 2}$ predicted by a simple, spherically symmetric, equilibrium
211: model for the halo and its gas. The simulations follow this relation
212: well over most of the mass range plotted in the figure. The X-ray
213: properties of the largest halos in our simulations have been studied by Pearce
214: \etal (2000).
215:
216: \section{The properties of galaxies}
217:
218: We now consider the properties of the population of galaxies in the
219: simulations, focussing on their abundance, ages, merger histories, and
220: individual star formation rates.
221:
222: \subsection{Demographics of the galaxy population}
223:
224: We identify ``galaxies" in the simulations with dense clumps of cold
225: gas. Locating them is straightforward except in the small
226: number of cases when they are undergoing a merger or are being tidally
227: disrupted within a large cluster. To find galaxies, we use the same
228: friends-of-friends algorithm that we employed for finding dark halos,
229: but with a linking length of only $b=0.02$ for SCDM and $b=0.0134$ for
230: $\Lambda$CDM, values that are 10\% of those required to obtain virialised
231: halos in these cosmologies.
232: When calculating the mass of each galaxy, we consider only particles
233: with temperature below $12000\K$, a condition that
234: rejects only very few particles. In practice, as the cold clumps of
235: gas that make up the galaxies have contracted by a large factor the
236: galaxy catalogues and the masses
237: and positions of each are insensitive to these choices. We find a
238: total of about 2000 resolved galaxies in each of the two simulations at the
239: present day.
240:
241: The evolution in the abundance of galaxies is shown in Fig~8. Here, we
242: characterize each galaxy by its cold gas mass and plot the cumulative
243: abundance at various epochs in our two simulations. In the \LCDM model, the
244: first substantial galaxies begin to form at a redshift $z\simeq 5$,
245: although some fairly massive objects are present even before this. The
246: massive galaxy population builds up rapidly between $z\simeq 5$ and
247: $z\simeq 3$ when it increases approximately ten-fold. A similar increase
248: occurs between $z\simeq 3$ and $z\simeq 1$. As time progresses, the mass of
249: the largest galaxies also increases. After $z=0.5$, the growth in the
250: population slows down somewhat because the rate at which new galaxies are
251: forming is more or less balanced by the rate at which existing galaxies are
252: destroyed by mergers and tidal disruption. At $z=0$, the abundance of
253: galaxies with gas mass $M>10^{11} \Msun$ approximately matches the observed
254: abundance of $L_*$ or brighter galaxies of $\sim 0.003$ per
255: $h^{-3}\Mpc^{3}$. The evolution in the SCDM model is qualitatively similar,
256: but shifted towards lower redshifts.
257:
258: The creation sites of galaxies at recent times are
259: illustrated in Fig.~9 which shows the locations where new
260: gas cools between $z=0.3$ and the present in the SCDM model. (Broadly similar
261: behaviour is seen in the \LCDM model.) The radius of the circle around
262: each of these locations is scaled by the mass of new cold
263: material. Some gas cools onto preexisting objects (note the small
264: circles at the centre of some clusters), but 383 new resolved
265: galaxies, about 15\% of the total, formed during this
266: period. The new galaxies are
267: small and tend to form along filaments or in low density regions, away
268: from the larger clusters.
269:
270: \begin{figure}
271: \psfig{file=figs/newgal.ps,height=8.5cm}
272: \caption{The circles indicate the
273: locations where new gas cools after $z=0.3$ in the
274: SCDM model, scaled in radius by the mass of new
275: cold material. The grey shading indicates
276: the underlying dark matter density at $z=0$.
277: \label{newgal}}
278: \end{figure}
279:
280: Galaxies are destroyed preferentially in large
281: clusters, where tidal forces are largest.
282: For example, about $20\%$ of the cold material within the virial
283: radius of the largest halo at $z=0.3$ is reheated before $z=0$, and is
284: associated with the destruction of about half the small (around
285: 1/3$L_*$) galaxies in this halo. As discussed earlier,
286: tidal disruption of galaxies in clusters is likely to be enhanced by
287: the limited spatial resolution of the simulations which results in
288: artificially low binding energies for small objects. Larger simulations
289: are required to investigate this important problem in detail.
290:
291: The major consequence of ``decoupling'' the hot gas from the cold
292: phase, as discussed in the preceeding section, is that no supermassive
293: objects form. The biggest galaxy in the \LCDM model contains $5\times
294: 10^{12} \Msun$ of cold gas. The turnover in the abundance of galaxies
295: at small masses in the two simulations is a numerical artifact. At
296: least 32 particles (corresponding to a mass of $6.4\times10^{10}
297: \Msun$) of gas must be present for cooling to be efficient; objects
298: below this are in the process of forming or are being ablated as they
299: move in a hot halo. The effective mass threshold in
300: both simulations is actually slightly higher than the nominal value
301: and corresponds to a mass threshold around $10^{11}\Msun$, or 50
302: gas particles.
303:
304: The number of galaxies within the virial radius of each halo at the present
305: day is plotted against the halo mass in Fig.~10. The largest cluster
306: contains nearly 30 large galaxies and has a virial radius of over
307: $1.5h^{-1}\Mpc$. Its galaxy content is consistent with those of $10^{15}
308: h^{-1} \Msun$ clusters which typically contain 30 - 100 $L_*$ galaxies.
309: This is, in fact, the only large Abell-type cluster that formed in our
310: simulations. All dark halos in the simulations that contain more than a
311: few hundred particles have at least one galaxy within them. There are no
312: {\it empty halos} of mass greater than $10^{13}\Msun$ in either cosmology.
313:
314: \begin{figure*}
315: \psfig{file=figs/ngal.ps,height=8.5cm,width=17.5cm}
316: \caption{The number of galaxies with mass above
317: $6.4\times10^{10}\Msun$ found within the virial radius of each dark
318: matter halo. In the \LCDM model, the largest halo contains 29 galaxies
319: (24 in SCDM) and there are 6 clusters of 10 or more galaxies (7 in
320: SCDM). The largest clusters have suffered
321: significant depletion of their galaxy population due to ablation of
322: galaxies moving through the cluster halo.
323: \label{ngal}}
324: \end{figure*}
325:
326: \subsection{Star formation and the age of galaxies}
327:
328: Although, for simplicity, we have not included any prescription for star
329: formation in our simulations, we can still derive some general conclusions
330: regarding the properties of the stars expected to form. Stars form from gas
331: that has cooled into galactic dark halos. The approximate time at which an
332: individual gas particle is available for star formation is easy to estimate
333: in the simulation. We take this to be halfway between the output time at
334: which the particle first has an overdensity greater that 10 and a
335: temperature smaller than $12000\K$, and the previous output time. These
336: criteria have previously been shown to pick out solely baryonic material
337: that has cooled into galactic objects (e.g. Kay \etal 2000). We assume that
338: as soon as a particle becomes available for star formation, a mass of stars
339: equal to the mass of the particle forms.
340:
341: By calculating a star formation time for each particle that ends up in
342: galaxy, we can derive a mean, mass-weighted stellar age or formation time,
343: $t_{\rm form}$, for each galaxy as
344: \begin{equation}
345: t_{\rm form} = \sum_{i=1}^{N} \frac{t_i}{N},
346: \end{equation}
347: where the sum is over all particles found within the final galaxy. This
348: definition of formation time is based exclusively on the age of the stellar
349: population of a galaxy and takes no account of when the galaxy was
350: assembled. An alternative definition of formation epoch is the time at
351: which half the final galaxy mass was assembled into a single progenitor. As
352: illustrated in Fig.~11, these two times can be very different. For the
353: largest galaxy in the SCDM simulation volume (which has a gas mass of $2.4
354: \times 10^{12}\Mo$ and for which we have enough resolution to follow many
355: progenitors), the average formation time of its stars corresponds to
356: redshift $z\simeq 2.5$, but half of the stellar material is only assembled
357: into a single distinct object at $z\simeq0.7$. The formation history of
358: this galaxy bears little resemblance to the classic spherical top-hot
359: collapse model. Rather, the galaxy is assembled through repeated mergers of
360: sub-units, many of which have a mass close to the resolution threshold of
361: the simulation and which are originally spread over quite a large comoving
362: volume. These fragments tend to line up with filaments of the dark matter
363: distribution, giving rise to an anisotropic fragment distribution which is
364: particularly visible at $z=3$ in Fig.~11. The mass distribution of the
365: progenitors of this massive galaxy at various epochs is shown in Fig.~12.
366:
367: \begin{figure*}
368: \psfig{file=figs/rollgal.ps,height=22.5cm}
369: \caption{The positions of the progenitor fragments that make up
370: the largest galaxy at $z=0$ in the SCDM cosmology. Objects containing more
371: than 10 particles are shown as open squares. At each redshift, the circle
372: indicates the virial radius of the largest progenitor of the dark matter
373: halo hosting the largest galaxy, and the grey shading shows the underlying
374: dark matter density. Each frame illustrates a comoving cubic region of side
375: $12h^{-1}\Mpc$. Half the mass has cooled by $z\simeq 2.5$, but it is not
376: assembled into a single object until around $z\simeq 0.7$.
377: \label{galform}}
378: \end{figure*}
379:
380: \begin{figure}
381: \psfig{file=figs/histo.ps,height=7.4cm,width=7.4cm}
382: \caption{The mass distribution of progenitors of the most massive
383: galaxy in the simulation. Each histogram refers to a different redshift as
384: shown in the legend.
385: \label{progenitor}}
386: \end{figure}
387:
388: Fig.~13 shows the distribution of the mean stellar ages (as a fraction of
389: the lookback time) for galaxies in the SCDM simulation. Contrary to the
390: naive expectation for a hierarchical scenario, there is a weak trend for
391: the largest galaxies to be the oldest. This may be an artificial
392: consequence of the resolution limit of our model which is at a fixed
393: mass. Models with more realistic feedback and no lower mass limit produce a
394: weak trend in the opposite sense for galaxies in the field and no trend at
395: all for galaxies in clusters (Kauffmann 1996).
396:
397: \begin{figure}
398: \psfig{file=figs/age.ps,height=8.0cm}
399: \caption{The age distribution of mean stellar ages of galaxies in
400: the SCDM simulation. The age is shown as a
401: fraction of the lookback time (i.e. 0 corresponds to the present and 1
402: to the Big Bang). The largest galaxies are the oldest because their
403: stars formed preferentially in small fragments at
404: at high redshift (see fig.~11).
405: \label{age}}
406: \end{figure}
407:
408: \begin{figure}
409: \psfig{file=figs/bigsfr.ps,height=8.5cm,width=8.5cm}
410: \caption{The star formation rate for the 20 largest galaxies by
411: mass selected at $z=0$ in the SCDM simulation as a function of cosmic
412: time.
413: \label{bigsfr}}
414: \end{figure}
415:
416: An estimate of the star formation rate (SFR) in our simulated galaxies can
417: be derived from the gas mass that cools between successive output
418: times. This crude measure (we only have 10 output times) is shown for the
419: 20 largest galaxies selected at the end of the SCDM simulation in
420: Fig.~14. Clearly, the inferred aggregate SFR within each of these objects
421: is very high but, as Fig.~11 shows, these galaxies are broken up into many
422: small precursor objects at early times each of which has a modest SFR. The
423: aggregate SFR for the most massive objects today is sharply peaked at
424: $z\simeq 3$. This contrasts with the star formation history of more
425: typical galaxies, illustrated in Fig.~15. This shows SFRs for every
426: $50^{\rm th}$ galaxy ordered by mass. These curves are more extended in
427: time and the mean SFR for the average galaxy is much lower than the
428: aggregate star formation rate of the largest objects. The SFRs generally
429: decline towards low redshift as galaxies consume the gas that is able to
430: cool by the present day. The SFR integrated over all the galaxies in the
431: simulation is shown in Fig.~4.
432:
433: \begin{figure}
434: \psfig{file=figs/sfrall.ps,height=8.5cm,width=8.5cm}
435: \caption{The star formation rate for every 50th galaxy ordered
436: by mass selected at $z=0$ in the SCDM simulation as a function of
437: cosmic time.
438: \label{allsfr}}
439: \end{figure}
440:
441: \begin{figure*}
442: \psfig{file=figs/kcount_comp.ps,height=8.5cm,width=17.5cm}
443: \caption{The $K-$band luminosity function. The left panel shows
444: the \LCDM model and the right panel the SCDM model. The open
445: squares show our model results, while the filled squares are
446: observational data from Gardner \etal (1997). Poisson error bars are
447: shown in both cases.
448: \label{kband}}
449: \end{figure*}
450:
451: By convolving the SFR in each galaxy with the Bruzual \& Charlot (1993)
452: stellar population synthesis model, we have calculated the $K-$band
453: luminosity of each galaxy. We have assumed a Salpeter Initial Mass
454: Function with a low mass cutoff of $0.1\Msun$ and solar metallicity. (This
455: is higher than our assumed metallicity of 0.3 solar for the ambient gas,
456: since the galactic material will have been enriched by reprocessed metals.)
457: As shown in Fig.~16, the model luminosity function in both models has a
458: similar shape to the $K-$band luminosity function measured by Gardner \etal
459: (1997). Furthermore, the model luminosity function can be made to agree
460: quite well with the data by assuming that only a fraction $1/\Upsilon$ of
461: the stellar mass is luminous (and the rest is in brown dwarfs), with
462: $\Upsilon =2.8$ for \LCDM\ and $\Upsilon=0.83$ for SCDM. The former is
463: somewhat larger and the latter somewhat smaller than the values inferred
464: from observations (Cole \etal 2000 and references therein) because, as
465: discussed earlier, we have used the same mass per gas particle in the two
466: simulations and, as a result, more gas cools globally in the \LCDM model
467: than in the SCDM case. Due to limited resolution, only the bright end of
468: the luminosity function ($L \gsim L_*$) is accessible in our simulations.
469:
470:
471: