1: \subsection{Temperature--Mass relation}
2: \label{sec:tm}
3:
4: The most direct way to compare the simulated
5: temperature--mass\footnote{We use temperature--mass rather than
6: mass--temperature because we are complete in mass rather than
7: temperature.} relation between simulated and observed clusters is to
8: use the thermal (mass-weighted) temperature of the gas within a small
9: region that is well-observed in X-rays. We have already done this
10: and presented the results in a short paper (Thomas et~al.~2002),
11: in which we compared the simulations described in this paper to
12: observations of 5 relaxed clusters using the {\it Chandra} satellite
13: (Allen, Schmidt \& Fabian 2001). In particular, we
14: compared the normalisation of the temperature--mass relation for matter
15: within $r_{2500}$ (where $r_\Delta$ is the radius of the sphere that
16: encloses a mean density of $\Delta$ times the critical density). The
17: {\it Non-radiative} simulation agrees with previous simulations of
18: that kind (e.g.~Mathiesen \& Evrard 2001) in predicting temperatures
19: that are too low for a given mass, whereas both the {\it Radiative}
20: and {\it Preheating} simulations reproduce the observations.
21: Unfortunately, spatially-resolved temperature data are as yet
22: available only for a few bright clusters and so the overlap in mass
23: between the observed and simulated clusters is small; nevertheless,
24: there is no reason to expect that our simulated results should not
25: extend up to higher temperatures.
26:
27: A much larger body of data exists for emission-weighted temperatures
28: of clusters, with poor spatial resolution. In this case some form of
29: modelling is required in order to derive the mass. The
30: emission-weighted temperature, being dominated by the high
31: surface brightness, central regions of the clusters, does not change
32: very much with $r_\Delta$, but the mass does. Generally, one wants to
33: choose as small a radius as possible in order to minimise the
34: extrapolation outside the region that is well-observed in X-rays. On
35: the other hand, theoretical predictions are for the mass within the
36: virial radius ($\Delta=111$ in this cosmology). In this paper, we
37: compromise and use $\Delta=200$ as this is the overdensity used in two
38: observational papers that we wish to compare with: Horner, Mushotzky
39: \& Scharf (1999) and Xu, Jin \& Wu (2001). A third, Finoguenov,
40: Reiprich \& B\"ohringer (2001) uses $\Delta=500$ but is easily
41: extrapolated to $\Delta=200$ (using $M\propto\Delta^{-1/2}$ for the
42: isothermal-$\beta$ model at radii much greater than the core radius).
43:
44: \begin{figure}
45: \psfig{file=tm6610.eps,angle=270,width=8.5cm}
46: \psfig{file=tm6720.eps,angle=270,width=8.5cm}
47: \psfig{file=tm6730.eps,angle=270,width=8.5cm}
48: \caption{The X-ray temperature versus mass relation for (a) the {\it
49: Non-radiative}, (b) the {\it Radiative} and (c) the {\it Preheating}
50: simulations. The open squares show the total soft-band X-ray
51: temperature whereas the filled circles exclude emission from (a) gas
52: with short cooling times, or (b,c) gas within the cooling radius. The
53: different lines are from observations (Horner et al. 1999) using mass
54: estimates from galaxy velocity dispersions (dashed line), X-ray
55: temperature profiles (dot-dashed line), the isothermal-$\beta$ model
56: (dotted line), and emissivity profiles derived by surface brightness
57: deprojection (triple-dot-dashed line).}
58: \label{fig:tm}
59: \end{figure}
60:
61: $kT_{X}-M_{200}$ relations for the clusters are presented in
62: Fig~\ref{fig:tm}. We use the cooling table of Raymond \& Smith (1977)
63: to calculate the X-ray temperature in the soft band (0.3-1.5 keV), as
64: described in Section~\ref{eq:entropy}. The open squares show the
65: temperature calculated using all the particles in the clusters.
66: However, the presence of cold, high-density gas in the cluster cores
67: (and also in infalling subclumps) gives emission-weighted
68: temperatures that are well below the virial temperature of the
69: cluster. Accordingly, we also show as filled circles, for the {\it
70: Radiative} and {\it Preheating} runs, the `cooling-flow corrected
71: temperature' obtained by omitting emission from within the cooling
72: radius, as defined in Section~\ref{sec:coolflow}. The change is most
73: important for high-mass clusters, for which it has the effect of both
74: tightening the relation and bringing it closer to the predicted slope
75: of two-thirds. The cooling-flow correction does not work very well
76: for clusters in the {\it Non-radiative} simulation which have cool,
77: dense gas at all radii: for this run, therefore, we show instead the
78: effect of omitting all gas particles with cooling times shorter than
79: 6\,Gyr.
80:
81: Most of our clusters are smaller than those for which X-ray masses
82: have been determined. Hence, to facilitate comparison with
83: observations, we fit a power-law to the temperature-mass relation for
84: clusters more massive than $5\times10^{13}\hMsol$.
85: Table~\ref{tab:tm} lists the normalisation, $A$, and slope,
86: $\alpha$ of the relation
87: \begin{equation}
88: kT_{X} = A \, (M_{200}/3\times 10^{14} h^{-1}{\rm M_{\odot}})^{1/\alpha}
89: \, {\rm keV},
90: \label{eqn:tmrel}
91: \end{equation}
92: where we have chosen to normalise at a mass scale
93: of $3\times10^{14}\hMsol$---towards the upper end of our
94: simulated catalogue but the lower-end of most observed ones.
95:
96: We have divided the Table into three parts. In the first, we list
97: results from previous simulations: EMN96---Evrard, Metzler \& Navarro
98: (1996); BN98---Bryan \& Norman (1998); T2001---Thomas et~al.~(2001);
99: ME01---Mathiesen \& Evrard (2001). These use various cosmologies, but
100: fortunately the results do not seem to be very sensitive to this.
101: Much more important is the resolution of the simulation. Thus ME01
102: have higher resolution than previous simulations (unfortunately they
103: do not state their precise mass-resolution in the paper) and find a
104: lower normalisation; our current simulations have a higher resolution
105: again and lower the normalisation still further. The reason why the
106: increasing resolution lowers the emission-weighted temperature of the
107: gas is the presence of cold, dense gas in subclumps. When we exclude
108: gas with a cooling time of less than 6\,Gyr from the calculation then
109: we find that the emission-weighted temperature rises once more to a
110: similar value to that found in the earlier, low-resolution simulations.
111:
112: The middle section of the Table shows results for the clusters
113: described in the current paper for emission in the soft-band,
114: 0.3--1.5\,keV, both with and without the inclusion of gas within the
115: cooling radius. The cooling flow correction has little effect on
116: clusters in the {\it Non-radiative} simulation, mainly because of the
117: presence of cool gas in infalling subclumps (the effect of removing
118: all gas with short cooling times is much larger and was shown in the
119: first part of the Table). On the other hand, gas with a short cooling
120: time in the {\it Radiative} and {\it Preheating} simulations resides
121: primarily in the cores of large clusters and its omission does
122: significantly raise the emission-weighted temperature. In a previous
123: paper, Pearce et~al.~(2000), it was shown that radiative cooling
124: \emph{raises} the temperature of the of the intracluster medium and we
125: confirm that result. Unfortunately, this is cancelled by the lower
126: temperature obtained by moving to higher resolution, so that the net
127: effect is to give temperature normalisations that are little changed
128: over earlier, non-radiative, low-resolution simulations
129:
130: In the lower portion of Table~\ref{tab:tm}, we present some
131: observational determinations of cluster temperatures from:
132: HMF99---Horner, Mushotzky \& Scharf (1999); FRB01---Finoguenov,
133: Reiprich \& B\"ohringer (2001); XJW01---Xu, Jin \& Wu (2001). It is
134: noticeable that various methods provide very different scaling
135: relations, both in normalisation and slope. The two methods that
136: provide the best agreement with the simulations are those that combine
137: optical velocity dispersions either with {\it ASCA} temperatures or with
138: surface-brightness deprojection of {\it Einstein} data to create
139: emissivity profiles (White, Jones \& Forman 1997). Unfortunately,
140: these are the least reliable as mass estimates from velocity
141: dispersions are prone to projection effects (e.g.~van Haarlem, Frenk
142: \& White 1997). Also, the deprojection method requires a
143: large extrapolation from $\Delta\approx2000$ out to $\Delta=200$.
144: The highest normalisation is provided by using resolved temperature
145: profiles. In principle this should be the most accurate method but as
146: yet the temperature profiles are poorly-determined and a high degree of
147: modelling is required. In two papers by Nevalainen, Markevitch \&
148: Forman (1999, 2000), for example, the enclosed gas mass fraction in
149: the clusters A\,401 and A\,3571 can be seen to be steeply rising at
150: the virial radius, contrary to expectation.
151:
152: The greatest degree of consensus is given by different authors using
153: the isothermal-$\beta$ model, which is another way of saying that
154: different groups measure the same relationship between $\beta_{\rm
155: fit}$ and temperature. In this model, the mass profile is
156: \begin{equation}
157: M(<r) \approx 1.11\times10^{14}\beta_{\rm fit}\,{kT\over\keV}\,
158: {r^2\over r_c^2+r^2}\,
159: \left(r\over h^{-1}\Mpc\right)\,\hMsol,
160: \end{equation}
161: which gives a mass within $r_{200}$ of
162: \begin{equation}
163: M_{200} \approx 7.69\times10^{13}
164: \left(\beta_{\rm fit}\,kT\over\keV\right)^{3\over2}
165: \left(r_{200}^2\over r_c^2+r_{200}^2\right)^{3\over2}\hMsol.
166: \label{eq:tmisobeta}
167: \end{equation}
168: In most cases $r_{200}\gg r_c$ so that the correction term in the above
169: equation for the finite core radius is approximately unity.
170:
171: \begin{table}
172: \caption{Power-law fits to the observational and simulated
173: temperature-mass relations of X-ray clusters: cluster sample; slope of
174: relation, $\alpha$; value of $kT/$keV at $3\times10^{14}\hMsol$, $A$.}
175: \label{tab:tm}
176: \begin{tabular}{llcc}
177: \hline
178: Sample&& $\alpha$& $A$ \\
179: \hline
180: EMN96 & soft band & 1.50& 3.6\\
181: BN98 & bolometric & 1.50& 2.9\\
182: T2001 & bolometric & 1.50& 3.6\\
183: ME01 & broad band, high res. & 1.51& 2.6\\
184: {\it Non-radiative}& bolometric, & 1.50& 2.1\\
185: & bolometric, $t_{\rm cool}>6$\,Gyr & 1.51& 3.3\\
186: \hline
187: {\it Non-radiative}& soft-band & 1.69& 2.0\\
188: & cooling-flow corrected& 1.69& 2.2\\
189: {\it Radiative} & soft-band & 1.96& 2.8\\
190: & cooling-flow corrected& 1.64& 3.4\\
191: {\it Preheating} & soft-band & 1.67& 3.2\\
192: & cooling-flow corrected& 1.61& 3.5\\
193: \hline
194: HMS99 & velocity dispersions & 1.53& 3.8\\
195: & temperature profiles & 1.48& 5.1\\
196: & emissivity profiles & 2.06& 3.7\\
197: & isothermal-$\beta$ model & 1.78& 4.4\\
198: FRB01 & isothermal-$\beta$ model & 1.67& 4.1\\
199: & polytropic-$\beta$ model, low-$T$ & 1.87& 4.3\\
200: & polytropic-$\beta$ model, high-$T$ & 1.48& 4.4\\
201: XJW01 & NFW-model & 1.81& 4.9\\
202: & isothermal-$\beta$ model & 1.60& 4.3\\
203: \hline
204: \end{tabular}
205: \end{table}
206:
207: The most extensive analysis of this kind was performed by Finoguenov,
208: Reiprich \& B\"ohringer (2001) for the HIFLUGCS (Highest X-ray Flux
209: Galaxy Cluster Sample) from the ROSAT All-Sky Survey. They used
210: $\beta_{\rm fit}$ values taken from fits to the ROSAT PSPC data and
211: temperatures mostly from ASCA. Where information on temperature
212: gradients was available they generalised the $\beta$-model to allow a
213: polytropic Equation of State and this gave very similar results. A
214: similar result is found by Xu, Jin \& Wu (2001) using a smaller sample
215: and by Horner, Mushotzky \& Scharf (1999) using ASCA data on 38
216: clusters from Fukazawa (1997). The slopes of the observed
217: temperature-mass relations are slightly steeper than 1.5,
218: reflecting the fact that the measured values of $\beta_{\rm fit}$ are
219: a slowly increasing function of mass, as we found for the simulated
220: clusters in Section~\ref{sec:sbprof}.
221:
222: \begin{figure}
223: \psfig{file=tmbeta6610.eps,angle=270,width=8.7cm}
224: \psfig{file=tmbeta6720.eps,angle=270,width=8.7cm}
225: \psfig{file=tmbeta6730.eps,angle=270,width=8.7cm}
226: \caption{The X-ray temperature versus mass relation for (a) the {\it
227: Non-radiative}, (b) the {\it Radiative} and (c) the {\it Preheating}
228: simulations, where the mass is estimated from the isothermal-$\beta$ model.
229: The dotted, dashed and dot-dashed lines are observed masses using
230: isothermal-$\beta$ model from HMS99, FRB01 and XJW01, respectively.}
231: \label{fig:tmbeta}
232: \end{figure}
233:
234: To test whether the temperature-mass relation using the
235: isothermal-$\beta$ model is consistent with the observations, we
236: estimate the masses of the simulated clusters using
237: equation~\ref{eq:tmisobeta} and the values of $\beta_{\rm fit}$ from
238: Section~\ref{sec:sbprof}. The resulting scaling relation, shown in
239: Figure~\ref{fig:tmbeta}, is fully consistent with the observations.
240: We do not want to over-interpret this result as the method of
241: determining $\beta_{\rm fit}$ in Section~\ref{sec:sbprof} is
242: far removed from the analysis that is carried out on real X-ray data.
243: Ideally, one would create mock observations from the simulations and
244: analyse them in the same way but that is a complex procedure that is
245: beyond the scope of the present paper. Nevertheless, we tentatively
246: conclude that the isothermal-$\beta$ model underestimates cluster
247: masses and that there is no disagreement between the masses of
248: simulated and real clusters.
249:
250:
251:
252: