1:
2: %\documentstyle{article}
3: %\documentstyle[aaspp4,flushrt]{article} \input epsf
4: %\documentclass[12pt]{article}
5: %\usepackage{aaspp4}
6: %\documentstyle[emulateapj,astrobib,epsfig]{article}
7: \documentstyle[astrobib,epsfig]{mn}
8: %\def\baselinestretch{2.0}
9:
10: \topmargin-1cm
11: \begin{document}
12:
13:
14: \def\bi#1{\hbox{\boldmath{$#1$}}}
15:
16: \newcommand{\beq}{\begin{equation}}
17: \newcommand{\eeq}{\end{equation}}
18: \newcommand{\beqa}{\begin{eqnarray}}
19: \newcommand{\eeqa}{\end{eqnarray}}
20:
21: \newcommand{\lexp}{\mathop{\langle}}
22: \newcommand{\rexp}{\mathop{\rangle}}
23: \newcommand{\rexpc}{\mathop{\rangle_c}}
24:
25: %\newcommand{\MNRAS}{Mon. Not. Roy. Astron. Soc.}
26: \def\bi#1{\hbox{\boldmath{$#1$}}}
27:
28:
29: \title{Constraints on galaxy halo profiles
30: from galaxy-galaxy lensing and Tully-Fisher/fundamental plane relations}
31:
32: \author[U. Seljak]{
33: U.~Seljak\thanks{E-mail: uros@feynman.princeton.edu} \\
34: Department of Physics, Princeton University, Princeton, NJ 08544,
35: USA}
36:
37: \pubyear{2001}
38:
39: \maketitle
40:
41: \begin{abstract}
42: Observations of galaxy-galaxy lensing from
43: Sloan Digital Sky Survey (SDSS) are combined with the Tully-Fisher
44: and fundamental plane relations to derive constraints on galactic halo
45: profiles.
46: We show that both for early and late type galaxies around
47: $L_*$ the
48: rotation velocity decreases
49: significantly from its peak value at the optical radius
50: to the virial radius $r_{200}$, $v_{\rm opt}/v_{200} \sim 1.8$ with about 20\%
51: uncertainty.
52: Such a decrease is expected in models
53: in which the halo profile is very concentrated, so that it
54: declines steeper than isothermal at large radii. This large decrease
55: can be explained as a result of both a concentrated dark matter profile
56: and a significant stellar contribution to the rotation velocity
57: at the optical radii.
58: We model the stellar component with a thin rotationally supported disk or
59: a Hernquist profile and use adiabatic dark matter response model to place
60: limits on the halo concentration as a function of the stellar
61: mass to light ratio. For reasonable values of the latter
62: we find concentrations $c_{200} $ consistent with
63: CDM predictions, suggesting there is no evidence for low concentrations
64: for the majority of halos in the universe.
65: We also discuss the origin of Faber-Jackson relation $L \propto \sigma^4$
66: in light of $L\propto v_{200}^{2.5}$ relation found for early type galaxies
67: above $L_*$
68: from galaxy-galaxy lensing. This leads to a decrease in $v_{\rm opt}/v_{200}$
69: with luminosity above $L_*$, so that at $7L_*$ the ratio is 1.4.
70: This is expected from the fundamental plane relation as a result of
71: a reduction in the baryonic contribution to the total mass at the optical
72: radius and a decrease in optical to virial rotation velocity in dark matter
73: profile.
74: These results imply that relations such as Tully-Fisher and
75: Faber-Jackson are not simply those between the
76: mass of dark matter halo and galaxy luminosity, but are also significantly
77: influenced by the baryonic
78: effects on the rotation velocity at optical radii.
79:
80: \end{abstract}
81:
82: %\keywords{large-scale structure of universe}
83: \begin{keywords}
84: cosmology: theory -- dark matter -- galaxies: haloes -- galaxies.
85: \end{keywords}
86:
87:
88: \section{Introduction}
89:
90: The study of dark matter profiles around galaxies has been an active
91: area since the original discovery of dark matter from the flat rotation
92: curves in spirals (see \citeNP{2001ARA&A..39..137S} for a review).
93: These show that there must be dark matter in the outer parts of the
94: galactic halos, but its extent is uncertain because of the limited
95: range probed by observations.
96: More recent observational studies of rotational velocity
97: in dwarf and low surface brightness galaxies have suggested
98: that the amount of dark
99: matter in the central regions is smaller than predicted for the average
100: galactic population by
101: CDM models (e.g. \citeNP{1994Natur.370..629M},
102: \citeNP{1994ApJ...427L...1F}, \citeNP{2000ApJ...543..704D},
103: \citeNP{2001ApJ...552L..23D}). The CDM models
104: predict very concentrated dark matter haloes. This is usually parametrized
105: using a universal dark matter profile such as NFW \cite{1997ApJ...490..493N}
106: \footnote{Although
107: other profiles have been proposed that differ significantly from NFW
108: in the inner parts of the halo, they agree well with NFW in the
109: outer parts \shortcite{2001ApJ...554..903K}. Since for the present work we are concerned predominantly
110: with the outer parts we only use NFW profile.}.
111: For a given virial mass
112: of the halo, in this paper defined as the mass enclosed within a sphere of
113: radius $r_{200}$ within which
114: the density is 200 times critical density, the
115: profile has a characteristic shape which depends only on a single parameter.
116: For NFW profile, in which the slope continuously
117: changes from the inner value of -1 to the outer value of -3,
118: the free parameter is often
119: defined as the concentration $c_{200}$, which is the
120: ratio between the radius $r_{200}$ and the scale radius
121: $r_s$ where the slope is close to $-2$. Higher concentration parameters
122: imply higher densities at the scale radius $r_s$. On galactic scales CDM
123: models predict $c_{200}\sim 7-15$ depending on the halo mass,
124: matter density, shape and
125: amplitude of the power spectrum.
126: %In general, models with higher amplitude
127: %of fluctuations, less negative effective slope or higher matter density
128: %predict larger $c_200$ \cite{2001ApJ...554..114E}.
129:
130: However, observational case for low concentrations
131: is not clear and different
132: conclusions have been reached by other studies (e.g.
133: \citeNP{2001MNRAS.325.1017V}).
134: Even if the observational results of low dark matter density in the inner
135: regions are confirmed, they do not necessarily indicate a problem
136: for the CDM models.
137: The variation
138: in the profile shapes between halos
139: is large and it is not clear whether the dwarf or
140: low surface brightness galaxies, which show strongest evidence
141: for low density cores,
142: can be associated with the average
143: halo population.
144: For example, if these galaxies are associated with a
145: population that had a major
146: merger in the recent past,
147: this may lead to a significantly flatter and less concentrated
148: halo structure than the average population (e.g. \shortciteNP{2001astro.ph..8151W}).
149: Another possibility is that astrophysical processes such as bar rotation
150: redistribute the dark matter in the inner parts of the halo \cite{2001astro.ph.10632W}.
151: However, if the case of low concentrations is extended to the population
152: of galaxies as a whole and to the outer parts of the halo, which cannot
153: be affected by astrophysical processes, then the CDM crisis would become
154: much more severe.
155:
156: It would thus be useful to have some information on the halo structure
157: for the mean population of galaxies. While many observationally
158: determined rotation velocity
159: profiles exist, the main uncertainty has always been the relative
160: contribution between the baryonic component in the gas, bulge/ellipsoid
161: and disk
162: and the dark matter in the halo. The baryonic and dark matter
163: components are difficult
164: to separate because the conversion from the star light (and, to a lesser
165: extent, gas density) to baryonic mass is still uncertain from the
166: stellar population synthesis models and other studies.
167: The situation is further
168: complicated by the fact that baryonic component is not dynamically
169: negligible and during its condensation
170: it induces a response of the dark matter halo changing
171: its original distribution in the inner parts. While these problems
172: are ameliorated for the low surface brightness galaxies, they are
173: almost never negligible.
174:
175: Since the modelling of relative disk or spheroid and
176: halo contributions is rather
177: difficult in the inner parts it is useful to concentrate on the
178: outer parts of the halo, where the baryon influence is less important.
179: However, optical
180: rotation curves typically extend only out to
181: $10-20$kpc and even the
182: ones based on HI measurements do not extend beyond 30-50kpc.
183: Similarly, velocity dispersion studies in early type galaxies also
184: do not extend past a few effective radii and even X-ray studies of
185: large ellipticals do not extend beyond 10 effective radii \cite{1999ApJ...518...50L}.
186: Over this limited range the data on disk $L_*$ galaxies
187: indicate that the rotation curves rise
188: in the inner parts of the disk and then stay approximately
189: constant or decline somewhat
190: out to the outer limit of observations. This flatness indicates
191: that approximating the density profile as isothermal, $\rho(r) \propto r^{-2}$,
192: is a good approximation to the matter profile over this range.
193: On the other hand,
194: theoretical CDM models predict that the velocity profile of the
195: dark matter increases out
196: to twice the scale radius $r_s$
197: and then slowly declines beyond that as the slope gradually decreases
198: towards -3. The fact that we do not observe a decrease in rotation
199: velocity at radii below $2.15r_s \sim 30-50$kpc
200: is presumably due to the baryonic effects, which increase the mass in
201: the center.
202: The decrease at large radii is however a robust prediction of these models.
203:
204: While, until recently, no accurate determination of the mass profile
205: at large radii has been available, recent galaxy-galaxy lensing observations
206: by the SDSS team \shortcite{2001astro.ph..8013M} have improved the
207: situation significantly by obtaining the morphology and luminosity
208: dependence of the signal. Theoretical analysis of lensing data
209: must take into account not just galactic halos, but also those from groups
210: and clusters, which dominate on large scales above 200-300$h^{-1}$kpc
211: \cite{gs02}.
212: These results show that a late type $L_*$
213: galaxy, with $L_I=2.7\times 10^{10}h^{-2}L_{\sun}$ (where we have applied
214: a 30\% internal extinction correction in $I$ band),
215: has a mass $M_{200}\sim (3.4\pm 2.1)\times 10^{11}h^{-1}M_{\sun}$, corresponding to the
216: circular velocity at the virial radius $v_{200} \sim 115$km/s.
217: While the error is still quite large it is gaussian distributed in mass, so
218: an increase in $v_{200}$ by 35\% to 150km/s
219: is excluded at 95\% confidence level.
220: This number can be
221: compared to the maximum rotation velocity for such a galaxy, which is
222: 208km/s with a small scatter \shortcite{1997ApJ...477L...1G}, the well known Tully-Fisher relation.
223: Comparing the two values shows that a decrease in rotation velocity from
224: the optical to the virial radius indeed occurs for the average
225: population of spiral galaxies. The decrease is large, almost a factor
226: of 2 and even if we push the virial velocity up by 2-$\sigma$
227: to $v_{200}= 150$km/s the
228: decrease is still around 40\%.
229: As shown in this paper, a similarly large decrease in rotation velocity
230: is obtained also
231: for early type galaxies which are not rotationally supported.
232:
233: These results are interesting, since they are in the direction predicted by
234: the CDM models and demonstrate that the halo profiles indeed become
235: steeper than -2 in the outer parts of the halo.
236: They set tight limits
237: both on the structure formation models,
238: by limiting the acceptable range of concentration parameters, and on the
239: disk/spheroid formation models, by constraining the stellar mass to light
240: ratio. Only for specific values of stellar mass to light ratio and
241: halo concentration can one satisfy these constraints.
242: The purpose of this paper is to investigate the constraints in detail
243: using halo and disk/spheroid formation models.
244:
245: Previous work on this subject has
246: explored the constraints from
247: the rotation velocity at optical radius given by the zero point of
248: Tully-Fisher relation
249: (\citeNP{2000MNRAS.318..163M},
250: \citeNP{2000ApJ...538..477N}, \citeNP{2001ApJ...554..114E}).
251: In the absence of virial mass information
252: hese models must rely on additional assumptions
253: to derive the constraints on cosmological models.
254: The advantage of the
255: additional information from lensing is that it provides another
256: dynamical constraint at large radii, which can
257: remove some of the modelling
258: uncertainties present in previous modelling. In addition,
259: while previous work only explored the constraints from late type
260: galaxies, in this paper we also investigate the constraints from
261: the early type galaxies. We find these are more robust both because the
262: virial masses are more accurately determined and because the velocity
263: dispersions at optical radii are obtained from the same SDSS
264: sample.
265:
266: \section{Late type galaxies}
267:
268: The average rotation velocity at optical radii can be obtained from the
269: Tully-Fisher (TF) relation, which in $I$-band is given by \shortcite{1997ApJ...477L...1G}
270: \begin{equation}
271: L_I=2.7\times 10^{10}\left({v_{\rm opt} \over 208{\rm km/s}}\right)^{3.1}h^{-2}L_{\sun},
272: \label{vrot}
273: \end{equation}
274: where we used $I-5\log h=-21-7.68[\log(2v_{\rm opt})-2.5]$
275: and $I_{\sun}=4.15$.
276: We denote with $v_{\rm opt}$ the maximum rotation
277: velocity typically achieved at the optical radius (roughly 3 times the
278: scale radius of the disk $R_d$, which is not to be confused by the
279: scale radius of the halo $r_s$). At this radius the rotation curve still has
280: a significant contribution from the disk.
281: Rotation curves at larger radii show that, for this range of
282: luminosities, the rotation curve at $L_*$ are flat or decline slightly
283: out to the largest radius observable, typically a few optical radii
284: (\citeNP{1991AJ....101.1231C},
285: \citeNP{1996MNRAS.281...27P},
286: \citeNP{2001astro.ph..8225V}).
287:
288: We would like to compare the dark matter velocity in the inner
289: parts of the halo to the
290: SDSS galaxy-galaxy lensing results. For a late type
291: $L_*$ galaxy with $i^*-5\log h =-21.26$
292: one finds $M_{200}=3.4\times 10^{11}h^{-2}M_{\sun}(c_{200}/10)^{-0.15}$ \cite{gs02}.
293: Using $i^*_{\sun}=4.52$ \cite{2001AJ....121.2358B},
294: applying an average 0.3 magnitude internal extinction correction
295: \cite{2001astro.ph..8225V}
296: and converting from the virial mass to the virial velocity using
297: the relation $GM_{200}/r_{200}=v_{200}^2$ we find
298: \begin{equation}
299: L_I=2.7\times 10^{10}\left({v_{200} \over 115{\rm km/s}}\right)^3
300: \left({c_{200}\over 10}\right)^{-0.05}h^{-2}L_{\sun}.
301: \label{v200}
302: \end{equation}
303: The dynamical range of galaxy-galaxy lensing is still rather small
304: and for late type galaxies luminosity dependence of the virial mass
305: cannot be established using the present sample, so we focus on $L_*$ galaxies,
306: which dominate the late galaxy galaxy-galaxy (g-g) lensing signal.
307:
308: By combining the two equations above
309: we can determine the best fitted ratio
310: \begin{equation}
311: {v_{\rm opt} \over v_{200}} \sim 1.8.
312: \label{vrat}
313: \end{equation}
314: Since the virial mass of late type galaxies is consistent with 0 at
315: 2-$\sigma$ level there is no upper limit to the velocity ratio, while
316: the lower limit is given by $v_{\rm opt}/v_{200}>1.4$ at 2-$\sigma$,
317: where the error budget
318: includes the statistical errors on the zero point of Tully-Fisher relation
319: \shortcite{1997ApJ...477L...1G}, but is dominated by the error
320: on the virial mass from galaxy-galaxy lensing \cite{gs02}
321: and we ignored the small concentration dependence.
322: In addition to the statistical errors there are also possible
323: systematic differences between the different Tully-Fisher
324: zero point determinations (\shortciteNP{1997ApJ...477L...1G},
325: \shortciteNP{1996ApJ...457..460W}, \citeNP{1997MNRAS.290L..77S}) which can be up to 0.2 magnitude.
326: Similarly there could be systematic differences between color selected
327: late type galaxies
328: \cite{2001astro.ph..8013M} and those selected for
329: rotation velocity studies, although morphological studies indicate that
330: the late type sample is
331: dominated by Sb/Sc morphological type used also in TF studies
332: \cite{2001astro.ph..7201S}.
333:
334: What constraints does equation \ref{vrat} imply on the structure formation
335: models? As discussed above, a decrease in rotation velocity implies that the
336: mass profile is steeper than isothermal in the outer parts of the
337: halo. This can be either because the dark matter profiles are steep or
338: because the stellar disk has a significant
339: contribution to the rotation velocity (or both).
340: An NFW profile is given by
341: $\rho(x)=\rho_s x^{-1}(1+x)^{-2}$, where $x=r/r_{s}$.
342: The rotation velocity is $v_c(r)^2=GM(r)/r$, where
343: $M(r)=4\pi \rho_sr_s^3A(c)$ and $A(c)=\ln(1+c)-c/(1+c)$. For NFW profile the rotation velocity
344: increases at small $x$ up to
345: the peak at $x=2.16$ and then declines gradually to the virial radius.
346: The ratio between the maximum and virial velocity is
347: $v_{\rm max} / v_{200}=0.46[c/A(c)]^{1/2}$
348: \cite{2001MNRAS.321..559B}.
349: For the range of concentration parameters predicted by CDM models
350: ($c<20$) one finds $v_{\rm opt} / v_{200}<1.4$.
351: It is thus unlikely that the dark matter halo can explain the decrease
352: by itself, since
353: even for very concentrated halos the velocity decrease is less than observed.
354: We must therefore include the contribution from the disk.
355:
356: We will use a model with a thin exponential
357: disk in a dark matter halo (\citeNP{1998MNRAS.295..319M},
358: \citeNP{2000ApJ...530..177V}).
359: We model the disk as a thin exponential surface density
360: profile,
361: \begin{equation}
362: \Sigma(R)=\Sigma_0\exp(-R/R_d),
363: \end{equation}
364: with the total disk mass given by $M_d=2\pi \Sigma_0R_d^2$.
365: The disk mass is related to the disk luminosity using disk
366: mass to light ratio $\Upsilon_I=M/L$. We will neglect
367: the bulge contribution both to the luminosity
368: and to the rotation curve (see \shortciteNP{1998MNRAS.295..319M}
369: for a discussion of this assumption).
370: We will use observations to provide
371: the typical scale length $R_{d}$
372: of the galactic disks. For an $L_*$ galaxy observations give $R_d\sim
373: 3.5h^{-1}$kpc (\citeNP{1997AJ....114.2402C}, \citeNP{1996A&A...313...45D}).
374: The scatter for this quantity at a given luminosity
375: is rather large, since galaxies come with a range of surface brightnesses.
376: Theoretically, the scatter has been
377: linked to the scatter in the spin parameter of haloes
378: \shortcite{1998MNRAS.295..319M}.
379: However, since in this paper we are primarily concerned with
380: the average properties of galaxies and not in the scatter around
381: the mean
382: we will assume $R_d= 3.5h^{-1}$kpc in the analysis as a mean value
383: for a typical $L_*$ galaxy.
384:
385: Disk gravity contributes to the measured rotation velocity.
386: In addition, disk gravity also induces a response of the dark
387: matter halo in the inner regions. The standard approach to
388: model this is to assume an adiabatic contraction of the halo,
389: which remains spherical, so that the angular momentum of individual
390: particles is conserved (see \citeNP{1984MNRAS.211..753B}, \shortciteNP{1986ApJ...301...27B} and \citeNP{2000ApJ...538..477N}
391: for more details and numerical tests on
392: the validity of this model). This leads to an implicit equation for the
393: final radius $r_{\rm f}$ of dark matter mass as a function of initial
394: radius $r_{\rm i}$
395: \begin{equation}
396: M_{\rm DM}(r_{\rm i})r_{\rm i}=[M_{\rm DM}(r_{\rm i})(1-f_*)+M_{\rm s}(r_{\rm f})]r_{\rm f},
397: \end{equation}
398: where $M_{\rm DM}(r)$ and $M_{\rm s}(r)$ are dark matter and stellar mass,
399: respectively and $f_*$ is the stellar mass fraction of the halo.
400: Assuming that the baryons which do not end up in the disk have the same
401: distribution as the dark matter allows one to solve the system completely for
402: a given disk mass and scale length and for a given halo profile.
403: Flattened nature of the disk is used when obtaining stellar rotation
404: velocity from the mass profile \cite{BT87}.
405: As a free parameter we will use the stellar mass to light ratio $\Upsilon_I$ (expressed
406: in solar units), which
407: from the known virial mass and luminosity of $L_*$ galaxy (and ignoring
408: bulge contribution to the luminosity) can be
409: related to $f_*$ as
410: \begin{equation}
411: \Upsilon_I=M_d/L_*={3.4\times 10^{11}h^{-1}M_{\sun}f_* \over 2.7\times 10^{10}h^{-2}L_{\sun}}
412: \approx 12f_* {hM_{\sun} \over L_{\sun}}.
413: \end{equation}
414: Typical values are $\Upsilon_I=(1-2)h$ (e.g. \citeNP{1997A&A...328..517B}), giving
415: $f_*=0.1-0.2$. Note that $f_*$ should not exceed $\Omega_b/\Omega_m \sim 0.04/\Omega_m$, since only the
416: baryons within the virial radius can condense to make stars. This implies $\Omega_m<0.4$ for the fiducial value of $M_*$ and $\Omega_m<1$ for the 95 \% c.l.
417: on $M_*$. This is in agreement with other determinations that give
418: $\Omega_m<0.4$, so
419: the fraction of baryons converted to stars does not
420: exceed the available supply, although it comes quite close to this limit and
421: is an argument against high values for $\Upsilon_I$.
422:
423: We can solve for $v_{\rm opt}/v_{200}$ for any given $f_*$ (or $\Upsilon_I$)
424: and $c_{200}$.
425: An example with $\Upsilon_I=1.7h$ and $c_{200}=12$ is shown in figure \ref{fig1}.
426: One can see that the velocity ratio of 1.8 can be naturally
427: obtained in a model with
428: concentration parameter in the range predicted by CDM models, $8<c_{200}<15$
429: (\citeNP{2001ApJ...554..114E}, \shortciteNP{2001MNRAS.321..559B}), and
430: with the expected stellar mass to light ratio, $1h<\Upsilon_I<2h$.
431: The rotation velocity is reasonably
432: close to flat over the optical region,
433: but decreases
434: by 10-20\% out to the largest range observable in HI,
435: in agreement with the observations for this range of
436: luminosities (\citeNP{1991AJ....101.1231C},
437: \citeNP{2001astro.ph..8225V}).
438: The disk and dark matter contributions to the mass within the optical radius
439: are comparable, each contributing 50\% in this example. This implies
440: that this model
441: satisfies the requirement that
442: the zero point of TF relation is independent of the disk surface brightness,
443: which requires about
444: 50\% dark matter contribution to the rotation velocity at optical radius
445: \cite{2000MNRAS.318..163M}.
446: Note that the adiabatic response
447: of dark matter is quite significant and dark matter
448: would be subdominant if it were not
449: compressed by baryonic condensation.
450: Because of this the
451: dark matter contribution to the rotation curve is never negligible,
452: even in the inner parts of the galaxy. However, our predictions may
453: not be reliable inside the optical radius, where bulge makes a
454: significant contribution and adibatic approximation may not be
455: valid \cite{2001astro.ph.10632W}.
456:
457: \begin{figure}
458: \begin{center}
459: \leavevmode
460: \epsfxsize=3.0in \epsfbox{fig1_tf.ps}
461: \end{center}
462: \caption{Rotation velocity contributions from the disk (short dashed),
463: dark matter (dotted) and total (solid) out to the virial radius.
464: Also shown is the
465: dark matter without adiabatic response to baryon contraction (long dashed).
466: Insert plots the same over the radii accessible with HI data,
467: showing that the rotation curve is close to flat over this range.
468: We do not model the bulge/bar, so the predictions within inner few kpc
469: are not reliable.
470: }
471: \label{fig1}
472: \end{figure}
473:
474: The example above was chosen based on the typical values for concentration and stellar
475: mass to light ratio. More generally for any choice of one there will be a particular value
476: for the other that satisfies the constraint in equation
477: (\ref{vrat}). This is shown in figure \ref{fig2} for
478: a family of stellar mass to light ratios.
479: Low values of $\Upsilon_I$ cannot be made compatible with the constraint
480: in equation \ref{vrat},
481: unless the concentrations are unreasonably high (e.g. $c>20$ for
482: $\Upsilon_I<1h$).
483: Low concentrations are also not acceptable unless
484: the observed $v_{\rm opt}/v_{200}$ is decreased by
485: 2-$\sigma$ to 1.4. In this case
486: one must adopt either
487: very low concentrations for reasonable $\Upsilon_I \sim 1.5h$ or very low
488: $\Upsilon_I<0.7h$ for reasonable $c \sim 10$. On the other hand, a positive
489: deviation in $v_{\rm opt}/v_{200}$ can be explained by using a somewhat higher
490: values for $\Upsilon_I$.
491: We conclude that the results are just what is expected
492: from the CDM models with standard concentrations and standard stellar
493: mass to light ratios. Only if the virial masses deviate by 2-$\sigma$
494: in the positive direction
495: from the mean value does one run into problems with the standard stellar mass
496: to light ratios and one requires $\Upsilon_I=0.5h$.
497: The majority of late type galaxies
498: therefore show no
499: evidence for shallow density profiles in the outer parts of the halo.
500:
501: \begin{figure}
502: \begin{center}
503: \leavevmode
504: \epsfxsize=3.0in \epsfbox{fig2_tf.ps}
505: \end{center}
506: \caption{
507: Lines of constant $\Upsilon_I$ (times $h$)
508: as a function of $c$ versus
509: $v_{\rm opt}/v_{200}$. Thick dashed horizontal line
510: is the observationally
511: determined value $v_{\rm opt}/v_{200}=1.8 $.
512: A 2-$\sigma$ (98\% confidence level)
513: lower limit is $v_{\rm opt}/v_{200}>1.4$ (thin dashed line), while currently
514: the data provide
515: no upper limit since the virial masses are consistent with 0.
516: }
517: \label{fig2}
518: \end{figure}
519:
520: \section{Early type galaxies}
521:
522: Early type galaxies have some advantages
523: in constraining the outer parts of the halo profiles.
524: The main advantage is that they
525: show a stronger g-g lensing signal, so that both $M_*$ and
526: its scaling with luminosity are reliably determined from the current data
527: and the errors associated with it are significantly smaller.
528: Moreover, one can study their early
529: type dynamics using the same spectroscopic SDSS sample also used for g-g lensing
530: (the actual samples used in the analysis here
531: are not completely equal since the analysis for the
532: early type dynamics in \citeN{2001astro.ph.10344B} was using a larger sample than the lensing
533: analysis in \citeN{2001astro.ph..8013M}, but the
534: statistical properties of the two samples should be very similar). The
535: fundamental plane relations as derived by
536: \shortciteN{2001astro.ph.10344B} are $L \propto \sigma^4 \propto R_{\rm e}^{1.5}$, where $\sigma$ is the
537: central stellar velocity dispersion and $R_{\rm e}$ is the effective
538: radius of de Vauculeurs profile. The values at $L_*$ in $i'=-21.26$ are
539: $\sigma=177$km/s and $R_{\rm e}=2.7h^{-1}$kpc (note that we are using $L_*$ for
540: the luminosity function of the whole sample given in
541: \shortciteNP{2001AJ....121.2358B},
542: not just the early type).
543:
544: For early type galaxies
545: the conversion from the stellar velocity dispersion to matter circular velocity
546: is less straightforward. In principle one can obtain it by solving
547: the Jeans equation, which however depends on the unknown anisotropy of velocity
548: dispersion.
549: If the circular velocity of the matter does not
550: change very much over the optical region then virial theorem guarantees that
551: $v_{\rm opt}=3^{1/2}\langle \sigma \rangle$, if $\langle \sigma \rangle$
552: is luminosity weighted line of sight
553: velocity dispersion.
554: In practice luminosity average is difficult to achieve and the velocity
555: dispersion decreases with radius, so more often observers report the
556: central velocity dispersion $\sigma_{\rm central}$. Based on detailed
557: kinematic analysis and on comparison between strong lensing and
558: stellar dispersions \citeN{1994ApJ...436...56K} argues
559: that in this case the relation is closer to $v_{\rm opt}=2^{1/2}\sigma_{\rm central}$.
560: In SDSS the aperture is determined by SDSS fibers and is $r_{\rm fiber}=1.5"$,
561: implying
562: that for nearby galaxies only central parts of the galaxy are detected,
563: while for distant galaxies most of the light is observed. \citeN{2001astro.ph.10344B}
564: attempt to correct for this using an empirical fit, $\sigma_{\rm central}/
565: \sigma=(8r_{\rm fiber}/R_{\rm e})^{0.04}$,
566: to obtain the central velocity dispersion. The correction is empirical and
567: appears
568: to be somewhat small to account for the suggested 22\% difference between the
569: central and luminosity weighted velocity dispersion, since one needs
570: very small $R_{\rm e}=0.15"$ to achieve this. We will use an intermediate
571: value of $v_{\rm opt}=1.5\sigma$ to convert from the central velocity
572: dispersion to the rotation velocity at the optical radius. This
573: should have at most a 10\% systematic uncertainty attached to it.
574: This conversion
575: agrees well with the studies of slowly rotating elliptical
576: galaxies where both $v_{\rm opt}$ and $\sigma_{\rm central}$ have been
577: measured \cite{2001AJ....121.1936G}.
578:
579: Another ingredient in the modelling is the dark matter response to
580: baryonic contraction.
581: We model it again using the adiabatic model.
582: In principle there is no reason why such a model would be appropriate
583: for elliptical galaxies, where
584: stars are not on circular orbits, but numerical
585: simulations of galaxy formation have found that adiabatic compression model
586: works remarkably well even for such systems \cite{gotbrath02}, perhaps
587: as a consequence of conservation of radial action in such systems.
588: We use the Hernquist
589: profile \cite{1990ApJ...356..359H}, which
590: has been shown to give a light profile very close
591: to the de Vaucouleurs profile and has an analytic 3-d radial distribution,
592: to model the star distribution \cite{2001ApJ...561...46K}.
593: Our canonic value for the stellar mass to light ratio in $i'$ is
594: a factor of 2
595: higher than that of late type galaxies in the same band,
596: $\Upsilon_{i'}=3h M_{\sun}/L_{\sun}$.
597: This is based on $K-i'$ color difference between early and late type
598: galaxies in SDSS, which is around 0.2-0.3 magnitudes (without
599: internal extinction correction, \shortciteNP{2001astro.ph.11024I}) and
600: the fact
601: that in K band luminosity to stellar mass conversion only depends
602: on the age of population and differs by less than a factor of 2 between
603: early and late type galaxies for reasonable IMF and assuming ages above
604: 3Gyr
605: \shortcite{2002astro.ph..1207D}.
606: This stellar mass to light ratio is again rather uncertain and significantly
607: higher values have been suggested in the literature. Most of the direct
608: studies are done in B band, but even after correcting for a factor of 2-3
609: difference between B and i' luminosity for early type galaxies it is on
610: the low side based on the dynamical studies of central regions using
611: the minimal halo models \shortcite{2001AJ....121.1936G}. The simplest
612: explanation is that part of the mass is actually due to dark matter, as
613: discussed further below.
614: Note that for $\Upsilon_{i'}=4h M_{\sun}/L_{\sun}$ the stellar to
615: virial halo mass ratio for early and late type galaxies become equal and are
616: approaching the maximal baryon to dark matter ratio still allowed by
617: the observations. It is thus unlikely that the average stellar mass to light
618: ratio can significantly exceed this value if the virial masses
619: from g-g lensing are correct.
620:
621: At $L_*=2\times 10^{10}h^{-2}L_{\sun}$
622: the virial mass for early type galaxy is
623: $(9.3 \pm 2.2)\times 10^{11}h^{-1}M_{\sun}$ \cite{gs02}, which translates to
624: $v_{200}=(160 \pm 15)$ km/s
625: at $r_{200}=160h^{-1}$kpc. At the effective radius
626: $R_{\rm e}=2.7h^{-1}$kpc the rotation velocity from optical velocity
627: dispersion is
628: $v_{\rm opt} \sim 1.5\times 177{\rm km/s}=
629: 265$km/s with a small error \cite{2001astro.ph.10344B}, leading to
630: \begin{equation}
631: {v_{opt} \over v_{200}} =1.68 \pm 0.2.
632: \end{equation}
633: The error is dominated by stellar to
634: dark matter velocity dispersion conversion and
635: virial mass uncertainty.
636: Here again we have a very large
637: decrease from the optical to the virial radius, which is inconsistent
638: with the flat rotation curve at more than 3-$\sigma$ level. This decrease is
639: similarly large to the one observed for the late type galaxies.
640: Such a
641: decrease cannot be explained by the dark matter alone unless halos are
642: extremely concentrated. More realistic models must include baryons,
643: which make
644: a significant contribution to the rotation velocity at optical radii,
645: both by direct contribution and by compressing the dark matter.
646:
647: The resulting velocity profiles are shown in top of figure \ref{fig3}
648: for the canonic values $\Upsilon_I=3h$ and $c_{200}=10$ (we use
649: a somewhat lower $c_{200}$ than for late types
650: since concentration is expected to decrease with halo mass).
651: The maximum rotation velocity peaks very close to the optical radius and
652: has a value of $v_{max}=270$km/s. This is in a close agreement
653: with the observed value $v_{\rm opt}=265$km/s and is well within the
654: estimated error of 30km/s, indicating that this model has
655: no problem explaining the observed ratio of
656: the optical to virial rotation velocity. Note that
657: at the optical radius the baryon and dark matter contributions are
658: comparable, while at somewhat
659: larger radii dark matter dominates. The resulting profile is much closer
660: to a constant velocity
661: SIS profile than if just light was contributing to the mass.
662: This is in a good agreement with the conclusions from strong
663: lensing \cite{1995ApJ...445..559K}, rotation velocity studies of ellipticals
664: \cite{2001AJ....121.1936G} and X-ray studies of ellipticals \cite{1999ApJ...518...50L}, which indicate that
665: flat rotation curve over the optical region ($r<10h^{-1}$kpc)
666: is a better fit to the data than a constant stellar mass to light
667: ratio with no dark matter contribution.
668:
669: \begin{figure}
670: \begin{center}
671: \leavevmode
672: \epsfxsize=3.0in \epsfbox{fig3_tf.ps}
673: \end{center}
674: \caption{Rotation velocity contributions from the spheroid (short dashed),
675: dark matter (dotted) and total (solid) for an $L_*$ early type
676: galaxy. Also shown is the
677: dark matter without adiabatic response to baryon contraction (long dashed).
678: Insert plots the same over the radii accessible with optical data,
679: showing that the profile is close to isothermal over this range.
680: }
681: \label{fig3}
682: \end{figure}
683:
684: A general exploration of $c$ versus $\Upsilon_i$ is
685: shown in figure \ref{fig6} for $L_*$. Within 1-$\sigma$ of the best fitted value
686: for $v_{\rm opt}/v_{200}$ one has $2h<\Upsilon_i<4h$ assuming $c=10$,
687: whereas at 2-$\sigma$ level this is extended to $1h<\Upsilon_i<5h$.
688: While the range of stellar mass to light ratios suggested in the
689: literature is rather large and extends even above $\Upsilon_i>5h$, such
690: high values are typically found for minimum halo models and are
691: thus an upper limit. Our analysis suggests that $\Upsilon_i<5h$
692: both because of the dynamical constraint and because of the limited baryon
693: supply. Very high values of stellar mass to light ratios are also
694: not compatible with the stellar population synthesis models using the
695: observed IMF (e.g.
696: \citeNP{1998MNRAS.294..705K}).
697:
698: \begin{figure}
699: \begin{center}
700: \leavevmode
701: \epsfxsize=3.0in \epsfbox{fig6_tf.ps}
702: \end{center}
703: \caption{
704: Same as figure \ref{fig2} for early type galaxies at $L_*$. Also shown
705: (thick dashed) is the best
706: fitted value $v_{\rm opt}/v_{200}=1.68$ together with
707: 2-$\sigma$ (95\% confidence level) lower and upper limits (thin dashed).
708: }
709: \label{fig6}
710: \end{figure}
711:
712: Since the scaling of virial velocity with luminosity $L\propto v_{200}^{2.5}$
713: \cite{gs02}
714: differs from the scaling
715: of optical velocity with luminosity $L \propto v_{\rm opt}^4$
716: above $L_*$ we must also compare at
717: a higher luminosity. We choose $L=7L_*$, which corresponds to the
718: highest luminosity bin in SDSS g-g lensing analysis \shortcite{2001astro.ph..8013M} and is dominated by
719: early type galaxies. At this luminosity
720: one finds $M_{200} \sim 10^{13}h^{-1}M_{\sun}$ \cite{gs02}, which
721: corresponds to $v_{200}=(310 \pm 30)$km/s.
722: The corresponding central velocity dispersion is
723: $\sigma_{\rm central}=290$km/s, implying $v_{\rm opt}=435$km/s
724: and $R_{\rm e}=10h^{-1}$kpc \cite{2001astro.ph.10344B}. Here the ratio is
725: $v_{max} / v_{200} =1.4 \pm 0.2$.
726: The dominating error is the virial mass at this luminosity.
727:
728: Results of the adiabatic model calculation are shown in figure \ref{fig4}
729: using $\Upsilon_{i'}=3h$.
730: We have used an even lower concentration value $c=8$ for this
731: higher mass halo, since
732: numerical simulations find $c \propto M^{-0.14}$ \cite{2001MNRAS.321..559B}.
733: We again find the rotation velocity at the optical radius
734: exceeding that at the virial radius, but the excess is significantly
735: smaller now.
736: The rotation curve is very flat for $r>5h^{-1}$kpc with the value
737: at the optical radius
738: around 445km/s, in good agreement with the observed value of 435km/s.
739: The peak value of the dark matter rotation curve alone is around 400km/s
740: and is close to the value at the optical radius
741: given that the rotation curve is so flat.
742: Note that at this luminosity the dark matter dominates already at the
743: optical radius and it is only below 5$h^{-1}$kpc that baryon mass
744: exceeds that of the dark matter. However, the baryons do have a significant
745: effect on the dark matter through the adiabatic compression, so that
746: even though the dark matter dominates at the optical radius it would
747: have been comparable to baryons if there was no dark matter compression.
748: Our analysis suggests
749: that the data are consistent with
750: brighter ellipticals being more dark matter dominated at the
751: optical radii than the fainter ones and the stellar mass to light
752: ratio not varying with luminosity.
753:
754: \begin{figure}
755: \begin{center}
756: \leavevmode
757: \epsfxsize=3.0in \epsfbox{fig4_tf.ps}
758: \end{center}
759: \caption{Same as figure \ref{fig3} for 7$L_*$ early type galaxy.
760: }
761: \label{fig4}
762: \end{figure}
763:
764: It is worth noting that the model can be extended by yet
765: another order of magnitude in virial mass, to cluster masses
766: with an elliptical cD galaxy at the
767: center. A typical example is Virgo cluster with M87 at the center.
768: For M87/Virgo the rotation velocity
769: increases from the optical radius $r_{\rm E}\sim 5h^{-1}$kpc
770: value $v_{\rm opt}\sim 450$km/s \cite{2001ApJ...553..722R},
771: to around 950km/s deduced from the observed
772: X-ray temperature around 3keV.
773: An example that can reproduce these constraints is shown in
774: figure \ref{fig5}, where a halo with $M_{200}=2\times 10^{14}h^{-1}M_{\sun}$,
775: $c=6$ and spheroid with
776: $M_{\rm stellar}=2\times 10^{11}h^{-1}M_{\sun}$ have been used
777: \cite{1999ApJ...524L..15G}. In this case
778: one has a ratio in
779: rotation velocity from optical to virial radius $v_{\rm opt}/v_{200}\sim 0.5$.
780: This is caused primarily by the cluster halo profile
781: becoming shallower than $r^{-2}$ in the inner parts. While baryons
782: in cD
783: increase the rotation velocity from the pure dark matter value in the inner
784: parts of the halo they cannot
785: pull in outer parts of the dark matter halo, which remain unaffected.
786: The trend of optical to virial rotation velocity
787: ratio decreasing with halo mass observed in elliptical galaxies
788: is therefore expected as a consequence of more
789: massive halos being less concentrated and less steep than isothermal in
790: the central region, although for quantitative predictions baryons and
791: their effect on dark matter must also be included.
792:
793: \begin{figure}
794: \begin{center}
795: \leavevmode
796: \epsfxsize=3.0in \epsfbox{fig5_tf.ps}
797: \end{center}
798: \caption{Same as figure \ref{fig3} for $2\times 10^{11}h^{-1}M_{\sun}$
799: early type galaxy in
800: a $2\times 10^{14}h^{-1}M_{\sun}$ halo as a model for M87/Virgo.
801: }
802: \label{fig5}
803: \end{figure}
804:
805:
806: \section{Discussion}
807:
808: In this paper we compare average properties of
809: rotation velocities between optical and virial radii based on
810: optical and galaxy-galay lensing measurements, respectively.
811: Our model is statistical in nature, since
812: it is not based on analysis of individual galaxies, but on their
813: average properties as a function of luminosity. On the other hand,
814: the galaxies in our sample were not chosen on the basis of any
815: selection criteria, so our results should apply
816: to the galaxy population as a whole. Moreover, the large dynamical range
817: between optical region (of order a few kpc) and virial region
818: (of order a few hundred kpc) allows one to measure slow changes in rotation
819: velocity which are not possible to detect from each of the observations
820: individually.
821:
822: Our main conclusion is that the rotation velocity
823: in galaxies
824: decreases significantly from optical radii to virial radii.
825: Such a decrease is theoretically expected since
826: the dark matter profiles in CDM models are steeper than
827: isothermal at large radii and contribution from
828: stars further increases the rotation
829: velocity at the optical radius. It has however
830: not been clearly demonstrated previously
831: because of a narrow range of scales observed. This demostrates the power of
832: g-g lensing measuring the mass at large radii
833: combined with the more traditional methods that measure mass at
834: smaller radii (see also \citeNP{2001ApJ...555..572W} for
835: a similar conclusion for early type galaxies at higher redshifts).
836: The concentration parameters obtained using reasonable stellar mass to
837: light ratios are around $c\sim 10$ and are
838: in a good agreement with predictions of
839: $\Lambda$CDM model with $\Omega_m=0.3$ and $\sigma_8=0.8$, which
840: gives $c_{200} \sim 10$ at $M \sim 10^{12}h^{-1}
841: M_{\sun}$ \shortcite{2001ApJ...554..114E}.
842: We find no evidence for low halo concentrations in the main galaxy
843: population both for early and late type galaxies.
844:
845: The theoretically predicted
846: stellar mass to light ratios have considerable theoretical uncertainties.
847: Assuming CDM halo profiles and concentrations we find $\Upsilon_I \sim
848: 1.5h$ for late type galaxies and $\Upsilon_i \sim 3h$ for early type
849: galaxies. These values are in a good agreement with the stellar population
850: synthesis models (e.g. \citeNP{1998MNRAS.294..705K},
851: \shortciteNP{2002astro.ph..1207D}). While higher stellar mass to light ratios are often
852: quoted in the literature (e.g. \shortciteNP{2001AJ....121.1936G}) these are based on minimal halo
853: models and are thus often an upper limit. It is interesting to note that
854: in our models the dark matter contribution to rotation velocity is comparable
855: to that of stars even below effective radius. This is because of the
856: adiabatic response, which compresses the dark matter in the center. While
857: this model may be oversimplified it nevertheless suggests
858: that it may be difficult or
859: impossible to separate the two components on the
860: basis of dynamical studies. Using the minimal halo assumption the
861: stellar mass to light ratio in early type galaxies may be overestimated
862: up to a factor of two.
863:
864: While the optical to virial velocity ratio
865: is above unity both for early and late type galaxies,
866: it is decreasing with halo mass from 1.8 at $3\times 10^{11}h^{-1}M_{\sun}$
867: to 1.4 at $10^{13}h^{-1}M_{\sun}$ (this statement is valid only for galaxies
868: above $L_*$ and it is possible that the trend is reversed towards the
869: low luminosity galaxies). This trend continues further into the
870: cluster halo masses, where the ratio falls below unity.
871: Such a trend is expected in models where halo profiles are less
872: concentrated for higher halo masses, implying that the turnaround
873: from an increase to a decrease in rotation velocity occurs at a larger
874: radius relative to the virial radius. In addition, in more massive halos
875: stars play a less important role both as a direct contribution to
876: the rotation velocity and through their effect on the
877: dark matter.
878: The rotation velocity-luminosity scalings at optical
879: radii, such as Tully-Fisher and Faber-Jackson relations,
880: are not directly related to
881: the properties of dark matter, but also require a proper modelling of
882: baryons and dark matter response to baryonic contraction (see also a
883: related discussion in \shortciteNP{2000ApJ...528..145G}, \citeNP{2001astro.ph..8160K} and \citeNP{2001astro.ph.12566V}).
884: While there are still uncertainties in the modelling
885: of these processes, the simple models presented here reproduce
886: well the constraints from the data both for early and late
887: type galaxies.
888:
889: How do our results compare to previous work?
890: The zero point of TF relation problem (\shortciteNP{2001ApJ...554..114E},
891: \citeNP{2000MNRAS.318..163M})
892: is the closest to the TF
893: analysis done here. In the absence of virial mass information
894: the value of rotation velocity at a given luminosity
895: does not suffice to make any general conclusions, so
896: in general one has to make additional
897: assumptions and/or modelling.
898: For example, the stellar mass fraction in the halo $f_*$ obtained
899: in previous work was lower, which lead to a higher virial mass
900: for a given luminosity,
901: which in turn requires lower concentrations and/or stellar mass to light
902: ratios. By increasing $f_*$ close to its maximum value
903: the virial mass can be reduced and this
904: alleviates the problem. The same solution also solves the suggested
905: overprediction of dark matter at the solar radius in our own galaxy
906: \cite{2001ApJ...554..114E},
907: since again if the stellar fraction is higher the virial mass
908: can be lower (there may be
909: additional problems for CDM profiles in the inner parts of our galaxy;
910: e.g. \citeNP{2001MNRAS.327L..27B}).
911:
912: For early type galaxies
913: it has been suggested that the concentrations
914: are low from the strong lens statistics \cite{2001ApJ...561...46K}, since very
915: concentrated halos would overpredict the expected number of lenses. This is
916: a difficult method to use since the expected number of lenses is very
917: sensitive to the assumed luminosity function
918: for early type galaxies as a function of redshift, which still has
919: considerable uncertainty. Additional uncertainty arises from the
920: adopted values for stellar mass to light ratios, which again can change
921: the lensing statistics significantly. The lensing results are
922: still compatible with low density CDM models, suggesting
923: there is no discrepancy with our results, although more work is required to
924: study this in detail.
925:
926: There are other problems that have been suggested as troublesome for CDM,
927: such as detailed shapes of velocity profiles in the optical region
928: \cite{2001ApJ...552L..23D}, bar rotation \cite{2000ApJ...543..704D}
929: or the halo
930: structure of the Milky Way \cite{2001MNRAS.327L..27B}. These probe inner
931: regions of the galaxy where complicated physical processes may be
932: taking place, so there is considerable more uncertainty in their
933: theoretical predictions. For example,
934: there are processes such as bar rotation that can disrupt dark matter
935: cusps \cite{2001astro.ph.10632W}.
936: It has recently been shown that CDM profiles can
937: fit most of the rotation curves for normal galaxies \cite{2002astro.ph..1352J}.
938: The galaxies that appear to be a problem for CDM belong to
939: one of the specific subsamples,
940: such as low surface brightness, dwarf or barred galaxies.
941: It is possible, although not necessarily easy to arrange, that
942: these samples are qualitatively different from the main
943: population, for example by forming later and thus being less concentrated.
944: Yet another possibility is that problems arise only
945: below $L_*$, since our analysis is only valid for galaxies around
946: and above $L_*$. Clearly, more
947: work is required to resolve these issues. However,
948: if the g-g lensing masses are correct, then for the main population of galaxies
949: around and above $L_*$
950: the CDM model predictions for the
951: amount of dark matter outside the inner few kpc
952: do not exceed the observations,
953: suggesting that the problems for CDM may not be as fundamental as previously
954: suggested.
955:
956: The author acknowledges the support of NASA, David and Lucille
957: Packard Foundation and Alfred P. Sloan Foundation.
958: I thank Mariangela Bernardi, \v Zeljko Ivezi\' c, Guinevere Kauffmann, Ravi
959: Sheth and
960: Tommaso Treu for useful discussions and Matthias Steinmetz for providing
961: their manuscript prior to publication.
962:
963: \bibliography{apjmnemonic,cosmo,cosmo_preprints}
964: % \bibliographystyle{apj}
965: \bibliographystyle{mnras}
966:
967: \end{document}
968: