astro-ph0503397/ms.tex
1: \documentclass[12pt,preprint]{aastex}
2: \newcommand{\hst}{{\sl HST}}
3: \newcommand{\vis}{$V_{606}$}
4: \newcommand{\nir}{$I_{814}$}
5: \slugcomment{ApJL accepted on 17 March 2005}
6: \shorttitle{The the WDs Cooling sequence in NGC~6791}
7: \shortauthors{Bedin et al.}
8: \begin{document}
9: \def\subr #1{_{{\rm #1}}}
10: \title{THE WHITE DWARF COOLING  SEQUENCE IN NGC\ 6791}
11: 
12: \author{ Luigi R.\ Bedin\altaffilmark{1,2},  
13:          Maurizio Salaris\altaffilmark{3},  
14:          Giampaolo Piotto\altaffilmark{2}, 
15:          Ivan R.\ King\altaffilmark{4}, 
16:          Jay Anderson\altaffilmark{5}, 
17:          Santi Cassisi\altaffilmark{6}, 
18:          \and 
19: 	 Yazan Momany\altaffilmark{2} 
20:          }
21: 
22: \altaffiltext{1}{European             Southern            Observatory,
23: Karl-Schwarzschild-Str.\ 2, D-85748 Garching, Germany; lbedin@eso.org}
24: 
25: \altaffiltext{2}{Dipartimento  di Astronomia, Universit\`a  di Padova,
26: Vicolo dell'Osservatorio 2, I-35122 Padova, Italy; piotto-momany@pd.astro.it}
27: 
28: \altaffiltext{3}{Astrophysics Research Institute, Liverpool John Moores
29: Univ., 12 Quays House, Birkenhead, CH41 1LD, UK; ms@astro.livjm.ac.uk}
30: 
31: \altaffiltext{4}{Department of Astronomy, University of Washington,
32: Box 351580, Seattle, WA 98195-1580; king@astro.washington.edu}
33: 
34: \altaffiltext{5}{Department of Physics and Astronomy, Mail Stop 108,
35: Rice University, 6100 Main Street, Houston, TX 77005;
36: jay@eeyore.rice.edu}
37: 
38: \altaffiltext{6}{INAF-Osservatorio Astronomico di Collurania,
39: via M. Maggini, 64100 Teramo, Italy;
40: cassisi@te.astro.it}
41: 
42: \begin{abstract}
43: % 
44: In the  old, populous, and metal-rich  open cluster NGC\  6791 we have
45: used deep  \hst/ACS images to  track the white dwarf  cooling sequence
46: down  to  $m_{\rm   F606W}\simeq28.5$.   The  white  dwarf  luminosity
47: function shows a well defined peak at $m_{\rm F606W}\simeq27.4$, and a
48: bending to  the blue  in the color--magnitude  diagram.  If  this peak
49: corresponds  to  the end  of  the  white  dwarf cooling  sequence  the
50: comparison with theoretical isochrones provides a cluster age estimate
51: of $\sim 2.4$ Gyr, in sharp contrast with the age of 8--9 Gyr inferred
52: from the main-sequence turn-off.  If the end is at fainter magnitudes,
53: the  peak at  $m_{\rm F606W}\simeq27.4$  is even  more  enigmatic.  We
54: discuss possible causes, none of them very convincing.
55: % 
56: \end{abstract}
57: 
58: \keywords{open  clusters and associations:  individual (NGC\ 6791) ---
59:   white dwarfs}
60: 
61: %%%%%%%%%%%%%%%%%%%%%%
62: %
63: \section{Introduction}
64: %
65: %%%%%%%%%%%%%%%%%%%%%%
66: 
67: NGC\ 6791 is  one of the richest open clusters, with  age $>8$ Gyr and
68: ([Fe/H]  $\sim$ +0.4).  Because  of these  unusual properties,  it has
69: been the target of several studies, the most recent by Carney, Lee, \&
70: Dodson (2005).
71:  
72: NGC\ 6791 is close enough that  \hst/ACS imaging can hope to reach the
73: bottom  of  the  white  dwarf  (WD) cooling  sequence,  and  therefore
74: estimate  its age  independently  of the  main-sequence (MS)  turn-off
75: (TO).  For  clusters older than  3 Gyr this  kind of age  estimate has
76: previously been attempted only in the $\sim$4-Gyr-old open cluster M67
77: (Richer et  al.\ 1998),  and the globular  cluster M4 (Hansen  et al.\
78: 2002, 2004).  In  the latter case the end of  the cooling sequence was
79: not detected, and the age estimate  was based on modeling the shape of
80: the observed part of the WD luminosity function (LF).
81:  
82: The data on which this letter  is based were originally taken to reach
83: the bottom of  the MS in NGC\ 6791,  but when we first saw  the CMD it
84: was clear that the WD cooling sequence (WDCS) was equally exciting.  A
85: companion paper will describe  the color--magnitude diagram and the MS
86: LF of  NGC 6791,  and compare the  observations with models.   Here we
87: focus on the WD sequence.
88: 
89: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
90: %
91: \section{Observations and Measurements}
92: %
93: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
94: 
95: The  data were  taken on  July 16--17  2003, with  \hst's  ACS/WFC (GO
96: 9815), and  consist of 6 long  ($\sim$1150s) and 3  short exposures of
97: $\sim$30s through  each of  the filters F606W  and F814W.   A detailed
98: discussion  of  the   data  reduction,  photometric  calibration,  and
99: artificial star  tests is presented in our  companion paper.  Briefly,
100: the images in  each filter were corrected for  distortion according to
101: Anderson  (2002),  and combined  into  a  stack  in which  pixels  are
102: sub-sampled  by  a  factor   of  two  to  remove  undersampling.   The
103: photometry  was  performed with  DAOPHOT  (Stetson 1987).   Artificial
104: stars were  added in order  to determine the completeness  curve.  The
105: calibration to the ACS Vega-mag  system was done following the recipes
106: in Bedin et al.\ (2005).
107:  
108: Figure 1  summarizes our observational results.  Panel  $a)$ shows the
109: color--magnitude  diagram   (CMD),  while  panel   $b)$  displays  the
110: completeness-corrected WD  LF.  There  is a clear  peak of  density at
111: $m_{\rm  F606W}\sim27.4$, after  which the  LF drops,  but  stays well
112: above  zero down  to $m_{\rm  F606W}\sim28.2$, where  our completeness
113: becomes $\le  50$\%.  The peak in  the LF coincides with  a bending in
114: the CMD of  the WDCS.  At $m_{\rm F606W}\sim28.2$ there  seems to be a
115: second peak,  but its significance remains to  be investigated.  Panel
116: $c)$ shows  the input  (continuous line) and  the output CMD  from the
117: artificial-star  experiments, and panel  $d)$ the  corresponding joint
118: (two filters) completeness curve for WDs.  Panels $e)$ and $f)$ show a
119: blow-up  of  the CMD  focused  on  the  observed and  artificial  WDs,
120: respectively.
121:  
122: It  is evident from  panels $b)$  and $d)$  that at  the level  of the
123: density peak  in the WD LF  the completeness is  still $\sim85$\%, and
124: therefore the peak and the bending to the blue are real.
125: %
126: At $\ell\sim70^\circ$, $b\sim+11^\circ$  we expect the contribution of
127: field WDs in our small field to be negligible, and because we rejected
128: all non-point-like sources by  visual inspection, we expect the number
129: of contaminating galaxies also to be negligible.
130: 
131: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
132: %
133: \section{Comparison with theory}
134: %
135: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
136: 
137: In Fig.~2,  DA WD isochrones for  ages of 2, 4,  6, 8, and  10 Gyr are
138: compared with the observed WD  sequence.  The isochrones came from the
139: WD models by Salaris et al.\ (2000), transformed into the ACS Vega-mag
140: system  using   the  most  up-to-date   on-orbit  transmission  curves
141: (Sirianni et al.\ 2005), and color transformations generously provided
142: by  P.\ Bergeron,  based on  the model  atmospheres used  in Bergeron,
143: Leggett, \& Ruiz  (2001).  In addition to the  cooling models, we have
144: adopted  the empirical relation  between WD  mass and  progenitor mass
145: initial-final   mass  relation,  IFMR)   by  Weidemann   (2000).   The
146: progenitor lifetimes come from the Pietrinferni et al.\ (2004) grid of
147: stellar models.
148:  
149: In the WD model-fitting we adopted the distance and reddening found in
150: our companion  paper, where we showed that  the best fit to  the MS TO
151: and  subgiant branch  (reproduced  also in  Fig.~2)  is obtained  with
152: [Fe/H] = 0.26,  an apparent modulus $(m-M)_{\rm F606W}=  13.44$, and a
153: reddening $E(m_{\rm F606W}-m_{\rm F814W})= 0.14$.  These correspond to
154: $(m-M)_V$=13.50 and $E(B-V)$=0.15 in the Johnson band-passes (Bedin et
155: al.\ 2005).   Notably, the age derived  from the fitting of  the TO is
156: between 8  and 9  Gyr, as  discussed in more  detail in  our companion
157: paper.   All these  values  are  in broad  agreement  with the  recent
158: analysis by Carney et al.\ (2005).
159: 
160: The brighter parts of our  WD isochrones are populated by objects with
161: mass  of  $\sim$0.54 $m_{\odot}$---i.e.,  WDs  produced  by the  stars
162: presently at  the asymptotic giant  branch (AGB) tip.   The isochrones
163: reproduce  well the  main body  of  the observed  WD sequence  between
164: $m_{\rm F606W}\sim 24$  and $m_{\rm F606W}\sim 26$.  A  few objects in
165: this magnitude  range lie  to the red  side of  the main locus  of the
166: WDCS, most probably a mixture  of field contamination, He-core WDs and
167: possibly unresolved binary WD-WD systems.
168:  
169: The bottom of  the WD isochrones shows  a turn to the blue  due to the
170: appearance  of larger  WD masses,  from  the more  massive stars  that
171: evolved off the AGB early  in the cluster's history. The brightness of
172: this turn to the blue is a function of cluster age (see, e.g., Salaris
173: et al.\ 2000, Richer et al.\ 2000).  It is clear from Fig.~2 that even
174: the bottom of the 6 Gyr WD isochrone is at a brightness where the peak
175: in the WD population has already dropped.
176:  
177: Figure 2  highlights {\it a  potentially serious mismatch  between the
178: MSTO  and WD  ages for  NGC\ 6791}.   For comparison  we also  show in
179: Fig.~2  the  location  of  the   sample  of  cool  WDs  in  the  solar
180: neighborhood  (open  squares) analyzed  by  Bergeron  et al.\  (2001),
181: translated into the same  ACS Vega-mag system with the transformations
182: adopted  in this  paper.  As  expected,  the local  WDs reach  fainter
183: magnitudes than  the bulk  of the cluster  WDCS, down to  a brightness
184: consistent with our 8 Gyr isochrone.
185:  
186: To investigate further this age  mismatch, we consider again the LF of
187: the cluster WDCS  (see Fig.~3 {\it top} panel),  with its well-defined
188: peak at $m_{\rm  F606W}\sim$ 27.4.  From a theoretical  point of view,
189: one can determine  a LF for a given age  starting from the appropriate
190: WD  isochrone and  adopting an  initial  mass function  (IMF) for  the
191: progenitors (see, e.g., Richer et al.\ 2000); all theoretical LFs show
192: a peak at the faint end  that moves toward fainter luminosities as the
193: cluster  age  increases, reflecting  the  behavior  of the  underlying
194: isochrones.  This peak is caused by the ``piling up'' at the bottom of
195: the WD  isochrone of WDs  of all masses,  due to their  finite cooling
196: age.   If one uses  a power-law  IMF ($dn/dm\propto  m^{-(1+x)}$), the
197: exact value of  the exponent $x$ ($x=1.35$ for  Salpeter IMF) does not
198: affect the position  of the LF peak.  The position of  the peak of the
199: observed WDCS can  be reproduced only by assuming  an age below 3~Gyr,
200: in  clear disagreement with  the estimated  TO age  of about  8--9 Gyr
201: ($\sim$1.5~mag  fainter).   A  sub-population  of stars  with  age  of
202: $\sim$3 Gyr  can be excluded  by the fact  that we do not  observe the
203: corresponding TO.
204:  
205: Further  evidence that we  are facing  a serious  problem is  given in
206: Fig.~3 ({\it middle}  and {\it bottom} panels), which  compares the WD
207: LF of  the solar-metallicity  open cluster M67  (Richer et  al.\ 1998)
208: with our NGC 6791 WD LF.   The NGC\ 6791 WD data have been transformed
209: into  the   Johnson-Cousins  system  from  the  ACS   one,  using  the
210: transformations by  Sirianni et al.\ (2005), which  should be accurate
211: to a  few hundredths  of a magnitude.   The distance moduli  have been
212: subtracted from both LFs (we used values in Percival \& Salaris 2003),
213: and  the NGC\ 6791  LF has  been rebinned  in order  to match  the M67
214: binning.  The  M67 LF also shows  a peak, which  most probably denotes
215: the bottom of  the DA WD sequence.  Surprisingly  enough, the peaks in
216: the two clusters  are at the same absolute magnitude.   But the TO age
217: of M67 is
218: %
219: only $\sim$4 Gyr  (see, e.g., Salaris, Weiss \&  Percival 2004), i.e.,
220: about half the NGC\ 6791 age.  With our WD isochrones, using lifetimes
221: for solar-metallicity progenitors, we  can reproduce the observed peak
222: in the M67  WD LF adopting an  age of 3.5 Gyr, consistent  with its TO
223: age. The same  result is obtained by Richer et  al.\ (1998, 2000) with
224: their own WD  isochrones.  The fact that the TO and  WD ages agree for
225: M67 makes the disagreement in NGC\ 6791 even more puzzling.
226: 
227: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
228: %
229: \section{Discussion}
230: %
231: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
232: 
233: As shown in  the previous paragraph, the nature of the  the WD LF peak
234: at  $m_{\rm F606W}\simeq27.4$  is enigmatic.   We are  faced  with two
235: possibilities:\ 1) either the peak  represents the bottom of the WDCS,
236: and in this  case we need to  understand why the WD cooling  age is in
237: sharp contrast with the age inferred from the MS TO; or 2) the peak is
238: not due  to the classical  DA WDs.  In  the following we  will further
239: discuss the consequences of these two hypotheses.
240:  
241: Let us  first examine  the possibility that  the observed  peak indeed
242: represents the bottom  of the WDCS of NGC\ 6791.   A wrong cooling age
243: could simply come from having adopted a wrong distance modulus, but an
244: apparent distance modulus smaller by 1.5 mag would imply a TO age well
245: above 20  Gyr.  On the other hand,  we cannot exclude that  the TO age
246: could be  wrong because the  adopted metallicity is wrong,  though the
247: most  likely  metallicity  interval  (Taylor 2001)  cannot  solve  the
248: discrepancy.
249: 
250: Another possibility  is an  IFMR radically different  from the  one we
251: used.  However, as we have verified with various numerical tests, none
252: of the proposed IFMR in Fig.\  11 of Hansen et al.\ (2004) helps solve
253: the problem.  As noted by the referee (H.\ Richer), the WD LF shows an
254: apparent gap  (of low  statistical significance, however),  at $M_{\rm
255: F606W}  \simeq12.2$,  which  could  correspond to  a  similar  feature
256: identified  in  M4  at   $M_{\rm  F606W}\simeq13.2$  (Hansen  et  al.\
257: 2004)---a full  magnitude fainter. Is  this further evidence  that the
258: cooling models  specific for NGC\  6791 fail?  In the  magnitude range
259: covered  by the  bottom  of the  observed  WD sequence  all modern  WD
260: cooling models---including  the Salaris et al.\ (2000)  models we have
261: adopted  here---provide   very  similar  cooling   times  (see,  e.g.,
262: Prada-Moroni \& Straniero 2002).
263: 
264: Another possibility  of explaining  this age mismatch  is to  invoke a
265: wrong CO  chemical stratification in the  WD core of  the models (that
266: underestimates  the  energy  available  during  the  cooling  and  the
267: crystallization process) and/or a wrong atmospheric H-layer thickness.
268: However,  the most  recent  determination of  the $\rm  C^{12}(\alpha,
269: \gamma)O^{16}$ reaction  rate would not  solve the problem,  nor would
270: the largest possible value of $\Delta m_{\rm H}$ for the H-layer.
271: 
272: The CMD of Fig.~1 clearly shows  that NGC\ 6791 has a large population
273: of binaries.   It might be that  the LF peak  is made of DA  WDs whose
274: progenitors  were  produced by  merging  (or  stripping) in  binaries,
275: possibly during their MS phase.  If these mergers occur about 2--3~Gyr
276: after the birth  of the cluster and the  resulting objects have masses
277: of $\sim$1.5~$m_{\odot}$,  their WD progeny would  populate the region
278: of the LF  peak for a cluster age of about  8~Gyr.  The counterpart of
279: these merged progenitors might  be the blue stragglers (BSs).  Davies,
280: Piotto, \& De Angeli (2004), based on observational evidence by Piotto
281: et  al.\ (2004),  show that  the number  of BSs  coming  from evolving
282: binaries should have been much larger in the first few Gyr of the life
283: of  the cluster.   We  also  note that  the  HB of  NGC\  6791 has  an
284: anomalous morphology (Yong  et al.\ 2000). The origin  of this anomaly
285: is not at all clear, and it  is even less clear how it propagates into
286: the WD regime.
287: 
288: Let us  now explore the case  that the LF peak  is not due  to DA WDs.
289: WDs with a degenerate He-core and masses of 0.3--0.4 $m_{\odot}$, from
290: progenitors of $\sim$1.3--1.5 $m_{\odot}$,  can populate the region of
291: the observed peak if we adopt  a cluster age of $\sim$8 Gyr, according
292: to  the He-core WD  models by  Serenelli et  al.\ (2001).   However, a
293: substantial contamination by He-core  WDs can be excluded because, due
294: to their lower mass, they are much redder than the bulk of the WDCS we
295: observe.  We  explicitly verified this point by  transforming into the
296: ACS Vega-mag photometric system the  He-core WD models by Serenelli et
297: al.\  (2001) for  masses of  0.32 and  0.41 $m_{\odot}$  (see Fig.~2).
298: Moreover, using as  normalizing factor the number of  stars on the MS,
299: and adopting a Salpeter IMF, we  estimated that the total number of CO
300: WDs in our  field should be $\sim$430, which  compares reasonably well
301: with the $\sim$600  objects observed.  Though ignorance of  the IMF at
302: higher  masses,  and  mass   segregation  effects,  make  this  number
303: extremely  uncertain,  there  seems  little  room  for  a  substantial
304: population  of  He-core  objects.    For  DA  WD  luminosities  around
305: $\log(L/L_{\odot})\sim -4.0$, non-DA objects are slightly fainter than
306: DAs ($\sim$0.5 mag,  but the exact value depends on  the WD mass), and
307: the  difference  increases  fast  with decreasing  luminosity.   These
308: objects may  contribute to the possible second  peak around $\sim$28.2
309: but, again,  only for ages  around 3~Gyr.  Also, the  possibility that
310: the peak comes from a cluster  of far, unresolved blue galaxies can be
311: excluded,  as the  spatial distribution  of  the observed  WDs is  not
312: significantly different from that of the cluster MS stars.
313:  
314: We also note that the WD LF  of Fig.~1b never drops to zero. This fact
315: could indicate that we have not  reached the end of the WDCS.  In this
316: case the presence of a peak in the WDCS may also indicate the onset of
317: an  unknown  energy  release  at  this specific  age  and  metallicity
318: (sedimentation/separation  of CO,  pycnonuclear reaction,  or whatever
319: can cause a temporary slowdown in the cooling).
320: 
321: Finally,  it is  worth  remarking that  NGC\  6791 is  a cluster  with
322: several peculiarities (high binary  fraction, anomalous HB, very metal
323: rich,  dynamically  old, etc.)   that  may  have  altered the  stellar
324: populations in  the cluster,  and hence  it is too  early to  call for
325: major  revisions in  the WD  formation  and evolution  models.  It  is
326: surely of great interest to  extend the present investigation at least
327: down to the  magnitude where we expect  to see the peak of  a 8--9 Gyr
328: WDCS.
329: 
330:  
331: \acknowledgments
332: 
333: We are grateful to Pierre  Bergeron and to Alvio Renzini.  I.R.K.\ and
334: J.A.\  acknowledge support by  STScI grant  GO-9815.  S.C.,  G.P., and
335: Y.M.\ acknowledge support by MIUR under PRIN2003.
336: 
337: \clearpage
338: 
339: \begin{thebibliography}{}
340: 
341: \bibitem[]{} Anderson, J. 2002, HST Calibration  Workshop 2002, ed.\ S.\ Arribas, A.\ Koekemoer, B.\ Whitmore (Baltimore:\ STScI), p.\ 13
342: \bibitem[]{} Anderson, J., \& King, I.\ R. 2000, \pasp, 112, 1360 
343: \bibitem[]{} Bedin, L.\ R., et al. 2005, \mnras, 357, 1038 
344: \bibitem[]{} Bergeron, P., Leggett, S.\ K., \& Ruiz, M.\ T. 2001, \apjs, 133, 413 
345: \bibitem[]{} Carney, B.\ W., Lee, J-W., \& Dodson, B. 2005, AJ, 129, 656
346: \bibitem[]{} Davies, M., Piotto, G., \& De Angeli, F.  2004, \mnras, 349, 129
347: \bibitem[]{} Hansen, B.\ M.\ S., et al.\ 2002, \apjl, 574, 155
348: \bibitem[]{} Hansen, B.\ M.\ S., et al.\ 2004, \apjs, 155, 551 
349: \bibitem[]{} Percival, S.\ M., \& Salaris, M. 2003, \mnras, 343, 539 
350: \bibitem[]{} Pietrinferni, A., Cassisi, S., Salaris, M., \& Castelli, F. 2004, \apj, 612, 168 
351: \bibitem[]{} Piotto, G., et al. 2004, \apjl, 604, 109
352: \bibitem[]{} Prada-Moroni, P.\ G., \& Straniero, O. 2002, \apj, 581, 585
353: \bibitem[]{} Richer, H.\ B., et al.\ 1998, \apjl, 504, 91
354: \bibitem[]{} Richer, H.\ B., et al.\ 2000, \apj, 529, 318 
355: \bibitem[]{} Salaris, M., et al. 2000, \apj, 544, 1036 
356: \bibitem[]{} Salaris, M., Weiss, A., \& Percival, S.\ M. 2004, \aap, 414, 163
357: \bibitem[]{} Serenelli, A.\ M. et al. 2001, \mnras, 325, 607
358: \bibitem[]{} Sirianni, M., et al. 2005, \pasp, submitted
359: \bibitem[]{} Stetson, P.\ B., 1987, PASP, 99, 191
360: \bibitem[]{} Taylor, B. \ J. 2001, \aap, 377, 473
361: \bibitem[]{} Weidemann, V. 2000, \aap, 188, 74
362: \bibitem[]{} Yong, H., Demarque, P, \& Yi, S. 2000, \apj 539, 928
363: \end{thebibliography}
364: 
365: \clearpage
366: 
367: \begin{figure}
368: \epsscale{1.00}
369: \plotone{f1.eps}
370: \caption{The  CMD and  WD  LF  of NGC\  6791  and the  artificial-star
371: results (see  text).  The broadening in  color of the  observed CMD at
372: the  level of the  WD peak  is $\sim60\%$  larger than  for artificial
373: stars.}
374: \end{figure}
375: 
376: \begin{figure}
377: \epsscale{1.00}
378: \plotone{f2.eps}
379: \caption{The MS and the WDCS of NGC\ 6791, fitted with 2, 4, 6, 8, and
380: 10 Gyr isochrones, with the  same distance modulus and reddening.  The
381: lines on  the red side of the  WD isochrones indicate the  loci of WDs
382: with pure helium cores of 0.32 and 0.41 $m_\odot$.}
383: \end{figure}
384: 
385: \begin{figure}
386: \epsscale{1.00}
387: \plotone{f3.eps}
388: \caption{ ({\it  Top} panel)  The completeness-corrected LF,  with the
389: theoretical LFs  for 2.4 and  8 Gyr overplotted.  The  vertical dotted
390: line is the  50\% completeness level.  The WD LF  of M67 ({\it middle}
391: panel) and that NGC\ 6791 ({\it  bottom} panel) have peaks at the same
392: absolute magnitude, corresponding to an age $<$4 Gyr.}
393: \end{figure}
394: 
395: \end{document}
396: