0801.2598/ms.tex
1: %                                                                 aa.dem
2: % AA vers. 6.1, LaTeX class for Astronomy & Astrophysics
3: % demonstration file
4: %                                                 (c) Springer-Verlag HD
5: %                                                revised by EDP Sciences
6: %-----------------------------------------------------------------------
7: %
8: %\documentclass[referee]{aa} % for a referee version
9: %\documentclass[onecolumn]{aa} % for a paper on 1 column
10: %\documentclass[longauth]{aa} % for the long lists of affiliations
11: \documentclass[rnote]{aa} % for the research notes
12: %\documentclass[letter]{aa} % for the letters
13: %
14: %\documentclass[rnote]{aa}
15: %
16: \usepackage{graphicx}
17: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
18: \usepackage{txfonts}
19: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
20: %
21: 
22: 
23: \begin{document}
24: %
25: 
26:    \title{Is the anti-correlation between the X-ray variability amplitude and black hole mass of AGNs intrinsic?}
27: 
28:    %\subtitle{anti-correlation between the X-ray variability amplitude and the luminosity}
29: 
30:    \author{Yuan Liu
31:           \inst{1}
32:           \and
33:           Shuang Nan Zhang\inst{1}
34:           }
35: 
36:    %\offprints{G. Wuchterl}
37: 
38:    \institute{Physics Department and Center for Astrophysics, Tsinghua University, Beijing, 100084,
39:                 China\\
40:               \email{zhangsn@tsinghua.edu.cn, yuan-liu@mails.tsinghua.edu.cn}
41:                       %\thanks{}
42:              }
43: 
44:    \date{}
45: 
46: % \abstract{}{}{}{}{}
47: % 5 {} token are mandatory
48: 
49:   \abstract
50:   % context heading (optional)
51:   % {} leave it empty if necessary
52:    {}
53:   % aims heading (mandatory)
54:    {Both the black hole mass and the X-ray luminosity of AGNs have been found to be
55:    anti-correlated with the normalized excess variance ($\sigma _{\rm rms}^2 $) of the X-ray light curves.
56:    We investigate which correlation with $\sigma _{\rm rms}^2 $
57:  is the intrinsic one.}
58:   % methods heading (mandatory)
59:    {We divide a full sample of 33 AGNs (O' Neill et al. 2005) into two sub-samples.
60:    The black hole masses of 17 objects in sub-sample 1 were determined by
61:    the reverberation mapping or the stellar velocity dispersion. The black hole masses
62:     of the remaining 16 objects were estimated from the relationship between broad
63:     line region radius and optical luminosity (sub-sample 2). Then partial
64:      correlation analysis, ordinary least squares regression and K-S tests are performed on the full sample and the sub-samples, respectively.}
65:   % results heading (mandatory)
66:    {We find that $\sigma _{\rm rms}^2 $ seems to be intrinsically correlated with the black hole
67:    mass in the full sample. However, this seems to be caused by including
68:    the sub-sample 2 into the analysis, which introduces an extra
69:    correlation between the black hole mass and the luminosity and strengthens any correlation with the black hole mass
70:     artificially. Therefore,
71:    the results from the full sample may be misleading. The results from the
72:     sub-sample 1 show that the correlation between $\sigma _{\rm rms}^2 $ and the X-ray luminosity
73:     may be the intrinsic one and therefore the anti-correlation between
74:     $\sigma _{\rm rms}^2 $ and the black hole mass is doubtful.}
75:   % conclusions heading (optional), leave it empty if necessary
76:    {}
77: 
78:    \keywords{X-rays: galaxies -- galaxies: active -- methods: statistical }
79: 
80:    \authorrunning{Yuan Liu and Shuang Nan Zhang}
81:    \titlerunning{Anti-correlation between AGN X-ray variability amplitude and luminosity}
82:    \maketitle
83: %
84: %________________________________________________________________
85: 
86: \section{Introduction}
87: 
88:    X-ray emission from active galactic nuclei (AGNs) exhibits variability on time scales from
89:    minutes to days. This indicates that X-rays are likely to be emitted from the inner most
90:     regions of AGNs and the variability may be related to the important properties of the central
91:     engine. Lawrence \& Papadakis (1993) utilized long term \emph{EXOSAT} observations to investigate
92:     the power density spectra of 12 AGNs. They found the power density spectra could be described
93:      as a power law $P \propto \nu ^{ - \alpha } $ with a mean index $\alpha  = 1.55$, and the amplitude was anti-correlated with the X-ray
94:       luminosity. Detailed studies of the power density spectra have been performed using
95:       \emph{RXTE}
96:       and \emph{XMM-Newton} data. The universal relation between the black hole mass and the "break time"
97:       in the power density spectra was found both in stellar mass and supermassive black holes
98:       (e. g. Uttely \& McHardy 2005). A tighter relation was discovered when the bolometric luminosity
99:       was involved, i.e. $T_B  \propto M^2 /L $ (McHardy et al. 2006). However, due to the
100: limited observation data,
101:        the accurate power density spectra are only available for a small number of AGNs. As an alternative,
102:        the normalized excess variance ($\sigma _{\rm rms}^2 $) can be easily calculated, and it was found that
103:        $\sigma _{\rm rms}^2 $ is anti-correlated with X-ray luminosity (e.g. Almaini et al. 2000).
104: 
105: 
106:        As a result of the progress in determining the black hole masses in AGNs, the relation between
107:        the variability and black hole mass has also been investigated. Lu \& Yu (2001) found the
108:        anti-correlation between the black hole mass and $\sigma _{\rm rms}^2 $, and suggested this correlation was an
109:        intrinsic one, rather than the apparent anti-correlation between the X-ray luminosity and $\sigma _{\rm rms}^2 $.
110:        Several authors also addressed this problem following Lu \& Yu's work. (e.g. Bian \& Zhao 2003;
111:        Papadakis 2004; O' Neill et al. 2005). These authors confirmed the anti-correlation between the
112:         black hole mass and $\sigma _{\rm rms}^2 $, and some models have been subsequently constructed to explain this correlation.
113: 
114:        In this paper we revisit this problem, by studying the sample of O' Neill et al. (2005), which
115:        includes 33 AGNs and uses nearly the same time scale for all these objects. The black hole
116:        masses of 17 objects in this sample were determined by the reverberation mapping or the stellar
117:         velocity dispersion (we denote these objects as sub-sample 1 in the following). The black hole
118:         masses of the remaining 16 objects were estimated from the relationship between broad line region
119:         radius and optical luminosity (we denote these objects as sub-sample 2 in the following).
120:         We find that the optical luminosity (which is used in determining the black hole
121:         mass for sub-sample 2) has obvious correlation with the X-ray luminosity in 2-10 keV
122:         band, as shown in Figure 1 (a). The values of the correlation coefficients between optical luminosity and X-ray luminosity for sub-sample 1 and sub-sample 2 are
123:          0.907 and 0.910, respectively. Even if the datum of NGC 4395 is excluded from sub-sample 2,
124:          the value of the correlation coefficient is still 0.819. It is well known that there is a strong correlation
125:          between the optical luminosity and the black hole mass (Kaspi et
126:         al. 2000). Thus if $\sigma _{\rm rms}^2 $ is intrinsically correlated with one of them, an artificial correlation
127:         with the other will appear. However, there will be an additional artificial correlation for sub-sample 2 due to
128:         the utilization of the optical luminosity in determining the black hole mass, whereas the black hole mass of sub-sample 1
129:         is independently obtained by the reverberation mapping or the stellar velocity dispersion. Therefore, the result from the full sample
130:         (especially for that from the sub-sample 2) may be misleading. Furthermore, due to the less reliable black hole mass of sub-sample 2, some
131:          unclear systematic biases may be introduced into the analysis. To find out which
132:           correlation is intrinsic, we perform the partial correlation analysis to sub-sample 1 and sub-sample 2 separately in \S2.1. The
133:          ordinary least squares regression results are shown in \S2.2 as another approach
134:          to this problem. K-S tests are performed in \S2.3 to test whether sub-sample 1 and
135:          sub-sample 2 are drawn from the same parent population. In \S3 we discuss our results and make conclusions.
136: 
137: 
138: 
139: 
140: 
141: %__________________________________________________________________
142: 
143: \section{Data analysis}
144: 
145: 
146:       \subsection{Partial correlation analysis}
147: The partial correlation analysis is an appropriate method to disentangle the
148: correlation between variables. The definition of the first order partial correlation
149: coefficient between variables $x$ and $y$ is (Kendall \& Stuart 1977), $r_{xy.z}  =
150: \frac{{r_{xy}  - r_{xz} r_{zy} }}{{\sqrt {(1 - r_{xz}^2 )(1 - r_{zy}^2 )} }}$, where
151: $z$ is the controlled variable and $r_{xy}$ is the correlation coefficient between
152: variables $x$ and $y$.
153: 
154: We adopt the data of the black hole mass, the X-ray luminosity and
155: $\sigma _{\rm rms}^2 $ from O' Neill et al. (2005) and present them
156: in Figure 1 (b) and (c) for clarity. The correlation analysis is
157: performed on the full sample and the sub-samples, respectively. The
158: results are shown in Table 1.
159: 
160: 
161: 
162:    \begin{figure}
163:    \centering
164:    \includegraphics[width=6cm]{1.eps}
165:       \caption{(a): The correlation between X-ray luminosity (2-10 keV) and optical
166:       luminosity ($\lambda  = 5100\;\mathop {\rm{A}}\limits^ \circ$ ). The point in the left-down corner is NGC 4395, which
167:       is a dwarf Seyfert galaxy. The value of the correlation coefficient is not significantly
168:        influenced by whether this point is included or not (see the text for details).
169:        (b): The data of the normalized excess variances and the X-ray luminosities (2-10 keV).
170:        (c): The data of the normalized excess variances and black holes masses.
171:        The values of black hole masses are in units of ${\rm{M}}_ \odot$. The data are obtained
172:        from O' Neill et al. (2005) and the references therein.}
173:          \label{FigVibStab}
174:    \end{figure}
175: 
176: 
177: 
178: For the full sample, both the black hole mass and the X-ray
179: luminosity show strong apparent anti-correlations with $\sigma _{\rm
180: rms}^2 $ (Figure 1 [b] and [c]). However, it appears that after the
181: black hole mass is controlled, the correlation between the
182: luminosity and $\sigma _{\rm rms}^2 $ is not significant. On the
183: contrary, the correlation between the black hole mass and $\sigma
184: _{\rm rms}^2 $ is still significant after the luminosity is
185: controlled. These results seem to support Lu \& Yu's suggestion that
186: the correlation between the black hole mass and $\sigma _{\rm rms}^2
187: $ is the intrinsic one. However, as discussed in \S1, since any
188: correlation between the black hole mass may be strengthened
189: artificially by the sub-sample 2, we should exclude them when
190: investigating the intrinsic correlation. For sub-sample 1, the
191: correlation between the black hole mass and $\sigma _{\rm rms}^2 $
192: disappears when the luminosity is controlled, whereas the
193: correlation between the luminosity and $\sigma _{\rm rms}^2 $ is
194: still significant. The results of sub-sample 2 are consistent with
195: those of the full sample. Both of them indicate $\sigma _{\rm rms}^2
196: $ is intrinsically correlated with the black hole mass, rather than
197: the luminosity. However, the analysis of the more reliable
198: sub-sample 1 shows the contrary results. Due to the limited size of
199: the present sample, we conclude that the results of sub-sample 2 is
200: doubtful and maybe the correlation between luminosity and $\sigma
201: _{\rm rms}^2 $ is the intrinsic one. More robust conclusion will be
202: deduced when a larger sample is available.
203: 
204: \begin{table*}
205: \caption{Results of partial correlation analysis.}             % title of Table
206: \label{table:1}      % is used to refer this table in the text
207: \centering                          % used for centering table
208: \begin{tabular}{c c c c c}        % centered columns (4 columns)
209: \hline                 % inserts double horizontal lines
210:  & $r_{L\;\sigma } $& $r_{M\;\sigma } $ & $r_{\;L\;\sigma .M}$ & $r_{M\;\sigma .L} $ \\    % table heading
211: \hline                       % inserts single horizontal line
212: Full sample & -0.636 ($>$99.99\%)   &  -0.697 ($>$99.99\%)   &  -0.277 (87.6\%)  & -0.452 (99.1\%)\\
213: Sub-sample 1  &   -0.856 ($>$99.99\%) &    -0.560 (98\%)   &  -0.781 ($>$99.99\%)   &  -0.003 (0.8\%)\\
214: Sub-sample 2  &   -0.605 (98.7\%) &  -0.784 ($>$99.99\%)  &   0.238 (60.8\%) &   -0.653 (99.2\%)\\
215: \hline                                   %inserts single line
216: \end{tabular}
217: \end{table*}
218: 
219: 
220: \subsection{Ordinary least squares regression}
221: To verify the results obtained in \S2.1, we perform the ordinary least square
222: regression to the full sample, sub-sample 1 and sub-sample 2, respectively. The
223: regression equation is, $\log (\sigma _{\rm rms}^2 ) = A\log M + B\log L + C.$
224: 
225: 
226: The results of the regression for the three samples are summarized
227: in Table 2. In Figure 2, we show the comparison between the values
228: of $\sigma _{\rm rms}^2 $ predicted by the results of the
229: regression and the observed ones.
230: 
231: \begin{table*}
232: 
233: \caption{Results of ordinary least squares regression. The significance of the linear
234: correlation is obtained by the $F$ statistic. The value of $\chi ^2 $ is calculated
235: from the regression result to estimate the goodness of the regression. The errors corresponding to 95\% confidence intervals are shown.}             % title of Table
236: 
237: 
238: \label{table:2}      % is used to refer this table in the text
239: \centering                          % used for centering table
240: \begin{tabular}{c c c c c c}        % centered columns (4 columns)
241: \hline                 % inserts double horizontal lines
242:     & A &  B &  C &  $F$ statistic & $\chi ^2 {\rm{(dof)}}$ \\    % table heading
243: \hline                       % inserts single horizontal line
244: Full sample & -0.38$\pm $0.28 &  -0.22$\pm $0.30 & 10.4$\pm $11.4 & 16.6 ($>$99.99\%)  & 708.6 (30)\\
245: Sub-sample 1  &   0.00$\pm $0.45 & -0.71$\pm $0.33 & 28.3$\pm $12.2 & 19.2 ($>$99.99\%) &  46.4 (14)\\
246: Sub-sample 2   &  -0.56$\pm $0.39 & 0.19$\pm $0.45 & -6.06$\pm $17.3 & 11.4 (99.86\%)  &  92.6 (13)\\
247: \hline                                   %inserts single line
248: 
249: \end{tabular}
250: 
251: \end{table*}
252: 
253: 
254: 
255: 
256: 
257: 
258: The results of $F$ statistic demonstrate the high significance of
259: the linear correlation. However, the values of $\chi ^2 $ are still
260: large, especially for the full sample. We should notice that the
261: dependence on the black hole mass and the luminosity seems to be
262:  different for the two sub-samples. For sub-sample 1, $\sigma _{\rm rms}^2 $ appears to depend
263:  weakly on the black hole mass,
264: whereas it depends more strongly on the X-ray luminosity. Due to the
265: small sample, the difference between values of $A$ and $B$ are not
266: very significant (they are coincidence within the 95\% confidence
267: interval). However, if the sub-sample 2 is included, the dependence
268: on the black hole mass is strengthened and the goodness of the
269: regression decreases dramatically. The value of the total $\chi ^2 $
270: of the sub-samples is 139 (27); therefore, the probability of the
271: improvement by chance is only about $10^{ - 9} $ (obtained by
272: $F$-test). Thus the above results indicate that the sub-sample 1 and
273: 2 are likely to obey different correlation relationships and it is
274: not appropriate to combine them into one sample.
275: 
276: 
277: 
278: 
279: 
280: \subsection{K-S tests}
281: 
282: 
283: To investigate whether the two sub-samples are drawn from the same
284: parent distribution, we first perform the 1D K-S test to the two
285: sub-samples. The cumulative distribution functions of the two
286: samples are calculated first. Then the maximum value of the absolute
287: difference
288:  between two cumulative distribution functions is used as the statistic to obtain the
289:  significance of the difference (see details of the K-S test in Press et al. [1992]).
290:  The significances of the differences are 89\%, 25\% and 96\% for the distributions of
291:  the black hole mass, the luminosity and $\sigma _{\rm rms}^2 $, respectively (the cumulative distribution
292:  functions are shown in Figure 3). Clearly except for the X-ray luminosity distribution,
293:  both the black hole mass and $\sigma _{\rm rms}^2 $ for the two sub-samples are not likely drawn from
294:  the same parent population. There is no obvious reason accounting for the differences, therefore this result is likely to be due to some unknown selection effects, which
295:  should be investigated further in the future.
296: 
297: Since it seems visually that the difference between the
298: distributions of sub-sample 1 and sub-sample 2
299:   in Figure 1 (b) is more significant than that in Figure 1(c), we perform the 2D K-S test to investigate this problem. The
300: results of the 2D  K-S test show that the significance of the
301: difference in Figure 1 (b) is 89.4\%, whereas significance of the
302: difference in Figure 1 (c) is 97.6\%. This unexpected result is due
303: to the existence of the point of NGC 4395. After this point is
304: removed,
305:   the 2D  K-S test results of Figure 1 (b) and 1 (c) are 96.5\% and 92.1\%, respectively. Although the
306:   significance of the difference in each figure is high, the visual difference
307: between two the figures is not significant.
308: 
309: 
310:  \begin{figure}
311:    \centering
312:    \includegraphics[width=7cm]{2.eps}
313:       \caption{The comparison between the values of $\sigma _{\rm rms}^2 $ predicted by the results of the regression
314:       and the observed ones. The results from the full sample are shown in (a). (b): The same
315:       as (a) but for the sub-sample 1. (c): The same as (a) but for the sub-sample 2.}
316:          \label{FigVibStab}
317:    \end{figure}
318: 
319: 
320: 
321:     \begin{figure}
322:    \centering
323:    \includegraphics[width=7cm]{3.eps}
324:       \caption{The cumulative distribution functions (CDF) of the black hole mass (a),
325:       the X-ray luminosity (b) and $\sigma _{\rm rms}^2 $ (c). The solid lines are the data of the sub-sample 1,
326:       and the dashed lines are the data of the sub-sample 2. The value of black hole mass is in units of ${\rm{M}}_ \odot$.}
327:          \label{FigVibStab}
328:    \end{figure}
329: 
330: \section{Discussions and conclusions}
331: 
332: 
333: In \S2, we have performed the partial correlation analysis and the
334: regression on the sample and found that the apparent intrinsic
335: correlation between $\sigma _{\rm rms}^2 $ and the black hole mass
336: is likely to be caused by including the sub-sample 2 into the
337: analysis. Because the black hole masses of AGNs in sub-sample 2 were
338: estimated from their optical luminosity which in turn is positively
339: correlated with their X-ray luminosity, an extra correlation between
340: the black hole mass and X-ray luminosity will be introduced by the
341: sub-sample 2. If the X-ray luminosity is the primary quantity, then
342: this will artificially strengthen any correlation with black hole
343: mass. We therefore should exclude them when investigating the
344: intrinsic correlation with $\sigma _{\rm rms}^2 $. According to the
345: results from the sub-sample 1, we conclude that the correlation
346: between $\sigma _{\rm rms}^2 $
347:  and the X-ray luminosity may be the
348: intrinsic one, whereas the apparent correlation between $\sigma _{\rm rms}^2 $
349:  and the
350: black hole mass is doubtful. Our K-S tests also suggest that
351: sub-samples 1 and 2 are not likely drawn from the same parent
352: population.
353: 
354: As discussed in Lu \& Yu (2001), several mechanisms may be responsible for the
355: correlation between $\sigma _{\rm rms}^2 $
356:  and the X-ray
357: luminosity, such as the hot-spot model, the obscurative variability
358: and so on. After the apparent correlation between $\sigma _{\rm
359: rms}^2 $ and the black hole mass was discovered, some models
360: accounting for this correlation were proposed (e.g. O' Neill et al.
361: 2005, Pessah 2007). However, it needs to be verified whether the
362: correlation is intrinsic. Although the black hole masses of about
363: three dozen AGNs have been determined by the reverberation mapping
364: method, the size of our sample is still limited due to the lack of
365: long enough and high quality observation data of these objects. More
366: conclusive results could be obtained when more and higher quality
367: data become available.
368: 
369: 
370: 
371: \begin{acknowledgements}
372: We thank the referee, Andy Lawrence, for valuable comments and suggestions. SNZ
373: acknowledges partial funding support by Directional Research Project of the Chinese
374: Academy of Sciences under project No. KJCX2-YW-T03 and by the National Natural Science
375: Foundation of China under project no. 10521001, 10733010 and 10725313.
376: \end{acknowledgements}
377: 
378: \begin{thebibliography}{}
379: \bibitem[1966]{baker}Almaini, O., et al. 2000, MNRAS, 315, 325
380: \bibitem[1966]{baker}Bian, W., \& Zhao, Y. 2003, MNRAS, 343, 164
381: \bibitem[1966]{baker}Kaspi, S., et al. 2000, ApJ, 533, 631
382: \bibitem[1966]{baker} Kendall, M., \& Stuart, A. 1977, The Advanced Theory of Statistics (London: Griffin)
383: \bibitem[1966]{baker}Lawrence, A., \& Papadakis, I. 1993, ApJ, 414, L85
384: \bibitem[1966]{baker}Lu, Y., \& Yu, Q. 2001, MNRAS, 324, 653
385: \bibitem[1966]{baker}McHardy, I., et al. 2006, Nature, 444, 730
386: \bibitem[1966]{baker}O' Neill, P., et al. 2005, MNRAS, 358, 1405
387: \bibitem[1966]{baker}Papadakis, I. 2004, MNRAS, 348, 207
388: \bibitem[1966]{baker}Pessah, M. 2007, ApJ, 655, 66
389: \bibitem[1966]{baker}Press, W., et al. 1992, Numerical Recipes in C, Cambridge University Press, Cambridge
390: \bibitem[1966]{baker}Uttely, P., \& McHardy, I. 2005, MNRAS, 363, 586
391: 
392: \end{thebibliography}
393: 
394: \end{document}
395: