1: %\documentclass[12pt,preprint]{aastex}
2: \documentclass{emulateapj}
3: \usepackage{apjfonts}
4: \usepackage{epsfig}
5: \usepackage{amsmath}
6: \usepackage{amssymb}
7: \usepackage{amsfonts}
8: \usepackage{lscape}
9: \usepackage{natbib} % This is required to use the apj.bst
10: \citestyle{aa} % and be able to use the AASTEX 5.x
11: % citations (\citep, \citet, etc.) - bph
12: \bibliographystyle{apj}
13:
14: \shorttitle{VARIABILITY IN THE LMXBS OF NGC~4697}
15: \shortauthors{SIVAKOFF ET AL.}
16:
17: \defcitealias{JCF+2004}{J2004}
18: \defcitealias{BIS+2006}{B2006}
19: \defcitealias{SJS+2007}{S2007}
20:
21: \defcitealias{ISB2000}{I2000}
22: \defcitealias{SIB2000}{Paper~I}
23: \defcitealias{SIB2001}{Paper~II}
24: \defcitealias{SSJ2005}{Paper~III}
25: \defcitealias{SJJ+2008a}{Paper~IV}
26: \defcitealias{SJJ+2008b}{Paper~V} % THIS PAPER
27:
28: \newcommand\cennodata{\multicolumn{1}{c}{\nodata}}
29:
30: \begin{document}
31:
32: \title{Measurements of Variability of Low Mass X-ray Binary Candidates
33: in the Early-Type Galaxy NGC~4697 from Multi-Epoch Chandra X-ray Observations}
34: \author{
35: Gregory R. Sivakoff\altaffilmark{1,2},
36: Andr\'{e}s Jord\'{a}n\altaffilmark{3,4},
37: Adrienne M. Juett\altaffilmark{5},
38: Craig L. Sarazin\altaffilmark{1},
39: Jimmy A. Irwin\altaffilmark{6}
40: }
41:
42: \altaffiltext{1}{
43: Department of Astronomy,
44: University of Virginia,
45: P. O. Box 400325,
46: Charlottesville, VA 22904-4325, USA;
47: sarazin@virginia.edu}
48: \altaffiltext{2}{
49: Current Address:
50: Department of Astronomy,
51: The Ohio State University,
52: 4055 McPherson Laboratory
53: 140 W. 18th Avenue, Columbus, OH 43210-1173, USA;
54: sivakoff@astronomy.ohio-state.edu}
55: \altaffiltext{3}{%
56: Clay Fellow,
57: Harvard-Smithsonian Center for Astrophysics,
58: 60 Garden Street,
59: MS-67, Cambridge, MA 02138, USA;
60: ajordan@cfa.harvard.edu}
61: \altaffiltext{4}{%
62: Departamento de Astronom\'{\i}a y Astrof\'{\i}sica,
63: Pontificia Universidad Cat\'olica de Chile,
64: Casilla 306, Santiago 22, Chile}
65: \altaffiltext{5}{%
66: NASA Postdoctoral Fellow,
67: Laboratory for X-ray Astrophysics,
68: NASA Goddard Space Flight Center,
69: Greenbelt, MD 20771, USA;
70: adrienne.m.juett@nasa.gov}
71: \altaffiltext{6}{
72: Department of Astronomy,
73: 909 Dennison Building,
74: University of Michigan,
75: Ann Arbor, MI 48109-1042, USA;
76: jairwin@umich.edu}
77:
78: \begin{abstract}
79: Multi-epoch {\it Chandra} X-ray observations of nearby massive early-type
80: galaxies open up the study of an important regime of low-mass X-ray binary
81: (LMXB) behavior --- long term variability. In a companion paper, we report on
82: the detection of 158 X-ray sources down to a detection/completeness limit of
83: $0.6/1.4\times 10^{37} {\rm \, ergs \, s}^{-1}$ using five {\it Chandra}
84: observations of NGC~4697, one of the nearest ($11.3 {\rm \, Mpc}$), optically
85: luminous ($M_B < -20$), elliptical (E6) galaxy. In this paper, we report on the
86: variability of LMXB candidates measured on timescales from seconds to years. At
87: timescales of seconds to hours, we detect five sources with significant
88: variability. Approximately $7\%$ of sources show variability between any two
89: observations, and $16\pm4\%$ of sources do not have a constant luminosity over
90: all five observations. Among variable sources, we identify eleven transient
91: candidates, with which we estimate that if all LMXBs in NGC~4697 are long-term
92: transients then they are on for $\sim 100 {\rm \, yr}$ and have a 7\% duty
93: cycle. These numbers are consistent with those found for brighter LMXBs in M87
94: and NGC~1399, which suggests that there does not appear to be a measurable
95: difference between the outburst durations of long-term transient neutron star
96: LMXBs and black hole LMXBs. We discuss in detail a transient supersoft source,
97: whose properties are not easily explained by standard explanations for supersoft
98: sources.
99: \end{abstract}
100: \keywords{
101: binaries: close ---
102: galaxies: elliptical and lenticular, cD ---
103: galaxies: star clusters ---
104: globular clusters: general ---
105: X-rays: binaries ---
106: X-rays: galaxies
107: }
108:
109: \section{Introduction}
110: \label{sec:n4697x_intro}
111:
112: With the angular resolution of the \textit{Chandra X-Ray Observatory} low-mass
113: X-ray binaries (LMXBs) in nearby early-type galaxies can be detected. For
114: early-type galaxies with a relatively low X-ray--to--optical luminosity ratio,
115: known as X-ray faint galaxies, the emission from LMXBs is the dominant source
116: \citep[e.g.,][hereafter Papers I and II]{SIB2000,SIB2001}.
117: In typical {\it Chandra} observations of such galaxies, tens to a few hundred
118: LMXBs are resolved at bright X-ray luminosities, $L_X \gtrsim 5$--$10\times
119: 10^{37} {\rm \, ergs \, s}^{-1}$. By stacking multi-epoch {\it Chandra}
120: observations of early-type galaxies, larger numbers of fainter LMXB candidates
121: can be studied, and brighter LMXB candidates can be studied in greater detail.
122: Such observations open up the study of an important regime of behavior --- long
123: term variability. Low mass X-ray binaries exhibit a wide range of variability,
124: including transient outbursts, type I bursts, and milder fluctuations. Our
125: understanding of LMXB variability in the Galaxy drives models for accretion and
126: LMXB evolution. However, there are several limitations in studying Galactic
127: LMXBs: source distances are known for only a small subset, it is difficult to
128: observe the whole Galaxy at once, and the size of the observed sample is very
129: limited. Observations of nearby early-type galaxies can overcome these
130: limitations of Galactic studies.
131:
132: Examples of extragalactic LMXB candidates with extreme behaviors, such as
133: luminous supersoft sources (SSs) and flaring sources, have also been examined
134: with {\it Chandra} in greater detail. In our Galaxy and M31, SSs have very soft
135: X-ray spectra $(\lesssim 75 {\rm \, eV})$ and are generally believed to be
136: accreting white dwarfs (WDs). Since the bolometric luminosities of extragalactic
137: SSs can exceed the Eddington luminosity for a Chandrasekhar mass WD, an
138: alternate hypothesis that these sources contain intermediate-mass ($\sim
139: 10^2$--$10^3 \, M_{\sun}$) accreting BHs has been put forth \citep{SGS+2002}.
140: Due to the degradation of the soft response of the Advanced CCD Imaging
141: Spectrometer (ACIS) on {\it Chandra}, the study of SSs has been limited. Some
142: LMXBs exhibit another extreme behavior, namely relatively short timescale
143: (seconds to minutes) flares. Recently, {\it XMM-Newton} detected type I X-ray
144: bursts in M31 \citep{PH2005} and we reported a more extreme example of flares in
145: NGC~4697 \citetext{\citealp{SSJ2005}, hereafter \citetalias{SSJ2005}}.
146:
147: In this paper, we report on the variability of LMXB candidates measured from
148: multi-epoch {\it Chandra} X-ray observations of NGC~4697, one of the nearest
149: ($11.3 {\rm \, Mpc}$), optically luminous ($M_B < -20$), elliptical (E6) galaxy.
150: In a companion paper (Sivakoff et al 2008b, hereafter Paper IV) we
151: report on the properties of LMXB candidates excluding variability, as well as
152: the connection to globular clusters (GCs) revealed by a {\it Hubble Space
153: Telescope} Advance Camera for Surveys ({\it HST}-ACS) observation near the
154: galaxy center.
155:
156: In \S~\ref{sec:n4697x_obs}, we recap key aspects of the observations and data
157: reduction of NGC~4697 discussed in Paper IV. Our observations of NGC~4697
158: provide five approximately equal chances to detect variability within a given
159: $\sim 40 {\rm \, ks}$ observation (\S~\ref{sec:n4697x_src_var_intra}), as well
160: as variability between different observations on timescales of $11 {\rm \, d}$
161: to $4.6 {\rm \, yr}$ (\S~\ref{sec:n4697x_src_var_inter}). We summarize our
162: conclusions in \S~\ref{sec:n4697x_conclusion}. Unless otherwise noted, all
163: errors refer to $1 \sigma $ confidence intervals, count rates are in the
164: $0.3$--$6 {\rm \, keV}$ band, and fluxes and luminosities are in the $0.3$--$10
165: {\rm \, keV}$ band.
166:
167: \section{Chandra Observations and Data Reduction}
168: \label{sec:n4697x_obs}
169:
170: Since detailed descriptions of the {\it Chandra} observation and
171: data reduction are reported in Paper IV, we only recap key aspects here.
172: This analysis covers five {\it Chandra} ACIS observations
173: (Observations
174: \dataset[ADS/Sa.CXO#obs/00784]{0784},
175: \dataset[ADS/Sa.CXO#obs/04727]{4727},
176: \dataset[ADS/Sa.CXO#obs/04728]{4728},
177: \dataset[ADS/Sa.CXO#obs/04729]{4729}, and
178: \dataset[ADS/Sa.CXO#obs/04730]{4730}) taken on
179: 2000 January 15, 2003 December 26, 2004 January 06, February 02, and August 18.
180: The final flare-filtered live exposure times for the five observations were
181: 37174, 39919, 35601, 32038, and $40044 {\rm \, s}$. There are differences in the
182: detector setup and instrument properties, such that the response of the Cycle-1
183: observation (0784) differs most from the Cycle-5 observations (4727, 4728, 4729,
184: 4730). Key differences include the gain files used
185: (acisD1999-09-16gainN0005.fits for Observation 0784 versus
186: acisD2000-01-29gain\_ctiN0001.fits with time-dependent gain and charge transfer
187: inefficiency corrections for the Cycle-5 observations), smaller background due
188: to Very-Faint mode cleaning in Cycle-5 observations, and increasing quantum
189: efficiency (QE) degradation with time.
190:
191: All {\it Chandra} observations were analyzed using {\sc ciao 3.1}%
192: \footnote{See \url{http://asc.harvard.edu/ciao/}.}
193: with {\sc caldb 2.28} and NASA's {\sc ftools 5.3}%
194: \footnote{See
195: \url{http://heasarc.gsfc.nasa.gov/docs/software/lheasoft/}%
196: \label{ftn:heasoft}.}. Source positions and extraction regions were
197: refined using ACIS Extract 3.34%
198: \footnote{See \url{http://www.astro.psu.edu/xray/docs/TARA/%
199: ae\_users\_guide.html}.}. All spectra
200: were fit using {\sc xspec}\footnotemark[\ref{ftn:heasoft}].
201:
202: The finding charts for the sources are displayed in Figures 1 and 2 of Paper IV.
203: In that Paper, the source properties from stacking the multi-epoch observations
204: are listed in Tables 1 and 7, and the instantaneous luminosities for each
205: observation are given in Table 6.
206:
207: Throughout Paper IV and this paper, we refer to the Analysis Sample, which
208: includes 126 sources that have photometric count rates determined at the $\ge 3
209: \sigma$ level when stacking all five observations. This corresponds to having
210: at least 18 net counts, a minimum detected count rate of $1.0 \times 10^{-4}
211: {\rm \, counts \, s}^{-1}$, and a minimum luminosity of $1.4 \times 10^{37} {\rm
212: \, ergs \, s}^{-1}$.
213:
214: The spectral properties of sources can be crudely characterized by hardness
215: ratios or X-ray colors (e.g., \citetalias{SIB2000,SIB2001}). We defined hardness
216: ratios of $H_{21} \equiv (M - S)/(M + S)$, $H_{31} \equiv (H - S)/(H + S)$, and
217: $H_{32} \equiv (H - M)/(H + M)$, where $S$, $M$, and $H$ are the total counts in
218: the soft (0.3--1 keV), medium (1--2 keV), and hard (2--6 keV) bands
219: \citep{SSC2004}.
220:
221:
222: \section{Intraobservation Variability}
223: \label{sec:n4697x_src_var_intra}
224:
225: After applying barycentric corrections to the times in event files, we
226: applied three techniques to search for intraobservation variability:
227: the Rayleigh statistic \citep[e.g.,][]{MBB+2003}, the K-S test
228: \citepalias[e.g.,][]{SIB2001}, and a newly developed flare detection technique
229: \citepalias{SSJ2005}.
230: These three techniques consider all photons in the source apertures,
231: and thus are likely to include some background photons.
232:
233: \tabletypesize{\footnotesize}
234: \input{4697_pervar_dlxtbl}
235:
236: We used the Rayleigh statistic to search for periodic signals with
237: frequencies between $5.0 \times 10^{-5} {\rm \, Hz}$ and $1.0 \times
238: 10^{-1} {\rm \, Hz}$ (periods of $10$--$20,000 {\rm \, s}$), testing
239: every $2.5 \times 10^{-5} {\rm \, Hz}$. Table~\ref{tab:n4697x_pervar}
240: summarizes the results for $> 2
241: \sigma$ level variable sources (i.e., the chance the observed source
242: variability is due to a statistical fluctuation is $< 4.6$\%),
243: listing the observation, probability that the detected periodicity is due to a
244: statistical fluctuation, $P_{\rm R}$, and detected period, $\tau_{\rm
245: R}$. Between zero and two sources appear periodic at the $> 2 \sigma$
246: level in any given observation. Since the expected number of false
247: detected sources at this level is $\sim 7$, we concentrate our
248: discussion on the one periodic source detected at the $> 3 \sigma$
249: level
250: (probability that the variability is due to a fluctuation is $< 0.27$\%).
251: The probability that the apparent periodicity among the 26
252: counts of Source 143 in Observation 0784 is due to a statistical
253: fluctuation is only $4.6\times10^{-8}$.
254: Unfortunately, this source is located near an edge of the S3 chip,
255: and the period of $\sim 1000 {\, \rm s}$ is consistent with a
256: periodicity induced by the yaw of the satellite aspect motion.
257: Thus, it is likely that this periodicity is instrumental, and we have no cases
258: of detected periodic variations
259: for sources in NGC~4697.
260:
261: \begin{figure}
262: \plotone{f1.eps}
263: \caption[X-ray Light Curve of Source 156 in NGC~4697]{
264: Binned lightcurve using 1350 ${\rm \, s}$ bins of Source 156
265: (the foreground star BD-05 3573) in Observation 4727.
266: Both the K-S test and our flare detection technique classify this source as
267: variable within this observation. This star flared shortly after the
268: observation began.
269: The average count rate during the observations is
270: indicated with the dotted line.
271: The count rate during the flare determined by our flare detection
272: algorithm is indicated by the dashed line and the extent of the flare is
273: indicated by the gray region.
274: \label{fig:n4697x_lng_flr}}
275: \end{figure}
276:
277: \tabletypesize{\footnotesize}
278: \input{4697_ksvar_dlxtbl}
279:
280: By comparing the cumulative fraction of events received from a source
281: to a constant rate, the K-S test can identify intraobservation variability.
282: This method has been applied to LMXB candidates in
283: previous papers \citepalias[e.g.,][]{SIB2001}. We summarize the
284: results in Table~\ref{tab:n4697x_ksvar} for sources which are variable
285: at the $> 2 \sigma$ level.
286: We list the
287: probability the detected variability is
288: due to a statistical fluctuation for each observation, $P_{\rm KS}$.
289: Between one and nine sources appear variable in any given observation.
290: Again, we concentrate on the one variable source detected at the $> 3 \sigma$
291: level.
292: Source 156 is the foreground star \object{BD-05 3573}
293: which undergoes a clear flare
294: ($P_{\rm KS} = 6.9\times 10^{-7}$) soon after Observation 4727 begins.
295: We display the binned lightcurve of this source in
296: Figure~\ref{fig:n4697x_lng_flr}.
297: We note that source 137 appears as a $2 \sigma$
298: variable source in both
299: Observations 0784 and 4727.
300: Following equation 3 of
301: \citetalias{SSJ2005}, the joint probability that this source is constant
302: is $4.6\times 10^{-3}$, which is still less than $3 \sigma$ significant.
303:
304: Finally, we applied a new method developed to search for flaring sources. This
305: technique is based on the arrival times of individual events compared to the
306: Poisson distribution at a constant rate, and was presented previously in
307: \citetalias{SSJ2005}. This flare detection uses the ONTIME values (37651, 40447,
308: 36072, 32462, and $40574 {\rm \, s}$, respectively for observations 0784 and
309: 4727--4730) as opposed to the live exposure times. \citetalias{SSJ2005} gave the
310: results for the most significantly flaring sources; note that Sources A, B, and
311: C in that Paper correspond to Sources 71, 75, and 57 here, respectively. We
312: summarize the results of these tests in Table~\ref{tab:n4697x_flarevar} for
313: sources which are variable at the $> 2 \sigma$ level in at least one
314: observation. For each observation of a flare-candidate source, we list the
315: observation number, the total number of photons in the observation, $N$, the
316: number of photons in the flare, $n$, the duration of the flare, $\delta t$, and
317: the Modified Julian Date (MJD) of the arrival time of the first photon in the
318: flare, $t_0$. The value of $P_{\rm constant}$ indicates the mathematical
319: probability of the given flare occurring due to a statistical fluctuation in a
320: constant rate source; however, we must account for the number of non-independent
321: flare searches undertaken. After applying corrections to $P_{\rm constant}$
322: based on the number of photons in the observation, $P_{\rm constant}^{\prime}$
323: is the probability a flare in a given observation is not real. The number of
324: sources with a statistically significant flare ranges from one to seven in the
325: different observations.
326:
327: \begin{figure}
328: \begin{center}
329: \epsfig{file=f2.eps, angle=90, width=0.85\linewidth,
330: clip=, bbllx=0, bblly=65, bburx=540, bbury=700}
331: \end{center}
332: \caption[Photon Impulse Diagram of Source 57 in NGC~4697]{
333: Impulse diagram indicating the time of arrival of photons in
334: Source 57 (Observation 0784). This LMXB candidate undergoes
335: short, bright flares in multiple observations
336: \citepalias{SSJ2005},
337: which were detected only with our new flare detection technique.
338: \label{fig:n4697x_sh_flr}}
339: \end{figure}
340:
341: Given the number of sources expected to show a flare from a statistical
342: fluctuation, we only consider the flaring sources in individual observations
343: detected at the $> 3 \sigma$ level, which are Sources 57, 155, and 156. These
344: sources have flares ranging from a few photons over tens of seconds to nearly a
345: hundred photons over thousands of seconds.
346: Source 57 was discussed earlier in \citetalias{SSJ2005} as Source C; note that
347: the luminosity conversion used in this Paper is about 10\% smaller than in
348: \citetalias{SSJ2005} due to updated calibrations. We display an impulse diagram
349: of the photon arrival times for Observation 0784 of Source 57 in
350: Figure~\ref{fig:n4697x_sh_flr}. The observed count rate of the flare is over 100
351: times the average count rate. Source 57 has no optical counterpart in the
352: HST-ACS image, suggesting it is most likely a flaring Field-LMXB.
353:
354: \tabletypesize{\footnotesize}
355: \input{4697_flarevar_dlxtbl}
356:
357: \begin{figure}
358: \plotone{f3.eps}
359: \caption[Accumulated Fractional X-ray Light Curve of Source 155 in NGC~4697]
360: {
361: Cumulative lightcurve of Source 155 (Observation 4727) overlaid by the
362: expected distribution for a constant source (dotted line). Our flare
363: detection technique classifies this source as variable within this
364: observation; however, the K-S test misses the variability of this source.
365: The flare interval, which lies between the dashed lines, has a clearly increased
366: rate (larger slope).
367: \label{fig:n4697x_med_flr}}
368: \end{figure}
369:
370: A flare of several hundred seconds with 17 counts is detected in Source 155, but
371: was missed by the K-S test. This source was just above the criteria used in
372: \citetalias{SSJ2005} ($P_{\rm constant, joint} < 0.32\%$) to identify a flaring
373: source. The observed count rate of the flare is about five times the average. As
374: shown in Figure~\ref{fig:n4697x_med_flr}, the flare is marked by a clearly
375: increased rate (larger slope in the cumulative light curve) that straddles the
376: cumulative light curve for a constant rate source. Since the K-S test looks for
377: the largest deviation from a constant rate, it is not surprising that the K-S
378: test fails to detect this type of flare. Follow-up of the optical counterpart to
379: Source 155 is necessary to determine if this source is an flaring LMXB in a GC
380: or a source unrelated to NGC~4697. As discussed earlier, Source 156 is the
381: foreground star \object{BD-05 3573} that undergoes a clear flare with 99 counts.
382: In Figure~\ref{fig:n4697x_lng_flr}, we overlay the flare count rate determined
383: by the flare detection algorithm for Source 156, which is about twice the
384: average rate.
385:
386: If we consider searching for flares over multiple observations, one can
387: calculate the joint probability the flares are not real, $P_{\rm constant,
388: joint}$. Thirteen sources in the Analysis Sample have a $P_{\rm constant,
389: joint}$ that indicate a $> 2 \sigma$ detection of flaring across multiple
390: observations; however, the detected flares in two of those sources (137 and 146)
391: have disparate enough durations that their separate flare detections probably
392: should not be combined. Only 4 sources (57, 71, 75, and 156) have flaring which
393: is significant at the $> 3 \sigma$ level across multiple observations; these are
394: the sources discussed in \citepalias{SSJ2005}, plus the foreground star
395: \object{BD-05 3573}. After correcting for the number of sources expected from
396: statistical fluctuations, we find that $4.2^{+3.3}_{-2.5}\%$ of the Analysis
397: Sample sources exhibit flaring which is detectable at the $> 2 \sigma$ level
398: with our new algorithm. This is consistent with the $3.1^{+2.4}_{-1.5}\%$ of
399: the Analysis Sample sources that exhibit flaring which is detectable at the
400: $> 3 \sigma$ level.
401:
402: \section{Interobservation Variability}
403: \label{sec:n4697x_src_var_inter}
404:
405: Our observations of NGC~4697 are among the first designed to measure
406: interobservation variability with strong coverage over a wide time baseline.
407: Some variability results have been presented before (e.g., Centaurus A:
408: \citealt{KKF+2001}, M87 \citealt{JCF+2004,Ir2006}, NGC 1399
409: \citealt{LAM2005,Ir2006}, and NGC 4636: \citealt{PRF+2008}); however, these
410: results did not have the same quality temporal coverage as our NGC~4697
411: observations. More recently, variability results were discussed for NGC~3379
412: \citep{BFK+2008}, with a similar coverage of time baselines as in our NGC~4697
413: observations. We searched for luminosity variability between any two
414: observations, as well as luminosity or spectral variability over the entire set
415: of observations.
416:
417: \subsection{Luminosity Variability Between Two Observations}
418: \label{sec:n4697x_src_var_lum2}
419:
420: \begin{figure}
421: \plotone{f4.eps}
422: \caption[Percentages of Variable Discrete X-ray Sources in NGC~4697]{
423: Percentage of Analysis Sample sources that appear variable (at the
424: $> 2 \sigma$ level) between two observations against the time period
425: between the observations.
426: Note that the error bars are not independent as the pairs contain the
427: same five observations.
428: The dotted horizontal line represents the
429: percentage of sources that would appear variable just due to
430: statistical fluctuations.
431: Although there is a rough tendency for more variability on longer
432: timescales, a constant variability percentage of 11.3\% is statistically
433: consistent with the data.
434: \label{fig:n4697x_2pt}}
435: \end{figure}
436:
437: We first searched for variability between pairs of observations by calculating
438: the value of $\chi^2$ assuming a constant flux. Sources were considered possibly
439: variable if $\chi^2 >4$, i.e., a $2 \sigma$ detection of variability. This
440: method differs slightly from \citet{PRF+2008}, which compares the individual
441: luminosities to each other. The probability that the variability is due to a
442: statistical fluctuation is listed in columns (2)--(11) of
443: Table~\ref{tab:n4697x_intervar} for the ten pairs of observations. In
444: Figure~\ref{fig:n4697x_2pt}, we display the percentage of variable sources as a
445: function of the timescale between two observations for all of the Analysis
446: Sample sources that are in the FOV for both observations in each pair. We also
447: display the percentage of variable sources expected due to statistical
448: fluctuations at the $2 \sigma$ level. There is a slight tendency for a higher
449: percentage of variable sources at longer timescales; however, this is not
450: statistically significant, based on a $\chi^2$ test. (There are two caveats with
451: this: the error bars in Figure~\ref{fig:n4697x_2pt} are not independent, and
452: there may be a slight excess variability on the four longest timescales due to
453: the improper correction of QE degradation for SSs. This latter effect comes
454: about because we have assumed an average spectrum when correcting for QE
455: degradation; however SSs have a much softer spectrum that is more strongly
456: affected by the QE degradation and the differences due to QE degradation will be
457: most evident in comparisons to the first observation. We estimate that this
458: latter effect is well within the displayed errors that arise from counting
459: statistics.) If we treat the error bars as independent, we find that
460: $11.3\pm1.0\%$ of the Analysis Sample sources are variable on average at the $2
461: \sigma$ level. Given the expected number of falsely identified variable sources,
462: $6.7\pm1.0\%$ of the Analysis Sample sources are intrinsically variable between
463: any pair of observations. If we only consider sources variable at the $3 \sigma$
464: level, the fraction of variable sources is $6.0\pm0.7\%$. Thus, over timescales
465: of $\sim 10$--$2000$ days, $\sim 6\%$ of sources are significantly variable. This
466: number is roughly consistent with that found for Centaurus A
467: ($9.8^{+4.5}_{-3.3}\%$) for a variability timescale of $164 {\rm \, d}$
468: \citep{KKF+2001}; although they derive a much higher percentage when they
469: include a generous definition of transient candidates ($14.2^{+2.6}_{-2.3}\%$).
470: Our derived variability percentage is lower than that found for a $1115 {\rm \,
471: d}$ timescale in NGC~4636 \citep[$23.8^{+3.3}_{-3.0}\%$;][]{PRF+2008}; however,
472: it is consistent with the variability for a day timescale
473: ($4.6^{+3.5}_{-2.2}$\%).
474:
475: \subsection{Long-Term Luminosity Variability}
476: \label{sec:n4697x_src_var_lum5}
477:
478: We can apply a similar $\chi^2$ technique to search for sources that show any
479: luminosity variability over the entire period of five observations. We refer to
480: these sources as Long-Term Luminosity Variables (LTLVs), and the probability
481: $P_{L,{\rm All}}$ that the variability arises from statistical fluctuations is
482: given in column 12 of Table~\ref{tab:n4697x_intervar}. The difference in
483: $\chi^2$ compared to a single merged luminosity and combined with the
484: appropriate degrees of freedom (one less than the number of observations a
485: source is in the FOV) is used to calculate the probability an observed LTLV is
486: due to a statistical fluctuation. Sources 148, 154, and 157 are excluded from
487: these calculations as they are only in the FOV of one observation. From the
488: Analysis Sample, 26/124 sources ($21.0^{+4.3}_{-3.8}\%$) are LTLVs at the $2
489: \sigma$ significance level, while $11.3^{+3.6}_{-2.9}\%$ of Analysis Sample
490: sources are LTLVs at the $> 3 \sigma$ significance level. Statistically
491: correcting for falsely identified LTLVs, we find $16.4^{+4.3}_{-3.8}\%$ of the
492: significantly detected sources show LTLV behavior at the $> 2 \sigma$
493: significance level. Three of these sources are not associated with NGC~4697:
494: Source 45 has an optical ID that is not a GC and thus is probably due to a
495: background AGN, Source 117 is a known background AGN, and Source 156 is a known
496: foreground star.
497: Two sources (25 and 155) are associated with optical counterparts that may or
498: may not be GCs in NGC~4697. Sources 39, 41, 51, 77, and 79 are GC-LMXBs that
499: exhibit LTLV behavior. (Sources 98 and 114 are associated with Kavelaars GC
500: candidates; new HST-ACS data will be used to test whether these sources are
501: really GC-LMXBs.) After correcting for the likely percentage of falsely
502: identified LTLVs, $10.1^{+8.7}_{-6.2}\%$ of GC-LMXBs and $<8.5\%$ of Field-LMXBs
503: exhibit LTLV behavior at the $> 2 \sigma$ significance level.
504:
505: The recent analysis of sources in NGC~3379 found that $42\pm5\%$ of their LMXBs
506: exhibited LTLV behavior \citep{BFK+2008} using a $\chi^2$ test similar to the
507: one used in this paper; however, they only required a source have a reduced
508: $\chi^2 > 1.2$. For sources in 2--5 observations, this corresponds to a
509: $\sim30\%$ false detection rate. After correcting for these false detections,
510: the rate for LTLVs in NGC 4697 ($\sim 11\%$) and NGC 3379 ($\sim 12\%$) are
511: comparable.
512:
513: Since some Galactic LMXBs are known to undergo transitions between
514: distinct luminosity states, we developed a method that groups observations
515: together into possible luminosity states.
516: For every LTLV, we first considered all possible combinations of observation
517: groupings that divide into two separate states.
518: That is, we assumed that the luminosity had two distinct values, and that
519: one or the other of the two values applied to each observation of that
520: source.
521: The two luminosity values were allowed to vary, and each observation was
522: assigned to one of the two luminosity states so as to minimize $\chi^2$.
523: Some sources still had a much larger $\chi^2$ than the number of dofs.
524: Given the new $\chi^2$, we could calculate the probability that
525: any deviations from our two-state hypothesis were due to statistical
526: fluctuations, given the new number of dofs and the number of combinations
527: of states. If this probability was low ($<4.6$\%), then we
528: tried to divide the observations into three luminosity states.
529: No source required more than three states to be fit acceptably.
530: We list the states in Table~\ref{tab:n4697x_state}.
531: For each
532: state, we list the observations included in a state (column 2), and
533: the combined luminosity and hardness ratios of the state (columns
534: 3--6). Unless otherwise noted, we placed no requirement on the
535: temporal pattern of the states; LTLVs whose states
536: are interspersed temporally may be switching between those states or
537: the intrinsic variability may be poorly characterized by a simple transition
538: between states.
539: There are several categories of luminosity state
540: transitions that indicate potentially interesting transitions and a
541: few sources that bear individual scrutiny.
542:
543: Sources 25, 78, and 110 are SSs that exhibit possible LTLV.
544: As SSs, they are more affected by QE degradation than the typical
545: source. Among these sources, the first two have a luminosity that
546: drops precipitously in the last four observations compared to the
547: first observation. Since the last four observations all have a similar
548: amount of QE degradation, we reran the LTLV detection
549: method only considering the last four observations; neither of the
550: sources remained LTLVs. Therefore, we do not
551: consider Sources 25 and 78 as LTLVs.
552: Since Source 110 does not turn on until the last observation, the QE
553: degradation does not complicate its identification as an LTLV.
554:
555: Sources 68 and 117 were the only sources broken into three luminosity states.
556: The individual luminosities of Source 68 are
557: qualitatively described by an increase over the first three
558: observations, followed by the source turning off, and perhaps
559: beginning to turn on again in the last observation. Its two-state mode
560: is consistent with this qualitative picture, containing a
561: higher-luminosity state of the first three observations and a state
562: where the source is not significantly detected in the last two
563: observations; however, the probability the two-state description is
564: correct is $6.1\times10^{-3}$. In its best
565: three-state mode,
566: Observations 0784 and 4730 are grouped together in a lower-luminosity
567: state that is still significantly detected; we note that the
568: three-state set where the first two observations share a
569: lower-luminosity state, the third observation is the higher-luminosity
570: state, and the last two observations are a state where the source is
571: not significantly detected has a $\chi^2$ only 0.46 higher than the
572: best three-state set. The individual luminosities of Source 117 vary
573: significantly with no clear temporal pattern. In its two-state set,
574: Observation 4729 is in the lower-luminosity state and the rest of the
575: observations are in the higher-luminosity state; however, the
576: probability the two-state description is correct is $4.2\times10^{-5}$.
577:
578: Some sources had a lower-luminosity state that was consistent with no
579: emission at the $3 \sigma$ level, and a higher-luminosity state, whose
580: observations were sequential and had emission detected at the $> 3
581: \sigma$ level. We labeled sources following this pattern of states
582: as transient candidates. For Sources 68 and 117, we considered their two-state
583: properties for discussions of transience. Eleven of the Analysis Sample sources
584: ($8.9^{+3.3}_{-2.6}\%$) are transient candidates (Sources 11, 37, 41, 45, 51,
585: 67, 68, 77, 94, 110, and 121). For the eleven transient candidate sources we
586: also calculated the minimum ratio of fluxes at the $3 \sigma$ level. Only the
587: flux ratio of Source 110 is above 10. We believe Source 110 is a truly transient
588: source and in \S~\ref{sub:n4697x_source110} we discuss it in detail. The other
589: transient candidates are either transient sources or LTLVs that have a lower
590: luminosity state that is below the detection limit. All eleven transient
591: candidates appeared to either turn on or turn off during the 4.6 years over
592: which we observed; no source turned on and then off, or off and then came back
593: on again. If we assume all the LMXBs in NGC~4697 are long-term transients, and
594: that our detected transient sources are just those which turned off or on during
595: our observation period, then the mean outburst duration for LMXBs in NGC~4697 is
596: $103^{+42}_{-28} {\rm \, yr}$. Under this hypothesis, the duty cycle of LMXBs
597: would also depend on their recurrence timescale. Since our observations find no
598: recurring transients, we have adopted the crustal heating recurrence timescale
599: for the Galactic NS-LMXB KS~1731$-$260 of $\sim 1500 {\rm \, yr}$
600: \citep{RBB+2002}. (The adoption of a crustal reheating timescale also implies
601: that we are assuming all the LMXBs in NGC~4697 are NS-LMXBs. While this is not
602: necessarily true, we note that the recurrence timescale of Galactic long-term
603: transient BH-LMXBs is unknown. As such, we present this calculation for
604: illustrative purposes only.) We note that this implies a duty cycle of $\sim
605: 7\%$, which is comparable to the duty cycle obtained comparing active and
606: quiescent LMXBs in Galactic GCs \citep[$\sim 12\%$;][]{HGL+2003}. We also note
607: that the number of transient candidates and possible transient candidates in
608: NGC~3379 yields similar duty cycles \citep{BFK+2008}, although their study
609: uses a different definition of transience.
610:
611:
612: Six of the LTLVs have a $> 2 \sigma$ difference in at least one of
613: their hardness ratios between their states (Sources 11, 45, 117, 121,
614: 134, and 156; Table~\ref{tab:n4697x_state}, columns 4--6). These
615: sources appear to exhibit luminosity/spectral state transitions. Since
616: the lower-luminosity state in Sources 45 and 121 is not even a
617: 1$\sigma$ significant detection, we exclude these sources from this
618: discussion. Although Source 11 is a Field-LMXB that is a transient
619: candidate, it may also be a luminosity/spectral state transition
620: source. The luminosity of the lower-luminosity state (Observation
621: 4730) is more than $2 \sigma$ significant and appears to be
622: 3.4$\sigma$ softer in $H_{21}$, than the higher-luminosity state.
623: Although this softer/fainter to harder/brighter transition is opposite
624: of the conventional relationship in BH state transitions, it is
625: possible that this discrepancy results from the softer bands used to
626: calculate the hardness with {\it Chandra} (Juett et al.\ 2008 in
627: preparation). Alternatively, this source may be transitioning from the
628: thermal state (formerly called high/soft) to the steep power law state
629: (formerly called very high). For example, in its 1996--1997 outburst,
630: the Galactic LMXB \object{GRO J1655-40} is softer and fainter in its
631: thermal state compared to its steep power law state \citep{RM2006}.
632: Source 117, a known AGN, appears to display the more typical
633: harder/fainter to softer/brighter transition \citep[e.g., the spectral
634: correlation of Seyfert galaxy NGC~4151 in][]{PPA+1986}; Observation
635: 4729 is $\sim 4$--5$\sigma$ harder in $H_{31}$ and $H_{32}$ than the
636: two higher luminosity states. Its spectrum seems to change most in the
637: 2--6 keV band. Source 134, a suspected AGN based on its hardness
638: ratios, appears to undergo a change in its 1--2 keV band. The
639: lower-luminosity state is $2.2 \sigma$ harder in $H_{21}$, but $2.0
640: \sigma$ softer in $H_{32}$ than its higher-luminosity state. Finally,
641: source 156, a known foreground star, undergoes a transition similar to
642: source 11. The lower-luminosity state of source 156 is $2.4 \sigma$
643: softer in $H_{21}$ than its higher-luminosity state.
644: This behavior is not surprising given that stellar flares generally heat their
645: corona \citep{Gu2004} and the higher-luminosity state occurs in the observation
646: where this source flares.
647:
648: \subsection{Long-Term Hardness Ratio Variability}
649: \label{sec:n4697x_src_var_hr}
650:
651: Our method for identifying LTLV can also be applied to
652: hardness ratios;
653: we define sources whose hardness ratios vary as Long Term Hardness Variables
654: (LTHVs).
655: Sources that have variations in the hardnesses ratios that are significant
656: at the $2 \sigma$ level have the probabilities that their variations are
657: due to statistical fluctuations listed in Table~\ref{tab:n4697x_intervar},
658: columns 13--15.
659: We identify fewer LTHVs than LTLVs
660: because the errors in hardness ratios are larger than in
661: luminosities.
662: Therefore, we only summarize the results for
663: the 83 sources in multiple observations and with luminosities detected
664: at the 5$\sigma$ level. We detect $6.0^{+3.9}_{-2.6}\%$,
665: $7.3^{+4.1}_{-2.9}\%$, and $6.1^{+3.9}_{-2.6}\%$ sources that exhibit
666: $H_{21}$, $H_{31}$, and $H_{32}$ variability at the $> 2 \sigma$
667: significance level, respectively. Corrections for falsely identified
668: variable sources reduce the percentages of hardness variable sources
669: to $<5.3\%$, $<6.9\%$, and $<5.5\%$.
670: That is, the numbers of sources selected purely based on hardness ratio
671: variations could result from statistical fluctuations.
672:
673: We also applied the same grouping of observations into states
674: as in \S~\ref{sec:n4697x_src_var_lum5}, using individual hardness ratios
675: as opposed to luminosities
676: (Table~\ref{tab:n4697x_state}, lower portions).
677: Here we discuss sources that were
678: identified as LTHVs in at least two hardness ratios,
679: or identified as LTHVs in at least one hardness
680: ratio and also as LTLVs. Sources 10, 66, and 102 are in
681: the first category, while Sources 11, 117, 134, and 156 are in the the
682: second. We discuss attempts to resolve the difference in their states,
683: and the implications of those attempts.
684:
685: For Source 10, the selections by $H_{31}$ and $H_{32}$ agree, pointing
686: to a probable decrease in the 2--6 keV count fraction for
687: Observation 4729. Although the best-fit states disagree for Source
688: 66, matching the $H_{32}$ selected state to the $H_{31}$ selected
689: state leads to a $\chi^2$ that is only 0.7 higher than the best
690: $H_{32}$ selected state. The fraction of 2--6 keV counts is
691: higher for Observation 4729. Similarly, matching the $H_{31}$
692: selected state to the $H_{32}$ selected state leads to a $\chi^2$ that
693: is only 0.2 higher than the best $H_{31}$ selected state for Source
694: 102; the 2--6 keV count fraction is higher for Observations 4727
695: and 4729.
696:
697: Sources 11, 117, 134, and 156 were all identified as
698: luminosity/spectral state transitions using the LTLV
699: selection. The hardness ratio two-state sets in Source 117 are
700: consistent with the two-state luminosity set; the 2--6 keV count
701: fraction is higher in the lower-luminosity state (Observation 4729). The
702: $H_{21}$ and luminosity states match for Source 156, where the
703: lower-luminosity state is softer in $H_{21}$ than the
704: higher-luminosity state.
705: Sources 11 and 134 are more difficult to resolve as clear
706: luminosity/spectral state transitions as the the grouping of observations
707: into states by luminosity and hardness ratios do not match. If we match the
708: $H_{21}$ selected state to the luminosity selected state for Source 11, the
709: $\chi^2$ is 4.1 higher;
710: however, both sets of states are consistent with the lower-luminosity state
711: being softer in $H_{21}$. The resolution for Source 134 is even more
712: problematic. This source appears best described by a non-monotonic evolution of
713: hardness with luminosity. As the source luminosity increases, it first becomes
714: softer and then becomes harder in all three hardness ratio bands; however, the
715: softest observation changes between 4728 or 4729 depending on the hardness
716: ratio.
717:
718: \subsection{Source 110: A Transient Supersoft Source}
719: \label{sub:n4697x_source110}
720:
721: We consider Source 110 to be a clear transient source. In the sum of the first
722: four observations, there were 5 counts in the source aperture and 19 counts in
723: the (three times) larger background aperture. After correcting for QE
724: degradation of a typical source in NGC~4697, the corresponding rate is $(-1.3
725: \pm 3.8) \times 10^{-5} {\rm \, cnt \, s}^{-1}$; the source is clearly
726: undetected. We also calculated the corrected rate including observations
727: 4727-4729, which have a similar level of QE degradation, to be $(-1.0 \pm 4.6)
728: \times 10^{-5} {\rm \, cnt \, s}^{-1}$. In the final observation, there were 87
729: counts in the source aperture and 4 counts in the background aperture,
730: corresponding to a QE degradation corrected rate of $(32.6 \pm 4.0) \times
731: 10^{-4} {\rm \, cnt \, s}^{-1}$. The ratio of the count rate in Observation 4730
732: to the other observations is $>27$ at the $3 \sigma$ level. Therefore, we only
733: consider data from Observation 4730 hereafter. Within the 0.3--$6.0 {\rm \,
734: keV}$ band, only two of the source aperture photons were above $1{\rm \, keV}$
735: (1.03 and $3.9 {\rm \, keV}$). Source 110 is clearly supersoft. Its $3\sigma$
736: limits on observed hardness ratios are $H_{21} < -0.89$ and $H_{31} < -0.88$.
737: Analysis of our flanking-field HST-ACS observations indicates Source 110 is not
738: associated with an optical source, although it is about $1\farcs9$ away from a
739: GC candidate. From simulated point sources at the location of Source 110 we
740: placed 90\% confidence limits on magnitudes for a single HST-ACS image of $g >
741: 26.3$ and $z > 25.3$ in AB mag. Since Source 110 is in an overlapping region of
742: two fields, these magnitudes correspond to the 99\% confidence limit on an
743: optical counterpart to Source 110.
744:
745: Since Source 110 is both transient and SS, we explored its
746: spectrum in greater detail. We examined both the raw spectrum and a
747: binned spectrum. Since Source 110 has only 87 gross counts in Observation
748: 4730, we created five bins with at least 16 counts in the 0.3--$1 {\rm \, keV}$
749: range and a bin with two counts from 1--$6 {\rm\, keV}$.
750: These bins were weighted by Gehrels errors.
751:
752: Luminous SSs in our Galaxy and M31 are generally believed to be
753: accreting WDs. Such sources are often fit by a blackbody (bb) model.
754: Compared to the best-fit for this model, the spectrum
755: appears to rise too quickly within the 0.3--$1 {\rm \, keV}$ band to
756: be consistent with Galactic absorption in the raw spectrum. In
757: the binned spectrum, the best-fit model with Galactic absorption was
758: strongly rejected, $\chi^2 = 16.78$ for 4 dof ($kT_{\rm bb} = 129\pm16
759: {\rm \, eV}$). This model is also hotter than typical fits to SS-WDs
760: \citep[15--$80 {\rm \, eV}$;][]{KH1997}. If we allow the absorption to
761: vary, we do get a high absorbing column and lower temperature,
762: $N_H = 1.2^{+1.0}_{-0.6} \times 10^{22} {\rm \, cm}^{2}$ and
763: $kT_{\rm bb} = 55^{+26}_{-19} {\rm \, eV}$
764: (90\% individual confidence limits), for
765: an acceptable fit with $\chi^2 = 2.04$ for 3 dof.
766: However, the implied X-ray and bolometric luminosity in the
767: 0.3--$10 {\rm \, keV}$,
768: $L_X = 3.0 \times 10^{41} {\rm \, ergs \, s}^{-1}$
769: and
770: $L_{\rm bol} = 1.6 \times 10^{42} {\rm \, ergs \, s}^{-1}$
771: are much too large for a WD at the distance of NGC~4697.
772: Even the lowest X-ray and bolometric luminosities allowed by the uncertainties
773: at 90\% confidence for jointly varying absorption and temperature, $L_X = 1.6
774: \times 10^{39} {\rm \, ergs \, s}^{-1}$ and $L_{\rm bol} = 3.0 \times 10^{39}
775: {\rm \, ergs \, s}^{-1}$ , are much too luminous for a SS-WD in NGC~4697. These
776: luminosities come at hotter temperatures, $90 {\rm \, eV}$, than seen in SS-WDs.
777: Although more detailed WD atmosphere models can lead to luminosities lower by an
778: order of magnitude than derived from the bb model
779: \citep[e.g.,][]{HTK1994,BKO1998,OZM+2007}, they also tend to increase
780: the temperature. Thus, we consider it unlikely that Source 110 is a SS-WD in NGC
781: 4697. On the other hand, our optical limit suggests that it is unlikely that
782: Source 110 is a SS-WD within the Milky Way. Of the likely companions for a
783: SS-WD, main sequence stars earlier than F5, red giants overflowing their Roche
784: lobe, and asymptotic giant branch stars with winds \citep{KH1997}, an F5
785: main-sequence star is the faintest with $M_{V} \sim 3.5$ \citep[pg.\
786: 107][]{BM1987}. We would detect such a source out to $\sim 300 {\rm \, kpc}$ in
787: our HST images. Thus, Source 110 is unlikely to be a SS-WD, either as a
788: foreground Galactic source or a source in NGC~4697.
789:
790: A very soft, luminous X-ray transient with an excess absorption column
791: has been observed in M101 \citep[M101~ULX-1;][]{KD2005}. The
792: combination of spectra, outburst luminosity, and transience was used
793: to suggest that source is an intermediate-mass BH (IMBH). The disk
794: blackbody model \citep[hereafter diskbb,][]{MIK+1984} is a multi-color
795: blackbody model that is often used to represent BH soft X-ray
796: transients (SXTs), which have $kT_{\rm diskbb} < 1.2 {\rm \, keV}$
797: \citep{TS1996}.
798: As with the simple blackbody model, the spectrum is poorly fit by a
799: model with Galactic absorption; $kT_{\rm diskbb} = 163\pm27 {\rm \,
800: eV}$ yields $\chi^2 = 20.05$ for 4 dof. If we allow the absorption to
801: vary, the absorption and temperature are strongly correlated and span
802: a fairly large range, but the fit is much improved with a
803: $\chi^2 = 2.02$ for 3 dof. We find $N_H = 1.2^{+1.0}_{-0.6} \times 10^{22}
804: {\rm \, cm}^{2}$ and $kT_{\rm diskbb} = 59^{+28}_{-22} {\rm \, eV}$
805: (90\% individual confidence limits). The 0.3--$10 {\rm \, keV}$ X-ray
806: luminosity of acceptable spectral models (within the 90\%
807: two-dimensional confidence interval for varying absorption and
808: temperature) ranges from $3.5 \times 10^{39}$ -- $6.8 \times 10^{46}
809: {\rm \, ergs \, s}^{-1}$.
810: The bolometric luminosity range is even
811: larger, $1.4 \times 10^{40} < L_{\rm bol} < 4.8 \times 10^{48}
812: {\rm \, ergs \, s}^{-1}$.
813: We use the normalization of the diskbb
814: model to estimate a BH mass following equation (3) in \citet{FRM2004}.
815: Following \citet{FRM2004}, we also assume a correction factor of $f_0 = 1.35$,
816: an inclination angle of $i = 60\degr$,
817: and an innermost disk radius of $6 \, G \, M_{\rm BH}/c^2$, where
818: $M_{\rm BH}$ is the BH mass.
819: Dividing our derived bolometric luminosity by the Eddington luminosity,
820: $L_{\rm Edd} = 1.26 \times 10^{38} ( M_{\rm BH} / M_{\odot} ) {\rm \, ergs \,
821: s}^{-1}$,
822: gives the Eddington efficiency of the accretion disk, which we use
823: later in evaluating the plausibility of the model.
824: The value of the disk blackbody temperature also provides a second estimate
825: of the black hole mass, which is roughly
826: $M_{\rm BH} / M_{\odot} \sim ( kT_{\rm diskbb} / 1 \, {\rm keV} )^{-4}$
827: \citep{MKM+2000}.
828: Considering
829: models within the 90\% two-dimensional confidence interval for varying
830: absorption and temperature, these two estimates of BH mass agree to
831: within a factor of two for absorption columns between 0.5 and $1\times
832: 10^{22} {\rm \, cm}^{2}$.
833: Within this region, our source is consistent
834: with BH masses of
835: $\sim ( 7.2 \times 10^{3}$--$1.1\times 10^{6} ) \, M_\odot$
836: accreting at $\sim 1.6$--4.7\% of the Eddington limit.
837: Thus, this source is consistent with a rather massive IMBH or a small
838: supermassive black hole about a factor of 2--3 smaller in mass than the
839: one at the center of the Milky Way \citep{GMB+2000}.
840: Understanding the origin of such a large BH outside of the center of
841: a galaxy or of a GC is a formidable challenge to our current theories of BH
842: formation.
843:
844: Some AGNs have ultrasoft spectra. For example, \citet{PMC+1992}
845: identified 53 ultrasoft AGN candidates with {\it Einstein
846: Observatory}. These sources had 0.16--$0.56 {\rm \, keV}$ count rates
847: nearly three times that of their 0.56--$1.08 {\rm \, keV}$ count
848: rates. This corresponds to steep power-law photon indices, $\Gamma
849: \gtrsim 3$, with Galactic absorption. In a study of variability in
850: soft X-ray selected AGNs, \citet{GTB2001} found that AGNs with steeper X-ray
851: spectra had larger short-term variability than those with flatter spectra. Two
852: of the transient sources, \object{IC 3599} and \object{HE 0036-5133} (also
853: called WPVS~007) were also previously reported as ultrasoft. These sources may
854: be analogs to Source 110. A variety of complicated models have been used to
855: explain the spectra of AGNs with excess soft emission compared to the power-law
856: models derived from harder bands \citep[e.g.,][]{GD2004}. Given the relative
857: paucity of our data on Source 110, we chose to only consider simple power-law
858: and blackbody models. The power-law model with Galactic absorption, $\Gamma =
859: 4.23^{+0.86}_{-0.58}$, is an exceptionally poor fit, $\chi^2 = 30.20$ for 4 dof.
860: Even with a free absorption, power-law models with reasonable photon indices,
861: $\Gamma < 10$, did not fit our data well, $\chi^2 > 12.25$ for 3 dof. From
862: above, we know that the disk blackbody model is a good statistical
863: representation of the spectra. However, our derivation of the BH masses in the
864: above paragraph was predicated on the source being at the distance of NGC~4697.
865: As we allow the source to lie farther behind NGC~4697, the agreement between the
866: mass derived from the normalization and the mass derived from the temperature
867: begins to break down. Thus, it is unlikely that the disk blackbody model fully
868: represents the physical mechanism behind the emission. Regardless of the model
869: we choose to represent the emission, we measure an absorbed X-ray flux
870: (0.5--$8.0 {\rm \, keV}$) of $F_X
871: \sim 7
872: \times 10^{-15} {\rm \, ergs \, s}^{-1} {\rm \, cm}^{-2}$.
873: At that flux level, we only expect $\sim 3.6$ sources
874: unrelated to NGC~4697 for galactocentric radii interior to that of
875: Source 110.
876: Thus, the probability Source 110 is a
877: background AGN is $\sim 3.3\%$ from the X-ray data alone.
878: If we assume that the transient behavior observed was not an extreme flare that
879: began between 2004 February 02 and August 18, but ended before the HST-ACS
880: flanking field observations began in 2005 December 19, we can place an additional
881: constraint using the logarithm of its X-ray--to--optical flux ratio, $\log (F_X/F_{\rm opt})$,
882: where $F_{\rm opt}$ is the optical flux in the $z$ band.
883: Based on our non-detection of a counterpart to Source 110, its
884: $\log (F_X/F_{\rm opt})$ is greater than 0.9 at the 99\% confidence level.
885: Such a large ratio is expected only from the most
886: obscured AGNs. There are two problems with suggesting Source 110 is
887: an obscured AGN.
888: First, the most obscured AGNs can have columns one
889: to two orders of magnitude larger than the estimated column for Source
890: 110.
891: In fact, its column is actually near the division point typically
892: used for obscured and unobscured AGN
893: \citep[$10^{22} {\rm \, cm}^{2}$;][]{TUC+2004}.
894: Second, very few sources are predicted to have such high X-ray--to--optical flux
895: ratios. In \citet{TUC+2004}, a model based on the AGN unification paradigm
896: predicts only $\sim13\%$ of AGNs will have $\log (F_X/F_{\rm opt}) > 0.8$. In
897: summary, although some of its X-ray properties may be consistent with transient
898: ultrasoft AGN behavior, the combination of the expected number of background
899: AGNs with the high X-ray--to--optical flux ratios makes it unlikely (probability
900: $\sim 0.4\%$) that Source 110 is a transient ultrasoft AGN.
901:
902: \section{Conclusions}
903: \label{sec:n4697x_conclusion}
904: Our multi-epoch {\it Chandra} observations were planned to shed light on LMXB
905: variability in NGC~4697, the nearest, optically luminous, elliptical galaxy. We
906: measure variability on timescales of seconds to years. Although we do not detect
907: any periodic LMXBs at significant levels, we detect five sources with
908: significant variability on timescales of seconds to hours
909: (Source 57, 71, 75, 155, and 156).
910:
911: When we compare the luminosities of sources between every pair of observations,
912: we determine that $6.7\pm1.0\%$ of the Analysis Sample sources are variable over
913: timescales of $\sim 10$--$2000$ days. There is a slight increase in the fraction
914: of variable sources with the variability timescale, but it is not found to be
915: statistically significant. Our number is roughly consistent with that found for
916: Centaurus A for a variability timescale of $164 {\rm \, d}$
917: \citep{KKF+2001} and for NGC~4636 for a day timescale, but is lower
918: than that found for a $1115 {\rm \, d}$ timescale in NGC~4636 \citep{PRF+2008}.
919: New multi-epoch observations of NGC~1023, NGC~3379, NGC~4278, NGC~4365, and
920: Centaurus A will help fill in more variability timescales. As this phase space
921: is populated, we will be able to test if more sources are variable on longer
922: timescales than shorter timescales. Increased knowledge of the temporal
923: evolution of variability will impact models that explore the nature of the
924: observed LMXBs \citep[e.g.,][]{PB2002}.
925:
926:
927: We are more sensitive to variability when we consider all five observations
928: simultaneously. We find that $\approx16\pm4\%$ of Analysis Sample sources do not
929: have a constant luminosity over all five observations. If we apply a similar
930: requirement for a source to be considered variable to the \citet{BFK+2008}
931: analysis of LMXBs in NGC~3379 as in this study, we find that the number of
932: variable sources is $12\pm5\%$, consistent with what we find for NGC~4697. The
933: fraction of GC-LMXBs that are variable is statistically
934: indistinguishable from the fraction of variable Field-LMXBs in our sample due
935: to the small number of variable sources. To better test for any difference
936: between the two populations, multi-epoch variability from multiple galaxies will
937: need to be combined.
938:
939: For the detected sources we group observations into sets of luminosity and
940: hardness ratio states. Among these variable sources, we identify eleven
941: transient candidates. If we assume all the LMXBs in NGC~4697 are long-term
942: transients, we estimate that they are on for $\sim 100 {\rm \, yr}$ and have a
943: 7\% duty cycle. The latter is comparable to the duty cycle obtained comparing
944: active and quiescent LMXBs in Galactic GCs \citep{HGL+2003}.
945:
946: The mean outburst duration for LMXBs in NGC~4697 of $\sim 100 {\rm \, yr}$ is
947: consistent with a lower limit of $\sim 50 {\rm \, yr}$ found for 33 luminous
948: LMXBs in NGC~1399 and M87 that had X-ray luminosities exceeding $8 \times
949: 10^{38}$ ergs s$^{-1}$ \citep{Ir2006}. These 33 sources were assumed to be
950: BH-LMXBs because of their sustained super-Eddington luminosities over several
951: year timescales, and most likely represent extragalactic analogs of the Milky
952: Way X-ray binary GRS1915+105. Since a majority of the sources in NGC~4697 are
953: probably NS-LMXBs, there does not appear to be a measurable difference between
954: the outburst durations of long-term transient NS-LMXBs and BH-LMXBs.
955:
956: We also find four sources with a combined luminosity/spectral transition, only
957: one of which is likely associated with NGC 4697. One limit to our ability to
958: detect such sources is the duration of our single observations. Although
959: observations of NGC~3379 and NGC~4278 have longer single observation durations,
960: this issue is likely to be best addressed by a recently completed set of six,
961: 100 ks observations of Centaurus A, which is at least 2.5 times closer than any
962: other massive early-type galaxy.
963:
964: Source 110 is clearly a supersoft transient source with no detected
965: counterpart in {\it HST} observations. Although it only has $\approx
966: 85$ X-ray counts, our spectral analysis rules out the possibility that
967: this source is a luminous supersoft WD. The high X-ray--to--optical flux
968: ratio and relative small number of expected background AGNs argue
969: against the source being a transient ultrasoft AGN. Our spectral fit
970: could be consistent with an accreting BH with a mass in the range of
971: $\sim$$( 7 \times 10^{3}$--$10^{6} ) \, M_\odot$;
972: however, the origin of such a large BH
973: outside of the galaxy center or a GC is a formidable challenge
974: to our current theories of BH formation. Further observations of
975: NGC~4697 could increase the information we can extract from the
976: spectra of Source 110; however, more work must also be done to
977: increase our understanding of the X-ray spectra of non-stellar-mass
978: black holes.
979:
980: \acknowledgements
981:
982: We thank Kris Beckwith, Marina Orio, and and Roseanne
983: di Stefano for very helpful discussions. Support for this work was
984: provided by NASA through {\it Chandra} Award Numbers GO4-5093X,
985: AR4-5008X, and GO5-6086X, and through HST Award Numbers
986: HST-GO-10003.01-A, HST-GO-10582.02-A, HST-GO-10597.03-A, and
987: HST-GO-10835.03-A.
988: G.~R.~S.\ acknowledges the receipt of an ARCS fellowship and support
989: provided by the
990: F. H. Levinson Fund.
991:
992: \bibliography{ms}
993:
994:
995: %%% TABLES
996:
997: \clearpage
998: \tabletypesize{\footnotesize}
999: \begin{landscape}
1000: \LongTables
1001: \input{4697_2var_dlxtbl}
1002: \clearpage
1003: \end{landscape}
1004:
1005: \clearpage
1006: \tabletypesize{\footnotesize}
1007: \LongTables
1008: \input{4697_state_dlxtbl}
1009:
1010:
1011: %%% FIGURES
1012:
1013:
1014:
1015:
1016:
1017:
1018:
1019: \end{document}
1020: