astro-ph0401212/analysis.tex
1: 
2: The {\it Standard 2} PCA data for the observations were extracted using the 
3: FTOOLS 5.0.2 package. Since the count-rates are small during the beginning and 
4: end of the outburst, we have used data from the observations when the maximum 
5: number PCUs were turned on. In order to avoid looking at the earth, we 
6: considered data only when the elevation angle was $10^{\circ}$ or more. Any 
7: data with (i) offset pointing more than $0^{\circ}.02$; or (ii) within 30 mins 
8: of SAA passage; or (iii) trapped electron contamination more than the 
9: prescribed value (i.e $ELECTRON2>0.1$) were rejected. Appropriate faint/bright 
10: background models were used when the source was fainter/brighter than 40 
11: counts/s/PCU.
12: 
13: \subsection{Fluxes and Colors} 
14: Following \citet{gd2002}, we extracted {\it standard 2} lightcurves for four 
15: energy bands, viz. 3-4, 4-6.4, 6.4-9.7 and 9.7-16 keV (corresponding to binned 
16: Std2 channels 5-6, 7-11, 13-19 and 20-33 respectively). Two bursts were 
17: detected during the rising phase and were excluded from analysis \citep{y2003}. 
18: The hardness-ratio for each observation was calculated from the ratio of counts
19:  in the hardest band (9.7-16 keV) to the softest band (3-4 keV; see 
20: Fig.~\ref{hardnesstime}). For typical spectral fits the softest band (3-4 keV) 
21: contains mostly thermal photons whereas the hardest band (9.7-1.6 keV) is 
22: dominated by powerlaw photons. Hence the (9.7-16 keV)/(3-4 keV) hardness ratio 
23: is an indicator of the ratio of powerlaw photons to the disk photons. As 
24: expected, the spectra remain hard during the low state, but make a very rapid 
25: transition to the soft state between MJD $\sim$ 51820 and 51822. The source 
26: remains soft until MJD$\sim$ 51864 after which we see a smooth transition to 
27: the hard state. However the transition back to the LH state from the HS state 
28: during the decaying phase occurs more slowly. However it is to be noted that 
29: since we only consider photons within certain energy-bands, the numerical 
30: values of the hardness-ratios do not say much about the physical processes nor 
31: is the hardness ratio equal to disk/powerlaw luminosity.
32: 
33: As seen in Fig.~\ref{xcmd}, the source traces a hysteresis loop over the 
34: entire outburst cycle in the hardness-luminosity plane [(9.7-16)keV/(3-4)keV
35: hardness-ratio vs. (6.4-9.7)keV count-rate]. Aql X-1 and few other SXT's are 
36: now known to show this feature \citep{mc2003,mrc2002,rmkf2000,c1998}. 
37: The transition from hard-to-soft state takes place at a (6.4-9.7 keV) 
38: luminosity $\sim10$ times higher than the luminosity at which it returns back 
39: to the hard state. The observed trend in evolution is very similar to what was 
40: observed by \citet{mc2003} using ASM data during the 1998 outburst.
41: 
42: A particularly interesting feature of the spectral evolution emerges when one 
43: considers the evolution in the color-color diagram. Soft color was defined as 
44: a ratio of 4-6.4 to 3-4 keV count rates, and hard color as a 9.7-16 over 
45: 6.4-9.7 keV ratio. We plotted each observation (dwell) as one point on the 
46: color-color diagram. Since the evolution of the source on the color-color 
47: plane during an observation is small compared to the entire space it spans 
48: during the entire outburst cycle, this is a reasonable procedure to track the 
49: time evolution during the entire outburst (Fig.~\ref{colcol}, top panel). We 
50: find that the overall soft state (hard color $\le$0.4) has two sub-states. 
51: Once the source makes a transition from hard state to the soft state, it 
52: occupies the lower right of the CC diagram (shown by open circles in the 
53: bottom panel of Fig.~\ref{colcol}) and stays there as long as the luminosity 
54: is close to that of peak outburst (plateau in Fig.~\ref{exposure}). Once the 
55: decay starts appreciably, the source moves slightly to the left in CC plane 
56: (filled circles in Fig.~\ref{colcol}), corresponding to the TSS, and stays 
57: there until it makes another sharp transition to the initial hard state. The 
58: HS state and the TS state occupy disjoint (though neighboring) regimes of the 
59: color-color diagram.
60: 
61: \subsection{Spectral deconvolution}
62: Spectral analysis was carried out with the XSPEC software (v.11.0.1). We used 
63: the PCA {\it Standard 2} data in the energy range of 3-20 keV where PCA 
64: calibrations are the best. The spectral model we adopted consist of two 
65: additive components: a multi-colored disk (MCD) or disk blackbody (DBB) 
66: \citep{m1984}, plus a powerlaw. Galactic X-ray absorption was accounted for 
67: using the multiplicative photoelectric absorption model {\it wabs} 
68: \citep{mm1983}, with a constant hydrogen column density of $3.4\times10^{21}$ 
69: atoms/cm$^2$, the galactic value along the line of sight towards Aql X-1. It 
70: was found that the inclusion of another multiplicative component, a smeared 
71: edge ({\it smedge} in XSPEC terminology) greatly improved the spectral fits. 
72: The threshold energy of the edge was constrained to lie between 7-8 keV. 
73: Assuming a systematic error of 1\% for all PCA channels \citep{h2001,mc2003}, 
74: the above model gave good fits with $\chi^2/dof$ close to 1. Including an Iron 
75: line near 6.5 keV improved the fits only marginally and hence we have omitted 
76: the feature. 
77: 
78: The results of the fits are tabulated in Table~\ref{tab:obslog} and plotted in 
79: Fig.~\ref{spectral}. As expected, both the disk and powerlaw normalizations 
80: rise/fall during the rise/decay of the outburst. In the hard state, the powerlaw 
81: photon index $\Gamma$ is flatter ($<2$) compared to its value in the soft state 
82: when it is always steeper than 2. Remarkably, all but the disk normalization 
83: remains very constant throughout the soft state. The only change when the 
84: source moves from the HS to the TSS is that the disk normalization falls as 
85: shown in Fig.~\ref{fits}.
86: 
87: \subsection{Temporal Analysis}
88: We performed a Leahy-Normalized fast Fourier transform \citep{l1983} for 
89: every 64s segment using {\it Event} mode data of each observation for the full 
90: PCA ($\sim$2-60 keV) energy band. The {\it powspec} tool in XRONOS package 
91: was used to calculate the power density spectrum (PDS). Tha data was binned 
92: into $2^{-6}$s time bins, corresponding to a Nyquist frequency of 32 Hz. The 
93: error bars in the average power spectrum were calculated by evaluating the 
94: standard deviation of the average power for each frequency. These indivudual 
95: PDSs were then added to obtain average power-density spectrum for each dwell. 
96: The PDS were rebinned logarithmically to increase S/N. Finally white Poisson 
97: noise was subtracted and the PDS renormalized by the mean 
98: count rate, as discussed by \citet{vdk1995}. Thus the final PDS has 
99: the units of (rms/mean)$^2$ Hz$^{-1}$. The total time variability in a certain 
100: frequency range was estimated by fitting the corresponding PDS with a 
101: power-law or a combination of power-law and Lorentzian, and integrating over 
102: the frequency range. Most of the power comes from frequencies less than 30 Hz, 
103: beyond which the spectra is essentially dominated by white Poisson noise.
104: 
105: \subsubsection{Low Frequency Power}
106: During the LH rising phase, the power spectra in the range of 0.1-30 Hz exhibit
107:  a shallow powerlaw continuum up to a break frequency $f_{br}$. Beyond the 
108: break frequency the noise falls as $f^{-1}$. To describe the continuum 
109: behavior we fitted a broken power law model of the form:
110: 
111: \[{\rm PDS}(f) = \left \{ \begin{array}{r}
112:         C\left( \frac{f}{f_{br}} \right)^{\alpha_{1}}, f < f_{br} \\
113:         C\left( \frac{f}{f_{br}} \right)^{\alpha_{2}}, f \ge f_{br} \\
114: \end{array}
115: \right .
116: \]
117: 
118: where $\alpha_{1}$ and $\alpha_{2}$ are power law indices, and $f_{br}$ is
119: defined above.  The constant $C$ is chosen so that the integral of the
120: continuum model gives the correct power, expressed as fractional rms
121: variability.
122: 
123: Broad QPO-like humps were also observed in some of the spectra 
124: (see Fig.~\ref{pdsfig}). With the progress of the outburst the break frequency 
125: as well as the QPO frequency appears to increase \citep{y2003} whereas the 
126: total continuum power level decreases. This trend was also observed during the 
127: 1999 outburst of the same source \citep{j2001} and for Cyg X-1 
128: \citep{bh1990,b2002}. Table~\ref{tab:pds} lists the break frequency, powerlaw 
129: indices and the relevant QPO parameters for these dwells, averaged over the 
130: exposure times in each dwells.
131: 
132: The soft state power spectrum is relatively featureless, the rms amplitudes 
133: are small and can be modelled by a flat powerlaw at frequencies lower than 
134: $\sim$ 10 Hz, and dominated by Poisson noise residuals at higher frequencies. 
135: Occasionally the signature of a break or a hump is observed, but the errors 
136: are too large to identify any feature conclusively, or to trace the evolution 
137: of any feature.
138: 
139: The variation of total power within 0.1-30 Hz frequency range (expressed in 
140: percentage rms variability), over the entire outburst is shown in 
141: Fig.~\ref{rmstime}. The strong correlation of the rms variability with 
142: spectral hardness (Fig.~\ref{hardnesstime}) strongly suggests that most of the 
143: time variability indeed comes from the hard, nonthermal photons presumably 
144: Compton upscattered in the corona above the accretion disk. The rms energy 
145: spectra for a particular dwell (Fig.~\ref{rmsE}) also supports the argument 
146: that most of the time variability comes from hard photons.
147: 
148: \subsubsection{kHz QPOs}
149: We also performed a Leahy-Normalized fast Fourier transform for every 128s 
150: segment of the {\it Event} data sets, extending up to 4 kHz in frequency. The 
151: frames were weighted and co-added to obtain average power-density spectrum for 
152: each observation. We found nine kilohertz QPOs in various segments of our 41st 
153: and 47th-51st observations. All the QPOs were observed during the decaying 
154: phase of the outburst, when the source was in the TSS. Near the kHz QPOs, the 
155: continuum is essentially flat with variations solely due to Poisson 
156: fluctuation in the data. The QPO profile is modelled by a simple Lorentzian of 
157: the form
158: \[PDS(f)=\frac{(rms_{qpo})^2}{\pi}\frac{\Delta f/2}{(f-f_{qpo})^2+(\Delta f/2)^2}\]
159: where $rms_{qpo}$ is the integrated fractional rms amplitude of the QPO, 
160: $f_{qpo}$ is the center of the Lorentzian and $\Delta f$ is its full width at 
161: half maximum (FWHM)\citep{c1997,c1998,z1998}. Best fit estimates for the 
162: observed QPOs are given in Table~\ref{tab:kHz}. In Fig.~\ref{4kpow} we show a 
163: sample time-averaged PDS from 0.01 Hz to 4 kHz for the dwell starting MJD 
164: 51864.17. There is no observable change in the powerspectrum or QPO frequency 
165: during the dwell. The last QPO occurs just before the source makes the 
166: transition back to hard state from the soft state. We searched the frequency 
167: regime up to 4kHz but there were no evidence of any other harmonics.
168: 
169: Previous work on kHz QPOs has shown an interesting correlation between source 
170: luminosity and QPO frequency \citep{vdk2001,k1998}. It 
171: is observed that on a smaller timescale of hours (e.g. within a single 
172: observation dwell), QPO frequency is well correlated with the flux. However 
173: this correlation is lost in larger timescales of days (e.g between 
174: observations). This leads to formation of ``parallel tracks'' in 
175: $\nu_{QPO}-Flux$ plane. This has been related to existence of two accretion 
176: flows rather than one. Owing to comparatively large number of kHz QPOs 
177: observed during the 2000 outburst of Aql X-1 and relatively high photon count 
178: rate, we could carry out FFT of the data for every 128s segment and track the 
179: QPO evolution (Fig~\ref{parallel}). We find that at lower count rates the 
180: tracks are somewhat more correlated and show the ``parallel tracks'' more 
181: clearly. At higher count-rates, however, even the short-term correlation seems
182:  to break down. This is in agreement with the observations of kHz QPOs in 
183: 4U 1608-52 by \citet{m1999} (see their Fig.2).
184: