All Figures

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f1.eps} \end{figure}$ Figure 1: Top panel: PCU2 light curve of the full outburst of GX 339-4, with one point per observation. Bottom panel: corresponding hardness ratio (see text). The dotted lines indicate major state transitions (see text). The horizontal dashed line shows the hardness level corresponding to the transition from the right branch to the left branch in the HID (see text and Fig. 2).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f2.eps} \end{figure}$ Figure 2: Hardness-Intensity diagram corresponding to the points in Fig. 1. Each point corresponds to one observation. Dotted lines correspond to those in Fig. 1. Observation #0 is indicated by an empty circle. The gray symbols mark the observations included in the groups B (stars), A (diamonds), U1 (squares), U2 (circles) and weak (triangles) (see Sect. 6). The gray arrows indicate the general time evolution along the outburst.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f3.eps} \end{figure}$ Figure 3: Four average power spectra from the right branch and four from the top branch. From top to bottom, they correspond to observations #2, 17, 24, 25, 27, 28, 30, 33. The power spectra are plotted in the $\nu P_\nu$ representation (Belloni et al. 1997) and are shifted vertically for clarity. The increasing trend at high frequencies that can be seen in some power spectra is made of 3 $\sigma$ upper limits, shown as inverted triangles.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.5cm,clip]{2457f4.eps} \end{figure}$ Figure 4: First three panels: time evolution of the characteristic frequencies of the L₁, L₂ and L₃ components for the top/right-branch observations at the beginning of the outburst. Fourth panel: $\nu _2$ vs. $\nu _1$ correlation. In all panels, the dotted line separates the right-branch points from the top-branch points.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.5cm,clip]{2457f5.eps} \end{figure}$ Figure 5: Phase lags for the right and top branches. a) Time evolution of the lags of the 1-64 Hz continuum. b) Time evolution of the QPO lags. c) Correlation between the lags of the 1-6 Hz continuum and the characteristic frequency $\nu _1$ of the L₁ component. d) Correlation between the lags of the QPO and its characteristic frequency.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.5cm,clip]{2457f6.eps} \end{figure}$ Figure 6: Phase-lag spectra for three representative observations from the right (Obs. #8), top (Obs. #29) and left (Obs. #34) branches. The vertical dashed lines mark the centroid frequency of the QPO.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.5cm,clip]{2457f7.eps} \end{figure}$ Figure 7: Evolution of the $\nu _1$ frequency as a function of count rate ( top panel) and hardness ( bottom panel) for the observations of the right and top branches (filled circles) and of the bottom branch (empty circles). The X axes of both panels are the same as on the corresponding axis in Fig. 1.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.7cm,clip]{2457f8.eps} \end{figure}$ Figure 8: Top panel: correlation between $\nu _1$ and $\nu _Q$ (large circles) and points from the WK correlation (Wijnands & van der Klis 1999; Belloni et al. 2002). The upper points are our detections of $\nu _0$ from the top branch. Middle panel: correlation between $\nu _Q$ and $\nu _2$ (large circles) and points from the PBK correlation (Psaltis et al. 1999; Belloni et al. 2002). Bottom panel: correlation between $\nu _3$ and $\nu _2$ (large circles) and points from the PBK second correlation (Psaltis et al. 1999; Belloni et al. 2002).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.7cm,clip]{2457f9.eps} \end{figure}$ Figure 9: Top panel: a 500-s segment of the light curve of Obs. #30 (1 s bin size); count rate is total count rate for all PCUs. Bottompanel: corresponding spectrogram (time integration 4 s, time step 1 s), where darker points correspond to higher power. Inset: average power spectrum from the full observation.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.1cm,clip]{2457f10.eps} \end{figure}$ Figure 10: Plot of integrated fractional rms versus hardness for the observations of the right, top, and upper left branch. Different symbols correspond to: right branch (open circles), top branch (asterisks), type-B QPO (diamonds), weak QPO (filled circles), LF noise (squares), red noise (plus signs) and no signal (dots). For the the definitions of the different elliptical regions, see Sect. 6.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f11.eps} \end{figure}$ Figure 11: The Poissonian-subtracted power spectrum of the five observations with a type-B QPO (see text), vertically shifted for clarity. From top to bottom: Obs. #34, 41, 42, 43, 97.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.5cm,clip]{2457f12.eps} \end{figure}$ Figure 12: Average power spectra of the observations classified as ( from top to bottom) type A, weak signal, U2 and U1 (see Fig. 10). The top power spectrum is in power units, all the others are shifted downwards by a factor of 10 from each other. The lines are the best fit models described in the text.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f13.eps} \end{figure}$ Figure 13: Plot of integrated fractional rms versus hardness for the observations of the bottom. Open symbols represent the same points as in Fig. 10: right branch and type C (circles), diamonds (type B), triangles (U1/U2), squares (type a), and dots (weak signal). Filled symbols correspond to different power-spectrum shapes observed in the bottom branch (see text): flat-top noise (circles), weak noise (triangles), 1 Hz QPO (diamond).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f14.eps} \end{figure}$ Figure 14: Top panel: 900s of the light curve of Obs. #149h (16 s bin size). Bottom panel: corresponding spectrogram (time integration 16 s, time step 4 s), where darker points correspond to higher power. Inset: average power spectrum from the full observation.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f15.eps} \end{figure}$ Figure 15: Four representative power spectra from the bottom-branch observations. The time sequence here is bottom to top: observations #161, 158b, 155, 149g. The power spectra are plotted in the $\nu P_\nu$ representation (Belloni et al. 1997).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.2cm,clip]{2457f16.eps} \end{figure}$ Figure 16: Same as Fig. 4, but for the bottom-branch observations.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f17.eps} \end{figure}$ Figure 17: WK correlation (see Wijnands & van der Klis 1999) for the observations on the bottom branch of the HID (filled circles for $\nu _b$ and empty circles for $\nu _h$ ). The small points are from Belloni et al. (2002).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f18.eps} \end{figure}$ Figure 18: Same Hardness-Intensity as in Fig. 2, but areas marking the regions corresponding to the source states described in the text.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=8.4cm,clip]{2457fA1.eps} \end{figure}$ Figure A.1: Same as Fig. 9 for Obs. #41.

In the text

$\begin{figure} \par\includegraphics[width=8.4cm,clip]{2457fA2.eps} \end{figure}$ Figure A.2: Same as Fig. A.1 for Obs. #42.

In the text

$\begin{figure} \par\includegraphics[width=8.4cm,clip]{2457fA3.eps} \end{figure}$ Figure A.3: Same as Fig. A.1 for Obs. #43.

In the text

$\begin{figure} \par\includegraphics[width=8.5cm,clip]{2457fA4.eps}\end{figure}$ Figure A.4: Same as Fig. A.1 for Obs. #97.

In the text

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f1.eps} \end{figure}$	Figure 1: Top panel: PCU2 light curve of the full outburst of GX 339-4, with one point per observation. Bottom panel: corresponding hardness ratio (see text). The dotted lines indicate major state transitions (see text). The horizontal dashed line shows the hardness level corresponding to the transition from the right branch to the left branch in the HID (see text and Fig. 2).
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.5cm,clip]{2457f4.eps} \end{figure}$	Figure 4: First three panels: time evolution of the characteristic frequencies of the L₁, L₂ and L₃ components for the top/right-branch observations at the beginning of the outburst. Fourth panel: $\nu _2$ vs. $\nu _1$ correlation. In all panels, the dotted line separates the right-branch points from the top-branch points.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.5cm,clip]{2457f5.eps} \end{figure}$	Figure 5: Phase lags for the right and top branches. a) Time evolution of the lags of the 1-64 Hz continuum. b) Time evolution of the QPO lags. c) Correlation between the lags of the 1-6 Hz continuum and the characteristic frequency $\nu _1$ of the L₁ component. d) Correlation between the lags of the QPO and its characteristic frequency.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.5cm,clip]{2457f6.eps} \end{figure}$	Figure 6: Phase-lag spectra for three representative observations from the right (Obs. #8), top (Obs. #29) and left (Obs. #34) branches. The vertical dashed lines mark the centroid frequency of the QPO.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.5cm,clip]{2457f7.eps} \end{figure}$	Figure 7: Evolution of the $\nu _1$ frequency as a function of count rate ( top panel) and hardness ( bottom panel) for the observations of the right and top branches (filled circles) and of the bottom branch (empty circles). The X axes of both panels are the same as on the corresponding axis in Fig. 1.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.7cm,clip]{2457f9.eps} \end{figure}$	Figure 9: Top panel: a 500-s segment of the light curve of Obs. #30 (1 s bin size); count rate is total count rate for all PCUs. Bottompanel: corresponding spectrogram (time integration 4 s, time step 1 s), where darker points correspond to higher power. Inset: average power spectrum from the full observation.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f11.eps} \end{figure}$	Figure 11: The Poissonian-subtracted power spectrum of the five observations with a type-B QPO (see text), vertically shifted for clarity. From top to bottom: Obs. #34, 41, 42, 43, 97.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.5cm,clip]{2457f12.eps} \end{figure}$	Figure 12: Average power spectra of the observations classified as ( from top to bottom) type A, weak signal, U2 and U1 (see Fig. 10). The top power spectrum is in power units, all the others are shifted downwards by a factor of 10 from each other. The lines are the best fit models described in the text.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f14.eps} \end{figure}$	Figure 14: Top panel: 900s of the light curve of Obs. #149h (16 s bin size). Bottom panel: corresponding spectrogram (time integration 16 s, time step 4 s), where darker points correspond to higher power. Inset: average power spectrum from the full observation.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f15.eps} \end{figure}$	Figure 15: Four representative power spectra from the bottom-branch observations. The time sequence here is bottom to top: observations #161, 158b, 155, 149g. The power spectra are plotted in the $\nu P_\nu$ representation (Belloni et al. 1997).
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.2cm,clip]{2457f16.eps} \end{figure}$	Figure 16: Same as Fig. 4, but for the bottom-branch observations.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f17.eps} \end{figure}$	Figure 17: WK correlation (see Wijnands & van der Klis 1999) for the observations on the bottom branch of the HID (filled circles for $\nu _b$ and empty circles for $\nu _h$ ). The small points are from Belloni et al. (2002).
Open with DEXTER

$\begin{figure} \par\includegraphics[width=7.4cm,clip]{2457f18.eps} \end{figure}$	Figure 18: Same Hardness-Intensity as in Fig. 2, but areas marking the regions corresponding to the source states described in the text.
Open with DEXTER