We know that the phase space of nonlinear dynamical systems consists of periodic, quasi-periodic and chaotic trajectories. Chaotic trajectories visit resonance islands, remain there for some time and then escape to the chaotic region during its evolution. To know the time interval of resonance trapping and resonance transition and to visualize the phenomenon of transient chaos are some important questions which compel us to study more about the application of different chaos indicators. Over the last few years, several chaos indicators have been introduced to study those aspects. Moreover, for dynamical system of two degrees of freedom, there are several chaos indicators such as the Poincare Surface of Section (PSS), Largest Lyapunov Characteristic Exponent (LLCE), Smaller Alignment Index (SALI). Fast Liapunov Indicators (FLI), the Generalized Alignment Index (GALI) and the Correlation Dimension (CD) (see [1] -[3] ). Visualization of the Poincare surfaces of section (PSS) is very useful for the study of the dynamical system of two degrees of freedom. But in case of three degrees of freedom, it becomes four-dimensional which is difficult to analyze. On the other hand, other indicators mentioned above require the solution of equations of motion and the first order variational equations whose computation is not easy in case of higher dimensional systems. Also, they do not tell about additional qualitative features like whether a trajectory is resonant or non-resonant. Recently in [4] , a new indicator (i.e. Total angular momentum) is introduced to study the ordered and chaotic motion of the asymmetric triaxial galaxy, including the effect of spherical dark halo components. (TAM) is used to describe the nature of orbits in this potential and is proved to be a fast and reliable indicator in comparison to a Lyapunov characteristic exponent and P(f)-indicator. The method of total angular momentum can distinguish regular trajectory from chaotic. But there is no clarity about the other aspects such as stickiness, resonance trapping and transition and transient chaos, which must be investigated. In [4] , the effect of dark halo component in the chaotic region is shown with the help of the Poincare surface of section by reducing the three-dimensional galactic model to two dimensions. Time-frequency analysis based on the extraction of instantaneous frequency from wavelet transform is computed via two ways. The first method is the computation of instantaneous frequency of the phase of continuous wavelet transform (CWT) and the second method is based on the computation of instantaneous frequency from the amplitude of the continuous wavelet transform. The amplitude-based method is very well described in [5] . Characterization of the phase space of standard map and the Hamiltonian of the hydrogen atom (moving in crossed magnetic and elliptically polarized microwave fields) is presented using this method. On the other hand, phase-based approach is already explained in [6] - [9] . In [10] , we can see the application of the phase-based approach to the circular restricted three-body problem (Sun-Jupiter system) to explain the phenomenon of resonance transition and transport condition. The phenomenon of resonance trapping of chaotic trajectory, resonance transition and transient chaos (time dependent system causes an orbit move from regular to chaos and vice-versa (see [11] ) can be computed and visualized with the help of TFA. Identification between regular and chaotic, periodic and quasi-periodic, chaotic sticky and chaotic non-sticky is possible with the help of TFA. Also, the computational time for TFA is negligible as compared to other chaos indicators. The aim of this paper is two folded. The first is to establish the method of time-frequency analysis in comparison to total angular momentum. The second is to investigate and characterize the phase space structure of asymmetric triaxial galaxies in the presence and absence of spherical dark halo components.

We have organized the paper as follows:

In Section 2, we have given a brief description of asymmetric triaxial galaxy enclosed by spherical dark halo component (3D). In Section 3, a brief description of TFA based on the phase of CWT and its implementation in Matlab are given. Results and discussion based on the application of TFA to the three-dimensional galactic model are shown in Section 4. The conclusion is given in Section 5.

The potential for triaxial galaxy enclosed by spherical dark halo (see [4] ) component is given by

where

and

Equation (2) denotes a triaxial galaxy with a bulge and a small asymmetry introduced by the term. The parameters and denote the flattening of the galaxy and denotes the scale length of the bulge of the galaxy. The parameter is used for consistency of galaxy units. Equation (3) presents a spherical dark halo component. Here and are the mass and the scale length of the dark halo com- ponent, respectively. Equations (1), (2) and (3) together represents the three-dimensional galactic model. Now we can write the equations of motion as follows,

The Hamiltonian of the potential given by Equation (1) can be expressed as

where, and are the momenta corresponding to coordinates x, y and z respectively. Also, denotes the numerical value of the Hamiltonian or the energy constant. In Equation (5), if we take, we get the potential for two dimensional triaxial galaxy which can be expressed as

where stands for numerical value of Hamiltonian in two dimension. In this paper, we use a system of units which is defined as follows:

Unit of length = 1 kpc;

Unit of mass = 2.325 ´ 10⁷ M_e;

Unit of time = 0.97748 ´ 10⁸ yr;

Unit of velocity = 10 km×s^-1;

Unit of energy (per unit mass) = 100 km²×s^-2;

G = 1 (gravitational constant).

While integrating the equations of motion in (4) for the computation of all the orbits, we use the fixed value of and where as and are taken as parameters. We calculate the trajectory by integrating the equations of motion given in Equation (4). It is done using Runge Kutta variable step size Integrator. Accuracy of calculations is maintained up to eight significant figures.

In this Section, we analyze the data of Table 1 and Table 2 and discuss the plots obtained by integrating the

equation of motion (4) at the given initial conditions using Runge-Kutta (4/5) variable step-size Integrator. Phase-portrait and -plot is drawn with the help of software Mathematica and Ridge-plot is drawn in Matlab.

At first, we discuss the results obtained using the Poincare surface of section in Figure 2. “Section condition” taken for the Poincare plot is and. Initial conditions are taken as on x-axis on the interval (−10, 10) with step-size 0.1 and integrated at each initial condition up to 2500 time units. is calculated with the help of Equation (5). Both results of PSSs are in confirmation of the Poincare plots given in [4] . In the absence of the dark halo component, almost all phase plane is filled up with a hazy collection of points except few small regular regions which consist of a set of islands. On the other hand, in the presence of the dark halo component we notice the substantial reduction in chaotic regions and increase in regular regions. Thus, we can say that the dark halo component serves as chaos controller in triaxial galaxies with small asym- metries. These observations confirm the result of [4] .

We have selected a sample of five representative orbits for both cases. Time interval considered for the TFA is (1, 32768) except two figures (Figure 3 and Figure 4). For Figure 3 and Figure 4, the interval of time unit is taken as (1, 65536) to present the results in a better way. It adds an additional expense of eighteen seconds in the C.P.U time for the execution of the program. In Figure 5 and Figure 6, we consider the orbits at initial conditions (−9.55, 0, 0.1, 0, 5.8029, 0) and (−7.5, 0, −1, 0, 7.4008, 0) respectively. In Ridge-plots of both figures, we notice a completely flat Ridge throughout the motion. We know that frequency remains constant for the periodic orbit and hence both orbits are periodic.

Note: In [4] , at same initial condition (−9.55, 0, 0.1, 0, 5.8029, 0) the orbit is termed as quasiperiodic on the basis of -plot. But, according to Ridge-plot this trajectory is periodic. Also, the phase portrait and -plot in Figure 5 is exactly similar to the figure of [4] .

Sample of orbits considered in Figure 7 and Figure 8 is at initial conditions (5.5, 0, 0.6, 0, 20.4251, 0) and (3, 0, 0.5, 0, 20.5996, 0) respectively. We observe a very little variation in Ridge curve (or instantaneous frequency) of both figures. In the plots of, we notice symmetric peaks indicating regular motion. On the basis of both Ridge-plots, we term both trajectories as quasi-periodic. Also plots of both orbits are confirming the result of -plot in [4] .

Now we consider the orbits at initial conditions (−0.5, 0, 0.85, 0, 27.9848, 0) and (0.1, 0, 0.5, 0, 24.5701, 0) are shown in Figure 9 and Figure 10 respectively. If we look at both Ridge-plots, we can say that the trajectories are chaotic in nature. Also, highly asymmetric features and large deviation between maxima and between minima of -plot confirms the chaotic behavior of trajectories. But in addition to this, we find very little variation in both Ridge-plots in the time intervals (0, 20000) and (0, 12000) respectively. It means that both trajectories are trapped on resonance islands for long time intervals and then escape to the chaotic region. This phenomenon is known as stickiness. Thus, we term both trajectories as chaotic sticky. But in case of plots, it is not clear.

Now, we consider the sample of trajectories at initial conditions (−0.9, 0, 0.85, 0, 27.6788, 0) and (1.7, 0, −1, 0, 22.2880, 0) presented in Figure 11 and Figure 12 respectively. If we look at -plots of both figures, we find similar results as in the previous case. But, if we look at the Ridge-plots of both figures we notice a continuous variation in the instantaneous frequency throughout the motion and hence we call these trajectories

as chaotic non-sticky.

At last, we consider sample of two orbits at initial conditions (0.7, 0, 0.85, 0, 27.8624, 0) and (0.01, 0, 0.1, 0, 24.8801, 0) shown in Figure 3 and Figure 4 respectively. Again in the plots of, we observe the same

phenomenon as in the previous case and hence we can say that these trajectories are chaotic. But in case of Ridge-plots scene is little different. We observe following things in the Ridge-plots:

1) Resonance trapping: In Figure 3 at time intervals (0, 13000), (19000, 25000) and (50000, 60000) approx we find a little variation in the frequency and besides that there is rapid change in frequency. Similarly, in Figure 4 at time intervals (0, 5000), (10000, 15000), (25000, 32000) and (41000, 60000) there is very little change in instantaneous frequency and besides that rapid change in frequency takes place. This is the case of resonance trapping which can be visualized and also accurately determined with the help of TFA.

2) Resonance transition: In Figure 4, we notice that at time unit 42,000 approx there is a jump from one type of resonance to another type of resonance and remains there for approx 18,000-time units. This phenomenon is known as resonance transition which is clearly observed in the Ridge-plot.

3) Transient Chaos: We know that in a time-dependent system a chaotic trajectory move from regular to chaos and vice versa (see [11] ). We call this phenomenon as transient chaos. We can visualize this phenomenon in the Ridge-plots.

As we have already discussed, the aim of the present work is to show the advantage of TFA in comparison to TFA and to explore some additional information of phase space structures of asymmetric triaxial galaxies in the presence and absence of the spherical dark halo component. Based on the discussion of Section 4, we can conclude that TFA has several advantages in comparison to TAM. We conclude following things:

1) TFA based on wavelets is comparatively fast and more reliable in comparison to TAM (C.P.U time taken for the computation of Ridge-plots for 32,768-time units is 15 seconds (maximum) whereas the time taken by TAM for 150-time units is 5 seconds (maximum)).

2) TFA can identify between periodic and quasi-periodic, chaotic sticky and non-sticky, and ordered and chaotic motion.

3) With the help of TFA, we can accurately determine and also visualize the event of trapping of a chaotic trajectory around resonance island (see Figure 3 and Figure 4).

4) The phenomenon of resonance transition and transient chaos can also be explained with the help of Ridge-plot (see Figure 3 and Figure 4).

5) Computational effort needed for programming of TFA based on wavelets is not easy in comparison to TAM. This is an important drawback of TFA based on wavelets. But once it is done, we can perform other computational works in comparatively negligible time.

6) We always search for an indicator which is applicable to higher-dimensional nonlinear dynamical systems. TFA is independent of the degree of freedom and requires the only solution of equations of motion which can be computed. Our present work is also an important example of the application to higher dimensional systems.

Thus, we can say that Time-frequency analysis based on wavelets can be given preference for the study of nonlinear dynamical systems for two or more degrees of freedom.