In this paper we adopt a method to analyse absorption line spectra from elliptical galaxies that determines the dynamics of the galaxy and at the same time offers a way to study the stellar populations in that galaxy by a direct fit to the spectra. The result of the modelling is a distribution function for each stellar population that is considered in the fit. The method is described in detail in an accompanying paper. This paper reports on a dynamical stellar population study in NGC3258, based on spectra in two different wavelength regions, the near-IR Ca II triplet around 8600A and the Ca H and K lines around 3900A. These absorption lines have discriminating power toward various stellar types. The dynamical model shows an increase in dwarfs (represented by G2V stars) toward the centre. Most of the rotation in the model is delivered by the giants (represented by M1III stars). Moreover, the different models that were considered indicate that establishing a potential for a galaxy is dependent on the wavelength range used for the modelling.

In this paper we present a method to analyse absorption line spectra of a galaxy designed to determine the stellar dynamics and the stellar populations by a direct fit to the spectra. This paper is the first one to report on the application of the method to data. The modelling results in the knowledge of distribution functions that are sums of basis functions. The practical implementation of the method is discussed and a new type of basis functions is introduced. With this method, a dynamical model for NGC 3258 is constructed. This galaxy can be successfully modelled with a potential containing 30% dark matter within 1r_e with a mass of 1.6x10^11 M_o. The total mass within 2r_e is estimated as 5x10^11 M_o, containing 63% dark matter. The model is isotropic in the centre, is radially anisotropic between 0.2 and 2 kpc (0.88 r_e) and becomes tangentially anisotropic further on. The photometry reveals the presence of a dust disk near the centre.

Current error estimates on kinematic parameters are based on the assumption that the data points in the spectra follow a Poisson distribution. For realistic data that have undergone several steps in a reduction process, this is generally not the case. Neither is the noise distribution independent in adjacent pixels. Hence, the error estimates on the derived kinematic parameters will (in most cases) be smaller than the real errors. In this paper we propose a method that makes a diagnosis of the characteristics of the observed noise The method also offers the possibility to calculate more realistic error estimates on kinematic parameters. The method was tested on spectroscopic observations of NGC3258. In this particular case, the realistic errors are almost a factor of 2 larger than the errors based on least squares statistics.

We use a set of Stackel potentials to obtain a local approximation for an effective third integral in axisymmetric systems. We present a study on the feasibility and effectiveness of this approach. We have applied it to three trial potentials of various flattenings, corresponding to nearly ellipsoidal, disky and boxy density isophotes. In all three cases, a good fit to the potential requires only a small set of Stackel potentials, and the associated Stackel third integral provides a very satisfactory, yet analytically simple, approximation to the trial potentials effective third integral.

Using the photometry and the kinematic data, part of which acquired with the NTT, we constructed a 2-integral and a 3-integral distribution function for NGC 4697. We detected a nuclear dust lane at $3.4''$ or 0.4 kpc from the centre. A comparison of the Lucy-deprojection method and the multi-gaussian expansion method showed that the latter offers a number of advantages, among which a qualitatively better fit and better controllable convergence. Futhermore, we show that a 2-integral model cannot be excluded for as far as the Satoh parameter $k$ is allowed to vary. The 3-integral model is based on a St\"ackel third integral. The original potential is retained where appropriate, in order to minimise errors due to the St\"ackel approximation. On the whole, the 3-integral model produces more satisfying results than the 2-integral model, and is better in handling some peculiarities in the kinematic data. For both models we show detailed contour plots for all deprojected moments up to order 2. The mass-to-light ratio appears to be a fairly well determined $(M/L)_B=4.8h_{50}$, and is much better constrained than would be the case for spherical anisotropic models with comparable data.