Towards a consistent model of the hot quadruple system HD 93206 = QZ Carinæ: II. N-body model
HD 93206 is early-type massive stellar system, composed of components resolved by direct imaging (Ab, Ad, B, C, D) as well as a compact sub-system (Aa1, Aa2, Ac1, Ac2). Its geometry was already determined on the basis of extensive photometric, spectroscopic and interferometric observations. However, the fundamental absolute parameters are still not known precisely enough. We use an advanced N-body model to account for all mutual gravitational perturbations among the four close components, and all observational data types, including: astrometry, radial velocities, eclipse timing variations, squared visibilities, closure phases, triple products, normalized spectra, and spectral-energy distribution (SED). The respective model has 38 free parameters, namely three sets of orbital elements, component masses, and their basic radiative properties ($T$, $\log g$, $v_{\rm rot}$). We revised the fundamental parameters of QZ Car as follows. For a model with the nominal extinction coefficient $R_V \equiv A_V/E(B-V) = 3.1$, the best-fit masses are $m_1 = 26.1\,M_{\rm S}$, $m_2 = 32.3\,M_{\rm S}$, $m_3 = 70.3\,M_{\rm S}$, $m_4 = 8.8\,M_{\rm S}$, with uncertainties of the order of $2\,M_{\rm S}$, and the system distance $d = (2800\pm 100)\,{\rm pc}$. In an alternative model, where we increased the weights of RV and TTV observations and relaxed the SED constraints, because extinction can be anomalous with $R_V \sim 3.4$, the distance is smaller, $d = (2450\pm 100)\,{\rm pc}$. This would correspond to that of Collinder 228 cluster. Independently, this is confirmed by dereddening of the SED, which is only then consistent with the early-type classification (O9.7Ib for Aa1, O8III for Ac1). Future modelling should also account for an accretion disk around Ac2 component.