The distribution of additives in a metal halide lamp is examined through numerical modelling. A model for a lamp containing sodium iodide additives has been modified to study a discharge containing dysprosium tri-iodide salts. To study the complex chemistry the method of Gibbs minimization is used to decide which species have to be taken into account and to fill lookup tables with the chemical composition at different combinations of elemental abundance, lamp pressure and temperature. The results from the model with dysprosium additives were compared with earlier results from the lamp containing sodium additives and a simulation of a pure mercury lamp. It was found that radial segregation creates the conditions required for axial segregation. Radial segregation occurs due to the unequal diffusion of atoms and molecules. Under the right conditions convection currents in the lamp can cause axial demixing. These conditions depend on the ratio of axial convection and radial diffusion as expressed by the Peclet number. At a Peclet number of unity axial segregation is most pronounced. At low Peclet numbers radial segregation is at its worst, while axial segregation is not present. At large Peclet numbers the discharge becomes homogeneously mixed. The degree of axial segregation at a Peclet number of unity depends on the temperature at which the additive under consideration fully dissociates. If the molecules dissociate very close to the walls no molecules are transported by the convective currents in the lamp, and hence axial segregation is limited. If they dissociate further away from the walls in the area where the downward convective currents are strongest, more axial segregation is observed.

Diffusive and convective processes in the metal-halide lamp cause an unwanted non-uniform distribution of the radiating metal additive (Dy in our case), which results in colour separation. The axial segregation has been described by Fischer (1976 J. Appl. Phys. 47 2954) for infinitely long lamps with a constant axis temperature. However, for our lamps this is not valid. We propose a semi-empirical extended model. The density inhomogeneity gives a measure for the non-uniformity of the Dy density distribution in the lamp. As an example, this parameter is calculated for some measurements obtained by imaging laser absorption spectroscopy.

The results from optical emission spectroscopy experiments of metal-halide lamps under the micro-gravity conditions on board the international space station are compared with the results of a numerical LTE model constructed with the platform Plasimo. At micro-gravity there is no convection which allows for easier modelling and for a separate study of the diffusion-induced radial segregation effect, undisturbed by convection. The plasma parameters that were experimentally determined and compared with the model were the Dy atom and ion density, the Hg ion density and the temperature.The model and experiments applied to a reference lamp burning on a plasma mixture of DyI 3 and Hg were found to be in reasonable agreement with each other. The cross-section for electron-Hg collisions was studied, it was found that the Rockwood values give the correct results. Experimental results guided a sensitivity analysis of the model for the Langevin cross-sections. The ratio of the ion densities Hg + /Dy + was found to be extremely sensitive for the cross-section of the elastic interaction σ (Hg, Dy + ) between the Dy ion and the Hg atom. The sensitivity analysis suggests that equating σ (Hg, Dy + ) to a value that is 10% higher than the Langevin cross-section is the best choice. We also found deviations from LTE in the outer regions of the plasma for relative radial positions of r/R > 50%.

The effect of the competition between convection and diffusion on the distribution of metal halide additives in a high pressure mercury lamp has been examined by placing COST reference lamps with mercury fillings of 5 and 10 mg in a centrifuge. By subjecting them to different accelerational conditions the convection speed of the mercury buffer gas is affected. The resulting distribution of the additives, in this case dysprosium iodide, has been studied by numerical simulations and measurements of the density of dysprosium atoms in the ground state using imaging laser spectroscopy. The competition between axial convection and radial diffusion determines the degree of axial segregation of the dysprosium additives.

The radial temperature profiles of a commercial metal-halide lamp and of lamps containing its individual components have been acquired using x-ray absorption of Hg density distribution. The temperature profiles were determined by combining the measured absorption of the spatially resolved Hg density with the wall temperature. The lamps studied were a commercial lamp, i.e. the Philips CDM-T 70 W/830, and identically shaped lamps containing NaI, DyI3 and TlI separately. It was found that the element Dy contracts the arc, Na broadens the arc while Tl causes the arc to have so-called ‘shoulders’. Combined in the commercial lamp this leads to a wall-stabilized arc without contraction and ‘shoulders’. A reproducibility test with identical lamps was also made and showed that the uncertainty in the temperature profile is about 8% for the absolute temperature and only 1.6% for the actual shape of the profile.