A numerical simulation of the travelling heater method (THM) process in the growth of HgTe is carried out. The whole system (furnace, ampoule and charge) is taken into account in the frame of a quasi-steady-state model. The mass conservation condition for the solute in the liquid zone permits the determination of the rate of advance of the crystallisation isotherm as a function of the heater position. We claim to study the evolution of different magnitudes along the growth process, searching for the physical reasons which could be at the origin of defects in the form of thin layers observed in some growing experiences. To solve the governing equations of fluid flow, heat transfer and mass transport we have made use of a commercial code which can run in a PC. The simulation is made by using a three-level strategy, which allows the reduction of the computational effort. In the first level, heat transport is assumed to be by conduction, convection and radiation between the furnace and the ampoule, and by conduction through the ampoule wall, coating, solid and liquid zones. The temperature calculated at this level in the air/ampoule boundary is used as boundary condition for the second and third level. In these two levels the ampoule and its content are studied in detail. Convection in the liquid zone is considered at the second level and thermosolutal convection is finally included at the third level. The analysis of the incoming/outcoming heat flux per second through the ampoule for the whole system shows that the lower part of the ampoule exhibits some ineffectiveness for the heat evacuation at certain positions of the growth run, depending on thermal properties of the whole system and the particular material to be grown. As a consequence, the growth rate suffers a significant variation just for these positions of the heater. From these considerations a plausible interpretation has been proposed to understand the apparition of solvent inclusions in the form of thin layers when a material as HgTe is grown by the THM method. (C) 2002 Elsevier Science B.V. All rights reserved.