I am trying to model a liquid metal melt pool which is created by a laser. I am trying to model transient heat transfer, the solid melting, flows in the surrounding gas, and flows in the melt pool. I am using 3 phases VOF (one for solid, one for liquid metal, one for gas), DO radiation for the laser from a point source, and surface tension at the gas/liquid metal interface is given a temperature-dependent gradient. However I find that it is too easy to make the convection rotate in an anti-clockwise fashion, which is the opposite of what is expected, when normal surface tension values/gradients with temperature are used.