A simple answer is a HUGE difference in the melting temperatures of pure Al and Al2O3. Namely, we have Tm,1=660C for the former and Tm,2=2054C for the latter. To understand how this difference brings about such a strong effect, let us consider the dependence of the surface tension G on the temperature. With quite a good accuracy (the expansion near the melting point) : G(T)=Gm,1+(dG/dT)(T-Tm,1), where Gm,1 is the surface tension at T=Tm,1. That is, for pure Al drops, we have Gm,1=G(Tm,1), by definition. What is important, the derivatives dG/dT at their melting points are both NEGATIVE and practically EQUAL (empirical fact), namely: (dG/dT)m,1= dG/dT)m,2=-0.155x10^{-3}N/mK. Now, given all the above, let us estimate the surface tension of Al2O3 at its melting point. We obtain that Gm,2=G(Tm,2)=Gm,1-0.155x10^{-3}*(Tm,2-Tm,1). Which gives Gm,2-Gm,1=-0.2N/m for the difference between surface tensions of Al2O3 and Al. For exp. values of Gm,1=1.05N/m, we indeed obtain Gm,2=0.85 N/m for the observed value of the surface tension of the oxide. Of course, a more interesting question is WHY these melting temperatures are so different. The answer on this question will take us to Quantum Chemistry. Returning to your original question on surface tension of oxides, any true explanation (including the origin of a rather strong dependence of Gm,2 on oxygen concentration) would require a detailed analysis of surface-interface-surface configuration for liquid Al2O3 based on Laplace equation and much much more. If you are really interested in this stuff, here is a very informative review (not just on Al but on many other metals; by the way, for Fe the above difference can reach 40%).

Thank you for your response Sergei! I still have doubts but I'll try to summarize these doubts and ask again!

Also thanks for the article!

In general, whenever you add to a pure liquid a second component, the surface tension of the liquid changes as a function of the surface segregation of this second component to the surface (Gibbs adsorption equation). This is due to the interaction potentials between the two species and the real effect can be predicted by various theoretical treatments. In the case of liquid metals, oxygen, when in contact with liquid metals, nearly often plays the role of surface active element, segregating to the surface. If you plot the surface tension value vs the log of oxygen concentration in the bulk, you obtain, also for Al, an S shaped curve, starting from a plateau region at low oxygen concentration and dropping down until a new plateau is found at higher oxygen concentration. (However, some specific points are to be discussed at VERY low oxygen conc.).

You may find exactly the same behaviour in all systems, be them water-based systems, other organic fluids or molten alloys, whenever and "active" component is added to the pure substance. I think you can easily find reference papers on this subject. Ask me if needed!

regards