If you have worked in a lab or industrial plant with a crucible or smelting furnace containing a liquid metal, you probably observed a trail of fume or smoke escaping over the melt. This fume can have serious consequences if it is injurious to your health, or if deposition of the fume causes contamination or damage to surrounding equipment. But have you wondered about the nature of the fume: what it is or how we can understand and control its formation?
In a classic paper, Turkdogan, Grieveson, and Darken of the US Steel Corp. Research Center provided experiments and interpretations to clarify the nature of the metal fuming process [J. Phys. Chem. (1963) 67 (8), 1647]. Experimentally, the rate of mass loss of a pure metal was measured as a stream of argon plus oxygen was passed over the liquid metal. Both the oxygen partial pressure and the gas flow rate were varied systematically for experiments involving six pure metals: Cu, Mn, Fe, Co, Ni, and Cr. For each of these liquid metals, the rate of mass loss was directly proportional to the oxygen partial pressure in the streaming gas phase, with little dependence on the moderate variation in gas flow rate. Perhaps surprisingly, for a given oxygen partial pressure and Ar flow rate, each of these six metals vaporized at the same rate, although the vapor pressures of the individual metals differed greatly. Upon increasing oxygen partial pressure in the inert gas, however, a critical value was reached where vaporization stopped and an oxide skin formed on each of the metals.
The interpretation of these experiments left little doubt about the process particulars. In the flowing inert gas stream, the surface of each liquid metal is clean (unoxidized) and, thus, the metal evaporates (as atoms) as the surface establishes the equilibrium vapor pressure for the given metal. In every case, however, the metal vapor pressure is decades in magnitude lower than the partial pressure of added oxygen. Close to the metal surface, collisions of the metal atoms with oxygen molecules nucleate tiny oxide particles. Growth and agglomeration of these oxide particles creates the visible fume or smoke. The rate of arrival of oxygen molecules via diffusion across an effective stagnant boundary layer of Ar, and not the rate of metal vaporization, is the rate-limiting step. Therefore, the vaporization of all metals depends directly on the oxygen partial pressure, and all metals react at the same rate for a given oxygen partial pressure and gas flow rate. The critical maximum oxygen pressure at which enhanced vaporization stops correlates perfectly with the condition that the rate of oxygen arrival equals the maximum possible (Hertz-Langmuir) rate of metal evaporation. For this condition, the oxide fume precipitation site is pushed against the metal surface so that a passivating oxide film is formed. Obviously, to minimize the rate of fume formation, one should minimize the oxygen content in the surrounding gas phase. On the other hand, the observation of fume is proof that the metal is not dissolving oxygen, because oxygen never reaches the surface. Argon-shielded arc welding is based on this concept. Air, which nominally consists of 79% nitrogen and 21% oxygen, is also an example of a mixture of an oxidant and an inert diluent. Without the nitrogen shroud provided by oxidation in air, reactions in pure oxygen tend to run away (burn up) as the exothermic heat of reaction raises the temperature and thereby the kinetics without any means to limit the oxidant arrival rate. The fiery loss of the Apollo 1 spacecraft and crew on the launch pad served notice about the danger of pure oxygen.
This excellent paper by Turkdogan-Grieveson-Darken (TGD) left one uncertainty, because, without experimentation, silicon was also added to the plots describing their mechanism. Earlier, Wagner had explained that enhanced vaporization of Si in an oxygen plus inert diluent gas occurs because oxygen does reach the surface and the volatile SiO molecule is formed and desorbs [J. Appl. Phys. (1958) 29 (9), 1295]. In fact, these kinetics are also limited by oxygen arrival through a stagnant boundary layer, giving only a factor of two difference in the predicted rates for the differing TGD and Wagner mechanisms. This predicament was nicely resolved by Hinze and Graham, who showed that the Wagner model is correct for Si [J. Electrochem. Soc. (1976) 123, 1066]. So, in general, most metals have higher vapor pressures than their oxides, and the TGD mechanism applies, but a few metals exhibit higher vapor pressures for their oxides, and they follow the Wagner mechanism.