Crevice corrosion occurs when the protective oxide layer that is present on stainless alloys is degraded. The phenomenon is due to a combination of factors. A small leakage current, called the passive current, causes metal dissolution that is initially balanced by the reduction of oxygen in the crevice. Eventually, the crevice becomes deoxygenated because the rates of oxygen consumption in the crevice and diffusion from the bulk solution differ. Once deoxygenated, the cathodic reaction occurs mainly on the bold surface of the metal and the metal ions in the crevice hydrolyze to form an acidic solution. Chloride ions also accumulate in the crevice due to migration. The culmination of these processes is either the breakdown of the passive film or the attainment of a passive steady state condition.

Several mathematical models have been developed to predict the concentration and potential inside crevices. They are often restricted to modelling the process at one temperature and do not consider the effect of transient variations in the bulk solution composition. The majority of models require initial conditions for both the concentration and potential in the crevice.

This thesis describes advances in the mathematical modelling of the initiation period of crevice corrosion. Mass transport was treated rigorously and based on infinitely dilute solution theory. Chemical reactions were decoupled from transport processes by the assumption of equilibrium. The models require the prescription of an initial concentration distribution in the crevice and bulk solution, and boundary conditions in terms of specified concentrations or fluxes.

The first simulation describes the processes that occur during initiation of crevice corrosion on a titanium crevice, immersed in an oxygenated, aqueous sodium chloride environment. For mass transport purposes, the crevice was assumed to be unidimensional and the bulk solution composition was invariant. The model predicted a rapid decrease in pH in the crevice due to the hydrolysis of the dissolution product. The crevice mouth did not deoxygenate which caused a slight increase in the pH. A significant amount of chloride ion was predicted to migrate into the crevice, with the highest concentration occurring at the crevice tip. A study on the effect of the size of the crevice gap, demonstrated that narrower crevices initially have a greater ohmic potential drop, which decreases as the crevice solution becomes more concentrated. The influence on the maximum time step that was prescribed to the model was studied. It was determined that the value of the steady state pH changed as this parameter was increased. This difference was a result of the algorithm reintroducing oxygen into the crevice at larger time steps.

In a second model, the computational domain was extended further into the bulk solution. Mass transport and chemical reaction were simulated in both the crevice and bulk solutions, and charge transfer processes were determined for the crevice walls and bold surface. The crevice and the bold surface were electrostatically coupled using Mixed Potential theory. A simulation was performed for a titanium crevice immersed in deaerated sodium chloride solution. As expected, the predictions showed very little change in pH and chloride ion concentration. The model was initially unstable but the fluctuations in concentration quickly subsided, as the conductivity of the solution increased. A grid independence study showed that the predicted crevice solution composition was relatively independent of the size of the bold surface. The major limitation of the use of this model is the protracted simulation times.

The extended model was then used to simulate the crevice corrosion on a titanium crevice in a neutral oxygenated chloride solution. The extended model predicted a rapid decrease in pH, once the crevice was deoxygenated, to a value lower than that predicted by the one dimensional model. The predicted levels of chloride ion in the crevice with the extended model were also lower. The concentrations of the bulk solution species were determined and it was found that there were small changes in the levels of all species. The value of the bulk solution pH at the bold surface initially increased slightly and this increase proceeded to spread through the solution. The most significant change in concentration in the bulk solution occurred with oxygen. The levels of oxygen decreased at the bold surface in sufficient quantities to deoxygenate the crevice mouth. This resulted in a pH decrease at that location. It was concluded from this study that the constant bulk solution boundary condition, that is implemented in all other crevice corrosion models, is adequate except for the oxygen concentration.

The numerical simulation of the initiation of crevice corrosion on type 304 stainless steel was simulated using a one dimensional model, with constant bulk solution composition. An initial simulation was performed for a temperature of 298 K. The predicted pH was low except at the crevice mouth. The initial pH decrease was attributed to the production of Cr(OH)_{3 <​aq>​} but at steady state the dominant species in the crevice were the unhydrolyzed cations. The chloride ion concentration in the crevice attained a steady state value of approximately 2.8 M at the crevice tip. The model was validated by comparison with the experimental results of others and there was good agreement. A parameter study was performed to determine the relative effects of the passive current, the transport properties and chemical equilibrium on the propensity for crevice corrosion at elevated temperatures. The value of passive current significantly changed the deoxygenation time and the pH in the crevice. Changes to the equilibrium constants at elevated temperature also reduced the pH in the crevice considerably. However, the effect of higher temperatures on the diffusion coefficient and mobility had the opposite effect;​ the pH was found to be less acidic due to increased rates of mass transport. For type 316 stainless steel crevices, the simulations predicted that the solution pH was dependent upon the redox potential prescribed to the model. A lowering of this parameter resulted in pH values that were less acidic than for simulations where the vi potential was higher. However, the results are inconclusive as the simulations were terminated abruptly. The problem was deduced to be in the simulation of the electrochemical reactions of molybdenum. It is recommended that a model for the redox potential be applied to eliminate this difficulty.