Probabilistic Corrosion

The concept of probability distribution was first applied in a corrosion context in the 1930's. In their landmark paper, U.R. Evans and R.B. Mears described a simple arrangement of metallic coupons subdivided in a multitude of small squares by masking with paraffin. From these the influence of sixteen basic variables on the probability and relative velocity of corrosion attack was estimated.

Schematic representation of the experiment carried out by Mears and Evans to demonstrate the probabilistic nature of corrosion phenomena.

The "Probability" of Corrosion

by R.B. Mears and U.R. Evans, Transaction of the Faraday Society, Vol. 31, 1935, 527-542.

The probability of corrosion developing on a small area is controlled by laws quite different from those governing the " conditional velocity " (the average velocity measured in cases where attack occurs at all). The authors have already found that one factor (oxygen-concentration) may shift "probability" and "conditional velocity" in different directions. From the purely scientific standpoint, measurements of probability, which give information regarding the mechanism of breakdown of passivity, are as important as measurements of velocity, which indicate the mechanism maintaining the attack. From the practical standpoint, also, probability has importance ; it may, indeed, be more important to know whether-at a microscopic pore in a protective coat-corrosion is likely to occur at all than to know how quickly it will develop.

The measurements recorded below, however, are concerned with the pure-science aspect, and seek to establish the influence of sixteen "external factors" upon (a) the Probability of attack developing and (b) the early value of the Conditional Velocity, as measured over a 22-hour period. In each series of experiments, all the factors except one were kept constant, thus isolating the effect of varying this one factor. The external factors studied were:

The time of duration of the experiment.
The area of metal exposed to the liquid.
The oxygen-concentration in the atmosphere during the experiment.
The quality of distilled water.
The time of pre-exposure (to dried air).
The temperature of the experiment.
The temperature of pre-exposure.
The concentration of salt used (potassium chloride).
The effect of anion (comparison with other potassium salts).
The concentration of inhibitor (potassium carbonate in mixtures with potassium chloride).
The effect of other Inhibitors.
The effect of additions of acid or alkali.
The effect of sulfur dioxide or carbon dioxide in the gas-phase.
The nature of abrasive treatment and preliminary washing with various liquids.
The character of the Scratch line.
The effect of corrosion in the neighbourhood (protective action of a recent scratch-line for a less recent one).

Summary

(1) Measurements have been made of the influence of sixteen external variables on (a) the Probability and (b) the Conditional Velocity of corrosion by drops. Oxygen-concentration depresses the probability but increases the conditional velocity ; the same is true of certain recognized inhibitors (e.g. potassium carbonate). Pre-exposure to oxygen depresses the probability of attack by certain liquids (0.07 M. sodium bicarbonate), but not by others. Arise in the temperature of pie-exposure tends somewhat to depress the probability, whereas a rise in the temperature of experiment greatly increases the probability, as does also the presence of sulphur dioxide in the gas phase ; carbon dioxide does not appreciably increase the probability, but stimulates the conditional velocity. With increase of potassium chloride, the probability rises, while the conditional velocity first rises and then falls again. The presence of neighboring corroding points depresses the probability ("mutual protective effect"). A comparison has also been made of different drop sizes, different irons and steels, different varieties of distilled water, different anions, different inhibitors, different PH values, different abrasive treatments, different cleaning treatments, and different forms of scratch lines.

(2) Assuming that inhibition occurs only when the corrosion-product formed at a pore in an oxide film is precipitated in physical contact with the metal, thus stifling attack, it is shown that the oxygen-concentration needed to prevent corrosion in a drop is a suitable measure of the size of the biggest pore. From the data connecting probability and oxygen, concentration, it is shown that the pores grow continuously more frequent as the "size" under consideration is reduced, thus indicating that the film-substance possesses an essentially porous character-a conclusion analogous to that reached, by a different argument, after experiments on the growth of silver iodide films.