You are on page 1of 2

Evidence for inhibition between pitting events on carbon steel

T. P. HUTCHINSON
In data published by Cheng and Luo ( British Corrosion Journal, 2000) , the variance to mean ratio of the number of pitting events on carbon steel is less than 1. It is here argued that this is evidence for inhibition between events, not for promotion. T he branch of probability known as renewal theory is useful for analysing counts of events occurring in time, the intervals between them, and the intensity of events. BCJ /1913 T he author is in the Department of Psychology, M acquarie University, Sydney, N SW 2109, A ustralia (phutchin@bunyip.bhs.mq.edu.au). M anuscript received 17 September 2001; accepted 9 July 2002.
2002 IoM Communications Ltd. Published by Maney for the Institute of M aterials, M inerals and M ining.

INTRODUCTION
Cheng and Luo1 reported experiments on metastable pitting of A51670 carbon steel. In some conditions, pitting (as indicated by the occurrence of a current transient) was random, in the sense of appearing to be a Poisson process: the variance to mean ratio of the number of events in a given time interval was close to 1. In other conditions (e.g. data for an immersion time of 5 h in Fig. 4 of Cheng and Luo, a potential of 50 mV in Fig. 7, and a chloride ion concentration of 01M in Fig. 10), the Poisson model failed. Reading Ref. 1, the impression gained is that when the Poisson model was unsatisfactory, this was because of cooperation (promotion) between the pitting events: that is, the occurrence of one event was associated with an increased probability of another event. It is argued in this paper that, on the contrary, the data indicate that when the Poisson model fails, this is because of inhibition between the pitting events: that is, the occurrence of one event is associated with a reduced probability of another event.

VARIABILITY OF COUNTS AND OF TIME INTERVALS


Considering the number of metastable pitting events per time interval, the ratio of variance to mean is seen to be either close to 1 or less than 1 in Figs. 4, 7, and 10 of Cheng and Luo. The latter indicates an inhibitory process, not a cooperative one. (In contrast, simulated data in Fig. 5 of Lunt et al.2 show a variance to mean ratio of greater than 1; the model that was used to generate these data was a cooperative one.)3 It is common to consider the variability of the time intervals between successive events, rather than the variability of counts. Issues studied include the distribution and the mutual dependence of these time intervals. When the intervals are independent and identically distributed and have an exponential distribution, the process is a Poisson one. The intervals being independent and identically distributed is a common nding in experimental work and a common assumption in theoretical work. The term renewal process is then used. Relaxation of the assumption of the exponential distribution is the usual way of generalising beyond the Poisson process. Probabilists are capable of stating precisely and proving theorems that connect the properties of the intervals between successive events with those of the numbers of events in a time period. For example, asymptotically, the variance to mean ratio of the count of events equals the square of the coeYcient of variation of the lengths of the intervals. (The coeYcient of variation is the ratio of standard deviation to mean.) These two ratios are both 1 for the Poisson process.
DOI 10.1179/000705902225004428

There is no necessary connection between the mean and the coeYcient of variation of the intervals, in the sense that the same external conditions (e.g., as regards immersion time, potential, and chloride ion concentration) that lead to a small mean (and thus a high rate of events) could also lead to a small coeYcient of variation (and thus the appearance of inhibition between events). Reference 4 is a standard work. The Encyclopedia of Statistical Sciences is an excellent starting point for many topics within its scope; relevant articles include that by Smith.5 Let r be the intensity of events, meaning that for an in nitesimal length of time dt, r . dt is the probability of an event. This intensity may be written r(t), to indicate possible dependence on the time that has elapsed since the previous event. Assume the interevent intervals to be independent. (If r(t) is constant, we have the Poisson process.) Example 1: suppose r(t) is an increasing power function t a (with (a> 0). Example 2: suppose r(t) is 0 for t < t, and is constant for t > t, where t is some positive length of time (either a constant, or if a random variable, then with a coeYcient of variation less than 1). It seems fair to describe both examples as exhibiting inhibition: the intensity is 0 immediately following an event, and is non-decreasing until the next event occurs. For both examples, the coeYcient of variation of the intervals is less than 1, and thus the variance to mean ratio of the counts is less than 1 also. (Notice that memory is unnecessary for inhibition or cooperation to be observed, except in as much as t being measured from when an event last occurred can be said to constitute memory. The model proposed by Wu et al.3 and simulated by Lunt et al.2 was not a renewal process, it had memory and there was an implied positive correlation between successive intervals.)

DISCUSSION
It is concluded that the data in Figs. 4, 7, and 10 of Cheng and Luo are evidence for inhibition between metastable pitting events. (The author believes the autocorrelations shown in Figs. 5, 8, and 11 to arise from the length that an event lasts. The timescale is a matter of seconds, whereas interevent intervals are tens or hundreds of seconds.) I have no specialist knowledge of electrochemistry to enable me to say what the inhibitory process might be, but at page 129, Cheng and Luo themselves describe one: The close spatial proximity of two or more metastable pits forming within a short time interval can lower the chance of occurrence for subsequent pits, because the former pits serve as an anodic area with respect to the surrounding region and inhibit the initiation of subsequent pits. Cheng and Luo give no reason why this process of inhibition should require two or more pits; perhaps one is suYcient.
British Corrosion Journal 2002 Vol. 37 No. 3 239

240 Hutchinson

Inhibition between pitting events on carbon steel 3. b. wu, j. r. scully, j. l. hudson, and a. s. mikhailov: J. Electrochem. Soc., 1997, 144, 16141620. 4. d. r. cox: Renewal theory; 1962, London, Methuen. 5. w. l. smith: in Encyclopedia of statistical sciences, Vol. 8, 3036: 1988, New York, NY, Wiley.

REFERENCES
1. y. f. cheng and j. l. luo: Br. Corros. J., 2000, 35, 125 130. 2. t. t. lunt, s. t. pride, j. r. scully, j. l. hudson, and a. s. mikhailov: J. Electrochem. Soc., 1997, 144, 16201629.

British Corrosion Journal

2002 Vol. 37

No. 3

You might also like