Development and innovative use of corrosion inhibitors is of ongoing interest to the corrosion community. This interest is shown by the large number of papers submitted to Corrosion associated with this subject. Such submissions are encouraged and welcomed. Throughout my 15 years on the Editorial Board of Corrosion, I have handled several hundred papers, many of which involve corrosion inhibitor development and evaluation. During this time, I have noticed areas of data evaluation and reporting that might be improved to better serve the corrosion community. These areas are the use of “inhibitor efficiency” as almost the sole method to characterize inhibitor efficacy, the dearth of replicate experiments or inappropriate exposure times for determining corrosion rates especially from mass-loss experiments, the sometimes automatic use of curve-fitting of polarization data for estimating corrosion rates without critical evaluation of the calculated terms, and the inappropriate use of significant digits in reporting results. My hope is to stimulate a dialogue among researchers on these and other subjects pertaining to the handling and presenting of scientific information so that the articles that appear in Corrosion best serve the corrosion community.
One often-used way of reporting inhibitor effectiveness is through a term called “inhibitor efficiency.” This quantity is usually defined as the quotient of the difference between the corrosion rate in the uninhibited and inhibited solutions divided by the corrosion rate in the uninhibited solution, often multiplied by 100 to make the final value a percent. While using such a measure is not inherently wrong, taken alone it has only limited value. The calculated efficiency provides little information about the actual corrosion rate or mass loss, the quantity that is important for making practical decisions about the effectiveness of the inhibitor or deciding to pursue follow-up work with that inhibitor. For example, efficiencies between 90% and 95% are sometimes attributed to compounds that are labeled as effective. But, that range would be consistent with the following two cases: a change in corrosion rate from 100 mpy without the inhibitor to 5 mpy to 10 mpy with the inhibitor and from 1,000 mpy without the inhibitor to 50 mpy to 100 mpy with the inhibitor. Both cases reflect corrosion inhibition; that is, the use of the compound decreases the corrosion rate. But, under many circumstances, the molecule creating the first result would be of more immediate practical interest than the molecule creating the second result. Corrosion rates of 50 mpy to 100 mpy are usually not acceptable from a performance standpoint and are often of little further interest. One might even question the value of the time spent writing a paper discussing such a compound. I am not criticizing the reporting of the inhibitor efficiency, because that measure can be useful especially when comparing or ranking members of a class of compounds. What I am suggesting is that corrosion inhibitor papers should always include actual corrosion rates for all reported conditions along-side the efficiencies so the reader can easily discern the usefulness of the molecule. In addition, if the corrosion rate is not really practical, even though the efficiency is high (along the lines of the second case above), the paper should include additional scientific or other information (not supposition) supporting why that particular molecule or family is important to the corrosion community.
Sometimes corrosion rates are reported from mass-loss experiments under conditions in which either the exposure time was extraordinarily short or only one experiment was run under each condition. Several years ago, an article was published (R.A. Freeman and D.C. Silverman, “On Error Propagation in Coupon Immersion Testing,” Corrosion, Vol. 48, No. 6 [1992], p. 463) that attempted to examine how experimental errors in measurements of time, weight, and dimensions can propagate through the calculation and the percentage error that would result in the corrosion rate calculation. The results showed that the final errors are interrelated. The paper provides an estimate of the minimum error for several experimental conditions when the only sources of error are these three experimental parameters. One implication is that, especially at lower corrosion rates, a one-day exposure could be inadequate in terms of the amount of inherent error created even assuming that the corrosion rate was at steady state during the entire exposure. In this case, performing many replicate experiments under identical conditions would not increase the reliability of the numbers. A greater awareness of this type of consideration seems to be needed when designing the protocol for and reporting results from mass-loss experiments.
Additional sources of error beyond those related to measuring time, weight, and dimensions exist, e.g., presence of nonuniform penetration, variation in corrosion rate as steady state is approached, projected and actual surface areas being different because of surface roughness (the calculation is based on projected surface area, not actual surface area), incomplete corrosion product removal, change in overall surface area as with high corrosion rates, and corrosion or penetration rate changing with time. The implication is that more than one exposure is needed. The task of deciding on the number of replicate experiments required to be confident in the corrosion rate estimate is difficult. Some guidance might be obtained in ASTM Standard G-31, “Standard Practice for Laboratory Immersion Corrosion Testing of Metals” and ASTM Standard G-16, “Standard Guide for Applying Statistics to Analysis of Corrosion Data.” But, the implications are that the experimental conditions must be defined clearly, appropriate immersion times must be chosen to minimize the inherent errors, and more than one experiment at each condition should be standard procedure to provide some idea of the variability.
Often results are presented from both mass-loss and electrochemical measurements. Use of more than one experimental approach to estimating corrosion rates and mechanisms is appropriate because no one technique can provide the true answer (if such an answer even exists). But, deficiencies have been observed in how the results are used and related to each other. A common approach is to generate a polarization scan and utilize software to estimate Tafel constants and the polarization resistances by nonlinear regression. These values are used to estimate the corrosion current. The problem is that the values can be reported without further analysis. Though different software packages have different algorithms, they tend to be based more or less on the assumption that the corrosion rate is dominated by one anodic and by one cathodic reaction, each being expressed in terms of an activated process. Beyond the issues with using nonlinear regression for curve fitting as, for example, being stuck on a local minimum and not finding the true minimum in the sum of squares of errors, a number of assumptions about the corrosion process itself are inherent in the approach and sometimes ignored by the authors. Some of these requirements are the absence of uncompensated resistance, the rate-controlling processes being under activation control, corrosion potential and reversible potential being well separated, each site functioning as both anode and cathode, and additional electrochemical reactions being absent. An extensive discussion of the historical development, theory and assumptions, and pitfalls for the polarization resistance technique can be found in F. Mansfeld’s, “The Polarization Resistance Technique for Measuring Corrosion Currents,” in Advances in Corrosion Science and Engineering, eds. M.G. Fontana and R.W. Staehle, Vol. 6 (New York, NY: Plenum Press, 1976), p. 163. Granted assessing how well these assumptions are fulfilled requires some knowledge of the corrosion process, which may be difficult to ascertain. But, at the very least, the corrosion rates estimated from the electrochemistry should be compared to those estimated by mass loss. Both values should be reported in the submitted paper. Substantial disagreement between these two results should act as a “red flag” to the author that something is amiss and further investigation is required, possibly even before the results are constructed into a manuscript.
The last area concerns the number of significant digits that are reported in papers. I have seen papers that report, for example, corrosion rates and Tafel slopes to four decimal places implying four significant digits. Computers can provide almost any number of digits. The question is what number of digits is appropriate and relevant. Too often this question is ignored. The place to start is probably with the experimental measurement. For example, the calculated value can have no more significant digits than the component of the measurement that has the least number of significant digits, possibly fewer. When performing subsequent calculations, the results should be rounded off to the least number of significant digits in the calculation. Some internet sites that might be helpful:
These ideas are my thoughts only and in no way reflect the opinions of the other members of the NACE International Publications Administrative Committee or the Corrosion Editorial Board. Once again, my hope is that this letter stimulates a dialogue among researchers on these and other subjects pertaining to the handling of corrosion information so that the articles that appear in Corrosion best serve the corrosion community. Perhaps this space could become a forum for such a discussion.
EDITOR’S NOTE: This editorial by David Silverman is the first of a series. Readers are invited to contribute to the series by joining in this discussion or other issues of interest to the corrosion research community. (Comments may be submitted online at http://www.nace.org/CJ_PaperTracker or mailed to Dr. J.B. Lumsden, FNACE, Corrosion Technical Editor, Rockwell International, 1049 Camino Dos Rios, Thousand Oaks, CA 91360.) —Jesse B. Lumsden, Technical Editor
