THE ELECTROLYTIC THEORY O F CORROSION’ BY WILDER D. BANCROFT
About every so often in the development of a theory, it is necessary to stop and take account of stock. The original propounder of the theory may have been considering a,special case and may not have worded his theory in the best possible way, or a new set of facts may have been discovered which had not been foreseen. The supporters of the theory may have over-emphasized certain applications of the theoryuntil people mistake these for the theory itself. Unless people are on their guard, which they rarely are, the theory will change in passing through many hands until it is scarcely recognizable. This is the more likely to happen when, as is usually the case, people do not bother themselves with the historical development. Some or all of the things have happened to the electrolytic theory of corrosion and Messrs. Bengoiigh and Stuart2 concluded that corrosion may be either chemical or electrochemical. Since many of their conclusions are based on what seem to me misapprehensions of fact and misunderstandings of the electrolytic corrosion t h e ~ r y , ~ it has seemed worth while t o formulate the electrolytic theory of corrosion as I understand it, and to show how it should be applied in a number of special cases. This is the more necessary because it must be admitted that the supporters of the electrolytic theory of corrosion have too often played into the hands of their 0pponent.s and have not made the most of the material available. The first formulation of the theory of electrolytic corrosion was made by WhitneyS4 “Practically the only factor which limits the life of the iron is oxidation, under which are included all the chemical processes whereby the iron is corroded, eaten away, or rusted. I n undergoing this change, the iron always passes through or into a state of solution, and, as we have no evidence of iron going into aqueous solution except in the form of ions (probably electrically charged atoms), we have really t o consider the effects of conditions upon the potential difference between iron and its surroundings. The whole subject of corrosion of iron is therefore an electrochemical one, and the rate of corrosion is simply a function of electromotive force and resistance of circuit. If now we apply Nernst’s conception of the source of electromotive force between a metal and a solution, we must conclude from the measured potential difference that iron in contact with an aqueous solution tends to A report to the Corrosion Comrnittce of the National Research Council. The cost of some of the ex eriments on which this report is based was met by a grant from the National Research 8ouncil and b a grant from the Heckscher Foundation for the Advancement of Research, established %yAugust Heckscher at Cornell University. * “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 31 (1922). I ) Dunstan and Hill: J. Chem. SOC. 99, 1857 (191 I). J. Am. Chem. SOC. 25, 394 (1903).
786
WILDER D. BANCROFT
dissolve, ionize, or oxidize with a force expressed as equivalent, to about 10,000 atmospheres’ pressure at ordinary temperature. “In other words, iron should tend to dissolve in any aqueous solution1 until the concentration of the electrically charged iron ions reaches such a concentration that the osmotic pressure is equal to the above value. This means that the saturated iron solution must be at least 4.50-fold normal, which is a concentration not practically obtainable. Thus far the theory requires that iron should tend to oxidize in any aqueous solution. Whether it will do so or not depends on other conditions. Something may here be gained by a study of analogies. The dissolving zinc electrode of the Daniell or gravity battery, although possessing an enormous electrolytic solution pressure, does not dissolve when the electric circuit is broken, but begins oxidizing immediately when connected through any external resistance to t8he copper pole. It is not enough then for oxidation or solution of the metal, that it have a tendency t o dissolve; it must be in metallic connection with some other material capable of acting as an electrode, and this second electrode, if a positive element, must have a lower electrolytic solution pressure than the iron. Iron in contact with zinc and an aqueous solution will therefore not dissolve; but if copper replace the zinc, the iron will dissolve, the velocity of solution i n these cases being determined by the resistance of the complete electric circuit. These two cases are often met with in practice. In marine boilers, zinc plates are sometimes suspended from the boiler tubes in the water, that they may be attacked instead of the iron. On the other hand, scrap iron is used very commonly t o recover copper from solution in mine waters and other copper liquors in which case the iron rapidly dissolves. “Hydrogen acts as a metal and is electrolytically classified in the group with copper when compared with iron and zinc. That is, if a cell were made up on the Daniell model, iron being used instead of zinc, and hydrogen in place of copper, the cell would generate a current when the iron and hydrogen electrodes were connected. Iron would then dissolve with a velocity dependent on the total resistance of the circuit. So also, and for the same reason, iron when placed in a solution containing hydrogen ions will dissolve as the hydrogen precipitates, just exactly as when placed in the copper salt solution. That iron does oxidize or dissolve in all solutions containing appreciable quantities of hydrogen ions is well known. This electrochemical relationship between iron and hydrogen is the primal cause of rusting.” These paragraphs contain the substance of the theory of electrolytic corrosion and are just as true today as they were twenty-one years ago, except as to absolute values. In two respects, however, the wording is not as happy as it might have been. At the time that Whitney wrote, it was still believed by many that we knew the single-potential difference between a metal and a solution with a fair degree of accuracy, whereas nowadays we know that we do not. Consequently we cannot calculate the absolute value of the solution pressure; but it is immaterial for the argument whether this value is equivalent to 5000, 10000,or 20000 atmospheres pressure. The value is larger than 1
[In pure water the limiting solubility will be that of hydrous ferrous oxide. W. D. B.]
T H E ELECTROLYTIC THEORY O F CORROSION
787
that for hydrogen, which is all that is of importance. While it is true that iron should tend to dissolve in any aqueous solution until the concentration of the electrically-charged iron ions reaches such a concentration that the osmotic pressure is equal to the solution pressure, Whitney has neglected t o state that this will only be true so long as the iron is kept electrically neutral. If the metallic iron becomes charged negatively because of the formation of positively charged ferrous irons, there will be set up an electrostatic field which will prevent further formation of ferrous ions. This may happen even when the concentration of ferrous ions in the mass of the solution is practically negligible. This point was brought out clearly by Nernst’ and by Whitney2 in his translation of LeBlanc’s book. “In order to explain the production of a potential difference through the contact of a solid substance with a liquid, imagine a metal dipped into pure water, and that a certain amount of metal ions is produced owing t o the electrolytic solution pressure. The metal at the same time becomes negatively electrified, since both kinds of electricity must be simultaaeously produced whenever electrical energy comes into existence. The solution is thus positively electrified and the metal negatively, and there is formed a so-called double layer. The ions sent into the solution with positive charges and the negatively charged metal attract each other; in other words, a potential difference is produced. The solution constantly tends to send more ions into solution, while the electrostatic attraction of the electrical double-layer opposes this action, and eviden tly equilibrium is reached when the opposing tendencies are equal. Since the ions have very high charges of electricity, this condition of equilibrium occurs before weighable quantities of the ions have passed into the water.” Whitney undoubtedly did not discuss this point in detail because he assumed that it was familiar to everybody; but people nowadays are holding it up against the theory that measurable quantities of some metal are not always found in solution. The second point in which Whitney’s formulation has been confusing to some of his successors is that he did aoh lay sufficient stress on over-voltage. Since the solution pressure of iron is higher than that of hydrogen even in presence of practically pure water, iron should corrode in pure water as indeed it does. I n regard to this, Whitney3 says: “Iron dissolves in pure water qualitatively just as in a solution of copper sulphate, hydrogen being deposited in place of copper. The velocity with which this process proceeds will depend on the temperature and on the hydrogen-ion concentration in the water. When this concentration is so great that the potential difference exceeds a certain value, the hydrogen will be evolved as gas, separating from the liquid at the surface of the iron as bubbles. This potential value depends on the state of tEe surface so that i t is usually higher than the theoretical value for polarization by hydrogen when the gas is in equilibrium with the solution. a
Z. physik. Chem. 4, 152 (1889). LeBlanc: “A Text-Book of Electrochemistry,” 177 (1907). J. Am. Chem. SOC.25, 399 (1903).
788
WILDER D. BANCROFT
Below this value, the hydrogen, which is nevertheless being deposited upon the iron, but at a concentration below that corresponding to atmospheric pressure, slowly dissolves in the water, forming an ordinary solution of it, and escapes by diffusion.” There is a reference to over-voltage here; but it is not very clear-cut and it has been misunderstood repeatedly. Nowadays, we know that hydrogen over-voltage is due to polarization by electrically-neutral, monatomic hydrogenl and hydrogen will be given off as gas or be dissolved by the solution only as the infir.itely small amounts of monatomic hydrogen, presumably adsorbed on the surface, react to form molecular hydrogen. This is a very different thing from the hypothesis postulated by Bengough and Stuart2 and absolutely nullifies their criticism. “On the electrochemical view, the conclusion would be that a minute trace of hydrogen would adhere to the metal for long periods of time, entirely preventing corrosion in absence of oxygen. Yet when metal is attacked by very dilute acids (e.g. zinc in I :so00 acetic acid), the attack proceeds steadily in spite of the evolution of hydrogen, visible under the microscope, a t many points over the metal. It seems clear, therefore, that a hydrogen film is not a good, or even a moderate, protective against corrosion.” They have forgotten all about polarization and are merely consideriag the ohmic resistance of a film of gas. Whitney’ says that “there is no doubt that iron, even at ordinary temperatures reacts with pure water, in accord with this conception. The experiments on this point, carried out by Deville a t high temperatures only, showed a balanced condition at various temperatures for the reaction. 3 Fe +4 H20=Fe304+4H2. He found experimentally that water vapor in contact with iron must produce a certain concentration of hydrogen gas to be in equilibrium; in other words, if the hydrogen was continually removed, the iron could continually oxidize. Within the range of temperature employed (200’ to IOOO’C),he found that the pressure of the hydrogen produced by the action between iron and water increased as the pressure of the water vapor was increased. As the temperature rose, the hydrogen concentration (or pressure) at equilibrium diminished; in other words, the lower tEe temperature i n his experiments, the greater the tendency for oxidation of the iron. According t o these results we should expect water to act on iron to generate hydrogen even at ordinary temperatures, and it is a well-known fact that very finely divided iron such as is obtained by dry reduction of iron salts, reacts with pure water and generates hydrogen. “This fact, that pure water causes solution of iron, is in accord with other experimentally discovered facts. Mr. G. 0. Adams, in connection with a thesis presented to this institute in 1900, made analyses of various samples of ‘Ostwdd: Z. Elektrochem. 6, 40 (1889); Muller: Z. anorg. Chem. 26, 1 1 ( 1 9 0 1 ) ; Tafe1:Z. physik. Chem. 34, zoo ( I 00);50, 641, 713 (1905); Lewis and Jackson: Proc. Am. Acad. Arts Sci. 41,399; 8.physik. 8hem. 56,207; Brunner: 331 (1906); Bennett and Thompson: J. phys. Chem. 20, 296; Bancroft: 396 ( 1 9 1 6 ) ;Aten: Proc. Acad. Sci. Amsterdam, 18, 1379 (1916). 2
“Sixth Report t o the Corrosion Research Committee of the Institute of Metals.”
J. Inst. Metals, 28, 88 (1922). 3
J. Am. Chem. Soc? 25, 396 (1903).
’ I
T H E ELECTROLYTIC THEORY OF CORROSION
789
gases collected from radiators in different houses where the hot-water gravity heating system is used, and where different water supplies are concerned, and always found a large quantity of hydrogen gas. In most cases, on opening the air-cock of the radiators the gas could be lighted with a match. A number of such mixtures were analyzed. These varied in composition, but were mainly mixtures of nitrogen and hydrogen with usually no oxygen. The hydrogen content varied from 44 to 78 percent by volume. Measurable quant,ities of carbon dioxide or of hydrocarbons were not usually present .” The results of Mr. Adams have been confirmed in recent years by Mr. Speller of the National Tube Company. Bengough and Stuart1 believe that when zinc corrodes in pure distilled water, no evolution of hydrogen can be observed although considerable local corrosion (type B) takes place. ’“This is in sharp contrast with the behaviour of zinc in very dilute acid, where the evolution of hydrogen is quantitative, . . . The first observation, i.e. that no hydrogen is involved in the local corrosion of zinc, is of great importance in connection with theories of corrosion; and further observations have shown that no hydrogen appears when many other metals undergo corrosion in distilled water-in fact, it is only in the case of highly electro-positive metals (e.g. calcium) that hydrogen gas appears during local corrosion. Yet the electro-chemical theory assumes that the first action in such corrosion is the passage of metal ions into solution, and the displacement of hydrogen in the case of all metals. The assumption really is that all corrosion by water and salt solutions is of the acid type. “There is, however, no direct evidence for this assumption, but some against it. Lambert and Cullis2 found no trace of lead in pure water which had been in contact with pure lead in a vacuum for several months. On the electro-chemical theory, which states that corrosion is due t o electrical currents set up between portions of the metal of different potentials, Lambert explains the above result by assuming polarization of the areas of low potential by displaced hydrogen, which stopped corrosion so rapidly and completely that neither metal nor the hydrogen responsible for the stoppage could be detected. A similar explanation would doubtless be applied to the fact, observed by many investigators, that local corrosion (type B) even of ordinary commercial metals never takes place in absence of oxygen. Thus Friend detected very little change in Swedish iron kept for twelve years in pure water jn absence of air.”3 This is not helpful at all because we come out with the same analytical result whether we postulate that oxygen acts as a depolarizer or reacts direct with the metal. Consequently experiments of this sort are distinctly a waste of time. It is probable t,hat with Friend’s Swedish iron he had either an especially high over-voltage, or the formation of an adherent film of ferrous oxide. Whitney4 obtained quite different results with other samples of iron “Sixth Report t o the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 87 88 (1922). Lambert and 6ullis: J. Chem. SOC.107, 214 (1915). 8 Carnegie Scholarship Memoirs, 1922, 125. J. Am. Chem. Soc. 25, 398 (1903).
7 90
WILDER D. BANCROFT
in some experiments which he made for another purpose. “To learn whether carbonic acid was necessary to iron-rusting, a clean bottle was steamed out for a time to remove soluble alkali from the glass and was then filled with pure distilled water which was kept boiling by passing steam through it for fifteen minutes. While still boiling, a bright piece of iron was placed i n the bottle. A stopper (in some cases rubber and in others cork) carrying a tube open in a capillary several inches above the stopper, was inserted into the bottle and firmly fastened in place, the water being kept boiling. Finally, the glass capillary was heated hot by means of a blowpipe and sealed by squeezing the walls together. The bottle was then allowed to cool to a temperature of about 80’ C; and the neck of the bottle was finally covered with paraffin to prevent leaking. It was thought that in this way the oxygen, carbonic acid, and other gases in the water were completely removed. Bottles containing iron and sealed in this manner have stood without any visible change for weeks. I n some cases a little air was subsequently admitted to bottles which had stood in this way with the iron apparently unaffected, and within a few minutes the water became cloudy and assumed a yellow color. Ordinary rust rapidly deposited upon the glass and in spots upon the metal. I n fifteen or twenty minutes the production of rust throughout the bottle was perfectly evident. It seemed plain from the rapidity of formation of oxide and its precipitation on the glass, that the iron had dissolved in the water before the addition of the air, and that the latter simply permitted tho formation of the insoluble oxide. “Mr. J. A. Collins, in connection with his thesis of 1898, performed a similar experiment which shows that the iron is dissolved in the water and that its appearance as rust is a secondary phenomenon due t o the action of oxygen on the solution. A cleaned iron pipe 0.5 inch by 15 inches, sealed at one end and having a screw cap t o fit the other, was filled with boiling distilled water, and the boiling continued by heating the pipe until half the water had boiled away. While still rapidly distilling, the cap was screwed on tightly and the tube heated to about 125’C for a n hour. On cooling and removing the water from the pipe, it was found to be perfectly clear and colorless, but, OD exposure to air in a glass vessel, it rapidly precipitated rust. The pure water had dissolved iron in some form from the clean metallic surface, and this remained in solution until precipitated by the oxygen of the air. I n experiments with air-free water in contact with iron, in glass bottles and flasks, Mr. Collins let into the flasks containing only the pure water together with its vapor and the bright iron, air which had been freed from carbonic acid by being exposed to the action of a barium solution in a closed bottle for twelve hours, the bottle being repeatedly shaken to hasten the absorption. In case of this treated air, the production of rust in the flasks was evidently as rapid as with ordinary air. “A similar experiment with purified oxygen gave the same result. One is forced to conclude from such results that if the rusting is due in any way t o carbonic acid it is rapidly brought about by such a quantity of this gas as is left in air or oxygen after treatment with a barium hydroxide solution; in
THE ELECTROLYTIC THEORY OF CORROSION
791
other words, by an inappreciably small quantity. As this acid could owe its activity solely t o it.s hydrogen ions, because of the fact that the carbonate ion has no rusting or dissolving action on the iron, it is interesting to note that in the case of recently boiled water, the hydrogen-ion content due to carbonic acid may well be as low as the hydrogen-ion content due to the dissociation of pure water.” Dunstan, Jowett and Goulding’ have questioned Whitney’s statements of fact. “An attempt has been made by Whitney2 to apply the theory of electrolytic dissociation to the explanation of Ihe rusting of iron. Water being assumed to be slightly dissociated into hydrogen and hydroxyl ions, it should be capable of dissolving minute quantities of metallic iron, owing to the formation of an electric circuit containing iron as the positive and hydrogen as the negative element. If this is the case, the rusting process would be explained, the ferrous hydroxide fir,qt formed absorbing oxygen from the air. Substances such as alkaline salts interfere with, or altogether prevent, rusting by hindering the accumulation of hydrogen ions, whilst acids and certain salts tend to accelerate rusting by increasing the accumulation of hydrogen ions. When the concentration of the hydrogen has attained a certain maximum, the hydrogen, according to Whitney, is evolved as a gas. This is, however, contrary to fact, no hydrogen being liberated in ordinary circumstances during the rusting of iron. Careful experiment has also failed to confirm Whitney’s statement that iron dissolves to a slight extent in water, whilst the theory in question is shown to afford no explanatioa of the fact, established during the course of t,his investigation, that substances other than alkaline salts, m c h as chromic acid and potassium dichromate, prevent the rusting of iron.” “It has been assumed by Whitney (loc. cit.) that water, on the electrolytic hypothesis, being slightly dissociated, is capable of dissolving iron in the absence of oxygen or of carbon dioxide owing to the formation of a voltaic couple in which the iron acts as the positive element, whilst the negative element consists of the few hydrogen ions which the water normally contains. On this supposition he has founded a theory of rusting. He shows that theoretically those substances, such as acids, which permit of the concentration of hydrogen ions accelerate rusting, whilst those substances, such as alkalis, which diminish the concentration of the hydrogen ions inhibit rusting. The experimental evidence adduced by Whitney in support of the fundamental assumption that iron dissolves appreciably in pure water in the absence of oxygen or carbonic acid is slender and unsatisfactory. The theory also involves the assumption that hydrogen is liberated during the rusting of iron, but no evidence of the formation of free hydrogen is produced by Whitney, and we have never noticed its production in any of the experiments we have conducted. As regards the solubility of iron in water, Whitney describes an experiment in which a piece of bright iron was placed in a bottle of boiling water, and, while the water was still boiling, a stopper carrying a glass tube was firmly inserted in the neck of the bottle and the end of the glass tube J. Chem. SOC.87, 1551(1905).
* J. Am. Chem.
Son. 25, 394 (1903).
792
WILDER D. BANCROFT
sealed in the blowpipe; the stopper and the neck of the bottle were afterwards coated with paraffin wax. In experiments thus made, he states that the iron remained without change for weeks, but, on admitting air, rust was formed in a few minutes, the water becoming cloudy and assuming a yellow colour. I n 15 or 2 0 minutes, rust was produced throughout the bottle, being deposited on the glass as well as on the metal. From these results, Whitney concludes that the iron had dissolved in the water before the admission of air, and that the oxygen admitted reacted with the dissolved iron with the formation of rust. “We have repeated this experiment in the following manner. A flask of 600 cc capacity, filled with distilled water, was boiled for 15 minutes; two pieces of purified iron each about 1 - 1 / 2 inches square were then placed in the flask, and an india-rubber stopper carrying a glass tube which projected 7-8 inches above the stopper and ended in a capillary was fitted into the neck of the flask, the water being kept boiling continuously. The water was allowed to boil for five minutes longer, when the capillary was sealed and the stopper coated with paraffin wax. This flask was left a t the ordinary temperature for three weeks, in the course of which no visible change occurred. It was then opened, when one-half of the liquid was quickly poured into a beaker, the other half being left in contact with the iron in the flask. The liquid in the beaker on exposure to the air showed no cloudiness, no yellow coloration, and no separation of rust. In fact, on testing the liquid for iron by the extremely delicate thiocyanate reaction not a trace could be detected. The pieces of iron in the open flask after an hour showed signs of rusting, just, as in ordinary cases, but the phenomena described by Whitney were not observed. We are therefore unable to confirm Whitney’s statement that liquid water alone is capable of dissolving even an infinitesimal quantity of iron. This being the case, the theory based on this statement becomes untenable. “It having been proved that iron rusts in the presence of oxygen and water without t,he aid of carbon dioxide, it follows that the inhibitive action of alkalis on the process of rusting must find some other explanation than that hitherto accepted, which assumes that alkalis remove carbon dioxide, in the absence of which rusting cannot occur.” The whole difficulty is in the tacit assumption that iron is iron and that the sample used by Whitney behaved exactly like the sample used by Dunstan. If Whitney used a somewhat impure iron, as he undoubtedly did, the impurities might well have caused such a decrease in the over-voltage that a perceptible amount of corrosion took place and a perceptible amount of hydrogen was set free in the one case, whereas that did not happen with Dunstan’s samples of iron. It must be borne in mind that the theory only calls for the actual corrosion of iron and the setting free of hydrogen in so far as the overvoltage does not interfere. Under favorable conditions, the actual corrosion may be very small and may well be negligible in an experiment lasting only three weeks. As I shall take up later the question of the action of alkalies and bichromates in inhibiting rusting, we need not, worry for the present over Dunstan’s
T H E ELECTROLYTIC THEORY O F CORROSION
793
assumption that the electrol-vtic theory of corrosion cannot account for these phenomena. I n later paragraphs Whitnay‘ pointed out that carbonic acid may act as a carrier under certain conditions. “Free alkali added to a boiler of water not only greatly reduces the concentration of the hydrogen ions, from whatever source, thus diminishing the electrochemical reaction of interchange between hydrogen and iron and in the boiler itself, but also produces another valuable effect. If a steam or hot-water heating system is fed with a water which is not naturally alkaline, a part at least of the carbonic acid, which it always contains, will be driven from the water on boiling and pass t o the cooler portion of the system to be redissolved in condensing water. Thus the return pipes of the system will be subjected to the action of this acid or hydrogen-ion solution. No protecting scale of salts from the water being produced in these return pipes, such as is almost always produced in the boiler itself, the corrosive action will be most marked in the return pipes, and especially where the pipes are exposed to the action of continuous supplies of the water in motion. This will prevent the establishment of equilibria and the iron will be continually removed. It ought also to be removed according to the previous discussion, though more slowly, even if no carbonic acid or other acid were present, because of the hydrogen ions of the pure water. This reaction could again be reduced by the presence of volatile alkali in the condensed water, but in practice this latter is usually a negligible effect compared with the effect of volatile acid. “I wish now to show that the effect of the carbonic acid is actually a cyclic one, the same molecule of acid doing unlimited corrosive work, and that the very harmful corrosion of return pipes in many heating systems may be directly attributed t o this usually inconsiderable and unnoticed ingredient of the water. T o make this point clear, let us imagine a steam-heating system made up of a boiler, with steam pipes leading to various heating stacks and radiators from which return pipes bring the condensed sieam back to the boiler below the water-level. For simplicity, we assume that the plant is run without the addition of water after the boiler has been originally charged. I n other words, no steam is blown out into the air and the plant is not used, as some are, to incidentally supply hoh water for foreign uses, which thus requires a continual water feed. Our closed system usually contaim, when in actual running condition, a number of dead-ends where gases have accumulated and where the pipes are cold. This may be observed in many radiators of common type. Into this colder portion of the system, the gases such as oxygen, nitrogen, and carbonic acid, which were originally in the feed water, will collect. Here they will dissolve ip the condensed water which is to return t o the boiler, the carbonic acid being especially soluble. The carbonic acid or its active hydrogen will cause the eolution of iron from the return pipes and this iron will be carried back towards the boiler as bicarbonate of iron, being held in solution just as is
* J. Am. Chem. SOC.25, 401 (1903).
794
WILDER D. BANCROFT
calcium carbonate in water containing carbonic acid. This may be represented by the reaction Fe+zC02+2Hz0 =HzFe(C0~)2+H2. “In the case of the bicarbonate of calcium solution, it is well known that boiling decomposes the salt and liberates half of the carbonic acid, leaving a precipitate of calcium carbonate. I n case of the soluble bicarbonate of iron, as will be shown later, the decomposition of the compound by heat, liberates all of the carbonic acid instead of half of it and produces the insoluble oxide or hydroxide of iron. Whether this decomposition of the soluble iron salt takes place in the boiler after the solution has returned to it, or in the return pipe where the proximity to the boiler produces a sufficiently high temperature, is of no immediate importance. In either case, the carbonic acid is all set free and must immediately return with steam to the cooler parts of the system; there again it will dissolve in the condensate, again render soluble some iron and carry it towards the boiler and so forth. In each cycle of this kind hydrogen will be set free which will remain in the cooler parts of the system, as it is but slightly soluble in water. It seems necessary, therefore, that in common practice a very small quantit’v of carbonic acid must often cause an unlimited amount of corrosion, without in any way losing its power to continue the process. The process of corrosion of the iron in this case will amount in loto to the union of iron with theoxygen of water and liberationof hydrogen, the carbonic acid acting merely as a catalyzer, where the mechanics of its action is apparent. This peculiar condition of affairs has been observed by us in certain large heating systems where we have found, first the carbonic acid of the feed water; secondly, much carbonic acid mixed with hydrogen, nitrogen, and oxygen in deadends or cold parts of the system; thirdly, water in the return pipes, where veiy rapid corrosion of the piping was taking place, in which much dissolved bicarbonate of iron was found; and finally, much precipitated oxide FIG. 1 of iron in the boiler and hottest partsof thereturn pipes. “It has been possible also to reproduce these phenomena in the laboratory in various ways and in some cases in glass apparatus where the complete cycle becomes practically visible. I n connection with his thesis, Mr. C. L. Wright arranged an apparatus of which a sketch is shown. Pure distilled water was boiled in the flask A, and various quantities of air and carbonic acid mixtures were enclosed in the system which was kept nearly at atmospheric pressure by the liquid seal made by the bottles B connected with the condenser C. The steam condensed and took up a little of the carbonic acid; this solution came into contact with a piece of cleaned pipe, P, from which the outer layer had been removed in the lathe. This iron was supported in an extractor, E,
---
THE ELECTROLYTIC THEORY O F CORROSION
79s
between the condenser and the boiler in such a way that it was alternately covered with the water and uncovered by the intermittent siphon action of the extractor. “In the first few experiments a precipitate was soon formed in the boilei which was not analyzed, but which was evidently a mixture of hydroxide of iron and silica from the glass. Subsequent experiments in the same apparatus differed in result from the first one and showed that, in the absence of free oxygen, the boiler water simply became black and opaque but showed little or no precipitate. The glass tube leading from the condenser to the boiler was also quickly coated with a black deposit of iron oxide. It was evident that iron was being dissolved and the resulting compound decomposed in the boiler, and in the hot return tube, but it was at first thought peculiar that the precipitate which appears when the glass or flask is a new one, was not produced in subsequent experiments, with the same apparatus. This led to attributing the actual formation of a precipitate i r the first case to the presence of dissolved glass. The dark color could only be due to the iron. The clear, black, filterable solution obtained from such previously-used apparatus was then shown t o be a colloidal solution of iron oxide by the following method : Small quantities of such salts as sodium and barium chloride were added to portions of the clear, black water, and they caused an immediate coagulation and a consequent heavy precipitate, coupled with complete decolorization of the solution. This precipitate, well washed by decantation with pure water, was treated with dilute sulphuric acid in a closed vessel and pure air passed through this and then through a vessel containing a barium hydroxide solution. The failure to produce a precipitate of barium carbonate i n this barium hydroxide solution showed that the original compound did not contain a carbonate. The precipitate was shown to contain iron by dissolving it in acid, oxidizing and precipitating the ferric hydroxide by ammonia. The formation of this colloidal solution of ferrous oxide is in exact accord with the principles which determine hhe formation of colloidal solutions in general. It is a general principle that whenever aoy substance which is by nature insoluble, is formed in water, it will tend to remain in a colloidal or suspended state until coagulated by electrolytes. “In the production of most precipitates in common laboratory reactions, there are always sufficient electrolytes present to account for the coagulation of the insoluble substances, if we may judge by the concentration usually necessary where measurements have been made. Where this is not the case, a colloidal state usually results. I n the case at hand, there are practically no electrolytes present when the soluble ferrous bicarbonate is decomposed by heat, as this process requires the presence of but exceedingly small quantities of soluble salts in the solution at any one time. The insoluble ferrous oxide consequently remains in the colloidal state. This colloid may be precipitated by salts dissolved from the glass vessel, if of sufficient concentration, as was rhe case with new glass apparatus, and in this respect this colloid is like many others, such as platinum and silica.
.
796
WILDER D. BANCROFT
“The ease with which the soluble iron salt is decomposed by heat was well shown in the glass tube connecting the condenser and boiler, the return pipe of the above experiment. Above the stopper of the flask, where this tube was fairly cool, the glass remained clear, but below the stopper where the tube was heated by the steam of the flask, it was covered deeply with a black deposit, probably ferrous oxide. The deposition of this substance at this part of the return tube, commenced almost immediately on starting the experiment. ‘(An experiment carried out in this way where pure water and carbon dioxide were used, where analysis showed the gaseous mixture t o contain eleven percent carbon dioxide, produced such rapid corrosion of the iron that within a few days nearly a third of the exposed surface had been eaten away t o depths of several hundredths of an inch, a t which rate an ordinary pipe would not last more than a few months. It is not surprising that carbonic acid should dissolve iron under these conditions, but the fact that this corrosive action is a cyclic one, in which under suitable circumstances even a trace of carbonic acid may dissolve an unlimited quantitiy of iron without losing its corrosive power, has not received sufficient attention.” Whitneyl was quite clear as t o the effect of scale, though it is quite possible that he did not emphasize this factor as much as seems desirable twenty years later. The polarization by hydrogen “should greatly limit the velocity of solution of the iron, even if there is no other complicating effect due to the production of an insoluble compound, an oxide or hydroxide, in case air be present. This is usually the case in practice. The production of a compact adherent coating of oxide on the surface of the iron generally retards the corrosive action. Especially is this true at temperatures of steam where the magnetic oxide is formed, This oxide always forms as an adherent solid coating on the iron and seems to be interrupted only by cracks caused by its unequal coefficient of expansion compared with the iron. The red oxide or rust is always flocculent and spongy and, besides not protecting the iron, actually seems to increase the velocity of corrosion in its vicinity. It is common to attribute to the red oxide or rust a catalyzing effect on the corrosion of iron.” That Whitney’s attitude in regard to films was inadequate is shown i~ the next paragraph.2 “If the primary rate of corrosion of iron, independent of subsequent formation of insoluble substances, is simply dependent upon the concentration of the hydrogen ions of the water, anything which reduced this concentration should also reduce the corrosion. . , . This reduction of the hydrogen ion concentration may be brought about by the addition of any alkali t o the water. That the corrosion is thereby diminished is a well-known fact and one that already receives many practical applications. Iron and steel tools in process of manufacture, between the roughly ground state and the final polished condition, are often kept under water saturated with lime. This prevents the rusting which would quickly take place if they were left in moist air. The effect of the lime must be attributed solely to the hydroxyl ions which in turn reduce the concentration of the hydrogen ions of the water. J. Am. Chem. SOC.25, 400 (1903). Whitney: J. Am. Chem. SOC.25, 400 (1903).
T H E ELECTROLYTIC THEORY OF CORROSION
797
Neutral salts of calcium do not exert this effect and hence it can not be attributed t o the calcium compound nor to the calcium ion. Other soluble alkalies do it equally well and these contain, in common, only the hydroxyl ion.” It would be a bold man nowadays who would question the formation of a surface film on iron in alkaline solutions and consequently the effect of the lime cannot be due solely to the reduction of the concentration of the hydrogen ions of the water in the sense that Whitney means. Whitney was right as far as he went and it is quite possible that he went as far as was wise twenty years ago. As Whitney was discussing the corrosion of iron, it was not necessary for him to consider the corrosion of sodium by water, especially as this case had been discussed explicitly by Nernst.l “Let zinc be brought in contact with acids, or sodium with water: then the electrostatic charge is obviously great enough to drive the positive hydrogen ions out of the solution and into the metal in which they dissolve; the hydrogen is able to separate from the metal in an electrically neutral form as soon as its concentration in the metal shalI have reached a sufficient amount, Le. as soon as its vapor pressure shall amount to the pressure of one at,mosphere.” There is nothing in this paragraph or in Whitney’s paper to the effect that a pure metal cannot corrode and yet Bengough and Stuart2 say that the characteristic conception of the electrolytic theory of corrosion “is that of a non-uniform distribution of anodic and cathodic areas over the surface of the metal, such areas being caused by local differences in solution pressure . . . . An ideally pure and homogeneous metal is assumed to be incorrodible.” I think that every chemist would expect a pure and theoretically uniform sheet of sodium t o be attacked by water, though he could not, on that statement of facts, predict at what point the first bubble of hydrogen would appear. Similarly he would admit that copper would be corroded by concentrated nitric acid containing nitrous acid regardless of how pure or how uni- ’ // form the metal was. V Fifteen years ago Walkers said that “every metal, when placed in water or under such conditions that a film of water may condense upon it, tends to dissolve in the water, or, in other words, to pass from its atomic or metallic condition into its ionic condition. This escaping tendency of the metals varies from that shown by sodium or potassium, which is so great as to cause instant and rapid decomposition of the metal and water, to gold or platinum where such tendency is zero. Between these two extremes we find the other common metals, including thereunder the element hydrogen, which may be considered as a metal.” It is quite true that Messrs. Bengough and Stuart could justify their statement by quotations; but their attitude should not be that of special pleaders trying t o make a point. “Theoretical Chemistry,” 612 (1895). “Sixth Report t o the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 54 (1922). 3 J. Iron Steel Inst. 1, 70 (1909). 1
a
798
WILDER D. BANCROFT
In 1906 I pointed out1 that “the most striking characteristic of an electrolytic reaction is that it occurs in two places- at the anode and the cathode. This peculiarity can be made less marked by bringing the electrodes nearer and nearer together. When the distance between them vanishes we have a chemical reaction in the ordinary sense of the word and not an electrochemical reaction. Any chemical reaction therefore which can be made to take place electrolytically must consist of an anode and a cathode process.2 Considering the matter in this light we see that in the chemical reaction there is a possibility of the anode and cathode processes interfering, and of one perhaps masking the other. “In some cases it is easy enough to tell what the anode and the cathode processes are. If we dissolve zinc in sulphuric acid, the formation of zinc sulphate is the anode process and the evolution of hydrogen is the cathode process. Now we know that pure zinc does not dissolve readily in sulphuric acid. Consequently we should expect to find a difficulty of some sort if we electrolyze sulphuric acid between zinc electrodes. We find this in the form of the so-called ‘excess voltage’ a t the cathode; and in the electrolytic process we can obtain a more or less quantitative measurement of the phenomenon though we are still far from knowing the cause of it. A less simple case is that of copper in dilute nitric acid. Copper reacts chemically with dilute nitric acid, setting free nitric oxide. The formation of copper nitrate must be the anode process and the reduction of the nitric acid the cathode process. When we start to test this we find difficulties. Everybody knows that we get ammonia instead of nitric oxide if we electrolyze dilute nitric acid, using a copper cathode. We have here an apparent contradiction, the chemical reduction yielding nitric oxide and the electrochemical one ammonia. Mr. Turrentine was good enough t o solve the mystery for me. When copper reacts chemically with nitric acid, the anode product, copper nitrate, is formed at the same spot that the reduction takes place. In the electrolytic reduction of nitric acid with a copper cathode, the reduction takes place in a solution practically free from copper salt. The conditions are therefore not the same in the two cases. Mr. Turrentine therefore electrolyzed a solution of nitric acid and copper nitrate using a copper cathode. A gas was evolved a t the cathode which proved on analysis to be chiefly nitric oxide. This experiment can be done in another form which is more striking. If dilute nitric acid be electrolyzed between copper electrodes, there will at first be no evolution of gas at the cathode. Gas will begin to appear as soon as the blue solution formed at the anode comes in contact with the cathode. A corollary t o this is that ammonia would be formed in the chemical reaction between copper and nitric acid if the concentration of the copper salt could be kept sufficiently low. There did not seem to be any salt which one could add to the solution without introducing more complications than were eliminated. The difficulty was overcome by Mr. Turrentine in a distinctly ingenious manner. Strips of copper were hung vertically in a tall vessel. The copper nitrate 1 2
Bancroft: Trans. Am. Electrochem. SOC. 9, 13 (1906). Cf. Traube: Ber. 26, 1473 (1893); Haber: 8. physik. Chem. 34, 514 (1900).
T H E ELECTROLYTIC THEORY O F CORROSION
799
flowed to the bottom of the vessel and the copper was removed by electrolytic precipitation in the form of cupric hydroxide. No current flowed through the copper strips and there was no copper cathode; but ammonia was formed. “These experiments were performed to prove that the difference between the electrochemical and the chemical corrosion of copper by nitric acid was an apparent one only and due to an unsuspected difference in the conditions. I n addition they illustrate the superior flexibility of the electrochemical method over the chemical method. I n the electrochemical method there is no difficulty in varying the concentration of the copper salt at the cathode between any desired limits, while this is very difficult to do in the case of the chemical method. This is in addition to the advantage, which the electrochemical method always has, of permitting a wide variation in the rate of reaction for constant temperature and constant concentration. If we are ever to have a thorough knowledge of the chemical reactions between nitric acid and the metals we must study the problem electrochemicwlly. “When metals are acted on slowly by oxygen in presence of moisture, it is known that half the oxygen reacts with water to form hydrogen peroxide, this hydrogen peroxide then often reacting with the metal. By shaking a zinc amalgam with a solution containing sodium and calcium hydroxides, Traubel was able t o isolate the insoluble calcium peroxide. So far as we now know hydrogen peroxide is formed at the anode only under special conditions, such as electrolysis of a concentrated sulphuric acid with a high anode density. Even under these conditions it is by no means certain that hydrogen peroxide is not a secondary product resulting from the decomposition of persulphuric acid. On the other hand hydrogen peroxide is the first reduction product a t an oxygen cathode. In the reaction studied by Traube the oxidation of the zinc is the anode process and the formation of hydrogen peroxide is the cathode process. A consequence of this is that if we electrolyze a caustic soda solution between zinc electrodes and bubble in air round the cathode, we ought to get a corrosion at the cathode due to the secondary reaction of the hydrogea peroxide with the zinc cathode. It has been found that the cathode does corrode under these circumstances; but it is a little difficult to tell whether this corrosion is due to hydrogen peroxide or to the oxygen of the a+. With a n iron cathode the corrosion has never been anything like as great as was obtained by simply bubbling air against the plate when no current waspassing. This whole question calls for much more study than has yet been given to it. “The slow oxidation of metals in contact with solutions means the corrosion of these metals by these solutions, a very important problem. If we can substitute electrochemical methods for chemical ones, it means an enormous saving in time and a corresponding increase in the number and in the quality of the data which we can accumulate. A single experiment on chemical corrosion may easily last six weeks. This means great difficulty in keeping conditions constant. By the time that experiment has been repeated several times with the variations which each repetition suggests, six months or a year have passed. On the other hand an experiment in electrolytic corrosion can Ber. 26, 1471 (1893).
WILDER D. BANCROFT
800
be run through in a few hours, seven or eight at the outside. The advantage is obvious provided the results are the same in the two cases. We can see in a moment that the two phenomena must usually be the same. Suppose we have a copper plate’ in a sulphate solution which contains dissolved oxygen and suppose that the concentration of oxygen is not the same over the entire surface of the plate. We shall then have an oxygen concentration cell with copper electrodes and a current will tend to flow through the solution from the place of lower oxygen concentration to the place of higher oxygen concentration. Since the dissolved oxygen will never be absolutely uniform in Concentration, its effect will always be to start a miniature electrolytic cell. The only real difference between the chemical corrosion and the electrolytic corrosion will be in the magnitude of the current. A special case occurs when the current makes a metal anode passive. In that case the metal does not corrode in that solution. Iron becomes practically passive when made anode in caustic soda solution or in concentrated nitric acid, and iron does not rust to any appreciable extent in these solutions. An iron anode is not attacked in a carbonate solution and is attacked in a bicarbonate solution. Iron does not rust readily in the first solution and does in the second. Iron dissolves quantitatively as anode in sulphate or chloride solutions and rusts with surprising rapidity in these solutions. Nickel becomes passive in sulphate solutions and does not rust. Nickel does not become passive in sodium chloride solutions and it corrodes in these solutions, though not rapidly. In general we may say that any addition which makes a metal passive i n any solution will prevent the metal from corroding in that solution. “You have been told today that electrolytic corrosion and chemical corrosion are two fundamentally different things? I should prefer to word that somewhat differently and I should say that electrolytic and chemical corrosion are fundamentally one and the same. Any apparent differences will be found to be due to special differences in the conditions, as in the case of copper and nitric acid to wEich I have already referred. It may also happen that an insoluble salt will be formed as an adherent crust in one case and as a non-adherent powder in another. This will of course have a material effect on the rate of corrosion whether electrolytic or chemical; but such cases are perfectly simple if examined carefully. One apparent discrepancy has occurred in our own work. When pure manganese is made anode in caustic soda solution, permanganate is formed. A caustic soda solution reacts with metallic manganese chemically, forming a hydroxide of manganese and setting free hydrogen. Mr. White soon found that a manganese anode forms permanganate in caustic soda solution only when the anode current density exceeded a certain limiting value. With lower current densities manganous hydroxide or manganese dioxide is formed. The apparent discrepancy is therefore an imaginary one due to the artificial difference in the rate of corrosion. “We can now pass to the question of the corrosion of alloys, under which heading the corrosion of iron could have been taken up. Very little carefuI ’ Cf. Haber: Z. Elektrochem. 12, 32 (1906). 2
Toch: Trans. Am. Electrochem. SOC.9, (1906).
T H E ELECTROLYTIC THEORY O F CORROSION
80 I
work has been done on the electrolytic corrosion of alloys and the accepted theory of the phenomena is not at all in accord with the facts. In one of the latest text-books on electrochemistry we read1 ‘that obviously the potential of a mixture of two metals determines its behavior as anode.’ This means that the less noble metal or the less noble phase will dissolve first. In the case of copper-tin alloys annealed just above 2 0 0 ° , we have five possible solid phases. From IOO to 87 percent copper we have a series of solid solutions known as the a-crystals. From 74.5 to 67 percent we have the homogeneous &crystals, also a series of solid solutions. These are the crystals which Heycock and Neville believed to be the compound CurSn. At 61.3 percent copper we have the compound Cu3Sn, the only compound in the series. From 41 to 40 percent we have a new series of solid solutions, the €-crystals, formerly supposed to be CuSn. Lastly, we have pure tin. The experiments of Herschkowitsch2 and unpublished work by Shepherd show that the €-crystals have a potential differing from that of pure tin by only a few millivolts, while the Cu3Sn, 6, and a-crystals differ in potential from pure copper by an even smaller amount. The order of solubility should therefore be tin, E, CuaSn, 6 and u, the last being the least readily corroded. As a matter of fact, in most solutions tin and the a-crystals are the most readily attacked, while the e-crystals are the least readily ~ o r r o d e d . ~ “The cause for this discrepancy between theory and experiment is that we are reasoning from static to dynamic experiments, from a stationary state to a changing one. An electromotive force is measured on open circuit with no current flowing. Elect,rolytic corrosion takes place on closed circuit with a current flowing. “Reasoning from electromotive force measurements to current phenomena is permissible only when no surface change takes place and when equilibrium conditions are satisfied. We know now that neither of these conditions is satisfied with the bronzes, and there is no reason for supposing that these conditions will be fulfilled except in special cases. A number of 1 he bronzes become passive owing to the formation of a surface film of stannic oxide. In both the copper-rich bronzes and the copper-rich brasses, the reaction between the alloy and a copper sulphate solution is so very slow that no reversible equilibrium is reached during corrosion. “While we cannot predict the actual way in which an alloy will corrode, from electromotive force measurements of the phases, we can predict that any change iq the current efficiency with a given alloy as anode will coincide more or less closely with the appearance or disappearance of a phase. Some experiments on the behavior of bronzes by Mr. Curry bring this point out very clearly. In sodium sulphate solutions the only phases to dissolve are pure tin and the copper-rich crystals known as the a-crystals. The current efficiency decreases as the percentage of these two sets of crystals decrease and is practically zero for the 6, Cu3Sn, and e-crystals. In sodium acetate solutions we Foerster: “Elektrochemie wasseriger Losungen,” 208. hysik. Chem. 27, 123 (1898). a [ T l e argument does not depend at all on the assumptions made as to the nature of the solid phases. W. D. B.]
* Z.
WILDER D. BANCROFT
802
have a similar behavior, but the a-crystals are much less readily corroded in the sulphate solutions and the current efficiency becomes practically zero as soon as any &crystals are present. I n sodium nitrate solutions the 8-crystals corrode until the copper content has fallen to about one-half the maximum. There is practically no corrosion with the tin-rich 6 crystals, with Cu,Sn, and with the e-crystals, while alloys containing free tin corrode readily. I n acidified ammonium oxalate solution, a and &crystals dissolve with one hundred percent current efficiency, while CuaSn is less readily attacked and the current efficiencydrops t o about twenty-five percent. Still other relations are found in alkaline sodium tartrate solutions, while with sodium chloride solutions there is no tendency for the alloys to become passive and the current efficiency is approximately one hundred percent over the whole range from pure copper to pure tin. “We thus see that the electrolytic corrosion varies with the nature of the solution and that the changes in the current efficiency stand in a clearlymarked relation to the changes in the nature of the crystals present in the solid alloy. Special experiments on the chemical corrosion of the bronzes in persulphate and in chloride solutions show that the results are identical with those obtained for electrolytic corrosion i n sulphate and in chloride solutions respectively. Since the identity of the chemical and the electrolytic corrosions has been proved for these two typical solutions, it is fair to assume that a corresponding identity exists in the case of other solutions. The simple relations which have been obtained in a relatively short time by a study of the electrolytic phenomena would have taken an immense amount of time if we had attempted to obtain them from a study of chemical corrosion alone. I n fact it is doubtful whether any satisfactory result would have been obtained without the electrolytic corrosion. This method i s obviously applicable to all other alloys and should lead to important results, especially in the case of the steels.” Cushman’ seems to have been the first t o have emphasized the part that two phases may play in promoting electrolytic corrosion. He suggested that unequal distribution of manganese sulphide in fence wire might be one of the reasons why modern fence wire corrcdes rapidly. This point of view was developed more in detail2 as time went on. “In considering the corrosion of iron it is important to remember that iron is a metal which readily combines with or dissolves nearly all the other elements. With possibly one or two exceptions, there are no elements that do not either dissolve in or combine readily with iron. It is also unique in the fact that very small quantities of impurities suffice to entirely change its physical characteristics. On account of this fact metallurgists scrutinize the hundredths of a percent of some of the principal impurities that are generally associated with this metal. This is particularly true, for instance, of the element phosphorus. It is so important to modern metallurgy that the amount of phosphorus should be controlled in certain forms of steel that an animated discussion is going on at the present 1 2
Department of Agriculture. Farmers’ Bulletin No. 239 (1905). Cushman: Trans, Am. Electrochem. SOC. 12, 403 (1907).
THE ELECTROLYTIC THEORY OF CORROSION
803
time between certain interests as to the control of the amount of phosphorus that steel shall carry, and the question at issue amounts to no more than a few hundredths of one percent. “Manganese is also an element which is nearly always associated in modern metallurgy with iron and steel. Manganese decreases the electrical conductivity of iron, and as the percentage of manganese, starting from zero, rises, the electrical resistance increases up t o a certain specific maximum. You will see that if the presence of manganese in iron raises the electrical resistance, any change in t h e distribution of the manganese means that there will not be a constaqt electrical conductivity throughout its mass, or on any given surface. One who is familiar with the methods of modern metallurgy,.that the manganese is added for certain specific purposes, not as a rule quantitatively, but in accordance with the views of the iron master who has control of the mill or furnace. Moreover, the manganese is usually added by throwing lumps of ferro-manganese into the molten metal, either in the furnace itself, or in the ladle into which it has been poured. Chemists know the extreme care that has to be taken in order to get uniform mixtures of substances in the course of chemical operations. In the large scale of metallurgical processes, even if it were possible to take great care in the mixing, it still happens that when iron is cooled from che molten state, segregation takes place; that is to say, the impurities, although they may have been thoroughly mixed in the molten mass, do not remain homogeneously distributed after the metal is cooled. “For these reasons we must remember that in studying iron and steel from the standpoint of their stability, under the conditions of service, we are not dealing with homogeneous pure metal. It is not difficult for an electrochemist to believe that when such material is immersed in, or even brought into contact with, an electrolyte, electrolysis will take place upon the surface, and thereby induce rapid corrosion. It is probable that the corrosion of all metals is more 01: less due to electrochemical action. Before metals can be attacked at ordinary temperatures in the presence of water they must first pass into soluLion, and ip passing into solution become ionized. This is true of copper, zinc, lead and the other metals that suffer corrosion. “In accordance with the conceptions of Nernst and the modern theory of solutions, all metals have a certain solution pressure which will operate until counterbalanced by the osmotic pressure. Iron, however, appears to differ from the other metals in one important respect. The corrosion of iron does not take place evenly and uniformly over the surface. On the contrary, it is a matter of common observation that iron corrodes rapidly a t certain weak points, the effect produced being known as pitting. That this effect produced by local electrolysis would hardly be doubted, even if it were not possible to demonstrate it by the use of the ferroxyl indicator. “Early in this investigation the writer observed that whenever a specimen of iron or steel is immersed in water or a dilute neutral solution of an electrolyte t o which a few drops of phenolphthalein indicator has been added, a pink color is developed. If the solution is allowed to stand perfectly quiet, it will be ’
t
“
I
804
WILDER D. BANCROFT
noticed that the pink color is confined to certain spots or nodes on the surface. The pink color of the indicator is a proof of the presence of hydroxyl ions and thus indicates the negative poles. “Since phenolphthalein shows only the nodes where solution of iron and subsequent oxidation can not take place, Dr. W. H. Walker suggested the addition of a trace of potassium ferricyanide to the reacting solution, in order to furnish an indicator for the ferrous ions whose appearance marks the positive poles. If iron goes into solution, ferrous ions must appear, which, with ferricyanide, form the well-known Turnbull’s blue compound. Going a step fart”her,Walker suggested stiffening the reagent with gelatin or agar-agar, so as to prevent diffusion and preserve the effects produced. For this combined reagent, which indicates a t one and the same time the appearance of hydroxyl and ferrous ions at opposite poles, the writer has suggested for the sake of brevity the name “ferroxyl.” If the reagent has been properly prepared the color effects are strong and beautiful. In the course of a few days the maximum degree of beauty in the colors is obtained, after which a gradual deterjoration sets in. “In the pink zones, as would naturally be expected, the iron remains quite bright as long as the pink color persists. I n the blue zones the iron passes into solution and continually oxidizes, with a resulting formation of rust. Even the purest iron develops the nodes in the ferroxgl indicator, but impure and badly segregated metal develops the colors with greater rapidity and with bolder outlines. This result would of course be expected, as i n pure iron the formation of poles would be conditioned by a much more delicate equilibrium than in impure iroq, where changes in concentration of the dissolved impurities would stimulate the electrolytic effects. “In the writer’s opinion these effects which are produced in the ferroxyl indicator constitute a visible demonstration of electrolytic action taking place on the surface of iron, and causing rapid corrosion at the positive nodes.” One difficulty with the ferroxyl indicator is that it is too sensitive. Walker1 points out that “if a piece of chemically pure iron free from mechanical strains and without evident crystallization, be immersed in the ferroxyl indicator, the positive and negative zones will be apparent after a few moments. There appears to be an unequal concentration of oxygen or segregation of oxygen upon the surface, which cannot be explained by discernible differences in the character of the surface. If the ferroxyl indicator be removed, the surface cleaned by rubbing with a dry towel, and the indicator again applied, the same separation into zones is seen, though in an entirely different configuration. The indicator is apparently a very delicate one and susceptible to changes in equilibrium, which up t o the present have not been detected by other means.” Walker goes on to say that “since corrosion is manifestly an electrochemical action, it seemed probable that if two specimens of iron were selected, one of which had proven itself in practice as especially resistant to corrosion, and another which had shown itself to be very susceptible to corrosion, certain l
J. Am. Chem. SOC. 29,
1262 (1907).
T H E ELECTROLYTIC THEORY O F CORROSION
80 5
differences in the electrochemical behavior should be discernible. TWOsets of such specimens were obtained. The first consisted of two pieces of sheet nietal which had been used as culverts in road building, one of which had given way in but a short time while the other was practically intact. The second set consisted of two strands of wire from a piece of ordinary barbed wire fencing, one of which was badly corroded upon an exposure of but six months while its neighbor was apparently in its original condition. Nothing is known of the history of the two wires; the good piece of sheet was known to be from an open hearth steel ingot containing but a trace of manganese, and in the heating and rolling of which extra precautions had been taken to prevent segregation. The poor piece was known to be from an ingot of ordinary Bessemer steel. “If corrosion be an electrochemical phenomenon depending upon the formation of an anode portion and a cathode portion and the passage of a currcn t between these two, the rate of this corrosion may be assumed t o be proportional to the difference of potential between the two surfaces.” The results showed that “areas having marked differences in potential exist in far greater number upon the surface of a piece of iron prone to corrosion than upon iron which is resistant to corrosion.” Walker was careful t o add, however, that ‘lit is as yet too early t o decide that measurements of this kind indicate the tendency of iron to corrode; but we hope to obtain definite informahion on this point by investigating a large number of specimens of known resistance t o corrosion.” Since differences of homogeneity will tend to cause local voltaic cells and will therefore tend to cause corrosion, the natural corollary is that, if other things are equal, the most homogeneous metal will corrode the least rapidly. This is absolutely sound; but then people assumed, perhaps unconsciously, that other things were equal and that therefore the most homogeneous metal would corrode the least rapidly. This unwarranted assumption has done a great deal t o retard the effective study of corrosion, because attention has been concentrated unduly on the question of homogeneity to the comparative exclusion of more important factors; and many people have believed that the electrolytic corrosion theory was justifiable only in case homogeneity was the most important factor in preventing corrosion. In 1917 E. A. Richardson and L. T. Richardson’ report on experimevts on the relative corrosion of cast iron, steel, and pure iron. They conclude that “the results of this test agree with other tests and add to the evidence already accumulated, tbat the resistance of an iroti to corrosion does not necessarily depend upon its purity or homogeneity, as would be indicated by the electralytic theory of corrosion. The theory, in its present accepted form, does not explain the great resistance offered by cast iron to corrosion. While it may be true that the initial rusting is largely electrolytic in character, other factors, such as adherence of the rust and the protection thereby given to the metal, come into operation and outweigh any electrolytic corrosion-a conclusion that has also been arrived at by other observers.” 1
Trans. Am. Electrochem, SOC.31, 195 (1917).
806
WILDER D. BANCROFT
There is nothing about the electrolytic theory of corrosion which precludes the formation electrolytically of insoluble and protecting coatings. The authors have substituted for the theory of electrolytic corrosion the narrow views of some of its protagonists. As was pointed out by Speller‘ in the discussion of their paper, their results do not in any way “throw discredit upon the electrolytic theory. Copper-steel has been mentioned, and here the longer life seems to be due to the formation of an oxide which adheres very tenaciously t o the surface of the iron. Copper-steel and ordinary steel in the atmosphere act very much the same for the first few weeks, after which the corrosion of the copper-carrying steel is retarded very materially, and in some cases actually seems t o stop.” We cannot criticize the Messrs. Richardson very severely for the attitude they took, wrong though it is, because even Walker,2 four years later, seems to consider the effect due t o homogeneity as the most important feature of the electrolytic theory of corrosion. “The mechanism of corrosion is now fairly well understood. Whitney in 1900 showed it to be an electrochemical phenomenon, obeying the laws of electrochemistry. It was the privilege of the writer to demonstrate in 1907 the function of oxygen in corrosion, and t o apply the experimental facts to the preservation of many engineering structures. The engineering public, however, has been slow in availing itself of the published work on corrosion probably for a number of reasons. First, the entire literature of corrosion is full of apparent contradictions and obvious mis-statements. This is occasioned by the fact that corrosion is a most complex phenomenon and that factors which may be controlling in the results are frequently overlooked. No two men have the same sample of iron and steel; the heat treatment is generally different, the analysis imperfectly known; the state of the surface and the finish is neglected, and a host of other important conditions completely ignored. No wonder the results present discrepancies, and the public fails t o show an interest in them. Second, certain conclusions which in the early days seemed t o follow logically from the electrochemical theory of corrosion have been shown to be erroneous, and have led to disappointing results. Possibly the logic was false; but a t any rate we cannot crown the theory with any farreaching beneficial results. If I may paraphrase, “the evil that a theory does lives after it; the good is oft interred with its bones.” The theory that the earth was flat delayed the discovery of America by a number of centuries; the phlogiston theory guided experimenters to fruitless fields of investigation and wasted much valuable time and energy. And so with the doctrine that homogeneity of structure in iron and steel, insures an absence of corrosion. The public is still told through advertising propaganda that the old Newburyport bridge withstood Atlantic Ocean storms for over one hundred years because the iron with which it was built was pure and homogeneous, when, as a matter of fact, the majority of the links of the great chains were very impure and extremely heterogeneous. The further fact that when these very pure links were ‘Trans. Am. Electrochem. Soc. 31, 199 (1917). Trans. Am. Electrochem. SOC.39, 53 (1921).
2
THE ELECTROLYTIC THEORY OF CORROSION
807
rolled down to sheet form and exposed to the weather, they rusted like modern iron and disappeared in a few years, is conveniently ignored. The old blacksmith’s forge apparently introduced an element of protection which has not yet been predicted by the electrochemical theory nor realized by laboratory experimentation. “Corrosion may be divided into two classes-that taking place in the atmosphere with free access of the oxygen of the air, and that in positions from which the air is or can be excluded. In the first field the work of Mr. D. M. Buck stands pre-eminently the greatest contribution made in the art of preventing corrosion since the introduction of hot galvanizing. Impartial evidence is now legion that no commercial iron or steel so well withstands atmospheric corrosion as does s tee1 containing approximately 0.2 percent copper. When thus alloyed eve I the much maligned Bessemer steel excels in resistance to corrosion the widely-heralded, pure, open-hearth iron. Unfortunately, copper-bearing steel is not a child of the electrolytic theory, although we have no doubt that when adequate explanation of the wonderful effect of this small amount of copper is finally suggested, it will be found to be essentially electrochemical. With the unbroken record of successful achievement now possessed by copper-bearing steel, and the low cost of metallic copper, there seems to be no excuse for the fabrication of those structures which must withstand exposure t o the weather, from any other material. “But a copper-bearing steel produces its own protective covering only when the oxygen concentration is high. When the amount of oxygen is limited, as when the structure is immersed in water, other protective means must be found. If all of the oxygen in a system could be removed, corrosion would cease. This fact was early recognized by Mr. F. N. Speller, and the careful experimentation carried on by him during the last twelve years in the application of this principle to the preservation of engineering struchres, is now being crowned with success. Enough operating data from large installations are now available to demonstrate the effectiveness of this method of conserving iron and steel.” This statement by Walker called forth a rather bitter reply from Cushman.’ “It is of course perfectly natural that Dr. Walker should have become a strong proponent of the copper alloy theory in corrosion resistance, and i t is noteworthy that he now deserts the so-called electrolytic theory of corrosion and seems to point out that homogeneity of structure of iron and steel is not worth while, from a corrosion standpoint. My own recollection is, and I think the record will show, that Dr. Whitney neither formulated a theory nor attempted to prove anything. He published, as I recollect il, a short paper, in which he suggested, as the result of a passing observation, that the corrosion of iron might be explained or linked up t o the Arrhenius, van’t Hoff, and Nernst conceptions of the theory of solutions. Dr. Whitney stated that this suggestion might be sustained by careful research which he thought worth while. He did not, however, pre-empt the field, but put his suggestion out as an invitation to men who had more time to enter this field of investigation, ‘Trans.Am. Electrochem. Soc. 39, 57 (1921).
808
WILDER D. BANCROFT
Dr. Walker has cited a number of authorities who sustain his views on copperbearing steel, but he chooses to ignore by citation any workers in the field who disagree with his opinions. Dr. Walker states that in 1907 he had the privilege of demonstrating the function of oxygen in connection with corrosion. I may state that in 1905 and 1906 I had already demonstrated, by the use of phenolphthalein the formation of electropositive and negative nodes whenever iron or steel is rusting. This led to thc development of the ferroxyl test which proves electrolytic action. Now a thing that is proved is not a theory. The electrolytic explanation of corrosion is not to be evaded or avoided, no matter whether it coincides with an individual bias or prejudice or not.” It is t9 bc noticed thal Cushman’s statement that “the electrolytic explanation of corrosion is not t o be evaded or avoided” is perfectly true but does not show chat homogeneity is necessarily the important factor or even a desirable factor. In the discussion Walker1 says that “it is obvious from a reading of my paper that not only have I not abandoned the electrolytic theory of corrosion, but that I predict that when the explanatioa for the remarkable behavior of small amounts of copper in reducing corrosion is explained it will be found to be essentially electrochemical. Certain conclusions which at first seemed t o follow from the electrolytic theory of corrosion have now proven themselves erroneous.” It is rather curious that the delusion as to homogeneity being the most important point should have taken hold of people so firmly, because Walker2 pointed out years ago that a continuous coating would protect iron even though the coating were a material which would accelerate corrosion when placed in contact with the iron. “I have already shown that mill scale, or magnetic oxide of iron, is strongly electronegative to iron. Since mill scale is insoluble in water and cannot of itself enter into the reaction, its only function can be analogous t o that of platinum or other insoluble conductor of this kind, viz, to furnish a surface on which the hydrogen liberated by the dissolving iron can separate and be catalytically oxidized to water again. This is also true of the black oxide protective coatings sometimes used upon iron and steel, as, for example, that of the Bower-Barf process. Just as is the case in mill scale, these coatiogs are very serviceable so long as the whole coating is intact. But so soon as a portion of the metallic iron is exposed, this portion corrodes all the moie rapidly an account of the presence of the surface of scale on which the oxidation of the hydrogen and consequent depolarization can go on. The inevitable result is that a ‘pit’ forms at the exposed point and grows deeper and more marked in proportion as the scale is dense and closely adherent to the iron surface. Hence, if it were possible to remove the mill scale entirely from steam boiler tubes, for example, pitting would be largely eliminated, and the life of the tube prolonged.” Attention seems t o have been focussed on the disadvantage of a break in the protecting coating of this type and people have rather overlooked the 1 Trans. Am. Electrochem. SOC.39, 59 (1921). 2Trans. Am.Electrochem. SOC.14, 181 (1908).
T H E ELECTROLYTIC THEORY OF CORROSION
809
possibility of a self-healing coating such as we actually have on aluminum and on nickel. The new stainless steel owes its properties to a coating of this type. Lambert,l in England, has made out an astonishingly good case for the non-corrodibili ty of a homogeneous metal, “The impurities contained in the best commercial iron must, from a chemical point of view, be regarded as considerable, and, i n the light of our present knowledge of the great modifications capable of being produced in the properties of substances by the presence of even minute traces of impurities, it cannot be contended that experiments with impure iron afford trustworthy grounds for a sat,isfactory theory of the oxidation of iron. “The aim of the present investigation, was to bring together, under the simplest possible conditions, the purest obtainable water, oxygen, and iron, in vessels which would be least likely to be acted on by any of these substances. . . . “The choice of the kind of vessel in which t o carry out the experiments was the cause of much difficulty. It was finally decided to use vessels made of transparent fused silica as being least likely to be affected by either water, iron, or oxygen. After many trials and experiments, a simple form of glasd vessel was devised, which, with a tube of clear fused silica, gave all the advantages of an apparatus made entirely of silica, since the water which collected in the silica tube and came in contact with the iron must have condensed on the inside of the silica tube itself. . . . “The material employed i n the preparation of pure iron was a pure speci: men of ‘Kahlbaum’ ferric chloride. The salt was found to be free f y m sulphate, arsenic, alkali, or alkaline earth metals. A solution of the sa!t as made in conductivity water and electrolysed between electrodes of-pure iridium foil. This method is made possible.by the fact that pure iridium is not attacked by chlorine, which is evolved at the anode. The m e t a i c iron which was deposited on the cathode was then t,horoughly wasiid with conductivity water, and dissolved in pure dilute nitric acid. This solution of ferric nitrate in excess of nitric acid was concentrated on the water-bath, and the salt crystallized from the solution in concentrated nitric acid. The crystals were separated from the mother liquor, washed with pure concentrated nitric acid, and recrystallized four or five times from this solvent. The crystals so obbained were colourless, or white when seen in bulk. It is to be noticed that ferric nitrate, prepared from ordinary pure iron, has, when seen in bulk, a pale violet colour like that of iron alum, and tha; the colour cannot be removed by repeated crystallisation from pure nitric acid. “The ferric nitrate crystals were transferred by means of a spatula of iridium foil to a pure iridium boat. The boat was then heated in air on a thick tile, so that the flame gases did not come in contact with it. The ferric nitrate was thus converted into the oxide or basic nitrate. The boat containing the flakes of oxide was then placed inbo n transparent silica tube, and heated in an electric resistance furnace to a bright red heat (just above IOOO’), while 1 J. Chem. SOC. 97, 2426 (1910); 101, 2056 (1912); 107, 210, 218 (1915);Trans Faraday SOC.9, 108 (1913).
810
WILDER D. BANCROFT
a stream of pure hydrogen was passed through the tube.l . . . The hydrogen was prepared by the electrolysis of a solution of pure barium hydroxide. . . . The electrolytic cell contains two pairs of large platinum electrodes and is capable of producing a steady stream of hydrogen. The gas was passed through a U-tube containing lumps of pure sodium hydroxide, in order to remove excess of water vapour, and then through another U-tube containing tightly-packed glass wool. “The metallic iron so obtained, by direct reduction of the flakes of oxide or basic nitrate, had a distinct metallic lustre and a light grey colour. If the flakes of oxide were ground in an agate mortar before being reduced, the iron produced by reduction was light grey in colour, but had little or no lustre. The properties of the two kinds of iron were the same.” “It was found that pule iron, prepared exactly as described above, did not undergo any visible oxidation when treated with pure water and pure oxygen in vessels made of clear fused silica, a r d that there was no change even after several months. “If, however, ferric nitrate, prepared from ordinary pure iron, was used, even after ten recrystallisations, and iron made from it by precisely the same method, the iron invariably showed signs of oxidation in two or three hours, and, after twelve hours, there was always a considerable deposit of reddishyellow ferric oxide on parts of the metal. Oxidation also took place even when the oxide prepared from the nitrate was strongly heated in a stream of pure oxygen for several hours before being reduced t o the metal. “It is impossible that iron prepared in this way can contain anything more than a very slight trace of impurit,y, and that impurity, whatever it may be, cannot be of such a nature that it is acid, or will give an acid on oxidation. “Again, if platinum vessels were used, particularly if a platinum boat was used in which to reduce tbe iron, the iron produced readily underwent oxidation in two or three hours, and oxidation invariably took place at those parts of the metal which had been heated in contact with the platinum boat. “Richards,” in his work on the atomic weight of iron, prepared iron i n somewhat the same way as we have done, but he distilled the nitric acid used from a platinum retort, and employed platinum vessels throughout for his preparation. He states that the iron always contained slight traces of platinum, and that, when it was dissolved in acids, a small black speck of platinum remained. “This small trace of platinum, which may be merely attached to the iron, or may be present in the form af a solid solution, would seem to be enough to cause oxidation t o take place.. 1 In some experiments the oxide was heated in a stream of pure oxygen for several hours before being reduced, in order to remove the occluded nitrogen which is contained in most oxides formed from nitrates. This operation, however, was found to be unnecessary, since the pro erties of the resulting iron were exactly the same as when the oxide or basic nitrate was Jrectly reduced in hydrogen. The slight oxidation undergone by the iridium did not seem to affect the iron. The occluded nitrogen was undoubtedly removed by heating in hydrogen to the high temperature of the furnace. This temperature was between the melting point of silver and that of copper. 2Proc. Am. Acad., 35, 253 (1900).
T H E ELECTROLYTIC THEORY OF CORROSION
811
“All kinds of commercial iron which were used readily rusted under the same conditions of experiment, as also did iron made with the most scrupulous care by many other methods. “A specimen of commercial electrolytic sheet iron (99.9 percent of iron), which had been polished and treated with a I percent solution of chromic acid for three months, and afterwards washed with pure water and quickly dried, readily rusted under the same conditions of experiment. This method of treating ordinary iron is said by Moody t o remove the impurities from the surface of the iron. It seems probable that other reasons must be sought for the non-rusting of the commercial iron used by Moody under his precise conditions of experiment. “It would seem to be proved from these experiments that pure iron will not undergo visible oxidation in contact with pure water and pure oxygen, but that a small trace of impurity i n the iron is sufficient to cause oxidation under exactly the same conditions of experiment, even if this impurity be not of an acid nature or likely t o produce an acid during the reaction.” In the second article Lambert1 says that “pure iron has been kept in contact with pure water and pure oxygen, under atmospheric pressure, for more than two years without showing any signs of corrosion or alteration of any kind. Further, ordinary air can be substituted for pure oxygen and ordinary tap-water for pure water, and the result is the same-the surface of the metal remains brighc and untarnished for an apparently indefinite time. The explanation of this fact is to be sought in the greater homogeneity of the iron, due to its purit’y. If such iron be really homogeneous, and all parts of it have the same solution-pressure, then, on the above theory, a piece of metal placed in contact with an electrolyte will not furnish the conditions for the production of an electric current; since there is nothing to bring about the passage of an electric current, the iron cannot therefore pass into solution and no corrosion can take place. It does not necessarily follow from the fact that. the iron does not rust that there must be perfect homogeneity in the metal. It may be that differences of electric potential do exist on the surface of the iron, but they are so much smaller in this iron than in any commercial varieties of the metalthat,in the presenceof water andoxygen the electric current which passes between the points of different solution-pressure is so small that the amount of iron passing into solution is not sufficient for the formation of a perceptible amount of rust even after long periods. “A striking confirmation of the truth of this argument is afforded by the following experiment: Some pieces of pure iron which bad been exposed t o water and air for several months without showing any signs of corrosion were carefully dried between Swedish filter papers. Some of the pieces were then placed in a polished agate mortar and pressed strongly with an agate pestle, whilst others were left untouched. All of them were then again put in contact with water in silica tubes and exposed to air. The pieces which had been subjected to pressure in the agate mortar showed signs of corrosion in less than an hour, and after several hours a golden-yellow deposit of rust had formed J. Chem. SOC.101, 2068 (1912).
812
WILDER D. BANCROFT
on those parts of the iron which had not been pressed, whilst the parts which had been pressed by actual contact with the agate remained quite bright. Those pieces of iron which had not been pressed showed no signs of corrosion, proving that the piocess of drying had not been the cause of the striking alteration in property of the pieces which had been put under a strain. “The pieces of metal which had been subjected to pressure consisted of two modifications of iron, a pressed part and an unpressed part. These two varieties of iron would have a different solution-pressure, and so, when placed in contact with the electrolyte water, would constitute a self-contained galvanic element. The fact that rust formed on the unpressed part showed that iron passed into solution at that part, which was therefore relatively electropositive to the pressed part. “That the pure iron is not absolutely homogeneous-that it does possess electrically different parts-is indicated by its behaviour toward acids and towards solutions of salts of the alkali metals in the presence of air or oxygen. These reactions are still under investigation, and can only be considered very briefly at present. “It may be said generally that cold dilute dulphuric and nitric acids have very little visible action on the metal, but that even very dilute cold hydrochloric acid causes the slow evolution of bubbles of hydrogen. The metal readily dissolves in all three acids on warming, but, again, hydrochloric acid is much more vigorous in its actidn than the others. “The action of solutions of alkali salts on the metal in contact with air shows many irregularities which still await investigation. Pure iron which had been exposed to the action of water and air for many months without showing any signs of rusting, underwent corrosion in a few hours when transferred to a normal soluticn of sodium chloride in air. “That othcr constituents of the air besides oxygen play no part, in this reaction is shown by the fact that corrosion took place just a$ readily when piire oxygen was used. “The chlorides of potassium and ammonium seem to have a similar artion, hut, curiously, the corresponding sulphabes and nitrates behave differently. The pure iron may bc exposed to concentrated solutions of the sulphates and nitrates of sodium, potassium and ammonium in the presence of air, often for many days, wilhout undergoing much corrosion. “It seeins quite possible that the study of the action of salt solutions on pure iron may bring to light some definite evidence for the view held by the author that a considerable factor in determining the corrosion iron is the alteration in the electrical character of the different parts of the metal brought about by the action of salts and acids. “It is well known that all ordinary forms of iron, when placed in contact with solutions of copper salts, are immediately covered with a deposit of metallic copper. It is only when the metal has been rendered ‘passive’, by one of the many processes which produce this condition, that iron can be made to remain unaffected in a solution of a copper salt. The copper salt solution must be
THE ELECTROLYTIC THEORY OF CORROSION
813
very dilute, and even then the passive iron recovers its ordinary behaviour after several hours, and causes copper t o be deposited on it. “Pure iron will withstand the action of a saturated solution of copper sulphate or copper nitrate, at the ordinary temperature, for an apparently indefinite time, without losing any of its metallic lustre and without any perceptible trace of copper being deposited. ‘‘Some specimens of iron which had been exposed for several months to a concentrated solution of copper sulphate were removed, washed, and dried, and examined under the microscope. The curious structure of the surface was exactly the same as it was before the iron was placed in contact with the copper sulphate solution. “Pure copper sulphate, free from iron, was used, and the solution after being in contact with the iron for many months failed to give any test for the presence of iron. “If the temperature of the copper sulphate solution is raised to that of boiling water, deposition of copper on the ircn slowly occurs, and finally, after some hours, the iron passes into solution completely, and copper is left behind in the form of very small crystals. “Copper is also deposited on the iron if it is pressed in an agate mortar before being put in the solution of copper sulphate, or if it is pressed with s quartz rod while under the copper sulphate solution. “The behaviour of the pure iron towards copper chloride is, however, quite different. If a concentrated solution of copper chloride is used, the iron becomes coated with copper immediately it is put into the solution, and, within a few minutes, the iron all disappears, and only finely dividcd copper remains. “If a very dilute solution (less than one percent) of copper chloride is used, the action is slower. For a few seconds the iron retains its brightness; then dull spots are seen at some points on the surface of the metal; these quickly spread over the whole surface of the iron, and the reaction proceeds to an end, as before, in a remarkably short space of time. “The experiment is perhaps even more striking if carried out in a vacuum. The finely divided copper, which is left after the first reaction, slowly dissolves, and fipally white, insoluble cuprous chloride is left. “Similar results are obtained if a aolution of sodium chloride is added to solutions of the sulphate and nitrate of copper. “Ordinary metallic aluminum behaves in much the same way towards solutions of copper saIbs. The metal is not affected by solutions of copper sulphate or copper nitrate, but, if copper chloride is used, or if sodium chloride is added to the solutions of copper sulphate or nitrate, precipitation of copper on the aluminum immediately follows. “This behaviour of aluminum has been attributed to the fact that the metal, under many conditions, becomes coated with a protective film of bydroxide or basic salt . “It is assumed that this protective film is more readily soluble in the hydrochloric acid produced by the hydrolysis of the copper chloride than in the sulphuric or nitric acids from the sulphate or nitrate respectively. This argu-
814
WILDER D. BANCROFT
ment is strengthened by tlhefact that aluminum is readily dissolved by hydrochloric acid, whilst it is practically unaffected by sulphurk or nitric acids eveD
at roo’. “Experiments have shown that the same arguments cannot possibly hold good in the case of pure iron. “It is very improbable that iron prepared by the reduction of the oxide by hydrogen at a high temperature, and allowed to cool in the gas, would have a film of oxide on the surface, and, on account of the irregularity in the shape of the pieces of iron, it is unlikely that such an oxide film, if it existed, would form a complete, unbroken protective coating. “Nevertheless, the presence of some kind of protective coating on the iron, capable of being dissolved by cold dilute hydrochloric acid or nitric acid under the same conditions, would explain why copper is deposited from a solution of copper chloride and not from copper sulphate oncopper nitrate solutions; it would explain, too, why iron when subjected to pressure under copper sulphate and copper nitrate solutions causes the deposition of copper on it, for pressure might bring about a disruption of such a film; and further, it might be considered as an explanation of the fact that rise of temperature will cause copper to be deposited on the iron from copper sulphate and nitrate solutions, on the ground that such a film would be more soluble in hot, tkan in cold acids. “There are two possible kinds of protective films which might conceivably be present on the surface of the pure iron, namely, (a) an oxide film produced by the reversible decomposition of small traces of water in the hydrogen used for the reduction of the iron oxide t o iron, and (b) a protective layer of some hydride of iron produced by the absorption of the hydrogen as the metal cooled down in the gas. “Careful experiments have been made to test these possibilities. The hydrogen used for the reduction of the iron oxide was dried by passing it through a long tube containing phosphoric oxide, so as to remove all but the most minute traces of water, and then the iron which was produced was brought in contact with copper sulphate solution whilst it was still in the atmosphere of hydrogen, The copper sulphate solution itself had been previously saturated wich pure hydrogen. There was no deposition of copper on the iron, which remained quite untarnished in the solution. “Other specimens of iron were prepared and allowed t o come in contact with the air before they were cold. A thin, yellow layer of oxide was formed on the metal, but this slightly oxidised metal caused the immediate deposition of copper when placed in contact with a solution of copper sulphate. “It is probable that, as suggested above, the oxide film does not form a complete protective coating; in such a case there would be differences of potential on the surface of the iron, and we should expect copper to be deposited . “It must be concluded, then, that the non-deposition of copper by the pure iron from copper sulphate and copper nitrate solutions cannot be accounted for by the preseoce of a protective oxide film on the metal.
T H E ELECTROLYTIC THEORY O F CORROSION
815
“That the same argument holds good in the case of a possible protective film of hydride is proved by the fact that specimens of the metal which had been heated in a clear silica tube for several hours, a t about 1000’in a vacuum, until spectroscopic tests showed that all the hydrogen had been removed, behaved in all respects as before-they remained quite bright and unaffected in contact with saturated solutions of copper nitrate and copper sulphate, but caused the immediate deposition of copper when placed in very dilute snlutions of copper chloride. “The behaviour of the iron under different conditions towards solutions of the sulphate and nitrate of copper can be readily explained on the basis of an electrolytic theory, but the extraordinary behaviour of the metal in copper chloride solutions is mysterious. ((Wehave seen that the chlorides of the alkali metals have a very remarkable action in starting and promoting the rusting of the pure iron in air, and it may be that soluble chlorides have the power, in some way or other, of increasing the electrical differences which exist in the iron. If the chloridior has such a property it would also explain this curious behaviour of copper chloride. For the present this must be left an open question. Cushmanl has stated that no iron has been found of such purity that it gives no trace of positive and negative nodes in the ferroxyl indicator; but Lambert’s “pure iron remains quite bright, and causes no formation of positive and negat,ive nodes when left in contact with the reagent for an indefinite time. If a piece of the pure metal is subjected to pressure, however, it behaves like other kinds of iron when put in contact with ferroxyl reagent. A pink colour develops in the jelly around the pressed part, and Turnbull’s blue is formed round the parts which have not been subjected to pressure.” While this is a very good piece of work, the author has proved too much. On any point of view copper will precipitate on iron unless the iron is covered with a film of some sort. All the experiments indicate the existence of a film. On the other hand the experiments to determine the nature of the film gave negative results. Consequently they are either faulty in some way or they do not cover the ground. It would have been interesting if Lambert had determined the electromotive force of his pure iron. This is not essential however and the film which keeps the copper sulphate solution from acting is quite sufficient to keep the metal from corroding. It seems practically certain that the film is an oxide film formed in the reduction experiments when the copper sulphate solution came in contact with the iron. It must be admitted, however, that it is not clear why rusting should take place a t the unbruised portion rather than at the bruised portion. The experiments with lead2 are also very interesting even though one cannot accept the author’s conclusions. (‘Many methods were employed to prepare lead in a high state of purity previous to its distillation in the manner described above; but it was found very difficult to remove all traces of iron. “Corrosion and Preservation of Iron and Steel,” 49 (1910). *Lambert: J. Chem. SOC.107, 210 (1915).
816
WILDER D. BANCROFT
Finally, lead was prepared by the method described by Stas.’ I t was cut into small chips by means of a pure indium spatula on to which a cutting edge had been ground.2 The lead was then distilled in clear quartz tubes, which were about 7-8 mm bore and 2 5 cm long, and were divided into three compartments by narrowing the tube down to capillaries. The tubes were exhausted while the metal was kept in the molten state, and finally the lead was distilled at about 1 2 0 0 ~ . The first fraction was collected in the first compartment which was then sealed off. The middle fraction was collected in the second compartment, whilst the residue was left behind in the third compartment. It is worthy of note that the globules of lead obtained in this way adhered very closely on the under side to the quartz, producing a very fine reflecting surface. It was found that if the distilled lead was exposed to air a short time after distillation, but when quite cold, the brilliant surface was quickly dimmed by a film of oxide. If, however, the metal was kept in the vacuous tube for several monthe, exposure to the air did not then cause any appreciable decrease in its brilliant, lustre for several days. It was only the middle fraction of lead that WRS used in t,he experiments described above.” I n another experiment water was distilled into the vessel within twentyfour hours after t,he distillation of the lead and then the vessel, containing only pure lead and pure water in a vacuum, was sealed off and set aside for twelve months. “After the lead in the first vessel had remained in contact with water in a vacuum for twelve months the metallic lustre of the metal was quite unaffected. The point of one of t)hecapillaries was then broken off in an atmosphere of pure oxygen, and the gas thus allowed to come into contact with the lead and water. For more than a week there mas no visible tarnishing of the brilliant metallic surface of the large globule of lead which was under the surface of the water, but after that time the metallic lustre was gradually dimmed by the formation of a dark-coloared coat,ing of oxide on its surface. The rate of corrosion was, however, so slow that after more than six months’ exposure to t8hecombined acticr, of water and oxygei the layer of oxide was thin enough to display interference colours. After exposure for nearly a year the globule of lead is covered with a layer of a dark brown oxide, and minute crystals of hydrated lead oxide can be seen in the water.” Lambert considers that lead probably exists in different modifical ions, the more instable forms predominating in the freshly distilled metal and g:ving it a heterogeneous character, which will tend to disappear with time owing t o the conversion of the instable forms into the stable one? “When the metal is kept for a long time in a vacuum, it will become more homogeneous as far as its physical character is concerned, and consequently the rate at which it will pass into solution in water in the presence of oxygen will be greatly decreased.” ‘Bull. Acad. roy. Belg. 10, 295 (1860).
* Pure iridium is so hard that it is possible to prind a sharp cutting-edpe on to a plate
of the metal. 3 Since room temperature is an annealing temperature for lead, the instable form may have been lead under strain.
THE ELECTROLYTIC THEORY OF CORROSION
817
Lambert’ draws the following conclusions from his work:“(I) Pure lead which has been distilled in a vacuum undergoes very rapid corrosion when subjected to the action of pure water and pure oxygen within a short time of the distillation of the metal. Chemical tests fail t o show the presence of lead in solution before the addition of oxygen. “(2) If the lead prepared in this way is kept for long periods in water, in a vacuum, before the oxygen is added, the rate of corrosion is enormously decreased. “(3) Pure lead which has been distilled in a vacuum, and kept for some months, can be exposed t o ordinary air for many days without any appreciable diminution of its brilliant metallic lustre, but more prolonged exposure causes the gradual formation of a dark-coloured oxide on the surface of the metal. The pure, distilled lead resembles silver or mercury in appearance, and shows none of the blue or bluish-grey colour usually associated with metallic lead. “(4) The electrolytic theory of corrosion is applicable to lead. The passing of the metal into solution, which precedes corrosion, is due to electrolytic action between the electrically different parts of the mass of lead. In the case of chemically pure lead, the physical heterogeneity (due t o the presence of different allotropic modifications of lead in the mass of metal) causes parts of the mass to be electrically different from other parts, and these electrical differences persist for a long time after the preparation of the metal.” I n this work Lambert has violated the first canon of research. It is legitimate t o postulate allot’ropic modifications of lead. I n view of the experiments of Baker2 on the behavior of dry mercury and dry benzene, it is perhaps legitimabe t o assign any properties one pleases to pure lead. These are assumptions ad hoc, however, and must be justified independently. The author has not done this. He is also in error in saying that his results “afford very strong evidence of the electrolytic theory of corrosion.” There is nothing in this theory, when applied properly, which requires that a homogeneous metal should be inert in the presence of a depolarizer like oxygen. The POtentistl difference between silver and silver nitrate may be, and doubtless is, modified by the presence of impurities in the silver; but nobody would claim that we did not have reversible equilibrium between a silver nitrate solution and an absolutely pure and homogeneous sheet of silver. It is much more reasonable t o assume that, on long standing in an alleged vacuum, a film forms on the lead which is more resistant than the oxide film which forms more rapidly. Lambert has overlooked another point. I n presence of water and oxygen his lead oxidizes slowly while his iron does not. For this to be true the iron must be more noble than the lead or it must be covered with a more resistant film. Very pure iron may be a meha1 as noble as silver; but this seems distinctly improbable and certainly requires proof. If pure iron has as low a solution pressure as silver, impurities must increase the solution pressure enormously which is very interesting if true. We must reject Lambert’s conclusions on the grounds that he has given no independent confirmation of his J. Chem. SOC. 107, 217 (1915).
* J. Chem. SOC. 121, 568
(1922).
818
WILDER D. BANCROFT
postulates, that his conclusions are mutually inconsistent, and that his results are not in accordance with the electrolytic theory of corrosion though he says they are. With these obvious errors of omission and commission the probability of his also going astray in some less obvious manner becomes painfully great. In any event his results are purely of academic interest so far as the problem of corrosion is concerned, because he finds that iron and lead of very high, though not the highest, purity behave as they should and corrode in the presence of water and oxygen. The function of oxygen as a depolarizer was pointed out by Walker1 sixteen years ago. “One fact regarding the corrosion of iron appears to be undisputed, viz., that oxygen is necessary for a continued action. This corrosive action can cease from two causes, viz., the osmotic pressure of the dissolved iron may increase until it neutralizes or compensates the solution pressure of the metallic iron; or the action may be stopped by the separation of a film of molecular or gaseous hydrogen2 upon the metal, which, owing to its resistance, prevents the flow of an appreciable current. From the fact that iron possesses, even in re1at;vely concentrated solutions of iron salts, a very appreciable potential, it would seem highly improbable that the solution of the iron in water should be stopped by the osmotic pressure of that already dissolved, and therefore, although the oxidation and subsequent precipitation of the already dissolved iron is the most striking function of the oxygen, it is probably by far the least important, and its real accelerating action lies in the destruction of the hydrogen film already separated out on the surface of the metall$ iron.” In a later paper Walker8 says that the “accelerating action of metals such as platinum, copper, and lead, and materials such as coke and millscale upon the corrosion of iron is undoubtedly due alone to the aid these bodies offer to the elimination of the hydrogen produced by the corrosion reaction, owing both to the low over-voltage of hydrogen apon these substances and to the high catalytic action of these materials on the hydrogen-oxygen reaction by which ’ the deposited hydrogen film is removed. “While in specific cases the corrosion of iron can be absolutely controlled by first one and then another of these factors influencing it, too little attention is generally paid to the last one. The retarding action of the hydrogen film which is stahle in the absence of oxygen and any acid-forming compound is SG controlling that, from this point of view, oxygen (or air) may be said to be the cause of the corrosion. If oxygen be removed completely from boiler feedwater, the boilers will not pit or corrode. If oxygen be separated from the feed of ordinary hot-water supply lines, the ‘red-water plague’ and other corrosion troubles will disappear. “By this removal of the cause of the disease, not only are all its ill effects avoided; but the necessity of drugging with alkali, removal of the products of corrosion and such curative measures, with their attendant evils, are elimTrans. Am. Electrochem. Soc. 14, 179 (1908). Walker has overlooked the polarization effect due to over-voltage (monatomic hydrogen).] “Trans. Am. Electrochem. Soc. 29, 436 (1916). 1
[2
THE ELECTROLYTIC THEORY O F CORROSION
819
inated, and in this, as in so many other cases, experience has proved prevention to be better than a cure. If electrical and mechanical engineers will only take more closely into their confidence their brother electro-chemists, we can together more quickly make available the knowledge on the subject of corrosion already a t our command for the elimination of the difficulties which corrosion introduces into commercial practice.” I n line with this Speller’ says that “the electrolytic theory of corrosion as formulated in I903 by Dr. Whitney has led to the development of certain protective systems which are based on the removal of dissolved oxygen from water. Careful experiments in the Research Laboratories of the Massachusetts Institute of Technology and the National Tube Company have demonstrated that the amount of corrosion found is almost directly proportional to the amount of oxygen in solution and varies directly as the temperature. The predominating influence of free oxygen in water was suspected before this as a result of the early study of pipe corrosion, for the most striking fact in practical pipe experience is that hot-water heating systems invariably show no corrosion to speak of after thirty-five or forty years use whereas frequently hot-water supply systems operating a t the same average temperature with the same water last only six or eight years. That this was independent of whether the material was iron or steel was fully demonstrated by many service tests which were conducted for a period of over ten years, in which representative pipes of each class were installed alternately in hot water lines.” Quite remarkable results have been obtained when all the oxygen is removed from a hot water system either by a vacuum process or by chemical combination.’ “If the water space of the heater is filled with thin steel lathing or if the water be made to flow through another tank containing steel scrap in this form, the residual oxygen will be fixed in a few minutes. Water so treated is practically inactive towards iron, and may be used in boilers or steel economizers without fear of corrosion so far as the water is concerned. Other metals may be used for oxygen removal, such as zinc; but iron is the most economical, not only because of its low first cost but for the reason that the ferrous hydrate formed removes an equivalent of oxygen in addition to oxygen taken up by the hydrogen.” W. D. Richardson states that “oxygen is of the utmost importance in promoting the continuance of corrosinn by acting as a depolarizing agent, and this action is particularly prominent in the corrosion of iron and its alloys. If, however, the hydrogen active in corrosion is liberated at corrosion cathodes with low over-voltage, polarization supervenes less readily and corrosion proceeds at a greater velocity than in the case of a compars tively pure metal containing no suitable corrosion cathodes in its composition. Oxygen is more necessary to, and causes greater acceleration of, corrosion in pure iron than in an impure alloy such as cast iron, for in the absence of mucb oxygen relatively pure iron corrodes at an extremely low rate, but at a comparatively rapid rate Trans. Am. Electrochem. SOC.39, 141 (1921);J. Franklin Inst. 193,515 (1922). Though not specifically so stated, dissolved carbon dioxide must also be removed. Trans. Am. Electrochem. SOC.39, 64 (1921).
WILDER D. BANCROFT
820
in the presence of oxygen. The opposite behavior of relatively pure iron when compared with cast iron under different conditions of corrosion is one of the most st’rikingphenomena connected with the subject, and will probably afford illuminating data when correctly interpreted.” Messrs. Bengough and Stuart’ dispute the depolarizing action of oxygen. “It is a well-tested experimental fact that no evclution of hydrogen gas from distilled water can be detected in the case of any of the metals mentioned above a t the ordinary temperature, with the possible exception of magnesium. Even when the most careful microscopic search is made, none can be observed in the case of copper and zinc, as has been already reported2 and recent work shows similar results for nickel and lead. With aluminum the evolution of hydrogen can be detected t o a very limited extent in London tap-water, and quite readily if the metal be treated with mercury and then immersed in distilled water; in the latter case, the hydrogen evolved is chemically equivalent to the aluminum corroded. In this case, therefore, Reaction 1appears t o take place; but with these exceptions no direct evidence of the evolution of hydrogen gas has ever been found by the present authors, or been recorded in the literature on the subject, in the case of corrosion of ordinary massive metals from aluminum downwards in the electro-chemical list, in distilled water, conductivity water, or the specially purified water used by Lambert in his experiments on iron and lead. Nevertheless, under certain circumstances, all these metals may be very considerably corroded by the action of distilled water in the presence of atmospheric oxygen. There are two possible explanations for the non-appearance of hydrogen. The first and more usually accepted explanation is that the hydrogen is displaced by the metal electrolytically, but immediately oxidized by the oxygen of the atmosphere with formation of water. The conception is that of electrolytic cells dispersed over the surface of the metal, from the anodes of which metallic ions pass into solution displacing hydrogen, which appears at the cathodes, and is immediately oxidized hy the atmospheric oxygen. A second possible explanation is that either no hydrogen is displaced, or that though displaced from solution it never reaches the gaseous form. “TOthe first view, grave objections may be urged. I n the first place, when hydrogen is actively evolved, as in the cases of magnesium and aluminum already quoted, it comes off from the seat of corrosion, such as a pit, and not elsewhere at a more or less distant cathode, as demanded by this theory. I n the second place, A. M. Williams3 has measured the depolarizing power of mixtures of oxygen and nitrogen a t the ordinary temperature in the case of a silver-zinc cell, the mixture being allowed to play up the surface of tho cathode of the cell, As compared with air, a gg percent oxygen mixture gave an increased depolarization of 0 . 2 7 volt, and 55 percent oxygen of 0.165 volt; it is clear, therefore, that if air is a depolarizer, it is a very imperfect one. I n the 1
“Sixth Report t o the Corrosion Research Committee of the Institute of Metals.”
J. Inst. Metals, 28,
72 (1922).
Kahlenberg: J. Am. Chem. SOC., 25 (1903). 3 J. SOC.Chem. Ind. 39 (1920). 2
T H E ELECTROLYTIC THEORY O F CORROSION
821
third place, hydrogen can be collected quantitatively when displaced from acid by zinc in the presence of air, by magnesium from salt solutions, and by aluminum amalgam from distilled water, and this shows that little or none of the gas has been directly oxidized by the oxygen of the atmosphere to form water. The depolarizing action of atmospheric oxygen must be so slow in such cases, as compared with the rate of hydrogen production, that it is practically zero. I n the fourth place, it has been found, in the case of zinc, that when the metal is placed in very dilute acetic acid ( I acid: 5000 water), the rate of corrosion by which it is comparable to that by distilled water, bubbles of hydrogen can be detected coming off from the surface of the metal; hence it is clear that at least part, and probably the whole, of the displaced hydrogen gas is not directly oxidized in the presence of water. But if the same sample of zinc be subsequently placed in distilled water, no evolution of hydrogen can be detected, yet the amount of zinc oxidized is approximately the same in the two cases if calculated over the whole area of the specimens, and much greater at certain localities in the case of distilled water than at any area in the acetic acid; hence the hydrogen should be more easily detected, if it be really evolved at such areas, in the distilled water than in the acid. “Many cases could be quoted to illust,rate the slowness with which cathodic polarization is removed in the absence of any depolarizer other than air, such for instance, as the necessity for the presence of potassium bichromate or nitric acid in the bichromate and Grove types of primary cell respectively, even though the cathode be of carbon, which should assist the oxidation of the hydrogen by adsorbed oxygen. “If oxygen functions solely as a depolarizer, rapid circulation of the liquid over the metal should assist corrosion by increasing the rate of removal of the hydrogen film. Actually, Newton Friend has shown that an increase of water speed over the surface of the metal increases the rate of corrosion of iron only up to a limiting speed. If the speed be increased beyond this limit, the rate of corrosion rapidly falls. The latter result seems quite incompatible with the idea that the function of oxygen in corrosion is solely that of a depolarizer.” Mr. Speller informs me that hydrogen is found in their deoxygenated hot water systems and that iron will react with water, though slowly, to give hydrogen at lower temperatures. If zinc reacts with water in the absence of air, one of the products must be hydrogen. If air is present, there is no necessary reason why hydrogen should be given off. The criticism as to t’heevolution of hydrogen from aluminum and magnesium is not well taken because the anode and cathode reactions will always occur as close together as possible so as to make the resistance in the circuit as small as possible, so that speaking about “a more or less distant cathode” is either absurd or deliberately misleading. Nobody denies that oxygen under m’ost conditions is a slow and unsatisfactory depolarizer. Even under the most favorable circumstances we can only draw very small currents from a hydrogen-oxygen gas cell because of the rapid polarization. In the experiments with zinc and dilute acetic acid, it is probable that the evolution of hydrogen prevents the oxygen from coming in contact with the metal after the initial depolarization. With zinc and water, the rate
WILDER D. BANCROFT
822
of corrosion is a function of the rate of diffusion of the oxygen among other things. The solubilit’yand other properties of the corrosion products are also possible important. factors. One cannot prove anything with experiments like this. The paragraph in regard to the use of bichromate or nitric acid in primary batteries is quite pathetic. People draw from these batteries currents which are enormously greater than any which would correspond to the corrosion of ordinary metals under ordinary conditions. One ampere means over a gram of zinc per hour which would be an unheard of rate of corrosion. There is some question as t o the accuracy of Friend’s work on the effect of stirring on the rate of corrosion of iron,’so that a discussion of it is unnecessary at the present time. There is plenty of positive evidence as to the depolarizing action of oxygen, some of it coming from the opponents of the electrolytic theory of corrosion. Moritz Traube2 has shown that the first reduction product of oxygen at the cathode is hydrogen peroxide and this has been confirmed by Richarz and Lonnes? though one must not have too much oxygen4 if one is to isolate the product. This is not disputed by Messrs. Bengough and Stuart5because they say specifically that it is very doubtful whether depolarization by atmospheric action takes place at all at atmospheric temperatures, except in special circumstances, such as the presence of platinized platinum. They are willing to concede some depolarization in presence of platinized platinum; but they deny its occurrence with any other metals. Unfortunately for this rather ingenuous view, Schonbein6showed that when lead amalgam is shaken with dilute sulphuric acid and air, hydrogen peroxide is formed in an amount equivalent to the corroded lead. Traube obtained similar results with zinc in the presence of water. Dunstm, Jowett and Goulding’ state that “the conclusion is inevitable that, although hydrogen peroxide cannot be actually detected during the rusting of iron, this compound is probably formed as an intermediate product of the change. . . . The conclusion that hydrogen peroxide is formed in the process of rusting receives strong support from the evidence accumulated by other observers that this compound is frequently produced in those chemical changes which involve spontaneous oxidation through the agency of the oxygen of the air. “The formation of hydrogen peroxide thus appearing t o be a necessary part of the chemical process of rusting, the nature of this process required investigation. The formation of hydrogen peroxide during various processes of oxidation has been explained, notably by Hoppe-Seyler in connection with physiological processes, by the supposition that the substance is oxidized by one atom of a molecule of oxygen, the other atom of which attaches itself to a Wilson: Ind. Eng. Chem. 15, 129 (1923). 2Ber. 15, 243 (1882). * Z.physik. Clem. 20, 145 (1896). 4 Bornemann: Z.anorg. Chem. 34, I (1903). 6 “Sixth Report to Corrosion Research Committee of the Institute of Metals.” . I nst: Metals, 28,75 (1922). 0 Mellor’s “Treatise on Inorganic Chemistry,” 1, 926 (1922). 7 J. Ch em. SOC.87, 1,548(1905). 1
THE ELECTROLYTIC TIIEOAY OF CORROSION
823
molecule of water forming hydrogen peroxide. Thus in the case of iron the initial change would be Fe Oz+HzO = FeO +HzOz. “On the other hand, Traube has supposed the oxygen is taken, not from molecular oxygen, but from one molecule of water, the liberated hydrogen combining with a molecule of oxygen to form hydrogen peroxide. Thus, in the case of iron, Fe+OH2+02 = FeO+H202, Traube’s view involves the assumption that hydrogen peroxide is not oxidised water, but ‘reduced’ or hydrogenised oxygen, and in support of this contention he has brought forward a considerable body of evidence. “The two possible modes of formation of hydrogen peroxide have been experimentally investigated so far as they relate to the rusting of iron. Positive evidence has been obtained in support of the theory involving the decomposition of water, whilst negative evidence was forthcoming against the view that oxygen is taken directly from dissolved oxygen. The results distinctly support the conclusion that water is decomposed by the iron and that the liberated hydrogen goes t o form hydrogen peroxide with the dissolved molecular oxygen. If the existence of hydrogen peroxide is prevented by the introduction of a soluble substance capable of destroying it, little or no action between the iron and the water takes place at ordinary temperature. The oxidation process appears therefore to be a part of a definite cycle of chemical change, the energy of which is partly derived from the combination of the hydrogen formed. It has been found that rusting of iron can occur in the absence of free oxygen provided that certain oxidising agents are present with which the hydrogen of the water can interact. “Another possible explanation of the formation of hydrogen peroxide in the rusting process may be noticed here. It has often been suggested that hydrogen peroxide may be formed from oxygen dissolved in water, especially under the influence of light. If this were proved to be the case, the formation of hydrogen peroxide by the direct oxidation of water would be established and an extremely simple explanation afforded of the phenomena of the oxidation of iron. There is, however, no satisfactory recorded evidence that hydrogen peroxide is ever produced in a solution of oxygen in water, whilst rusting commences and proceeds without interruption in the dark.” “If the theory [of Hoppe-Seyler], which supposes direct action of iron on the oxygen is correct, it is probable that the oxygen could be replaced by nitrous oxide which, as is well known, is readily separated into nitrogen and oxygen. The reaction might proceed thus: Fe+ON2+Hz0 = FeO+N2+H20, and rusting should therefore take place in the absence of free oxygen. Two experiments in which pure iron was left in contact with water and nitrous oxide in the absence of free oxygen were carefully carried out in the manner described, but no rusting occurred in either case.” “On the other hand, if it is supposed with Traube that the metal first attacks the water liberating hydrogen, it ought to be possible to replace the
+
824
WILDER D. BANCROFT
oxygen by some reducible substance capable of reacting with the liberated hydrogen. Rusting should then proceed in the absence of free oxygen. “Experiments were made in which potassium ferricyanide, nitroethane, nitrobenzene, methyl alcohol, free hydroxylamine, and potassium nitrate, respectively, were included in the tube containing pure iron and pure water, the remainder of the tube being full of hydrogen. In order to prove the complete absence of oxygen the hydrogen, before being allowed to enter, was passed through a mixture of pure iron and water only, contained in a tube which was afterwards sealed up; since no rusting took place in this tube, the absence of oxygen was verified. The results of the experiments were as follows: I n the case of potassium ferricyanide, the liquid assumed a yellowish-green colour, whilst the surface of the iron became coated with a blue substance; after a time the action ceased and no further change was observed. With nitroethane, ordinary rust was produced on the iron and the liquid became dark in colour. I n the case of hydroxylamine, oxidation of the iron occurred and bubbles of gas were evolved. I n another experiment in which the iron was exposed to the action of the hydroxlamine and water in a vacuum, the same action was noticed. With potassium nitrate and nitrobenzene, the iron remained quite bright . “The results of these experiments show therefore that the free oxygen can be replaced by potassium ferricyanide, nitroe thane, or hydroxylamine, and that under these conditions rusting of iron takes place in the absence of free oxygen .” Special experiments proved that rusting of iron does not occur at ordinary temperatures “in the presence of either dry or moist oxygen, carbon dioxide, or in a mixture of these gases if the temperature is constant. When, however, the temperature fluctuates and liquid water is deposited on the metal, rusting occurs in the presence of oxygen alone or mixed with carbon dioxide, but not in carbon dioxide alone or in oxygen mixed with ammonia. . . Experiments were also made in which pure iron and oxygen were left in contact with dry ether instead of water. In this case no rusting occurred. It is therefore concluded from the results of all these experiments that liquid water is essential for the rusting of iron, and that the chemical action involved is the reduction of the water by the iron, thc hydrogen thus formed going to produce hydrogen peroxide, which, reacting with ferrous oxide fir& formed, produces the form of ferric hydroxide known as iron rust,.” Dunstan and the others refer t o Whitney’s electrolytic theory of corrosion but reject it, in favor of what they call the hydrogen peroxide theory, apparently not seeing that, if oxygen acts as a depolarizer in an electrolytic cell, hydrogen peroxide must be formed by the reduction of the oxygen. The electrolytic theory of corrosion in presence of air is merely that we have a metal-oxygen cell M HzO 02.Dunstan himself points out that the free oxygen can be replaced by potassium ferricyanide or nitroethane, which shows that a depolarizer is all that is necessary and that neither oxygen or hydrogen peroxide is essential. It is not clear just bow the hydroxylamine functioned. If it acted solely as a reducing agent, it should not have been
I
I
T H E ELECTROLYTIC THEORY O F CORROSION
825
beneficial and if it was reduced to ammonia, there is no reason why gas bubbles should have been given off. The phenomena described by Dunstan might occur if hydroxylamine lowered the over-voltage for hydrogen; but we do not know that it does that. While everybody has talked about the depolarizing action of oxygen, it does not seem to have dawned on them that a self-contained electrolytic action involves the existende of a short-circuited voltaic cell. In fact Lambert’ says that (‘hydrogen peroxide, which is produced in the wet oxidation of lead, is the product of a subsidiary action, and has no direct bearing on the process of corrosion. It plays a part, however, in oxidizing the monoxide t o higher oxides.” When a man throws away tricks in this way, it is not surprising that he does not win. Bengough and Stuart2 evidently do not understand at all what is meant by a depolarizer. “It appears t o be common ground with practically all investigators of corrosion that the presence of oxygen is necessary if any appreciable amount of corrosion of the common metals by water is to take place at the ordinary temperature. There is at present, however, a wide difference of opinion as to the function of oxygen. Its rBle has been explained in two different ways. On one view its function is direct and primary; on the other (and this is a more widely accepted view), its function is considered t o be that of a depolarizer, and consequently secondary. These two views may be expressed as follows in the case of a divalent metal:
I.
+ +
M HzO O*M(OH) 2HpO
I1.
{ z o
2
+M(OH)z+H2
+HzO
1
Reaction I can conceivably take place whether or not a metal can displace hydrogen from solution. Reaction I1 can only take place if a metal is able to do so.” The second reaction should have been written to show the formation of hydrogen peroxide; but that is a minor matter. The serious error is the statement that Reaction I1 can only take place in case a metal is able to displace hydrogen from solution. The effect of the depolarizer is to lower the cathode voltage and t,o make it possible for the reaction to occur at a measurable rate when it could not otherwise do so. Apparently Bengough and Stuart would explain what happens a t the cathode in the electrolysis of caustic soda by saying that sodium is first set free and then reacts with water, setting free hydrogen. Potential measurements would have shown Bengough and Stuart that their statement was wrong. This error leads them naturally into another in the next paragraph. ‘(Copper can only displace hydrogen from hot concentrated hydrochloric acid, and then only to a very limited extent. I n the case of copper, it is important t o notice that hydrochloric and sulphuric acids in the cold will only attack the metal in the presence of oxygen, and there is little J. Chem. SOC.107,218 (1915). “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 70 (1922).
826
WILDER D. BANCROFT
doubt that the explanation is t o be found in the direct oxidation of the copper and the subsequent solution of the oxide.” It is practically certain that t,his is not the true explanation. If we make up a cell with copper, sulphuric acid, and an oxygen electrode, the copper will corrode readily even though the oxygen is not in contact with it. The ordinary reaction is the same thing in a condensed form and the reaction goes through the hydrogen peroxide stage as shown by Dunstan; Jowett, and Goulding. Bengough and Stuart’ wish to limit electrochemical action of cases in which there is a meamrable distance between t’heanode and the cathode; but the advisability of this as a working hypothesis is very doubtful. “On such a view the electrochemical process (except the colloid precipitation) operated entirely inside the pit, and the process and the result were practically indistinguishable from those associated with the oxidation of the metal; it was merely a question whether or not there was an intermediate and momentary formation of hydrogen-a question that eluded experimental test. I n the limiting case in which the anodes and cathodes were indefinitely close together, the difference between chemical and electrochemical action disappeared, and this appeared to be the usual case. I n default of evidence the authors preferred the broader term chemical. Subsequent to the initial action, whatever its type might be, corrosion proceeded precisely as on the authors’view. . . . “It might be pointed out that in certain cases electrochemical action became indistinguishable. Thus if hydrogen were being displaced by a metal in acid, and the hydrogen were being evolved uniformly as far as even microscopic observation could ascertain, then there was no means of distinguishing between spatial separation and contact between the reacting bodies, since any cathodes and anodes must be indefinitely close. The authors, therefore, preferred to distinguish as electrochemical those cases in which a perfectly clear distincticln could be made between the cathodes and anodes. It might be noted that the adherent,s of the electrochemical view of corrosion claimed that such distinchion would be readily made by means of the ferroxyl test.” This seems t o explain what is otherwise a meaningless paragraph.I “The oft-repeated general statement, that metals of a high degree of commercial purity are less readily corroded by neutral media than more impure metals (a statement that is usually put forward as a proof that corrosion is electrochemical, though it is also explainable on other grounds), can only be accepted with great reservations. It is true that Ramsay and. Reynolds’ highly purified zinc was only slightly attacked by acid, unless a trace of platinum black was added t o it. The platinum may, however, have affected the rate of corrosion by facilitating the chemical evolution of hydrogen at the surface of the metal owing t o the lowered over-voltage rather than by the formation of a couple.” These two things are identical and not antagonistic as Bengough and Stuart would have us believe. The platinum acts by forming a couple which 1 “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 128 (1922). 1 Bengough and Stuart: ‘[Sixth Report to the Corrosion Research Committee of the Institute of Metals,” J. Inst. Metals, 28, 59 (1922).
THE ELECTROLYTIC THEORY O F CORROSION
827
permits the evolution of hydrogen at the platinum cathode because of the low over-voltage there. The proof of the pudding is in the eating. The Sixth Report does not incline one to accept the views of Bengough and Stuart. Some experiments on the action of metals on nitric acid are now being carried on a t Cornell. When these are finished, there will be available some very interesting evidence of the effectiveness of the electrolytic theory of corrosion as a working hypothesis. Bengough and Stuart will scarcely claim that the consideration of this problem as a purely chemical one has been successful in giving us a satisfactory theory of the phenomena. Dunstan, Jowett, and Gouldingl made some Experiments to determine what metals gave hydrogen peroxide when corroding in presence of oxygen. “The action of oxygen and water on several metals in the presence of a trace of sulphuric acid, which would promote the liberation of hydrogen, was now examined in the following manner. The metal was placed under distilled water containing a trace of sulpburic acid, and oxygen was bubbled through the liquid for some time. The mixture was then shaken and tested for hvdrogen peroxide from time to time. It was found in all these instances except in that of iron. A parallel series of experiments was carried out omitting the sulphuric acid, and, except in the case of zinc, no hydrogen peroxide could be detected. The results of the experiments made in the presence of acids were as follows : Metal. Result. Copper. After 48 hours, the liquid was of a faint blue colour, and gave a distinct reaction for hydrogen peroxide. Mercury. After 2 hours, a distinct reaction for hydrogen peroxide. Silver. After 24 hours, a trace of hydrogen peroxide was found. Lead. The liquid became milky at once, and gave a well-marked reaction for hydrogen peroxide. Bismuth. A well-marked reaction for hydrogen peroxide. Tin. The reaction was not so well marked as in the case of bismuth. Zinc. A well-marked reaction for hydrogen peroxide. Iron. Although examined from time to time, no hydrogen peroxide could be detected. “Thus, hydrogen peroxide was found in every case except in that of iron. If hydrogen peroxide were produced in presence of iron, it would be at once decomposed. “In order to ascertain whether hydrogen peroxide would attack iron in the presence of a substance which inhibits ordinary rusting, metallic iron was introduced into solutions of borax and lime respectively, and hydrogen peroxide was then added. The iron in these cases remained unattacked although the peroxide was decomposed and oxygen was evolved. If a plate of bright and highly polished steel is immersed in a strongly alkaline solution of hydrogen peroxide, decomposition of the hydrogen peroxide is extremely rapid and J. Chem. Soo. 87, 1560 (1905).
828
WILDER D. BANCROFT
bubbles of oxygen are liberated on the surface of the steel, yet no rusting occurs if the alkaline solution is fairly strong.” All this is grist to the electrolytic mill and this is not the only slip that Dunstan, Jowett, and Goulding made. One of their reasons for rejecting the electrolytic theory of corrosion was that ih would not account for the fact that potassium bichromate solution prevents the rusting of iron. Bengough and Stuart1 make much of the same fact. “The action of bichromate solutions in passivifying metals is not readily explained on electrochemical lines, since bichromate is a powerful depolarizer, and would be expected greatly t o enhance any electrochemical action that took place in distilled water (as indeed it does in certain acid solutions). , . . An interesting point arises in connection with the function of known depolarizers, such as potassium bichromate. In electrochemical corrosion, as, for instance, in the case of the primaries, these substances powerfully promote corrosion of the anode. But when metals such as iron, zinc, and copper, and even magnesium, are placed in distilled water containing bichromate, corrosion is inhibited .” Evans2 has objected t o the second part of the quotation as misleading because the bichromate is not added in contact with the corroding metal in the primary cell whereas it is in the corrosion experiments, “It was true that the presence of chromates at the cathodic portions of the corrosion couples would be favorable t o the reaction; but at the anodic portions they would stop the reaction altogether. Since the cathodic action could not proceed without the anodic action, corrosion would cease. Many workers had shown that if an electrolytic cell fitted with an iron anode dipping in a neutral non-oxidizing salt solution were joined to an external battery iron was dissolved at the anode; but the addition of a little chromate to the solution stopped the anodic dissolution of the iron altogether. To this comment the only reply that Bengough and Stuart made, (p. 127), was that “the writers had been very interested in Mr. Evans’ explanation of the behaviour of bichromate in inhibiting corrosion. He suggested that it was a case of anodic polarization produced by the formation on the anodic surface of some form of oxide. But surely this was a case of direct oxidation such as was postulated all along by the authors.” Unfortunately, Bengough and Stuart did not explain what they meant by their last sentence. As a matter of fact we shall see that Bengough and Stuart were wrong and that Evans was right. It would have been interesting if Evans had carried the war into Africa and had pointbedout that bichromate is not a strong depolarizer at the concentrations used in the corrosion experiments. One would think, t o read the comments by Rengough and Stuart, that the upholders of the electrolytic theory of corrosion either knew nothing about the action of bichromate or avoided all mention of the painful fact. Nothing 1 “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” Jour. Inst. Metals, 28, 52, 74 (1922). 2 Discussion of “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 129 (1922).
T H E ELECTROLYTIC THEORY OF CORROSION
829
could be further from the truth. Cushman and Gardner’ pointed out that “as has already been noted, solutions of chromic acid and potassium bichromate inhibit the rusting of iron. I n order to determine the concentration necessary to produce complete protection, a number of polished strips of two different samples of steel were immersed in bichromate solutions of increasing concentration, contained in tubes which were left quite open to the air. There were twelve tubes in each series, ranging by regular dilutions from tenth-normal down t o ten-thousandth normal. At the end of two months the last four tubes showed graded rusting with accumulations of ferric hydroxide. No rusting had occurred in any of the solutions above tube No. 8, which contained six-hundred-and-fortieth normal bichromate, a strength corresponding t o about 8 parts of the salt in IOO,OOO parts of water, or about 2 pounds t o 3,000 gallons. Since solutions of bichromate do not hydrolyze with an alkaline reaction, but on the contrary are usually slightly acid, some other explanation must be found for this remarkable phenomenon. On first thought it would seem a paradox that a strong oxidizing agent should have the effect of preventing the oxidation of iron, and yet this is precisely the case. If, however, the initial cause of rusting is the hydrogen ion, it is possible to believe that under certain conditions oxygen would prove the most effctive inhibitors. “One of the authors has observed that, if a rod or strip of bright iron or steel is immersed for a few hours in a strong ( 5 to I O percent) solution of potassium bichromate, and is then removed and thoroughly washed, a certain change has been produced on the surface of the metal. The surface may be thoroughly washed and wiped with a clean cloth without disturbing this new surface condition. No visible change has been effected, for the polished aurfaces examined under the microscope appear t o be untouched. If, however, the polished strips are immersed in water it will be found that rusting is inhibited. An ordinary, untreated, polished specimen of steel will show rusting in a few minutes when immersed in the ordinary distilled water of the laboratory. Chromated specimens will stand immersion for varying lengths of time before rust appears, but the induced passivity gradually disappears. “The passivity which iron has acquired can be much more strikingly shown, however, than by the rusting effect produced by air and water. If a piece of polished steel is dipped into a one percent solution of copper sulphate, a Io-second immersion is sufficient to plate it with a distinctly visible coating of copper which cannot be wiped off2. A similar polished strip of steel which has been soaked over night in a concentrated solution of bichromate and subsequently well washed and wiped will stand from six to ten IOsecond immersions in I percent copper sulphate before a permanent coating of copper is deposited. Even a momentary plunging of the metal into the bichromate will induce a certain passivity, but the maximum effect appears to require a more prolonged contact with the solution. . . . It has been already pointed out in a previous chapter that in many cases a stimulative “Corrosion and Preservation of Iron and Steel”, I I I (1910). This experiment has failed in hands of certain experimenters who have not been careful t o use the copper sulphate solution as dilute as directed.
830
.
WILDER D. BANCROFT
and an inhibitive tendency may be at work a t one and the same time. This assertion is well brought out by the following experiment in which an inhibitoi and stimulator are literally ‘pitted’ against one another. Samples of bright steel wire were immersed in IOO cubic centimeters of a very dilute one-thousandth normal solution of potassium bichromate in a series of shallow dishes. The wire test pieces were suspended in the solution so that they did not come in direct cont,act with the glass surfaces of the dishes. This precaution should never be omitted in experiments of this kind, as owing to the absorption of air by glass, rusting is always stimulated at the point of contact between glass and iron. The first dish was left as a blank, the second received one drop equal to 1/20 cubic centimeter of a dilute tenth normal copper sulphate solution. The third dish received two drops of the solution, and so on, each dish getting an increased amount of copper sulphate until twenty-five dishes had been prepared. “Now it is apparent that we have in this system two contending forces at work. Iron has a higher solution tension than copper, and therefore tends to pass into solution, the copper tending 60 plate out on the iron. Chromate ions, on the other hand, put the surface of iron in a condition in which it cannot pass into solution. In the solution system iron-chromate-copper we have an equilibrium to be decided between two contending forces acting in opposite directions. It was interesting and instructive to note the results of this struggle, which was known to be going on, although the actual conflict could not he watched. In the first dish, in which no copper was present, no corrosion took place; in the second, also, no action was visible; in the third, however, minute specks of iron rust appeared. Theee were larger and more frequent in the immediately succeeding dishes, the test-pieces showing rust tuberculation with the well-known pitting effect. As the middle of the series of dishes was approached, both iron rust and precipitated copper began to appear side by side on tho surface of the iron, and from thereon in the series more and more copper separated, while less and less rust formed, until in the end dishes copper and iron were changing places evenly over the surface without apparent hindrance. These experiments, and other of a similar nature, were repeated many times with the same results, and there seems to be no escape from at least the following two conclusions to which they obviously lead: . “(I) If the surface of iron is subject to the action of two contending influences, one tending t o stimulate corrosion and the other to inhibit it, the result will be a breaking down of the defensive action of the inhibitor at the weakest points, thus localizing the action and leading to pitting effects. “ ( 2 ) While the concentration of an inhibitor may be strong enough to prevent the electrolyte exchange between atom and ion, it must be still stronger to prevent entirely the solution of iron and the subsequent oxidation which leads to the formation of rust-spots.” Dunstan and Hill1state that all agents which inhibit rusting render passive the metals studied by them, Among the solutions acting in this way was a one percent solution of potassium bichromate. This was checked qualitatively J. Chem. SOC.99, 1835, 1855 (1911).
THE ELECTROLYTIC THEORY O F CORROSION
83 1
by Miss Soudersl at Cornell who suspended strips of magnesium, zinc, iron, and copper in a bichromate solution for several weeks. The surfaces of all these metals remained practically untarnished. If the electrolytic theory of corrosion is correct, this should mean that these metals would not undergo corrosion when made anode in a bichromate solution under suitable conditions. The solution used in the electrolytic experiments was M/40 K2Crz07which is a little less than one percent. Voltage curlqent curves were determined for the cell MI Kz Crz07[ Pt. With iron there was visible evolution of oxygen and hydrogen at 2.2 volts and the decomposition voltage was found by extrapolation to be about 2.0 volts. The evolution of hydrogen at the cathode shows that potassium bichromate is not a powerful depolarizer at this concentration. The hydrogen was not determined experimentally, so we cannot say definitely that there was no depolarization; but the curve was absolutely normal in every respect so it seems that the amount of hydrogen oxidized was negligible. This matter is now under investigation a t Cornell. The iron anode did not corrode perceptibly, so there is perfect agreement between the ordinary and the electrochemical phenomena. In order to study the effect of the addition of sodium chloride, varying amounts of M / ~ NaCl o solution were substituted for the M/40 K2Cr20i solution. The decomposition volbage decreased with increasing relative amounts of sodium chloride and the corrosion of the anode also increased. The iron anode did not corrode perceptibly in the solution containing g g parts bichromate to one of salt and the form of the decomposition voltage curve changed but slightly. With 75 parts bichromate to 25 parts salt, the decomposition voltage is roughly 0.5 volts by extrapolation. No experiments were made with a bichromate solution diluted with water alone; but a whole series of experiments is planned for different concentrations and temperatures. In order to compare the anodic corrosion with ordinary corrosion, strips of iron were immersed in solutions identical with those used in the decomposition-voltage experiments. At the end of a week the iron had corroded in all the mixtures except the one containing g g parts bichromate and one part of salt, and the amount of corrosion varied directly with the concentration of the sodium chloride. Copper, zinc and magnesium corroded when made anodes in the potassium bichromate solution, which was distinctly not according to Hoyle. It was noticed, however, that the cell Cu KZCrz07 Pt had an electromotive force of approximately 0.3 volts when measured on open circuit with a potentiometer; but no current flowed when the cell was short-circuited on a milliammeter. No corrosion took place when a platinum wire was wrapped tightly round a copper wire and the whole was immersed in a bichromate solution, while marked tarnishing was perceptible in a few hours when a similar pair of wires was immersed in a M/4o Na2SOl solution. This showed that copper became passivc for low current densities and active when a sufficienb voltage was impressed. Miss Souders measured the anode potentials of iron, copper, zinc and magnesium in M/4o K~Cr207at the decomposition point and found that lhey
I
I
Unpublished thesis.
I
83 2
WILDER D. BANCROFT
were all much more noble than they should be. The actual data will be published some day in her thesis. At low voltages, which is what we deal wit’h in corrosion experiments, the ordinary and electrolytic corrosions run parallel; but the passivity breaks down a t higher voltages for copper, zinc, and magnesium, just as it breaks down in the aluminum rectifier though the phenomenon is much more striking in the latter case. The difficulties imagined by Beagough and Stuart have vanished into thin air and we see that Evans was right in saying that the anode reaction was the important one in these cases. Alkalies are also inhibitors of corrosion; but the explanation of this action given by Cushman and Gardner,l Walker, and others is inadequate. “All substanc3es in solution which contain hydrogen ions, such as acids, stimulate the corrosion of iron. This is also true of salts of strong acids and weak bases, which, though perfectly stable in a dry condition, hydrolyze in solution t o an acid reaction; or which, though neutral in fresh solutions, undergo slow decomposition under the action of light, with the formation of acid salts or free acid. With certain exceptions, salts which are perfectly neutral in solution do not prevent oxidation but appear to aid it by increasing the electrolytic action. All substances which develop hvdroxyl ions in solution, such as the alkalies or salts of strong bases with weak acids, to a certain extent inhibit, and, if the concentration is high enough, absolutely prohibit, the rusting of iron ? “Under the electrolytic theory the explanation of the protection afforded by hydroxyl ions is a simple one. Owing to the small dissociation of water, hydrogen ions cannot exist in a solution i n which the hydroxyl ions are in excess.3 As hydrogen ions cannot exist or be locally formed in sufficiently st,rong alkline solutions, no attack is made upon the iron, which remains permanently unaltered. If, however, the concentration of the hydroxyl ions is not sufficiently great, electrolysis can go on with an apparent stimulation of the pitting effects similar to that produced by perfectly neutral electrolytes, such as sodium chloride.” This view is not consistent with the later generalization of Dunstan and Hill that all substances which inhibit rusting render the metal passive. Miss Souders finds that with increasing anodic polarization of zinc in a one percent caustic soda solution, the zinc is first active, then passive, and finally active again. I n a sodium chloride solution there is of course no evidence of passivity; but marked evidence of passivity is to be seen with a 0.9% NaCl+o.Iyo NaOH solution and the curve for 0.5% NaCl+o.5yo NaOH does not differ appreciably from that for 1% NaOH. An iron anode polarizes in 170NaOH to practically the same value as in M/40 Kz Crz07,about 0.7-0.8volts cathodic when the hydrogen electrode is taken as zero. Zinc is less noble in caustic “Corroaion and Preservation of Iron and Steel,” 110 (1910). Under exceptional conditions this statement may require modification. Iron has & considerRble solution tenRion in strong boiling alkaline solutions: but in such a case the equilibrium is reversed and the metal acts t,he part of a negativc radical in conjunct’ion with oxygen. This point is a t present being investigated. 8 [Cushman did not say exactly what he meant, which is that hydrogen ions cannot occur except in very low concentration.] 1 2
THE ELECTROLYTIC THEORY O F CORROSION
833
soda than in bichromate, while copper and magnesium are much more noble when polarized anodically in caustic soda than in bichromate. Bengough and Stuart1 are quite impressed with their observations on corrosion in distilled water and in sea-water. “Support for the view that the course of corrosion is not always determined by the conductivity of a solution can easily be found, however, even in acknowledged electrolytes. Thus metallic zinc is corroded from four t o six times as rapidly by stagnant distilled water, as by sea-water or 2.5 percent sodium chloride, both of which are excellent electrolytes. Nickel and lead are also more rapidly corroded in stagnant distilled water than in sea-water. It is clear, therefore, that the conductivity of the corroding liquid may be quite a minor factor in many corrosion phenomena, whereas the nature of the scale may be a major factor. . . . “This statement [in regard to over-voltage], however, does not take into account certain relevant experimental facts which are that distilled water will readily corrode zinc, even when highly purified, and no hydrogen is evolved; 2 . 5 percent sodium chloride solution when stagnant, will only attack the same sample of zinc much less slowly; extremely dilute weak acid, such as acetic acid, will also attack it readily, but with quantitative evolution of hydrogen. It follows from these facts that neither the over-voltage nor the depolarization of hydrogen by atmospheric oxygen have any important effect on the result, since the former is not able to stop corrosion and the latter cannot assist it. The important facts determining the rate of corrosion are of quite another kind. Over-voltage phenomena, in fact, afford little assistance in elucidating corrosion by water and salt solutions, except in those comparatively few cases in which hydrogen is displaced from solution, e.g. by magnesium from sodium chloride solutions.” Only reckless men would draw any conclusions from experiments on the corrosion of zinc in distilled water and in sodium chloride solution. Miss Souders made some experiments with ( I ) aerated distilled water, ( 2 ) aerated sodium chloride solution, (3) stagnant distilled water, and (4) stagnant sodium chloride solution. At the end of five days the order of corrosion from highest to lowest was: aerated sodium chloride; aerated distilled water; stagnant water; stagnant sodium chloride. At the end of fifty days the order was stagnant sodium chloride; aerated sodium chloride; aerated distilled water; stagnant distilled water. I n all cases a film forms and a t the end of fifty days the corrosion ceased practically in the cases of aerated sodium chloride, aerated water, and stagnant water, even if one substituted fresh liquid. The stagnant sodium chloride was not carried t o an end; but it looked a t the end of eighty-five days as though the corrosion were nearly ended. The amount of corrosion depends on the way in which the film forms and that is probably a function of many factors, many of them probably almost insignificant ones. I offer no explanation why the stagnant sodium chloride did so much better than any of the others; but I do feel t h a t data like these should not be put forward seriously as proving anything. “Sixth Report t o the Corrosion Research Committee of the Institute of Met.als.”
J. Inst. Metals, 28, 53, 80
(1922).
834
WILDER D. BANCROFT
Another phenomenon which seems convincing t o Bengough and Stuart’ is the behavior of magnesium. “A very formidable difficulty in the way of the electrochemical theory of corrosion is found in the effect of ions of the corroded metal on the rate of corrosion of a liquid containing them. The presence of these ions, in the case of such metals as magnesium, zinc, aluminum, iron, nickel, and tin, should decrease the rate of corrosion according tG the theory, by increasing the osmotic pressure which opposes the solution tension. It has, however, long been known that magnesium chloride is an extremely powerful corrosive agent for magnesium,2far more so than distilled water or the chlorides of the alkaline earths or heavy metals, and hydrogen is freely evolved from the metal and does not protect it a t all. Dhara has shown that the presence in solution of the ions of the corresponding metal increases the potential difference between the me tal and water of a neutral salt solution instead of lessening it, as required by the theory in the cases of magnesium, aluminum, zinc, and iron; nickel and tin, on the other hand, show a decreased potential difference.” Since magnesium cannot be precipitated from an aqueoua solution, a magnesium electrode in an aqueous solution of a magnesium salt is not a reversible equilibrium and one cannot apply the Nernst theory to it. This is the same sort of mistake that Carhart4 made over twenty years ago when he wished to consider nickel in nickel sulphate solution as a reversible electrode. This is, however, a minor matter and the really important thing is why magnesium corrodes more rapidly in a magnesium chloride solution than in distilled water. There is fortunately no dispute in regard to the facts. The explanation was suggested by Bryant6 the same year that Tommasi published his experiments. Bryant studied the action of alcohol, water, and sodium sulphate upon magnesium and found that the rate of reaction increased from alcohol t o sodium sulphate. “Magnesium has itself, I submit, the power t o decompose water a t all temperatures above O’C, and the action is stopped by an insoluble coating of oxide forming in the metal. This would e lain the change of color seen after the action has ceased, the renewal of action when the surface is cleaned, and the abundant evolution of gas from a solution of sodium sulphate, since magnesium oxide is more soluble in that than in pure water.” If Bryant had only known the magic word, peptization, he could have explained everything. Miss Souders immersed t,wo rods of magnesium, each IO cm long and I cm in diameter, in pure water and in one percent magnesium chloride solution respectively. The rod in the magnesium chloride solution became covered with white flakes6 which were easily removed; the solution was milky, and
v
1 “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 61 (1922). 2 Tommasi: Bull. 21, 885 (1899); Kahlenberg: J. Am. Chem. SOC.25, 380; Roberts
and Brown: 841 (1903); Getman: 38, 2596 (1916); 39, 596 (1917). * Z. anorg. Chem. 118, 75 (1921). 4 Trans. Am. Electrochem. SOC. 1, 105; 2, 122 (1922). 6 Chem. News, 80, 21 I (1899). Cf. Evans: J. Inst. Metals, 28, 117 (1922).
T H E ELECTROLYTIC THEORY O F CORROSION
83 5
a considerable precipitate settled to the bottom. The rod which had been in the water acquired a dull lusterless surface which was moderately smooth; the liquid was clear. The rods were then transposed, the one that had been in the water being placed in the magnesium chloride solution, and the one that had been in the magnesium chloride solution being placed in the water. The rod which was now in the magnesium chloride solution began to flake and become rougher, the film originally produced by the water being obliterated entirely. The rod which had been transferred to the water became more like the one which had started in the water, the surface seeming smoother and the coating more adherent. It never got really back because the original pitting had been too severe, There is no question, however, but that the two films are quite characteristic and that the film in water retards corrosion. For this discussion it is immaterial whether one says that magnesium chloride peptizes magnesium oxide or not. That is the simplest explanation because we know that zinc chloride peptizes zinc oxide, ferric chloride peptizes ferric oxide, and chromic chloride peptizes chromic oxide. The essential thing is that there is nothing mysterious about the corrosion of magnesium in a magnesium chloride solution. Any sceptic can repeat the experiments and satisfy himself. One reason why Bengough and Stuart1 have considered corrosion experiments in distilled water so important is that they believe that the conductance of the liquid medium should be a very important factor in determining the rate of corrosion. They say that a fact which is “difficult t o explain on a purely electrochemical theory” is that “the conductivity of electrolytes is not directly connected with the amount of corrosion.” There is some justification for this attitude. Walker2 stated that “the third factor [influencing the rate of corrosion] is the ease, due t o the lack of resistance, with which the electric current, generated by the solution of the metal a t one point and the separation of the molecular hydrogen a t another, can pass from one of these points to the other. They may be infinitely close together, or separated by quite a distance. Therefore, this theory demands that solution or corrosion should take pla& more rapidly in water in which there is dissolved a trace of an electrolyte, than in absolutely pure water.” Taken by itself, this paragraph might seem to support the contention of Messrs. Bengough and Stuart; but this is one of the statements that everybody understands t o mean that the rate of corrosion increases with the conductance provided other things remain equal-which they rarely do. I n the preceding paragraph Walker had mentioned that caustic soda inhibits the corrosion of iron and that he had shown the inadequacy of Friend’s experiments by showing the presence of caustic soda on the iron by means of conductance experiments. Walkel.3 has covered the case explicitly himself in a later paper. The factors which influence corrosion are: “third, the ease with “Sixth Report to the Corrosion Research Committee of the Institute of Metals.”
J. Inst. Metals, 28, 32 (1922).
2Trans. Am. Electrochem, SOC.14,178 (1908). Trans. Am. Electrochem. SOC.29, 436 (1916).
a
836
WILDER D. BANCROFT
which the hydrogen ion reaches the iron. For example, the wonderful resistance of sheet steel containing small quantities of copper is unquestionably due to the adherent nature of the film of oxide produced when corrosion starts. This film, as in the case of the alumina coat on metallic aluminum, effectually msulates the metal and practically prevents further corrosion.” This seems to be an appropriate time to introduce two quotations from W. D. Richardson,‘ just to show that people have recognized the existence of factors which would modify the conclusions which one might draw from the electrolytic theory of corrosion in its simplest form. “The solution tension of a metal which determines its position in the electrochemical series, one might suppose would be a dominant factor in determining the rate of solution in an acid, as well as the rate of corrosion. Abundant evidence shows that it does neither of these things. It represents only a tendency to go into solution, a tendency which under the ordinary conditions of the experiment is so quickly brought to a halt by various influences, on the other hand so readily accelerated by catalysts, and so seldom realized, that the actual order of corrodibility and solution of metals is quite at variance with the order in the electrocheniical series. It is so easily overbalanced by polarization with hydrogen, by the influence of oxygen with or without the resultant formation of a passive layer, by the effects of impurities in the metal, and by the influence of the products of corrosion, that it is rarely the determining factor.’) “Matters do not occur according to the simple idea or the simple conception of the electrolytic theory of corrosion as advanced in the early days, and bhat is what I meant when I spoke of the older view of it. The recent view, and ib is coming more and more into prominence, is that while this older view of a simple electrolytic theory is at the basis of all corrosion phenomena, there are so many modifying, upsetting and antagonizing influences that it does not take a straight course as we might expect if we had no further knowledge of it. Let me call your attention to one upsetting influence, rust, which I have emphasized over and over again, may accelerate in a general way, the corrosion of iron. If it forms a closely adherent layer of the physical nature of a paint film over the entire surface, it may stop corrosion by preventing the access of water and oxygen, and if it is entirely adhesent in patches, it may cause deep pitting. Now here we have the same general influence acting in different ways, in all instances upsetting the ordinary action of the corrosive substances and elements.” “I believe we have another influence to contend with which may act in different and antagonistic ways, namely oxygen. We know that oxygen is the chief disturbing influence in ordinary corrosion; it is the principal substance which causes corrosion t o go in some other way than we might expect from the simple theory. We know that the corrosion of iron goes on at practically the zero rate in the absence of oxygen and in the presence of water, and we know it may go on very rapidly in the presence of pure oxygen and less rapidly in air. We know, too, that the influence of oxygen or an oxidizing agent on the 1
Trans. Am. Electrochem. Soc. 38, 248
(1920);
39, 162 (1921).
THE ELECTROLYTIC THEORY O F CORROSION
83 7
surface of iron may be such as to produce the passive condition, and when this condition supervenes, the iron behaves as monel metal and there is no corrosion a t all. I have been led t o suspect by some phenomena, that I have mentioned in an earlier paper in connection with some work on cast iron, that a modified passivity may supervene in instances where we have not suspected it. I do not know that this theory explains it, but in the case of cast iron we have some very peculiar rust-resisting properties, and it may be under certain conditions the iron coupled with graphitic carbon and under oxidizing conditions may assume, temporarily at least, or from time to time, or intermittently, a passive condition. I n discussion with Dr. Burgess last evening on the subject of copper steel, I spoke with him about the experiments which I had made, adding copper to normal nitric acid, with the result that the corrosion rate was lower, and he suggested that possibly in that case passivity was induced. I then spoke to him about the effect of graphitic carbon or cast iron, and it may be that in the case of the copper steel, we have an induced passivity, or intermittent passivity, which may account for the action of copper in that instance. I think none of us can afford to be at all dogmatic in our discussions of corrosion, and particularly when we are dealing with a metal like copper of proved peculiar behavior.” At the risk of appearing dogmatic, it seems that Bengough and Stuart’ are still living in the era of Cushman and Lambert, besides being handicapped by a very rudimentary knowledge of electrochemistry. “If gold be placed in a beaker of chlorine water it will be slowly attacked, and pass into solution forming gold chloride. Since gold cannot displace hydrogen from water or hydrochloric acid, and is not attacked by the chlorine ions existing in such easily reducible solutions as cupric chloride, it is reasonable to assume that the reaction is not ionic but molecular and that the non-ionized chlorine molecule which exists in the chlorine water directly attacks the gold. z Au+3C12+2 uCb liq.(+54338 cal.). This view of the action is suggested by the position of gold and chlorine in the electrochemical series; they occur together a t one end, and would therefore tend t o combine by virtue of their non-ionized valencies rather than by their ionized valencies. The fact that the resulting solution of the gold contains complex ions such as (AuCl~O)” in water, and (AuCl,)’ in dilute hydrochloric acid also suggests that the initially formed molecular compound does not entirely rearrange itself so as to lorm the simple Au‘ ‘ ‘and C1’ ions, which would be expected from the ionic formation of AuCb. “Thus, it seems reasonable t o assume that the gold has passed into solution by purely chemical action. It is possible, however, t o arrange that the gold shall pass into solution electrochemically. If the chlorine water be replaced by sodium chloride solution, no gold will dissolve; but if the gold be electrically connected to a piece of platinum placed in another beaker containing sodium chloride, and the two beakers be connected by a siphon and chlorine water be poured over the platinum, then the gold will dissolve and a
Er
“Sixth Report to the Corrosion Research Committee of the Institute of Metals.”
J. Inst. Metals, 28, 44
(1922).
83 8
WILDER D. BANCROFT
current will pass. The action is now carried out electrochemically, since there is spatial separation of the reacting bodies and free energy has appeared as an electric current. Gold has been oxidized at the anode, thus: 2 Au-2 Au‘ ‘ ‘, and the chlorine has been reduced at the cathode thus: ClZ---+Z C1’ A similar result has been reached as in the chemical process, by a different route and at a different speed. “A similar series of experiments could be carried out with platinum and it seems fair to assume that certain metals, at any rate, may be corroded either chemically or electrochemically, and that the actual conditions of the experiments might determine in which way the action would proceed.” Bengough and Stuart have been fair in admitting explicitly that it is only an assumption that gold reacts chemically with chlorine under these circumstances. What their experiments prove is that gold can react with chlorine electrochemically and that the same reaction takes place wheu there is no spatial separation. If one is t o assume, as they do, that the whole type of reaction changes when the anodes and cathodes are very close together, one would like to know at what point the change in type occurs. It is quite possible t8hatBengough and Stuart would say that the change came when the chlorine was in contact with the gold and not with the platinum; but that will not do, because the same results would have been obtained with a gold cathode provided the resistance of the circuit was kept low enough. There is nothing to show whether they italicized the words “the platinum”, because they were surprised that the chlorine should be added there or because they thought that their readers would be surprised; but there seems to be no justification for either thing because the general principle was explained in detail by Ostwaldl over thirty vears ago. “As is well-known, nmalgated zinc is not attacked by dilute acids, but it dissolves in acid with evolution of hydrogen if a piece of platinum wire is wrapped round the zinc. Zinc wrapped with platinclm does not dissolve in solutions of neutral salts but does if a few drops of acid, sulphuric acid for instance, be added to the salt solution. For the platinum to have this effect it is only necessary for it t o be in contact with the zinc at a single point. If one makes a staple out of zinc and platinum with the two ends only a little way apart and dips this into potassium sulphate solution so that a diaphragm of porous material, such as earthenware or parchment paper, separates the two ends, it is a simple matter t o determine which of the metals, zinc or platinum, must come in contact with the acid in order t o make the zinc corrode. “At first sight, the question seems absurd; for if the zinc is to corrode, it seems self-evident that the acid should be added to it. If one tries the thing out, however, the opposite proves t o be true. When one adds acid to the potassium sulphate solution in contact with the zinc, no zinc dissolves except traces which would have gone into solution anyhow; it dissolves rapidly and there is a plentiful evolution of hydrogen when the liquid round the platinum 12;.
physik. Chem. 9,540 (1892).
THE ELECTROLYTIC THEORY O F CORROSION
83 9
is made acid. The hydrogen comes off at the platinum as is always the case when zinc is in contact with platinum. I n order to dissolve sinc under these conditions, it is therefore necessary to add the attacking substance to the platinum which is connected with the zinc and not directly to the zinc which is to be attacked.” Bengough and Stuart state that it is reasonable to assume “that the nonionized chlorine molecule which exists in the chlorine water directly atta.cks the gold.” They have evidently forgotten that Wohlwill‘ has shown that a gold anode does not corrode in a neutral auric chloride solution or in a solution of chlorauric acid, HAuC14. Chlorine is evolved at the anode, which seems to dispose pretty effectually of the assumption that chlorine reacts direct with massive gold. The anode does corrode when there is an excess of hydrochloric acid so that chlorauric acid can be formed. While Bengough and Stuart are quite right in saying that gold cannot react with hydrochloric acid, displacing hydrogen, they have overlooked entirely the fact that gold can and does react with hydrochloric acid when chlorine is present as a depolarizer. Just as the corrosion or non-corrosion of iron in presence of oxygen is to be considered as the action of a cell, Fe I solution 02,so the corrosion is to be considered as the action of a cell, Au HC11 Clz. Coehn and Jacobsen2 have studied the anodic behavior of gold in gold chloride and in hydrochloric acid solutions and find that the gold anode becomes passive a t current densities which become higher as the concentration of hydrochloric acid increases. Bengnugh and Stuart3 devote nearly six pages to the behavior of copper electrodcs in sodium chloride and cupric chloride solutions, so they presumably consider these experiments as significant. They placed a half-normal cupric chloride solution inside a porous cup and a three percent sodium chloride solution outside. In these two solutions they placed two similarly shaped copper electrodes (99.96 percent copper) and connected them metallically so as t o form a concentration cell, the electrode in the sodium chloride solution being of course the anode and the one in the cupric chloride solution the cathode. At the end of twenty-four hours, the electrodes were removed, examined, dipped momentarily into strong hydrochloric acid t o remove cuprous chloride, and then weighed. The anode lost 6.93 grams and the cathode had gained half a gram which they saw as crystals of copper. They do not realize that this is conclusive proof of bad experimentation. Under the conditions which they describe and thought they bad, no copper could precipitate on the cathode until practically all of the cupric chloride had been reduced to cuprous chloride and that they say was not the case. It is quite clear that a film of cuprous chloride formed over the cathode and that for a portion of the time no cupric chloride was in contact with the cathode. At the time when they started the electrolytic experiment, they also placed similar pieces of copper, not electrically connected, in each of the solutions.
1
I
1
Z. Elektrochem. 4, 381 (1898).
3
“Sixth Report to the Corrosion Research Committee of the Institute of Metals.’’
* Z. a n o g . Chem. 55, 330 (1907). J. Inst. Metals, 28, 45 (1922).
840
WILDER D. BANCROFT
At the end of the twenty-four hours the results were just what everybody would have predicted, a thin film of cuprous chloride on the piece of copper in the sodium chloride solution and a thick film of cuprous chloride on the piece of copper in the cupric chloride solution. After being cleaned, the copper in the sodium chloride solution had lost 0.13 grams and the one in the cupric chloride solution 3.0 grams. Most people would consider the reaction between copper and a cupric chloride solution as essentially electrolytic; but not so Bengough and Stuart. Having assumed that the corrosion of copper in a cupric chloride solution is chemical, partly because no current is generated and partly because the corrosion is fairly uniform over the surface of the copper, they say that “it seems fair to draw the conclusion that if copper is to be corroded electrochemically without the existence of an external electromotive force, it must either not be placed in the cupric chloride at all, or else it must be so placed that a concentration cell effect may be produced. If it be placed in a homogeneous solution of cupric chloride, so that the essential condition of chemical action, namely, contact, be fulfilled, then it may be predicted that it will be initially corroded chemically, and energy will be set free as heat, since the action is exothermic. If the reaction products be not uniformly distributed, concentration cells may arise in course of time. From these experiments, it is clear that a metal can be corroded either chemically or electrochemically according to the conditions prevailing in its neighbourhood-conditions which are independent of the metal itself. When conditions specially favour the electrochemical type, the action may be more rapid than with the chemical type; when, however, they do not, then the chemical action may be more vigorous. For instance, if a considerable resistance be introduced into either or both of the conducting circuits, then the electrochemical action may be reduced approximately to aero.’’ The experiments prove nothing in regard to the chemical corrosion of copper. It was assumed that the corrosion of copper in cupric chloride solutions is chemical corrosion. If one is to assume that, there is no point in the experiments because everybody admits that metals can be corroded electrolytically. The only point worth noticing is that the electrolytic and the socalled chemical corrosions give identical products when the experiments are done carefully, though not when done by Bengough and Stuart. A point of view which does not enable one to detect obvious errors does not have much pragmatic value as a working hypothesis. After this unfortunate episode with cupric chloride Bengough and Stuart1 say that “many other cases might be quoted which are difficult or impossible of explanation on the assumption that no direct chemical action can take place between the metal and the liquid in which it is placed; only two or three will be mentioned. Thus, if copper be electrically connected with aluminum in nitric acid (I of acid: 3 of water) the electromotive force of the couple is about 0.47 volt, the aluminum being the anode and the copper the cathode. Never1
“Sixth Report to the Corrosion Research Committee of the Institute of Metals.”
J. Inst. Metals, 28, 5 1
(1922).
THE ELECTROLYTIC THEORY O F CORROSION
841
theless, the copper rapidly reduces the nitric acid in its neighbourhood and passes into solution with solution of oxides of nitrogen. The aluminum is also slightly, but much less attacked. In such circumstances it seems clear that the nitric acid is chemically reduced by the copper, which is simultaneously oxidized or corroded. If brass be electrically connected t o metallic iron and the two be placed in strong cupric chloride solution, the iron will be the anode and the brass the cathode. The action, however, is not merely a displacement of copper by iron, since the cathode is gradually corroded away, or converted into a mass of loose copper crystals. It is not possible t o explain either of the foregoing examples of the corrosion of the cathode on the assumption that such corrosion is due to local couples formed on the surface of the cathode, since the potential of every point of the surface of the cathode is below that of t h e liquid, and consequently no metal can pass into solution electrochemically; moreover, the action can be obtained with highly purified copper substitut~edfor the brass, the corrosive attack on which can be shown to be independent of the distribution of any impurity present This is the same fallacy, that a pure metal is incorrodible, plus the explicit assumption that local action is always chemical. If the corrosion of copper by nitric acid is chemical and not electrolytic, the case is proved and there is no need of going any further. If the corrosion of copper by nitric acid is electrolytic, it is still electrolytic even though we superpose another electrolytic action. This corrosion of the cathode used t o be trotted out as an argument against the chemical theory of the voltaic cell’ nearly a hundred years ago and it bas rather lost its effectiveness as a bugbear. Fechner2 set up a cell, Zn/HzO/Cu/acid/Zn. Although the acid attacked the impure zinc of those days wit’h violent evolution of hydrogen, the zinc in the water solution was the anode. Fechner considered this a conclusive experiment against the chemical theory of the voltaic cell and now we have a cell involving practically the same principle cited as a conclusive experiment against the electrolytic theory of corrosion. Bengough and Stuart3 state that “the fact that local action is quite independent of the existence on metallic surfaces of cathodes and anodes is shown by the fact that any selected portion of a metal can be caused to suffer heavy local corrosion if the conditions external t o the metal are suitably controlled. A simple way of showing this is to tie a piece of ordinary string round a piece of copper or brass and immerse the whole in sea-water. Active local corrosion will take place beneath the string in spite of the fact that access of oxygen t o the corroded area is apparently greatly lessened. Local corrosion at any selected spot can also be produced beneath cotton-wool, coke, glass fragments (if not in too fine a state of division), paraffin wax (whenever liquid can penetrate beneath the wax,) and many other bodies. No such action takes place, however, beneath a deposit of red lead, which is an excellent depolarizer. Any purely electrochemical explanation of these phenomena is ruled out by the
.”
Ostwald: “Elektrochemie,” 485 (1896). Schweigger’s Jour. Chem. Physik, 57, g (1829). [‘Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 66,go, 98 (1922). 1
a a
842
WILDER D. BANCROFT
fact that they can be brought about by such insulators as glass and paraffin wax.” The authors’ explanation of this (p. 98) is very obscure. In the first paragraph they say that “the preceding theory indicates that the ordinary locaI corrosion essentially consists in direct oxidation of the metal by dissolved oxygen. The reaction may be simply oxidation of the massive metal; but possibly the metal surface is then disintegrated by the corroding liquid to form colloidal metal and then oxidized. The resulting colloid acquires a charge under conditions whereby the metal holds the equivalent opposite charge, and these charges are only neutralized in presence of electrolytes. The colloid is thereby precipitated by the anion, and the corresponding cation discharged on the metal surface.” I understand the first sentence of t,hat paragraph and believe it to be wrong. I do not understand the last sentence but it sounds like reversed electrolysis and neither the authors nor I believe in that. While we have never done any work on this type of localized corrosion a t Cornell, I am quite willing t o accept the explanation offered by Evans (p. 118) that the oxygen-free portion is normally the anode and the oxygen-rich portion is normally the cathode. Bengough and Stuart’ have produced corrosion “without the use of any substance other than water and air.” They let a jet of water or salt solution carrying entangled air bubbles impinge on the metal. They find that a strip of iron corrodes on the side away from the jet, which means that the oxygenpoor portion becomes anode and the oxygen-rich portion becomes the cathode. This is exactly what one would expect on the basis of the electrolytic theory of corrosion. If we are dealing with a gas cell, the oxygen must form the cathode. The only difficulty with this is that, copper behaves in the opposite way, the side where the jet impinges becoming anodic and not, cathodic as in the case of iron. It is not surprising that Bengough and Stuart should say that “clearly, it would be very difficult to form a general theory of corrosion on the lines of an oxygen concentration cell which took into account both these sets of facts,” and yet it is a pity that they did not study the matter further. When Bengough and Stuart showed me these experiment8 in London last year, I admitted, because I had to, that I had no explanation t o offer for the phenomenon; but the electrolytic theory of corrosion is right when properly applied and consequently I made the prediction that they had overlooked something, though I did not know what it was. On my return to Ithaca we started some experiments, varying the conditions. We placed two similar electrodes in a beaker containing M/I or M/z sulphuric acid, short-circuited the cell on a Weston milli-ammeter, and bubbled air against one of the electrodes. We confirmed the results of Bengough and Stuart in every respect. The electrode, against which air is bubbled, becomes cathode in the case of nickel and iron and anode in the case of copper. The differences of potential, 1 “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 66, 125 (1922).
T H E ELECTROLYTIC THEORY OF CORROSION
843
as measured roughly with a milli-voltmeter, were of the same general order 0.5-2 .o millivolts. Mr. Vieweg then had the brilliant idea of bubbling hydrogen against one electrode. Distinctly to my surprise, the results were qualitatively the same, the electrode, against which the gas bubbled, being cathode in the case of nickel and iron, and anode in the case of copper, the observed voltages being somewhat lower than in the case of air. Since the phenomenon appeared not to depend primarily on the nature of the gas, similar experiments were made with nitrogen and with illuminating gas with similar results. A few quantitative corrosion experiments were then made using a solution of 28 cc. ethyl alcohol and 60 cc concentrated sulphuric acid diluted to one liter with water. The two short-circuited copper electrodes were separated by a porous cup and the gas was bubbled against one of them. With nitrogen, the electrode against which the gas was bubbled lost 48 mg while the unexposed electrode lost 5 mg in seventeen hours. With oxygen the exposed electrode lost 2 0 mg and the unexposed one 2 mg in three hours. With hydrogen the corresponding losses were 2 0 mg and 3 mg in nine hours. The corrosion is of the same order of magnitude in all three cases; but the actual rate of loss in oxygen is two to three times as much as in either of the other gases. It is evident that we are dealing with frictional electrification, a phenomenon which has not hitherto been considered in corrosion research. It is worth noting that these results agree satisfactorily with Lenard’s observations on the electrification of drops of water in different atmospheres. One wonders whether the method of preparing the standard hydrogen electrode really gives. accurate values. It is of course possible that this effect may be negligible with platinized platinum; but the mere possibility of an error of the order of a millivolt is a bit startling. It is to be hoped that some of the people who specialize in accurate measurements will look into this. Having eliminated the electrolytic theory of corrosion to their own satisfaction in many cases, Bengough and Stuart’ state that ‘% remains to determine the nature of the reaction which does produce corrosion. The following considerations deal with the corrosion of single metals, and couples of which both metals corrode very much as when separated. Such couples in general show only small potential differences. It must be clearly understocd that electrochemical action does predominate in the corrosion of other couples, the metals of which corrode in a manner different from that of the separated metals, and which show considerable d;fferences of potential. “Leaving aside the latter, it is clear that there is some factor present in local corrosion the effect of which has become much more important than difference of potential. This factor is undoubtedly the ‘scale’. The term ‘scale’ has been defined as the mixture of solid, gelatinous, and colloidal substances which is formed by corrosive action and remains in the neighborhood of the metallic surface. Corrosion has already been defined as the oxidation of a metallic substance, using the term ‘oxidation’ in its general sense. Two “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 85 (1922).
844
WILDER D. BANCROFT
well-defined types of corrosion, differing markedly in their effects on the metal surface, can now be distinguished from one another: “(A) All corrosion products, except hydrogen and displaced gas or metals, completely soluble in the corroding liquid, giving true solutions. In such cases the metal surface is comparatively evenly attacked, and the solutiontension theory gives a reasonably correct account of the phenomena observed, though a strictly electro-chemical application of it is not necessary. Examples: sodium in water and zinc in hydrochloric acid. “(B) One or more corrosion products comparatively insoluble in the corroding liquid and closely adherent t o the metal surface. Metal surface usually attacked locally, giving isolated ‘pits’. “It is important to notice that to obtain well-defined corrosion of type B the product must adhere closely to the metal. Thus aluminum amalgam in distilled water shows the f i s t type of attack (type A), since the aluminum hydroxide, though insoluble, does not adhere to the metal surface. “With some metals, e.g. sodium, aqueous solutions always produce the first type of attack, since sodium hydroxide and sodium salts are readily soluble. Most of the metals in commercial use, however, can be made t o undergo either type of corrosion by varying the nature of the corroding liquid. Thus zinc is fairly evenly attacked by dilute acids, even by very dilute acetic acid (type A), but in distilled water becomes deeply pitted and covered with gelatinous zinc hydroxide. I n neither type of corrosion is there any definite evidence that the action at the metallic surface is electro-chemical in character.’, “The points in connection with local corrosion have already been fully considered, and will now be briefly summarized: “The corrosion of zinc in distilled water may be taken as an example of local corrosion (type B), and t,he following facts should be noted: “I. No evolution of hydrogen can be observed, although considerable corrosion takes place. This is in sharp contrast to the behaviour of zinc in very dilute acid, where the evolution of hydrogen is quantitative. “ 2 . Pits rapidly develop, and the surface of the metal round each pit becomes covered with gelatinous adherent deposits of zinc hydroxide, which often protect the surrounding metal from corrosion for considerable periods. “3. If the pits be examined under the microscope while the metal is still wet, they will appear to contain nothing but clear transparent liqnid. The bright metallic surface at the bottom of the pit is clearly visible. “The most obvious feature of such corrosion is that intense action takes place a t a particular spot, producing a pit in the metal, while the surface immediately surrounding this pit remains comparatively unat tacked. Superficially considered, this observation seems to support the electro-chemical view very strongly, the pit being regarded as an anode, and the surrounding metal as the corresponding cathode. But it has been noted that this view will not bear close investigation, since, for instance, no such local action occurs in dilute acids, although the action at the pit in distilled water may be more rapid than that occurring anywhere over the surface in acid.”
T H E ELECTROLYTIC THEORY O F CORROSION
’
845
We can agree heartily with Bengough and Stuart that the question of surface film or scale is all-important i n the study of corrosion. After the smoke of battle has cleared, that will be the point on which we ought all t o get together, to study the conditions affecting the formation and properties of surface films on metals. The work must be done definitely and specifically in each particular case and we must keep clear of the vague phrases which are painfully common a t present. While agreeing absolutely with Bengough and Stuart that the study of film formation is the vital thing in the problem of corrosion, it must be recognized that there is nothing strikingly new about this, though most of the research in the past has been along different lines. Cushman and Gardner’ have a few pages on the protection of iron by the production of surface films of magnetic oxide or other substances. I have myself called attention to the bearing of surface films on corrosion2and this has been emphasized by Curry? Walker4 cited the case of aluminum and of copper-bearing steels. E. A. Richardson and H. K. Richardson5 point out that “while it may be true that the initial rusting is largely electrolytic in character, other factors, such as the adherence of the rust and the protection thereby given to the metal, come into operation and outweigh any electrolytic corrosion-a conclusion that has also been arrived at by other observers. W. D. Richardson6 has recognized three types of rust. My experience’ at the same meeting was that people were only too glad to take up the detailed study of protective films and, in fact, it was as a result of that symposium* that the present Corrosion Committee of the National Research Council was formed and that this report was written. While we know very little about the properties of surface films, there is one case which calls for special mention. Bengough and Stuartg point out that “as regards both copper and brass, the usual effect of cold-working the metal by punching, drawing, or rolling is t o retard slightly the rate of corrosion both local and general, provided t h e finished surface is smcoth, owing to the formation of a thin film of flowed material that is somewhat more resistant to water corrosion than the underlying crystalline material. . . . The protective effect of the flowed layer is best seen on smooth surfaces; on roughly cut surfaces and edges, such as are formed by a hacksaw, the layer is not sufficiently uniform and continuous for good protection of the underlying metal.g In acid the effect of flowed layers is not so conspicuous, and they probably depend for their effects, partly at any rate, on oxide films formed during the period of mobility.” It is a matter of general knowledge that highly polished surfaces show a surprising resistance to corrosion. Whether one considers the polished surface “Corrosion and Preservation of Iron and Steel,” 158 (1910). Bancroft: Trans. Am. Electrochem. SOC.9, 17 (1906). 3 J. Phys. Chem. 10, 84 (1906). Trans. Am. Electroctern. SOC.29, 436 (1916). Trans. Am. Electrochem. SOC.39, 69 (1921). Bancroft: Trans. Am. Electrochem. SOC.39, 211; Fink: 259 (1921). Corse: Trans. Am. Electrochem. SOC.39 2 8 (1921). “Sixth Report to the Corrosion Research 6ommittee of the Institute of Metals.” J. Inst. Metals, 28, 68 (1922). 9 This matter has been more fully discussed in the Fourth and Fifth Reports. 1
846
WILDER D. BANCROFT
to consist of very fine crystals or of an amorphous layer as Reilby does, it must be anode against a normal surface and should therefore corrode more readily if everything else were equal. I n this case one must assume, as Bengough and Stuart1 do in another case, that a more coherent film forms on a very smooth surface. “If perfectly pure metal be exposed t o water, oxygen, and electrolyte, a continuous protective gel will probably be formed over the whole surface of the metal. No appreciable corrosion will take place as long as the gel does not change. This is doubtless the explanation of‘ the fact that TJambert’s pure iron remained apparently uncorroded, even when exposed to tap-water and ordinary air.2 The greater “homogeneity of the iron” ensured a continuous protective film, and the inhibition of corrosion throws no light on the question of differences in solution pressure over the actual metal surface. Such iron would necessarily give no reaction under the ferroxyl test, as charged colloid is not being given off at any point on the metal. The presence of such an extremely thin gel film over the metal would be extremely difficult to detect. Thus freshly cut aluminum is supposed to become instantsly covered with a film of oxide; but the metal remains perfectly bright. The corrosion of Lambert’s pure iron in sodium chloride solutions would seem to be due to subsequent chemical or physical alteration of the gel film by the salt solution .” It is a pleasure to find myself in agreement with Bengough and Stuart in the opinion that himbert’s pure iron was covered with a film of some sort. Since an annealed metal becomes more crystalline and therefore probably mugher cn the surface, this may account for the more rapid corrosion of the annealed metal in many cases. It is quite possible that some of the conflicting data in regard to the effect of strain may be due to differences in surface under the different conditions. Bengough and Stuart8 consider that non-metallic impurities in the metal surface may prevent the formation of a film at those points and that such a perforated film may not prevent corrosion, which will develop at the pore. There is no doubt that these things may work this way and there is no doubt that a film with pin-holes in it will not give effective protection. T h e canners have had plenty of experience along those lines. On the other hand it is not safe to make the generalization that this will always happen. It has already been pointed out that cast iron shows an unexpected resistance to corrosion under certain conditions and nobody could have predicted the behavior of the copper-bearing steels or of cerlain of the copper-tin bronzes. It is true that the impurities in the last two cases are not sonims; but it will be much safer not to indulge in too many predictions along this line for the present. There is no question but that the problem of the properties of the surface film is a problem in colloid chemistry. On the other hand we have not explained anything when we say colloids; we have merely indicated a line of ‘‘Sixth Report to the Corrosion Research Committee of the Institute of Metals.”
J. Inst. Metals, 28, 9 107 (1922). Lambert: J. 8hem. SOC.101,2068 (1912).
a “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 95 (1922).
T H E ELECTROLYTIC THEORY O F CORROSION
847
attack. This is a point which is often overlooked. Thcre is no doubt, for instance, but that the plasticity of clay is due to colloids; but we have not explained anything by that. It is still necessary t o show how and why the colloidal material acts, and we must show it in detail. One cannot accept the suggestion of Friend1 that bichromates inhibit the corrosion of iron because they coagulate the ferric hydroxide sol since we know that they act by making iron passive and we know that the passivity of iron is not due t o ferric oxide. I n another paper Friend2 gives a new theory of the corrosion of iron which seems to be that ordinary iron reacts with oxygen in presencc cf water to form a ferrous hydroxide hydrosol which is particularly reactive chemically, and which oxidizes, producing ferric hydroxide hydrosol “in the most favorable circumstances. This higher hydrosol is now alternately reduced in contact with the iron and oxidized again by atmospheric oxygen, thus catalytically accelerating the oxidation of the metal. When the sol flocculates or precipitates out, it yields rust.” So far as I can see, one could substitute the word copper for iron in this theory and prove that copper rusts in the same way as iron does. Also, the peroxide plate of the storage battery would be quite impossible to maintain, because the lead peroxide would react with t’he underlying me tal to form lead oxide. I rejoice that Friend has emphasized the colloidal side of the problem but I cannot see that his papers indicate any real progress. I object very strongly to his calling it a new theory and to his saying lhat the electroljdic theory of corrosion is inadequate to account for the facts. All he really means is that we must consider the properties of the co-roaion products and that these corrosion products are oflen colloidal; but, we don’t question that . If he had said what he really meant, it would have done some good; but, as it is, he has done actual harm, though not much, in an attempt to make his paper seem more of a step forward than it really is. Bad advertizing does not pay. There is nothing to he gained by criticizing Bengough and Stuart’s applications of colloidal chemistry. They are very weird and the best one can say for them is that they do call attention t o the colloidal side of the problem. Bengough and Stunrt3 summarize their hypotheses as follows : “A metal immersed in water sends positively charged metal ions into the licruid, and becomes itself negatively charged. In the case of ordinary commercial metals, the metal also becomes superficially oxidized if dissolved oxygen is present. The hydroxide produced by this oxidation can take up the ions given off by the metal, and the hydroxide thereby passes into the state of a positively charged colloid. Some of this colloid will diffuse away, permitting further reaction between the oxygen and the metal surface, and thereby re-forming the hydroxide film over the latt8er. Oxidation is then stopped till this hydroxide can pass into the colloidal state by acquiring positively charged metal ions. This, in general, does not take place till the colloid J. Chem. SOC.119, 937 (1921). Trans. Am. Electrochem. SOC.39, 63 (1921). 3 “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 33 (1922).
848
WILDER D. BANCROFT
initially formed has diffused into the presence of electrolyte, when it is precipitated by the anion of the dissolved salt, the cation neutralizing the charge on the metal corresponding to that on the colloid. This allows the metal to send more ions into solution, and the uncharged hydroxide thereby acquires a charge. If the colloid so produced can diffuse away, the process can continue and corrosion develop. “For steady corrosion, therefore, the colloid must be produced under conditions which allow it to diffuse some distance from the metal before precipitation. If it precipitates directly on the corroding surface it will, in general, adhere to the latter and stop corrosion. In the case of a corrosion pit, the first condition is fulfilled, since no precipitation occurs inside the pit. It is only when the colloid diffuses through an aperture (generally very small), in the gel-deposits at the mouth of the pit, that it meets electrolyte and is then precipitated. Such precipitation merely thickens the external geldeposits. These gel-deposits adhere directly to, and protect, the metal surrounding the pit, and thereby emphasize the local nature of the corrosion.” Rengough and Stuart’ justify their contention of direct oxidation of the metals by the statements that “at temperatures at which no liquid water can be present, it is clear that such corrosion can and does take place, since no electrolyte is present; as an instance of this type of action t8heformation of temper colours on steel at, say 23ooC, may be quoted, and this is clearly a molecular and not an ionic reaction. . . . Kuster2 has shown that liquid water is not necessary for the corrosion of sodium at the ordinary temperature, since water vapour will readily attack it.” They might also have added that sodium burns in air to sodium peroxide, which it does not do in presence of water.8 I admit frankly that I do not know whether the oxidation of iron at 230’ is or is not a molecular reaction. I know that it does not take place in perfectly dry air and that one can drop dry sodium iqto a flask containing boiling ’ bromine without any reaction taking place, provided the bromine is dry. Until we know more about the theory of combustion than we do now, I am very reluctant to base conclusions on such premises until we are forced to. I have purposely avoided any reference to non-aqueous solutions, though some instances have been cited triumphantly by Bengough and S t ~ s r t , ~ because I believe in attacking the relatively simple problem first. Of course, we must struggle with the non-aqueous solutions some day; buc I do not see how to attack them profitably at present. So far as aqueous solutions are concerned, I believe now, as I have for years, that any special corrosion prolnlem can be accounted for satisfactorily and quickly if the investigator makes use of the electrolytic theory of corrosion, the facts of chemistry, and common“Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 41, 52 (1922). 2 Z. anorg. Chem. 41,474 (1904). 8 Dry oxidation is differentiated sharply from wet oxidation by Haber: Z. Elektrochem. 7,445 (1900). 4 “Sixth Report to the Corrosion Research Committee of the Institute of Metals.” J. Inst. Metals, 28, 52 (1922).
THE ELECTROLYTIC THEORY O F CORROSION
849
sense. These problems are not mysterious in any way; it is we who make them seem so. Since the first draft of this report was written, there hasappearedar, article1 on “The Electrochemical Character of Corrosion” by Evans which also controverts many of the points raised by Bengough and Stuart. “It seems likely, indeed, that the oxidation of metals in air at high temperatures, which appears to proceed in the absence of any moisture, must be regarded as a direct chemical attack. On the other hand, at ordinary temperatures the observed phenomena of corrosion, as far as they are known a t present, correspond closely to those which could be predicted from fundamental electrochemical principles. Consequently here theories of ‘pure chemical corrosion’ are at present superfluous, although it is possible that real cases of ‘pure chemical corrosion’ a t low temperatures may be discovered by future research. Those who at present assert that the ordinary corrosion of common metals, like zinc and lead, are cases of simple chemical oxidation, may be asked why the electrochemical corrosion which would be predicted from our accurate knowledge of electrode potentials fails to occur. Only when they have explained the absence of electrochemical corrosion are they entitled to construct theories regarding chemical corrosion.’’ Evans2 is rather sympat,hetic with Lambert’s point of view. “We are now in a position to approach the subject of electrochemical corrosion. Let us consider a piece of metal immersed in a liquid. If the metal is absolutely uniform, both chemically and physically, the potential between metal and liquid will be the same a t all points, and no current can be set up. No electrochemical corrosion is therefore to be expected. It is significant here t o recall that, in the case of iron and lead, Lambert and his coworkersa have actiially succeeded in preparing materials so pure and uniform that they do not corrode under conditions which allow rapid attack of materials of merely ‘ordinary purity.’ ‘‘If, on the other hand, the metal is not uniform, varying from place to place either in chemical or physical character, it is certain that at the moment of immersion there cannot be equilibrium. The potential will vary from point t o point, and a current, (momentarily at least) must be set up between the different points. At certain places (the anodic areas) the metal will commence t o pass into the ionic condition; at other places (the cathodic areas) hydrogen ions will be discharged (assuming no other metallic ions are present in solution), and elementary hydrogen will commence to alter the potential of the cathodic areas in a negative sense, thus making it more nearly equal to the potential at the anodic areas. If this process of hydrogen accumulation continues until the potentials at the originally anodic and cathodic areas become equal, equilibrium will be established, and in such a case there can be no appreciable electrochemical corrosion. 2
J. Inst. Metals, 30,239 (1923). J. Inst. Metals, 30, 244 (1923). Lambert, Thomson, and Cullis: J. Chem. SOC. 97, 2426 (1910); 101, 2056 (1912);
107, 218 (1915).
850
WILDER D. BANCROFT
“Two occurrences, however, may prevent the establishment of equilibrium. Before the potential at the cathodic areas becomes equal to that of the anodic areas, the cathodic areas may become so supersaturated with hydrogen that the gas begins to stream off in bubbles, in this case, corrosion of the metal at the anodic portions will continue indefinitely, the rate being equivalent t o the rate of production of gaseous hydrogen a t the cathodes; we may call this the Hydrogen-Evolution Type of Corrosion. If, however, the conditions are not such as t o allow the evolution of gaseous hydrogen in bubbles, a relatively slow removal of hydrogen from the cathodic areas may take place if oxygen (or an oxidizing agent) is present in the solution; this will allow the corrosion of the metal a t the anodic areas to cont.inue, but the rate will be limited by the rate of diffusion of dissolved oxygen (or the oxidizing agent) across the layer of liquid next to the metal; we may term this kind of corrosion t8heOxygenDiffusion Type.” Evans points out that “magnesium, for instance, scarcely reacts with pure water; but the presence of a soluble chloride in the water causes the formation of the hydroxide in a ‘loose form’, and has thus an ‘activating’ effect on the metal, and allows a vigorous evolution of hydrogen.” In another passage he points out, that this loosening effect is probably akin to peptization. “If now we consider the reaction of a metal in a solution without contact with platinum-black, we have to take considerations of over-potential into account. The condit,ion needed for the evolution of hydrogen in bubbles will become 7T*--4--Th,> 0, where 4 is the overpotential of the cathodic areas. Thus metals like tin, lead, and nickel, which stand close to hydrogen in the potential series, at ordinary temperatures do not readily displace hydrogen in bubbles from dilute acids; they may do so, however, if the solution is warmed, so as t o diminish the overpotential.. Even zinc does not displace hydrogen readily from acids when pure, in spite of its position near the “reactive” end of the potential series; but if it contains noble impurities of low overpotential value, the gas is evolved more rapidly. Commercial zinc evolves hydrogen slowly a t first, hut after a period of induction” the reactien becomes quite vigorous, owing t o the accumulation of a black deposit of the noble impurities on the metallic surf ace.” “If the conditions are such that hydrogen cannot be evolved as a gas, it may yet be removed from the cathodic areas slowly if dissolved oxygen is present in the solution a t these points. This is found to be the case; metals like iron, nickel, cadmium, and lead, which do not displace hydrogen in bubble form when placed in B neutral solution, nevertheless are corroded slowly by a neutral solution (say, a solution of sodium chloride) containing oxygen; likewise copper, which cannot cause the evolution of hydrogen gas from dilute acids, dissolves slowly in these acids in the presence of oxygen. “This type of corrosion will always be slow; for rough purposes, we may say that a maximum rate is fixed by the rate of diffusion of oxygen across the layer next to the metallic surface. Consequently, within certain limits, the ((
’
THE ELECTROLYTIC THEORY O F CORROSION
851
rate of corrosion will be nearly independent of the conductivity of the solution. There is no reason why oxygen should diffuse more quickly through a highly conducting liquid than through a badly conducting liquid. If anything, oxygen will actually diffuse less quickly through a solution of a salt than through pure water, because the presence of a salt generally depresses the solubility of oxygen in water.” “The electrochemical prediction of corrosion would lead us to regard as possible the attack of metal at points to which dissolved oxygen has no direct access. The presence of oxygen is only needed at the cathodic areas; at the anodic areas its presence will, if anything, be unfavourable to the reaction. It is interesting to find that specially marked corrosion often occurs at points to which oxygen can only diffuse very slowly-for instance, at the bottom of pits or over areas covered with porous substances, such as string or fabric. “The electrochemical corrosion of a met,al by, say, sodium chloride, should produce the metallic chloride at the anodic portions, whilst the discharge of hydrogen ions on the cathodic portions will leave the liquid alkaline at these points ; where the alkali and the metallic chloride diffuse together, the metallic hydroxide will generally be precipitated, often as a membranous web or a gelatinous precipitate; this phenomenon has been observed in practice. “If, however, the metal is immersed in a solution containing an anion which forms an insoluble salt with Lhe metal in question, the primary product at the anodic areas will be an insoluble substance; under certain circumstances this may adhere to the anodic areas and protect or ennoble them, thus causing corrosion to cease. Numerous examples of this effect are known; perhaps the most important case is that of lead immersed in waters rich in carbonate. “The formation of a protective film, is, however, not confined to cases of this kind. It is well known that when an external e.m.f. is applied to an electrolytic cell fitted with an iron anode immersed, say, in a sodium sulphate solution, a low current density will give rise t o the formation of soluble iron sulphate. But there i s always the possibility that instead of the discharge of Sod” ions, the discharge of OH‘ ions may occur. If the current density is raised unduly, something in the nature of an invisible oxide film (possibly a layer of adsorbed oxygen atoms) is produced, and the iyon becomes “passive”; the layer retards the anodic dissolution of the iron, whilst permitting the production of oxygen gas; I have discussed this subject e1sewhere.l With an aluminium anode the effect is even more striking; a porous oxide film containing oxygen gas in the pores is produced, and this prevents the passage of current altogether in one direction; in other words, aluminium refuses to funct.ion as an anode, unless certain substances (notably chlorides) are present, which cause the film to break down.” Evans2 makes the same mistake that Bengough and Stuart did in postulating that potassium bichromate and potassium chromate are powerful oxidizing agents. The discussion of the oxidation of apparently dry metals is apparently sound and certainly interesting. “It has been stated that in U. R. Evans: Trans. Faraday SOC., 18, J. Inst. Metals., 30,251 (1923).
I (1922).
WILDER D. BANCROFT
8j2
general the conductivity of the liquid should be only a minor factor in determining the rate of corrosion. In one case, however, the conductivity becomes of great importance. This is in the corrosion of “apparently dry” metals by air. If a metal is truly dry, it does not corrode in dry air at low temperatures, but if an invisible film of adherent water is present, corrosion may occur. The thinness of the film will restrict the current flowing between anodic and cathodic areas, and if the film is composed of pure water the attack must be extremely slow. If, however, the atmosphere contains volatile electrolytes, such as hydrogen chloride or sulphur dioxide, these may be expected t o diasolve in the water and increase the conductivity of the film, and thus accelerate the corrosion; if, further, the corrosion product has a hygroscopic character, it may be expected to absorb further water from the at,mosphere, and by increasing the thickness of the film, increase still more the rate of attack. I have shown experimentally that both these expectations are realized. “The work of Pilling and Bedworth2 has shown that a t high temperatures the direct oxidation of metals by air, even in the absence of water, is possible; but, as the oxide film grows thicker, the rate of diffusion of oxygen through it becomes slower, until finally a thickness is reached at which further oxidation becomes negligibly slow. The thickness attained will of course depend on the temperature. On iron the oxide film produced at zgo’c. is capable of extinguishing (by interference) the yellow rays of light, causing a blue “temper colour”; at 230’ the film is thinner and the blue rays are extinguished, the temper colour being therefore yellow; below 200OC.the film is too thin to cause the extinction of any visible rays, and clearly at ordinary temperatures the thickness of the film caused by direct oxidation will be negligible. It is evident,, therefore, why the presence of a moisture film and an electrolyte is needed for rapid corrosion at ordinary temperature; t h e hydroxide or oxide formed by interaction between the soluble salts from the anodic areas and the alkli from the cathodic areas will not necessarily form a protective coating over the whole surface, and will t’hereforenot necessarily bring the action to a standstill.” Onp. 254Evans says that “the type of corrosion where hydrogen is liberated in bubbles is already generally admitted to be electrochemical in character. Consequently only a few experiments were conducted on this type, t o investigate certain points which seemed in doubt. . . . Cadmium was found t o be practically unaffected by normal hydrochloric acid in the absence of other metals, n o doubt owing to its high overpotential value. Contact with nickel ---or even with iron-caused the evolution of hydrogen from the second metal. Contact with copper caused only a slow evolution of hydrogen from the copper; whilst contact with lead-a metal of high overpotential-had no effect.” This is interesting, because apparently an alloy of cadmium and lead would not corrode readily in spite of the surface being heterogeneous. “The type of corrosion wherein no hydrogen is liberated in bubbles is considered by some authors, notably by Bengough and his c o - ~ o r k e r s , ~ ‘U.R. Evans: Trans. Faraday SOC., 19 (1923). 3
N. B. Pilling and R. E. Bedworth; J. Inst. Metals, 29, 529 (1923). J. Inst. Metals, 21, 37 (1919);28, 31 (1922).
THE ELECTROLYTIC THEORY OF CORROSION
853
t o be due to the direct chemical oxidation of the metal by oxygen, and not t o electrochemical action. Consequently, extensive experiments were performed, with a view t o ascertaining whether t.he acti,on was electrochemical or not. . . . “Bengough and Hudson,l in 1919,found that when cast zinc was suspended in “dist,illed water” certain white vertical striations, formed of zinc hydroxide, were produced upcn the surface. I obtained rather similar results with cast zlnc, and also obtained well-marked striae upon sheets of roZZed zinc placed in a vertical or nearly vertical position in glass tumblers half-filled with liquid. The plate could be obtained with zinc of variety “A”, but is better developed on the less pure varieties of zinc. It could be produced best in a N/2 potassium chloride solution ; but a somewhat similar phenomenon was obtained in ordinary laboratory distilled water, or even in “conductivity water” having a conductivity of 2 x 1 0 mho-cms. ~ ~ (Le. the same quality of water as that used by Bengough and Hudson; it is, of course, far from being “pure water”, which has a conductivity of about 4x10-*mho-cms,). The striations appear within a few hours of the immersion of the metal and continue to develop during several days. The generad appearance of t h e striae in the early stages of corrosicn often recalls the tracks made by ra.in-drops running down a window-pane, and suggests that they are formed by drops of some heavy liquid sinking down along the surface of the zinc; the tracks are bounded on each side by a thin membranous wall of zinc hydroxide. There is R more or less continuous barrier of zinc hydroxide separating a nearly uncorroded area near the water-line from the corroded portion below; in places this wall may stand out for a distance of I t o 2 mm. from the surface of the zinc. Where potassium chloride solution is used there is a great deal of flocculent zinc hydroxide a t the bott$omof the tumbler, in addition to that attached to the metallic surface; it is evident that the attack upon the zinc is very much more pronounced when potassium chloride is present than when “distilled water” is used-a fact which has been confirmed by experiments with weighed specimens. “If after the formation of these striae the metal is removed from the liquid, allowed to drain, and is then tested with phenolphthalein, it is found that alkali is present. The walls of zinc hydroxide become pink;’ but where potassium chloride solution has been used, a much greater amount of alkali is present on the uncorroded portion close to the water-line; very little alkali seems to be formed below the main hydroxide “barrier.” “The electrochemical view of corrosion suggests a very simple explanation of these phenomena. The oxygen-rich portions near the water-line become cathodic, and the test of the zinc, where the supply of oxygen cannot be renewed, becomes anodic. Alkali is produced at the cathodic portions, and The production of the pink colour where “distilled water” has been used to develop the striations may be explained either by the supposition that pure zinc hydroxide is alkaline, or by the presence of minute quantities of sodium or potassium salts in the distilled water. Preliminary experiments with water condensed in quartz show that the pink colour is still produced with this water, but it is possible that small amounts of sodium salts were introduced with the zinc or from the emery-cloth. It is hoped to carry out further work on this subject.
854
WILDER D. BANCROFT
diffuses outwards into the body of the liquid; zinc chloride (or some other zinc salt) is produced at the anodic points lower down, but being heavy it tends to sink down along the metallic surface. The membranous walls of hydroxide simply represent the surfaces where the alkali and zinc salt have come into contact and have reacted together.” “Bengough and Hudson1found, in 1919,that if two pieces of copper placed in a divided cell were joined to the poles of a galvanometer, and if air was bubbled over one piece, that piece became anodic; thus copper shows a behaviour which at first sight i s the exact converse of that displayed by zinc, cadmium, iron, and lead. “It seemed most important to investigate this anomaly, and, after numerous experiments, it became clear that the potential of copper immersed in potassium chloride solution was affected t o an important extent by three types of treatment : (a) Aeration.-The presence of oxygen tends to render the potential more positive, as in the case of all other metals. (b) Stirring.-The removal of an accumulation of copper ions from the metallic surface alters the potential in the negative direction, in accordance with the principle explained in Part I. of this paper. (c) Abrasion.-The renewal of the surface by energy or scraping removes any protective film of cuprous chloride (or hydroxide) that may exist on it, and thus renders the metal “active,” and shifts the potential in the negative direction. “It was not easy at first t o separate these three effects. They can best be demonstrated by using an alkaline solution made by adding I O C.C. of normal sodium carbonate to 90 C.C. of N/2 potassium chloride. I have, however, satisfied myself that they also occur in a N/2 solution of potassium chloride t o which no alkali has been added. If two copper electrodes are joined to a galvanometer and are immersed in the alkaline solution for some minutes, and one electrode is then taken out, dried, ground with emery and replaced, a current is produced, the treated electrode being the negative pole. The current produced in this way is comparatively large, but very fugitive; over 2 milliamps. were obtained momentarily between electrodes of immersed area 4x4 cm. If, on the other hand, two electrodes composed of the same type of copper, and having a similar electrochemical history, are immersed in N/2 potassium chloride until any current produced on first immersion dies away (this may sometimes require fifteen to thirty minutes), so that both electrodes are known to be covered with a film to the same extent, and if in this state the electrodes are alternately taken out into t’he air and replaced, a very much smaller current (usually only a few micro-amperes) is produced, the aerated electrode being always positive. The effect of stirring is seen if the electrodes are immersed in N/2 potassium chloride in a divided cell, one electrode being stationary, whilst the other is moved about; the electrode in motion becomes the negative pole. If both electrodes are stationary, but air is bubbled into one compartment, it is evident that both aeration and stirring will be pro1
G. D. Bengouph and 0. F. Hudson: Jour. Inst. Metals, 21,
122 (1919).
THE ELECTROLYTIC THEORY OF CORROSION
855
duced. It is found, however, that Lhe effect of stirring greatlv predominates over the effect of aeration, and that the current produced flows in such a direction as to make the copper over which bubbles are passing the negative pole. Since Bengough and Hudson used a stream of bubbles t o “aerate” their copper, it is clear why they obtained a current passing i n what at first sight may be regarded as the ‘(wrongdirection.” “If instead of using potassium chloride solution we use normal sulphuric acid, the abrasion effect ceases to be important, for obvious reasons; but the stirring effect becomes very pronounced and tends to mask the aeration effect. The effect of aeration can, however, be demonstrated by the following way. Two copper electrodes joined to a galvanometer are immersed in normal sulphuric acid. One is taken out, wiped dry with filter-paper, so as to expose the metal directly to the air, and replaced; a momentary current, sometinies reaching 3 milliamps. for electrodes of the dimensions given above, is produced, the aerated electrode being positive. “It seems clear, therefore, that copper behaves towards aeration exactly in the same way as other metlals, but that the effect may be masked by the effects of stirring and of abrasion. The effect of abrasion has been noticed in the case of other metals which are liable to become covered with “protective films”, and needs no further discussion; but the influence of stirring requires special consideration at this point. Evidently the stirring removed the accumulation of copper ions from the neighbourhood of the electrode; such a removal will clearly shift the potential in a negative direclion. It may be asked why the power of stirring to shift the potential in the negative direction is not notviceablein the case of zinc, cadmium, iron, and lead; in these metals stirring-in so far as it has any effect-tends to move the potentinl in the positive direction, by renewing the supply of dissolved oxygen at the electrode surface. A little consideration will show, however, that the effect of any accumulation of ions which is likely to be realized in practice will be negligible in the case of the more reactive metals, but will be very pronounced in the case of the noble metals; for at any givev potential, the ionic concentration needed to put a stop to the passage into the ionic state is very high in the case of reactive metals, but very low in the case of noble metals.” The chief difficulty with this explanation is that, as has been mentioned, nitrogen and hydrogen act like oxygen. It is therefore a gas effect and not a question of oxidation at all. “It seemed important t o ascertain whether the general laws regarding the effect of aeration on the potential apply also to aluminum. It has often been asserted by critics of thp electrochemical view of corrosion that thc stable behaviour of aluminum towards corrosive agencies was in sharp contrast with the highly negative potential ( - 1.337 volt) usually assigned to the metal. One has to remember, however, that the experimental determination of the potentinl of aluminium is carried out with amalgamated aluminiuma material which oxidizes with great, rapidity in damp air, and which has none of the stability associated with ordinary aluminium. Ordinary aluminium is clearly unsuited for the determination of the reversible potential, because it
856
WILDER D. BANCROFT
cannot function freely as an anode; such measurements of potential as have been made with ordinary aluminium give it a value very different from that attributed t o aluminium in Table I.’ If studied in the light of the extensive work of Schulzel on the valve action of aluminium produced by an externally applied e.m.f., the behaviour of aluminium immersed in different liquidsthe knowledge of which we owe mainly t o the work of Seligman and Williams2 -becomes quite intelligible, and the facts seem to indicate that here also we are dealing with a case of valve action, but produced by an internally generated electromotive force. “The rapid formation of a protective film on aluminium renders its study somewhat difficult. I found, however, that if two pieces of rolled aluminium were ground with emery, and were at once placed in a divided cell containing N/2 potassium chloride solut,ion and joined to a microammeter, they show the aeration effect just as easily as any other metal. The withdrawal a r d replacement of either electrode produces a current, the aerated pole being always positive. After a few minutes, however, the current produced in either direction by aeration begins t o be much smaller, although the effkct of aeration can still be detected after five minutes. I t is evident that the formation of obstructive films is responsible for the falling off in the sensitiveness, for if one electrode is treated afresh with emery-paper, it always becomes strongly negative (anodic) towards the other; boring or cutting of an electrode-processes which likewise expose a fresh, unoxidized surface-produce the same effect, the electrode so treated becoming negative. The conclusion may be drawn, therefore, that aluminium behaves towards aeration just in the same way as other metals, but the effect is complicated and often hidden by the formation of an obstructive skin. “This matter is of some importance, since it serves to afford an explanation of certain phenomena observed by Seligman and william^.^ These authors came to the conclusion that the blistering of aluminium in hard industrial waters was connected with the presence of small cavities in the metal. They succeeded in reproducing the effect experimentally by making a number of small depressions in the metal and closing them up by hammering; on immersion typical “pits” and “blisters” were formed. Presumably the interior of the cavities to which oxygen could not diffuse readily became anodic, and consequently serious attack took place within the cavities. A simlar explanation can be extended to the specially marked corrosion occurring over areas covered up by poroud materials-a fact also established by Seligman and Williams. “In my own experiments, it was found thab if holes were bored in an aluminium surface, and the metal was then immersed in N/2 potassium chloride, a considerable amount of aluminium hydroxide usually formed over the mouth of the holes. But I am inclined t o tbink that this was due t o the See, for instance, T. Heyrovsk?: J. Chem. SOC.107, 27 (1920). Schulze: Ann. Physik, 21, 929 (1906); 22, 543 (1907); 26, 372 (1908). R. Seligman and P. Williams: J. SOC.Chem. Ind., 35, 88, 665 (1916): 37, 159 T (1918); J. Inst. Metals, 23, 159 (1920). 3 R. Seligman and P. Williams: J. Inst. Metals, 23, 166-168 (1920). 1
1
2
T H E ELECTROLYTIC THEORY O F CORROSION
857
presence of bubbles of air within the holes. . , , Thus, the phenomenon produced artificially by me is probably the exact converse of that which causes blistering in aluminium articles containing natural holes. “Certainly, wherever air bubbles could be seen clinging to the aluminium surface, much flocculent aluminium hydroxide was produced in the liquid around these points. This was observed not only with the aluminium sheet, but also with utensils made of aluminium. Some aluminium teaspoons were immersed in N/z potassium chloride solution, and air was bubbled through the solution for a few minutes; the spoons were then allowed to stand in the solution for two days. The greater part of the surface became tarnished and covered with a brownish-white deposit ; but around and below the adherent air bubbles it remained comparatively bright, and at these places loose white aluminium hydroxide appeared in the liquid, as though produced by inheraction between the alkali (from the cathodic areas around the bubbles) and the aluminium chloride (from the anodic areas elsewhere).” Evans1 draws a number of conclusions regarding the mechanism of corrosion. “If the experiments described above, dealing with eight different metals, are considered as a whole, it may be claimed that each part of the electrochemical mechanism of corrosion has been demonstrated. Where the metal is immersed in a neutral salt solution, it has been possible to show: I . The production of a current. 2 . The production of a metallic salt at the anodic portions. 3. The production of alkali at the cathodic protions. 4. The precipitation of hydroxide where they meet. “If the aeration is uniform, it is apparently the distribution of impurities that determines which portions shall act as cathodes and which as anodes. If, however, the aeration is not uniform, then-provided there is a sufficient amount of noble impurities distributed throughout the metal as a separate phase-the cathodic and anodic areas are determined by considerations of the supply of oxygen, the “aerated” areas being cathodic and the unaerated anodic. Experiments on pure cadmium and lead seem to show that some impurities are actually needed on the aerated portions to act as cat.hodic particles; mere selective aeration of one area will not cause it to become cathodic, unless a suitable noble impurity is present. If, however, there are noble impurities scattered throughout all parts of the metal, the mechanism of selective aeration must be as follows. When one part of a piece of metal is aerated, and one is protected from aeration, the foreign particles in the protected area cannot (after t h e first moments) function as cathodes, because there is no supply of oxygen for the cathodic reaction. The foreign particles on the aerated portion act as cathodes, therefore, to the dominant metal in both areas; but in the aerated area the accumulation of insoluble corrosion product soon begins to interfere with the anodic function of the metal surrounding the foreign particles; and accordingly we are left with the foreign particles in the aerated area acting as cathodes, and the bright unobstructed metal of the unaerated area acting as anode. The anodic attack would be J. Inst. Metals, 30, 278 (1923).
858
WILDER D. BANCROFT
expected to be most pronounced in the portions of the unaernted area closest to the aerated mea, as has been found t o be t’he case. “We thus obtain a raticnal explanation of the dangerous corrosion occurring at the bottom of pits and crannies, and other inaccessible places, as well as in areas covered wit,h porous debris-a phenomenon well known to practical men. At such places, which are wholly anodic, owing t’othe absence of oxygen, t’he product of corrosion is generally a soluble metallic salt, which does not interfere with the subsequent course of corrosion.. Over the accessible parts of the surface of a metallic article we may indeed temporarily experience electrochemical corrosion; but here the cathodic and anodic areas are close together, and insoluble metallic hydroxide is produced close against the metallic surface, and tends to protect it,, and thus st’op the action. The electrochemical explnnat’ion of the in tense corrosion occurring at shekered places is the only satisfactory one. Those theories, which attempt to account for the special corrosion at inaccessible poin tJ by assuming that flocculst,ing elect1olytes cannot reach these points (or that hydrogen peroxidn is retained at t,hese points and aids subsequent corrosion), involve many difficulties. If, for instlance, diffusion through the mouth of a pit is assumed to be so slow tha.t electzolytes cannot diffuse inwards (or t8hat hydrogen peroxide cannot diffuse outwards), how can we account for the constant replenishment of oxygen wit’hin the pit? The electrochemical view of corrosion, which does not demand t,he presenm of oxygen in the pit, is free from these difficulties. “The produc,t.ion of a prot,ective film on t,h.e anodic areas of the met,al oftmenintroduces a comp1icatin.g factor; this is most serious in t’he case of aluminium, but it was noticed in experiments on ot,her metals that the current produced sometmimesten.ded to drop off with the time. Indeed, the “influence of the corrosion product” is probably the most vit’alfactor in determining the course of the later stages of corrosion. In the series of experiments described above, this complicating factor has been eliminat’edas far as possible by starting with specimens havin.g a freshly ground surface. I have discussed the factors governing the question of the adhesion of the corrosion product elsewhere;’ small chan.ges in the composition of either the metallic phase or of the liquid may completely alter the degree of adhesion. The formation of a compact, protective film over the anodic portions of t,he metal will tend to “ennoble” the latter and to reduce the at,tack. On the other hand, the presence of a porous, gelatinous precipitate over these areas may act as a diaphragm, allowing the current t o pass, but hindering the diffusion of oxygen to those areas; this will actually favour the reaction. In cases where a hydroxide precipitat,e is produced at the junction of the anodic and cathodic areas, it will presumably depend largely on chance convection currents in the liquid as to whether the hydroxide tends to move on to ihe anodic or cat.hodic area; in the former case, it may favour corrosion; in the latter case, by hindering the diffusion of oxygen to the cat.hodic area, it will tend t’oinberfere with it.” Evans also says that “it has been proved that the unequal aeration of a metallic surface tends to produce intense corrosion at cert
