The Mechanism of Rusting By U. R. Evans F.R.S. U N I V E R S I T Y OF CAMBRIDGE Introduction Scaling and Rusting.-The rusting of iron (the formation of hydrated oxide in presence of oxygen and water) must be distinguished from scaling (the formation of anhydrous oxide). In scaling the oxidation-rate falls off as the thickness of the oxide-scale increases. At high temperatures the thickening rate may be inversely proportional to the thickness. At room temperature the rate falls off more abruptly; recent measurements1 record 16 8 after a day and 35 8 after a year. The latter thickness is insufficient to produce the interference colours seen on iron after brief heating in air within the range 250-350"~. Films formed on iron by unpolluted dust-free air at room temperatures generally produce no change in appearance; they can diminish reactivity towards certain reagents.In contrast the rate of rusting sometimes remains almost constant over con- siderable periods. The reason for the difference is that in scaling the oxide is formed upon the metallic surface whilst in rusting it may happen that oxygen is reduced at one place iron passes into solution at a second place whilst the iron oxide appears in hydrated form at a third place where it cannot interfere with continued attack. Conditions for Rusting.-A horizontal iron plate fully immersed in any natural water (fresh or saline) with air above the water surface suffers rusting usually slowly. More rapid rusting occurs if the plate (placed vertically or sloping) is only partly immersed oxygen being readily replenished where the plate cuts the water-line; as explained later there is often an unattacked zone at or just below the meniscus.The rust formed on atmospheric exposure is usually more adherent than that produced by immersion. Exposure outdoors (when the surface is alternately wet and dry) soon sets up rusting but in highly polluted atmospheres sheltered surfaces sometimes rust more rapidly than those kept clean by rain. Indoors an unpainted iron surface often remains bright for some time except where salt particles can settle on it or where the air contains appreciable amounts of sulphur compounds derived from fuel. Mechanism.-The corrosion of iron (or zinc) immersed in water or salt solution is undoubtedly electrochemical; the currents have been measured by two in- ID. Gilroy and J. E. 0. Mayne Corrosion Science 1965 5 55.29 Quarterly Reviews dependent methods2p3 and correlated with the corrosion-rate in the sense of Faraday's law; the good agreement between observed and calculated rates pro- vides quantitative evidence for the electrochemical mechanism. For outdoor atmospheric rusting the evidence is less direct although Russian measurements* on plates covered with a shallow layer of water show that electrochemical action is proceeding. For indoor conditions also it is difficult to explain the facts except by assuming an electrochemical mechanism. Immersed Conditions Rusting in Potassium or Sodium Salt Solution.-The contrast between scaling and rusting is seen (Figure 1) on comparing (A) the oxidation of steel plates 225' w . x 175" X -.-e-r-Q- o-grav ime t ric x- electrometric k' - 0 2 4 6 8 1 0 1 2 1 4 Time (hr) 0 0 2 N -K2 SO /Om"- KC' O'OO 25 5 0 75 100 (hr.) Fig.I. 0.0 215 s'o ;5 I Time Chr.) exposed to dry oxygen at 175" and 22505 with (B) the corrosion of such plates partly immersed in salt solutions.2 In (A) the oxidation-rate falls off with time; in (B) after initial irregularities the corrosion velocity attains a constant value. This depends only slightly on the solution used. A twenty-fold reduction in the concentration of potassium chloride produces only a small reduction in the corrosion-rate nor is there much change when sulphate is substituted for chloride. Other measurements have shown that the corrosion-rate on pure electrolytic iron is of the same order of magnitude as the rate obtained on steel. The fact that the corrosion-rate is so little influenced by either metallic or liquid phase is due to the fact that it is largely controlled by happenings at the meniscus the only region where oxygen can readily be replenished from the air.Here oxygen is reduced by a cathodic reaction which occurs in steps but can be summarised 0 + 2H,O + 4e = 40H- U. R. Evans and T. P. Hoar Proc. Roy. SOC. 1932 A 137,343. J. N. Agar quoted by U. R. Evans J. Iron Steel Inst. 1940 l41,219P esp. p. 2 2 1 ~ . L. L. Rosenfel'd First internat. Congress met. Corrosion 1961 Report (Butterworths) 1962 p. 243; N. D. Tomashov Collected papers on 'Corrosion of Metals and Alloys' 1963 (Moscow) English Translation by A. D. Mercer ed. C. J. L. Booker. D. E. Davies U. R. Evans and J. N. Agar Proc. Roy. SOC. 1954 A 225 443. 30 Evans This requires four electrons which are supplied by iron entering the liquid as cations at points lower down the specimen by an anodic reaction which also occurs in steps but can be summarised 2Fe = 2Fe++ + 4e Since the main ions in a potassium chloride solution are K+ and C1- the cathodic and anodic products can be regarded as potassium hydroxide and ferrous chloride; where they meet they will interact to give a precipitate of a yellow- brown rust consisting of hydrated ferric oxide Fe,O,,H,O or FeO,OH it being assumed that plenty of oxygen is present; if a narrow vessel is used restricting the oxygen-supply a ferroso-ferric compound may be formed.All these facts have been established by direct observation. The alkali and ferrous salt can be detected by simple chemical tests; the precipitation and settlement of the rust are obvious to the eye.The electric current representing the upward movement of electrons through the iron from anodic to cathodic zone can be measured directly on a meter if the specimen is cut along the line dividing the expected anodic area from the expected cathodic area [Figure 2(A)]; this is however a less accurate method of measuring the total current flowing than others w. L MENISCUS Fig. 2. The ‘corrosion pattern’ varies with the physical character of the metal surface. On carefully rolled steel the anodic attack may occur only along the cut edges [Figure 2(B)] where the internal stresses left by the shearing probably prevent the maintenance of a protective film. With less carefully prepared steel corrosion starts at defects on the face and spreads downwards and sideways over arch- shaped areas.These are first seen at points low down [Figure 2(C)] and later at points higher up [Figure 2(D)]. Arch-shaped ‘mantles’ of membranous rust grow out roughly at right-angles to the surface representing the surfaces separating regions wherein OH- and Fe++ respectively predominate. In the end [Figure 2(E) and (F)] nearly the whole surface is corroding except the zone just below the water-line where any iron ions escaping from the metal would find them- selves in alkaline liquid so that precipitation must occur in contact with the metal; this area suffers no appreciable corrosion but usually develops a film thick enough to display interference colours when the specimen is taken out washed and dried. 31 Quarterly Reviews The main body of the rust is precipitated well away from the metal.Its com- position and appearance depends on the cross-section of the containing vessel which controls the oxygen supply. It is not impossible that Fe++ OH- and 0 can interact to give hydrated ferric oxide as the first solid phase but it is generally assumed that ferrous hydroxide or a basic salt is first precipitated and then becomes oxidised. Such oxidation has been studied.g Given plenty of oxygen the product can be a- or 7-FeO,OH according as the pH value is high or low. With deficiency of oxygen green ferroso-ferric compounds may appear; these have been studied at Beme,' and appear to be not simple hydroxides but to contain C1- or SO,- groups. With some geometric arrangements black magnetite is formed.Since however a solid formed out of contact with the metallic surface cannot protect the composition and structure of the rust is not of primary importance. The rate of attack will remain almost constant unless the liquid is one capable of forming a sparingly soluble substance as cathodic or anodic product. Such cases will now be considered. Rusting in Magnesium Salt Solution.-The non-protective character of rust formed in a potassium or sodium salt solution is due to its precipitation at a distance from the metal; if formed in contact with the metal surface rust provides a modicum of protection as is shown by comparing behaviour of steel in magnesium and potassium sulphates. Steel partially immersed in magnesium sulphate has been found* to corrode at slightly more than half the rate caused by distilled water (of the quality used in preparing the solution); the rate was remarkably constant over the concentration-range 0.01 - 0 .4 5 ~ . In contrast potassium sulphate corroded the steel much more quickly than the distilled water. Observations of the attack made the reason quite clear. In a potassium sulphate solution ferrous sulphate is formed below the mantles and potassium hydroxide above them. The potassium hydroxide being soluble is soon dis- persed into the liquid and except on the interference-tint zone just below the meniscus there is no protection. With magnesium sulphate there is a similar formation of arch-shaped areas first low down [Figure 2(C)] and then higher up [Figure 2(D)]. But here the magnesium hydroxide formed by cathodic action just above the lowest arch-shaped mantles does not disperse being sparingly soluble.It is deposited on the metal and when the attack becomes directed on higher points the ferrous ions formed by anodic action underneath the adherent magnesium hydroxide convert it into an iron compound. First a green ferroso- ferric layer appears and then a brown rust; these clinging deposits contain magnesium and the green is far brighter and the brown distinctly lighter than corresponding deposits containing no magnesium; probably in the green sub- stance some Mg2+ ions replace Fe2+. In the end the whole area is covered with clinging rust except for a narrow horizontal strip of white magnesium hydroxide 1 to 2 mm. broad at the meniscus. The important feature of the situation is that J. E. 0. Mayne J. Chem.SOC. 1953 129. ' W. Feitknecht and G. Keller 2. anorg. Chern. 1950 262 61. U. R. Evans J. SOC. Chem. Ind. 1928 47 S ~ T esp. p. 581. 32 Evans the solid matter is deposited upon the metallic surface instead of being precipi- tated at a distance so that there is appreciable interference with attack. Rusting in Sea-water.-Magnesium and calcium are invariably found in sea- water. At one time ~ / 1 0 sodium chloride was often used to represent sea-water in laboratory corrosion tests but generally it corrodes more rapidly. At first the difference may not be great but it becomes important after times sufficient for clinging rust to be formed; in experiments lasting 128 days sea-water collected from the English Channel caused only one third of the corrosion of ~ / 1 0 sodium ~hIoride.~ In certain circumstances however sea-water may be more corrosive than sodium chloride solution.Harbour waters containing organic sulphur com- pounds such as cystine can be very dangerous whilst the mud of estuaries may contain sulphate-reducing bacteria which render the oxygen of SO:- available for the cathodic process. Atmospheric rusting near the coast is probably favoured by magnesium chloride in the salt which prevents drying. Rusting in Hard Fresh Waters.-Many waters used for supply purposes contain calcium bicarbonate as main constituent. A pure solution of calcium bicarbonate is unstable and would deposit solid calcium carbonate if an excess of carbonic acid were not present; the amount of carbonic acid needed for stability depends on the Ca(HCO& concentration and the relationship between stabilising carbonic acid and hardness has been established by German work;l0J1 the results accord well with calculations based on the law of mass action if allow- ance is made for activity coefficients.If water carries more carbonic acid than is needed for stability the excess is known as ‘Aggressive Carbonic Acid’. Such water will dissolve solid calcium carbonate the pH rising as a result. The old ‘marble test’ useful for providing a rough forecast of the corrosive character of a water consisted in determining the pH value before and after contact with marble dust the difference between two values being known as the Langelier Index; its significance as a measure of corrosive power has perhaps been overestimated since other constituents of a water affect behaviour but any water in which marble produces an appreciable pH rise is open to suspicion and should be subjected to more reliable tests based on iron or steel specimens.If a water is free from aggressive carbonic acid (i.e. if it would fail to dissolve calcium carbonate) it will quickly deposit on iron or steel a layer of chalky rust (or rusty chalk) possessing some protective character. If for instance such a water is run through a pipe any incipient corrosion will raise the pH by the cathodic reaction and a layer of chalk later converted in part into ferrous carbonate and thereafter to clinging rust will be deposited gradually spreading C. A. J. Taylor quoted by U. R. Evans ‘Corrosion and Oxidation of Metals’ Arnold London 1960 p. 165. lo G. Bodliinder 2. phys. Chem. 1900,35,23.J. Tillmans and 0. Heubleim Gesundheits-Zngenieur 1912 35 669. 33 QuartprZy Reviews over the surface as in the experiments with magnesium sulphate; this will be equally true whether the cathodic points are particles of mill-scale remaining on the surface or cementite particles within the steel or again areas where the oxygen is preferentially renewed. The protection is not complete but in absence of complicating factors the corrosion will be much less severe than that produced by water containing aggressive carbonic acid which can build no continuous ‘chalky film’. A non-aggressive water may become aggressive if softened by base-exchange since the replacement of Ca(HCO,) by NaHCO, even if only partial lessens the requirements for stabilising carbonic acid so that some of the H2C03 present becomes ‘aggressive’.The danger could be removed by passing the water over limestone or (better) calcined dolomite but this would re-introduce some hardness. Much of the excess of carbonic acid can be removed by simply passing the water down a series of cascades or even allowing it to overflow from the top of a vertical pipe. Such treatment has been found useful at water-works for avoiding corrosion complaints not only from softened water but also from water which contains carbonic acid derived from the vegetation of the collecting area especially at certain seasons of the year. Rusting of Totally-immersed Iron under Stagnant Conditions.-Long-period studies on horizontal discs fully immersed at a known distance below the surface have been carried out at Teddington,12 the reaction being followed by measuring the disappearance of oxygen from the air-space above the liquid generally a solution of potassium chloride.Under stagnant vibration-free conditions the rate of attack is usually slow being controlled by the rate at which oxygen reaches the disc. It is greatest when the upper surface of the disc is very close to the water surface. On a disc of fixed diameter the corrosion-rate rises with increasing breadth of the containing vessel which provides a larger area of water surface; if it is assumed that only a small fraction of water molecules striking a water surface pass into the body of the liquid this would be expected. Corrosion is accelerated if the oxygen pressure in the gas space above the liquid is increased but at a certain level (about 25 atm.) protective-film formation begins to retard attack; even at lower pressures (2-5 atm.) corrosion is diminished if the specimen is exposed dry to oxygen before the liquid is intr0d~ced.l~ Under conditions where oxygen-supply is limited ferroso-ferric products appear as well as ferric rust generally black magnetite rather than the green compounds mentioned above.In the Teddington experiments in ~ / 1 0 potassium chloride the corrosion-rate was constant for twenty days and then declined owing to an ‘alteration in the physical character of the anodic film’. For the first 500-600 days the corrosion product consisted of a thin layer of magnetite over- lain by a loose mass of hydrated ferric oxide which could be shaken off without la G. D. Bengough A. R. Lee and F. Wormwell Proc.Roy. SOC. 1931 A 134,308; 1933 A 140 399; Third Report of Corrosion Committee 1935 p. 123 (Iron and Steel Inst.). l3 G. D. Bengough and F. Wormwell quoted by U. R. Evans ‘Metallic Corrosion Passivity and Protection’ Arnold London 1946 p. 297. 34 Evans OH OH t affecting the corrosion rate. Later a change to a more coherent form was accom- panied by a steady decrease in the corrosion-rate. A French study14 of iron in distilled water or soft water also revealed fairly adherent magnetite covered with brown ferric rust. Apparently the magnetite formed the cathode of a corrosion couple and the iron the anode for the cor- rosion of iron was found to be stimulated by contact with magnetite whereas contact with rust had little effect. Rust Prevention by Inhibitors.-Whilst magnesium or calcium salts merely slow rusting certain inhibitors like sodium hydroxide sodium phosphate or potas- sium chromate can prevent it altogether.Instead of loose rust an invisible film is produced; the film formed on iron placed in a solution of sodium hydroxide or phosphate has been identified by electron diffraction15J6 as y-ferric oxide;* that formed in chromate generally contains chromium.17 In sodium hydroxide the film-formation can be ascribed to OH- ions driven up to anodic points by the current or attracted by adsorption forces; the anodic reaction consists in the removal of the H from the oriented hydroxyl leaving oxide on the surface [Figure 3(A)]; the process can be repeated [Figure 3(B)] until - + METAL LIQUID 1 OH I ME METAL @ LlWlD OH OH CI CI - CI C I OH the oxide becomes sufficiently thick for protection (or perhaps until the store of loosely bonded iron atoms becomes exhausted).If the liquid contains C1- in excess of OH- [Figure 3(C)] there will be places where C1- is adsorbed instead of OH- and the anodic reaction will there be the movement of a metal ion into the * The so-called y-ferric oxide often contains hydrogen and is sometimes regarded as magnetite with the ferrous ions replaced by pairs of protons (M. C. Bloom and L. Goldenberg Corrosion Scieuce 1965 5 623); this view is not held universally. l4 E. Herzog Bull. SOC. chim. France 1936 3 1530; 1938,5 187; Corrosion et Anti-corrosion 1964 12 No. 5. l5 J. E. 0. Mayne J. W. Menter and M. J. Pryor J . Chern. SOC. 1950 3229. 1' R. M. Brasher and E. R. Stove Chem. and Ind. 1952 171. J. E.0. Mayne and J. W. Menter J. Chem. SOC. 1954,99 103. 35 Quarterly Reviews liquid not the detachment of H from OH; the two reactions involve the same electron transfer but only if C1- is present will there be corrosion and rusting. If to a chloride solution OH- is added in quantity just insufficient to prevent corrosion at the most susceptible spots (e.g. where the iron atoms are most loosely packed) it will attack iron locally but intensely; the immune area pro- vides a large cathode for oxygen-reduction and since the attack is concentrated on small anodic spots the corrosion per unit area will there be rapid and the rate of penetration serious. A small anodic area surrounded by a large cathodic area generally represents a dangerous combination. The criteria deciding between (1) film-formation and (2) passage of ferrous ions into the liquid with rust formation where they interact with alkali from the cathode have been discussed elsewhere.l* Under nearly reversible conditions the preferred reaction will be that which leads to the greatest drop of free energy.But where a high anodic current density is imposed by happenings at the cathodic area there may be an insufficient supply of ions possessing the necessary activa- tion energy to pass through the positive zone and the iron may become ‘passive’ even when the free-energy criterion would predict continuing attack. A reducible substance capable of stimulating the cathodic reaction may thus favour passivity; under certain circumstances oxidising agents inhibit corrosion although the sit- uation is complicated.Pertechnetates although less strong oxidising agents than chromates as judged by the redox potention are better inhibitors as shown by work at Oak Ridge;lg they possess however only academic interest since the supply coming from certain atomic energy plants is limited. Atmospheric Conditions The R6le of Sulphur Dioxide.-Iron exposed to humid air in absence of dust and sulphur dioxide suffers little or no rusting. The rapid formation of rust in indus- trial and urban districts has long been attributed to acidity but earlier inves- tigators regarded carbon dioxide as responsible. Work at South Kensington and Teddington20 showed that sulphur dioxide was the main cause of atmospheric corrosion at inland places and that carbon dioxide could even retard attack. The opening stage of rusting in moist air containing sulphur dioxide proceeds rapidly and is probably analogous to the opening stage of the ‘fogging’ of nickel which was found in South Kensington work21 to require sulphur dioxide (or suspended sulphate particles) and a relative humidity exceeding a critical value (about 70 x).Sulphur dioxide is adsorbed and then combines at catalytically active points with atmospheric oxygen and water giving sulphuric acid which collects more water; this attacks the metal by well-established reactions today regarded as electrochemical producing iron (or nickel) sulphate. When once ferrous sulphate has appeared on an iron or steel surface the specimen can be l8 U. R. Evans First internat. Cong. met. Corrosion 1961. Report (Butterworths) p. 1. G. H. Cartledge J. Phys.Chem. 1955,59,979; 1956,60,28 1057; 1957,61,973; Corrosion 1965 21 217. 2o W. H. J. Vernon Trans. Faraday SOC. 1924 19 886; 1927 23 159; 1935,31 1678; Chem. and Ind. 1943 p. 318; J . Roy. SOC. Arts 1949 97 589. 21 W. H. J. Vernon J. Inst. Metals 1932 48 121. 36 Evans exposed to an atmosphere containing moisture but no sulphur dioxide and rust- ing can nevertheless continue; in fact an iron specimen carrying lines of ferrous sulphate crystals develops rust along those lines when exposed to a moist atmo- sphere. One atom of sulphur can cause the transformation into rust of many atoms of iron a point brought out by detailed researches at Stuttgart.22 Any theory of the formation of rust by ferrous sulphate must embody a regeneration mechanism; it must also explain the fact that whilst the outer part of the rust is loose (it can be wiped off on filter paper) the inner part is adherent and resists vigorous scrubbing.One suggestion is that the ferrous sulphate is oxidised to the ferric state and that hydrolysis produces ferric rust and sulphuric acid which then attacks a further quantity of iron. This ‘oxidative hydrolysis’ may indeed be responsible for the loose outer portions of the rust (the detection of a trace of ferric sulphate supports the idea) but it can hardly account for the tightly adherent layer. If the acid liberated by hydrolysis were continually attacking the metallic surface to which the rust was attached adhesion would be impossible. It is more likely that there is electrochemical action with the anodic attack on the iron in pits and reduction of oxygen on the face outside (Figure 4).It has been dern~nstrated~~ (A) (e> A l R Al R RUST FC S v 4 HzO RUST METAL METAL Cathodic bu rfacc pit Fig.4. that such action can indeed produce highly adherent rust at a well-arated cathodic zone. Since as many ferrous ions are produced by the anodic attack as are used up in cathodic rust-deposition complete regeneration is provided and the mechanism explains a curious fact that although soluble salts are easily detectable in the opening stages of rusting they seem largely to disappear later; presumably the anions migrate into the pits thus hiding themselves ‘under- ground’. Although in the early stages the pits may be crevices [Fig. 4(A)] they seem to extend laterally and after a time become saucer-shaped [Fig. 4(B)]. At that stage the ferrous sulphate becomes accessible.On steel exposed outdoors for some years it has been identified independently at N~rthwich~~ and Batter~ea~~ as FeS0,,4H20. The appearance of the tetrahydrate is significant and confirms the fact that the ferrous sulphate is an anodic product; towards 22 G. Schikorr Werkstofe u Korrosion 1963 14 69; 1964 15 457. 23 U. R. Evans Nature 1965 206 980. ** R. S. Thornhill personal communication 1952. 25 A. G. Tanner Chem. and Ind. 1964 p. 1027. 37 2 Quarterly Reviews pure water the heptahydrate is the stable phase but anodic action is to develop acidity and the presence of sulphuric acid would favour the for- mation of a lower hydratemZ8 It would be expected that sooner or later tho lateral extension of the anodic area would undermine the rust layer and this indeed occurs.The periodical shredding of atmospheric rust in large flakes has been noted by several investigators. In the early stages the cathodic reaction may be the reduction of 0 to OH-. When once rust has accumulated however the cathodic reaction can be an entry of ferrous ions into the ferric rust to form a ferroso-ferric compound which is then reoxidised to the ferric state by atmospheric oxygen.* This two-stage (catalytic) mechanism may well proceed more smoothly than the direct teduction of oxygen; the matter is under investigation. Whatever the details the electro- chemical mechanism of atmospheric attack is today generally recognised. In the advanced stages of outdoor rusting formation of fresh rust is seen preferentially at points where nests of ferrous sulphate exist in the existing rust; this supports the catalytic mechanism.The effect of ferrous sulphate nests is shown by work at S t ~ t t g a r t ~ ~ and also in connection with paint-breakdown at Cambridge.3o The presence of ferrous sulphate or chloride in rust is the main reason why painting over rust gives poor protection. Owing to their strong adhesion rust traces are difficult to remove; if shut in below a paint-coat they cause local formation of fresh voluminous rust which pushes away the coat and causes it to break. Corrosion Probability.-Common observation shows that when steel is exposed to the atmosphere rusting starts locally and parts may still be bright after long periods. It is necessary therefore to consider the probability of the inception of corrosion. One method of measuring probability31 consists of placing drops of distilled water on a horizontal surface and counting the proportion which developed rust.When the atmosphere is an oxygen-nitrogen mixture the probability declines as the oxygen content is raised although the 'conditional velocity' of the corrosion produced by those drops which corrode at all in- creases. Below the drops which produce no obvious change the iron probably * Omitting combined water we can write the cathodic reaction Fea+ + 4Fe,0 + 2e = 3Fe30,; this destroys the same amount of Fee+ as is liberated by the anodic reaction Fe I Fea+ + 2e. Since the oxidation of 3Fe30 by air will yield 4.5 Fe,O instead of the original 4-0 Fe,O, it is evident that the iron destroyed in the pits is appearing as rust on the face.Two- stage experiments now being carried out by C. A. J. Taylor (with iron exposed first to moist air containing SO and then to moist air without SO,) show that the rusting continues as quickly as it does when SO is present in the second stage; if between the two stages the iron is immersed in water there is practically no rusting in the second stage. This shows that it is the soluble iron sulphate and not the rust which causes the rusting to continue in the absence of SO,. Other experiments at Cambridge have provided evidence of the electro- chemical mechanism of atmospheric corrosion. *' C. Edeleanu and U. R. Evans Trans. Faraday Soc. 1951,47 1121. 28 T. P. Hoar personal communication 1965. e9 H. Schwarz Werkstofe und Korrosion 1965 16 93 208. 80 J. E. 0. Mayne J. Appl. Chem. 1959 9 673.31 R. B. Mears and U. R. Evans Trans. Furaduy SOC. 1935 31 527. T. P. Hoar Trans. Faraday SOC. 1937 33 1152. 38 Evans develops a protective film. It has been founds* that the presence of sulphur in the steel and also its presence in air (generaIly as SO,) increases both corrosion probability and also conditional velocity This largely explains the bad behaviour of unpainted steel in modern times and also the surprisingly good behaviour of ancient iron in unpolluted atmospheres. The influence of the sulphur dioxide content of the air on the rate of atmo- spheric rusting was clearly brought out in the British corrosion tests33 by a com- parison of behaviour at exposure stations representing different degrees of pollution and equally in the Germana and Indianss tests by a comparison of behaviour at different seasons.Oriental Iron.-Iron produced in Eastern countries in early times presumably with charcoal as fuel must have contained very little sulphur; some of the old iron beams pillars or chains are in places still remote from industrialisation where the air is almost free from sulphur dioxide. The Delhi pillar probably erected in the fourth century and moved to the present site in the twelfth century provides an example. Three analyses quoted in an Indian publicationM record only a ‘trace’ of sulphur in the portion above ground and 0.008 % in the under- ground part. The upper portion of the pillar has remained rust-free and most of it is described as bronzy or bluish by different writers; evidently of the two alternative reactions oxide-formation has prevailed and during sixteen cen- tunes exposure at temperatures periodically elevated by the sun the film has reached visible thickness.The Reviewer is inclined to ascribe the immunity to low probability rather than low velocity; in the lowest part (possibly owing to salts from the soil) corrosion has occurred. However numerous other views have been expressed. One author3’ attributes the good behaviour simply to the absence. of atmospheric pollution; others to the absence of manganese36 or the presence of pho~phorus.~~ Mediaeval British Wrought Iron.-The good performance of wrought iron in our mediaval buildings is responsible for the belief that if only the traditional process of making wrought iron could be revived rusting would be avoided. It is however probable that the absence of sulphur compounds from the air in early days was responsible for the comparative absence of corrosion trouble.Exposure tests38 carried out about 1931-1938 showed that wrought iron suffered 32 R. B. Mears Carnegie Schol. Memoirs (Iron and Steel Inst.) 1935,2A 69. 33 J. C. Hudson Sixth Report of Corrosion Committee 1959 p. 9 (Iron and Steel Inst.). 34 G. Schikorr Werkstoffe und Korrosion 1964 15 457. 35 B. Sanyal and P. D. V. Bhadwar J. SOC. Ind. Res. (Kanpur) 1951,18A 69; B. Sanyal G. K. Singhania and V. K. Nigam Lardev J . Scf. Tech. India 1965 3 104. 36 National Metallurgical Laboratory Technical Journal (Jamshedpur) Feb. 1963 (Delhi Iron Pillar number). The paper by W. E. Bardgett and J. F. Stanners on p. 24 appeared also in J. Iron Steel Inst. 1963 201 3. 37 J. C.Hudson Nature 1953 172 499. 38 J. C. Hudson ‘Corrosion of Iron and Steel’ (Chapman and Hall) 1940 p. 82; J. C. Hudson and J. F. Stanners J . Iron Steel Inst. 1953 180 27. 39 Quarterly Reviews corrosion only slightly more slowly than ordinary mild steel and at a rate rather similar to steel containing 0.2% of copper. It is right to note that the wrought iron used for these tests probably contained more sulphur than the iron made in medizeval times when charcoal was used as fuel both in the puddling furnace and the forge. But Swedish iron made with charcoal corroded more rapidly than the British iron. The subject is complicated by the fact that British wrought iron after puddling and piling contains much slag and consists of a series of parallel layers some very susceptible to attack and others resistant;39 as a result the corrosion velocity parallel to the layers is much faster than in the normal direc- tion.It would however not be possible when beating out iron ornaments at the forge to arrange that the resistant layers are always parallel to the surface. There is little doubt that it is the increased sulphur content of the air and not that of the metal which explains the enhanced rusting of modern times. Low-alloy Steels.-It has been mentioned that the addition of 0.2 % of copper to mild steel reduces the corrosion-rate roughly to the level of wrought iron. Small amounts of nickel chromium aluminium and molybdenum (often in combina- tion with copper) provide even better results as shown by extensive tests in the U.S.A.40 and the U.K.38 These ‘low-alloy’ steels are not non-rusting and must be distinguished from the stainless steels containing 13 % of chromium (or in the austenitic type 18 % chromium and 8 % nickel).They do however corrode more slowly than unalloyed steel and are relatively cheap; one commercial product with 0.5 % copper 1.0 % chromium 0.16 % phosphorus and 0.8 % silicon corrodes at about one-third of the rate of ordinary mild That chromium gives protection will cause no surprise but the action of copper is less easy to explain. An American authority,4O who observes that the rust on the more resistant steels is darker more adherent and of finer texture than that on the less resistant believes that the presence of copper or nickel locks up the sulphate in insoluble form as a complex basic salt. The Role of Salt.-It is common knowledge that atmospheric rusting occurs quickly near the sea.At places where both temperature and humidity are high corrosion is particularly rapid ; probably the presence of hygroscopic magnesium chloride along with the sodium chloride here helps corrosion retarding evapora- tion of droplets that strike a steel surface.4l Careful study in West Africa42 has shown how the amount of air-borne chloride and with it the corrosion-rate decline as the distance from the sea increases; later trials conducted in the U.K. have brought out a similar relationship. Even far inland dry salt particles may be present in dust; if these are hygro- scopic they may set up corrosion where they settle on steel. The fact that dust 8s J. P. Chilton and U. R. Evans J . Iron SteeZInst. 1955,181 113; 1957,185,497; 1957,186 98.40 H. R. Copson Proc. Amer. SOC. Testing Materials 1945 45 554; 1948 48 191 ; 1952 52 1005. I1 U. R. Evans and S. C. Britton J . SOC. Chem. I d . 1930,49 173T. 4a H. R. Ambler and A. A. J. Bain J . Appl. Chem. 1955 5 437; 1960 10,213. 40 Evans can cause rusting has long been known and classical work at South Ken~ington~~ has shown that different types of particle behave very differently; silica was found to be harmless but particles of ammonium sulphate initiated rust spots. When a steel surface was exposed to dust-free air for eleven weeks it was found to become relatively immune from rusting even if later dusty air was admitted. It should not of course be assumed that it requires eleven weeks to produce a film over the surface; optical measurements at TrondheimM show that an invisible oxide-film 20 A thick can appear within 10 minutes; probably the need for the long exposure is due to the fact that at first the film keeps cracking and only after internal stresses have been used up in repeated cracking followed by repair can the invisible film contribute effectively to protection.Particles of hygroscopic salts are the most dangerous at least indoors and an analytical study of dust in factories might assist corrosion control. Early work at South Kensington and Teddingt0n~O3~~ has shown the importance of the relative humidity (R.H.) of the air. Interesting results have also been obtained at Saturated sodium bromide solution stands in equilibrium with air of 59 % R.H. at 20”c; thus a solid particle of sodium bromide will become damp if the air humidity exceeds 59%.It was found that steel inoculated with sodium bromide remained unrusted in air of 50 % R.H. and developed rust at 60 % R.H. But the passage from immunity to rusting was not always sharp. Saturated sodium chloride solution is in equilibrium with air of 78% R.H. Steel carrying sodium chloride becomes strongly rusted at 80% R.H. and remains bright at 60%; at 70% the salt particles become brown and there is some attack. Among hygro- scopic salts liable to set up rusting even in unusually dry air may be mentioned the chlorides of zinc lithium magnesium and calcium; their saturated solutions stand in equilibrium with air of lo% 15 % 32 % and 32-3 % R.H. respectively. The rusting set up at a point where a salt-particle rests on iron or steel may spread if humidity conditions are favourable.Uniform spreading gives expanding circles but frequently if through some irregularity advance commences more quickly at certain points on the periphery than others thread-like growths develop. This ‘filiform’ corrosion cannot be discussed here. The reader is referred to an excellent paper from Teddington,46 in which an electrochemical mechanism is favoured. Final Remarks General Conclusions.-The place of formation of rust is all-important. Under immersed conditions a hard water which can throw down a sparingly soluble compound by the cathodic reaction so that the rust is formed in contact with the metal will cause less rapid corrosion than a water causing loose rust. Under atmospheric conditions the closely adherent character of the rust is a less welcome feature making removal difficult; the rust usually contains soluble 43 W.H. J. Vernon Trans. Faraday SOC. 1924 19 886; 1927 23 159. 44 A. B. Winterbottom Trans. Electrochem. SOC. 1939 76 326. 46 A. Bukowiecki Schweizer Archiv. angew. Wiss. 1957 33 97. 46 R. St. J. Preston and B. Sanyal J . Appl. Chem. 1956 6 26. 41 Quarterly Reviews ferrous salts paint coats applied over rust-traces may be pushed away at the situations of salt-nests suffering breakdown. Special paints designed to con- vert ferrous sulphate to harmless compounds are described elsewhere?' Thanks are due to F. Wornwell R. S. Hudson IT. F. Stanners and J. E. 0. Mayne for useful discussions and to C. A. J. Taylor who drew the diagrams. 47 U. R. Evans and C. A. J. Taylor Trans. Insr. Metal Finishing 1962,39 188; 1965,43 169. 42