SECOND EXPERIMENTAL REPORT TO THE ATMO-SPHERIC CORROSION RESEARCH COMMITTEE ASSOCIATION). BY W. €3. J. VERNON. (Received I 6th December 1926 and read before THE FARADAY SOCIETY on 30th March 1927.) TABLE OF CONTENTS. INTRODUCTION . . . . . . . . . . . . . I 15 PAGE PART I. INDOOR EXPOSURE TESTS AND LABORATORY EXPERIMENTS. SECTION I. COPPER. A. Introduction . . . . . . . . . . . . I 17 B. Experimental. Atmosphere and specimens employed in normal exposure tests. Method of weighing. Apparatus employed in tests with synthetic at-Effects Chemical cleaning ; reagents employed in tests, mospheres . . . . . . . . . . . I 17 C. Influence of Surface Condition upon Rate of Tarnishing. of Chemical Cleaning. Mechanical cleaning. immediate and subsequent effects of treatment .. . . . 1x9 D. Influence of Purity of the Metal upon Rate of Tarnishing . . 121 E. Influence of Sulphur Content of Atmosphere upon Rate of Tarnish-Determination of and influence of “reactive sulphur content ” of atmo-ing. Composition of Tarnish Films. sphere. Determination of sulphide content of tarnish filrns . . 122 F. Relation between Weight-increment and Time. (I) Oxidation at the ordinary temperature with production of tarnish colours. Theoretical considerations ; “ short-period ” and 6‘ long-period ” tests ; constancy of parabolic relationships under fluctuating -atmospheric conditions . . . . . . . . . 124 (2) Oxidation at the ordinary temperature without development of colours 127 (3) Oxidation at temperatures above normal with development of ‘6 temper colours .. . . . . . . . . . . . . 129 G. Properties of Oxide Films developed in Relatively Pure Atmosphere at the Ordinary Temperature . . . . . . . . 130 H. Properties of Oxide Films developed at Temperatures above Normal 130 I. Properties of Sulphide Films. The properties of initially-pure sulphide films compared with those of J. Effects produced by ‘‘ Smoke Films.” . . . . . . . 1x3 9 oxide and oxy-sulphide films. The mechanism of tarnishing . I33 13 = I 4 TABLE OF CONTENTS PAGE SECTION 11. ZINC. A. Atmosphere Type I . ‘‘ Basement.” (I) Short-period intensive tests ; relation between weight-increment and time . . . . . . . . . . . . . 135 (2) Long-period tests; relation between weight-increment and time . 137 (3) Note upon effect of surface condition .. . . . . 138 (4) Note upon visual results I39 (5) Microstructure of surrosion films ’ 139 . . . . . . . . . . . . . €3. Atmosphere Type 2. ‘‘ Tank-room.” Relation between weight-increment and time. Rate of attack as com-pared with rate of attack in Atmosphere Type I. Conclusions . 139 SECTION 111. BRASS. A. Relation between Weight-increment and Time. Influence of relative humidity; conclusions . . . . . . 140 B. Relation between Intensity of Corrosion and Composition of Alloy 142 D. Properties of Oxide Films developed at the Ordinary Temperature 142 E. Properties of Oxide Films developed at Temperatures above Normal 143 F. Use of Lanoline for protecting Brass exposed to Atmosphere of C. Microstructural Changes on Atmospheric Exposure .. . 142 relatively High Humidity Optical and gravimetric results . . . . . . . . 144 SECTION Iv. RELATIVE BEHAVIOUR OF COPPER ZINC AND BRASS EXPOSED TO SEVERAL TYPES OF INDOOR ATMOSPHERE. A. EXPOSURE IN ATMOSPHERE TYPE I . Varying but mainly low relative humidity . . . . . . 146 Varying but mainly high relative humidity . . . . . . 146 Analysis of Corrosion Products . . . . . . . . . 146 B. EXPOSURE IN ATMOSPHERE TYPE 2. C. Exposure in Domestic Kitchen. SECTION V. ALUMINIUM. A. B. C. D. Introduction . . . . . . . . . . . . I 50 Experimental. Atmosphere specimens and method of weighing . . . . . 150 Relation between Weight-increment and Time. ‘‘ Short-period ” and ‘( long-period ” tests ; special type of weight-incre-Thickness of Surrosion Film.Influence of Purity of Atmosphere, ment curve . . . . . . . . . . . . 152 . . . and of Purity and Physical Condition of Metal * 1-54 SECTION VI. LEAD. A. Experimental . . . . . . . . . . . 1 5 6 B. Relation between Weight-increment and Time . . . . 156 D. Tarnishing of Lead in Atmosphere contaminated with traces of E. Protective Properties of Invisible Oxide Films upon Lead . . 158 C. Oxidation Values of Aluminium and Lead compared . . - I57 Vapours from Drying Paint ‘57 . . . . . . . SECTION VII. IRON. A. Experimental * ‘59 . . . . . . . . . . . B. Behaviour in Indoor Atmosphere of Low Relative Humidity con-taining Suspended Solid Particles. Effect of filtering air and of screening the specimen . . . . 15 115 W. H. J. VERNON PAGE C. Behaviour in Atmosphere Saturated with Water Vapour .. 161 D. Behaviour in Indoor Atmosphere of High Relative Humidity. I. The iron is initially free from rust . . . . . . . 162 2. The ironis initially covered with dry rust . . . . . . 162 E. Formation of Invisible Protective Oxide Films. Analogy with copper and with lead. Note on passivity . . . . 164 F. Influence of Purity of Metal upon Rate of Rusting . . . . 165 G . Behaviour of Iron compared with that of Non-ferrous Materials . 165 PART 11. OPEN-AIR EXPOSURE TESTS. INTRODUCTION . . . . . . . . . . . . . . 166 SERIES A. COPPER ZINC AND BRASS. Experimental. DISCUSSION OF RESULTS. 100 WEEKS’ EXPOSURE (INTEN-SIVE TESTS). Specimens employed ; method of conducting tests . . . . . 166 ( I ) “ Erosion ” values.Metal attacked and removed in rain. Influence of surface ‘‘ skin ” from manufacture of physical condition of metal and of mode of suspen-sion of specimen. Relative erosion of materials ; relation between (2) ‘‘ Surrosion ” values. erosion and time. Dezincification of the brasses . . . . 167 corrosion products; dezincification . . . . . . . 172 Metal attacked but remaining in situ. Computation ; analyses of residual (3) Total corrosion. Computation ; relation between loss in weight and total corrosion . . 174 SERIES B. COPPER (VARIOUS GRADES) ZINC AND BRASS. 225 WEEKS’ EXPOSURE. Analysis of residual corrosion products . . . . . . . 176 Determination of weight-losses and computation of total corrosion . . 177 Relative behaviour of materials. Influence of size of specimen .. . . . . . . . 178 (I) Co#fav.-Influence of modification of composition. Possible con-nection between loss of reflectivity in early stages and ultimate resistance to corrosion. Relation between surrosion and total corrosion . . . . . . . . . . . 178 (2) Zinc and Brass . . . . . . . . . . 178 SUMMARY AND APPENDIX . . . . 179 ZiVTXOD UCTIOA? The ‘ First Experimental Report ’ to the Atmospheric Corrosion Re-search Committee was presented and discussed a t a meeting of the Faraday Society on December 17th 1923.1 The earlier work was occupied with the systematic examination of the behaviour of the commoner metals on ex-posure to several types of indoor atmosphere and to the open air employing both loss of reflectivity and increase in weight to estimate the changes taking place at the metal surface.Probably the outstanding result lay in the recognition of three main types of curve connecting weight-increment with time the significance of which in relation to the function of the reaction-product was also discussed. I n the work recorded in the present report the examination of the various materials has been carried to a further stage; additional (and Trans. Faraday Soc. 1924 19 839 116 SECOND EXPERIMENTAL REPORT generally more intensive) experiments have been instituted. These have been concerned chiefly with the mechanism involved in the formation of various films on exposure to air with particular reference to the conditions under which protective oxide films may be obtained. To this end attention has been mainly directed upon copper partly because of its intrinsic import-ance and partly because of its special predilection to tarnishing.I t was thought that information so gained might be usefully applied to the ex-amination and treatment of the many alloys in which copper is the major constituent. At the same time zinc and brass have received more critical attention than was the case in the First Report. A new series of tests upon the metals aluminium and lead has necessi-tated the recognition of another distinctive type of weight-increment curve. The experiments upon iron originally included purely for comparative purposes have been continued and have brought to light certain factors in the atmospheric corrosion of iron which appear hitherto to have escaped attention.The arrangement of matter in the present communication differs from that previously adopted. The Discussion on the First Report showed that some confusion had arisen through the use of the term ‘‘ Field Tests.” I t was pointed out that in most cases these had been conducted under purely laboratory conditions. In later work the ‘‘ indoor exposure tests ” and ‘( laboratory experiments ” have been even more closely identified and distinction between them has become increasingly difficult. Accordingly, they are now grouped together in Part 1. of the Report whilst Part 11. is devoted to the consideration of tests conducted upon specimens exposed freely to the open air. Even in the case of the open-air tests laboratory methods (for example optical measurements in the early stages and chemical examination of products both during and after exposure) have been em-ployed as far as possible and have enabled definite conclusions to be reached.In the lntroduction to the First Report references to the earlier literature were given and their paucity was commented upon. A list of cognate papers which have been published subsequently is given below from which it would appear that an increasing amount of attention is being devoted to the behaviour of the non-ferrous metals on exposure to the atmosphere. U. R. EVANS. “ Relation between tarnishing and corrosion,” Trans. Amer. Electro-chem. SOC. 1924,46,247. G. W. VINAL and G. N. SCHRAMM. “The tarnishing and de-tarnishing of silver,” Met. Ind. (N.Y.) 1924 22 15 110 151 231.D. H. BANGHAM and J. STAFFORD. “The velocity of oxidation of the metals and the structure of coioured oxide films,” Nature 1925 115 83. U. R. EVANS. The colours due to thin films on metals,” Proc. Roy. Soc. 1925 (A), 1q,228. U. R. EVANS. “ The production of oxide films on copper at the ordinary temperature,” 3’. Chem. SOC. 1925 127 2484. E. WILSON. ‘‘ The electrical conductivity of certain light aluminium alloys and copper conductors as affected by atmospheric exposure.” J. Inst. Elect. Eng. 1925, 63 1108. J. S. DUNN. The high temperature oxidation of metals,” Proc. Roy. Soc. 1926 (A), 111 203. J. S. DUNN. “The low temperature oxidation of copper,” Proc. Roy. Soc. 1926 (A), 111 210. U. R. EVANS. ‘ I Temper colours tarnish-colours and other tints on metals,” Chem.and Ind. 1926 45 211 W. H. J. VERNON W. H. J. VERNON. E. WILSON. b L The formation of protective oxide films on copper and brass by exposure to air at various temperatures,” y. Chem. SOC. 1926 128 2273. (‘The corrosion products and mechanical properties of certain light aluminium alloys as affected by atmospheric exposure,” Proc. Physical Soc. 1926, 3 9 9 15-PART L-INDOOR EXPOSURE TESTS AND LABORATORY EXPERIMENTS. SECTION L-COPPER. (a) Introduction. In previous work (First Report Zoc. cit. 897) the tarnishing of copper under ordinary conditions (i.e. as initiated by traces of sulphur compounds in the atmosphere) appeared to be favoured by a low relative humidity and hindered when the concentration of water vapour was excessively high.The mechanism was concluded to be of a simple chemical character gas and metal reacting directly and moisture playing a secondary part (ibid. 927). On the other hand experiments with an atmosphere charged with hydrogen sulphide yielded apparently contrary results increasing humidity then con-sistently increasing the rate of attack ($id. goo). I t has since been shown by U. R. Evans however (Trans. Amer. Electrochem. Sol. 1924 &,.26>), that z;f the concentration of hydrogeen sur’phide is suflcieently Zow tarnishing is actually hindered by the presence of a film of moisture on the metal surface a result which brings the two sets of phenomena into agreement. The present work has shown that although gaseous sulphur impurities may still play an essential part when present in excessively low concentra-tions under these circumstances the resulting tarnish film consists mainly of oxide.Examination of the behaviour of films consisting initially of sul-phide has thrown useful light upon the mechanism of tarnishing in air. The properties of oxide films resistant to tarnishing have also been studied in relation to the conditions under which they may be obtained. (b) Experimental. In the main series of tests specimens have been exposed to that indoor atmosphere which was described in the First Report as “Atmosphere Type I. (Base-ment) ’’ (loc. tit. p 853). I t will be recalled that a pronounced change in hygrometric conditions takes place in May when the artificial heat supply is cut off from the building and again in October when heating is re-started.This is brought out by the following figures; it is noteworthy, however that the mean temperature alters very little. Atmosphere Employed i?z the Normal Exposure Tests. Period S. Period W. May-September. November-April. Mean temperature . . 64O F. 67’ F. Mean relative humidity . . 68 per cent. 43 per cent. Mean daily difference between minimum temperature and dew -poin t . . 8” F. 22” F. Under a later heading the influence of gaseous sulphur impurities in very low concentrations is discussed and a method for their determination is described. For the moment attention is drawn to the numerical result 118 SECOND EXPERIMENTAL REPORT which are plotted in Fig. I representing the variation in “ reactive sulphur content” throughout the year.This diagram is comparable with Fig. I of the First Report in which the hygrometric values were plotted. As before, a vertical line marks the approximate division between the two periods but the contrast is now considerably greater. During ‘ I Period W ” not only is the reactive sulphur content generally high but the relatively wide limits within which it varies should be especially noted. In the tests to be described the specimens have been taken from high-grade electrolytic copper sheet 1-2 mm. in thickness. Their dimensions have been either 10.0 x 10.0 cm. as in the earlier tests or 10-0 x 5.0 cm. in some of the more recent work. The method of abrasion described in the First Report has been relied upon for the preparation of the specimens used in the main series of tests the final surface having been obtained by means of Hubert No.I emery paper as before. For the final cleaning however the use of Specimens EmpZoyed in the Nornzal Exyosure Tests. I 1 I I 16.0 I I 1 I I -PERIOD W ADP~L MAV 23 23 6 14.0 -IZ.0 -10.0 -8.0 -6.0 - PERIOD S 4.0 -2.0 -MAY JUNE Src. OCT. Nsv.’ DEC. JAN. FEB. 6 3 18 5 16 17 I2 Date of Sulphur Estimation. Fig. 1.-Indoor atmosphere employed in tests (“ ttmosphere Type I.”). Variation in ‘ reactive sulphur content throughout the year. N.B.-The dotted portion of the curve represents values which are below the limits of satisfactory measurement ; they may be regarded howezer, as maximum values so far as the actnal “ reactive sulphur content is concerned.organic reagents has been discontinued and dry cleaning with pure cotton wool has been substituted. The method consists in rubbing the specimen repeatedly until no further mark can be obtained upon a piece of the clean wool. Immediately after this treatment the specimens are transferred to a vacuum desiccator and allowed to stand in vacuo for weighing and ex-posure the following day. has been followed throughout the work and has proved of the utmost value. A complete weighing includes four partial weighings each of which in turn involves four readings of the pointer oscillating entirely to one side of the zero. Between partial weighings the specimens and the weights are in-terchanged. I n the course of the present work it has been found ad-vantageous to repeat the whole set of readings with the pointer swinging entirely to the other side of the zero thus making eight partial weighings as one complete “ unit.” The method has enabled specimens such as those The author has enjoyed the advantage of the personal advice of Professor Conrady whose kindness he gratefully acknowledges.Methodof We&-ing. The system of weighing evolved by A. E. Conrady PYOC. Roy. SOC. 1922 A 102 211 W. H. J. VERNON 1 19 described above to be weighed-consistently correct to 0-01 mg. and in favourable circumstances correct to 0.005 mg. A certain amount of work has already been done with synthetic atmospheres and it is hoped to develop this line of investigation more extensively in the im-mediate future. The apparatus consists essentially of two desiccators each of approximately 4300 C.C.capacity fitted with ground-in stoppers and connecting tube with tap. There are also connections to (i) a purification train through which air can be led (ii) a small gas-burette from which addi-tions of known amounts of H,S or other gas can be made (iii) a barometer tube and (iv) a vacuum pump. After passing through the purification train the air can be led either through several drying-towers containing phosphorus pentoxide or through wash-bottles in which it becomes saturated with water vapour thus enabling any desired humidity to be obtained the volume of air entering the desiccators being obtained from readings of the Sarometer tube. One of the desiccators contains the specimens under examination ; the other is used as a preliminary mixing chamber for the air.In starting an experiment both desiccators are evacuated and the tap between them is closed. The gases necessary to make up the tarnishing atmosphere are then led into the mixing chamber. After an interval for diffusion the mixture is allowed to divide itself between the two vessels and finally is brought up to atmospheric pressure by the admission of purified air. Apparatus EmpZoyed in Tests with Synthetic Atmospheres. (c) Influence of Surface Condition upon Rate of Tarnishing. Effects of Chemical Cleaning. MecAanical Cleaning I n the First Report it was recorded that surfaces prepared by abrasion with fine emery paper were more susceptible to attack from tarnishing than those in the brightly polished condition ; the difference was capable of estimation optically or gravimetri-cally although not obvious to visual in~pection.~ The inference that the attack upon emeried surfaces should increase with the coarseness of the emery employed has been confirmed by experiment.Specimens were prepared with fine emery paper (Hubert No. I) and coarse emery cloth (FF) respectively and were allowed to tarnish under identical conditions. After 36 days’ exposure the fine and rough surfaces had increased in weight by 0.93 and 1-42 mg. respectively (the figures are the means of closely agreeing duplicate results) ; again however there was very little to choose between the appearance of the specimens. Throughout the main tests the method of abrasion has been relied upon for the preparation of specimens since it was felt that any method of chemical attack would be open to objection on account of the possibility of specific effects due to the reagents employed.Neverthe-less the advisability of obtaining information concerning any such effect has been kept in mind and to this end a limited number of experiments have been carried out. Reagents employed in the Tests and Immediate Efects of Treatment. Used boiling this reagent removes slight stains and very thin oxide films. It has practically no etching effect and no effect on the colour of the copper. ChmicaZ Cleanifig. Ammonium ChZooride 10 per cent. Solution. 3 Gravimetric results from ‘I bright ” and dull ” specimens ‘after 1300 days’ exposure together with results 1 from comparable zinc specimens will be found on p.138 I 2 0 SECOND EXPERIMENTAL REPORT Nitric Acid (Conc. HN03 I vol ; HzO I vol.) Used cold or warm, this removes relatively thick films of oxide. I t etches the metal deeply and, unless the time of treatment is extremely short leads to severe pitting. Copper cleaned in this way is slightly darker than that cleaned by abrasion alone. (Conc. HN03 I vol ; Conc. H,SO,, 2 vols.) Of the reagents tried this gives much the best surface to the metal. I t does not etch nearly so deeply as the nitric acid alone and does not lead to pitting. Chromic Acid SoZutions. (I) Chromic Acid 20 per cent. aqueous solution. (2) Chromic Acid 10 per cent.; Ferric sulphate 5 Fer cent.* (3) Potassium Dichromate 12 per cent. ; Sulphuric Acid 12 per cent.5 These solutions each of which has chromic acid as its effective constituent, are approximately equivalent in respect to their action upon copper.They dean the surface quite thoroughly but tend to leave a somewhat patchy appearance; moreover the colour is definitely darker than that yielded by the other methods. The best effect is produced by a previous treatment with the sulphuric-nitric solution finishing with the chromic mixture (3), whereby " patchiness " is avoided and a slightly paler colour obtained. Subsequent E'ects of CZeaning Treatment. Specimens cleaned as above (due precautions being taken to remove all traces of reagent by sub*equent washing and drying) were exposed to the basement atmosphere during April 1926. The results obtained are summarised below.Ammonium Chloride Solution. Visually there is nothing to choose between the behaviour of specimens prepared with emery (Hubert No. I ) only and similar specimens which have also been treated with ammonium chloride solution. Whilst specimens so treated have given slightly greater weight increments the difference has been too small to justify any signifi-cance being attached thereto. This result should be compared with the effects of ammonium chloride upon specimens which have previousIy been treated with chromate solutions (see below). Again there is practically no differ-ence in the subsequent behaviour of chemically cleaned specimens and those prepared with fine emery. Moreover specimens finished by the two methods respectively have given remarkably good agreement in weight-increment.Specimens cleaned with any of the chromic acid solutions have yielded anomalous results in that they have displayed considerable immunity from tarnishing on subsequent exposure.6 Thus, after 5 0 days' exposure the mean weight-increment of such specimens was 0.36 mg. per sq. dm. whereas the mean weight-increment of emeried specimens exposed simultaneously was 0.95 mg. per sq. dm. Whilst these results are not surprising in view of the known passifying action of chromic acid they are especially significant in connection with results to be described later where a similar and even more marked protective effect has been obtained by a preliminary heating of the metal in air with the production of an almost invisible oxide film.Nitvic-Su&wic Acid Solution. Also it has practically no effect upon the colour. Nitric-Sz@huyic Acid Solutions. Chromic Acid Sulz~tiuns. The use of ferric sulphate was suggested by Mr. J. M. Stuart M.A. tj See paper by E. A. Bolton ('L The Removal of Red Stains from Brass "1 7. Insst. It should be noted that Bolton in the work already referred to (ibid. 155) remarks that brass which had been treated with the acid dichromate solution was L L resistant to atmospheric tarnishing." Metals 1925 331 143 W. H. J. VERNON I21 Interest attaches to the effect of following up the chromic acid treat-ment with immersion in hot ammonium chloride solution. A copper specimen 10.0 x 10.0 cm. was treated with chromic acid solution (2), washed thoroughly and dried.I t was then accurately halved and one por-tion was immersed for two minutes in boiling ammonium chloride solution (10 per cent.). A similar specimen was cut in two after cleaning with fine emery only and one half was treated with the ammonium chloride solution. The weight-increments yielded by the four specimens on subsequent exposure to tarnishing conditions for 50 days are given below :-Weight-increments. Mgms. per Sq. Dm. (I) No further treatment . . . . . 1-02 chloride solution . . . . . 1.10 Specimen initially emeried only (2) Subsequent treatment with hot ammonium-cia solution. I' ' chlohde solution . . . . . 1-13 I t will be seen that whereas the ammonium chloride solution has had practically no effect upon the emeried specimen it has entirely destroyed the immunity conferred by the chromic acid treatment.Again the result is not surprising since the activating effect of ammonium chloride is well-known ; it confirms however the existence of an oxide film upon specimens cleaned with chromic acid and also shows that while in the presence of a very thin oxide film the effect of ammonium chloride may be extremely marked in the absence of such a film it is practically without influence. (d) Influence of Purity of the Metal upon Rate of Tarnishing. Representative specimens of various grades of copper including copper containing respectively 0.45 per cent. arsenic 0.9 per cent. tin 2 - 5 per cent. nickel exposed in the basement atmosphere during period W have dis-played no difference whatever in appearance during the course of tarnishing.(For a description of the visible changes see First Report Zuc. cit. 855). The following weight-increments were obtained after a month's exposure :-Actual. Mean. H.C. Copper . . . . . {El! ) 1-89 mg. per sq. dm. Arsenical Copper (0.45 per cent. As) . { ~ ~ } I W ~ , , ,, Tin Copper (0'9 per cent. Sn) . . { ::$}1*,4 , , ,, Nickel Copper (2.5 per cent. Ni) . . { ::E!}r*68 , , ,, In view of the similarity in appearance it is not surprising that the differences among the mean weight-increments are very small ; nevertheless some definite conclusion can be drawn. Clearly 0.45 per cent. arsenic has had no measurable influence. The effect of 0.9 per cent. tin is only just appreci-able that of 2 - 5 per cent. nickel is distinctly more pronounced.In each case the effect is in the direction which would be expectedfrom the ordinary behaviour of the respective elements in the same atmosphere and is of a magnitude commensurate with the proportion present in the ~opper.~ I t 7 For example nickel is unaffected in the present type of atmosphere and it was shown in the First Report (Zoc. &. 882) that the rate of tarnishing of copper-nickel alloys is a function of the copper content. On the other hand in a more humid atmosphere copper-nickel alloys undergo a characteristic condensation or '' fogging " which is a I 2 2 SECOND EXPERIMENTAL REPORT may be concluded that the relative purity of the metal as ordinarily under-stood is without any appreciable influence upon the rate of tarnishing3 (e) Influence of Sulphur Content of Atmosphere upon Rate of Tarnishing.Composition of Tarnish Films. Determination of ‘‘ Reactizie Sudphur Content.” - The influence of the relative purity of the atmosphere upon rate of tarnishing was indicated in general terms in the First Report (loc. cit. pp. 897-8). For the purpose of obtaining more exact information a series of tests has since been carried out at representative periods of the year. Each test has consisted in the determination of the “ reactive sulphur content ” of the atmosphere whilst at the same time a freshly-cleaned specimen has been exposed and its weight-increment determined after a suitable i n t e r ~ a l . ~ By ‘‘ reactive sulphur content ’’ is understood sulphur in such a form that it responds to the “ alkaline lead acetate ” (sodium plumbite) test.The method actually adopted was as follows. The delivery tube of a Dreschel wash-bottle is drawn out to a jet leaving a clearance of about inch between the end of the jet and the bottom of the bottle. A bulb filled with cotton wool is sealed to the outer end of the tube and acts as an air filter. In the bottom of the bottle immediately under the jet is fitted a disc of filter paper saturated with alkaline lead acetate solution. Air is drawn through the apparatus at a rate of 18 cubic feet per hour and the volume required to produce a faint brown stain upon the paper is noted from an experimental gas-meter in series; this volume is assumed to be inversely proportional to the reactive sulphur content.The test has been calibrated under optimum conditions. Thus when 10 cub. ft. of air is required to produce the first stain the re-active sulphur content is equivalent to I volume of H,S in 35 million volumes of air a concentration which represents the most intensive tarnish ing conditions experienced during the present ‘‘ normal ” tests. Probably the lowest limit of sulphur content to which the method can be satisfactorily employed is about I part in 600 million for the estimation of which the passage of I 70 cub. ft. is necessary. Beyond this the incrustation of sodium carbonage seriously interferes with the detection of stain and hence vitiates the result. Fortunately however the range indicated above is sufficient for the purpose under discussion. Results of tests covering a period of 12 months are plotted in Fig.J . Influence of “Reactive Su&huu Content.”-The results of a series of sulphur determinations together with corresponding tarnishing results over a period including the maximum and minimum values referred to above, are set out in Table I. It may be noted at this point that during the months of July August and September (represented by a dotted line in Fig. I) tarnishing ceases i.e. the metal no longer passes through any colour changes. As to the corresponding reactive sulphur content all that can be said at present is that it is less than I volume in 600 million the lowest limit of estimation with the method described above. function of the nickel content. I t is significant that after exposure to the atmosphere of the ‘‘ tank room ” (Zoc.tit. 853) for a sufficient time the weight-increment of copper containing 2+ per cent. nickel is appreciably higher than that of the pure metal. * This is in striking contrast with the conclusion reached in connection with the open-air corrosion of copper (see Part z ) where 0-45 per cent. arsenic is shown to have a profound influence in reducing the total loss by corrosion. This procedure is justified by results described under the next heading which show that the rate of attack upon the metal even after comparatively long periods is de-termined very largely by the conditions prevailing at the time of the first exposure W. H. J. VERNON 123 Reactive Sulphur Content (Vols. H S Per 600 Milliot From the results in Table I.(see second and last columns) the interesting result emerges that if the rate of tarnishing be taken as proportional to the square of weight-increment per unit time lo then rate of tarnishing is direitly jro$ortionaZ to the reactive su@hur content of the atmosphere. Actual (Mg. per Sq. Dm. per Day)? TABLE J. Vols. Air). I 6.0 13'5 9'0 4'4 1'3 1'0 RELATION BETWEEN REACTIVE SULPHUR CONTENT OF ATMOSPHERX AND RATE OF TARNISHING OF COPPER. 0.177 0-1 65 0.1 18 0'042 0.015 0.0107 Atmosphere. Specimen : Rate of Tarnishing. Volume of Air Taken to Produce Date. I Stain (Cub. Ft.). Composition of Tarnish FiZms. The fact that the rate of tarnishing is directly proportional to the reactive sulphur content of the atmosphere renders it particularly expedient to enquire into the sulphur content of the resulting tarnish film.ll Copper specimens were exposed on Nov.21st 1923 and allowed to tarnish in the usual way. On July 16th 1925 the weight increment was 5.4 mg. per sq. dm. and determinations showed that of this increment 0.58 mg. was due to sulphur as su&hide i.e. 10.7 per cent. On Sept. Ioth 1926 the weight-increment was 6.1 mg. per. sq. dm. of which 0.61 mg. was due to sulphur i.e. 10-0 per cent. Other specimens were exposed on Dec. 19 1925 and on Sept. gth, 1926 the weight-increment was 2-6 mg. per sq. dm. of which c-4 mg. or I 5-7 per cent. was due to sulphur again as sulphide. These results show clearly that instead of the film consisting essentially of sulphide as was thought in the early days of the work sulphide enters only to a minor degree into its composition.I t is true of course that it is an essentiaZ constituent and the proportions found by experiment are certainly more in accord with the concentration of atmospheric sulphur upon which the tarnishing process depends. Nevertheless it is remarkable that such excessively small traces of sulphur in the atmosphere can determine the oxidation of the metal at the ordinary temperature to such relatively gross extents. I t may be noted that the foregoing results throw considerable light upon lo A little consideration will show that this is actually a sound method of computing the rate of attack where a parabolic relationship obtains between weight-increment and time. l1 The method ultimately employed for the estimation of the very small amounts of sulphur is stated in the Appendix.For the particular purpose of these experiments, specimens in the form of foil were employed as distinct from the thin plates used in the bulk of the tests SECOND EXPERIMENTAL REPORT the “edge effect” which characterises the tarnishing of copper in the atmosphere. (A description of this effect i.e. the propagation of tarnishing from the edges inwards was given in the First Report doc. cit. p. 856). Thus the formation of the tarnish film containing the percentages of sulphur noted above through exposure to an atmosphere in which the sulphur content is so excessively low requires not merely .contact with the atmosphere but requires also that the air in contact with the specimens shall be continuously replenished by diffusion.Previous results have shown how quickly and exclusively the ‘‘ tarnishing constituents ” are removed from air impinging upon copper (First Report duc. cit. 895). I t may readily be understood that the bulk of the diffusion of air over a given specimen will take place from the edges inwards and hence will account for the effects observed. (9 Relation betweem Weight-Increment and Time-r. Oxidation at the Ordinary Temperature with Production of Tarnish Colours. Thoreticad Considerations. The rate of formation of the films which give rise to tarnish colours,” in relation to period of exposure at constant temperature is a matter of some theoretical importance. It has for example an immediate bearing upon the question whether such colours are due to interference or to diffraction of light concerning which there has been divergence of opinion.According to the alternative views either the film takes the form of a continuous transparent envelope reflection from the upper and lower interfaces of which gives rise to interference or it has a granular structure and light falling upon its constituent opaque particles is diffracted as from a diffraction grating. On hypothesis the magnitudes involved i.e. thickness of film or spacing of particles must have dimensions comparable with those of the wave-lengths of light. Clearly if interference is the correct explanation the metal must be covered by a continuous film from the earliest stages and the process of tarnishing must be regulated by the diffusion of the atmosphere through this film which itself thickens con-tinuously as the result of the process.I n this case a parabolic relationship between thickness of film and time of exposure is to be expected, where W is the increase in weight of the specimen in time t and K is a constant.12 In the earlier work (First Report Zoc. cit., 863) equation ( I ) was found to hold true for the tarnishing of copper at ordinary temperatures subject to the qualification that the first point plotted was always at an appreciable distance from the origin (11 days was in fact the earliest period at which weight-increment was determined). Having regard to the improvements in technique effected since the publica-tion of the First Report it was felt desirable to determine the actual form of the curve in the very early stages.Accordingly a series of careful weigh-ings has been conducted (at daily intervals during the first week or so) upon specimens exposed at various times. From the curve given in Fig. 2, IZ The matter has been summed up recently by U. R. Evans (“ The colours due to thin films on metals,” Proc. Roy. SOC. A. 107 1925 228) chiefly however with reference to the cognate phenomena of “ temper colours ” due to oxidation at tempera-tures above normal. Evans shows that the evidence strongly supports interference as the getzeral explanation. During the formation of the film the colours appear in their correct order ; moreover by reducing the thickness of the oxide film upon iron electro-lytically he has succeeded in reproducing earlier colours in their appropriate order.i. e. W 2 = K t . * (1) Inteiisive Short-Period Tests iV. H. J. VERNON 125 which is quite typical of the results obtained it will be seen that the early part is very nearZy LZ stra2ht Zzne. The squares of the weight-increments show that from a certain point onwards ( “ P ” in Fig. 2 ) the curve is accurately parabolic and moreover the position of this point marking the transition from the straight line to the parabola is fairly sharply defined. A simple explanation of this state of affairs is readily forthcoming and is doubtless connected with the fact already noted that in the case of a flat specimen the attack does not begin at all parts of the surface simultaneously, but proceeds from the edges inwards.Clearly the method of calculating 60 80 100 I20 I40 I60 Time in days. Fig. 2.-Oxidation (tarnishing) of copper at ordinary temperature in (See Fig. 1, Square specimens. Reration between weight-increment = Squares of weight-increments (values taken from smooth atmosphere containing traces of sul hur compounds. Period W). and time in the very early stages. curve) x 0.4. the actual weight-increment upon the total area cannot be expected to yield a parabola during the period when only a portion of this area is actually undergoing attack. The matter is rendered clearer if it be re-membered that equation (I) as it stands is only true for a surface the whole of which is being simultaneously attacked unless indeed it is re-garded as representing only a small element of the surface.Otherwise it should be written (z)2 = Qt. . where A is the actual area which has suffered the change represented by the increase of weight W at any particular time t. But in the exampl I 26 SECOND EXPERIMENTAL REPORT under consideration where the process depends upon the steady diffusion of the atmosphere over the surface a little consideration will show that the area undergoing attack must itself bear an approximately parabolic re-lationship with time i.e. substituting this value in (2) we get A2 = Rt . - (3) or W = K t . * (4) W2 = (QR)P 1.e. a linear relationship in agreement with the experimental result. I t should be remembered of course that in the present instance this equation has no fundamental significance such as attaches to the same expression in the case of zinc but that it is merely the result of compounding two equa-l I 1 I I I I I I b - ' I 1 1 1 I I I I I 1 I I 1 .H 100 tM) UK) 4.00 So0 600 700 800 900 1000 1100 I200 ifm Time in days.Fig. 3.-Oxidation (tarnishing) of copper at ordinary temperature in atmosphere periodically contaminated with traces of sulphur compounds. (See Fig. 1). Relation between weight-increment ; and time up to 1300 days exposure. 1:' = Squares of observed weight-increments ( x 0.5). tions one only of which is fundamental. Moreover it can only apply to the restricted period-resembling a '( period of induction "-during which the specimen is becoming covered with tarnish e g . in Fig 2 the period immediately preceding the point P.Beyond that point the primary equa-tion W2 = Kt applies without qualification since A is then constant and equation (2) becomes identical with equation (I). This expression there-fore may be accepted as representing the fundamental relationship between weight-increment and time when copper is tarnishing and hence the view which attributes the accompanying colour changes to interference-already on general grounds almost certainly the correct view-receives quantitative confirmation. Loq-Period Tests.- Consfany of ParaboZii ReZations.$$ under fihctuating Atmos@ric Conditions. Turning from the intensive examination of the very early portions of the curve yielded by copper it is of interest to enquire into its disposition after considerably longer periods than those hitherto considered.The results plotted in Pig. 3-representing a total period o W. H. J. VERNON ' 2 7 exposure of I 300 days-were obtained from the identical specimens whose weight-increments up to 114 days formed the subject of Fig. 8 of the First Report. I t is important to note that in each of the four years represented in Fig. 3 the atmosphere has passed through the changes in type which have already been discussed giving rise to two periods in each year (' W ' and S ' respectively) the approximate durations of which are indicated by vertical lines in the diagram. From the squares of the weight-increments it will be seen that the parabolic relationship has been followed very closely for the whole time in spite of the fluctuations in external conditions.13 When it is recalled how profoundly different are the intrinsic effects upon copper during the winter and summer periods respectively the significance of this result will be appreciated.One is indeed led to the conclusion that the rate of attack is determined by the condifionsprevaihhg at the time of the j r s t exposure and that external conditions may subsequently vary within extremely wide limits without disturbing the course of the attack. The important corollary follows that the process must be very largely controlled by the properties of the initially-formed film. Relation between Weight= Increment 'and Time-(2). Oxidation at the Ordinary Temperature without Development of Colours. It has already been mentioned that during the summer period when the concentration of atmospheric sulphur falls below a certain excessively small limit copper specimens cease to undergo any appreciable visible change on exposure.Nevertheless the formation of a film of sufficient thickness to confer protection upon the underlying metal is demonstrated by the fact that on emerging into the winter period such specimens stilt? retain their inzmuni2y from tarnishing whiZst fresh&-cZeaned specimens tarnish in the usuaZ way. Arising from this observation the attempt was made to trace the development of the protective film gravimetrically ; the results of a typical series of weighings are plotted in Fig. 4. The experiment was started and most of the weighings conducted during Period S; the last point on the curve however falls well within Period W.To facilitate comparison the early part of the curve (Fig. 2) previously discussed as typical of Period W is plotted on the same diagram to the same scale and in its correct chronological position. Apart from the great disparity in the respective rates of attack interest chiefly attaches to the disposition of the curves in the early stages. I t is significant that in the case of exposure during the summer period the curve is parabolic from the start as distinct from the straight line which characterises the early part of the other curve, the meaning of which has already been discussed. From similar considera-tions it follows that in the present instance the initial attack takes place simultaneously over the whole surface.l* I t will be observed that the curve which started in period S is undeflected from its original course on emerging i n to period W notwithstanding the excessive increase in the activity of the atmosphere.Thus of the two alternatives presented the continuance of the type of film already started is preferred One additional and important point remains to be noted. 13 Results up to 1600 days will be found in Fig. 12. 14 It is for this reason that it is still possible to measure the weight-increment after the first day's exposure notwithstanding that the intrinsic rate of attack is so very much less; indeed in the examples plotted the value for one day's exposure was actually greater for curve S than for curve W I I J I I 1 I I I I I ?q 0.6 PERIOD S 10 20 30 40 50 Time in days.Fig. 4.-Pcriod S (Curve A).-Oxidation of copper at ordinary temperature weight-increment and time. Period W.-Attack upor freshly-cleaned copper (Curve B). (Continuation of Curve A), z= Squares of weight-increments (values taken frum smooth Immunit W. H. J. VERNON 129 already been enunciated (from consideration of curves started in period W), that the rate of attack is largely determined by the conditions prevailing at the time of the first exposure. Relation between Weight- Increment and Time-(3). Oxidation at Temperatures above Normal with Development of “Temper Colours. ’ ’ The phenomenon of ‘‘ temper colours,” i.e. the characteristic sequence of colours developed during the heating of metals at suitable temperatures in air is clearly sufficiently cognate to warrant some measure of attention in an enquiry of the present kind.Between these colours and the tarnish colours produced at normal temperatures there is sufficient general resemblance to suggest at once a similar physical cause i.e. the interference of light in the film of reaction product. From previous considerations, ’c 0 a 1.00 z 0 4 0 6 12 18 2A 30 36 42 48 5.-Oxidation of copper at 150OC. (with development of “temper colours”). Relation between weight-increment and time. #- = Squares of observed weight-increments. Time in hours. Fig, therefore a parabolic relationship between thickness of film and period of exposure at any given temperature would again be expected and since complications due to any ‘‘ edge effect ” are now absent (i.e.the colours associated with any particular combination of temperature and time are exhibited uniformly over the surface) the weight-incrementltime curve should be parabolic from the start. To put this reasoning to experimental test a series of determinations was carried out at I ~ o O C . the results of which are plotted in Fig. 5.15 The times of heating varied from one hour, giving a bluish-purple film to 47 hours when a dull reddish-brown film (a second order colour) was obtained. From the squares of the weight-increments it will be seen that the foregoing considerations are satisfactorily borne out. The absence of any initial straight line suggests at once that 15 Attention should be called to a recent paper by J. S. Dunn (Proc.Rojt. SOC. 1926, A. 1x1 210). By noting the sequence of temper colours on copper heated in air, and plotting ‘‘ equivalent air thickness ” against time Dunn obtained well-defined parabolas for temperatures between 184O and 444’ C . At the latter temperature the parabolic relationship was confirmed by measuring the fall in electrical conductivity. I 130 SECOND EXPERIMENTAL REPORT traces of sulphur compounds do not play the essential part which they do in tarnishing ; experiments have confirmed that within the limits which have already been discussed (see Fig. I) relative purity of the atmosphere is practically without influence in the development of temper colours. (g) Properties of Oxide Films Developed in Relatively Pure Atmos-phere at Ordinary Temperature.The evidence described under previous headings leads to the conclusion that the film which develops upon copper at the ordinary temperature in an atmosphere of relatively high purity is totally different in properties from that which forms also at ordinary temperature in the presence of traces of sulphur compounds. In contradistinction to the latter the former shows no interference colours; it is not only independent of the presence of traces of sulphur compounds in its formation but wher. once formed it is relatively impervious to their presence. I t seems reasonable to believe that the explanation lies in a fundamental difference in the original space lattice of the two types of film respectively. Thus it can readily be imagined that the one in which sulphur does not appear will resist the intrusion of sulphur atoms into the lattice and hence, will protect the underlying metal.On the other hand when the clean metal surface is exposed to an atmosphere containing traces of sulphur it is probable that sulphur atoms take an essential part in the formation of the first lattice which is therefore of such a type as to permit of the diffusion of more sulphur atoms to the underlying metal; hence the build-ing up of the same lattice is facilitated. The imperviousness of the sulphur-free oxide film is clearly a matter of importance in view of the protection afforded to the underlying metal. The protective effect is shown graphically in Fig. 4; it may also be illustrated by the following figures. Specimens exposed for a month during period W increased in weight by 0.8 mg.per sq. dm. During the same time specimens which had been previously exposed for a month during period S increased by 0.06 mg. per sq. dm.; moreover during their whol’eperiod of exposure they had increased by only 0 - 3 mg. per sq. dm. without it may be added the accompaniment of any colour changes. (h) The Properties of Oxide Films Develloped at Temperatures above Normal. The protective character of the film obtained by exposing copper to a relatively pure atmosphere at the ordinary temperature suggests at once that a similar degree of protection might be conferred more rapidly by ex-posure at higher temperatures. Actually it was found in 1923 that very similar effects could be obtained in as short a time as I hour by simply raising the temperature to IOOO C.Furthermore the formation of the (almost invisible) protective film at 100’ C. was found to be independent of the time of year at which the treatment was conducted and no special purification of the air was necessitated. I t was felt desirable however to carry the matter to a further stage and investigate the effects produced by heating specimens at both lower and higher temperatures. The results of this enquiry have formed the subject of a separate paper.16 Specimens were heated for I hour at temperatures extending from soo to 1sW. H. J. Vernon 7. Chem. SOL 1926 128 2273 W. H. J. VERNON 131 250' C. ; below 100' the films obtained were practically invisible but at higher temperatures the usual sequence of interference colours was observed.Weight-increments were determined and their relationship with temperature was found to conform exactly to an equation developed by J. S. Dunn,17 but previously verified for much higher temperatures only. This equation is identical in form with the Van't Hoff isochore i.e.-d log w p ~ = -The agreement will be seen by reference to Fig. 6 where logarithms of weight-increment are plotted against reciprocals of absolute temperature. 1 I 1 1 1 1 1 -06 0.0 -z a x Yo's -P L L o -5 * n D 3 ? I S -C Y s" D Fig. 6.-Influence of temperature on rate of direct oxida-tion (Copper and 60/40 brass). Subsequent exposure of heat-treated specimens during Period W yielded results of particular interest as will be seen from the following description .of the visible effects after 20 days' exposure :-I.Unheated specimen . Tarnished a reddish-purple colour. 2. Heated at 50" (similarly for 55") . Reddish-purple (exactly as I). 3. Heated at 75" . . No change since heat-treatment i.e. piwtically unaltered copper. 4. Heated at 100' . . No change since heat-treatment i.e. slightly darker than 3 as result of heat-treatment but otherwise unaltered. S.-IO. Heated at 125-250" . No change since heat-treatment i.e. sqies of interference colours (Zoc. cit.). Considerable significance attaches to the fact that an hour's heating in air at a temperature as low as 75'-a treatment practically without any visible effect upon the metal-is sufficient entirely to inhibit the develop-ment of interference colours on subsequent exposure to tarnishing conditions.The conclusion is inevitable that the protection is brought about by a film of oxide either so thin as to be almost invisible (as by heating at 757 or, while still transparent sufficiently thick to refract light (as by heating at temperatures above 100~). l7 Proc. Roy. SOC. 1926 A. 1x1 203 = 32 SECOND EXPERIMENTAL REPORT The very sudden break in the behaviour of specimens heated at 5s0 and 7 5 O respectively suggested that a film of minimum thickness was necessary for protection and an explanation was attempted based on the estimated thickness of the film. Assuming the first product of interaction of copper and oxygen to be cuprous oxide (density 6-0) and taking from Fig. 6 the weight of oxygen absorbed,.the figure obtained for the thickness of the film produced at 55' was 10-5 Angstrom units.In the calculation the nominal area of the specimen was assumed but if a value more nearly approaching the true area had been employed (bearing in mind the emeried nature of the surface) undoubtedly the result would have approximated much more closely to the value usually accepted for the lattice dimensions of cuprous oxide i.c. 4.3 A. u. The suggestion was made therefore that the minimum thickness o f $ h necessary forprotection may be such fht the unit Zaffice of the oxide is comjZeted for the whole of the surface. TABLE 11. COPPER WEIGHT-INCREMENTS SUBSEQUENT TO HEAT-TREATMENT. Temperature of Heating.: .C. (Not heated) Period of Exposure and Increase in Weight (mg.per sq. dm.). Series A. 6 Days. 34 Days. 100 Days. Series B. 152 Days. 206 Days. Weight-increments (subsequent to heat-treatment) of the specimens discussed above are given in Table 11. (Series A) and are plotted in Fig. 7. Gravimetric data for other (less complete) series after longer exposure are also included in Table 11. Although protected specimens slowly increase in weight they do not appear to undergo any further colour change. Actually the gravimetric method does not give a satisfactory response to visible changes and the measurement of loss of reflectivity as employed in the earlier work (First Report Zuc. cit. 851) is in this respect much to be preferred. The results recorded above were obtained with specimens that had been polished with fine emery paper (Hubert No.I). Tests conducted upon a limited number of specimens brightly polished with mechanically driven '' mops " yielded results which although quite convincing gravimetrically (the figures comparing favourably with those given in Table II.) were not so satisfactory from the point of view of the suppression of visible changes. I n these tests however no other temperature than xooo was employed for the heat-treatment and further work seems to be necessary in the case of copper polished in this way W. H. J. VERNON I33 (i) Properties of Sulphide Films. In view of the characteristic properties of oxide films and of c r ~ x y -sulphide ” films which have been discussed above interest attaches to the behaviour of films consisting initially of pure sulphide on exposure to the atmosphere.To obtain comparable information upon this point specimens of copper foil each measuring 10.0 cm. x 5.0 cm. were cleaned in the usual way and submitted to initial treatment as follows :-A. Heated in air at 1 5 0 O C. for 20 minutes. The thickness of the resulting oxide film was such as to give an orange “temper colour,” corresponding with that obtained by heating for one hour at I 25’. (See I - ” ’ ’ ” ’ ” I ’ I ’ Ho (1) a“’“ e Fig. 7.-Weight-increment due to oxidation (tarnrsh-ing) of copper and 60140 brass at the ordinary temperature in relation to tem-perature of previous heat-treatment (for one hour). Fig. 6.) Increase in weight = O~*IO mg. per sq. dm. (very nearly). Approximate film thickness = 150 A.u. ‘( Temper colour ” = fine rose. Increase in weight =00-20 mg. per sq. dm. (very nearly). Approximate film thickness = 300 A. u. C. Exposed to atmosphere of hydrogen sulphide for short period until an orange interference colour practically identical in appearance with A, was obtained. Then submitted to vacuum desiccation several times in order to remove any associated hydrogen sulphide. Increase in weight = 0.10 mg. per sq. dm. D. Exposed to similar atmosphere of hydrogen sulphide for a longer period until the almost exact colour equivalent of B was obtained i.e. “rose”; then vacuum desiccation as before Increase in weight = 0.25 mg. per sq. dm. B. Heated in air at 150’ for one hour 134 SECOND EXPERIMENTAL REPORT Weight of Oxygen or of Sulphur in Initial Film.E and F. No initial treatment other than cleaning. Besides the untreated ‘‘ blanks,” the specimens thus included two typical oxide films (one twice the thickness of the other) together with two corresponding sulphide films ; after preparation as above the whole were exposed to the basement atmosphere during c L period W.” Interesting visual results quickly followed. In striking contrast with the heat-treated specimens which showed no sign of change,ls those initially covered with sulphide proceeded (at of course a modified rate) through fh same succession of coZour changes which they would have done zjc th treatment with hvdrogen suphide had been continued. I n two days the “thin ’’ sulphide film had progressed to the stage initially represented by the “thick” film i.e.a “ rose ” colour whilst the thick film was then deep blue ; in three days the colours were purple and steely blue respectively. The rate of change of specimens initially covered with sulphide was visibly greater than those exposed in the freshly-cleaned condition ; thus in three days the c c blanks ” had not progressed beyond a deep orange border. After one month’s exposure the specimens were !weighed and the sulphur (sulphide) content of the films was determined. Weight-Increment Subsequent to Exposure. TABLE 111. BEHAVIOUR OF SULPHIDE AND OXIDE FILMS ON EXPOSURE TO THE ATMOSPHERE. Nature of Initial Film. -1 Sulphide 51 9 9 Weight of Sulphur Absorbed Subsequent to Exposure. Mg. per Sq. Dm. < 0’002 19 0’43 0.13 Proportion of Sub sequent Weight In-crement Accounted for by Sulphur.< 1.0 per cent. Is 26.2 ,, 14’8 I , In Table 111. the increases in weight subsequent to the initial treatment are given together with the corresponding increment in sulphur content ; in the last column the proportion of sulphur which is responsible for the subsequent weight-increment is expressed as a percentage of that incre-ment. I t will be seen that the presence of the initial sulphide film has not only caused a definitely greater attack upon the underlying metal but also it has materially increased the selective absorption of sulphur from the atmosphere. On the other hand the powerful restraining influence of the initial oxide film is confirmed; moreover the figures quoted for sulphur content demonstrate the virtual impenetrability of this film by ordinary atmospheric sulphur.Incorporating what appear to be reasonable conclusions with statements of fact the position may be summed up as follows. Films of pure oxide upon copper are relatively impervious to oxygen atoms at the ordinary temperature and excessively impervious to sulphur atoms. Pure sulphide 18Similar heat-treated specimens have now been exposed for 12 months and l9 These values were obtained from examination of similar heat-treated specimens have as yet showed no visible alteration. which had been exposed for nearly 12 months W. H. J. VERNON I35 films on the other hand are relatively pervious both to oxygen and sulphur atoms ; moreover compared with a freshly cleaned copper surface they favour the selective absorption of sulphur.Exposed to an ordinary atmosphere containing traces of sulphur as impurity sulphide films increase in thickness mainly through absorption of oxygen ; the optical homogeneity of the film is maintained however (i.e. there is no break in the sequence of interference colours) which leads to the suggestion that isomorphous mixtures of sulphide and oxide are produced. In the case of clean copper surfaces exposed to a similar atmosphere films of this isomorphous type are produced at the outset ; these films contain a preponderance of oxide over sulphide but their permeability is excessively greater than that of the pure oxide and is not far removed from that of the pure sulphide.The function of traces of atmospheric sulphur in this respect appears to be akin to if not identical with that of catalysis ; the conclusion is reached that it is the (c activation ” of the first oxide lattice through the impingement of sulphur atoms which entirely determines the subsequent course of attack. (j) Effects produced by “Smoke Films.” In the course of the present work some experiments were carried out by Mr. L. Whitby some time ago with the object of ascertaining the effect of smoke (particularly tobacco smoke) upon a clean copper surface. Specimens prepared in the usual way were placed in tubes which were then filled with smoke and sealed. Contact with the smoke produced very little visible effect upon the metal and hence a negative result was recorded.Later however the surprising fact was noted that specimens which had received this treatment still retained their practically unaltered appearance when exposed to conditions such that untreated specimens tarnished rapidly. It was found that immunity from tarnishing was conferred by exposure to tobocco smoke and the smoke from cellulose, either in the form of pure cotton wool or in the less pure form of paper or rag.2o The best results however were obtained from the pure cellulose smoke since this had the least visible effect upon the metal in the first place whilst yielding equally good results on subsequent exposure. Similar protective effects were observed upon silver (both “ fine ” and “sterling ’7 as the result of this treatment ; the actual extent of the protection can be gauged from the fact that the attack of an atmosphere highly charged with hydrogen sulphide (in which the untreated metal became severely tarnished in a few minutes) was resisted for as long as 24 hours.The phenomenon is one which would probably repay further study but for the present it is merely recorded as an interesting observation. A. ATMOSPHERE TYPE 1.-BASEMENT.” (I) Short-period (Intensive) Tests Relation between Weight-Increment and Time. In previous work the relationship between weight-increment and time in the case of zinc exposed to the “ unsaturated ” type of atmosphere was shown to be consistently linear (First Report Zuoc. tit. 863). Although, 20The smoke from wool however so far from protecting has an intrinsic and considerable tarnishing effect due of course to the sulphur content of the wool.21 For description see First Report loc. cit. 553 and the present report p. 117 Time in days (from respective origins). Fig. 8.-Oxidation of zinc at the ordinary temperature. Relation between early stages. Dates of Starting.-A Jan. 29; B April 24; C May 12; D W. H. J. VERNON I37 Results obtained from tests started on four separate occasions are represented in Fig. 8 ; the figures plotted are the means of closely-agreeing duplicates and with one exception include the weight-increment from the first day’s exposure onwards. Two interesting features will be noted: (I) from the first day the points fall with striking precision upon a straight line thus confirming the linear relationship for very short exposure which was the main object of the tests ; (2) the straight line does not quite pass through the origin but cuts the vertical axis indicating a greater initial rate of attack which quickly assumes a lower and constant value.Time in days. Fig. g.-Oxidation of zinc at the ordinary temperature. Relation be-tween weight-increment and time. “ Short Period.” (Con-tinuation of Curve B Fig. 8). The very open scale upon which Fig. 8 is plotted should be borne in mind. After longer periods of exposure the discrepancy at the origin becomes relatively smaller and therefore difficult to detect on a close-scale diagram. Thus in the case of curve B the last point plotted was at 12 days; no further determinations were made until 112 days and the results at that stage are shown in Fig.9. I t will be noted how exactly the original straight line has persisted and also how insignificant the original defection from the origin now appears; clearly after more prolonged exposure a “fair curve ” through the points would fail to distinguish it at all. (2) Long-period Tests. Relation between Weight= Increment and Time. From the intensive short-period tests described above it is of interest to turn to the other extreme and consider the behaviour of zinc expose 138 SECOND EXPERIMENTAL REPORT during the longest periods over which records are available. That the linear relationship is still maintained will be seen by reference to Fig. 10, where results extending over a period of 3+ years are plotted. Consider-ing the neutral or pervious character of the scale it may appear surprising that the seasonal fluctuations in atmospheric conditions (see p.I I 7) have not affected the long-period tests to a greater extent. The explanation lies in the fact not only that the room temperature is fairly equable throughout the year but also that variations in relative humidity and in “reactive sulphur content,” within the limits already described have little if any 00.0 E: e .r( 4 li.0 ii ‘2.0 s 3 l s .U $ kl Q) a Q) 5 -9 4.0 J Time in days. Fig. lO.-Oxidation of zinc at the ordinary tyFperature. Flation be-tween weight-increment and time. Long Period. effect upon the surrosion of zinc. Thus the reason for the continuity ot the curve is radically different from that which obtains in the case- of copper.(3) Note upon Effect of Surface Condition. I n the First Report (doc. clt. Table I.,.p. 860) it was shown that up to 205 days’ exposure the difference in behaviour between zinc specimens with a dull emeried surface and those which had been brightly polished to a “mirror finish,” was extremely small. I t may now be noted that after an exposure of I 300 days dull ” and “ bright ” specimens showed weight-increments of 20.8 and 20.05 mg. per sq. dm. respectively. On the other hand copper specimens exposed simultaneously gave for dull ” and “ bright ” surfaces respectively weight-increments of 5 ‘4 and 4-3 mg. per sq. dm W. H. J. VERNON 139 (4) Note upon Visual Results. The absence of interference colours from the surface of zinc which is undergoing c L surrosion ” is clearly in harmony with the linear nature of the weight-increment curve which indicates a product having a granular or discrete structure as distinct from the continuous envelope which gives rise both to interference colours and to a parabolic weight-increment curve, Nevertheless it will be observed that the possibility of dzfracfion colours is not excluded.It is of interest to note that the only colour effects which have been observed upon zinc whilst quite inexplicable on the interference view become completely intelligible when regarded from the point of view of diffraction. I n the First Report (doc. cit. 857) the development of a ‘‘ sky-blue ” border was noted ; specimens exposed more recently have shown a distinct blue tinge over the whole surface.Now in the case of a film which is gradually increasing in thickness and causing interference colours to be produced light waves will be extinguished in the order of their lengths, the shortest being the first affected; hence emergent light will at first correspond to the colours of the longer waves (yellow orange and red) and therefore blue will not be expected till this sequence has been passed. But in the case of diffraction the colour will depend entirely upon the spacing of the granules in the reaction-product and hence would be expected to be relatively independent of the period of exposure. There can be little doubt therefore that the blue colour associated with the surrosion of zinc is a diffraction effect such as would normally be expected from the gravimetric data.(5) Microstructure of Surmsion = Films upon Zinc. Hitherto the view that the film of reaction-product upon zinc partakes of a ‘‘ granular ” or (‘ cellular ” nature has been inferred from the linear dis-position of the weight-increment curve. Direct evidence has now been obtained however by microscopical examination of the film. The photo-micrograph reproduced in Fig. 11 was obtained from a zinc specimen which had been exposed in the ‘( basement ” atmosphere for 18 months. I t will be seen that the granular structure is actually quite distinct. I n addition it may be noted that there is now a slight indication of the underlying crystal structure some crystal grains appearing darker than others, suggesting selective rates of attack due to differences in orientation.I t is instructive to compare this photomicrograph with Fig. 2 2 of the First Report which shows the appearance of tarnished copper under the microscope. The comparison brings out clearly the fundamentally different nature of the two films upon which the whole course of the oxidation process in each case depends. B. ATMOSPHERE TYPE 2.-“ TANK ROOM.” ’’ Relation between We&-hf-Increment and Time.-The results available at the time of the First Report suggested the L L possibility ’’ of an acceleration in the rate of attack upon zinc exposed in the “tank room” quite apart from any acceleration due to increasing corrosivity of the atmosphere (First Report doc. cit. 869). More extended observations however have failed to confirm this suggestion.On the contrary the results now plotted in Fig. 13 (representing 1200 days’ exposure) and also results yielded by other specimens for similar periods have shown that the straight line relationship For description see First Report loc. tit. p. 853 140 SECOND EXPERIMENTAL REPORT Mean Total Weight Increment. Mg. per Sq. Dm. has been followed extremely closely having regard to the relatively severe fluctuations in external conditions. Rate of Attack as compared with Rate of Attack in Atmosphere Type I.-Tile linear relationship between weight-increment and time having been shown to apply to exposure in both types of atmosphere even after com-paratively lengthy periods comparison of the respective rates of attack is obviously facilitated.Typical data are included in Table IV. Rate of Atiack. Mg. per Sq. Dm. per 1000 Days. TABLE IV. RATE OF ATTACK UPON ZINC EXPOSED IN ‘‘ BASEMENT ” AND “ TANK-ROOM ” RESPECTIVELY. Atmosphere. Series No. ‘( Basement ” Typer *I 4B. 9B. IOB. Type 2 4T. ‘( Tankroom ” * I roT. Period of Exposure. Days. I I I I60 1300 830 18’7 20.6 13.8 X6’I 15.9 I 6-5 7‘5 9’5 I t is somewhat surprising that under the relatively dry conditions of the c L basement ” the rate of attack upon zinc is considerably greater than that which obtains under the relatively humid condition of the ‘( tank-room.” The most probable explanation however is that the divergence in behaviour is a temperature effect (see First Report Fig.I) and that the influence of humidity within the limits represented by the present conditions, is a neutral SECTION 1.L-BRASS. (a) Relation between Weight - Increment and Time. Atmosphere Tyfe I . ~ ~ - - I ~ the First Report (Zoc. a?. p. 864) the follow-ing statement was made. “ Results of tests upon 70/30 and 60/40 brass for a period up to 450 days the first weighing having been conducted 144 days from the start have given points lying almost exactly upon a parabola.” In Fig. 1 2 are now given the results obtained from these same specimens during a period of 1660 days the squares of weight-increment being plotted separately ; the graphs yielded by a single copper specimen exposed simultaneously are also included. The statement quoted above remained true in the case of 70/30 brass for about a further ICO days after which weight-increments were in excess of those required by the parabola.In the case of the 60/40 brass the defection from the parabola appears already to have commenced when the statement was made. In each case the subsequent course followed was within experimental error a straight line. On the other hand even after this lengthy period of exposure copper has still adhered closely to the parabola. a3Definite results obtained in connection with the rusting of iron (see Section 7) suggest that for each corrosion product there is a critical atmospheric humidity above which moisture condenses upon the specimen and below which variation in the content of water vapour is without appreciable influence.a4 For description see First Keport foe. cit. 853 present report p. 117 Fig. 11.-Zinc after exposure to indoor atmosphere as described in text. x 150 diams. (To facepage 140 Fig. 14.-60/40 Brass after exposure to indoor atmosphere as described in text. x 150 diams. [To face page 141 Weight-incremcnts. Mgms. per sq. decim. " P g ~ ~ e - $ a ~ 0 0 0 0 0 0 0 0 tr! op Weight-Mgms ? 142 SECOND EXPERIMENTAL REPORT The position to date therefore is that while in the present type of atmosphere zinc and copper have given no indication whatever of any de-parture from the straight line and parabola respectively 70/30 and 60/40 brass behave anomalously. Although for a time they resemble copper and yield a curve which approximates to the parabola after more extended ex-posure they revert to the straight line which is characteristic of zinc.Atmosjhre Type 2.24-Under the more humid conditions of the (( tank-room” the tendency referred to above for brass to follow the parabolic behaviour of copper seems entirely to disappear. So far as the observations have gone there is at no stage any indication of a falling off in the rate of attack ; on the contrary there appears to be definite evidence of acceleration. This is shown in Fig. .13 where results from specimens of */o/30 and 60/40 brass after 1200 days’ exposure are plotted together with results from zinc exposed simultaneously. The approximately linear relationship maintained by the zinc suggests that there has been no appreciable increase in the corrosivity of the atmosphere.Possibily the increased rate of attack upon the brasses may be accounted for by the selective deposition of moisture upon the respective corrosion products. (b) Relation between Weight - Increment (Corrosion) and Composition. Atmosphere Type I. Atmosphere Type 2. Domestic Kitchen. See Section 4(a) p. 146. See Section 4(6) p. 146. See Section 4(c) p. 146. (c) Microstructural Changes on Atmospheric Exposure. The microscopical examination of 60/40 brass at various stages of ex-posure brings out several interesting points. -In the early stages of exposure in atmosphere Type I the attack is directed almost entirely upon the alpha constituent which develops an apparently continuous tarnish film in the same way as copper. During this time indeed the alloy behaves in a similar manner to copper and tends to follow a parabolic curve.Later on, however (coinciding approximately with the appearance of the ‘‘ fogging effect” described in the First Report Zoc. cit. p. 856) an attack of quite a different type develops upon the beta constituent ; the effect is discontinuous over the surface and gives rise to an apparantly duplex structure. This is shown in the photomicrograph reproduced in Fig. 14. There can be little doubt that this phenomenon marks the early stages of the departure from the parabola to the straight line which has been referred to in connection with the gravimetric results (“ a ” above). (d) Properties of Oxide Films Developed in Relatively Pure Atmos-phere at Ordinary Temperatures. In view of the remarkable protective properties of the film developed upon copper during the “summer period,” interest attaches to the be-haviour of brass when exposed under similar conditions.I t will be recalled that intensive weighings upon copper specimens showed that the parabola of relatively low rate constant obtained during the summer or non-tarnishing period was entirely undeflected on emerging into the period of relatively high tarnishing capacity. The specimens from which the result W. H. J. VERNON I43 plotted in Fig. 15 were obtained including representatives of 70/30 and 60/40 brass and also of copper were first exposed on June rgth (1923) and hence had the maximum time for the development of any protective film. (In the figure the approximate division between the summer and winter period is indicated by a vertical line.) During the summer period it so happened that the weight-increments yielded by the various specimens were almost identical a circumstance which lends additional interest to the remaining portion of the diagram.I t will be seen that although agreeing so closely in weight the respective films differ consider-ably in properties. Only in the case of copper does the parabola continue undeflected through the period representing the more drastic conditions of 4 . 0 40 80 I20 160 200 240 280 320 Time in days. Fig. 15.-Copper 70130 brass 60/40 brass. Atmosphere Type 1. Relative protection afforded during Period W by thin films of oxide developed during Period S. (Compare Fig. 1) exposure a result which confirms those previously recorded as to the im-perviousness of the film.In the case of each of the brasses there is a very considerable increase in the rate of attack showing that the earlier film is not proof against the altered atmospheric conditions i.e. the efficiency of any protective effect must be extremely small. (e) Properties of Oxide Films Developed at Temperatures Above Normal. Experiments have been conducted to determine whether by exposure to air at temperatures above normal an oxide film could be developed upon brass sufficiently impervious to afford protection against tarnishing on sub-sequent exposure. A preliminary set of heat-treatment experiments upon 70/30 and 60/40 brass showed that very similar visible effects were produced on each alloy for the same temperature of heating; in order to try out the more extreme case first the experiments were continued upon 60/40 brass only.These experiments were projected upon exactly the same lines as those described for copper and the results to date were included in the communication previously referred to (Zoc. cif.). Immediate Efecfs of Heat-Treatment.-A uniform time of heating of on I44 SECOND EXPERIMENTAL REPORT hour was employed throughout. For equivalent thickness of film con-siderably higher temperatures were required than was the case with copper ; e.g. to produce a darkening of the metal approximately equivalent to that obtained by heating copper at IOO' brass specimens required to be heated at zoo' after which treatment although slightly darkened they were still pale yellow in colour.Increasing the temperature of heat-treatment pro-gressively to 350" merely increased the depth of yellow but between this temperature and 400' a neutral or steely tint was obtained. At 425' a second-order yellow appeared accompanied however by specks of zinc oxide indicating the end of the interference colour range.25 Subsequent E'ects of Heat- Treatment.-Available evidence as to the pro-tective effect of oxidation is not so conclusive as that which has been recorded for copper. This is particularly true of the gravimetric results as will be seen by reference to Fig. 7 where subsequent weight-increments of brass specimens after IOO days' exposure are plotted against the tempera-ture of heat-treatment ; corresponding data for copper are included on the same diagram.Although the first heat-treated member of the series (zooo) has suffered definitely less attack than the unheated " blank," the reduction is considerably less than that which obtained for copper heated at 75'. Moreover the weight-increment progressively increases as the temperature of heat-treatment rises (in contradistinction to the behaviour of copper) until with the specimen heated at 250' the increase is very little less than that suffered by the unprotected " blank." Nevertheless the visible results were much more satisfactory than these figures would suggest. Thus after 3 8 days' exposure untreated specimens had darkened considerably with development of light-brown patches whilst the specimen heated at zooo had apparently undergone no change.As a result the treated specimen appeared much lighter than the untreated blank although when first ex-posed it was a shade darker owing to the effects of the heat-treatment ; the remainder of the specimens appeared to have undergone no further change. After I 00 days untreated specimens had become yellowish-brown gener-ally; there was marked contrast between these and the specimen heated at zoo' which was still pale yellow whilst the remaining specimens had altered but very little in appearance since the first exposure. I t will be seen therefore that although oxide films upon brass are not so efficacious as those upon copper there is definite evidence that the pro-tection extends to brass the effects being confined mainly to the suppression of further colour changes.26 (f) Use of Lanoline for Protecting Brass Exposed to 6 b Dew-Point '' Conditions.The tests to be described were started as far back as November 1922, since when the specimens have been exposed to the atmosphere of the '' tank-room.'' The results are of interest in view of the conclusions already discussed as to the susceptibility of brass to attack by this type of atmosphere (see p. 142). In the present instance however the material employed 25 For weight-increments and other details the original paper may be referred to. ac The experiments described in the text were suggested by the results which had already emerged from the systematic work upon copper. The earlier work of Bengough and Hudson (English Patent 1917 y.SOC. CJtem. lnd. 1919 640 A.) may be recalled, however in which protection against attack by sea-water was obtained by a preliminary oxidation of the brass in air at a temperature not exceeding 4joo C . Alternatively im-mersion in a solution of permanganate or an mated solution of carbonate was recommended W. H. J. VERNON I45 Reflectivity Per Cent. Before After Treatment. Treatment. contained only 10 per cent. ~inc.2~ Direct results upon 70/30 or 60/40 brass would no doubt have been advantageous. Actually however the results yielded by the (unprotected) 90/10 brass were similar in type to those obtained from the standard brasses under similar conditions ; there can be little doubt that the treatment would be equally effective if applied to these materials.Method of Treatment and IntiiaZ h'fects of LanoZina-Two specimens (each c' dull-polished " and in the hard and annealed condition respectively) were immersed in a bath of molten lanoline at 100' C. They were allowed to drain in a steam-oven for some hours28 and finally as much as possible of the lanoline removed by wiping with cotton-wool. The weight of the lanoline remaining was 4.7 and 5 '0 mg. on the hard and annealed specimens respectively. This treatment had the effect of darkening the plates rather comiderably as will be seen from the following reflectivity values :-Loss Per Cent. TABLE V. EFFECT OF TREATMENT WITH LANOLINE AT zooo C. UPON REFLECTIVITY OF BRASS SPECIMENS. Hard specimens Annealed specimens Front Back Front Back I I 68'0 73'0 42'8 47'8 3 7'0 I 34'5 Subsequent Protective Efects.(a) YisuaZ and Optical.-Although when first exposed the '' lanolined " specimens were much darker than the rest, on subsequent exposure they appeared to undergo no further change and at the end of six weeks they appeared far more like the original brass than any of the other specimens. The protection afforded by the lanoline is also shown by the following reflectivity figures :-Mean Zoss of ReJectivifY during 6 weeks' exposure. Untreated specimens . . 47-2 per cent. Treated specimens . ' 17'8 9 9 9 9 (b) Gravimetric.-At the end of the 1400 days the weight-increment of untreated specimens showed little variation and gave the mean value of 15.0 mg. per sq. dm. On the other hand the weight-increment of the (hard and annealed) lanolined specimens were 6.8 and 7-9 mg.per sq. dm. re-spectively. The conclusions reached in the early stages as to the protective value of the treatment with lanoline thus receives adequate confirmation. 27 At the time the experiments were started this was the only alloy available in sufficient quantity to satisfy the complete objects of the tests which included not only the effect of lanoline treatment but also the effect of various methods of preparing the original surface. In the latter respect the results showed remarkable uniformity as will be seen by reference to Appendix B of the First Report (loc. cii. p. 893) where the optical results at 14 days are given in full for the whole series-excluding those which had been treated with lanoline.28 Actually in the attempt to drain off the whole of the lanoline ; this however was found to be impracticable. I SECOND EXPERIMENTAL REPORT SECTION IV. TNE RELATIVE BENAVIOUR OF COPPER ZINC AND BRASS EXPOSED TO SEVERAL TYPES OE INDOOR A TMOSPBERES. (a) Exposure in Atmosphere Type I (“ Basement”). In the First Report (Zoc. cit. Fig. 7) the weight-increments yielded by representatives of the zinc-copper series at 15 36 and 205 days’ exposure were plotted against zinc content. Results up to 916 days are now plotted in Fig. 16 but for the sake of simplicity only one set of curves (those which refer to emeried specimens) is included. The diagram brings out clearly the increasing disparity in the surrosion of copper and zinc respe~tively.2~ The explanation lies of course in the fact that during the whole time copper has followed the parabola whilst zinc has followed the straight line.That the high-zinc brasses have approached relatively nearer to the zinc in respect to their position on the diagram is also accounted for by the fact that after prolonged exposure these alloys leave the parabola upon which they started in the early stages and follow the straight line as for zinc. (b) Exposure in Atmosphere Type 2 (“Tank-room”). I t will be recalled that the chief feature of the atmosphere of the ‘‘ tank-room,” as distinguished from that of the ‘( basement,” is its relatively high, and fluctuating humidity. In addition however for the greater part of the year the mean temperature is appreciably lower than that which obtains for the “ basement ” (see First Report Fig.I). Increasing humidity naturally favours condensation of moisture but it should be remembered that whether moisture is actually precipitated or not will also depend upon specific pro-perties of the corrosion product. Results obtained from copper zinc and the two standard brasses after an exposure of I 2 0 0 days in the (‘ tank-room,” are plotted in Fig. 17. I t will be seen that the difference in behaviour between copper and zinc is less marked than was the case in the “ basement.” Moreover the intrinsic attack is considerably less an effect presumably attributable to the reduced temperature ; clearly however any effect due to increased humidity must have been extremely small.On the other hand, the attack upon the brasses is now greatly in excess of that upon either copper or zinc and also (in spite of the lower temperature) in excess of that undergone by the same materials exposed to the “ basement ” atmosphere for a similar period. In this case therefore it is highly probable that the increased attack has been brought about by a selective precipitation of moisture. (c) Exposure in Domestic Kitchen.30 I t is evident that under the conditions of a domestic kitchen relative humidity will be subject to considerably wider fluctuations than is the case with either of the atmospheres previously considered ; precipitation of moisture upon the specimens will inevitably take place at more or less fre-quent intervals. The position is rendered still more complex by the products 2s1 I t is instructive to compare this diagram with the earlier one bearing in mind the 30 See First Report Zoc.cit. p. 855. great difference in the vertical scales W. H. J. VERNON I47 of combustion (in this case principally those of coal-gas) which are emitted directly into the room. 16.0 14.0 12.0 10.0 8.0 6.0 4-0 2-0 10 20 30 40 50 60 70 80 90 100 Composition per cent. Zinc. periods. Atmosphere Type 3.-Fig. 16.-Zinc-copper Series. Weight-incrementa a t various lotted (on a more open vertical scale) in Fig. 7 of the First geport. N:B.-Curves A B and C were Under such conditions specimens of copper 70/30 and 60/40 brass Each material and zinc have been exposed for a period of 1230 days 148 SECOND EXPERIMENTAL REPORT Copper Bright (5:;) 7.2 Dull 150.6) 164'2 * 9 Dull 2;:;) 33.5 , Dull Brass 70130 Bright 177-8 d)40 Bright was represented by two specimens in the emeried and brightly-polished condition respectively their dimensions being the same as in (a) and (6) above i.e.10.0 x 10.0 cm. For the purpose of comparison a similar pair of iron plates (ordinary ingot iron) 31 was included but for a rather shorter period (1150 days). No corrosion product was observed to leave any of the non-ferrous speci-mens but for the greater part of the time there was a steady fall of rust from the iron. When dismantled the copper and zinc specimens showed little if any sign of moisture but the brasses more particularly 70/30 were distinctly moist owing no doubt to the hygroscopic nature of the corrosion-products.Each specimen was then carefully scraped with a blunt tool for the purpose 20.0 Zinc Bright 30-81 27.8 , Dull 24.81 Iron Bright 739-2 , Dull 626*6}682*9 8 12.0 c 10.0 M 10 -CC; 30 40 50 60 70 80 90 100 Composition per cent. Zinc. Fig. l?.-BeIat:ve corrosion (weight-increments) of copper 70/30 brass 60/40 brass and zinc, after 1200 days exposure in " tank-room. W. H. J. VERNON I49 represent loss of weight in Fig. 18 and the usual increase in weight in the other cases; the very great numerical difference in the vertical scales should also be noted. As would be expected there is a broad general resemblance to the “ tank-room ” curve 70/30 occupying the ‘‘ peak ” in each case. The attack upon the 70/30 brass however is now excessively greater than that upon any of the other materials an effect which pre-sumably may be attributed to the incidence of products of combustion.32.Analysis of Corrosion ProtEzccts.-The complete analyses are set out in Table VI. The ‘‘ dezincification ” of the brasses is shown by the prepon-derance of zinc in the products and is excessive in the case of 60/40 brass. The high content of sulphate more particularly in the products from copper and 7 0 / 3 0 brass will be noted. An interesting feature is the LO 20 30 40 50 60 70 80 90 100 Composition per cent. Zinc Fig. 18.-Relative corrosion (losses in weight) of copper 70/30 brass 60/40 brass and zinc, after 1200 days exposure in domestic kitchen. appearance of combined organic acid radicdes.They were determined by extraction of the fatty acid with ether after hydrolysis with hydrochloric acid. (Only in the case of 60/40 brass was any free fatty matter present, as shown by treatment with ether before hydrolysis.) The further interest-ing result may be noted that the proportion of combined organic radicles increases with the zinc content suggesting that they are present in combina-tion with zinc. A rough determination of the mean molecular weight of the associated fatty acids gave a value of 133 a relatively low value in-dicating a correspondingly high vapour pressure as would of course be expected. 32 In interpreting Fig. 18 the dezincification of the brasses (see analyses of corrosion products) should be borne in mind. As will be discussed more fully in Part II.this represents a definite amount of corrosion which is not included in the figure for loss in weight ; hence the actual attack upon the brasses no doubt exceeds that upon copper or zinc to a greater extent than is shown by Fig. 18. Moreover in view of the excessive dezincification of 60140 brass the curve in Fig. 18 probably exaggerates the disparity in the actual corrosion of 70130 and 60140 brass respectively 150 SECOND EXPERIMENTAL REPORT 1 Copper. I 70/30 Brass. 60140 Brass. 14.5 16.1 1.5'7 1 '5 1'0 1'5 TABLE VI. Zinc. 13'4 2'4 COPPER ZINC 70130 AND 60140 BRASS. ANALYSIS OF CORROSION PRODUCTS AFTER A. Moisture driven off at 100' C. 176 WEEKS' EXPOSURE IN A DOMESTIC KITCHEN. Matter insoluble in acids. Percentage values.H,O . . . . Insolublematter . . B. Metallic and non-metallic radicles in product after drying. Percentage values, excluding insoluble matter. c u . . . . . Zn . . CSO*l -P I ' CCOJ Combined radicles Free fatty fatty matter 31.4 SECTION K AL UMINIUJf. (a) Introduction. The primary object in undertaking these experiments was to determine the precise form of weight-increment curve yielded by aluminium on oxida-tion at the ordinary temperatures so that the information might be collated with that already obtained for other metals. Possible difficulties not met with hitherto were anticipated at the outset. Thus it is generally believed that a freshly-cleaned surface of aluminium becomes rapidly covered with a film of oxide which tends to retard further oxidation.I n the main how-ever the evidence is of an indirect character and so far as the author is aware there has been no previous attempt to trace the formation of the film at ordinary temperatures by direct weighing.33 (b) Experimental. Aimosphre.-The tests to be described were conducted entirely in Atmosphere Type I. (see p. I I 7). Specimens.-For the bulk of the experiments high-grade metal contain-ing 99'6 per cent. aluminium was employed.34 Tests have also been con-ducted upon aluminium of exceptional purity prepared by the Hoopes electrolytic refining process.35 Specimens having the uniform dimensions of 10-0 x 10.0 cm. have been used throughout. The preparation of the surface occasioned some preliminary investigation since the treatment with 33 The high temperature oxidation of aluminium has been investigated by Pilling and Bedworth (y.Inst. Metals 1923 29. 573) who found that at 600' C. the oxide film increased in thickness during sixty to eighty hours after which no further appreciable weight-increment occurred even after prolonged heating. Assuming a normal density, the thickness of the film was then approximately 0-ooooz cm. 34 Kindly supplied by the British Aluminium Company. %See J. E. Edwards (Trans. Amer. Electrochenl. SOL. 1925 47) where the typical figure of 99'98 per cent. aluminium is given W. H. J. VERNON 151 Correc-tion.38 mg. fine emery paper which had served so well for copper was found to be quite unsuitable; particles of emery embedded themselves in the surface and could not be removed by ordinary methods.The treatment decided upon really consisted of '' dry scratch-brushing." Using a steel wire brush it was found that any pre-existing scale could be thoroughly and rapidly removed and quite a fine "silky" surface imparted to the metal. Immediately following the final treatment with the wire-brush the specimen was brushed quickly with a flat camel-hair brush to remove particles of abraded metal and then transferred as expeditiously as possible to a desiccator which was at once evacuated. In addition to the vacuum (oil) pump connection was also made to a nitrogen cylinder and by alternately filling with nitrogen and evacuating several times finishing with a vacuum (approx. I mm.) the removal of all air from the desiccator was ensured.Each experiment was performed in duplicate but usually a separate desiccator was used for each specimen. We&4ing.-The Conrady system of weighing was employed as in the copper experiments (see p. 118). On account however of the very ap-preciable " buoyancy effect " introduced by the difference in density be-tween the specimens and the brass weights the effect of fluctuations of barometric pressure and also of temperature had now to be taken into con-sideration and an appropriate correction was applied to each weighing.36 I n order to illustrate the data involved a portion of a table of results is re-produced in Table VII. The desiccators were left overnight at the side of the balance. TABLE VII. Corrected Weight. Grams.ALUMINIUM GRAVIMETRIC RESULTS CORRESPONDING TO PORTION OF FIG. 19. EXAMPLE OF EXPERIMENTAL DATA. Weight-Increment mg. (actual) 0'20 (I9 36 The importance of this procedure will be appreciated when it is stated that it was by no means unusual for the apparent weight of the specimen after exposure (even for as long as ten days) to be less than the apparent initial weight. 37 Al/H4 = Specimen containing 99-61 per cent. !A1 (see p. 150). Hard-rolled. Alp4 = , annealed. AM/H2 = Specimen containing 99-98 per cent. A1 (see p. 150). Al& = , annealed. weight in a vaciium by 12.0 mg. throughout. maintaining a constant load throughout. Hard-rolled. 38 The correction as used is short of the correction necessary to bring to the true s9 These weights were added to the specimen side of the balance for the purpose o 152 SECOND EXPERIMENTAL REPORT (c) ReIation between Weight = Increment and Time.6cShort-$eriod” Tests.-From the results plotted in Fig. 19 up to 27 days’ exposure it will be seen that an entirely new type of weight-increment curve is now represented; whilst at the start there is a resemblance to the parabola this quickly disappears in a “flattening” of the curve towards the horizontal axis.40 A consideration of the squares of the weight-increments leads indeed to the conclusion that at no time does the curve actually con-ferm to the parabola i.e. the process depends upon an essentially different mechanism from that which has obtained in any of the cases previously considered. The elucidation of this mechanism however presents greater difficulty than heretofore.It is probably not merely a coincidence that the present curve is similar in type to that obtained by various workers in connection 2 4 6 8 10 I2 14 16 18 20 22 24 26 28 30 Time in days. Fig. 19.-Oxidation of aluminium at the Grdinary temperature. Relation hetween weight-increment and time in the very early stages. A-AMJH2 and AM/A2 (see Table 4). B-Al/H4 and A1/S4 ( , , ). 9 = Squares of weight-increments ( x 2.01 values taken from Curve B. with veZocity of adsor-tion-e.g. of gases and vapours upon charcoal,41 celluloid,42 gla~s,4~ and (perhaps most significant) upon alumina. 44 Various equations have been proposed increasing in complexity the more nearly they agree with the experimental curve.43 I n discussing t6e complexity of the factors involved Freundlich 45 points out that the curve of adsorption velocity cannot be represented by a simply mathematical function and, 10 If the total weight-increment at ten days’ exposure be considered it is of interest to note that almost exactly half of this took place during the first day and rather more than one-third during the first eight hours.41 Giesen Ann. Physik. rgo3 10 838 ; Bergter ibid. 1g12 37,472 ; Harned J. Amer. Chem. SOC. 1920,42,920. “Lefebure J. Chem. Soc. 1914 105 328. 48 Bangham and Burt PYOC. Roy. SOC. 1924 IoSA 481 ; J. Physical Chem. 1925, 2 9 7 113. UMunro and Johnson J. Physical Chem. 1926 30 172. 6 Kapillarchemie 1923 148 Time in days. Fig. 20.-Oxidation of aluminium at the ordinary temperature.A and B.-Continuation of curvee A and B of Fig. 19. C.-Mean weight-increments from similar specimens started positions). Weight-increment a late = 54 SECOND EXPERIMENTAL REPORT clearly the same remark may be extended to the curve now under discus-sion. The similarity in the curves whilst strongly suggesting that adsorp-tion is common to the two processes does not of course permit of a really definite conclusion in regard to the oxidation of aluminium. Possibly the mechanism consists in the formation of an initial film of alumina with con-comitant adsorption of water vap0ur.4~ In this case the final product would be expected to consist of the hydrated oxide but experimental veri-fication is difficult on account of the thinness of the film.In one respect at least the disposition of the curve leads to a perfectly definite interpreta-tion i.e. the rapid flattening towards the time axis leaves no room for doubt that the resulting film is extremely impervious in character. The straight Zine film formation is regulated by gaseous diffusion only i.e. of the atmosphere through the interstices of the granular film. ThejaraboZa film formation is regulated by diffusion through a solid envelope. The present curve: film formation probably does not depend upon diffusion at all but upon adsorption. The resulting film is characterised by limited thickness and extreme imperviousness. Long-period Tests.-In the course of the experiments upon aluminium, a somewhat puzzling feature was encountered. Thus a series of determina-tions would be carried out the “flattening ” of the curve would be ob-served and thenceforward weighings would be conducted at less frequent intervals.The apparent constancy of the curve would then encourage leaving the specimens for a still longer interval but on returning a weight-increment would be obtained appreciably in excess of that required by extrapolation of the curve. The curves reproduced in Fig. 2 0 representing exposure up to 91 days, illustrate this difficulty but at the same time they include sufficient deter-mined values to supply a probable explanation. I t will be observed that whenever an increase in the weight-increment is obtained in excess of that expected the curve continues for a time without suffering further .alteration-i.e.the ‘‘flatness ” of the curve is retained-until another increase takes place when again there is a further period without change. I t is note-worthy that the “subsidiary increases” are of the same type as the primary increase i.e. the attack takes place very rapidly but equally rapidly slows down. Finally the conclusion is reached that when the primary film has practically ceased to thicken (after about two weeks under the conditions of the experiments) any further appreciable change takes place through the occurrence of cracks or fissures in the primary film. This explanation would satisfactorily account for the sudden breaks in the weight-increment curve ; it would be expected moreover that any exposure of the underlying metal would result in the formation of new film to a smaller extent but by a similar process to that which was formed originally.The position may be summed up as follows. (d) Thickness of Surrosion Film-Influence of Purity of Atmosphere and of Purity and Physical Conditions of Metal. In@uence of Purity of Atmos9here.- Thickness of Surrosion Fih.-For the purpose of computing the relative intensity of attack it is useful to take the value of the weight-increment when the curve has reached the ‘‘ nearly flat” stage some 10 or 14 days after the first exposure. The following 46 Munro and Johnson (Ind. Eng. Chem. 1925 29 256) found that alumina ad-sorbed water more readily than any other vapour W. H. J. VERNON I55 values were obtained at various times of the year the date of the first ex-posure being given in each case :-March 5th 0.17 mg.per sq. dm. April 24th 0.17 1 71 Aug. 14th 0.14 1 ? , 26th 0.18 1 > l Nov. 2nd 0.15 99 19 , 17th 0-18 0.24 , 9 , Dec. 11th 0x5 0.22 , 9 9 The extremely small limits within which the attack on the metal varies, renders it difficult to attach any significance to the differences observed. I t is true that the highest values were obtained at a period of minimum atmospheric purity (Nov. 17th and Dec. ~ ~ t h ) but in each case specimens exposed at the same time gave average values. I t is probable indeed that the differences observed are really due to differences in the relative freedom from oxide at the time of the initial weighings (since clearly the metal will never be absolutely free from oxide in spite of precautions).On the whoIe the constancy of the values observed is somewhat surprising. Including the abnormally high value of 0.24 for Nov. 17th (curve B in Fig. IS) the values range from 0.14 to 0.24 mg. per sq. dm. Accepting a mean value of 0.19 an estimate of the thickness of the corresponding film may be attempted; the result will of course depend upon the composition (and density) which the film is assumed to have but the probable alterna-tives cannot appreciably affect the order of the value obtained. Thus, taking the alternatives to be anhydrous or hydrated AI,O respectively on the one hand the density of the crystalline oxide is 3.9 ; on the other hand the commonest form of the hydrated oxide (i.e. the mineral bauxite-pro-bably a mixture of coIloidal hydroxides) is usually assumed to have a formula .of A$03 2H,O and a density of 2-55.Introducing these values the following relationships obtain :-A1,0,. A1,0,. aH,O. Increase in weight mg. per sq. dm. = 0'19 Weight of film mg. per sq. dm. = 0'404 0'313 0.19 Volume of film C.C. per sq. dm. = 1-04 x 10-4 1.35 x I O - ~ Hence thickness of blm cm. = 1-04 x I O - ~ 1-35 x 10-6 Thus in either case the thickness of the film is of the order of a millionth of a centimetre. The meaning to be attached to this value is that it re-presents the thickness of the " primary film," i.e. before any '' subsidiary increases '' have taken place. InJuence of PuriQ and Physical Condition of Metal.-As in the case of relative purity of the atmosphere it is only a negative result which has to be recorded.Experiments conducted with the greatest care employing the two materials described on p. 150 and starting on November 17th (see above, also Fig. 19) showed that the less pure material gave the higher weight-increment. In experiments conducted with equally great care a little later, however (Dec. I Ith ; above) the position was reversed the highly pure aluminium then showing the greater increment. Similarly with physical condition ; sometimes the hard and sometimes the annealed specimen has appeared to have suffered the greater change and no definite conclusion can be drawn. As before the differences actually observed may be due to differences in the initial stage of oxidation 156 SECOND EXPERIMENTAL REPORT 4 SECTION VL-LEAD.The object of these tests was precisely the same as that stated for aluminium i.e. the determination of the form of the weight-increment curve concerning which there appeared to be no available information. (a) Experimental. The experiments were conducted in the same atmosphere and precisely the same methods were followed as for aluminium. The specimens were taken from a sample of exceedingly pure sheet lead.47 Their dimensions were uniformly 10-0 x 10.0 em. in area and their thickness in most of the tests approximately I mm. giving a weight of just over zoo grams For the preparation of the specimens abrasion with emery was clearly even more out of the question than was the case with aluminium but fortunately the treatment with the wire-brush was found to serve the purpose admirably.Weighings were conducted at similar frequent intervals as for aluminium ie. from the first 8 hours onwards. Tables of results were identical in form to those for aluminium (see Table VII.) and an example is therefore not reproduced. I n weighing the buoyancy correction was again necessary. (b) Relation between Weight - Increment and Time. Shout-Period Tests.-The curve plotted in Fig. z I representing the means of closely-agreeing values from duplicate specimens up to 2 5 2 4 6 8 10 I 2 14 16 18 20 22 24 26 Time in days, Fig. 2l.-Oxidation of lead at the ordinary temperature. Relation between weight-increment and time in the very early stages. a = Squares of weight-increments (values taken from smooth curve) x 4.0. days' may be taken as quite typical of the results obtained.The similarity between this curve and the aluminium curve of Fig. 19 will be at once apparent ; a comparison of the squares of the weight-increments (plotted on each diagram) confirms that they are really of the same type. Clearly therefore the considerations already discussed under aluminium apply with equal effect to lead and again suggest that adsorption plays some part in the mechanism of film formation. Similar properties in regard to limited thickness and imperviousness may again be attributed to the film. 47 Supplied by Messrs. Cookson Newcastle. 48 First exposed on April Ioth 1925. Analysis 99.995 per cent. lead W. H. J. VERNON I 5 7 Lon,.-Period Tests. -The determination of the weight-increment over compsratively long periods is attended with greater difficulties than in the case of aluminium and no really satisfactory curves are available.The same phenomenon of unexpected weight-increments has been observed, however and it is highly probable that the behaviour of the " primary film " on more prolonged exposure is similar for each metal. (c) Comparison of Oxidation Values of Aluminium and Lead. As in the case of aluminium the intensity of the attack upon the metal may be conveniently estimated by noting the weight-increment when the nearly flat portion of the curve has been reached for which a somewhat shorter period-about 7 days-is usually required. The value so obtained is appreciably less than in the case of aluminium the figures ranging from 0-10 to 0.15 mg.per. sq. dm. as compared with 0.14 to 0.24 mg. per sq. dm. for aluminium. Owing to the difficulty of assigning a probable formula to the resulting oxide no attempt has been made to estimate the thickness of the film in the case of lead but it is probably of a similar order of magnitude to that which obtains for aluminium. (d) Anomalous Behaviour of Lead in Atmosphere Contaminated with Traces of Vapour from Drying Paint. To appreciate what follows it should be borne in mind that in none of the tests hitherto described (either on lead or aluminium) has any visible change been observed upon the metal. These tests had been in progress for about 7 months however when a somewhat extraordinary phenomenon was observed. Specimens of lead which had been recently cleaned and exposed were found to be undergoing a gross attack the metal rapidly developing a blue colour,4Q and actually becoming detp bZue in little more than a week.The cause was quickly traced to painting operations in the same building. Exploratory tests upon small specimens suspended over vessels containing the paint etc. showed that whereas the paint in bulk was without effect rapid tarnishing was induced by paint which was in the process of drying. These experiments suggest that it is the turpentine content which is ultimately responsible for the effects observed, but no attempt has been made to follow the matter exhaustively. The phenomenon is entirely peculiar to lead in the freshly-cleaned condition. Thus none of the other materials (aluminium copper zinc, etc.) which were exposed in the same room were in the slightest degree affected as proved by the intensive weighings then in progress; neither indeed were specimens of lead which had already been exposed for a sufficiently long time (this point is discussed below).On the other hand the magnitude of the attack upon freshly-cleaned lead can be gauged from the fact that in 3 weeks50 such specimens had increased in weight by 3-5 mg. per sq. dm. in striking contrast with the 0.15 mg. per sq. dm. which characterises the attack upon lead under normal conditions during a similar period and following a similar initial treatment. 49 Presumably a diffraction effect. LH) The attack then rapidly slowed down on account of the suspension of painting operations shortly after which the attack upon freshly-cleaned lead ceased altogether.1 58 SECOND EXPERIMENTAL REPORT (e) Protective Properties of Invisible Oxide Films upon Lead. In connection with the tarnishing of lead through the causes described above considerable interest attaches to the fact that if the previous exposure to the unpolluted atmosphere has been sufficiently prolonged no visible effect is produced upon the metal. In the absence of the actual specimens the phenomenon is best illustrated by reference to Fig. 22. Here Curve A represents the mean weight-increment of two specimens first exposed on April Ist 1925. Curve B represents to the same scale and in its correct chronological position (commencing Aug. I 3th) the attack upon specimens during the tarnishing period.The enormously greater rate of oxidation exhibited by Curve B will at once be noted but the point to which Time in days. Fig. =.-Rapid oxidation of lead on exposure to atmosphere contaiiiina traces of vapours from drying paint. (Curve BI. Immunity ot lead which has previously developed a thin film of oxide by exposure to nncontaininated air (Curve A). attention is specially drawn is the fact that Curve A in spite of the smallness of the ordinates has passed through the tarnishing period without having suffered the slightest deflection. This provides,an interesting sequel to the parallel case already discussed for copper (see Fig. 4) and constitutes an even more convincing example of the protection which may be conferred by a thin film of oxide.Once again the condition underlying the production of the non-protective type of film is determined by the incidence of a (‘ foreign constituent ” in the atmosphere in excessively small concentration. The nature of this constituent is different in the two cases and the effect is entirely selective upon the metal concerned but there can be little doubt that the mechanism of the process is essentially the same. The suggestion again arises that the difference between the protective and the non-protectiv W. H. J. VERNON I59 type of film lies in a relatively small difference in the original lattice. Although no analytical figures are available in the present instance it is indeed highly probable that the two films although differing so enormously in physical properties are really but little different in their chemical composition.SECTION VIL-IRON. The tests described in Section 5 of the First Report were started purely for comparative purposes and were admittedly strictly limited. Nevertheless the results appeared to possess sufficient interest to justify some extension of the original experiments. (a) Experimental. Throughout the recent work specimens measuring 10 cm. x 5 cm. havetbeen employed i.e. half the size of those used originally. They have been prepared by grinding to remove all traces of pre-existing rust finishing with the same grade of emery paper (Oakley 0) to ensure uniformity of surface. Particles of emery have been removed by rubbing with pure cotton wool as described for copper. Finally the specimens have been submitted to an over-night vacuum desiccation and weighed.Unless otherwise stated the material may be assumed to be the same-as that used in the original tests i.e. a commercial ingot iron.51 (b) Behaviour of Iron in an Indoor Atmosphere of Low Relative Humidity containing the usual Suspended Solid Particles-Effect of Filtering the Air and of Screening the Specimen behind Muslin. I n Fig. 23 Curve B (up to 92 days) represents a typical series of results obtained from the exposure of iron specimens in Atmosphere Type I, Period W.52 As already stated 53 the hygrometric conditions were such that the dew-point was exceeded by a wide margin (some 22’ F. or 12’-C.) during the whole time. The development of rust took place in the characteristic discontinuous manner which was described in the First Report (lot.cit. p. 887 Fig. 22). Subject to the proviso that each speck of rust was surrounded by unaltered iron (evident to visual inspection especially if aided by a lens) at the end of period W the whole surface was c c covered with rust. ” Considering the relative dryness of the atmosphere this result, although perfectly typical appears to be contrary to the prevailing impression that “ iron does not rust above the dew-point,’’ an impression which seems to have originated through a too literal application of a principle established 54 under academic conditions for the simple system of air and water-vapour. Clearly however ordinary air is a more complex case ; in addition to water-vapour and other incidental gaseous impurities it carries suspended d i d impurities although normally invisible in relatively large quantity.Solid particles in the atmosphere are in general hygroscopic through adsorption 61 The influence of composition (essentially negative under the conditions of the tests) is dealt with later. 62 The approximate position of the change from ‘‘ Period W ” to ‘‘ Period S ” is indicated on t h e diagram by a vertical line. 53 See page 117. 54 Dunstan Jowett and Goulding y. Cheni. SOC. 1905 87 1548 160 SECOND EXPERIMENTAL REPORT of hygroscopic matter55 and in the case of town air many of these are definitely acid through adsorption of acidic matter.66 I t would be expected, therefore that the deposition of such particles upon a clean iron surface would give rise to corresponding rust centres and hence would account for the rusting of the metal above the dew-point.To test this two small specimens of iron were polished and placed re-spectively in glass tubes one of which was preceded by a filter of pure cotton wool ; room air was then drawn through each tube at a similar rate. The experiment was conducted contemporaneously with the exposure of specimens yielding curves as in Fig. 23 and was discontinued after the passage of IOO cubic feet of air (7 days). The specimen over which the filtered air had passed was then still perfectly bright whilst the other was slightly but definitely rusted-again in the form of isolated ‘‘ dots ” over the surface. 57. 20 40 60 8c I00 120 140 160 Time in days. Fig. 23.-Rusting of Iron in indoor unsaturated atmosphere containing the usual suspended solid particles (Curve B).lmmunity of iron screened behind muslin (Curve A). The experiment just described although conclusive as to the rusting effect of suspended solid matter clearly suffers from the disadvantage that the motion of the air through the tube lessens the chance of particles ad-hering to the metal. Accordingly it was sought to demonstrate the effect by another method. Simultaneously with the exposure of the specimens represented by Curve B. in Fig. 23 (and immediately adjacent to them), three specimens were suspended inside a muslin c c cage.” 58 The arrange-ment was such that the specimens were entirely surrounded by a single thickness of muslin at a distance of several inches.The ((cage ” was only opened when necessitated by the periodical weighing of specimens. The somewhat extraordinary result then emerged that the rusting of the sjecimens 55 See contribution by Dr. Simpson to the discussion on the First Report Zoc. cit., 56 See Cohen and Ruston’s (‘ Smoke A Study of Town Air ’’ (Arnold). Shaw and 57 In contradistinction to this result it is of interest to recall that filtering the air 58 These experiments were started on January 3oth 1925. p. 912. Owens’ ‘‘ Smoke Problem of Great Cities ” (Constable). is without influence upon the tarnishing of copper. Also J. S. Owens Proc. Roy. SOC. A 1926 110. See First Report loc. cit. p. 895 W. H. J. VERNON 161 behind the muslin was entire4 inhibited so that at the end of period W 59 they were apparently as bright as when first exposed whilst the unscreened specimens had rusted in the manner already described.Although visibly unaffected the protected specimens increased slightly in weight during ex-posure and their weight-increments are plotted in Fig. 23 (Curve A). The disposition of the curve for the normally exposed specimens may now be explained. Rusting is due to the precipitation of solid particles from the atmosphere. Each particle is associated with sufficient adsorbed moisture or adsorbed acid to account for the attack upon the metal in the immediate vicinity and no further. At first when the whole area of the specimen is available for attack the resulting weight-increment is relatively large but the rate of change slows down as the available area becomes less.The falling off in the rate of attack is clearly due to fundamentally different causes from those which have obtained in any of the non-ferrous examples, and any resemblance in the curves is therefore purely superficial. So far from being (‘ protected ” in the ordinary sense the surface of the metal at the end of Period W is potentially in an extremely active condition. This, however will be discussed later. .(c) Behadour of Iron in Atmosphere Saturated with Water-Vapour (‘6 Dew-point ’’ Conditions) but Practically Free from Suspended Solid Particles. The specimens from which the curve of Fig. 24 was obtained were sus-pended on a wooden frame (approximately 2 ft. sq.) surrounded by muslin; I 24.0 1 2 3 4 5 6 7 8 Time in weeks.Fig. 24.-Rusting of iron in highly-humid atmosphere ( ‘ I dew-point conditions ”) practically free from suspended solid particles. this was placed in a large earthenware vat containing a layer of water into which the lower part of the muslin dipped. The whole was covered by a glass plate and exposed to conditions away from artificial heating so that XI Subsequent results obtained after passing through Pdriod W are discussed later-(e) p. 164. I 162 SECOND EXPERIMENTAL REPORT diurnal fluctuations in temperature would undoubtedly result in the pre-cipitation of moisture. The stand carried 16 specimens of which two were removed weekly and replaced by two others. Fig. 24 represents the mean total weight-increments for each week during 8 weeks’ exposure.I t will be seen that the curve is fundamentally different from that discussed under the previous heading (Fig. 23). As would be expected from the virtual absence of solid particles the attack is relatively slow at first but accelerates later, owing no doubt to electro-chemical action rendered possible by the presence of a film of moisture. The relative purity of this moisture film.presumably keeps down the rate of acceleration and under the conditions of the ex-periments a longer time of exposure would have been preferable; never-theless the curve is sufficiently advanced to bring out the effect quite clearly. (d) Behaviour of Iron in Indoor Atmosphere of High Relative Humidity. The newer tests described below were carried out in atmosphere Type I during Period S.The (approximate) mean relative humidity during this period is 68 per cent. as compared with 43 per cent. for Period W. Sus-pended solid matter is most probably smaller in total amount and is certainly less active in character than that which obtains during Period W. Case I. The iron is initial4 free from rust.-This will be recognised as an intermediate case between (b) and (c).above both in respect to the re-lative humidity of the atmosphere and also its content of active solid particles. It would be expected therefore that the effects upon a clean iron surface would be intermediate in type between those observed in the examples already given ; this is borne out by experiment. Thus in one season’s observations the curve was slightly concave about the vertical axis (i.e.tending to follow Fig. 24) whilst in another the type of Fig. 23 was followed-with however an intensity of attack only half that during period W (0.6 as compared with 1.2 mg. per sq. dm. per IOO days). I n the latter case clearly the observed effect was still due to active solid particles only and the increased humidity was without influence during the period of the test.6O. Case 2. The iron is iEitial4 covered with dry rust.-Previous con-siderations have shown that the function of suspended solid particles is the initiation of rusting whilst the function of water-vapour (when present in sufficiently high concentration) is to supply the necessary moisture for the acceleration of this primary attack. The acceleration is effected by the juxtaposition at each rust spot of iron rust water and oxygen.The presence of liquid water under conditions above the dew-point is accounted for in two ways (I) the hygroscopic character of the ‘‘ rust-spot,” due both to the presence of the initiating “solid particle,” and to the rust itself; (2) the increase in the concentration of atmospheric water-vapour above a certain critical value. If this concentration is not reached a curve of the type of Fig. 23 is obtained. If it is reached and the surface of the iron is initially clean the rate of rusting will still depend upon the rate at which solid particles can be precipitated upon the surface and this may be quite 60 In the light of results which immediately follow it would seem that the accelera-tion of the curve front the start above the dew-point depends upon the precipitation of active solid particles in conjunction with at the outset a sufficiently high humidity.The ’ 4 critical humidity ” necessary for this purpose appears to be distinctly higher in the early stages when the rust centres are widely separated than later on when the specimen is almost entirely covered with rust W. H. J. VERNON 163 slow. (It seems probable indeed that they are not precipitated so fast when the humidity is high.) An interesting question arises as to what will happen if a surface already covered with rust (by the first of the processes con-sidered) is presented to an atmosphere of approximately the critical humid-ity. Clearly this state of affairs is reached when specimens which have been exposed during '( Period W7" emerge into (' Period S." The answer was at first supplied with unexpected and somewhat startling emphasis.Thus in the tests represented in Fig. 23 the mean weight-increment was 20.0 I I I I I I 1 I 1 I PERIOD W 2.0 -Q - B I I 1 I I I 20 40 60 80 100 100 160 180 200 Time in days. Fig. 25.-A-Effect of increase in atmospheric humidity (Period S) upon iron which has already rusted in a relatively dry atmosphere (Period W). B-Intrinsic effect of the more humid conditions upon freshly cleaned iron. determined practically at the end of Period W (i.e. 9 2 days); it was then 2-2 mg. per sq. dm. No other determination was made until a further 67 .days had elapsed i.e. well within Period S ; the weight-increment was then found to have the astonishing value of 48-5 mg.per sq. dm.61 This result was of such interest that it was decided at the next oppor-tunity to trace the behaviour of the metal systematically during the transi-tion period. Again the vertical line indicates approximately the position of change from Period W to Period S. (The values plotted are the means of figures obtained from six In Fig. 23 this corresponds to an ordinate (at 159 days) more than 12 times the The result is shown in Fig. 25 Curve A. total height of the diagram 164 SECOND EXPERIMENTAL REPORT specimens each of which would have yielded the same type of curve.) The almost immediate eKect of entering the more humid period is now apparent, whilst the extraordinary acceleration is brought out clearly.Additional point is given to this result by consideration of the effect of the same at-mosphere upon freshly cleaned specimens as shown by Curve B in Fig. 25. (e) Formation of an Invisible Protective Oxide Film on Iron by Ex-posure to Air (Free from Suspended Solid Particles). The inhibition of rusting by the screening of specimens behind muslin has already been described; it will be recalled that specimens exposed in this way during Period W showed no visible alteration at the end of 9 2 days during which time normally exposed duplicates had become covered with rust. I t remains now to add that the altered hygrometric conditions consequent upon emerging into Period S which have been seen to produce 1.6 E' 4 .& 20 40 60 80 I00 I20 140 Time in days.Fig. 26.-ReIative behaviour on exposure to room atmosphere of A iron which has been previously exposed for 11 months behind muslin (without rusting) and 13 freshly-cleaned iron. extraordinary effects upon rusted specimens and relatively slight effects upon those which have been freshly cleaned were st2l witAout any visible efect ujon specimm screened behind mudin. Neither as will be seen by reference to Fig. 23 was there any appreciable alteration gravimetrically ; the very small weight-increment already recorded had merely increased at almost exactly the same rate; the total increment at 159 days was 0.2 mg. per sq. dm. On the other hand as already noted the tota1:weight-increment suffered by unprotected specimens was then 48.5 mg. per sq. dm.62 After 314 days' exposure behind the muslin one of the specimens (still apparently unaltered) was withdrawn and exposed in the ordinary way to-gether with freshly-cleaned specimens.This was done early in Period w, so that the atmospheric conditions coincided exactly with those discussed under (c) above. The interesting result then emerged that fur a time t k oZd specimen stiZZ preserved its brightness whilst the rusting of the new ones 62At this stage the disparity was increasing rapidly and a little later rust was actually falling away from the unprotected specimens thus ruling out any further deter-minations of weight-increment W. H. J. VERNON 165 proceeded ilz the usuaZ way. Later tiny specks of rust appeared but the subsequent course of attack was then quite different from that which ob-tained in other cases and consisted in an intensification of existing rust-spots rather than in the formation of new ones.From the weight-increments plotted in Fig. 26 the increasing rate of attack upon the pre-exposed specimen will be noted in contradistinction to the falling-off in the attack upon the normal specimens. At the same time the relatively large disparity in the ordinates of the two curves confirms the suggestion that a definite degree of protection has been conferred by previous exposure. Apart from the actual breakdown of the film in the present instance there is an obvious an-alogy with results already recorded for copper (p. 130) and for lead (p. I 58). Although the nature of the attack has differed in each case the resistance to this attack has undoubtedly originated in a common factor throughout, namely the formation of a film of oxide during previous exposure of the metal to uncontaminated conditions.Incidentally the results would ap-pear to have some bearing upon the wider phenomena of passivity.63 Finally it may be observed that the local acceleration in the rate of attack upon the protected specimen is quite in harmony with the view that the protection is due to the formation of an invisible oxide film. At any rust-spot conditions are favourable for an ‘‘ oxygen concentration cell,” the underlying metal tending to become anodic. When however the rust is formed by the perforation of an initially-continuous oxide film at any point, clearly the ennoblement of the surrounding cathodic area will then accen-tuate any electrochemical effect and hence lead to increased attack.(9 Influence of Purity of Metal upon Rate of Rusting. In the bulk of the tests described above specimens representing each of (I) Ordinary Ingot Iron (cold-rolled close-annealed) as used in the ( 2 ) EZectrol‘ytic Analysis Carbon 0.04 per cent. ; Mn Si S P, ( 3 ) “ Armcu” Ingot ( 4 ) Spring SteeZ. The ordinary ingot iron only was exposed to dew-point conditions. So far as conditions above the dew-point are concerned it can be said quite definitely that the influence of purity and of composition (within the limits shown above) is nil. Specimens representing all materials have been ex-posed on various occasions and even where comparatively wide differences have been displayed the results have been definite only in the sense of showing that the composition of the metal had taken no part.the following materials have been included :-earlier tests. Ni Cr (total) 0.167 per cent. ; Iron (by difference) 99’833 per cent. 99.84 per cent. Iron. 9.5 per cent Carbon. (g) Behaviour of Iron compared with that of Non-ferrous Materials. In the First Report (Zoc. tit. 886-890) results were quoted showing the relative behaviour of iron and non-ferrous metals in the atmosphere of the ‘‘ basement ” and the “ tank-room,” and also on exposure to the open air. Comparable results relating to exposure in a domestic kitchen are included in Section IV. of the present Report. 63 Note for example the similarity in the effects produced upon copper by treatment with chromic acid (p.120) and by exposure to air (p. 130). 6J Kindly supplied by Dr. W. H. Hatfield to whom the author is also indebted for a complete analysis and for a photomicrograph of the material. Presented by the Whitecross Company Ltd. whose courtesy the author acknow-ledges 166 SECOND EXPERIMENTAL REPORT PART IX-OPEN-AIR EXPOSURE TESTS. Introduction. In the First Report tests conducted upon specimens exposed completely to the open-air were described for copper zinc and brass (lac. cit. pp. 975-80) and for iron (pp. 890-2). The results then discussed were confined to those yielded by measurement of the increase in weight of the specimen (up to 6 days) measurement of loss of reflectivity (various periods up to a maximum of little over 2 months) and ordinary visual inspection (up to the total period of just over 12 months at the time the report was written).In the tests to be described (from which however iron has been ex-cluded) the period during which the course of corrosion may be quantita-tively followed has been extended by a system whereby the rain-water passing over specimens has been collected and analysed at convenient intervals. Ordinary loss of weight tests (supplemented by the analysis of corrosion products) are also described representing a total period of exposure of over 4 years. Square specimens of one decimeter edge and 3 mm. thickness have been employed throughout and several types of surface condition have been investigated. (Series A.) The exposure of the specimens has been conducted entirely at South Kensington on the roof of the Royal School of Mines building (see First Report doc.cit. p. 8 5 ~ ) . ~ ~ (Series A,) (Series B.) SERIES A.-COPPER ZINC AND BRASS. Systematic Tests during Course of Exposure up to 100 weeks In-Computation Analysis volving Estimation of Metal Removed by Rain. of ‘‘ Erosion,” ‘‘ Surrosion,” and “Total Corrosion.” of Corrosion products. Experimental. In the case of indoor exposure the changes taking place on the metal surface have been followed systematically by observation of the increase in weight of the specimen. On complete exposure to the open air how-ever serious limitations are imposed. I t is true that here again the action results in a scale which of itself must contribute to the weight of the specimen-indeed measurements of weight-increase in the early stages have actually been carried out.But superimposed upon this effect-i.e. the surrosion 67 effect-the corrosion product must now be carried away to a greater or less extent through the agency of rain. To distinguish the latter from the former process a special term is clearly desirable and fortunately an appropriate one is forthcoming in the word (First Report Zoc. cit. 879.) 66 Tne task of extending the open-air tests to different localities-ultimately it is hoped to different climates-was entrusted by the Committee in March 1925 to Mr. J. C. Hudson M.Sc. To meet the special circumstances involved Mr. Hudson has de-veloped a method by which the electrical resistance of small coils is accurately deter-mined before and after corrosion ; a statement on this method in relation to the problem in hand was made by him in the discussion recently of a paper by Professor E.Wilson-Proc. Physical SOL 1926 39 23. See also First Report loc. ctt. 851. J. Newton Friend Cavnegie Scholarship Memoirs 1922 XI I W. H. J. VERNON 167 erosion.68 This expression will therefore be employed to denote that type of corrosion which results in the removal of material from the metal surface-as distinct from surrosion which adds to it. Thus atmospheric corrosion of the open-air type may be regarded as being made up of two components, surrosion and erosion respectively. It is evident moreover that the loss in weight of the specimen after prolonged exposure cannot be accepted as a criterion of the extent of the total alteration of the metal surface since it fails to distinguish between the two opposing factors of which it is the re-sultan t.69 In the present tests (which were started on July 21st 1922) an attempt has been made to obtain a direct measure of erosion by determining chemically the amount of metal removed in the rain-water falling over the specimen. This method has the advantage of enabling information to be obtained from time to time as to the course of erosion without interfering in any way with the specimen itself. If however the loss in weight is determined at any given time sufficient data are at once available to determine from the total erosion the extent of the surrosion with which i t is accompanied.This was done for the whole series of specimens at IOO weeks’ exposure. Specimens Employed in the Tests.-Specimens (dimensions as given on p. 166) of copper zinc 70/30 and 60/40 brass were duplicated in each of the physical conditions enumerated in Table VIII. Method of Carrving Out the Tests.-The general arrangement of the specimens during exposure will be clear from the photograph which was reproduced in Fig. 3 of the First Report. Specimens were suspended both vertically and horizontally the rain being collected in suitable vessels and transferred to storage bottles for subsequent evaporation and analysis. In the method as carried out the solutions have been filtered previous to evaporation and the residue treated separately with the object of differ-entiating between corrosion product in suspension and that in solution.Actually however the proportion held in suspension has been found to be extremely small copper specimens yielding from 2 to 3 per cent. insoluble matter brasses approximately I per cent. and zinc less than I per cent., showing that by far the greater amount of metal carried away in the rain-water is in a state of true solution. The figures given in Table VIII. incorporate in all cases the very small weights of metal found in the insoluble portions. Discussion of Results. (I)--“ Erosion ” Values (Metal Attacked, and Removed in Rain). With reference to the results set out in Table VIII. the following com-ments may be offered. Influence of Surface “Skin ” from Manufacture.-Specimen No.I in each group was exposed in exactly the same surface conditions as received from the works i.e. without polishing treatment of any kind. I t is perhaps 6B The author is aware that in corrosion literature the word is now generally used to signify the removal of metal by purely mechanical agencies. There appears however, to be a very good case for using the terms surrosion and erosion in the way adopted in the present report as originally suggested by Friend. fly I t is true that in practice the specimen is scraped to remove the product of sur-rosion but a t best the method is unsatisfactory on account of the virtual impossibility of effecting a clean separation of product and metal. The difficulty is especially great in the case of a material having a duplex structure e.g.60/40 brass where one constituent may be attacked to a much greater extent than the other I 68 SECOND EXPERIMENTAL REPORT significant that the only really definite result is that yielded by copper, which shows an appreciably lower rate of erosion for the untreated specimen. TABLE VIII, COPPER ZINC 70/30 AND 60/40 BRASS OPEN-AIR EXPOSURE (SOUTH KENSINGTON). " EROSION " VALUES METAL IN RAIN-WATER COLLECTED FROM BRNEATH SPECIMENS. Reference No. '074 '824 A. = Annealed. W. = Surface as received from works (i.e. riot polished). D. = Surface polished finishing with fine emery paper (Hubert No. I). B. = Surface polished finishing with " mops." (Bright '' mirror " finish.) "To convert to grams per sq. dm. values should be halved throughout W.H. J. VERNON 169 There is some slight indication of a similar effect in the case of 70/30 brass, but with 60/40 brass and with zinc no distinction whatever can be drawn. In no case do the figures justify any conclusion being reached either as between hard and soft specimens or those in the dull and bright condition respectively except that any effects due to physical condition cannot have been appreciable under the conditions of the experiments. Indeed in the case of copper the values present an agreement which appears to be some-what remarkable. Influence of Physical Condih’oPt.-(Nos. 2 to 5 in each group). 20 40 60 a0 100 Time in wPeekq. Fig. 27.-8‘ Erosion ” of copper 70/30 brass 60/40 brass, end zinc on exposure to the open air (South Kensington).Weight of metal removed in rain. water in relation to period of exposure, Ifizuence of Mode of Suspension.-(Nos. 6 and 7 Copper; No. 6 in other groups). Specimens suspended horizontally have yielded appreciably higher results than those suspended vertically. The effect is most marked with 60/40 brass. ReZaative Erosion of MaterzaZs.-ReZaation between Erosion and Time.-The average total values obtained from polished specimens at the end of 26 7 2 and IOO weeks’ exposure respectively are plotted in Fig. 27. This serves both to show the relationship in behaviour among the different materials and also the relationship between erosion and time in each case. I t will be seen that the rate of erosion is excessively greater for zinc than for copper while the brasses occupy an intermediate position.Although on account of fluctuations in atmospheric conditions undu 170 SECOND EXPERIMENTAL REPORT significance cannot be attached to the disposition of the curves individually, it is noteworthy that the first two determined points fall almost exactly upon a straight line. In the case of copper the linear relationship is maintained, approximately up to the last point at IOO weeks; in comparison however, there is evidence of acceleration in the case o the other materials more especially the brasses. “ Dezinczj5cation ” of the Brasses.-The predominance of zinc in the corrosion products of brass is a quite general phenomenon to which the 300 200 c 9 ZOO i3 3 --70:30 - BRASS -20 A0 60 80 100 Time in weeks.Fig. 28.-“ Dezincification ” of 70130 and 60/40 brass on exposure to the open-air (South Kensington). Respective weights of copper and zinc removed in rain-water in relation to period of exposure. term dezincification is normally applied. In the case of sub-aqueous cor-rosion however the work of G. D. Bengough has shown that almost certainly the brass is attacked first as a whole and that the appearance of excess of zinc in the product is due to a secondary reaction in which copper is re-deposited. Particular interest attaches to the values for the brasses in Table VIII. as showing the excessive preponderance of zinc over copper in the rain-water solutions. I n Fig. 28 the values yielded by each constituent metal are plotted separately for each alloy. I t will be seen that although the total erosion is approximately the same the extent of (( dezincification TABLE IX.Copper Zinc 70/30 and 60/40 Brass Open-air Exposure Corrosion Values after 100 Weeks. (All Values in Grams as Actually Determined.) Reference No. 79 To convert to Grams per Sq. Dm. values shoul = 7 2 SECOND EXPERIMENTAL REPORT or ‘‘ copper re-deposition ” is very much greater for 60140 brass than for An interesting sequel arises in connection with the “ surrosion ” results, 70/30-to which reference will be made later. Discussion of Results. (2)-“ Surrosion ’’ Values (Metal Attacked, but remaining in situ). Computation of Surrosion. Analysis of Residual Corrosion - Products. When the specimens were dismantled at IOO weeks’ exposure they were each washed many times with hot distilled water and the metal deter-mined in the washings ; the value obtained (actually quite small) was added to the erosion figures previously obtained.The object of this treatment was to obtain a figure for the maximum erosion including not only metal already removed by rain but also that still associated with the specimens in soluble form and therefore ready for such removal. The value for the total erosion (E) is given in Table IX. After washing the specimens were dried, thoroughly desiccated and weighed. This weighing (W,) together with the initial weighing (W,) is also in Table IX. I t will be seen that in no case is the loss in weight equal to the weight of metal actually removed. An interesting situation is presented by 60/40 brass where with the one excep-tion of the horizontal specimen theJinaZ we@ of the specillten is consistently greater than the initial we&& in spite of the relatively large amount of metal which has been removed.The discrepancy is clearly due to the weight of the non-metallic portion of the scale which still adheres firmly to the underlying metal. I t follows that an expression for the weight of material associated with the metal in this way is given by S = W2 + E - W,, where W is the final weight of the specimen W the initial weight and E the weight of metal carried away by rain. S is clearly the (net) surrosion of the specimen-calculated values of which are given in Table IX. The abnormally high surrosion of 60/40 brass is perhaps significant in view of the excessive ‘‘ dezincification ” displayed by this material (vide supra).In work published in 1 9 2 2 ~ ~ R B. Abrams after confirming the earlier conclusions of Bengough as to the initial mechanism of dezincifica-tion showed that redeposition of copper does not take place unless there is some means of holding the dissolved metal in contact with brass usually effected by the presence of some kind of “membrane.” I t is of interest to note that in the present instance the membrane is forthcoming in the relatively thick layer of ‘‘ surrosion product.” The complete analysis of the product was carried out for each metal; the results are given in Table X. Generally no attempt was made to ascer-tain the solubility of the (powdered) product in water and in most cases in view of the smallness of the total amount and the very thorough washing which it received before removal there was very little reason to suppose that any soluble matter was left.I n the case of 60/40 brass however the pos-sibility was suggested by the much larger amount of residual product; accordingly a separate estimation was made the results of which are given below. 5z Trans. Ainer. Electrochem. SOC. 1922 42 39 W. H. J. VERNON I 7 3 TABLE X. COPPER ZINC 70/30 AND 60140 BRASS. SERIES A. ANALYSIS OF RESIDUAL COR-ROSION PRODUCTS AFTER IOO WEEKS' EXPOSURE TO THE OPEN AIR (SOUTH KENSINGTON). A. Analyses of the Products after drying at 105~ C. Values as determined-metallic and non-metallic radicles and matter insoluble in acids.Oxide hydroxide By 26.3 and combined H20 } difference 1 1 42'8 1 Anaiysis of Residual Corrosion-product from 60/40 Brass. 8.7 22.6 per cent. 1 Aq tteOus Edrmt- c u trace Zn Insoluble Residue- c u Zn [SO41 nil 10.8 L77-4 per cent. [O] [OH]! Comb. H20. Matter insol. in acids.} 24'g1 The presence of 2 2 per cent. of soluble matter (practically entirely it may be noted in the form of zinc sulphate) appears remarkable and testifies tn the nercictentv with arhirh cnliihlP matter m a v he retained hv the ccrnem-brane " in spite of drastic washing of the specimen. a similar weight of [O] and [OH] radicles. 74 For the immediate purpose it is assumed that a given weight of iron accounts fo I74 SECOND EXPERIMENTAL REPORT AnaZysis of Residual Corrosion Products (" Surrosion Products ")from the Whob Series.75-Following upon the weighing (W2) of the (dried) specimens, they were scraped with a blunt tool to remove the residual corrosion pro-duct as completely as possible with the minimum disturbance of underlying metal. The specimens were finally weighed (see below) and the collected products submitted to analysis. The results of the complete analyses in Table X. are given in three ways :-Individual radicles as actually determined together with insoluble matter (carbonaceous and siliceous). This insoluble matter is obviously *derived from external sources and carried on to the specimens by wind; iron which appears in all the analyses must also be due to wind-borne contamination (largely ' no doubt from iron railings in close proximity) since all the materials were initially free from this element.The relatively large amount of material available in the case of 60/40 brass suggested a separate analysis of the product from the specimen exposed '( as received )) ; apart from indicating a somewhat reduced dezincification the result differs but very little from that given by polished specimens. Mean values of essential radicles excluding extraneous matter (in-soluble and rust). The figures for iron were eliminated by assuming a given weight of iron to account for a similar weight of [O] or [OH] radicles. This gives the results in the form most convenient for comparison. Probably the most striking feature is the high content of sulphate and sulphide particularly in the case of the brasses with correspondingly low carbonate content.The production of sulphide appears to be especially favoured by 60/40 brass as will be seen by the high figure of approximately 42 per cent. for the total sulphide in the product from this material. A. B. C. Mean values of main constituents. Discussion of Results. (3)-Computation of Total Corrosion. Relation between Loss in Weight and Total Corrosion. Computation of Total Corrosion.-It is evident that an expression for the total corrosion of the specimens must include values for erosion and surrosion respectively. A difficulty arises however as to the form in which the separate values are expressed since in Table IX. E represents actual weight of metal removed whilst S corresponds to the weight of non-metallic radicles in the residual corrosion product.The analyses however now enable a value to be obtained for the metal attacked but remaining in sit%. This has been done for each material and the results (S') are given in Table IX. Clearly these figures are directly comparable with E and therefore the sum of the two values represents the total weight of metal attacked i.e. C = E + S', where C is the total corrosion. This value is given in Table IX. for the complete series. ReZation between Loss of We&& and Total Corrosion.-It is of interest to compare the final loss in weight of the scraped specimen with the value for complete corrosion obtained by the methods described above bearing in mind that it is loss in weight which is usually accepted as the criterion of corrosion.I t will be seen that with copper and zinc the difference between the two values is not very serious the loss in weight accounting for approximately The figures are given in Table IX. 75 For methods of analyses see Appendix W. H. J. VERNON I 7 5 90 per cent. of the total corrosion in each case. The discrepancy is com-paratively large (67 per cent.) for 70130 brass and in the case of 60140 brass reaches the seriously low figure of 56 per cent. The explanation of this discrepancy is well brought out by observing the behaviour of 60140 brass after the corrosion product has been removed ap-parently completely. On polishing and exposing to tarnishing conditions, the '' original beta )) constituent tarnishes extremely rapidly owing to the fact that it now consists very largely of redeposited copper.If examined under the microscope after this treatment the boundaries of the tarnished beta and the still bright alpha show up very ~learly.7~ Moreover the cor-roded beta constituent contains numerous cavities offering lodgment for 900 800 ui so0 E L o d 1 300 - 9 0 30 40 Composition per cent. zinc. Pig. 29.-Corrosion Values of Ccpper 70/30 brass 60/40 brass, and zinc after 100 weeks' exposure to th+ Opexi Air (South Kensi ngton). corrosion product which thus fails to contribute to the observed loss in weight of the specimen. In Fig. 29 the several quantities which have been discussed are plotted against composition; it will be observed how very definitely erosion or surrosion or loss in weight fail in themselves to supply a true estimate of the total attack upon the different materials under conditions of complete exposure.SERIES B.-COPPER ( VARIOUS GRADES) ZINC AND BRASS (70130 AND 60140). 225 WEEKS EXPOSURE. Specimens of this series of which the various details are given in Table XII. were exposed on the roof of the Royal School of Mines building, 76 Normally under the same conditions it is the copper constituent which tarnishes first. See p. 14 176 SECOND EXPERIMENTAL REPORT 6.4 56'5 31.1 6.0 from July 21st 1922 to Nov. nnnd 1926 supported on a stand a photograph of which was reproduced in Fig. 2 of the First Report. On dismantling they were dried scraped with a blunt tool as before and the loss in weight was determined; the weight of product removed from each specimen was also obtained and the products were finally submitted to analysis.The pro-cedure differed from that in Series A in that the rain-water solutions were not collected during exposure. Nevertheless the information gained from the more intensive earlier series was utilised in the interpretation of the later results and in the computation of the approximate total corrosion. Analysis of Residual Corrosion Products.-The results set out in Table XI. may be compared with the results already given for the shorter period of exposure in Series A (Table X.). I t will be seen that whereas the sulphide content of the products from copper and 70/30 brass has fallen considerably in the case of 60/40 brass and zinc it has remained about the same; hence the product from 60140 brass is characterised by an even greater relative preponderance of sulphide.TABLE XI, COPPER ZINC 70/30 AND q 4 0 BRASS. SERIES B. ANALYSES OF RESIDUAL COR-ROSION PRODUCTS AFTER 225 WEEKS' EXPOSURE TO THE OPEN AIR (SOUTH KENSINGTON). A. Analyses of the products after drying at 105' C. Values as determined-metallic and non-metallic radicles and matter insoluble in acids. B. Mean values of metallic and non-metallic radicles excluding iron (with associated radicles) and insoluble matter. [Ol Cow; and' BY C. Mean estim,.ec Carbonate . . . . Sulphate (anhydrous) . . Sulphide . . . . . Oxide hydroxide By and combined H,O}dXerence. The sulphate content of the copper product having greatIy increased, sulphate is now highest for that metal and lowest for zinc; also it is much higher for 70/30 brass than for 60140.The carbonate content is low throughout and there is little to choose in this respect among the materials. Generally the proportion of oxide or hydroxide is much higher than in Series A (i . the products are more basic) ; in both series it is highest for zinc. Determination of Loss in Wezght of Specimen and Computation of TotaZ Corrosion.-It should be explained that when the superincumbent corrosion product has been removed there remains a @m which appears to have entirely different properties and which cannot usually be dislodged wTithout removing underlying metal. The figures given for loss in weight do not necessarily include the weight of this film.For example in the case of arsenical copper the film is exceedingly tenaceous and smooth and is virtually a part of the metal itself. I t can only be removed by the use of a sharp instrument with consequent abrasion of metal. The values for approximate total COYYQS~OH given in Table XII. were obtained by applying the appropriate factor to the loss in weight figure (see Table IX.). As explained above the correction is not serious for copper and zinc but is very appreciable for the brasses. TABLE XII. COPPER (VARIOUS COMPOSITIONS) 70/30 BRASS 60140 BRASS AND ZINC. OPEN-AIR EXPOSURE (SOUTH KENSINGTON). LOSSES IN WEIGHT AND COMPUTED TOTAL CORROSION AFTER 225 WEEKS. Mark. Description. Copper. 77 Values for " approximate total corrosion " are obtained from values for loss in 78 Note that these figures (grams per sq.decim.) require to be doubled to be com-weight by application of appropriate factor ; see last column of Table IX. (Series A). parable with those in Table IX. I 178 SECOND EXPERIMENTAL REPORT Inzaehce of Size of Sjechen.-Zn Table XII. specimens which are marked * are of smaller area each side being 0.4 sq. dm. as compared with I -0 sq. dm. in other cases. Comparison of results from the first two “ pairs ” (all H.C. copper) brings out the much greater intensity of corrosion on the smaller specimens loss in weight per unit area exceeding that of normal-sized specimens by about 65 per cent. Relative Behaviour of Materials ( I ) Cojper.-In Table XII. the relative corrosion values of the materials are given that of H.C.copper being taken as 100. Comparing the different varieties of copper it is significant that corrosion is most severe on the purest material. The material which has resisted attack by far the best is arsenical copper, the corrosion value of which is only 43 per cent. that of H.C. copper. Nickel copper ( 2 8 per cent. Ni) comes next with a corrosion value of 62 per cent. In this connection it is of interest to recall the description of the visual results in the very early stages of exposure (First Report, Zoc. cit. p. 875).-“With certain exceptions the appearances observed are practically the same for various grades of copper the exceptions so far experienced being copper containing nickel (2.5 per cent.) and arsenic (0.45 per cent.).I n each case a more rapid darkening of the original ‘ light-brown patina ’ is observed ; whilst in the case of nickel-copper however the effect is comparatively slight with arsenical-copper the greater rapidity of darkening is extremely pronounced.” In connection with the optical results it was stated (doc. cit. p. 878) that “among the copper specimens those of arsenical copper are always the first to have their reflectivity reduced beyond the limits of the optical bench.” I t would seem therefore that the formation of this initial film is actually a criterion of subsequent resistance. The amount of corrosion product removed by scraping may be taken as indicating very approximately the ‘‘ surrosion values ’’ of the specimens. Relative values obtained in this way are given in the last column of Table XII.from which it will be seen that there is a rough parallel between surrosion and total corrosion ; the explanation no doubt lies in the fact that through the accumulation of corrosion product water is held for longer periods in contact with the metal. ReZative BeAaviour of Materials ( 2 ) Bmss and Zinc.-It will be seen $hat the present results confirm those already recorded in Series A. and hence call for little further comment. Interest chiefly attaches to 60/40 (brass which shows evidence of accelerated corrosion in the high value of 227 (H.C. Copper IOO) actually by far the highest of the series. The definite results yielded by copper of various compositions (above), suggest the desirability of obtaining comparable information upon zinc.In a new series of tests which is now being planned it is hoped to amplify the work upon zinc particularly from the point of view of influence of com-position. The author desires gratefully to acknowledge the invaluable encourage-ment and advice he has at all times received from Professor H. C. H. Carpenter F.R.S. under whose general supervision the research is being conducted. To Dr. R. S. Hutton and Dr. 0. F. Hudson of the British Non-Ferrous Metals Research Association his best thanks are due for their sustained interest and readiness to help in every way. His colleague Mr. J. C. Hudson M.Sc. although fully occupied with the special investigation of another branch of the work the results from which will be reported in due course has also co-operated most helpfully W.H. J. VERNON I79 I t is again the author’s pleasant duty to acknowledge the valuable services of Mr. L. Whitby (Research Assistant) and he would like especially to record his appreciation of Mr. Whitby’s skill in connection with the analytical work. Finally the author desires to thank the Atmospheric Corrosion Com-mittee and the Council of the British Non-ferrous Metals Research Associ-ation for permission to publish the Report in its present form. APPENDIX. Methods used in Analysis of Corrosion Products. The following notes do not attempt to give more than the briefest statement as to the methods ultimately relied upon for the estimation of the principal radicles. Copper.-The ordinary 6 6 iodide method.” Zinc.-(a) Relatively Large Amounts.-Titration with potassium ferrocyanide.(b) Relatively Small Amounts.-Precipitation as sulphide reaction with iodine and titration back with thiosulphate. SuZfhide [S].-Evolution as hydrogen sulphide absorption in cadmium acetate solution and titration with iodine and thiosulphate. It was found however that in the presence of’ excess of copper salt hydrogen sulphide is not evolved on simple treatment with acid. The difficulty was overcome by the addition of stannous chloride in the evolution flask thus maintaining copper in the cuprous condition. By this means a quantitative yield of hydrogen sulphide can be obtained even when the amount of sulphide in the original mixture is exceedingly small. Carbonate [COJ-Evolution as CO by means of phosphoric acid and gravimetric absorption after passing through acid potassium permanganate solution to remove any accompanying H,S.Sulpliate [SO,].-Precipitation as barium sulphate in the ordinary way followed (for very small amounts) by centrifuging. SUMMARY. The report is concerned with the behaviour of typical metals and alloys on exposure to the atmosphere ; it is divided into two parts dealing respectively with ‘‘ indoor ” and ‘ 6 open-air exposure tests the former including associated laboratory experiments. It is shown that conclusions which hold good for a given set of conditions do not necessarily apply when those conditions are changed ; this is exemplified by the influence of “impurities ” in copper. indoors under conditions such that only tarnishing has to be considered a given element exercises an effect which is either neglible or in simple proportion to the amount of element present.Exposed to the open air however the same element may exert an effect out of all proportion to its concentration. It would appear that protection against indoor tarnishing should be sought by methods other than modification of composition alone ; promising results have been obtained in connection with the formation of protective surface films. The main conclusions relating to the two parts of the work are summarised below. Part 1.-Indoor Exposure Tests and Laboratory Experiments. Copper. DeveZojment and Inhibition of Tarnishing Phenomena.-Znfluence of Surface Condition. -The intensity of attack as measured by weight-increment per unit area increases with the coarseness of the emery employed in abrasion.Inftzlence of Chemica Z CZeaning. -Treatment with solutions containing nitric acid is without appreciable influence upon the behaviour of the metal during subsequent exposure. Solutions containing chromic acid on the other hand, produce a considerable degree of immunity from tarnishing ; this immunity may be destroyed by treatment with ammonium chloride I 80 SECOND EXPERIMENTAL REPORT Injuence of Purity of MetaZ upon rate of tarnishing appears to be negligible (although modifications of composition may have pronounced effects upon the corrosion of copper under conditions of complete exposure to the open air Influence of SuZ$hur Content of Atmosphere. -The tarnishing of copper is initiated by atmospheric sulphur in such form as to respond to the alkaline lead acetate test ; the minimum concentration necessary for tarnishing however is extremely small (equivalent approximately to I volume of H,S in 600 million volumes of air).Above this value tarnishing commences and proceeds at a rate proportional to the “reactive ” sulphur content of the atmosphere. When once a tarnish tilm is formed increase in thickness follows accurately a parabolic relationship with time though the atmospheric sulphur content may fluctuate within extremely wide limits ; i.e. rate of tarnishing is determined by the conditions prevailing at the time of the first exposure. Com$osition of Tarnish FiZms and Mechanism of their Formation.-The sulphur (sulphide) content of tarnish films produced on copper in ordinary room air has not exceeded I 5 per cent.Films consisting initially of pure sulphide on exposure to ordinary room air increase in thickness and continue through the same sequence of interference colours which they would have done if the original treatment had continued but they do so mainly through absorption of oxygen. It is concluded that tarnish films consist essentially of isomorphous mixtures of oxide and sulphide and that in the case of an initially-clean surface under ordinary conditions of exposure the surface is activated by the impingement of sulphur atoms which enter into the formation of the first (mainly oxide) lattice ; their relative distribution in this primary lattice then determines the rate of diffusion and the whole subsequent course of tarnishing.There is reason to believe that if the first oxide lattice is completed before sulphur atJms can impinge (in the form and in the concentration in which they commonly occur in the atmosphere) their subsequent entry in any appreciable amount IS prevented and hence tarnishing is inhibited (see below). Formation of Protective Oxide FiZms a t the Ordinary Temperature.- When the content of atmospheric sulphur falls below the limit necessary for tarnishing, the formation of a thin film of oxide takes place the progress of which has been followed gravimetrically. This film is unaffected by subsequent increase in the atmospheric sulphur content (within the usual limits) i.e. it is resistant to tarnishing.boymation of Protective Oxide FiZms a t Temjeratures above NorwzaZ.-Whilst at the ordinary temperature there are two alternative types of film possible, i.e. the ‘‘ tarnish film ” and the “protective oxide film,” according to the relative purity of the almosphere at higher temperatures the latter type is developed to the exclusion of the former. A curve has been plotted which connects the rate of film formation with temperature. There appears however to be a critical thickness of film necessary for protection to be afforded against subsequent tarnishing. (For one hour’s heating this film is produced between 5 5” and 7 5” C.) The evidence leads to the suggestion that the minimum thickness of film necessary for protection is such that the unit lattice of the oxide is completed for the whole of the surface.Protective Eflects of Smoke FiZms.-It has been found that close contact with smoke from cellulose produces a considerable degree of immunity from tarnishing on subsequent exposure. -p.V.). Zinc. ReZation between Thickness of FiZm and Period of Ex#osure.-The linear relationship between weight-increment and time found in the earlier work has been confirmed for relatively long periods of exposure ; it holds true also in the early stages with the probable exception however of the first day. Profierties of Oxide FiZms DeveZoped @on Zinc a t the Ordinary Tem#era-t u r ~ -The linear relationship between weight-increment and time which charac-tenses the oxidation of zinc at the ordinary temperature leads to the conclusio W.H. J. VERNON I81 that the resulting film must have a granular structure through the interstices of which gaseous diffusion of the atmosphere takes place. Direct evidence of this has now been obtained by examination under the microscope. Moreover any colour effects should be due to diffraction and not to interference of light ; this is confirmed by the regular appearance of a blue tint without passage through the sequence of colours which is characteristic of interference. Brass. ReZation between Weight Increment and Time. -In an atmosphere of varying but mainly low relative humidity both 70/30 and 60/40 brass for a considerable time simulate the behaviour of copper and yield curves which approximate very closely to the parabola. After more extended exposure however they revert to the straight line which is characteristic of zinc.In a more humid atmosphere the resemblance to the behaviour of copper appears to disappear altogether. Efect of Atmospheric Exposure u#on Microstmccture (60/40 Brass).- In an atmosphere of low relative humidity containing traces of sulphur compounds characteristic of winter-time town air the attack is at first chiefly directed upon the alpha constituent which tarnishes in a similar manner to copper. Subse-quently (and more rapidly the greater the humidity) the beta constituent is attacked but the film is no longer '' continuous " and takes the form of isolated pits giving rise to an apparently duplex structure. It is suggested that this phenomenon is connected with the departure from the parabola to the straight line.Formation of Protective Oxide FiZms upon Brass.-The film of oxide produced at the ordinary temperature by exposure to a relatively pure atmosphere, while affording considerable protection to copper has relatively little effect upon the tarnishing of either 70/30 or 60/40 brass. Films produced at higher temperatures however exert a greater protective influence. Protection by means of LanoZine. - In the case of brass exposed to a relatively humid atmosphere a considerable degree of protection may be obtained by treat-ment with lanoline combined with heat-treatment at 100'. Behaviour of Brass Compared with that of Copper and Zinc in Several Types of Indoor Atmosphere.-Results from tests extending over approximately 3+ years may be summarked as follows the materials in each case being arranged in ascending order of the total corrosion which they have undergone.A. Atmosphere of varying but mainly low relative humidity. (I) Copper ( 2 ) 70/30 Brass (3) 60/40 Brass (4) Zinc. The brasses tend to approach nearer to zinc with increasing time. B. Atmosphere of varying but mainly high relative humidity. C . Domestic Kitchen. The order is nominally the same as in B but the attack on 70/30 brass is now very much greater than that on 60/40. Analysis of corrosion products from specimens exposed in the domestic kitchen show that a certain amount of the attack is due to organic fatty acids ; these appear to act preferentially upon zinc and give rise to their zinc salts in the corrosion products. (I) Copper ( 2 ) Zinc (3) 60/40 Brass (4) 70/30 Brass.( I ) Copper ( 2 ) Zinc f3) 60/40 Brass (4) 70/30 Brass. Aluminium. The formation of the primary film of oxide in the very early stages of exposure has been followed by careful gravimetric measurements which have led to the recognition of a weight-increment curve of an essentially different type from any of those previously obtained. Closely resembling a parabola at the outset it quickly '' flattens " toward the time axis indicating an extremely rapid retardation in the rate of attack. The type of curve suggests that adsorption may play an essential part in the mechanism of oxidation in the present instance. In 10 or 14 days the primary film has practically ceased to thicken 182 SECOND EXPERIMENTAL REPORT Assuming the film to consist of alumina either anhydrous or hydrated the weight-increment then corresponds to a thickness of the order of IOO A.U.Subsequent increases in weight are probably due to the formation of cracks or fissures in the primary film. Lead. The weight-increment curve is of the same type as that obtained for aluminium. The nearly flat portion of the curve is reached however in about 7 days and the weight-increment is then only a little more than half that which obtains for aluminium. As in the case of aluminium the film is completely invisible. It has been found that in the presence of traces of vapours from drying paint freshly-cleaned lead undergoes rapid oxidation and in a week or so becomes deep blue in colour. (Of all the metals examined the effect is peculiar to lead.) If however the metal has been exposed for a sufficient length of time to the uncontaminated atmosphere it is immune from such attack owing no doubt to the formation of a protective film of oxide.Iron. Atmosphere of ReZativeb Low Humidity. -In an ordinary room atmosphere of low relative humidity such as obtains under conditions of artificial heating the process of rusting is controlled entirely by suspended solid impurities in the atmosphere. The weight-increment curve is concave about the time axis i.e. the rate of attack falls off with increasing time. Rusting may be stopped entirely either by filtering the air or by screening the specimen behind a single thickness of muslin. During such time as the iron is exposed to an atmosphere screened from solid particles it develops a protective film in an analogous manner to other metals (i.e. copper and lead). On subsequent (normal) exposure this film definitely resists attack for a time but then breaks down at certain points, following which the localised attack proceeds at an accelerating rate. Atmosp/tere of ReZatively High Humidity (such as may obtain in an ordin-ary room which is not artificially heated).-In the presence of suspended solid particles as before and if the relative humidity is sufficiently high the rate of attack accelerates from the start owing to the deposition of particles and con-comitant precipitation of moisture. (Rusting may still be prevented by screen-ing behind muslin.) If iron already covered with (dry) rust is exposed to an atmosphere of the necessary relative humidity (actually not more than 70 per cent.) an extraordinary acceleration in the rate of attack at once takes place. These phenomena have been observed equally upon ingot iron of com-mercial quality highly purified iron and upon steel containing 0.5 per cent. C. Atmos#kere Saturated with Water Va#our. -Under conditions prevailing at the dew-point rusting of the iron takes place in the absence of suspended solid particles. The rate of attack accelerates from the start yielding a curve which is concave about the weight-axis. Part 11. Open-Air Exposure Tests. Intensive tests have been carried out upon specimens of copper 70/30 brass, 60/40 brass and zinc in various physical conditions. The rain-water has been collected from beneath these specimens and subjected to analysis at intervals. Information has thus been obtained concerning the weight of metal removed in relation to the period of exposure. For the whole of the time in the case of copper and for the major part of the time in the case of other materials a linear relationship was found. The order in which the materials were affected by this type of attack (i.e. “erosion,” as defined in the text) was as follows Zinc 70/30, 60/40 copper. In the case of the brasses there was excessive preponderance of zinc over copper in the soluble products W. H. J. VERNON From the weight of metal removed and the initial and final weights of the specimen a value was obtained for the L(surrosion,” represented by the residual products adhering to the specimens. Analyses of these products were carried out in all cases. The order of ‘( surrosion ” for the different materials was as follows : 60/40 70/30 copper zinc. Finally by the summing of erosion and surrosion respectively values were obtained representing the total corrosion of the speci-mens in which respect the materials came out in the following order ofcorrodi-biZity 60/40 70/30 zinc copper. Further tests representing over four years’ exposure have been conducted upon various grades of copper together with 70/30 brass 60/40 brass and zinc. The order of resistance to corrosion is given below the figures indicating the approximate total corrosion that of H.C. copper being taken as 100. Copper with 0.45 per cent. arsenic Copper with 2-5 per cent. nickel . Copper with 0.8 per cent. tin . Ordinary copper. . . . H.C. copper . . . . 70130 brass . . . . Zinc . . . . . . 60/40 brass . . . . . . . . In connection with the high resistance displayed by arsenical and nickel copper respectively it was noted that within the first few days of exposure these materials, among the copper specimens showed the most rapid loss of reflectivity