Corrosion of Aluminum in High Pressure Steam at Temperatures above and below 320°C By Tatsuo Kondo*, Cho Akutsu* and Masayuki Kawasaki* ‘Sues of tho corrsia of aluninam in high temperature stat, inching reseton kets and structural observations of corrosion products, were care out using specially desizned autoclaveseapable of obtaining high temperate and pressure ‘condition simaliamaouly. There was a romarkablo difference in corsaion behaviour hetween the temperature ranges above ‘nd below 350°C, due tothe transition of the oxide formed on the saface of the motl- Tn the present experiment, the coession process in the low temperature cange up 9 320°C had a parabolic time dependence represented by the Fenmale JW2=kp-t or 4W2—=C +kp-t fram which the activation energy, 4H, was caleulated tobe about 32 kealimsle, which approximately ageoes with the value abtsined fram the aqueous eoroion rest in the same temperatixe range. From the expetiment on the transition by hydzothrmal treatment of bulk oxide it was suggested thatthe ‘corrson of aleinom in the Bhi forming range is chractried by the inwar-hydration process of the oxi. ‘Athighoe temperatures, at which steam is superheated, the corcsion was negligibly small. However, a localized pent: tion stack occured which proceeded very rapidly, and was quit ravdomly distributed. ‘The relation betwomn penstation {induction time and steam pressre wan determined tobe =Ci-p~4, from which the resturs dependence of the corosin rate wos derived k=Ca-pt, ‘This elation ageoed with that forthe low temperature range repted By other authors "The very thin oxide film vas ramved from tbe mstal sucfaco and. analyzed with an eloctan wierseope twas deter ‘inet be snporphous oxide with dispersed mack end grown whiskers of a-ALOs- ‘A mimber of hydrogen bubbles were observoi:io be agulameratod at the grain boundaries of the metal by the rmatallagaphic teciaue. The mechani of the penetration attack is thoroughly discussed with emphasis onthe whisker for ‘mation attributed to the sts: by the hydrozen agglomeration inthe underlying metal (Received April 18, 198) I, Introduction several countries in connection with its nuclear Aqueous coorrosion of aluminum at elevated tem- application, and several explanations have been peratures has been investigated for a long time in reported as to the corrosion mechanism,O~® ~* Metallurgical Laboratory, Japan Atomic Energy Research (2) K.M. Carlen J. Blestrochem. Soc. 108 (2957), 147. Istituto, Tokai Mura, Japa (4) TL Dilla and V. H. Teoutner: HW-21849, Sot, (1957) (5) D. Altenpohl :Z, Metall, 48 (1957), 908. (1) Jc. Daley and W. - Ruther : ANL-5658, Appr, (1957). (6) M. Kewanaki, S. Nomura and T. Kondo: J. Japan Inet. (2) RA. Us Huddle :A/Gonf. P/411, (1959) 403. ‘Metal, 25 (1961, 76,20. ‘Trans. JIM, 1962 Vol. 3Tatsuo Kondo, Cho Akuisu and Mesayuli Kaseasaki au The present study was intended as an extension of the study on aqueous corrosion of aluminum at elevated temperatures. The corrosion study in steam hhas several advantages, that is, the influences of various ions dissolved in the water, dissolution of aluminum or its oxide, electrochemical polarization and others, which must be taken into account when solutions are used, can be ignored. Summerizing recent knowledge?~©®) of the cor- rosion of aluminum in steam; the corrosion behaviour jin saturated or slightly superheated steam up to 370°C are apparently the same as in the case of water in which the surface film is so that a large weight gain is observed. ‘The dependence of the corrosion rate on the steam pressure is very large, as derived experimentally by Wilkins and Wanklyn, In the higher temperature range, 400° to 600°C, a kind of localized penetration attack occurs, usually starting at the edge of the plate with quite a random distribution, then spreading until the entire specimen is destroyed. This phenomenon was explained by Draley and his coworkers as being based on the absorbed hydrogen in the metal which was quantita- tively analyzed for exposed specimens. From the phase diagram of the system Al,Os-H,0 it is expected that there will be a considerable change in the corrosion behavior of aluminum due to the transition of the oxide in the temperature range from 300° to 400°C. ‘Therefore, it is believed that there still remains a number of questions about the influence of the surface oxide during the corrosion process. IL, Experimental Method 1. Apparatus In earlier experiments much error was caused for the kinetic data obtained as well for the structure of oxides because of the use of slowly heated autoclaves, more specifically, by the heating process in which the steam must once pass through the saturated state. Thus the specimens inevitably take on a pre- film prior to the film formation in the experimental ‘econdition, Therefore, it was considered most important to provide an improved procedure capable of simulta- neously supplying steam with well controlled tem- peratures and pressures to the preheated specimens, ‘The final procedure adopted consisted of two auto- claves, with a connecting spiral pipe heated to the test temperature. One of the autoclaves, with a capacity of 2 liters, was used as a test section and the other, with a capacity of 5 liters, was used as a steam generator, as shown in Fig. 1. ‘The specimens were hung in the 2 liter autoclave (7) H. Corio, al: AlGanf, 5 (1950), 128. (8) C£ Groot and RE Wien: HW-A1797, May, (1956). (9) NJ.M, Witkins and JeN, Wanklyn J. int Metals, 88 (2950-60), 134. (0) J.E.Dmaley, W.E. Ruther and S. Greenberg: ANL-6207, ‘April (1961 (11) Ein and F. F, Osborn : J. Geo. 89 (1951), 38 with aluminum hooks for exposure to the homo- geneous heat radiated from the wall; the hooks were designed to divert condensed drops of water away from the specimens. Fcc cos Ce Fig. 1 Schematic drawing of tet apparatus 2. Procedure Before heating, the 2 liters test autoclave was filled and evacuated with pure argon gas several times, as preliminary test runs showed that the air in the 2 liters test autoclave had an adverse effect. When the 2 liters test autoclave was brought to the test temperature two valves were opend to let heated steam into the test autoclave. The desired condition was usually reached within 3 minutes or less after the valves were opened. After the steam was released from the test auto- clave, cold argon gas was fed into the test autoclave to cool the specimens, after which the autoclave lid was removed. 3. Specimens Table 1 shows the chemical composition of the 99 percent Al and the 99,99 percent Al used. The corrosion behaviours of these two types of Al in water at high temperature are well known, having been ‘Table 1 Chemical composition of specimens. Impucities | | | mer | iu | a | se | e| am] me | pra | | 99% AL | 99450] 018 | 0.64 | 002 | om) we | 99.99 5 al | 99992 | 0005) 0001] 0002) te. | te | te reported previously by two of the present authors® the 99 percent Al specimen acquired a homo- geneous protective oxide before its rate-transition time while the 99,99 percent Al specimen corroded intergranularly. The specimens were cut from rolled Al sheets 1000 mm wide by 1 mm thick and vacuum-annealed for 1 hour at 400°C, All specimens were chemically polished for 3 minutes at 85°C in a solution of nitric acid and phosphoric acid, By removing surface contaminants or irregularities and obtaining a bright surface the kinetic measurement can be made more accurately. After being thoroughly washed with he effect of atmosferic gees was found to be appreciable Jn soveral preliminary tests.ne, Corrosion oF Aluminum in High Presure Steam at Temperatures above and below 320°C distilled water the specimens were kept in « desiccator for at least seven days in order to stabilize‘? the natural oxide film. ‘Samples of the oxide powder were first wrapped in fine mesh stainless steel and then exposed to steam for the hydrothermal treatment. 4. Examination after exposure ‘The gravimetric method is aceurate enough for the ‘measurement of the corrosion rate of steam corrosion. ‘The specimens were dried in a desiccator for 30 minutes during which time the desiccator was evacuated by a vacuum pump, after which the speci- mons were weighed, After weighing the results were calculated by taking the mean value of an area of at least 240 em? ; the accuracy was about 1.3 yg/em*. Weighed specimens were turned for electron micro- scopy or metallography and other weighed samples were dissolved in a solution of iodine-methanol® to separate the corrosion produet for X-ray diffraction analysis. A chromium shadowed carbon replica was used for electron microscopy of the surface profile and the thin oxide film on the metal surface was removed with a solution of mercuric chloride for direct observation. In addition to observation of the film, the micro- electron diffraction method was used to determine the erystal structures. Finally the specimens were examined with an ‘optical microscope for hydrogen agglomeration using ‘the taper-section method after electropolishing in a solution of ethanol and perehloric acid. After anodiz~ ing in a solution of ethanol, water and flouric acid the specimens were by examined with polarized light for grain boundaries, IIL. Experimental Results and Discussion 1, Corrosion in saturated steam at temperatures up to 320°C (1) Kinetic observation 99 percent Al was used chiefly for a kinetic study with 99,99 percent Al being used for comparison. ‘The apparent manner of corrosion in this temperature range closely resembled that of aqueous corrosion in the formation of heavy oxide film and the results obtained by the gravimetric method were obviously parabolic in time dependence, as shown in Fig. 2. ‘According to Cabrera-Mott’s theory®) on the parabolic oxidation Jaw for thick oxide film the result obtained above corresponds to the diffusion-controlled parabolic rate process. Then the parabolic constant ‘was obtained for the equation AW? ist a (12) M, S. Hunter and P. Fowle: J, Elentrochem, Sos, 108 (1956), 482. (03) M.J. Per and D.S. Keir: J. Hletrochem, Soe, 192 (1955), m0, (04) CE. His end W. Evans : AECL-1286, May (1961). (5) N. Cabrera and NF. Mott: Rep. Prog. Physy 12 (1948.4), 183. where JW, kp and t denote the weight change, para- bolic constant and exposure time respectively ; the parabolic constant ky has in general the temperature dependence as given by the equation Fig. 2. Corrosion of 9 Al in saturated Moa wp to 820°C, y= A-exp —JH|RT 2 where 4 is the constant which refers to the frequency factor and 4H, R, T represent activation energy, gas constant and absolute temperature respectively. Fig. 3 is a logarithmic plot of k, for 99 percent Al against reciprocal temperature. From this plot the activation energy, 4H, is calculated to be about 32 keal/mole using equation (2) e022 , J N | ig.3. Effet of temperature on pasball constants of conion. The calculated value 4H~32 keal[mole is found to be very close to the value caleulated from the result of the break-away-induction time, assuming the induction process to he parabolic against time, for corrosion of an aluminum alloy in water and steam in the same temperature range; they are about 31 and 35 Keal/mol respectively. Tt was also observed that there was an evidence of an extremely large dependence of the corrosion rate on steam pressure, which can not be expected from Wagner's mechanism‘ for the pressure dependence of oxidation. Wilkins and Wanklyn'® observed that there existed a relation between the amount of cor rosion and pressure represented by the formula W7= k-p* at 325°C, Fig. 4 shows corrosion data for superheated steam at 260°C, which also shows an (16) G. Wagner : Z, Phys, Chem, B 22 (1983), 212.Tatu Kondo, Cho Akuen and Masayuki Kowosoki us extremely diminished value of k, compared to the value in Fig. 2. Considering the facts described above, itis rather 4iffioult to understand the corrosion mechanism by HOF aweceit SY kas oo °o 20 3040-80 60 e080 WO 4 Cons of 80% Alm supe sean 0 te the simple diffusion of aluminum ions through the oxide layer. Thus another process controlled by migration of protons or some other hydrogen-bearing carriers such as OH or H,0 may be thought to contribute mainly to the inward-hydration and recrystallization of the oxide, which will tend to reduce the protective nature of the oxide film. In general the diffusion controlled oxidation proceeds with increasing thickness of the protective film; whereas in the present case it has a tendency to decrease the barrier thickness, although the distance of the diffusion path is increased in the outer hydrated portion, This is considered to cause an increase of the total amount of corrosion or of the frequency factor, 4, in equation (2). ‘As generally known aluminum has a double-layered film in the thermally formed oxide? In the present experiment, as will be described later, « distinet double layer was observed. ‘Then the following explanation is attempted based on the double layer model. Dillon and Troutner pointed ‘out that the outer layer, which is hydrated and inner layer just as in dry corrosion, and it should be noted that the growth of the inner layer is restricted by the inward hydration process, in other words, as the direction of both processes are opposite to each other the destruction of inward hydration, which is controlled by the migration of some species, supposed to be some hydrogen-bearing carriers through the outer layer, will accelerate the oxidation process by decreasing the thickness of the inner protective layer. From this point of view the characteristic feature of aluminum corrosion in the boehmite-form- ing range can be understood. (2) Oxide film By X-ray diffraction all the films produced in steam up to 320°C were determined to be bochmite, a- AIOOH, whether or not the steam was saturated. Tn most ctses they had distinct layer structures which are shown in Fig. 5; the outer white layer was crystalline boehmite and the inner grey layer was amorphous or slightly hydrated alumina, presumably p-alumina as indicated by the single broad peak in Fig. 5. (a) outer film, Boehm a-AIOOH () inner fim, alumina? ALOs-ZH,0 Fig. 8 Xray diffraction of the inner and the oater oxide ‘focmedoring corrosion in saturated star at 200°C, On the other hand, the electron micrographs as shown in Photo. 1 reprsent the crystal growth of bochmite in saturated steam with increasing exposure 1e. is also noted that the grain growth of bochmite 1b) 96 hous - (6) 212 hour Photo. 1. Cepstal growth of bnchmste (-AIOOH) film formed on 999% Al saturated stem at 260°C, crystallized, is porous but occupies almost alll the thickness of the oxide, so that it would be natural to ‘think that the diffusion process in the outer thick layer more probably controls the corrosion process ‘than that in the inner layer, which is anhydrous and amorphous, although the latter has a higher specific resistance, However, the oxidation reaction itself is, usually restricted by the activation process through the (07) M.S. Hunter end P. Lowe: J. Blestrchem. Soc, 108 (1981), 139. is considerably rapid in the case of 99.99 percent aluminum which was destroyed rapidly by inter- granular attack as shown in Table 2. 2. Corrosion in superheated steam in higher temperature ranges (1) Kinetic consideration of the penetration attack As the result of several experimental series per- formed, in which the steam pressure was varied froma4 15 to 100 kgjem* at 350° and 400°C, the following, conclusions were drawn. ‘The weight changes were negligibly small and the time before the occurrence of the penetration attack ‘Table 2 Corrosion of 99 Se and 99.99 % Al by short time ‘expomie in strated steam at 260°C ‘was so short that it was difficult to determine the corrosion rates. As reported by other investigators 0, the penetration attack observed had also a peculiar nature, They started at a few spots, dis- ‘tributed randomly on the surface, and spread ina cireular shape to destroy the whole specimen, a3 shown in Photo, 2. oe os 9 Photo. 2 Macro view of penetration attack 99% Al exposed or 45st siperiatal steam at 390°C, The reason they often originated at the edges of the specimen and extended only parallel to the sur- face plane was understood after several experimental trials, to be attributable to the rolling-annealing textures of aluminum plate which were also thought to be related with the oriented diffusion of hydrogen, as will be deseribed later. It is to be noted that the induction period to the initiation of attack depends largely on steam pressure rather than on temperature. From the results given in Fig. 6 the relation (3) is derived. erean Pestone ua Fig, 6 Effect of steam prosure on pnetraion induction tine of 99 Se A at 400°C w= CP @) where ti, G, and P represent the penetration indue- tion time, constant and steam pressure respectively. Corrsion of Aluminum in High Pressure Stoam at Temperstares above and Below 320°C ‘The equation (3) ean be converted to k= CPt 4 where k is the rate constant, when assuming. (i) The penetration attack is initiated by a definite amount of corrosion reaction, in other words, by the critical amounts of the corrosion products. (Gi) The dependence of the corrosion rate law on time is not altered by any pressure as in the case of lower ‘temperatures as shown in Fig. 4. ‘This relation is quite similar to the experimental equation established by Wanklyn and Wilkins‘® for the case of lower temperature (at 325°C). Tt is noteworthy that there exists the same pressure dependence of the corrosion rate in both higher and lower temperature ranges although they have quite different surface films. ‘These facts also suggest that there may be a controlling factor largly related to the steam state, The diffusing species, some hydroge bearing carriers, might have an important influen at low temperatures, as already described, when assuming the inward hydration of the oxide, and at higher temperatures, as will be written later, a large amount of hydrogen is absorbed into the metal. ‘An additional result, considered as important, was obtained by the comparison of two kinds of specimens exposed to superheated steam at 400°C; which indicated that the 99.99 percent specimen twice as long as the 99 percent specimen and was not attacked intergranularly. This is an entire reversal of the results obtained for the low temperature range. Therefore, a conclusion was arrived at that pure aluminum is in inferior in aqueous or steam corrosion as is its susceptibility to intergranuiar attack, which fare due to the nature of the oxide which is less resistant to hydration in the bochmite-forming range. Since the grain boundaries of high purity aluminum are supposed to have a larger boundary energy, the surface energy, which maintains the stability of the amorphous barrier layer of oxide adjacent to the metal, will be (decreased: by the interaction with boundary energy, resulting in preferential hydration at the boundary area ; thus it is natural for the cor- rosion to proceed preferentially at the grain bounda- ries in the boehmite-forming range. ‘The considerations stated above are supported by the experimental results in section II-3. (2) Corrosion product oxide ‘The corrosion products produced by penetration attack were all found to be a-Al,O by X-ray diffrac tion with the one exception at 350°C which showed slightly hydrated amorphous alumina (referred to as p-alumina). ‘This type of alumina was observed in the inner film of the specimen exposed to saturated steam at 300°C, as already mentioned. It should be noted that very little corrosion occurred before the penet= ration attack, but the original surface brightness remained and no interfering color appeared. Making transmiscon electron microscopy of the surface film was possible. Photo, 8 (a), (b, (c) and (d) indicate thatTatsxo Kondo, Cho Abutse ond Masayuki Kawasaki us amount of oxidation this large absorption of hydro- gen was not expected, so that almost all of the hydrogen produced by the reaction must have been absorbed, or as in the ease of post-transition corrosion the thin film consisted of fine nuclei of a-AlO; dispersed in the amorphous matrix. In some cases, there were whiskers of a-AlO; observed on the surface of the thin film, as shown in Photo. 3 (c) Photo. Transmission electron microgsaphs and diffraction pattems of sinftce fle Forint in team at 400°C, (a) lei of o-ALOs in smorphoss matrix, 90.90% Al in 35 kglem? steam, 3 ra) (b) micro diffraction of (shows pattern of «-ALOs. {c) whisker growth on the surface of oxide fl. (99% Alin 18 kglent tam, 3 hes) (@) mera diffraction of (e) shows patter of «-Al,Os including lager snge eta If the whiskers were formed probably by the stresses. in the underlying metal as generally believed, the stresses should be attributed not to the volume change in the oxide-metal interface, but to the hydrogen-blister formation in the metal, because extremely thin films do not have very great strength and the mechanical strength is further reduced at higher temperatures. ‘The growth of bochmite whiskers during aqueous corrosion has also been reported by other investiga torsW®. ‘The formation of a-Al,03 at 350°C deviated alittle from the equilibrium phase diagram of the ALO;-H,0% as shown in Fig. 7, (3) The agglomeration of hydrogen at grain boundaries After exposure to superheated steam at 400°C the ‘ross section of the specimens was examined with an ‘optical microscope. As seen in Photo. 4(a) and and Photo. 5 (a) and (b), there appeared at the grain houndaries a mumaber of pores, which increased both jn size and number with increasing time and pres- sure, especially just before the initiation of the penetration attack, Considering the extremely small (19) LK. Hct and W-E. Ruther: ANI. 6230, April (196), of zirconium it might be attributed to the therma decomposition of H,0° z lz z Pi i ‘Ehrcaatuae Get Fig. 7 Eiulbsium dng fos the ALOsHO systom (Ervin, Osbara By the observation of the eross sections there were found several blisters at the grain boundaries near the interface, which were thought to cause penetra- yn attacks as shown in Photo. 6 (a) and (b). Wh the penetration attack began the absosption and agglomeration increased considerably as shown in ‘and the penetration reaction was accelerated by the destructive action of hydrogen. The size of “(@D) FHL Kronz: Private commanication16 Ceorrecion of Aluminum in High Presure Stam at Temperasares above and below 320°C the pores was made larger by annealing for 1 hour at 500°C, According to the analysis of post-exposure speci- mens by Draley and others, the hydrogen was temperatures as mentioned above. However, chained micro pits were observed along the subgrain boundary network of the underlying metal, as seen in the electron micrograph in Photo. 8. (b) poled ght Photo. 4 Microphotographs ofthe eros secton of 9 Al expos far 20 rs to supeshented seam ia ° e ‘ ° (a) ondinery light at 400°C, 20 kee. en zi ay ° eo —.. ~(@)oslineyy light (©) poled ight Photo. 5 Miexsho:ngraphs of the ors section of 90.99 5 Al exposed fo: 20k. to aupecheated steam at 400°C, 20g? (a ondinary Light hhydiogen. (170) absorbed in the initial stages of exposure. If the penetration attack was induced by hydrogen as above, the induction time would be taken to be the same 8s the time required for the hydrogen absorbed to agglomerate at the boundaries. However it is not clear whether or not the hydrogen was absorbed continuously after the film formation, as the amount ble was too small to measure. The inter- granular attack could not be observed at lower () polarized light Photo. 6 Blister Forsation at the grain boundary beneath the surace due to the agelomeration of It is considered that the solid solubility of hydrogen in the metal will increase with the defect portions such as the substructure networks or grain boundaries will serve as the inlet paths for atomic hydrogen. emperature and that 3, Hydrothermal treatment of a-Al,03 formed during the penetration attack It has been mentioned above that there was @Tatsno Kendo, Cho Ahutsu and Mosayuki Kaveaseki ‘characteristic feature in the corrosion of aluminum in water or steam in the bochmite-forming range. Thus the following test was attempted by exposing anhydrous a-Al,03 to steam at 300°C in order to 7 Intergranolar penetration of hydrogen atthe coxidental interfaes, (170) Photo, 8 Trenton lectamiengaph of surface film fozmenlon 99% Al by exposute in sper heated steam at 400°C, I8kgicn®. Micrpits ace heecved along the sbatractar-net work ‘examine the hydration process. ‘The test used two Kinds of powdered oxide formed by the penetration, attack at 500°C and SOkg/em? on 99.99 percent aluminum, Table 3. Intention of Xray diffraction specta from bulk «-ALOs hydrothermaly rested in stam at 300°C, 90 kom uz In this process the oxide formed from the 99.99 percent specimen hydrated much faster than the 99 percent specimen. As far as the experiments performed are concerned, the hydration process is Ah Fig. 8 X-ray diffraction of oxide powder hydrotherally treated in steam at 300°C, 90 kom? (98.99% Al) (a) e-ALDs trated for 3 hours isp the most orninant. ()aALO) treated for 20 hous, bochnite isthe ost dominant. VaaeAOy, de-dinspore, be Doshit. found to be largely influenced by the impurities in the oxide. Here it should be noted that the experi- ment using the a-AlO; powder may not represent perfectly the kind of hydration process in actual corrosion. However, if the hydration reaction or crystallization of the hydration product is rapid compared to the diffusion rate of the related species the latter would be the controlling step, so that the experiment would be an estimation. ‘The presence of impurities in bochmite has been reported by other investigators to influence the diffusion of deutron in the erystal?, IV. Conclusions According to the kinetic results in the bochmite- forming range, the corrosion process is thought to be characterized by the inward hydration, which in turn is controlled by the diffusion process of some Tatami | te Diigoss A100H Bodhnie a-AIOOH) | ayant) 3.99 a0) 2.131 (121) 1.850 (020) a | 3 8 x a oO wen | ao 2 a wok wae |_2 3 5 ~ | a | 1% amen | 3 | 2 Ge Rot: sale mm (3) = ‘At 300°C and 90 kglem?, the a-Al,03 could easily be hydrated to monohydrates, whiel from Table 3 and Fig. 8 to proceed in the following steps @-Al,03- ++ -8-AIOOH (diaspore) ‘ean be seen ++ s@-A100H (boehmite) hydrogen-bearing carriers, such as OH” or H,0, in the Dochmite erystals. From the above point of view the corrosion behavious of aluminum and its alloys in water or steam in the boehmite-forming temperature range can be successfully explained in combination (Gi) S. Mori, JE. Draley and R. B, Bornstoin : ANI-S889 Add.us Influence of $75°C Embrisiement of High Chromium Perritie Stes on their Corvasion Resistances with the result of hydrothermal treatment of bulk oxide. [At higher temperatures the corrosion produet was identified with a-Al,05 with both electron and X-ray diffraction methods, and the corrosion amount was negligibly small. However, the relation between corrosion rate and steam pressure was derived from the experimental data on the penetration attack, and it was almost equal to the data obtained for the low temperature range. ‘The penetration attack was attributable to the hydrogen agglomeration and bubble formation at the grain boundaries, The bubbles increased both in size and number with time and steam pressure, and finally they induced the penetration attack accom- panied by cracking and penetration of high pressure steam, Acknowledgement ‘The authors wish to acknowledge the meny in- structive discussions by Dr. 8. Nomura, also the many useful suggestions offered by Professor Dr. Saburo Shimodaira and Dr. K. Sakamoto. ‘The authors are also deeply indebted to Mr. K. Shiozawa who took all of the electron micrographs.