Thermally-Formed Oxide on Aluminum and Magnesium

Home , Diaspore, Magnesium oxide, Thermal oxidation

Materials Transactions, Vol. 47, No. 5 (2006) pp. 1347 to 1353 #2006 The Japan Institute of Metals

Teng-Shih Shih*1 and Zin-Bou Liu*2

Department of Mechanical Engineering, National Central University, Chung-Li, Taiwan 32001, R. O. China

Pure aluminum and magnesium cube samples were heated at a specific temperature for a given time after being polished by abrasive papers. The amorphous layer that formed on the sample surface was transformed into a crystallized oxide film during the heating and holding period. We discuss how this thermally-formed oxide film progressively developed on the surface of the cubic samples. Gibbsite (Al(OH)3) forms on pure aluminum during the initial stage of heating, after which it is then transformed to complex oxides, diaspore and -alumina. After an extended holding time at 883 K, the thermally-formed oxide film will be comprised of gibbsite, diaspore, -Al2O3 and -alumina. This thermally-formed oxide film is compact and contains evenly-distributed microchannels. With pure magnesium, the transformation of periclase from brucite is associated with the formation of microcracks. In this study we use TGA (thermo-gravimetric analysis) to describe the progressive development of complex oxides and periclase films on pure Al and Mg respectively. [doi:10.2320/matertrans.47.1347]

(Received November 7, 2005; Accepted March 28, 2006; Published May 15, 2006)

Keywords: thermally-formed oxide, Al2O3, periclase

1. Introduction characteristics of transformed Al2O3 and MgO films. Oxide growth was discussed based on thermogravimetric analysis, Aluminum and magnesium are active metals when they X-ray powder diffractomer analyses and SEM observations. come into contact with air or oxygen. An oxide film can be This thermally-formed oxide film was eventually entrap- readily deposited on their surfaces during heating. Several ped in the melt during melting and could also be introduced researchers have studied the thermal oxidation of aluminum during filling process. It also could possibly be entrapped in and found that an air-formed oxide film develops when the matrix during deforming the hot workpiece. A better aluminum is heated in air or oxygen to a temperature above basic understanding of the effect of an entrapped thermally- 723 K. An amorphous oxide film is formed initially and then formed oxide film on the mechanical properties of aluminum -alumina crystals develop at the metal-oxide interface.1–4) and magnesium alloys is of great industrial interest. As the oxidation temperature increases, the crystalline - alumina is gradually transformed from the amorphous 2. Experimental Procedure aluminum oxide by the inward diffusion of oxygen. Magnesium alloys can become a fire hazard and are subject High purity aluminum (99.999 mass%) and the pure to surface degradation when they are machined or melted. As commercial magnesium (Mg: 99.96%; Si: 255 ppm and a result, one must take different measures to avoid this Mn: 108 ppm) were used in this study. The material was surface degradation and to ensure quality products. Shih and machined into 10 10 10 mm3 cubes. The cubic samples coworkers have studied the combustion of an AZ61A alloy were polished on one surface by abrasive papers (from p400 5) given different gases. Decreasing the ratio of CO2/Ar to p2000). The pure aluminum samples might be further decreases the amount of heat produced by oxidation reaction. polished by a fine rayon cloth and fine silicon carbide powder Czerwinski used thermogravimetric analysis to observe the (0.3 mm), prior to carrying out the different surface prepara- oxidation and evaporation behavior of AZ91D magnesium tions. alloys. He found that non-protective oxidation occurred at The pure aluminum and the pure magnesium cubes were 683 K. He also schematically illustrated the development of heated in a muffle furnace at 883 K and 700 K respectively, MgO morphologies from the early compact uniform layer to for two time spans: 3:6 103 s and 9 104 s. Thermal oxide the final loosely scaled structure.6) films formed on all surfaces of the sample cubes. The oxide Nordlien and coworkers carried out a seminal study of how film on the polished surface was observed by an optical corrosion-resistant oxides formed on magnesium alloys. In microscope and a scanning electron microscope. Auger air a film formed immediately on a fresh surface, after being electron spectroscopy was used to analyze the composition of polished by abrasive paper. This film is thin and dense with the surface oxide. The relation between the composition and an amorphous structure. Furthermore, in humid air, a the sputtering time was observed and recorded using a 3 keV hydrated layer will form as a result of water ingression and 1 mAmm2 argon ion beam. Thin pieces of pure through the initial layer and metal oxidation.7) aluminum were prepared and heated in a muffle furnace for In this study we used extra-pure aluminum (99.999 different time periods. An X-ray powder diffractometer mass%) and pure commercial magnesium to prepare samples equipped with glancing incident angle (G.I.A.) was utilized with a thermally-formed oxide film on their polished surface. to test the constituents of thin oxide films on the heated Micrographic observations were made to examine the samples. Cu target was used in the test and the wave length is 0.154056 nm. The theta angle is 1 degree. *1Corresponding author, E-Mail: [email protected] For the thermogravimetric analysis a Perkin Elmer (TGA- *2Graduate Student, National Central University, Taiwan, 32001, R. O. 7) apparatus was used to record the weight variation of heated China samples in an air atmosphere, at a flow rate of 50 cm3/min. 1348 T.-S. Shih and Z.-B. Liu

Microchannel

Al substrate Al O 2 3 Resin film

(a) Fig. 2 The sectional view of a cube sample (99.999 mass%Al), after being polished by abrasive papers, heated at 883 K for 9 105 s; note the bumpy interface.

1000 0.02 TGA 5N Al air : 50 cm3/min

0 800

A 885 K

-0.02 832 K 600

B Temperature, T / K Weight change, M / mg -0.04 I II III IV 400 (b) D C

Fig. 1 The surface morphologies of different heated samples (99.999 -0.06 mass%Al), after (a) being polished by abrasive papers, (b) followed by a 0 5000 10000 15000 20000 25000 fine rayon cloth and SC powder; the cube samples were heated at 883 K for Time, t / s 9 105 s. Fig. 3 TGA analysis showing the weight change of a sample (99.999 mass%Al) at different heating times; heating history is also included; note The heating rate was 8.3 K/min for pure aluminum and 5 K/ the categories of four stages. min for pure magnesium. For the latter material, the temperature was held at 423 K for 1800 s and then heated to 700 K and held for 2:16 104 s. For the former material, 0.1–0.2 mm microchannel penetrates through the film. the specimen was heated to 883 K and held for 2:16 104 s. Thermogravimetric analyses show the weight change of pure aluminum during heating and holding at 883 K up to 4 3. Experimental Results 2:16 10 s; see Fig. 3. Gibbsite Al(OH)3 film would form first on the polished surface and increases slightly in sample’s 3.1 Thermally-formed alumina weight in a short heating time. This sample then loses weight A pure aluminum cube was heated at 883 K for 9 105 s continuously due to gibbsite dehydration. The film may after being polished by abrasive papers. The heated cube experience a phase change at about 830 K, as indicated by a showed some striations and polish marks on its surface. Some significant weight gain at 832 K; see Fig. 3. When the heating orderly distributed white spots were also visible as shown in time (or holding) is extended, dehydration of the film persists Fig. 1(a). If the sample was further polished by a fine rayon and the weight change gradually becomes stable in stage III; cloth and water, then heated, the striations became less see Fig. 3. In stage IV the tested sample rapidly increases in apparent but the white spots grew in size, as shown in weight. The thermally-formed oxide film increases in thick- Fig. 1(b). ness. The white spots shown in the optical micrograph have Figure 2 illustrates a sectional view of the sample cube been confirmed to initiate at the triple junctions of grains and with the thermally-formed oxide film. The sample was first inclusion particles; in each spot there resides one micro- polished by abrasive papers then heated to 883 K for channel. 9 104 s. The interface between the substrate and the oxide The measured peaks in the XPS test show that the film is bumpy and the film thickness is about 0.8–1.2 mm.A thermally-formed oxide film is comprised of Al and O. The Thermally-Formed Oxide on Aluminum and Magnesium 1349

5N Al 70 O-1h Al-1h

40 Atomic concentration ( at% )

30 (a) 0 100 200 300 400 500 Sputtering time, t / s (a)

70 5N Al O-25h Al-25h

Micro-crack

40 Atomic concentration ( at% ) (b)

Fig. 5 The sectional SEM observations for a sample (99.956 mass%Mg), 30 after being polished by abrasive papers, heated at 700 K for 3:6 103 s; 0 100 200 300 magniﬁcations of (a) 20,000X, showing a microchannel near an inclusion Sputtering time, t / s particle,(b) 35,000X, showing a micro-crack. (b)

Fig. 4 Relation of the tested Al 2p and O 1s versus the sputtering time; the stoichiometric Al O concentration, compared with pure aluminum heated at 883 K for (a) 3:6 103 s; (b) 9 105 s. 2 3 Fig. 3.

3.2 Thermally-formed magnesium oxide ratio of the signal intensities between the O 1s and the Al 2p Salas describes the structure of MgO as being so loose that spectra is close to that of sapphire.5) Figures 4(a) and (b) it can not act as a barrier to prevent the further oxidation of display the atomic concentration versus the oxide film magnesium.9) A sectional view of this thermally-formed sputtering time for samples heated at 883 K for 3:6 103 magnesium oxide on a pure magnesium sample heated at and 9 105 s, respectively. The surface layer is obviously far 700 K for 3:6 103 s after being polished by abrasive papers rich in oxygen than the stoichiometric Al2O3 concentration, (from P400 to P2000) is shown in Fig. 5. The oxide film is probably due to the development of O–H bonds on the about 0.5–0.8 mm in thickness and is indeed seen to be loose; surface layer of the oxide film. For a sample heated for a short including microcracks (Figs. 5(a) and (b)). The partially- time, 3:6 103 s, the measured oxygen content in the film filled microcrack can also be seen from the sectional view of 4 was less than the stoichiometric Al2O3. Referring to Fig. 3, a the 9 10 s heated sample. The dehydroxylation of brucite sample heated for 3:6 103 s was approaching the end of is accompanied by a 55% reduction in the matrix volume stage I, where the thermally-formed oxide film would be which leads to crack and void formation.10) Microcracks on comprised of complex oxides Al(OH)3, AlO(OH) and Al2O3 the thermally-formed magnesium oxide film can be readily oxide. Santos noted that monohydroxide boehmite is the only seen both after short and long heating times in this study. material to produce -Al2O3 by dehydration. Boehmite can The relation of atomic concentration versus sputtering be produced from aluminum trihydroxide by hydrothermal time is shown in Figs. 6(a) and (b) for samples held for 3:6 synthesis at 453–523 K.8) Consequently, for samples held at 103 and 9 104 s, respectively. The measured composition 883 K for 3:6 103 s the oxygen concentration is less than profiles for both samples are similar. The outer layer, from 1350 T.-S. Shih and Z.-B. Liu

0.06 800 pure Mg TGA pure Mg Mg-1h air : 50 cm3/min 100 O-1h 0.04 700

80 0.02 600

60 0 500 Temperature, T / K Weight change, M / mg 40 -0.02 400 I II III IV

-0.04 300 Atomic concentration ( at% ) 20 0 10000 20000 30000 Time, t / s

0 Fig. 7 TGA analysis showing the weight change of the heated sample 0 400 800 1200 1600 2000 (99.956 mass%Mg) at diﬀerent heating times; heating history is also Sputtering time, t / s included; note the categories of four stages. (a)

pure Mg period of time; see stage III in Fig. 7. Note the inclusion Mg-25h 100 particle located near a partially-filled microcrack in Fig. 5(a). O-25h The formation of the microcrack may be related to the entrapment of an inclusion particle. The microcrack is 80 irregular and not as open as the microchannel in the oxide film on the aluminum. The growth mechanism for Al2O3 on 60 aluminum, and perilcase on magnesium, is different based on the measured atomic concentration and the microscopic observations of oxide film. 40 3.3 Discussion

Atomic concentration ( at% ) 3.3.1 Gibbsite, Diaspore, -Al O and -Al O 20 2 3 2 3 Jeurgens and coworkers calculated the thermodynamic stability of amorphous oxide films on metals. For sufficiently 0 thin oxide films on a metal substrate, the amorphous state can 12) 0 1000 2000 3000 be preferred over the crystalline state. The amorphous film Sputtering time, t / s becomes unstable during heating due to its high bulk energy. (b) The oxidation of Al up to a temperature of 823 K and an 3 oxygen pressure of PO2 ¼ 1:33 10 Pa led to formation of Fig. 6 Relation of the tested Mg 2p and O 1s versus the sputtering time; an amorphous Al O film. After a long period of annealing pure magnesium heated at 700 K for (a) 3:6 103 s, (b) 9 105 s. 2 3 (>60 h) at 823 K, nucleation and growth of -Al2O3 was observed.13) Thin pieces of pure aluminum were heated in the muffle the surface to the position where the magnesium/oxygen furnace for 2:1 103, 6:2 103, 1:4 104 and 2:2 104 s, ratio is about 1, is rich in oxygen. This ratio then increases as respectively, after which they were cooled in air, then the sputtering time (or the depth of the oxide film) increases removed for running X-ray diffractometer analyses. The indicating that the metallic magnesium in the inner layer has samples were heated for times corresponding to ‘‘A’’, ‘‘B’’, increased. ‘‘C’’ and ‘‘D’’ on the TGA curve; see Fig. 3. During heating, The thermogravimetric analysis is recorded in Fig. 7. The pure Al hydrated to form gibbsite, which contributed to the cube sample was preheated to 423 K for 1800 s, then heated weight gain in stage I, as confirmed by sample ‘‘A’’ in Fig. 8. up at 5 K/min to 700 K for 2:16 104 s in an air atmosphere, With increasing temperature, gibbsite is subject to the 3 with an air-flow rate of 50 cm /min. Brucite Mg(OH)2 following reactions; Al(OH)3 ! AlO(OH) þ H2O and formed when the magnesium sample was heated to and held 2AlO(OH) ! Al2O3 þ H2O. The weight loss is due to at 423 K. The brucite dehydration process is reversible dehydration; the weight change is about 6:55 105 mg/ (brucite , periclase þ water); this reaction can be shifted in K. Gamma alumina can be expected to form on the top layer either direction by increasing or decreasing the amount of of the thermally-formed oxide film; see Fig. 8, sample ‘‘B’’. water vapor at the appropriate temperature.10,11) The TGA Lippens et al. found that under hydrothermal conditions measurements indicate that the weight decreased rapidly due below 107 Pa, boehmite AlO(OH) can be formed from to the dehydroxylation of brucite; the sample’s weight gibbsite and boehmite remains in a thermaldynamically stabilized after temperature reached 700 K and was held a stable phase up to 571 K.14) If the hydrothermally-formed Thermally-Formed Oxide on Aluminum and Magnesium 1351

-- aluminum thermally-formed oxide film do not change much with -- diaspore 4000 -- gibbsite extended heating time; compare sample ‘‘C’’ to ‘‘D’’ in -- α−alumina -- γ−alumina Fig. 8. In stage IV however the oxide film increases in 2000 thickness leading to an apparent weight gain. Hydration "A" persists at the interface. Noted that some gibbsite still exists 1000 at stage IV; see Fig. 8. The test data in Fig. 8 indicate no or few detectable boehmite peaks; AlO(OH): 2 theta = 14.5 or 0 28.2 degrees. During heating the formation of diaspore 2000 insead of boehmite can only be interpreted as due to the "B" 1000 pressure effect. The mechanism for the formation of microchannels during 0 the dehydration process is unknown, but they are indeed Intensity 400 "C" initiated at all triple-junctions of grains and inclusion particles. In stage IV gibbsite continues to form at the 200 interface of oxide film and substrate. The dehydrated oxide has a limited chance to rehydrate during holding at 883 K in 0 400 the muffle furnace due to lack of hydrogen at this high "D" temperature. If hydration continues to occur, hydrogen can

200 only come from the bulk of the sample. The HORIBA EMGA-621W machine was utilized to test the hydrogen

0 content of pure Al samples, with and without heating. The 3 0 102030405060708090 hydrogen content of pure Al is 0.077 cm /100 g of Al and 3 2 Theta, degree/θ that of heated sample (25 hrs at 883 K) is 0.044 cm /100 g of Al. Hydrogen diffuses from the bulk and moved along grain Fig. 8 X-ray powder diffractometer analyses (intensity versus two theta) boundaries to the interface, forming gibbsite during heating. for four samples, ‘‘A’’, ‘‘B’’, ‘‘C’’ and ‘‘D’’ (as shown in Fig. 3) with Gibbsite thus persistently forms at the interface, then 3 3 4 4 different heating time, 2:1 10 , 6:2 10 , 1:4 10 and 6:2 10 s, dehydrates, releasing H or O–H into the surrounding respectively; heating rate 8.3 K/min holding at 883 K. 2 atmosphere via channels through the complex oxide film. Therefore it can be seen that the microchannels are fully open if gibbsite can persistently form and dehydrate at the boemite is further heated in air, then dehydration to -Al2O3 interface. Oxygen diffuses inward from the atmosphere to occurs. The boehmite will decompose into -Al2O3 after the the interface to react with the aluminum there, increasing the gibbsite is heated in air for 24 hrs at 803 K.15) If the gibbsite is thickness of the oxide film and the weight of the sample, as hydrothermally-treated at a higher pressure, diaspore indicated by the TGA tests. If hydrogen is available at the AlO(OH) is likely to form. Chen et al. prepared pure interface, such as in the area close to the triple-junction of the diaspore samples by the dehydration of gibbsite at a grains, gibbsite may form first, then dehydrate to form 7 temperature of 623 K and a pressure of 1:5 10 Pa diaspore, -Al2O3 or -Al2O3, depending on the surrounding (1 atm ¼ 1:0132 105 Pa).16) In this study, we detected the pressure during the reaction. Consequently, the thermally- existence of diaspore in sample ‘‘B’’; see Fig. 8. Presumably, formed oxide film will be comprised of complex oxides, in -Al2O3 covered the thermally-formed oxide film, and conjunction with an oxygen rich top layer and evenly- dehydration persistently occurred within the oxide film. distributed microchannels. 3 The molar volume of water (H2O) is 18.069 cm and the 3.3.2 Brucite and periclase steam (ideal gas) is 24789.2 cm3.17) The pores generated from Brucite can remain balanced with periclase (MgO) plus the phase transformation of Al(OH)3 to AlO(OH), may water at different temperatures with different pressures. The subject to a high local pressure, since rigid -Al2O3 covers dehydration of brucite liberates the water, changing its over the top layer. The molar volume of diaspore differs from volume in the volumometer system.10) In this study, a pure that of boehmite; the former is much more compact than the magnesium sample was preheated at 473 K for 1800 s, with latter. Both diaspore and boehmite have formula weight of an air flow rate of 50 cm3/min used in the thermogravimeter 59.989 g/mol, but molar volume 17.76 vs. 19.54 cm3.17) An analysis. When the sample was heated for a short time, the apparent weight gain occurs at 832 K due to the trans- magnesium hydrates changes to brucite slightly increasing formation of gibbsite to diaspore and it is possible that the sample’s weight as shown in Fig. 7. During heating, the hydration still persists at the interface between oxide film and dehydration of brucite reduced the weight of the sample at a substrate as noted by the existence of gibbsite. Tsuchida used rate of 1:843 104 mg/K, which was far greater than that TG-DTA and a high-temperature X-ray diffractomer to study for gibbsite (6:55 105 mg/K) in Fig. 3. A heating rate the dehydration of diaspore in air. The heating rate was 5 K/ of 8.3 K/min was used to heat the aluminum and 5.0 K/min min. He found that diaspore decomposed directly into - to heat the magnesium. 18) Al2O3 at 773 K. It can be seen that in stage III the A protective behavior occurs in stage III (Fig. 7), due to a thermally-formed oxide film will therefore be composed of lack of easy-paths for fast Mg transport.6) Figures 5(a) and complex oxide (-Al2O3, gibbsite and a minor amount of - (b) show the sectional morphologies of a sample heated at 3 Al2O3); see sample ‘‘C’’ in Fig. 8. The constituents of the 700 K for 3:6 10 s. Microcracks are visible, but these 1352 T.-S. Shih and Z.-B. Liu

vapor to form brucite (G ¼711:8 KJ/mol) instead of periclase (G ¼525:8 KJ/mol)17) which heals the microcracks in a very short period of time, Fig. 5(b). The fresh brucite on the walls and in the valleys of the micro-crack can be transformed to periclase by dehydroxylation at the high temperature of 700 K. The above reactions proceed periodi- cally during heating, forming an irregular interface between the oxide film and the substrate; magnesium containing species can also remain within the oxide film, Fig. 6(a). When most of the brucite film is transformed to MgO (stage III in Fig. 7) the water vapor is almost completely depleted. The great difference in the thermal expansion of magnesium and MgO may introduce cracks in the film, which can then act as channels for transporting magnesium vapor outward to the surrounding atmosphere. Ridges form and the sample gains weight significantly, as shown in stage IV, (a) Fig. 7. For sample held at 700 K for 9 104 s, oxygen mapping indicates that a ridge penetrates into the substrate; see Figs. 9(a) and (b).

4. Conclusions

During heating, gibbsite forms and covers the surface of an aluminum sample. Gibbsite dehydrates to form -Al2O3 on the top layer of the thermally-formed oxide film. Diaspore then forms within the thermally-formed oxide film. Hydra- tion persists at the interface. After an extended heating, the thermally-formed oxide film will be comprised of complex oxides; gibssite, diaspore, -Al2O3 and possibly -Al2O3. This film also possesses evenly-distributed microchannels. During preheating and/or low temperature heating, brucite forms to cover the surface of a magnesium sample. This is transformed into MgO when dehydroxylation occurs at a high temperature. The dehydroxylation of brucite brings out water (b) vapor which causes cracking in the film. The Mg-containing Fig. 9 (a) The sectional SEM observation for the sample (99.956 species are transported from the metal substrate via the open mass%Mg), after being polished by abrasive papers, heated at 700 K for crack to react with this water vapor, forming brucite on the 9 105 s, (b) an oxygen mapping (EPMA) showing a ridge. walls and valleys of the cracks. Open microcracks can thus be healed by the water vapor supplied from the dehydroxylation of brucite. The water vapor becomes depleted at the end of cracks do not penetrate the oxide film; in other words, no easy the protective period. The magnesium diffuses outward from paths are available for the magnesium transport. the substrate via the cracks to react with the air, forming Brucite dehydrated and formed lamella MgO during ridges. heating. Concurrently, dehydroxylation also brings water vapor to the surface of the brucite film. The crystalline MgO was readily rehydroxylated as the water vapor pressure REFERENCES increased.10) Dehydroxylation during heating is energetically 1) P. F. Doherty and R. S. Davis: J. Appl. Phys. 34 (1963) 619–628. favorable for transforming brucite to be MgO. Rehydrox- 2) A. F. Beck, M. A. Heine, E. J. Caule and M. Yor: Corros. Sci. 7 (1967) ylation may also occur locally when a high vapor pressure 1–10. exists. Sharma observed that rehydroxylation in the heavily 3) M. J. Dignam, W. R. Faucett and H. Bohni: J. Electrochem. Soc. 113 dehydroxylated brucite matrix proceeds via nucleation and (1966) 656–662. 10) 4) J. I. Eldridge, R. J. Hussey, D. F. Mitchell and M. J. Graham: Oxid. the growth of Mg(OH)2. Met. 30 (1988) 301–328. McKelvy used in-situ nanoscale observations to study 5) T. S. Shih and I. C. Chen: Mater. Trans. JIM 46 (2005) 1868–1876. lamellar nucleation and growth of MgO in magnesium 6) F. Czerwinski: Corros. Sci. 46 (2004) 377–386. hydroxide dehydroxylation. MgO oxide lamella nucleates 7) J. H. Nordlien, S. Ono, N. Masuko and K. Nisancioglu: Corros. Sci. 39 and grows at the expense of the parent hydroxide lamella 8 (1997) 1397–1414. until dehydroxylation is complete. Nanoscale cracking and 8) P. Souza Santos, H. Souza Santos and S. P. Toledo: J. Mater. Res. 3 19) (2000) 104–114. fragmentation occur due to high cumulative strain. Due to 9) O. Salas, H. Ni, V. Jayaram, K. C. Vlach, C. G. Levi and R. Mehrabian: their high mobility during heating the magnesium containing J. Mater. Res. 6 (1991) 1964–1981. species may fill up the cracks. They react with the steam 10) R. Sharma, M. J. McKelvy, H. Bearat, A. V. G. Chizmeshya and R. W. Thermally-Formed Oxide on Aluminum and Magnesium 1353

Carpenter: Philos. Mag., A. Phys. Condens. Matter. Defects Mech. 15) C. Novak, G. Pokol, V. Izvekov and T. Gal: J. Therm. Anal. 36 (1990) Prop. 84 (2004) 2711–2729. 1895–1909. 11) J. A. Schramke, D. M. Kerrick and J. G. Blencoe: Am. Mineral. 67 16) Q. Chen, W. Zeng, S. Gu, G. Yang, H. Zhou and Z. Yin: J. Chem. (1982) 269–276. Thermodyn. 27 (1995) 443–446. 12) L. P. H. Jeurgens, W. G. Sloof, F. D. Tichelaar and E. J. Mittemeijer: 17) R. A. Robie, B. S. Hemingway and J. R. Fisher: Thermodynamic Phys. Rev. B 62 (2000) 4707–4719. Properties of Mineral and Related Substance at 298.15 K and 1 bar (105 13) K. Shinohara, T. Seo and H. Kyogoku: Z. Metallkd. 73 (1982) 774– Pascals) pressure and at Higher Temperatures, (Washington D.C.: U.S. 780. Govt. Print. Oﬀ., 1978) 18, 19, 184–185. 14) B. C. Lippens and I. Steggerda: Physical and Chemical Aspects of 18) T. Tsuchida: Solid State Ion. 63–65 (1993) 464–470. Adsorbents and Catalysts; ed. B. G. Linsen (Academic Press, London- 19) M. J. McKelvy, R. Sharma, A. V. G. Chizmeshya, R. W. Carpenter and New York, 1970) 171. K. Streib: Chem. Mater. 13 (2001) 921–926.