April 5, 1960 PHOSPHIDESOF ALUMINUM, GALLIUM AND INDIUM 1537 tained when the hydrogen atom concentration near an activation energy of approximately 1-2 kcal./ the wall was lowered by increasing the path between mole.8,9 filament and wall. The necessity for providing It should be emphasized that the composite a very cold surface and the stability problem were H/AI ratios in Table I do not necessarily represent shown by the drastic reduction in reaction product the compositions of the hydride molecules because when a bath of -78” was substituted for the - 195” these ratios also include unreacted aluminum bath. These observations lead to the hypothesis atoms and aluminum atoms which resulted from that at least the last, if not all, of the reaction steps the decomposition of initially formed hydrides. leading to the formation of the hydride occur at the Since aluminum distilled to the wall of the reactor wall and that the continuous evaporation of alumi- where the surface was constantly covered by suc- num serves the purpose of providing a constantly cessive layers, it is extremely unlikely that all renewed film surface. However, the data do not aluminum atoms could react, keeping in mind that eliminate the possibility that this is essentially a covered atoms were lost to the reaction. The gas phase rather than a heterogeneous reaction. drastic reduction in product when the wall bath The experimental determination of half-order was at -78”, a temperature at which the hydride kinetics for this reaction is in line with similar is normally quite stable, proves that the hydride kinetics that previously were found when hydrogen molecules at the constantly changing surface were was dissociated into atoms at a glowing tungsten exposed to a disruptive environment. This con- filament.6,7 Roberts and Bryce have explained sisted of constant attack by hydrogen atoms with their kinetics in terms of the mechanism of dis- H recombination reactions liberating 103 kcal./ sociation of hydrogen at the tungsten surface over mole while hydride molecules with an activation a temperature range that coincides with the present energy for decomposition of 14.4 kcal./mole served study. It may thus be reasonably assumed that as third bodies; in addition radiation and conduc- the rate determining step in the present process is tion energy was transferred from the glowing the production of hydrogen atoms at the filament. filaments. Since a composite value of H/Al = The geometry of the apparatus supports this con- 1.02 was achieved under these conditions, it is clusion since the probability of hydrogen collisions reasonable to postulate that for the actual hydride, with the filament is considerably lower than gas AlH,, x was greater than unity.lc phase or wall collisions. Acknowledgments.-The author gratefully ac- That the hydride formed was a chemical com- knowledges the interest and advice of Dr. A. pound rather than physical adsorption of hydrogen Wheeler, Dr. E. Mishuck and D. Seedman and is was demonstrated by the decomposition kinetics. indebted to Dr. Seymour Siege1 for critically The activation energy of 14.4 kcal./mole establishes reading the manuscript. The Bureau of Ordnance that the bonding is in the class of chemical com- of the U. S. Dept. of the Navy sponsored this in- pounds.8 A physical desorption process would have vestigation under Contract NOrd-18386. (6) J. K. Roberts and G. Bryce, Proc. Cambridge Phil. SOC.,33, 653 (9) A. Eucken and W. Hunsmann, Chem. Abst., 34, 666 (1940). (1936). (10) The instability of AIHz at ambient temperatures and the fact (7) I. Mochan, Chem. Abst., 36, 3091 (1942). that the hydride was isolated admixed with surface-active aluminum (8) S. Glasstone, K. Laidler and H. Eyring, “The Theory of Rate have made its chemical characterization difficult; for example, both Processes,” chapter Heterogeneous Processes, McGraw-Hill Book on AIH, and aluminum deposited in the present apparatus react com- Co., New York. N. Y., and R. Gomer and C. S. Smith, “Struc- 1941, pletely with water to liberate H2. ture and Properties of Solid Surfaces,” chapter on Chemisorption, Univ. Chicago Press, Chicago, Illinois, 1952. AZUSA,CALIFORNIA

[CONTRIBUTION FROM THE GENERALELECTRIC COMPAXY, LAMPDEVELOPMENT DEPARTMENT] On the Preparation of the of Aluminum, Gallium and Indium BY ARRIGOADDAMIANO RECEIVEDAUGUST 26, 1959

The phosphides of aluminum, gallium and indium can be prepared by reaction between those metals and , ZnaPz,in a neutral atmosphere (argon, hydrogen) at about 800-900’. The reactions proceed according to the scheme 2Me + ZnaPt +.2MeP + 3Zn t (Me = Al, Ga, 2x1) with formation of the phosphides of the third group elements and evolution of zinc vapors. X-Ray and analytical data are gven. The yield of reaction is a maximum for Alp, while some metal is left unreacted in the preparation of GaP and InP.

Introduction perhaps the phosphide of aluminum, while gallium The phosphides of aluminum, gallium and in- and have been prepared in the dium all have been prepared by direct synthesis form of single ~rystals~~~Jand considered for appli- from the elements.’-’ Of them the least known is cations such as electroluminescence and high- (1) W. E. White and A. H. Bushey, THIS JOURNAL, 66, 1666 temperature rectification. It seemed interesting (1944). (5) G. A. Wolff, R. A. Hebert and J. D. Broder, “ (2) E. Monlignie, Bull. SOL. chim. France, 276 (1946). and Phosphors,” M. Schon and H. Welker, Editors, Interscience Pub- (3) J. W. Mellor, “Inorganic and Theoretical Chemistry,” Vol. VII1, lishers, he., New York, N. Y., 1958, p. 547. Longmans, Gteen and Co., London, second impression, 1947, p. 645. (6) G. A. Wolff, L. Toman, Jr., N. J. Field and J. C. Clark, ibid., p. (4) G. Wolff, P. H. Keck and J. L). Broder, Pkys. Rev., (2) 94, 753 463. (1954). (7) R. J. Guire and K. Weiser, U.S. P. 2,671,100(1959). 1538 ARKIGOADDAMIANO VOl. 82 to prepare these compounds by a method which cases where zinc phosphide was in excess the un- would avoid the hazards of working with dangerous reacted zinc phosphide was driven off at the end of materials such as or and the reaction by raising the temperature to 1000- which would lend itself to continuous production of 1100' for about 30 minutes. the phosphides. The reaction between the metals, The reaction products, both those remaining in aluminum, gallium and indium, and a commercially the boat and those which had condensed down- available product, zinc-phosphide, ZngPZ,proved stream on the silica tube, were identified by X-ray satisfactory to this end and will be described here. diffraction. It was found that between 400 and GOOo no reaction occurs. In 6 hr. at 700' incom- Experimental and Results plete reaction takes place. Five hours at 800-900' The general schema of reaction for the prepara- results in an almost quantitative formation of alu- tion of the metal phosphides in question is minum phosphide and zinc. At about 1000° zinc phosphide begins to sublime and its presence in the 2Me + ZnaPz --f 2MeP + 3Znt (1) sublimation products can be detected by X-ray where Me = Al, Ga, In. The reaction can be ex- diffraction. pected to lead to quantitative formation of the X-Ray powder photographs of different samples phosphides of the third-group elements if the zinc of aluminum phosphide are consistent with the is subtracted from the reaction as soon as it is existence of a cubic, zinc-blende structure. The formed, for instance by sublimation. This can be measured value of the cell-edge is a0 = 5.451 A., accomplished by operating at temperatures close to in good agreement with an earlier determination by the b.p. of zinc (907'). In fact the b.p. of alumi- Passerini. lo um, gallium and indium are very high (ca. 2000°), The intensities of the lines, however, were found the Sublimation rate of zinc phosphide becomes ap- to be quite different from the data reported by preciable only around 110Oo,* and the melting the ASTM File Index. A recalculation of the in- points of the phosphides of aluminum, gallium and tensities indiumsvgare above 1000°. In practice a tempera- ture of 800' circa proved convenient for the prepa- ration of all three of the phosphides. where IF1 = structure factor modulus, fi = multi- Aluminum Phosphide.-Only two papers1,* on plicity, 8 = Bragg angle of reflection, gave good the preparation of aluminum phosphide have been agreement with the experimental values, which published in recent years. Although in both cases are reported in Table I. the direct synthesis from the elements was used, the properties ascribed to the products obtained TABLEI (such as body color, thermal stability, resistance to X-RAYDATA FOR A1P attack by water, acids and alkalies, etc.) are en- (CUI<, ratlistioii, ha, = 1.5405 A.; a. = 5.451 A,) tirely different. hkl sin 0 lobed. lenlod. x lo-' The zinc phosphide and aluminum powder used 111 0 213 V.V.S. 866 in my experiments were commercial products, 200 , '7x3 ... 3 from Fisher Scientific Company. The aluminum II'?a0 399 V.S. 568 powder was assumed to be pure: a trace of alunii- 311 ,469 S. 339 num oxide and perhaps smaller amounts of other 222 ,489 ... 53 impurities were probably present. As for the zinc 400 ,565 V.W. 88 phosphide, which was known to contain relatively 33 1 ,616 m.s. 136 large amounts of water and other impurities, 420 ,632 ... <1 purification was deemed necessary. This was ac- 422 602 m.s. 250 complished by subliming the compound in a stream 511 + 333 . 734 1v. 83 + 28 of argon at 1100'. In agreement with Jolibois8 410 ,799 V.V.W. 72 it was found that zinc phosphide sublimes unde- 531 ,836 m.s. 146 composed. The compact mass obtained by subli- 600 + 442 ,847 .I. 0 + 44 mation, which often contained beautiful single 020 ,894 ins. 157 crystals of zinc phosphide, was reduced once more 533 ,926 V.W. 95 to a fine powder (minus 200 mesh size), and this 622 ,937 ... 64 powder was used for the preparations. 444 ,979 W. 143 Weighed amounts of zinc phosphide and alumi- a V. = very; s. = strong; w. = weak; m. = moderately. num were mixed thoroughly and transferred to a boat of silica or spectrographically pure graphite. Quantitative analytical data were obtained by measuring the change in weight of the contents of The boat was placed in a silica tube and heated the reaction boats upon reaction. They are con- for several hours at the desired temperature in a combustion tube furnace. A continuous flow of sistent with a yield of reaction of 94% or more. argon through the silica tube prevented oxidation of Samples of A1P also have been weighed, dissolved the sample and also displaced the evolved zinc through acid attack and the aluminum determined vapor. after precipitation by %-hydroxyquinolinein alka- A stoichiometric mixture of aluminum and zinc line media. For a preparation at 850°, for instance, phosphide was used for most experiments. In those the data are: found A1 46.8170, theor. A1 46.54%. The body color of aluminum phosphide obtained (8) P. Jolibois, Compl. rend., 147, 801 (1908). is pea-green, and its physical and chemical proper- (9) J. van den Boomgaard and K. Schol, Philips Research Repfs., 12, 127 (1857). (10) L. Passerini, Gam chim. ifai., 68, 662 (1928). April 5, 1960 PIlOSPHIDES OF ALUMINUM,G.4LIAIUhf AND INDIUM 1539 ties are in agreement with those described by White TABLEI1 and Bushey.' Thus the substance has a low vapor X-RAYDATA FOR GaP pressure at 1000G,has a faint smell of phosphine, (CuK, radiation, A,, = 1.5405 A.; a0 = 5.4504 A.) reacts violently with concd. or dil. HC1, and with h kl sin 0 Iobed.o Icdod. x lo-' explosion with 1:l H2S04, aqua regia, etc. Some 111 0.245 V.S. 2933 differences, more of a quantitative than of a quali- 200 ,283 V.V.W. 399 tative nature, were observed, however, in the sta- 220 ,399 S. 1688 bility in presence of water and alkalies. It was ob- 311 ,469 m.s. 1136 served that after some weeks immersion in water 222 .489 ... 138 samples of A1P still gave the diffraction pattern of 400 .565 V.V.W. 281 the compound together with lines due to products of 331 ,616 W. 442 hydrolysis. Also, alkaline solutions at room tem- 420 .632 ... 104 perature attack aluminum phosphide very slowly, 422 .692 w . 604 without fast development of gases and inflammation 511 + 333 ,734 V.W. 281 + 94 of the vapors, as is the case with oxidizing acids. 440 .:99 V.V.W. 225 These differences of behavior probably are due to 53 1 ,836 W. 523 the absence, in the aluminum phosphide prepared 600 f 442 .848 ... 17 + 111 from zinc phosphide and aluminum, of free phos- 620 .894 W. 520 phorus and aluminum, which are always present in 533 ,926 V.W. 351 the products obtained by direct synthesis. 622 ,937 ... 160 Gallium Phosphide.-The preparation of gallium 444 .979 V.V.W. 425 phosphide is very similar to that of aluminum Ov. = very; s. = strong; w. = weak; m. = moder- phosphide. Either pure gallium (99.99% pure), ately. or a Ga-Zn eutectic (m.p. 25') were used. Difficul- TABLE ties are encountered in obtaining a homogeneous I11 mixture of the metal and zinc phosphide. By long X-RAYDATA FOR InP shaking of a mixture of liquid gallium and zinc (CuK, radiation; A,, = 1.5405 A.; a0 = 5.8687 A.) phosphide, the gallium can be dispersed in the form hkl sin B Iobd. 6 loalod. x lo-' of minute droplets. However, in the process of 111 0.227 V.S. 8952 heating, large drops of gallium metal form again 200 .262 m.s. tow. 2236 and these react only on the surface, greatly de- 220 .371 s. 4931 creasing the yield of reaction. Occasional shaking 311 .435 s. 4040 of the reaction system helps somewhat but a con- 222 ,455 V.W. 684 tinuous stirring is recommended. Operating with 400 .525 w. 79 1 an excess of zinc phosphide increases slightly the 33 1 .572 m.s. 1606 yield of reaction. Zinc phosphide in excess can 420 .587 w. 778 be separated easily by hydrochloric acid treatment, 422 .643 m.s. 1694 as gallium phosphide dissolves in hydrochloric 511 + 333 .682 m.s. 939 + 313 acid only with difficulty. 440 .742 w. 579 The reaction occurs at SOO', when evolution .of 53 1 .776 ms. 1488 zinc vapors becomes appreciable. It is, however, 600 + 442 .787 w. 98 + 436 slower than in the case of aluminum phosphide and 620 .830 m.s. 1089 reaction times of 12-24 hr. are required. Even so 533 .860 w. 777 not all the gallium metal reacts. What is left un- 622 .870 V.W. 471 reacted can be separated mechanically and by 444 .909 V.V.W. 459 centrifugation and/or sedimentation techniques. 711 + 551 .937 m.s. 1081 + 1081 The product obtained after purification is a powder 640 .946 V.V.W. 627 of yellow or yellow-green color, rather than an 642 .982 s. 6040 Ov. = very; w. = weak; s. = strong; m. = moder- orange one, probably because of traces of gallium ately. metal still present. The powder pattern of it, however, contains only the lines of gallium phos- with values by Iandelli.12 Table I11 contains the phide and the intensities are in good agreement with relevant X-ray data. the calculated values, as shown in Table 11. Conclusions Indium Phosphide.-Pure (99.999%) indium metal and zinc phosphide in excess over stoichio- The method just described for the preparation of metric requirements, produce InP with low yields the phosphides of aluminum, gallium and indium, after 1-2 days at 700-800°. Excess zinc phos- is relatively simple and danger free. The reactions phide can be eliminated by HC1 treatment, and occur at low temperature and, at least for alumi- much unreacted indium metal is eliminated me- num phosphide, the yield is the theoretical one. chanically, very much as in the case for Gap. The An interesting feature of the process is also the product obtained, a dark gray or black powder, ease with which excess unreacted zinc phosphide gives the expected X-ray pattern'l for a substance can be eliminated at the end of the reactions, by with zinc-blende structure and cell edge a. = 5.87 sublimation in the case of aluminum phosphide, The relative intensities of the lines agree with which melts at high temperature but is easily at- A. tacked by acids, and by hydrochloric acid treat- the calculated values, though they are at variance (12) A. Iandelli. Gam. chim. ild.,71, 58 (1941). These data contain (11) G. Giesecke and H. Pfister. Acta Cryst., 11, 369 (1958). an absorption correction with pi - - (private communication). 1540 P. J. HERLEYAND E. G. BOUT Vol. 82 ment for the phosphides of gallium and indium, industrial production of the phosphides of alumi- which are stable in the presence of non-oxidizing num, gallium and indium. acids. Acknowledgments.-The author wishes to thank The X-ray photographs of the products ob- several colleagues of the Lamp Development tained, after purification, contain no lines due to Department for their contribution at different impurities or to phosphides of different compo- stages of the research: R. P. Taylor for the col- sition. Unreacted metals (Al, Ga, In), if present in lection of analytical data on AlP and a procedure the end products, are therefore in trace amount to separate excess gallium from gallium phosphide: concentration. While in preparations on a labora- Miss J. R. Cooper for obtaining X-ray powder tory scale the zinc obtained in the reactions is not photographs of different samples; F. Kuhlman for worth saving, a continuous process can be en- calculating the theoretical intensities of the X-ray visaged, in which the zinc is retransformed into reflections on our Bendix G-15-D computer. zinc phosphide and this reacts again with the Helpful discussions with D. M. Speros also are third group elements. This possibility makes the gratefully acknowledged. process interesting from the standpoint of the CLEVELAND,OHIO

[CONTRIBUTION FROM THE DEPARTMENTOF CHEMISTRY, RHODESuNIVERSITY, GRAHAMSTOWVN,SOUTH AFRICA] The Thermal Decomposition of Silver Oxide

BY P. J. HERLEYAND E. G PROUT RECEIVEDJULY 23, 1959

The results of the thermal decomposition of silver oxide, freed from silver carbonate by preheating at 280' for 3 hr. in vacuo, are highly reproducible. Study of the kinetics of the reaction in the range 330-380" has shown that the pressure- time curves over the acceleratory region are represented by the equation P'/z = kt + c. Reaction proceeds by the growth of two-dimensional nuclei with an activation energy of 28 3 kcal./mole. The decay stage follows a unimolecular law and the activation energy is 29.6 kcal./mole. Pre-irradiation by ultraviolet light, cathode rays, WOy-rays, thermal neutrons and fast neutrons has no effect on the subsequent thermal decomposition

The thermal decomposition of silver oxide has Benton and Drake4 decomposed silver oxide been investigated by various workers. Their prepared by the combination of silver and oxygen. results show discrepancies between the activation The decomposition temperature was 150" lower energies, the form of the pressure-time curves than that of Lewis and the activation energy was and the mathematical relationships describing 35-36 kcal.,/mole. such plots. hverbukh and Chufarov5 decomposed silver Lewis' prepared silver oxide from' silver nitrate oxide on a spring balance. The rate of decompo- and barium hydroxide (or sodium carbonate). sition for a constant degree of dissociation was de- The oxide was heated in oxygen at atmospheric termined and an activation energy of 29.0 kcal./ pressure at 320-330". The decomposition, in the mole was calculated. case of the first specimen (Merck silver oxide), Garner and Reeves6 repeated the preparations was preceded by very long induction periods. of Lewis and decomposed the oxide at 300-330" None of the prepared specimens showed this prop- in vacuum. The pressure-time plots showed no erty. In all cases the equation induction period but gave a maximum rate at the dx/'dt = k~(l- X) (1) beginning of the reaction. The curves were ir- regular and not of an autocatalytic type. As well, where x is the fractional number of moles of silver specimens of silver oxide were annealed in oxygen oxide decomposed, was obeyed. The activation at 28-38 atm. for 8-10 days at 200-300" to yield a energy for the reaction was 31.8 kcal. /mole. dense coarsely crystalline oxide which decomposed Hood and Murphy,2 using a similar procedure, in zacuo to yield sigmoid curves for the pressure- confirmed the results of Lewis. The integrated time plots. The autocatalytic relation of Lewis was form of equation 1 not valid, but it was considered that the plots were In x/(l - x) = kt + c (2) described by the power law was applicable. P'l3 = kit + c1 (3) Pavlyuchenko and Gurevich3 prepared silver It was not possible to determine the activation oxide in darkness, and in the presence of air, by energy in the usual manner due, presumably, to the adding an aqueous solution of KOH to an aqueous irreproducibility of results from sample to sample. solution of AgN03 at 5'. They obtained complex Instead split runs on the same specimen of oxide at pressure-time curves and a low activation energy of different temperatures were performed. The acti- 10.2 kcal./mole. Decomposition was carried out vation energy for the reaction was 45.6-46.2 kcal.:' in vacuo in the relatively low temperature range of mole. 118-220". (4) A. F. Benton and L. C. Drake, THISJOURNAL, 66, 255 (1934). (1) G. N. Lewis, Z. 9hysi.k. Chem., 62, 310 (1905). (5) B. D. Averbukh and G. I. Chufarov, Zhur. Fiz. Khim., 23, 37 (2) G. C. Hood and G. W. Murphy, J. Chem. Educ., 26, 169 (1949). (1949). (3) M. M. Pavlyuchenko and E. Gurevich, J. Gen. Chem., U.S.S.R. (6) W. E. Garner and I.,. 1%'. Reevrs, Trans. Furaday .Tor., 60, 254 (Eng. Transl.), 21, 511 (1951). (1:I.j4),