INORGANIC CHEMISTRY.1. NON-STOICHEIOMETRIC COMPOUNDS.RECENT work in inorganic chemistry has raised the question of the validityof the law of constant proportions, as applied to solid compounds. Theexistence and classification of “ Berthollide ” compounds have been notedpreviously in these Reports,l and it appears opportune to review our presentstate of knowledge of the subject.It is generally recognised that the intermediate phases in metallic systemsmay exist over a range of composition, not necessarily including a rationalchemical formula, ; an idealised chemical formula can usually be assigned,however, based on the composition of the unit cell of the crystal.2 Betweenintermetallic and ionic compounds there is a transition, rather than anabrupt demar~ation,~ depending on the difference in electronegativity ofthe combining atoms, and whereas the elements of Groups VB, VIB, andVIIB form true salts with the most electropositive metals (Groups IA, IIA),their compounds with the transition and B-sub-group metals display acomplete transition from the ionic to the quasi-metallic type.Variabilityof composition runs broadly parallel t o sub-metallic properties, but is byno means limited to compounds of obviously sub-metallic character.The distinction between solid solutions, interstitial compounds and non-stoicheiometric compounds is, in the last analysis, rather arbitrary.N. S. Kurnakow first proposed the term “ Berthollide ” (as distinct from“ Daltonide ”) to describe homogeneous phases in systems where themaxima or minima of properties-melting point, conductivity, lattice order,etc.-do not coincide with a rational atomic ratio of the components. Forthe purpose of this report it is convenient to follow W.Schottky andC. Wagner in considering the familiar “ Daltonide ” type as a specialcase of the “ ordered mixed phase ”, a 2-component (or multicomponent)system with statistically regular lattice array.Our present knowledge of crystal structure confers precise meaning onthe term “solid solution” as applied to crystals of atomic lattice types.I n a crystal phase of ideal formula AB,, a stoicheiometric excess of elementB can be accommodated structurally in only three ways : (i) Substitutionalsolid solution : B atoms replace A atoms on lattice sites proper to A.(ii)Interstitial solid solution : additional B atoms are located in inter-latticepositions. (iii) Xubtractive solid solution : all B atoms occupy proper Blattice sites, but a number of A lattice sites is left untenanted.Since (ii) increases and (iii) decreases the average weight per unit cell,distinction between them is possible by combining density and X-ray cellAnn. Reports, 1933, 30, 381; 1935, 32, 211.A. Westgren, Angew. Chem., 1932, 45, 33.Cf. E. Zintl, ibid., 1939, 52, 1.2. physikal. Chem., 1930, B, 11, 163.a 2. anorg. Chem., 1914, 88, 109ANDERSON : NON-STOICHEIOMETRIC COMPOUNDS. 106dimension measurements. I n this way it was shown that in pyrrhotite,FeS-FeSl.l,,6 ferrous selenide, FeSe-FeSe,.,,' and wiistite, FeOl.m-FeO,.,, the stoicheiometric excess of non-metal represents a cationdeficiency, the anion lattice being substantially complete.Thus a pyr-rhotite Fe,S, is properly represented Fe,,.,,S; it cannot be regarded as asolid solution between two Daltonide compounds, and is a true non-stoicheio-metric compound. In the €-phase of the Fe-Sb system (ideally FeSb),increase in cell dimensions with increasing iron content above the idealformula indicates that the excess of iron is accommodated interstitially .gSubstitutional solid solution is likely only in intermetallic compounds,where ionic repulsions mould not be involved. Thus, in the @-phase of theNa-Pb system (27-35 atoms % Na; ideal composition, NaPb,), thestoicheiometric phase lies outside the iange of homogeneity ; the stablephase has 4-9% of the Pb atoms replaced by Na.lo According to M.J.Buerger,ll in the marcasite-type FeSb,, FeAs,, FeS,, the stoicheiometricexcess of iron usually present is substituted for a proportion of the non-metal. Such substitution in a metallic sulphide seems improbable, andthis series could profitably be reinvestigated.The conditions of equilibrium of lattice defects in a real crystal wereworked out by W. Schottky and C . Wagner?? l 2 I n a stoicheiometriccrystalline compound MX, displacement of atoms from their regular latticepositions would be an endothermic process ; the resulting interstitial atomsor vacant lattice sites could be distributed a t random amongst any of theavailable positions of the crystal lattice.The defects therefore contributesubstantially to the configurational entropy, as well as raising the totalenergy of the crystal, and it emerges that at all temperatures above 0" K.the free energy G (= H - TX) is a minimum for certain finite concentrationsof lattice defects of each kind (depending on the energy involved in creatingthe defects). If interstitial M atoms, interstitial X atoms, vacant M sites,and vacant X sites are all present in significant CQncentrations, the equi-librium conditions are rather complex. However, if-as is reasonable forkT <energy of defect formation-it can be assumed that all types ofdefect are not equally probable, two simple limiting cases arise : (i) Equalconcentrations of vacant cation sites and vacant anion sites (Schottkydefects); l3 believed valid, e.g., for NaC1.(ii) Interstitial atoms of one orG. Hagg and G. Siicksdorf, 2. physikal. Chem., 1933, B, 22, 444; Nature, 1933,131, 167.G. Htigg and A. L. Kindstrom, 2. physkal. Chem., 1933, B, 22, 453.* E. R. Jette and F. Foote, J . Chem. Physics, 1932, 1, 29.0 A. Oftedal, 2. physikal. Chem., 1927, 128, 135; G. HBgg, 2. Krist., 1928, 68,lo E. Zintl and A. Harder, 2. physikal. Chem., 1931, A , 154, 63.l1 Amer. Min., 1934, 19, 37.12 W. Schottky, 2. physikal. Chem., 1935, B, 29, 335; R. H. Fowler and E. A.13 W. Schottky, Naturwiss., 1935, 23, 656; 2. physikal. Chem., 1935, By 29,470.Guggenheim, " Statistical Thermodynamics ", Cambridge, 1939, paras. 1302, 1303.335106 INORGANIC CHEMISTRY.other kind, with a corresponding number of vacant lattice sites (Frenkeldefects) ; l4 example, AgBr.For thecrystal in contact with the vapour of one of its components (e.g., a diatomicnon-metal, such as O,, I,, etc.), we must consider the possible addition orremoval of X- ions a t the surface of the originally stoicheiometric crystal,by such processes as (I) or (11).(1) M++M++ + eAddition of supernumerary X- ions to the crystal involves ( a ) an increasein valency of a corresponding number of M+ ions and (b) the creation ofvacant cation sites which will ultimately distribute themselves by diffusionthroughout the lattice.(11) X- + Mf (on lattice site) + 4X2 (gas) + M+ (interstitial) + e (trappedRemoval of X- ions from the lattice involves ( a ) effective conversion of thesame number of M+ cations into M atoms and (b) creation of interstitialatoms or (for crystals with Schottky defects) vacant anion sites.Addition or removal of X ions will accordingly involve changes bothin total energy and in configurational entropy as compared with thestoicheiometric crystal.The minimum free energy for any temperatureand any given pressure px of the vapour X, corresponds (for a crystal withFrenkel defects) with concentrations of defects given * byThe stoicheiometric compound is, however, a limiting case.&X2 + e (at surface) ---+ X- (on lattice site) + cation holenear interstitial atom)Nh = Nl . px). exp. - Ex/kT . . . . . . . (1)Ni = Nl . p x - t . exp. - (Eh + Ei + E x ) / k T .. . (2)where Nh, Ni = number of M holes and interstitial M atoms in a crystalcontaining NZ cations ; Eh, Ei = energy expenditure to produce one vacantM site or interstitial M ?tom in the stoicheiometric crystal, Ex = expenditureof energy in adding one additional X atom to the crystal. The concen-trations of holes and interstitial atoms are not independent, being related by(3) defines in effect the equilibrium constant of a quasi-chemicaldissociation :Lattice site occupied by ion ---+ Interstitial ion + lattice holeFor the stoicheiometric crystal,6 is the intrinsic disorder of the stoicheiometric crystal, which is in equi-librium with one particular partial pressure of the component X only.l4 J. Frenkel, 2. Physik, 1926, 35, 652.* Approximately, the.contributions to the vibrational energy being neglected (cf.Nrt*lN1 = Ni*/NI = 6 = exp. -(Eh + Ei)/2kT . . (4)Mott and Gurney, " Electronic Processes in Ionic Crystals ", Oxford, 1940, p. 29)ANDERSON : NON-STOICHEIOMETRIC COMPOUNDS. 107For any other pressure of X the crystal will contain at equilibrium astoicheiometric excess (or deficiency) of X given by Nh - Ni, which dependsupon the intrinsic disorder of the stoicheiometric phase. Writing (Nh -Ni)/Nl = n, and the pressures of X in equilibrium with MX1.OOO, MX, + 12as p(O), p(n), respectively, we have. . . ( 5 )Variation of stoicheiometric defect for two values of 6 is illustrated in thefigure. If the stoicheiometric crystal is almost perfectly ordered, the40 1 6 = 0 - 0005I 6= U405equilibrium pressure must increase very steeply to produce even smallchanges in composition.The essential consequences of this thermodynamic theorem are asfollows.(i) It is perfectly general, suggesting potential variability of compositionfor all ionic, semimetallic, or intermetallic compounds.(ii) Unless the degree of lattice disorder in the stoicheiometric compoundis appreciable, variation of composition under experimentally accessibleequilibrium conditions may be imperceptibly small.The “ Daltonide ”compound thus appears as a special case.15(iii) The intrinsic disorder 6 will be small unless production of defectsis not too endothermic, as compared with the thermal energy. Eh, Ei aresmaller than the lattice energy of the crystal by a factor depending l6 onthe polarisation and distortion of the crystal lattice around each defect.Non-ionic interactions ( e . g ., van der Waals forces) between the atoms areparticularly important in stabilising defects and in favouring the locationof atoms or ions in interstitial positions. Compounds of the transition orl6 A. &mder, 2. physikal. Chern., 1933, A , 165, 65.l6 W. Jost, J . Chem. Physics, 1932, 1, 466; W. Jost and G. Nehlep, 2. physikal.Chem., 1936, B, 32, 1 ; N. F. Mott and M. J. Littleton, Trans. Faraday Soc., 1938,34,485; W. Jost, ibid., 1938, 34, 860108 INORGAXIU CHEWSTRY.B-sub-group metals are therefore likely to possess a higher degree of intrinsicdisorder than are structures bugt up from the inert-gas-like ions.Estimatesof 6 in typical compounds have been made by J. Addink l7 and by E. Kochand C. Wagner.l*(iv) Deviation from stoicheiometry involves a valence change. Excessmetal may be incorporated by converting some cations effectively intoneutral atoms, or into cations of lower valency. This decrease in valencyis possible for any cation, including the inert-gas-like cations. The alkalihalides, heated in the vapour of the corresponding metal, will take up afew atoms per thousand of excess metal.19 Excess non-metal involves thepresence either of cations of higher valency (energetically permissible onlyin compounds of the metals displaying variable valency), or of anions oflower valency. The low binding energy renders the latter source of stoicheio-metric variation less favoured, but potassium iodide (e.g.) will incorporatea stoicheiometric excess of iodine (up to a few atoms per million I- ions).lg* 2OIt seems likely that supernumerary S, molecules may be built into the pyritestructure on lattice sites proper to S2,- groups (vide infru, NiS,, CoS,).For a compound to be stable over an appreciable range of composition,certain conditions must evidently be fulfilled.,l The energy expenditure toproduce defects must not be too large; the energy difference between thetwo valency states involved must also be small; the difference in sizebetween the ions in the two valency states must be small, so that the latticemay not be distorted to the point of collapse.In all these respects thecompounds of the heavier metals occupy a special position, and it is amongstthese that marked variations from stoicheiometric simplicity have beenencountered.The Schottky-Wagner theory makes no reference to factors limiting therange of existence of a crystal phase, and is strictly valid only where theconcentration of lattice defects is very small.The tolerance of a crystallattice for excess of its components is actually limited. The positions ofall atoms adjacent to an interstitial atom, lattice hole or ion of highervalency must undergo adjustment, and if the concentration of such latticedisturbances exceeds some limiting value, the crystal lattice may breakup t o give a second phase and a structure " saturated " with defects.Anattempt to include this within the scope of the theory has been made22by considering not only the energy expenditure to produce lattice defects,but also an energy of interaction of defects when adjacent to each other.The effect of this is that the distribution of defects through the crystallattice is no longer completely random. Below a critical temperature1 7 Nature, 1946,157, 764.by, the effects of secondary structure, 6 being much smaller still.18 2. physikal. Chem., 1937, B, 38, 295.19 Cf. R. W. Pohl, Proc. Physical SOC., 1937, 39, Extra part, 1.20 E. Mollwo, Ann. Physik, 1937, 29, 304.2 1 W. I<Iemm, Atti X Cony. int. Chiin., Rome, 1938, 2, 696; Die Chemie, 1943,56, 6.22 J. S. Anderson, Proc. Roy. SOC., 1946, A, 185, 69.These estimates include, and are probably dominateANDERSON : NON-STOIUHEIOYETRIU UOMPOUNDS.109(dependent on the interaction energy) the lattice is stable only when theconcentration of defects is less than a limiting value; if this is exceeded,the non-stoicheiometric phase breaks up into a 2-phase system. As thesaturation concentration of (e.9.) vacant cation holes and interstitial cationswill in general be different, the range of accessible compositions on themetal-poor and the metal-rich side of the ideal formula can differ widely.I n particular, the maximum permitted concentration of interstitial atomscould be less than would correspond to the intrinsic disorder 6 of the latticeof the stoicheiometric compound. The ideal composition would then fallwithin the 2-phase region, and the stoicheiometric compound would beunstable : only the phase with a stoicheiometric excess of non-metal wouldexist.I n a number of well-established instances (see below) this is foundt o be the case. This model is over simplified but reproduces some typicalfeatures of equilibria involving non-stoicheiometric phases.Occurrence of Non-stoicheiometry in Binary Compounds.-Apart fromsome compounds of variable composition long known from their mineraloccurrence (e.g., pyrrhotite), evidence for the existence of non-stoicheio-metric compounds has come principally from studies of phase equilibriain binary systems, and needs critical examination. I n metallic and quasi-metallic systems, the standard methods of thermal analysis may revealthe range of stability of intermediate phases.Where one component isvolatile (e.g., in oxide and sulphide systems), study of the ( p , X ) , equilibriumis convenient, as used in the long series of memoirs from the attingen andHanover schools of W. Biltz.23 If a solid phase has a range of composition,the system is bivariant over the same range; the equilibrium pressurevaries with the composition of the solid phase instead of changing abruptlyfrom one univariant equilibrium to another at the composition of eachstoicheiometric solid compound. However, (p,X) isotherms or ( T , X )isobars of similar shape can result from a completely different cause:formation of the product of a reaction in a stoicheiometric but " active "state-e.g., imperfectly crystallised or having high surface energy owing toits state of subdivision24-whereby false equilibria are set up. This hastoo frequently been overlooked, and conclusions drawn from the shape ofdegradation curves have in several instances (cf.lead and antimony oxides,below) subsequently been found incorrect. Where, as in such systems asNiS-NiS, 25 and CoS-CoS, 26 the experimental measurements have beenmade at temperatures high enough for recrystallisation and true equili-bration (diffusion is appreciable above 0.5 x absolute melting the29 Key papers to the experimental method are : W. Biltz and H. Muller, Z. anorg.Chem., 1927, 163, 257; W. Biltz and R. Juza, ibid., 1930, 190, 162; F. Wiechmann,M. Heimburg, and W. Biltz, ibid., 1939, 240, 129.24 R.Fricke, Maturwws., 1943, 31, 469; Q. F. Huttig and F. Kolbl, 2. anorg. Chem.,1933, 214, 289.25 W. Biltz, A. Voigt, K. Meisel, F. Weibke, and P. Ehrlich, ibid., 1936, 228, 273.1 6 0. Hiilamann and W. Biltz, ibid., 1936, 224, 73.'7 G. Tammann, ibid., 1926, 149, 67110 INOWANIC CHEMISTRY.evidence of ( p , X ) isotherms is significant. Less weight attaches to evidenceobtained similarly for oxide systems. The melting points of oxides areusually so high that “active” states hamper the attainment of realequilibria.X-Ray studies have been widely used alone or in conjunction with( p , X ) p or ( T , X ) , measurements. Where the diffraction lines of the minimumdetectable amount of a new phase must be sought, the tendency is toexaggerate the range of existence of a phase.A narrow, but finite, rangeof existence may equally well be overlooked. Most conclusive, but usedhitherto in too few instances, is the precise measurement of cell dimensions.Occurrence of a non-stoicheiometric phase can generally be unequivocallydetected, and the mode of incorporation of the excess of one componentcan be d e d ~ c e d . ~ ~ 28, 2 9 ~ 30 With paramagnetic or ferromagnetic compoundsof the transition metals, magnetic measurements can be used to determinethe phase 32Minute departures from stoicheiometric balance, below any limit ofanalytical detection, may still be detected through the electronic semi-con-ducting properties they coder on otherwise non-conducting crystals.33 Eachatom of excess metal represents a supernumerary cation + trapped electronin the lattice, and constitutes a filled impurity level from which, by thefluctuations of thermal energy, the electron may be excited to the con-duction band of the crystal.A cation of higher valency, in a crystal withexcess non-metal, is a site of electron deficiency, an empty impurity levelt o which an electron may be excited from the originally filled valency bandof the crystaLN Electrons in the first case, or “positive holes’’ in thesecond case, are thereby rendered mobile, giving rise to electronic con-ductivity but differing in respect of the sign of certain consequential effects(Hall effect, thermoelectric effect) .35 The relation of semiconductingproperties to variability of composition is revealed, for metallic oxides,by the effect of the oxygen pressure.Diminution of oxygen pressureincreases the conductivity of oxides derived from the highest valency stateof a metal (stoicheiometric excess of metal increased), and decreases the(positive hole) conductivity of oxides derived from lower valency states(stoicheiometric excess of oxygen de~reased).~~ Even the most refractoryoxides, like Al,O, and CaO, become metal-excess conductors at high tem-peratures ; 37 colour changes such as those of ZnO, In,O,, CeO, are associatedwith the reversible loss of oxygen atoms from the crystal lattice. C. Wag-28 G. Hagg and G. Soderholm, 2. physikal. Chem., 1935, B, 29, 88.29 H. Haraldsen, 2. anorg. Chem., 1937, 234, 372.W.Klemm and N. Fratini, ibid., 1943, 251, 222.31 H. Haraldsen, ibid., 1937, 231, 78; 1941, 246, 169, 195.32 H. Haraldsen and F. Mehmed, ibid., 1938, 239, 369.33 C. Wagner, 2. physikal. Chem., 1933, B, 22, 181.34 For a general review, cf. F. Seitz, J. AppZ. Physics, 1945, 16, 553.35 For a general review, cf. R. J. Maurer, ibid., p. 563.86 E . Friederich, 2. Physik, 1925, 31, 813; W. Meyer, ibid., 1933, 85, 278.97 W. Hartmann, ibid., 1936, 102, 709ANDERSON : NON-STOIUHEIOMETRIO OOMPOUNDS. 111ner38 has sought to follow the ( p , X ) equilibria in the ZnO and the Cu,Osystems on the basis of plausible assumptions as to the relation betweenconductivity and stoicheiometric excess, but his assumptions are in doubtfulaccord with the whole range of experimental facts.39The following survey is not exhaustive, and includes only those com-pounds for which explicit evidence has been cited ; intermetallic compoundsare omitted.Non-stoicheiometric phases are indicated (cf. Klemm) 21 bya bar above the idealised formula.(I) Nydrides.-The interstitial semimetallic hydrides of Zr, Th, Ta andthe rare-earth metals approach stoicheiometric compositions only as anupper limit of hydrogen content .40 The essentially Berthollide Pd-Hequilibria were discussed from the statistical thermodynamic viewpoint byJ. R. Lacher,4l and it has recently been shown4, that the Zr-H system hassimilar characteristics when the complicating effect of oxygen, present ininterstitial solution in the metal-which vitiated much of Sieverts’s work-is avoided.The complete equilibria might be interpreted along the linesindicated in ref. (22).43(11) Su.lphides, Selenides, Tellurides.-The table collects data for MX andMX, compounds of the first transition series, but omits the (quasi-metallic)subsulphides, etc., some of which (e.g., pentlandite Ni,S,44 and the Co4S3phase45) undoubtedly have a range of existence. In every case, the MXcompounds with NiAs type structure exist over a wide range of composition(contrast MnS and the low-temperature forms of FeSe and NiS) throughthe omission of cations from the structure. In some cases at least (e.g.,C-) the stoicheiometric compound’is unstable. In the light of the theor-etical discussion (equation 5 ) it is important that the proportion 6 of vacantsites of both kinds in the stoicheiometric phase may apparently be as highas 5-’7%.49 The complete transition between VSe and VSe,, CoTe andCoTe,, NiTe and NiTe, is particularly noteworthy.The NiAs and theCd1,structures are so related that the former is transformed into the latterby the ordered omission of half the cations. Thus, in the V-Se system,38 H. H. von Baumbach and C. Wagner, 2. physikal. Chem., 1933, B, 22, 199;H. Diinwald and C. Wagner, ibid., p. 212; J. Gundermann, K. Hauffe, and C. Wagner,ibid., 1937, By 37, 148.39 Cf. B. Gudden, Ergebn. exakt. Naturwws., 1934, 13, 222; J. S. Anderson andM. C. Morton, Proc. Roy. SOC., 1945, A , 184, 83; Trans. Faraday SOC., in the press.4O A. Sieverts et al., 2. anorg.Chem., 1926, 153, 289; 1930, 187, 155; 1928, 172,1 ; 1931,199, 384.dl Proc. Roy. SOC., 1937, A , 161, 525.42 M. N. A. Hall, S. L. H. Martin, and A. L. G. Rees. Trans. Paraday SOC., 1945,43 Dr. A. L. G. Rees, private communication.44 Cf. J. E. Hawley, G. L. Colgrove, and H. F. Zurbrigg, Econ. Geol., 1943, 38, 335.46 0. Hulsmann and F. Weibke, 2. anorg. Chem., 1936, 227, 113.46 W. Biltz, P. Ehrlich, and K. Meisel, ibid., 1937, 234, 97.4 7 W. Klemm and E. Hoschek, 2. anorg. Chem., 1939, 242, 49.48 W. Biltz and A. Kocher, ibid., 1939, 241, 324.49 H. Haraldaen, ibid., 1937, 234, 372; H. Haraldsen and A. Neuber, ibid., p. 337.41. 306TiVCrMnFecoNiCorn- Struc-pound. ture.TiSTIS ,. 6TSa C6VS (a) BS m., (8)MnS (a) B1 $4 B3MnS2FTS B8S S , C2, C18G S B8CoSl.3a H11cos, c2 -NiS B13E S B8NiS, c2n.1.0-1*11.1-1.51.5-2.0 *1*0-1*161.17-1.531.0-1.1 71.22-1.481.002-001 SO-1 * 141.95-2'051.05-1.251.3331 * 9- > 2 -01 *oo1.0-1.22->3Ref.4647484951525354526564558Compounds MX, of Transition Metals.Corn- Struc-pound.ture. n.V S e (a)SS 0.98-1-2Vs.6 (p)M 1.2-1.6VSe, ( y ) C6 1.6-2.0C x e (a) B8 1.0-1.15CrSe,.,, (8) M 1.20-1.33CrSe,15 ( y ) H 1.44-1.50FeSe T 1 -00FeSe, C18 ?FTSG (a) B8 1.0-1.13(8) M 1.13-1.31case, c 2 ?NiSe, C2 ?Ref.473265757* At high temperatures; range of existence narrower a t231, NaCl type. B3, ZnS type. B8, NiAs type. B13, trigonal millerite type.C6,tme. H11, spinel type. M , H , T, monoclinic, hexagonal, and tetragonal phases of otheANDERSON : NON-STOICHEIOMETRIC COMPOUNDS. 113the a-phase is stable with up to 17% of the cation sites vacant, the " holes "being distributed at random. A corresponding proportion of V2+ ions isreplaced by V3+ or V4+. Further increase in the concentration of cationholes initiates an ordering process which lowers the crystal symmetry@-phase). Finally, with 38--50% of the original cation sites vacant, theholes are segregated largely into alternate cation sheets of the originalstructure. Another hexagonal, Cd1,-type structure results, ideally VSe,,but including up to 20% extra cations (partial replacement of V4+ by V2').Only in exceptionally favourable cases can the range of existence be as wideas this, and the gaps of miscibility as narrow, but a similar sequence ofchanges is met with in some oxide systems.The high-temperature modifications of these, witha random distribution of cations,60 are stable with a wide range of cationdeficiency ; certainly up to Cul.,S, Cul.,Se, Cu,.,Te.Certain properties-conductivity, self-diffusion, etc.-have a maximum value close to thecomposition Cu1.,X.61, 6, The C q phase is of considerable mineralogicalinterest; N. W. Buerger G3 considers that cubic chalcocite has the idealcomposition Cul.,S (i.e., 10% of cations missing) and is distinct from theCu,S phase proper. CuS does not appear t o have a measurable range ofexistence.64 As befits the instability of higher valency states of silver, Ag,Sis not stable over any wide range of composition, but does take up a measur-able excess of sulphur.65 Equilibrium compositions a t 300" areVapour pressure of S, mm................... 0 0-6 5.2 21Composition.. ..................................... . Ag2*000S Agl'BB92S Agl*BB7S Agl*9B6S-I____ Cu,S, Cu,Se, Cu,Te.(111) Arsenides, etc.-Transition metals form compounds MX, MX,,with NiAs and (mostly) marcasite structures respectively ; few systemshave been investigated thoroughly. Lollingite, FeAsz, usually contains an60 H. Haraldsen and A. Neuber, 2. anorp. Chem., 234, 353.61 W. Biltz and F. Wiechmann, ibid., 1936, 228, 268.62 R. Juza and W. Biltz, ibid., 1932, 205, 275; H. S. Roberts, J. Amr. Chem. Soc..1935, 57, 1034; H.Haraldsen, 2. anorg. Chem., 1937, 231, 78; 1941, 246, 169, 195;2. Elektrochem., 1939, 45, 370; E. Jensen, Amer. J. Sci., 1942, 240, 695; J. J. Lukes,C. F. Prutton, and D. Turnbull, J . Arner. Chem. Soc., 1945, 67, 697.63 Ref. (11).64 F. G. Smith, Amer. M i n . , 1942, 27, 1.65 A. Oftedal, 2. physikal. Chem., 1928, 132, 208.6 8 H. Haraldsen, 2. anorg. Chem., 1935, 224, 8 5 ; M. Heimbrecht, W. Biltz, and67 S. TengnBr, ibid., 1938, 239, 127. m Ref. (30).6o P. Rahlfs, 2,physikaZ. Chem., 1936, B, 31, 157.61 H. Reinhold and H. Mohring, ibid., 1937, B, 38, 221 ; H. Reinhold and H. Seidel,62 H. Reinhold and H. Brauninger, ibid., 1938, B, 41, 397.G3 J . Chem. Physics, 1939, 7 , 1067; Econ. Beol., 1941, 36, 19.64 A. M. Bateman, {bid., 1932, 27, 62; R.Juza and W. Biltz, 2. anorg. Chem.,6s H. Reinhold and K. Schmitt, 2. physikd. Chem., 1939, B, 44, 76.K. Meisel, ibid., 1939, 242, 229.68 Ref. (25).ibid., p . 245.1930, 190, 161114 INORQANIC CHEMISTRY.excess of iron.66 FeSb exists a t the ordinary temperature only over therange Fe,.,,Sb to Fel.,,Sb,67 and the maximum melting point correspondsroughly to Fel.,5Sb.68 The ideal formula, however, is established by theNiAs lattice type. NiSb 69 and NiBi 7O similarly have an existence rangewith excess of metal. Stoicheiometric FeSi, is non-existent; 7 1 the phaseof maximum melting point has a 20% deficiency of cations, due possiblyto substitutional solid solution.72,(IV) Oxides.-Non-stoicheiometric phases have been reported in thesystems listed below but, as will be evident, finality has not been reachedin a number of instances.Possible reasons for over-estimating ranges ofexistence have already been indicated.Titanium. 73 a-Phase, TiO,, TiO, .oo-TiO,.,, ; p-phase, lower symmetrythan rutile, TiOl.80-TiOl.70 ; y-phase, Ti203, corundum type, TiOl.56-TiO,.,, ; &phase, TYO, NaCl type, Tiol.35-TiOo.6 ; in addition, the metaltakes up about 42 atoms yo of oxygen in interstitial solid s0lution.~4 TheTX phase is of interest as showing how, in'a structure with a very highdegree of Schottky lattice disorder, stoicheiometric variation arises fromthe unbalance of anion and cation holes.--__Composition. Ti01.33. TiOl.12. TiOl.oo. TiOo.6s.Lattice sites occupied : Ti, yo ............74 81 85 960, Yo ............... 98 91 85 66Zirconium. Oxides have not yet been investigated, but the metaltakes up oxygen interstitially to a t least Zr0,.,.75Vanadium. V203, corundum type, extends from VO,.,, to VO,.,approx. ; vG NaCl type, from VO,., to VO,.,, though the range of existenceis much narrower a t low temperatures.',Oxides have very limited ranges of composition (cf. the lessready variability of valency of Nb as compared with V). Nb,O, has aprobable range NbO,.,-NbO,., ; NbO,, NbO (cf. T O and m) no detectableranges, though the NbO structure is of a unique defect lattice type.77Results for chromium oxides are conflicting, and need revision.A. Cameron, E. H. Harbard, and A. King 78 found bivariant equilibria in the6 6 L.H. Bauer and H. Bermann, Amer. Min., 1927, 12, 39; M. J. Buerger, ibid.,1934, 19, 37.8 7 Ref. (9).0 8 R. Vogel and W. Dannohl, Arch. Eisenhiittenw., 1934, 8, 39.69 E. 8. Makarov, Ann. Sect. d'Anal. Phys.-Chim., 1943, 16, No. 1 ; A . , 1943, I, 15.70 G. HZigg and G . Funke, 2. physikal. Chem., 1929, B, 6, 272.7 1 G. Phragmen, J . Iron SteeE Inst., 1926, 114, 397; M. Bamberger, 0. Einerl,72 J. L. Haughton and M. L. Becker, J . Iron Steel Inst., 1930, 121, 315.v 3 P. Ehrlich, 2. Elektrochem., 1939, 45, 362.74 Idem, 2. anorg. Chern., 1941, 247, 53.75 J. H. de Boer and J. Fast, Rec. Trav. chim., 1940, 59, 161.7 6 W. Klemm and L. Grimm, 2. anorg. Chem., 1942,250, 42.7 7 G. Brauer, ibid., 1941, 248, 1.Niobium.and J. Nussbaum, Stahl u.Eisen, 1925, 45, 141ANDERSON : NON-STOICHEIOMETRIC COMPOUNDS. 115ranges Cr01.7-Cr01.9, C~O,.,-CTO,.~, covering the range of complex oxidesreported by earlier workers.79 A. Michel and J. B6nard,so however, donot find these phases, but report that Cr203 has an upper limit of com-position about Cr01.56.Lower oxides of molybdenum have long been a matter of disagreement.According to G . Hagg and A. Magnhli,81 the system is similar to that of thetungsten oxides, with p- and p'-phases, roughly MoO~.,,-MOO~.,, ; y-phaseMoO,.,-MOO~.~, ; &phase MOO,.0. Glemser and H. Sauer 82 find : a-phase W0,-WO,.,,;p-phase W02.,2-W0,.88 ; y-phase W0,.,,-W0,.6, (W,O1,, with X-raydiagram identical with the W,O,, of F. Ebert and H. Flasch); 83 %phaseWO,.O,-WO,.,.Hagg and Magnhli substantially confirm these results.81Closely related are the interesting tungsten bronzes, Na,WO,, etc., of whichthe stoicheiometric compounds (x = 1) apparently do not exist; the cubicsodium bronzes (x = 0.95-0.30) are of defective perowskite type, givingplace (for x = 0.3-0.2) to structures of lower symmetry.84 The tungstenblues, and the hydrogen-containing compounds studied by Ebert andFlasch and by 0. Glemser and H. Sauer 82, 85 appear similar inconstitution.Uranium. It seems clearly established that a t elevated temperaturesstoicheiometric UO, is unstable, and the U,08 phase is of widely variablecomposition.86Manganese. Numerous oxides intermediate between Mn20, and MnO,have been reported, but their individuality is questionable : if they are notmixtures, a non-stoicheiometric phase seems likely.A. Simon and F. Feller 87inferred the existence of such a phase from tensimetric studies, but otherworkers 88 appear agreed that pyrolusite has, at the most, only a smallcomposition range, although it usually contains less oxygen than corre-sponds to MnO,.oo. However, at least three modifications of MnO, appearto e x i ~ t . 8 ~ ~ ~ 0 Wet methods of preparation [e.g., oxidation of Mn(OH), orMnO-OH] can produce hydrous oxides of variable composition (but fewstructural defects) through double substitution of Mn3+ for Mn4+ and OH-7 8 J., 1939, 55; S. S. Bhatnagar, A. Cameron, E. H. Harbard, P. D. Kapur,79 Cf. A. Simon and T. Schmidt, 2. anorg. Chem., 1926, 153, 191.Tungsten.A.King, and B. Prakash, J., 1939, 1433.Bull. Xoc. chim., 1943, 10, 315.Arkiv Kemi M i n . Geol., 1944, 19, A , No. 2; A., 1946, I, 144.82 2. anorg. Chem., 1943, 252, 144.83 Ibid., 1934, 217, 95; 1935, 226, 65.8 p G. Hiigg, Nature, 1935, 135, 874; 2. physikal. Chem., 1935, B, 29, 192.s5 2. anorg. Chem., 1943, 252, 160.8 7 2. Elektrochem., 1932, 38, 137.W. Biltz and €1. Muller, ibid., 1927, 163, 257.M. Le Blanc and G. Wehner, 2. physikal. Chem., 1934, A , 168, 59; C. Druckerand R. Huttner, ibid., 1928, 131, 237; P. Dubois, Ann. Ckirn., 1936, 5, 411.89 P. Dubois, loc. cit., ref. (88); 0. Glemser, Ber., 1939, 72, 1879.90 W. F. Cole, A. D. Wadsley, and A. Wslkley, private communication116 INORGANIC CHEMISTRY.for 02-.90, 91 It is reported that one modification of Mn203 takes up excessoxygen to MnO,.,, a t least.92Iron.Two ranges of non-stoicheiometric oxides are of importance inmetallurgy. (I) R. Schenck and T. DingmannS3 first reported that theFeO (wiistite) phase, stable only above 580", invariably contains a stoicheio-metric excess of oxygen which represents an excess of vacant cation sites.8At 1400" the range of composition extends from FeO,.,, to Fe0,.,9.94Although his interpretation cannot be accepted, the work of J. BBnardS5has shown that in the oxidation of iron a t high temperatures the primaryproduct is the K O phase, with a continuous composition gradient from theiron-rich limit to the oxygen-rich limit. A means is thereby provided fora continuous diffusion of iron through the oxide film to the FeO-02 inter-face, as envisaged by K.F i s ~ h b e c k . ~ ~ (11) G. Haggg7 showed that Fe30,and the y-Fe,03 defective spinel structure represented the limits of onephase of variable composition. At high temperatures, where y-Fe,03 isunstable, the phase relations are still uncertain. R. C. Sosman and J. C.Hostetter9* concluded that a-Fe203 and Fe304 had extended ranges ofexistence towards lower and higher oxygen contents respectively, andJ. C. White g9 appears to confirm this substantially. Later work by Sosmanet d 1 O 0 indicated that neither Fe203 nor Fe,O, was appreciably variable incomposition.Cobalt. M. Le Blanc and E. Mobius lol and M. Watanabe lo2 report thatCOO and Co304 can each take up a substantial stoicheiometric excess ofoxygen.Black NiO was found by M.Le Blanc and H. Sachse lo3 tocontain an excess of oxygen, although a homogeneous phase. Accordingto W. Klemm and E. Haw, stoicheiometric NiO is metastable, breaking upinto NiO,., + Ni.lo4CuO is apparently stoicheiometric, but the much studied semi-conducting properties of Cu,O depend on a small excess of oxygen, whichis present in true equilibrium with the gaseous phase a t high temperatures,as envisaged by theory. Measurements by C. Wagner and H. Hammen lo5give, in equilibrium with 0.7 mm. of 0, at lOOO", C~2Ol.000~2; with 33 mm.Nickel.Copper.of 02, C~201-rn114-91 W. Feitknecht and W. Marti, Helv. Chim. Acta, 1945, 28, 129, 149.92 ( a ) M.'Ulumenthal, Bull. SOC.chim., 1933, 53, 1418; ( b ) C. B. Holtermann, Ann.O3 2. anorg. Chem., 1927, 166, 113.94 L. S. Darken and R. W. Gurry, J . Amer. Chem. Soc., 1945,67, 1398.g5 Ann, Chim., 1939, 12, 5 ; Compt. rend., 1943, 217, 7 7 .9 7 2. pihysikal. Chem., 1935, B , 29, 95.89 Iron and Steel Inst., Carnegie Schol. Mem., 1938, 27, 1.100 Amer. J . Sci., 1935, 30, 239.102 Sci. Rep. Tdhoku, 1934, 23, 89; A., 1934, 699.Io3 2. Elektrochem., 1926, 32, 68, 204.m4 2. anorg. Chem., 1934, 219, 82.Chim., 1940, 14, 121.2. Metallk., 1932, 24, 313; Metallwirts., 1935, 14, 733.98 J . Amer. Chem. SOC., 1916, 38, 807.lol 2. physikal. Chern., 1929, A , 142, 151.lo6 2. phyeikaE, Chem., 1938, B, a, 197ANDERSON : NON-STOICHEIOMETRIC COMPOUNDS. 117Zinc. Although the composition of ZnO is not measurably variable,its semiconducting properties show that the colour change on heating isassociated with a loss of oxygen and the presence of a minute excess ofzinc.lo6 The red ZnO obtained by A.Kutzelnigg lo7 has been shown byA. Greenstone and W. Ehret lo8 to contain up to 0.02% excess of zinc, butmay well be thermodynamically highly unstable. Cadmium oxide probablyhas at least a similar range of composition,10s and the reversible colourchanges of other oxides (In,03, CeO,) can probably be interpreted similarly.Finality as to the oxides of lead has certainly not yet been reached. Pb304seems to be a closely stoicheiometric compound, but M. Le Blanc andE. Eberius concluded that PbO, PbO,, and another intermediate oxidewere all non-stoicheiometric.It now appears ll1 that PbO, has only anarrow range, perhaps from PbO,.,, to PbO,.,; it is probably not obtain-able without constitutional water and the limited stoicheiometric variabilitycould arise from replacement of 202- by 20H-, Pb4+ by Pb2+, in the idealstructure (.cf. MnO,). By degradation of PbO,, or by reaction of PbO withoxygen, two definitely non-stoicheiometric intermediate oxides may beformed, but there is no agreement as t o their nature. Bystrom’s a-PbO,with the range Pb01.,-Pb01.,7 may correspond with the non-stoicheio-metric 112 or stoicheiometric 113 Pb,O, or Pb701, 92b phases of other workers(although there is no agreement as to the symmetry of this phase). Bystrom’sP-PbO, (Pb01.4,-Pb01.,1) probably corresponds to G.L. Clark andR. Rowan’s PbO,. However, the discrepancies between different workersare not to be reconciled.The oxides provide an instructive instance of false conclu-sions drawn from tensimetric measurements, indicative of a phase of con-tinuous composition between Sb203 and Sb,O, 114 and apparently confirmedby X-ray measurernents.ll5 Later work 116 has put a completely differentAntimony.interpretation on the facts, and provides no evidence of stoicheiometricallyvariable antimony oxides.(V) Halides.-The electrical conductivity of cuprous iodide is stronglydependent on the pressure of iodine in equilibrium with the solid compound.This, as first shown by K. Badeker,l17 takes up a stoicheiometric excess of1013 Ref.(38).108 J . Amer. Chern. SOC., 1943, 65, 872.lo* R. Faivre, Ann. Chim., 1944, 19, 58; H. H. v. Baumbach and C. \Va,nner,ll1 A. Bystrom, Arkiv Kemi Min. Qeol., 1945, 20, A , No. 11;112 G. L. Clark and R. Rowan, J . Amer. Chem. SOC., 1941,63, 1305.113 F. Fischer and H. Ploetze, 2. anorg. Chem., 1912, 75, 1.114 A. Simon and E. ThaIer, ibid., 1927, 162, 253.116 U. Dehlinger, 2. physikal. Chem., 1929, B, 6, 127; U. Dehlinger and R. Glocker,116 K. Dihlatrom and A. Westgren, ibicl., 1937, 235, 153; K. Dihlstrom, ibicl., 1938,lL7 Ann. Physik, 1907, 22, 749; 1909, 29, 566; Physikal. Z . , 1908, 9, 431; 1912,1°7 2. anorg. Chem., 1932, 208, 23; 1934, 221, 116.2. physikal. Chem., 1933, B, 22, 199. 110 Ibid., 1932, A , 160, 60.2. anorg. Chem., 1927, 165, 41.239, 51.13, 1080; K. Nagel and C.Wagner, 2. physikal. Chem., 1933, B, 25, 71118 INORGANIC CHEMISTRY,iodine, up to the composition Cu11.M)45. The CUI system is one of the fewfor which ( p , T , X ) equilibrium data can be correlated properly with measure-ments of semiconducting properties.ll8(VI) Ternary Compounds.-Distinction between non-stoicheiometriccompounds and mixed-crystal phases is here more arbitrary, since in additionto subtractive and interstitial types of solid solution, there is the possibilityof " anomalous " solid solutions-also involving the creation of vacantlattice sites or interstitial atoms-f the kind exemplified by the y-A120,-MgA120, phase,28 and the defective fluorite-type solid solutions studied byZintl et al.l19 However, certain classes of ternary compound have beendescribed which are inherently non-stoicheiometric, e.g., the tungstenbronzes already mentioned.84 Sillkn and his co-workers have recentlydescribed a number of double oxides and oxy-halides of bismuth withbivalent metals in which, by variation in the M3' : M2+ cation ratio, either(i) the cation lattice remains complete, but a variable proportion of oxygensites is vacant, or (ii) the anion lattice is perfect, but the number of cationsin the structure is variable. Rational formulz cannot always be assigned to" idealised " compounds.cadmium and the alkaline earths.122 The oxyhalides of bismuth with cal-cium,lZ3 cadmium,124 and other bivalent metals are of type (ii), and exemplifysome very interesting structural principles ; the inherently non-stoicheio-metric phases M~~2-wBil+2z02X3, M112-3zBi3+2z04X5, MI12 -3zBi5+2z06X7(X = C1, Br) have been described.According to C. Brosset,125 potas-sium cryolite, ideally K3A1F5, may vary in composition through replace-ment of AIF,3- groups by CAW5( H20)l2- groups, with corresponding omissionof K+ cations (up to 3%) from the structure. A range of homogeneity hasalso been assigned to the alkali tantalates and niobates.126 It is likelya priori that the ternary sulphides, etc., will be variable in composition, butfew systems have been closely studied. Chalcopyrite appears definitelynon-stoicheiometric, with the limiting composition CuFeS,.,,. 127R6le of Non-stoicheiometric Phases in the Reactions of Solids.-Reactionsbetween solids, or between solid and fluid substances, take place a t theinterface between the reactants.Transport of reactant to this interfacemust take place, in general, by diffusion through the solid product of reac-tion, and this may be the rate-determining process in the reaction. TheOf type (i) are the double oxides iMI12zBi2 - 2z03 ---f2.g., Pb1.2Bi0.802.40 to Pbo.64Bil.3602.68,120 and analOgOUS compounds Of118 R. J. Maurer, J . Chem. Physics, 1945, 13, 321.119 2. anorg. Chem.; 1939, 240, 145, 150; 1939, 242, 79.120 L. G. Sillen and B. Aurivillois, Naturwiss., 1939, 27, 388; 2. Krist., 1939, 101,n1 L. G. Sillen and B. Sill&, 2. physikal. Chem., 1941, B, 49, 27.lZ2 B. Aurivillois, Arkiv Kemi Min.Cfeol., 1943, 16, A , No. 17.lZ3 L. G. Sillen and A. S. Gjorling-Husberg, 2. anorg. Chem., 1941, 248, 121, 135.lZ4 L. G . Sill&, ibid., 1941, 248, 331.125 A ~ k i v Kcini Min. Geol., 1946, 21, A , No. 9.lZ6 F. Halla, A. Neth, and F. Windmaisser, 2. Krist., 1942, 104, 161.127 H. E. Merwin and R. H. Lombard, Econ. Geol., 1937, 32. 203.483ANDERSON : NON-STOICHEIOMETRIC COMPOUNDS. 119mechanism of diffusion, and of ionic conduction, in polar solids can beinterpreted in terms of the presence and migration of lattice defects,33 anddeparture from stoicheiometry, by controlling the concentration of inter-stitial ions or vacant sites, affects the diffusion coefficient D. This appliesboth to ionic conductivity or diffusion along a concentration gradient andto self-diffusion, 128 as has recently been shown e~perimentally.~~~ Whereonly one ion (most frequently the cation) is mobile, D is a minimum for thestoicheiometric crystal, in the case of a crystal with Frenkel defects, orincreases monotonically with stoicheiometric excess of non-metal for thecase of Schottky defects.I n reactions, stoicheiometric variations enable acomposition gradient to be set up through the layer of reaction product.C. Wagner 130 has derived a quantitative theory for " tarnish " reactions,which proceed by continuous migration of cations t o the solid-gas interface[cf. ref. (96)], and has extended it to include reactions between solids-e.g.,double salt, spinel, and silicate f0rmati0n.l~~ The acceleration of suchprocesses by conditions producing small deviations from stoicheiometry hasbeen demonstrated for the formation of MgA1204.132 Both Al,O, and MgOare metal-excess conductors a t high temperatures, and their union proceedsmarkedly faster in vacuum or in hydrogen than in air.Such factors mayhave considerable significance in ceramic processes.133There is some evidence that non-stoicheiometric phases can be formedunder non-equilibrium conditions, as intermediate stages in the formationor dissociation of solid compounds. For instance, brucite formed by slowoxidation of magnesium in moist oxygen,lM or partially dissociated silveroxide 135 is stated to contain an excess of metal. Analogous cases are onrecord. lo7$ l36In a broad sense, the non-stoicheiometric character of a solid may beassociated with the mechanism of catalysis in heterogeneous reactions.Thus, C.Wagner and K. Hauffe 137 have deduced the rate-determining stepin the nickel-oxide-catalysed reactions 2CO + 0, = 2CO,, 2N,O = 2N2 + O,,from the composition of the oxide catalyst (as shown by its electronic con-ductivity) in the stationary state. Similar observations have been madefor the H, + S + H2S reaction catalysed by silver suiphide,138 andthe mechanism of the catalysed water-gas reaction 139 and the ammonia128 C. Wagner, 2. physikal. Chem., 1931, Bodenstein Festb., 177.12@ J. S. Anderson and J. R. Richards, J., 1946, 537.l30 Z.physika1. Chem., 1933, B, 21,25; 1936, B,32,447; Angew. Chem., 1936,49,735.131 C. Wagner, 2.physikal. Chem., 1936, B, 34, 309, 317.132 H. C. Castel, S. Dilnot and M. Warrington, Nature, 1944, 153, 653.133 Cf. J. A. Hedvall, Die Chemie, 1942, 55, 334; Trans. Chalmers Uniu. Technology,134 R. Faivre and A. Michel, Compt. rend., 1939, 208, 1008.135 R. Faivre, ibid., 1940, 210, 398.136 E. I. Mokeeva and N. I. Mokeeva, J . Physical Chem. RUSS., 1941, 15, 686.137 2. Electrochem., 1938, 44, 172.138 H. Reinhold, W. Appel, and P. Frisch, 2. physikal. Chem., 1939, A , 184, 273.13* E. Doehlmann, 2. Elektrochem., 1938, 44, 178.Goteborg, 1942, No. 15120 INORGANIC CHEMISTRY.synthesis have been discussed from a similar standpoint. A further influenceupon the surface properties of solids is shown by the dependence of adsorptiveproperties of metallic sulphides upon small variations of stoicheiometriccomposition.140J.S. A.2. COMPLEX COMPOUNDS OF THE PLATINUM METALS.it was observed that platinum metal complexes have recentlyreceived special attention particularly by Russian workers, but it wasfound necessary to defer discussion of their work to a later report, and itis this work which forms the bulk of the present review.In 1944Platinum.OZe$n Complexes.-The complexes formed by metallic salts and olefinswere discovered before 1830 but no satisfactory structure has yet beenassigned to them. Since the last review of this subject in 1936 a com-prehensive account of these compounds by R. N. Keller and a considerablevolume of work by Hel’man and his co-workers have been published.Platinum salts form the best-known and most stable complexes, so workhas been limited almost entirely to the platinum series.Palladium com-plexes are less stable and recent attempts to obtain cobalt and nickel com-plexes were unsucce~sful.~ Typical members of the series are K[Pt C2H4 GI3],[(Pt C,H4 Cl,),], [Pt C,H, py CI,] and the most recently added member[Pt C2H4 NH, py C1]N0,.5 The anionic complex is very much more stablethan the cationic complex, and attempts to obtain two mono-olefins attachedto one platinum atom have so far failed.6, l 6Unsaturated molecules behave similarly to ammonia and occupy onlyone co-ordination place round the platinum atom, but they differ fromammonia and pyridine in their directing influence on substituents enteringa complex which already contains an olefin.They labilise the group inthe trans-position so that, in preference to the cis-groups, it is replaced bythe entering substituent. This difference is illustrated by comparing thereactions (A) and (B) with the analogous reactions (C) and (D).’, *Whilst the products (I) and (11) are identical, but different from Jor-gensen’s [Pt(NH,) py Cl,], products (111) and (IV) are isomeric. (Ij, (11),and (111) are claimed to be cis-isomers, but (IV) is claimed to be the truns-l40 J. A. Hedvall and S . Nord, 2. Elektrochem., 1943, 49, 467.1 Ann. Reports, 1944, 41, 98. a W. C. Zeise, Mag. Pharm., 1830, 35, 105.Chern. Reviews, 1941, 28, 229.A. D. Hel’man and I. B. Litvak, Ann. Secteur phtine, Inst. chim. gin.(U.S.S.R.),ti A. D. Hel’man and E. A. Meilakh, Compt. rend. Acad. Sci. U.R.S.S., 1946,51, 207.’ I. I. Chernyaev and A. D. Hel’man, Ann. Secteurplatine, I n s t . chim. ge’n. (U.S.S.H.),1039, 16, 29.A. D. Hel’man, ibid., 1939, 23, 532.1935, 15, 5.A. D. Hel’man, Compt. rend. Acad. Sci. U.R.S.S., 1939, 22, 107.S. M. Jorgensen, J . pr. Chem., 1886, 33, 489CHATT: COMl’LEX COMPOUNDS OF THE PLATINUM METALS. 121isomer, and the analogously prepared pyridine complex [Pt(C,H,) py Cl,]reacts with pyridine to give trans-[Pt py2 C1,].lo(B) K[Pt py + NH3 = [Pt py (NH3) C121 + KCl(11.)(c) K[Pt(NH3)C&] + c2H4 = [Pt(NH3)(C,H&C&I KC1(111.)(N.)(D) qPt(C2H4)C13] -k NH3 = [Pt(C2H&(NH3)C1,] + KC1Reactions (C) and (D) have been shown to be general and have beenapplied to obtain similar isomers containing other olefins and carbon mon-further adaption isomers containing four different groups attached to theplatinum atom have been obtained, e.g., [Pt(C2H,)(P\TH3)C1Br].13It was shown by J.S. Anderson 14* 15 that the stability of the ethylenecomplexes of PtCl, was altered markedly by substitution of other univalentradicals in place of chlorine and also by substitution in the ethylene moleculeitself. This work has been repeated and extended by Russian workers,who have found that the amine substituent in [Pt(C,H,) am Cl,] causes adecrease of stability in the order : lo am = quinoline > pyridine > am-monia > thiourea. They confirm that the stability decreases as chlorineis replaced in the order : C1 > Br > I > NO, > CNS > CN, but differfrom Anderson in placing styrene higher than ethylene in the stability ofits complexes.They find the following order of stabilities : l 6 ~ l7 NO >CO > styrene > butadiene - C,H, > C3H6 - C,H,.This work is somewhat qualitative as no allowance is made for therelative volatilities and solubilities of the various olefins. The replacementsof unsaturated molecules were effected by reaction of the appropriateunsaturated substance Un with a dilute acid solution of the salt K[PtUn’C13],then trying the reverse replacement Un‘ into K[PtUnCl,]. Sometimes i twas found that both replacements occurred, e.g., with propylene andbutylene.Besides the above series of stabilities two interesting facts emerged ;when attempts were made to replace CO by NO in py H[Pt(CO)Cl,] in veryacid solution by passing NO through it for two months, the platinum wasoxidised and (py H),[PtC16] was isolated ; and also the reaction of ethylenewith the [PtCl,]“ ion in dilute acid solution is catalysed by propylene.1710 I.I. Chernyaev and A. D. Hel’man, Ann. Secfeur platine, Inst. chim. gdn. (U.S.S.H.),1937, 14, 77.11 A. D. Hel’man, Con@. rend. Acad. Sci. U.R.S.S., 1937, 16, 351.l2 A. D. Hel’man and M. Bauman, ibid., 1938, 18, 645.l3 A. D. Hel’man, ibid., 1943, 38, 310.l8 A. D. Hel’man, Compt. rend. Acad. Sci. U.R.S.X., 1041, 32, 347.17 A. D. Hel’man, ibid., 1938, 20, 307.oxide instead of ethylene as well as bromine instead of chlorine.l’, l2 BYl4 J., 1934, 971.15 J., 1936, 1042122 INORGANIC CHEMISTRY.By passage of a mixture of propylene and ethylene into an acid solution ofpotassium chloroplatinite for four days, a 52% yield of Zeise’s salt free fromthe propylene salt was obtained, whereas pure ethylene in the same timewould have given only a 15:L yield. It is suggested that the greatersolubility of propylene leads to a more rapid reaction with the [PtCl,]- ion,yielding [Pt (C,H,)Cl,]- as intermediate from which the propylene is rapidlyevicted by the ethylene.It is interesting that, although CO readily replaces all the olefins fromions of the type [Pt Un Cl,]-, the [Pt(CO)Cl,]- ion produced is comparativelyunstable, being decomposed by water except in strongly acid solution.Onthe other hand, the NO complexes are exceptionally stable, yet NO replacesthe olefins only very slowly.It appears that the diolefins, butadiene and diallyl, do not form chelatecomplexes but each double bond reacts with a different platinum atom.6316Attempts to obtain chelate compounds by using ethylenediamine ledto no greater success : l82K[Pt(C,H4 )C&I + C2H4(NH2 )2 [{ (C2H4 )PtC12 ,NHz*CHz*}2]The structure of these complex compounds remains unsolved.* Hel’man 19has suggested that the ethylene molecule undergoes an electromeric changein the presence of the platinum-containing ion, and the carbon atom with adeficiency of electrons accepts two electrons from the platinum atom, pre-sumably from a 5d orbital, thus raising the platinum to the platinic state.The other carbon atom now donates its electrons to the platinum atom,forming a four-electron bond between the ethylene molecule and the platinumatom.In support of this the electrometric titrations of K[Pt(NH,)Cl,],NH, [ Pt ( C2H4) Cl,] , K2 [ Pt C1, ( C4H6) Pt Cl,] , and NH,[Pt (NH,) C1 5] were com -pared in acid solution using 0-1x-permanganate to effect the oxidation.Oxidation occurred only when K[Pt(NH,)Cl,] was titrated ; 2o also theinitial potential of the solution of K[Pt(C,H,)Cl,] was 650 mv., and of theabove butadiene analogue 700 mv. as compared with 660 mv. forNH,[Pt(NH,)Cl,] and 520 mv. for NH,[Pt(NH,)Cl,]. These results supportthe suggestion that platinum in the olefin complexes is in the platinic state,but this point deserves further investigation for it would appear that we18 Doklady Akad.Nauk. S.S.S.R., 1943, 38, 272.19 A. D. Hel’man, Compt. rend. Acad. Sci. U.R.X.S., 1939, 24, 549.20 A. D. Hel’man and D. I. Ryabchikov, ibid., 1941, 33, 462.* Since this report was completed, A. D. Walsh (Nature, 1947, 159, 165; J . , 1947,89) has pointed out that the ionisation potential of the ?T electrons of ethylene is 10.45 v.as against 10.8 v. for the lone pair electrons of ammonia and suggested that the n elec-trons should be capable of donation to suitable atoms or groups thus binding togetherthree nuclei. This view is closely allied to that of Winstein and Lucas ( J . Amer. Chem.SOC., 1938, 80, 836) who proposed a resonance of the three structures:>c=c<Ag +after their study of the silver ion complexesCHATT : COMPLEX COMPOUNDS OF THE PLATINUM METALS.123have the unusual and somewhat loose combination of two reducing sub-stances to yield a product resistant to oxidation by permanganate.Bokii 21, 22, 23 has attempted to obtain the structure ofcis-[Pt (NH3)(CzH4)Cl2]by X-ray methods and claims that the substance is dimeric, with a Pt-Ptbond of length 1.4 A. Each platinum atom is surrounded in a distortedoctahedron by the other platinum atom, two carbon atoms, two chlorineatoms, and a nitrogen atom. Again i t would appear that the platinum isin the platinic state, but it must be remembered that the weight of previouschemical evidence has pointed to the platinum being in the bivalentstate.Acetylene compounds analogous to the ethylene complexes have notbeen obtained, and attempts lead only to brownish intractable substances.However, the substituted acetylene CMe,(OH)*CIC*CMe,( OH) ( = Un) hasyielded a compound [PtUn py Cl,] similar to the corresponding trans-ethylene complex.24Aminopyridine Complexes.-As might be expected from its stereo-chemistry, 2-aminopyridine (apy) does not form chelate compounds withplatinous chloride,25, 26 but the compounds formed [apy2PtCl,] and[apy4Pt]C1, are more stable than the corresponding pyridine or ammoniaderivatives.The former, obtained by direct action of 2-aminopyridine onpotassium chloroplatinate, has been asigned a cis-configuration, which isto be expected, and to account for its greater stability, A.M. Rubinshtein 27suggests that the co-ordination takes place through the tertiary nitrogenatom whilst the amino-hydrogen atoms take part in hydrogen-bond form-ation with the adjacent chlorine atoms. More highly substituted pyridines,e.g., 5-iodo-2-aminopyridineY react directly with potassium chloroplatinite toyield, in this case, truns-[iapy,PtCl,], probably because steric hindranceprevents formation of the cis-compound. The iodoaminopyridine is readilyreplaced by pyridine to yield trans-[py,PtCl,] .28, 29Thiosulphate Complexes.-Surprisingly little research into platinumthiosulphate complexes had been done until D. I. Ryabchikov 30 started avery thorough study of them in 1938 and found that the thiosulphate ionis co-ordinated very strongly to PtII.Previously, P. Shottlander 31 hadobtained Na6[Pt(S20,)4],10H,0 by action of excess of sodium thiosulphateZ1 G. B. Bokii and E. E. Baishteii, Doklady Akad. Nauk. S.S.S.R., 1943, 38, 323.22 G. B. Bokii and E. E. Vainshtein, Compt. rend. Acad. Sci. U.R.S.S., 1943,38, 307.zs G. B. Bokii, N. I. Usikov, and G. L. Trusevich, Bull. Acad. Sci. U.R.S.S., Classe24 A. D. Herman, S. Bukhovetz, and E. Meilakh, Compt. rend. Acad. Sci. U.R.S.S.,25 A. M. Rubinshtein, ibid., 1938, 20, 575.2 G A. M. Rubinshtein, Bull. Acad. Sci. U.R.S.S., Classe sci. chim., 1944, 42.27 Compt. rend. Acad. Sci. U.R.S.S., 1944, 43, 59.2 8 A. M. Rubinshtein, Bull. Acad. Sci. U.R.S.S., Classe sci. chim., 1944, 216.29 A. M. Rubinshtein, Compt.rend. Acad. Sci. U.R.S.S., 1944, 44, 277.30 Ibid., 1938, 18, 39.sci. chim., 1942, 413.1945, 46, 105.31 Annalen, 1866, 140, 200124 YXOBQANIU OHElKCSTRY.on potassium chloroplatinite, but the co-ordinating a M t y of the thio-sulphate radical is such that the halogen atoms of the [PtCI,]= ion can bereplaced two a t a time by action of sodium thiosulphate in theoreticalproportions,32 yielding the ions [Pt(S,O,)CI,]', [Pt(S,O,),]=, [Pt(S,03),]4-,and finally [Pt(S203)4]6-. These complexes are very stable; even hothydrochloric acid fails to produce elementary sulphur or other appreciablechange in them and the thiosulphate ion occupies either one or two co-ordination places.The extraordinary co-ordinating affinity of the thiosulphate ion isstrikingly illustrated by the action of sodium thiosulphate on [Pt (NH3),]C1,.33Normally, the replacement of the ammonia molecules by acid radicalsrequires an excess of reagent and does not proceed easily, but thiosulphatein theoretical quantity reacts in hot solution with evolution of ammonia toproduce [Pt(NH,),S,O,] or trans-Na,[Pt(NH,),(S,03),],6H20 according tothe proportions of the reagents, and excess of thiosulphate yieldsNa2[Pt(S203)4].Even thiourea is completely evicted from [Pt(CS(NH,),},]+fby excess of tliiosulphate.Particular interest attaches to the ion [Pt(S,O,),]= which has beenobtained in cis- and trans-form~,~~ an isomerism very common amongst thecationic and neutral platinous complexes but very rarely observed in anioniccomplexes.The two ions are produced together when the chloroplatiniteand thiosulphate (1-8 mols.) react in aqueous solution, and are readilyseparated by the great difference in solubility of their potassium salts.Ethylenediamine reacts differently with the two salts, and on the basisthat oxygen co-ordination places are attacked in preference to sulphur,the isomers have been orientated by the following reactions : 34Soluble isomer.,Sparingly soluble isomer.Hence the soluble isomer is cis-, and the less soluble is trans-.Ryabchikov also finds that the group trans- to the sulphur atom in thethiosulphate complexes is labilised in the same way as it is by thiourea.This fact strengthens his argument regarding the orientation of [Pt (S,03),]=and is well illustrated by comparison of the reaction between thiosulphateand cis- and trans-[Pt(NH,)2C12],35 which he has suggested as useful todistinguish cis- and truns-isomers of platinum diammines.36 Both isomers32 D.I. Ryabchikov, Conzpt. rend. Acad. S c i . U.R.S.S., 1940, 27, 349.33 Idem, ibid., p . 690. a4 Idem, ibid., 1943, 41, 208.36 Idern, ibid., 1940,28, 231. 36 Idem, ibid., 1941, 32, 344OHATT: UOMPLEX UOMPOUWDS 03 THE PLATINUM METALS. 125react with one molecule of thiosulphate to give sparingly soluble precipitatesbut these have different metal contents.cis-[(NH,),PtX,] + Na,S,O, 4 [(NH3)2 Pt S,O,] J.The labilising influence of the thiosulphate radical causes the X radicalin the trans-position in the unstable intermediate to be replaced by a watermolecule.Two molecules of sodium thiosulphate substitute both acid radicals inboth isomers, but the products (V) and (VI) differ markedly in their stabilitytowards excess thiosulphate, for, whilst (V), having both ammonia moleculeslabilised by trans-thiosulphate radicals, reacts with any slight excess of(V.) (VI.) (VII.)thiosulphate, yet (VI) is stable to 2 4 molecules excess of thiosulphate.Even the acid (VII), prepared from the barium salt by means of sulphuricacid, is stable in aqueous solution and is a strong acid.37 Larger excess ofthiosulphate replaces all the ammonia from both isomers.The converse eviction of thiosulphate ions by amines is p0ssible,~8 buteven thiourea in hot aqueous solution can replace only three thiosulphateradicals from K,[Pt(S,O,),] to give [Pt{CS(NH,),),S,O,]. Ammonia re-places only two to yield cis-K,[ (NH3),Pt(S203),], and ethylenediaminebehaves similarly, whereas pyridine, presumably through the intermediateformation of cis-K,[py,Pt(S,O,),], removes two thiosulphate ions, butbecause of the labilising effect of the trans-thiosulphate ions, the pyridine islost and the final product is K,[Pt(S,O,),] ; trans-K,[py,Pt(S,03),] is, ofcourse, quite stable.33Thiosulphate complexes of quadrivalent platinum could not be obtainedeither directly39 or by oxidation of the platinous c0mplexes.4~ In theformer case the platinum was reduced to the bivalent state by the thio-sulphate, and in the latter the oxidation occurred in three stages, the firstof which was oxidation of the thiosulphate with deposition of elementarysulphur.Palladium forms similar thiosulphate complexe~,*~ but as would be37 D.I. Ryabchikov, Compt. rend. Acad. Sci. U.R.S.S., 1940, 28, 236.Idem, ibid,, 1943, 40, 229.40 Idem, ibid., 1941, 33, 233.D. I. Ryabchikov end A. P. Imkova, Dokludg dead. Nauk. S.S.S.B., 1943, 41,The platinum salt was then oxidised and finally the sulphur.Idem, ibid., 1944,42, 178169126 INORGANIC OHEMISTRY.expected no isomerism was observed, and the equimolecular reaction ofthiosulphate and palladochloride produced PdS and PdS,O, but notNa,[Pd(S,O,)Cl,], which is in keeping with the lower stability of palladiumcomplexes usually observed.Hydroxylamine Complexes.-The lower stability of palladium complexesis well illustrated by the reaction of halogen acids with [Pt(NH,*OH)4](OH)2and its palladium analogue described by Goremykin and his co-workers intheir comparative study of these complexes.42~ 43Products from-Acid. [Pt(NHz'OH)4] (OH) 2.[P~(NH,*OH)~I(OH)Z.~ ~ ~ ~ ~ H " : : ~ ~ ~ ~ ~ ~ ~ ~ ~ HF [ P ~ ( N H z . O H ) ~ I ( ~ ~ Z ) ZHCl [Pt(NH,*OH)4]Cl,HBr [Pt(NH,-OH)4]Br, + [Pt(NH,*OH),Br,] [Pd(NH,*OH),Br,]HI [Pt(NH,*OH)4]12 + [Pt(NH,*OH),I,] PdI,Direct oxidation of the platinous hydroxylamine complexes by chlorineor bromine does not yield the platinic complexes, the hydroxylamine beingoxidised in preference to the platinum,44 but they have been prepared in avery interesting way.45¶ 46When a [Pt(NH,*OH),]++ salt is heated on a water-bath with 2 0 4 8 %hydrobromic acid the platinum is oxidised, presumably by the hydroxyl-amine liberated from the complex, and bright orange insoluble derivativesof PtIV can be isolated, e.g., [Pt(NH,*OH),Br,].Starting from mixed cis-tetramines, e.g., [Pt(NH,*OH),py,]++, mixed derivatives of type[Pt(NH2*OH) PY Br4Iare obtained.Unlike the platinous hydroxylamine complexes, the platinic complexesdecompose without explosion when they are heated, and also the hydroxyl-&mine can be replaced by pyridine. In the latter reaction the liberatedhydroxylamine reduces the platinum to the bivalent state again.Iridium.S'ulphito-compZexes.-Lebedinsky and Gurin have made a study ofchlorosulphitoiridites and aminosulphitoiridites in which iridium is tervalent.By heating sodium chloroiridite with an excess of sodium bisulphite threechlorosulphitoiridites are obtained according to the time of reaction.47 Ifthe heating is stopped when the olive-green solution has become light red,yellow Na,Ir(S0,),Cl2,7H,O crystallises out together with redNa,Ir(S0,),C14,7H,0.If the reaction is continued until the solution is dark red then only thered salt crystallises out and the yellow salt appears slowly from the42 v, I.Goremykin, Compt. rend. Acad. Sci. U.R.S.S., 1938, 18, 341.43 Idem, ibid., 1941, 32, 633.4 4 Idem, Bull. Acad. Sci. U.R.S.S., Classe sci. chim., 1944, 185.45 V. I. Goremykin and K. A. Gladyshevskaya, ibid., 1943, 108.4 6 Idem, ibid., p. 338.4 7 V. V. Lebedinsky and M.M. Gurin, Compt. rend. Acad. Sci. U.R.S.S., 1942,86, 22c u m : COMPLEX COMPOUNDS OF THE PLATINUM METALS. 127cold mother-liquor, whereas if the mother-liquor is evaporated by boiling,yellow Na,Ir( S03),C1,,5H,0 separates from the hot solution.48 This penta-hydrate retains one molecule of water up to 170°, whereas the heptahydrateloses all its water at loo", but both salts yield Na,Ir(S03)3(NH3)3,7H,0with ammonia.Conductivity measurements indicate that this ammine is only tri-ionicand it is considered that one of the sodium atoms is covalently linked inthe complex.49 By double decomposition with a zinc salt only two sodiumatoms are replaced by zinc.The parent salt Na,Ir(S03),C1,,5H,0 is also unusual in its reaction withdilute acid, which replaces two sodium atoms to yield a non-acidic crystallinesubstance Na5H,Ir(S0,),C1,,10H,0.This has a very poor conductivity 48and must have the hydrogen atoms and perhaps some sodium in the com-plex ion. The hydrogen atoms can be replaced by bases to produce againa neutral salt, and it seems probable that the above compounds are morecomplex than the simple formuh would indicate.One of thechlorine atoms is remarkably labile, conductivity measurements indicatedissociation into more than six ions at temperatures of over about 30°, andeven in the cold rubidium chloride reacts to yield NaRb3[Ir(S03)2C13],6H20.47The ammonium and potassium salts of the above chlorosulphitoiriditescould not be obtained,48 the tendency being to obtain salts of the ion[Ir(S0,),C13]4- which seems to be identical with that originally describedby C.Claus 50 in the salt K4[Ir(S03),C13],6H20.The prolonged action of large excess of ammonium bisulphite on am-monium chloroiridite did not replace all the chlorine atoms, but a new salt(NH4),[Ir(S0,),C1,], which could readily be converted into the sodium orpotassium salt by the alkali hydroxide, was formed.51Although suggestions have been made regarding the configurations ofmost of the compounds studied, it was not possible to assign a configurationto any of them with certainty.Organic Arsine CompZexes.-Continuing their investigation of the com-pounds formed by the platinum metals in their lower valency states,F. P. Dwyer and R. S. Nyholm have prepared a number of complexesof iridium dichloride and trichloride with aryldialkyl- and diarylalkyl-ar~ines.~,, 53 These are consistent with six-fold co-ordination of iridium inboth valency states.The simpler complexes [IrC1,,3AsPh2Me] (VIII),[IrC12,4AsPh,Me] (IX), and [IrC13,3AsPh2Me] (X) are all less stable thanthe corresponding rhodium compounds 54 and smell of the free arsine. Thehalogen, on the other hand, is strongly bound and not readily removedeven by silver nitrate. The compounds of type (IX) are not well defined48 V. V. Lebedinsky andM. M. Gurin, Compt. rend. Acad. Sci. U.R.S.S., 1943,38, 128.Qg Idem, ibid., 1941, 33, 241.61 M. M. Gurin, Compt. rend. Acad. Sci. U.R.S.S., 1944, 44, 100.62 J . Proc. Roy. SOC. N.S.W., 1944, 77, 116.64 Ann.Repork, 1944, 41, 101.NaSIr(S0,),C1,,7H,0 also shows unexpected properties.6o J . p r . Chem., 1817, 42, 348.63 Ibid., 1946, 79, 121128 IN0IK)BNIO (YHEMISTRY.and emit a strong odour of the fiee arsine; they are transformed into (VIII)by shaking with light petroleum. Compounds (VIII) and (IX) were theonly complexes obtainable from iridium dihalides and diphenylmethyl-arsine. They were prepared by reduction of the tervalent iridium com-plexes in presence of different quantities of the arsine with hypophosphorousacid in acid aqueous-alcoholic solu-( AsPh ,Me),XIr ‘IrX( AsPh,Me), tion. The complexes of type (VIII)were well defined and have beenassigned a bridged structure (XI), buteven in freezing benzene solution they are highly dissociated.The simple complex (X) is obtained by direct action of the arsine onthe halide IrX, in weakly acid aqueous-alcoholic solution.In boilingstrongly acid solution, however, a number of unexpected and interestingreactions occur.Iridium trichloride yields a yellow, slightly soluble compound isomericwith (X), probably [IrC1,,4AsPh,Me][IrC14,2AsPh,Me] (XII), and from themother-liquor from which this compound has separated an acidH[IrC14,2AsPh,Me] (XIII) can be isolated. It is acid to litmus and givespink ammonium and pyridinium salts but is insoluble in sodium hydroxidesolution.The analogous reaction in the bromine series yields the analogue of(XIII) but not of (XII); however, a complex containing bivalent iridiumis precipitated, viz., [IrBr2,2AsPh,Me].This reduction is unexpected and isprobably facilitated by the low solubility of the complex formed. A similarreduction does not occur in the iodine series except with aryldialkylarsines,but the iodides are quite soluble, and it appears that the instability of theiridium complexes is also important in helping this reduction which doesnot occur in the rhodium series although RhIII is generally easier to reducethan I+.X‘Xfl(XI.)Rhodium.Dimethylglyoxime Complexes .-In their search for square complexes ofhodium, F. P. Dwyer and R. S. Nyholm have prepared dimethylglyoximecomplexes of bi- and ter-valent s6 With dimethylglyoxime,rhodic chloride readily yields a sparingly soluble substance, all the propertiesof which are consistent with the structure (XIV).0- N N 4 Q ! IH I ‘ H Cl----Rh--Cl dl $/s b J .Proc. Roy. SOC. N.S.W., 1946,H,C-CH,I II H(XV.)78, 266. 66 Idem, aid., 1946, 79, 126WELCH : INORGANIC CHEMISTRY OF METALLURGICAL PROCESSES. 1%The complex is a strong acid of pronounced monobasic character. Itforms stable soluble salts and the halogen atoms cannot be removed evenby boiling silver nitrate or chelating acid groups. The silver salt is insolublein water but soluble in dilute nitric acid. The chlorine atoms are almostcertainly in the trans-position, particularly as the dimethylglyoxime canbe reversibly replaced by ethyleneiminebis-salicylaldehyde to yield a com-plex (XV) which must have the nitrogen and oxygen atoms in one plane.This latter complex also yields a violet sodium salt which would indicatebenzenoid-quinonoid resonance of the ion.The rhodic complex Rh(C,H,N,O,), was ultimately prepared in pooryield as an insoluble powder from rhodic sulphate.It dissolved in hydro-chloric acid to give a reddish solution, possibly of the cis-form of (XIV),which lightened in colour and finally deposited the stable trans-isomerPure rhodous complexes were not isolated, although evidence for theirformation by reduction of compound (XIV) with sodium formate wasfound.Continuing his study of the polarographic reduction of the platinummetal complexes, J. B. Willis 57 finds that, of the metals ruthenium, osmium,iridium, palladium, and platinum, only palladium complexes give a satis-factory polarographic step.This corresponds to the reduction of PdII toPd, and the half-wave potential of the ammino-complexes of palladiumbecomes more negative with increasing basic strength of the amine whilstthe reduction also becomes less reversible.Finally, attention should be directed to an excellent review of thestereochemistry of square complexes by D. P. M e l l ~ r . ~ ~(XIV).J. C.3. THE INORGANIC CHEMISTRY OF SOME METALLURGICAL PROCESSES.The enforced development of special metallurgical processes during waryears has necessarily involved new advances in, and applications of,fundamental inorganic chemistry, and several of these appear to merit reviewin these Reports. The topics selected for discussion are the extraction ofmagnesium (particularly from sea water) , the production of highly electro-positive metals by thermal reduction processes, the extraction of alumina andaluminium from clay, and the extraction chemistry of beryllium andzirconium.Magnesium from Sea Water.-The most successful of the commercialsea water processes is the Dow process, operated on a very large scale a tVelasco, Texas.l Here the filtered sea water (containing about 0.13% ofmagnesium) is treated with a controlled excess of calcium hydroxide (pre-pared by slaking lime obtained by calcining oyster shells), and magnesiumhydroxide is precipitated; by thickening the hydroxide is obtained as aW.P. Schambra, Trans. Amer. Inst. Chem. Eng., 1945, 41, 35; C. M. Shigley,57 J . Arner. Chem. Soc., 1945, 67, 547.5 8 Chene. Reviews, 1943, 33, 137.Amer. Inst. Min. Met. Eng., Tech. Publ. No. 1845 (1945).REP.-VOL. XLIII. 130 INORGANIC CHEMISTRY.slurry containing 12% of Mg(OH),. After filtration, the filter cake [25%Mg(OH),] is treated with hydrochloric acid solution containing a littlesulphuric acid (to aid precipitation of calcium as sulphate), and the resultingcrude 15 yo magnesium chloride solution is concentrated by submergedcombustion of natural gas, a controlled gas-air mixture being burned underthe liquid surface. This direct heating is necessary because calcium sulphatewould cause serious scaling of any ordinary form of evaporator. Afterevaporative cooling under vacuum, a 35% magnesium chloride solution isobtained ; a calculated quantity of magnesium sulphate, sufficient to pre-cipitate the unwanted calcium, is added, and the solution allowed to stand.Filtration from precipitated sodium chloride and calcium sulphate then givesa magnesium chloride solution of high purity, from which a solid salt of theapproximate composition MgC12,1-5H20 is obtained by a two-stage evapor-ation process.This salt is suitable for direct feed to the electrolytic mag-nesium cells, in which the electrolyte consists of molten magnesium, calcium,and sodium chlorides at 700-750°.2 The chlorine evolved at the cell anodesis converted into hydrochloric acid (for re-use in the process) by reaction withsteam and natural gas. The molten magnesium is ladled from the cells andcast into ingots of purity at least 99-9%.An interesting variant of the Dow process uses calcined and slaked dolo-mite (comprising a mixture of magnesium and calcium hydroxides) insteadof slaked lime in the initial treatment of the sea water; the magnesiumcontent of the dolomite is then retained with the hydroxide precipitatedfrom the sea water, and the process affords an economic means of utilisingboth sources of magne~ium.~The success of these processes is basically dependent on the very lowsolubility of magnesium hydroxide, which permits its precipitation fromextremely dilute solutions of magnesium salts.Although the engineeringproblems involved in the treatment of large volumes of sea water are con-siderable, and each stage of the process requires careful control, both methodshave been successfully applied.Magnesium from Dolomite and Silicate Minerals.-The abundance ofdolomite (MgCO,,CaCOJ in nature immediately suggests that its use as asource of magnesium should be economic, but the difficulty of separatingmagnesium from large amounts of calcium is considerable.The use ofdolomite in conjunction with sea water has been outlined above; anothertypical dolomite process has been described re~ently.~ The dolomite is firstcalcined and slaked with water to give a mixture of calcium and magnesiumhydroxides, which is boiled with ammonium chloride solution ; calcium thengoes into solution as the chloride, whereas magnesium hydroxide remainssubstantially unaffected : Mg(OH), + Ca(OH), + 2NH4C1 + Mg(OH), +CaC1, + ZNH, + 2H,O.The magnesium hydroxide may be separated by2 R. M. Hunter, Trans. Electrochem. SOC., 144, 86, Preprht 30, 343.4 J. M. Avery and R. I?. Evans, Amer. Inst. Min. Met. Eng., Tech. Publ. No. 1829See P. L. Teed, Bull. Inst. Min. Met., 1946, No. 479, 25.( 1945)WELCH : INORGANIC UHEWSTRY OF METALLURGICAL PROCESSES. 131thickening and filtration and converted into oxide by ignition,or the slurry fromthe previous stage may be treated directly with carbon dioxide to precipitatecalcium carbonate and leave magnesium chloride in solution : Mg(OH), +CaC1, + CO, + MgC1, + CaCO, + H,O. Purified magnesium chloridemay then be obtained from the solution by methods similar to those used inthe Dow process. Economic application of the process just described isensured by linking it with the ammonia-soda process, so that the reaction ofammonium chloride with slaked dolomite calcine provides the necessarymeans of recycling ammonia gas.Olivine, (Mg,Fe),SiO,, and other silicate minerals of magnesium are anattractive source of the metal in some localities.Such minerals are con-veniently attacked by hydrochloric acid, which extracts magnesium and ironas chlorides and leaves the silica substantially insoluble; the use of 20%acid at 90-110" ensures separation of silica in a form that settles well onstanding. Impurities in the acid extract (mainly iron) are precipitated ashydrated oxides by adding the requisite quantity of magnesia, either as suchor in the form of a sludge from electrolytic magnesium cells, containingmagnesium chloride and oxide.Magnesium chloride of sufficient purity forcell-feed is obtained from the solution by evaporation.Eledropositive Metuls by Thermal Reduction.-Until quite recently thedifficulty of reducing oxides or salts of metals such as magnesium, calcium,and potassium has necessitated the production of these metals by electro-lytic methods. Considerable use is now made of direct thermal reductionprocesses, particularly for magnesium, their industrial application havingbeen promoted by development of plant operating under high vacuum.The reduction of magnesium oxide by carbon a t temperaturesapproaching 2000" has for some time been known to be possible; the use ofthis reaction (MgO + C Mg + CO) is hindered by its rapid reversal atsomewhat lower temperatures, the magnesium vapour produced tending toreact with carbon monoxide before it can be condensed.The equilibriumpressures of magnesium vapour and carbon monoxide are calcqlated to reachone atmosphere at 1851", but they fall to less than 0-1 atmosphere at 1 6 0 0 O . 6Recent success with the carbon reduction process has depended on very rapid" quenching " of the hot product gases with hydrogen, natural gas, or a sprayof mineral oil.' This serves the double purpose of cooling the gases to atemperature at which the back-reaction occurs to a negligible extent, and ofslowing down this reaction by extensive dilution of the reactants with inertgas. The magnesium, contaminated with oxide and free carbon, is recoveredas a pyrophoric powder.In one typical application of this process8 themagnesium oxide (prepared from dolomite and sea water) is compressed intopellets with petroleum coke, and the pellets are fed continuously into anelectric-arc furnace lined with carbon blocks. As soon as the product gasesti E. C . Houston, Amer. Inst. Min. Met. Eng., Tech. Publ. No. 1828 (1945).6 K. K. Kelley, see ref. (8).7 See P. L. Teed, Bull. Inst. Min. Met., 1946, No. 479, 25.* T. A. Duncan, Amer. Inst. Min. Met. Eng., Teoh. Publ. No. 1671 (1944)132 INORGANIC CHEMISTRY.leave the reaction zone they meet a cold blast of natural gas issuing fromcooled jets mounted annularly round the exit pipe, and the average gastemperature quickly falls to about 250".The condensed magnesium dust iscollected in a large drum through which the gases pass and (mainly) inwoollen bag filters; it is then collected, without exposure to air, made into apaste with asphaltic material, or into briquettes, and transferred to sublim-ation retorts. I n these retorts, heated to 800°, the pressure is reduced to0.2 mm. or less, and magnesium of high purity sublimes on to a cylindricalsteel liner placed in the cooled head of each retort. After admission ofhydrogen and cooling, the liners are removed, and the magnesium is strippedOff.Reducing agents other than carbon have been widely used in the thermalreduction of magnesia, and successful use of ferrosilicon, calcium carbide, oraluminium is reported. The ferrosilicon process is notable for its simplicity,and for the fact that calcined dolomite may be used directly to supply part orall of the magnesium, the reaction being as follows : 2Mg0 + CaO + Si(from ferrosilicon) -+ 2Mg + ZCaO,SiO,.Use of this reaction a t readilyaccessible temperatures depends on the maintenance of a high vacuum;in practice, briquettes of calcined dolomite and ferrosilicon, containing a littlecalcium fluoride, are charged into steel retorts, which are heated to about1150" and pumped down to 0-05 mm. pressure. Magnesium condenses in thecooled head of each retort. Traces of alkali-metal salts in the charge givea small condensate of alkali metal, which may set fire to the magnesium whenthe retort is opened; this danger is minimised by condensing the morevolatile alkali metal in the retort cap, which is quickly removed when air isadmitted.Calcium carbide and aluminium are used as reducers in a very similarmanner,' the reactions involved being MgO + CaC, + Mg + CaO + 2C ;3Mg0 + 2A1+ 3Mg + A1,0,.The use of small additions of calciumfluoride to the charge appears to be general in most of the processes described,although the mechanism by which it promotes the reaction is admittedlyobscure.It has been found that calcium can be produced from lime convenientlyand economically in plant designed for the ferrosilicon reduction ofmagnesia,lO if aluminium is employed as the reducer a t about 1200". Thereaction is 6Ca0 + 2A1+ 3Ca + 3Ca0,Al,03. Since other alkaline-earthand alkali metals present in the charge distil with the calcium, the use ofhigh-purity lime is important.The regular production of potassium metal by thermal reduction isreported from Germany.11 Potassium fluoride is reduced a t 1000-1 150"with calcium carbide (2KF + CaC, + 2K + CaF, + 2C) or silicon, limebeing added in the latter case to combine with silica formed in the reactionQ L.M. Pidgeon, Canad. Mining and Met. Bull., 1944, No. 381.10 P. H. Staub, Chem. and Met. Eng., 1945, 52, No. 8, 94; C. c. Loomis, Trans.11 F.I.A.T. Final Report, No. 695.Electrochem. Xoc., 1946, 89, Preprint 9, 119WELCH : INORGABTO CHEMISTRY OF METALLURGICAL PROCIESSES. 133(4KF + Si + 4Ca0 + 4K + 2CaF2 + 2Ca0,Si02). The process is carriedout in steel retorts; the metal distils out of the reaction mixture, and iscondensed and collected under petroleum.Part of the potassium fluoridemay be substituted by potassium carbonate [2K2C0, + 3Si + 6Ca0 --+4K + 2C + 3(2CaO,SiO,)] or silicate [2K,SiO, + Si + 6Ca0 4 4K +3(2Ca0,Si02)] without appreciable loss of yield. All the reactants must bethoroughly dried; explosions are said to occur if moisture is present whenpotassium carbonate is used in the reaction.Aluminium from Clay and High-silica Bauxite.-The extraction ofalumina or aluminium metal from clay, its most abundant and accessiblenatural source, has been investigated over a long period, and a voluminousliterature of the subject exists.12 The objectionable impurities likely tooccur in alumina derived from clay are silica and iron, and each of the twomain types of process generally proposed deals effectively with dne only ofthese impurities; " acid " extraction methods applied to clay lead t o rapidseparation from silica, elimination of iron being difficult, whereas " alkali "processes, generally depending on an extraction of soluble sodium aluminateby water, cause difficulty with removal of silica.A recent careful study l3 of a sulphuric acid process for treatment of clayclearly indicates the difficulties associated with this method.The calcinedclay is leached with sulphuric acid, and iron is precipitated from the resultingaluminium sulphate solution by treatment with manganous sulphate andozone. After partial concentration, a clay residue is added to promoteprecipitation of silica.The purified solution is evaporated (by submergedcombustion), and the aluminium sulphate dehydrated and calcined toalumina; the sulphur oxides evolved in the calcination are recovered assulphuric acid for use in the first stage of the process.An alternative to the leaching of clay with sulphuric acid is roasting withammonium sulphate l4 (or in some similar processes, ammonium hydrogensulphate 15). Leaching of the product with water gives a solution from whichammonium alum may be crystallised; this is converted into hydratedalumina by treatment with ammonia evolved in the roasting stage.Ammonium sulphate can be recovered and re-cycled through the process.An interesting recent " acid " process l6 uses a combination of sulphuricand sulphurous acid leaching of clay, alumina being recovered by precipit-ation of a basic aluminium sulphate from the leach liquor.The basic saltis afterwards dissolved in sodium hydroxide solution (giving sodiumaluminate), and hydrated alumina is precipitated by controlled addition of12 See a recent bibliography by R. J. Woody, Washington Stale Coll., Electromet. Res.1s J. H. Walthall, P. Miller, and M. M. Striplin, jun., Trans. Amer. Inst. Chem. Eng.,14 H. W. St. Clair, S. F. Ravitz, A. T. Sweet, and C. E. Plummer, ibid., 1944,159,255.l5 Anon., Mining World, 1945, 7, No. 10, 22.16 0. Redlich, C. C. March, M. F. Adams, I?. H. Sharp, E. K. Holt, and J. E. Taylor,Lab., Bull. E-1 (1943).l345, 41, 55.Ind. Eng. Chem., 1946, 38, 1181134 INORUANIC CHEMISTRY.sulphuric acid.The sodium sulphate solution remaining is electrolysed togive sulphuric acid and sodium hydroxide solutions for use in earlier steps ofthe process.The most useful of the “ alkali processes ” for treatment of clay appears tobe the “ lime-soda sinter ” process, long known but largely investigated in theUnited States during the War.17 If clay is sintered a t a moderately hightemperature with controlled quantities of sodium carbonate and lime, itsaluminium content is converted into sodium aluminate (NaAIO,), and itssilica into an insoluble calcium silicate, probably ZCaO,SiO,. I n theory,extraction of the sinter with water should give a sodium aluminate solutionsubstantially free from silica, from which hydrated alumina could be pre-cipitated by “ seeding ” with the hydrate, or by treatment with carbondioxide; in practice, however, the extract is found to contain appreciableamounts of silica, a t least part of which is precipitated with the alumina, andthe product requires re-processing by the usual Bayer procedure before it canbe used for electrolytic production of aluminium.The lime-soda sinter process has been more usefully applied to low-gradebauxites containing much silica.18 I n the usual Bayer process bauxite istreated with hot sodium hydroxide solution under pressure, to give a silica-free sodium aluminate solution from which a pure alumina hydrate is pre-cipitated. If the bauxite contains much silica, uneconomic amounts of bothaluminium and sodium are retained in the Bayer treatment residue ((‘ redmud ”) as an insoluble sodium aluminium silicate.Recovery of the sodiumand aluminium may be effected by sintering the red mud from a high-silicabauxite with sodium carbonate and lime in suitable proportions, and leachingthe product with water. The extract contains sodium aluminate with a littlesilica, but if the solution is added to the alkali liquor used in a succeedingBayer treatment, this silica is precipitated during the pressure digestion.This ingenious addition to the well-established alumina process has extendedits useful application to poor-quality ores.“ Lime-sinter ” processes have also been investigated recently.19 Inthese, the clay is sintered a t a high temperature with lime, and the silica andalumina contents are converted into calcium silicates and aluminates.Ontreating the sinter with sodium carbonate solution, the silicates are unchanged,but the aluminates undergo double decomposition, calcium carbonateremaining in the residue and sodium aluminate being formed in solution.Alumina can be precipitated from the aluminate extract by the usual methods,the residual alkali being returned to the process as sodium carbonate.In a somewhat similar German process,20 clay has been sintered with cokel7 Univ. Kansas Publ., Kansas Xtate Geol. Survey Bull., 1943,47, 114.Chem. and Met. Eng., 1945, 52, No. 1, 106; J. D. Edwards, Amer. Inst. Min.Met. Eng., Tech. Publ., No. 1833 (1945).R. L. Copson, J. H. Walthall, and T.P. Hignett, Trans. Amer. Inst. Min. Met.Eng., 1944, 159, 241; F. R. Archibald and C. F. Jackson, Amer. Inst. Min. Eng.,Tech. Publ., No. 1700 (1944).2o C.I.O.S. Report, No. XXXII-21WELCH : INORGANIC CHEMISTRY OF METALLURGICAL PROOESSES. 135and anhydrite (CaSO,) to give a product apparently consisting of calciumaluminate and silicate ; on treatment with sodium carbonate solution thisaffords sodium aluminate solution and a residue (calcium carbonate, silicate,etc.) which can be recalcined to a cement of good quality.Some clay processes of a quite different type have also been investigatedin Germany.20 I n these the first step is a direct reduction of the alumina-silica ore with carbon in an electric-arc furnace, giving an alloy (" silumin ")containing about 60% of aluminium, with silicon and a little iron.Aluminium is then extracted from the finely divided alloy by treatment withmercury or molten magnesium. At about 600" and 22 atmospheres pressuremercury gives a solution containing nearly 40% of aluminium; on cooling,almost all the aluminium is precipitated and can be removed by a filtrationprocess, adhering mercury being removed subsequently by vacuum dis-tillation.If magnesium is used for the extraction it is recovered directlyfrom the alloy by vacuum distillation. The silicon-iron residue may ineither case be used for thermal reduction of magnesia, an economic advantageof the process. Other methods suggested for the processing of silumin areextraction with molten lead and volatilisation of aluminium subAuoride.21Beryllium.-The extraction of beryllium and its compounds is com-plicated by the difficulty of attacking beryl (3Be0,A1,0,,6Si02), the only oreavailable in large quantities, and by the somewhat difficult separation ofberyllium from the accompanying aluminium.The simplest method ofattack is to fuse the beryl in a carbon-lined electric arc furnace at not less than1500-1600" and quench the melt in cold water ; the resulting vitreous mass,after crushing, reacts readily with concentrated sulphuric acid, the berylliumand aluminium being converted into partly hydrated sulphates.22 The massof sulphates, containing insoluble silica, is leached with water, andammonium sulphate is added to the solution to precipitate the bulk of thealuminium as ammonium alum, (NH,),Al(S0,),,12H20, which has a very lowsolubility in beryllium sulphate solution containing ammonium sulphate.Crude hydrated beryllium sulphate is crystallised from the filtrate, andsubsequently purified by a recrystallisation procedure.High-grade beryl-lium oxide is obtained from the sulphate by ignition at 1350".A process used in Germany23 is similar in many respects. The beryl isfused with calcium oxide a t about 1500", and the quenched melt '' sulphated "by treatment with sulphuric acid; aluminium is removed as ammoniumalum, as before, and the crude solution of beryllium sulphate is freed fromiron (present initially as ferrous salt) by addition of hydrogen peroxide andcalcium carbonate, which precipitate hydrated ferric oxide. On passinggaseous ammonia into the resulting beryllium sulphate solution, thehydroxide, Be(OH),, is precipitated; this is finally ignited to the oxide atabout 1000".Attack of beryl by fusion or sintering with fluoride or complex fluoride,21 C.B. Willmore, U.S.P. 2,184,705.22 B. R. F . Kjellgren, Trans. Electrochem. Soc., 1946, 89, Preprint 6 , 83.z3 B.I.O.S. Final Report, No. 158; P.I.A.T. Final Report, No. 522, p. 46136 INORGANIC CHEMISTRY.originally a favourite method,24 still persists in recent processes ; the aim ofthe older fluoride methods of attack was usually to convert beryllium intosoluble sodium beryllium fluoride, Na2BeF4, and aluminium into insolublecryolite, Na,AlF,, so that a beryllium salt of moderate purity could beleached directly from the reaction product.Modern variants of the fluorideprocess are designed to extract the beryllium as a soluble fluoride and leavethe aluminium oxide and silica from the ore substantially unattacked. In anItalian process 25 the beryl is sintered at about 800" with sufficient sodiumhydrogen fluoride, NaHF,, to convert all the beryllium present into a sodiumberyllium fluoride, presumed (probably wrongly) to be 3NaF,2BeF2, whichcan be extracted with water from the residue of oxides. A more novelprocess 26 uses sodium ferric fluoride, Na3FeF6, as the attacking reagent ;this reacts preferentially with beryllium oxide (3Be0 + 2Na3FeF6 +3Na,BeF, + Fe203) and leaves alumina, silica, and iron oxide (impurity)unattacked.The sodium beryllium fluoride solution from either method ofattack is treated with alkali to precipitate beryllium hydroxide, preferablyby adding an excess of sodium hydroxide to redissolve the initial precipitate,and then pouring a further quantity of sodium beryllium fluoride solutioninto the hot liquid; this procedure gives a granular, crystalline hydroxidewhich is convenient to filter off.26 The filtrate contains sodium fluoride, andthe economic use of the process demands recovery of its fluoride content ;this is conveniently effected by adding ferric sulphate to precipitate sodiumferric fluoride [12NaF + Fe,(SO,), --+ 2Na,FeF6 + 3Na2S04], which maybe re-used in the attack of the ore.26 The sodium ferric fluoride mode of attackis stated to be applicable with advantage to a new low-grade beryllium orecomprising magnetite with small quantities of helvite, a beryllium ironaluminium silicate.Processes of this type, in which the desired ore con-stituent is selectively attacked without consumption of reagents by unwantedmaterial, are of special value in the utilisation of low-grade mineral deposits.The isolation of beryllium metal was formerly effected by electrolysis of afluoride melt, usually containing alkali or alkaline-earth fluoride withberyllium fluoride or oxyfluoride. This method required the use of relativelyhigh electrolyte temperatures, and highly toxic fluorine gases evolved a t theanodes were troublesome. Preference is now given to electrolysis of a meltof sodium and beryllium chlorides at about 350".23 Satis€actory applicationof this process depends on a supply of pure anhydrous beryllium chloride,now prepared by chlorinating briquettes of beryllium oxide and carbonat 700-800" (Be0 + C + C1, + BeC3, + CO) ; the beryllium chloridesublimes out of the furnace, and is subsequently purified by fractionalsublimation in it current of hydrogen.Beryllium powder has been satisfactorily prepared from the chloride byreduction by sodium vapour at reduced pressure.27 Beryllium fluoride may24 See, e.g., R. Rimbach and A. J. Michel, " Beryllium ", New York, 1932.26 P.I.A.T. Final Report, No. 522, p. 62.z 6 H. C. Kawecki, Trans. Electrochem. SOC., 1946, 89, Preprint 11, 133.27 J. M. Tien, ibid., Preprint 19, 223WELCH : INORGANIC CHEMISTRY OF METALLURGICAL PROCESSES, 137also be reduced with magnesium to give a pure Alloys of beryllium(particularly with copper or nickel) can be prepared by reducing berylliumoxide with carbon in presence of the free alloying meta1,22t28 but the berylliumcontent obtainable in the alloy is limited; beryllium fluoride may similarlybe reduced with magnesium in presence of base metal to give an alloy.28Zirconium-Considerable interest in zirconium, particularly in the metalin its ductile state,29 has been evident recently, and the publication of hreview 3o of much scattered information is timely.I n a typical recent process,31 zircon sand (mainly ZrSiO,) is heated withcarbon in an electric resistor furnace at about 2000", and arnixture of zirconiumand silicon carbides is formed; some of the silicon is said to be volatilisedfrom the charge as the monoxide, SiO. The mixture of carbides is heated inchlorine, and the tetrachlorides of zirconium and silicon are formed in astrongly exothermic reaction ; the zirconium tetrachloride can be fractionallycondensed from the mixture in a vessel held above 80", the condensatecontaining only 0.05--0.3:/, of silicon, with up to 0.5% of iron and otherimpurities. The chloride is purified by sublimation in hydrogen, whichreduces ferric chloride t o the relatively non-volatile ferrous chloride ; thesublimate contains only 0.05y0 of iron. The purified zirconium tetrachlorideis reduced with magnesium in a special furnace so designed that reactionoccurs between the tetrachloride vapour and molten magnesium, a heliumatmosphere in the furnace ensuring exclusion of air. The mixture of zir-conium metal, magnesium chloride, and excess of magnesium produced in thereaction is vacuum-distilled a t up to 900" in a second furnace ; the residue ofzirconium remaining may still contain up to 1% of magnesium chloride,magnesium, and hydrogen, which are removed by heating the metal in avacuum induction furnace a t 1500". Finally, the zirconium is melted downinto small ingots in an arc furnace in an atmosphere of helium at lowpressure. The product is ductile and can readily be rolled into sheet.Another method of reducing zirconium tetrachloride with magnesium hasbeen de~cribed.~2 A. J. E. W.J. S. ANDEBSON.J. CHATT.A. J. E. WELCH.28 W. J. Kroll. U.S. Bur. Mine.?, Inf. Circ. No. 7326 (1945).28 See, for example, D. B. AInutt and C. L. Scheer, Tians. Electrochem. SOC., 1945,30 W. J. Kroll and A. W. Schlechten, 77.8. Bur. Mines, Inf. Circ. No. 7341 (1 946).31 W. J. Kroll, A. W. Schlechten, and L. A. Yerkes, Trans. Electrochem. SOC., 1946,s2 R. von Zeppelin, Metal2 u. Em, 1943, 40, 252.88, Preprint 30, 357.89, Preprint 29, 365