American Mineralogist, Volume 70, pages 443454, 1985
Crystal chemistry:past, present, and futurer
CrHRrns T. Pnrwrrr Departmentof Earth and SpaceSciences StateUniuersity of New York StonyBrook, New York 11794
Abstract Crystal chemistry is an important part of the scienceof mineralogy and describesthe relationships of mineral crystal structures with the correspondingphysical and chemical properties.Crystal chemistryhas evolved over a period of fifty years or more from a purely qualitative endeavorto one that is quantitative and that can provide essentialinsight to the behavior of minerals under varying conditions of pressure,temperature, and chemicalenvi- ronment.Increased interest in the chemicaland physicalproperties of mineralsand how these are related to important problemsin the geologicaland materialssciences makes the future look verypromising.
Introduction and technologyas well. The subject of this paper, crystal chemistry,embodies aspects of all three areasin Figure 1 Role of crystal chemistry and, therefore,should be consideredto lie in the overlap The study of crystal chemistry is of fundamental impor- regionamong the differentcircles. tance to the science ofmineralogy, as well as to other fields Evans (1952)defined crystal chemistry as "the study of such as materials science and solid-state chemistry. How- the relationship of the internal structure of a body to its ever, it is my perception that mineralogy itself means differ- physical and chemical properties.It aims at interpreting ent things to different people, as do crystallography and the propertiesof any substancein terms of its crystal struc- petrology, disciplines that are also considered to be oflicial ture, and, conversely,at associatingwith any structural interests of the Mineralogical Society of America and the characteristica correspondingset of physicaland chemical Society often regulates its activities according to these cate- properties.Ideally crystal chemistry should enable us to gories. For example, awards are given each year to prom- predict and synthesizechemical compounds having any de- ising young scientists,but alternately to crystallographers siredcombination of propertieswhatsoever." one year and mineralogists/petrologists the next. I believe The early developmentof crystal chemistrywas the pri- that these labels are not really representative of the way in mary result of experimentsand interpretation of W. L. which we apportion our scientific efforts and that another Bragg and coworkers at Manchesterin the 1920'sand by kind of division is more realistic. This division is illustrated V. M. Goldschmidtand coworkersin Oslo in the 1930's. in Figure I and is characterized by three major parts, de- Their efforts were characterizedby an empirical approach scriptive mineralogy, mineralogical applications, and phys- to the understandingof crystal structure through simple ics and chemistry of minerals. Descriptive mineralogy, in- models basedon observedinteratomic distances.Later in volving identification and classification of minerals is really the 1930's,Pauling showedthat it was important to ac- classical mineralogy and describes virtually all mineral- count for chemicalvalence and introduced the conceptof ogical activity before the 1930's,and still is of great interest electronegativityto help account for varying degreesof to many scientists.Since the 1930's,applications of miner- covalencein chemical bonds. Subsequently,the approach alogical theory and experiment have become more impor- to crystal chemistry has developedalong essentiallytwo tant in the understanding of and application to petrology, tracks, one still largely empirical and based on observa- geochemistry, and even geophysics. This constitutes the tions of interatomic distancesand incorporating relatively second division in Figure I and is also the one in which simple ideas about valence,coordination, and electronic most people are involved today. The third category, the structure.The secondapproach is one basedon quantum study of physical and chemical properties of minerals, is mechanicsand has been used mostly for interpretation of one that has seen substantial growth in recent years and spectraor for treatment of molecular structuresor hypo- one that is becoming increasingly important, not only for thetical clustersthat simulate configurationsfound in real the earth sciences, but for many other aspects of science minerals. A major problem for crystal chemistsinterested in min- I Adapted from Presidential Address at the annual meeting of erals is that most of the important mineral structuresare the Mineralogical Society of America, November 6, 1984, Reno, relatively complex and are composedof infinite periodic Nevada. arraysof individualions rather than of molecules.Thus, it 0003-{)o4xl85/050@43$02.00 443 PREWITT: CRYSTAL CHEMISTRY: PAST, PRESEN? AND FUTURE
Principles of crystal chemistry Empirical crystal chemistry is based on the concept of Oescriplrve periodic structures containing ions of specific size, charge, Minerology and cobrdination number, with one or two other parame- ters such as electronegativity helping to account for cova- lence or metal-metal bonding effects. The coordination groups are combined with each other, resulting in an over- all symmetry that can be defined through the unit cell and A ppI icotion Physicol8 space group. If the coordinations of the cations by anions ChemicoI are defined, then the coordinations ofanions by cations are fixed, as illustrated by the relation for compound PAoaBo *X., ,p:pp+ee
PHYSICALA whereP and Q are the CN's of the cationsand R is the CN DESCRIPTIVE CHEMICAL of the anion. The size of an ion is an arbitrary value de- APPLICAT IO N MI NERALOGY PROPERTIES pending on what fraction of the spacingsbetween the ion Pelrology Identificolion Theory and its neighborsis allocatedto the central ion. In mineral- Geochemi slry Clossificoiion Modeling ogical crystal chemistry, it is most common to use the Geophysi c s Experimenl "hard sphere"model where eachion is consideredto be a Fig. 1. Circles representing the fields of mineralogy. Crystal sphereof a specificsize, although molecularchemists often chemistry falls in the area of overlap among the circles. assigna bonding radius and a van der Waal's radius to representa non-sphericalion. has been diflicult to apply the quantum mechanical theory The chargeassigned to ions is usually the formal valence, developed by physicists for simple cubic structures or although even high school chemistry studentsknow that quantum chemistry that has been applied to molecules by most bonds have a significantdegree of covalenceand the chemists. Because of this, the empirical approach of Bragg, absolutevalue of the effectivevalence in real crystalsis less Goldschmidt, and Pauling is still being used by many in- than the formal valence.Various schemeshave been pro- vestigators although there has been substantial progress posed for specifyingthe actual charge on an ion or the recently in applying quantum mechanical techniques to degree of ionic character of a bond. Probably the most minerals. A good summary of the latter was made by G. V. well-known of theseschemes is that of Pauling (1960)who Gibbs (1982) in his MSA Presidential Address and today related the degreeof ionicity with the electronegativitydif- Gibbs, J. A. Tossell, A. Lasaga, their students, and others ferencebetween the two atoms involved. For example,his are making good progress toward combining the cluster relation calculations with considerations of long-range forces in :23(xt - crystals. A(Si-O) xa)2 My purpose is to discuss some of the basic concepts of indicatesthat the Si*O bond is about 50% i
Structural parameters as functions of radii The fundamental assumption of ionic radii tables is that one can predict the auerage interatomic distance in a coor- dination polyhedron and that the average interatomic dis- tances will be the same for all similar polyhedra from struc- ture to structure, as long as the cation remains the same. However, Shannon and Prewitt (1968)found that there are additional factors that influence average polyhedral intera- tomic distances, such as the linkages between polyhedra, relative electronegativity of other cations in the structure, 54 5 !a 40 42 44 45 .40 5O !2 14 16 5a the degrees of distortion of the polyhedra, and metal-metal Fig. 2. Plots of cell volume (Iz) vs. 13for cations in the zircon bonding across shared polyhedral elements. The linkage and olivine structures.From Shannonand Prewitt (1970a). problem was solved by determining the radii of oxide and fluoride ions as functions of coordination of these ions by cations. Thus, the average interatomic distances in poly- CaTiO. does not fall on the smooth trend for titanates. We hedra increase as the linkages between octahedra increase. have re-refined the CaTiO3 structure (Sasaki et al., ms.) Polyhedral distortion will affect the average distance, as and find that the new values move the point for CaTiOr to was shown by Shannon et al. (1975b).These authors plot- the linear trend. ted average interatomic distances for Mn-O polyhedra vs. Metal-metal bonding polyhedral distortion showing that the average distances increase with increased polyhedral distortion. Whenever the formula of a compound does not balance One surprising result of our work was the demonstration according to the usual valences of cations and anions (e.g', that plots of a vs. r or V vs. 13 are usually found to be Ni.S, or FeS2) or when interatomic distances are unusu- linear and that any significant departure from linearity in- ally short, cation-cation or anion-anion bonding is an im- dicates an error in the data or an interesting crystal chemi- portant feature of the structure. This may also affect the cal phenomenon. Figure 2 contains plots of Z vs. 13 for cation-anion bonds and the result is that ionic radii may compounds with the olivine and zircon structures (Shan- not predict the actual interatomic distances in the struc- non and Prewitt, 1970a). These structures are ionic in ture. An interesting example of this is seen in compounds character and the substituting cations do not exhibit any with the corundum structure (Fig. 5). Here, each octa- unusual interactions. Figure 3 is an example ofa departure hedron shares one face and three edges with other octa- from linearity. We interpreted the curvature in the plots hedra containing identical cations. Pauling's 3rd rule state relating cell volume to 13 as a result of the changing ef- that "the presence of shared edges, and particularly of fective coordination number of the rare-earth A ion in per- shared faces. in a coordinated structure decreasesits stabil- ovskite structures,i.e., as the size of the A ion increases,the ity; this effect is large for cations with large valence and effective coordination number increases from eight nor- small coordination number, and is especially large in case mally assigned for the orthorhombic or rhombohedral the radius ratio approaches the lower limit of stability of structure toward the twelve found for cubic perovskite. the polyhedron." In ionic structures that do contain shared Figure 4 indicates another type of relation, that of the polyhedral elements, the cations repel each other and the Goldschmidt tolerance factor applied to the orthorhombic O-O distances across the shared element tend to be shorter perovskites where bond length distortion is plotted vs. the than others in the structure. Exceptions to this for the tolerance factor, to6., calculated from observed interatomic corundum structure are shown in Figure 6, a plot of the distances or from radii tables. Note that the value for cation-cation distances as a function of cation radii for 446 PREWITT: CRYSTALCHEMISTRY: PAST,PRESENT AND FUTURE
./ ./ , 4 '-' '// // -/ lt/ a.a ,a' J ../
i.z' z-
Fig, 5. Corundum structure showing shared faces between oc- tahedra along c, and edges shared between octahedra in the layers normal to c. From Prewitt et al. (1969).
severalcompositions containing aluminum and transition metal cations. Ti3+ contains one d electron and there is strong interaction (contraction)across the sharedface. V3 + has two d electronsand there is interaction acrossboth the face and the sharededges. Cr3+ has three d electronsand there is relatively more shorteningacross the face than the |.I YDIi €. Ylb edge.These interactions 60atrr$tmtpt2o becomemore complicatedfor ca,t- ions with morc d electronsand for structurescontaining Fig. 3. Cell volumes(I/) vs. 13 of rare-earthions in the A site more than one kind of cation. However,this exampledoes for othorhombic and rhombohedral perovskites.The nonJinear illustrate the importance of metal-metal bonding for cer- featuresare thought to be a result of changingeffective coorcii- tain structures and compositions.A similar situation is nation number ofthe A ions. From Prewitt and Shannon(1969). found for the rutile structure, shown in Figure 7. Here
1.00 0 98 0,96 o 94 o 92 0bserved ToleranceFactor tobs Fig' 4. Bond length distortion vs. Goldschmidt tolerancefactor for orthorhombic perovskites.Recent re-refirrement of the CaTiO, structurebrings to CaTi point to the linear trend. From Sasakiet al. (19g3). PREWITT: CRYSTAL CHEMISTRY: PAST,PRESENT AND FUTURE 447
lrj o 2
F al, o G &. NUMBEROF d ELECTRONSPER CATION Fig. 8. Plot of cfa vs. number of d electronsfor first, second, and third row transition elementsin the rutile structure.ApPar- ently, the metal-metalinteractions are much strongerfor the ions with partly-filled4d and 5d shells.From Rogerset al. (1969). EFFECTIVE IOI{IC RADIUS (AI
Fig. 6. Cation-cation distances across the octahedral sharcd delafossite structure taken from Shannon and Prewitt face and edges plotted vs. effective ionic radii for compounds with (1970a). One point illustrated here is that it is often possi- the corundum structure. From Prewitt et al. (1969). ble to detect faulty structural data by making these plots. The plus signs represent literature data that was improved octahedrashare edges along c. Figure 8 is a plot of cfa vs. the no. of d electronson the cation. There are sharp vari- ations in cfa a.sa function of metal-metal interactionsthat are somewhat different from those in the corundum- structureexample because the cations in rutile are divalent rather than trivalent. Figure 9 is a plot of I/ vs. 13 for compoundswith the
o
I I I I r|, Fig. 9. Delafossite,cell volume(I/) vs. r3 for the M'(octahedral) ______.,1 o ions in the delafossite+ypestructures. The dashedlines show cor- Fig. 7. Therutile structure showing how octahedra share edges rections made when cell volumes were re-determinedon well' in columnsparallel to c. FromRogers et al.(1969). characterizedsamples. From Shannonand Prewitt (1970a). 448 PREWITT: CRYSTALCHEMISTRY: PAST,PRESENT AND FUTURE
@tt Oco @o
CuCoO2
I F.- o---l 't1 Fig. 10. Delafossite structure for the composition ptCoOr. pt Fig. ll. V vs. rt for the linearly-coordinated ions in the is linearly-coordinated by two oxygens and Co is octahedrally- delafossite-type structures. The cell volumes for the compounds coordinated. From Prewitt et (1971). al. containing Pd are anomalously low because pd has an effective valenceof2* rather than l*. From Shannon and prewitt (1970a). by synthesizingpure crystals and refining their crystal structures.The delafossite(CuFeOr) structure is shown in varied as a function of the Shannon and prewitt (196g) Figure 10.Delafossite is a mineral that containsCu coordi_ ionic radii. Figure 13 showsa Pc-Ir diagram for an olivine nated linearly by two oxygen ions and Fe coordinated oc- crystal in a fine-grained groundmass.Except for Zn, all ions show parabolic distributions with the maxima at
a @ PYBITE-STBUCTUBEDISULF IDES c,j pounds containing Pd and pt. Becausethe octahedral Co-O distancesare exactly right for Co3+ and because J theseions cannot exist in a 1+ state,our conclusionis that N Pd and Pt are really divalent and that thereis a delocalized electron,which is confirmed by the very high conductivity G in the metal layer, as opposedto delafossiteitself. o Another Iru examplewhere the application of radii has not F produced (J consistentresults is for the transition-metaldisul_ z fides with the pyrite structure (King prewitt, LL.l and 1979). __l Figure 12 is a plot of r vs. atomic number for the seriesof o9 disulfidesfrom Mn through Cu. This plot is entirely consis- z^-. tent with Cu2* in CuSr, but no sulfide has ever beencon- co firmed to contain Cu2+ and photoelectron spectroscopy (Nakai et al., 1978)shows that Cu in this phase is consis- O tent with monovalentCu, the only speciesthus far reported O for N any copper sulfide.Clearly, this apparentcontradiction 5.00 6.00 7.00 8.00 9.00 needsfurther investigation. NUMBER OF 3d ELECTRONS FOR M2+ Trace and minor element distributions between phases Fig. 12. Effectiveionic radius vs. atomic number for disulfides with the pyrite structure.Although the Cu-S bond length in CuS, Severalyears ago, Matsui et al.(1977) showedhow parti- is consistentwith Cu2+,other evidence shows that it is reallvCu+. tion coellicientsof elementsbetween a liquid and a crystal Data from King and Prewitt (1979). PREWITT : CRYSTAL CHEMIS.|RY : PAST, PRESENTAND FUTU RE 449
Figure 16 shows a plot of log P" (partition coeflicient) vs. Olivine- Grm (r, - do-o * re), where r, is the ionic radius of the trace ion M iyale -sima Mg X substituting for the major alkali cation, da-o is the polyhedron, and LO average bond length for the alkali cation ro is the radius of oxygen. This is a very valuable tool for identifying actual sites occupied by trace elements and should be used extensively by geochemists. High pressure Crystalstructures at highpressure c O Oneof my primaryinterests is in thecrystal chemistry of o high-pressureand high-temperaturephases. Crystal- o chemicalprinciples can help us to understandhigh- c ul pressure phenomena and are useful as a guide to syn- = phases. In recent years the diamond anvil cell d thesizing new (L has been used to obtain extensive information about crys- tal structures and phase transformations at high pressure' both with single crystals and powder samples. I will not discuss diamond-cell technology because this has been cov- ered well in reviews by Hazet and Finger (1982), Skelton (1984), Jayaraman (1984),and others. One of the most sig- nificant results from single-crystal work is that the same regular kinds of trends we have have seen at room pressure are applicable at high pressure. For example, Hazen and
o00l 04 05 08 r0 12 t1 16 r0 Shannon-PrewrttRaorus (A) Augite- Grm Fig. 13.Partition coeffrcient vs. effectiveionic radius(PC-IR) Taka-sima for olivine-groundmasssystem in tholeiitebasalt. From Matsuiet al.(19771. about the radius of the major cation, Mg. For bronzite, Matsui et al. showedthat there are two maxima indicating two M polyhedrathat are more differentin sizethan the M polyhedra in olivine, and Figure 14 givesthe resultsfor an o augite where the peaks are even more pronounced.Evi- o dently, the problemswith Zn and Cr are a result of strong (Jo tetrahedralpreference for the former and strong octahedral 6 0r crystal field stabilizationfor the latter. Thesediagrams are ! striking examples of how significant information about q sizesand chemicalcharacteristics of ions can be obtained. Recently,Yurimoto and Sueno (1984)demonstrated that similar curvesfor anionscould be obtainedrelatively easily using secondary-ionmass spectrometry(SIMS) as well as curvesfor cations.One of their plots, reproducedin Figure 15,implies that the ideal sizesfor ions in silicatestructures (the maxima on the parabolic curves)may not correspond with any actual ionic radii. This observationmay not be valid, however,because the radius ratios under conditions of crystallizationare probably quite different from thoseat room temperature. Another kind of display of partition coefficientswas 04 05 08 1.0 1,? Radius (A) given by Volfinger and Robert (1980)where PC was plot- Shannon-Prewitt ted vs. the differencein size between a polyhedron that Fig. 14. PC-IR diagramfor augite-groundmasssystem showing would be preferredby the substituting ion with the poly- two maxima, one correspondingto each M polyhedron in the hedron occupiedby the major ion in the mica structure. structure.From Matsui et al.(197'll. 450 PREWITT: CRYSTAL CHEMISTRY: PAST, PRESENT AND FU.TURE
t0,
lo0 Pr
AI \ \ IF 9cl
Fz
I
o e lo-l vs I I
F I AS F t & -08 C -06 -o4 -o2 04 05
Fig. 16. Partition coeffrcient between hydrothermal solutions n-e and micas plotted vs. d where 6 is the difference in size between the alkali polyhdron in the mica structure and the polyhedron that normally would be preferred by the substituting ion. The circles on the left-hand branch of the "V" represent cations smaller than the major alkali cation, and those on the righfhand branch represent larger cations. The circles to the far left of the diagram represent ions that occupy smaller interlayer cavities rather than l0-t g the large alkali site. From Volfinger and Robert (1980). 2s r0 6e 88 rr 120 rre 150 t8c 2a IONIC RF0IUS zpn similar conclusions. Fig- 15. PC-IR diagram for an olivine-groundmass system con- However, Prewitt (1982)pointed out structed using SIMS data. The positions of the maxima corre- that it is hard to reconcile these conclusionswith recent spond to the sizes of cavities for the tetrahedral and octahedral determinationsof crystal structure parametersdetermined cations, and for the oxygens. From Yurimoto and Sueno 09g4). on oxide structures at high pressureusing the diamond- anvil cell. This work showsthat cation-oxygenbonds com- press and expand more than those with larger bond
Prewitt(1977) found that a linearrelationship exists for the compressibilityvs. size and charge of ions in oxide struc- 400 = (d3/zt tures (Fig. 17).A generalstatement is that polyhedra con- E 3t x t0-6Kbor_r taining large, low-chargeions are more compressiblethan those containing small ions with high charge.Hazen and I 7 Finger (1979)extended this conceptin a comprehensive ts 300 compilation .o of data. Y A curiousconsequence ofour work on structuresat high o pressureis contrary to the popular belief that anions are o more compressiblethan cations and that most of the com- x 200 rq pressionin a structure is taken up by the anions. For ex- ample, Ida (1976) calculated compressibilitiesof cations nol"* .-\ '8"n.. and anions in alkali halidesusing Born lattice theory com- ,/ bined with experimental values of interatomic distances roo oB..-.1 and bulk moduli. This analysis indicates that the halide anions are substantiallymore compressiblethan the alkali ions. Ida extrapolatedthis result to abundant ions in the Fsi4* Earth'smantle, namely Mg2*, Fe2+, Sia+,and O2-, and oo' 2468tO concluded that the compressibilitiesof these cations are negligiblysmall and that only 02- is effectivein determin- d31, ti!/es"l ing compressibilitiesof common silicates. Several other Fig. 17. Linear relationship between compressibility and workers(Miyamoto and Takeda,1980; Matsui et al., l9g2) volume of cation-oxygen polyhedra. From Hazen and prewitt performed molecular dynamic calculations and reached (r9771. PREWITT: CRYST,ALCHEMISTRY: PAST,PRESENT AND FUTURE 451 strength; for example,Ca-O bonds compressmuch more ionic radii, and (c) high electronegativities.These parame- than do Si-O bonds. In fact, when compressibilitiesof ters are intimately related. The electronegativityneed be many different cation-oxygen bonds are compared, one consideredonly when the size and charge of A ions are could concludethat most compressionis taken up by the similar, e.g., for the pairs Mg2*-Co2*, Sc3*-In3*, or cationsrather than by oxygen,even though this is contrary Ca2*1d2*. This rule is relatedto Pauling'sionic bond to the prevailingidea that oxygenions are rather large and strengthmodified by the electronegativityfactor. "fluffy". Examination of the changesin radius with coordi- Rule I can be illustrated with transformationsof ger- nation shows that cation radii increasemore rapidly with manatesand silicateshaving formulas ABO3 (G : 1) and increasingcoordination number than do oxygenradii. This A2BO4 (C:2). For example, MgSiO3 (clinoenstatite) leads to the idea that, as far as a single ion is concerned, transformsto a structure with six-coordinatedSi (ilmenite) oxygen is more rigid than at least the large cations. An at a lower pressure(20 GPa) than does Mg2SiOn(spinel), exampleof this relationship is shown in Figure 18 where which disproportionates to perovskite plus periclase at effectiveionic radii for Ca2* and 02- are plotted against almost 30 GPa (Ito, 1977).Rule 2 easilyexplains why Si in coordination number. From this diagram, it is apparent P2SiO?is six-coordinatedat low pressures,but does not that the radius of Ca2* changesat a significantly higher explain completely the transformation of alkaline-earth rate than does the radius for O2-. Although this may not metagermanatesto the perovskite structures. SrGeOt be sulficient evidenceto prove that oxygen ions are rela- transformsto cubic perovskiteat 5.0 Gpa, and BaGeO. at tively harder than cations,the changein radii with coordi- 9.5 Gpa. Nevertheless,this kind of approachto the crystal nation must be explained by any model for the detailed chemistry of high-pressurephases can provide interesting structuresof oxides. and valuableinsight as to why a particular transition takes I have been interested in mantle mineralogy for a place as it does.It appearsthat much more of this kind of number of years and also in the synthesisof other phases analysiscan be donewith dataalready available. that give interesting crystal chemical information. In a search for rules to guide such synthesis,Shannon et al. Spin crossouer (1975a) developed two general rules based on bond One interestingpossibility for a phasetransition in the strengthconsiderations that extendthe applicability of the mantle that has beendiscussed by a number of authors.but analogstructure concept. Given a compositionA*B'O", the never observedexperimentally for iron oxides,involves a two rulesare: transition from high-spin to low-spin iron. Such a change 1. When the ratio of A cationsto B cations(G) is greater has been called "spin crossover" by chemists (Gtitlich, than a certain critical value, G,, the coordination of B is 1981)and has beenwell-documented for molecularorgano- generallytetrahedral. When this ratio is lessthan - 1.0,the metallic compounds,but no good review for this effect in tendencyfor B to have a CN of five or six is greatly in- solid-statematerials has beenmade. Several authors have creased.This rule gives an approximate measure of the suggestedthat increasingpressure can induce a transition probability of finding two B cations bonded to the same from high spin to low spin in oxides and sulfides,and oxygen ion. The lower the value of G, the gteater is this Ohnishi (1978) has given a theoretical treatment of the probability and the more unstablethe tetrahedral coordi- temperatureand pressureeffects on spin crossover.Figure nation. 19, a plot of radius vs. atomic number for the divalent 2. The greater the A-O bond strength, the greater the first-row transition elementstaken from Shannonand Pre- tendency for six-coordination of B. High A-O bond witt (1968),illustrates the reason for this transformation. strengthsimply for A: (a) high formal charges,(b) small Here, the upper branch of the curve follows the radii of high-spinions and the lower branch the low-spin ions. Be- causepressure favors higher density, spin crossoverfrom Cotions. O OO3-O OOeizCf,t /4, \ high spin to low spin can occur as pressureincreases. Table \ACNi o2- ooorizcw I adapted from Prewitt and Rajamani (1974)shows how increasing pressure, temperature, and ligand field can inducespin crossoverin transition-metalions. With regard to minerals,spin crossovercould, for exam- r (A) ple, result in a l3oh increasein zero-pressuredensity for FerSiOo.However, no suchtransitions have been observed and proven for iron in an oxide material. Apparently, at high pressurethe oxygencoordination around Fe increases before the crystal field becomeslarge enough to force a changeto the low-spin state. Severalyears ago, Yagi and Akimoto (1982)reported a phasechange in Fe2O3at about Coordinotionnumber (CN) 55 GPa and from the X-ray powder Pattern it looked as Fig. 18. Plots ofionic radius vs. coordination number for Ca2* though the high-pressurephase volume was consistentwith and 02-. The slopes of these curves may be related to the com- LS Fe. However, subsequentMdssbauer work indicates pressibilities of the ions. From Prewitt (1982). that there are two or more crystallographicsites in this 452 PREWITT: CRYSTAL CHEMB.TRY: PAST, PRESENT AND FUTURE
were performing significant high-pressureresearch related to geology and geophysics.Only a few of the laboratories were working at pressuresin excessof 40-50 kbar, e.g., thoseof Kennedyat UCLA, Boyd at the GeophysicalLab- oratory, and Sclar at Lehigh University. Severalother lab- oratoriesnot directly involved with earth scienceneverthe- less produced results that were of considerablehelp in understandingmineralogical problems, e.g., those at Bell Telephone, Du Pont, Air Force Cambridge Research Center, Brigham Young University, and Cornell Univer- sity. I was involved in high-pressurecrystal chemistryin the 1960'sand was involved with two experimentsthat have Fig.re. Effective,""::;,Tr::,-,Joi, erectronsrornrst- beenof lasting importance to the understandingof mantle row transition elements. The two branches indicate different radii mineralogy.One for high-spin and low-spin ions. From Shannon and prewitt. 196g. of theseexperiments was the synthesisof single crystals and structure determination of CdGeO, with the garnet structure (Prewitt and Sleight, 1969). high-pressurephase and it is possiblethat it containsFe in Powder samplesof this phasehad beenmade previouslyby a highercoordination (Syono et al.,1984). Ringwood and Seabrook(1963), but they were unable to It is possiblethat experimentalistshave beenlooking at index the completeX-ray pattern. We found that the cell the wrong material and that a silicateolivine or spinelwith was tetragonaland that Cd was orderedinto the dodecahe- a small amount of Fe would be the most likelv to exhibit dral and octahedral sites, and that Ge was ordered into spin crossover.This is in analogywith the experimentof octahedral and tetrahedral sites. The structural formula Bargeronet al. (1971)wlterc 2o/o Fe in MnS, showedcross_ can be written showing each crystallographically-distinct over at about 40 kbar. It is possiblethat a HS to LS ion species: transition would occur in olivine or spinel at very high vru[Cd4cd2]vl[cd2Ge2] rv[GeceGenl pressuresand it might be instructive to examine analog phasescontaining HS Co where the transition would take rv[ooo4o4ononon] place at much lower pressures.Thus far, the only reported This formula indicatesa wide rangeof possibilities experimentalwork on a cobaltoxide exhibiting spin cross- for sub- stitution overis for CorO. (Chenavaset al, l97ll. of different ions in the tetragonalgarnet structure without changingthe overall symmetry.To my knowledge, Cubic-anuilcell no systematicinvestigations have been made of thesepos- sibilities. In the 1950'sand 1960'sthere was greatinterest among A silicatewith this kind geochemistsand geophysicistsin high-pressuresynthesis of of distorted garnetstructure was new phases that provided important new information subsequently found in meteorites and named majorite (Smith about mineralsand conditionsin the Earth'smantle. This and Mason, 1970).The other phasesynthesized by interest was also shared by materials scientistswho were Prewitt and Sleight (1969) was CaGeO. with the per- engagedin synthesisof new phasesfor the purpose of ovskite structure. This had also been synthesizedpre- making new materialswith unique properties(e.g., synthet- viouslyby Ringwoodand Major (1963),and attemptshad ic diamond). However, two things happenedthat reduced beenmade to index powder patternsof both samplesusing the amount of activity (exceptfor Japan and possibly the cubic unit cells. Later, we obtained single crystals and USSR) in large-volume,high-pressure research: (l) Very found that the material was orthorhombic and with a few new materials were found that could be produced on structuresimilar to that of MgSiO, perovskite(Sasaki et an economicbasis for commercialapplications, and (2) the al., 1983).It is clearthat high pressurepermits the synthesis diamond-anvil cell was invented and perceivedby many of an almost unlimited number of new phasesthat can scientistsand administrators as an economicalsubstitute that could perform most of the experimentsformerly car- ried out with the large-volumeapparatus. Nevertheless, we Table l. Effect of temperature, pressure, and ligand field on spin learned an enormous amount about systematic crystal crossover of first-row transition elements chemistryand physicalproperties of materialsthrough the synthesis of quenchable high-pressurephases with the P T Ligand Ni3* latter equipment. Foremost in the earth sciencesat that time were the Ringwood group at Australian National F ES ES ES HS o ES University and Akimoto's laboratory at the University of a+ HS us/Ls LS Tokyo where many germanate and silicate phaseswere s RS HS/LS LS LS synthesizedand used to interpret the structure of the $2 ES/LS LS LS LS Earth's mantle. In the USA, a number of laboratoriesalso PREWITT: CRYSTALCHEMIS.TRY: PAST, PRESENT AND FUTURE 453 provide insightsto mineral chemistrythat cannot be made been able to achievepressures and temperaturesin ex@ss in any other way. In the late 1970'sand 1980's,many of the of 20 GPa and 2000"Cin his laboratory and we expectto Americanhigh-pressure facilities in geosciencedepartments do as well. We believethat this facility will be an extremely becameinactive, and although the diamond cell was used powerful too for investigatingmantle materials,for experi- extensivelyin the USA and abroad,most of the significant mental petrology, and for synthesizingmany previously- large-volumesyntheses were done in Japan.Figure 20 con- unknown phasesof varying composition.It will also pro- trasts the capabilities of the cubic-anvil deviceswith the vide the opportunity for expandingour knowledgeof crys- diamond-anvil cell. There are at least nine ultra-high- tal chemistry phenomenaat high pressuresand temper- pressurelaboratories operating in Japanand two more are atures. under consideration.Except for laboratoriesin'the USSR, the only other such equipment in the World is in a new Summary and conclusions facility in Ringwood'slab in Australia. The successof the In conclusion,I am optimistic about the future of crystal Japaneseprogram is illustrated by the recent book edited chemistry and how it fits in with other disciplinesin the by Akimoto and Manghnani (1982) containing papers earth sciences.There needsto be more coordination and given at the U.S-Japan Conferenceon High Pressurein cooperation among the various people involved to make Geophysics,and by Sunagawa's(1983) book on Materials really significantprogress, but I think this can be done.It is Scienceofthe Earth's Interior. Thesereported the resultsof clear that momentum is building for more support of min- many important investigationsof the chemicaland physi- eral physics and mineral chemistry efforts on a national cal propertiesof materialssynthesized at temperaturesand leveland I think this will be of benefitto us all. One further pressuresthat cannot be duplicatedwith current apparatus comment:Returning to mineralogyand MSA, I have an availablein the USA. impressionfrom this meeting,more than in any before,that At Stony Brook, with support from NSF and SUNY, we very substantial changesare in store for us during the are undertaking the construction of a new ultra-high- 1980'sand 1990's.Greater understandingof mineral chem- pressurefacility featuring a Sumitomo 2000 ton pressand istry and mineral physics, improved experimental tech- cubic anvils of the split-spheretype, similar to equipment niques, and, above all, the revolution in microcomputers now being used at the Institute for Thermal Spring Re- and supercomputersare going to introduce major changes search in Misasa, Japan (cf., Ito, 1982).This apparatus in the way that we teach and conduct researchin the must be used in a two-stage arrangementwhere the six mineralogically-relatedsciences. These changes are already steelanvils presson the faceof a cube that is itself divided beginningto take place and it is up to us to adapt and to into eight tungstencarbide subcubes, each ofwhich is trun- be sure that we and our Societyare participantsand lead- catedon one interior corner to form an octahedron.Ito has ersin this revolution.
Acknowledgments Much of the research reviewed here was done in collaboration with several different colleagues and students, to whom I am very grateful for their inspiration, hard work, and friendship over the years. I particularly would like to acknowledge R. D. Shannon of LASFR - HEATED whose ideas contributed so DIAMONDANVIL E. I. du Pont de Nemours and Co. much to my thinking about crystal chemistry. This work was supported in part by a grant from the National Science Foundation. EAR83 19504. T("C) MA-8 CUBICANVIL References 2000 Akimoto, S. and Manghnani,M. H. (1982)Advances in Earth and Planetary Sciences,Vol. 12, High-PressureResearch in Geo- physics.Center for AcademicPublications, Tokyo. 'I. Bargeron,C. 8., Avinor, M. and Drickamer, G. (1971)The effectof pressureon the spin stateof iron(Il) and manganesdlV) sulfide.Inorganic Chemistry, 10, 1338-1339. Bragg, W. L. (1920)The arrangernentof ions in crystals.Philo- sophicalMagazine,,t0, 169-1 89. EXTERNALLY- HEATED (1984) DIAMONDANVIL Burdett, J. K. and Mclarnan, T. J. An orbital interpreta- tion of Pauling'srules. American Mineralogist,69, ffil-621. Chenavas,J., Joubert,J. C., and Marezio, M. (1971)Low-spin - > IOOGPo high-spin transition in high pressurecobalt sesquioxide.Solid 9, 1057-1060. 50 StateCommunications, p (Gpo) Evans,R. C. (1952')An Introduction to Crystal Chemistry.Cam- Fig. 20. Approximatepressure-temperature regimes accessible bridge University Press. with different kinds of high-pressure apparatus. The DIA-6 is a Gibbs, G. V. (1932)Molecules as model for bonding in silicates. single-stage cubic anvil device and the MA-8 has two stages. AmericanMineralogist, 67, 421450. 454 PREWITT: CRYSTAL CHEMISTRY: PAST,PRESENT AND FUTURE
Giitlich, P. (1981)Spin crossoverin (Il)-complexes. iron Structure chemistry of oxides having the corundum structure. and Bonding, Inorganic 44,83-185. Chemistry,8, 1985-1992. Hazen, R. M. and Finger, L. W. (1979) Bulk modulus_volume Prewitt, C. T. and Sleight,A. W. (1969)Garnetlike structuresof relationshipfor cation-anionpolyhedra. Journal of Geophysical high-pressurecadmium germanateand calcium germanate.Sci- Research,84, 67 23-67 29. ence,163,386-387. Hazen, R. M. and Finger, L. W. (1982) Comparative Crystal Ringwood, A. E. and Major, A. (1967)Some high pressuretrans- Chemistry.John Wiley and york. Sons,New formations of geophysicalinterest. Earth and PlanetaryScience Hazen,R. M. and Prewitt, C. T. (lgj7) Effectsof ternperatureand Letters,2, 106-110. pressure on interatomic distancesin oxygen_basedminerals. Ringwood, A. E. and Seabrook,M. (1963) High-pressurephase AmericanMineralogist, 62, 309-315. transformations in germanate pyroxenes and related com- Ida, Y. (1976) Interionic repulsive force and compressibilityof pounds.Journal of GeophysicalResearch, 68,46014609. ions. Physicsofthe planetary Earth and Interiors, 13,97_104. Rogers,D. 8., Shannon,R. D., Sleight,A. W., and Gillson, J. L. Ito, E. (1977) The absenceof oxide mixture in high-pressure (1969) Crystal chemistry of metal dioxides with rutile-related phases of Mg-silicates.Geophysical Research Letterc, 4,72_74. structures.Inorganic Chemistry,8, 841-849. Ito, E. (1982) Ultra-high pressurephase relarions of the system Sasaki,S., Prewitt, C. T., and Liebermann,R. C. (1983)The crystal MgO-FeO-SiO, and their geophysicalimplications. In Materi_ structureof CaGeO, perovskiteand the crystal chemistryof the als Scienceofthe Earth's Interior, I. Sunagawa,Ed. Terra Scien_ GdFeOr-type perovskites.American Mineralogist, 68, 1189- tific PublishingCo., Tokyo. l 198. Jayaraman,A. (1984)The diamond-anvilhigh-pressure cell. Scien_ Shannon, R. D., Chenavas,J., and Joubert, J. C. (1975a)Bond tific American,25O. 54-62. strength considerations applied to cation coordination in King H. E. Jr. and Prewitt, C. T. (19?9)Structure and symmetry normal and high-pressureoxides. Journal of Solid State Chem- of CuS, (pyrite structure). American Mineralogist, U, 1265_ istry,12,16-30. 127r. Shannon,R. D., Gumerman,P. S. and Chenavas,J. (1975b)Etrect Matsui, Y., Kawamura,K., and Syono,y. (19g2)Molecular dy_ of octahedral distortion on mean Mn3+ distances,American namicscalculations applied to silicatesystems: Molten and vit_ Mineralogist,@,'l l+7 16. reous MgSiO, and MgrSiOu under low and high pressures. Shannon, R. D. and Prewitt, C. T. (1981) This week,scitation Advances in Earth and planetary Sciences,Vol. 12, High_ classic: Effective ionic radii in oxides and fluorides. Current Pressure Researchin Geophysics,S. Akimoto and M. H. Contents,p. 18,No. 2l,May 25. Manghnani,Eds. Center for Academicpublications, Tokyo. Shannon,R. D. and Prewitt, C. T. (1968)Effective ionic radii in Matsui,Y., Onuma,A., Nagasawa,H., Higuchi,H., and Banno,S. oxidesand fl uorides.Acta Crystallographica,825, 925445. (1977)Crystal structure control in trace elementpartition be- Shannon, R. D. and Prewitt, C. T. (1969) Coordination and tween crystal and magma. Bulletin de la Soci6t6francaise de volume changesaccompanying high-pressure phase transforma- Min6ralogieet de Cristallographie,100, 311-324. tions of oxides.Materials Research Bulletin,4, 57-62. Miyamoto, M. and Takeda, H. (1980)An interpretation of the Shannon,R. D. and Prewitt, C. T. (f970a)Effective ionic radii and structure of mantle minerals at high pressuresin terms of in_ crystal chemistry.Journal of Inorganic and Nuclear Chemistry, terionic forces.Geochemical Journal, 14,243-249. 32.1427-1441. Nakai, I., Sugitani, y. K., Nagashima,K., and Niwa, (197g)X-ray Shannon,R. D. and Prewitt, C. T. (1970b)Revised values of ef- photo-electron spectroscopicstudy of copper minerals.Journal fectiveionic radii. Acta Crystallographica,826, 1046- 1 048. of Inorganic and Nuclear Chemistry,40, 7gg_791. Skelton, E. F. (1984) High-pressureresearch with synchrotron Onishi, (1978) S. A theory of the pressure-inducedhigh_spinJow- radiation. PhysicsToday, 37,4-52. spin transition of transition-metal physics oxides. of the Earth Smith,J. and Mason, B. (1970)Pyroxene-garnet transformation in and Planetary Interiors, 17,l3l}_139. Coorara meteorite.Science, 168, 832-833. Pauling, L. (1960)The Nature of the Chemical Bond. 3rd. Ed. Sunagawa,I. (1983) Materials Scienceof the Earth's Interior. Cornell University Press, Ithaca,Ny. Terra ScientificPublishing Co., Tokyo. Prewitt, C. T. (1982)Size and compressibility of ions at high pres_ Syono, Y., Ito, A., Morimoto, S., Suzuki, T., Yagi, T., and Aki sure.In S. Akimoto and M. H. Manghnani, Eds., Advancesin moto, S. (1984)Mossbauer study on the high pressurephase of Earth and Planetary Sciences,yol. 12, High-pressureResearch FerO.. Solid StateCommunications, 50, 97-100. in Geophysics,p. 43J438. Center publications, for Academic Volfinger, M. and Robert, J.-L. (1980) Tokyo. Structural control of the distribution of trace elements between silicates and hy- Prewitt, C. T. and Rajamani,V. (1974)Electron interacrionsand drothermal solutions. Geochemicaet Cosmochimica Acta, 44, chemicalbonding in p. sulfides.In H. Ribbe,Ed., SulfideMiner- 1455-1461. alogy, Vol. 1, Reviewsin Mineralogy. Mineralogical Societyof Yagi, T. and Alimoto, S. (1982)Rapid X-ray measurernentsto 100 America,Washington, D.C. GPa range and static compressionof -FerO.. In S. Akimoto Prewitt, C. T. and Shannon, (1969) R. D. Use of radii as an aid ro and M. H. Manghnani, Eds.,Advances in Earth and Planetary understandingthe crystal chemistry of high_pressurephases. Sciences,Vol. 12, High-PressureResearch in Geophysics,p. Transactionsof the American CrystallographicAssociation, 5, 433438. Centerfor AcademicPublications, Tokyo. 5l-60. Yurimoto, H. and Sueno,S. (1984)Anion and cation partitioning Prewitt, C. T., Shannon,R. D. and Rogers,D. B. (1971)Chernistry betweenolivine, plagioclasephenocrysts and the host magma: of noble metal ptCoOr, oxides. II. Crystal structuresof A new application of ion microprobe study. GeochemicalJour- PdCoOr, and AgFeOr. Inorganic Chemistry,10,7lg_723. nal, 18.8!-94. Prewitt, C. T., Shannon,R. D., Rogers,D. 8., and Sleight,A. W. Manuscript receioeil,February 4, l9g5; (1969)The C rare earth oxide-corundum transition and crvstal acceptedforpublication, February g, 19g5.