Yu. S. Genshaft
Schmidt United Institute of Physics of the Earth, Russian Academy of Sciences
I. P. Ilupin
Central Research Institute for Base and Noble Metals
Spinel is a typomorphic mineral in mafic and ultramafic rocks of various depth facies. As a rule, spinel, a mineral that forms a variety of solid solution series with end members such as noble spinel, hercynite, magnetite, ulvospinel, picrochromite, and chromite, has a range of proportions of the following cations: Al+3, Cr+3, Mg+2, Fe+2, Fe+3, and Ti+4. The principal microcomponents are Mn and Zn. In deep-seated rocks of ultramafic composition, the most common varieties are chromian spinels. Rocks with spinels of such composition are represented by alpine-type ultramafites, nodules in alkaline basalts, kimberlites, and lamproites, as well as inclusions in deep-seated minerals, of which diamonds are of the greatest interest. Accordingly, one can discuss 5 groups of chromian spinel compositions corresponding to the rock types just listed.
We studied chemical compositions of chromian spinels from the above 5 groups by means of mathematical statistics methods--primarily, hierarchic cluster analysis using nearest neighbor search in Euclidean space and principal-component factor analysis. Minerals having more than 2% Cr2O3 were classed with chromian spinels. Besides, a chemical screen of TiO2<8 % was introduced. High-Ti chromites are rare in terrestrial rocks and common in Lunar rock samples; they are crystallized under highly reducing conditions [Agrell et al., 1970]. This type of chromian spinels is beyond the scope of our study. Our data set, totaling 1821 analyses, was represented by spinels from rocks of ultramafic massifs (alpine-type ultramafites, n = 318), nodules from basalts (n = 340), lamproites (n = 106), kimberlites (n = 928), and from diamond assemblages (n = 129). Overall, the spinels have broad ranges of all the components selected for analysis - Ti, Al, Cr, Fe, Mg, and Mn oxides. Inasmuch as the overwhelming majority of analyses are X -ray spectral microanalysis (PCM) data, the processing involved total iron as ferric iron, FeO . When considering spinel compositions averaged within the established clusters, the proportion of ferric iron, calculated stoichiometrically using the well-known MINFILE program, was taken into account. Mathematical processing of the sets of analyses was carried out using the Statistica 5.0 program [Borovikov and Borovikov, 1997].
When compiling the data base for spinel chemical compositions, it caught our attention that different authors cite appreciably dissimilar chromite compositions for the same kimberlite pipes. Thus, [Bobrievich et al., 1964; Garanin et al., 1980; Sobolev et al., 1975] listed partial and complete compositions of spinels for the Mir, Aikhal, and Udachnaya pipes of the Yakutian kimberlite province, which we compared to I. P. Ilupin's unpublished data (Table 1). Grains to be analyzed were taken from concentrates or from hand samples; Bobrievich et al.  give results obtained by wet chemistry from a large number of grains, while the rest of the data come from PCM measurements. It is readily apparent that Al and Ti have appreciable ranges of variation. Sobolev et al.  surmised that deviations from the data of [Bobrievich et al., 1964] are due to differences in beneficiation and spinel separation techniques. However, the same sampling procedures, when used by other workers, yielded strongly dissimilar results. The cause of these discrepancies, to our mind, remains unclear.
A study on chromian spinel compositions in ultramafites of the Kempirsai massif was used to demonstrate that in the grain size range between < 40 mcm to 400 mcm, Cr and Ti contents decrease with grain size [Pustotsvetov et al., 1992]. This effect is especially conspicuous in grains of < 200 mcm. In peridotites of the Miamori massif and in nodules from basalts of Ichinomegata crater, the Mg-number of spinels was shown to depend on grain size, with spinel Cr-number remaining virtually constant [Ozawa, 1983]. In order to further verify the grain-size dependence of spinel compositions, we correlated data obtained for grains ranging 0.2-0.4 mm (n = 26) and 0.6-0.9 mm (n = 20) from the Yakutskaya (eastern part), Aerosyemochnaya, Ukrainskya (Daldynskoe field) and Zapolyarnaya (Munskoe field) pipes. Average values and standard deviations for each oxide proved to be closely similar in both data sets (Table 2). Hence, for grains in the range 0.2-0.9 mm, no grain-size dependence for spinel composition has been established.
A review of published data shows that the main parameters used in compositional classifications for chromian spinels are their #Cr = Cr/(Cr+Al), #Mg = Mg/(Mg+Fe+2 ), and proportion of ferric iron #Fe+3 = Fe+3/(Fe+3 +Cr+Al) [Basu and Mac Gregor, 1975; Bazylev, 1995; Dick and Bullen, 1984; Irvine, 1967; Rasskazov and Genshaft, 1987; Roeder, 1994; Sack and Ghiorso, 1991]. Diagrams plotted on these parameters as a reference frame discriminate spinel fields for deep-seated nodules, alpine-type and stratiform peridotites [Irvine, 1967], and alpine-type and abyssal ultramafites [Dick and Bullen, 1984], also demarcating compositional fields of spinels from kimberlites and alkaline basalts [Basu and Mac Gregor, 1975]. A variety of studies have shown spinel compositions, primarily #Cr and #Mg, to be defined largely by the type of their host ultramafites. According to Dick and Bullen , spinel composition depends first of all on melt composition and crystallization pressure--i.e., it is a function of the degree of mantle melting. These workers identified three rock groups as to spinel composition, with #Cr < 0.6 (Group I), #Cr > 0.6 (Group III), and Group II, with #Cr transitional between the values characteristic of Groups I and III. An essential distinction between Group I and Group III peridotites is whether or not the restite contains clinopyroxene. Based on this criterion, abyssal peridotites belong to Group I. Group III comprises alpine-type peridotites only, albeit some of them can be classed with Group I. It is shown that 70% of abyssal peridotites dredged from mid-ocean ridges are plagioclase-free spinel harzburgites and lherzolites. Mineralogically and texturally, these resemble the pertinent alpine-type peridotites and are interpreted as crystalline residues after partial melting of ultramafic mantle material. Judging by CIPW norms for clinopyroxene, which point to the melting degree, alpine-type peridotites are more refractory than abyssal peridotites.
Different genetic types of peridotites, including those from ophiolite assemblages metamorphosed in the amphibolite facies, have been discussed in [Bazylev, 1995; Bazylev and Kamenetsky, 1998]. Spinel composition is indicative of formative conditions for these rocks. Thus, a case study on Macquarie Island ophiolite, SW Pacific, exemplifies variations in #Cr, #Fe +3, and TiO 2 in spinels from spinel- and plagioclase harzburgites, plagioclase cumulate peridotites, and olivine orthopyroxenites. In terms of spinel #Cr, the rocks encompass all three groups delineated in [Dick and Bullen, 1984]. Bazylev and Kamenetsky  analyzed the effect of redox conditions on spinel composition at different stages of metamorphic recrystallization of ultramafic rocks.
Spinel compositions specific to kimberlitic sources are delimited in [Bushueva et al., 1991], and for a broader range of rocks, including non-kimberlitic sources (ultramafic nodules in basalts), in [Matsyuk et al., 1989]. It is shown that, overall, with an appreciable overlap of spinel compositions in rocks of the series dunite-harzburgite-lherzolite, increase in the proportion of pyroxenes (especially, clinopyroxene) leads to a decrease in #Cr and an increase in #Mg of the mineral in point [Matsyuk et al., 1989; Roeder, 1994]. This is particularly notable when correlating different types of ultramafites from the same kimberlite deposit (e.g., Udachnaya pipe). Dawson  reports inverse relations between ultramafite types and spinel compositions, with spinel #Cr being highest in lherzolites and lowest in harzburgites.
Spinels associated with diamonds make the highest-Cr group [Fipke, 1994; Sobolev, 1974]. A recent study [Vaganov et al., 1999) was dedicated to chromian spinels as indicators for the appraisal of diamond potential of rocks based on analysis of a data bank for spinel compositions from kimberlites, lamproites, komatiites, meymechites, certain types of basalts, and alpine-type peridotites. The scope of that study was limited to the discussion of data presented in terms of #Fe+3 -#Fe+2, #Cr-#Fe+2, Cr2O3 -MgO, and some other diagrams; it focused solely on the character of variation trends for chromites from different rock assemblages without identifying the genetic nature of these trends. It is shown that the use of a number of diagrams affords the delineation of the fields for spinels from kimberlites and lamproites, including the discrimination between diamondiferous and diamond-free types of rocks.
A number of studies [Bagdasarov, 1983; Bagdasarov and Poroshin, 1980; Bagdasarov et al., 1979] address compositional features of chromian spinels from various mafite-ultramafite rocks, including ultramafites of Kamchatka and the Maimecha-Kotui region, Kilauea basalts, Archean komatiites, Guli dunite-peridotite intrusion, Uralian and Sikhote Alin picrites, layered intrusions, alpine-type massifs, and kimberlites. On #Cr-#Mg and #Fe +3 -#Mg plots, various rock assemblages are shown to have both overlapping fields of spinel compositions and distinctions in terms of their fields and variation trends. Relatively limited analytical data are used to draw important conclusions on how spinel compositions depend on bulk rock composition, crystallization depth and temperature, and oxidation conditions. The tenet of the "early igneous" variation trend for spinel compositions, as inferred from experimental data [Hill and Roeder, 1974] on positive correlation between #Cr and #Mg, is further supported. Negative correlation, according to [Bagdasarov et al., 1979], is inherent in spinels from alpine-type ultramafites, peridotite xenoliths in kimberlites, and constituent minerals of kimberlites proper.
Noteworthy is the review of Roeder , who analyzed a voluminous data bank comprising some 17,000 spinel analyses from all types of terrestrial rocks and meteorites. This study addresses the nature of variability of spinel compositions and certain traits of spinel compositions from igneous and metamorphic rocks.
Compositions of chromian spinels were used to classify mantle peridotites [Palandzhyan and Dmitrenko, 1989; Poroshin, 1988] and a large group of volcanic rocks and peridotites from a range of tectonic structures [Young Il Lee, 1999] with respect to tectonic setting.
Clusterization of compositions in each of the five groups shows that, alongside large clusters, i.e., those incorporating large numbers of analyses, there are small clusters, which comprise 1 to 10 compositions. For this reason, clusterization was carried out so as to obtain at least three large clusters. In each group, thus, a specific clusterization step was selected. Small clusters were rejected from further consideration. Overall, in each of the five groups, proportions of those analyses that entered small clusters are 6% for spinels from massifs, 17.6% for spinels from "basalts," 5% for spinels from "kimberlites," 23.6% for spinels from "lamproites," and 29.4% for "diamond" spinels. The proportion of large to small clusters is as follows: spinels from massifs, 7 to 10; spinels from "basalts," 10 to 22; spinels from "kimberlites," 9 to 12; spinels from "lamproites," 4 to 5; and "diamond" spinels, 3 to 11.
Cluster and factor analyses enabled us to identify the most significant compositional components responsible for spinel variability in each particular group and in individual clusters. In this context, the groups of spinels of the kimberlitic, lamproitic, and diamond types can be contrasted with spinels of alpine-type massifs and the basalt type. In the former group, compositional variability is due mainly to changes in #Mg and abundances of iron (ferric, ferrous, and total), whereas the latter is characterized primarily by variations in Cr an Al contents. This feature is clearly seen from the factor load diagram in the first/second factor coordinates, where joint contribution of both factors to compositional variability of spinels is 85% (Figure 7). Table 3 lists factor analysis results from each of the 5 groups, including large and small clusters alike. One can see that each group has its distinctive features of spinel composition variations, with compositional components having different contributions to these variations. Overall, it can be ascertained that all the spinels in point are characterized primarily by isomorphic replacements of Al for Cr and of Mg for Fe, which is in perfect agreement with the previously published data of various workers. Spinels from kimberlitic and lamproitic sources, often including high-Ti varieties, are distinguished by their compositional variability being largely due to contribution from Ti variations.
Clusterization of the entire set of compositions (n = 1821) yielded 40 clusters, of which 6 accommodate over 100 analyses each, and 10, between 10 and 100. Diamond-related spinels entered chiefly the 5 large clusters (89% of the entire sample). In none of the clusters, however, did they constitute a majority of analyses. The relationships are as follows: Cluster 1, a total of 166 analyses, including 64 diamond-related, 61 kimberlitic, 23 from massifs, 15 lamproitic, and 3 basaltic; Cluster 2, a total of 39 analyses, including, in the same order, 18, 11, 5, 3, and 2; Cluster 3, a total of 148 analyses, including 11, 86, 39, 15, and 7; Cluster 4, a total of 246 analyses, including 26, 118, 86, 8, and 8; and Cluster 5, a total of 41, including, respectively, 6, 9, 19, 0, and 7. A more appreciable contribution to certain large clusters comes from kimberlitic and basaltic types of spinel. Thus, the cluster with n = 367 has 321 kimberlitic spinels (29 lamproitic, 4 diamond-related, 11 basaltic, and 2 from massifs); in the cluster with n = 16, all the spinels are kimberlitic in type (the highest-Ti ones having 4.58% TiO2 ); both clusters are distinguished by high TiO2. In the cluster with n = 185, the basaltic type is represented by 160 analyses (in addition, 17 are from massifs, 7 are kimberlitic, and 1 is lamproitic type). One cluster (n = 32) consists almost wholly of spinels from massifs (n = 30; and 2 are kimberlitic-type).
Average spinel compositions of the 5 groups, standard deviations, and minimum and maximum values of each component (Table 5), all indicate that the most uniform group of spinels is the one associated with diamonds. Their closest counterpart in terms of Cr2O3, FeO, and MgO contents is spinels from lamproites. Spinels from nodules in basalts differ most strongly from the rest of the groups in having the highest #Mg and Al and the lowest Cr abundances.
Experimental data point to a strong dependence of spinel #Cr on oxygen fugacity [Hill and Roeder, 1974]: increase in pO2 causes a decrease in Cr2O3, a slight decline in Al2O3, and a strong increase in the proportion of ferric iron and #Mg of spinel crystallizing from basaltic melt at T = 1200o C. With increasing crystallization pressure, Cr content decreases and Al increases in spinel crystallizing at a constant composition of the melt [Bazylev, 1995; Dick and Bullen, 1984]. It has been demonstrated using case studies on basaltic and kimberlitic melts [Roeder, 1994] that the composition of crystallizing medium and changes in oxygen fugacity during crystallization bear strongly on the resultant spinel composition.
According to the spinel classification proposed in [Palandzhyan and Dmitrenko, 1989], a large group of clusters of basaltic spinels represents subcontinental mantle rocks (lherzolites and diopside harzburgites), least depleted in basaltic components. The suboceanic type (#Cr = 0.25-0.5), or Group I--the main body of abyssal peridotites [Dick and Bullen, 1984]--is represented virtually by a single cluster of spinels from alpine-type massifs. The rest of the clusters, or the bulk of alpine-type peridotites of Group III of the same authors, belong chiefly to the island-arc type (#Cr = 0.6-0.9), comprised of extremely depleted, olivine-rich harzburgites. The transitional type from oceanic to island-arc deep seated material (#Cr = 0.5-0.6) is portrayed by 2 basaltic clusters, 1 kimberlitic cluster, and 1 cluster for spinels from massifs. Therefore, spinel compositions from ultramafic rocks of continental structures imply that at depth beneath these structures, there exist physicochemical conditions of crystallization typical of all the principal tectonic features of the earth--continents, oceans, and transition zones (island arcs).
In conclusion, we must emphasize once again (i) the detached character of most spinel-bearing rocks associated with basalts and (ii) the marginal position of spinel compositions from diamonds in the majority of diagrams. In terms of bulk chemical composition, the closest counterpart to diamond-related spinels is spinels from lamproites. The greatest contribution to compositional variability of spinels of different groups comes from isomorphic replacement of Al for Cr and of Fe for Mg.
Afanasiev, V. P., N. P. Pokhilenko, A. M. Logvinova, and E. S. Yefimova, Problem of false indicators for kimberlites and lamproites (on the example of chromites), 7 Int. Kimberlite Conf., Extend. Abstracts, 7-8, Cape Town, 1998.
Agrell, S. O., A. Peenett, F. R. Boyd, et al., Titanian chromite, aluminian chromite and chromian ulvospinel from Apollo 11 rocks, Proc. Apollo 11 Lunar Sci. Conf., 1, 1970.
Bagdasarov, E., Comparative characteristics and peculiarities of compositional variations of accessory chromian spinels, in: Composition and properties of deep-seated crustal and upper mantle rocks of gratons, pp. 191-221, Nauka, Moscow, 1983.
Bagdasarov, E., and E. Poroshin, Peculiarities of compositional evolution and crystallization conditions for garnets and chromian spinels of ultramafic, mafic, and related rocks (in Russian) (abstract), All-Union Symposium on Mantle Nodules and the Issues of Ultramafic Magmas, 69-70, Nauka, Novosibirsk, 1980.
Bagdasarov, E., E. Landa, and B. Markovsky, Chemical composition and conditions of crystallization of chromian spinels from ultramafic volcanites and other rocks of the ultramafite-mafite series (in Russian), Zap. Vses. Mineral. O-va, 106, (5), 524-535, 1979.
Basu, A. R., and I. D. Mac Gregor, Chromite spinels from ultramafic xenoliths, Geochim. Cosmochim. Acta, 39, (6/7), 937-945, 1975.
Bazylev, B., Compositions of clinopyroxene and spinel from residual ultramafites as indicators of formative conditions and composition of cognate primary mantle magmas (in Russian), Geokhimiya, 7, 915-924, 1995.
Bazylev, B., and V. Kamenetsky, Origin of peridotites of the Macquarie Island ophiolite assemblage, SW Pacific (in Russian), Petrologiya, 6, (4), 363-380, 1998.
Bobrievich, A., I. Ilupin, I. Kozlov, et al., Petrography and Mineralogy of Kimberlitic Rocks of Yakutia, 191 pp., Nedra, Moscow, 1964.
Borovikov, V., and I. Borovikov, Statistica: Statistics Analysis and Data Processing in Windows Environment, 608 pp., Filin, Moscow, 1997.
Bushueva, E., V. Garanin, and G. Kudryavtseva, Data Banks, Chemical Characteristics, and Classification of Spinellids from Kimberlites, Moscow State University, Moscow, available from VINITI, 2879-V92, 1991.
Dawson, J. B., Kimberlites and Their Xenoliths, 250 pp., Springer-Verlag, Berlin, 1980.
Dick, H. J. B., and T. Bullen, Chromian spinel as a petrogenetic indicator in abyssal and alpine-type peridotites and spatially associated lavas, Contrib. Mineral. Petrol., 86, (1), 54-76, 1984.
Fipke, C. E., Significance of chromite, ilmenite, G5 Mg-almandine garnet, zircon and tourmaline in heavy mineral detection of diamond bearing lamproite, in Kimberlites, Related Rocks and Mantle Xenoliths, SPRM Spec. Publ. 1A, 1, 366-381, 1994.
Garanin, V., G. Kudryavtseva, and T. Posukhova, Composition, Micromorphology, and Some Physical Properties of Chromian Spinels from the Udachnaya Kimberlitic Pipe (in Russian), paper presented at the 7th Conf. of Young Scientists (Geochemistry and Mineral Deposits), Geology Department, Moscow State University, Moscow, 1980, available from VINITI, no. 3144, 1980.
Genshaft, Yu., Structure of deep-seated nodules as a reflection of dynamic processes in the lithosphere (in Russian), Fiz. Zemli, (10), 44-59, 1993.
Genshaft, Yu., A. Saltykovsky, and G. Budzinsky, Compositional characteristics of Tertiary basalts and their deep-seated nodules from Thuringia and Saxony, the German Democratic Republic, in: Properties and State of Mineral Matter in the Earth's Interior, pp. 8-35, Nauka, Moscow, 1989.
Hill, R., and P. Roeder, The crystallization of spinel from basaltic liquid as a function of oxygen fugacity, J. Petrol., 82, (6), 709-729, 1974.
Ilupin, I., Chemical composition of chromian spinels and inhomogeneity of the Daldynskoe kimberlitic field (in Russian), Geokhimiya, 6, 603-608, 1997.
Irvine, T. N., Chromium spinel as a petrogenetic indicator, Part 2: petrologic applications, Canad. J. Earth Sci., 4, 71-103, 1967.
Matsyuk, S., A. Platonov, E. Polshin, et al., Spinellids of Mantle Rocks, 212 pp., Naukova Dumka, Kiev, 1989.
Ozawa, K., Evaluation of olivine-spinel geothermometry as an indicator of thermal history for peridotites, Contrib. Mineral. Petrol., 82, (1), 52-65, 1983.
Palandzhyan, S., and G. Dmitrenko, Typization of mantle peridotites based on composition of accessory chromian spinels (in Russian), Dokl. Akad. Nauk SSSR, 307, (3), 688-691, 1989.
Paneyakh, N., Igneous ultramafite suites identified from chromian spinel composition, in: Continental Crust in Space and Time: Magmatism, pp. 236-243, Nauka, Moscow, 1989.
Poroshin, E., High-Cr spinels and issues of origin of basaltic magmas (in Russian), Geol. Geofiz., 8, 39-47, 1988.
Prikhodko, V., Compositional variations of chromian spinels: Implications for ultramafite petrology (in Russian), Geol. Geofiz., 7, 67-73, 1980.
Pustotsvetov, A., E. Mitina, and A. Ukhanov, Inhomogeneity of accessory chromian spinel as a possible geothermometer (in Russian), Geokhimiya, (10), 1412-1422, 1992.
Rasskazov, S., and Yu. Genshaft, Spinels from deep-seated nodules in Late Cenozoic basaltic rocks from Transbaikalia (in Russia), Min. Zhurnal, 9, (6), 22-30, 1987.
Roeder, P. L., Chromite: from the fiery rain of chondrules to the Kilauea Iki lava lake, Canad. Mineral., 32, (4), 729-746, 1994.
Sack, R. O., and M. S. Ghiorso, Chromian spinels as petrogenetic indicators: thermodynamics and petrological applications, Am. Mineral., 76, 827-847, 1991.
Sobolev, N., Deep-seated Inclusions in Kimberlites and the Issue of Upper Mantle Composition, 264 pp., Nauka, Novosibirsk, 1974.
Sobolev, N., N. Pokhilenko, Yu. Lavrentyev, and L. Usova, Compositional peculiarities of chromian spinels from Yakutian diamonds and kimberlites (in Russian), Geol. Geofiz., 11, 7-24, 1975.
Vaganov, V., I. Ilupin, and A. Kocherov, Chromian spinels as indicators of diamond potential (in Russian), Rudy i Metally, 3, 35-47, 1999.
Young Il Lee, Geotectonic significance of detrital chromian spinel: a review, Geosci. J., 3, (1), 23-29, 1999.
Zhelyazkova-Panayotova, M., On a rational classification of the spinel group from ultrabasic rocks, Mineral. Soc. Japan. Spec. Pap., 1, 174-179, 1971.