The adaptation of Pearce element ratio diagrams to complex high silica systems

Pearce element ratios (PERs, of Pearce 1968) express geochemical data in a form where variations in absolute compositions of an igneous suite can be evaluated. Generally the denominator value in the ratio is taken as a major element abundance, but it is argued here that Zr provides a more suitable choice. Zr remains incompatible in magmatic systems up to 68 wt.% SiO₂ because zircon fractionation can be suppressed by high melt temperatures and increased volatile contents. The use of Zr thus permits PER modelling to be extended to much higher levels of silica than previously investigated. However, such systems are more complex than those just involving simple basaltic magmas. Besides fractionation, the processes of magma mixing, combined assimilation and fractional crystallization, and the initial degree of partial melting in the mantle source must also be considered. To distinguish and evaluate these processes a set of example suites are investigated from a complex synextensional calc-alkaline province in the western USA. Samples within most individual suites can be modelled by fractionation, however a significant trend orthogonal to the main fractionation vector is also apparent, and open system processes are inferred. Successful modelling is achieved on an inter-suite basis using diagrams with axis functions of ([4(Ca+Na)+0.5(Fe+Mg)]/Zr versus (Si+Al)/Zr). Potential open system evolution paths between mafic end members and crustal contaminants are also displayed and evaluated on these same diagrams. The encouraging results suggest that such PER diagrams may be employed as a versatile tool for investigating the systematics of related igneous suites over a wide area.     

The statistics of Pearce element diagrams and the Chayes closure problem

Pearce element ratios are defined as having a constituent in their denominator that is conserved in a system undergoing change. The presence of a conserved element in the denominator simplifies the statistics of such ratios and renders them subject to statistical tests, especially tests of significance of the correlation coefficient between Pearce element ratios. Pearce element ratio diagrams provide unambigous tests of petrologic hypotheses because they are based on the stoichiometry of rock-forming minerals. There are three ways to recognize a conserved element: 1. The petrologic behavior of the element can be used to select conserved ones. They are usually the incompatible elements. 2. The ratio of two conserved elements will be constant in a comagmatic suite. 3. An element ratio diagram that is not constructed with a conserved element in the denominator will have a trend with a near zero intercept. The last two criteria can be tested statistically. The significance of the slope, intercept and correlation coefficient can be tested by estimating the probability of obtaining the observed values from a random population of arrays. This population of arrays must satisfy two criteria: 1. The population must contain at least one array that has the means and variances of the array of analytical data for the rock suite. 2. Arrays with the means and variances of the data must not be so abundant in the population that nearly every array selected at random has the properties of the data. The population of random closed arrays can be obtained from a population of open arrays whose elements are randomly selected from probability distributions. The means and variances of these probability distributions are themselves selected from probability distributions which have means and variances equal to a hypothetical open array that would give the means and variances of the data on closure. This hypothetical open array is called the Chayes array. Alternatively, the population of random closed arrays can be drawn from the compositional space available to rock-forming processes. The minerals comprising the available space can be described with one additive component per mineral phase and a small number of exchange components. This space is called Thompson space. Statistics based on either space lead to the conclusion that Pearce element ratios are statistically valid and that Pearce element diagrams depict the processes that create chemical inhomogeneities in igneous rock suites.     

Effects of non-conserved denominators on pearce element ratio diagrams

Pearce element ratios (PER's) have conserved denominators which have not participated in the material transfer processes that cause chemical variations in rocks. Theoretically, there is no truly conserved element (constituent) which can be used as a PER denominator because in every material transfer process all constituents have non-zero concentrations in the phases that are being transferred. Thus, constituents used as denominators of PERs may have undergone at least a small amount of material transfer. This communication investigates the degree to which a non-conserved PER denominator changes the trend of data produced by a material transfer process from that produced by the same process but plotted on a PER diagram with a truly conserved denominator. An equation is developed that utilizes the partition coefficient as the measure of the degree of involvement of the denominator constituent in the phase undergoing transfer. This equation is examined to determine how the magnitude and direction of a PER diagram data trend change with increasing involvement of the denominator constituent in the transferring phase. A set of plagioclase fractionation examples are presented which use different elements as PER denominators and consider the effects that small amounts of these elements in the plagioclase structure will have on the data trend, as a function of the element partition coefficient between crystal and melt. Results demonstrate that the direction of change in slope of a material transfer data trend is a function of the initial relative magnitudes of the numerator constituents on the PER diagram. Additionally, if the amount of involvement of a PER denominator in a separating phase is very small relative to the amount of the numerator constituents in the separating phase, there is no significant change in the data trend caused by material transfer on a PER diagram. Moreover, if the denominator constituent substitutes for a numerator constituent in the phase undergoing transfer, the intercept of the trend of the data may not converge to zero when there is a large partition coefficient, as would be expected from theory. Thus, statistical tests to determine if a PER denominator is conserved, which evaluate whether the intercept is significantly different from zero, may not be very powerful because a large amount of denominator variation is necessary before the intercept of a data trend is forced through the origin, if at all.     

Analysis of petrologic hypotheses with Pearce element ratios

Pearce element ratios can test whether the members of a rock suite are comagmatic and can illustrate the causes of chemical diversity in comagmatic suites. Comagmatic rocks have constant ratios for elements conserved in the system during changes that led to the chemical diversity. In basaltic systems, the incompatible elements, Ti, K, and P, are often conserved. The slope of the trend on a Pearce element ratio diagram is sensitive to the stoichiometry of the crystallizing and segregating phases. A judicious choice of ratios as axes for the diagram provides a signature for the phases involved and estimates of their compositions. In basaltic rocks, diagrams with Ti/K vs P/K can provide a test of the comagmatic hypothesis. Diagrams with 0.5 [Mg + Fe]/K vs Si/K have trends that are distinct for each comagmatic suite and different mineral assemblage. Different suites are distinguished by the intercepts in diagrams, whereas mineral assemblages are recognized by the slopes of the trends. For example, if olivine is the sole crystallizing and segregating phase, the trend will have a slope of 1. Diagrams with [2Ca + Na]/K vs Al/K distinguish plagioclase from augite assemblages and, in conjunction with 0.5 [Mg + Fe]/K diagrams, unravel the crystallization sequences of suites that have suffered three phase crystallization and segregation. Analyses from the Uwekahuna laccolith, Kilauea, the 1955 and 1967–68 eruptions of Kilauea, Diamond Craters Volcanic Field, Oregon, and experimental data on MORB glasses provide illustrations of the interpretations that can be obtained from Pearce element ratios.     

Modelling of igneous fractionation and other processes using Pearce diagrams

Geochemical data can be quantitatively modelled by means of Pearce diagrams. These are graphs of A/Z vs B/Z where A, B and Z are compositional abundances (e.g. wt.% SiO₂, wt.% MgO, and ppm La) and Z has the additional property of having constant absolute abundance. In the terminology of igneous petrology, Z (the common denominator variable) could be an incompatible element. The numerators (A and B) may be complex algebraic combinations of elements, or even CIPW normative abundances. The utility of Pearce diagrams lies in the fact that slopes of data distributions equal the bulk AB ratio of minerals lost or gained from a suite of cogenetic rocks. There is no distortion because these plots correct for data closure. Terms of the form A_i·Z₀/Z_i (where Z₀ is the abundance in a reference sample) remove the scaling to A_i caused by the abundance of a particular choice of Z. Subtraction of these terms for different samples (e.g. A_i·(Z₀/Z_i)-A_j· (Z₀/Z_j)) quantifies mineral losses and gains. Mathematical analysis shows that limited compatibility of the denominator variable is permitted. A bulk partition value (D) of 0.1 introduces an error of only 10% in values of A_i***-Z₀/Z_i, and 10° in slope-angle on Pearce diagrams over a crystallization interval of 50%. For D0.01 the error is minimal for a crystallization interval over 90%.     

An empirical hypothesis of deformation rate ratio criterion

Summary Based on the experiences obtained from more than 10 years' monitoring of tunnel excavation in soft ground, and in order to solve the problem that the currently-used surrounding rock stability criteria have nothing to do with the rockbolts-shotcrete support operation, thus they are hard to be applied by in-situ engineers, the author puts forward a new empirical hypothesis of criterion for surrounding rock stability assessment. Then five case histories are described as references. Furthermore, the open complex giant system methodology is introduced to explain the theory and practice of empirical hypothesis.     

The snowball Earth hypothesis: testing the limits of global change

The gradual discovery that late Neoproterozoic ice sheets extended to sea level near the equator poses a palaeoenvironmental conundrum. Was the Earth's orbital obliquity > 60° (making the tropics colder than the poles) for 4.0 billion years following the lunar‐forming impact, or did climate cool globally for some reason to the point at which runaway ice‐albedo feedback created a `snowball' Earth? The high‐obliquity hypothesis does not account for major features of the Neoproterozoic glacial record such as the abrupt onsets and terminations of discrete glacial events, their close association with large (> 10‰) negative δ<sup>13</sup>C shifts in seawater proxies, the deposition of strange carbonate layers (`cap carbonates') globally during post‐glacial sea‐level rise, and the return of large sedimentary iron formations, after a 1.1 billion year hiatus, exclusively during glacial events. A snowball event, on the other hand, should begin and end abruptly, particularly at lower latitudes. It should last for millions of years, because outgassing must amass an intense greenhouse in order to overcome the ice albedo. A largely ice‐covered ocean should become anoxic and reduced iron should be widely transported in solution and precipitated as iron formation wherever oxygenic photosynthesis occurred, or upon deglaciation. The intense greenhouse ensures a transient post‐glacial regime of enhanced carbonate and silicate weathering, which should drive a flux of alkalinity that could quantitatively account for the world‐wide occurrence of cap carbonates. The resulting high rates of carbonate sedimentation, coupled with the kinetic isotope effect of transferring the CO₂ burden to the ocean, should drive down the δ¹³C of seawater, as is observed. If cap carbonates are the `smoke' of a snowball Earth, what was the `gun'? In proposing the original Neoproterozoic snowball Earth hypothesis, Joe Kirschvink postulated that an unusual preponderance of land masses in the middle and low latitudes, consistent with palaeomagnetic evidence, set the stage for snowball events by raising the planetary albedo. Others had pointed out that silicate weathering would most likely be enhanced if many continents were in the tropics, resulting in lower atmospheric CO₂ and a colder climate. Negative δ¹³C shifts of 10–20‰ precede glaciation in many regions, giving rise to speculation that the climate was destabilized by a growing dependency on greenhouse methane, stemming ultimately from the same unusual continental distribution. Given the existing palaeomagnetic, geochemical and geological evidence for late Neoproterozoic climatic shocks without parallel in the Phanerozoic, it seems inevitable that the history of life was impacted, perhaps profoundly so.     

Low-strain integrity testing of drilled piles with high slenderness ratio

A low strain integrity test is adopted to assess the quality of cast-in situ reinforcement concrete piles with high slenderness ratios. The disadvantage of traditional NDT equipment and signal post-process is that it cannot detect and interpret any structural faults that exist in long bored piles with high slenderness ratios. Due to insufficient impact energy, testing signal decay, soil–pile interaction and insufficiently homogeneous pile concrete material, etc., only the lower frequency portion of the test signal was used in nondestructive evaluation. In order to overcome these shortcomings, the testing devices have to first be adjusted for acquiring a lower frequency signal. Secondly, numerical signal process techniques (i.e., smoothing and amplifying) should be performed to enhance the reflection signals from the pile tip. The experimental results indicate that the testing signal identification abilities can be improved by both proper device adjustments and some signal process skills. A bored pile of up to 58 m in length with a slenderness ratio of 38.6 can be identified, even with a pile cap. Finally, a new proposed numerical signal process method, i.e., joint time–frequency analysis, is applied to explore the time–frequency component of the testing result which shows better resolution than traditional methods.     

Discriminant diagrams for iron oxide trace element fingerprinting of mineral deposit types

Magnetite and hematite are common minerals in a range of mineral deposit types. These minerals form partial to complete solid solutions with magnetite, chromite, and spinel series, and ulvospinel as a result of divalent, trivalent, and tetravalent cation substitutions. Electron microprobe analyses of minor and trace elements in magnetite and hematite from a range of mineral deposit types (iron oxide-copper-gold (IOCG), Kiruna apatite–magnetite, banded iron formation (BIF), porphyry Cu, Fe-Cu skarn, Fe-Ti, V, Cr, Ni-Cu-PGE, Cu-Zn-Pb volcanogenic massive sulfide (VMS) and Archean Au-Cu porphyry and Opemiska Cu veins) show compositional differences that can be related to deposit types, and are used to construct discriminant diagrams that separate different styles of mineralization. The Ni + Cr vs. Si + Mg diagram can be used to isolate Ni-Cu-PGE, and Cr deposits from other deposit types. Similarly, the Al/(Zn + Ca) vs. Cu/(Si + Ca) diagram can be used to separate Cu-Zn-Pb VMS deposits from other deposit types. Samples plotting outside the Ni-Cu-PGE and Cu-Zn-Pb VMS fields are discriminated using the Ni/(Cr + Mn) vs. Ti + V or Ca + Al + Mn vs. Ti + V diagrams that discriminate for IOCG, Kiruna, porphyry Cu, BIF, skarn, Fe-Ti, and V deposits.     

Systematic use of trace element distribution patterns in log-log diagrams for plutonic suites

Trace element modelling has been widely used for petrogenetic interpretation of basaltic systems. This paper indicates how to select trace element pairs having very different bulk distribution coefficients (D) which when plotted on simple log-log diagrams permit the identification of the main magmatic process (magma mixing, partial melting, fractional crystallization) involved in the genesis of plutonic rocks. Fractional crystallization gives a straight line on such a diagram with a strong decrease of the compatible element whereas the concentrations of the element with D ⪡ 1 increase slowly. A similar evolution of the solids in equilibrium is observed and when data of at least one of the cumulates are directly available, it is possible to calculate the D and F parameters of the sequence of fractional crystallization.An example of this procedure is shown for a French Hercynian plutonic suite: the basic suite of Variscan Corsica.     

Structural consequences of cohesion in gravitational instabilities triggered by fluid overpressure: Analytical derivation and experimental testing

The critical taper theory of Coulomb wedges has been classically applied to compressive regimes (accretionary prisms/fold-and-thrust belts), and more recently to gravitational instabilities. Following the initial hypothesis of the theory, we provide an alternative expression of the exact solution for a non-cohesive wedge by considering the balance of forces applied to the external surfaces. Then, we use this approach to derive a solution for the case of cohesive wedges. We show that cohesion has conspicuous structural effects, including a minimum length required for sliding and the formation of listric faults. The stabilizing effect of cohesion is accentuated in the foremost thin domain of the wedge, defining a required Minimum Failure Length (MFL), and producing sliding of a rigid mass above the detachment. This MFL decreases with less cohesion, a smaller coefficient of internal friction, larger fluid overpressure ratio, and steeper upper and basal surfaces for the wedge. Listricity of the normal faults depends on the fluid overpressure magnitude within the wedge. For moderate fluid overpressure, normal faults are curved close to the surface, and become straight at depth. In contrast, where fluid overpressure exceeds a critical value corresponding to the fluid pressure required to destabilize the surface of a noncohesive wedge, the state of stress changes and rotates at depth. The faults are straight close to the surface and listric at depth, becoming parallel to the upper surface if the wedge is thick enough. We tested some of these structural effects of a cohesive wedge on gravitational instabilities using analogue models where cohesive material was subjected to pore-fluid pressure. The shape of the faults obtained in the models is consistent with the predictions of the theory.     

Systematics of modal phenocryst assemblages in submarine basalts: Petrologic implications

Phenocryst assemblages in ocean-ridge basalts generally show an increasing proportion of plagioclase as the total amount of phenocrysts increases. The variations in phase assemblages, as well as most crystal-liquid K_d's, are similar to variations (equimodal trends) predicted by low-pressure laboratory experiments, suggesting that many of these basalts have experienced varying degrees of low-pressure cyrstallization prior to quenching, with little sorting of crystals and liquid. Important exceptions include moderately to highly phyric basalts enriched either in plagioclase or olivine which lie well off the experimental trends. In these basalts, megacrysts and xenocrysts usually cited as evidence for magma mixing commonly represent a small proportion of the total crystalline phase assemblage. However, phase proportions for many of these basalts lie well outside the range that could be produced by simple mixing; selective gravitative sorting either prior or subsequent to mixing appears to be the likely explanation for these phyric basalts. A relation between spreading rate and phase proportions is neither supported nor refuted by the data, which as yet do not adequately represent fast-spreading ridges. Pyroxene-phyric varieties are especially common among LIL-element enriched (Group 2) basalts, and these basalts also show the greatest abundance of olivine-enriched (picritic) samples. Selective enrichment in plagioclase is more common among LIL-element depleted (Group 1) basalts, and pyroxene appears in Group 1 basalts only at relatively high degrees of crystallinity. These differences are consistent with expected compositional effects (including volatiles) on phase boundaries, as well as likely differences in depth (pressure) of mantle melting and magma fractionation. Sparsely to moderately phyric basalts tend to contain only olivine (±spinel) as phenocrysts, and lie in the olivine field in the projection from plagioclase in the CMAS tetrahedron. This is consistent with the concept that these magmas approach low-pressure equilibrium by olivine fractionation from a more picritic parent. The origin of these basalts, and relationships between them, remains an important fundamental problem. Phenocryst phase assemblages are consistent with the low-pressure phase saturation indicated by the projected positions of the associated glasses in CMAS. It is suggested that, in contrast to the classical practice of classifying basalts according to phase proportions, a classification based on presence and/or first appearance of each crystalline phase is both practical and petrogenetically significant for water-quenched submarine basalts.     

The identification and assessment of spurious trends in Pearce-type ratio variation diagrams: a discussion of some statistical arguments

Spurious trends in Pearce-type ratio variation diagrams are constrained to go through the origin and the average composition of the data set. A valid trend has a significant (nonzero) y-intercept and a meaningful slope. The two types of trends are thus easily distinguished. Numerical simulation is a useful, perhaps even necessary, method of checking the effects of random variations. Calculations of probability suggest that the odds against purely random data mimicking an olivine trend in komatiites are great — perhaps as much as 100 00 0000 to 1. That such original igneous trends are descernible through the veil of metamorphic overprinting is a powerful demonstration of the usefulness of this technique.     

K-Rb ratio in high grade metamorphism: A confirmation of the hypothesis of a continual crustal evolution

The K:Rb ratio (R) in a metamorphic sequence with increasing metamorphism of the West Alps has been analysed. By a regression analysis a comparison is made with the trends observed for R byShaw (1968) in many magmatic processes. The K:Rb ratio in the metamorphic rocks shows a marked increase with increasing metamorphism in relation to a removal of the K content. The arithmetic averages of R in the rocks of amphibolite and granulite facies are 231 and 505 respectively. In some samples of granulite facies the R values are higher than those of the continental alkalic basalts and approximate to the oceanic tholeiites. The variations in the K:Rb ratio may be imputed, according toShaw, to an anatexis process which took place under high grade metamorphic conditions. The marked increase of the K:Rb ratio in the deep material would be caused by partial separation between K and Rb in the granulite assemblages and by an upward migration of Rb, relative to K, from the deep crustal levels of granulite facies. These considerations seem to confirmShaw’s hypothesis of the possibility of a marked crustal stratification and consequently of a continual crustal evolution.     

Estimation of Probabilities of Three Kinds of Petrologic Hypotheses with Bayes Theorem

Physical-chemical explanations of the causes of variations in rock suites are evaluated by comparing predicted to measured compositions. Consistent data turn an explanation into a viable hypothesis. Predicted and measured values seldom are equal, creating problems of defining consistency and quantifying confidence in the hypthesis. Bayes theorem leads to methods for testing alternative hypotheses. Information available prior to data collection provides estimates of prior probabilities for competing hypotheses. After consideration of new data, Bayes theorem updates the probabilities for the hypotheses being correct, returning posterior probabilities. Bayes factors, B, are a means of expressing Bayes theorem if there are two hypotheses, H ₀ and H ₁. For fixed values of the prior probabilities, B > 1 implies an increased posterior probability for H ₀ over its prior probability, whereas B < 1 implies an increased posterior probability for H ₁ over its prior probability. Three common problems are: (1) comparing variances in sets of data with known analytical uncertainties, (2) comparing mean values of two datasets with known analytical uncertainties, and (3) determining whether a data point falls on a predicted trend. The probability is better than 0.9934 that lava flows of the 1968 eruption of Kilauea Volcano, Hawaii, are from a single magma batch. The probability is 0.99 that lava flows from two outcrops near Mount Edziza, British Columbia, are from different magma batches, suggesting that the two outcrops can be the same age only by an unlikely coincidence. Bayes factors for hypotheses relating lava flows from Volcano Mountain, Yukon Territory, by crystal fractionation support the hypothesis for one flow but the factor for another flow is so small it practically guarantees the fractionation hypothesis is wrong. Probabilities for petrologic hypotheses cannot become large with a single line of evidence; several data points or datasets are required for high probabilities.     

Damage characterization during laboratory strength testing: A 3D-finite-discrete element approach

A combined finite-discrete element approach is used to simulate the complete 3D fracture process during conventional laboratory testing, including Brazilian indirect tension and uniaxial and biaxial compression. A typical granite rock type (based on the Lac du Bonnet granite) was simulated to investigate the fracture pattern and mechanical strength of brittle rock in the laboratory. Damage intensity parameters (D₂₁ and D₃₂) are introduced and utilized to characterize the induced damage in the models. These parameters provide an improved representation of the cumulative associated damage and facilitate a quantitative characterization of crack intensity during testing. The numerical simulations included both 3D and 2D models, and show that there is a good agreement between the strength response derived from simulations both in 3D and 2D and the considered rock material. A good correlation also exists between the fracture pattern in 3D and the equivalent 2D models. The influence of confinement on the biaxial strength and the associated damage in compression is investigated. While axial splitting is the dominant failure mode at low confinement, finite-discrete element simulations show that a shear failure mode tends to dominate as the confinement increases. The dependency of dilation upon the confining pressure is also demonstrated, the dilation angle decreasing with increased confinement.     

Do major oxide tectonic discrimination diagrams work? Evaluating new log‐ratio and discriminant‐analysis‐based diagrams with Indian Ocean mafic volcanics and Asian ophiolites

Many geochemical diagrams exist that classify old volcanic terranes of ambiguous provenance into various modern plate tectonic settings, with variable success. Recently proposed diagrams, based on log‐ratios and linear discriminant analysis with large datasets of major oxides, were tested here with data for ocean island, arc and mid‐ocean ridge lavas from the Indian Ocean. Success rates are 45–100%, with misclassifications potentially caused by alteration, although alteration demonstrably need not cause misclassification. The diagrams were further applied to some Asian ophiolites, representing Tethyan and Indian ocean crusts, to see if the diagrams confirm their tectonic setting inferred from trace and isotopic data. Lower success rates (30–60%, but 75–100% for specific suites) are not surprising in view of the ubiquitous and complex alteration in ophiolites. Log‐ratio transformation and linear discriminant analysis appear to be powerful methods when discrimination diagrams are based on major oxide data alone.     

识别洋陆转换的岩石学思路——洋内弧与初始俯冲的识别

阐述了洋陆转化形成的洋内弧与初始弧的岩石组合序列及其地球化学特征,提出岩浆弧是由洋陆转化以及底侵的壳幔转化共同作用形成的认识,前弧环境是洋陆转化形成初生大陆的场所,由特征的类似洋中脊的洋内弧前弧玄武岩类构成。大陆的形成过程如下:从地幔中生长出洋壳,从洋壳中的洋陆转化生长出不成熟的弧陆壳,最后从弧陆壳底侵的壳幔转化中长出成熟的陆壳。这样,地壳的生长和形成主要通过岩浆增生作用来实现。     

岩石稀土元素分布曲线图绘制的VB方法

岩石稀土元素分布曲线图是地质研究领域中经常涉及的图件之一,用传统的方法制做,因涉及对数坐标绘制起来较为繁琐,而精度也较低.利用基本的VB语言可设计出这类图的制作程序,大大提高工作效率. 1 设计思路及程序设计 先设计坐标系统, X坐标为各稀土元素,一般每一样品分析14个稀土元素,Y坐标为元素的含量(刻度值为对数值),再将数值输入计算机中并标准化,然后绘制图形,最后保存图形.以下为VB6程序,在WIN98和WINXP下运行通过[1]. 1.1 坐标系统的建立 首先建立一标准EXE工程,设窗体Form1的Scale Mode为厘米,在窗体内放置一图片框,大小…     

Petrologic evolution of partially molten cumulate: the Atok section of the Bushveld Complex

The postcumulus evolution of a portion of the Bushveld Complex that includes the Merensky reef is inferred from the study of a continuous 56 m drill core. The core penetrated the basal orthopyroxenites of the Merensky and Bastard units, the massive anorthosites overlying the two pyroxenites and about 10 m of norite underlying the Merensky pyroxenite. Detailed profiles of major, minor and rare earth element (REE) contents of clinopyroxene and orthopyroxene were determined by electron and ion probe. Good correlations exist between textural and lithological variations and the REE contents of the pyroxenes. Specifically, enrichments in pyroxene REE abundances are observed in the basal pyroxenites of the Merensky and Bastard units relative to the underlying and overlying rocks. In the pyroxenites the Nd content of clinopyroxene is typically over 12 ppm and reaches nearly 40 ppm (≈90 × chondrite), and Nd/Yb ratio is in the range 8 to 25. These extreme enrichments in REE are not accompanied by large variations in major element contents. Computations of the compaction parameters relevant to the conditions of crystallization of the Bushveld Complex combined with a consideration of cooling history confirm the importance of compaction as a post-cumulus process. This analysis indicates that the geochemical variability is a result of redistribution of interstitial melt driven by compaction and cannot reflect variations in the initial porosity of the accumulating crystal pile. A model for the development of the Atok section is developed. The Merensky anorthosite is interpreted to have served as a barrier to the upward porous flow of late-stage, hydrous and incompatible-element enriched melt, which was thus trapped in the underlying Merensky pyroxenite. The flow was driven by compaction of the 350+ meter-thick section of predominantly norite beneath the anorthosite. The introduction and accumulation of this melt in the pyroxenite and subsequent cooling resulted in partial dissolution, recrystallization and REE enrichment of the rock forming minerals, and in the formation of the main lithologic features of the Merensky pyroxenite. Further upward percolation of the interstitial melt through the Merensky anorthosite was restricted to channels due to the relatively impermeable nature of the cemented anorthosite. This melt accumulated in and metasomatized the Bastard pyroxenite in the same manner as the Merensky pyroxenite, having again been trapped by the overlying Bastard anorthosite. Received: 10 July 1996 / Accepted: 27 February 1997