Terminology for Layered Intrusions

Transcription

1 Terminology for Layered Intrusions by T. N. IRVINE Geophysical Laboratory, Carnegie Institution of Washington, Washington, D.C (Received 3 April 1981; in revised form 19 October 1981) WITH EIGHT PLATES 'Words, as is well known, are the foes of reality' Joseph Conrad (1911), Under Western Eyes ABSTRACT The terminology of igneous cumulates was originally formulated on the concept that the rocks were concentrations of minerals accumulated by crystal settling. It is now well established for several major intrusions that plagioclase that had been typed as cumulus should, in fact, have floated in its parental magmatic liquid. Nevertheless, the terminology has been valuable in the classification of the rocks of these and other intrusions, and it has been useful in certain aspects of their genetic interpretation. The name 'cumulate' is therefore redefined so that crystal settling is a possible but not essential process in the origins of the rocks to which it is applied. The overall nomenclature of layered intrusions is then reviewed, and definitions of names are refined and systematized to clarify their applications. Careful distinction is made between descriptive and genetic names. A cumulate is defined as an igneous rock characterized by a framework of touching mineral crystals and grains that evidently were concentrated through fractional crystallization of their parental magmatic liquids. A layer is a sheetlike unit of cumulate that is a distinctive entity in its compositional and (or) textural features. A lamina is a thin (2-5 cm), sharply defined layer. Igneous lamination is a general name for planar structure that is pervasive through cumulates on the scale of their grain size. Layering is the overall structure and fabric manifest through combinations of layers, laminae, and lamination. The difference between layers and layering is essentially the difference between individuals and a population, and the terms that describe one are not necessarily appropriate to the description of the other. But the fundamental variants of all of layering, lamination, grading in layers, and layer contacts can essentially be described by one or more of the four modifiers, 'grain-size', 'modal', 'textural', and 'cryptic', where the last refers to cumulus-mineral compositions. A system for subdividing layered intrusions is proposed in which distinction is made between general divisions ranked as series, zones, and subzones, and prominent units that are called groups, members, rhythmic units, and cyclic units. Applications to the Skaergaard, Bushveld, Muskox, Stillwater, and other intrusions are briefly described. INTRODUCTION In 1960, Wager, Brown, & Wadsworth proposed the term 'cumulate' as a group name for igneous rocks formed by crystal accumulation, particularly in layered intrusions. They also coined 'orthocumulate', 'mesocumulate', and 'adcumulate' for application to distinctive variants. The accumulated crystals were called 'cumulus crystals', and the pore liquid that was inferred to be trapped among them was termed 'intercumulus liquid'. Wager & Brown (1968) subsequently expanded the nomenclature, and Jackson (1967) independently added other names, among them the major adjective, 'postcumulus'. Most of the terms have come to be widely used, not only for layered intrusions, but also in investigations of lunar rocks and in studies of ultramafic and gabbroic nodules from volcanic rocks and kimberlite pipes. Petrologists have obviously found them valuable. Uouraal of Petrology, VoL 23, Put 2, pp ,19821

2 128 ' T. N. IRVINE The original definitions of cumulate, cumulus crystal, and intercumulus liquid were essentially genetic: the underlying tenet was that the crystals accumulated by settling. Wager et al. (1960, p. 73) described cumulus crystals as 'discrete crystals successively separated from magma as a result of their greater density'. Jackson (1967, table 2.1) defined a cumulus crystal as 'a crystal that came into existence outside of, and previously to, the magmatic sediment of which it now forms a part'. There may, however, be no better example of the perils of genetic classification. For Wager and his colleagues, as judged by their choice of examples, the type occurrence of cumulates was the Layered Series of the Skaergaard Intrusion, and the type cumulus mineral was the dominant phase in this series, plagioclase. But it is now known that most, if not all of Skaergaard plagioclase was less dense than its parental liquid and, therefore, rather than settling, it should have floated (Bottinga & Weill, 1970; Campbell et al., 1978; McBirney & Noyes, 1979; see also Morse, 1972, 1979a, b, for similar conclusions about plagioclase in the Kiglapait Intrusion). For this reason, the concept of cumulates and cumulus processes has been seriously questioned (e.g. Campbell, 1978; McBirney & Noyes, 1979; Morse, 1979a, b), and in a recent paper on the Skaergaard Intrusion, McBirney & Noyes (1979, p ) chose to avoid cumulate nomenclature altogether. The feeling obviously was that, because of their connotations, the words were 'foes of reality'. Much of the basis for the early acceptance of the terminology stemmed from success achieved in applying it to matters of (1) subdividing and classifying the rocks of layered intrusions, and (2) defining patterns of mineral fractionation. In both aspects, it is obvious simply from textures that there is great value in distinguishing the minerals that have traditionally been called cumulus from those that are usually typed as postcumulus. The terminology clearly is not all bad. There is no question that the common mafic igneous minerals such as olivine, the pyroxenes, chromite, magnetite, and ilmenite (as well as less abundant phases such as apatite and immiscible sulfide liquid) are denser than any pertinent basic or ultrabasic magmatic silicate liquid. It might be argued, therefore, that the cumulate terminology should be restricted to rocks in which only these phases appear to have settled. In such case, application would effectively be limited to ultramafic rocks. As a rule, however, there is little difference between the textural characteristics of the above minerals in layered ultramafic rocks that one might call cumulates and their habits in associated layered gabbroic rocks. Moreover, the interlayering of the ultramafic and gabbroic types is commonly such that if one is called a cumulate, then logically so should be the other. An alternative approach, and a primary objective of the present paper, is simply to redefine the name 'cumulate' so that crystal settling is & possible but not essential process in the origin of rocks to which it is applied. There is nothing in the meaning of the word 'cumulus' itself that negates this approach. In meteorology, for example, its use in 'cumulus cloud' does not imply settling. The present aim is not just to guide future use of cumulate terminology; it is also to be conservative with respect to past applications. The intention is that most rocks that have previously been called cumulates should continue to be called by that name, but without the stigma that they have necessarily formed by crystal settling. From this starting point, an attempt is made to improve the overall terminology of layered intrusions. The existing nomenclature has evolved from many sources, both intrusions and authors. Inevitably, therefore, the same features have sometimes been given different names, and different features have been called by the same name further examples of words being the foes of reality. The objective in this regard is to refine the existing nomenclature and systematize its application. A few new names are proposed, but it is also recommended that several existing names be abandoned some because they are inherently ambiguous; others

3 TERMINOLOGY FOR LAYERED INTRUSIONS 129 because they do not accurately describe the features to which they are supposed to apply; and a few because they are poorly constructed grammatically. On the other hand, several valuable terms have not been used in a consistent way, and in these cases, guidelines are suggested that should improve their application. Table 1 is a listing of the subjects and names that are considered. Definitions are not tabulated, but the listing is arranged as an outline of the text with page numbers so that it can be used as an index. In the text, important names are placed in quotation marks where they TABLE 1 Listing of subjects discussed and terms defined in the text in respect to their application to layered intrusions The table headings correspond to the text headings and thus serve as an index. CUMULATES AND THEIR COMPONENTS Page Definitions and distinctions 131 Cumulate, cumulus framework, cumulus crystals Postcumulus material, intercumulus (pore) liquid Primary or cumulus crystallization stage, postcumulus processes Cumulate stratigraphy Primocryst, phenocryst Precipitation cumulate, flotation cumulate Depositional cumulate, accretion cumulate, crescumulate Intercumulus space filling, overgrowth, reaction replacement Discrete postcumulus materials, oikocrysts Classification by cumulus minerals 133 Order of decreasing mineral abundance, cotectic proportions Classification by postcumulus features 137 Orthocumulate, mesocumulate, accumulate Poikilitic (ortho-, meso-, ad-) cumulate Initial porosity, residual porosity Modification by postcumulus replacement 138 Postcumulus metasomatism, recrystallized rocks Replacement rocks, cryptically metasomatized rocks LAYERS, LAMINAE, AND LAMINATION Definitions 138 Layer, horizon, lamina Lamination: planar, lineate, imbricate, grain-size, modal, textural Characterization 139 Thickness: thin, medium-thick, thick Form: planar, tapered, lenticular, lensy, discontinuous, etc. Internal constitution: uniform, stratigraphically or laterally variable, isomodal leucocratic, mesocratic, melanocratic poikilitic, fragmental, pegmatitic Graded layers 144 grain-size graded, modally graded texturally graded, cryptically graded normal and reverse grading continuous or seriate grading delayed, discontinuous, or interrupted grading symmetric grading, multiple or recurrent grading Layer contacts 144 phase, modal, grain-size, textural, cryptic cumulus appearance, cumulus termination Form: planar, smooth, scalloped, etc.

4 130 T. N. IRVINE TABLE 1 (continued) LAYERING AND BANDING Definitions 145 Layering, stratification, banding Characterization 146 Layers vs. layering LJthology: modal, grain-size, textural cryptic Demarcation: prominent, inconspicuous, vague, diffuse, etc., schlieren layering, comb layering Regularity: thickness, lithologic variations, repetition Structure: trough layering, cross-bedded, codoform, corrugated, slumped, convoluted, wavy, deformed Repetition: rhythmic, macrorhythmic, microrhythmic, two-by-two layering Distribution and continuity: local; laterally continuous, discontinuous; stratigraphically continuous, intermittent Mode of origin: crystallization layering, current layering, pegmatitic layering Discontinuities 150 Progressive discontinuity Regressive discontinuities: intrusive contact, unconformity, paraconformity SUBDIVISION OF LAYERED INTRUSIONS Approach 151 General names (divisions, units), specific names Formal and informal names Divisions and subdivisions 15 2 Series: layered, marginal border, upper border, etc. Zones, subzones Units 153 Layering unit Member, Group Rhythmic unit, macrorhythmic unit, microrhythmic unit Cyclic unit, megacyclic unit, microcyclic unit Applications 154 Muskox, Great Dyke, Stillwater, Bushveld Slcaergaard, Kiglapait are introduced for discussion and are italicized where they are defined. Other terms that are useful for descriptive purposes, but probably not as names, are simply listed. DESCRIPTIVE VERSUS GENETIC NOMENCLATURE Before any names are defined, a few comments are in order concerning methods of classification, because classification is inevitably involved in nomenclature. It is often argued (especially by students) that the classification of rocks and their features should be on purely descriptive bases. In practice, however, such an approach is virtually impossible, and indeed, in many cases, not even desirable. Even the most fundamental typing of a rock as igneous, sedimentary, or metamorphic is interpretative. This particular scheme happens to be accepted, not because it can always be implemented without question (frequently it cannot), but because its basis is well established. The contention here is that the principal problem with cumulate terminology has been, not that it is genetic, but that the appropriate basis was not defined. In the present paper, names are defined on descriptive bases as much as possible, but genetic definitions are not avoided and are used whenever they are considered necessary, more appropriate, or more effective. On the other hand, a successful genetic classification might even be viewed as a desirable goal. As is well known, one reason for classification is to establish order in the hope that with order will come understanding. Conversely, understanding is a basis for improved classification: the sophistication of a successful genetic classification reflects the knowledge of its subject. But because interpretative foundations sometimes crumble, it is critical, first to Page

5 TERMINOLOGY FOR LAYERED INTRUSIONS 131 realize that a system is genetic, and then always to remember that it is. In the pages to follow, genetic names are noted, but readers (especially users) are advised to scrutinize all names, definitions, and applications for hidden connotations. An interesting phenomenon in this regard is that an advance in the understanding of a feature may restrict the applicability of its name, even though the latter was formulated descriptively. Thus, two features may qualify for the same descriptive name within the limits of its definition, but immediately the origin of one is understood, that origin is reflected on the other feature by the name, whether it is appropriate or not. An example described below concerns the term 'cyclic unit'. The chosen solution to the problem in this case is to give cyclic unit a genetic definition and to define another, descriptive name (rhythmic unit) for similar features whose origin is more doubtful. Definitions and distinctions CUMULATES AND THEIR COMPONENTS The principal terms of interest here will first be defined, then the basis of the definitions and the application of the terms will be discussed. A cumulate is defined as an igneous rock characterized by a cumulus framework of touching mineral crystals or grains that were evidently formed and concentrated primarily through fractional crystallization. (Note that only a framework of touching crystals is required. Not all crystals have to touch.) The fractionated crystals are called cumulus crystals. They typically are subhedral to euhedral, and generally they are cemented together by a texturally later generation of postcumulus material that appears to have crystallized from intercumulus liquid in the interstices or pores of the cumulus framework. The formation of the cumulus minerals is referred to as the primary or cumulus crystallization stage; the processes involved in solidification of the intercumulus liquid are termed postcumulus processes. Cumulates typically are layered, and given sufficient exposures, it is generally possible to demonstrate that the layers formed in succession. The lithologic variations through such a succession (from oldest to youngest) constitute its stratigraphy. Cumulates commonly also have features suggesting that gravity was a principal controlling factor in the fractional crystallization process, but an origin by crystal settling is not specified in the definition of cumulate, and it should not be inferred from the name. The basis of the definitions is as follows: Although cumulate terminology has been applied in various igneous environments, it was developed for, and has been particularly useful in, the description and discussion of layered intrusions. Intrusions such as Skaergaard, Stillwater, Bushveld, Muskox, Kiglapait, and Duke Island differ greatly in various aspects of composition and structure, but their rocks also have undeniable similarities, particularly in texture. Wager et al. (1960) believed that such similarities were related to crystal settling. Given now that this inference cannot be completely correct, it appears that the more pertinent process that the intrusions had in common was fractional crystallization. The exact mechanics of the fractionation may have differed from intrusion to intrusion (and from time to time and place to place in any one intrusion), but the evidence that it occurred is virtually unequivocal in all occurrences that have been well studied. Moreover, it is equally clear that the main fractionated minerals are those that form the frameworks of the rocks. They are the crystals that have usefully been called cumulus crystals in the past, and the argument here is that they should continue to be called by that name, whether the fractionation occurred by the crystals separating from the liquid (as by settling), by the liquid separating from the crystals (by convecting away), or by some other mechanism (as through the intermediary of a gas phase).

6 132 T. N. IRVINE In principle, most applications of the terminology should be the same as before, but if anything, they should be more definite because rocks can be shown more conclusively to have formed by fractional crystallization than crystals can be shown to have settled. (Witness the Skaegaard experience!) In general, the definitions will not be applicable on the basis of a single hand specimen or thin section*. To demonstrate that an individual rock has formed by fractional crystallization, one would generally have to show that the possible cumulus minerals are present in concentrations greater than would obtain if they had simply crystallized from their parental liquid with no fractionation and, or course, other possible origins, such as by metasomatic processes, would also have to be excluded. The usual circumstance, however, is that the parental-liquid composition is not even known. The demonstration therefore will generally require an examination of as many rock types as possible to see if their differences can be accounted for by fractional crystallization models based on their mineralogy and textural relations, relevant phase-equilibrium data, crystalliquid element partitioning coefficients, and so forth all in the light of their distribution in the intrusion, layering structures, and other field relations. The task is demanding, but one consolation deriving from empirical observations is that, in intrusions in which fractional crystallization is established, most rocks qualify as cumulates by the definitions given here. In such circumstances, the problem effectively reverses, and it becomes a more critical matter to identify the rocks that are not cumulates. An aspect of the definitions that places important limitations on the use of the name 'cumulate' is that the fractionated crystals should form a framework. This feature will usually distinguish cumulates from igneous rocks such as porphyritic dike rocks that may contain accumulated crystals (phenocrysts) but that would not have been called cumulates by the original definitions of Wager et al. (1960) and Jackson (1967). In this regard, it may be helpful to recall the name 'primocryst' originated by Wager & Brown (1968). By their definition, primocrysts are simply early-formed crystals in magma. If these crystals are isolated from one another and much coarser than the later minerals crystallized around them when solidification of the magma is completed, then they are phenocrysts. But if they are fractionated and concentrated to the degree that they touch one another, then they qualify as cumulus crystals. Cumulates can be extrusive as well as intrusive rocks. Komatiite flows, for example, commonly have cumulate zones (e.g. Pyke et al., 1973; Arndt et al., 1977). On the other hand, the fact that crystal settling is not a specified origin in the definition of cumulate usefully broadens the applicability of the name. The features of layered intrusions that have traditionally been regarded as principal evidence of crystal settling are: (1) the layers in the many intrusions appear to have been horizontal or near horizontal when they formed, and they typically have formed from the bottom up; and (2) certain of the more distinctive layers are graded from base to top either in grain size from coarse to fine, or in modal composition from dense to less dense minerals as though there had been gravity sorting. In a few intrusions, however, cumulate-type layering has evidently formed in near vertical disposition. Layered rocks in the Mt. Johnson Intrusion in the Monteregion Hills * In principle, this feature of the definitions is valuable in that it requires more judicious use of the word cumulate than has commonly been the practice. A rock should not be called a cumulate, for example, just because it 'looks like a cumulate in thin section'. As a case in point, the author would cite certain rocks in the Duke Island, Skaergaard, Stillwater, and other layered intrusions that have evidently (on the basis of structural relations) formed by postcumulus replacement (see Irvine, 19806, fig. 21). Given only thin sections for comparison, one would generally be hard pressed to distinguish these rocks from well-layered rocks in the same intrusions that were evidently formed by primary fractional crystallization. But in the realities of practice, much of the judgement on whether a rock is a cumulate will usually be based on comparisons with type examples in major intrusions. Perhaps the principal lesson to be learned from all this is that fractional crystallization commonly yields rocks in which the mineral grains appear to have settled, whether they have or not.

7 TERMINOLOGY FOR LAYERED INTRUSIONS 133 (Philpotts, 1968) constitute one example; the marginal border group rocks of the Skaergaard Intrusion (Plates 1, 2A; Irvine, 1979, fig. 9-14A; McBirney & Noyes, 1979, plate 5C) are another. Some cumulate units of ophiolite complexes appear also to have formed with their layering steeply inclined, vertical, or even overhanging (e.g. Casey & Carson, 1981). It is probably debatable in all these examples whether the crystals in the steep layers formed in situ on cooling walls or whether they were plated onto the walls by density or convection currents. But given that fractional crystallization was a principal aspect of the process, then by the definitions given here, the rocks can legitimately be called cumulates. Ah" the above notwithstanding, just a change in definition is not likely to remove the association of crystal settling with the name cumulate from the minds of most geologists. Nor should it, but one further step can be taken in an attempt to strike a better balance that is, to define genetic cumulate names that specifically distinguish different modes of origin. Four possibilities are precipitation cumulate, flotation cumulate, depositional cumulate, and accretion cumulate. The first pertains to crystal settling, although it is a little ambiguous*. The second, which was used by Ferguson (1964), is self-explanatory. The third covers the.possibility that the cumulus minerals were actually deposited through the mechanical action of magmatic density or convection currents (cf. Irvine, 1980a). Such deposition probably occurs mainly on horizontal or gently inclined surfaces, but as indicated above, it might also occur by the plating of crystals on steep walls, or even under the roofs of magma bodies. Accretion cumulate applies when the cumulus crystals have grown in place. Special examples are crescumulates (Wager & Brown, 1968, p. 554) wherein some of the minerals occur as coarse growths normal to the layering. The type crescumulate is the 'harrisitic' peridotite of the Rhum Layered Intrusion, which features coral-like, upward-branching growths of platy olivine crystals (Plate 2B; Wadsworth, 1961). Some of the marginal rocks of the Skaergaard Intrusion are also crescumulates (Plate 2A; also Wager & Brown, 1968). Jackson (1967) noted that 'postcumulus material' has three principal habits: (1) simple intercumulus space-filling, (2) overgrowth on cumulus crystals, and (3) reaction (peritectic) replacement of cumulus minerals. Forms (1) and (3) are particularly characteristic of discrete postcumulus minerals that is minerals that are not cumulus phases. These minerals tend also to be poikilitic and commonly form prominent oikocrysts containing dozens to hundreds of cumulus crystals (e.g. Jackson, 1961). It is emphasized, however, that not all poikilitic rocks in layered intrusions are cumulates. In the Duke Island Complex, for example, wehrlitic peridotite with augite oikocrysts has generally formed by replacement (Irvine, 1974, p. 43). Classification by cumulus minerals Wager et al. (1960) proposed that cumulates be named according to their cumulus minerals, listed in order of decreasing abundance. The names simplify the matter of delineating fractionation sequences, and they are helpful in tracing or correlating layers that have the same cumulus assemblage throughout but that show local postcumulus variations sufficient to change the rock name in a traditional system (Jackson, 1967). This last problem is especially common in the Stillwater Complex, and the cumulate nomenclature has been invaluable in unravelling its stratigraphy (e.g. Jackson, 1961; Todd et al., 1979; McCallum et al, 1980). This ambiguity is probably fitting, because concrete evidence of crystal settling in layered intrusions is exceedingly rare. There are unequivocal examples of current deposition of crystals (Irvine, 1974, plate 27) and of crystals grown in situ (plate 2). But to the author's knowledge, all evidence of crystals settling (or floating) in magmas is circumstantial and might alternatively derive from these other processes. He knows of no case of actual evidence that crystals descended through magmatic liquid from one level to another because of their individual gravitational body forces.

8 134 T. N. IRVINE PLATE 1. (A) Modal banding in the Skaergaard marginal border gabbro with couoform structures resembling stromatolites, Ivnarmuit Island. Pegmatitic recrystallization and replacement are locally superimposed. The contact of the intrusion is about 100 m to the left; younging is to the right. The bulbous structures characteristically point away from the cooling surface, and it is apparent that they represent in situ crystallization. (B) Larger view of the same kind of irregular banding (note man for scale at upper right); younging is to the right. Although it is not obvious here, the couoform structures are generally more prominent in horizontal than in steep cross-sections (see also Wager & Brown, 1968, p. 123), indicating that they are actually a near-vertical corrugation.

9 TERMINOLOGY FOR LAYERED INTRUSIONS 135 PLATE 2. (A) Crescumulate, perpendicular-feldspar gabbro from near the margin of the Skaergaard Intrusion, Mellemo Island. The contact of the intrusion is only about 10 m away, to the lower right; hence it is apparent that the plagioclase grew In situ away from a cooling surface. (B) Peridotitic crescumulate layering in the western part of the Rhum Intrusion. The crescumulate consists of vertically oriented olivine crystals; it alternates with granular olivine cumulate showing grain-size (and probably slight modal) layering. The contact between the two variants could be called a textural contact. The locality is far distant from the edge of the intrusion, so the crescumulate olivine presumably grew upward into supercooled liquid (Wager & Brown, 1968, p. 294).

10 136 T. N. IRVINE The cumulate naming system has the unfortunate feature, however, that the name is cumbersome when the rock contains several cumulus phases. One solution to this problem is the use of one-letter abbreviations for the mineral names. For example, in Stillwater investigations, plagioclase-olivine cumulate is denoted poc; plagioclase-bronzite-augite cumulate is pbac. This method is very convenient for mapping, drill-core logging, thin-section descriptions, and other research activities. But it does require some initiation, and there can be difficulties with duplication of abbreviations and changes of mineral variety (as from bronzite to hypersthene). Thus, it is less satisfactory for published work. Ultimately, it seems, the most effective approach is to use both cumulate and traditional names. The cumulate names serve for classification and genetic analysis at the research stage. Then familiar traditional names are formulated in accordance with the main cumulus minerals, and both types of names are used (as appropriate) for descriptions and discussions. TABLE 2 Conventional rock names for the more common types of ultramafic and gabbroic cumulates The cumulus minerals are listed in order of decreasing abundance; accessory minerals are in parentheses. A maximum of only three cumulus phases is considered; additional minerals would be varietal phases in the conventional rock names. Cumulate type Peridotitic cumulates Olivine-{chromite) Olivine-chromite Olivine-augitic clinopyroxene Olivine-orthopyroxene Olivine-clinopyroxene-orthopyroxene Pyroxenitic cumulates Clinopyroxene Clinopyroxene-olivine Orthopyroxene Orthopyroxene olivine Clinopyroxene-orthopyroxene Clinopyroxene-orthopyroxene-olivine Clinopyroxene-orthopyroxene-plagioclase Orthopyroxene-chromite Clinopyroxene-magnetite Gabbroic cumulates Plagioclase Plagioclase-olivine Olivine-plagioclase Plagioclase-augitic clinopyroxene Plagioclase-cHnopyroxene-olivine Plagjoclase-orthopyroxene Plagioclase-clinopvroxene-orthopyroxene Plagioclase-orthopyroxene-clinopyroxene Plagioclase-clinopyroxene-magnetite Plagioclase-clinopyroxene-apatite Oxide cumulates Chromite Chromite-olivine Magnetite Ilmenite Conventional rock name Dunite, peridotite, picrite* Chromite dunite, peridotite' Wehrlite Harzburgite Picritic websterite Clinopyroxenite Olivine clinopyroxenite Orthopyroxenite Olivine orthopyroxenite Websterite Olivine websterite Gabbroic websterite Chromite orthopyroxenite Magnetite clinopyroxenite Anorthosite Troctolite Picritic troctolite Gabbro Olivine gabbro Norite Two-pyroxene gabbro Noritic gabbro Magnetite gabbro Apatite gabbro Chromitite Olivine chromitite (Cumulate name preferred) (Cumulate name preferred) Choice of name depends on the amount of postcumulus material. In some ultramafic cumulates, an abundance of postcumulus material can be indicated by the adjective 'feldspathic'. If the cumulus plagioclase is more sodic than about An 43, then 'diorite' should be substituted for 'gabbro'.

11 TERMINOLOGY FOR LAYERED INTRUSIONS 137 A comparative listing of names for common ultramafic and gabbroic cumulates that is useful in this jegard is given in Table 2. When two or more cumulus minerals are present in a cumulate, they tend on the average to occur in cotectic proportions that is, in the ratios that would be expected if they coprecipitated by fractional crystallization (e.g. see McCallum et al., 1980; Irvine, 1970; also Morse, 1979a, b, for exceptions). The cotectic abundance order for the minerals in Table 2 is qualitatively plagioclase, clinopyroxene, orthopyroxene, olivine, ilmenite, magnetite, apatite, chromite. The simpler traditional rock names in Table 2 are appropriate to this order; the mixed names (such as picritic troctolite, picritic websterite, and noritic gabbro) are designed to apply when the mineral proportions are strongly exceptional (noncotectic). Note that, even in the mixed names, the essential name identifies the cumulus mineral assemblage. Thus a plagioclase orthopyroxene-clinopyroxene cumulate is called noritic gabbro rather than gabbronorite (which would be ambiguous). Classification by postcumulus materials Many cumulates are not adequately characterized just by a listing of their cumulus minerals, not even a quantitative listing. Consideration must also be given to the nature and abundance of the postcumulus materials as a secondary basis of classification. Wager et al. (1960) were endeavoring to do this when they introduced 'orthocumulate', mesocumulate', and adcumulate', but the original definitions of these names were couched in a genetic discussion, and they will be redefined here somewhat more descriptively. In an orthocumulate, postcumulus materials should be abundant (and mostly interstitial) and the cumulus minerals ideally should exhibit much of their original crystallization forms. Mesocumulates have less postcumulus material, and the cumulus grains should adjoin in part along mutual interference boundaries developed through overgrowth. Adcumulates have only minor discrete postcumulus material, and mutual interference boundaries are the norm for the cumulus phases. The exact percentage limits for postcumulus material that one might choose for each type may vary somewhat, depending on the number and kinds of cumulus minerals and the grain size and texture of the rock, but as a rule, postcumulus minerals make up per cent (by volume) or orthocumulates, 7-25 per cent of mesocumulates, and 0-7 per cent of adcumulates. These definitions also are genetic in that the identification of postcumulus material involves interpretation of textures and modal relationships. Moreover, the interpretation becomes more tenuous the greater the number of cumulus phases, because overgrowth is more difficult to distinguish than discrete material. The definitions are reasonably practical, however, and the distinction may be facilitated by using the whole-rock concentrations of 'excluded' or 'incompatible' elements (that is, elements that do not readily enter the structures of the cumulus phases) as indicators of the amount of the trapped pore liquid from which the postcumulus minerals are supposed to have crystallized (e.g. Morse, 1979a, b; Irvine, 1979, 19806). The latter practice has been criticized on the grounds that a name denoting texture is being assigned on the basis of chemical composition, and the criticism is justified. Nevertheless, the feature of ultimate interest usually is the amount and composition of the trapped liquid, which should be better indicated by the chemical data than by the texture. (For example, McBirney (1975) has attempted to define the composition of successive Skaergaard liquids through experiments in which he remelted the intercumulus liquid trapped in orthocumulates from the intrusion.) In assigning the name, however, the textural relations should be the essential basis, and the chemical data only a guide when the textures are not definitive. Wager et al. (1960) also proposed 'heteradcumulate', to which they gave a genetic

12 138 T. N. IRVINE definition of questionable applicability. A principal feature of the rocks for which the name was intended is that they have prominent oikocrysts. The rocks, therefore, are probably better called simply poikilitic (prtho-, meso-, or ad-) cumulates. Among poikilitic cumulates, some oikocrysts have apparently formed by simple interstitial crystallization; others are evidently in part the products of peritectic replacement of cumulus grains (Jackson, 1961, p ; Von Gruenewaldt & Weber-Diefenbach, 1979). No names have been found to distinguish these modes of origin, but they are probably better covered by a few words of description. It commonly appears that differences in the percentage amounts of postcumulus materials reflect processes that have acted to remove pore liquid from the cumulus framework. For genetic considerations, therefore, it is commonly necessary to distinguish between initial porosity and residual porosity (e.g. Morse, 1979a). Modification by postcumulus metasomatism It is becoming increasingly evident that many cumulates are slightly to extensively modified by processes of postcumulus metasomatism (e.g. Irvine, 1974, 19806). The exact circumstances of these processes are probably varied, but essentially they appear to be related to movements of intercumulus liquids and, probably in some cases, associated gas. The conspicuous products are either recrystallized rocks (e.g. recrystallized peridotite and pyroxenite) or replacement rocks (e.g. replacement dunite or anorthosite). The former, by definition, have roughly the same modal compositions as their primary predecessors, but their textures are different, typically coarser. In the latter, the original modal compositions have been substantially modified, most commonly toward assemblages of fewer minerals. There are also cryptically metasomatized rocks, as in an example documented in the Muskox Intrusion (Irvine, 19806) in which the textures of cumulus olivine and pyroxene do not appear to have been much altered, but their compositions have been significantly changed. The most common evidence of postcumulus metasomatism is that it transgresses layering without appreciably displacing or distorting it thus, the visible recrystallization and replacement appear to have occurred more or less on a volume-for-volume basis. Relict traces of primary textures and structures may also be preserved. In the cryptic metasomatism in the Muskox Intrusion, primary mineral-composition trends or discontinuities have evidently been displaced. Most of the occurrences of postcumulus metasomatism that have been documented to date are relatively local. The Onverwacht hortonolite dunite pipe in the Bushveld Complex described by Wagner (1929; see also Cameron & Desborough, 1964) is a classic example. Indications are, however, that the processes can also occur on broader scales: thus, in the Muskox Intrusion, the modal compositions of parts of some layers appear to have been significantly modified over several kilometers. Metasomatic rocks of the above types constitute a principal case where there is a danger of misapplication of cumulate terminology. Thus, before a cumulate name is used, the question should be asked whether the rock in question might be metasomatic, and evidence for or against this possibility should be sought. Evidence is not likely to be found in every thin section or outcrop, but if definite primary textures or structures cannot be demonstrated somewhere in some reasonably large unit of the rock, then it is advisable not to use the cumulate names. Definitions LAYERS, LAMINAE, AND LAMINATION In the present context, a layer is a sheetlike cumulate unit that is a distinctive entity in its compositional and (or) textural features. A horizon is any conformable reference plane in a layer or sequence of layers. Note that a horizon has no thickness and so is not itself a layer.

13 TERMINOLOGY FOR LAYERED INTRUSIONS 139 Lamina is a name that can be usefully applied to thin, sharply defined layers. In sedimentary petrology, laminae are usually described as being less than 1 cm thick; for igneous cumulates, 2-3 cm (about an inch) is probably a more useful upper limit It might be emphasized that laminae are distinctive individual units. By contrast, igneous lamination pertains to various more subtle features that are pervasive through the cumulate fabric on the scale of the grain size. One variant is planar lamination, a term introduced by Wager & Deer (1939) to refer to the platy alignment of tabular minerals (such as plagioclase) parallel to cumulate layers. Jackson (1967) proposed lineate lamination to cover linear alignment in the plane of the layering. To these names, the author would add imbricate lamination, grain-size lamination, modal lamination, and textural lamination. The first obviously applies where platy cumulus minerals have an imbricate arrangement (see Philpotts, 1968, fig. 6, for an example); the second is even more self-explanatory (Plate 4A; also Irvine, 1974). Modal lamination is a subtle, planar segregation of two or more cumulus minerals on approximately the scale of their grain size. Clearly defined layers and laminae may be associated, but they are not essential features. Modal lamination is perhaps even more common than planar lamination, and it tends to be particularly conspicuous in gabbroic and syenitic cumulates because it is enhanced by the color contrast of mafic minerals and plagioclase (Plate 3A). It can, however, be seen in almost any multiphase cumulate (e.g. Plate 3B). 'Textural' as used here and elsewhere below is intended to cover features of fabric and mineral habit other than grain size. (Planar and linear grain orientation are also excluded in the present case.) An example of textural lamination could be the alternate development of poikilitic and granular laminae. The name, however, is likely to be the least useful of the lamination terms. Some cumulates show vertical alignment of cumulus minerals (e.g. Wadsworth, 1961; Irvine, 1979), but they do not usually exhibit lamination (see later discussion of'crescumulates'). Characterization Layers can be referred to in terms of their thickness, form, composition, texture, and internal structure. Whatever name is chosen, it should be remembered that it is only for convenience of reference and is not a substitute for a specific description or quantitative measurements. General working limits for thickness might be 0-5 cm for thin (Plates 4A, 6B), 5 cm-1 m for medium-thick (Plates 4, 5), and more than a meter for thick (Plates 6A; 8A, B). Different limits may be desirable for different intrusions, depending on the population of layer thicknesses. The forms of layers can be referred to by any number of familiar terms planar, tapered, lenticular, lensy, discontinuous, extensive, trough-shaped, basinform, bowl-shaped, dishshaped, synformal, antiformal, arched, domed, convoluted, folded, faulted, and so forth. Ambiguous words such as 'irregular' and 'uneven' should be specifically applied (e.g. irregular in thickness). If a layer has the same composition, texture, and structure throughout, it is uniform (Wager, 1963). Other layers are stratigraphically variable or laterally variable (or both), these names referring respectively to upward and horizontal directions through the layer if it were flat-lying. The variations may be in any or all of modal composition, grain size, texture, internal structure (such as lamination or deformation), and chemical composition (both of mineral solid solutions or of the rock as a whole). Jackson (1967) introduced the name 'isomodal layer' for layers with constant mineral proportions. The name is attractive, but if the layer is variable in respects other than its mode, then possibly they should be mentioned; otherwise, the layer is better described as 'uniform'. Most features that might be mentioned in reference to composition are covered by the rock or cumulate name, but it frequently is useful additionally to distinguish leucocratic and

14 140 T. N. IRVINE PLATE 3. (A) Modal lamination in olivine gabbro in the Axelgold Intrusion (central British Columbia), near the center of the body. Thin, concordant dark plates like that on the left at the level of the hammer tip are common in the intrusion. They consist mainly of pyroxene, but some have cores of quartz, and it is evident that they have formed by magmatic reaction replacement of slabby fragments of the quartzitic schist that is a principal country rock of the intrusion. (B) Rude modal lamination in olivine clinopyroxenite in the Judd Harbor part of the Duke Island Ultramafic Complex. This structure was previously called inch-scale layering (Irvine, 1974), but modal lamination clearly is a better name.

15 TERMINOLOGY FOR LAYERED INTRUSIONS 141 PLATE 4. (A) Rhythmic grain-size layering in the Hall Cove part of the Duke Island Ultramafic Complex, featuring grain-size graded layers and sharply defined grain-size laminae. The rock is a clinopyroxene-olivine cumulate, and both cumulus minerals are size sorted. Olivine is generally slightly finer on the average; consequently there is some modal sorting as well. (B) Graded fragmental layers in the Younger Intrusion of the Hall Cove part of the Duke Island Complex. The grading is largely restricted to the granular upper part of the layer (i.e. it is 'delayed grading")- The fact that the fragments are distributed is evidence that the layering is current layering and that the rock is a depositional cumulate.

16 142 T. N. IRVINE PLATE 5. (A) Rhythmic layering in Skaergaard Lower Zone (LZc) gabbro featuring a continuous succession of modally graded layers. (B) Rhythmic layering in Upper Zone ferrodiorite (UZa) comprised of alternate uniform and modally graded layers.

17 TERMINOLOGY FOR LAYERED INTRUSIONS 143 PLATE 6. (A) Areal view of macrorhythmic layering in the Skaergaard Layered Series, in a northwest spur of Wager Peak. The layers are draped over a large block of anorthositic gabbro that fell from the"top of the magma chamber. The block is about 50 m long, and macrorhythmic units (each consisting of a light plus a dark layer) are mostly about 3-8 m thick (see Irvine, 1980c, fig. 7). (B) Microrhythmic (inch-scale) modal layering in the Skaergaard Middle Zone, Pukagaqryggen Peak. Some thick layers like those in (A) are composed of this type of thinly layered rock (cf. McBirney & Noyes, 1979, plate 5B).

18 144 T. N. IRVINE melanocratic variants (or even mesocratic variants on occasion). Some layers are outstanding because they are 'mottled' by virtue of the cumulate being poikilitic. A few intrusions contain fragmental layers, these having conspicuous contents of distributed fragments, most commonly cognate xenoliths of older differentiates (Plate 4B; Irvine, 1974, Plates 11, 12; Irvine, 1980c, fig. 4; McBirney & Noyes, 1979, plate 4B). Layers can also be 'pegmatitic', but the use of this adjective (rather surprisingly) requires some care, as explained in the discussion of layering below. Graded layers Among variable layers, those that are systematically graded stratigraphically can be singled out for special attention. Jackson (1967) proposed that such layers be classed as 'size graded', 'mineral graded', or 'chemical graded', depending on whether the grading is in the grain size, modal abundance, or chemical composition of the cumulus minerals, but none of these names is really appropriate. In the first case, it is not the size of the layer that is graded, it is the size of the contained cumulus crystals or grains. In the second, it is not the minerals that are graded, it is their modal proportions. And the third term is ambiguous: the layer could be graded chemically because of modal grading as well as because of mineralcomposition grading. Preferable names are grain-size graded layer (Plate 4), modally graded layer (Plate 5), and cryptically graded layer (e.g. Irvine, 19806, figs. 8, 9). The last stems, of course, from the familiar 'cryptic layering', originated by Wager & Deer (1939) for reference to stratigraphic cumulus-mineral composition variations. Texturally graded layer might also be used for example, if there was gradation from poikilitic to granular but like its lamination counterpart, it probably has limited usefulness. The preferred names might be criticized as being unnecessarily long, but it is noted that as a rule in practice, after the grading in the layers of an intrusion has been described, it is sufficient for most subsequent purposes to refer to the layers simply as 'graded layers'. The stratigraphic grading of layers can in some cases be typed as normal or reverse (e.g. Parsons, 1979, fig. 5). In the most familiar cases, normal grading is upward either from coarse to fine in grain size, or from dense to less dense (dark to light) minerals in mode. One should be mindful, however, of a tendency to regard these arrangements as 'normal' because they accord with the process of crystal sorting by settling. There are indications in some intrusions (e.g. Skaergaard and Stillwater) of alternative (crystallization) processes that 'normally' yield light-to-dark upward grading. Thus, the basis for using normal and reverse should always be explained. It will be appreciated also that, even when the grading in layers is undirectional, it is not necessarily continuous or seriate. To borrow terms used in the descriptions of graded beds in sediments, there can also be delayed grading in which only the upper part of a layer is graded (Plate 4B), or discontinuous or interrupted grading in which some intermediate stage is sparsely represented (e.g. Dzulynski & Walton, 1965). Alternations of normal and reverse grading can also be found, in which case sedimentary terms such as symmetric grading and multiple or recurrent grading might be applied. Layer contacts In the description of layer contacts, features of primary interest are whether they are sharp or gradational and concordant or discordant. A 'sharp' contact is one that can be defined within a few grain diameters, but a larger limit may occasionally be appropriate, particularly for very thick layers. The thickness of gradational contacts should be noted, and if discordance is evident, it should be described. The contacts themselves are distinguished by changes in any or all of mode, grain size,

19 TERMINOLOGY FOR LAYERED INTRUSIONS 145 texture (e.g. poikilitic vs. nonpoikilitic), and mineral composition (a contact can be 'cryptic'). Jackson (1967) suggested that contacts defined by changes in mode be called either 'phase' or 'ratio' contacts, depending on whether they are marked by a change in the cumulus-mineral assemblage or by a change in mineral proportions. For contacts marked by changes in grain size or texture, he suggested the name 'form contact'. Phase contact is certainly useful (see Plate 7A), but ratio contact is not in itself very meaningful, and form contact stands to be confused with the 'form' of a contact, which is something very different (see next paragraph). The names recommended here, besides phase contact, are modal contact (Plate 5), grain-size contact (Plate 4), textural contact (Plate 2B), and cryptic contact (Irvine, 19806, figs. 8, 9). Modal contact would cover any change in mineral proportions; phase contact applies to the special case of a change in assemblage. The other terms are self-explanatory (an example of a textural contact is shown in Plate 3B). It should be emphasized that contacts are commonly marked by changes in several features, in which case no single adjective is sufficient. As a rule in practice, therefore, a few words of description are essential before any name can be implemented. Phase contacts are particularly important, and the term cumulus appearance* is useful in relation to their description and discussion. It pertains to the appearance of a cumulus mineral with stratigraphic height in a cumulate pile above a significant interval in which it is absent Note that the appearance may be in gradually increasing amounts. A cumulus mineral can also have a cumulus termination, this term referring to the top or end of its stratigraphic range. Such terminations can occur because of peritectic reactions, for example, or because of more catastrophic changes in magma composition. ('Cumulus disappearance' would mean the mineral had been present as a cumulus phase but ceased to exist. An example has been described in the Bushveld Complex in which small cumulus pigeonite grains were resorbed during the postcumulus growth of large orthopyroxene oikocrysts; Von Gruenewaldt & Weber-Diefenbach, 1979.) Jackson (1967, p. 34) suggested that 'phase contacts' usually reflect supply of crystals through primary crystallization and that 'ratio contacts' commonly are due to mechanical sorting of crystals in the magma. The first of these statements is almost certainly true, but the second is now much more questionable (even as a generalization) than it seemed in 1967, and in neither case is the contact relationship itself necessary or sufficient evidence of the mode of origin. The forms of a contact are analogous to those of a layer, but some of the possibilities differ. Familiar adjectives are planar, smooth, curved, scalloped, wavy, corrugated, rippled, convoluted, folded, and faulted. Some layer contacts are load-casted with flame structures (e.g. Parsons, 1979, fig. 6). Definitions LAYERING AND BANDING Layering is here defined as the overall structure and fabric of cumulates manifest through combinations of layers, laminae, and lamination. Stratification is synonymous. Like many words in geology, layering can be used in other contexts, but in igneous petrology at least, it seems desirable to restrict its application as much as possible to cumulate stratification in order to avoid confusion. This name is fashioned after the term 'cumulus arrival' used by Morse (19796) for the same purpose. His term is not unattractive, but as a reviewer has emphasized, neither is it accurate. On one hand, it suggests that the mineral was introduced, whereas the mineral might have crystallized in situ. On another, if crystallization is the reference, then rather than the arrival being that of the mineral, it more probably would be that of the liquid composition reaching the mineral's saturation limit

20 146 T. N. IRVINE Unfortunately, there are not many names of similar connotation appropriate for other types of planar structures in igneous rocks. In fact, the only alternative appears to be 'banding'. Its use is commonly criticized on the grounds that a 'band' is two-dimensional, not three-dimensional like a layer, but given the shortage of suitable words, some compromise seems essential. Banding is defined, therefore, as a general name for three-dimensional structures characterized by alternation of planar rock units (bands) of contrasting appearance. There are no limitations of scale, and the name is intended to be free of genetic undertones, but it may be assigned genetic modifiers, as in 'flow banding' and 'metasomatic banding'. As applied to cumulates, banding is perhaps most useful for identifying structures that differ appreciably from the common types of layering. For example, Wager & Brown (1968) used it effectively in the Skaergaard Intrusion to distinguish the characteristically irregular to colloform 'banding' of the marginal rocks (Plate 1) from the typically smooth, almost planar 'layering' in the interior (Plates 5, 6). On the other hand, the 'trough banding' in this intrusion would better have been called 'trough layering'. Characterization By present definitions, the difference between 'layers' and 'layering' is the difference between individuals and a population. The terms that describe one are not necessarily appropriate to the description of the other. Jackson (1967, 1970) was particularly careful in making this distinction, but many other authors have not been. Words such as 'thick', 'graded', and 'isomodal' properly apply to individual layers, whereas 'rhythmic', 'cryptic', and 'modal' are strictly correct only for layering. If graded or thick is to be applied to layering, it must be by some arbitrary definition based on frequency of occurrence of graded or thick layers. In the author's experience such definitions are generally unnecessary and frequently misleading. For example, despite the abundance of graded layers in the Duke Island Ultramafic Complex, there was no need in its description for 'graded layering' (cf. Irvine, 1974). And the common circumstance with respect to thickness is that thick layers of one style alternate with (or even consist of) thin layers of another. The distinction emphasized here may seem a matter of splitting hairs, but the vocabulary of our language is remarkably comprehensive, and there is no need to be inaccurate. The lithologic variations that characterize layering can, nevertheless, generally be described by the same modifiers used for lamination, graded layers, and layer contacts. Thus, we may have any of modal (Plate 5), grain-size (Plate 4), textural (Plate 2), or cryptic layering (Wager & Deer, 1939). Even more commonly than with individual layers, however, layering tends to reflect two or more of these features. Thus, in most cases, before a layering name is assigned, it is essential to describe the kinds of layers involved and the ways they are interstratified. Even then, blanket names should be avoided unless they are frequently useful. The distinctness or demarcation of layering is a feature of interest in descriptions, but it is not generally the basis for a name. Of concern is whether the layering is prominent or inconspicuous, sharply defined or rudely developed, vague, diffuse, or faint. 'Wispy' is a handy word for describing layering in which thin, vaguely defined layers inobtrusively appear and fade out. It is not very appropriate, though, for the name of any fundamental layering type. Moore & Lockwood (1973) have used schlieren layering for layering of this type, and it seems a better choice (see Bates & Jackson, 1980, for a definition of schlieren). Moore & Lockwood also introduced comb layering for successions of layers in which the main minerals occur in elongate crystals oriented perpendicularly (or at least at high angles) to the layering planes. This name is preferable to the 'Willow Lake layering' applied to this type of stratification by Taubeneck & Poldervaart (1960). The regularity of layering is also of interest descriptively. The words 'regular' and

21 TERMINOLOGY FOR LAYERED INTRUSIONS 147 'irregular' are themselves ambiguous, however, and their application must be specified for example, whether they pertain to the thickness or form of layers, patterns of lithologic variation, or mode of repetition. The words 'even' and 'uneven' are similarly ambiguous. The structure of layering is occasionally the basis for a name. A famous case is the 'trough layering (banding)' of the Skaergaard Intrusion. Other possibilities are cross-bedded (Plate 7B), colloform (Plate 1), slumped, convoluted, wavy, and deformed layering. Mode of repetition is probably the most common feature mentioned in names for layering. The type example, of course, is rhythmic layering, originated by Wager & Deer (1939) in the classic work on the Skaergaard Intrusion. It is characterized by conspicuous, systematic recurrence of distinctive layers or sequences of layers of the same kind. In the Skaergaard Intrusion, the most familiar distinguishing layers are modally graded units, 5-50 cm in thickness (Plate 5), but there seems no reason why the term cannot be used for any kind of layer as long as the repetition is regular and prominent (e.g. Plate 4). If the layers are meters in thickness, then the name macrorhythmic layering is appropriate; if they are only laminae, less than 3 cm in thickness, then microrhythmic layering might be employed. The former is exemplified by prominent thick layering in the central part of the Skaergaard Intrusion (Plate 6A; also Wager & Brown, 1968, plate IV; McBirney & Noyes, 1979, plate 8; Irvine, 1980c, fig. 7) and by layering in the Himaussaq Intrusion (e.g. Ussing, 1912, fig. 7; Wager & Brown, 1968, fig. 268). Fine-scale layering in the Skaergaard Intrusion (Plate 6B; also McBirney & Noyes, 1979, plate 5B) would qualify as microrhythmic layering, as would the Stillwater inch-scale layering (Hess, I960; Wager & Brown, 1968, figs. 187, 188). The word 'cyclic' has also been used to name layering, but it has much the same connotation as 'rhythmic', and to avoid confusion, it is better reserved for 'cyclic unit', to be discussed below. In rare cases, patterns of repetition can be described numerically for example, some of the Stillwater inch-scale layering features two-by-two layer repetition (Wager & Brown, 1968, fig. 187). In general, though, the details of layer repetition cannot be adequately (or accurately) conveyed in a simple name. A few brief words of description are much more valuable, as in 'the layering formed by alternation of modally graded and uniform layers' or 'the layering comprised of a continuous succession of modally graded layers'. The continuity and distribution of layering can be described on various scales from that of an outcrop to the size of the whole intrusion. Familiar terms are 'laterally continuous or discontinuous' and 'stratigraphically continuous or intermittent' but they are too long for names. McBirney & Noyes (1979, p. 511) proposed 'intermittent layering' for reference to the alternation of modally graded and uniform layers that is a common feature of Skaergaard rhythmic layering. The name, however, does not properly distinguish between layers and layering. In a typical outcrop, the graded layers could be said to be intermittent, but the layering (as a whole) is continuous (Plate 5B). Although the names for layering should generally be descriptive, two fundamentally genetic terms are useful for discussion purposes. One is current layering, which applies to sequences of layers deposited by magmatic currents (Plate 4). The other is crystallization layering, which is manifest through Variations in assemblages or proportions of minerals caused by crystallization processes (Plates 2, 6, 7B, 8). This second name is recommended over 'phase layering', originated by Hess (1960) for the same purpose, because it is more specific with respect to the process and less apt to be confused with 'phase contact'. The latter point is significant, because not all crystallization layering is marked by phase contacts, and probably not all phase contacts are due to crystallization layering. A prerequisite to the application of either current or crystallization layering is, of course, evidence that it is appropriate.

22 148 T. N. IRVINE PLATE 7. (A) Phase contact between a uniform olivine-cumulate layer above and a modally laminated gabbro layer below in the Stillwater Banded Series, Picket Pin Mountain area. The contact is a paraconformity in that is essentially conformable and represents a modal regression in the overall course of the Stillwater differentiation. (B) Cross-bedded modal layering in the Lower Zone of the Skaergaard Layered Series, Uttentals Plateau. One prominent angular unconformity and several less conspicuous examples are present

23 TERMINOLOGY FOR LAYERED INTRUSIONS 149 PLATE 8. (A) Thick layers in the Eastern Bushveld Complex. In the middle foreground are resistant layers of orthopyroxene cumulate alternating with eroded olivine cumulate layers. Cameron (1978) has identified three cyclic units in this section. The distant hills are underlain by the Critical Zone of the complex and feature alternate layers of noritic, anorthositic, and pyroxenitic rocks. According to Cameron (1963). layering units in this section extend for 60 km, but there is little or no regularity in the layer repetition. On the basis of the scale of its development, the layering genetically must be crystallization layering, but it locally has features that appear to represent current layering. (B) Cyclic units in the eastern part of the Rhum Intrusion. In each unit, a thick layer of olivine cumulate is overlain by a thinner layer of troctolitic and gabbroic cumulates that are more resistant to weathering.

24 150 T. N. IRVINE A rather special case of a name that is essentially descriptive but whose use requires careful genetic consideration is 'pegmatitic layering'. 'Pegmatite' was originally a name for graphic granite, but as explained by Jahns (1955) in an authoritative review, its application has been greatly broadened over the years. Jahns chose to apply it to 'holocrystauine rocks that are at least in part very coarse grained, and whose major constituents include minerals typically found in ordinary igneous rocks, and in which extreme textural variations, especially in grain size, are characteristic'. On this basis, the word 'pegmatoid', which essentially means 'like pegmatite', is superfluous. Pegmatite has also been used as a name for bodies of pegmatitic rock, but Jahns recommended that these be specifically identified, as in pegmatite 'dike' or 'mass'. (Consider also the difference between 'pegmatitic layer' and 'layered pegmatite'.) The critical question that must be asked in application of'pegmatitic' to layering is whether the rock units under consideration are properly called layers in the cumulate sense. In parts of the Skaergaard Layered Series, for example, well defined bodies of pegmatite follow layering planes so closely that they appear to be layers (see Wager & Brown, 1968, fig. 51), but locally they are transgressive with dilational relationships that leave no doubt that they are actually intrusive bodies distinctly younger than the cumulate layering (see McBirney & Noyes,-1979, plate 5E). In other places, particularly the marginal parts of the intrusion, anchiconcordant lenses and tabular bodies of pegmatite have evidently formed by volume-for-volume replacement of primary layering. Similarly in the Duke Island Ultramafic Complex, pegmatitic rocks characteristically are the products of recrystallization and replacement of layered cumulates (Irvine, 1974). In all these cases, the name 'pegmatitic layering' is appropriate in a sense, but it is also misleading (in effect, it is a foe to reality) in that the pegmatite is not primary in relation to the layering. Thus, the fact that the concordant pegmatite bodies in the Skaergaard Layered Series crystallized in situ does not mean that the layers that they follow were formed that way. And the fact that pegmatite lenses in the marginal parts of the intrusion are internally differentiated with perpendicular growths of crystals is not indicative of the origins of the primary marginal banding. In general, therefore, rather than using the name 'pegmatitic layering', it is better simply to describe how the pegmatite occurs and, if possible, how it most probably formed for examples, whether it is in anchiconcordant intrusive lenses, anchiconcordant replacement veins, irregularly shaped replacement pods, patchy to pervasive replacement, or (perhaps) primary layers. Discontinuities Many layered intrusions have internal contacts of special significance in that, in one sense or another, they are 'discontinuities' in the layer succession. Most commonly these contacts are defined structurally by discordance or truncation of layering. In some cases, however, they are evident only lithologically or even chemically through cumulus mineral compositions. Like other contacts, they may be sharp or gradational; some are themselves banded or layered. But even where a discontinuity is physically obvious, its identification involves enough interpretation that the term is essentially genetic. Broadly speaking, discontinuities are either 'progressive' or 'regressive'. A progressive discontinuity is developed between distinctive parts of an intrusion through normal crystallization processes that is, without apparently involving catastrophic events such as intrusion of new magma or tectonic disturbance. Type examples might be the contacts in the western and upper parts of the Skaergaard Intrusion between the Layered Series and the Marginal and Upper Border Groups (cf. Wager & Brown, 1968). These divisions evidently grew concurrently, and their contacts appear to have evolved gradually as the three divisions solidified in their individual ways. Other examples of progressive discontinuities are the junctions in the Muskox Intrusion formed by the onlap of the Layered Series over the Marginal Zones that line the footwall contacts (cf. Smith & Kapp, 1963).

25 TERMINOLOGY FOR LAYERED INTRUSIONS 151 A regressive discontinuity is marked by a hiatus or reversal in the succession of rocks that would be expected to develop through undisturbed crystallization. One type is an internal intrusive contact between plutons of a composite intrusion. The contacts between the main and younger intrusions of the Duke Island Ultramafic Complex are principal examples familiar to the author (Irvine, 1974). Unconformities are a second type of regressive discontinuity. Along them, early layers typically appear eroded or truncated. Unconformities abound in the 'cross-bedded belt' of the Skaergaard Intrusion (Wager & Brown, 1968; see also Plate 7B), and they have been described in many other intrusions, notable among them, the Duke Island Complex (Irvine, 1974, plates 23, 24; Irvine, 1980c, fig. 3) and several syenitic intrusions in Greenland (see Wager & Brown, 1968, p. 475, 478). A major angular unconformity occurs in the western part of the Bushveld Complex where the magnetite-rich rocks of its Upper Zone discordantly overlie earlier layers (including the Pt-rich Merensky Reef; cf. Wager & Brown, 1968, fig. 206). A third type of regressive discontinuity might be called a paraconjormity. Like its counterpart in sedimentary rocks, it is an apparently conformable contact, but rather than being an inconspicuous surface of erosion or representing an extended time period of nondeposition, it is defined by a regression in the succession of cumulates that would be expected for the crystallization path of the parental magma. Examples are the basal contacts of the cyclic units in the Stillwater Complex (Jackson, 1961), the Muskox Intrusion (Irvine & Smith, 1967), and the Rhum Intrusion (Brown, 1956). Paraconformities are commonly modal contacts marked by the reappearance of a mineral phase or assemblage that had undergone a cumulus termination at some lower level in the stratigraphic column. In some cases, they are cryptic regressions evident through the recurrence of mineral-composition trends (e.g. Irvine, 19806, figs. 8, 9). Approach SUBDIVISION OF LAYERED INTRUSIONS As a rule, layered intrusions have to be divided into parts for purposes of description and genetic analysis. The problem is comparable to stratigraphic subdivision of sequences of sedimentary and volcanic rocks, where consideration is given both to observed lithostratigraphic and biostratigraphic units and to inferred time-stratigraphic units. In a layered intrusion, consideration must be given to structure as well as stratigraphy, and the value of a subdivision may be enhanced through genetic considerations of crystallization relations based on textural features and cumulus-mineral sequences and compositions. There has, in fact, been some attempt to apply standardized stratigraphic nomenclature to layered intrusions. Inevitably the same names arise because of the limited availability of suitable words principal examples are 'series', 'zone', 'group', and 'member'. But most of the ways these names have been used in layered intrusions do not conform with stratigraphic conventions. For example, the four names just mentioned have all been employed in much the same lithostructural sense, whereas in sedimentary stratigraphy, series is time-stratigraphic, zone is biostratigraphic, and group and member are lithostratigraphic. On the other hand, the important lithostratigraphic term 'formation' is almost unheard of in studies of layered intrusions, apparently because it has connotations that are not appealing in this context. This author's overall impression is that rigorous application of the conventional stratigraphic rules would require extensive and frequently unappealing revisions for many intrusions. On the other hand, a more definite nomenclature is desirable, and in the pages to follow a system is proposed based on those parts of existing classifications that appear to have been most successful. The system is first described with brief illustrative examples; then its applicability is examined in more detail with respect to several major intrusions.

26 152 T. N. IRVINE The system consists of two sets of terms one for the overall subdivision of an intrusion, the other for the identification of distinctive parts or combinations of parts. Each set includes specific names plus a general name that covers them all. For the overall subdivision of an intrusion, the general name is 'division' (or 'subdivision'); the specific names are 'series', 'zone', and 'subzone'. For distinctive parts of an intrusion, the general term is 'unit' or 'layering unit'; the principal specific names are 'group', 'member', and 'layer' (or 'band', if appropriate), and two others are 'rhythmic unit' and 'cyclic unit'. All these names are basically informal, but any of them can be incorporated into ajormal name. Formal names should be capitalized but may be abbreviated with symbols. They apply when the division or unit in question is uniquely identified. For example, the Skaergaard Layered Series (abbreviated LS) has been specifically defined and therefore is a formally named division (even though Wager and his colleagues did not use capitalization). (It might be emphasized that the distinction between formal and informal terminology is a matter of definition, not just capitalization. Thus, 'the Bushveld Lower Zone' is necessarily a formal name, because it requires that the zone be specifically defined, whereas 'the lower part of the Stillwater Complex' and 'the main dunite zone in the Duke Island Complex' are informal references because they are sufficient in themselves to identify the features in question (as on a geologic map)). Divisions The names series, zones, and subzones will be recognized as those used by Wager (1960 and elsewhere) for the Skaergaard Layered Series. In fact, his nomenclature is practically a type example for the system suggested here. The proposed use of 'series' will undoubtedly draw criticism because it conflicts with the time-stratigraphic specifications developed for conventional stratigraphy. According to the American Committee for Stratigraphic Nomenclature, the term pertains to rocks formed during a geologic epoch, and it should not be applied to successions of lithostratigraphic formations or to successions of volcanic eruptions or igneous intrusions (cf. Bates & Jackson, 1980). In igneous petrology, however, series traditionally has been applied to lineages such as 'rock series' and 'magma series', wherein the features in question form a natural sequence. And in applications to layered intrusions, it very appropriately connotes (on the basis of general dictionary definitions) succession and progression, characteristics that are commonly represented in these intrusions in all of space (stratigraphically), time (geologically), and rock and mineral compositions. In actual application, however, the word is usually modified in such a way that it refers to structural divisions (as in the Skaergaard Layered Series). Thus, as defined here, the name series, appropriately modified (examples: layered, marginal, and border series), applies to structural divisions embodying major stratigraphic successions of cumulates. At the top of the hierarchy of divisions, a series can stand alone in that it is not part of anything less than the intrusion itself. 'Zone' is probably the most popular name that has been used for formal subdivisions of layered intrusions, but it has a host of informal applications as well, as in 'alteration zone' and 'zone of deformation'. For formal purposes, zones and subzones are here defined as first- and second-rank stratigraphic subdivisions of series. In the Skaergaard Intrusion, Wager (1960) delimited zones on the basis of the stratigraphic range of a key cumulus mineral (olivine), and subzones were marked by the cumulus appearance of other principal minerals. This method has now been used in several other intrusions, but it is not always practical for example, in the Bushveld and Stillwater Complexes, there are simply too many layer repetitions for it to be applied rigorously. The basis for choosing zones and subzones must therefore depend on the intrusion. The essential feature is that their limits be definite, and if possible, zones at least should be mappable.

27 TERMINOLOGY FOR LAYERED INTRUSIONS 153 A suggested practice for application of the division names is that a division does not have to be subdivided, but if it is, the subdivision should be complete. Thus, in the Skaergaard Layered Series, the Middle Zone is not divided, but the Lower and Upper Zones are completely broken down into subzones (cf. Wager, 1960). Also, it is intended that a subzone should always be part of a zone, and a zone part of a series. Units By contrast, units are here defined as compositionally distinctive features that can be singled out and named or numbered for specific reference. They are not necessarily part of anything except the intrusion. For layered rocks, the general term is layering unit. It is intended to be nongenetic and can be applied to any layer or combination of layers that can be usefully isolated for mapping, description, or discussion. Members and groups are meant to have formal names. A member is an outstanding single unit; a group is a prominent combination of several similar units (which may or may not be members). The Skaergaard Triple Group could be regarded as a type example of both. It consists of three layering units or members that might be denoted TGI, TG2, and TG3, each a distinctive (relatively thick) combination of a melanocratic (gabbroic) layer above a leucocratic (anorthositic) layer. These three members are close together stratigraphically, but they are not consecutive, and they form only a small part of the Middle Zone of the Layered Series. There have been significant differences of opinion concerning the definition and application of the names 'rhythmic unit' and 'cyclic unit'. In fact, their past use includes prime examples of the problems of the same feature being called by different names and of different features being called by the same name. As mentioned earlier, the present choice has been to define rhythmic unit as a descriptive term and cyclic unit as a genetic term. The history of the problem and the bases for this choice are complicated, however, and so are placed in an Appendix. Only the definitions are given here. A rhythmic unit is defined as a succession of layer types that is repeated stratigraphically in a regular way (e.g. Plate 6). If the unit is relatively thick (say, more than 5 m), it is a macrorhythmic unit (Plate 7B); if it is very thin (say, with layers only 1 cm thick) it may be called a microrhythmic unit (Plate 7B). The origins of the units and their repetition are not pertinent to the use of these names. The three prominent layering units of the Skaergaard Triple Group are regarded as type macrorhythmic units (Wager & Brown, 1968, plate IV). A cyclic unit (Jackson, 1961, 1970) is defined as a rhythmic unit (usually a macrorhythmic unit) in which the cumulus-mineral sequence can be identified as the fractional crystallization order of the magma (Plate 8). Among occurrences familiar to the author, cyclic units are particularly well developed in the Muskox Intrusion, and in this case, there is strong evidence also that the repetition reflects the repeated addition of fresh magma to the intrusion while it was solidifying (Irvine & Smith, 1967; Irvine, 1980). This interpretation is also favored for the repetition of cyclic units in several other major intrusions, such as the Stillwater Complex. (Raedeke & McCallum, 19806) and the Rhum Layered Intrusion (Dunham & Wadsworth, 1978), but it is probably not sufficiently well established to be specified in the definition. The upper part of the Stillwater Complex appears to comprise complex major cyclic units, m thick, that have been called megacyclic units (Todd et al., 1981). Cyclic units developed on the scale of a thin section would be appropriately called microcyclic units. Where cyclic units occur, it is common that some have fewer layers than others, apparently because their parental liquid did not progress through as many stages of crystallization. Jackson (1967, 1970) referred to these units as 'beheaded', but the term is poorly chosen because it implies that layers have been removed, whereas all indications (such as no evidence of erosion) are that the extra layers never were developed. The term might also be taken to imply that some cyclic units are 'complete', but few if any exhibit the full gamut of

28 154 T. N. IRVINE crystallization stages that might be expected of their parental liquid. In fact, almost by definition, cyclic units are inherently incomplete. Applications In the discussion to follow, the actual divisions and units currently recognized in the intrusions examined are accepted as appropriate to the present state of knowledge. Any suggestions for revisions pertain only to their names. The author has no illusions that these suggestions will be accepted, however the real purpose is to demonstrate the above system and some of the ways it can be used to deal with problems. In addition to being consistent with Skaergaard Layered Series terminology, the proposed system is essentially in accord with existing terminology for the Rhum and Aberdeenshire Intrusions in Scotland (Wadsworth, 1961, 1970), the Kap Edvard Holm and Kaerven Intrusions in Greenland (Wager & Brown, 1968), the Kapalugulu Intrusion in Tanzania (Wadsworth, 1963), and the Kaapmuiden Intrusion in South Africa (Viljoen & Viljoen, 1970). As is discussed below, it is also consistent with existing classifications of the Muskox Intrusion and Great Dyke, it has been applied to the Stillwater Complex, and the author believes that it could usefully be applied to the Bushveld Complex. The Muskox Intrusion is currently divided into a Feeder Dike, two Marginal Zones (East and West), a Layered Series, and a Granophyric Roof Zone (Irvirie, 19806). The Marginal Zones have potential subzones (still named only by rock type) and together could be said to comprise a Lower Marginal Series. A Roof Marginal Series could also be established, consisting of a Granophyre Zone and a Roof Breccia Zone. The Layered Series contains 42 layers that are formally identified by number (± letter suffixes), and they are grouped into 25 cyclic units, also identified by number. The Layered Series could additionally be divided into a lower, 'Ultramafic Zone' and an upper, 'Gabbroic Zone', but to date there has been little call for such a breakdown. In the case of the Great Dyke, each of the four major segments or 'complexes' identified by Worst (1960) could have its own Layered Series, each with a lower, Ultramafic Zone and an upper, Gabbroic Zone. The Ultramafic Zones are clearly divisible into cyclic units (Jackson, 1970). The application of the suggested classification system to the Stillwater Complex was made by Todd et al. (1981). This intrusion has long been divided into Basal, Ultramafic, and Banded Zones (Peoples, 1939), but two contrasting systems of more detailed subdivision had evolved in recent years. In one system, developed by a succession of investigators, the Basal Zone has a Basal Norite Member overlain by a Basal Bronzitite Member (Page, 1977); the Ultramafic Zone has a Peridotite Member containing at least 15 cyclic units, plus an Upper Bronzitite Member (Jackson, 1961, 1967; Page, 1977); and the Banded Zone has ten members named according to their contents of norite, gabbro, and anorthosite (Segerstrom & Carlson, 1977, 1979, 1980; Carlson & Segerstrom, 1978). In the other system (McCallum et at., 1980; Raedeke & McCallum, 1980a), the divisions of the Basal and Ultramafic Zones are essentially the same, except they are called subzones rather than members (and a few more cyclic units are identified). The Banded Zone is divided into 12 subzones with names (and, in some cases, boundaries) different from the members in the other system. The revisions proposed by Todd et al. were principally designed to accommodate the complicated local stratigraphy associated with a major Pt-Pd ore zone called the J-M Reef. These authors elevated the Basal, Ultramafic, and Banded Zones to series and divided them into finer zones (roughly equivalent to the previous members or subzones) and subzones. The stratigraphy of the Banded Series was simplified by grouping its zones into six megacyclic units based on broad patterns of lithologic repetition, and ten thin olivine-bearing units associated with the

29 TERMINOLOGY FOR LAYERED INTRUSIONS 155 J-M Reef are identified as members. Within the framework of this system, important chromitite 'zones' in the Ultramaflc Series (e.g. the G and H zones; Jackson, 1967, 1969) could be called 'members', thereby eliminating the nomenclatural inconsistency associated with their original definitions. The stratigraphy of the Bushveld Complex has recently been reviewed by the South African Committee for Stratigraphy (SACS, 1980), who attempted to bring it into accord with the stratigraphic code. The Committee chose to discredit a traditional, general classification scheme (declaring it 'informal') in favor of a system based on local place names. The change is unfortunate in several respects, however, and the following comments include some suggestions for a compromise based on the system proposed here. The name 'Bushveld Complex' itself warrants comment. As explained by the SACS (1980), it was originally applied to a large array of plutonic and volcanic 1 rocks underlying an area of some 65,000 km 2 in the central part of Transvaal Province. The SACS (1980) specifically considered three major intrusive divisions: (1) the ultramafic and mafic rocks that make up the main layered intrusion (which, on the basis of its enormous size alone is almost certainly derived from multiple injections of magma); (2) the granitic rocks contained in several large plutons that cut the layered intrusion; and (3) various granophyric rocks that mostly occur in the volcanic and sedimentary remnants of the roof of the layered intrusion. Multitudinous basic sills in the layered intrusion floor, and volcanic rocks in both the roof and floor, were considered part of the complex but were not discussed. The layered intrusion rocks were called "The Rustenburg Layered Suite', but this name is rather awkward in that two of their three major exposure areas are more than 200 km from the city of Rustenburg*. An alternative approach, based on the practice of many geologists and followed here, is to call the layered intrusion itself the Bushveld Complex (short for Bushveld layered intrusive complex). Although this practice is probably more common outside South Africa than inside, it has certainly been encouraged by the title (and foreword) of Geological Society of South Africa Special Publication 1, 'Symposium on the Bushveld Igneous Complex and Other Layered Intrusions'. The intrusion is clearly the central feature involved, and to call it the Bushveld Complex would also be consistent with the names of other major layered intrusive complexes (notably those at Stillwater and Duluth) and would avoid the question of whether it is actually one intrusion or several. Moreover, the total array of plutonic and volcanic rocks in the region deserves a more imposing (and specific) name, such as 'the Bushveld Magmatic Province'. With regard to stratigraphic classification, it is particularly desirable for the Bushveld Complex, because of its size and complexity, to have a system that is amenable both to generalization in the one extreme and to very detailed subdivision in the other. The traditional classification scheme is an ideal base in these regards. Among the three main areas of the complex, the northern is still not well known, but the western and eastern had come to be broadly divided into Marginal, Lower, Critical, Main, and Upper Zones (see SACS, 1980, for a historical summary). The general locations of the zones can be readily appreciated from this simple list, and the lithologic successions for the two major outcrop areas are sufficiently similar to justify the use of the same division names for both. In addition, numerous conspicuous layers or layering units have been distinguished as stratigraphic 'markers'. The best known of these are the famous, Pt-rich, Merensky Reef and the closely similar, but barren, Bastard Reef, both contained in well developed cyclic units at the top of the Critical Zone in both the east and the west. Other examples are Lower, Middle, and Upper Groups of The word suite is also somewhat unfortunate. It has had various applications in igneous petrology and stratigraphy (cf. Bates & Jackson, 1980, p. 626), but the definition chosen for Bushveld is the most recent and least familiar (if not the least acceptable) of the lot.

30 156 T. N. IRVINE chromitite units in the western part of the Critical Zone. These layers have been assigned symbols (LG1-LG7; MG1-MG4; UG1, UG2; Cousins & Feringa, 1964) and several of them have also been definitely identified in the eastern part of the complex. Main Zone contains distinctive pyroxenite and anorthosite markers, and Upper Zone has a succession of prominent magnetite-rich layers (e.g. Willemse, 1969; Molyneux, 1970). The SACS (1980) condemned the use of such markers for stratigraphic subdivision on the grounds that many of them do not crop out well enough to be mappable. Its remains, however, that (1) in almost all discussions of Critical Zone stratigraphy, the chromitite layers and Merensky Reef are standard reference units, and (2) many of the details of Bushveld stratigraphy will necessarily have to be resolved through drill-hole and mining exposures. In this author's view, the markers (which would be 'members' in the system proposed here) have fundamental stratigraphic importance simply because they are the units that can be correlated with least doubt, whatever the mode of their exposure. The divisions proposed by the SACS (1980) cover separately the entire stratigraphies of the western and eastern parts of the complex. The divisions are described as lithostratigraphic equivalents of 'formations' but are not actually called by that name. Each is identified by the names of a type locality and one or two principal rock types. Most are only parts of traditional zones (some are equivalent), but in general they are still lithologically complicated successions of layers. This author believes that a more satisfactory classification would be achieved if these divisions were defined as parts of the traditional 'zones', rather than being presented as alternatives to them. But first, on the basis of experience with the Stillwater Complex where similar problems have been encountered, he would rename the traditional zones, calling them the Basal, Lower, Critical, Main, and Upper Series. The divisions with place names could then be called 'zones', and eventually, where necessary, they could be divided into 'subzones'. Given further that the names of each of the above series begins with a different letter, their subdivisions might also be conveniently identified by symbols modeled after those commonly used in North American stratigraphic mapping. For examples: the Lower Series Jagdlust Harzburgite Zone could be denoted LSjh, and its subzones might be LSjh,, LSjh 2,...; the Critical Series Mooihoek Pyroxenite Zone could be CSmp; and so on. In other applications of the classification system proposed here, two principal problems concern the use of 'series' and 'group'. The first case involves the classical nomenclature of the marginal divisions of the Skaergaard Intrusion. Wager & Deer (1939) called these divisions the Marginal and Upper Border 'Groups' and showed that they have subdivisions approximately equivalent to the zones and subzones of the Layered Series (see also Wager & Brown, 1968; Naslund, 1976, 1980; Hoover, 1978). In view of this equivalence, however, the 'groups' would more logically be called 'series'. Then, by the present system, the subdivisions of the Marginal Border Series (MBS) might be termed the Marginal Contact Zone (MCZ; preferable to the present name Tranquil Zone, which has genetic connotations that may not be appropriate) and the Outer, Middle, and Inner Banded Zones (OBZ, MBZ, IBZ). The subdivisions for the Upper Border Series (UBS) could be the Upper Contact Zone (UCZ; preferable to the present name Outer Border Zone), and Upper Border Zones Alpha, Beta, and Gamma (UBZa, UBZ/?, UBZy, preferable to the present UBG subdivision notation). The obvious alternative to the above is to change 'Layered Series' to 'Layered Group', leaving the nomenclature of the marginal divisions unchanged. However, the connotation of progression would be lost, and there would be the inconsistency of the Triple Group occurring within the Layered Group. The second group-m.-series problem relates to the Kiglapait Intrusion (Morse, 1969). Original terminology defined for this body is essentially in accord with the nomenclature

31 TERMINOLOGY FOR LAYERED INTRUSIONS 157 proposed here, but Morse (1972) recently substituted Layered Group for Layered Series 'in conformance with standard stratigraphic nomenclature' (where group is lithostratigraphic and series, time-stratigraphic). The logical extension of this argument, if a standard nomenclature is indeed in order, is that this change should be made in all intrusions in which 'Layered Series' has been used. Perhaps so, but strictly applied, such a change would affect the nomenclature not only for these intrusions (which are of substantial number) but also for the many intrusions in which 'Group' has been used. It might also be argued that, with the standard nomenclature, it would be possible to make time-stratigraphic subdivisions of layered intrusions. But although such subdivisions are certainly of interest, a specific nomenclature hardly seems necessary. Also, in some cases at least, the combinations of rocks that might comprise such divisions would be poorly described by the standard terms. For example, the Skaergaard Middle Zone (MZ) and its Marginal and Upper Series equivalents (MBZ and UBZ/7) appear to have formed concurrently, but they do not constitute a 'series' by any dictionary definition of the word in fact, they would be better described as a 'group'. Thus, the contentions here are that layered intrusions.are sufficiently distinctive in their idiosyncracies and problems to warrant their own system of classification, and that the Skaergaard Layered Series has now been used as a model enough times that, effectively, it is a well established standard in its own right. CONCLUDING REMARKS Although the above analysis touches on a large number of features, terms, and concepts, it is by no means complete. Also, other approaches are possible, other names can be found for some features, and other definitions can be formulated for certain names. From previous experience with classification and nomenclature, the author's impression is that the interested audience can generally be divided into two notably distinct 'groups' (companies of individuals of similar viewpoint). The members of the one group simply want a reasonable and practical system that they can use. The degree of their acceptance is one measure of the success of a system. The members of the other group are personally interested in taxonomy and are moved to design their own system*. The success of their alternatives is a measure of a system's shortcomings. In time, of course, all systems have to be updated. The terminology defined above will be no exception to the last statement, but it is believed to be reasonable and practical at the present state of knowledge of layered intrusions. A principal hope is that at least the logic of the nomenclature will be appreciated, even if all the names and definitions (and the author's opinions) are not. With regard to additions or changes, three guidelines are suggested for choosing a new name: (1) It should be grammatically correct. (2) It should describe as accurately as possible within the limitations of its length the feature to which it applies. An inaccurate or misleading name is worse than no name. (3) It should be sufficiently distinctive not to be confused with any name that has been widely used for a different feature. An ideal new name is one that satisfies requirements (1) and (2) but originates outside the current vocabulary of its subject like 'cryptic' in 1939 and 'cumulate' in Another guideline is do not introduce a new name unless it is to be used! Such emphasis on terminology may seem overly academic. Words are, however, our principal means of communication. They convey our observations and insights and our prejudices and ignorance. They frequently have connotations that go far beyond any In addition, there are always a few individuals who chose not to follow any system.

32 158 T. N. IRVINE dictionary definition. Their choice is important not only for addressing an audience, but also in influencing one's own thinking. If not used carefully, they will indeed be foes. ACKNOWLEDGEMENTS This paper represents an attempt to maintain order in a subject in which concepts have been changing rapidly. A consequence that I regret is that the reevaluation of terminology has resulted in some criticism of the choices made by the geologists who have been most innovative in developing the terminology L. R. Wager, G. M. Brown, W. J. Wadsworth, and E. D. Jackson. I would emphasize, therefore, that the primary objective has been to preserve the cumulate terminology that they all advocated. This goal reflects my respect for their contributions; the criticisms are trivial by comparison. Many of the observations and thoughts expressed derive from associations and discussions with colleagues interested in layered intrusions. They deserve some credit for what may be good, but no blame for what is not. In my personal chronology, they include H. D. B. Wilson, J. A. Noble, H. H. Schmitt, H. P. Taylor, H. V. Tuominen, C. H. Smith, D. C. Findlay, N. D. MacRae, A. J. Naldrett, J. A. V. Douglas, T. P. Thayer, E. D. Jackson, N. J. Page, A. B. Ford, E. N. Cameron, G. Von Gruenewaldt, W. J. Wadsworth, R. D. Ridler, A. R. McBirney, H. R. Naslund, J. D. Hoover, I. H. Campbell, S. G. Todd, D. W. Keith, R. R. Carlson, and many others. I thank L. M. Irvine for literary guidance. The manuscript was improved through constructive reviews by F. Chayes, I. H. Campbell, L. J. Hulbert, D. W. Keith, D. Twist, S. G. Todd, W. J. Wadsworth, and H. S. Yoder, Jr. Some of the field work was supported by National Science Foundation grant EAR APPENDIX Cyclic unit versus macrorhythmic unit The name 'rhythmic unit' was first used in the present context by Brown (1956) in application to a layer succession that is repeated ten times in the Rhum Intrusion. The Rhum units typically are m thick and consist of a relatively thick layer of peridotite (olivine cumulate) overlain by a thinner troctolitic (plagioclase olivine cumulate) layer and, in some cases, by a third, gabbroic (plagioclase-augite-olivine cumulate) layer. Brown attributed the rock succession to crystallization differentiation, and he ascribed the repetition to formation in a subvolcanic chamber that was periodically replenished through concurrent removal of fractionated magma and addition of fresh liquid. The name 'cyclic unit' was originated by Jackson (1961) for repeated layering units in the Ultramafic Zone of the Stillwater Complex. The typical Stillwater example has a layer of peridotite (olivine cumulate) overlain by a modally graded layer of harzburgite (olivinebronzite cumulate) capped with a layer of bronzitite (bronzite cumulate). Jackson recognized 15 such units, most of them m thick. He proposed that they formed by fractional crystallization effected by crystals settling from a stagnant bottom layer of magma. The repetition was thought to arise because this process was periodically interrupted by episodes of convective overturn. Jackson believed that the Stillwater units differed significantly from the Rhum units in having sharper contacts between layers, and he cited this difference as the reason for the different name. Wager & Brown (1968) did not agree with this assessment, however. They argued that in both cases the repetition had a 'rhythmic nature', abc... abc, rather than a 'cyclic pattern', abcbabc, and they called the units 'macrorhythmic units'. They also reinterpreted the differentiation of the Rhum units, attributing it to a non-equilibrium process controlled by crystal nucleation.

33 TERMINOLOGY FOR LAYERED INTRUSIONS 159 In the meantime, Irvine & Smith (1967) had delineated 25 repeated layering units in the Muskox Intrusion. The most completely developed examples feature layer successions of peridotite, pyroxenite, and gabbro. Thicknesses range from 10 to 125 m, and layer contacts typically are sharp. Irvine & Smith showed that the cumulus mineral successions could be closely simulated by fractional crystallization models based on equilibrium phase diagrams, and they provided quantitative evidence derived from chemical and structural data and fractionation models that the layer repetition was due to repeated replenishment of a subvolcanic magma chamber. From their own comparisons, they recognized that the Muskox units were similar to those in the Stillwater Ultramafic Zone, and accordingly, they called them cyclic units. Jackson (1970) then applied their phase-diagram analysis method to the Stillwater units and used it also in demonstrating the probable existence of cyclic units in the Great Dyke and Bushveld Complex. Irvine (1970) noted that the cumulus-mineral sequences in the Rhum units could similarly be predicted, and W. J. Wadsworth (personal communication, 1980) has since detected appropriate Mg/Fe fractionation trends in these units, which Dunham & Wadsworth (1978) have referred to as 'cyclic units'. From personal observations, the present author feels no doubt that the units of the Stillwater Ultramafic Zone, the Muskox Intrusion, the Great Dyke, the Rhum Intrusion, and some of the harzburgitic Bushveld units are all essentially the same phenomenon. (Some of the Rhum layer contacts may be more gradational, but the difference is scarcely noticeable and does not seem very fundamental.) He has also seen this kind of unit in two intrusions in Ontario the Kakagi Lake sill (Ridler, 1966) and the Centre Hill Intrusion (MacRae, 1969) and he would venture to guess that repeated layering units in many other intrusions are the same for examples, units in the Kapalugulu Intrusion (Wadsworth, 1963), the Kaapmuiden Intrusion (Viljoen & Viljoen, 1970), and the Jimberlana Intrusion (Campbell, 1977). In all the above examples, the cumulus mineral succession is reasonably identified with the crystallization order of the parental magma, and in most cases the unit repetition has been attributed to magma replenishment (even in Stillwater, Raedeke & McCallum, 1980ft). A principal example of an apparently different type of repeated layering unit is the three units of the Skaergaard Triple Group. They are descriptively similar in that they are repeated, laterally extensive (c. 4 km 2 ) layer successions, tens of meters (20-60 m) thick. But from available information at least (Wager & Brown, 1968; Maaloe, 1978; and observations and sampling by the author), the cumulate succession does not appear to follow the Skaergaard crystallization order. The rock sequence is a plagioclase-rich, almost anorthositic layer overlain gradationally by a darker, gabbroic layer. In the topmost unit, the dark layer locally contains cumulus olivine (Wager & Brown, 1968, p. 71; Maaloe, 1978), even though the Group as a whole occurs in the nominally olivine-free Middle Zone of the Layered Series, but in the author's samples from all three units, the cumulus assemblages of the light and dark layers are,the same; only the proportions of plagioclase are different. Moreover, the Triple Group is closely associated with several dozen thinner layering units (typically 3-7 m thick) of similar lithology and comparable lateral extent that appear to be thinner versions of the same phenomenon (Plate 5; also Wager & Brown, 1968, plate IV). McBirney & Noyes (1979, p. 512) categorized this layering as 'cyclic layering', but in the present author's view, the whole association is very unlike the cyclic units of the Stillwater, Muskox, tjjushveld, Great Dyke, and Rhum Intrusions. Also, from past accounts of its petrogenesis, it seems highly unlikely that the Skaergaard Intrusion was open to several dozen additions of fresh magma. Another example of repeated layering units that appear fundamentally different from the Stillwater-Muskox-Rhum units is the 'black, red, and white' units of kakortokite described by Ussing (1912) in the Ilimaussaq Syenitic Intrusion (see Wager & Brown, 1968,

34 160 T. N. IRVINE p. 489). These typically are m thick, and their physical appearance is strongly reminiscent of the Skaergaard Triple Group and macrorhythmic layering. The difference between 'rhythmic' and 'cyclic' cited by Wager & Brown (1968) appears to be a rather personal distinction. Their 'cyclic pattern' implies an element of reversibility that is not mentioned in standard dictionary definitions, including discussions in the Glossary oj Geology (Bates & Jackson, 1980). These definitions emphasize only 'recurrence' of items or events. Thus, in familiar applications in geology, as in cyclic sedimentation or cyclic erosion, the standard pattern is youth, maturity, old age and then rebirth, not reversal. Notable also is that, if there is a significant difference between rhythmic and cyclic in these applications, it is that rhythms tend to be small and cycles large. This contrast is evident in the present case, in that the choice is between cyclic unit and macrorhythmic unit. In the light of all the above, the author believes that a reasonable compromise is to use both terms, but in different ways. He chooses to retain 'cyclic unit' for units of the Stillwater-Muskox-Rhum type, but with a genetic definition based on considerations of crystallization order. Given, on the other hand, that Skaergaard is the type locality of 'rhythmic layering', it seems fitting to call its successions of thick layers 'macrorhythmic layering' and to take the Triple Group units as type 'macrorhythmic units'. Given further that the origin of these units is still unknown, it is proper to define macrorhythmic unit as a descriptive name (see text). The Triple Group macrorhythmic units may someday prove to be cyclic units, but for the moment their name involves no genetic commitment If, on the other hand, they should prove to have a different origin, then some appropriate new genetic name should be improvized, preserving macrorhythmic as a descriptive adjective. REFERENCES Arndt, N. T., Naldrett, A. J., & Pyke, D. R., Komatiitic and iron-rich tholeiitic lavas of Munro Township, northeast Ontario, J. Petrology, 18, Bates, R. L., & Jackson, J. A., Glossary of Geology. Falls Church, Virginia: American Geological Institute. Bottinga, Y., & Weill, D. F., Densities of liquid silicate systems calculated from partial molar volumes of oxide components. Am. J. Scl. 269, Brown, G. M., The layered ultrabasic rocks of Rhum, Inner Hebrides. Phil. Trans. R. Soc. Lond. B, 240, Cameron, E. N., Structure and rock sequences of the Critical Zone of the eastern Bushveld Complex. Spec. Pap. Miner. Soc. Am. 1, The Lower Zone of the Eastern Bushveld Complex in the Olifants River Trough. J. Petrology, 19, & Desborough, G. A., Origin of certain magnetite-bearing pegmatites in the eastern part of the Bushveld Complex, South Africa. Econ. Geol. 59, Campbell, I. H., A study of cumulate processes and macrorhythmic layering in the Jimberlana Layered Intrusion of Western Australia. Part I: The upper layered series. J. Petrology, 18, Some problems with the cumulus theory. Ltthos, 11, Roeder, P. L., & Dixon, J. M., Plagioclase buoyancy in basaltic liquids as determined with a centrifuge furnace. Contr. Miner. Petrol. 67, Carlson, R. R., & Segerstrom, K., Preliminary geologic map of the East Boulder Sector of the Stillwater Complex, Sweetgrass Counties, Montana. Open-File Rep. U.S. geol. Surv Casey, J. F., & Carson, J. A., Magma chamber profiles from the Bay of Islands Ophiolite Complex: implications for crustal-level magma chambers at mid-ocean ridges. Earth planet. Scl. Lett, (in press) Conrad, J., Under Western Eyes. London: Methuen. Cousins, C. A., eft Feringa, G., The chromite deposits of the western belt of the Bushveld Complex. In: Haughton, S. H. (ed.), The Geology of Some Ore Deposits in South Africa, II. Johannesburg: Geological Society of South Africa. Dunham, A. C, & Wadsworth, W. J., Cryptic variation in the Rhum Layered Intrusion. Miner. Mag. 42, Dzulynski, S., & Walton, E. K., Sedimentary Features of Flysch and Greywackes. Amsterdam: Elsevier Publishing Co. FeTguson, J., Geology of the llimaussaq Alkaline Intrusion, South Greenland. Meddr. Grtnland, 172, Hess, H. H., Stillwater Igneous Complex, Montana: A quantitative mineralogical study. Mem. geol. Soc. Am. 80.

35 TERMINOLOGY FOR LAYERED INTRUSIONS 161 Hoover, J. D., Petrologic features of the Skaergaard Marginal Border Group. Yb. Carnegie Instn. Wash. 77, Irvine, T. N., Crystallization sequences in the Muslcox Intrusion and other layered intrusions. I. Olivine-pyroxene-plagioclase relations. Spec, Publ. geol. Soc. S. AJr. 1, Petrology of the Duke Island Ultramafic Complex, southeastern Alaska. Mem. geol. Soc. Am Rocks whose compositions are determined by crystal accumulation and sorting. In: Yoder, H. S., Jr. (ed.), Evolution of the Igneous Rocks: 50th Anniversary Perspectives. Princeton, N. J.: Princeton University Press, a. Magmatic density currents and cumulus processes. Am. J. sci. 280A, Magmatic infiltration metasomatism, double-diffusive fractional crystallization, and adcumulus growth in the Muskox Intrusion and other layered intrusions. In: Hargraves, R. B. (ed.), Physics oj Magmatic Processes. Princeton, N. J.: Princeton University Press, c. Observations on layering in the Skaergaard Intrusion. Yb. Carnegie Instn. Wash. 79, & Smith, C. H., The ultramafic rocks of the Muskox Intrusion. In: Wyllie, P. J. (ed.), Ultramafic and Relaxed Rocks. New York: John Wiley & Sons, Inc., Jackson, E. D., Primary textures and mineral associations in the Ultramafic Zone of the Stillwater Complex, Montana. Prof. Pap. U.S. geol. Surv Ultramafic cumulates in the Stillwater, Great Dyke, and Bushveld Intrusions. In: Wyllie, P. J. (ed.), Ultramafic and Related Rocks. New York: John Wiley & Sons, Inc., Chemical variation in coexisting chromite and olivine in chromitite zones of the Stillwater Complex. Monogr. Econ. Geol. 4, The cyclic unit in layered intrusions a comparison of the repetitive stratigraphy in the ultramafic parts of the Stillwater, Muskox, Great Dyke and Bushveld Complexes. Spec. Publ. geol. Soc. S. AJr. 1, 391^*24. Jahns, R. H., The study of pegmatites. In: Bateman, A. M. (ed.), Fiftieth Anniversary Volume, Economic Geology. Lancaster, Pa.: Economic Geology Publishing Co., Maaloe, S., The origin of rhythmic layering. Miner. Mag. 42, McBirney, A. R., Differentiation of the Skaergaard Intrusion. Nature, 233, & Noyes, R. M., Crystallization and layering of the Skaergaard Intrusion. /. Petrology, 20, McCallum, I. S., Raedeke, L. D., & Mathez, G. A Investigation of the Stillwater Complex: Part I, Stratigraphy and structure of the Banded Zone. Am. J. Set. 280-A, MacRae, N. D., Ultramafic intrusions of the Abitibi area, Ontario. Can. J. Earth Sci. 6, Molyneux, T. G., The geology of the area in the vicinity of Magnet Heights, Eastern Transvaal, with special reference to the magnetitic iron ore". Spec. Publ. geol. Soc. S. Afr. 1, Moore, J. G., & Lockwood, J. P., Origin of comb layering and orbicular structure, Sierra Nevada Batholith, California. Bull. geol. Soc. Am. 84, Morse, S. A., The Kiglapait Layered Intrusion, Labrador. Mem. geol. Soc. Am., The feldspar/magma density paradox. Univ. Mass. Geol. Dept. Contr. 11, a. Kiglapait geochemistry. I: Systematics, sampling, and density. J. Petrology, 20, Kiglapait geochemistry. II: Petrography. Ibid. 20, Naslund, H. R., Mineralogical variations in the upper part of the Skaergaard Intrusion, East Greenland. Yb. Carnegie Instn. Wash. 75, Petrology of the Upper Border Group of the Skaergaard Intrusion, East Greenland. Ph. D. thesis, University of Oregon. Page, N. J., Stillwater Complex, Montana: Rock succession, metamorphism, and structure of the complex and adjacent rocks. Prof. Pap. U.S. geol. Surv. 999, Parsons, I., The Klokken gabbro-syenite complex, South Greenland: cryptic variation and origin of inversely graded layering. J. Petrology, 20, Peoples, J. W., The Stillwater Igneous Complex, Montana. Trans. N.Y. Acad. Sci. H.I, Philpotts, A. R., Igneous structures and mechanism of emplacement of Mount Johnson, a Monteregion Intrusion, Quebec. Can. J. Earth Sci. 5, Pyke, D. E., Naldrett, A. J., & Eckstrand, O. R., Archean ultramanc flows in Munro Township, Ontario. Bull. geol. Soc. Am. 84, Raedeke, L. D., & McCallum, I. S., 1980a. A comparison of fractionation trends in the lunar crust and the Stillwater Complex. In: Papike, J. J., & Merrill, R. B. (eds.), Proc. ConJ. Lunar Highlands Crust, Modal and chemical variations in the Ultramafic Zone of the Stillwater Complex. Abstr. Progrm. geol. Soc. Am. 12, 505. Ridler, R. H., Petrographlc Study of the Crow Lake Ultrabaslc Sill, Keewatin Volcanic Belt, Northwestern Ontario. M. A. Sc. thesis, University of Toronto. Segerstrom, K,, & Carlson, R. R., Preliminary geologic map of the upper zones of the western end of the Stillwater Complex, Park and Sweet Grass Counties, Montana. Open-File Rep. U.S. geol. Surv Preliminary geologic map of the Picket Pin to Mountain View sector of the Stillwater Complex, StillwateT County, Montana. Ibid Preliminary geologic map of the eastern end of the upper zones of the Stillwater Complex, Stillwater County, Montana. Ibid

36 162 T. N. IRVINE Smith, C. H., & Kapp, H. E., The Muskox Intrusion, a recently discovered layered intrusion in the Coppermine River area, Northwest Territories, Canada. Spec. Pap. miner. Soc. Am. 1, South African Committee for Stratigraphy (SACS), Stratigraphy of South Africa. Part 1 (Comp. L. E. Kent). Lithostratigraphy of the Republic of South Africa, South West Africa/Namibia, and the Republics of Bophuthatswana, Transkei and Venda: Handbook geol. Sure. S. Ajr. 8. Taubeneck, H., & Poldervaart, A., Geology of the Elkhorn Mountains, northeastern Oregon: Part 2. Wfllow Lake Intrusion. Bull. geol. Soc. Am. 71, Todd, S. G., Schissel, D. J., & Irvine, T. N., Lithostratigraphic variations associated with the platinum-rich zone of the Stillwater Complex. Yb. Carnegie Instn. Wash. 78, Keith, D. W., Leroy, L. W., Mann, E L., & Irvine, T. N., The J-M Platinum-Palladium Reef of the Stillwater Complex, Montana. I. Stratigraphic relationships. Econ. Geol. (in press) Ussing, N. V., Geology of the country around Julianehab, Greenland. Meddr. Gronland, 38, Viljoen, R. P., & Viljoen, M. J., The geology and geochemistry of layered ultramafic bodies of the Kaapmuiden area, Barberton Mountain Land. Spec. Publ. geol. Soc. S. AJrica, 1, Von GruenewaUt, G., & Weber-Diefenbach, K., Coexisting Ca-Poor pyroxenes in the Main Zone of the Bushveld Complex. Contr. Miner. Petrol. 65, Wadsworth, W. J., The ultrabasic rocks of southwest Rhum. Phil. Trans. R. Soc. Lond. B244, The Kapalagulu Layered Intrusion of western Tanganyika. Spec. Pap. miner. Soc. Am. 1, The Aberdeenshire Layered Intrusion of north-east Scotland. Spec. Publ. geol. Soc. S. Ajr. 1, Wager, L. R., The major element variation of the layered series of the Skaergaard Intrusion and a re-estimation of the average composition of the hidden layered series and of the successive residual magmas. /. Petrology, 1, The mechanism of adcumulus growth in the layered series of the Skaergaard Intrusion. Spec. Pap. miner. Soc. Am. 1, & Brown, G. M., Layered Igneous Rocks. Edinburgh: Oliver & Boyd, Ltd. & Wadsworth, W. J., Types of igneous cumulates. J. Petrology, 1, & Deer, W. A., Geological investigations in East Greenland, Part III. The petrology of the Skaergaard Intrusion, Kangerdluqssuaq, East Greenland. Meddr. Grtnland, 105, Wagner, P. A., The Platinum Deposits and Mines of South AJrica. Edinburgh: Oliver & Boyd. Willemse, J., The geology of the Bushveld Complex, the largest repository of magmatic ore deposits in the world. Monogr. Econ. Geol. 4, Worst, B. G., The Great Dyke of Southern Rhodesia. Bull. geol. Sun. S. Rhodesia, 47.