Analysis of compositional variations along profiles from tholeiitic sills provides insights into syn- and post-emplacement magmatic differentiation processes. We present here 18 whole-rock compositional profiles sampled from a saucer-shaped sill emplaced in the Karoo Basin (South Africa), the Golden Valley Sill. We show that different compositional profile patterns previously described in basic-ultrabasic sills may be found in different parts of a single saucer-shaped sill. The detailed examination of the mineral grain assemblage and compositions suggests that processes taking place in hundred-meter-thick sills relate to early and late fractional crystallization. Our observations in the Golden Valley Sill suggest that a significant part of fractionation takes place at a late stage of cooling when a crystalline skeleton or mush zone is formed. We show that porous flow of interstitial melt driven by forces related to the particular geometry (saucer-shaped) of the sill may result in a post-emplacement compositional evolution. We propose that the process of post-emplacement melt flow regionally overprinted compositional patterns produced by earlier crystal segregation from the cooling magma at fluid-like stages during the emplacement.

Bulk composition profiles of basic-ultrabasic sills have been investigated for a long time in order to understand the magmatic differentiation processes associated with their emplacement and crystallization. The most recurrent geochemical profiles observed in such sills are I-, D-, and S-shaped (Fig. 1; e.g., Gibb and Henderson, 1992; Latypov, 2003a). The nomenclature is based on the variations in the whole-rock Mg# (= cation proportion 100 × Mg/[Mg + Fetotal]) from floor to roof of the sills. The term I-shaped profile is used to describe sills showing little field, petrographic, and geochemical evidence of differentiation throughout the height of the sill (Mangan and Marsh, 1992; Marsh, 1996). The name D-shaped profile was introduced to characterize sills with the least-differentiated composition at the sill center (i.e., highest Mg#) and most evolved composition at the sill margins (i.e., lowest Mg#). Finally, an S-shaped profile shows S-shaped variations in Mg# upwards through the sill (Frenkel et al., 1988, 1989; Fujii, 1974; Gray and Crain, 1969; Marsh, 1989).

D- and S-shaped profiles observed in relatively thin sills (100 m) pose a problem because they are “inversed” from what is expected of cooling magma. Large layered mafic intrusions have typically mafic margins and felsic cores (normal zoning) forming C-shaped geochemical profiles (Fig. 1; Skaergaard intrusion, Naslund, 1984). C-shaped zoning or compositional profile is interpreted as the result of in situ processes of fractional crystallization (Rice, 1981). The lack of C-shaped profiles in relatively thin sills is significant and suggests that processes occurring in a large magmatic body are different from those in thin sills (100 m). An additional problem is that available data are generally restricted to one to three profiles in a sill (e.g., Richardson, 1979). Consequently, we usually do not know if a single sill contains different types of compositional profiles and if there are variations in type of profile along the boundary of sills. Such information is, however, of first-order importance in order to understand the magmatic differentiation processes that occur in sills.

Our strategy in this study was to document the vertical and lateral compositional and textural variations within a single well-exposed saucer-shaped sill in order to reveal the magmatic differentiation processes within this sill. Our study object was the Golden Valley Sill in the Golden Valley Sill Complex (GVSC), Karoo Basin, South Africa (Aarnes et al., 2008; Galerne et al., 2008; Polteau et al., 2008a). The excellent exposure of the Golden Valley Sill allowed systematic sampling along vertical profiles all along the inclined sheets/outer sill. We report here whole-rock analyses on 18 profiles (Fig. 2A; five samples per profile in average) together with mineral chemistry on representative profiles.

This paper has three main sections: The first section presents the geology and geochemistry of the Golden Valley Sill followed by sampling strategy and methodology. The next section gives a systematic overview of significant textural, petrographic, and chemical variations along the profiles. The profiles are classified on the basis of the established nomenclature, i.e., I-, D-, and S-shaped profiles. Profiles that did not fit into the established nomenclature were termed X-shaped profiles. Mineral chemical variations are illustrated for one profile. The last section gives a review of the main results and discusses the various types of compositional profiles found in the Golden Valley Sill in terms of cooling and crystallization processes.

The GVSC consists of a group of sills and dykes emplaced in an undeformed sequence of sandstone and shale (Fig. 2). A concordia age of 182.7 ± 0.3 Ma of a sill in the GVSC (Svensen et al., 2007) implies that these sills are contemporaneous with the Lesotho Flood basalts which are dated to ca. 183 Ma (Duncan et al., 1997). Sill complexes in the Karoo Basin represent a significant portion of the Jurassic Karoo igneous province (Galerne et al., 2008; Marsh and Mndaweni, 1998). Saucer-shape sills are common structures in sill complexes (Hansen and Cartwright, 2006; Hansen et al., 2004; Hansen et al., 2008; Polteau et al., 2008b; Thomson and Hutton, 2004). A saucer-shaped sill is characterized by a flat inner sill connected outward and upward to a ring of discordant inclined sheets, which are often terminated in a flat outer sill.

The GVSC has intruded the Beaufort Group in the Karoo Basin sedimentary sequences. Like numerous sills in the Beaufort Group, the GVSC's main sill morphology is saucer-shaped. Four major saucer-shaped sills, all elliptical in shape, constitute the GVSC; the Golden Valley Sill, Glen Sill, Morning Sun, and Harmony Sill (Fig. 2). Additionally, a minor subcircular saucer-shaped sill, the MV Sill, is attached to the northwestern boundary of the Golden Valley Sill. A major dyke (15 m thick and 17 km long, strikes NW-SE), the Golden Valley Dyke, has also been identified in the western part of the GVSC (Fig. 2).

The dolerites of the GVSC are basalts to basaltic andesites with TiO2 < 2 wt% and Zr < 180 ppm, which is characteristic of the low Ti-Zr (LTZ) basalts of the Karoo-Ferrar igneous province (e.g., Cox et al., 1967; Erlank et al., 1988; Marsh et al., 1997; Sweeney et al., 1994). A statistical analysis of the GVSC geochemical data set (Galerne et al., 2008) showed that the saucer-shaped sills exhibit three distinct geochemical signatures (shown in different colors in Fig. 2). The identical geochemical signature, in addition to the physical connection of the MV Sill and the Golden Valley Sill, lead the authors to suggest that the MV Sill resulted from lateral overflow of the major Golden Valley Sill. Also the Glen Sill has the same geochemical signature as the Golden Valley Sill (beige in Fig. 2). The different geochemical signatures imply that at least three separate magma pulses were involved in the formation of the sills in the GVSC (Galerne et al., 2008). The 100-m-thick Golden Valley Sill is the best exposed sill in the GVSC.

A main objective in this study was to get an overview of the variation in type of profile along the border of the sill rather than the finer variations in the sill structure. We therefore collected samples in 18 profiles along the boundary of the Golden Valley Sill. In each profile rock samples were taken every 5–10 m, from the chilled bottom contact to the roof (Fig. 3). Some samples were taken more than 10 m from the next because fresh samples could not be obtained in between, or because some parts of the cliff in between were inaccessible to sampling. Because of these sampling problems some profiles are represented by only three or four samples (e.g., P8, P9). The maximum number of samples along a single profile is ten (P11, P15). This means that the finer variations in the sill structure will be missed and the classification of some profiles may be dubious. However, in spite of these limitations we found a systematic distribution of profile types along the boundary of the Golden Valley Sill. We regard this result as very important and an indication that our sampling strategy served our main purpose.

The roof of the sill is usually removed by erosion and thus is not available for sampling. However, the horizontal entablature characteristic of cooling features in the top contact of magmatic bodies can be observed in the upper parts of the sill (Fig. 3B). In addition, sediment dykes (∼40–50 cm thick) have been observed at the surface of the exposed parts of the saucer-shaped sill. These observations suggest that the top of the sill is located close to the original roof. In many profiles the bottom contact with the host rock is exposed (example in Fig. 3C), but in several profiles the bottom contact was covered by debris. However, extrapolating the slope of the country-rock beneath the debris toward the cliff shows that the bottom contact is not more than 10 m below the lowermost exposures. We thus believe that most of the original thickness of the Golden Valley Sill is preserved and that we have sampled the full thickness of the sill along profiles.

The coordinates (latitude and longitude) acquired by GPS during sampling are reported with the whole-rock analyses in Tables 1 and 2 (and Supplemental Table 11). The distances between the samples were measured directly using a decameter, and controlled by GPS. These measurements were positioned with respect to the bottom contact, and corrected to give the real distance of the sample normal to the dip direction of the sill bottom contact. These values are reported in Tables 1 and 2 and Supplemental Tables 1, 22 and 33 under the acronym D.r.B. for distance with respect to the bottom contact. The corresponding values in height normalized to 100 for each sample in a given profile are also reported in Tables 1 and 2 and in Supplemental Table 1 (see footnote 1), and are illustrated in Figures 4–7,567.

Galerne et al. (2008) showed that the southern part of the Golden Valley Sill has a complicated relationship with the underlying Morning Sun sill. Based on field observations and differences in geochemistry the authors showed that profiles 17 and 18 (Figs. 2, 7, and 8) from the Golden Valley Sill west limb were composite. The lower part (10 m above floor contact) of these profiles corresponds to the continuation of the Morning Sun sill northwestward below the Golden Valley Sill. The profiles P17 and P18 were aimed to capture compositional variations in the two overlying sills (Golden Valley Sill and Morning Sun). In order to investigate more precisely the compositional variations along profiles from the underlying Morning Sun saucer-shaped sill, three additional profiles from the Morning Sun South segment are presented (green cylinders, Fig. 2A).

Powders for geochemical analyses were prepared from 30 g of fresh rock using a stone mill. Eighty-eight samples were analyzed by inductively coupled plasma–atomic emission spectrometry (ICP-AES) and inductively coupled plasma–mass spectrometry (ICP-MS) at the University of London, Royal Holloway. We weighed 0.2 g of powdered sample into a graphite crucible and added 1.0 g of LiBO2. The powders were carefully mixed and fused at 900 °C for 20 min. The resulting mixture was dissolved in 200 ml of cold 5% nitric acid. Ga was added to the flux to act as an internal standard. This solution was then analyzed for Si, Al, and Zr by ICP-AES using a Perkin Elmer Optima 3300R. The instrument was calibrated with natural and synthetic standards. The solution was also used to analyze for (Cs, Nb, Rb, Ta, Th, Tl, U, Y), and the rare earth elements (REE) using ICP-MS. The instrument used was a Perkin Elmer Elan 5000 and the instrument was calibrated with natural and synthetic standards. We also dissolved 0.2 g of powdered sample in 6 ml of HF and HClO4 (2:1 mixture). This was then evaporated to dryness, cooled, and dissolved in 20 ml of 10% HNO3. This solution was analyzed by ICP-AES for Fe, Mg, Ca, Na, K, Ti, P, Mn, Ba, Co, Cr, Cu, Li, Ni, Pb, Sc, Sr, V, and Zn. The analytical precision is 1% for Si, Al, Fe, Mg, and Ca and 2% for Na, K, Ti, P, and Mn. Detection limits for measured trace elements are reported in Table 1. The rest of the samples (47) were analyzed for major and trace elements (Rb, Sr, Nb, Zr, Y, Co, Cr, Cu, Ni, V, Zn; Table 2) by standard X-ray fluorescence (XRF) techniques at the University of Bergen (Norway). Standard deviations are reported in Table 2. Tables 1 and 2 present representative analyses of characteristic selected profiles from the Golden Valley Sill and Morning Sun sill. Supplemental Table 1 (see footnote 1) provides the vertical position and group of samples belonging to profiles for whole-rock analyses published by Galerne et al. (2008).

The minerals were analyzed for major elements using a CAMECA SX100 electron microprobe (EMP) at the University of Oslo. The instrument was fitted with an integrated energy-dispersive spectrometer and five wavelength-dispersive crystal spectrometers. Accelerating voltage was 15 kV and counting times were 20–30 s. Minerals were analyzed using a beam current of 15 nA and a focused beam with 5 μm diameter. The detection limit of the analysis is on average 0.05 wt% (Supplemental Tables 2 and 3 [see footnotes 2 and 3]).

5.1. Texture

Like the other sills in the GVSC, the Golden Valley Sill is a nonlayered intrusion, but does present some textural variations. Also finer variations may exist (subjected to sampling resolution, see Sec. 3), two types of textural variations from bottom to top in the sill have been consistently observed and identified. Both types exhibit fine-grained texture at the sill margins (within the uppermost and lowermost 5 m; Fig. 4A). The first textural type is homogeneously medium grained over the sill height (0.2–0.8 mm; Fig. 4B, log A). This type of textural log is common, particularly at the Golden Valley Sill east limb at its south and northern ends. In the second type of textural variations, the fine-grained texture at the sill margins shifts through medium-grained zones, to coarser-grained zones near the center of the sill (Fig. 4C, log B). This coarsening is partly made up by the progressive development of clinopyroxene oikocrysts forming a “clinopyroxene-rich domain.” The large pyroxene oikocrysts (0.5–2 mm up to 1 cm in diameter) enclose randomly oriented small plagioclase laths as well as small olivine grains. Many clinopyroxene oikocrysts are zoned. The Fe-enriched rims contain closely spaced exsolution lamellae of pigeonite. Exsolution lamellae of clinopyroxene may also be present in pigeonite. The clinopyroxene oikocrysts are surrounded by large plagioclase laths that increase in size from the sill margins toward the central parts where they form “plagioclase-rich domains” dominated by large euhedral plagioclase and clusters (both can attain 6 mm) and Fe-Ti oxides. Clinopyroxene can also be found in plagioclase-rich domains as small interstitial grains (0.8–1.0 mm) or in groups of several grains (1.0–2.0 mm).

In detail, the second type of texture (log B), with large pyroxene oikocrysts, is often not well-developed (e.g., P14, Golden Valley Sill west limb). The clearest expression of this textural pattern has been observed in the profile P5 from the Golden Valley Sill east limb. The log B–type profile is asymmetrical in the profile P5 with the coarser-grained part halfway to the sill center (Fig. 4C). A similar asymmetrical texture variation has been observed in profile P19 from the Morning Sun South limb.

5.2. Petrography

The main phases in the Golden Valley Sill are plagioclase, clinopyroxene, Fe-Ti oxides, and olivine. Pigeonite is present in small proportions (<2%). Accessory apatite, pyrite, biotite (<0.5%), and very rare fluorine and sphene may be present. The relative proportions of plagioclase:clinopyroxene:Fe-Ti oxides:olivine are roughly 50:40:5:5. The matrix is dominantly made of euhedral plagioclase laths (0.5 mm), larger euhedral plagioclase grains may occur (6 mm). Also found in the matrix are euhedral to subhedral olivine grains (0.4–2.4 mm) and anhedral clinopyroxene. Plagioclase is also found as crystal aggregates (clusters; Fig. 4A) of larger subhedral grains (2.0–3.6 mm) that form a plagioclase-phyric texture. These plagioclase clusters are present all the way from the chilled margins toward the center of the sill and are usually made of zoned crystals. Large plagioclase single laths found in the coarser regions are also usually zoned. Olivine is usually subhedral and is found as single grains and grain clusters, as inclusion in clinopyroxene oikocrysts, and occasionally in plagioclase clusters. Olivine can be partly altered and replaced by minerals such as iddingsite, carbonate, and serpentine.

Figures 5–7,67 show compositional profiles from north to south along the east limb of the Golden Valley Sill (Fig. 5), the west limb (Fig. 6), and the southern part of the Golden Valley Sill together with the Morning Sun south limb (Fig. 7). These figures report the variations of strongly compatible elements (i.e., Mg#, CaO, Ni, and Cr) and incompatible elements (i.e., TiO2, P2O5, Zr, and Y). TiO2 and P2O5 are incompatible in these rocks as mineral phases in which these elements are compatible (i.e., Fe-Ti oxides and apatite) only appear late on liquidus. In Figures 5–7,67 the chemical components are plotted in different scales in order to better compare the trends they form. In order to simplify the comparison between the different profiles the vertical height has been normalized to 100 (Figs. 5–9,6789). The distance of the samples with respect to the bottom contact (D.r.B.) and their conversions in height normalized to 100 (nd), are reported in Tables 1 and 2 and in Supplemental Table 1. True distance from the bottom contacts are given for each profile on the right side of the diagrams (Figs. 5–7,67, 9).

Characterization of profiles into type of shape was based on consistent variations of strongly compatible elements (i.e., Mg#, CaO, Ni, and Cr). Profiles that corresponded to established nomenclature, i.e., I-, D-, and S-shaped have been marked in Figures 5–7,67. Those that did not fit in any established nomenclature are marked as X-shaped profiles (Figs. 5–7). For profiles with few samples (e.g., P9) or large distances between some of the samples (P3, P8, P9, P16, P18, and P20) the low resolution makes the classification questionable. For these profiles the classification problem is acknowledged by a question mark (Figs. 5–7,67). Figure 9 presents the most characteristic I-, D-, and S-shaped profiles identified in the Golden Valley Sill. These will be used as reference profiles for further interpretation.

6.1. I-Shaped Profiles

The profiles P1–P3, P8, and P9 (Fig. 5) are classified as I-shaped. Each of these profiles shows homogeneous compositions. However, the compositions vary somewhat from profile to profile: e.g., Mg# = 49.4–52.6, CaO = 10.3– 10.7 wt%, Ni = 62–69 ppm, Cr = 251–301 ppm, TiO2 = 0.97–1.17 wt%, Zr = 91–109 ppm, and Y = 25–31 ppm.

Profiles at the southern end of the Golden Valley Sill west limb, P17 and P18, are composite. They show a chilled contact between the lowermost part, which has been identified as the Morning Sun, and the upper part which is part of the Golden Valley Sill (Galerne et al., 2008). The upper parts of these profiles, which represent the Golden Valley Sill, are I-shaped (Figs. 7 and 8).

6.2. Complex Profiles

In the Golden Valley Sill D-, S-, and X-shaped profiles comprise P4–P7 and P10–16. The widest variations are found in the S-shaped profile P5 from the central part of the Golden Valley Sill east limb: Mg# = 43.7–56.6, CaO = 9.4–11.0 wt%, TiO2 = 0.72–1.30 wt%, P2O5 = 0.11–0.20 wt%, Ni = 40–67 ppm, Cr = 110–285 ppm, Zr = 79–133 ppm, and Y = 23–36 ppm (Figs. 5 and 9). The S-shape of this profile is also reflected in other chemical parameters, with positive correlations between Mg# and compatible elements and negative correlations between Mg# and incompatible elements.

The profile P14 presents significant variations in concentration along the thickness of the sill in a symmetrical manner forming a “D-shaped” profile on the basis of strongly compatible elements (Mg#, CaO, Ni, and Cr). It varies from lower values at the sill margins (e.g., P14; Figs. 6 and 9: Mg# = 51.9–52.2, Ni = 63–69 ppm) to higher values at the sill center (e.g., P14; Figs. 6 and 9: Mg# = 54.8, Ni = 84 ppm). Incompatible elements show a clear opposite behavior, changing from higher values at the sill margins (e.g., P14; Figs. 6 and 9: P2O5 = 0.16–0.15 wt%, Zr = 96–101 ppm) to lower values at the sill center (e.g., P14; Figs. 6 and 9: P2O5 = 0.13 wt%, Zr = 87 ppm). Profile 13 shows more restricted but similar variations (Figs. 6 and 9).

Another example of a D-shaped profile is P19 in the underlying Morning Sun sill (Figs. 7 and 8) which shows Mg# = 55.8–55.4 at the roof and floor margins, respectively, and Mg# = 57.4 at the sill center. However, this profile is asymmetrical with the most mafic composition well below the center of the sill (Fig. 7). Additionally, a reversal in the composition is observed in the last meters from the roof contact (Mg# = 50.8–56.4). Unlike in a typical D-shaped profile of the Golden Valley Sill, CaO shows a relatively constant concentration between 10.7 and 11.0 wt%. The strongly compatible trace elements (Ni, Cr), however, confirm the D-shaped behavior of the profile, and the incompatible major elements show a clear negative correlation with the strongly compatible elements (P2O5, Y).

The remaining complex profiles are called X-shaped profiles as they do not fit to any particular end-member type of shapes, i.e., I, D, and Sshape. Unlike the end-member type of shapes, some of the X-shaped profiles do not show clear negative correlations between strongly compatible and incompatible elements (e.g., P4, P7, and P10; Figs. 5 and 6). Other X-shaped profiles, however, show positive correlations between strongly compatible elements and Mg#, and negative correlations between Mg# and incompatible elements (e.g., P6, P11, and P15; Figs. 5 and 6).

The profile P15 presents the widest range of variation for the Golden Valley Sill west limb with Mg# = 49.3–54.6, CaO = 10.2–11.1 wt%, Ni = 57–82 ppm, Cr = 247–355 ppm, TiO2 = 0.92–1.10 wt%, P2O5 = 0.14–0.17 wt%, Zr = 82–101 ppm, and Y = 24–28 ppm. The profile P15, sampled on the inclined part of the Golden Valley Sill west limb, is located eastward from the profile P14 location. It presents a shape that could be described as an inverse S shape where compatible and incompatible elements present perfect mirror images of one another. However, the bottom contact of this profile was not confidently identified (the only profile reported with this problem). Thus the profile P15 cannot be further interpreted in terms of significant profile shape to the magmatic differentiation process.

The mineral analyses presented here are aimed at characterizing the chemical variations along the well-defined D-shaped profile P19 in more detail than obtained from whole-rock chemistry. The profile P19 also shows the clearest textural variations defined by the log B type (Fig. 4). We systematically probed plagioclase, clinopyroxene, pigeonite (when present), and olivine in the texturally clinopyroxene-rich and plagioclase-rich domains. The result indicates that: (1) the clinopyroxene-rich domain contains Mg-rich clinopyroxene and olivine and CaO-rich plagioclase (Fig. 10), and (2) the plagioclase-rich domains show wider variations in olivine chemistry and extend to much lower Mg# values (Fo80-Fo30); clinopyroxene is more iron-rich, and plagioclase shows a wider range in CaO content (6–18 wt%; Fig. 10; Supplemental Tables 2 and 3 [see footnotes 2 and 3]).

Variations in mineral chemistry along profile P19 are shown in Figure 11A. In Mg# versus normalized height only olivine and clinopyroxene show significant compositional variation associated with clinopyroxene-rich and plagioclase-rich domains (Figs. 11B and 11C). Pigeonite presents a more scattered composition and does not show any particular trend (Supplemental Tables 2 and 3 [see footnotes 2 and 3]). Olivine in both domains indicates a general decrease in Mg# inward in the sill (Fig. 11B). Olivine is most magnesium-rich in clinopyroxene-rich domains (Fig. 11B) with Mg# varying from 80 at the floor to 51 in the center of the sill. Olivine in a plagioclase-rich domain shows Mg# values from 80 at the floor to 31 in the center of the sill (Fig. 11B). The clinopyroxene forming large oikocrysts is homogeneous with Mg# = 80 along the D-shaped region of the profile (Fig. 11C). Clinopyroxene from the plagioclase-rich domain is more iron-rich with Mg# values similar to those of the whole-rock chemistry (56–58).

The chemistry of plagioclase clusters and plagioclase laths in the two distinct textural domains are represented in Figures 11D–11F. The relatively high anorthite values at the core of the clusters (An60-An65) are regarded as a representative composition of early fractionated plagioclase, before the magma is emplaced and forms the Morning Sun saucer-shaped sill (CaO = 16–17 wt%; Fig. 11D). However, the rims are much less CaO-rich (CaO = 12–13 wt%; Fig. 11D) and have compositions similar to plagioclase laths in the plagioclase-rich domain (Fig. 11E). Finally, the plagioclase found as inclusions in clinopyroxene-rich domains shows a D-shaped profile with respect to CaO, ranging from CaO = 12.5 wt% at the floor and near the upper part of the D-shaped region to 17 wt% approximately in the central part of the D-shaped region (Fig. 11F). These are similar values to those found in the cores of the plagioclase clusters (Fig. 11D).

Figure 12 shows the distribution of the main compositional profiles (Mg# versus normalized height) around the Golden Valley Sill and along the Morning Sun south segment. It shows that: (1) all types of compositional profiles, I-, D-, and S-shaped, except for C-shaped, are present in the single Golden Valley Sill; and (2) differently shaped profiles are systematically distributed along this elliptical sill; I-shaped profiles are more abundant at, and close to, the northern and southern tips of the Golden Valley Sill, whereas D and Sshapes are restricted to the central parts of the western and eastern flanks (Fig. 12).

8.1. I-Shaped Profiles

With the exception of the northern part of the Golden Valley Sill west limb all I-shaped profiles are located at the northern and southern ends of the Golden Valley Sill where the curvature is strongest (Fig. 12). The absence of I-shaped profiles along the northern part of the west limb may be due to the presence of the minor MV Sill, which is a direct continuation of the Golden Valley Sill (Figs. 5 and 12). Although some variations along single profiles do occur (e.g., P1, P2, and P17), most components are constant along the sill height. The number of samples in profiles P8 and P9 are low (four and three samples, respectively) for this part of the sill that is significantly thinner (P9 = 28 m). Thus the low number of samples makes the classification of these profiles as I-shaped uncertain.

8.2. D- and S-Shaped Profiles

Profiles collected along the central parts of the western and eastern limbs in the Golden Valley Sill show considerable compositional variations giving rise to D-, S-, and X-shaped profiles (Fig. 12). D-shaped compositional profiles are found at the center of the Golden Valley Sill west limb, whereas the S-shaped and other complex profiles are found at the center of the Golden Valley Sill east limb. On the east limb the S-shaped profile P5 is located near the highest parts of the inclined sheet, whereas, on the west limb a particularly well-defined D-shaped profile (P14) is located on the flat outer sill. Finally, the Morning Sun south limb has been followed from a position directly below the Golden Valley Sill west limb's southern region (Figs. 7 and 8), down to progressively lower altitudes. In this part, the Morning Sun south limb becomes distinct from the Golden Valley Sill west limb. The westernmost Morning Sun profile is D-shaped (P19; Figs. 7 and 8).

8.3. Complex or X-Shaped Profiles

Complicated types of shapes (X-shaped profiles) occur in four regions of the Golden Valley Sill (Fig. 12), that is between I-shaped and D- or S-shaped profiles along both limbs, including the MV Sill. The widest range of variation in the MV Sill occurs in the profile P11, which has the most complicated shape observed. Yet, this profile also shows a negative correlation between some of the strongly compatible (Mg#, CaO) and incompatible (Zr, Y; Fig. 6) elements. The profile P10 at the very northern end of the Golden Valley Sill west limb presents an interesting pattern with a gradual decrease in strongly compatible elements toward the upper part of the profile (e.g., Mg#, Figs. 6 and 12). Incompatible elements, on the other hand, do not show any particular variations along the sill height (Fig. 6).

Only one explanation has so far been proposed for the origin of sills with I-shaped compositional profiles. It consists of the injection of a phenocryst-poor magma; crystallization and crystal growth predominantly occurs within the solidification front. This will prevent any evolution of the magma because the interstitial liquid will be captured, inhibiting differentiation (Mangan and Marsh, 1992; Marsh, 1996). In contrast, there is a wide range of processes that can explain the formation of D- and S-shaped profiles. Examples range from multiple or prolonged continuous magma influxes (S-shaped: Gorring and Naslund, 1995), a convective flux of refractory components within the crystal-liquid mush of the boundary layer during in situ differentiation or compositional convection (S-shaped: Tait and Jaupart, 1996), and gravitational settling (S-shaped: Frenkel et al., 1989). Soret fractionation (Latypov, 2003a, 2003b; Latypov et al., 2007) was recently proposed to explain the origin of the often observed marginal reversals in layered intrusions. This model was combined with the in situ crystallization in thermal boundary layers of Tait and Jaupart (1996; Latypov et al., 2007). Another process recently proposed to explain the formation of D-shaped profiles and marginal reversal is the post-emplacement melt flow induced by thermal stresses (Aarnes et al., 2008), which is further detailed in Sec. 9.2.1. Finally, gravity-induced settling of crystals present in the magma at the time of the emplacement (Fujii, 1974; Gray and Crain, 1969; Marsh, 1989) and/or new in situ grown minerals (Frenkel et al., 1988, 1989) may result in complex compositional profile shapes.

These processes refer to quite different stages of crystallization of igneous sills: from a liquid state where convection will dominate, to near solidus conditions where mush-related processes will be dominant. Before comparing the Golden Valley Sill profiles with theoretical profiles based on the existing models outlined above, we present a review of the main findings.

9.1. Review of Main Results

Our documentation of compositional profiles in a single saucer-shaped sill shows the following.

(1) A variety of compositional profiles, I-, D- and S-shaped (Fig. 9), have been found at different locations around a single, ∼100-m-thick tholeiitic saucer-shaped sill (Figs. 5–7,67, 12). Additionally, there is a systematic distribution of the various types of profiles around the Golden Valley Sill (Fig. 12). I-shaped compositional profiles are observed at the northern and southern parts of the sill. D and S shapes and more complex compositional profiles are observed in the central regions of the conjugate limbs. An exception to this general pattern is profiles in the area of the minor saucer-shaped sill, the MV Sill, in the northwestern part of the Golden Valley Sill, where complicated compositional profiles break the symmetry in the Golden Valley Sill (Figs. 6 and 12).

(2) There is a consistent negative correlation between strongly compatible and incompatible elements in profiles that show significant variations (i.e., D- and S-shaped profiles and some of the X profiles, Figs. 5–9,6789). These observations are consistent with a fractional crystallization process.

(3) No textural variations were observed along I-shaped profiles from the North and South region of the conjugated limb's tips of the Golden Valley Sill (Fig. 4, log A).

(4) Textural analysis of the D- and S-shaped profiles (P14, P5) in the central region of the Golden Valley Sill limbs, and one profile from the Morning Sun south limb (P19), show clear textural variations (Fig. 4, log B) that correlate with the chemical variations. In these profiles the central part of the D-shaped region exhibits two distinct textural domains. One domain is characterized by circular ophitic clinopyroxene. The second domain surrounds the first one and mostly contains plagioclase and Fe-Ti oxides.

(5) The profile with the clearest textural variation is the D-shaped profile P19 in the Morning Sun sill. EMP analyses of the two different textural domains showed that clinopyroxene-rich domains contain Mg-rich minerals, whereas plagioclase-rich domains contain more Fe-rich minerals (Figs. 11B and 11C). Similar, but less evident, results have been found in the high-Mg# part of the S-shaped profile P5 (Supplemental Table 3 [see footnote 3]).

9.2. Compositional Profiles and Processes in the Golden Valley Sill

All together, these results suggest that the magmatic differentiation processes may differ in different parts of a single cooling sill to the extent that contrasting compositional profiles are produced (I-, D-, and S-shaped, and more complex). A recent in-depth analysis of available mechanisms to explain the formation of the various types of compositional profiles conventionally classified as I-, D-, and S-shaped in sills has been made by Latypov (2003a, 2003b). In these papers the author pointed out the presence of marginal compositional reversals in terms of modal, phase, and cryptic layering representing a mirror image of the large Layered Series (C-shaped in our nomenclature). A number of distinctive features of marginal compositional reversals have been described by Latypov (2003a). All together, it appears that the available processes reviewed in Latypov (2003a, 2003b, and references therein) briefly mentioned above are feasible, under some circumstances. However, they are unlikely to represent the dominant explanation for a single sill yielding differently shaped compositional profiles (I-, D-, and S-shaped), such as the Golden Valley Sill. Furthermore, they are unable to explain the formation of basal and top reversals that form D- and S-shaped profiles.

Petrographic observations from the Golden Valley Sill show that two types of textural logs can be distinguished (Fig. 4). Although log B indicates a coarsening inward, the sill marked by the development of large oikocryst clinopyroxene, the Golden Valley Sill is characterized by relatively homogeneous medium-grained texture over its height (log A). Furthermore, we show that numerous profiles indicate I-shaped compositional profiles associated with homogeneous textural profiles (Fig. 4, log A; Fig. 12). The only available mechanism proposed by Mangan and Marsh (1992) and Marsh (1996) to explain the formation of I-shaped compositional profiles involves the injection of phenocryst-poor magma, thus implying that the Golden Valley Sill was formed through the injection of such phenocryst-poor magma. Flow inducing crystal segregation and concentration of larger phenocrysts in the center of a flow channel has been shown to be particularly inoperative for sills with a thickness of 100 m or more (Barriere, 1976). Thus flow segregation involving a Bagnold effect is unlikely to be a major mechanism in the presently studied sill.

There is no evidence of multiple injections in the Golden Valley Sill. The samples are all homogeneous, showing no abrupt changes in the texture. Furthermore, a structural and geochemical investigation of the whole Golden Valley Sill Complex indicated that each individual saucer-shaped sill resulted from a single impulse of magma, corresponding to an individual geochemical signature amongst the six distinct magma batches geochemically discriminated (Galerne et al., 2008).

Fractional crystallization embraces a wide range of processes that occur in a cooling mass from the earliest to the latest stages of crystallization. It corresponds to any process that prevents a solid and a melt originally at equilibrium, to continuously reequilibrate during physicochemical changes (e.g., cooling). This leads to chemical changes. In detail, fractionation (i.e., segregation) processes may differ (e.g., see review by Latypov [2003b] and references therein). During the early stages of magma crystallization, newly grown crystals (or crystals brought through the feeding conduct) may be segregated by processes such as convective fractionation (e.g., Sparks et al., 1984), or crystal settling (Wager and Brown, 1968; Wager et al., 1960). Thus early formed minerals may collect in the calm part of a convecting magma body or at the cooling margins. During late stages of crystallization the magma body consists of a continuous crystal mush and convection has stopped. In this regime melt/solid separation can only occur through flow of the remaining melt fraction through the porous crystal framework (Aarnes et al., 2008). Below, various magmatic processes of melt-crystal segregation leading to fractional crystallization are examined at the early and late stage of crystallization. The resulting theoretical profiles are compared to the compositional profiles obtained in the Golden Valley Sill.

9.2.1. Compositional Profiles and Processes at Early Stage of Crystallization

Gravitational ordering or segregation is historically the first mechanism postulated to result in fractional crystallization (Bowen, 1915, 1928). It is based on the idea that mineral phases in magma chambers will settle or be buoyant with respect to the ambient melt. This will result in compositional evolution. Profiles expected to form as the result of this process are sketched in Figure 13. The expected profiles will be more enriched in Mg# toward the base of the profile as heavy minerals such as olivine will tend to sink. According to this model, plagioclase would be slightly more buoyant and collects in the upper part of the magma body where the CaO concentration will consequently be highest. This will most likely result in asymmetrical but linear profiles (Fig. 13).

The profile P10 at the north end of the Golden Valley west limb presents an asymmetrical linear profile for strongly compatible elements, particularly Mg#, Ni, and Cr (Fig. 6). Mg# in the profile P10 is higher at the base and lower near the top of the sill. This suggests that early formed olivine may have settled down. Although CaO shows a similar tendency suggesting settling of clinopyroxene, a slight belly shape toward the bottom contact is observed. Sr, in contrast, shows a negative correlation to CaO (table 2 in Galerne et al., 2008). Sr is strongly compatible with plagioclase and its relative depletion in the lowermost part of the profile suggests a possible effect of plagioclase buoyancy. The combined effect of clinopyroxene settling and plagioclase buoyancy potentially results in the shape of the CaO profile. This type of compositional profile is unique in the Golden Valley Sill. Thus gravitational separation cannot be the dominant mechanism for magmatic differentiation in the Golden Valley Sill, but may have occurred in the northern end of the west limb.

An important alternative to settling and gravitational ordering is convective fractionation (Rice, 1981; Sparks et al., 1984). Both are complex processes that explain the C-shaped profiles (in our nomenclature) observed in large magma chambers (e.g., Wager and Brown, 1968). It is based on the consideration that the inward cooling from the roof and bottom of a magma chamber will preferentially remove the phases that have the highest melting points from the melt through solidification. This leaves a melt relatively enriched in the components that go into phases with low melting points at the solidification front (Rice, 1981). No C-shaped profile has been found in the Golden Valley Sill. However, the profile P21 (Figs. 7 and 8) in the Morning Sun sill may be regarded as an asymmetrical C-shaped type. Unfortunately, its sampling resolution was too low to be ascertained as a C-shaped type and was left unclassified. This suggests that even if C-shaped profiles may exist in ∼100-m-thick sills, they are rare. This implies that convective fractionation is unlikely to be among the dominant processes resulting in differentiation in a sill such as the Golden Valley Sill.

Latypov (2003a, b) revived the hypothesis of Soret fractionation to explain the marginal reversals producing D- and S-shaped compositional profiles in basic-ultrabasic sills. Soret fractionation (thermal diffusion) is a process that causes heavy components (e.g., Fe) to migrate toward the colder end of a thermal gradient, and the lighter components (e.g., Mg) to migrate toward the hotter end. Although Soret fractionation is a potential process that can explain D-shaped profiles, the geochemical data of the Golden Valley Sill show strong evidence against it. If heavy elements migrate toward the cooling margins, we would expect the same trend for all elements of differing molecular weight. In contrast, Ca increases toward the center of the sill and Na and Cr are enriched toward the margin, although Ca is heavier than Na (D-shape in Fig. 9). Additionally, Ni and Cr are also enriched in the center even though their molecular weights are in the same range as Fe (Fig. 9).

9.2.2. Compositional Profiles and Processes at Late Stage of Crystallization

Largely motivated by the negative correlation between compatible and incompatible elements observed in D-shaped, S-shaped, and some complex profiles, we suggest the examination of a post-emplacement melt flow and separation from the crystalline mush. This requires that crystallization has proceeded so far that a continuous porous crystal network has formed, from which the remaining melt may flow from one region to another.

In a schematic representation of a binary system with a complete solid solution (Fig. 14A), the removal of a melt fraction at a constant temperature implies a shift from the bulk composition toward higher values of Mg# for the remaining drained region (Fig. 14B). The flow of the iron-rich melt fraction to other parts of the crystalline mush will change the bulk rock composition toward lower Mg# values for the region where the fractionated melt is added (Fig. 14B). Figure 14C sketched the case of a marginal melt flow from Aarnes et al. (2008).

9.2.2.a. Post-emplacement melt flow induced by thermal stresses.

Aarnes et al. (2008) showed through numerical modeling that thermally activated stresses associated with the cooling and crystallization of sills were sufficient driving forces to produce melt fractionation from the crystal-solid network originally of uniform composition throughout. The authors concluded that the thermal stresses induce large under-pressure near the cooling margins, where melt will be sucked in by porous flow. As a result, the margins will be enriched in incompatible elements and Fe, while the center will be relatively depleted in incompatible elements and enriched in Mg, thereby producing a D-shaped profile (Fig. 14C).

9.2.2.b. A lateral source-sink model for post-emplacement melt flow in saucer-shaped sills.

We present here a further step in developing the melt-separation model introduced by Aarnes et al. (2008). Our aim is to develop a model that can explain the systematic geographical distribution of compositional profiles around the Golden Valley Sill. The new model also takes into account the potential effect of the three-dimensional (3-D) geometry of a saucer-shaped sill as compared to a horizontal sill, which is the basis for the model of Aarnes et al. (2008).

The 3-D geometry of a saucer-shaped sill has the potential to allow both vertical and lateral flow, where distinct segments related to the saucer-shaped geometry will act as source and sink. We propose that a porous melt flow may occur in a post-emplacement regime. Cooled from the outside, the inward progress of the top and bottom solidification fronts in a horizontal sill will, with time, create a mid-plane porosity region (Fig. 15A). The porosity profile will have a similar Gaussian distribution across the sill height as temperature, because the crystal-melt proportion (i.e., porosity) is controlled by cooling and crystallization (i.e., higher porosity and temperature in the mid-plane of the sill and lower porosity and temperature near the sill margins, Fig. 15C). Thus any forces acting on the porous mush are likely to initiate melt flow and segregation from its equilibrium solid crystalline mush (Figs. 15B and 15C). This will induce a bulk differentiation through melt segregation from a “source region” which supplies melt to a “sink region” in the sill (Fig. 15D).

Such a system can be represented by a binary solid solution in a T-X diagram (Fig. 15B). This diagram displays the stability field of a given phase (e.g., olivine) at solid (S) and liquid (L, i.e., melt) states and their onset representing the porosity (L+S). Readily, the liquidus represents the upper bound of the solid-liquid mixture (Φ = 1) and the solidus represents the lower bound or transition between the mixture and solid state (Φ = 0). Thus the mass balance at equilibrium for a given isotherm (e.g., T0, Fig. 15B) and known phase proportion (e.g., Φ = 0.5, Fig. 15B) can be written as:

To be considered as a continuous mush (the condition required for a porous flow to occur in the entire sill) the higher porosity will be close to the random packing of crystals to form a mush (crystallinity = 60%), i.e., 0.4 porosity. This mush domain is defined as a comprise between 90% < N < 55%, where N stands for crystallinity (Marsh, 2002). The mush zone is bounded by the capture front that is defined as the boundary between a crystal-free zone and a mush zone (i.e., 55% < N < 25%).

In the presently studied system we consider the stage where both solidification fronts from the top and bottom of the sill meet in its mid-plane. Thus the maximum porosity at the sill center will lie within the range defined by the capture front (0.45 < Φ < 0.75), but close to 0.45 porosity that is the upper bound for the mush zone. Under the assumption that the starting composition over the sill height is homogeneous, a phase rule similar to the above equation (1) can be applied to an entire profile (Fig. 15C). Specifically for each isotherm (from T1 at the sill center to T5 at the sill margin), the corresponding phase proportion is given by the starting porosity profile (Φ ∼0.45 at the sill center, and 0.1 > Φ at the sill margins).

Figure 14 shows that melt which is extracted from the center and supplied to the margins of the sill will generate increasing Mg# in the melt-drained region and decreasing Mg# at the melt-supplied region. The source-sink model implies that melt extraction will take place from floor to roof in the melt-drained region of the sill, but will, because of the difference in porosity (Fig. 15A), be most effective in the hot central region. The melt fraction available for extraction will decrease as permeability and porosity decrease, toward the margins of the sill. This is represented in Figure 15D, where, for each mass balance at constant temperature T1–5 and corresponding initial porosity Φ, the melt fraction extracted (Φout) is to be subtracted from the original melt fraction available in the source region, or to be added (melt fraction supplied: Φin) to the original melt fraction in the sink region; both melt fractions can be assumed equal to each other (Φout ∼Φin) but smaller than the original porosity Φ. Applying the above equation (1) with the newly obtained melt fraction (ΦNew), the resulting profiles will be D-shaped in the drained region and C-shaped in the sink region (Fig. 15D). The process described above assumes a dominance of mechanical processes over diffusive processes. By this, we imply no significant change in the temperature along profiles during the flow event.

All the profiles sampled along the Golden Valley Sill are from either the transgressive sheets or the flat outer sill. Hence we do not have any profiles in the flat inner sill. The above process of post-emplacement melt flow implies the presence of a sink region in the saucer-shaped sill where we expect to observe C-shaped profiles.

As none of the sampled region of the Golden Valley Sill yield C-shaped profiles the only possible sink region in the saucer-shaped sill is the nonsampled inner sill. Thus for the above model to be correct we need to explain a melt flow counteracting gravity, as the melt density should be smaller than the solid density thereby being relatively buoyant.

A possible reason for a melt flow counteracting gravity is a pressure gradient triggering fluid flow. Specifically, a large enough downhill pressure gradient with a positive pressure in the high part and a negative pressure in the low part of a saucer-shaped sill (outer sill and flat inner sill, respectively) may result in melt expulsion from the high part and flow downward to low parts of the saucer-shaped sill.

9.3. Model Description for Lateral Source-Sink Model: The Case of a Saucer-Shaped Sill

During the emplacement stage of a saucer-shaped sill, the melt is actively supplied by the feeding channel. Minerals can be brought through the feeding channel and newly nucleated crystals may grow inside the saucer-shaped sill. The hydrofracturing process associated with the formation of the saucer-shaped sill (Galland et al., 2009) potentially combined with early related magmatic processes (e.g., unstable convection cells, crystal settling; Fig. 16A) eventually explains the even distribution of plagioclase clusters over the whole profile sections all around the saucer-shaped sill (Fig. 16A, 1a). Eventually, in the thicker part of the saucer-shaped sill, settling of crystals behaving as effective dense plumes may occur (Brandeis and Jaupart, 1986; sketched inset in Fig. 16A, 1a).

The post-emplacement regime refers to the stage where the saucer-shaped sill has reached its final size and shape and is no longer supplied by melt through the feeding channel. During this stage, the sill progressively crystallizes. As the solidification front progresses inward in the sill (Fig. 16B) a porosity gradient will arise with the maximum porosity in the mid-plane region of the saucer-shaped sill (Figs. 15A and 16B). The growth and nucleation rate of plagioclase will be controlled by the cooling rate of the sill (Fig. 16B). Thus the solidification front will grow inward, using the preexisting plagioclase clusters as initial building blocks. As defined by Jerram et al. (2003), the initial building blocks consist of a mixed population of clusters (in our case plagioclase clusters). The resulting crystal framework or mush will control and precondition later textural responses (Cashman, 1993).

When all conditions required for the flow are met, an adiabatic process (see Sec. 9.2.2.b) of porous melt flow-induced stresses on the mush zone triggers a single flow event. We propose that these conditions are met at the moment when both solidification fronts meet in the mid-plane of the sill. At this stage the overall continuous saturated porous mush is relatively buoyant over the entire saucer-shaped sill. Specifically the buoyancy of the porous mush integrated over the height between the higher outer sill and lower inner sill will dynamically generate an overpressure ahead of the porosity mush located in the high region of the saucer-shaped sill, i.e., the outer sill. This overpressure is caused by the buoyancy of the continuous mush zone in the mid-plane of the sill (Fig. 17A). Due to the motion of the crystalline mush, an associated dynamic underpressure region is expected to be located at the transition between the inclined-sheet and the inner sill (Fig. 17A). Thus the melt is expected to flow downward following the pressure gradient (Fig. 17B). The dynamically overpressured region will expel the saturated melt fraction that will flow against gravity down to the dynamically underpressurized region where it will be sucked in (Fig. 17B). Finally, because a limited amount of remaining melt fraction will be available, the melt flow will stop as soon as it has been expelled from the overpressure region. Thus the region that has been affected by melt expulsion and flow dries up and is expected to cool down faster. The feasibility of this process will be tested in the near future through a numerical code using the fully coupled set of equations developed by Tantserev et al. (2009) solving the scenario of differentiation due to post-emplacement melt flow.

We suggest that the plagioclase domain identified in our study represents the fossil trace of such a crystalline framework. This region that sustains the compressive stresses applied on the porous network was previously believed to be difficult to identify at the scale of a thin section (Philpotts and Carroll, 1996). Philpotts and Carroll (1996), however, showed that in plagioclase-rich tholeiitic basalt with only 30% crystals, a relatively strong network of crystals exists. They showed that the compressive strength of the mush increases dramatically from 336 to 4333 Pa as the percentage of crystal increases only from 33% to 37%. Like Marsh (2002), we emphasize that this result is relevant to plagioclase-rich basalts. It is, however, closely tied to the specific phase equilibria of the considered magma and the crystallinity will vary somewhat with magma type.

Thus the presently described textural pattern is likely to result from different local states of pressures at a common temperature for the studied magma. In the present model, this texture pattern is interpreted as the result of the forced draining of the melt associated with the melt fractionation process that reveals the trace of the saturated porous mush zone. The proposed mechanism is expected to be “instantaneous” as compared to the period of time it takes for a 100-m-thick sill to be fully crystallized.

The symmetrical D-shaped profiles from the Golden Valley Sill west limb P14 also show slight coarsening in the central region of the sill associated with increasing growth of interstitial clinopyroxene. These grains, however, have not developed to the extent of the circular oikocrysts observed in the D-shaped profile from the Morning Sun sill P19 and the S-shaped P5 profile from the Golden Valley Sill in the central region of the east limb. Thus mineral chemistry differences between the two textural domains could not be tested in the profile P14. Nevertheless, published analyses from Aarnes et al. (2008) from the same profile P14 indicate a homogeneous increase in Mg# in olivine composition at the center of the sill. Hence, mineral chemistry as well as whole-rock geochemistry show symmetrical profiles and support a process of porous melt flow as a mechanism for differentiation in this region of the Golden Valley Sill.

The mineral chemistry in the true D-shaped part of the profile P19 from the Morning Sun sill suggests that a more complex process has occurred. Although whole-rock chemistry and mineral compositional profiles from both textural domains are symmetrical, the olivine composition does not follow this general trend (Figs. 11 and 18). The olivine composition in the two distinct textural domains shows a gradual decrease in Mg# from the bottom contact to the upper part of the D-shaped profile (upper boundary at 70 m height [nd]). Because olivine has a higher liquidus temperature than clinopyroxene and plagioclase, the olivine profile may be regarded as representing earlier magmatic processes than the post-emplacement melt flow event. The asymmetry eventually suggests that early fractionated olivine was settling from the upper region of the sill (Fig. 16A, 1a). This may explain the decrease in Mg# in the upper margin of the sill, observed, for example, in the upper part of profile P19. On the other hand, the symmetrical profile shown by the bulk, plagioclase, and clinopyroxene compositional profile are consistent with a post-emplacement related melt flow through a continuous crystalline mush.

Finally, plagioclase clusters appear to be distributed over the whole-thickness of the examined profiles, including P19. The core composition in plagioclase clusters is significantly different from their rim composition. As shown in Figure 11, CaO content of the rims is identical to the plagioclase laths (Figs. 11D and 11E). It suggests that the plagioclase clusters were brought through the feeding conduct and were homogeneously distributed across the sill. During the in situ post-emplacement crystallization of the sill, plagioclase clusters eventually grew rims of identical composition to the plagioclase laths formed in situ.

Our model suggests that the differently shaped profiles and their distribution around the Golden Valley Sill results from varying magmatic differentiation processes at different stages of the magmatic evolution of the saucer-shaped sill. Similarly to Mangan and Marsh (1992) and Marsh (1996), the I-shaped profiles are interpreted to be the result of the early stage of crystallization associated with crystal growth predominantly occurring within the solidification front. During the early stage of crystallization isolated gravitational ordering in the calm part of the cooling magma at fluidlike stages may develop, possibly explaining the X-shaped profile P10.

The D-shaped profiles are interpreted as the results of a post-emplacement related melt flow. The Golden Valley Sill west limb is more developed than its conjugate neighbor. It is formed by a continuous inclined-sheet to outer sill that is characterized by two D-shaped profiles (e.g., P13 and P14). The P15 profile located at the lower part of the inclined-sheet of the sill demonstrates a complicated pattern (Fig. 6). Unfortunately, there are no exposed parts of the Golden Valley Sill inner sill that allow profile sampling and would complete all the profile shapes continuation to support our theoretical model. However, our model predicts that C-shaped profiles may be found close to the transition between inclined-sheet and inner sill. This may be confirmed in a future study of the Morning Sun south limb, analyzing in greater detail the vertical variations in the texture and chemistry. The profile P21 indeed suggests an asymmetrical C-shaped type of profile. Furthermore, the location of this profile in respect to the saucer geometry is at about the transition between the climbing-sheet and the inner sill.

The X-shaped and S-shaped profiles show complexities in processes, possibly resulting from the combined effect of processes including settling and post-emplacement melt flow at different stages. The regional preservation of the differently shaped profiles related to early magmatic processes indicates that different degrees of post-emplacement melt flow occurred within the sill, with the most pronounced effect in the central region of the conjugated limbs of the Golden Valley Sill.

Our systematic analysis of the Golden Valley Sill Complex along profiles indicates the following. (1) All types of compositional profiles described in mafic sills (I-, D-, S-, and possibly also C-shaped) may form within a single sill. We also show that a systematic space distribution of compositional, textural, and petrological variations along profiles laterally exists around an example of a saucer-shaped sill, the Golden Valley Sill. (2) The formation of I-, D-, and S-shaped profiles in the GVSC results from a shift in types of fractionation processes from the early to the latest stages of cooling and crystallization. (3) The early stage of emplacement and crystallization of phenocryst-poor magma induced compositional homogenization that is later preserved during solidification in the post-emplacement stage and reflected in I-shaped profiles. (4) We propose that the D- and S-shaped profiles were generated in a post-emplacement porous melt flow regime. This flow was driven by a buoyancy-related over- and under-pressure in a crystalline porous mush. For a saucer-shaped sill the buoyancy of the porous mush zone located in the mid-plane of the sill triggers melt expulsion and flow from the overpressured outer-sill region, down to the underpressured region located near the transition between the inner sill and inclined-sheet.

The authors acknowledge the careful reviews of Julian S. Marsh and Rais M. Latypov who improved the original paper with their suggestions. We thank Yuri Podladchikov for suggestions and constructive discussions, and Timm John and Sergei Medvedev for their suggestions and criticism of earlier versions of this paper. Muriel Erambert and John Nicholas Walsh are acknowledged for help with the analytical work. Thanks are due to the Norwegian Research Council (NFR) for financial support through the project 159824/V30 “Emplacement mechanisms and magma flow in sheet intrusions in sedimentary basins,” including a doctoral fellowship to Christophe Galerne. The work has also been supported by a Centre of Excellence grant from the Norwegian Research Council to P.G.P.

1Supplemental Table 1 is a Microsoft Word document. If you are viewing the PDF of this paper or reading it offline, please visit or the full-text article on to view Supplemental Table 1.
2Supplemental Table 2 is a Microsoft Excel spreadsheet. If you are viewing the PDF of this paper or reading it offline, please visit or the full-text article on .org to view Supplemental Table 2.
3Supplemental Table 3 is a Microsoft Excel spreadsheet. If you are viewing the PDF of this paper or reading it offline, please visit or the full-text article on .org to view Supplemental Table 3.