A whole-rock data set for the Skaergaard intrusion, East Greenland

We report a compilation of new and published whole-rock major and trace element analyses for 646 samples of the Skaergaard intrusion, East Greenland. The samples were collected in 14 stratigraphic profiles either from accessible and well-exposed surface areas or from drill core

To examine the petrology and ore bodies of the Skaergaard Intrusion, East Greenland, we have collected hundreds of samples during six field expeditions between 1993 and 2017. In addition, we have collected samples from drill core material housed at the Natural History Museum of Denmark www.geusbulletin.org (University of Copenhagen) and the Harker Collection of the Sedgwick Museum (University of Cambridge). Here, we report a compilation of 646 whole-rock analyses for these samples attached as one Excel spreadsheet (Supplementary Data File 1). All samples were analysed by X-ray fluorescence (XRF); most of these data (n = 409) were published in Tegner (1997), Tegner et al. (2009), Salmonsen & Tegner (2013), Holness et al. (2015Holness et al. ( , 2022 and Thy et al. (in press). The remaining analyses (n = 237) are reported here for the first time apart from P 2 O 5 data (n = 167), which were reported in Holness et al. (2017). A subset of 271 samples were analysed by inductively coupled plasma mass spectrometry (ICP-MS). About half of these (n = 130) were published in Tegner et al. (2009) and Thy et al. (in press); the remaining analyses (n = 141) are reported here for the first time. The analysed samples mainly represent mafic cumulate rocks (n = 623) but also include gabbropegmatite and granophyric pods and layers (n = 23). Details of subsets of the present bulkrock data set have been described and discussed in a number of publications (e.g. Tegner 1997;Tegner et al. 2009;Thy et al. 2009, in press;Tegner & Cawthorn 2010;McKenzie 2011;Salmonsen & Tegner 2013;Namur et al. 2013Namur et al. , 2014Holness et al. 2015Holness et al. , 2017Holness et al. , 2022Nielsen et al. 2015;Keays & Tegner 2016;Pedersen et al. 2021).
The samples were collected in 14 stratigraphic profiles as shown on the geological map ( Fig. 1), in a schematic cross-section (Fig. 2), listed in Table 1 (Fig. 3) shows an example of layered rocks (Layered Series). Further selected outcrops are shown in Fig. S1 (Supplementary Data File 2). Previous work has shown that the intrusion represents the result of prolonged, uninterrupted differentiation of a tholeiitic magma (Wager & Deer 1939;Wager & Brown 1968;Naslund 1984;Hoover 1989;McBirney 1996;Irvine et al. 1998). Palladium and gold mineralisations have been identified in the LS Nielsen et al. 2015). A key point is that the LS, MBS and UBS appear to represent continuous and synchronous crystallisation on the floor, margins and roof and SH the products of the most evolved, last drops of magma in the interior of the intrusion. These rocks, therefore, allow evaluation of the processes resulting in igneous differentiation and ore formation in opposite positions relative to gravity (e.g. McBirney 1995).

Sample profiles
The stratigraphic profiles (n = 14) are summarised in Table 1 and illustrated in Figs 1 and 2. The profiles cover most regions of the intrusion and were collected either from accessible and well exposed surface areas or from drill core. The details of the sample profiles are described in Supplementary Data File 2.

Sampling strategy
The sampling was directed to obtain mainly average rock compositions in systematic stratigraphic sections. Additional samples of outcrop features such as gabbropegmatites, subzone boundaries and layered structures were also included, for example, the 'wavy pyroxene rocks' and colloform banding of the MBS (Humphreys & Holness 2010;Namur et al. 2013). The sample positions were recorded by GPS and altimeter readings (Supplementary Data File 1). For LS and UBS, the stratigraphic thicknesses were calculated relative to the local strike and dip of layering as described previously Salmonsen & Tegner 2013) and are listed in Supplementary Data File 1. Within the limitations of outcrops, we aimed to sample at regular stratigraphic intervals. For LS and UBS, the average stratigraphic interval was 12 ± 13 m (1 s.d.). For MBS, the lateral distance from the contact is recorded and the average spacing between samples was 18 ± 6 m (1 s.d.; Holness et al. 2022).

Calculation of fraction of melt remaining (F)
The box-like appearance of the intrusion with onionring distribution of zones and subzones ( Fig. 2) implies that stratigraphic thickness, and mass proportions are not proportional (Nielsen 2004). Based on mass proportions estimated for each subzone in the floor, wall and roof series (Nielsen 2004), the mass fraction of melt remaining in the chamber, F, can be estimated for subzone boundaries. For the 'reference profile' of LS, a second-order polynomial was fitted to subzone boundaries to relate F to stratigraphic height (H) and given below as equation 1 ): The F values for the 90-10 and 90-23 drill core samples were also estimated using equation 1 and tied to the 'reference profile' at the H of the UZa/b boundary www.geusbulletin.org  [1] [2] [1] [1] [

Analytical methods
All samples were collected, prepared and analysed in the same way. From surface outcrops, we collected samples weighing 1-4 kg and avoiding alteration veins. The samples were trimmed for surface weathering by sawing, and an aliquot (100-400 g) was crushed to small aggregates (<2 cm) in a hydraulic steel press. This was followed by splitting, pre-contamination of a corundum shatterbox, cleaning and finally powdering of c. 30 g. The drill core material was prepared in the same way with one exception. The powders of drill core 90-22 (n = 51) were prepared using a steel jaw crusher and a tungsten-carbide shatterbox. All samples were prepared at Aarhus University as described in Tegner et al. (2009).

www.geusbulletin.org
The major and trace element data were obtained by a combination of XRF at Aarhus University and ICP-MS at the University of California, Davis and AcmeLabs as described in Tegner et al. (2009), Thy et al. (in press) and Tegner et al. (2019), respectively. Concentrations of FeO were determined by titration with potassium dichromate. The mass lost on ignition (LOI) was determined by heating the powder in air in a muffle furnace at 950°C for 3 h. The values obtained for certified reference materials, BHVO-1 and BIR-1, are reported in Supplementary Data File 3. XRF analyses of BHVO-1 and BIR-1 (n = 53-63) demonstrate that the relative variation of repeated analyses is less than 4.5% (1 s.d./average value) for most major element oxides. However, in BIR-1, the relative variation is higher (14%) for K 2 O, which has a relatively low concentration (0.027 wt%; Jochum et al. 2016). For the trace elements measured by XRF, the relative variation of the transition metals (V, Cr, Ni, Cu, Zn) and Sr are within 5%. In BHVO-1, the relative variation is moderate for Rb, Y, Zr, Nb and Ba (4-12%) and higher for Ce (20%). In BIR-1, which is depleted in these elements relative to BHVO-1 (Jochum et al. 2016), the relative variation of repeat analyses is somewhat higher for Rb, Y, Zr and Nb and Pb (7-28%) and much higher for Ba and Ce. The accuracy or relative deviation from the preferred values for BHVO-1 is within 11% for all oxides and trace elements. Similar values were obtained for the accuracy of BIR-1, except for Rb, Nb and Ce, which have sub-ppm preferred values.
In conclusion, the XRF data can generally be viewed as accurate down to a few ppm. Repeat ICP-MS analyses of standards at AcmeLab (n = 14) and University of California (n = 6) deviate less than 7 and 18%, respectively, from the preferred values for all trace elements reported in this study (Supplementary Data File 3).

Data description and main features
The bulk compositions of cumulate rocks, such as those reported here, represent a mix of accumulated liquidus crystals (cumulus) and interstitial material (intercumulus) derived from crystallisation of interstitial melt (Wager et al. 1960;Irvine 1982). The present data set thus tracks changes in the compositions Example compositional data available in the data set. a: Whole-rock FeO total versus SiO 2 (wt%) for the Skaergaard intrusion. One outlier at 17 wt% SiO 2 and 50 wt% FeO is not shown. b, c: Stratigraphic variations in whole-rock FeO total and Rb contents. Data from Supplementary Data File 1. Importantly, the bulk-rock compositions do not directly represent liquid compositions. The data set can therefore be used to evaluate igneous processes during crystallisation and ore formation. Figure 4a, for example, shows that bulk-rock FeO total and SiO 2 vary considerably and display a negative correlation. These two oxides also show systematic variations between zones (HZ, LZ, MZ, UZ). The compositions generally overlap between LS, MBS and UBS rocks although the most FeO-rich rocks occur in LS. In the UZ equivalents, the UBS rocks are enriched in SiO 2 relative to LS and MBS rocks. Not surprisingly, the highest SiO 2 and the lowest FeO values are seen for granophyres sensu lato. Figure  4 also shows two examples of stratigraphic variations plotted against the calculated fraction of melt remaining (F). In Fig. 4b, FeO generally increases from LZ to UZ equivalents and displays a marked increase across the LZb/LZc boundary, reflecting accumulation of magnetite and ilmenite. In the lower and middle part of the stratigraphy (HZ-UZa), the FeO contents are comparable in the floor, wall and roof rocks. However, in UZb' and UZc' of the roof (UBS), FeO is markedly lower compared to LS and MBS rocks. Figure 4c shows the stratigraphic trends of the incompatible element Rb. In the lower and middle parts (HZ-UZa), the trends are relatively flat and display comparable values in rocks from LS and MBS, while higher values are found in UBS rocks. Closer to SH (UZb and UZc equivalents), Rb increases exponentially in the cumulate rocks and shows the highest values in the melanogranophyres. The compiled whole-rock data set can, for example, be used to constrain processes of igneous differentiation and ore formation.