The 1.87–1.84 Ga Black Hills dike swarm of the Kalahari craton (South Africa) is coeval with several regional magmatic provinces used here to resolve the craton's position during Columbia assembly. We report a new 1850 ± 4 Ma (U-Pb isotope dilution–thermal ionization mass spectrometry [ID-TIMS] on baddeleyite) crystallization age for one dike and new paleomagnetic data for 34 dikes of which 8 have precise U-Pb ages. Results are constrained by positive baked-contact and reversal tests, which combined with existing data produce a 1.87–1.84 Ga mean pole from 63 individual dikes. By integrating paleomagnetic and geochronological data sets, we calculate poles for three magmatic episodes and produce a magnetostratigraphic record. At 1.88 Ga, the Kalahari craton is reconstructed next to the Superior craton so that their ca. 2.0 Ga poles align. As such, magmatism forms part of a radiating pattern with the coeval ca. 1.88 Ga Circum-Superior large igneous province.

The Paleoproterozoic to Mesoproterozoic supercontinent (referred to as Nuna, Hudsonland, or Columbia) formed by assembly of Archean cratons starting at 1.9 Ga and was fully amalgamated as late as 1.65–1.58 Ga (Meert, 2012; Pisarevsky et al., 2014; Pourteau et al., 2018). The position of the Kalahari craton (i.e., here the pre–1.2 Ga conjoined Kaapvaal and Zimbabwe cratons, or the proto-Kalahari craton, sensu stricto; see de Kock et al., 2021) in Columbia is obscured by limited paleomagnetic data. It does not appear in some reconstructions (e.g., Evans and Mitchell, 2011; Pehrsson et al., 2016), but in other reconstructions it is placed at Columbia's periphery (e.g., Zhao et al., 2003) or is regarded as a “lone craton” (Pisarevsky et al., 2014). A craton's position in Columbia is typically evaluated through 1.8–1.3 Ga apparent polar wander path comparison (Evans and Mitchell, 2011). For the Kalahari craton (simply Kalahari hereafter), post–1.8 Ga paleomagnetic data remain sparse (de Kock et al., 2021) and its tenure in Columbia remains unconstrained. Paleomagnetic data from 1.89 to 1.83 Ga magmatic provinces, however, can provide constraints on Kalahari's role in Columbia's assembly.

The current 1.89–1.83 Ga paleomagnetic record of Kalahari can be enhanced by studying its magmatic provinces (Fig. 1). The 1.87–1.84 Ga northeast- to north-northeast–trending mafic Black Hills dike swarm (BHDS; South Africa) is one such province (Olsson et al., 2016; Wabo et al., 2019). We present new paleomagnetic data from 36 BHDS dikes together with existing data from the 1.89–1.87 Ga Mashonaland sill province (Söderlund et al., 2010; Hanson et al., 2011) and the 1.88–1.87 Ga post-Waterberg sill province (Hanson et al., 2004). The BHDS is spatially, geochemically, and temporally associated with <1.83 Ga Soutpansberg Basin magmatism (Geng et al., 2014; Olsson et al., 2016) and the ca. 1.8 Ga Mazowe dike swarm of Zimbabwe (Fig. 1B) (Hanson et al., 2011). Paleomagnetic data from the Mashonaland and post-Waterberg sill provinces differ significantly and require large tectonic displacement between the Kaapvaal and Zimbabwe cratons (Hanson et al., 2011). Currently there is no geological support for such displacement. Another explanation may be rapid true polar wander (TPW) (Mitchell et al., 2010; Antonio et al., 2017). Unfortunately, few precisely dated Mashonaland intrusions also have paleomagnetic constraints (Table S3 in the Supplemental Material1). Furthermore, the BHDS was emplaced in the immediate aftermath of the proposed TPW. This limits careful evaluation of the discrepancy. Our data, however, provide a high-resolution paleomagnetic record to test Kalahari's 1.87–1.84 Ga paleogeography.

The BHDS is a >300-km-wide swarm of mainly northeast- to north-northeast–trending mafic dikes intruding Archean basement, the 2.68–2.06 Ga Transvaal Supergroup, and the 2.06–2.05 Ga Bushveld and Phalaborwa Complexes (Olsson et al., 2016; Fig. 1). Paleoproterozoic magnetizations that pass reversal and baked-contact tests were reported for the BHDS (Letts et al., 2005, 2011; Lubnina et al., 2010) and are confirmed by combined U-Pb geochronology, geochemistry, and paleomagnetic data sets (Olsson et al., 2016; Wabo et al., 2019). Baddeleyite U-Pb crystallization ages of 12 dikes range between ca. 1.87 and ca. 1.84 Ga (Olsson et al., 2016; Wabo et al., 2019). However, only two dated dikes were also studied paleomagnetically (Lubnina et al., 2010; Wabo et al., 2019). Dike ages do not discriminate between trend or location but subdivide the swarm into an older, 1.88–1.86 Ga, more-primitive group and a younger, 1.85–1.84 Ga, more chemically enriched group (Olsson et al., 2016). The combined BHDS, Mashonaland, and Post-Waterberg sill provinces, and Soutpansberg Basin magmatism (Klausen et al., 2010; Lubnina et al., 2010; Olsson et al., 2016) form an ~50 m.y. record.


Baddeleyite grains extracted from an 8–10-m-thick north-northeast–trending dike (dike MAD01 in Fig. 1C) were divided into three fractions and dated by isotope dilution–thermal ionization mass spectrometry (ID-TIMS) at the Department of Geosciences, Swedish Museum of Natural History in Stockholm. The full methodology and data, including geochemistry (Table S2), are provided in the Supplemental Material. Free regression yields upper and lower intercepts at 1850 ± 4 Ma and 200 ± 250 Ma (mean squared weighted deviation [MSWD] = 1.13), respectively (Table S1; Fig. S1). The upper intercept is interpreted as the dike's crystallization age.


We sampled 61 dikes for paleomagnetic study using standard methods (see the Supplemental Material; Table S3). Directly dated dikes were explicitly targeted. Each paleomagnetic site sampled a distinct dike and corresponds to a unique cooling unit. High-temperature magnetizations identified in 36 dikes are interpreted as Paleoproterozoic (Fig. 1; Fig. S2). The remaining 25 dikes had heterogeneous demagnetization behavior and are not discussed here (for more detail, see the Supplemental Material). High-temperature remanence components are grouped as northwest (downward) or southeast (upward), showing moderate to steep inclinations. Site means with radii of 95% confidence (α95) >16° were excluded (i.e., 2 of our 36 dikes). The northwest and southeast groups share common precision and are ~180° apart, illustrating a class-C reversal test (Fig. S6). Sampling thus spans one or more reversals of the geomagnetic field. Two dated ca. 1.85 Ga dikes (Olsson et al., 2016), one north trending and one northeast trending (dikes CDL and BJL in Fig. 1C), have positive baked-contact tests supporting the primary Paleoproterozoic nature of the magnetization (see the Supplemental Material). Where directly dated, the crystallization age is assumed also to be the timing of remanence acquisition.

The 1.89–1.83 Ga Paleomagnetic Record

Paleomagnetic Data by Magmatic Province

Our data (34 dikes) are combined with published BHDS results (29 dikes; Table S3) to define two polarities (Fig. 2A; see the Supplemental Material) and a mean paleopole for the BHDS at 15.3°N, 14.9°E and A95 (radii of 95% confidence) = 5.6° (Table S4; quality, Q, = 7; after Van der Voo, 1990). This pole represents ~30 m.y. and is indistinguishable from the 1.88–1.87 Ga Post-Waterberg sill province and <1.83 Ga Soutpansberg Basin poles (Table S4).

Magmatic Episodes and Magnetostratigraphy

Some units from regional magmatic provinces are coeval, and we calculated weighted mean crystallization ages (from U-Pb ID-TIMS baddeleyite dates; Fig. 2B) at 1873 ± 1 Ma (MSWD = 0.62), 1860 ± 2 Ma (MSWD = 0.7), and 1848 ± 2 Ma (MSWD = 0.7). These episodes refine the groupings of Olsson et al. (2016), who named the 1860 Ma episode as “old” and the ca. 1848 Ma episode as “young” BHDS. The crystallization age uncertainty of dike LDH of Wabo et al. (2019) allows it to be a member of either the 1873 Ma or 1860 Ma episode. Dike CDU of this study (sample BCD5-85 of Olsson et al., 2016) can similarly be a member of either the 1860 Ma or 1848 Ma episode. Units of the ca. 1873 Ma episode all have positive inclinations (Fig. 2C). Negatively inclined dike LDH thus likely belongs to the ca. 1860 Ma episode, which is exclusively represented by dikes of negative inclination (Fig. 2C). Besides polarity, the chemical composition provides another discriminator between episodes (Olsson et al., 2016). We note that dikes MAD01 (herein dated at 1850 ± 5 Ma) and LDH (dated at 1867 ± 10 Ma by Wabo et al. [2019]) have MgO contents of 4.9 wt% and 6.0 wt%, respectively (Table S2), which are comparable to those of the “young” and “old” groups, respectively.

Only two paleomagnetically constrained units of the Mashonaland sill province are directly dated. One sill has an age corresponding to the ca. 1873 Ma episode, and it has a positive inclination as expected. An older ca. 1883 Ma sill also has positive inclination. The Mashonaland data thus span at least two reversals, and mixed polarity is otherwise assigned to the province (Fig. 2C).

Mean poles are calculated for episodes based on dated units. Such poles are temporally better defined than magmatic province poles (Fig. 2D; Table S4). Dated BHDS dikes were exclusively of negative inclination during the ca 1860 Ma and ca. 1848 Ma episodes and of positive inclination during the ca. 1873 Ma episode (Fig. 2C). This defines a late Paleoproterozoic magnetostratigraphic record for Kalahari. All undated BHDS dikes that recorded positive inclinations were used to calculate a ca. 1873 Ma episode pole (11.8°N, 10.6°E and A95 = 11.8°; Table S4; Q = 7). Poles at ca. 1860 Ma and ca. 1848 Ma have larger uncertainties but are statistically indistinguishable from the ca. 1873 Ma pole (Table S4).

Kalahari and Columbia's Assembly

The timing of Columbia's assembly and its final configuration is debated (Meert, 2012), but there is consensus that Baltica, Laurentia, and Siberia formed the core around which other continents were accreted (e.g., Evans and Mitchell, 2011; Pehrsson et al., 2016), while the suture between Laurentia and Australia occurred as late as 1.6 Ga (Pourteau et al., 2018). At 1.88 Ga, Baltica and Laurentia were not fully assembled. Sarmatia and Volgo-Uralia collided with Fennoscandia (i.e., the Kola craton, Karelia craton, and Svecofennian crust) at 1.82–1.80 Ga to form Baltica. Overlap of 1.89–1.79 Ga Fennoscandian poles suggests that it was fairly stationary (Klein et al., 2016; Fig. 3D). Laurentian assembly occurred between ca. 1.91 Ga and 1.81 Ga (e.g., Mitchell et al., 2014), and its data are discordant. For Superior poles, some discordance is resolved by straightening the Kapuskasing zone (Evans and Halls, 2010). For the Slave and Superior cratons (Figs. 3B and 3C), TPW may account for the remaining dispersion (Mitchell et al., 2010).

A comparison of absolute polarity can constrain the relative positions of the Superior, Fennoscandia, and Kalahari cratons. In the late Paleoproterozoic, absolute polarity is assigned to Laurentia based on trade-wind orographic patterns across the Slave craton (Hoffman and Grotzinger, 1993). Driscoll and Evans (2016) followed this rationale and assigned positive inclinations from the Superior craton as normal polarity. In this vane, Fennoscandia showed normal polarity before ca. 1.88 Ga (Fig. 2D). Normal polarity is constrained in Kalahari by a single dated Mashonaland sill. Mixed polarities between 1.89 Ga and 1.87 Ga are otherwise recorded by the Mashonaland sill province (Fig. 2C). Between ca. 1.88 Ga and ca. 1.86 Ga, dual magnetic polarities are described from the ca. 1870 ± 9 Ma Svecofennian Keuruu dikes (paleopole 34 in Table S4; Fig. 2D). The reversal(s) recorded by these dikes can be correlated to either the 1873 Ma or pre–1873 Ma reversals of Kalahari. Normal polarity is reported from the ca. 1863 ± 7 Ma Eastern Murmansk sills of the Kola craton (paleopole 35 in Table S4; Fig. 2D). Given current age constraints, this normal polarity cannot be separated from the ca. 1873 Ma normal polarity chron of Kalahari. The ca. 1840–1837 Ma Haukivesi intrusion (Svecofennia) provides an isolated normal polarity data point for Fennoscandia (paleopole 36 in Table S4; Fig. 2D), but there are no data to compare with Kalahari. On Kalahari, several sites from the <1.83 Ga Soutpansberg Basin yield reverse polarity (Fig. 2C). The shared normal polarity before ca. 1.88 Ga and shared record of a reversal at ca. 1.87 Ga suggest that Fennoscandia and Kalahari were in the same hemisphere, although this should be tested as new data become available.

On the Superior craton, the ca. 1.88 Ga Molson dikes record dual polarities as does the ca. 1.87 Ga Haig-Flaherty-Sutton mean (paleopoles 20 and 21 in Table S4; Fig. 2D). Reversals represented by these data may correlate to the post–1873 Ma Kalahari reversal. It is, however, important to note potential disagreement with this model in terms of 1850–1830 Ma normal polarities recorded by the ca. 1850 Ma but poorly dated Sudbury irruptive and the ca. 1838 Ma Boot-Phantom pluton (Hood, 1961; Symons and McKay, 1999). Both these results were, however, excluded from quality-filtered Laurentian data (Swanson-Hysell, 2021).

At 1.88 Ga, Kalahari restores to 45°–59° paleolatitude, Fennoscandia to 11°–28°, and the Superior craton to 32°–58° (Fig. 3E). The Superior-Nain block can be placed in a position northeast of Fennoscandia preceding the Northern Europe–North America (NENA) configuration (Klein et al., 2016) by overlapping the 1.88 Ga Molson dikes pole with 1.89–1.84 Ga Fennoscandian poles (Figs. 3B and 3D). The 1.89–1.83 Ga Kalahari poles are overlapped with the 1.88 Ga Molson dikes pole. Kalahari is placed to the west of the reconstructed Superior craton adjacent to the Penokean orogen (Fig. 3E). The modern southwestern margin of the Superior craton is occupied by the 1.77–1.60 Ga Central Plains, Yavapai, and Mazatzal orogens (Whitmeyer and Karlstrom, 2007). Placing Kalahari to the east of the Superior craton is permissible in paleolatitude, but the area is occupied by the Manikewan Ocean (Fig. 3E). Restoration of Kalahari to the west of the Superior craton, in contrast to “lone-craton” configurations, aligns the BHDS and the Mazowe dike swarm into a radiating pattern around the ca. 1.88 Ga Circum-Superior large igneous province magmatic center (Minifie et al., 2013). This event thus likely included the intraplate magmatism of Kalahari as well as that of Fennoscandia (Fig. 4). In this position, overlap is also achieved between the ca. 2.0 Ga Kalahari and Superior craton poles (Fig. 3E). This suggests a longer-lived link, and it is interesting to note that the Kaapvaal craton at ca. 2.43 Ga was similarly reconstructed (Gumsley et al., 2017). If this is correct, it implies that a loop in apparent polar wander (defined by paleopoles 6 and 7 in Table S4; Fig. 3A) has gone unrecognized in the Superior craton. Furthermore, in our model, there is alignment between the Kapuskasing zone of the Superior craton and the Thabazimbi-Murchison lineament of Kalahari, as well as possible continuation between the Limpopo and Penokean orogens (Fig. 4). On the opposite side of the Manikewan Ocean, the Slave-Rae craton is reconstructed at low paleolatitude using the ca. 1.88 Ga Ghost dike swarm and mean Kahochella–Peacock Hills poles (paleopoles 23 and 25 in Table S4; Fig. 3C). The Ghost dike swarm pole is considered more reliable than similarly aged poles from the Slave craton (Swanson-Hysell, 2021).

After ca. 1.88 Ga, Kalahari and Fennoscandia remained fairly fixed as formation of the Russian belt concluded the assembly of Baltica. At the same time, the Superior craton is envisaged to have rotated clockwise into a classic NENA configuration, leading to possible extension between it and Kalahari and the ultimate closure of the Manikewan Ocean. Unfortunately, a 1.83–1.40 Ga lacuna in Kalahari's paleomagnetic data prevent evaluation of its position throughout the existence of Columbia, but our 1.88 Ga reconstruction does suggest a peripheral position for the craton early during Columbia's assembly.

We thank D.A.D. Evans, R.N. Mitchell, and an anonymous reviewer. C. Djeutchou, M. de Kock and H. Wabo acknowledge support of the South African Department of Science and Innovation–National Research Foundation Centre of Excellence for Integrated Mineral and Energy Resource Analysis (DSI-NRF CIMERA) and a National Research Foundation development grant (RDYR14080787934). A. Gumsley acknowledges support through a Polish National Science Centre grant (POLONEZ grant UMO-2016/23/P/ST10/02423), funded from the European Union's Horizon 2020 research and innovation program under the Marie Skłodowska-Curie Actions CO-FUND-2014 (grant 665778). A. Gumsley also acknowledges the support of the EPOS-PL project (No. POIR.04.02.00-14-A003/16), co-financed by the European Union from funds of the European Regional Development Fund (ERDF) to the laboratory facilities at IG PAS used in the study.

1Supplemental Material. Geochronology, geochemistry, and paleomagnetic datasets. Please visit to access the supplemental material, and contact with any questions.
Gold Open Access: This paper is published under the terms of the CC-BY license.