Timing and tempo of the Great Oxidation Event

Significance

We present U-Pb ages for the extensive Ongeluk large igneous province, a large-scale magmatic event that took place near the equator in the Paleoproterozoic Transvaal basin of southern Africa at ca. 2,426 Ma. This magmatism also dates the oldest Paleoproterozoic global glaciation and the onset of significant atmospheric oxygenation. This result forces a significant reinterpretation of the iconic Transvaal basin stratigraphy and implies that the oxygenation involved several oscillations in oxygen levels across 10⁻⁵ present atmospheric levels before the irreversible oxygenation of the atmosphere. Data also indicate that the Paleoproterozoic glaciations and oxygenation were ushered in by assembly of a large continental mass, extensive magmatism, and continental migration to near-equatorial latitudes, mirroring a similar chain of events in the Neoproterozoic.

Abstract

The first significant buildup in atmospheric oxygen, the Great Oxidation Event (GOE), began in the early Paleoproterozoic in association with global glaciations and continued until the end of the Lomagundi carbon isotope excursion ca. 2,060 Ma. The exact timing of and relationships among these events are debated because of poor age constraints and contradictory stratigraphic correlations. Here, we show that the first Paleoproterozoic global glaciation and the onset of the GOE occurred between ca. 2,460 and 2,426 Ma, ∼100 My earlier than previously estimated, based on an age of 2,426 ± 3 Ma for Ongeluk Formation magmatism from the Kaapvaal Craton of southern Africa. This age helps define a key paleomagnetic pole that positions the Kaapvaal Craton at equatorial latitudes of 11° ± 6° at this time. Furthermore, the rise of atmospheric oxygen was not monotonic, but was instead characterized by oscillations, which together with climatic instabilities may have continued over the next ∼200 My until ≤2,250–2,240 Ma. Ongeluk Formation volcanism at ca. 2,426 Ma was part of a large igneous province (LIP) and represents a waning stage in the emplacement of several temporally discrete LIPs across a large low-latitude continental landmass. These LIPs played critical, albeit complex, roles in the rise of oxygen and in both initiating and terminating global glaciations. This series of events invites comparison with the Neoproterozoic oxygen increase and Sturtian Snowball Earth glaciation, which accompanied emplacement of LIPs across supercontinent Rodinia, also positioned at low latitude.

The early Paleoproterozoic is characterized by dramatic changes in Earth’s atmosphere and oceans, with the transition from anoxic to oxic conditions commonly referred to as the Great Oxidation Event (GOE) (1). It is generally thought that the onset of the GOE was a singular event (2), an assumption rooted in the perceived bistability of atmospheric oxygen (3). However, this inferred bistability in oxygen was challenged through additional modeling (4), allowing for multiple oscillations in atmospheric oxygen during the onset of the GOE. Geological evidence has also established that this transition was broadly coincident with emplacement of numerous large igneous provinces (LIPs) (5) on extensive continental landmasses positioned at low latitudes (6) and glaciations interpreted to reflect global Snowball Earth conditions (7). Models linking these events have been hampered, however, by uncertainties in local and global stratigraphic correlations and age constraints (2). Evidence from the Huronian Supergroup on the Superior Craton in Canada, which hosts three Paleoproterozoic glacial intervals, indicates that the GOE is bracketed in age between ca. 2,460 and 2,308 Ma (8–10). New observations from the critical Transvaal Supergroup in southern Africa indicate that the GOE may have occurred by ca. 2,310 Ma (9) but requires a fourth glaciation at ≤2,250–2,240 Ma (9, 11) before the development of the oldest widely accepted oxygenated paleosols (12) and red-bed sandstones (13). Discrepancies in correlations between these glacially influenced stratigraphic successions and the age of the GOE hinge on a disputed age for the Ongeluk Formation basalts (14–16). By using high-resolution in situ secondary ion mass spectrometry (SIMS) on microbaddeleyite grains coupled with precise isotope dilution thermal ionization mass spectrometry (ID-TIMS) and paleomagnetic studies, we resolve these uncertainties by obtaining accurate and precise ages for the volcanic Ongeluk Formation and related intrusions in South Africa. These ages lead to a more coherent global perspective on the timing and tempo of the GOE and associated global glaciations and LIPs.

Transvaal Supergroup

The Neoarchean to Paleoproterozoic Transvaal Supergroup overlies the Kaapvaal Craton and is preserved in two main subbasins. In the Griqualand West subbasin, the Makganyene Formation consists of a series of glaciomarine diamictites (Fig. 1) (17). Paleomagnetic data for the Ongeluk Formation, which conformably overlies and interfingers with the Makganyene Formation, indicate that these glacial sedimentary rocks were deposited at low latitude, implying a glacial event of global extent (6, 7). The age of the Ongeluk Formation basalts has long been accepted at 2,222 ± 13 Ma based on a whole-rock Pb-Pb isochron date (14) and correlation with the noncontiguous basalts of the ≤2,250–2,240 Ma Hekpoort Formation in the Transvaal subbasin (11, 12, 14, 18), although this correlation has been questioned recently (15, 16, 19). The Ongeluk Formation is overlain by banded iron and manganese deposits of the Hotazel Formation, which in turn, are followed by carbonate rocks of the Mooidraai Formation (Fig. 1). Pb-Pb and U-Pb dating of the Mooidraai Formation carbonates has yielded dates of 2,394 ± 26 and 2,392 ± 23 Ma (15, 20), respectively, in conflict with the 2,222 ± 13 Ma Pb-Pb date on the stratigraphically lower Ongeluk Formation (14). The Hotazel Formation hosts giant Mn deposits with a negative Ce anomaly that are unequivocally interpreted to reflect deposition after the onset of the GOE (7, 21), whereas the Koegas Subgroup underlying the Makganyene Formation contains detrital pyrite and uraninite grains signifying deposition before the GOE (22). In the Transvaal subbasin (Fig. 1), the start of the GOE has been placed in the middle of either the Duitschland Formation or the Rooihoogte Formation by different authors (23, 24), with the age of the upper Duitschland Formation constrained by detrital zircon to ≤2,424 ± 24 Ma (11). A 2,316 ± 7 Ma Re-Os age for diagenetic pyrite (25) and a 2,309 ± 9 Ma U-Pb age for tuff in the lower Timeball Hill Formation (9), which conformably overlies the Rooihoogte Formation, suggest that the GOE began by ca. 2,309 Ma (24). All chronological and redox records for the Transvaal Supergroup are provided in Tables S1 and S2.

Fig. 1.

Stratigraphic synthesis (SI Methods, Stratigraphic Synthesis) of the Transvaal Supergroup as preserved in its two main subbasins: (A) Griqualand West in the southwest and (B) Transvaal in the northeast. Dated samples and results (bold) are shown in stratigraphic context and collectively, unlock the long-held correlation between the basalts of the Ongeluk and Hekpoort formations (12, 14, 18). A selection of previously published ages is schematically shown (Table S1) along with redox indicators and ranges of carbon isotope values in carbonates (Table S2). The redox records within the two subbasins tracks the rhythm of GOE and reflects at least two O₂ oscillations back through 10⁻⁵ PAL after the onset of GOE (bold pink arrows and variable intensity pink background shading). Redox indicators requiring more detailed studies are denoted with question marks. All ages are quoted at 2σ uncertainty.

Table S1.

Compilation of ages for the Transvaal Supergroup and the Huronian Supergroup

Stratigraphic unit	Rock type	Method	Mineral/rock	Age ±2σ error	Refs.
Transvaal Supergroup—Griqualand West subbasin
Mooidraai Fm.	Carbonate	U-Pb TIMS	Whole rock	2,392 ± 23	20
Mooidraai Fm.	Carbonate	Pb-Pb TIMS	Whole rock	2,394 ± 26	21
Ongeluk Fm.	Basalt	Pb-Pb TIMS	Whole rock	2,222 ± 13	14
Ongeluk Fm. (sill)	Dolerite	U-Pb SIMS in situ	Baddeleyite	2,397 ± 22	This study
Ongeluk Fm.	Basalt	U-Pb SIMS in situ	Baddeleyite	2,424 ± 32	This study
Makganyene Fm.	Clastic	U-Pb SHRIMP	Zircon, detrital	2,436 ± 14	19
Asbestos Hills Sg. (Westerberg Sill)	Dolerite	U-Pb TIMS	Baddeleyite	2,428 ± 4 (2,441 ± 6)	28
Asbestos Hills Sg. (Westerberg Sill)	Dolerite	U-Pb TIMS	Baddeleyite	2,426 ± 1	28
Asbestos Hills Sg. (Kuruman Fm.)	Tuff	U-Pb SHRIMP	Zircon	2,460 ± 5	72
Campbellrand Sg. (Gamohaan Fm.)	Tuff	U-Pb TIMS	Zircon	2,521 ± 3	73
Campbellrand Sg. (Nauga Fm.)	Tuff	U-Pb SHRIMP	Zircon	2,552 ± 11	74
Campbellrand Sg. (Nauga Fm.)	Tuff	U-Pb SHRIMP	Zircon	2,588 ± 6	75
Schmidtsdrif Sg. (N-trending Dike)	Dolerite	U-Pb TIMS	Baddeleyite	2,421 ± 3	This study
Transvaal Supergroup—Transvaal subbasin
Hekpoort Fm.	Clastic	U-Pb SHRIMP	Zircon, detrital	2,250–2,240	11
Timeball Hill Fm.	Tuff	U-Pb SHRIMP	Zircon	2,256 ± 6	9
Timeball Hill Fm.	Clastic	U-Pb SHRIMP	Zircon, detrital	2,324 ± 35	11
Timeball Hill Fm.	Tuff	U-Pb SHRIMP	Zircon	2,310 ± 9	9
Timeball Hill Fm.	Shale	Re-Os TIMS	Pyrite	2,316 ± 7	25
Duitschland Fm.	Clastic	U-Pb SHRIMP	Zircon, detrital	2,424 ± 24	11
Malmani Sg. (Penge Fm.)	Tuff	U-Pb SHRIMP	Zircon	2,480 ± 6	76
Archean basement (sheet)	Dolerite	U-Pb TIMS	Baddeleyite	2,423 ± 7	This study
Huronian Supergroup
Gordon Lake Fm.	Tuff	U-Pb SHRIMP	Zircon	2,308 ± 8	9
Mississagi Fm. (sill)	Dolerite	U-Pb TIMS	Zircon	2,215 ± 1	10
Thessalon Fm.	Rhyolite	U-Pb TIMS	Zircon	2,453 ± 6	8
Thessalon Fm.	Rhyolite and granite	U-Pb TIMS	Zircon	2,459 ± 7	10

Fm., formation; Sg., subgroup; SHRIMP, sensitive high-resolution ion microprobe.

Table S2.

Compilation of carbon isotope data and evidence for the redox state of surface environments during deposition of the Transvaal and Huronian Supergroups

Stratigraphic unit	Evidence	Refs.
Griqualand West subbasin
Mooidraai Fm.	No Ce anomaly	21
Mooidraai Fm.	No MIF-S?	23
Mooidraai Fm.	Carbon isotopes	15, 77
Hotazel Fm.	Giant Mn deposit	7
Hotazel Fm.	Ce anomaly	21
Koegas Sg.	Detrital pyrite and uranite	22
Koegas Sg.	No MIF-S?	23
Koegas Sg.	MIF-S?	71
Koegas Sg.	Red beds?	65
Campbellrand Sg.	MIF-S	69
Campbellrand Sg.	Carbon isotopes	77
Transvaal subbasin
Dwaalheuwel Fm.	Red beds	13
Hekpoort Fm.	Oxidized paleosol	12
Timeball Hill Fm.	No MIF-S	70
Timeball Hill Fm.	Hematiitic ooliths and pisoliths	68
Rooihoogte Fm.	MIF-S	24
Rooihoogte Fm.	No MIF-S	24
Duitschland Fm.	MIF-S	23
Duitschland Fm.	No MIF-S	23
Duitschland Fm.	Carbon isotopes	31, 77
Tongwane Fm.	No MIF-S?	23
Tongwane Fm.	Carbon isotopes	31, 77
Huronian Supergroup
Gordon Lake Fm.	Carbon isotopes	78
Gordon Lake Fm.	Red beds	79
Gordon Lake Fm.	Sulfate deposits	80
Lorraine Fm.	Oxidized paleosol	81
Lorraine Fm.	No MIF-S?	82
Gowganda Fm.	No MIF-S?	82
Espanola Fm.	Carbon isotopes	83
Mississagi Fm.	Detrital pyrite and uranite	33
Espanola Fm.	No MIF-S?	82
Pecors Fm.	MIF-S?	82
McKim Fm.	MIF-S?	82
Matinenda Fm.	Reduced paleosol	84
Matinenda Fm.	Detrital pyrite and uranite	85

Fm., formation; Sg., subgroup.

Sampling

To test and resolve some of these critical correlations, we have dated by U-Pb isotopic methods a number of dolerite and basalt samples that are linked geologically and paleomagnetically to the Ongeluk Formation basalts (SI Methods, Sampling, Fig. S1, and Table S3). An N-trending dolerite dike from the Griqualand West subbasin (G02-B) (Fig. 1) (26) as well as an intrusive dolerite sheet from the southeastern Kaapvaal Craton (NL-13c) (Fig. 1) (27) were dated using U-Pb ID-TIMS on baddeleyite. Samples TGS-05 and OLL-2, a coarse-grained, thick basalt flow and a dolerite sill, respectively, from near the base of the Ongeluk Formation basalts (Fig. 1) (6) were dated by in situ U-Pb SIMS on microbaddeleyite grains. To couple geochronological and paleomagnetic records for these mafic units, complementary paleomagnetic studies were conducted on specimens from the TGS-05 and MDK-05 sample sites (Fig. 1) using conventional demagnetization techniques.

Fig. S1.

The Kaapvaal Craton in southern Africa with the three subbasins of the Transvaal Supergroup shown along with the Archean basement. Sample localities in the Ongeluk Formation and related intrusions from this study and that of the Westerberg Sill Province (28) are also shown.

Table S3.

Sampling locality descriptions of the Ongeluk Formation volcanic rocks and related mafic intrusions

Sample ID	Latitude, Longitude*	Lithology	Age (Ma)	Refs.	Attitude†
TGS-05	28.3061° S, 23.2705° E	Basalt	2,424 ± 32	This study	Local bedding: 349°/06° W; regional bedding: 031°/18° SE
OLL-2	28.8923° S, 23.0636° E	Dolerite Sill	2,397 ± 22	This study	Attitude unknown; no structural correction
GO2-B	29.2071° S, 23.4176° E	Dolerite Dike	2,421 ± 3	This study	No structural correction
NL-13c	28.3328° S, 31.3029° E	Dolerite Sheet	2,423 ± 7	This study	150°/20° NE
M03WA	29.4545° S, 22.6081° E	Dolerite Sill	2,441 ± 6 (or 2,428 ± 4)	28	Fold axis trend and plunge: 019°/19°; regional bedding: 011°/21° SE
TGS-01	27.8652° S, 23.4695° E	Dolerite Sill	2,426 ± 1	28	Regional bedding: 356°/08° W
MDK-05	28.2737° S, 23.2967° E	Dolerite Sill	Not dated; for paleomagnetic study only	This study	Regional bedding: 321°/18° SE

^*Coordinates obtained by global positioning system (WGS84 datum).

^†Bedding attitudes given in strike azimuth/dip angle format; strikes measured using the right-hand rule.

Results

Samples G02-B and NL-13c produce upper intercept baddeleyite dates of 2,421 ± 3 and 2,423 ± 7 Ma, respectively (Fig. 2, SI Methods, Geochronology—ID-TIMS Analysis, Fig. S2, and Table S4), whereas samples TGS-05 and OLL-2 yield upper intercept dates of 2,424 ± 32 and 2,397 ± 22 Ma, respectively (Fig. 2, SI Methods, Geochronology—SIMS Analysis, Fig. S2, and Table S5). The upper intercept dates of 2,424 ± 32, 2,423 ± 7, and 2,421 ± 3 Ma are interpreted as reliable crystallization ages and overlap within 2σ uncertainty, whereas the date obtained from OLL-2 is likely affected by a minor contribution of secondary zircon as indicated by SEM imaging. We, therefore, interpret the 2,397 ± 22 Ma date of OLL-2 as a minimum age. In support of this interpretation, one baddeleyite grain (spot F861b) (Fig. 2) yielded a ²⁰⁷Pb/²⁰⁶Pb date of 2,421 ± 22 Ma.

Fig. 2.

Weighted mean age of the Ongeluk LIP. Shown is a comparison of upper intercept dates with 2σ uncertainties (red columns) from five samples of the Ongeluk LIP, including the Westerberg Sill Province (samples TGS-01 and M03WA) (28), with a calculated weighted mean age of 2,425.5 ± 2.6 Ma (green bar). The result from a single analysis spot (F861b in OLL-2) is shown for comparison (blue column) (Fig. S2).

Fig. S2.

(A) ID-TIMS concordia diagrams for baddeleyite from samples NL-13c and G02-B, with data ellipses at 2σ uncertainty. (B) In situ SIMS concordia diagrams of microbaddeleyite in samples OLL-2 and TGS-05, with data ellipses at 1σ uncertainty. In addition, (B, Left) polished thin sections (with SIMS analysis points used in the age calculations) were mapped by SEM to locate and identify (B, Right) Zr-bearing accessory minerals (SEM imagery of baddeleyites and zircons analyzed with SIMS and used in the age calculations). Some baddeleyite grains (white in the back-scattered electron images) display minor alteration to zircon (light-gray rims; e.g., spot F755).

Table S4.

ID-TIMS U-Pb baddeleyite isotopic data for dolerite samples NL-13c and G02-B

Analysis no. (no. of grains)	U/Th	Raw*		[Corr.]†				[age (Ma)]				Conc.
		Pbc/Pbtot	206Pb/204Pb	207Pb/235U	±2σ (% Error)	206Pb/238U	±2σ (% Error)	207Pb/235U	206Pb/238U	207Pb/206Pb	±2σ (Abs. error)
NL-13c: 2,423 ± 7 Ma, 4-point upper intercept date (MSWD = 0.27)
1 (5)	14.2	0.140	448.5	8.4561	1.88	0.39858	1.87	2,281.2	2,162.5	2,389.4	8.5	0.905
2 (2)	14.5	0.174	362.7	9.7646	2.39	0.45182	2.38	2,412.8	2,403.3	2,420.8	10	0.993
3 (3)	13.3	0.052	1,180	9.2915	1.10	0.43323	1.00	2,367.2	2,320.2	2,407.9	8.3	0.964
4 (5)	15.1	0.055	1,156	9.4967	0.96	0.44075	0.96	2,387.2	2,354.0	2,415.7	4.0	0.974
GO2-B: 2,421 ± 3 Ma, 5-point upper intercept date (MSWD = 2.20)
1 (5)	22.0	0.045	1,426	8.8014	0.68	0.40873	0.68	2,317.6	2,209.1	2,414.7	2.9	0.915
2 (4)	16.7	0.039	1,638	9.1680	0.58	0.42523	0.58	2,354.9	2,284.2	2,416.8	2.5	0.945
3 (3)	12.7	0.029	2,185	9.3865	0.45	0.43548	0.45	2,376.5	2,330.4	2,416.3	2.0	0.964
4 (3)	14.5	0.094	664.2	8.2795	1.46	0.38645	1.47	2,262.1	2,106.4	2,406.0	6.3	0.875
5 (3)	11.3	0.024	2,621	9.1684	0.43	0.42498	0.43	2,355.0	2,283.0	2,417.9	1.9	0.944

Abs., absolute; conc., concordance; corr., corrected; Pbc, common Pb; Pb_tot, total Pb (radiogenic, blank, and initial).

^*Measured ratio corrected for mass fractionation and spike.

^†Isotopic ratios corrected for mass fractionation (0.1% per amu for Pb determined by replicate analyses of NBS reference materials SRM 981 and SRM 983), spike contribution (²⁰⁵Pb-²³³U-²³⁶U tracer solution), blank (1 pg Pb and <0.1 pg U), and initial common Pb. Initial common Pb corrected with isotopic compositions from the Pb evolution model at the age of the samples (54).

Table S5.

Baddeleyite and zircon in situ SIMS U-Pb isotopic data for samples TGS-05 and OLL-2

Spot	r206Pb (%)*	Ratios†				Rho‡	Ages (Ma)						UO2/U	Th/U	U (ppm)
		207Pb/235U	1σ (%)	206Pb/238U	1σ (%)		206Pb/238U	1σ (Abs.)	207Pb/235U	1σ (Abs.)	207Pb/206Pb	1σ (Abs.)
TGS-05: 2,424 ± 32 Ma, 8-point upper intercept baddeleyite date (MSWD = 0.76)
F069	97.9	7.052	6.5	0.3464	6.3	0.983	1,917	105	2,118	58	2,319	20	9.8	0.58	1,477
F086bz§	95.7	5.083	11.0	0.2674	10	0.992	1,528	142	1,833	90	2,201	23	7.0	0.91	2,997
F077b§	92.3	8.244	10.0	0.3985	3.5	0.563	2,162	65	2,258	90	2,346	145	16.8	0.09	1,010
F042§	94.0	8.042	7.3	0.3918	5.5	0.834	2,131	100	2,236	66	2,333	70	11.0	0.33	944
F013	98.7	10.58	6.5	0.4865	6.2	0.969	2,555	130	2,487	60	2,432	27	10.5	0.02	2,157
F107bz§	96.9	1.706	4.1	0.1271	3.3	0.882	771	24	1,011	26	1,574	37	18.4	2.48	39,504
F285§	83.8	7.163	23	0.3505	4.9	0.685	1,937	83	2,132	204	2,325	341	15.9	0.95	887
F219	98.3	8.289	4.8	0.3950	4.6	0.983	2,146	85	2,263	44	2,371	15	12.5	0.18	6,136
F263	99.1	9.048	4.4	0.4184	4.3	0.944	2,253	82	2,343	40	2,422	25	14	0.92	1,317
F298	98.9	8.372	3.6	0.4002	3.4	0.965	2,170	62	2,272	32	2,366	16	16.3	0.27	1,114
F280	98.2	7.920	7.0	0.3732	6.8	0.981	2,044	119	2,222	63	2,390	23	10.0	0.15	1,524
F279	98.9	10.51	5.3	0.4882	5.1	0.961	2,563	107	2,481	49	2,415	25	11.4	0.15	604
F210	97.9	10.60	6.1	0.4764	5.7	0.946	2,512	118	2,489	57	2,470	33	10.8	0.18	552
F277§	97.5	2.699	7.9	0.1886	7.7	0.974	1,114	79	1,328	59	1,694	34	8.5	1.63	7,929
OLL-2: 2,397 ± 22 Ma, 6-point upper intercept baddeleyite date (MSWD = 1.90)
F664§	92.7	9.009	21	0.4836	8.0	0.709	2,543	169	2,339	193	2,165	287	11.5	0.15	1,978
F691bz	99.5	6.324	9.2	0.3222	9.0	0.996	1,800	142	2,022	81	2,256	15	9.0	0.48	7,929
F431	99.1	8.698	8.4	0.4097	8.5	0.985	2,214	159	2,307	77	2,390	25	9.9	0.46	2,011
F539	99.8	8.782	7.1	0.4247	7.0	0.993	2,282	134	2,316	64	2,346	14	11.5	0.25	4,677
F252§	89.8	9.829	19	0.4437	6.2	0.748	2,367	122	2,419	173	2,463	248	22.5	0.3	249
F079	99.6	9.837	3.7	0.4654	3.6	0.973	2,463	73	2,420	34	2,383	14	18.7	0.76	1,671
F755	99.8	7.898	6.0	0.3762	5.9	0.996	2,059	104	2,219	54	2,371	9	12.1	0.39	2,931
F818	98.3	9.573	5.2	0.4528	5.1	0.990	2,408	103	2,395	48	2,384	12	13.4	0.11	2,841
F861_a§	92.6	13.16	11	0.5470	8.7	0.823	2,813	199	2,691	107	2,601	107	12.8	0.4	1,897
F861_b	99.7	9.823	7.2	0.4545	7.2	0.996	2,415	144	2,418	66	2,421	11	10.4	0.05	1,487
F352z	95.8	0.800	7.8	0.0779	2.4	0.566	470	19	584	38	1,056	135	11.2¶	2.94	6,773
F239z@2	99.1	0.502	4.1	0.0513	2.2	0.553	313	11	402	17	957	70	12.2¶	1.57	13,192
F239z@3	99.9	7.760	1.5	0.3626	1.4	0.954	1,931	41	2,170	22	2,404	8	15.1¶	0.17	2,164

All grains are pure baddeleyite, except as noted. Data acquired October 25, 2014 on the CAMECA ims1270 secondary ion microprobe (SIMS) at UCLA. Microbaddeleyite grains were analyzed in situ from polished thin sections using an aperture in the transfer section of the secondary beam column to reduce the effective sampling diameter from ∼20 to 4 μm. Sample chamber was flooded with oxygen (∼10⁵ torr) to enhance Pb secondary ion yields ∼10-fold; U/Pb relative sensitivity was calibrated by UO₂/U for baddeleyite using Phalaborwa reference material with values that ranged from 12.4 to 4.6 and UO/U for zircon using AS3 zircon standard with values that ranged from 10.4 to 8.23. U concentrations based on U/⁹⁴Zr₂O measurements compared with those of zircon standard 91500 (80 ppm U) during the same session. Pb values were corrected for common Pb using the ²⁰⁴Pb method for baddeleyite and zircon. Abs., absolute; bz, baddeleyite rimmed with zircon; z, zircon.

^*Radiogenic ²⁰⁶Pb in percentage.

^†Ratios corrected for common Pb using the measured ²⁰⁴Pb.

^‡Correlation coefficient between ²⁰⁶Pb/²³⁸U and ²⁰⁷Pb/²³⁵U ratios.

^§Rejected from age calculations because of low percentage radiogenic ²⁰⁶Pb.

^¶These values are UO/U and not UO₂/U.

A structurally corrected virtual geomagnetic pole (VGP) for the dated sample TGS-05 is located at −0.7° N and 288.2° E, whereas the VGP of sample MDK-05 is located at −0.9° N and 283.7° E (SI Methods, Paleomagnetism, Fig. S3, and Table S6). Samples OLL-2, G02-B, and NL-13c already have published shallow magnetization directions similar to those reported for the Ongeluk Formation (6, 26, 27).

Fig. S3.

(A, Inset) Representative Zijderveld diagrams for the demagnetization behavior of (a) TGS-05 and (c) MDK-05, with (b and d) corresponding equal area projections of the characteristic magnetic components. Equal area projections are in situ. In the Zijderveld diagrams, tick marks represent 1 μA⋅m⁻². (A) The magnetic components on which the Ongeluk LIP key paleomagnetic pole is based. (B) The Ongeluk LIP key paleomagnetic pole. The Ongeluk LIP is defined with error margins (red circle). In addition, the Paleoproterozoic apparent polar wander path is presented with previously published paleomagnetic poles and error margins (blue circles) (Tables S6 and S7).

Table S6.

Paleomagnetic data for TGS-05 and MDK-05, with a summary of published individual VGP data from the ca. 2,426 Ma Ongeluk Formation and associated intrusions, which were then combined to form a grand mean key paleomagnetic pole

Mafic unit and site	Age (Ma)	Site Lat. (°N)	Site Lng. (°E)	n/N	Structural correction	Geographic coordinates				Tilt-corrected coordinates				Tilt-corr. VGP Lat. (° N)	Tilt-corr. VGP Lng. (° E)	dp (°)	dm (°)
						Decl. (°)	Incl. (°)	α95 (°)	k	Decl. (°)	Incl. (°)	α95 (°)	k
Westerberg Sill Province (i.e., dolerite sills in Kuruman Iron Fm.) (28)
TKW-E	2,428 ± 4	−29.3	22.3	14/17	Fold axis azimuth and plunge: 19.1°/19.4°; bedding strike and dip: 344°/16° NE	296.6	−39.1	6.4	39.4	278.3	−35.1	6.6	37.2	16.3	278.9	4.4	7.6
TKW-A	2,428 ± 4	−29.3	22.3	2/10	Fold axis azimuth and plunge: 19.1°/19.4°; bedding strike and dip: 011°/21° SE	281.8*	−55*	34.6*	27.1*	309*	−50.1*	34.6*	27.1*	—	—	—	—
TKW-D	2,428 ± 4	−29.4	22.6	6/10	Fold axis azimuth and plunge: 19.1°/19.4°; bedding strike and dip: 242°/19° NW	300.2	−12.4	6.2	116.2	294.5	−19.0	6.2	115.9	26.1	296.0	3.4	6.5
FYL	2,426 ± 1	−27.8	23.4	7/8	Bedding strike and dip: 356°/08° E	257.2*	−53*	23.3*	6.6*	259*	−45.4*	23.2*	6.6*	—	—	—	—
MDK-03	—	−28.0	23.4	5/6	Bedding strike and dip: 138°/14° SW	282.2	−29.2	19.5	13.0	289.9	−36.7	19.5	13.1	26.5	283.2	13.3	22.8
MDK-05	—	−28.3	23.3	8/8	Bedding strike and dip: 141°/18° SW	263.2	−0.6	11.5	21.3	264.6	−15.8	11.5	21.4	−00.9	283.7	6.1	11.8
Sill (Ongeluk Fm.) (6)
OLL-2	2,397 ± 22	−28.9	23.1	5/8	No structural correction	110.0*	11.0*	34.0*	6.0*	110.0*	11.0*	34.0*	6.0*	—	—	—	—
C-.g. Ongeluk Fm. volcanic flow unit
TGS-05	2,424 ± 32	−28.3	23.3	5/8	Bedding strike and dip: 349°/06° NE	266.7	6.4	15.0	21.6	266.9	−8.3	15.0	21.6	−00.7	288.2	7.6	15.1
F.-g. Ongeluk Fm. volcanic flow units (6)
OL-C		−28.1	22.9	4/4	No structural correction	279.0	−17.0	18.0	20.4	279	−17.0	18.0	20.4	12.0	289.4	9.6	18.6
OWS-1		−27.1	23.0	5/7	Bedding strike and dip: 195°/05° W	239.6	−22.5	7.7	100.0	238.5	−26.9	7.7	99.1	−19.8	264.4	4.5	8.4
OWS-2		−27.1	23.0	6/6	Bedding strike and dip: 195°/05° W	246.2	−27.9	8.2	67.3	245.1	−32.4	8.1	70.0	−12.7	265.4	5.2	9.1
OV-P		−28.3	23.3	5/5	No structural correction	271.6	−26.0	8.1	89.8	271.6	−26.0	8.1	89.8	07.8	281.6	4.7	8.8
OL-B		−28.9	22.8	7/7	No structural correction	257.8	−16.9	5.6	118.0	257.8	−16.9	5.6	118.0	−06.3	279.3	3.0	5.8
OVG-1		−28.9	22.8	5/6	Bedding strike and dip: 023°/05° E	266.7	−26.2	10.4	54.9	267.5	−21.4	10.3	55.8	03.2	281.9	5.7	10.9
OVG-2		−28.9	22.8	6/6	Bedding strike and dip: 023°/05° E	265.8	−29.9	6.6	103.0	264.7	−25.4	6.6	103.0	01.9	278.6	3.8	7.1
OVG-3	>2,392 ± 23 or 2,222 ± 13 (14)	−28.9	22.8	6/6	Bedding strike and dip: 023°/05° E	261.5	−18.8	16.3	17.9	260.9	−14.1	16.3	17.9	−04.4	282.1	8.5	16.6
OLVW-1		−28.9	22.8	5/6	No structural correction	266.1	−5.8	13.2	34.5	266.1	−5.8	13.2	34.5	−02.0	288.4	6.7	13.3
OLVW-2		−28.9	22.8	6/7	No structural correction	270.0	−8.0	19.3	13.0	270.0	−8.0	19.3	13.0	01.9	289.3	9.8	19.4
OLVN-1		−28.9	22.9	6/6	No structural correction	259.8	−25.3	4.2	251.0	259.8	−25.3	4.2	251.0	−02.3	279.6	2.4	4.5
OLVN-2		−28.9	22.9	6/6	No structural correction	259.4	−21.9	5.3	159.0	259.4	−21.9	5.3	159.0	−03.6	277.8	3.0	5.6
OL-K		−28.9	22.9	6/6	Bedding strike and dip: 023°/05° E	261.3	−30.7	8.7	60.7	258.7	−24.3	8.8	59.4	−03.5	276.3	5.0	9.4
OLK-D		−28.9	22.9	8/8	No structural correction	264.2	−25.9	7.2	61.0	264.2	−25.9	7.2	61.0	01.6	278.2	4.2	7.7
OLV-P1		−28.9	22.9	4/4	Bedding strike and dip: 153°/10° SW	267.4	−24.9	15.4	36.5	270.3	−33.4	15.4	36.5	09.0	276.9	10.0	17.5
OLV-P2		−28.9	22.9	4/4	Bedding strike and dip: 153°/10° SW	269.8	−29.6	9.8	89.5	273.6	−37.8	9.8	89.8	13.1	275.7	6.8	11.5
OLV-P3		−28.9	23.0	6/6	No structural correction	266.8	−30.7	13.7	25.0	266.8	−30.7	13.7	25.0	05.2	276.9	8.5	15.2
OLVP-4		−28.9	22.9	6/6	Bedding strike and dip: 153°/10° SW	261.5	−16.3	19.6	12.7	262.0	−22.1	19.6	12.7	−01.3	279.0	10.9	20.7
OLR-2		−28.9	23.0	6/6	No structural correction	256.6	−20.6	14.9	14.9	256.6	−20.6	14.9	14.9	−06.3	277.1	8.2	15.6
OLL-1		−28.9	23.0	6/6	No structural correction	289.0	−3.7	17.9	14.9	289.0	−3.7	17.9	14.9	17.5	300.7	9.0	18.0
Dolerite dike (26)
GO2-B	2,421 ± 3	−29.2	23.4	4/11	No structural correction	289.5	24.8	17.1	22.3	289.5	24.8	17.1	22.3	10.0	314.5	9.8	18.4
Dolerite sheet (27)
NL-13c	2,423 ± 7	−28.3	31.3	10/10	Sheet strike and dip: 330°/20° NE	297.1	−32.0	6.8	51.1	289.2	−19.6	6.8	51.1	14.0	330.2	4.3	7.7
Mean of all sites (excluding sites with α95 > 20°)						269.2	−19.7	6.8	17.6	268.7	−21.0	6.2	20.8
														Pole Lat. (° N)	Pole Lng. (° E)	A95 (°)	K
Mean paleomagnetic pole for sites with α95 ≤ 20°														04.1	282.9	5.3	28.1

A₉₅, radius of 95% confidence about the mean pole; α₉₅, radius of 95% confidence cone about the mean direction; C.-g., coarse-grained; corr., corrected; dp and dm, semiaxes of 95% confidence about the mean; F.-g., fine-grained; Fm., formation; K, precision parameter for pole; k, precision parameter for direction; Lat., latitude; Lng.. Longitude; N, site directions obtained; n, site directions used.

^*Sites with α95 > 20° were excluded from the mean of all sites.

Discussion

The Ongeluk LIP.

All samples were collected stratigraphically from within or below the lower Ongeluk Formation (Fig. 1) and represent either the feeder system of dolerite dikes and sills to the Ongeluk Formation basalts or coarse-grained interiors of thicker basalt flows. We interpret the Westerberg Sill Province and the N-trending dolerite dike swarm (Fig. 1), both intruding into the Griqualand West subbasin stratigraphy, as parts of the same short-lived magmatic event based on their temporal, spatial, and stratigraphic proximities and similar paleomagnetic results. The age of the Westerberg Sill itself was defined by an upper intercept date of 2,441 ± 6 Ma composed of five discordant baddeleyite analyses (28). However, excluding the most discordant analysis from this result yields a more probable upper intercept date of 2,428 ± 4 Ma. This reinterpretation is supported by concordant baddeleyite analyses from a second sill dated at 2,426 ± 1 Ma as part of the same study (28). Combining all of these dates, we calculate a weighted mean date of 2,425.5 ± 2.6 Ma (Fig. 2) as the age of a single relatively short-lived magmatic event, which is now clearly distinguished from ≤2,250–2,240 Ma Hekpoort Formation volcanism with which it was previously correlated. Because sample NL-13c is located ∼1,000 km to the east of the other sample sites, our results indicate a defined craton-scale LIP, the Ongeluk LIP.

The Ongeluk Key Paleomagnetic Pole.

The VGPs presented in this study overlap with previously published paleopoles for the Ongeluk Formation and associated intrusions (6, 26–28). Collectively, the combined VGPs from all of these studies for the basalts from the Ongeluk Formation (6) and their intrusive feeders (26–28) define a grand key paleomagnetic pole for the Ongeluk LIP that is near-equatorial (29) at 4.1° N, 282.9° E (Table S7), with an A₉₅ of 5.3° that achieves five of seven on the quality scale by Van der Voo (30).

Table S7.

Compilation of paleomagnetic poles for the Neoarchean-Paleoproterozoic Kaapvaal Craton

Abbreviation	Rock unit	Age (Ma)	Age refs.	Latitude (°N)	Longitude (°E)	(dp, dm) or A95 (°)	Paleopole refs.
1	Allanridge Formation	Ca. 2,700	86	−69.8	345.6	5.8	52
2	Rykoppies–White Mfolozi dyke swarms	Ca. 2,700–2,660	86	−62.1	336	(3.5, 4.2)	27
3	Westerberg Sill Province	Ca. 2,426	28	16.8	279.9	(4.4, 7.7)	28
4	Ongeluk Formation	Ca. 2,426	This study	22	−0.5	5.3	6
				4.1	282.9	5.3	This study
5	Phalaborwa Complex	Ca. 2,061	87	27.7	35.8	6.6	88
6	Bushveld Complex	Ca. 2,056–2,055	89	19.2	30.8	5.8	90
7	Lower Waterberg Group	≤2,054	91	36.5	51.3	10.9	92
8	Vredefort Impact	Ca. 2,023	93	21.8	44.5	(11.3, 15.4)	94
9	Mapedi–Gamagara Unconformity	Ca. 2,200–2,000	64	2.2	81.9	(7.2, 11.5)	64

Revising the Transvaal Supergroup Stratigraphy.

The age for Ongeluk Formation volcanism demands the revision of stratigraphic correlations between the successions of the Griqualand West and Transvaal subbasins of the Transvaal Supergroup (Fig. 1). Previous studies have correlated the Postmasburg and Pretoria Groups using the 2,222 ± 13 Ma Ongeluk Formation and the ≤2,250–2,240 Ma Hekpoort Formation volcanic rocks (11, 12, 14, 18). This correlation is now shown to be incorrect based on the 2,426 ± 3 Ma age constraint provided by the Ongeluk Formation volcanic rocks and associated intrusions. In addition, lithologic and chemostratigraphic data for the Duitschland and Rooihoogte formations (15, 18, 23, 24, 31) coupled with recent age constraints (9, 11, 25) and arguments relying on basin architecture (16) indicate that the Duitschland and Rooihoogte formations are younger than the Postmasburg Group (Fig. 1). A more complete discussion on the Transvaal Supergroup stratigraphy and the proposed revisions is provided in the stratigraphic synthesis (SI Methods, Stratigraphic Synthesis).

Atmospheric Oxygen Oscillations.

This stratigraphic interpretation indicates a dynamic state of atmospheric oxygen levels during the early Paleoproterozoic glacial period. The onset of the GOE occurred in the immediate aftermath of the Makganyene Formation glaciation with deposition of the world’s largest manganese deposit, the Hotazel Formation, with a negative Ce anomaly, both indicative of oxygenation (7, 21). This oxygenation was followed by a return to anoxic atmospheric conditions as indicated by the lack of a Ce anomaly in foreslope carbonates of the Mooidraai Formation (Fig. 1) (21, 32) and a mass-independent fractionation of sulfur (MIF-S) signal recorded by early diagenetic sulfides from the lower Duitschland Formation (Fig. 1) (23). Subsequent oxygenation events occurred during deposition of the upper Duitschland Formation (23) and once again, in the middle of the Rooihoogte Formation (24) as indicated by the reappearance and disappearance of the MIF-S signal (Fig. 1).

Detrital pyrite grains persist above the oldest glacial diamictite in the Mississagi Formation in the Huronian Supergroup of the Superior Craton, Canada (33) and support oscillations in atmospheric oxygen herein inferred from the records of the Transvaal Supergroup (Fig. 3). Our geochronologic and stratigraphic framework argues against a simple, monotonic rise of atmospheric oxygen in the early Paleoproterozoic, a time period further characterized by four glaciations. Instead, the onset of the GOE was followed by oscillations in atmospheric oxygen content across the 10⁻⁵ present atmospheric level (PAL) threshold over an ∼200-My interval, adding empirical evidence to atmospheric modeling predictions (4).

Fig. 3.

A is a graph illustrating the approximate chronology (Table S1) of the glaciations and atmospheric oxygen oscillations according to the related redox indicators in the Huronian and Transvaal basins (Table S2) using the same symbols as used in Fig. 1. Also shown are dated mafic and felsic magmatic events as well as δ¹³C ranges for carbonates (Tables S1 and S2) and the extent of stratigraphic records in each basin denoting gaps in records at unconformities and disconformities. (B) The early Paleoproterozoic geography of the Superior, Kola–Karelia, Hearne, and Wyoming cratons as integral parts of the supercraton Superia (5, 36)* with the addition of the Kaapvaal and Pilbara cratons in the supercraton Vaalbara configuration (52). The early Paleoproterozoic basins developed on these cratonic fragments include both ca. 2.51–2.43 Ga volcanic rocks and glacial units, which can be correlated across the cratons. The glacial units in bold denote the glacial deposits likely recording the first glaciation. All of the cratonic fragments also contain dolerite dikes and sills emplaced between ca. 2.51 Ga and 2.43 Ga, showing the extent of the LIPs formed during this time. Available paleomagnetic studies indicate that the majority of the cratonic fragments (as part of supercraton Superia) were positioned near the paleoequator. The arrows denoting present-day true north in the crustal blocks illustrate the rotations necessary to make the reconstruction. (Inset) The hypothesized paleolatitude of these Archean cratons in the early Paleoproterozoic.

Linking Snowball Earth and the GOE.

The Makganyene Formation, deposited in the tropics (6), records the oldest known Snowball Earth event (7). Makganyene Formation diamictites are now constrained to be slightly older than ca. 2,426 Ma and likely correlate with glacial units of the Ramsay Lake Formation in the Huronian Supergroup, Canada and the Campbell Lake Formation from the Snowy Pass Supergroup on the Wyoming Craton in the United States. Other correlative glacial units worldwide include the ca. 2,435 Ma Polisarka Formation on the Kola–Karelia Craton in Fennoscandia (34) as well as possibly, the Meteorite Bore Member on the Pilbara Craton in Australia, defining the wide extent of the oldest Paleoproterozoic glaciation (Fig. 3). This early Paleoproterozoic glaciation is broadly coeval with the onset of the GOE; both events are now tightly bracketed between ca. 2,460 Ma, the age of volcanic rocks near the base of the Huronian Supergroup (8, 10), and ca. 2,426 Ma, the approximate age of the Makganyene Formation glaciation and near-coeval Ongeluk Formation volcanic rocks in the Transvaal Supergroup. In addition, the 2,442 ± 2 Ma Seidorechka Formation and the overlying 2,435 ± 2 Ma Polisarka formations (34, 35) overlying the Kola–Karelia Craton may tightly bracket the oldest Paleoproterozoic glacial event between ca. 2,442 Ma and 2,435 Ma, respectively, implying that it lasted less than 7 My (Fig. 3).

Our results add to the growing evidence for large low-latitude continental landmasses in the early Paleoproterozoic, including the Kaapvaal Craton (6) and the clan of cratons that defines supercraton Superia (5, 36, 37): Superior, Wyoming, Hearne, and Karelia–Kola. These landmasses, at least in part contiguous, record a series of LIP events between 2,510 Ma and 2,440 Ma and include the Mistassini, Kaminak, Baltic, Baggot Rocks, and Matachewan LIPs (Fig. 3) (5) before the ca. 2,426 Ma Ongeluk LIP of the Kaapvaal Craton. Cumulatively, these large juvenile volcanic provinces on extensive low-latitude continental landmasses are likely to have triggered near-equatorial glaciations via enhanced chemical weathering of aerially extensive, nutrient-rich continental flood basalts. This weathering resulted in increased carbon dioxide drawdown (38, 39) and an enhanced flux of phosphorus and other essential nutrients (40) onto extensive continental margins and into intracratonic basins. An enhanced nutrient flux would have greatly increased photosynthetic activity and oxygen production, temporally linked to higher net burial of organic carbon in accumulating sediments, as reflected by δ¹³C values of carbonates in the upper Tongwane Formation of the Transvaal Supergroup (31). The Tongwane Formation is locally preserved above the ca. 2,480–2,460 Ma iron formations (41) but below the major unconformity documented in the entire Transvaal basin (Fig. 1) (18). Thus, even an incipient rise of free atmospheric oxygen would have led to rapid oxidation of atmospheric methane (42), forcing catastrophic climate change and plunging Earth into a global glaciation (43). Importantly, the dated near-equatorial Ongeluk LIP, conformably overlying and interfingering with the uppermost Makganyene Formation glacial diamictites (Fig. 1), illustrates the dual role of LIPs in these global events, in that they would also have contributed carbon dioxide to rebuilding the greenhouse atmosphere that led to abrupt termination of the first Snowball Earth state of the early Paleoproterozoic.

Comparing the Paleoproterozoic with the Neoproterozoic.

In the Neoproterozoic, a remarkably similar sequence of events occurred, involving successive emplacement of multiple LIPs on the supercontinent Rodinia, a low-latitude position of this supercontinent, and incipient rifting and breakup (44). The massive Franklin LIP at ca. 717 Ma (45) immediately preceded the most dramatic and longest global glaciation of the Neoproterozoic, the Sturtian (46). This overall period is characterized by the second most dramatic change in surface redox conditions linked with Snowball Earth glaciations (47) and accompanied high rates of organic carbon burial (48, 49). Although substantial differences between the Paleo- and Neoproterozoic glacial periods might be expected, for instance, in the triggering mechanisms for the initial global glaciations, because methane was probably a more important greenhouse gas before the Makganyene glaciation (42, 43) than before the Sturtian glaciation (38, 39), there are also uncanny parallels. Examples include supercraton- or supercontinent-size continental landmasses that were capped by continental flood basalts, incipient rifting and/or breakup, and rapid transit to low latitudes. All of these factors enhanced chemical weathering of juvenile basaltic material and greatly increased the flux of bio-limiting nutrients to depositional basins, thus leading to a biotic response of higher organic carbon burial (40, 50) as well as deposition of giant iron and manganese deposits. It seems likely that these similar scenarios are not coincidental but that the critical factors (assembly of large landmasses, LIPs, incipient rifting, and relief enhancement—all resulting in a lithospheric mass anomaly and movement to low latitudes, high rates of organic carbon burial, surface oxygenation, and Snowball Earth glaciations) are mechanistically linked. In this case, a critical link might be true polar wander caused by the lithospheric mass anomaly that nudged the basalt-covered and rifting supercontinental landmasses to the equator (51), where chemical weathering and nutrient fluxes kicked into high gear and triggered the biotic, redox, and climatic responses.

SI Methods

Sampling.

Samples analyzed and interpreted as part of this study were taken from dolerite intrusions (i.e., sills and dikes) that intruded into basement of the western part of the Kaapvaal Craton and the overlying cover succession of the Transvaal Supergroup in the Griqualand West subbasin, South Africa, except for one sample that was taken ∼1,000 km to the east on the southeastern part of the craton (Fig. 1, Fig. S1, and Table S3). However, one sample (TGS-05) is tentatively interpreted as a coarse-grained basalt from the base of the Ongeluk Formation (Table S3). The sampled targets were selected using 1:250,000 map sheets of the Council for Geoscience (TGS-05 and MDK-05) or studied previously (e.g., OLL-2, G02-B, and NL-13c) (6, 26, 27). The previously studied intrusions yielded “Ongeluk-like” paleomagnetic directions. In addition, samples M03WA and TGS-01 (28) were reinterpreted in this study.

Sample TGS-05 (Fig. 1, Fig. S1, and Table S3) was identified in a road cut between Danielskuil and Postmasburg. It is interpreted as a medium-grained green basalt at the very base of the Ongeluk Formation. It is tentatively interpreted as a lava flow based on amygdales and nearby flow-top breccias. The transition into the volcanic breccias was not observed, however, and a shallow dolerite sill cannot be ruled out. As part of our study, it was drilled for complimentary paleomagnetic data. The Ongeluk Formation dips less than 10° in this location. Paleomagnetic samples were also collected from dolerite sill site MDK-05 (Fig. 1, Fig. S1, and Table S3) on the Groenwater Farm between Danielskuil and Postmasburg (near sample site TGS-05). It is a medium-grained green dolerite at the contact of the Asbestos Hill Subgroup and the overlying Makganyene Formation. Contacts and structural attitude were not observed because of poor outcrop, but the geological map indicates a conformable sill that may feed up into the Ongeluk Formation basalts. Nearby units of the Ongeluk Formation dip less than 10°. Sample OLL-2 (Fig. 1, Fig. S1, and Table S3) is from a sill within the lower Ongeluk Formation volcanic rocks exposed along the road from Griekwastad to Campbell. It is a coarse-grained green dolerite occurring within the lower Ongeluk Formation and was paleomagnetically studied previously (6). It is interpreted as a sill, but contacts were not observed. The attitude is unknown. Sample G02-B (Fig. 1, Fig. S1, and Table S3) was taken along the Orange River at the Gewonne Farm, where an exposure of dolomite and limestone of Schmidtsdrif Subgroup mapped within the river bed reveals an ∼80-m-wide subvertical, N-trending dolerite dike with a previously determined Ongeluk-like paleomagnetic direction (26). It is a medium- to coarse-grained green dolerite. The sample NL-13c (Fig. 1, Fig. S1, and Table S3) was taken from a shallow-dipping, medium-grained gray dolerite sheet intruding into Mesoarchean basement below the Pongola Supergroup ∼1,000 km away in the southeastern Kaapvaal Craton. It was described as a sheet intrusion (27), having a structural orientation similar to the regional strike and dip of the nearby basal Pongola Supergroup units. It cross-cuts an east–north–east-trending dyke, which in turn, cross-cuts an south–east-trending dolerite dike in the region. This dolerite sheet had been interpreted previously as a ca. 0.18 Ga intrusion based on a structurally uncorrected ‘”Karoo-like” magnetic remanence (27); however, structural correction reveals the Ongeluk-like magnetic remanence. Sample M03WA (Fig. 1, Fig. S1, and Table S3) is from the Westerberg Sill itself in the Griqualand West subbasin and was sampled for geochronologic and paleomagnetic studies previously along with TGS-01 (28), yielding a 2,441 ± 6 Ma age and Ongeluk-like paleomagnetic direction. M03WA was taken near the old asbestos mine village of Westerberg along the Orange River. It is a medium-grained gray–green dolerite sill intruding the Asbestos Hills Subgroup along the contact between contrasting rock units. TGS-01 (28), a medium-grained gray–green dolerite sample (Fig. 1, Fig. S1, and Table S3), intruded the Asbestos Hill Subgroup much farther northeast into the hinterland of the Kaapvaal Craton nearer to Kuruman. It is probably a sill, but no contacts were seen because of poor exposure. Regional dip of bedding is less than 10°. It was dated at 2,426 ± 1 Ma and also, has an Ongeluk-like paleomagnetic direction (28).

Geochronology—ID-TIMS Analysis.

Recovery of baddeleyite was attempted from all of the mafic rock samples using a water-based separation technique from crushed material at Lund University. Baddeleyite is a common accessory phase in mafic intrusions and coarse-grained basalt flows and a proven U-Pb geochronometer. This technique was successful in separating baddeleyite from the coarser-grained mafic intrusions NL-13c and G02-B. Consequently, these samples were dated by dissolution using U-Pb ID-TIMS (53).

The best-quality baddeleyite grains were transferred to Teflon dissolution capsules and rinsed repeatedly in ultrapure 7 M HNO₃ and H₂O to decrease the Pb blank. An ultrapure HF-HNO₃ mixture was then added to the dissolution capsules together with a ²⁰⁵Pb-^233–236U tracer solution. Capsules were then placed in an oven at 190 °C for 3 d until the grains were fully dissolved. After dissolution, the samples were dried down on a hotplate and redissolved in ultrapure 6 M HCl with the addition of ultrapure 0.25 M H₃PO₄ before being dried down again on a hotplate. The samples were then dissolved in 2 μL silica gel and loaded on outgassed Re filaments. The U and Pb isotopic compositions were measured at the Department of Geosciences at the Swedish Museum of Natural History using a Thermo Finnigan Triton Mass Spectrometer equipped with Faraday detectors and a Secondary Electron Multiplier. Pb isotope analyses were measured in dynamic (peak-jumping) mode using the Secondary Electron Multiplier. After completion of the ²⁰⁴Pb, ²⁰⁵Pb, ²⁰⁶Pb, and ²⁰⁷Pb isotopic measurements (typically in the 1,210 °C to 1,230 °C filament temperature range) for between 50 and 100 cycles, the ²³³U, ²³⁶U, and ²³⁸U isotopic measurements were made in dynamic mode at filament temperatures greater than 1,350 °C. Procedural blanks are typically 1.0 pg Pb and 0.1 pg U at the Swedish Museum of Natural History. Mass fractionation is 0.1% per mass unit for Pb determined by replicate analyses of National Bureau of Standards (NBS) reference materials SRM 981 and SRM 983. U fractionation was determined directly from the measured ²³³U/²³⁶U isotopic ratio. Initial Pb compositions were taken from the modeled common Pb evolution curve (54). The decay constants used were 1.55125 × 10⁻¹¹ (²³⁸U) and 9.8485 × 10⁻¹⁰ (²³⁵U), with the isotopic composition of U being ²³⁸U/²³⁵U = 137.88 (55, 56). The listed uncertainties in Pb/U ratios were calculated by propagating the within-run error for measured isotopic ratios with the uncertainties in fractionation (±0.04 for Pb; absolute uncertainties), Pb and U blank concentration (±50%), and Pb blank composition (2.0% for ²⁰⁶Pb/²⁰⁴Pb and 0.2% for ²⁰⁷Pb/²⁰⁴Pb). Analytical results were calculated and plotted using the Microsoft Excel Macro, Isoplot (57).

Four baddeleyite fractions were analyzed from NL-13c (Table S4), each composed of between five and eight moderately brown and minute (<40 μm in length) crystals. One fraction is concordant within error, whereas the other fractions plot between 2.5 and 9.5% discordant (Fig. S2). The upper and lower intercept dates are 2,423 ± 7 and 510 ± 180 Ma, respectively, with a mean square weighted deviation (MSWD) of 0.27. The date of 2,423 ± 7 Ma is interpreted to represent the crystallization age of this sample. For G02-B, five fractions were analyzed, each composed of between one and four grains (Table S4). The grains were dark brown and somewhat frosty because of the thin skins of secondary zircon, and results were between 4 and 12% discordant (Fig. S2). An upper intercept date is calculated at 2,421 ± 3 Ma (MSWD = 2.20), which we interpret as the crystallization age of this dyke. The lower intercept date of 132 ± 89 Ma indicates minor near-0 Ma age Pb loss or isotopic disturbance caused by ca. 0.18 Ga Karoo magmatism.

Geochronology—SIMS Analysis.

Physical separation failed for the two critical finer-grained samples, TGS-05 and OLL-2, in which baddeleyite and zircon grains were generally ≤25 μm in size. For the analysis of such grains, we applied in situ U-Pb isotopic measurements using SIMS (58). The first step in this in situ dating method of “microbaddeleyites” involves identifying baddeleyite grains (and minor zircon if present) in polished thin sections by locating Zr-bearing phases through imaging and mapping using back-scattered electron imaging with brightness triggers and follow-up energy-dispersive spectrometry on a scanning electron microscope (Fig. S2). Mapped accessory Zr phases (baddeleyite, zircon, and zirconolite) are then assessed and ranked based on their size, shape, content of inclusions, and degree of alteration (e.g., secondary zircon rims on baddeleyite). Selected regions of the polished thin section, hosting the best targets, are then cut out, mounted in an epoxy disk together with baddeleyite and zircon reference materials, and gold-coated for analysis. After gold coating, the mounts are reimaged in reflected light microscopy to assist in locating the target grains for SIMS analysis. U and Pb isotopic ratios were determined on in situ baddeleyite grains in thin sections using a 20-µm-diameter primary ion beam on the CAMECA ims1270 Mass Spectrometer at UCLA. The field aperture was adjusted to reduce the effective sampling regions to 4-µm areas to screen out ions from host phases. Analysis of grains as small as 3 μm is possible. Additional details of the operating procedures for the CAMECA 1270 at UCLA are given elsewhere (58). Beams were focused onto the Zr-bearing phases using a 20–25-μm-diameter spot, with the secondary ions extracted at 10 kV. The secondary Pb⁺ yield was enhanced by a factor of ∼10 through oxygen flooding (∼10⁵ torr) into the sample chamber. Isotopes of ²⁰⁴Pb, ²⁰⁶Pb, ²⁰⁷Pb, and ²⁰⁸Pb were counted in conjunction with ²³²Th, ²³⁸U, ²³⁸UO, and ²³⁸UO₂ using a Secondary Electron Multiplier for between 15 and 20 cycles, taking ∼15 min per spot. The instrumental fractionation of Pb/U was corrected using UO₂⁺/U⁺ vs. Pb/U for baddeleyite of the Phalaborwa reference material (values from 12.4 to 4.6) and UO⁺/U⁺ for zircon using the AS3 reference material (values from 10.4 to 8.2) (59). U concentration was calculated from measured U/⁹⁴Zr₂O of zircon reference material 91,500 with a concentration of 80 ppm U (×1 in zircon and ×2 in baddeleyite). Pb isotopes are not fractionated in SIMS analysis (59). Common Pb was corrected using the ²⁰⁴Pb method (60). Radiogenic ²⁰⁶Pb was typically >95%. Analytical results were also calculated and plotted using the Microsoft Excel Macro, Isoplot (57).

For sample TGS-05, 14 spots on different baddeleyite grains were analyzed (Fig. S2 and Table S5). Six analyses were rejected from the final age calculation because of low radiogenic ²⁰⁶Pb (<98%) and also, unacceptably high Th/U (>0.6), which has been observed to correlate with alteration or contamination. Linear regression of the eight remaining baddeleyite analyses resulted in an upper intercept date of 2,424 ± 32 Ma (MSWD = 0.76), which we interpret as the emplacement age of the volcanic lava flow (Fig. S2). The baddeleyites of OLL-2 suffered more alteration than those of TGS-05 (Fig. S2 and Table S5), which was indicated by secondary zircon rims of variable width around the majority of the grains. Three of nine baddeleyite analyses were rejected because of low radiogenic ²⁰⁶Pb (<98%). Linear regression of the remaining six analyses resulted in an upper intercept date of 2,397 ± 22 Ma with an MSWD = 1.90 (Fig. S2). Analysis of one baddeleyite and one coexisting zircon yielded ²⁰⁷Pb/²⁰⁶Pb dates of 2,421 ± 22 and 2,404 ± 16 Ma, respectively (2σ). Data from two other zircons were extremely discordant and suggest significant Pb loss during more recent alteration and metamorphism. The date of 2,421 ± 22 Ma from a single well-preserved baddeleyite grain (F861a) is interpreted to reflect better the true emplacement age of the sill in light of the evidence for alteration and metamorphism.

Paleomagnetism.

Core specimens from dolerite bodies TGS-05 and MDK-05 were collected using a portable gas-powered drill and oriented using both magnetic and sun compasses. Structural orientation (attitude) was derived from bedding of the Ongeluk Formation volcanic rocks or the Asbestos Hills Subgroup in the vicinity. All measurements of magnetic remanence were made using the superconducting rock magnetometer at the UJ (a vertical 2G Enterprises DC-4K Magnetometer) equipped with an RAPID Consortium Automatic Sample Changer. All specimens were exposed to stepwise alternating-field demagnetization from 2 to 100 mT. The process was stopped when sample intensity dropped below instrument noise level (∼1 pAm²) or samples started to show erratic behavior. Magnetic components were quantified via Zijderveld diagrams and least squares component analysis (61) using the software Paleomag 3.1b2 (62). Linear fits were included in subsequent analyses if they had a mean angular deviation ≤10°. Paleopole calculations are based on the assumption of a geocentric axial dipole field and a stable Earth radius throughout geological time that equals the present day radius. Visualization of pole positions and paleogeographic reconstructions was made with GPlates (63).

A summary of the demagnetization results is provided (Fig. S3 and Table S6). Specimens from sample site TGS-05 display a low-coercivity magnetic component that was removed during the first couple of demagnetization steps from 7.5 to 20 mT. This component is directed north and upward, with six of eight specimens displaying this component [declination (Decl.) = 346.3°, inclination (Incl.) = −46.3°, α₉₅ = 8.55, k = 51.98, n = 6]. At demagnetization steps from 20 to 60 mT, a consistent medium- to high-coercivity component was defined and oriented west and downward, with characteristic remanence directions decaying toward the origin on removal of erratic directions recorded above 60 mT. This measurement defines the characteristic magnetic component at Decl. = 266.7°, Incl. = 6.4°, α₉₅ = 15.0, k = 21.6, and n = 5, which on structural correction, yields Decl. = 266.9° and Incl. = −8.3° (Fig. S3 and Table S6). A structurally corrected VGP of −0.7° N and 288.2° E (Fig. S3 and Table S6) is defined for this characteristic magnetization (dp = 7.6°, dm = 15.1°). The oval area of confidence around the paleopole has principal axes aligned parallel (dp) and perpendicular (dm) to the great circle connecting the paleopole to the sampling site. Samples from site MDK-05 displayed randomly oriented lower-coercivity magnetic components that were removed below 50 mT and either shallow upward and westerly or shallow downward and easterly directed high-coercivity components above 50 mT. The high-coercivity components were identified in all eight samples and yielded a combined VGP mean at Decl. = 263.2°, Incl. = −0.6°, α₉₅ = 11.5, k = 21.3, and n = 8, which on structural correction, becomes Decl. = 264.6° and Incl. = −15.8° (Fig. S3 and Table S6). A structurally corrected VGP of −0.9° N and 283.7° E is revealed (dp = 6.1°, dm = 11.8°) (Fig. S3 and Table S6).

A paleomagnetic pole was calculated from combining 27 tilt-corrected site VGPs of ca. 2,426 Ma intrusive and extrusive units of the Ongeluk LIP, Kaapvaal Craton (Fig. S3 and Table S6), which represent ∼20 independent cooling units. Unit component means with α₉₅ values larger than 20°, and k values smaller than 10 were excluded from the pole calculation. This pole calculation includes the dated site OLL-2 as well as G02-B, TKW-A (M03WA), FYL (TGS-01), and NL-13c (6, 26–28). The resulting paleopole is located at 4.1° N and 282.9° E (with an antipole at −4.1° N and 103.9° E) with A₉₅ = 5.3° and K = 28.1, indicating a paleolatitude of 11° ± 6° (Table S7 and Fig. S3). This pole rates as five of seven possible in terms of quality criteria by Van der Voo (30), only missing criterions 7 (“no resemblance to paleopoles of younger age”) and 6 (“the presence of reversals”) (30). The ca. 2,426 Ma paleomagnetic pole resembles the paleomagnetic pole of a ca. 2.00 Ga paleoweathering surface that is exposed in the Griqualand West subbasin (64). Site OLL-2 did display an opposite polarity compared with all other sites but was not included in the paleopole calculation. The primary nature of the paleomagnetic pole is supported by a positive baked contact test (26) and positive conglomerate test (6). Our ca. 2,426 Ma paleomagnetic pole satisfies the more stringent criteria to be recognized as a key paleomagnetic pole (29).

Stratigraphic Synthesis.

Our samples and U-Pb dating results, together with all other relevant details, are summarized in Fig. 1, which represents our synthesis of the complex stratigraphy of the Transvaal Supergroup in its two main structural subbasins, the Griqualand West subbasin in the southwest and the Transvaal subbasin in the northeast (Fig. S1). These basin remnants occur on either side of a positive basement feature, the Vryburg Arch (16, 17). A third erosional remnant across the border in Botswana, the Kanye subbasin, can be considered as the western continuation of the Transvaal subbasin and is not further discussed.

The stratigraphy as shown in Fig. 1 generally follows the South African Committee on Stratigraphy and other reviews (9, 12) but with some important modifications. Our U-Pb age for the Ongeluk magmatism, at 2,426 ± 3 Ma, unlocks the long-held Ongeluk Formation basalts (Griqualand West subbasin) to Hekpoort Formation basalts (Transvaal subbasin) correlation (11, 12, 14, 18) and eliminates the ∼200-My hiatus at the base of the Postmasburg Group. With no major time hiatus at the base of the Makganyene Formation to contend with, we recognize that the transition from the Koegas Subgroup, characterized by alternating Mn-rich iron formations and sandstones (16, 17, 65), to the diamictites of the Makganyene Formation is essentially conformable (17) over distances of 50–200 km. This observation precludes significant deformation and basin inversion having occurred before deposition of the lower Postmasburg Group. Erosional effects and very low-angle disconformable relationships where Makganyene Formation diamictites directly overlie the upper Asbestos Hills Subgroup iron formations on the Ghaap Plateau (18, 66) can be largely attributed to a major sea-level fall associated with the low latitude and therefore, global (6, 7) Makganyene Formation glaciation. Basin upward shallowing started during final deposition of the iron formations. Less than 1° warping of the platform edge, perhaps associated with minor activity on syn-sedimentary basin-margin faults (e.g., the Griquatown fault system) (17, 65, 66), could also have contributed to the gradual disappearance of the Koegas Subgroup to the northeast. We thus consider the base of the Makganyene Formation to be a disconformity largely attributed to the glaciation (Fig. 1). Interestingly, the Tongwane Formation, which has a similar stratigraphic position with respect to the underlying iron formations as the Koegas Subgroup, also contains moderate Mn enrichment (41).

With no major time hiatus and considering the new ages that have emerged in recent years (15, 19, 20), a next logical step is to consider the Makganyene Formation diamictites and conformably overlying basalts of the Ongeluk Formation as a significant but temporary interruption of the carbonate- and iron formation-dominated deposition of the entire Griqualand West subbasin stratigraphy (17). After these climatic and magmatic perturbations decayed, Mn-rich iron formation deposition resumed with the Hotazel Formation (7, 17), continuing a theme first established in the underlying Koegas Subgroup. Deposition then transitioned once again to a carbonate platform, the foreslope Mooidraai Formation (15, 21, 32). Lack of compelling litho- and chronostratigraphic links of any of the Postmasburg Group formations with the upper Chuniespoort Group of the Transvaal subbasin strongly suggests that the entire Postmasburg Group is younger than the upper Chuniespoort Group (i.e., Penge and Tongwane formations) and simply has no preserved time-equivalent record in the Transvaal subbasin.

Sometime after deposition of the Mooidraai Formation carbonates, a significant tectonic disturbance led to the demise of the carbonate and iron formation platform in the Griqualand West subbasin, resulting in folding, uplift, and erosion. When deposition resumed, to the extent that it is preserved, the depocenter had shifted to the northeastern Transvaal subbasin, where the Duitschland Formation lies unconformably on folded iron formations of the Penge Formation and locally preserved iron-rich carbonates of the Tongwane Formation (31, 41), in a fundamentally different (successor?) basin with a different depocenter (18). Therefore, we consider the Duitschland Formation and the lower Pretoria Group (starting with the Rooihoogte Formation) of the Transvaal subbasin to be entirely younger than the depositional record of the Griqualand West subbasin as displayed in Fig.1.

This view is further supported by the different depositional and lithological character of the Duitschland Formation and lower Pretoria Group compared with strata of the Ghaap and Postmasburg groups and contrasting carbon isotope values for carbonates in these units (31, 67). Thick submarine volcanic rocks of the Ongeluk Formation overlying the Makganyene Formation diamictites have no equivalent in the Duitschland Formation above its basal diamictite, although a dolerite sheet of Ongeluk age is now recognized in the basement of the southeasternmost Kaapvaal Craton in this study; also, there is no lithological equivalent in the Duitschland Formation of the Hotazel Formation Mn- and Fe-rich deposits. Cap carbonates of the lower Duitschland Formation, with negative carbon isotope values, and carbonates of the upper Duitschland Formation, with positive carbon isotope values (31), also have no equivalents in the Postmasburg Group. Vice versa, carbonates of the Mooidraai Formation with near-zero carbon isotope values recording seawater composition (15) lack an equivalent in the Duitschland Formation. Our overall interpretation is fully compatible with all of the existing isotopic age constraints (Fig. 1 and Tables S1 and S2).

The Rooihoogte Formation has been considered by some authors (2, 11, 24) to be a time equivalent of the Duitschland Formation, with the Duitschland Formation restricted to the northeastern part of the Transvaal subbasin and the Rooihoogte Formation developed to the southwest. However, this interpretation is not universally adopted (18). There are important sedimentologic and provenance differences between the Duitschland and Rooihoogte formations that preclude their correlation. Although the Rooihoogte Formation is thin (up to ∼220-m thick and increases in thickness to the southwest) and has a gradational contact with the Timeball Hill Formation in the southwestern part of the Transvaal subbasin, the Duitschland Formation is up to ∼1,000 m in thickness and has a thick chert breccia on the contact with the Timeball Hill Formation, indicating exposure and an unconformity (18). In contrast, thickness of the Timeball Hill Formation does not change significantly along strike (18). There is also a dramatic change in provenance from the Duitschland Formation to the Timeball Hill Formation, with some Archean detrital zircon distribution modes disappearing, although early Paleoproterozoic sources appear for the first time in the provenance (11). The Rooihoogte Formation is, therefore, genetically linked with the Timeball Hill Formation, whereas the Duitschland Formation is unconformably overlain by the Timeball Hill Formation. The Rooihoogte and Timeball Hill formations were deposited in a different basin after a period of erosion that removed most of the Duitschland Formation and sourced from a distinctly different provenance. It is, therefore, reasonable to infer that the Rooihoogte and Duitschland formations are not correlative, with the Duitschland Formation being older.

We display the stratigraphy of the Transvaal subbasin up to the Dwaalheuvel Formation, which has the first widely accepted terrestrial red-bed sandstones in southern Africa, deposited in a fluvial to fluviodeltaic setting (12, 13). These red beds overlie oxic paleosols developed on the upper contact of the subaerial Hekpoort basalts (13) that are younger than ca. 2,250–2,240 Ma based on the age of a small number of detrital zircons in intercalated volcanoclastic sandstone (11) and the 2,256 ± 6 Ma U-Pb zircon age of a tuff in the underlying upper Timeball Hill Formation (9). Well-developed hematitic ooliths and pisoliths in fluviodeltaic deposits of the middle Timeball Hill Formation (68) constrain a significant rise in atmospheric oxygen level to have occurred well before extrusion of the Hekpoort Formation basalts and development of its overlying paleosol (12).

Fig. 1 also summarizes the record of mass-independent sulfur isotope fractionations being either present or absent (MIF-S vs. no MIF-S, respectively) as currently available in the literature (23, 24, 69–71). Combining this record with that of other redox indicators (Table S2), we conclude that the onset of the GOE (1) was approximately coeval with the oldest Paleoproterozoic glaciation and the submarine extrusion of the up to ∼800-m thick Ongeluk Formation basalts (i.e., ca. 2,426 Ma) and postdates the last robust record of well-rounded detrital pyrite and rare uraninite grains in the Koegas Subgroup (22). Interestingly and puzzlingly, the upper Koegas Subgroup might also contain the potentially oldest red beds with detrital grains coated by hematite (65), although this interpretation remains to be further tested. Given the notable recurrence of an MIF-S signature in the lower Duitschland and Rooihoogte formations (23, 24), we conclude that this record resolves at least two oscillations back through the threshold of 10⁻⁵ PAL of oxygen (42). Lack of a Ce anomaly in carbonates of the upper Mooidraai Formation, following a negative anomaly in Mn-rich deposits of the underlying Hotazel Formation (21), is also inconsistent with a simple monotonic rise in atmospheric oxygen levels after the initial onset of GOE.

If our stratigraphic interpretations are correct, the onset of GOE is now geochronologically well-constrained to have occurred between ca. 2,460 Ma (72) and our precise U-Pb age of 2,426 ± 3 Ma for the Ongeluk Formation in the Transvaal Supergroup. Adding constraints from the Huronian Supergroup in Canada, where detrital pyrite grains persist well above the now well-dated ca. 2,460–2,455 Ma lower Huronian volcanic rocks (8, 10), tighten the age constraint closer to our Ongeluk Formation age. The tempo and rhythm of the GOE were likely complex, perhaps driven by various nonlinear processes, with current data resolving at least two major oscillations of atmospheric oxygen levels through the MIF-S threshold. Global interbasinal correlations based largely on the assumption of a single MIF-S to no MIF-S transition (2) might be erroneous in the absence of independent stratigraphic and geochronologic constraints.

Methods

Sampling.

Samples in this study were taken from dolerite intrusions that intruded into the basement of the Kaapvaal Craton and the overlying cover succession of the Transvaal Supergroup in the Griqualand West subbasin, South Africa, except for one sample taken ∼1,000 km to the east (Fig. 1, SI Methods, Fig. S1, and Table S3). One sample is tentatively interpreted as a coarse-grained basalt from the base of the Ongeluk Formation.

Geochronology—ID-TIMS Analysis.

Water-based separation of baddeleyite was attempted for all rock samples at Lund University. Samples NL-13c and G02-B yielded baddeleyite grains using this method. Grains were selected and analyzed using U-Pb ID-TIMS on a Thermo Finnigan Triton Mass Spectrometer at the Department of Geosciences at the Swedish Museum of Natural History. Additional details of the operating procedures for the ID-TIMS analyses at the Swedish Museum of Natural History and results are given in SI Methods, Fig. S2, and Table S4.

Geochronology—SIMS Analysis.

Zr-bearing phases were imaged, mapped using SEM in conjunction with energy-dispersive spectrometry in samples TGS-05 and OLL-2, and then, ranked (SI Methods and Fig. S2). U-Pb isotopic data using SIMS were determined in situ on the microbaddeleyite grains in the mapped thin sections using the CAMECA ims1270 Mass Spectrometer at the University of California, Los Angeles (UCLA). Additional details of the operating procedures for the SIMS at UCLA and results are given in SI Methods, Fig. S2, and Table S5.

Paleomagnetism.

Measurements of magnetic remanence from samples TGS-05 and MDK-05 were made using the superconducting vertical 2G Enterprises DC-4K Rock Magnetometer at the University of Johannesburg (UJ). All specimens were exposed to stepwise alternating-field demagnetization. Additional details of the sampling and operating procedures for the magnetometer at the UJ and results are given in SI Methods, Fig. S3, and Table S6.