Trachytes have fine-grained groundmasses made of alkali feldspar laths and sparse diopside, Ti-magnetite, apatite ± interstitial analcime, and up to ∼25% macrocrysts of alkali feldspar, diopside ± Ti-magnetite, titanite, biotite and very rare kaersutite and ilmenite. Euhedral to subhedral alkali feldspar are the most abundant and the largest macrocrystals (0.5–5 mm; Figure 2). Trachytic texture and agglomerates made of Ti-magnetite, diopside and apatite, or multiple feldspar laths are common. Samples recovered by dredge D16 have abundant round amygdales (0.2–4 mm) filled with calcite and/or analcime and rare apatite microphenocrysts (0.12–0.44 mm).

Dredge D05 trachyandesites have ∼4%–7% macrocrysts of diopside, kaersutite and Ti-magnetite ± plagioclase, rare titanite and biotite. Their fine-grained groundmass has alkali feldspar ± plagioclase, Ti-magnetite, diopside, apatite, and interstitial analcime. Trachytic texture and sparse agglomerates formed by diopside and Ti-magnetite occur locally. Sample RGR1-D15-001 is strikingly different from others and has rare alkali feldspar, Ti-magnetite, and apatite microphenocrysts (0.05–0.25 mm), round amygdales (0.1–1.3 cm) filled with calcite, and a groundmass with alkali feldspar, apatite, acicular biotite and ilmenite.

Basalts from D05 and D08 are highly phyric (25%–30% macrocrysts) with fine-grained groundmass made of plagioclase, diopside, Ti-magnetite, apatite ± olivine pseudomorphs and interstitial analcime. The most abundant macrocrysts are of euhedral zoned diopside (0.25–8 mm) sometimes with colorless cores and brownish rims. A few relict olivine macrocrysts (0.5–4 mm) altered to iddingsite and/or bowlingite are observed, as well as smaller (0.13–0.8 mm) plagioclase and Ti-magnetite macrocrysts and microphenocrysts. Rare resorbed biotite macrocrysts are found in sample RGR1-D05-019, which also has sparse amygdales (0.4–0.8 mm) filled by zeolite.

The analyzed trachybasalt has plagioclase, diopside, Ti-magnetite, and rare olivine pseudomorph and apatite macrocrysts in a fine-grained groundmass made of plagioclase, diopside, Ti-magnetite, interstitial analcime, apatite, and sparse altered crystals which resemble olivine pseudomorphs. Most macrocrysts are of subhedral plagioclase (0.3–2.5 mm) with rounded borders rich in inclusions (Figure 2C). Rare glomerocrysts made of plagioclase, Ti-magnetite and diopside are also observed.

The basanite (RGR1-D08-001, Figure 2D) has ∼30% euhedral macrocrysts and microphenocrysts of olivine pseudomorphs replaced by bowlingite (0.13–2.5 mm) and zoned diopside (0.5–1 mm) in a fine-grained groundmass made mostly of prismatic diopside and interstitial analcime, with smaller amounts of Ti-magnetite, rare interstitial feldspar, and apatite.

The autobreccia (RGR1-D15-013) has subrounded to angular clasts (0.025 mm to >2 cm) of porphyritic trachyte and groundmass made of carbonate, iron oxide/hydroxide and clays. Feldspar macrocryst fragments are also found amid the groundmass. Trachyte clasts have variable grain size, with 0.04–0.2 mm alkali feldspar laths and rare Ti-magnetite (0.01–0.05 mm).

Weathering grade of the samples varies from relatively fresh to heavily altered and no fresh glass was found. However, the primary mineralogy of all studied samples is largely preserved. More detailed descriptions are available in Supplementary File SC.

Despite careful sample selection and preparation for analysis, the obtained LOI values are significant (1.02–9.88 wt%). Sea-water alteration may modify the contents of some major elements such as Si, Mg, Ca and K, as well as large-ion lithophile elements like Rb, Cs, Ba and Sr, which are fluid-mobile (e.g., Hart et al., 1974; Verma, 1981; Verma, 1992). Therefore, this section is more focused on trace-elements which are not as mobile, while major oxide and additional trace element results are presented in Supplementary File SD.

Detailed petrography and mineral chemistry analyses reveal a few differences between the chemical and QAPF modal classification of the studied samples. While the most evolved rocks are classified as phonolites according to diagrams such as the TAS (Le Bas et al., 1986; Figure 3A) and Zr/Ti versus Nb/Y (Pearce, 1996; Figure 3B), they would be classified as trachytes according to the QAPF modal classification of volcanic rocks recommended by the International Union of Geological Sciences (IUGS; Le Maitre et al., 2002), as they are essentially made of alkali feldspar, have less than 10% of analcime and no other feldspathoids. A sample chemically classified as a tephriphonolite also has less than 10% feldspathoids and would be considered a trachyandesite according to the mineralogical classification. In the case of the TAS diagram, this is probably due to the higher K₂O contents (8.57–10.34 wt%) of these samples, which were all recovered from dredges D15 and D16 (Area A2, Figure 1B). Also, sample RGR1-D05-001 is chemically classified as a basalt, but only alkali feldspar was found in its heavily altered groundmass and its mineralogy, with clinopyroxene and kaersutite macrocrysts, is suggestive of a composition closer to that of the studied trachyandesites than to the basalts.

The distribution of alkaline volcanic samples from the WRGR (Fodor et al., 1977; Hoernle et al., 2015; Homrighausen et al., 2018; Hoyer et al., 2022a; this study) on the TAS diagram suggests the possible existence of two magmatic series (Figure 3A). One of them would evolve from picro-basalt to at least trachyte, and the other would be more undersaturated, evolving from basanite to phonolite. However, on the diagram by Pearce (1996), many samples classified as basanites, tephrites and phonotephrites in the TAS diagram are classified as alkali basalts and trachyandesites instead. Therefore, it is possible that at least part of the scatter towards more undersaturated fields in the TAS diagram is due to alteration and enrichment in Na and K.

On the chondrite-normalized rare-earth elements (REE) diagram (Figure 4), four main types of patterns, all concave-upwards, are observed. The first one corresponds to the picro-basalt and trachybasalt, which have relatively low (La/Sm)_N (3.2–4.0) and (La/Yb)_N (∼17.0) compared to other samples. The second one corresponds to the trachyandesites, which are slightly more enriched in light REE, with (La/Sm)_N=4.0–6.1, (Tb/Yb)_N=1.7–2.5 and (La/Yb)_N=23–26. The third one corresponds to trachytes from dredges D08 and D13, which have the highest (La/Sm)_N (5.5–8.5), lowest (Tb/Yb)_N (1.3–1.5), and the most significant negative Eu anomalies (Eu/Eu*=0.80–0.94). The last one refers to the trachytes from dredge D16, which have the greatest (La/Yb)_N (49 and 51). All patterns observed are smooth, and none of the samples show negative Ce anomalies, suggesting that REE contents of the magmas and their sources were not affected by seawater alteration (e.g., Cotten et al., 1995; Price et al., 1991).

On the spidergrams (Figure 4), some differences between the rock groups can be observed. Trachytes generally have stronger Sr and Ti negative anomalies. The trachyandesite patterns are similar to that of the trachytes, but with lower normalized values for the more incompatible elements (with the exception of Ba in two samples), and less intense Sr, P and Ti negative anomalies, although one of them has a stronger Sr anomaly. The most mafic samples are poorer in incompatible elements and have positive Ti anomalies and negative K anomalies.

The results obtained are consistent with the incompatible element patterns, contents, and ratios (e.g., Nb/Zr, Nb/Th, Hf/Th, Nb/La) reported by Hoyer et al. (2022a); Hoyer et al. (2022b); Supplementary Files SD, A, Supplementary Figure S4.

Zircons found in a single trachyte (RGR1-D13-004) were dated by Sensitive High-Resolution Ion MicroProbe (SHRIMP). A total of 22 spots were analyzed in mostly subhedral 130–240 μm long crystals with aspect ratios between 2:1 and 3:1 (Supplementary File SA; Supplementary Figure S7). Sector zoning is observed in almost every crystal and oscillatory zoning is frequent. No inherited cores were observed. Average U and Th contents are 491 and 1,000 ppm respectively (U=168–1007 ppm; Th=187–2,629 ppm; Supplementary File SG). Results yielded a concordia age of 46.9 ± 0.3 Ma (2σ; MSDW=0.00052; probability=0.98; n=22; Figure 13), the first U-Pb age obtained for any WRGR magmatic rock. The result obtained is close to the K-Ar age of 46.3 ± 0.7 Ma (Bryan and Duncan, 1983) and the Ar-Ar age of 46.0 ± 0.1 Ma (Rohde et al., 2013b) reported on the literature for WRGR rocks.

Pb, Sr and Nd isotope ratios of the studied alkaline volcanic rocks, including trachyte RGR1-D13-004, dated in this work at 46.9 ± 0.3 Ma, are similar to results available in the literature (Hoernle et al., 2015; Homrighausen et al., 2023), and differ only for the higher ⁸⁷Sr/⁸⁶Sr ratios of three samples (Figure 14). The results lie along the Tristan compositional trend on the ²⁰⁷Pb/²⁰⁴Pb versus ²⁰⁶Pb/²⁰⁴Pb diagram like both WRGR and ERGR samples analyzed by Homrighausen et al. (2023), but WRGR alkaline rocks generally have less radiogenic signatures. The only analyzed sample dredged from the northern flank of the Cruzeiro do Sul Rift (trachyte RGR1-D16-003, Area A2, Figure 1), has more radiogenic Pb and Nd ratios than the rocks which were dredged from the southern flank (Area A1), and plots over Tristan-type samples from the Walvis Ridge and Guyot Province. This sample also differs from the others for its higher Nb/Zr and Nb/La ratios, even when compared to samples with similar MgO (Supplementary Figure S4).

The EMI-flavored signatures of the analyzed alkaline samples are similar to those from the tholeiitic basement drilled at DSDP Site 516 in terms of Pb ratios but have lower ¹⁴⁴Nd/¹⁴³Nd and higher ⁸⁷Sr/⁸⁶Sr. In contrast, the alkali basalt analyzed by Homrighausen et al. (2018), which was recovered by the same dredge as the phonotephrite dated by Rohde et al. (2013b), has strikingly different Pb, Sr and Nd signatures compared to other samples from the RGR, and plots on the Prevalent Mantle (PREMA) field (Figure 14). It also has higher Nb/La and Nb/Th ratios for a given MgO (Supplementary Figure S4).

The variable textural and compositional features observed in clinopyroxene are suggestive of an open system, as evidenced by complex zoning patterns which include oscillatory zoning, normal abrupt zoning on macrocrysts with resorbed and corroded rounded or irregular-shaped cores, and less frequent reverse zoning, all indicative of disequilibrium between crystals and host magmas (e.g., Cao et al., 2022; Jerram and Martin, 2008; Streck, 2008; Ubide et al., 2014a; 2014b). The normal abrupt zoning patterns observed probably result from reaction between more primitive crystals with host magmas in response to changes in temperature, pressure, and/or composition. The presence of crystals with reverse zoning is indicative that not only more primitive cores were remobilized by more evolved magmas, but also that more evolved crystals interacted with more primitive magmas. Moreover, frequent oscillatory zoning indicates that crystals recorded cyclic compositional variations while stored in magmatic chambers submitted to multiple episodes of more primitive magma recharge, also affected by magma differentiation by fractional crystallization. Crystals with sector zoning are found only in the alkali-basalt and indicate dynamic crystallization conditions.

The disequilibrium textures observed on clinopyroxene primitive cores and their contrasting composition when compared to rims and groundmass crystals indicate that they are not s.s. phenocrysts, i.e., they are not early-formed crystals which grew larger than others crystallized from the same magma. Furthermore, the abrupt zoning patterns are not consistent with subsolidus diffusive processes. These cores are probably antecrysts crystallized from magmas which were more primitive than their host liquid, and it is possible that some correspond to mantle xenocrysts, considering the partial overlap between the cores with the highest Mg# (>85) and compositional fields of clinopyroxene from mantle xenoliths (Figure 9). Many feldspar macrocryst cores might also be antecrysts. This is probably the case of the plagioclase cores from macrocrysts with alkali feldspar rims from trachyte RGR1-D08-003, of the resorbed bytownite cores found in the strongly zoned picro-basalt macrocrysts with labradorite rims, and of the plagioclase macrocrysts found in the trachybasalt, which are strongly corroded and rounded.

The infiltration textures commonly observed in clinopyroxene evidence the effect of the cryptic metasomatic processes to which they were submitted along their evolution, possibly in multiple stages, as suggested by the complex patterns observed in the Cr₂O₃ compositional maps of primitive cores (Figure 7). A few of the point-analyses made in these cores yielded the highest Cr contents of the dataset, what suggests that the superimposed cryptic metasomatism is partially masking the possible mantle-derived nature of xenocrystic cores.

Most of the groundmass clinopyroxene and feldspar, which represent quench low-pressure crystals, have similar composition to that of macrocryst rims from the host sample. This suggests that outer rims from macrocrysts frequently represent low-pressure overgrowths on host higher-pressure crystals. The only sample which has clinopyroxene groundmass crystals and macrocryst rims with significant compositional differences is trachyte RGR1-D13-004. The higher Mg# and Ti/Al and lower Na₂O from the groundmass crystals suggest that the clinopyroxene macrocrysts from this sample are also antecrysts.

Despite the multiple evidence of disequilibrium between more primitive macrocryst cores and their respective host liquids, a few euhedral clinopyroxene crystals with concentric cores and rims of contrasting composition are observed. This is indicative that in some cases there was progressive equilibrium of the primitive crystals with the magma which generated the rims with lower Mg#.

To further evaluate the equilibrium between crystals and host magmas from a compositional point of view, clinopyroxene from samples with lower proportions of macrocrysts were compared to the results of whole-rock geochemistry analyses using the Rhodes diagram (Rhodes et al., 1979) and predicted and observed clinopyroxene components (DiHd, EnFs, CaTs; Putirka et al., 1996; Putirka, 1999) Additionally, the composition of the very fine-grained groundmass of the picro-basalt (RGR1-D08-004) was estimated using a semi-quantitative EDS area analysis to further investigate this sample, which has the clinopyroxene with the strongest abrupt zoning patterns and the most primitive cores.

Most of the analyzed clinopyroxene are in disequilibrium with their respective whole-rock composition, which is probably biased even by smaller proportions of macrocrysts (Supplementary File SA, Supplementary Figure S8). Trachyte RGR1-D08-002 is the only sample with a few clinopyroxene crystals in equilibrium with the whole-rock, but most of its crystals plot below the equilibrium curve in the Rhodes diagram, suggesting that they were crystallized from a host liquid with lower Mg#. For these clinopyroxene-melt pairs, temperatures of 1023ºC–1040°C and pressures of 3.1–3.8 kbar were estimated using the thermobarometer from Masotta et al. (2013), which was calibrated for trachytic and phonolitic compositions and has standard errors of estimate (SEE) of 1.15 kbar and 18.2°C. A single ilmenite and Ti-magnetite pair found in the same trachyte was used to estimate temperature and oxygen fugacity using ILMAT (Lepage et al., 2003) and yielded temperatures of 826ºC–839°C and log₁₀fO₂ between −14.1 and −13.6 using the thermobarometer by Andersen and Lindsley (1985), indicating relatively high oxygen fugacity, close to the NiNiO redox buffer.

In the case of the picro-basalt, 24 clinopyroxene-melt pairs are in apparent equilibrium with the estimated groundmass composition. Among these, there are mostly point-analyses from evolved rims, one made on rare homogeneous macrocrysts with no zoning, a few analyses of infiltrated portions on the macrocrysts, and a single groundmass crystal. The remaining 4 point-analyses on groundmass crystals yielded K_DFe-Mg values which are suggestive of equilibrium with the melt estimative but did not satisfy the equilibrium criteria considering predicted and observed clinopyroxene components, probably due to the uncertainty associated with the estimated host liquid composition. The most primitive clinopyroxene cores would be in equilibrium with melts with significantly higher Mg# (>70).

Estimates for the picro-basalt clinopyroxene-melt pairs were made using the thermometer from Putirka et al. (2003) and Equation 32b barometer from Putirka (2008), which have SEEs of 59°C and 2.6 kbar. These were selected based on the work by Hammer et al. (2016), which evaluates the best models to estimate temperature and pressure of high-Al diopsides from Hawaiian ankaramites. The average temperature obtained is 1144°C. Results from the thermometer from Putirka et al. (2003) agree within error with estimates using Equation 33 from Putirka (2008) iteratively calculated with the barometer from Neave and Putirka (2017). Pressure estimates depend on the H₂O content of the liquid in equilibrium with clinopyroxene, which is unknown; thus, they were made considering an anhydrous composition and H₂O contents=0.5%, 1%, and 2%. The results are variable (averages between 2.2 and 3.1 kbar with increasing H₂O) but agree within error (Supplementary File SE). Despite the limitations, the calculations constrain the pressure to an interval between 0 and 5.7 kbar, considering the SEE.

Even though it is not possible to use thermobarometers to estimate pressure conditions for the crystallization of other macrocrysts, qualitative aspects can be assessed based on clinopyroxene composition, especially in the case of the strongly zoned clinopyroxene from the mafic rocks. For instance, the Al/Ti ratio increases significantly with pressure (e.g., Dobosi and Fodor, 1992; Thy, 1991). Trends towards higher Al^VI/Al^IV are also indicative of an increase in pressure crystallization conditions (e.g., Aoki and Kushiro, 1968; Dobosi and Fodor, 1992; Wass, 1979). The Al^VI/Al^IV value depends on the magma composition, but the trend of increasing Al^VI/Al^IV ratios with increasing pressure is observed independent of melt compositions and is not restricted to alkali basalts (Wass, 1979). Other trends observed in clinopyroxene composition with decreasing pressure are the increase in Ti+Al^IV with decreasing Si and increase in TiO₂ with decreasing Mg# (Wass, 1979).

Clinopyroxene macrocrysts and groundmass crystals from the picro-basalt and alkali basalt have similar composition and Al/Ti and Al^VI/Al^IV trends (Figure 15). Cores generally follow a trend of higher Al/Ti and Al^VI/Al^IV ratios than most analyzed rims, except for analyses made on the inner parts of rims with oscillatory zoning. Such trends suggest that cores and part of the oscillatory-zoned rims were formed under higher pressure in deeper parts of the magmatic system. This is also supported by a general trend of increasing Al^IV/Al^VI and Ti and decreasing Si with decreasing Mg# from cores to groundmass and rims without oscillatory zoning (Wass, 1979). Furthermore, macrocrysts with reverse zoning and microphenocrysts have similar Al/Ti and Al^VI/Al^IV to those of rims from macrocrysts with normal zoning, which suggests that they were formed under similar pressure conditions, in shallower depths when compared to the most primitive cores.

In contrast, cores and rims of crystals with normal abrupt zoning from the basanite have more similar Al^IV/Al^VI ratios and Al/Ti ratios (Figure 15), which are also close to those from the primitive intermediate zones of the mapped microphenocrysts (Figures 7C, D), suggesting that they were crystallized at similar depths. Although Al/Ti ratios from cores are slightly higher when compared to rims, the trend observed suggests that the possible difference in pressure conditions is not as significant as in the case of the basalts. The more evolved cores from the mapped basanite microphenocrysts have higher Al^VI/Al^IV and core #1 has much higher Al/Ti ratios than any other crystal. This evolved core might have been formed under high pressure conditions before it was remobilized by a more primitive magma, which indicates that interaction between more evolved macrocrysts and primitive magmas occurred under variable pressure.

Recent investigations of the RGR provided new constraints on the isotope compositional variability of its volcanic rocks. Perhaps the most relevant finding of this study is that trachyte sample RGR1-D13-004, dated at 46.9 ± 0.3 Ma, has a Tristan-type isotope signature like most other RGR alkaline lavas (this study; Hoernle et al., 2015; Homrighausen et al., 2023). Such signatures were previously found only in Eocene volcanic rocks on the African Plate (Rohde et al., 2013a; b; O’Connor and Jokat, 2015; Homrighausen et al., 2019). Sample RC11-2RD (analyzed by Homrighausen et al., 2018), which was recovered by the same dredge as the 46.0 ± 0.1 Ma phonotephrite dated by Rohde et al. (2013b), has contrasting isotope signatures and differs from other WRGR alkaline volcanics (Hoernle et al., 2015; Homrighausen et al., 2023; this work) for its considerably higher ²⁰⁶Pb/²⁰⁴Pb and ¹⁴⁴Nd/¹⁴³Nd and lower ⁸⁷Sr/⁸⁶Sr. This observed variability might be due to small-scale source heterogeneities.

In the uranogenic diagram in Figure 14, there seems to be a gap in the middle of the Tristan-type trend between samples from the Walvis Ridge and Guyot Province and most samples from the RGR, which are less radiogenic. Only two WRGR alkaline samples overlap with lavas from the Walvis Ridge and Guyot Province in this diagram, and these samples have higher Nb/Zr ratios compared to the others. This might also be a result of source heterogeneity, but the gap itself could also reflect the limited sampling of the plateau.

Hoyer et al. (2022a) suggested that the Eocene alkaline volcanism from the Walvis Ridge/Guyot Province and the Rio Grande Rise was not originated by the same processes because by ∼70 Ma they were already separated from each other according to paleogeographic reconstructions (Graça et al., 2019). The authors also mention the different isotope signatures of the single WRGR sample analyzed by Homrighausen et al. (2018) and late-stage Walvis Ridge lavas (Homrighausen et al., 2019). The WRGR alkaline volcanism might actually be related to lithospheric rifting along the Cruzeiro do Sul Rift as suggested by Hoyer et al. (2022a) and Mohriak et al. (2010). However, the compositional similarity between Walvis Ridge, Guyot Province and RGR lavas supported by more recent data (Homrighausen et al., 2023; this study), strongly suggests that at least the enrichment of the sources from their magmatism is related to the same processes. Moreover, even after the separation of the RGR and the Walvis Ridge, these sources still shared common characteristics.

Homrighausen et al. (2023) point out that the EMI signatures with the lowest ²⁰⁶Pb/²⁰⁴Pb ratios are found in the oldest Gough-type samples from the Walvis Ridge and the oldest Tristan-type samples from the Rio Grande Rise and conclude that both Gough and Tristan-type lavas follow an evolution trend towards PREMA compositions with time. However, until now, there is no evidence of alkaline volcanism in the plateau before the Eocene, and our work shows that signatures of Eocene alkaline rocks from the WRGR have low ²⁰⁶Pb/²⁰⁴Pb values close to those of the 80–88 Ma tholeiitic basement drilled at DSDP Site 516F. More geochronology investigations are essential to better understand the temporal distribution of the plateau’s enriched signatures, but it seems that in the case of the WRGR, there is no trend of evolution towards higher ²⁰⁶Pb/²⁰⁴Pb ratios with time.

The origin of the EMI component and the Tristan/Gough chemical zonation remains unclear, and the results obtained are not supportive of a specific model, but they have important implications as any attempts to explain such signatures in South Atlantic lavas need to consider that during the Eocene they were still manifested in both South American and African plates. According to the model proposed by O’Connor et al. (2018), it is possible that supply of enriched material from the Tristan-Gough plume persisted through time because of lateral fast flows caused by the horizontal deflection of upwelling material from the African superplume in the asthenosphere beneath the RGR. However, more investigations are needed to understand the mechanisms behind the manifestation of Tristan and Gough-type signatures, and new studies of the RGR have significant potential to shed light on this matter. Not only the complex spatial and temporal distribution of EMI-flavored lavas on the plateau needs to be better constrained, but also the duration and volume of its alkaline versus tholeiitic magmatism, as well as the spatial relationship between their volcanic products.