The elemental and isotopic chemistry of zircon from CAMP magmas varies within and between samples. The significance of these variations for understanding the origin and the conditions for zircon crystallization in these magmas are explored below.
When does zircon form in LIP magmas and under what conditions?
Zircon (ZrSiO4), baddeleyite (ZrO2), and zirconolite (CaZrTi2O7) are found in interstitial melt pockets within CAMP samples (Fig. 6). These melt pockets are typically at the scale of ~ 100 μm, although in the NMB, they can reach the 10 cm scale (Kontak and Dostal 2010). They contain K-feldspar, quartz, apatite, plagioclase, ilmenite, and minor sulfides (that are likely secondary). This mineral assemblage suggests highly fractionated SiO2-rich melts and is similar to the mineralogy of melt pockets previously identified in CAMP, and other LIPs (Kontak et al. 2002; Kontak and Dostal 2010; Ver Hoeve et al. 2018; Heimdal et al. 2019). The highly variable and enriched trace-element concentrations of the zircons are also consistent with crystallization in highly fractionated, silica-rich melts (Fig. 5). It is clear that none of the Zr-bearing minerals crystallized with the main fractionating assemblage of pyroxene, olivine, and plagioclase, which is not surprising, since zircon saturation conditions in basaltic (tholeiitic) magmas at temperatures > 1000 °C require > 7000 µg/g Zr (Boehnke et al. 2013), which is about 50 times more than the typical content of CAMP basalts. Also, the Ti temperatures of the zircon are mostly around 800 °C suggesting much lower crystallization temperatures (Fig. 4). For a more precise understanding of the zircon saturation conditions, the composition of potential primary magmas can be used (see, e.g., Callegaro et al. 2013; Merle et al. 2014; Whalen et al. 2015; Marzoli et al. 2019). If we use primary CAMP sample AN133 from Marzoli et al. (2019) as an estimate for a primary low Ti magma, with a Zr concentration of 119 µg/g, the zircon saturation conditions can be calculated based on the degree of polymerization of the melt, which can be determined based on the cation ratio (Na + K + 2Ca)/(Al × Si), referred to as M (Watson and Harrison 1983; Boehnke et al. 2013). Higher values of M indicate low degrees of polymerization and enhanced zircon solubility. The AN133 composition gives an M value of 2.77 resulting in an unrealistically low temperature of zircon saturation of 578 °C, which is far below the wet granite solidus ~ 700 °C (e.g., Ebadi and Johannes 1991), clearly indicating that zircon will not crystallize without significant modification through crystallization or assimilation. Conversely, this also means that antecrystic zircon will be readily dissolved in these melts, and therefore, old zircon ages should not normally be attributed to antecrysts.
The zircon saturation conditions in CAMP magmas can be refined through forward modeling using rhyolite-MELTS (version 1.2.x, Gualda et al. 2012; Ghiorso and Gualda 2015), and monitoring the changing composition of the magma as it undergoes fractional crystallization from ~ 1200 °C to ~ 700 °C at 1.5 kbar pressure, which is ~ 4 km depth. The Zr concentration of the melt can be calculated at each step using the mineral phase proportions along with mineral/melt Zr partition coefficients for each phase present (the partition coefficients used are reported in the Supplementary data). The Zr concentration in the melt is then compared to the zircon saturation conditions at every step, calculated using the equations of Boehnke et al. (2013). Where the Zr concentration and the zircon saturation conditions coincide is the likely point of zircon saturation (Fig. 7a). In Fig. 7, the model used sample AN133 as a starting composition (low Ti basalt; Marzoli et al. 2019) and different initial H2O concentrations to show the effect of initial water on the zircon saturation conditions. Note that the composition of the liquid from which the zircon crystallizes is granitic with ~ 73% SiO2 for each of the initial H2O concentrations (see TAS diagram showing the composition of the melt at zircon saturation in the supplementary information). When using the high Ti CAMP sample M13 from Merle et al. (2014) as starting material, zircon saturation was not achieved under any conditions, implying that zircon does not saturate in such a melt, unless they are significantly contaminated (see next paragraph). This is also the case for CS49, which is a low Ti primary CAMP basalt from South-Eastern USA (Callegaro et al. 2013).
It is clear from Fig. 7a that the initial H2O concentration of the melt is very important in determining when zircon saturates, this is because H2O suppresses silicate saturation (also see TAS diagram in supplementary information). With an initial H2O concentration of 0.5 wt. % (similar to the initial water concentrations estimated in Callegaro et al. 2013), the magma becomes zircon saturated at ~ 790 °C, after 85% crystallization. Increasing the initial H2O concentration of the melt decreases the needed amount of fractional crystallization and also the temperature for zircon saturation. Based on our modeling, zircon saturation is reached after 82% fractional crystallization at ~ 750 °C for an initial H2O concentration of 0.8 wt. % and at ~ 740 °C after 80% fractional crystallization with an initial H2O concentration of 1.5 wt. %. Comparing these modeled zircon saturation conditions with the newly determined zircon crystallization (Ti) temperatures of ~ 800 ± 50 °C based on the Ti in zircon thermometer (Fig. 5c) suggests that the 0.5% initial H2O model provides the most consistent saturation condition estimates. It should be noted that the zircon crystallization temperatures were calculated assuming a silica activity of aSiO2=1 and a Ti activity of aTiO2 = 0.6 both of which have an effect on the crystallization temperatures calculated. Reducing the silica activity or increasing the activity of Ti (see discussion on the NMB data above) results in a reduction in calculated zircon temperature, which would be more in agreement with higher initial H2O. Zircon crystallization under equilibrium conditions is also assumed, which may not be the case in these fast cooling, fractionated melt pockets, if kinetic effects (e.g., Albarède and Bottinga 1972) play a role in the trace-element compositions of the zircon. This has recently been theoretically demonstrated for Zr isotopes in zircon (Chen et al. 2020; Méheut et al. 2021), and hence, the Ti concentrations in our zircon may not only be dependent on temperature. High initial H2O concentrations have not been previously proposed for the CAMP source rocks, which is in agreement with the typically anhydrous mineralogy of CAMP tholeiites, which contain olivine, pyroxene plagioclase, and oxides without significant amphibole (although amphibole and biotite may be present in some of the evolved melt pockets, Heimdal et al. 2019). In a recent study, however, Capriolo et al. (2020) identified deeply sourced CO2-rich bubbles trapped in melt inclusions in CAMP minerals, and melt inclusions associated with these gas bubbles were also found to contain ~ 1 wt % H2O, suggesting that the source for CAMP melts likely had some initial volatile component.
Crustal contamination
Crustal contamination is also thought to play a role in the generation of the CAMP magmas, and this is also likely to affect the zircon saturation conditions, since it will affect the M parameter, and the Zr concentration of the melt. The changes in saturation conditions may even result in xenocrystic zircon being preserved. Examples of CAMP rocks with xenocrystic zircon are the Orange Mountain basalt (one of the CAMP basalts erupted in NE USA; Blackburn et al. 2013; also see NMB discussion below) along with a dyke from Morocco (sample AN733; Davies, unpublished data).
The amount of crustal contamination in the CAMP magmas is thought to be ≤ 10% based on Os isotope modeling (see Merle et al. 2011, 2014; Callegaro et al. 2014, 2017; Marzoli et al. 2018). To assess the impact of this on the zircon saturation conditions, we modeled the effects of assimilation of various crustal rocks, i.e., shale, lower continental crust (LCC), granite, and sandstone (SST; Fig. 7b) (the compositions of the assimilants are given in the supplementary data). The assimilation was simulated using the magma chamber simulator (Bohrson et al. 2014, 2020; Heinonen et al. 2019, 2020), which is a thermodynamic model that calculates the evolution of a composite magmatic system and its fractionally crystallizing minerals. It also contains a number of sub-systems that control the thermodynamics and chemistry of magma recharge, assimilation, stoping, and accumulation of cumulates. The magma chamber simulator uses rhyolite-MELTS to compute phase equilibria, and all of the different sub-systems thermodynamically interact, e.g., the intrusion of magma heats up the wall rock, possibly causing melting which mixes with the melt (see Bohrson et al. 2014, 2020 for a detailed explanation of the model). Here, the magma chamber simulator was used to constrain the effects of assimilating via stoping, different possible contaminants into CAMP primitive magmas, and determine the associated effects on zircon saturation. The bulk composition of the assimilated material was added to the CAMP magma as a liquid after a certain amount of fractionation (the stoping method, see Bohrson et al. 2020), using the magma recharge function of the magma chamber simulator. All assimilation models were run using AN133 as the starting composition to represent a primary CAMP melt, with 0.5% initial H2O. Also, most CAMP samples are not primary melts, and represent liquids after various degrees of fractional crystallization (10–50%) at low pressure (Marzoli et al. 2018). Therefore, the modeled assimilation proportions were 10 or 5% and assimilation occurred after 40% fractional crystallization. For reference, the effects of assimilating after 55 and 60% fractional crystallization are also shown for granite (Fig. 7b). To model the effects on zircon saturation, the Zr concentration of the melt needs to be known and this was calculated using partition coefficients (given in the supplementary information) between the melt and minerals crystallizing at any time step. It is clear from the modeling results that assimilation has a smaller impact on the zircon saturation conditions than increasing the initial H2O (note in 7a the position of the zoom shown in 7b), the exception is assimilating shale, which, in our example, had 5% H2O. Only shale assimilation had a significant effect on the zircon saturation conditions, and resulted in zircon saturation being reached after 79% or 75% fractional crystallization for the 5 and 10% assimilation models respectively. The M value for shale assimilation is also the lowest for all of the models (2 and 1.9 for the 5% and 10% assimilation models, respectively, whereas all other assimilation models have M values ~ 2.2), which explains the drastic change in saturation conditions. Other primary CAMP compositions, for example, CS49 (Callegaro et al. 2013), only reach zircon saturation after 10% assimilation of shale, and even then, it requires 91% fractional crystallization with an M value of 2.7 at saturation, indicating that the melt only just reaches zircon saturation. High Ti CAMP samples (M13, from Merle et al. 2011), which have high Zr, but are geochemically more depleted than the low Ti CAMP samples, reach zircon saturation only after > 15% assimilation of shale. Zircon saturation in these samples is reached after 82% fractional crystallization, with an M value of 2.32 (see insert in Fig. 7b).
Overall, most assimilants do not affect the zircon saturation conditions too much if the melt will saturate zircon on its own without assimilation. The saturation conditions with 0.5% initial H2O and some degree of assimilation are consistent with the thin-section evidence of zircon in highly fractionated melt pockets (Fig. 6), high trace-element contents in the CAMP zircon, and also the Ti temperatures, suggesting that the modeling may reflect real petrological processes (Fig. 5). Many primary CAMP melt compositions will not saturate zircon without significant contamination, and shale seems to be the contaminant that most easily creates the conditions for zircon saturation. The exact nature of the contaminant for each CAMP sample is likely to be different, since the samples cover the entire province (i.e., over ~ 8000 km distance from the NMB in Canada to the Tarabuco sill in Bolivia). However, it is clear that assimilation (especially of shale) played a role in enabling the magmas to reach zircon saturation. Also, it is crucial (for geochronology) for magmas to reach zircon saturation early, so that zircon has longer time to form, and larger crystals are more likely; only the assimilation of shale causes zircon saturation to occur significantly earlier, therefore, these are the cases where dateable zircon is more likely to be present. Assimilation after > 40% fractional crystallization also causes zircon saturation to occur earlier (see darker blue stars in Fig. 7b) promoting zircon growth.
What can the isotopic compositions of zircon tell us about the contamination processes present in CAMP melts?
To use the O isotopic data to help understand petrologic processes during CAMP magmatism, the zircon δ18O values need to be corrected for the effects of temperature-dependent fractionation, and converted to basalt values assuming equilibrium isotope fractionation between basalt and zircon. This is a large assumption, since the magma that the zircon crystallized from was not basaltic, and was likely modified by upper crustal contamination (Figs. 5, 7). However, given these caveats, δ18O values in zircon that are different from the mantle value after the effects of equilibrium isotopic fractionation between basalt and zircon have been accounted for, may be used, in combination with the εHf values and other available data, to understand the magmatic processes that occurred during emplacement. Most of the CAMP samples contain a homogenous population of zircon δ18O values, with a few outliers, mostly at higher values.
To calculate the equilibrium isotope fractionation between basalt and zircon, the fractionation factor needs to be known. Experiments to determine the equilibrium zircon-WR (whole rock) oxygen isotope factors have been attempted (e.g., Trail et al. 2009), along with empirical observations of a general relationship between WR δ18O value and that of zircon (e.g., Lackey et al. 2008). However, there currently is no consensus on the correct fractionation factors to use. We applied the 1000 × ln(αZrc-Fo) = 0.4 × 106/T2 from the first principals’ calculations of (Kieffer 1982) combined with 1000 × ln(αbasalt-Fo) = 1.4 × 106/T2 from Eiler (2001), to obtain 1000 × ln(αbasalt-Zrc) = 1.0 × 106/T2, with α being the fractionation factor and T being temperature in Kelvin. Using our equilibrium isotope fractionation factor, combined with the average Ti temperature determined for each suite of zircon crystals, we can determine a δ18O value for the basalt in equilibrium with the zircon. These basaltic values are shown in Fig. 8a along with a range of possible basalt oxygen isotopic compositions that could potentially be produced through fractional crystallization of a tholeiitic basalt at low pressure (after Bucholz et al. 2017). It is clear that some of the calculated basalt δ18O values are higher than would be expected through fractional crystallization of a primitive basalt alone; therefore, they are likely recording some other processes. Also, as expected, the higher δ18O values from some of the grains result in extremely high basaltic values of 7.5–8 ‰.
There are numerous models for the source of the CAMP basalts, and most of these involve mixtures of multiple mantle components and mixing with a small amount of either subducted material or lower continental crust in the case of the low Ti samples, or enriched metasomatic veins in the subcontinental lithospheric mantle in the case of the high Ti samples (Pegram 1990; Puffer 2001; Dorais and Tubrett 2008; Merle et al. 2011, 2014; Callegaro et al. 2013, 2014; Whalen et al. 2015; Marzoli et al. 2018). The different mantle end members are unlikely to explain the δ18O values seen here (Fig. 8a); also, melts would need to contain up to 100% lower continental crustal material to explain the δ18O enrichments, since the proportion of oxygen between mafic mantle-derived melts and lower continental crust is approximately equal. The most likely explanation for the high δ18O values recorded in the CAMP zircon is contamination from high δ18O upper crustal sediments during emplacement, which is also in agreement with the zircon saturation modeling (Fig. 7) and NMB εHf data (Fig. 9).
A simple mixing relationship between a mantle-derived basalt with an εHf value of + 3.5 (for the low Ti CAMP samples; Elkins et al. 2020), a Hf concentration of 2.88 µg/g (an average of Moroccan CAMP samples, Marzoli et al. 2019), and a δ18O value of 5.8 ‰ (to reflect a small amount of enriched component in the source), with upper crustal sediments, for example shales, that have εHf of − 10 with 6 µg/g Hf (e.g., Bayon et al. 2006), and a δ18O value of 16 (Bindeman et al. 2016), can easily explain the variation seen in zircon data. This simple mixing model plotted with 2% mixing increments (steep mixing lines in Fig. 8b) shows that only 10–20% of a sedimentary shale component is required to explain the most positive δ18O value of ~ 8‰ found in the Amelal sill. However, it is extremely unlikely that a single crustal rock is the contaminant for all of the CAMP magmas. Therefore, the mixing calculations shown in Fig. 8 involve various hypothetical sediment compositions and have a range in Hf concentration and εHf values which covers the entire range of values seen in the zircon data. These hypothetical sediments cover compositional ranges of sediments derived from mafic to felsic rocks, although all have elevated δ18O values (see compilations in Vervoort et al. 2000; Pettke et al. 2002; Bindeman 2008; Rickli et al. 2013; Bindeman et al. 2014, 2016; Greber et al. 2017). Also shown is a mixing curve with a hypothetical granitic sample with 7 µg/g Hf, εHf of -10, and δ18O of 8‰, with mixing increments of 2%, and up to a maximum of 50%. It is clear that contaminants with low δ18O values cannot explain the elevated δ18O values of zircon. It should be noted that the CAMP Hf and O end members chosen here do not explain all of the zircon data; for example, the NMB data require a CAMP source with lower εHf. The overall conclusion is that some amount of crustal contamination is required to explain the variable δ18O values in zircon.
This conclusion is corroborated by other independent proxies. For example, samples that show high δ18O values, notably the Kakoulima and Fouta Djalon intrusions and the Amelal sill, have whole rock and Nd, Sr, and Pb isotopic compositions that suggest significant crustal assimilation (up to 30%) (Deckart et al. 2005). The Amelal sill intrudes lacustrine silts (Marzoli et al. 2019) and the NMB has direct zircon evidence of sediment contamination (Fig. 9). The zircon δ18O data for the other samples do not require sediment assimilation to explain their isotopic composition, since they can be explained through high degrees of fractional crystallization of basaltic melt. However, it is clear that zircon saturation can be promoted by minor amounts of crustal contamination and this may not change the O-Hf isotopic compositions significantly. The high Ti CAMP samples do not require assimilation to explain their oxygen and hafnium isotope data. The more positive εHf values found for the high Ti sample can be explained using the same mechanisms as those described in Merle et al. (2014) and Callegaro et al. (2017), where a depleted mantle source is mixed with a small amount of highly enriched mantle melt, for example a lamproite (see horizontal mixing line on Fig. 8). However, as discussed above, contamination is likely required to explain the presence of zircon in these samples.
In conclusion, we suggest that late-stage addition of sediments into the melts can create the conditions for zircon saturation to occur and, in some cases, also explains their isotopic compositions. Upper crustal contamination is also consistent with bulk rock radiogenic isotope data. Therefore, zircon isotopic and elemental compositions from LIP magmas can help to understand the conditions in which these crystals grew, but such data are not well suited for determining the source of the melts, or their tectonic setting.
Why does zircon in the North Mountain basalt have such variable ages, trace element, and isotopic compositions?
Zircons from the NMB display more variable ages, Hf and O isotopic compositions, and higher Ti temperatures than zircon from all the other CAMP samples (Figs. 4, 9). The NMB has been dated by high-precision U–Pb techniques in multiple studies (Schoene et al. 2010a; Blackburn et al. 2013; Davies et al. 2017, and the current work), and the recommended eruption age of 201.498 ± 0.028 Ma is an average of these studies. New data obtained here (Fig. 9) including some analyses from Davies et al. (2017) suggest zircon crystallization over ~ 2.5 Ma, which seems highly unlikely given the zircon undersaturated nature of the CAMP basalts. The simplest explanation for the old ages is the presence of xenocrystic cores. Truly inherited grains are unlikely since the ages of the zircon from sediments the NMB intrudes and erupts onto are Carboniferous and older (> 360 Ma, Marzoli et al. 2017). However, it is possible for xenocrystic zircon to survive if they were incorporated just before emplacement and the magma was not able to fully resorb them before cooling below the solidus (e.g., Tang et al. 2014).
To identify potential xenocrystic cores in the NMB zircon, cathodoluminescence (CL) imaging was conducted (Fig. 6; Supplementary data); however, no discernable xenocrystic cores within the grains were detected despite imaging of > 100 grains. Many of the zircons contain melt channels, whereas most have relatively monotonous CL response without significant oscillatory or sector zoning, a typical feature of zircon from mafic magmas (e.g., Rioux et al. 2012; Grimes et al. 2013). Hafnium isotopic compositions can also help to identify xenocrystic zircons. For the NMB, most of the zircon with ages overlapping with the time of the eruption have εHf values of ca. − 0.5 (Fig. 9), similar to other low Ti CAMP magmas (Figs. 3, 4). The older grains have slightly more negative εHf values (around − 1.5 εHf), although this shift is minor and within error of the younger grains. These more negative values at older ages could be explained by the incorporation of tiny (referred to as nano) xenocrystic cores into the grains. The oldest zircon (204.2 ± 0.1 Ma) can be explained by 3% mixing between a xenocrystic core of 360 Ma with εHf − 16 (based on zircon data from similar sediments in the area (Marzoli et al. 2017; Waldron et al. 2019) and NMB magmatic zircon using the age of the eruption and an εHf − 0.5. The mixing trend is very steep with very small changes in the εHf value. Using this mixing proportion, further constraints can be placed on the hypothetical xenocrystic cores based on the 9 pg of radiogenic Pb present within the grain. Only 0.3 pg of (old) Pb is required from the xenocrystic core, suggesting that it was likely extremely small (potentially ≤ ~ 1 μm diameter), assuming that the Hf concentrations are approximately equal between the nano-xenocryst and magmatic zircon. The other grains with ages older than the eruption with similar εHf values require even smaller xenocrystic cores to explain their age and εHf values. There are much older zircon grains in the sediments under the NMB; however, mixing with these older grains cannot explain both the shift toward more negative εHf and the age shift seen in the NMB zircon. Given the fact that the Ti concentrations in the NMB zircon indicate an unusually high zircon crystallization temperature of ~ 900 °C, we suggest a scenario where the sandstones and shales underlying the NMB were entrained into the basaltic melt, quickly dissolved (e.g., Hansen and Nielsen 1999; Heimdal et al. 2019), and locally increased its zircon saturation temperature (see Fig. 7b, where assimilation of sandstone increases the zircon saturation temperature). After dissolution of almost all of the xenocrystic zircon, the last nano-scale zircon cores became the nucleus for magmatic zircon growth. These tiny cores are also unlikely to be present in the CL images due to their small size, resulting in a low chance of exposure during polishing of a grain mount.
There are also two grains with ages close to the eruption age that have εHf values of ~ + 4 and + 9 that cannot be explained by the presence of nano-xenocrystic cores (Fig. 9). It is impossible to add enough juvenile Hf from a xenocrystic core to change the εHf value of a CAMP zircon from -0.5 to an εHf value of + 4 or + 9 without significantly affecting the age. Hafnium isotope variability not associated with xenocrystic zircon is commonly observed in felsic I- and S-type magmas (e.g., Kemp et al. 2007; Villaros et al. 2012; Tang et al. 2014), and multiple potential explanations for this phenomenon have been proposed, including magma mixing (Appleby et al. 2010), antecrystic zircon (Annen et al. 2015), or disequilibrium melting (Tang et al. 2014). However, in mafic systems such as the NMB, the zircon saturation conditions should not allow the preservation of antecrystic or xenocrystic zircon, since they should readily dissolve in the magma (Fig. 7). Isotopic heterogeneities resulting from mixing of magmas would likely be small in low viscosity mafic melts. Therefore, a different justification is needed to explain these data. An alternative model has been suggested by Farina et al. (2014) and Tang et al. (2014) where xenocrystic zircons dissolve in a melt to produce local regions with Hf isotopic compositions that are a mixture between the original melt and the isotopic composition of the dissolved zircon (the composition is skewed toward that of the Hf rich dissolved zircon). The diffusion of Zr in a melt is relatively slow (Zhang 2010), and therefore, newly formed zircon in these local regions use the Zr from the dissolved grain (without retaining any of the radiogenic Pb) and also obtain a Hf isotopic composition close to that of the xenocryst rather than that of the primary melt. In the case of the NMB, a large portion of the zircons in the sediment through which the basalt intrudes had juvenile Hf (see Waldron et al. 2019) associated with island arc magmatism from terrains accreted onto Laurentia during the Appalachian orogen. The dissolution processes described by (Farina et al. 2014) could succinctly explain the variation toward positive εHf values without affecting the zircon age. This process was likely minor in the NMB magma since only two grains apparently record anomalous εHf values.
Magmatic processes such as the presence of isolated, chemically distinct magma batches, and the potential preservation of xenocrystic cores, are often found for zircon from felsic magmatic systems (Wotzlaw et al. 2014; Samperton et al. 2015; Schaltegger and Davies 2017). The data from the NMB suggest that these are also occurring in mafic magmas. However, the scale of these heterogeneous domains is likely much smaller (as indicated by the melt pockets in Fig. 6), especially since zircon is crystallizing so late (see Fig. 7).
Interpretation of LIP zircon U–Pb data
All of the data presented here indicate that the CAMP magmas were generally highly zircon undersaturated, and those that could crystallize zircon without help from assimilation required > 80% fractional crystallization before they reached zircon saturation, usually at a temperature of ~ 800 °C. Other CAMP samples required significant upper crustal assimilation to reach zircon saturation, and even then, this only occurred after similar amounts of fractional crystallization. This suggests that antecrystic zircon is highly unlikely to be present in CAMP (and LIP) samples, and U–Pb ages from LIP zircon should therefore reflect the final stages of crystallization, and be very close to the age of emplacement (or eruption). Following this logic, when a U–Pb dataset from LIP zircon produces an age distribution (not all of the ages overlap—see Fig. 2), the oldest ages in the distribution should be the closest to eruption or emplacement (given the above zircon saturation constraints). The younger portion of the distribution may be more likely to be influenced by Pb loss, especially given the extremely high U contents that can be present in some LIP zircon (Fig. 5b). This interpretation can be tested using the CAMP U–Pb ages produced here and a simple thermal model of a CAMP sill (the thermal model parameters used to solve the one-dimensional heat conduction equation are in the supplementary information). Samples RP136 and RP134 both have zircon age ranges of ~ 160 ka (not including the much younger zircon from RP134). If a cooling CAMP sill stays within the zircon crystallization window for ~ 160 ka (we define the crystallization window as between 900 and 700 °C), then the zircon age range reflects crystallization; if the sill cools much quicker than this, it is likely that the younger ages reflect small amounts of Pb loss. Figure 10 shows two cooling paths for 150 m-thick CAMP sills in the Amazonian basin, one which intrudes at 3 km depth and the other at 0.5 km depth. The thickness and depths were obtained from the maximum and minimum emplacement depths and the maximum thickness for CAMP sills found within drill core from the Amazon basin (Heimdal et al. 2018). Even at 3 km depth, a 150 m-thick sill should cool below the zircon crystallization window within ~ 500 years, which is ~ 300 times shorter than the measured zircon age range. As discussed briefly above, there are various possible reasons for the zircon crystallization temperatures to be inaccurate (unknown activities of Si and Ti, and also possible kinetic effects); however, changing the zircon crystallization window by tens of °C would not be enough to explain the measured zircon age ranges. It is possible that the thermal model significantly underestimates the geothermal gradient (25 °C/km) in the basin, due to the multiple sill intrusions occurring over ~ 300 ka; however, it seems unlikely that the temperature would be elevated enough to maintain the magma above its solidus for ~ 300 times longer than the model currently predicts. Based on our thermal modeling, and zircon crystallization constraints discussed in detail above, we prefer to interpret the older part of the age distribution as the emplacement and crystallization age of the sills. This method of interpreting zircon age distributions is opposite to those used in felsic rocks (intrusive or extrusive) (e.g., Keller et al. 2018), where the younger portion of the distribution is considered to be closest to the eruption or emplacement age.
Unfortunately, the argument above does not seem to be valid for the case of the NMB data, and neither explains the presence of xenocrysts in the Orange Mountain basalt and AN733. The NMB and Orange mountain basalt are both basalt flows, whereas all of the other samples are intrusives, which could indicate a much shorter time between assimilation and freezing of the magma for the aerially exposed basalts, hence the preservation of xenocrysts. We, therefore, argue that when complex age distributions are found in mafic rocks, combinations of elemental, isotopic and U–Pb data, along with modeling, may be required to correctly interpret the data.