The Cabo Ortegal Complex represents an allochthonous assemblage of continental- and oceanic-derived lithologies that record subduction-related metamorphism prior to their emplacement onto the Gondwanan margin during the Variscan orogeny (e.g. Martínez Catalán et al., 2019). Previous geochronological studies of the complex suggest that the magmatic protoliths (520–460 Ma) were metamorphosed during the Variscan at ca. 400–390 Ma followed by a rapid decompression to shallow crustal levels (Beranoaguirre et al., 2020). The demantoid andraditic garnet analysed here was described in detail by Madon et al. (1991). It occurs associated with chlorite and serpentine in fissures cutting across the foliation of the pyroxenite. Due to these cross-cutting relationships, it is believed that the demantoid-bearing veins postdate even the youngest regional granitoids (ca. 280 Ma; Rodríguez et al., 2007).

The first attempt to date the demantoid garnet crystals was made using a single-collector ICP-MS (the method is described in Millonig et al., 2020 and Beranoaguirre et al., 2022). The resulting data defined a regression line with a lower intercept at 284 ± 65 Ma (MSWD = 2, Fig. 3a), and only 43 out of 72 analytical spots had U or Pb concentrations above the detection limits. Subsequently, two sessions were performed with the MC-ICP-MS. The resulting lower intercept dates for the sessions were 254.7 ± 7.9/8.7 Ma (MSWD = 1.19, n = 47/64, Fig. 3b) and 255.3 ± 9.8/10.5 Ma (MSWD = 0.37, n = 31/40, Fig. 3c). The regression lines were anchored to a 207Pb/206Pb ratio of 0.870 ± 0.005, derived from the mean of the unanchored upper intercept of both sessions, providing a consistent baseline for comparison. Most of the analyses have extremely low U concentrations between 0.5 and 1 ngg-1, with rare exceptions up to 4 ngg-1 (Figs. 2a and 4; Tables S2 and S3).

The Orlica-Śnieżnik Dome is located on the northeastern margin of the Bohemian Massif in SW Poland. It comprises predominantly amphibolite-facies, partly migmatised, orthogneisses, hosting numerous inclusions of (ultra) high-pressure rocks (e.g. Walczak et al., 2017). Their granitic precursors intruded supracrustal sedimentary successions of Neoproterozoic to Ordovician age and were subsequently metamorphosed during the Variscan orogeny (Szczepański and Ilnicki, 2014).

The garnet analysed here corresponds to the felsic granulite G99-2 studied in detail by Anczkiewicz et al. (2007), who obtained garnet Lu-Hf and Sm-Nd ages of 387 ± 5 Ma and 320 ± 3 Ma, respectively. From the U-Pb analyses performed in our laboratory, an unanchored lower intercept age of 384.0 ± 10.0/11.5 Ma (MSDW = 0.52, n = 46, Fig. 3d) was calculated. The U and Pb contents are homogeneous, varying between 2 and 4 ngg-1, and between 0.3 and 3 ngg-1, respectively (Figs. 2b and 4; Table S4 in the Supplement).

Ultrahigh temperature (UHT) metamorphic crustal granulite xenoliths were found in four kimberlite diatremes from along the central axis of the Witwatersrand basin in the central Kaapvaal craton (Schmitz and Bowring, 2003). The xenoliths are K-free, Mg-Al-rich UHT granulites with abundant garnet, sillimanite, rutile, graphite and sulphides, and local occurrence of sapphirine, orthopyroxene, plagioclase and quartz. Metamorphic conditions have been estimated at temperatures above 1050 °C and pressures between 0.9–1.2 GPa (Dawson et al., 1997). Zircon and monazite gave U–Pb ages of approximately 2.7 Ga (Schmitz and Bowring, 2003). The xenoliths were brought to the surface by Cretaceous kimberlite magmatism at approximately 120 Ma.

The results of these samples have already been published and discussed in detail by Shu et al. (2024). The analyses of the garnet in that publication followed the method described here, and considering that the aim of this work is purely methodological, two examples of those rocks are shown: the one with the lowest U concentration (sample ST66) and the one with the highest precision (ST 70). The latter was achieved by anchoring the regression line to a sillimanite crystal analysis, assuming that both minerals are in equilibrium. The samples ST66 and ST70 define regression lines with lower intercepts at 3096 ± 63/70 Ma (MSWD = 2.18, n = 53/53, Fig. 5a) and 3096 ± 37/48 Ma (MSWD = 3.05, n = 53/58, Fig. 5b). As stated above, sample ST66 shows the lowest U content among the samples presented in this work, varying from 0.3 to 2.2 ngg-1, whereas sample ST70 has a more variable U content, between 1 and 17 ngg-1. Pb and U contents are positively correlated. Unfortunately, it was not possible to measure the 208Pb isotope, so the calculated Pb content is the sum of 204Pb + 206Pb + 207Pb. The resulting Pb concentrations do not exceed 1 ngg-1 for sample ST66 and 6 ngg-1 for sample ST70, except in two spots (Fig. 4; Table S5).

Multiple Cretaceous kimberlite pipes occurring on the Proterozoic Namaqua–Natal-Fold Belt, which surrounds the Archean Kaapvaal craton, have brought to the surface a significant amount of high-pressure eclogite, granulite and pyroxenite xenoliths (Le Roex et al., 2020). The eclogites are mainly bimineralic (garnet-clinopyroxene) with accessory rutile and coesite/quartz (and even diamond), although kyanite-bearing eclogites are also described. As for the Kaapvaal craton granulites, we have analysed several samples from different kimberlites of the area, but only two of them are discussed in the present manuscript. Data points from sample L02100 define a regression line with a lower intercept at 102.7 ± 3.6/3.9 Ma (MSWD = 1.05; n = 37/41; Fig. 5c), whereas the analyses of sample L212 define a regression line with a lower intercept at 105.5 ± 2.5/2.9 Ma (MSWD = 0.92; n = 34/48; Fig. 5d). Both samples have relatively homogeneous U contents of 14–16 ngg-1 on average. Sample L212 also shows a homogeneous Pb content between 0.5 and 4 ngg-1, whereas sample L02100 display a higher variability with up to 14 ngg-1 (Fig. 4, Table S6).

One of the main strengths of the LA-ICP-MS technique is its high spatial resolution. In contrast with the bulk dating method (TIMS), the laser ablation method only analyses a relatively small volume of the targeted mineral, and therefore, the bulk amount of U analysed is limited. Besides, not all the ablated material reaches the detector, and the useful yield for U (= ions detected/total number of atoms in sample volume for a species of interest, Schaltegger et al., 2015) is variable depending on the utilised instrument.

In the case of this study, the ablation of garnet with a 193 µm spot size, 15 Hz and ∼ 2 Jcm-2 for 18 s results in a ca. 15 µm deep ablation pit. The depth is variable depending on the composition of the garnet, typically between 8–22 µm. Figure 6 shows a representative ablation pit from sample L02100 together with its measured depth profile, and similar pit depth measurements were routinely performed on all analysed samples to estimate the average ablated volume as accurately as possible. Considering an average pit depth of 15 µm, each analysis corresponds to a total volume of ca. 600 400 µm3 or 2.3 µg of ablated material, assuming a mean garnet density of 3.9 gcm-3 (Table 1). Li et al. (2022) recently reported skarn garnet with up to 75 µgg-1U. Applying the above calculations, this corresponds to ca. 100 pgU per ablated spot, and ca. 8 pgU and 0.4 pgPb for a Mali garnet analysis (6 µgg-1U and 0.3 µgg-1Pb, Seman et al., 2017). However, the metamorphic garnet rarely exceeds U concentrations of ∼ 100 ngg-1 (Deng et al., 2022). For example, a regional metamorphic garnet with 90 ngg-1U (sample SS-1 of Millonig et al., 2020) would yield ca. 0.1 pgU per analytical spot. In the present study, sample ST66 has the lowest U concentration, with an average concentration of 1.1 ngg-1. The total U ablated during such a spot analysis is ca. 1 fg (femtogram; 10⁻¹⁵ g). This is ca. 3500 times less U than for a renowned GJ-1 zircon analysis (287 µgg-1Uavg; 26 µgg-1Pbavg, Jackson et al., 2004), which yields 3.5 pg of U and 0.30 pg of Pb per spot, calculated for a 20 µm spot size, 15 µm pit depth and ca. 10 ng of ablated material (considering a zircon density of 4.65 gcm-3) (Table 1).

Samples with even lower amounts of U could be analysed, if they are older and have higher 238U than 206Pb for the majority of the spots (as well as 206Pb > 207Pb), with the exception of the high common-Pb analyses. As an example, a 200 Ma old concordant analysis will have ca. 30 times more 238U than 206Pb and ca. 20 times more 206Pb than 207Pb (or 600 times more 238U than 207Pb). In turn, a 3 Ga sample intercepts the Concordia at ca. 1.5 (238U/206Pb) and 0.25 (207Pb/206Pb), the 238U being only ca. 6 times more abundant than 207Pb. Hence, for such old samples, it would be possible to date a garnet with even lower U concentrations.

Such low U and Pb concentrations in garnet correspond to a few hundred counts per second (cps), and therefore, the background level plays a vital role, especially for 207Pb, the least abundant of the measured isotopes. Background Pb may come from different sources, such as a memory effect in the tubing, contamination/dirt in gas-line connections, the cones or the laser cell. Although careful cleaning of those parts is effective in reducing their Pb contribution, Pb derived from the Ar and He gas supplies is unavoidable. This contribution varies from day to day (Fig. 7 shows the background level at the FIERCE laboratory observed over one and a half years), and it is a critical factor for the analytical performance. Following the approach proposed by Pettke et al. (2012), the limits of detection for average background signals of 10 cps and 400 cps are approximately 6 and 30 pgg-1, respectively (corresponding to net signals of 2 and 8 cps at the detection limit, considering the aforementioned sensitivity of 250 400 cps per µgg-1). Although the background intensity differs by a factor of 40 in these examples, the limit of detection increases by only about a factor of 4–5 because the background count rate enters the calculation under the square root. On the contrary, the limit of quantification (Currie, 1995), defined as 10 times the standard deviation of the measured background signal, drastically increases with increasing background intensity. For example, during low-background sessions, the standard deviation of the 207Pb background can be as good as 5–6 cps, corresponding to a limit of quantification of approximately 50–60 cps (Fig. 6). In contrast, during one of the highest-background sessions, the standard deviation is ca. 200 cps, increasing the limit of quantification to 2000 cps. Thus, a high Pb background prevents reliable analysis of low 207Pb samples, as their signal falls below the quantification limit. In practice, samples with ultra-low-U and -Pb garnet are only analysed during sessions in which the 207Pb background remains below 50 cps.

However, even though the Pb can be reliably detected, the precision of each analysis is also crucial. Figure 8 shows the relationship between the 207Pb signal intensity and the within-run precision of the 207Pb/206Pb ratio, over several sequences covering a large range of 207Pb concentrations. As could be expected from counting statistics, a lower 207Pb signal results in a lower precision. Analyses yielding 207Pb signals above 2500 cps commonly achieve uncertainties of less than 1 % (2 SE), whereas signals below 2500 cps typically result in uncertainties above 1 % (2 SE). Furthermore, as the signal gets closer to the limit of quantification (≤ 100 cps), there is a strong increase in the uncertainty, which highlights the limit of the technique.

Another potential source of uncertainty, although it is not considered in our calculations, is the variance of the U decay, or in other words, the statistical uncertainty coming from the decay of a small number of radioactive atoms. In the case of dating high-U minerals, as well as old samples that have been decaying for a long time, such uncertainty can be considered negligible and that is likely why it is not considered in geochronology. However, physicists studying slow decay processes have discussed this phenomenon (e.g. Opendak and Wildenhain, 1994). In the case of low-U and young minerals (< 1 ngg-1U or < 10 Ma old), this incertitude may be of importance and needs to be considered. Nonetheless, young samples with such a low content of U are currently not measurable with the analytical setup discussed in this study, as the amount of radiogenic 207Pb may be below the limit of quantification.

In addition to the analytical precision of each individual spot measurement, the precision of U-Pb dating for non-concordant minerals like garnet when using a Tera–Wasserburg diagram is also controlled by how accurately the regression line can be defined. This strongly depends on the variability of the 238U/206Pb and 207Pb/206Pb ratios; while samples with high common-Pb contributions tend to cluster near the upper intercept, a larger spread in these ratios yields a smaller uncertainty on the lower intercept with the Concordia curve and, consequently, a more precise age. The internal precision for the analyses of the Cabo Ortegal demantoid are ± 4.7 % and ± 6.3 % for the absolute garnet ages, even with only 40 analytical spots and an extremely low U concentration (mainly below 1 ngg-1, Fig. 4). In turn, U concentrations of 2–4 ngg-1, as for the G99-2 garnet (Orlica-Śnieżnik granulite), significantly improve the precision to ca. ± 3 % absolute age uncertainty. Furthermore, the analyses of sample ST70 resulted in an internal age precision of ca. ± 1 %. In this case, two factors concurred to result in such precise data: (I) higher U and Pb contents and (II) the Archean age of the sample, as older samples benefit the most from the technique due to their more abundant daughter isotopes 206Pb and 207Pb. However, in samples as young as the Cretaceous Namaqua–Natal belt eclogites, we have obtained precisions of ca. ± 3 % at U concentrations of ca. 14–16 ngg-1. The results obtained for the demantoid in two independent sessions are reproducible within uncertainty. Likewise, the U-Pb age of the garnet from G99-2, 384 ± 12 Ma, is in agreement with the Lu-Hf age reported by Anczkiewicz et al. (2007, 387 ± 5 Ma). Thus, the procedure described in this study produces accurate results within the internal precision of the technique.

U-Pb dating of garnet is becoming a relatively common tool, above all for skarn garnet, as they typically contain several µgg-1 of U (e.g., Reinhardt et al., 2022). However, the applicability of the technique is rather limited in metamorphic, typically almandine-pyrope, garnet due to their low U content (Millonig et al., 2020; Cerva-Alves et al., 2021; Schannor et al., 2021; O'Sullivan et al., 2023; Mark et al., 2023; Peillod et al., 2024; Manzotti et al., 2025), and the common presence of high-U mineral inclusions. The method described here allows for the analysis of very low-U garnet, or a reduction in laser spot sizes to avoid inclusions (cf. Walters et al., 2025) and thus has enormous potential to become a routine tool for petro(chrono)logical investigations.

Moreover, due to the high closure temperature of the U-Pb system in garnet (exceeding 1050–1100 °C; Shu et al., 2024), it appears to be the most reliable method for dating its formation and thus the associated mineral reactions. In many cases, these events cannot be accurately constrained by traditional geochronometers like zircon, apatite, rutile, or monazite due to inherent limitations in such settings. For instance, rutile exhibits relatively low U-Pb closure temperatures (< 600 °C, depending on grain size; Cherniak, 2000; Zack and Kooijman, 2017). Similarly, minerals like apatite and monazite are prone to fluid-assisted coupled dissolution–reprecipitation or recrystallisation during subsequent hydrothermal events (e.g., Harlov et al., 2002, 2011; Putnis, 2009). Conversely, highly resilient minerals like zircon frequently retain inherited domains that complicate both their geochronological analysis and geological interpretation (e.g., Rubatto, 2017). Furthermore, if such minerals occur as inclusions within garnet, they may remain shielded from fluid interaction, thereby recording pre-metamorphic ages. Finally, many of these accessory phases are scarce or absent in mafic and ultramafic rocks. The granulite and eclogite xenoliths from the Kaapvaal craton and Namaqua–Natal belt analysed here are good examples, similar to the peridotitic garnet dated by O'Sullivan et al. (2023). Even though scarce rutile has been found in a few samples, the most common mineral assemblage is formed only by garnet and clinopyroxene. In addition, the low U content and their small grain sizes do not necessarily make metamorphic rutile a better geochronometer. In this work, we have successfully dated garnet from those geological contexts, despite their very low U contents. Larger sample sets of both the Kaapvaal craton granulites and the Namaqua–Natal belt eclogites have been analysed, and the results are reported and discussed in Shu et al. (2024).

Likewise, at present, the only possible way to date the hydrothermal garnet-chlorite-serpentine veins crosscutting the pyroxenites at Cabo Ortegal is by U-Pb dating of the Cabo Ortegal demantoid. The rock is near-monomineralic garnet, and because the associated chlorite lacks K and Rb, traditional K–Ar or Rb–Sr geochronological techniques cannot be applied. As the garnet occurs in fissures cutting across the foliation of the pyroxenite, and likely the whole complex in depth, ages younger than the plutons of the area were expected (youngest plutons are ca. 280 Ma, Rodriguez et al., 2007). The results of ca. 255 Ma obtained here are similar to the Aia Pluton age in the westernmost Pyrenees (dated to 267 Ma, Denèle et al., 2012). Those authors interpreted the age as reflecting the earliest evidence of an extensive regime that led later to the formation of the Bay of Biscay rift. Thus, the garnet-bearing veins at the Cabo Ortegal Complex may have formed during the same event.

The speed of data acquisition provided by laser-ablation techniques also allows for obtaining a large dataset in a short time, in contrast to the more precise acid-digestion-based garnet dating methods (Lu-Hf and Sm-Nd; e.g., Münker et al., 2001; Anczkiewicz and Thirlwall, 2003). This gives the opportunity to investigate a larger number of samples, either with a higher sampling density in a given area of investigation, or by extending the investigation to a much larger area. In some cases, a larger dataset can reveal small-scale metamorphic processes that have been hidden due to sampling bias (including samples not analysed due to the appearance of alteration). For instance, the U-Pb age of garnet from the NE Bohemian Massif obtained here is within the uncertainty of the Lu-Hf age of this sample (Anczkiewicz et al., 2007). Walczak et al. (2017) dated other mesocratic and mafic granulites from the same outcrop sampled by Anczkiewicz et al. (2007) using the same analytical protocols. Their results pointed to a younger event of garnet formation at ca. 345–340 Ma and they attributed this inconsistency to a contribution of inherited Hf in the whole-rock analysis of Anczkiewicz et al. (2007). The in-situ U-Pb dating in garnet circumvents the inherited Hf issue as potential inclusions can be avoided by direct petrographic observations (i.e. setting spots in clean domains) or rejecting analysis with distinct U and/or Pb content (Fig. 2). In any case, the fact that the sample analysed by Walczak et al. (2017) is not exactly the same, may indicate that the geological evolution of the area is more complex (polymetamorphic?) than previously thought.