Rb–Sr geochronology was carried out via laser ablation inductively coupled plasma tandem mass spectrometry following the methods initially set out in Zack and Hogmalm (2016) and Hogmalm et al. (2017). Analyses were carried out in the Fipke Laboratory for Trace Element Research (FiLTER) at the University of British Columbia, Okanagan (UBCO), using a laser (ESL NWR193) with a TwoVol3 ablation cell paired with an Agilent 8900 triple quadrupole (QQQ) ICP-MS instrument as described in Larson et al. (2023a, b). A circular laser spot with a diameter of 65 µm, an approximate pit depth of ∼ 10–15 µm, a fluence of ∼ 3 J cm-2, and a repetition rate of 5 Hz was used for all materials. Analyses of the glass reference material NIST SRM 610 (Jochum et al., 2011) were carried out using both 60 and 30 µm diameter laser spots to ensure the analyses were measured in both pulse and analogue detector modes for cross-calibration (e.g. Zack and Hogmalm, 2016; Glorie et al., 2024). Instrument drift was corrected, and down-hole fractionation was minimized (e.g. Glorie et al., 2023), based on analyses of NIST SRM 610 using an in-house data reduction scheme (DRS; Larson, 2024) developed for iolite v.4.X (Paton et al., 2011) based on the freely available “U–Pb Python Example.py” DRS (Petrus, 2022), following the approach and error propagation detailed in Redaa et al. (2021), and using a covariance matrix to calculate error correlation. Matrix fractionation effects on 87Rb/86Sr were corrected based on repeated analyses of Mica 1O (phlogopite – 986 ± 5 Ma; Camacho et al., 2020) and verified based on the analyses of GA-1550 (Mt Dromedary biotite – 98.7 ± 1.9 Ma, Li et al., 2008). During run 1 (108F and 47-49F), GA-1550 returned an isochron date of 99.6 ± 3.5 Ma (mean squared weighted deviation (MSWD) = 1.9, n=14/15, initial 87Sr/86Sr = 0.7049 ± 0.0017). During run 2 (25F and 86F), GA-1550 returned an isochron date of 98.1 ± 2.3 Ma (MSWD = 0.93, n=17/17, initial 87Sr/86Sr = 0.7049 ± 0.0017), while an additional in-house reference material, Mica 1B (phlogopite – 990 ± 6 Ma; Camacho et al., 2012), returned an isochron date of 997 ± 6 Ma (MSWD = 3.7, n=17/20, initial 87Sr/86Sr = 0.7035 ± 0.0005). Full Rb–Sr data are provided in Tables S1 and S2.

In addition to Rb–Sr isotopes, Ti, Mg, and Fe concentrations were measured for each analytical spot via LA-ICP-MS with dwell times of 1, 0.5, and 1 ms, respectively. Concentrations were normalized to repeated measurements of the NIST SRM 610 glass reference material (Jochum et al., 2011) assuming stoichiometric Si (16.36 weight (wt.) % Si or 35 wt % SiO2), typical of metamorphic biotite from greenschist through granulite facies (e.g. Dyck et al., 2021). Titanium concentrations (ppm) were converted to molar-weight-per-oxide TiO2 equivalents and then normalized to calculated Si content based on an assumed 5.4 atoms per formula unit (a.p.f.u.) Si per 22 O as detailed in Eq. (1): 1Tia.p.f.u.=(CTi/WTTiO2/10000/MTiO2)×(5.4/ASi), where CTi is the concentration of Ti in ppm, WTTiO2 is the weight proportion of Ti in TiO2, MTiO2 is the molar weight per oxide, and ASi is the molar weight per cation Si assuming 35 wt % SiO2. Ti-in-biotite temperatures were calculated using the equation of Henry et al. (2005), which requires Ti a.p.f.u and Mg# (Mg/(Mg + Fe)) for each analysis. The values for Mg# were calculated using parts-per-million (ppm) concentrations of Fe and Mg. The 1σ uncertainty of the Ti-in-biotite temperatures is <24 °C at 480 °C decreasing to 12 °C at 800 °C (Henry et al., 2005).

To verify the LA-ICP-MS-derived Ti-in-biotite temperatures, the chemistries of the same mica grains were analyzed using the CAMECA SXFiveFE electron probe microanalyzer (EPMA), also housed in the FiLTER facility. Quantitative spot analyses were carried out with an acceleration voltage of 15 kV, a regulated beam current of 20 nA, and a spot size of 5 µm. Elemental x-ray data were collected using a dwell time of 30 s on peak and 15 s on background and were calibrated to the known composition of synthetic and natural mineral reference standards from Micro-Analysis Consultants Ltd. The standards used for calibration were spessartine for Si and Mn, titanite for Ti, almandine for Al and Fe, wollastonite for Ca, albite for Na, diopside for Mg, orthoclase for K, and apatite for P. The crystals used were LTAP for Si and Al; LPET for Ti, Ca, K, and P; LLIF for Fe and Mn; and TAP for Na, Mg. The EPMA data were converted to atoms-per-formula-unit biotite on the basis of 22 oxygen whereby 2cation (a.p.f.u.)=cation proportion×no. oxygen (a.p.f.u.) /sum of oxygen proportions,3oxygen proportion=cation proportion×(1/2)cation charge,4cation proportion=oxide wt %/oxide molecular wt. Given that the biotite analyzed was collected as detritus from the Bengal Fan and given the tendency of biotite to weather (e.g. Wilson, 2004), alteration was expected. To avoid the most altered material, EPMA data were filtered for totals >88 wt % and K concentrations >1.1 a.p.f.u. Plotting K a.p.f.u against Mg# or calculated temperature shows that analyses with K a.p.f.u. >1.1 rarely form outliers (Fig. 3). While it is recognized that an a.p.f.u of 1.1 K is significantly lower than that expected of unaltered biotite, K is expected to leave the mineral via hydrated cation exchange as part of the initial weathering stage/vermiculitization (Price and Velbel, 2014; Gilkes and Suddhiprakarn, 1979). In contrast, Fe, Mg, and Ti are relatively immobile until more advanced alteration of biotite to goethite or kaolinite (Gilkes and Suddhiprakarn, 1979). Exchange of hydrated cations with interlayer K will result in lower-oxide EPMA wt % totals, which, upon normalization (for a biotite with an XMg of 0.44), increases the temperature calculated using the Ti-in-biotite thermometer by ∼ 2.2 °C per 1 wt % oxide deficit. The sensitivity of Ti-in-biotite temperatures to EPMA oxide wt % totals, using the Henry et al. (2005) calibration, is shown in Fig. S1. To illustrate the effect of weathering on our data we plot all analyses, coloured by oxide wt % totals, in Fig. 3a and b. In general, filtering analyses with K concentrations >1.1 a.p.f.u rather than filtering solely on oxide wt % totals is the more robust approach to removing outliers with anomalously low Mg# (e.g. <0.15) and Ti-in-biotite temperatures (e.g. <400 °C). Full EPMA data are provided in Table S3.

All Rb–Sr results were filtered for Rb/Sr (>3), Fe + Mg (between 15  % and 26 wt %), and Ti (>0.1 a.p.f.u) to avoid spurious analyses. Of the 137 spot analyses of biotite from specimen 25F, 132 remained after filtering. The data spread between isochrons at 45 and 8 Ma and define an over-dispersed (MSWD = 12.9) ca. 13 Ma isochron (Fig. 1a). Similarly, 138 spots in biotite from 108F yield 102 viable analyses that spread between 5 and 30 Ma isochrons. The data define an over-dispersed regression (MSWD = 5.5) at ca. 14 Ma (Fig. 1b). Of 104 analyses in biotite separated from 49F, 96 spread between isochrons at 10 and 30 Ma, defining an over-dispersed (MSWD = 3.0) isochron at ca. 16 Ma (Fig. 1c). Finally, 113 of 117 biotite analyses of material from 86F spread between isochrons at 37 and 11 Ma and comprise an over-dispersed (MSWD = 6.88) regression at ca. 15 Ma (Fig. 1d).

For each specimen, spot dates were calculated as two-point dates based on a regression between each analysis and a specified initial 87Sr/86Sr value (see Tables S4–S7). The initial 87Sr/86Sr values used (0.71, 0.72, 0.73, 0.74, 0.75, and 0.76) cover the range of expected values for most plutonic and metamorphic sources (Rösel and Zack, 2022). The dates calculated for each specimen are depicted in Fig. 2 as kernel density estimations (KDEs; bandwidth = 3). In general, the KDEs have a single peak for >0 Ma dates (if a spot analysis has an 87Sr/86Sr < the initial 87Sr/86Sr, a negative date will be calculated, a situation exacerbated in young material that has not had time to accumulate radiogenic 87Sr; see Tables S4–S7). The KDE peaks for each specimen generally increase in width, decrease in height, and move to slightly younger positions with two-point dates calculated with progressively higher initial 87Sr/86Sr (Fig. 2).

Ti-in-biotite temperatures calculated via LA-ICP-MS data generally range between ∼ 650 and 725 °C for most samples (Fig. 3b, c). The SiO2 content used in the calculations has little effect on temperature. Increasing the SiO2 from 35 wt % to 40 wt % results in only an ∼ 5 °C temperature change (Fig. S2), well within the expected uncertainty of the method (Henry et al., 2005). While the precision of Ti data generated by LA-ICP-MS is lower than EPMA data, the second- and third-quartile temperatures overlap (within uncertainty) with those calculated using EPMA data for each specimen (Fig. 3c).

As with any isochron method, the value of the intercept can have a significant impact on the results when calculating two-point isochron Rb–Sr dates. That effect, however, is less significant with older material (higher radiogenic Sr) and more radiogenic analyses. As shown in Fig. 4, the effect of initial intercept selection can be demonstrated by generating three artificial datasets with one analysis every 500 87Rb/86Sr increments between 10 and 4510. One dataset defines a 25 Ma isochron, one defines a 250 Ma isochron, and one defines a 2500 Ma isochron, each with an initial 87Sr/86Sr of 0.70. Two-point spot Rb–Sr dates were calculated for each data point within each model for different initial 87Sr/86Sr ranging from 0.71 to 0.79, and the percent of absolute deviation in two-point isochron date for each modelled point from the “known” date (i.e. 25, 250, or 2500 Ma) was quantified (Fig. 4). Plotting the % change in two-point isochron date versus 87Rb/86Sr demonstrates that young, low-87Rb/86Sr material is most affected by changing the initial 87Sr/86Sr. Moreover, the plot further demonstrates that for high 87Rb/86Sr (∼ >2500), even in young material, the typical difference in initial 87Sr/86Sr encountered in most crustal rocks (i.e. 0.70–0.76; Rösel and Zack, 2022) will have ∼ <5 % effect on the date calculated (Fig. 4). For older material, the 87Rb/86Sr cut-offs to minimize the effect of initial 87Sr/86Sr on dates (i.e. <5 %) drop significantly (i.e. ∼ > 87Rb/86Sr = 500 at 250 Ma and ∼ > 87Rb/86Sr = 50 at 2500 Ma; Fig. 4).

The effect of initial 87Sr/86Sr relative to 87Rb/86Sr can be further investigated using the real-world data presented herein. Filtering the two-point isochron dates calculated based on 87Rb/86Sr demonstrates that the position of the main population of the data remains relatively invariant for different 87Rb/86Sr but that the half-width decreases and the density of the main population increases with higher 87Rb/86Sr cut-offs (bottom curves Fig. 2). These results indicate that filtering detrital data by 87Rb/86Sr may help ameliorate the complication of unknown initial 87Sr/86Sr values, though it should be noted that doing so would introduce bias toward more radiogenic populations and, as such, should not be done in isolation.

Najman et al. (2019) presented the results of various detrital geochronology methods employed on the samples examined in the current study and/or other proximal samples. These methods include 40Ar/39Ar geochronology on white mica. White mica was targeted in that study both because of the common problem of excess Ar associated with biotite (Stübner et al., 2017; Larson et al., 2023a) in the Himalayan system from which these samples were collected (i.e. Himalayan detritus shed into the Bay of Bengal) and because of the resilience of white mica, relative to biotite, to weathering (Wilson, 2004). Given the differences in child product retention between white mica 40Ar/39Ar geochronology and biotite Rb–Sr geochronology, it may not be expected that the data from the two systems would overlap. White mica has an estimated nominal closure temperature of ∼ 425–400 °C for Ar diffusion (100 µm radius grain, 10°C Ma-1 cooling rate, 5–10 kbar; Harrison et al., 2009), whereas closure to diffusion of Sr out of biotite is more varied, with estimates ranging from ∼ 300 °C (Jager et al., 1967; Armstrong et al., 1966) to ∼ 400 °C (Verschure et al., 1980; Del Moro et al., 1982) or even in excess of 400 °C depending on the specifics of the mineralogy and chemistries of the samples (Jenkin et al., 1995, 2001). This variability in chronometer “closure temperatures” (e.g. Dodson, 1973) may be reflected in variable offsets between the Rb–Sr and 40Ar/39Ar dates (Fig. 2).

The various two-point isochron Rb–Sr dates calculated for 108F and 86F define density peaks with a slightly younger (∼ 1 Ma) offset than the 40Ar/39Ar dates (initial 87Sr/86Sr = 0.71), whereas for 49F and 25F two-point isochron dates essentially reproduce the density peak of the 40Ar/39Ar dates (Fig. 2). Overall, the total Rb–Sr dataset, regardless of initial 87Sr/86Sr used, defines a similar dominant population of Early to Middle Miocene age (Fig. 5) to that noted by Najman et al. (2019). That peak was interpreted to demonstrate rapid exhumation of the eastern and central Himalaya with an ∼ <4 Myr lag time between exhumation through mica closure to child product diffusion and sedimentation (Najman et al., 2019). Given the similarities between the datasets, the detrital Rb–Sr geochronology would have led to the same conclusions as the 40Ar/39Ar data generated by Najman et al. (2019).

Comparison of EPMA and LA-ICP-MS thermometry results demonstrates that LA-ICP-MS data can yield results comparable with traditional methods (Fig. 3). The temperatures calculated for the spot data are consistent with derivation from amphibolite facies metamorphic rocks or from associated leucogranites (e.g. Waters, 2019), which dominate the inferred exhuming Himalayan midcrustal source (Najman et al., 2019). The temperatures also broadly overlap with Zr-in-rutile temperature derived from detrital rutile from the same samples (∼ 650–780 °C; Najman et al., 2019). Direct comparison of LA-ICP-MS- and EPMA-derived temperatures for the same grains from 25F and 86F defines mean Euclidean distances of 13.6 and 15.9 °C, respectively (Fig. 3d). Because this approximately 15 °C uncertainty likely reflects the analytical precision of LA-ICP-MS rather than being solely due to the uncertainties stemming from natural variations in biotite chemistry or the thermometer calibration, it should be regarded as additive to the ±12–24 °C 1σ uncertainty of Henry et al. (2005). The precision on the Ti-in-biotite temperature estimates, whether EPMA- or LA-ICP-MS-derived, does not facilitate quantifying a change across the sampled strata, as data from all samples overlap (Fig. 3c).

The viability of quantifying Ti-in-biotite temperatures via LA-ICP-MS allows each Rb–Sr spot analyzed to be associated with a unique temperature. This kind of petrochronologic association opens the possibility of many different types of studies, not unlike the linking of chemistry to spot dates did for U(–Th)–Pb geochronology in the early 2000s (e.g. Foster et al., 2004; Gibson et al., 2004; Rubatto, 2002).