the Greater Caucasus, along with select zircon U-Pb and Hf isotopic data from coeval igneous rocks, to link key magmatic and depositional episodes along the Caucasus convergent margin. Devonian to Early Carboniferous rocks were deposited prior to Late Carboniferous accretion of the Greater Caucasus crystalline core onto the Laurussian margin. Permian to Triassic rocks document a period of northward subduction and forearc deposition south of a continental margin volcanic arc in the Northern Caucasus and Scythian Platform. Jurassic rocks record the opening of the Caucasus Basin as a back-arc rift during southward migration of the arc front into the Lesser Caucasus. Cretaceous rocks have few Jurassic-Cretaceous zircons, indicating a period of relative magmatic quiescence and minimal exhumation within this basin. Late Cenozoic closure of the Caucasus Basin juxtaposed the Lesser Caucasus arc to the south against the crystalline core of the Greater Caucasus to the north and led to the formation of a hypothesized terminal suture. We expect this suture to be within ~20 km of the southern range front of the Greater Caucasus because all analysed rocks to the north exhibit a provenance affinity with the crystalline core of the Greater Caucasus.

Convergent plate margins serve a critical function as the locus of volcanic and seismic hazard on human timescales (e.g. Stern, 2002;Wirth et al., 2022) and as centers of lithospheric assembly and recycling on geological timescales (e.g. Cawood et al., 2013;Hawkesworth et al., 2017;Levander et al., 2014;Whitmeyer & Karlstrom, 2007). In theory, the transition from subduction of oceanic lithosphere to continent-continent collision along such margins is a conceptually straightforward process (Figure 1; e.g. Dewey & Bird, 1970;Dewey & Horsfield, 1970;Dickinson, 1971). In practice, long-lived convergent margins are notoriously complex due to episodic tectonic processes that create, destroy, and modify lithosphere along the margin prior to terminal collision. For example, the relatively buoyant and/or thickened crust of volcanic arcs, seamounts, or continental blocks may temporarily impede subduction, resulting in an episode of collisional deformation associated with accretion of these outboard terranes onto the continental margin (Figure 1; e.g. Dewey, 1977;Kapp & DeCelles, 2019;Tetreault & Buiter, 2012). Extension within the overriding plate of a subduction zone may create small basins in the forearc or back-arc region, with associated magmatic and sedimentary rocks that become incorporated into the margin upon basin closure either before or during terminal collision (Figure 1; e.g. Balázs et al., 2022;Cowgill et al., 2016;Karig, 1971).

Advancements in geochronology have increasingly delineated such episodicity in convergent margin processes, improving understanding of the processes by which lithosphere is built and modified over time in long-lived convergent margins (e.g. DeCelles et al., 2009;Ducea & Barton, 2007;Paterson & Ducea, 2015). Within relict convergent margins, large datasets of zircon U-Pb analyses compiled from igneous zircons in primarily intermediate to felsic magmatic rocks (e.g. Cecil et al., 2018;Gehrels et al., 2009) or from detrital zircons sourcing such magmatic rocks (e.g. Attia et al., 2021;DeGraaff-Surpless et al., 2002) can indicate the timing and location of elevated magmatic flux. These datasets have been interpreted to indicate that much of the magmatic activity in longlived convergent margins occurs in high-flux events as short as ~10 to 20 m.y., separated by equal or longer periods of relative magmatic quiescence. Proposed underlying causes for such episodicity range from external changes in plate motions or slab dynamics to internal cycles of lithospheric thickening and delamination (e.g. Cecil et al., 2018;Chapman et al., 2021;DeCelles et al., 2009;Paterson & Ducea, 2015). However, such studies have primarily focused on the North American Cordilleran convergent margin system, which did not culminate in terminal continent-continent collision. More histories of protracted convergent margin evolution from the geological record could advance a general understanding of how convergent margins evolve through time.

The active Alpine-Himalayan orogenic belt results from ongoing continent-continent collision of Arabia and India with Eurasia and provides the ideal setting for delineating episodic processes that build up a convergent margin prior to terminal ocean basin closure and collision (e.g. Dewey et al., 1973;Gansser, 1980;Şengör, 1986). The collisional orogens within this belt expose rocks recording the evolution of convergent margins that persisted throughout much of the Phanerozoic, but multiple, overprinting episodes of deformation, magmatism, and metamorphism often impede robust reconstructions of these margins (e.g. Adamia, Zakariadze, et al., 2011;Kapp & DeCelles, 2019;von Raumer et al., 2013). Detrital zircon geochronology provides an invaluable tool for investigating convergent margin evolution in such regions by allowing patterns of deposition within marginal basins to be linked to major phases of magmatism and erosional exhumation resulting from shifting tectonic regimes (e.g. Cawood et al., 2012;Gehrels, 2014;Sundell & Macdonald, 2022).

In the northernmost part of the Arabia-Eurasia collision zone, the Caucasus region records a protracted convergent margin history that spans the Phanerozoic and terminated in the late Cenozoic with closure of the Mesozoic to Cenozoic Caucasus Basin and formation of the Greater and Lesser Caucasus mountains (e.g. Adamia et al., 1977;Gamkrelidze, 1986;Mosar et al., 2022;Rolland et al., 2011;Vasey et al., 2020). Although the duration and complexity of this convergent margin system have long been acknowledged (e.g. Adamia, Alania, et al., 2011;Nikishin et al., 2011;Saintot et al., 2006;van Hinsbergen et al., 2020), much of the tectonic history has not been integrated into a framework of convergent margin evolution that can be exported and compared with other active and ancient margins throughout Earth history. Specifically, the underlying mechanism connecting Paleozoic terrane accretion and basin deposition with Jurassic formation of the Caucasus Basin has not been established. Additionally, although this convergent margin system culminated in late Cenozoic terminal collision, the location of a hypothesized suture zone representing closure of the Caucasus Basin and juxtaposition of the Greater Caucasus crystalline core to the north with the Lesser Caucasus volcanic arc to the south has not been clearly defined (Cowgill et al., 2016;Trexler et al., 2022).

Here, we present new detrital zircon U-Pb geochronology results from Paleozoic and Mesozoic (meta)sedimentary rocks in the Greater Caucasus Mountains that document episodic processes in the Phanerozoic evolution of the Caucasus convergent margin and identify spatial shifts in Caucasus Basin provenance consistent with the presence of a late Cenozoic suture. We supplement these analyses with zircon U-Pb and Hf isotopic analyses from igneous rocks in the Greater Caucasus that help characterize possible provenance sources for basin strata.

Our results indicate that detrital zircon maximum depositional ages (MDAs) in the Caucasus region generally match previously reported biostratigraphic depositional ages for Cretaceous and older units. However, many rocks previously mapped as Early Jurassic are likely Triassic or older, given the absence of Jurassic grains in these samples and the ubiquity of Jurassic grains in all other Jurassic and Cretaceous samples. We suggest that these Triassic and older rocks likely indicate a previously unrecognized forearc basin south of a long-lived continental margin arc that developed after Late Carboniferous terrane accretion. Jurassic opening of a back-arc basin along the Caucasus convergent margin transitioned the forearc basin to a back-arc setting. Overall, detrital zircon spectra delineate four distinct phases of Palaeozoic and Mesozoic deposition

Diagram showing the range of evolutionary paths and associated episodic convergent margin processes that influence lithosphere evolution along continental margins prior to terminal continent-continent collision. Center: In its most basic form, a convergent continental margin consists of a subducting slab beneath a continental margin and an associated continental volcanic arc. Top: Arcs, seamounts or continental fragments outboard of the continental margin may be accreted to the margin via subduction beneath the margin, subduction beneath the terrane, or double-sided subduction. Bottom: Slab rollback leads to extension within the overriding continental margin, creating marginal basins in the forearc or back-arc region that may be closed during subsequent slab advance. After terrane accretion or marginal basin closure, the system can return to a basic subduction zone until the ocean between two continents is consumed by subduction, resulting in terminal collision.

Back-arc Basin Extensional Forearc Basin ~50 km Slab Rollback Slab Rollback Arc Magmatism Collisional Magmatism/ Metamorphism Back-arc Magmatism Forearc/Back-arc Deposition Collisional Suture Subducting Slab Ribbon Continent, etc.) Continental Margin Convergent Margin Processes Re tu rn to B a s ic C o n v e r g e n t M a rg in Mode 3: Continent-directed Subduction Mode 1: Terrane-directed Subduction Mode 2: Terrane Accretion: Back-arc Closure: Forearc Closure:

Re tu rn to B as ic C o n v e r g e n t M a r g i n n i g r a M t n e g r e v n o C c i s a B o t n r u t e R Double-sided Subduction Terminal Continent-Continent Collision These results also indicate that all sampled units ~20 km north of the Greater Caucasus range front have a provenance affinity with the Greater Caucasus crystalline core to the north, rather than with the Lesser Caucasus arc and adjacent crystalline massifs to the south, suggesting that the hypothesized late Cenozoic suture lies along one or more structures to the south of this provenance break. Our study illustrates how interconnected episodes of terrane accretion, marginal basin opening during slab rollback and basin closure in the Caucasus convergent margin combine to influence the structure of the presentday collisional orogen.

The Greater Caucasus Mountains stretch ~1200 km between the Black and Caspian Seas and accommodate most of the orthogonal component of plate convergence at their longitude within the active Arabia-Eurasia collision zone (Figure 2a; Forte et al., 2010;Jackson, 1992;Mosar et al., 2010;Reilinger et al., 2006;Sokhadze et al., 2018). Eocene or Miocene onset of collision between Arabia and Eurasia along the Bitlis-Zagros suture (Figure 2a; e.g. Agard et al., 2005;Darin & Umhoefer, 2022;Hempton, 1985;Koshnaw et al., 2018;McQuarrie & van Hinsbergen, 2013) transferred stress northward to the Caucasus Basin, a Mesozoic to Cenozoic marine back-arc basin, inducing late Cenozoic basin closure and formation of the Greater Caucasus Mountains (Adamia et al., 1977;Cowgill et al., 2016;Gamkrelidze, 1986;Vincent et al., 2016;Zonenshain & Le Pichon, 1986). Although the evolution of the Caucasus Basin and the partitioning of strain within basin strata during basin closure had a major influence on the ongoing uplift of the Greater Caucasus Mountains, the basin structure and history remain poorly understood.

In the western Greater Caucasus Mountains, Mesozoic and Cenozoic strata deposited in the Caucasus Basin are broadly distributed between the Greater Caucasus crystalline core to the north and the Lesser Caucasus Mountains to the south (Figure 2b; e.g. Cowgill et al., 2016;Mosar et al., 2010;Saintot et al., 2006;Vincent et al., 2016). The Greater Caucasus crystalline core consists of heterogeneous Paleozoic igneous and high-grade metamorphic rocks typically interpreted as the basement rocks of the Greater Caucasus Mountains (e.g. Adamia, Zakariadze, et al., 2011;Somin, 2011;Vasey et al., 2020). The Lesser Caucasus Mountains comprise Paleozoic crystalline massifs, the Jurassic to Paleogene Lesser Caucasus volcanic arc, and the Paleogene Adjara-Trialeti back-arc basin (e.g. Gusmeo et al., 2021;Hässig et al., 2020;Mayringer et al., 2011;Rolland et al., 2011). Additional Mesozoic and Cenozoic strata form a north-dipping homocline in the Northern Caucasus region between the Greater Caucasus crystalline core to the south and the Scythian Platform to the north (Figure 2b; e.g. Cowgill et al., 2016;Ershov et al., 2003;Nalivkin, 1976).

Building on prior work studying the Caucasus Basin and related rocks (e.g. Forte et al., 2022;Mosar et al., 2022;Trexler et al., 2022;Vasey et al., 2020), we primarily organize our results into samples collected along major river corridors in the western Svaneti (SV) and the eastern Kazbegi (KZ) regions of Republic of Georgia, separated by ~200 km along-strike (Figures 2b and 3). We also include a sample from Jurassic strata of the Northern Caucasus in Russia (RU) in a region of primarily magmatic and metamorphic rocks known as the Bechasyn Zone (Figure 2b; e.g. Gerasimov et al., 2015;Hanel et al., 1992;Hess et al., 1995). These samples allow us to characterize the major phases of Paleozoic and Mesozoic marginal basin evolution, as described below.

The Dizi series: Predecessor to the

In Svaneti, Mesozoic strata of the Caucasus Basin are juxtaposed unconformably or by fault against reported Devonian to Triassic rocks of the Dizi series to the north (Figures 2b and 3; e.g. Adamia, Alania, et al., 2011;Gamkrelidze & Kakhazdze, 1959;Gudjabidze, 2003;Kazmin, 2006;Vasey et al., 2020). The Dizi series lies to the south of the Greater Caucasus crystalline core, consists primarily of tightly folded quartzite, slate, phyllite and marble metamorphosed to greenschist-facies conditions, and is intruded by plutonic rocks of reported Jurassic age (e.g. Adamia, Alania, et al., 2011;Gamkrelidze & Kakhazdze, 1959;Gudjabidze, 2003;Javakhishvili et al., 2021;Vasey et al., 2020). Although detrital zircon analyses of samples mapped as the Dizi series yield Devonian or Carboniferous maximum depositional ages (MDAs), Permian or Triassic zircons have yet to be reported (Vasey et al., 2020), suggesting the age range for this unit may be narrower (Devonian to Early Carboniferous) than previously interpreted (Devonian to Triassic).

The Dizi series is generally considered distinct from the Greater Caucasus crystalline core and represents a former marine basin that has been variably interpreted as a small back-arc basin (e.g. Adamia, Alania, et al., 2011;Vasey et al., 2020), the Paleotethys Ocean (e.g. Şengör et al., 1984), or the Rheic Ocean (Stampfli, 2013). This basin

Overview of the Arabia-Eurasia collision zone showing the location of the Caucasus region (Figure 2b) between the Black and Caspian seas, the main Bitlis-Zagros suture between Arabia and Eurasia, and the Scythian Platform between the Caucasus region and the core of Eurasia. (b) Simplified tectonic map of the Caucasus region showing locations of key components of the Greater Caucasus (GC -Greater Caucasus crystalline core, North Georgia Fault System, Main Caucasus Thrust, Dizi series, Bechasyn Zone) and the Lesser Caucasus (L -Lesser Caucasus Arc, Adjara-Trialeti, Palaeozoic crystalline massifs) described in the text. Location of Early Jurassic sample RU1 from the Northern Caucasus (NC) indicated by blue circle, locations of detrital zircon catchment analyses from the Lesser Caucasus (Cowgill et al., 2016;Tye et al., 2020) indicated by black polygons, and location of igneous zircon U-Pb and Hf isotopic sample GE0345 (Mayringer et al., 2011) indicated by yellow square. (c) Schematic cross-section of the western Greater Caucasus, after Vasey et al. (2020). (d) Overview of Triassic to Jurassic geodynamic evolution of the Caucasus region, in which Permian to Triassic northward subduction of the Tethys Ocean beneath Laurasia and associated arc volcanism was followed by Late Triassic collision of Iran with Laurasia, resulting in along-strike back-arc extension opening the Caucasus Basin and migration of the volcanic arc front south to the Lesser Caucasus (from Vasey et al., 2021).

(a) (b) (c) (d)

was deformed and either partially (e.g. Adamia, Alania, et al., 2011) or fully (e.g. Şengör et al., 1984) closed prior to the Jurassic onset of extension and deposition in the Caucasus Basin. The timing and mechanism of the transition from basin opening to closure remain poorly understood.

By Permian to Early Triassic time, northward subduction had initiated along the Laurasian margin beneath the Greater Caucasus, leading to arc magmatism within the present-day Scythian Platform and Northern Caucasus (Figure 2d; Alexandre et al., 2004;Gerasimov et al., 2015;Natal'in & Şengör, 2005;Okay & Nikishin, 2015;Tikhomirov et al., 2004;Vasey et al., 2021). The younger sedimentary cover of the Scythian Platform obfuscates the geological record of this phase, and only a few Triassic arc rocks have been directly dated in well cores from the Scythian Platform or outcrop samples in the Northern Caucasus (Alexandre et al., 2004;Gerasimov et al., 2015). Detrital zircons found in Caucasus Basin strata suggest that the main episode of arc magmatism took place from ~260 to 220 Ma (Tye et al., 2020;Vasey et al., 2020Vasey et al., , 2021)). Troughs in detrital zircon spectra imply a lull in arc magmatism from ~220 to 180 Ma, followed by a second episode of arc magmatism from ~180 to 140 Ma in the Lesser Caucasus (Tye et al., 2020;Vasey et al., 2021). This lull in magmatism has been interpreted to reflect southward migration of the arc front from the Scythian Platform and Northern Caucasus to the Lesser Caucasus arc due to slab rollback induced by the Late Triassic collision of Iran with Laurasia to the east (Vasey et al., 2021).

Back-arc basin deposition in the Caucasus Basin to the north of the Lesser Caucasus arc is generally considered to have begun in the Early Jurassic and continued until the onset of Cenozoic closure of the basin in Eocene or Oligocene time during Arabia-Eurasia collision (e.g. Adamia, Alania, et al., 2011;Nikishin et al., 2011;Saintot et al., 2006;Vincent et al., 2016;Zonenshain & Le Pichon, 1986). Caucasus Basin strata exposed in the western Greater Caucasus are dominated by siliciclastic, volcanic, F I G U R E 3 Simplified tectonic map of the study area in the Svaneti (SV samples) and Kazbegi (KZ samples) regions of the Republic of Georgia. Samples from this study are indicated by circles (detrital zircon) or squares (igneous zircon U-Pb and Hf isotopes). Published detrital zircon samples are indicated by diamonds for (meta)sandstone samples (Cowgill et al., 2016;Vasey et al., 2020) or catchmentdelineating black polygons for modern sand samples (Forte et al., 2022;Tye et al., 2020). Detrital zircon samples are coloured by depositional age as interpreted in the text. and volcaniclastic rocks in the Jurassic and mixed siliciclastic and carbonate rocks in the Cretaceous (Gudjabidze, 2003;Trexler et al., 2022;Vincent et al., 2016). A peak of ~180 to 160 Ma detrital zircons in Caucasus Basin rocks indicates a pulse of magmatism at this time, which likely reflects a major Jurassic episode of extension, rifting, and potentially seafloor spreading within the backarc basin prior to Cretaceous magmatic quiescence (Tye et al., 2020;Vasey et al., 2021). Reported Middle Jurassic plutonism (e.g. Okrostsvaridze et al., 2022) intruding Dizi series and Caucasus Basin strata north of the Lesser Caucasus arc is likely related to this ~180 to 160 Ma back-arc extension. Vasey et al. (2021) proposed that Triassic collision of Iran with Laurasia initiated both southward migration of the Caucasus arc front to the Lesser Caucasus and back-arc extension within the Caucasus Basin as the result of trench retreat and slab rollback along strike of the collision (Figure 2d). However, the temporal and spatial evolution of the basin throughout the Mesozoic is still largely unknown.

Closure of the Caucasus Basin and exhumation of the Greater Caucasus Mountains appears to have initiated by ~35 Ma (Avdeev & Niemi, 2011;Vincent et al., 2007Vincent et al., , 2011)), with a marked acceleration in exhumation and deformation at ~10 to 5 Ma (Avdeev & Niemi, 2011;Forte et al., 2022;Mitchell & Westaway, 1999;Mosar et al., 2010;Vasey et al., 2020;Vincent et al., 2020). The present-day orogen is singly vergent in the western part of the range and doubly-vergent in the central and eastern parts (Forte et al., 2014;Mosar et al., 2010Mosar et al., , 2022)), with active surface deformation largely concentrated within foreland fold-thrust belts to the south of the core of the range (e.g. Forte et al., 2010;Stahl et al., 2022;Sukhishvili et al., 2021;Tibaldi et al., 2017;Trexler et al., 2020;Tsereteli et al., 2016). However, the overarching style of basin closure remains contested, with end-member scenarios including inversion of a narrow Caucasus Basin primarily via reactivation of high-angle normal faults (Mosar et al., 2022;Vincent et al., 2016Vincent et al., , 2018) ) to subduction of a wider Caucasus Basin at least partially floored by oceanic or transitional crust and formation of a collisional orogenic wedge (e.g. Cowgill et al., 2016Cowgill et al., , 2018;;Mumladze et al., 2015;Trexler et al., 2022;Vasey et al., 2021;Zonenshain & Le Pichon, 1986). Given the role the geometry of the Caucasus Basin played in guiding the evolution of the Cenozoic Greater Caucasus orogen, resolving the precollisional structure and history of the Caucasus informs understanding of the present-day configuration of the mountain belt.

In the western Greater Caucasus Mountains, between the Greater Caucasus crystalline core and the foreland deformation front, Caucasus Basin strata are deformed by a series of dominantly north-dipping imbricate thrusts within the North Georgia Fault System (NGFS; Figure 2; Trexler et al., 2022). Previous detrital zircon analyses of Jurassic and Cretaceous Caucasus Basin strata indicate distinctly different provenance for samples north of the NGFS relative to those to the south (Cowgill et al., 2016;Tye et al., 2020), suggesting that one or more faults of the NGFS represent an enigmatic suture between rocks affiliated with the northern and southern margins of a relatively wide Caucasus Basin (Cowgill et al., 2016;Trexler et al., 2022). However, existing data are insufficient to determine if the boundary between the two provenance domains coincides with one or more faults within the NGFS and to assess the degree to which changes in provenance result from temporal rather than spatial changes within the Caucasus Basin.

We conducted U-Th-Pb analyses by laser ablation multicollector inductively coupled mass spectrometry (LA-MC-ICPMS) on 15 (meta)sedimentary rock samples and three igneous rock samples from the Greater Caucasus Mountains at the Arizona LaserChron Center (Tables 1, Tables S1 and S2), following procedures outlined in Gehrels et al. (2006Gehrels et al. ( , 2008)), Gehrels and Pecha (2014), Sundell et al. (2021), and the supporting information (Vasey et al., 2023). Grains were ablated with a Photon Machines Analyte G2 excimer laser with a spot diameter of 20-30 μm, depending on the size of the smallest zircons in the sample, and the ablated material was carried in helium to the plasma source of a Nu HR ICPMS to measure U, Th, and Pb isotopes (Table S4). For the three igneous samples (Tables S3 and S5), we then conducted Hf isotope analyses on zircons from the primary age populations using the same instrument with a spot size of 40 μm, with the spot placed directly on top of the spot used for the accompanying U-Th-Pb analysis, following analytical procedures outlined in Gehrels and Pecha (2014).

We reduced and filtered the U-Th-Pb data using the AgeCalcML software (Sundell et al., 2021) and the Hf isotopic data using an in-house LaserChron Excel workbook and present the resulting accepted data as kernel density estimates (KDEs; Figures 4 and 5). Univariate KDEs for U-Th-Pb ages and bivariate KDEs for combined age and Hf isotopic data are plotted using custom Python scripts (Vasey, 2023a(Vasey, , 2023b)).

T A B L E 1 Summary of zircon U-Pb and Hf isotopic analyses. Detrital zircon analyses: Sample Field number Lithology Region Latitude (°N) Longitude (°E) Youngest grain cluster age (Ma) a Maximum likelihood age (Ma) b Youngest single grain age (Ma) SV1 CT130918-2A Quartzite Svaneti 42.84498 42.85980 557.8 ± 6.0 391.6 ± 10.9 377.0 ± 10.6 SV2 100211-9A Quartzite Svaneti 43.00862 42.29912 229.5 ± 2.9 223.7 ± 3.4 222.1 ± 4.0 SV3 CT130918-9A Quartzite Svaneti 42.79157 42.74833 279.5 ± 3.4 216.6 ± 4.2 215.2 ± 3.3 SV4 CT130919-5A Quartzite Svaneti 42.74698 42.75903 174.7 ± 2.9 172.4 ± 4.5 171.4 ± 4.7 SV5 100411-5 Volcaniclastic Svaneti 42.81750 42.02909 173.2 ± 4.0 164.1 ± 3.6 164.1 ± 4.8 SV6 CT130919-8A Volcaniclastic Svaneti 42.66670 42.76710 165.4 ± 1.8 165.4 ± 1.5 162.7 ± 4.5 SV7 AB0926 Quartzite Svaneti 43.06635 42.69727 162.5 ± 1.9 161.5 ± 3.7 160.5 ± 4.0 SV8 100211-1A Quartzite Svaneti 43.03919 42.70195 132.3 ± 1.9 116.1 ± 3.8 116.0 ± 2.9 KZ1 CT15076 Quartzite Kazbegi 42.60419 44.57925 161.3 ± 1.4 159.6 ± 3.6 157.3 ± 3.4 KZ2 CT15082 Marlstone Kazbegi 42.54620 44.49082 319.0 ± 3.6 199.3 ± 11.5 199.2 ± 9.8 KZ3 CT15092 Sandstone Kazbegi 42.35320 44.70846 128.3 ± 1.1 128.3 ± 1.7 126.3 ± 2.0 KZ4 CT15127 Marlstone Kazbegi 42.21713 44.67327 303.0 ± 3.7 303.0 ± 3.4 302.5 ± 4.8 KZ5 CT15099 Marlstone Kazbegi 42.27469 44.68106 286.9 ± 3.8 286.8 ± 3.6 284.7 ± 5.1 KZ6 CT15113 Marlstone Kazbegi 42.16097 44.70196 208.6 ± 2.1 90.0 ± 2.1 89.9 ± 1.7 RU1 AB0913 Sandstone Russia 43.82004 42.02070 178.3 ± 2.3 181.8 ± 0.5 166.4 ± 3.7 Igneous zircon analyses: Sample Field number Lithology Region Latitude (°N) Longitude (°E) Crystallization age (Ma) c εHf i d SVGR V17043A Granite Svaneti 43.02871 42.92826 318.6 ± 5.2 -2.7 ± 2.5 SVGD C17069B Granodiorite Svaneti 43.07069 42.27934 169.6 ± 2.8 0.0 ± 4.6 SVAN 100311-3A Andesite Svaneti 42.98741 42.25975 166.3 ± 4.5 3.5 ± 2.6 Note: All errors reported at 2σ. All detrital zircon maximum depositional age metrics exclude individual analyses with >20% discordance or > 10% reverse discordance in relative age of the 238 U/ 206 Pb and 235 U/ 207 Pb systems. a Youngest grain cluster age calculated as weighted mean of youngest population of three or more grains overlapping within 2σ or two grains overlapping within 1σ (Dickinson & Gehrels, 2009). b Maximum likelihood age calculated using the maximum likelihood estimation method described in Vermeesch (2021a) and implemented in IsoplotR (Vermeesch, 2018). c Crystallization age reported as weighted mean of youngest population of ages, factoring in systematic error and overdispersion (Vermeesch, 2018). d Initial εHf calculated from initial 176 Hf/ 177 Hf values at time of crystallization compared with 176 Hf/ 177 Hf value at time of crystallization for chondritic uniform reservoir (CHUR) calculated using present-day 176 Hf /177 Hf value of 0.282785 and 176 Lu/ 177

Hf value of 0.0336 (Bouvier et al., 2008).

We filtered our U-Th-Pb results based on concordance, analytical uncertainty, and/or 204 Pb concentration. For zircons with 238 U/ 206 Pb ages greater than 600 Ma, we used a relative age discordance filter and rejected ages with more than 20% discordance or more than 10% reverse discordance between the 238 U/ 206 Pb and 207 Pb/ 206 Pb ages. We also rejected analyses with more than 10% analytical uncertainty at the 1σ level in either the 207 Pb/ 206 Pb age (unless the 238 U/ 206 Pb age was greater than 600 Ma) or in the 238 U/ 206 Pb age. Finally, we rejected analyses with high common Pb as indicated by concentrations of 204 Pb greater than 600 cps. Additional details on U-Th-Pb filtering and a discussion of possible alternative approaches, including the use of concordia ages and the log ratio distance to concordia composition (Vermeesch, 2021b), are provided in the Supporting Information and Figures S13 and S14 (Vasey et al., 2023).

Assessment of maximum depositional age (MDA) from a detrital zircon sample is a subject of active discussion, with methods continuing to be developed and debated (e.g. Copeland, 2020;Coutts et al., 2019;Sharman & Malkowski, 2020;Spencer et al., 2016; F I G U R E 4 Kernel density estimates (KDEs) of new detrital zircon samples reported in this study from the Svaneti and Kazbegi regions of Georgia and the Russian Northern Caucasus. Samples are coloured by depositional age as interpreted in the text and include the number of accepted grains vs. the number of total analysed grains, the youngest single-grain (YSG) age, the maximum likelihood age (MLA; Vermeesch, 2021a), and the youngest grain cluster (YGC) age (Dickinson & Gehrels, 2009) Vermeesch, 2021a). For this reason, we present and consider three possible metrics for assessing MDA (Figures 4 and 6; Figures S1-S12). The first is the errorweighted mean of the youngest grain cluster (YGC), which is either the youngest population of three or more grains that overlap within 2σ error, or, if no three grains overlap within 2σ, the youngest two grains that overlap within 1σ (Dickinson & Gehrels, 2009). The second is the maximum likelihood age (MLA) proposed by Vermeesch (2021a), which models the age data as a mixture of two components, with one being a minimum age population and the other containing the rest of the data. The third is the age of the youngest single grain (YSG). For each of these metrics, we only consider grains of any age with relative age discordance less than 20% and reverse discordance less than 10% between the 238 U/ 206 Pb and 235 U/ 207 Pb systems. A discussion of alternative discordance filters for grains used in MDA calculations is provided in the Supporting Information text and Table S6 (Vasey et al., 2023). Systematic error for each of these metrics is propagated after calculation of the metric using only random errors for individual analyses (Horstwood et al., 2016).

All these metrics for determining MDA have drawbacks. The YGC method can overestimate MDA by ignoring a valid youngest age that is not part of a population and may also erroneously assume that grains with ages that overlap within uncertainty have a common source (Copeland, 2020), whereas relying on the YSG age risks basing MDA entirely on a single grain that is too young due to Pb loss, metamorphic overgrowths, or analytical dispersion (Sharman & Malkowski, 2020;Vermeesch, 2021a). The MLA method provides a useful alternative that corrects for the tendency of the YGC, YSG, and other methods to artificially converge to younger and younger ages as the number of analyses increases (Vermeesch, 2021a). However, the MLA method may still artificially group similarage grains from multiple populations or rely exclusively on a single grain analysis that has not been replicated. Thus, our interpretation of MDA for a given sample varies depending on the additional geological context and the individual analytical results.

We present igneous crystallization ages as the errorweighted mean of the principal population of ages in the sample, with overdispersion factored in by multiplying the errors by the square root of the mean square weighted deviation (MSWD; Figure S17; Vermeesch, 2018). We calculate initial 176 Hf/ 177 Hf at the time of crystallization using the analysed U-Pb age and calculate the chondritic uniform reservoir (CHUR) 176 Hf/ 177 Hf value at the time of crystallization using a present-day CHUR 176 Hf/ 177 Hf value of 0.282785 and 176 Lu/ 177 Hf value of 0.0336 (Bouvier et al., 2008). The 176 Hf/ 177 Hf values for the sample and CHUR at the time of crystallization were then used to calculate initial εHf. We assign MDA and crystallization ages to geological time intervals using the Geological Society of America geological time scale (Walker et al., 2013).

F I G U R E 5 Bivariate KDEs of igneous zircon U-Pb and Hf isotopic data from rocks in the Greater Caucasus (in red, this study) and sample GE0345 from the Dzirula Massif (in blue, Mayringer et al., 2011). Crystallization ages are presented as the weighted mean of the youngest population of ages, with errors (2σ) multiplied by an overdispersion factor of the square root of the mean square weighted deviation (MSWD). Initial εHf at the time of crystallization is calculated using the U-Pb age and a present-day chondritic uniform reservoir (CHUR) 176 Hf/ 177 Hf value of 0.282785 (Bouvier et al., 2008). Initial εHf reported as the mean of the youngest population with 2σ error indicated. EAGE VASEY et al. Sample SV1 is a lithic-rich, medium-grained, poorly sorted quartzite with dominantly subangular grains from a unit north of the Dizi series in Svaneti that was previously assigned an Early Jurassic biostratigraphic depositional age (Figure 3; Gamkrelidze & Kakhazdze, 1959;Geguchadze et al., 1985d;Gudjabidze, 2003). Detrital zircon analysis of this sample yields a Proterozoic YGC age of 557.8 ± 6.0 Ma (5 grains), a Devonian MLA of 391.6 ± 10.9 Ma, and a Devonian YSG age of 377.0 ± 10.6 Ma (Figure 4), all of which are significantly older than the reported Early Jurassic age. This sample is dominated by Precambrian grains and includes significant populations in key age ranges from 2.8 to 2.4 Ga (6% of grains), 2.3 to 1.5 Ga (20%), and 700 to 550 Ma (36%), in addition to a population centred at ~1.0 Ga not observed in other samples from this study (Figure 4; Figure S15).

Two additional lithic-rich quartzite samples (SV2, SV3) that are medium-grained and poorly sorted with subangular grains were recovered from units south of the Dizi series in Svaneti that were previously assigned Early Jurassic biostratigraphic depositional ages (Figure 3; Gamkrelidze & Kakhazdze, 1959;Geguchadze et al., 1985aGudjabidze, 2003). SV2 has a YGC age of 229.5 ± 2.9 Ma (3 grains), an MLA of 223.7 ± 3.4, and a YSG age of 222.1 ± 4.0, all of which are Triassic. SV3 has a Permian YGC age of 279.5 ± 3.4 Ma (4 grains), a Triassic MLA of 216.6 ± 4.2, and a Triassic YSG age of 215.2 ± 3.3 Ma. Both samples have very similar age spectra (Figure 4; Figure S15), dominated by Paleozoic populations centred around ~450 Ma (SV2: 28% of grains; SV3: 29%) and ~300 Ma (SV2: 29%; SV3: 26%), with a range of Precambrian ages including populations from 2.8 to 2.4 Ga (SV2: 3%; SV3: 2%) and from 2.3 to 1.5 Ga (SV2: 12%; SV3: 8%).

Lithic-rich quartzite sample SV4 is fine-grained and moderately sorted with dominantly subrounded grains and lies ~5 km south of sample SV3 in Svaneti (Figure 3). This sample has an MDA straddling the Early and Middle Jurassic, with a YGC age of 174.7 ± 2.9 Ma (4 grains), an MLA of 172.4 ± 4.5, and a YSG age of 171.4 ± 4.7 Ma. All three MDA metrics are consistent with the Early Jurassic biostratigraphic age reported for this unit on previous maps (Figures 3 and 4; Gamkrelidze & Kakhazdze, 1959;Geguchadze et al., 1985e;Gudjabidze, 2003). Apart from a small population of Jurassic zircons centred at ~170 Ma (4% of grains) and a larger Permian to Triassic population centred at ~240 Ma (9% of grains), sample SV4 is otherwise characterized by a very similar age spectra to sample SV3 (Figure 4; Figure S15).

Samples SV5 and SV6 are both volcaniclastic rocks that lie to the south of other Svaneti samples and are dominated by volcanic clasts, with smaller amounts of quartz and plagioclase present. SV5 is more quartz-rich, finegrained, and well-sorted, whereas SV6 contains layers of variable grain size and more plagioclase. These samples yielded few zircons (39 accepted in SV5, 10 accepted in SV6) but still contain small populations of grains defining Middle Jurassic MDAs matching the reported depositional ages of the their respective units (Figures 3 and 4; Gamkrelidze & Kakhazdze, 1959;Geguchadze et al., 1985cGudjabidze, 2003). SV5 has a YGC age of 173.2 ± 4.0 Ma (4 grains), an MLA of 164.1 ± 3.6 Ma, and a YSG age of 164.1 ± 4.8 Ma, whereas SV6 has a YGC age of 165.4 ± 1.8 Ma (5 grains), an MLA of 165.4 ± 1.5 Ma, and a YSG age of 162.7 ± 4.5 Ma.

Quartzite sample SV7 lies within a unit on the north limb of a major synclinorium in Svaneti north of the Dizi series (Figure 3). This sample has Late Jurassic MDA metrics of (YGC: 162.5 ± 1.9, 4 grains; MLA: 161.5 ± 3.7; YSG: 160.5 ± 4.0) that match the depositional age for the unit on a recent map (Gudjabidze, 2003) but are slightly younger than the Early to Middle Jurassic ages reported on others (Figure 4; Gamkrelidze & Kakhazdze, 1959;Geguchadze et al., 1985b). This sample is dominated by Paleozoic and Mesozoic age populations centred at ~450 Ma (14%), ~300 Ma (19%), ~240 Ma (24%) and ~170 Ma (10%), along with scattered Precambrian grains (Figure 4; Figure S15).

Lithic-rich quartzite sample SV8 was recovered south of sample SV7 in the core of the synclinorium in Svaneti and is fine-grained and poorly sorted with dominantly subangular grains (Figure 3). This sample has Early Cretaceous MDA metrics, including a YGC age of 132.1 ± 1.9 Ma (3 grains), an MLA of 116.1 ± 3.8 Ma, and a YSG age of 116.0 ± 2.9 Ma. All three metrics match previously reported Late Jurassic to Early Cretaceous biostratigraphic depositional ages for this unit (Figures 3 and 4; Gamkrelidze & Kakhazdze, 1959;Geguchadze et al., 1985b;Gudjabidze, 2003). Though it contains Cretaceous and Jurassic zircons, this sample is dominated by Precambrian and Paleozoic ages (Figure 4; Figure S15), including a population from 2.3 to 1.5 Ga (26%), a population centred at ~450 Ma (9%), and a population centred at ~300 Ma (22%).

Sample SVGR is a granite in the Greater Caucasus crystalline core in Svaneti (Figure 3) with a Carboniferous EAGE VASEY et al.

crystallization age of 318.6 ± 5.2 Ma (30 dated zircons; MSWD: 2.1). In addition, SVGR yielded 4 inherited older Paleozoic zircons (Figure 5). The youngest population of zircons yields initial εHf values of -2.7 ± 2.5.

Sample SVGD is from a granodiorite intrusion into the Dizi series in Svaneti (Figure 3). This sample yields a Middle Jurassic crystallization age of 169.6 ± 2.8 Ma (33 dated zircons; MSWD: 2.1) consistent with previously reported ages for plutons in this region (Figures 3 and 5; Gamkrelidze & Kakhazdze, 1959;Gudjabidze, 2003;Okrostsvaridze et al., 2022). This population of grains has an initial εHf value of 0.0 ± 4.6, similar to the value of the CHUR (Figure 5; Bouvier et al., 2008). Andesite sample SVAN ~10 km to the south yields a similar Middle Jurassic crystallization age of 166.3 ± 4.5 Ma (27 dated zircons; MSWD: 5.3) but a more depleted initial εHf value of 3.5 ± 2.6 (Figure 5).

In Kazbegi (Figure 3), lithic-rich quartzite sample KZ1 is fine-grained and poorly sorted with dominantly subangular grains and has Late Jurassic MDA metrics, including a YGC age of 161.3 ± 1.4 Ma (7 grains), an MLA of 159.6 ± 3.6, and a YSG age of 157.3 ± 3.4 Ma (Figure 4). These ages are all slightly younger than the previously mapped Early to Middle Jurassic biostratigraphic depositional age for the unit sampled (Gudjabidze, 2003;Kandelaki & Kakhazdze, 1957). The age distribution for KZ1 contains a major population centered at ~300 Ma (46% of grains), minor populations centered at ~450 Ma (8%) and ~170 Ma (11%), and a few older Precambrian grains (Figure 4; Figure S16). Sample KZ2 comprises finegrained, well-sorted marlstone with subrounded grains collected ~10 km south of KZ1 that yielded only 10 zircon ages. These ages generally match the distribution in sample KZ1 but are insufficient in number to be used to determine MDA or characterize provenance. The youngest single-grain age is 199.2 ± 9.8 Ma, which is older than the previously reported Late Jurassic biostratigraphic age for this unit (Figures 3 and 4; Gudjabidze, 2003;Kandelaki & Kakhazdze, 1957), although it is unlikely that this difference is significant considering the low zircon yield.

Sandstone sample KZ3 contains substantial plagioclase and lithic fragments and is fine-grained and moderately sorted with subrounded grains. This sample has a YGC age of 128.3 ± 1.1 Ma (7 grains), an MLA of 128.3 ± 1.7, and a YSG age of 126.3 ± 2.0 Ma, all of which match the Early Cretaceous biostratigraphic depositional age reported for this unit (Figure 3; Gudjabidze, 2003;Kandelaki & Kakhazdze, 1957). This sample is dominated by age populations centred at ~450 Ma (15%) and ~300 Ma (44%) but also contains low numbers of Precambrian, Triassic, and Jurassic grains (Figures 4; Figure S16).

Marlstone samples KZ4 and KZ5 are very fine-grained and moderately sorted with rounded grains. These samples yielded very few zircons (n = 5 in KZ4, n = 10 in KZ5) with all ages ~300 Ma or older, preventing accurate characterization of MDA or age distribution for these rocks previously reported as Cretaceous (Figure 4; Gudjabidze, 2003;Kandelaki & Kakhazdze, 1957).

Marlstone sample KZ6 consists of angular, poorly sorted coarse clasts of calcite, quartz and volcanic fragments. This sample contains a single zircon with a Late Cretaceous age of 89.9 ± 1.7 Ma that matches the previously mapped age, as does the MLA based primarily on this age (90.0 ± 2.2 Ma), though the YGC age is 208.6 ± 2.1 Ma from four grains (Figures 3 and 4; Gudjabidze, 2003;Kandelaki & Kakhazdze, 1957). Other grains are Jurassic and older, with the greatest density of ages centred at ~300 Ma (35%).

Sandstone sample RU1 from the Northern Caucasus (Bechasyn Zone) yields an Early Jurassic YGC age of 178.3 ± 2.3 Ma (32 grains), an Early Jurassic MLA of 181.8 ± 0.5, and a Middle Jurassic YSG age of 166.4 ± 3.7 Ma. The YGC age and MLA match the previously reported biostratigraphic age of the sample (Figure 2b;Nalivkin, 1976;Potapenko, 1964). This sample consists of a nearly unimodal population of Early Jurassic ages centered at ~185 Ma (76% of grains between 190 Ma and 150 Ma), with a smaller population centred at ~500 Ma (Figure 4; Figure S16).

In conjunction with prior work (Figure S18), the new detrital zircon analyses reveal important relationships between detrital zircon maximum depositional ages (MDAs) and biostratigraphic depositional ages reported on prior maps. In general, detrital zircon MDAs mostly match and thus confirm reported biostratigraphic ages from existing maps (Figures 4 and 6), although important exceptions are discussed below. We found Jurassic zircons in most of the rocks that were previously reported to be of Jurassic age (RU1, SV4, SV5, SV6, SV7, KZ1) and Cretaceous zircons in all purported Cretaceous samples that yielded more than 10 zircons (SV8, KZ3, KZ6; Figures 4 and 6). Such correlation between MDA and mapped age was also reported in previous studies of samples from the Greater Caucasus crystalline core, Dizi series, and Jurassic portions of the Caucasus Basin (Cowgill et al., 2016;Trexler et al., 2022;Vasey et al., 2020).

However, we also find a systematic mismatch between prior biostratigraphic age assignments and MDA metrics for samples that were recovered from units previously assigned Early Jurassic biostratigraphic ages but have Triassic or older MDAs (Figure 6; Cowgill et al., 2016;Vasey et al., 2020). Specific examples are samples SV1, SV2, and SV3 in this study, as well as previously analysed samples NW-GC (Cowgill et al., 2016) and N5 (Vasey et al., 2020). We note that biostratigraphic age assignments in the Greater Caucasus generally result from extrapolating ages from fossil localities along-strike to rocks of similar lithology (e.g. Geguchadze et al., 1985b) and thus may not account for rocks that have similar lithology but different ages. Although it is possible for the MDA to be older than the true depositional age of a rock if no source of young zircon is available at the time of deposition (e.g. Gehrels, 2014;Nelson, 2001), we contend that these rocks are likely Triassic or older in age for the following reasons.

First, and most importantly, every other detrital zircon sample in this study of reported Jurassic or Cretaceous age with more than 10 dated grains (SV4, SV5, SV7, SV8, KZ1, KZ3, KZ6) contains Early Jurassic zircon (Figure 6). Even in Cretaceous samples where only a few Jurassic grains are present (SV8, KZ3), there are still no significant gaps in Jurassic zircon ages between larger peaks (Figures 4 and 6). Second, plutonic and volcanic rocks in the Svaneti region provide substantial local sources of Jurassic zircon (e.g. Okrostsvaridze et al., 2022), with Middle Jurassic crystallization ages confirmed in this study in samples SVGD and SVAN (Figure 5). Thus, it is unlikely that Jurassic rocks deposited in the region would be completely bereft of Jurassic zircon.

Finally, comparison of samples SV2, SV3 and SV4 is particularly telling: all three are lithic-rich quartzite with nearly identical age spectra dominated by peaks at ~450 and ~300 Ma, yet samples SV2 and SV3 have no zircon younger than 200 Ma, whereas sample SV4 has a Middle Jurassic MDA that matches its mapped age (Figure 4). We infer that these three quartzite samples had similar source rocks and depositional environments, based on their proximity to one another and the similarity of their pre-200 Ma provenance signatures. Thus, the simplest interpretation of the lack of Jurassic grains in SV2 and SV3 relative to SV4 is that it reflects a difference in timing of deposition rather than coeval Jurassic deposition in areas of different provenance. We conclude that SV2 and SV3 were deposited in the Triassic, followed by SV4 in the Middle Jurassic.

In sum, we infer that Triassic or older sedimentary rocks are more widespread in the western Greater Caucasus than previously reported. We discuss the implications of this finding below.

Of the samples reported here, SV1 has the oldest MDA, with a Proterozoic youngest population (~560 Ma) and a Devonian youngest single-grain age (~380 Ma). Given the similarity of this subangular, lithic-rich quartzite to rocks reported from the Dizi series, the absence of Late Carboniferous or younger detrital zircon grains, and the proximity of this sample to Dizi series rocks mapped to the south (Figures 3, 4 and 6; Gamkrelidze & Kakhazdze, 1959;Geguchadze et al., 1985d;Gudjabidze, 2003), we interpret SV1 as a Devonian rock deposited as part of the Dizi series, rather than an Early Jurassic rock deposited in the Caucasus Basin. The detrital zircon age spectrum of SV1 shares features with other Dizi series rocks (N2, N3; Vasey et al., 2020) including populations of zircon at ~2.6 Ga, ~2.0 Ga, ~1.0 Ga, and ~600 to 500 Ma (Figure 6). These populations are also well-represented in an amphibolite-facies paragneiss sample from the Greater Caucasus crystalline core (N1; Figure 6; Vasey et al., 2020).

Sample SV1 indicates that Devonian to Early Carboniferous Dizi series rocks comprise a subset of the older rocks that are likely more widespread in the Greater Caucasus than previously reported. Furthermore, we infer that the Dizi series may be a lower-grade, structurally shallower counterpart to Devonian to Early Carboniferous rocks of the Greater Caucasus crystalline core that were deformed and metamorphosed in Late Carboniferous time (Figure 7;Somin, 2011;Vasey et al., 2020), as opposed to representing a distinct ocean basin between continental blocks (e.g. Adamia, Alania, et al., 2011;Şengör et al., 1984). The major peaks in these samples also closely match the ages present in modern African rivers (Niger, Nile, Congo, Zambezi, Orange;Iizuka et al., 2013) and differ from those in modern rivers draining the East European Craton (Volga, Don, Dnieper; Safonova et al., 2010;Wang et al., 2011), indicating a distinctly Gondwanan, rather than Laurussian, affinity for both the Dizi series and the higher-grade Greater Caucasus crystalline core, as has been previously noted (Figure 7;Somin, 2011;Vasey et al., 2020). However, the provenance signature of the Dizi series does not represent the Greater Caucasus crystalline core as a whole, which contains a number of other Paleozoic rock packages and has an overall provenance signature captured in Caucasus Basin strata that is dominated by zircon ages at ~450 Ma from Ordovician metamorphic complexes and at ~300 Ma from Late Carboniferous plutonism (Figure 8;Somin, 2011;Tye et al., 2020;Vasey et al., 2020).

Diagrams showing our interpretations of how paleogeography and detrital zircon age spectra evolved in response to convergent margin processes in the Caucasus. Composite KDEs of detrital zircon data from this study and prior work in the Greater Caucasus for the four distinct depositional phases discussed in the text (left) are coupled with schematic maps and cross-sections of interpreted geodynamic depositional setting (right). Devonian to Early Carboniferous rocks (this study; Vasey et al., 2020) reflect deposition prior to accretion of the Greater Caucasus crystalline core onto Laurussia and contain a detrital zircon signature comparable to modern African rivers (Iizuka et al., 2013) and distinct from Russian rivers draining the East European Craton (EEC; Safonova et al., 2010;Wang et al., 2011). Permian to Triassic rocks (this study; Cowgill et al., 2016;Vasey et al., 2020) indicate deposition in the forearc of a continental arc in the Northern Caucasus and Scythian Platform. Jurassic rocks (this study; Cowgill et al., 2016;Vasey et al., 2020;Trexler et al., 2022) document deposition during active arc-back-arc magmatism, whereas Cretaceous rocks (this study) reflect back-arc basin deposition during a period of magmatic quiescence. Cenozoic closure of the Caucasus Basin resulted in suturing and juxtaposition of these Paleozoic to Mesozoic rocks in the North Georgia Fault System (NGFS). LC, Lesser Caucasus; GC, Greater Caucasus; B-P/ATA, Bitlis-Pötürge/Anatolide-Tauride-Armenian block.

As described above, both samples SV2 and SV3 have Triassic MDAs (Figure 6) that are most simply interpreted as indicating Triassic deposition, despite coming from units mapped as Early Jurassic (Gamkrelidze & Kakhazdze, 1959;Geguchadze et al., 1985aGudjabidze, 2003). Likewise, sample N5 from Vasey et al. (2020), also from the Svaneti region, has a Permian MDA that is significantly older than its Early Jurassic mapped age and likely reflects Permian deposition (Figure 6). As noted above, these samples suggest more widespread Permian to Triassic deposition in the Greater Caucasus than previously recognized.

It is possible that the Permian to Triassic deposits represent younger units within the Dizi series, which has previously been described as Devonian to Triassic in age (e.g. Adamia, Alania, et al., 2011). However, these rocks with Permian and Triassic MDAs are not found mapped Dizi series strata (Figure 3; Gamkrelidze & Kakhazdze, 1959;Geguchadze et al., 1985aGudjabidze, 2003), have very low metamorphic grade similar to strata in the Caucasus Basin, and have foliation and bedding orientations sub-parallel to those in strata of the Caucasus Basin (Trexler et al., 2022). Given that known Dizi series strata have only yielded Devonian or Early Carboniferous MDAs (Vasey et al., 2020), we propose, contrary to prior interpretations, that the Dizi series as presently mapped is restricted in age to the Devonian to Early Carboniferous and that many rocks previously considered Early Jurassic in the Caucasus Basin actually represent a separate and distinct phase of Permian to Triassic deposition (Figure 7).

This proposed distinct phase of Permian to Triassic deposition is consistent with the stratigraphy of the Northern Caucasus, where Tikhomirov et al. (2004) described intensely folded Devonian to Carboniferous strata unconformably overlain by Triassic siliciclastic, carbonate and volcanic strata. The volcanic component of these Triassic rocks has been interpreted as originating in a Permian to Triassic continental-margin arc that formed north of the present-day Greater Caucasus above a north-dipping (present orientation) subduction zone (Figures 2d and 7;Natal'in & Şengör, 2005;Okay & Nikishin, 2015 ;Vasey et al., 2021). We infer that Triassic deposition of samples SV2 and SV3 in the presentday Greater Caucasus, south of this arc, took place in a forearc basin and interpret the Permian to Triassic strata identified in this study as a more distal part of this forearc basin, rather than an early phase of deposition in the back-arc basin that subsequently opened in Jurassic time (Figure 7).

Note that in this interpretation, Triassic to Jurassic (meta)sedimentary rocks in the Greater Caucasus capture a critical transition from forearc to back-arc basin deposition during migration of the Caucasus arc front from north of the present-day Greater Caucasus to the present-day Lesser Caucasus due to slab rollback (Okay & Nikishin, 2015;Vasey et al., 2021). The poor presentday exposure of the Permian-Triassic arc and forearc basin could be the result of strike-slip dissection of the arc during a period of magmatic quiescence and oblique subduction (e.g. Natal'in & Şengör, 2005), a process that has similarly dissected volcanic arcs in the long-lived convergent margins of the North American Cordillera (e.g. Cecil et al., 2018;Gehrels et al., 2009).

Samples RU1, SV4, SV5, SV6, SV7 and KZ1 have Jurassic MDAs broadly matching their previously mapped depositional ages (Figures 2 and 3; e.g. Gamkrelidze & Kakhazdze, 1959;Gudjabidze, 2003;Kandelaki & Kakhazdze, 1957;Potapenko, 1964) and thus provide insight into the major phase of back-arc basin evolution 3 with underlying geology removed and samples coloured according to tectonostratigraphic packages outlined in Figure 8b. (b) Composite KDEs of detrital zircon rock (black text) and modern catchment (blue text) samples compiled for distinct tectonostratigraphic packages from this study and prior work. Most samples show an affinity with the Greater Caucasus crystalline core, with prominent peaks at ~450 Ma (green bars) and ~300 Ma (pink bars); additional peaks at ~240 Ma (orange bars) and ~170 Ma (blue bars) reflect Triassic arc and Jurassic back-arc magmatism, respectively. In Svaneti, such samples include siliciclastic-dominated strata north of the Dizi series and Triassic to Jurassic deposits, and in Kazbegi they include thrust sheets of Jurassic and Cretaceous strata. In contrast, samples in the southern volcaniclastic-dominated strata of Svaneti are dominated by a peak at ~170 Ma and lack the peaks at ~450 and ~240 Ma associated with the crystalline core and Triassic arc, respectively. We interpret these latter samples to lie south of a Late Cenozoic suture, based on their affinity with the zircon spectra from the Lesser Caucasus arc (bottom middle panel), which also lacks peaks at ~450 and ~240 Ma and has a slightly older Carboniferous peak. (c) Multidimensional scaling (MDS) plot of all samples discussed in this study, in which distance between two samples on the plot indicates their Kolmogorov-Smirnov (KS) similarity. Plot created using IsoplotR (Vermeesch, 2018). Samples coloured according to composite groupings in Figure 8b, except for Devonian to Early Carboniferous samples (red) and Northern Caucasus (dark blue) samples not included in Figure 8b. in the Caucasus Basin (e.g. Cowgill et al., 2016;Vasey et al., 2021;Vincent et al., 2016;Zonenshain & Le Pichon, 1986).

Sample RU1 in the Bechasyn Zone of the Northern Caucasus consists of a nearly unimodal peak centered around ~185 Ma (Figure 4), likely reflecting its proximity (<10 km) to Early Jurassic volcanic rocks in the Bechasyn Zone (Figure 2; Potapenko, 1964;Hess et al., 1995). The U-Pb zircon ages spanning from ~195 to 175 Ma within this peak are consistent with previously reported Early Jurassic 40 Ar/ 39 Ar integrated and plateau ages in plagioclase, K-feldspar, and biotite from Bechasyn Zone volcanic rocks (Hess et al., 1995). Given the lack of primary 40 Ar/ 39 Ar data reported in prior work and the susceptibility of these minerals to Ar diffusion at relatively low temperatures (McDougall & Harrison, 1999), the numerous U-Pb zircon ages in sample RU1 provide a useful constraint on the duration of Early Jurassic magmatism within the Bechasyn Zone. This new constraint supports the proposal in Vasey et al. (2021) that southward migration of the Caucasus arc front from Late Triassic to Early Jurassic time is indicated by a Triassic (~220 Ma) cessation of magmatism in the Scythian Platform, followed by Early Jurassic (~195 to 175 Ma) magmatism in the Bechaysn Zone to the south. Vasey et al. (2021) attributed such arc migration to forearc rotation during Triassic collision of Iran with Eurasia to the east and slab rollback in the continental-margin subduction zone to the west (Figure 2d). In the Early Jurassic, there may not have been a clear distinction between arc, forearc and back-arc volcanism/deposition, with the Bechasyn Zone effectively representing an extensional arc complex (e.g. Busby, 2011).

By the Middle Jurassic (~175 to 165 Ma), the arc front had migrated south to the Lesser Caucasus, with coeval arc magmatism in the Lesser Caucasus and back-arc basin deposition/volcanism in the Caucasus Basin to the north (Figure 7; e.g. Cowgill et al., 2016;Okrostsvaridze et al., 2022;Vasey et al., 2021;Vincent et al., 2016;Zonenshain & Le Pichon, 1986). Samples SV4, SV5 and SV6 all have Middle Jurassic MDAs reflecting deposition during this time, and samples SVGD and SVAN confirm Middle Jurassic magmatic activity (Figures 4 and 5). Late Jurassic (~165 to 145 Ma) MDAs in samples SV7 and KZ1 also confirm continued back-arc basin deposition into the Late Jurassic (Figure 4). As a whole, these Middle to Late Jurassic samples in the Caucasus Basin generally contain age peaks at ~450 and ~300 Ma, as well as a range of older Palaeozoic and Precambrian ages, indicating Jurassic erosion of material from the Greater Caucasus crystalline core (e.g. Somin, 2011;Tye et al., 2020;Vasey et al., 2020). They also tend to have minor Jurassic age peaks from ~180 to 160 Ma, consistent with the ages of SVGD, SVAN and other Jurassic plutonic rocks from Svaneti (Okrostsvaridze et al., 2022) that likely reflect back-arc basin magmatism (Figures 4,5 and 7). A few samples (SV4, SV7, KZ1) also contain Permian to Triassic zircon populations from ~260 to 220 Ma, which we interpret to result from Jurassic erosion of the Permian to Triassic arc in the Scythian Platform (Figure 7;Natal'in & Şengör, 2005;Okay & Nikishin, 2015;Vasey et al., 2021).

Cretaceous (~145 to 65 Ma) rocks indicate a shift in sedimentary source within the Caucasus Basin. The Greater Caucasus crystalline core continued to erode, remaining a sediment source for samples with Cretaceous MDAs (SV8, KZ3, KZ6), based on their age populations at ~450 and ~300 Ma along with Precambrian and older Palaeozoic grains (Figure 4). The sparse distribution of Jurassic grains in these samples suggest that neither the Jurassic arc nor Jurassic volcanic centers in the back-arc basin were major sediment sources for the Caucasus Basin during Cretaceous time.

Additionally, Cretaceous zircons are also sparse; samples SV8 and KZ3 contain only small populations very close to their MDAs, whereas sample KZ6 contains only a single Late Cretaceous grain out of 79 zircon ages, with all other ages either Middle Jurassic or older (Figures 4 and 7). We interpret this paucity of Cretaceous detrital zircons, in conjunction with the scarcity of Cretaceous magmatic rocks in the western Greater Caucasus (Figure 3; e.g. Gudjabidze, 2003), to indicate that Cretaceous magmatism, though not entirely absent in the Caucasus Basin, was much less significant than earlier Jurassic magmatism (Figure 7). The dominance of shallow-marine carbonates in Cretaceous Caucasus Basin strata indicative of lower rates of subsidence (e.g. Vincent et al., 2016;Cumberpatch et al., 2020;Trexler et al., 2022), as well as an overall scarcity of zircons of any age in Cretaceous strata (e.g. samples KZ4, KZ5), also suggest a period of limited extension, erosion, and deposition.

A lull in Cretaceous magmatism and extension following a pulse of Jurassic activity is consistent with observed episodicity in in both real-world (e.g. Tyrrhenian Sea; Faccenna et al., 2001) and model (Magni, 2019;Magni et al., 2014) back-arc basins, in which initial rapid extension and high magmatic flux induced by slab rollback gives way to a period of relative quiescence as rollback slows. In the Caucasus, the Cretaceous likely marks a period of such quiescence following the initial Middle Jurassic extensional and magmatic event that opened the Caucasus Basin (Figure 7; Vasey et al., 2021). The Cretaceous also coincides with a major phase of basin opening in the Black Sea (e.g. Okay et al., 1994;Stephenson & Schellart, 2010;Zonenshain & Le Pichon, 1986), so extension may have migrated westward from the Caucasus to the Black Sea region at this time.

Prior detrital zircon data indicate a provenance distinction between rocks associated with the Greater Caucasus crystalline core to the north and rocks associated with the Lesser Caucasus arc to the south. Those associated with the Greater Caucasus crystalline core generally have major populations of zircon ages at ~450 and ~300 Ma, in addition to numerous Precambrian and Paleozoic zircons, whereas those associated with the Lesser Caucasus arc tend to be dominated by near-unimodal populations of zircon ages at, or younger than, ~170 Ma (Cowgill et al., 2016;Trexler et al., 2022;Tye et al., 2020). Zircon age spectra within pre-Miocene Caucasus Basin strata tend to define one or the other of these end-member signatures with minimal mixing between them, which has formed the basis for the argument that there is a shift in provenance across a suture zone within the Caucasus Basin marking the area of Cenozoic closure of a relatively wide basin during Arabia-Eurasia collision (Cowgill et al., 2016(Cowgill et al., , 2018;;Trexler et al., 2022;Tye et al., 2020), though this view has been contested (Mosar et al., 2022;Vincent et al., 2016Vincent et al., , 2018)). Trexler et al. (2022) identified the North Georgia Fault System (NGFS) between these two provenance domains (Figures 2 and 3) and suggested that one or more of the major thrust faults within the NGFS represents the Cenozoic suture. Importantly, the NGFS juxtaposes regions with distinct lithostratigraphy, with siliciclasticdominated rocks to the north and dominantly volcaniclastic strata to the south.

Results reported here and in prior work are from samples that span and lie within the NGFS, further characterizing these provenance domains and providing additional constraints on the location of the hypothesized Cenozoic suture. In the western Svaneti region of Georgia, all samples in the northern, siliciclastic-dominated portions of the range (SV1, SV2, SV3, SV4, SV7, SV8) appear to share a provenance affinity with the Greater Caucasus crystalline core (Figure 3; Tye et al., 2020;Vasey et al., 2020), with all samples yielding major age peaks at ~450 and ~300 Ma, along with significant Precambrian and older Paleozoic grains (Figures 4 and 8). This result is consistent with recent detrital zircon data from modern river catchments that drain this region that also contain peaks at ~450 and ~300 Ma (Figures 3 and 8;Forte et al., 2022), suggesting that the major provenance break is south of all these samples.

Unfortunately, we are unable to determine if a suture lies between the siliciclastic-dominated rocks and volcaniclastic-dominated strata to the south because the two new samples within volcaniclastic strata contain too few detrital zircon analyses for robust provenance characterization (39 in SV5, 10 in SV6). Both samples contain Jurassic age populations at ~170 Ma (Figure 4), suggesting that this would be a major part of their detrital zircon signature, as it is for previous bedrock (Cowgill et al., 2016;Trexler et al., 2022) and modern river (Tye et al., 2020) analyses sampling volcaniclastic strata in this region (Figures 4 and 8). Sample SV5 also contains several zircons at ~450 and ~300 Ma potentially consistent with the Greater Caucasus crystalline core as a sediment source, though it is difficult to determine how significant these populations are with so few total ages. This sample lies to the south of samples from prior work with nearly unimodal age peaks at ~170 Ma (SW-GC and its replicate CT130924-9A; Cowgill et al., 2016;Trexler et al., 2022), making it unlikely that a significant provenance break lies to the south of SV5. As a result, we consider it plausible that the hypothesized provenance break lies between the siliciclastic-and volcaniclastic-dominated domains, while recognizing that the lack of robust analyses in the latter domain make this inference somewhat speculative.

In the eastern Kazbegi region, all samples with more than 10 dated zircons (KZ1, KZ3, KZ6) contain signatures indicative of a provenance affinity with the Paleozoic crystalline core of the Greater Caucasus and thus lie north of the inferred Cenozoic suture, with major age populations at ~450 and ~300 Ma, as well as numerous Precambrian and older Paleozoic grains (Figures 4 and 8; Cowgill et al., 2016;Tye et al., 2020). Even samples with few zircons (KZ2, KZ4, KZ5) still contain several grains of ~300 Ma age (Figure 4). This signature is also consistent with previously published detrital zircon spectra from modern river catchments that drain the Kazbegi region and are dominated by ages of ~450 and ~300 Ma (Figures 3 and 8;Forte et al., 2022;Tye et al., 2020). Thus, these data suggest that the Cenozoic suture lies along, or to the south of, the stratigraphic juxtaposition of Cretaceous rocks on Cenozoic strata (Figures 3 and 8; Kandelaki & Kakhazdze, 1957;Trexler et al., 2022). We are not aware of detrital zircon results from rocks immediately south of this juxtaposition. However, the presence of carbonate olistoliths in Cenozoic strata immediately south of this juxtaposition (Cowgill et al., 2016;Trexler et al., 2022) and the persistence of a Greater Caucasus crystalline core signature in all rocks north of the fault make this a plausible location for the hypothesized provenance break across a suture zone in the Greater Caucasus.

As a result, the new detrital zircon data presented here are consistent with the hypothesis of Trexler et al. (2022) that, regionally, the southernmost faults of the NGFS are the most likely candidate structures for the suture representing final Late Cenozoic closure of the Caucasus Basin (Figure 2). Although the precise locations of these structures across the western Greater Caucasus are not well constrained (Trexler et al., 2022) and detrital zircon data south of these structures are sparse, extant bedrock and modern river detrital zircon data suggest that this southern region is dominated by a zircon age population centred at ~170 Ma reflecting proximity to the Lesser Caucasus arc, with Precambrian and Paleozoic grains relatively sparse (Figure 8; Cowgill et al., 2016;Trexler et al., 2022;Tye et al., 2020). Notable exceptions to this pattern are modern river samples directly draining rocks of the Paleozoic crystalline massifs of the Lesser Caucasus, including the Dzirula, Khrami, and Loki massifs, which contain a major population of ages at ~300 Ma and a range of scattered Precambrian and older Paleozoic ages, similar to the Greater Caucasus crystalline core (Figure 8; Tye et al., 2020). However, Tye et al. (2020) note that these samples lack a major age peak at ~450 Ma that is common in samples affiliated with the Greater Caucasus core (Figure 8). As a result, although sources of ~300 Ma zircon may be present in rocks affiliated with the Lesser Caucasus and sources of ~170 Ma may be present in rocks affiliated with the Greater Caucasus core, the relative abundance of these ages in conjunction with the additional Greater Caucasus core peak at ~450 Ma still outline a provenance contrast within the relict Caucasus Basin.

Given the increased recognition of multiple sources of zircons of the same age in the Caucasus region, an important next step in understanding the evolution of the Caucasus Basin is combining detrital zircon U-Pb analyses with zircon trace element and/or Hf isotopic analyses. Pilot zircon Hf analyses in igneous rocks presented here illustrate the potential regional utility of these techniques. First, the largely positive initial εHf values in ~170 Ma samples from the Greater Caucasus (SVGD; SVAN) are consistent with a more depleted source for Jurassic back-arc magmatism in the Caucasus Basin, which was also determined from a recent review of the trace-element geochemistry of this region (Vasey et al., 2021). Although a zircon Hf signature remains to be reported for the ~170 Ma arc rocks of the Lesser Caucasus (Figure 5), we anticipate potentially lower initial εHf values in the arc owing to a more enriched mantle and/or crustal melt source relative to a more depleted mantle source in the back-arc basin.

Additionally, our Hf isotopic analyses from an ~300 Ma granite in the Greater Caucasus crystalline core (SVGR) yield negative initial εHf values (-2.7 ± 2.5) indicative of crustal melt sources consistent with our inference of collisional magmatism in the Late Carboniferous. Importantly, the initial εHf value of SVGR is still higher than that reported from a similarly-aged rock in the Dzirula Massif of the Lesser Caucasus (-6.9 ± 2.7; Mayringer et al., 2011), suggesting that zircon geochemistry may also provide a useful means for distinguishing Late Carboniferous zircons sourced from the Greater Caucasus from those derived from the Lesser Caucasus.

Our results provide new insight into the basin evolution that accompanies a protracted convergent margin tectonic setting. Based on the tectonic evolution described above, we contend that the Greater Caucasus appears to contain a record of at least four distinct periods of marginal basin evolution, including (1) Devonian to Early Carboniferous deposition of the Dizi series prior to Late Carboniferous accretion of the Caucasus crystalline core to Laurussia (present-day Eurasia), ( 2) Permian to Triassic deposition in a probable forearc position of a north-dipping subduction system with an arc in the Scythian Platform, (3) Jurassic deposition in a magmatically active back-arc basin that opened as the arc migrated southward to the Lesser Caucasus due to slab rollback, followed by a transition to (4) Cretaceous deposition during a magmatic lull within the same back-arc basin as slab rollback slowed (Figure 7). Such episodicity in convergent margin basin development mirrors the increasingly recognized episodicity of subduction-related arc magmatism in the North American Cordillera (e.g. Cecil et al., 2018;Chapman et al., 2021;DeCelles et al., 2009;Ducea & Barton, 2007;Gehrels et al., 2009;Paterson & Ducea, 2015).

In particular, the present study suggests that a transition from a forearc to back-arc basin during slab rollback, which was previously undetected in the Caucasus, may be more common than previously recognized and may be a key step in the formation of continental lithosphere within these systems (Figure 1). Migration and reactivation of arcs, forearc basins, and back-arc basins appear to be common processes in active Mediterranean arc systems (e.g. Balázs et al., 2022;Corradino et al., 2022), and similar transitions may yet emerge in additional relict arc systems. This study and prior work (Vasey et al., 2021) suggest that the forearc to back-arc transition in the Caucasus region occurred over an interval of ~40 m.y., between a ~40 m.y. phase of arc magmatism during Permian to Triassic time in the Scythian Platform (~260 to 220 Ma) and a subsequent ~40 m.y. pulse of arc magmatism during Early to Late Jurassic time in the Lesser Caucasus (~180 to 140 Ma). This general tempo is broadly consistent with the identification and timing of arc flare-ups in Cordilleran continental convergent margin systems. Those systems, including the Coast Mountains and Sierra Nevada batholiths, record lulls of relatively minimal magmatism separated by intervals of increased arc magmatism that lasted for <50 m.y., during which the bulk of magmatism was concentrated over only ~10 to 20 m.y. (e.g. Balgord et al., 2021;Ducea et al., 2015;Paterson & Ducea, 2015). Our results suggest that geodynamic triggers like slab rollback can serve as one potential underlying cause for such lulls between episodes of voluminous arc magmatism.

Detrital zircon U-Pb geochronology from Paleozoic and Mesozoic (meta)sedimentary rocks in the Greater Caucasus highlights four distinct episodes in the evolution of this protracted convergent margin. (1) Devonian to Early Carboniferous MDAs and significant age peaks affiliated with Gondwana in Dizi series and related rocks indicate deposition prior to the Late Carboniferous accretion of the Greater Caucasus crystalline core onto the southern margin of Laurussia. ( 2) Permian to Triassic MDAs in rocks from units formerly mapped as Early Jurassic indicate a previously unrecognized phase of deposition, likely in the forearc of a coeval continental-margin volcanic arc to the north in the Scythian Platform and Northern Caucasus.

(3) Jurassic rocks document deposition in an extending, magmatically active back-arc basin, and (4) limited Jurassic to Cretaceous grains in Cretaceous rocks indicate a period of magmatic quiescence. Maximum depositional ages (MDAs) in the Greater Caucasus generally match previously reported biostratigraphic depositional ages, except in the case of rocks from units mapped as Early Jurassic with Permian to Triassic MDAs.

In rocks ~20 km north of the Greater Caucasus range front in the North Georgia Fault System (NGFS), major age peaks at ~450 and ~300 Ma, as well as a minor peak at ~240 Ma and significant Precambrian grains, indicate an affinity with the Greater Caucasus Paleozoic crystalline core and the Permian to Triassic volcanic arc in the Scythian Platform and Northern Caucasus. As a result, a hypothesized Late Cenozoic suture in the Greater Caucasus from closure of the Caucasus Basin likely lies along the southernmost faults of the NGFS. Limited samples from the Greater Caucasus south of NGFS exhibit an affinity with the Lesser Caucasus volcanic arc and adjacent crystalline massifs, providing preliminary evidence of a provenance break across the NGFS. Our results highlight episodicity in the magmatic and depositional history of the Caucasus convergent margin in intervals of ~40 m.y., comparable to processes described in the North American Cordillera, illustrating the need for additional study of the episodic processes that have led to terminal continent-continent collision throughout the Alpine-Himalayan belt.