U–Pb zircon age and chronology of the Torfufell central volcano: implications for timing of rift relocation in North Iceland

The Late-Miocene Torfufell central volcano (ToCV), situated between the now extinct Snæfellsnes-Húnaflói rift zone and the presently active rift in North Iceland, provides an excellent opportunity to recreate the construction history of a volcanic edifice. We present new U–Pb zircon ages from six silicic units of the ToCV. The results range from 7.15 ± 0.12 to 6.76 ± 0.02 Ma, taken here to represent a ~ 400 kyr time-span for silicic activity at the volcano. Before that, the central volcano had produced basaltic lavas for 600–800 kyr, implying that it was active for ~ 1–1.2 Myr. A stratigraphically documented ~ 1 Myr hiatus above the volcano is contemporaneous with, but shorter than, a major unconformity in the Flateyjarskagi peninsula, considered to result from a major rift relocation in North Iceland. The new U–Pb ages show that silicic volcanism at the ToCV took place 1–2 Myr earlier than assumed previously and nearly synchronously with the rift relocation. As the age progression of the ToCV and the neighboring 5–6 Ma Tinná central volcano conflicts with the generally established geotectonic framework of central N-Iceland, we propose that these two volcanoes were formed at a leaky transform zone that developed to accommodate the rift relocation, with the ToCV formed at its junction with the embryonic rift zone, thus marking the initiation of the presently active rift in North Iceland. Since then, the two volcanoes have drifted away from the rift system due to plate spreading and migration of the plate boundary relative to the Iceland mantle plume.


Introduction
Iceland is situated at the intersection of the Mid-Atlantic Ridge (MAR) and the Iceland mantle plume and is constructed by ridge/plume interaction (e.g., Schilling 1973;Vink 1984).Today, volcanism in Iceland is confined to volcanic zones, each delineated by discrete volcanic systems (e.g., Jakobsson et al. 2008).The volcanic zones are further subdivided into rift-zones, oblique rift-zones (or "leaky" transform zones), propagating rifts and intra-plate belts (e.g., Einarsson 2008).Approximately 30 active volcanic systems are identified within the volcanic zones, and each system features either an elongate fissure swarm, a localized central volcano, or both (e.g., Thordarson and Höskuldsson 2008;Hjartarson and Saemundsson 2014).The fissure swarms trend subparallel to the respective volcanic zone and typically consist of monogenetic fissures erupting only basaltic magmas, while the central volcanoes are polygenetic (composite) and erupt basaltic to silicic magmas.While a typical lifetime of a volcanic system has been estimated around 0.5-1.5 Myr (Jakobsson et al. 1978;Jakobsson 1979;Saemundsson 1978;1979), a lifespan of 2.3-2.4Myr was evaluated in a recent study for a fossilized volcanic system in East Iceland featuring both a fissure swarm and a central volcano, where the central volcano within the system was active for ~ 1 Myr (Askew et al. 2020).However, central volcanoes in Iceland appear to have potentials for longer lifetimes than that, as longevities of silicic magmatism within individual central volcanoes have been recorded to reach > 1 Myr (Flude et al. 52 Page 2 of 17 2008) and ≥ 1.5 Myr (Banik et al. 2018), with ≥ 2.8 Myr documented as the longest lifespan of silicic activity within an Icelandic central volcano (Carley et al. 2017).Comparably, lifespans of composite volcanoes in other geodynamic settings, such as subduction zones and extensional basins, are commonly estimated around 1-1.3 Myr or less (e.g., Hildreth 2007;Tatsumi et al. 2020;Karaoğlu et al. 2010).
Arrangement of the rift zones in Iceland is explained by the west-northwestward migration of the mid-ocean ridge system in the North Atlantic relative to the Iceland mantle plume, which has induced repeated eastward relocations of the onshore part of MAR-rift segments that are accommodated by transform zones.The present distribution of the rifting segments in Iceland is established by three principal rift zones (Fig. 1a); the ~ N-S trending Northern Volcanic Zone (NVZ) in the north and two subparallel NE-SW trending branches in the south-the Eastern and Western Volcanic Zones (EVZ and WVZ) (e.g., Saemundsson 1979;Einarsson 2008).These three segments exemplify different stages in the evolution of a rift axis across Iceland, where the southern part of the EVZ is propagating through older crust to the SW, the WVZ is waning and receding, and the NVZ may be referred to as an established (or steady-state) rift as it accommodates the full spreading rate in North Iceland (e.g., Einarsson 2008).The division between the NVZ and the EVZ is nevertheless arbitrary, as these two domains form a single continuous structural identity.Two major transform zones cross Iceland with W-E and WNW-ESE trends: the Tjörnes fracture zone (TFZ) in the north and the South Iceland seismic zone (SISZ) in the south (e.g., Saemundsson 1979;Jóhannesson 1980).The NVZ is offset towards East from the WVZ along the transverse Mid-Iceland Belt (MIB) which has been described as a "leaky" transform fault (e.g., Óskarsson et al. 1985), i.e., comprising a small component of opening (extension), sufficient to induce local volcanism, although evidence of strike-slip movement in this belt is not apparent (e.g., Hjartardóttir and Einarsson 2021).The extension component has been proposed to be caused by the opposite sense of rotation of crustal blocks to the N and S (Karson 2017).Intra-plate volcanism occurs in the Snaefellsnes and Öraefi Volcanic Belts (SVB and ÖVB), with the former representing renewed volcanism through crust formed in the now extinct Snaefellsnes-Húnaflói rift zone (SHRZ) (Harðarson et al., 1997;Saemundsson 1979) and the latter considered as an embryonic rift (e.g., Thordarson and Höskuldsson 2002).These intra-plate volcanic zones, along with the propagating part of the EVZ, are typified by alkalic and/or transitional magmatism while the rift zone magmas are tholeiitic (e.g., Jakobsson et al. 2008).
In North Iceland, the last major rift relocation took place when the northern segment of the SHRZ shifted eastward and a new rift axis initiated at the NVZ.The classically quoted timing of the relocation, ~ 6-7 Ma (e.g., Saemundsson 1979;Jóhannesson 1980), appears to be further supported by new 40 Ar/ 39 Ar ages from lavas within the Tjörnes sedimentary sequence (Hall et al. 2023).The proposed alignment of the palaeo SHRZ runs along the axis of a syncline (Fig. 1a, b, c), interpreted to be induced by subsidence due to loading of erupted material, causing the volcanic successions to dip towards the then active rift (Saemundsson 1974;1979;Jóhannesson 1980;Harðarson et al. 2008).A major discontinuity (unconformity/hiatus) is recorded within the basaltic lava dominated stratigraphical succession along the Flateyjarskagi peninsula, west of the NVZ (Fig. 1b, c).This unconformity is interpreted to have formed because of the rift relocation, where lava flows beneath the hiatus were formed in the SHRZ and lavas above the hiatus were formed in the NVZ (Saemundsson 1974).Campaigns undertaken to constrain the timing and length of this hiatus indicate that it spans the time from ~ 10 to ~ 6 Ma in the north while the time gap decreases towards the south where it spans the time from ~ 7 to ~ 5 Ma (Jancin et al. 1985;Cotman 1979).These age determinations, along with a number of other datings in central North Iceland (Aronson and Saemundsson 1975;Saemundsson et al. 1980;Everts, 1975;Kristjánsson et al. 1992;Fig. 1b, c), were obtained by the K-Ar method which, although certainly useful, has proved to be problematic in Iceland, mostly due to low concentrations of K 2 O in Icelandic basalts in combination with problems arising from the mobility of K and resulting sensitivity to the thermal history of the rock.These limitations are largely overcome by the more recent 40 Ar/ 39 Ar technique (e.g., Lee 2015) which is now considered an optimal method for dating Icelandic rocks (e.g., Sigmarsson et al. 2012).This superseding method has been applied to numerous regional basaltic dykes in central North Iceland, dated in particular to provide spatio-temporal constraints on the ~ 6-7 Ma rift relocation (Garcia et al. 2003).A number of basaltic lavas have also been dated by the 40 Ar/ 39 Ar method in the central north region, including lavas in Mt.Hólafjall (Supakulopas 2018;Fig. 1b, c;Fig. 2) that provided indications of a major hiatus in the present study area.Age determinations of silicic portions of several of the fossilized and excavated central volcanoes in the central north region of Iceland have furthermore been published, i.e., K-Ar ages from two adjacent volcanoes in the Flateyjarskagi peninsula (Jancin et al. 1985), 40 Ar/ 39 Ar ages from rhyolite lavas within the Tinná (Hjartarson 2003) and Torfufell (Supakulopas 2018) central volcanoes, and U-Pb zircon ages from intrusive units of the Víðidalur and Laxárdalur central volcanoes (Carley et al. 2020).
The age data mentioned, compiled in Fig. 1b and c, have contributed significantly to the understanding of the volcano-tectonic evolution of central North Iceland, not least by providing constraints on the Flateyjarskagi hiatus and timing of the last major rift relocation in the region.However, a closer look at these ages shows that the distribution does not follow a straightforward progression away from the NVZ as assumed for mid-ocean ridge spreading centers.The ages record a general younging to the west, from 12 to 7-7.5 Ma, in the direction away from the Flateyjarskagi unconformity and towards the SHRZ syncline, while lavas above the unconformity are 4-5 Ma.Very young ages are recorded in the Skagafjörður region (Everts 1975;Hjartarson 2003) which have been interpreted as recommenced volcanism, long after extinction of the SHRZ, while lavas of Quaternary age in the Eyjafjörður region have been assumed to be remains of large regional lava flows reaching tens of kilometers out from the NVZ (Guillou et al. 2010).In the Eyjafjörður region (west of the Flateyjarskagi unconformity), a general younging is implied by lava ages from the coastline (i.e., the northernmost Tröllaskagi and Flateyjarskagi peninsulas) and into the land which appears to be nonexistent in the western part of the central North region.This younging in the Eyjafjörður region may possibly, at least to some extent, be explained by the general SE dipping of the strata pile (as measured by Jancin et al. 1985) below the Flateyjarskagi unconformity.The apparently non-existent younging from the coastline into the land in the western part of the central North region is unexplained, although it should be noted that in the west, the age data is scarcer than in the east while a large part of the available dates are from dykes instead of lavas.A relatively wide range of ages is recorded in some locations, especially in the Flateyjarskagi peninsula, which may largely be explained by sampling at different stratigraphic levels, while some may possibly be explained by inaccuracy in the K-Ar method.
The primary objective of this study is to address the timing and longevity of silicic (SiO 2 ≥ 65 wt%) magmatism and volcanism within the Late-Miocene Torfufell central volcano (ToCV), located in the region of the Eyjafjörður fjord, by the Tröllaskagi peninsula's southern end (Figs.1c  and 2), between the presently active NVZ and the northern segment of the palaeo SHRZ.This is a key area in the evolving understanding of rift transfer in North Iceland, allowing us to explore a potential continuation of the Flateyjarskagi unconformity to the south and to examine an inferred Basemap: Topography, lakes, rivers, roads and farms sourced from the National Land Survey of Iceland relation of the ToCV with the neighboring Tinná central volcano (TiCV), previously considered to be of similar age and potentially the same volcano (e.g., Hjartarson 2003;2005).As the ToCV is dissected by rivers and glaciers, it provides an excellent opportunity to recreate the construction history of a volcanic edifice.Contrary to basaltic lavas that may flow tens of kilometers to be observed far away from the eruptive fissure, central volcanoes are localized features, abundantly comprising rocks with restricted spreading such as silicic intrusions and rhyolite lavas of high viscosity and inability to flow long distances, thus providing ages at or very close to the eruptive site.Accordingly, addressing the timing of silicic magmatism of ancient volcanic centers provides geometric relationships that can significantly improve our understanding of the geodynamic evolution of Iceland.Of particular interest are the rhyolites of the Tröllaskagi peninsula which-being situated between the NVZ and the northern segment of the palaeo SHRZ-may hold the key to a more refined comprehension of rift transfer in North Iceland.
This paper details new U-Pb zircon ages from six silicic formations belonging to the ToCV via secondary ion mass spectrometry (SIMS; five formations) and chemical abrasion-isotope dilution thermal ionization mass spectrometry (CA-TIMS; one formation) methods.The SIMS method is a high-spatial resolution in situ technique and was employed in an early stage of the study to obtain, in one batch, a reliable idea about the age of the volcano and its chronological evolution.The CA-TIMS method is a mineral-dissolution technique that results in higher attainable precision than the SIMS method and was employed in a later phase to provide accurate age data for the youngest outcropping rhyolite lava and thereby a more precise conception of the absolute longevity of rhyolite activity at the volcano and, simultaneously, better constraints on the duration of a time gap in the strata pile, indicated by earlier research and supported but abridged by the SIMS age data.The present study was conducted synchronously with field mapping and palaeomagnetic measurements in the study area, with corollary sampling and whole-rock geochemical analyses (detailed separately).In the present paper, we couple the six new U-Pb zircon ages with these observations, as well as with the abovementioned 40 Ar/ 39 Ar ages from the study area (by Supakulopas 2018), to address the development and life-span of the ToCV and the volcano-tectonic evolution of central North Iceland.

Geological background
The region of the Eyjafjörður fjord is composed of sequences of Neogene flood basalts, commonly interbedded with thin layers of lithified sediments which are typically distinctively red and consisting largely of weathered, wind-blown tephra mixed with ancient soils (Saemundsson 1979;Saemundsson et al. 1980;Roaldset 1983).As the originally sub-horizontal lava pile now tilts regionally a few degrees to the south (SW in the northern part, turning to SE in the southern part; Saemundsson et al. 1980), the oldest lavas occur near the mouth of the fjord while strata from Early Pleistocene are present in the south at the head of the valley Eyjafjarðardalur.Several Miocene central volcanoes have been identified in the region (e.g., Hjartarson and Saemundsson 2014).Two of these, the Torfufell and Tinná central volcanoes, are ~ 15 km apart and have been proposed to be related and hence of similar age, 5-6 Ma (e.g., Hjartarson 2003;2005), while Jóhannesson (1991) suggested a higher age for the ToCV than the TiCV, i.e., 6-7 Ma.A stratigraphically constrained ToCV rhyolite yielding 40 Ar/ 39 Ar age of 7.12 ± 0.10 Ma in Mt.Hólafjall (Supakulopas 2018) indicated an even greater age than both proposals.
The present study area is approximately 40 km south of the largest town in North Iceland, Akureyri (Fig. 1a and Fig. 2).The main features of the bedrock in the area are illustrated in small-scale geological maps of Iceland (e.g., Hjartarson and Saemundsson 2014), delineating silicic volcanic formations-centered in Mt.Torfufell and enveloped by basaltic lavas-just beneath the boundary between the Miocene and Pliocene periods.Stratigraphic profiling and palaeomagnetic measurements have demonstrated the stratigraphy of Mt.Hólafjall in the eastern part of the area in more detail (Kristjánsson et al. 2004;Supakulopas 2018).Although only partly covering the ToCV, the study by Supakulopas (2018) indicated a ~ 1.5 Ma hiatus between the rhyolite lavas of the ToCV and the overlying basalts.

Field mapping and sample selection
The present study was a part of a field mapping (including stratigraphic and geomagnetic profiling) and geochemical investigation in the area around Mt. Torfufell (detailed separately).The field observations demonstrate that tectonic uplift and glacial erosion has exposed extrusive rocks and shallow intrusions of the ToCV.Outcropping silicic and intermediate extrusive units belonging to the volcano are thickest in the northern part of Mt.Torfufell (Fig. 2), where they form a ~ 400-m-thick succession.The lowermost ~ 30-60 m of this pile are composed of andesite and dacite lavas, which are overlain by thicker rhyolite lavas.Different palaeomagnetic polarities and sediments and individual basaltic to andesitic lavas separating the rhyolite lavas manifest that these are several discrete formations.Below the silicic/intermediate extrusive formations is a 200-300-m-thick stack of basaltic lavas, featuring unusually thin (individual layers typically < 5 m) ā´a lavas.This stack is separated from the silicic/intermediate volcanic 52 Page 6 of 17 products in several outcrops by a notable tephra-, smectite-, and fossil-bearing sediment horizon.According to our geochemical data, the composition of these unusually thin ā´a lavas is rather evolved and more evolved than underlying plateau basalts (~ 50-52 compared to ~ 48-49 wt% SiO 2 ; samples 24072017-0 to -3 and 20180903-1 in Table S1 of the Supplementary Material), indicating longer residence time of the magma in the shallow crust.Such unusually thin, phenocryst-free basaltic lava flows have been observed to be abundantly erupted from Icelandic central volcanoes (Saemundsson 1979) and interpreted to be abnormally thin as they were erupted onto a sloping landscape resulting from the buildup of central volcanism (Walker 1963).Therefore, we have a reason to assume that these thin ā´a lavas below the silicic/intermediate extrusive portion of the ToCV are sourced from the central volcano and erupted during its early phases.Later activity at the central volcano was typified by effusive and explosive rhyolitic eruptions and less frequent and smaller basaltic andesite to dacite events, ceasing with a Plinian eruption.The silicic volcanism was accompanied by silicic intrusions that are exposed at shallow levels (up to ~ 300 m below the base of the silicic/intermediate lavas), with evidence of magma ascending through transgressive sills and erupting through sub-vertical dykes as well as subcylindrical conduits.Basaltic lavas were formed concurrently with the rhyolites and continued to flow alongside the volcanic edifice after the end of silicic activity, eventually burying the highest rhyolitic peaks.A swarm of basaltic dykes is recorded north of the central volcano (north of the map in Fig. 2), but no clear evidence has been observed of whether this dyke swarm belongs to the volcanic system of the ToCV or not.While dykes belonging to this swarm generally trend NNE-SSW, similar to the regional trend of basaltic dykes in the Tröllaskagi peninsula (Saemundsson et al. 1980), silicic dykes are concentrated in the northern part of Mt.Torfufell, most often striking from ENE-WSW to ESE-WNW.Our strike and dip measurements indicate that the entire strata succession of the study area is dipping gently (from ~ 5 to 11°) towards southeast, while steeper (up to ~ 16°) and more southerly dips (SSE to SSW) are locally indicated in the central zone of the volcanic complex, i.e., in the northern and western slopes of Mt.Torfufell.
Six silicic rock samples from the ToCV were selected, on grounds of the geological mapping and in terms of stratigraphic relations, for U-Pb zircon dating by the SIMS and CA-TIMS methods.Four of the selected samples are from volcanic formations, located at different stratigraphic levels, including samples marking both the base and the top of silicic volcanic products as observed in outcrops.Three of these volcanic formations were formed in effusive eruptions (i.e., lavas) while one is the result of an explosive event (i.e., ignimbrite).The remaining two samples are from intrusive units; one of these is from a sill while the other is from the interiors of a volcanic conduit (plug).Sampling sites are presented in Fig. 2 in comparison with previously published age data from the study area.Descriptions of the six host samples and their stratigraphic levels are presented in Table 1.Our geochemical data, available from 5 of the host samples, indicate that they span a compositional range from ~ 65 to 75% SiO 2 .In order to provide background information on the geochemistry of the ToCV, we introduce geochemical analysis results of 27 additional rock samples from the study area (Tables S1 and S2 of the Supplementary Material; detailed separately by Árnadóttir et al., In Prep.).These samples, mostly collected from the central volcano itself while a few are obtained from enveloping lavas, demonstrate that the rhyolites of the ToCV are mildly alkalic, i.e., they have a transitional character towards alkaline rocks, and tend to have high silica contents.The geochemical data show strong similarities with the neighboring TiCV, and a comparison with data from the neovolcanic zones of Iceland indicates that basalt to rhyolite products belonging to these two volcanoes are in general more enriched than typical rift zone magmas (Fig. 3).
Zircons were identified in thin sections of four of the six rock samples; in three samples (07072015-2, 27072017-1, 26082015-25) as inclusions in plagioclase phenocrysts, and in the groundmass of one sample (10072017-3).Nevertheless, it turned out via mineral separation processes that all six host samples featured zircon crystals in different abundances.The highest zircon yield was obtained from samples 10072017-3 and 24072014-4 (~ 120-150 grains per kg rock) while samples 07072015-2, 20190823-3, and 26082015-25 yielded lower abundances (~ 40-70 grains per kg rock).The lowest yield was obtained from the dacite sample, 27072017-1 (~ 16 grains per kg rock).The zircons are commonly ~ 50-200 µm along the c-axis, elongate and prismatic, typically subhedral and less commonly euhedral, and often with pyramidal terminations (see examples in Fig. 4a).Common length-to-width ratios are from 2:1 to 3:1, while some crystals are very long (up to more than 300 µm in length) and thin, with ratios up to 6:1.Catholuminescence (CL) images, available from the CA-TIMS dated zircons (20190823-3), do typically not show very distinct brightness variations or strong CL intensities within the crystals as would be expected for well-developed zoning textures indicating distinct growth regions (Fig. S4 of the Supplementary Material).These images do, however, imply zoning in a few crystals and even rounded inherited cores; these crystals were avoided during dating (see examples of dated vs. avoided crystals in Fig. 4a).Only backscattered electron (BSE) images are available from the SIMS dated zircons (Fig. S4 of the Supplementary Material), which are not very helpful for illustrating internal zircon textures.Although test line scans (from energy-dispersive X-ray spectroscopy (EDS) analysis using a scanning electron  4b), it cannot be excluded that some of the SIMS dated zircons have inherited domains.In order to reduce the influence of such domains, the zircon cores were avoided as much as possible during dating.Inclusions of other minerals are rather commonly observed in zircons from all samples (see example in Fig. 4b); these were also evaded during dating.

Secondary ion mass spectrometry
U-Pb dating of zircons from host samples 20172017-1, 26082015-25, 07072015-2, 24072017-4, and 10072017-3 was carried out via in situ measurements using the CAMECA ims 1280 in the NordSIMS facility, Swedish Museum of Natural History, Stockholm.Zircons were separated from these five host samples using standard crushing, sieving, and density and magnetic methods by Geotrack in Australia.The zircon grains were mounted by hand onto a two-sided tape along with reference materials and then epoxy-casted and polished to expose grain interiors.The polished mount was imaged using backscattered electron signals in SEM (Hitachi TM3000 SEM at the University of Iceland) in order to carefully map the sample mount prior to dating, to confirm zircon identities of crystals, and to identify internal structures such as inclusions and cracks that were avoided during the dating.The zircon cores were also avoided during dating in order to exclude potential inheritance as possible.
The analytical and calibration procedures generally followed those described in Whitehouse and Kamber (2005) and Jeon and Whitehouse (2015), with a few differences noted in Supplementary Text S1.In total, 97 U-Pb analyses were carried out on 92 zircon grains from the five samples (all analyzed spots are presented in Fig. S4 of the Supplementary Material) and thereof, 7 analyses were excluded, resulting in a total of 90 analyzed spots.Eleven grains were selected from sample 27072017-1 (12 spots; thereof 2 spots were discordant and excluded (assumed as contamination due to abnormally high age; 529 and > 1200 Ma)); 19 grains from sample 26082015-25 (21 spots; thereof 2 spots were discordant and excluded (assumed as contamination due to abnormally high age; 73 and > 1200 Ma)); 22 grains from 07072015-2 (22 spots; no spot excluded); 21 grains from 24072017-4 (21 spots; no spot excluded); and 19 grains from 10072017-3 (21 spots; thereof 3 spots were excluded due to outstandingly high common Pb).

Chemical abrasion-isotope dilution thermal ionization mass spectrometry
Zircons from host sample 20190823-3 were dated at the Pacific Centre for Isotopic and Geochemical Research (PCIGR) at the University of British Columbia (UBC) by CA-TIMS method.The sample was prepared at the PCIGR, including separation of zircon crystals from the host sample and CL SEM imaging.Zircon crystals were subjected to a pre-treatment combining annealing and a modified version of Mattinson's (2005) chemical abrasion method for removing zircon domains that have lost lead from the interior of grains (see Supplementary Text S1).By this, the CA-TIMS method can isolate and analyze closed-system, high-quality domains within the zircon grains and exclude discordance due to loss of lead.The mineral-dissolution method furthermore samples larger volumes of material than the SIMS method, resulting in higher attainable precision.Prior to the chemical abrasion, 33 zircon grains were selected for in situ determination of trace element concentrations by LA-ICPMS (the LA-ICPMS conditions are given in Supplementary Table S7; the method is described in Supplementary Text S1).Six zircon grains were then selected for CA-TIMS analysis based on the LA-ICPMS results in combination with the prismatic morphology and the absence of inclusions, avoiding those crystals with rounded edges.U-Pb geochronology methods for isotope dilution thermal ionization mass spectrometry follow those previously published by Wall et al. (2018), detailed in Supplementary Text S1.Chemical abrasion in concentrated HF at 190° for 12 h resulted in only minor dissolution of the zircon crystals.

Results
Our U-Pb zircon ages (weighted averages) range from 7.15 ± 0.12 to 6.76 ± 0.02 Ma (error is in 2σ); that is Late Miocene.Weighted mean ages for each sample are Fig. 3 Geochemistry of the ToCV in comparison with samples from the neovolcanic zones in Iceland, the TiCV and the Neogene East and West fjords.The samples dated in this study are signified with bright red diamonds.Comparison data is from Clay et al. (2015), Debaille et al. (2009), Fitton et al. (1997), Flude et al. (2010), Fowler and Zierenberg (2016), Guðmundsdóttir et al. (2011), Gunnarsson et al. (1998), Harðarson (1993), Hardarson and Fitton (1997), Hardarson et al. (1997), Hjartarson (2003), Kuritani et al. (2011), Lacasse et al. (2007), Larsen et al. (1999;2001;2002), Maclennan et al. (2001), Martin and Sigmarsson (2007), Mattsson and Óskarsson (2005), McGarvie et al. (1990;2007), Meara et al. (2020), Nicholson et al. (1991), Óladóttir et al. (2011Óladóttir et al. ( ), Pollock et al. (2014)), Selbekk andTrönnes 2007, Sigvaldason (1979), Sinton et al. (2005), Slater (1996), Slater et al. (2001), Viccaro et al. (2015), and Zierenberg et al. (2013).The total alkali-silica diagram (TAS; uppermost left) shows root names from Le Bas et al. (1986).The TiO 2 vs Zr/(P 2 O 5 × 10,000) plot shows the discrimination between alkali basalts and tholeiite basalts after Winchester and Floyd (1976) ◂ 52 Page 10 of 17 presented in Table 2 and Fig S5-S6 of the Supplementary Material.Very similar ages were obtained for the two intrusive units, showing an age of 7.00 ± 0.06 Ma for the sill (sample 07072015-2) and 7.02 ± 0.05 Ma for the volcanic plug (sample 24072017-4).Slightly higher ages were obtained for the early silicic extrusive units: 7.03 ± 0.15 Ma for the dacite lava (sample 27072017-1) while the highest obtained U-Pb mean age is the ignimbrite age, 7.15 ± 0.12 Ma (Sample 26082015-25), although located slightly higher in the stratigraphy than the dacite.The ages of the upper rhyolite lavas are 6.85 ± 0.09 Ma for the Glerá lava (sample 10072017-3) and 6.76 ± 0.02 Ma for the uppermost rhyolite in Mt.Torfufell (sample 20190823-3).Very low 2σ and mean square weighted deviation (MSWD) were obtained for the CA-TIMS dated sample compared to the SIMS dated samples.Four SIMS analyses were rejected to get statistically meaningful MSWD (a value under 2) and probability (higher than 0.05 for 95% confidence level), 2 rejections in 07072015-2, 1 in 27072017-1 and 1 in 26082015-25 (blue bars in Fig. 5a and Fig. S2 of the Supplementary Material).However, while MSWD is < 2 for most SIMS samples, high MSWDs (> 4) were obtained for two samples (26082015-25 and 10072017-3).The weighted averages of these two samples also have zero probability, meaning that their data are too scattered to get a statistically grouped age-and the probability density plots for these two samples show more than one peak (while the plots for the other samples show a single peak), indicating that the data from each sample consists of more than one population (Figures S2 and S3 of the Supplementary Material).Since the zircons are young and fairly low in U, it is not reasonable to assign this to Pb loss, as might be the case for older and/or U-rich zircons.Instead, we rather assume that the dispersion in these two samples is related to recycling/mixing of older zircons into younger melts.To improve the situation, the data from these two samples were processed using the TuffZirc algorithm of Isoplot, which is largely insensitive to both Pb loss and inheritance and looks at the data statistically for a coherent group (Ludwig and Mundil 2002;Ludwig 2003).The ages calculated by TuffZirc are 7.03 + 0.18/ − 0.07 Ma for sample 26082015-25 (95.1% conf, from a coherent group of 12) and 6.93 + 0.08/ − 0.1 Ma for sample 10072017-3 (93.5% conf, from a coherent group of 11; Fig. 5a and Fig. S3 of the Supplementary Material).

Discussion
The silicic samples selected for this study are from volcanic formations, located at different stratigraphic levels, and intrusive units of the ToCV.By this, we were able to provide geochronological data that can be interpreted to represent the age range of silicic magmatism and volcanism at the volcano.We base the timing of the initiation of the ToCV on Supakulopas's (2018) 40 Ar/ 39 Ar 7.77 ± 0.15 and 7.98 ± 0.15 Ma ages, obtained from samples collected above and below the stratigraphic level of the base of the pile of thin basaltic ā'a lavas that we assume to define the onset of eruptions at the central volcano (Fig. 5a).However, as these two ages are inconsistent to the stratigraphic order (i.e., the higher age was obtained for the lava that is located higher in the strata pile while the lower lava gave younger age), the estimated timing of the establishment of the edifice base is quite imprecise (therefore given here as ~ 7.8-8.0Ma).While no age determinations are available for intermediate formations, correlation with Supakulopas (2018) and the geomagnetic time scale provides a rough estimate, indicating that intermediate volcanism began during geomagnetic subchron C3Br.2r (i.e., somewhere on the range from ~ 7.3 to ~ 7.5 Ma; Ogg 2020).
Our oldest weighted average U-Pb age (7.15 ± 0.12 Ma) was obtained for the ignimbrite sample (26082015-25), which appears to be located higher in the strata pile than the younger dacite sample (27072017-1; 7.03 ± 0.15 Ma).As the MSWD for the ignimbrite sample is high (4.3), and as the probability density plot for this sample indicates that the data consists of more than one population (Fig. S3 of the Supplementary Material), it may be speculated that there is an older (~ 7.4 Ma) component in a ~ 7 Ma melt that could explain the higher weighted average U-Pb age, suggesting that the ignimbrite and the dacite were emplaced in close temporal proximity.The 7.03 + 0.18/ − 0.07 Ma age of the ignimbrite, extracted from the most-coherent population by the TuffZirc algorithm of Isoplot (Ludwig and Mundil 2002;Ludwig 2003), supports this idea (Fig. 5a and Fig. S3 of the Supplementary Material).The older component in the ignimbrite may e.g., be derived from zircon inclusions in plagioclase phenocrysts, and/or from silicic lithics in the wholerock sample, which would imply silicic magmatism as early as 7.4 Ma at the ToCV.However, as Supakulopas's (2018) 7.12 ± 0.10 Ma 40 Ar/ 39 Ar age is from the lowermost rhyolite lava in Mt.Hólafjall, it can be considered to approximately mark the onset of silicic volcanism at the ToCV.
The two youngest weighted average ages, 6.85 ± 0.09 and 6.76 ± 0.02, are from lavas representing late silicic volcanism at the ToCV (samples 10072017-3 and 20190823-3, respectively).The weighted average of sample 10072017-3 has zero probability; MSWD for this sample is high (5.1) and the relative probability plot indicates that the data consists of more than one population (~ 7.0 and ~ 6.8 Ma;Figures S2 and S3 of the Supplementary Material).The mostcoherent group of zircon ages, determined using the TuffZirc algorithm of Isoplot (Ludwig and Mundil 2002;Ludwig 2003), provided an age of 6.93 + 0.08/ − 0.1 Ma (Fig. 5a and Fig. S3 of the Supplementary Material).While this sample is not confidently assigned to the youngest rhyolite lava at the volcano, sample 20190823-3 is collected at the highest rhyolite peak and has MSWD very close to 1 (0.98), showing decidedly that silicic activity continued at the ToCV until 6.76 ± 0.02 Ma.Field relations indicate that the rhyolite activity was succeeded by regional basaltic volcanism.Supakulopas (2018) provided a 5.73 ± 0.14 Ma age for a basaltic lava onlapping a rhyolite formation in Mt.Hólafjall which, according to his interpretation, was formed during geomagnetic subchron C3Bn (~ 7.21-7.10Ma), thereby identifying a ~ 1.5 Myr hiatus between these sequences (Fig. 5a).This hiatus is somewhat shortened by at least four lavas that are present between these two units, and due to the palaeo-landscape, the length of the hiatus may vary within the area, probably increasing in length at the highest rhyolite peak.Nevertheless, as we date the youngest rhyolite at 6.76 ± 0.02 Ma, it may be assumed that ~ 1 Myr passed from the extinction of the ToCV until basaltic lavas from nearby fissure swarms began to onlap the volcano.
The available age data indicate that the ToCV was active from ~ 7.8-8.0 to 6.76 ± 0.02 Ma, or for ~ 1-1.2 Myr, and that silicic magmatism and volcanism occurred during the last ~ 0.4 Myr of the volcano's life-span.Consistent with Supakulopas' (2018) conclusions, our age data show that the silicic rocks of the ToCV are 1-2 Myr older than suggested by earlier researchers (e.g., Hjartarson 2003;2005).The longest hiatus in the research area appears to be between the volcano's uppermost rhyolites and the younger, succeeding basalts.Although relatively shorter than the time gap recorded in the Flateyjarskagi peninsula, it may be speculated if the hiatus observed above the ToCV is potentially an extension of this major unconformity and thereby related to a plate boundary shift.If so, and according to the general Page 13 of 17 52 understanding of the Flateyjarskagi unconformity, the ToCV would be formed in the now extinct SHRZ, while the basalt lavas above the hiatus would be formed in the NVZ.The Flateyjarskagi hiatus appears to span the time from ~ 10 to ~ 6 Ma in the north (i.e., by the peninsula's SE end) and from ~ 7 to ~ 5 Ma in the south (i.e., in the southern part of the Fnjóskadalur valley; Jancin et al. 1985;Cotman 1979;Fig. 1b, c), thus spanning the Miocene-Pliocene boundary (~ 5.3 Ma), while the lavas onlapping the ToCV record ongoing volcanism across this boundary (Figs. 2 and 5a), at least until 5.05 Ma (via stratigraphic correlation with Supakulopas's (2018) 40 Ar/ 39 Ar ages).On the other hand, the hiatus above the ToCV may simply represent a local break at the end of the volcano's active period.Assuming that the geodynamic setting since the volcano's lifetime until present was dominated by spreading, the distance between the ToCV and the present NVZ (~ 70 km) is consistent with origin within the NVZ and a mean spreading half-rate of ~ 1 cm/ year (e.g., Árnadóttir et al. 2008).Furthermore, considering the general assumptions that regional lava dips in Iceland result from loading by voluminous lava flows in the respective volcanic zone (e.g., Pálmason 1973;Pálmason and Saemundsson 1974;Walker 1960), the regional southeast dip direction of the strata in the ToCV area might be taken to indicate origin within a zone that was roughly coincident with the present NVZ.This would agree with the hiatus representing only a local break and would indicate, as no major hiatuses are present, that a time gap related to a rift relocation is non-existent in the ToCV area.If so, the new U-Pb zircon ages from the ToCV may be considered to provide timing of silicic magma erupted soon after its parent rift relocated, and thereby constrains on the oldest silicic rocks formed at the incipient rift.Similar findings of appreciable silicic material erupting soon after relocation of the parent rift (at ~ 5.4-3.9Ma) are documented at the Hafnarfjall-Skarðsheiði central volcano in West Iceland which began forming nearly contemporaneously with the relocation of the southern segment of the SHRZ to the WVZ (Banik et al. 2018).However, the age progression (younging to the west) of the E-W aligned ToCV and the 5-6 Ma TiCV remains enigmatic as it conflicts with concepts of unidirectional spreading.Considering the age of the main formation of the TiCV, ~ 5.5 Ma (Hjartarson 2003;2005), it might be suggested that the TiCV originated inside the NVZ.The problem is that the volcano is 90 km west of the NVZ which would indicate a half spreading rate of 1.7 cm/year (Hjartarson 2009).Therefore, the TiCV appears to have formed outside the NVZ and to the west of the then extinct ToCV.As such, the results of our study do not align with the generally established geotectonic framework (e.g., Saemundsson 1979;Einarsson 2008).Therefore, to provide an explanation for this discrepancy, we propose a new hypothesis (Fig. 5b).
At ~ 8 Ma, the center of the mantle plume was located just east of the ToCV (an assumption based on Steinthorsson et al. (1985) and Martin et al. (2011)), after having reached a critical distance from the northern segment of the SHRZ, already weakening the lithosphere, resulting in partitioned spreading between the SHRZ and the incipient NVZ.To accommodate this rearrangement, the southern end of the NVZ was linked with the SHRZ by a transverse zone which was the precursor of the MIB.Similar to the present MIB, we propose that this proto-MIB was leaky, i.e., comprising a small component of opening, producing sufficient room for magma to erupt.Volcanic activity at this transverse zone was initially most effective at the east end, at the junction of the proto-MIB with the incipient NVZ, approximately by the center of the mantle plume, then gradually migrated westward-away from the center-resulting in the observed age progression of the Torfufell and Tinná central volcanoes.Lavas from the TiCV may have flowed from the west and reached the area around the ToCV and partly or totally covered the edifice along with lavas from the NVZ flowing from the east.This interpretation requires a ~ 50 km southward migration of the transverse zone with time, from the proto-MIB to the present location of the MIB.While such southward advancement could probably be explained partly by the WNW migration of the tectonic plates relative to the mantle plume, it may also be a result of southward propagation of the new rift zone, similar to the indicated southward migration of the SISZ with propagation of the EVZ (e.g., Einarsson 1991;Khodayar and Franzson 2007;Karson 2017) and northward migration of the TFZ with propagation of the NVZ (e.g., Karson 2017;Karson et al. 2019).
Our hypothesis does have some support from the wholerock geochemical data that indicate that the rock suite of the ToCV is transitional in nature (i.e., has a transitional character towards alkaline rocks) with basalts that are slightly higher in TiO 2 and K 2 O compared to typical rift zone basalts and rhyolites that have higher K 2 O and lower FeO than typical rift zone rhyolites (Fig. 3; Tables S1 and  S2 of the Supplementary Material).A comparison with the neighboring TiCV shows strong similarities between these two central volcanoes and indicates that basalt to rhyolite products of both volcanoes are, in general, more enriched than typical rift zone magmas (Fig. 3).More Ar/ 39 Ar ages and coupling of the schematic, composite stratigraphic column with the geomagnetic polarity time scale.The schematic stratigraphic column (same color legend as in Fig. 2) is based on original fieldwork, integrated with data by Kristjánsson et al. (2004) and Supakulopas (2018) from Mt. Hólafjall.The weighted average plots were made with Isoplot (Ludwig 2012).TuffZirc age plots for the two high-MSWD samples are presented in the upper right corner (generated using the TuffZirc algorithm of Isoplot by Ludwig and Mundil 2002; detailed in Fig. S3 of the Supplementary Material).b Schematic illustration of our new hypothesis; proposed to explain a discrepancy (younging towards west) in the ages of TiCV and ToCV ◂ 52 Page 14 of 17 importantly, these two volcanoes feature rhyolites similar to rhyolites associated either with the MIB or those in the propagating EVZ, or even the intraplate ÖVB, while their basalts are similar to those from the junction of the WVZ with the SISZ.Nonetheless, the new hypothesis awaits critical testing that optimally will involve extensive structural mapping along the hypothesized proto-MIB, to seek evidence of potential transform-related structures that may be compared with future collection of isotopic data from zircon crystals and whole rocks of the ToCV and TiCV and may be indicators of petrogenetic environments for silicic magmas at these volcanoes (see Carley et al. 2020).To obtain more robust geometric relationships in the central north region of Iceland, we consider dating of the remaining undated silicic centers in the region to be essential.Dating of additional samples from the TiCV and the two adjacent volcanoes in the Flateyjarskagi peninsula that can confidently confirm their ages would furthermore strengthen the dataset, and a closer look at the Víðidalur central volcano would be of particular interest as recent U-Pb zircon ages (6.3-7.3 Ma;Carley et al. 2020) document silicic magmatism at the SHRZ concurrently with silicic activity at the ToCV.

Conclusions
The U-Pb zircon ages (weighted averages) presented here range from 7.15 ± 0.12 to 6.76 ± 0.02 Ma, indicating that silicic magmatism and volcanism of the Torfufell central volcano extended over ~ 400 kyr.Correlations with previous work predict that the life-span of the central volcano is about or just over 1 Myr, commencing with basaltic volcanism at ~ 7.8-8.0Ma.A ~ 1 Myr volcanic hiatus is recorded between the uppermost rhyolites of the central volcano and the succeeding basaltic lavas, from 6.76 ± 0.02 to 5.73 ± 0.14 Ma, and broadly contemporaneous with (but shorter than) the Flateyjarskagi unconformity, we propose that this hiatus represents a local break at the end of the volcano's active period.The new U-Pb age data demonstrate that the silicic portion of the Torfufell central volcano was formed nearly synchronously with a major rift relocation in North Iceland and may provide constraints on the oldest silicic rocks formed at the incipient Northern volcanic zone.After its extinction, the Torfufell central volcano was buried beneath younger formations making up the lava succession above the hiatus.We propose that these younger lavas originated from the large Tinná central volcano in the west as well as from the newly formed NVZ in the east.Our results pertain to a key area in the evolving understanding of the development of crustal accretion in Iceland and underline the extensive room for additional high-quality radiometric dating on silicic activity in the central north region.

Fig. 1 a
Fig. 1 a Generalized geological map of Iceland (modified from Hjartarson & Saemundsson 2014; Thordarson and Höskuldsson 2008), highlighting active and fossil central volcanoes and active volcanic zones.See text for abbreviations.The approximate center of the Iceland mantle plume, as depicted by Wolfe et al. (1997), is indicated

Fig. 2
Fig. 2 Sampling sites, shown on a geological map, in comparison with 40 Ar/ 39 Ar ages in Mt.Hólafjall from Supakulopas (2018).Location of the study area is indicated with a blue box on the small image of Iceland.The geological map is based on original fieldwork, integrated with data from previous authors (stratigraphic and . 5a, while detailed results are listed in TablesS3-S4 and Figures S1-S4 of the Supplementary Material.Results of LA-ICPMS U-(Th-)Pb and trace element analysis of zircons from sample 20190823-3 are given in Tables

Fig. 4 a
Fig. 4 a CL images of zircons from sample 20190823-3.Numbered zircon grains were dated with the CA-TIMS method (with red circles indicating location of laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS) spots) while zircons without numbers (and without red circles) are examples of avoided grains.b Example of a random line scan test (from EDS analysis using a SEM) through

Table 1
Descriptions and stratigraphic position of the dated samples

Table 2
Summary of U-Pb radiometric data.Ages are given as weighted averages.Detailed results are listed in TablesS3 and S4in the Supplementary Material