The Evolution of the Silver Hills Volcanic Center, and Revised 40Ar/39Ar Geochronology of Montserrat, Lesser Antilles, With Implications for Island Arc Volcanism

Studying the older volcanic centers on Montserrat, Centre Hills and Silver Hills, may reveal how volcanic activity can change over long time periods (≥1 Myr), and whether the recent activity at the Soufrière Hills is typical of volcanism throughout Montserrat's history. Here we present the first detailed mapping of the Silver Hills, the oldest and arguably least studied volcanic center on Montserrat. Volcanism at the Silver Hills was dominated by episodic andesite lava dome growth and collapse, produced Vulcanian style eruptions, and experienced periodic sector collapse events, similar to the style of volcanic activity that has been documented for the Centre Hills and Soufrière Hills. We also present an updated geochronology of volcanism on Montserrat, by revising existing ages and obtaining new 40Ar/39Ar dates and paleomagnetic ages from marine tephra layers. We show that the centers of the Silver, Centre, and Soufrière Hills were active during at least ∼2.17–1.03 Ma, ∼1.14–0.38 Ma, and ∼0.45 Ma to present, respectively. Combined with timings of volcanism on Basse‐Terre, Guadeloupe, these ages suggest that ∼0.5–1 Ma is a common lifespan for volcanic centers in the Lesser Antilles. These new dates identify a previously unrecognized overlap in activity between the different volcanic centers, which appears to be a common phenomenon in island arcs. We also identify an older stage of Soufrière Hills activity ∼450–290 ka characterized by the eruption of hornblende‐orthopyroxene‐phyric lavas, demonstrating that the petrology of the Soufrière Hills eruptive products has changed at least twice throughout the volcano's development.


Introduction
Lava dome growth and collapse is a common process at arc volcanoes, e.g., as observed at Merapi volcano, Indonesia (Andreastuti et al., 2000), Mt St Helens, USA (Hoblitt et al., 1980), and Unzen volcano, Japan (Hoshizumi et al., 1999), but volcanic activity can be highly variable over the lifetime of an individual volcanic edifice (e.g., Hildreth & Lanphere, 1994;Komorowski et al., 2005;Myers et al., 1985;Pioli et al., 2015). This is exemplified in the Lesser Antilles arc, where volcanism is dominated by dome-forming eruptions but varies in composition and style both between discrete volcanic edifices on individual islands, and during the lifetime of an individual volcano. For example, eruptive activity at volcanic centers on Guadeloupe and Martinique has varied from effusive basaltic lava flows forming shield volcanoes, to andesitic dome eruptions to Plinian eruptions (Germa et al., 2011a(Germa et al., , 2011bKomorowski et al., 2005;Maury et al., 1990;Samper et al., 2007). Understanding what causes these changes in eruption styles is crucial to aiding predictions concerning the long-term behavior of a volcano.
Volcanism on Montserrat has been of particular interest since the onset of renewed volcanism at the Soufrière Hills volcano in 1995. As a consequence, it has become one of the most studied volcanoes on Earth. Since 1995 the eruption has had substantial social and economic impacts on the island, including destroying its capital, Plymouth (Kokelaar, 2002), and so gaining insight into the evolution of volcanic activity at Soufriè re Hills through time is of considerable interest. Both marine and terrestrial records show that the style of past Soufrière Hills activity has changed little throughout its known ca. 282 kyr history (e.g., Le Friant et al., 2008;Smith et al., 2007), but it may not necessarily continue to behave in this way. For example, Caricchi et al. (2014) suggest that larger eruptions tend to occur toward the end of a volcano's lifetime, Key Points: Silver Hills volcanism was dominated by andesite dome growth and collapse, Vulcanian style eruptions and periodic sector collapse New ages reveal overlap in volcanic activity between the Silver and Centre Hills, and Centre and Soufrière Hills An older, previously unreported stage of Soufriè re Hills activity at 450-290 ka erupted hornblendeorthopyroxene lavas Supporting Information: Supporting Information S1 Table S1  Table S2  Table S3  Table S4 Correspondence to: S. J. Hatter, sjh1e13@soton.ac.uk Currently, the Soufriè re Hills comprises a central dome complex surrounded by volcaniclastic deposits, which are andesitic in composition. The dome complex comprises four old domes: Chances Peak (<200 ka), Gages Mountain (151 6 8 ka), Galway's Mountain (112 6 18 ka), and Perches Mountain (24 6 4 ka) (Harford et al., 2002), and a dome within English's Crater created by the recent eruption (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010); Figure 1b). The modern dome covers a fifth older dome, Castle Peak (formed circa 1650 A.D.; Young et al., 1996) and has undergone multiple successive growth and collapse phases during the recent eruption. Indeed, the current eruptive phase has been dominated by a combination of Vulcanian explosions, dome growth, and dome collapse events (Kokelaar, 2002;Wadge et al., 2014), which is probably characteristic of Soufriè re Hills activity over the past 280 kyr (Rea, 1974;Roobol & Smith, 1998;Smith et al., 2007).
Large landslide deposits discovered offshore of Montserrat reveal that Soufriè re Hills has undergone multiple flank collapse events (Coussens et al., 2016;Deplus et al., 2001;Le Friant et al., 2004;Lebas et al., 2011). Additionally, a debris avalanche occurred during the recent eruption on 26 December 1997, resulting from dome growth over a hydrothermally weakened sector of the wall of English's crater Voight et al., 2002).

Field Work
Field mapping was undertaken in the Silver Hills, Montserrat in May 2014 and 2015 at a scale of 1:25,000, with samples collected for 40 Ar/ 39 Ar dating and geochemical analysis. The Silver Hills is highly eroded with a current subaerial area of 6.5 km 2 . The patchy nature of exposures coupled with the generally limited spatial extent of individual units means that radiometric dates are vital to constrain a stratigraphy for the Silver Hills deposits.

40 Ar/ 39 Ar Dating
Twelve lava and three pumice samples were prepared at the Scottish Universities Environmental Research Centre (SUERC) Argon Isotope Facility (AIF), where they were crushed in a jaw-crusher, sieved in to a 250-500 lm size fraction, leached in 20% HNO 3 , and passed through a Frantz magnetic barrier laboratory separator (model LB-1) to separate plagioclase phenocrysts from the groundmass. Plagioclase separates were then leached in 5% HF to remove any glass attached to the crystals. Two hundred milligrams and 1 g of Geochemistry, Geophysics, Geosystems 10.1002/2017GC007053 clean plagioclase phenocryst separates and groundmass (lavas only), respectively, were then hand picked for 40 Ar/ 39 Ar dating.
The samples were irradiated for 1 h in the Oregon State University reactor, Cd-shielded facility, and analyzed on a GVi instruments ARGUS V multicollector mass spectrometer using a variable sensitivity faraday collector array in static collection (nonpeak hopping) mode (Mark et al., 2009;Sparks et al., 2008) at SUERC AIF. Alder Creek sanidine (1.2056 6 0.0019 Ma, 1r; Renne et al., 2011) was used to monitor 39 Ar production and establish neutron flux values (J) for the samples. The reader is directed to Coussens et al. (2017) for more details of the method. The average total system blank for laser extractions, measured between each sample run, was 1.5 6 0.7 3 10 215 mol 40 Ar, 1.2 6 1.3 3 10 217 mol 39 Ar, and 8.5 6 3.6 3 10 218 mol 36 Ar for lava samples, and 1.7 6 0.1 3 10 215 mol 40 Ar, 1.6 6 0.5 3 1.1 217 mol 39 Ar, and 1.6 6 0.7 3 10 217 mol 36 Ar for pumice samples. All blank, interference and mass discrimination calculations were performed with the MassSpec software package (versions 8.058 and 8.16 for andesite and pumice samples respectively, authored by Al Deino, Berkeley Geochronology Center).
Plateau ages, or composite plateau ages for replicated samples, are chosen as the best estimates of the emplacement ages. Accepted plateau ages must be derived from a minimum of three contiguous steps (minimum 39 Ar content of each step is 0.1% of total 39 Ar release) which overlap in age within 2r uncertainty, and contains a minimum of 50% of 39 Ar released for the combined steps. The scatter between ages of the steps must be low with a MSWD (mean square weighted deviation) less than the 95% probability cutoff (Wendt & Carl, 1991). Further, the inverse isochron formed by the plateau steps must yield an age indistinguishable from the plateau age at 2r uncertainty, and the 40 Ar/ 36 Ar of the trapped component composition derived from the inverse isochron must be indistinguishable from the composition of air (298.56 6 0.61, 2r) at the 2r uncertainty level.
For fall deposits or ignimbrites, the single crystal approach is used to determine an eruptive age because this allows for discrimination of juvenile crystals and crystals derived from other sources, e.g., xenocrysts or antecrysts, but pumice samples yielded low-K plagioclase that could not be analyzed as single crystals. Instead, a multigrain, two-step heating approach was adopted to increase signal sizes allowing for more precise measurements. The presence of older xenocrysts in these multigrain aliquots could bias age results, an effect dependent on the age of the xenocryst and its relative potassium content. To account for this possibility, the experiment was repeated (n 15) and the youngest, Gaussian distributed population of ages was calculated from the set of analyses.
Each individual gas analysis represents an aliquot of 50-100 single grains, loaded into a 4 mm well in a copper planchette. An initial step at 0.5 W of power was used to drive off atmospheric gas and liberate argon from any alteration phases that might be present. A second fusion step at 15 W of power yielded age information associated with the plagioclase crystals.

Whole Rock Geochemistry
Nb and Y were analyzed on a Thermo Fisher Scientific X-Series 2 ICP-MS at the University of Southampton following the method of Coussens et al. (2017), in which 0.05 g of powdered sample is analyzed. Precision is better than 3%.
Pb and Nd isotopes were measured from 50 and 200 mg (for pumice and lava samples respectively) of hand-picked rock chips 0.5-1 mm in size. Samples were leached and digested following the method of Cassidy et al. (2012), and Pb was isolated from the matrix using AGX-1x8 200-400 mesh anion exchange resin, following the method of Kamber and Gladu (2009 Note. Ages from Harford et al. (2002) and Brown and Davidson (2008) presented here are recalculated ages (see text for details). BAF, block-and-ash flow deposit; PAF, pumice-and-ash flow deposit; S-BAF, submarine-BAF; PF, pumice fall deposit. Ages considered reliable are in bold. Ages viewed with caution: m maximum age; r Ar recoil; x excess Ar; e MSWD exceeds critical value; i isochron age. N, number of plateau steps used to determine age/total number of steps; % 39 Ar, percentage of total 39 Ar in N. a Harford et al. (2002). b Brown and Davidson (2008). c Coussens et al. (2017). d Unit name from Smith et al. (2007).
Geochemistry, Geophysics, Geosystems 10.1002/2017GC007053 long and provides a record extending back 4.5 Ma (Expedition 340 Scientists, 2013). This study investigates the tephra layers deposited in the past 2.35 Ma (the interval represented by the terrestrial record on Montserrat), which have been sampled for geochemical analysis. These samples have been previously analyzed for Pb isotopes by Palmer et al. (2016).  Table 1. All of these studies used the same strict criteria outlined above (section 3.2) to determine plateau ages. The ages of Harford et al. (2002) and Brown and Davidson (2008) have been recalculated with modern decay constants and standard ages using the Renne et al. ( , 2011 optimization model, resulting in slightly increased ages. All ages referenced from these authors are the recalculated ages (see supporting information Table S1 for original and recalculated ages). Here we review the dates from these previous studies to identify which ages are reliable, and which should be viewed with caution. Harford et al. (2002) obtained ages for all of the Montserrat volcanic centers ( Figure 1b and Table 1). Some of their ages, however, should be viewed with caution, as either their plateau profiles contain evidence of xenocrystic components or Ar recoil, they have a saddle-shaped plateau typical of lavas with excess argon or have a MSWD which exceeds the critical 95% confidence level. Of the dates considered reliable, two are EVOLUTION OF SILVER HILLS VOLCANISM from the Centre Hills, six are from the Soufrière Hills, and three are from the South Soufrière Hills, with none from the Silver Hills (Table 1).
Five 40 Ar/ 39 Ar ages were acquired for the Silver Hills by Brown and Davidson (2008), in the range of 1.52-1.40 Ma but only cover a small area of the Silver Hills center, four of which are part of the same stratigraphic sequence ( Figure 2 and Table 1). These ages were measured from groundmass (i.e., no xenocrystic material). Coussens et al. (2017) obtained five 40 Ar/ 39 Ar ages from Centre Hills deposits (0.79-0.48 Ma; Figure 1b and Table 1) and one from a fault block within the Centre Hills (1.31 Ma), which they interpret as originating from the Silver Hills volcanic center. These ages were obtained from plagioclase phenocrysts, so the presence of xenocrysts cannot be ruled out.
The questionable quality of the Silver Hills ages reported by Harford et al. (2002), and the limited spatial and stratigraphic extent of the Silver Hills ages reported in Brown and Davidson (2008) mean that the timing of volcanic activity is poorly known for the Silver Hills. This highlights the need for new 40 Ar/ 39 Ar dates, to better constrain the timing of Silver Hills activity, and aid the development of a detailed stratigraphy of the Silver Hills deposits.

New 40 Ar/ 39 Ar Dates
Here we present fifteen new 40 Ar/ 39 Ar ages for Montserrat (Figures 3 and 4 and Table 2; supporting information Tables S2 and S3): geographically eleven are from the Silver Hills (2.17-0.45 Ma; Figure 2) and four are from the Centre Hills (1.04-0.38 Ma; Figure 1b). All samples produced plateau ages that match our strict criteria. The first run of sample 17/76-AL did not yield a plateau, however, and therefore did not meet the criteria for a reliable age, whereas the second run yielded a release profile that meets our criteria and yielded an age of 1.18 6 0.30 Ma. Nevertheless, this age is viewed with caution as overall the sample shows significant heterogeneity and nonreproducible release spectra ( Figure 3). Harford et al. (2002) note evidence for xenocrysts in their ages obtained from plagioclase phenocrysts, leading them to interpret these ages as maximum ages, rather than true eruption ages. The influence of Geochemistry, Geophysics, Geosystems 10.1002/2017GC007053 xenocrysts in our sample ages can be assessed by examining the data from samples 05/03-AL and 15/70-AC, for which both groundmass and plagioclase ages were measured. For both samples the plagioclase age is systematically older than the groundmass age, suggesting that the plagioclase may contain a minor xenocrystic component, but there is overlap in the 2r errors of the two phases, which suggests that the plagioclase ages encompass the true eruption age within their 2r errors. The plagioclase ages obtained by Coussens et al. (2017) also likely represent the eruption ages, because they were obtained on similar material, and have similar scale 2r errors.
In the following sections, only the dates that are considered to be reliable are used in developing a detailed stratigraphy of the Silver Hills deposits, and when discussing the timing of volcanic activity.

Lithofacies of the Silver Hills Deposits
Here we adopt the nongenetic terminology of volcanic rocks proposed by White and Houghton (2006); lithofacies terms and abbreviations used are given in Table 3. For debris avalanche deposits we use the terms ''megablock facies'' (Voight et al., 2002) and ''matrix facies'' (Crandell et al., 1984;Voight et al., 2002).

Lava and Intensely Hydrothermally Altered Lava
The lava exposures are dominantly composed of plagioclasepyroxene-phyric andesite (30-60 vol % phenocrysts), with minor dacite and basaltic andesite with the same phenocryst populations. Lava thickness ranges from meters to hundreds of meters, with some exposures showing alignment of elongate crystals. The lava commonly locally transitions to breccia.
In some regions, the lava flows are extensively hydrothermally altered to an extent that no primary magmatic minerals are identifiable, and they exhibit a crumbly texture. The present day exposures may be white or orange and yellow in color, with the latter being associated with abundant centimeter-scale gypsum crystals. These altered exposures are surrounded by a halo of less hydrothermally altered rocks, with the degree of hydrothermal alteration decreasing outward.

Interpretation
The lavas form lava flows and domes, and the transitioning to breccia is interpreted as resulting from local autobrecciation of the lava along its margins. The zones of intensely hydrothermally altered lava are interpreted as the sites of fumarolic activity on the Silver Hills. The presence of gypsum crystals and pervasive variable discoloration make them comparable to the recent deposits of Galway's Soufrière (Voight et al., 2002), which also shows the intensity of hydrothermal alteration decreasing with distance from the fumaroles.

Massive to Diffusely Stratified, Lapilli-Tuff to Tuff-Breccia (m-dsLT-TBr)
The m-dsLT-TBr are poorly sorted, variably clast and matrix supported and consist of centimeter-scale to meter-scale subrounded to angular andesite clasts, up to 20% of which are hydrothermally altered. Some units of this facies display normal and inverse grading at their upper and lower boundaries, respectively.  n Age from nonreproducible release spectra. See supporting information Table 2 for decay rates, isotopic constants, and nucleogenic production ratios.

Interpretation
This facies is interpreted as block-and-ash flow deposits due to its massive structure, poor sorting, clast size range and grading. These characteristics are comparable to the block-and-ash flow deposits produced by dome collapses at Soufriè re Hills volcano during the 1995-2010 eruption (Cole et al., 2002;Stinton et al., 2014).

Massive Tuff-Breccia and Breccia, With Megablocks (mTBr-Br MB )
This facies is similar in appearance to the m-dsLT-TBr facies, but with three key differences: absence of grading, presence of jigsaw-fit fractures in some clasts, and the occurrence of megablocks. The megablocks are tens to hundreds of meters in scale, and can be split into two categories: lava megablocks and volcaniclastic megablocks. Lava megablocks vary from poorly to highly fractured and show variable hydrothermal alteration. Some of the megablocks are faulted. Volcaniclastic megablocks are blocks of other volcaniclastic facies (e.g., mLT) hosted within surrounding mTBr, and are either composed of a single unit or a layered sequence. Some megablocks are sheared, and others contain parts which appear to intrude into the surrounding breccia. Interdigitation of units within layered volcaniclastic megablocks is common. Most boundaries between megablocks and breccia are sharp, but some display signs of mixing.

Interpretation
The presence of megablocks within the mTBr suggests that they are parts of debris avalanche deposits; the megablocks are large coherent fragments of the original volcanic edifice that have been transported by debris avalanches with little internal deformation. Layered volcaniclastic megablocks may preserve their original structure and stratigraphy. The matrix facies consists of a poorly sorted massive breccia formed from the mingling of disaggregated megablocks and entrained material. Jigsaw-fit fracturing of clasts is a common feature of debris avalanche deposits (e.g., Crandell et al., 1984;Ui, 1983;Ui & Glicken, 1986), as are the fluidal textures at the edges of megablocks and deformation of units within individual volcaniclastic megablocks (e.g., interdigitation of units; e.g., Takarada et al., 1999;Ui & Glicken, 1986;van Wyk de Vries & Davies, 2015). Faulting is another well-documented feature of some megablocks (e.g., Glicken, 1996;Ui & Glicken, 1986;van Wyk de Vries & Davies, 2015).

Massive Lapilli-Tuff, With Regular Lapilli-Tuff and Tuff-Breccia Lenses (mLTlensLT-TBr)
The mLT are poorly sorted, matrix supported and consist of centimeter-scale rounded to subrounded andesite clasts, 10% of which are hydrothermally altered. The LT-TBr lenses are poorly sorted, clast supported and consist of centimeter-scale to decimeter-scale rounded andesite clasts, up to 5% of which are hydrothermally altered. They typically have meter-scale and decimeter-scale length and thickness, respectively, with decimeter-scale spacing.

Interpretation
This facies is interpreted as lahar deposits, on the basis that they contain multiple coarse lenses of rounded to subrounded andesite clasts in variably massive and parallel to cross-stratified coarse ash to coarse lapilli. They are similar to modern lahar deposits present in the Belham Valley, on the west flank of Soufriè re Hills (Barclay et al., 2007). 5.5. Pumiceous, Massive to Diffusely Stratified, Lapilli-Tuff to Tuff-Breccia (pm-dsLT-TBr) This facies can be split into two subfacies: well sorted and poorly sorted. Occurrences of the well-sorted subfacies are clast supported with centimeter-scale subangular to angular clasts of andesitic pumice and andesite. Clast proportions are typically 90% pumice and 10% andesite, with the pumice clasts being consistently coarser than the andesite clasts. Exposures of the poorly sorted subfacies are variably clast and matrix supported and consist of centimeter-scale to decimeter-scale rounded to angular clasts. The clast proportions are typically in the range of 65-99% pumice, 1-33% andesite, and 0-5% HA andesite. Locally, exposures of this subfacies have erosional bases. Note. Adapted from Branney and Kokelaar (2002). m-dsLT lensBr , for example, means massive to diffuse-stratified Lapilli-Tuff with lenses of breccia.

Interpretation
The well-sorted subfacies are interpreted as primary fallout deposits, due to the high degree of sorting and predominance of angular pumice with subordinate smaller lithics. The poorly sorted subfacies are interpreted as pumice-and-ash flow deposits, due to their poor sorting, rounded clasts and erosional bases, similar to pumiceand-ash flow deposits produced from recent Vulcanian eruptions of Soufrière Hills (Cole et al., 2002 6. Stratigraphy of the Silver Hills 6.1. Little Bay to South Drummonds, >2 Ma The oldest dated unit is an isolated exposure of dacitic mLT-TBr at south Drummonds. This unit contains meter-scale blocks, has a minimum thickness of 3 m, and is dated at 2.170 6 0.180 Ma (Figure 2). 6.2. Marguerita Ghaut and North Marguerita Bay, 1.7-1.6 Ma Marguerita Ghaut contains mTBr and coherent andesite and basaltic andesite lava exposures, with sharp boundaries between them. The mTBr has an age of 1.682 6 0.094 Ma (Figure 2). The eastern mouth of the ghaut contains a 35 m thick sequence of mTBr, and an andesite block from the base of this sequence has an age of 1.634 6 0.083 Ma (Figure 2). North Marguerita Bay contains a debris avalanche deposit consisting of a volcaniclastic megablock overlain by a mTBr MB ( Figure 5; see supporting information Text S1 for a detailed description).
6.3. Yellow Hole to Old Quaw, 1.5 Ma The Yellow Hole exposures consist of extensively hydrothermally altered andesite lava overlain by nonaltered lava dated at 1.520 6 0.051 Ma (Brown & Davidson, 2008). At the mouth of Old Quaw ghaut is a debris avalanche deposit (Figure 2), with a megablock of hydrothermally altered lava next to nonaltered mBr (matrix facies), and the boundary between them varies from sharp to transitional. There is an isolated block of the breccia within the hydrothermally altered lava, most likely resulting from internal deformation during transport (Siebert, 1984).
6.4. North West Bluff, 1.5-4 Ma North West Bluff is a 200 m high peak made of andesite lava with areas of local (<1 m 2 ) hydrothermal alteration, and an age of 1.493 6 0.098 Ma. Exposure of this lava continues as a ridge heading southeast from North West Bluff to Valentine Hill (Figure 2), which consists of hydrothermally altered lava overlain by nonaltered lava. In northwest Thatch Valley, the lavas have undergone extensive hydrothermal alteration, and locally becomes the 'intensely hydrothermally altered lava' facies.
North-central Thatch Valley contains a 45 m thick (minimum) mTBr which has an age of 1.450 6 0.160 Ma (Figure 2). Some clasts contain jigsaw-fit fractures, which supports an origin as a debris avalanche deposit, but the limited size of the exposure, coupled with the buried boundaries with surrounding units makes this difficult to confirm.
The andesite and dacite lava around and north of Rendezvous Bay is of a similar age, 1.424 6 0.080 Ma (Figure 2). There is exposure of 'intensely hydrothermally altered lava' midway between Rendezvous Bay and North West Bluff.
6.5. Silver Hill, 1.4 Ma Andesite lavas and mLT-Br's tens of meters thick are exposed alongside the road between Drummonds and Silver Hill and have been dated at 1.430 6 0.019 to 1.395 6 0.017 Ma (Brown & Davidson, 2008; Figure 2). Rendezvous Bluff consists of a 50 m high sequence of m-dsTBr overlying the lavas of Rendezvous Bay. Near Rendezvous they have a flow direction away from Silver Hill (Figure 2), and so are likely contemporaneous with the block-and-ash flow deposits between Silver Hill and Drummonds.
Further block-and-ash flow deposits from this period can be found overlying the basal hydrothermally altered lava of Valentine Hill (Figure 2). A ridge of this lava extends southeast from Valentine Hill, and on the southwestern side of this ridge the lava is overlain by 20 m of m-dsTBr's (units V1-4) infilling a paleovalley (supporting information Figure S1).
The lavas of Thatch Valley are locally overlain by mTBr, which locally has a hydrothermally altered matrix. These breccias are likely contemporaneous with the other mTBr units.  (Figure 1b). Due to its age, and location in a fault block, they interpret this deposit as originating from the Silver Hills. 6.7. Little Bay, 1.0-0.8 Ma The largest debris avalanche deposit of the Silver Hills is located in the Little Bay region, spanning Potato Hill, Davy Hill, and Little Bay (Figure 2 and supporting information Figures S2 and S3). It consists of lava-megablocks and volcaniclastic-megablocks, with surrounding matrix facies (see supporting information Text S2 for a detailed description). Lava from a megablock east of Potato Hill has an age of 1.430 6 0.120 Ma.   Figures 2 and 6a). Unit PH3 (primary fallout deposit) has been dated at 0.800 6 0.120 Ma (Figure 2) and has a Centre Hills origin (see section 8.1). This constrains the timing of the Little Bay debris avalanche to >0.8 Ma.

Geochemistry, Geophysics, Geosystems
6.9. Old Quaw Ghaut, 0.45 Ma The Old Quaw debris avalanche deposit is overlain by a horizontally bedded volcaniclastic sequence containing four units, three of which are separated by paleosols (units OQ1-4; Figure 6b) and were likely deposited in a paleovalley. The base of this sequence, a mLT lensLT-TBr (unit OQ1), is overlain by pumiceous deposits which display noticeable variation between two exposures 60 m apart (log sites 2-3; Figures 2 and 6b). Units OQ2a and 3a laterally transition between the well-sorted and poorly sorted subfacies of pmLT-TBr with small (centimeter-scale) erosional channels. At log site 2, they grade into coarser and finer grained poorly sorted pmLT-TBr, respectively, with higher proportions of lithic clasts and matrix. This grading of unit OQ2a is also displayed at log site 3, but here unit OQ3a is missing. Unit OQ2a has been dated at 0.450 6 0.130 Ma. The transition between the well-sorted and poorly sorted subfacies within the same unit suggests that units OQ2-3 are reworked fallout deposits. This interpretation is supported by the 40 Ar/ 39 Ar age and Pb isotope data, because the pumice clasts from unit OQ2a have a Soufrière Hills signature (see section 8.1). It is considered unfeasible for a pyroclastic density current to reach the eastern Silver Hills from the Soufriè re Hills, hence these pumice clasts are likely to be derived from fallout from an eruption column. This unit is the oldest known deposit from a Soufriè re Hills eruption.
7. Expanded Stratigraphy of the Centre Hills 7.1. East Coast, 1 Ma Sample 21/81-AC comes from a basaltic andesite mTBr on the east coast of Centre Hills (Figure 1b), which lies at the base of the stratigraphy of Coussens et al. (2017). The age of 1.040 6 0.250 Ma is within error of the oldest dates for this center (Coussens et al., 2017;Harford et al., 2002).

Southwest Centre Hills, 0.45 Ma
Pumice from a pumiceous mTBr in the southwest Centre Hills has an 40 Ar/ 39 Ar age of 450 6 170 ka (Figure  1b), and a Soufriè re Hills Pb isotope signature (see section 8.1). This provides further evidence for Soufriè re Hills volcanism at 450 ka.

Dry Waterfall and Spring Ghaut, 0.38 Ma
Andesite lava from Dry Waterfall (central Centre Hills) has been dated at 378 6 18 ka (Figure 1b) and has a Centre Hills Pb isotope value (section 8.1), providing the youngest known age for Centre Hills volcanism. A dacitic mTBr from Spring Ghaut (southwest Centre Hills) has been dated at 376 6 85 ka (Figure 1b) and has a Soufrière Hills Pb isotope signature. This demonstrates an overlap in activity between Centre Hills and Soufrière Hills volcanism.

Provenance
Many intraoceanic arcs display along-arc geochemical heterogeneity, caused by variations in source component contributions (e.g., Izu-Bonin-Mariana arc: Elliott et al. (1997) and Ishizuka et al. (2007); Tonga-Kermadec-Lau arc: Ewart et al. (1998); Lesser Antilles arc: Macdonald et al. (2000)). Such variation has also been observed on individual islands, allowing the eruption deposits of distinct volcanic centers to be distinguished using isotope and trace element ratios. For example, Pagan island in the Mariana arc contains two volcanic centers 8 km apart, Mount Pagan and South Pagan, whose eruption deposits can be distinguished using trace element ratios (e.g., Ba/Th, Ba/La, and Nb/Zr; Marske et al., 2011). The island of Hachijojima in the Izu-Bonin arc comprises two volcanic centers 7 km apart, Higashiyama and Nishiyama, the deposits of which can be distinguished using Pb isotopes, and Zr/Y and La/Sm ratios (Ishizuka et al., 2008). In the Lesser Antilles, there is a clear north-south gradient in the Sr, Nd, and Pb isotope ratios, and trace element patterns of volcanic rocks along the entire arc that has persisted for 5 Myr Lindsay et al., 2005). This geographical pattern is also observed on individual islands, such as Martinique, where 143 Nd/ 144 Nd can be used to separate deposits from Pitons du Carbet and Mont Conil-Mont Pel ee, which overlapped in activity 545-322 ka (Germa et al., 2011b;Labanieh et al., 2012). For Montserrat, Pb isotopes have been shown to clearly distinguish Soufriè re Hills deposits from those of the Centre and Silver Hills, and 143 Nd/ 144 Nd can be used to differentiate between Silver Hills and Centre Hills deposits (Cassidy et al., 2012).
For some samples, in this study, there is ambiguity over which volcanic center they are from, such as 16/72-PR, which is geographically within the Silver Hills, but has a Centre/Soufriè re Hills age. Here we use trace element and isotope geochemistry to help constrain their provenance (data is presented in supporting information Table S4). Pumice from a fault block in South Lime Kiln Bay (sample 11.1.4C), which is geographically part of the Centre Hills, has been dated at 1.31 Ma, leading to the interpretation that this deposit originated from Silver Hills volcanism (Coussens et al., 2017). This interpretation is supported by 143 Nd/ 144 Nd versus Nb/Y, in which this same sample plots within the Silver Hills field (Figure 8a).  (Hart, 1984). Geochemistry, Geophysics, Geosystems

10.1002/2017GC007053
The Silver Hills-Centre Hills boundary defined by Harford et al. (2002) follows Brimm's Ghaut, which runs through Marguerita Bay (Figure 2). Sample 12/13-AC from south Marguerita Bay (south of the boundary) has been dated at 1.33 Ma, suggesting a Silver Hills origin. This is supported by 143 Nd/ 144 Nd versus Nb/Y (Figure 8a), suggesting that the boundary between the Silver Hills and Centre Hills is further south than previously thought.
Sample 23/86-PF (unit PH3 on Potato Hill) is geographically within the Silver Hills but has an age of 0.8 Ma, suggesting this deposit is derived from the Centre Hills (Figure 2). This interpretation is supported by a Centre Hills geochemical signature (Figure 8a). The pumice sample from Davy Hill (24/88-PC) is also geographically within the Silver Hills but falls in the Centre Hills field in Figure 8a, suggesting a Centre Hills origin. Because this sample is part of the Little Bay debris avalanche deposit, this suggests that the debris avalanche occurred during the early stages of Centre Hills volcanism.

Marine Tephra Deposits
Marine sediment core U1396C contains fifteen tephra layers with thicknesses >1 cm within the 2.35-0.37 Ma time span covered by our new 40 Ar/ 39 Ar ages. The age assignments of the tephra layers are derived from paleomagnetic, biostratigraphic and foraminifera d 18 O correlations (Fraass et al., 2016;Hatfield, 2015). Pb isotope analyses show that thirteen of these layers are from Montserrat and the other two are from Guadeloupe . Of the Montserrat layers, eight contain enough pumice or fresh nonvesicular lava grains for geochemical analysis (Table 4). Of particular interest are the five tephra layers in the 1-1.3 Ma period (the gap in the terrestrial age record between the Silver and Centre Hills) with sufficient pumice. Figure 8b shows that the layers with ages of 1.10 and 1.03 Ma fall within the Silver Hills field, and 1.14, 1.08, and 1.04 Ma fall within the Centre Hills field (Figure 8b). The absolute uncertainty in the age of any individual tephra layer is 50 kyr, but the stratigraphic order of the layers is certain. Hence, we can be confident that there was overlap in activity between the Silver and Centre Hills for at least 130 ka.

Petrology
The lavas of the Silver Hills volcanic center are porphyritic (30-60 vol % phenocrysts) with a phenocryst assemblage of 65-75% plagioclase up to 4 mm, 15-20% orthopyroxene up to 6 mm and 5-15% clinopyroxene up to 5 mm. Quartz and amphibole are locally present in minor amounts. Zoning is observed in some clinopyroxene and orthopyroxene phenocrysts, and all plagioclase phenocrysts, with many of the latter displaying sieve textures. The groundmass is generally microcrystalline and composed of plagioclase, orthopyroxene, clinopyroxene, and Fe-Ti oxides (Figure 9), with some samples showing alignment of elongate crystals.  Enclaves are abundant within Silver Hills lavas, and comprise up to 20 vol %. They have coarser groundmasses (composed of plagioclase, orthopyroxene, clinopyroxene, and Fe-Ti oxides) than the host andesite, with a higher proportion of pyroxenes (Figure 9). Many of the enclaves have a diktytaxitic groundmass, with interlocking, randomly orientated elongate crystals, similar to enclaves erupted during the 1995-2010 eruption of the Soufriè re Hills .
The Centre Hills samples have a similar phenocryst assemblage as the Silver Hills lavas, closely matching the samples described by Coussens et al. (2017). The Soufriè re Hills samples are porphyritic (30 vol % phenocrysts) with a phenocryst assemblage of 65-75% plagioclase, 15-20% orthopyroxene, and 10-15% amphibole, with minor clinopyroxene and quartz, similar to the Soufrière Hills lavas erupted in the past 130 kyr (Harford et al., 2002).

Evolution of the Silver Hills Volcanic Center
Subaerial volcanic activity began in the Silver Hills prior to 2 Ma, with the oldest rocks present as lava domes around the Little Bay area and dome collapse deposits built up the south Drummonds region (Figure 10). At 1.7-1.6 Ma, activity migrated to the eastern part of Silver Hills, forming the lava domes and associated deposits of Marguerita Bay. Part of this edifice collapsed to form the Marguerita Bay debris avalanche deposit. Activity continued in this region up until 1.5 Ma, forming the lavas of Yellow Hole, with collapse of this edifice forming the Old Quaw debris avalanche. Next, activity shifted to the northwestern part of the Silver Hills, forming the lava domes of North West Bluff, Thatch Valley and Rendezvous Bay at 1.5-1.4 Ma, with concomitant hydrothermal activity producing the observed fumarole deposits in this region. At 1.4 Ma, activity shifted again to form lava domes at Silver Hill and Drummonds, which collapsed to form blockand-ash flow deposits throughout the north, west and center of the Silver Hills. Collapse of these edifices with part of the young Centre Hills edifice created the Little Bay debris avalanche deposit at around 1-0.8 Ma. The block-and-ash flow deposits of south Marguerita Bay suggest that lava dome growth was still active in the Drummonds area 1.35-1.30 Ma (Figure 10). The youngest deposits found in the Silver Hills region are reworked pumice fall deposits in Old Quaw, which Pb isotopes indicate to be derived from eruptions of the early stages of Soufrière Hills volcanism 0.45 Ma.

Comparisons With Soufriè re Hills and Centre Hills
The Silver Hills is the smallest volcanic center on Montserrat in terms of subaerial exposure, but bathymetric data may provide an insight into its original size. Montserrat is surrounded by a shallow submarine shelf, which extends up to 5 km wide around the Silver Hills (Figure 1b), and is interpreted as representing the original expanse of the Silver Hills (Le Friant et al., 2004). Assuming that the Silver Hills originally had a similar size to the Soufriè re Hills, the estimated minimum original volume of the Silver Hills above 100 m below sea level (the depth of the submarine shelf) is 17 km 3 ; the subaerial part of the Soufrière Hills is 12 km 3 (Le Friant et al., 2004).
Our work on the deposits of the Silver Hills volcano has shown its past volcanic activity to be similar to that of the Soufriè re Hills, with both characterized by lava dome growth and collapse coupled with Vulcanian eruptions and periodic flank/sector collapses. Soufrière Hills does, however, exhibit evidence of larger, sustained eruptions at 450 and 179 ka, whereas there is no record of large eruptions emanating from the Silver Hills. The Centre Hills activity is also characterized by lava dome growth and collapse with Vulcanian eruptions, but also produced multiple Plinian eruptions (possibly up to magnitude 5) throughout its eruptive history (Coussens et al., 2017;Harford et al., 2002). The consistency in the style of volcanic activity throughout the lifespan of the Silver Hills adds support to the hypothesis of Harford et al. (2002) that future activity at Soufriè re Hills is likely to continue in the same style as the 1995-2010 eruption.
It is worth noting, however, that the exposed massifs of the Soufriè re Hills and Centre Hills are both currently larger in size than Silver Hills, and that the Plinian deposits of both the former are situated along coastal exposures that are now missing from Silver Hills. Additionally, the current size of the subaerial portion of Silver Hills is approximately the same size as the central dome complex of the Soufriè re Hills which, like the central part of Centre Hills, is dominated by massive andesite dome-lava and their collapse deposits (Harford et al., 2002). Thus, it is feasible that any large sustained explosive eruptions that Silver Hills may Geochemistry, Geophysics, Geosystems have produced are not preserved on land, akin to the Basal Complex, the oldest volcanic center on Basse-Terre, Guadeloupe. This is another deeply eroded volcano whose subaerial record is dominated by effusive eruption products, with no evidence of large explosive activity. However, the study of a marine sediment core has revealed that this seemingly effusive center produced a large Plinian eruption (Volcanic Explosivity Index 6) toward the end of its life, at 2.4 Ma, which is believed to be the largest known eruption in the Lesser Antilles . While the Silver Hills may not have produced eruptions of this magnitude, this example serves to highlight a potential preservation bias within the Silver Hills' terrestrial record.

Comparisons With Other Arc Volcanic Centers
Many arc volcanoes dominated by andesitic and dacitic lava dome formation nevertheless also experience periodic Plinian eruptions, similar to that inferred for the Centre Hills (Coussens et al., 2017). For example, the 1902 Plinian eruption of Santa Maria, a dominantly andesitic (Rose, 1972) stratovolcano in Guatemala, was one of the 10 largest eruptions ever observed (Bennett et al., 1992), yet volcanic activity at Santa Maria has been dominated by eruptions of lava flows and domes, with their associated collapse pyroclastic deposits (Rose, 1972(Rose, , 1973(Rose, , 1987. Indeed, the active Santiaguito dome complex has built up inside the crater produced from the 1902 eruption. In Chile, Volc an Quizapu produced two of the largest historic eruptions of South America, which were of a similar composition, yet one was effusive (a 5 km 3 mingled andesitedacite lava flow in 1846-1847), and one was explosive (a 4.5 km 3 dacitic Plinian eruption in 1932; Hildreth & Drake, 1992). The transition between effusive and explosive eruptions at Volc an Quizapu has been linked to the extent of mafic magma recharge and mixing prior to eruption. The 1932 Plinian eruption had a relatively minor recharge component, while the recharge magma comprised <10-45 vol % of the lava erupted throughout 1846-1847 (Ruprecht & Bachmann, 2010). The resident magma temperatures for both eruptions is estimated to be 8708C (Ridolfi et al., 2010), but the greater volume of recharge magma of the 1846-1847 eruption led to an eruption temperature up to 1308C hotter than the 1932 eruption. It is thought that this higher magmatic temperature enhanced syneruptive magma degassing and led to effusive eruptive behavior, by accelerating volatile diffusion, lowering melt viscosity, and inhibiting brittle fragmentation (Ruprecht & Bachmann, 2010).
Magma recharge and mixing is also considered to be responsible for the long-term (500-0 ka) low explosivity of Mount Hood, OR, another andesite-dacite arc volcano with no evidence for large explosive eruptions (Koleszar et al., 2012;Scott et al., 1997). Viscosity models show that mafic recharge beneath Mount Hood may result in a fivefold to tenfold decrease in viscosity of the silicic resident magma, facilitating greater volatile diffusion, and delaying or precluding fragmentation during magma ascent. The continued low explosivity of activity at Mount Hood over 500 kyr is thought to be due to somewhat constant mixing proportions and time scales, whereas variability in mixing proportions and time scales leads to variation in eruption styles between effusive and explosive (Koleszar et al., 2012), such as observed at Volc an Quizapu (Ruprecht & Bachmann, 2010). Further, Koleszar et al. (2012) invoke this model to explain the long-term low explosivity (i.e., absence of Plinian eruptions) in other andesitic-dacitic arc volcanoes where magma mixing is a common process, such as Unzen volcano, Japan (Hoshizumi et al., 1999;Venezky & Rutherford, 1999), and Mount Dutton, AK (Miller et al., 1999).
On Montserrat, this model may also explain the low explosivity behavior of the Silver Hills and Soufriè re Hills, because the lavas of these centers contain abundant mafic enclaves, indicating magma mixing has consistently played a major role in controlling eruption styles at these volcanic centers (e.g., Murphy et al., 1998;this study). Furthermore, for the Centre Hills deposits, enclaves are only present within lavas, and not in the pumiceous (explosive) deposits (Coussens et al., 2017). This suggests that the absence of, or at least reduced extent of, magma recharge and mixing may have played a significant role in producing large explosive eruptions at the Centre Hills. However, this model relies on the recharge magma raising the temperature of the resident magma to prevent explosive eruptions, but geothermometry studies estimate that magma storage temperatures beneath the Centre Hills (810-10808C; Coussens et al., 2017) were similar to those of the 1995-2010 Soufrière Hills eruption (785-11008C; Barclay et al., 1998;Christopher et al., 2014;Devine et al., 1998, 2003. This suggests that increased temperature alone cannot be responsible for the explosive eruptions of the Centre Hills. Coussens et al. (2017) showed that the composition of Center Hills deposits remained constant, thus ruling out compositional control on eruption styles. They suggest instead that explosive eruptions may have been the result of local changes in magma storage conditions (variable magma temperature or viscosity) and preeruptive dynamics.

10.1002/2017GC007053
In the Lesser Antilles, Silver Hills volcanism is typical of that documented along the northern arc, in that andesitic dome growth and collapse-with small explosive eruptions-has dominated volcanic activity at Saba, St. Eustatius, St. Kitts, and Nevis, with no evidence for larger, sustained eruptions (Baker, 1984(Baker, , 1985Baker et al., 1980;Davidson & Wilson, 2011;Defant et al., 2001). Petrological evidence (e.g., reverse-zoned phenocrysts) indicate that magma recharge and mixing is a common process at Saba and St. Kitts (Baker et al., 1980;Toothill et al., 2007), which following the model of Koleszar et al. (2012), suggests that the characteristics of magma recharge and mixing have been constant for the northern islands, thus inhibiting the potential for large explosive eruptions and maintaining low explosivity behavior. The central and southern islands display similar activity to the northern islands, but with Plinian eruptions identified at La Soufriere, Guadeloupe (Komorowski et al., 2005), Morne Diablotins and Morne Trois Pitons-Microtrin, Dominica (Boudon et al., 2017), Mont Pel ee, Martinique (Roobol & Smith, 1976;Westercamp & Traineau, 1983), Qualibou, St Lucia (Wohletz et al., 1986), and Soufrière, St Vincent (Rowley, 1978;Wright et al., 1984), more akin to the Centre Hills volcanism (Coussens et al., 2017). In Dominica, mafic enclaves are present in lavas from multiple volcanic centers but are notably absent from pumiceous deposits (Howe et al., 2015), further supporting a link between magma recharge and mixing and eruption explosivity. Further detailed studies comparing evidence for magma mixing (e.g., enclave abundances, reverse-zoned phenocrysts) between eruption styles are required to further investigate the link between magma mixing and eruption explosivity in the Lesser Antilles.

Revised Geochronology of Volcanism on Montserrat
Here we present a revised geochronology of volcanic activity of Montserrat's three main volcanic centers (Figure 11), based on a review of existing ages, new 40 Ar/ 39 Ar dates, and paleomagnetic ages from marine tephra layers.
Previously, Silver Hills volcanism was dated to range from 2.6 to 1.2 Ma (Brown & Davidson, 2008;Harford et al., 2002), but as discussed some of these dates are unreliable. Silver Hill volcanism can now be reliably constrained to at least 2.17-1.03 Ma. Coussens et al. (2017) divided Centre Hills volcanism into two periods of activity, spanning >0.95 to 0.60 Ma and 0.60 to 0.40 Ma. The new Centre Hills dates presented in this study expand the timing of Centre Hills volcanism by 0.2 Myr, to 1.14-0.38 Ma.
Three stages of Soufriè re Hills volcanism have been previously recognized: >300-175 ka, 175-130 ka, and 112 ka to recent (Coussens et al., 2017;Harford et al., 2002). Our new ages identify a fourth, earlier stage, 450-300 ka, characterized by the eruption of hornblende-orthopyroxene andesites, similar to lavas erupted since 112 ka, and in contrast to the two-pyroxene andesites erupted 290-130 ka (Harford et al., 2002;Rea, 1974). This means that the petrology of the Soufrière Hills eruptive products has changed at least twice throughout the volcano's development. Garibaldi Hill, which contains a two-pyroxene andesite blockand-ash flow deposit dated at 290 ka (Harford et al., 2002), contains both two-pyroxene and hornblendehypersthene andesites, but the stratigraphic relationship between the two andesite types is not documented (Rea, 1974). A detailed stratigraphy of Garibaldi Hill may provide further details as to when the Geochemistry, Geophysics, Geosystems 10.1002/2017GC007053 change from hornblende-hypersthene to two-pyroxene andesite took place, and if it was an instant or a gradual change.
Our new ages show that Silver, Centre and Soufriere Hills were active for at least 1.14, 0.76, and 0.45 Myr, respectively. The most comparable island to Montserrat in the Lesser Antilles arc is the neighboring island of Basse-Terre, Guadeloupe. It also formed following tectonic adjustments during the Mid-Miocene and similarly shows a north-south migration in the locus of volcanism over a similar time interval. There is also high quality age data available for the volcanic centers on Basse-Terre; with 2.79-2.68 Ma, 1.81-1.15 Ma, 1.02-0.44 Ma, 0.56-0.47 Ma, and 0.21 Ma to present recorded for the Basal Complex, Septentrional Chain, Axial Chain, Monts Cara€ ıbes Massif, and Grande D ecouverte Volcanic Complex, respectively (Carlut et al., 2000;Ricci et al., 2015a, 2015b, Samper et al., 2007. With the exception of the oldest Basal Complex (which is heavily eroded and has limited subaerial exposure) and Monts Cara€ ıbes Massif (which has limited subaerial exposure and only two dates), these intervals of activity (0.66, 0.58, and 0.21 Myr) for the different centers on Basse-Terre are similar to those observed on Montserrat. These observations suggest that 0.5-1.0 Ma is a common lifespan for volcanic centers in the region. The question arises, therefore, as to whether this apparent time scale of volcanic center duration is purely stochastic or whether it has more fundamental significance. For example, the duration of a volcanic center may be related to the life cycle of volcanic edifices. Modeling of edifice loading indicates that progressive growth of a volcanic structure results in the confining pressure increasing to the extent that the center is forced to migrate laterally (Pinel et al., 2010). Alternatively, the duration of a volcanic center may be related to deeper processes. For example, it has been suggested that partial melting in the mantle wedge yields rising diapirs of partially molten material that rise through the mantle with frequencies of order 10 5 to 10 6 years (Hall & Kincaid, 2001;Stern, 2002). Further detailed geochronology of volcanic centers on other islands of the Lesser Antilles is required to assess these hypotheses in more detail.
Our new ages also provide evidence for a 130 kyr overlap in activity between the Silver and Centre Hills, and 70 kyr overlap between the Centre and Soufriè re Hills. There was also overlap between the Soufriè re and South Soufrière Hills (Cassidy et al., 2012(Cassidy et al., , 2014. Overlap in volcanic activity between two or more centers on an individual island has also been documented on Martinique, where Trois Ilets and Pitons du Carbet were both active from 998 to 345 ka. Here further coeval activity occurred at Mont Conil-Mont Pel ee during 545-345 ka (Germa et al., , 2011b. On Dominica, there has been coeval activity between at least up to four volcanic centers over much of the lifetime of the island (Smith et al., 2013, and references therein). The volcanic centers of Basse-Terre, Guadeloupe, have been extensively dated, and overlap in activity has only been identified between two centers: the Axial Chain and Monts Cara€ ıbes during 555-472 ka (Carlut et al., 2000;Ricci et al., 2015a, 2015b, Samper et al., 2007).
The overlap in volcanic activity between centers on these four islands suggests that it is a common phenomenon in the Lesser Antilles and, indeed, in other arc systems. For example, in the Izu-Bonin-Mariana arc, most islands contain a single volcanic center, but islands with multiple centers have experienced coeval activity at two neighboring volcanic centers. These are Hachijojima, where Higashiyama and Nishiyama (7 km apart) have been concurrently active for at least 10 ka (Ishizuka et al., 2008); Pagan island, where Mount Pagan and South Pagan (8 km apart) have experienced coeval activity for at least 64 ka (Marske et al., 2011); and Izu-Oshima, which has been simultaneously active with the neighboring Izu-Tobu volcanic field for at least 40-50 ka, which at their closest are only 4 km apart (Ishizuka et al., 2015 and references therein). The distance separating sites of coeval activity in the Izu-Bonin-Mariana and Lesser Antilles arcs (typically <10 km) relative to the depth to the melt generation zone above the subducting slab (of the order 100 km; Stern, 2002) suggests that a rising diapir or plume of partial melt may bifurcate to form separate magma chambers at intermediate depths (15 km;Odbert et al., 2014) beneath the individual volcanic centers.

Conclusions
We describe the rocks of the Silver Hills volcanic center, which are dominated by andesite lavas and breccias, formed from effusive lava dome eruptions and dome collapses. Pumiceous flow deposits are also present, providing evidence of explosive activity from Vulcanian style eruptions. There is evidence of widespread hydrothermal alteration, with three fumarole deposits identified, resulting from hydrothermal Geochemistry, Geophysics, Geosystems 10.1002/2017GC007053 activity concurrent with Silver Hills volcanism. Four debris avalanche deposits were also identified, indicating that the volcano experienced periodic sector collapses. These observations suggest that the Silver Hills was characterized by the same type of volcanic activity as has been observed at the Centre Hills and Soufriè re Hills. The notable absence of evidence of sustained explosive eruptions at Silver Hills, which occurred in the early stages of Soufrière Hills' development and throughout Centre Hills' activity, may be the result of consistent mafic magma recharge and mixing, which could raise magma temperatures and inhibit the potential for large explosive eruptions.
New 40 Ar/ 39 Ar dates, combined with ages from marine tephra layers and a review of existing ages, have yielded a revised geochronology of volcanic activity of Montserrat's three main centers, which were active during at least 2.17-1.03 Ma, 1.14-0.38 Ma, and 0.45 Ma to present for the Silver, Centre, and Soufrière Hills centers, respectively. Two key findings come from these new dates: the previously unknown overlaps in volcanic activity between both the Silver and Centre Hills and Centre and Soufriè re Hills, and the discovery of a new, older stage of Soufrière Hills activity 450-290 ka with the eruption of hornblendeorthopyroxene lavas. Combined with ages of volcanic centers on Basse-Terre, Guadeloupe, our ages suggest that 0.5-1 Ma is a common lifespan for volcanic centers in the Lesser Antilles. Furthermore, overlap in activity between closely spaced (<10 km) volcanic centers appears to be a common phenomenon in island arcs, as it has been observed on multiple islands in the Lesser Antilles and Izu-Bonin-Mariana arcs.