Plutonic record of a caldera-forming silicic eruption: The shatter zone of the Cadillac Mountain granite, coastal Maine

Petrologists and volcanologists increasingly recognize that a full understanding of crustal magmatism requires study of both plutonic and volcanic rocks (Bachmann and Huber, 2016; Bachmann et al., 2007; Keller et al., 2015; Lipman and Bachmann, 2015; Lundstrom and Glazner, 2016; Sparks and Cashman, 2017). As a result, shallow-level plutonic complexes associated with coeval volcanic rocks, which offer the opportunity to correlate contemporary processes and events in the plutonic and volcanic records, have increasingly attracted scientific attention (Karakas et al., 2019; Sewell et al., 2012; Watts et al., 2016). The earliest of such studies, conducted in the SW United States, examined the relations among caldera structures, erupted volcanic products, and shallow-level plutonic rocks (e.g., Lipman, 1984; Seager and McCurry, 1988). While these studies provided important insights, constraining an integrated magmatic history has proven difficult.

Perhaps because of this problem, a controversy emerged, with some workers suggesting that large ignimbrite eruptions need not be related directly to plutons and that volcanism and plutonism may not generally be linked (Bartley et al., 2005; Glazner et al., 2006; Mills and Coleman, 2013). On the other hand, most workers, often citing geophysical and petrologic evidence for the presence of large contemporaneous crystal mush zones beneath active calderas (e.g., Pritchard and Gregg, 2016; Bachmann and Huber, 2016, and references therein) argue that silicic volcanism is accompanied by pluton formation at depth (Cole et al., 2014; Kern et al., 2016; Lipman, 2007; Lipman and Bachmann, 2015; Wilson and Charlier, 2016). Nonetheless, numerous difficulties remain in connecting the plutonic and volcanic records of magmatic evolution. One major challenge is that the magmatic processes in silicic systems can operate on brief time scales that are below the resolution of isotopic dating methods available for plutonic rocks (Wilson and Charlier, 2016). There are also lingering questions regarding the type of physical record that a caldera-forming eruption might leave at depth (Lipman, 2007).

A useful approach in examining possible links between volcanism and plutonism is the study of strongly tilted and/or dissected terrains where both contemporaneous volcanic and plutonic rocks are exposed, and where it is possible to try to connect the stratigraphy of volcanic deposits to the history of plutonic systems (Colombini et al., 2011; Eddy et al., 2018; Faulds et al., 2008; Leigh et al., 2018; McDowell et al., 2014; Metcalf, 2004; Rioux et al., 2016; Schaen et al., 2018; Tang et al., 2017; Verplanck et al., 1999; Walker et al., 2008). Another approach is to focus on the plumbing system exposed within exhumed caldera ring dikes (Kennedy and Stix, 2007; McDonnell et al., 2004; Tomek et al., 2018). The Cadillac Mountain intrusive complex in Maine (USA; Wiebe, 1994; Wiebe et al., 1997a) is an example where tilting and erosion provide exposures of all of these facies, from the base of the intrusion up to its shallow-level margins and overlying volcanic deposits (Fig. 1), offering an exceptional opportunity to link coeval volcanic and plutonic processes.

The eastern and southern contacts of the Cadillac Mountain complex expose a unique magmatic breccia unit (“shatter zone”), consisting of intensely fractured and deformed country rock with clast size distributions indicative of explosive fragmentation (Roy et al., 2012). Inward from these margins, a chaotic zone of chemically and texturally diverse matrix hosts abundant blocks of country rock. The magma within this zone includes both silicic and more mafic hybrid compositions. These features provide a rare plutonic record of a major caldera-forming eruption, in which debris from the subsiding chamber roof and ring fault vents collapsed onto crystal mush beneath an eruptible layer of silicic magma.

To date, very little is known regarding the nature of and processes within such caldera-related marginal zones at plutonic depths, in contrast to higher structural levels (e.g., Kennedy and Stix, 2007) or at the surface (e.g., Rooyakkers et al., 2020). Here, we document in detail the structural, chemical, and textural features of the shatter zone and discuss its formation and implications for understanding links between plutonism and volcanism, and caldera-forming eruptions in particular.

The Silurian Cadillac Mountain intrusive complex and the associated Cranberry Island volcanic series (Seaman et al., 1999, 2019) are located on Mount Desert Island, Maine. They belong to the late Silurian to early Carboniferous Coastal Maine magmatic province, an ∼100 × 300 km zone of plutonic and associated volcanic rocks hosted in several NE-trending terranes of mainly pre-Devonian metavolcanic and metasedimentary units. The province reflects large-scale bimodal magmatism that included supervolcano-scale eruptions. Coeval volcanic successions ranging from 650 m up to >2.2 km in thickness (single ignimbrites thicknesses from 350 to 860 m) are preserved at three distinct intrusive complexes within the province (Seaman et al., 2019).

The Mount Desert Island complex consists of four main units in an oval area of ∼14 × 20 km (Fig. 1). All units are tilted moderately (10°–30°) to the SE, thus exposing progressively deeper structural levels of the intrusion to the west (Wiebe, 1994). The Cadillac Mountain granite, a hypersolvus hornblende granite, is the dominant unit and makes up most of the complex. The Cadillac Mountain granite is underlain by a gabbro-diorite unit consisting of interlayered gabbroic, dioritic, and minor granitic rocks. Gravity data (Hodge et al., 1982) and field relations suggest that the gabbro-diorite unit forms a saucer-shaped body, 2–3 km thick, beneath a lensoidal mass of Cadillac Mountain granite <3 km thick (Fig. 1). The younger Somesville biotite granite was intruded above the gabbro-diorite unit, forming a smaller body nested within the Cadillac Mountain granite. The older Southwest Harbor biotite granite is crosscut by these younger units in the southern part of the complex (Fig. 1; Wiebe et al., 1997a).

Extensive exposures of silicic volcanic rocks, including both lavas and ignimbrites, occur to the south of the intrusive complex on the Cranberry Islands (Seaman et al., 2019). The Cranberry Island series consists of an ∼1.8-km-thick section of felsic volcanic deposits (mainly rhyolitic ignimbrite) overlain by ∼750 m of basaltic tuffs and lavas (Gilman et al., 1988; Seaman et al., 1999, 2019).

Although radiometric data from these rocks are very scarce, zircon dating of the Cranberry Island rhyolites and the granites on Mount Desert Island by Seaman et al. (1995) placed them within a similar age range, providing further evidence for the plutonic-volcanic connection between them. The Cadillac Mountain granite and the Somesville Granite yielded U-Pb zircon ages of 419 ± 2 Ma and 424 ± 2 Ma, respectively, and the U-Pb zircon age for a single crystal from the Cranberry Island rhyolites was reported as 424 Ma ± 1 Ma (Seaman et al., 1995). The ages reported for the Mount Desert Island granites are, however, suspect because field relations demonstrate beyond doubt that the Somesville Granite is younger than the Cadillac Mountain granite (Wiebe et al., 1997a). Because the Cranberry Island volcanics are a series of units, it is most likely that the Cadillac Mountain granite complex generated more than one caldera-forming eruption over the course of its lifetime (possibly ranging from 1 to 3 m.y.). This suggests that it was a multicyclic system, like many, if not most, calderas.

In addition to voluminous crystal-poor rhyolite preserved on the Cranberry Islands, several further lines of evidence imply that the Cadillac Mountain granite reflects the cumulate base of a large upper-crustal silicic magmatic system (for details, see Supplemental Material S2: Cumulus granite and crystal mush¹). Emplacement at shallow depths, likely <5–7 km, is implied by: (1) abundant miarolitic cavities in the Upper Cadillac Mountain granite; (2) whole-rock compositions that plot close to the 500 bar minimum in the system Qtz-Or-Ab-H₂O (Wiebe et al., 1997a); (3) the presence of ferro-edenite, which is stable only at low pressures (Wiebe et al., 1997a); and (4) scarce zones of hydrothermal desilication in the Cadillac Mountain granite, which represent subsolidus reaction between granite and H₂O-rich vapor at pressures <800 bar as the Cadillac Mountain granite cooled (Nichols and Wiebe, 1998). The overall size and depth of the Cadillac Mountain granite magma system bear similarities to the crystal mush zones underlying large active silicic systems such as Long Valley caldera (Schmandt et al., 2019) or Laguna del Maule (Wespestad et al., 2019).

A wide zone of intrusive breccia is exposed for at least 25 km along the eastern and southern margins of the Cadillac Mountain granite, in contact with country rock (Braun, 2018). Because the outer edge of this breccia along the east coast is characterized by intense shattering of the country rock, Gilman et al. (1988) termed it a “shatter zone.” We retain this name here because of its long usage, but we recognize that only the outer edge along the east coast displays this shattered character. An explosive origin for the shatter zone was suggested by Roy et al. (2012) based on clast-size distribution analysis of country rock fragments >1.5 cm within the outer 50–100 m of its eastern margin. Fractal dimensions of clast populations generated in shear or via compressive fragmentation commonly tend to not exceed values of ∼2.5, whereas explosive fragmentation via decompression can be much more effective at generating finer grains and therewith higher fractal dimensions; see, for example, Sammis and King (2007), Kolzenburg et al. (2013), and Melosh et al. (2014), as well as references therein. The results of Roy et al. (2012) yielded surface fractal dimensions (D_s) >3, thus significantly exceeding the minimum value of 2.5 expected from explosive fragmentation. This supports the interpretation that the shatter zone formed during a major explosive eruption from the Cadillac Mountain granite magma chamber. The shatter zone provides a crucial piece of the plutonic-volcanic puzzle because it is the only unit of the Cadillac Mountain granite complex displaying evidence of fragmentation. Here, we describe the textural features of the shatter zone in detail and provide evidence for its explosive origin.

The shatter zone is best exposed along the eastern and southern margins of the Cadillac Mountain complex (Fig. 1). A small and poorly exposed lens of shatter zone material also occurs along a thrust fault that cuts the northern margin of the Cadillac Mountain granite (Braun, 2018), suggesting that the shatter zone likely also continued along the northern margin. Xenoliths of country rock within the Cadillac Mountain granite immediately above the Cadillac Mountain granite–gabbro-diorite unit contact to the north of the Somesville Granite (see Fig. 1) further support this observation. Lenses of country rock blocks, as well as lenses of gabbroic to intermediate rocks connected to the underlying gabbro-diorite unit that dips SE beneath the Cadillac Mountain granite, also occur along the southern margin of the Cadillac Mountain granite, south of the Somesville Granite, suggesting that the shatter zone extended along most of the Cadillac Mountain granite margin.

In this paper, we divide the shatter zone into two segments of distinct character: an eastern segment and a southern segment (Fig. 1). The inner boundary of the shatter zone is defined to enclose all country rock xenoliths in the Cadillac Mountain granite. Because the complex is tilted eastward, the eastern segment exposes the shallowest preserved structural levels of the intrusion and surrounding country rock, while the southern shatter zone segment exposes increasingly deeper structural levels to the west. Because gravity data (Hodge et al., 1982) suggest that an ∼3 km section of Cadillac Mountain granite overlies the gabbro-diorite unit, the western end of the southern shatter zone segment is likely 2–3 km deeper than the eastern segment. Hence, the changes in the character of the shatter zone between the two segments described below can be linked to variations with depth.

The eastern segment of the shatter zone occurs in contact with the Bar Harbor Formation, a metasedimentary unit intruded by large pre–Cadillac Mountain granite dikes and sills of diabase. This segment varies in width from <150 m to slightly >1 km. Its outer contact is steep wherever it is against diabase, seen most clearly at Great Head (Fig. 2C), where it cuts near vertically from sea level to ∼35 m elevation. Along wider exposures of shatter zone near the center of the eastern segment, the trace of the outer contact against Bar Harbor Formation suggests a moderate (10°–30°) westward dip toward the center of the intrusion. North of this, again against diabase, the trace of the contact passes straight across hills, implying a nearly vertical dip. The trace of the inner shatter zone boundary in the eastern segment suggests a gentle westward dip of ∼10°, toward the center of the intrusion; correcting for an eastward tilt of ∼20°, the original dip would have been ∼30°.

Four lenses of Cadillac Mountain granite that sit structurally above the shatter zone are exposed within the eastern segment as “islands” enclosed on all sides by outcropping shatter zone (Figs. 2B and 2C). The bases of the two northern patches are subhorizontal and occur at elevations both higher and lower than the adjacent inner boundary between shatter zone and Cadillac Mountain granite. The bases of the two southern patches dip moderately westward (Fig. 2C), indicating a wedge-like shape of the inner contact of the shatter zone against Cadillac Mountain granite.

Most of the eastern shatter zone segment can be divided into three distinct facies, here named shatter zone A through C (Fig. 2). Where in contact with Bar Harbor Formation, the outer facies, shatter zone A, consists of deformed and strongly fragmented Bar Harbor Formation invaded by thin irregular injections of aphanitic felsite. Initial felsite injections occur along sedimentary bedding planes, which are locally tightly folded and faulted (Fig. 3A), while later injections formed fractures that crosscut, deform, and/or brecciate the bedding (Fig. 3A). In some locations, initial injections along bedding have remobilized the beds and earlier felsite veins, resulting in complex contortions and folds (Fig. 3C). In contrast, where diabase forms the wall rock, shatter zone A is characterized by intense shattering; large angular blocks (1–5 m) of diabase are enclosed within and/or penetrated by much finer breccia of tightly packed fragments with 10%–15% felsite matrix (Fig. 3B). Locally, diabase and Bar Harbor Formation fragments are strongly deformed, forming a flow fabric that encloses larger blocks of massive diabase (Figs. 3D and 3E). Where both Bar Harbor Formation and diabase occur together along the outer edge of shatter zone A, their styles of deformation differ, reflecting a rheological contrast between the soft metasediments and stronger diabase. Diabase blocks are mostly angular and characterized by brittle fracturing, although some angular blocks of diabase were hot enough to deform and fold as the breccia settled downward (Fig. 3F). Aphanitic felsite commonly concentrates along the margins of these blocks. In contrast, Bar Harbor Formation shows pervasive veining and ductile deformation, and bedding planes are often wrapped around the stronger diabase fragments (Fig. 3E).

Shatter zone A grades over a few tens of meters to a central facies, shatter zone B, which is characterized by blocks of country rock that show marked variations in size, rounding, and degree of reaction or partial melting. The nature of shatter zone A changes from a clast-supported to dominantly matrix-supported facies across this transition, indicating both greater freedom of motion and higher temperatures due to the increased melt fraction. These blocks are hosted within ∼20%–60% fine- to medium-grained granitic matrix (Fig. 4). The blocks are dominantly Bar Harbor Formation (typically <1 m) and diabase (<1 m to tens of meters), along with scarce rhyolite, likely sourced from the overlying Cranberry Island volcanic series, locally <1–10 m (Figs. 2 and 4C). Diabase blocks in shatter zone B show highly variable degrees of reaction and digestion within the granite matrix; reaction intensity varies from prominent alteration rims up to 1 cm thick, with hornblende and biotite replacing clinopyroxene (20%–30% of blocks; Fig. 4D), to little or no reaction (Fig. 4F). Larger diabase blocks are typically rounded, and some are highly fractured (Fig. 4G), whereas small fragments range from angular to partly assimilated (Figs. 4B and 4C). The shatter zone B matrix is coarser than in shatter zone A and commonly contains many small fragments of partly digested or melted Bar Harbor Formation (Figs. 4A and 4C). Its character shows strong variations locally, ranging from felsite with scarce discrete country rock fragments (Fig. 4C) to felsite charged with partially melted Bar Harbor Formation (identified by a change in texture and mineralogy) and diabase fragments (Figs. 4E and 4F). Flow and compaction fabrics are common throughout the shatter zone B matrix, particularly beneath large blocks (Fig. 4B).

Shatter zone B typically grades over a few tens of meters to an inner, mostly matrix-supported facies, shatter zone C, which consists of significantly larger block sizes of diabase, Bar Harbor Formation, and rhyolite ranging from 10 to 80 m in dimension. These blocks are hosted in a medium-grained granitic matrix (50% to almost >95%), in which smaller fragments of country rock occur only locally. Two areas host large blocks of both pyroclastic and flow-banded rhyolite (Fig. 2). Inward, the shatter zone C matrix grades over a few tens of meters into typical hypersolvus Cadillac Mountain granite. Most exposures along the eastern margin are inland and covered by vegetation; the best are at Great Head (Fig. 2) and Otter Point (see Fig. 5C). Steep cylindrical schlieren occur locally in shatter zone C and within adjacent Cadillac Mountain granite. Cadillac Mountain granite within ∼100 m of the shatter zone typically contains scattered chilled intermediate enclaves (5–30 cm), which are otherwise scarce (Wiebe et al., 1997b).

Due to eastward tilting of the complex, the southern segment of the shatter zone (Fig. 5) provides a section along the shatter zone that varies in depth. The shallowest levels are exposed in the east at Otter Point, where the shatter zone cuts only Bar Harbor Formation and diabase dikes and sills. Gradually deeper levels are exposed westward, ending structurally ∼2–3 km below the eastern segment, approximately at the contact between Cadillac Mountain granite and the underlying gabbro-diorite unit. The shatter zone appears to thin out to the west, ending in two lenses of breccia within the Cadillac Mountain granite and away from the steep fault that elsewhere marks its outer margin (Fig. 5). Two good exposures of the fault against Bar Harbor Formation in the southern segment (at location 10–80 and south of location 10–90 in Fig. 5B) indicate a subvertical dip near the base of the outcrop, sharply cutting the gently dipping Bar Harbor Formation (Fig. 6B). The trace of the inner shatter zone boundary cutting across topography suggests that it dips northward beneath the Cadillac Mountain granite at 30°–40°.

Shatter zones B and C continue from the southern end of the eastern shatter zone segment across the eastern (structurally higher) half of the southern segment (Fig. 5A), although rounding and reaction of the diabase blocks are less pronounced. Shatter zone A is absent at the two southern exposures of the shatter zone (Fig. 5B), so it is uncertain how far shatter zone A extends along the south coast, because there are no exposures of this outer portion of the shatter zone between Great Head and location 10–80 (Fig. 5B).

The western half of the southern shatter zone segment (Fig. 5B) can be divided into two facies: an outer facies shatter zone D and an inner facies shatter zone E. Inward from the contact with country rock (Fig. 5B), shatter zone D is characterized by angular fragments of Bar Harbor Formation, diabase, and rhyolite (<1 to >10 m) hosted in >25% fine-grained granitic matrix, which generally lacks digested Bar Harbor Formation fragments and flow fabrics (Fig. 6B). Country rock blocks in shatter zone D are similar in size to those in shatter zone B of the eastern segment, but they typically lack the widespread reaction and rounding typical of shatter zone B.

The transition from shatter zone D to shatter zone E is gradational and well exposed along a N-S–trending coastal section (locations 10–90, 10–91, and 10–92; Fig. 5B). In this transition, angular blocks of Bar Harbor Formation and diabase are dominant, but rounded and reacted diabase blocks, which are typical of shatter zone B along the eastern segment, also occur along with rounded composite blocks (Fig. 6E) and blocks of highly deformed Bar Harbor Formation (Fig. 6F), which are characteristic of shatter zone A deformation in the eastern segment (see Fig. 3C). Shatter zone E is comparable to shatter zone C along the eastern segment, except that abundant smaller angular blocks also occur with blocks up to 80 m in size. As in the eastern segment, blocks of Bar Harbor Formation and diabase are still dominant in the south, but rhyolite blocks are more common and occur along the length of the southern segment in shatter zones C and E (Figs. 5C and 5D). Smaller blocks of Ellsworth schist and gabbroic blocks, resembling gabbro in the gabbro-diorite unit, are restricted to the westernmost 5 km of the shatter zone.

Whole-rock analyses were performed on Cadillac Mountain granite and shatter zone matrix samples to investigate the origin of the shatter zone matrix and its compositional variability. Major- and trace-element compositions were determined from analysis of fused glass disks by X-ray fluorescence spectroscopy (XRF) and laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS), respectively. Analytical precision was better than 5% relative (2σ) for most elements. Full details can be found in the Supplemental Material (Table S1: Analytical methods [^{footnote 1}]).

Zircon saturation temperatures (T_zirc) were calculated for all reported whole-rock compositions, using the revised zircon solubility model of Boehnke et al. (2013). All results are reported along with whole-rock geochemistry of Cadillac Mountain granite and shatter zone matrix in Tables S1 and S2. Inherited zircon cores were extremely scarce (David Hawkins, 2015, personal commun.), suggesting that these temperatures provide a minimum temperature estimate for the Cadillac Mountain granite magma prior to extensive crystallization (Miller et al., 2003).

Distinct zoning patterns in ternary feldspars of the shatter zone C matrix were analyzed by wavelength-dispersive spectroscopy with a Cameca SX-100 electron microprobe at the University of California–Davis, using an accelerating voltage of 15 kV, beam current of 20 nA, and spot diameter of 4 μm. Details of analytical conditions and uncertainties are reported in the Supplemental Material (Table S1: Analytical methods [^{footnote 1}]).

Typical Upper Cadillac Mountain granite is a homogeneous, massive, hypersolvus granite dominated by quartz and alkali feldspar (Fig. 7). Most major elements, as well as Ba, Sr, and Zr, are negatively correlated with SiO₂, which ranges from 71 to 75 wt% (Tables S1 and S2; Figs. 8–10). Lower Cadillac Mountain granite is petrographically similar to Upper Cadillac Mountain granite (Table 1) but distinguished by a population of hypersolvus feldspars with ternary feldspar cores and alkali feldspar mantles. Compositions range from 71 to 73 wt% SiO₂ and extend to slightly higher MgO, TiO₂, FeO^t, CaO, P₂O₅, Sr, and Zr and lower alkali contents than Upper Cadillac Mountain granite (Figs. 8–10).

The matrix of the shatter zone shows marked spatial variations in texture, mineralogy, and composition. Although some small-scale variations may reflect local contamination from assimilation of country rock blocks, systematic regional variations in shatter zone matrix characteristics are readily apparent and are interpreted as primary magmatic features. Based on these petrographic differences, we recognized four distinct types of shatter zone matrix, labeled I–IV for convenience (Table 1). Matrix types I, II, and III are chemically and petrographically distinct from Cadillac Mountain granite (Table 1; Figs. 8–10).

Two-feldspar biotite granite matrix (type I; Table 1) is characteristic of shatter zone A and outer shatter zone B in the eastern shatter zone segment and shatter zone D in the southern shatter zone segment. Compositions are scattered, ranging from 66 to 76 wt% SiO₂, commonly with higher TiO₂, Al₂O₃, MgO, P₂O₅, Sr, and Th, and lower K₂O, Ba, and Zr than typical Upper Cadillac Mountain granite (Tables S1 and S2; Figs. 8–10). Rare earth element (REE) contents also vary widely and extend to higher concentrations than all other samples (Fig. 10). Both positive and negative Eu anomalies occur (Eu/Eu* = 0.58 to 4.03). Fine-grained leucogranite matrix (type II; Table 1) occurs locally as decimeter- to meter-scale pods hosted within type I matrix, and it is enriched in SiO₂ (76–79 wt%) and depleted in TiO₂, FeO^t, and Zr relative to Cadillac Mountain granite. Alkalies, Ca, Sr, and Ba vary greatly, possibly reflecting hydrothermal alteration during cooling. The leucogranite matrix shows a much greater range in REE concentrations than Cadillac Mountain granite, and most samples have prominent negative Eu anomalies (Eu/Eu* = 0.08 to 1.14; Fig. 10).

The two-feldspar biotite granite (type I) matrix characteristic of shatter zones A and B gradually coarsens and is replaced by hypersolvus granite in shatter zone C (types III and IV; Table 1). Hypersolvus type III shatter zone C matrix along the eastern shatter zone segment is characteristically finer grained and more mafic than hypersolvus matrix (type IV) in the southern segment (Fig. 11). SiO₂ ranges from 65 to 73 wt%, with well-defined trends of decreasing TiO₂, Al₂O₃, FeO^t, MnO, CaO, Sr, and Eu (Tables S1 and S2; Figs. 8 and 9) with increasing SiO₂. At 73 wt% SiO₂, MgO in type III matrix averages ∼0.10 wt%, i.e., well below the Cadillac Mountain granite average (0.21 wt%), and increases only to 0.2 wt% at 65 wt% SiO₂. Ba, Sr, and Eu concentrations extend toward higher values than in typical Cadillac Mountain granite (Fig. 9), and most samples have well-defined positive Eu anomalies (Eu/Eu* = 0.88–2.15; Fig. 10). In contrast, type IV hypersolvus matrix along the southern shatter zone segment closely matches the compositions of typical Upper Cadillac Mountain granite (Figs. 8–10). Chondrite-normalized REE patterns are similar to but more variable than Cadillac Mountain granite, and all samples except one (with Eu/Eu* = 0.37) have only a slight Eu anomaly (Eu/Eu* = 0.79–1.19; Fig. 10).

Feldspars in type III matrix show a distinctive zonal sequence for more than 15 km along the eastern and southeastern shatter zone C (Fig. 12), at least as far as Otter Point (Fig. 5C) along the south coast. This zonal sequence consists of: (1) a homogeneous core of about An₁₀Ab₈₀Or₁₀ (where An-Ab-Or is anorthite-albite-orthoclase), (2) a middle section that is ∼0.5 mm in width with ∼10–15 oscillations (e.g., from Or₁₀ to Or₃₀), each averaging ∼30 μm in width, and (3) a nearly homogeneous rim of about An₃Ab₇₀Or₂₇ with variable width (up to a few tens of microns; Fig. 13; Table S3). Comparable zoning also occurs in some crystals in the outermost 100 m of the Cadillac Mountain granite along the eastern shatter zone segment. Figure 12 shows three representative examples of this zoning. Representative microprobe transects within a single crystal are shown in Figure 13. Ba concentrations are decoupled from Ab-Or zoning and are consistently below 0.34 wt%, averaging 0.28 wt% (Table S3).

Prior studies indicate that the Cadillac Mountain granite, and hence also the shatter zone in gradational contact with it, likely crystallized at shallow depth (P <2 kbar; Nichols and Wiebe, 1998; Wiebe et al., 1997a). Holtz et al. (1992) showed experimentally that the thermal minimum of the haplogranite system (Ab-Or-SiO₂-H₂O) at 2 kbar increases with decreasing a_H2O (from 690 °C at a_H2O = 1.0 to 840 °C at a_H2O = 0.5). Because Cadillac Mountain granite, shatter zone C, and shatter zone E matrix compositions plot near the minimum at pressures between 2 and 0.5 kbar, high zircon saturation temperatures (>900 °C; Table 1) thus imply that the Cadillac Mountain granite magma was relatively dry and H₂O-undersaturated. The presence of hornblende only as a minor interstitial phase in typical Cadillac Mountain granite and southern shatter zone C is consistent with low magmatic H₂O contents. The mineral assemblage (ternary feldspar, Fe-rich clinopyroxene, Fe-rich hornblende, and minor fayalite) and high T_zirc of the more mafic matrix in eastern shatter zone C also imply crystallization from a relatively dry magma.

The Cadillac Mountain shatter zone provides clear evidence for a major caldera-forming pyroclastic eruption preserved in the plutonic record. The continuous, arcuate geometry of this zone around the eastern and southern margins of the intrusion mirrors the geometry of shallower ring dikes and ring fault vent systems related to caldera-forming events elsewhere (Hildreth and Mahood, 1986; Miura, 1999), and we therefore interpret the shatter zone as a deeper, plutonic equivalent of these structures. In addition to the continuous arcuate geometry of the shatter zone, its width varies from several tens of meters to a kilometer or more, typical of ring fault systems mapped elsewhere (e.g., Ossipee; Kennedy and Stix, 2007). The pervasive evidence for fragmentation within this structure attests to the extreme violence of the processes involved. Such violence is exactly what should be expected in a ring fault system that is transporting fragmented magma and country rock both upward and downward, while at the same time acting as the principal zone of subsidence as the magma chamber roof drops down and the caldera is formed. Furthermore, a thick, 2.6 km sequence of Cranberry Island volcanic rocks, with very similar age to the Cadillac Mountain granite, is found several kilometers outboard of the southern shatter zone (Seaman et al., 2019). The thickness of this occurrence demonstrates that these volcanic rocks accumulated in an intracaldera environment between the ring fault (the shatter zone) and the topographic margin of the caldera. These various lines of evidence together indicate that the shatter zone was a ring fault linking a surface caldera (or calderas) to the underlying magmatic plumbing system.

We now examine each shatter zone facies and the overall structure of the shatter zone, demonstrating their link to explosive volcanic processes.

Shatter zone A records brittle fragmentation of country rock around the shallow margins of the Cadillac Mountain granite magma body in a style akin to explosive eruptive products (Fig. 3; Roy et al., 2012). We interpret this fragmentation as the result of decompression of the magmatic system during the onset of eruption and/or later opening of vents around a ring fracture (Druitt and Sparks, 1984; Palladino and Simei, 2005). It is likely that this “shattering” reflects not simply a single decompression event, but instead multiple recurring decompression and pressurization episodes, as occur in volcanic environments (Alidibirov and Dingwell, 1996). Indeed, they are expected during a caldera-forming eruption, as the first eruptive activity initially decompresses the system, followed by subsidence of the magma chamber’s roof, which alternately pressurizes and depressurizes the chamber (Kumagai et al., 2001). Such events are extremely efficient at generating pervasive fracture halos in rocks surrounding fluid-filled cavities in both brittle and ductile regimes (Kim et al., 2016; Velde et al., 1993) during compression and decompression (Alatorre-Ibargüengoitia et al., 2010; Kolzenburg et al., 2013). Such decompression-pressurization events also may be recorded in feldspar zonation throughout the shatter zone (see below). In the southern shatter zone segment, a lesser degree of deformation and veining of country rock along the ring fault relative to that seen along the eastern segment is consistent with increased depth and greater confining pressure to the west.

The deformation and fragmentation style for each type of country rock along shatter zone A reflect their differing thermomechanical properties: Diabase shows intense brittle fracturing with little or no rounding of fragments (e.g., Figs. 3B and 3D), while softer Bar Harbor Formation material is interspersed with irregular veins, and hence was injected with melt and possibly also partially melted, and mainly deformed plastically (e.g., Figs. 3A and 3C). The outer contact of shatter zone A is a steep, near-vertical ring fault, which probably acted as a conduit for eruptible Cadillac Mountain granite magma and fed ring fault vents at shallower levels. The only section where the outer contact appears to dip moderately inward is an ∼3 km length of the shallow eastern segment where Bar Harbor Formation alone is present at the contact (Fig. 2A). Theoretical models of ring fault formation predict changes in dip angle, and in some cases dip direction, at contacts between rock types with differing mechanical properties (Gudmundsson, 2007). Hence, we suggest that the local inward dip of the fault along this section reflects the weak mechanical properties of the Bar Harbor Formation and the absence of stronger diabase.

The nearly aphanitic grain size of felsic injections into wall rock at the outer edges of shatter zone A indicates rapid cooling against the cooler country rock at shallow depth. However, the veins are very finely dispersed in intricate patterns and display sharp boundaries against their host rocks (e.g., Fig. 3A). Due to its high viscosity, flow of granitic magma through such a narrow and intricate network would require very long time scales and/or extremely high temperatures, both of which would induce substantial reaction between melt and host rock. Aside from some degree of rounding of the fragments, we observed no indication of such reaction. Given the extremely high temperatures required to attain low melt viscosities, we consider it unlikely that these fracture fills were emplaced by magma flowing into the fractures. Instead, we suggest that the veins are composed of fragmented rhyolitic ash that was forced into the fractures as a gas-particle mixture, where it subsequently welded and crystallized. Similar welding of fragmented pyroclastic material is common in shallow volcanic systems (Kendrick et al., 2016; Kolzenburg et al., 2012; Paisley et al., 2019; Tuffen and Dingwell, 2004) and in dike complexes near plutons (Díaz-Bravo and Morán-Zenteno, 2011; Heiken et al., 1988) and has been linked to cyclical explosive behavior (Kolzenburg et al., 2019). While the emplacement temperatures of fragmented magma do not rule out postemplacement reaction between the rewelded melt and country rock fragments, particle rounding is most likely a result of attrition during emplacement of the gas-particle jet (Campbell et al., 2013).

Shatter zone B hosts country rock blocks that range widely in size, rounding, and degree of reaction/assimilation (Fig. 4). We interpret these as lithic fragments eroded from the unstable margins of a ring fault conduit/vent system during eruption and subsequently milled in a turbulent, fluidized gas-particle jet (Campbell et al., 2013; Jones and Russell, 2018). Some smaller blocks may have been erupted, but ones too large to be expelled collapsed downward, accumulating on crystal mush within the magma chamber and piling up against the brecciated chamber wall (shatter zone A) as the eruption waned, thus forming a wedge-shaped “magmatic debris slope,” which is preserved as shatter zone B (Fig. 2C). The lateral continuity of shatter zone B along the eastern and shallower parts of the southern segment (Fig. 2) implies that the ring fault vents were widely distributed around the complex (Díaz-Bravo and Morán-Zenteno, 2011; Hildreth and Mahood, 1986; Suzuki-Kamata et al., 1993). We interpret differences in the relative abundance of block types (Bar Harbor Formation, diabase, rhyolite, and Ellsworth schist) along the shatter zone, particularly between the eastern and southern segments, to reflect the country rock types forming the conduit walls, which varied with azimuth and depth. Similar azimuthal variations in lithic componentry are common in ignimbrite deposits and are inferred to reflect eruption from multiple vents around a ring fracture (Pittari et al., 2008; Suzuki-Kamata et al., 1993). Differences in the degree of rounding and reaction among adjacent blocks imply that the thermal, erosional, and transport histories of each block were variable; heavily rounded blocks were likely abraded for longer durations, whereas more angular blocks had shorter residence times and/or transport distances in the conduit (Campbell et al., 2013; Jones and Russell, 2018). Thermal erosion was particularly effective for blocks of Bar Harbor Formation, as shown by varying degrees of partial assimilation around their margins. Smaller angular fragments within the shatter zone B matrix may reflect autofragmentation of individual blocks (Roy et al., 2012), material spalled from the margins of the larger blocks (Campbell et al., 2013), and/or material locally derived from initial shattering of adjacent diabase wall rock.

Shatter zone D in the western half of the southern shatter zone segment is interpreted as a deeper equivalent of shatter zone B. The absence of intense shattering of diabase and deformation of Bar Harbor Formation material at the ring fault contrasts strongly with the shatter zone B along the shallow eastern shatter zone segment. The absence of intense rounding and the reaction of country rock blocks near the ring fault suggest that, at this depth, fractured blocks of country rock near the ring fault were derived locally from shattering of wall rock, which collapsed downward without experiencing milling in a conduit. This contrast with facies in the eastern shatter zone is consistent with higher confining pressure at depth. In the transition from shatter zone D to shatter zone E, the sporadic occurrence of rounded and reacted diabase blocks, composite blocks (Fig. 6E), and blocks of strongly deformed Bar Harbor Formation (Fig. 6F) that resemble milled blocks in shatter zone B clearly demonstrates that blocks were transported from vents near the surface to these deeper levels (2–3 km), collapsing onto a slope of angular breccia characteristic of shatter zone D.

We infer that the very large (up to 80 m) angular country rock blocks in the innermost shatter zone facies, shatter zone C and shatter zone E, were stoped from the base of the magma chamber roof, likely during caldera collapse. We therefore suggest that the formation of shatter zones C and E occurred after shatter zones A, B, and D had been emplaced. At this point, the blocks sank through rheologically weak upper levels of crystal mush at the base of eruptible magma in the chamber (Rooyakkers et al., 2018), eventually coming to rest where the strength of the mush was sufficient to support them (Bain et al., 2013; Wiebe et al., 2007). This sequence is supported by the internal structures of many blocks in shatter zones C and E, which document pervasive brittle deformation in the roof of the magma chamber. In Figure 6A, fine-grained fracture infill occurs both within a settled block as well as on its margin, and both are sharply cut by the host granitic matrix. This observation indicates that the roof disintegrated along preexisting fracture surfaces and provides evidence for least two brittle fracturing events: (1) the initial fracturing of the roof followed by infilling of the fractures by the fine-grained vein material (akin to those formed during syneruptive pressure oscillations in shatter zone A) and (2) subsequent exploitation of the preexisting fracture network to form the block that then settled through the Cadillac Mountain granite before coming to rest on strong mush. The latter process is documented by the continuation of the fine-grained vein infill along the outer boundary of the block.

Larger blocks would sink to deeper and stronger levels of the mush than smaller ones, and so the prevalence of very large blocks in shatter zones C and E may reflect a sorting process whereby smaller collapsed blocks were arrested at shallower (now eroded) levels of the intrusion. Because the western half of the intrusion is eroded, it is uncertain whether the country rock blocks in shatter zones C and E represent a narrow band of roof debris adjacent to the ring fracture, or the edge of a continuous sheet of large stoped blocks that extended across the entire intrusion beneath the collapsed roof. Stoped blocks comparable in size to those in shatter zone C can be traced for >15 km in the nearby, contemporaneous Vinalhaven intrusion and were inferred to record a similar process of chamber roof collapse (Hawkins and Wiebe, 2004; Wiebe and Hawkins, 2015).

The two-feldspar biotite granite matrix (type I; Table 1) is characteristic of eastern (shallow depth) exposures of shatter zones A and B, and it also occurs over a wider band at deeper structural levels in the southern segment in shatter zone D. The presence of this matrix at deep structural levels, alongside rather than above the Cadillac Mountain granite, indicates that type I matrix does not represent the remnants of an eruptible cap of crystal-poor magma extracted from Cadillac Mountain granite mush prior to eruption (Bachmann and Bergantz, 2004; Hildreth, 2004). Compositions of the type I matrix also rule out such a scenario, because most samples are less evolved than Cadillac Mountain granite, containing higher MgO and Sr contents and extending to lower K₂O and SiO₂. Instead, the increased width of type I matrix with depth implies a deeper origin. We thus interpret type I matrix as a fresh (likely syneruptive) injection of granitic melt from chambers at depth into the shatter zone. Transport of this magma from depth could have been the result of the pressure differential between the magmatic reservoir and decompressed regions below active vents. The intense syneruptive brittle fracturing events in the marginal areas of the shatter zone against country rock described above would greatly increase permeability through this zone, making it a preferential pathway for melt migration. We suggest that this high permeability allowed evolved, high-temperature magma from a deeper reservoir to ascend rapidly through the shatter zone, toward low-pressure zones beneath active ring fault/dike vents. The large compositional scatter of type I matrix likely reflects varying local contamination of this magma by assimilated country rock. Pockets of highly silicic leucogranite with strong negative Eu anomalies (type II matrix) appear to represent filter-pressed concentrations of interstitial liquid from type I matrix.

The extremely high Fe/(Fe + Mg) values of the eastern shatter zone C (type III) matrix indicate extensive fractionation of mafic silicates. The likely source for this more mafic type III matrix is fractionated hybrid magma from beneath the Cadillac Mountain granite and above mafic replenishments of the gabbro-diorite unit (Wiebe et al., 1997b). The strong positive Eu anomalies of this matrix in shatter zone C are typical of feldspar cumulates, suggesting that much of the melt carrying these crystals erupted to the surface, leaving a concentration of crystals. We infer that this hybrid magma ascended rapidly along the chamber’s eastern margin to compensate for volume loss as a felsic cap above the Cadillac Mountain granite mush was erupted, likely driven toward areas of low pressure beneath open vents. This type III matrix is absent along the southern segment of the shatter zone, suggesting that the upwelling was localized to the eastern margin of the magma chamber. Quenched intermediate enclaves with high Fe/Mg are common within the transition from shatter zone C matrix to homogeneous Cadillac Mountain granite along the eastern shatter zone segment (Wiebe et al., 1997b), supporting this connection to the underlying hybrid magmas. Similar syneruptive ascent of deeper-derived, more mafic (often hybrid) magma has been recently identified in several caldera-forming silicic eruptions (Allan et al., 2017; Rooyakkers et al., 2018; Matsumoto et al., 2018).

Although oscillations in compatible element concentrations (e.g., Ba) have been described in alkali feldspars and related to recharge processes (Gagnevin et al., 2005; Moore and Sisson, 2008), sequences of strong oscillatory Ab-Or zones in alkali feldspar are rare (Abart et al., 2009). Ab-Or zoning in ternary feldspar is most prominent in the type III matrix of shatter zone C along the eastern shatter zone segment. The wide spatial distribution of these crystals over an area extending >12 km along shatter zone C requires that their characteristic zoning pattern was produced by a process that operated on a regional scale and simultaneously affected a large portion of the magmatic system.

We suggest that the oscillatory zoning in the alkali feldspars records pressure oscillations (and potentially also concurrent variations in melt water content) associated with eruption(s) from the Cadillac Mountain granite magma chamber. Thus, the observed 10–15 oscillations may represent as many pressure cycles. Since the melt hosting the zoned alkali feldspars was hot and relatively dry (see above), pressure reduction would cause the Ab-rich loop of the alkali-feldspar phase diagram to shift to higher temperatures (Holtz et al., 1992), causing increased undercooling and initiating crystallization while shifting the equilibrium feldspar composition to higher Or values (Fig. 14). Pressure fluctuations during crystallization would thus cause oscillations in the Ab-Or composition of the crystallizing feldspars similar to those we observed here (Figs. 12 and 13). Oscillatory zoning near the rims of plagioclase phenocrysts from Shiveluch volcano (Kamchatka) was observed by Humphreys et al. (2006) and similarly interpreted to record magmatic pressure oscillations related to eruptions.

Large eruptions are commonly preceded by smaller precursory magmatic, phreatomagmatic, and/or phreatic events (e.g., precursory activity months to years prior to caldera-forming events at Askja, Mazama, Taupo, Yellowstone, and Santorini; Sigvaldason, 1979; Bacon, 1983; Wilson, 2001; Allan et al., 2012; Myers et al., 2016; Simmons et al., 2016). These events generate a series of decompression events within the underlying magma chamber; such a sequence of similar precursory events could have caused the pressure oscillations recorded by the feldspars. Alternatively, the pressure oscillations could relate to a single caldera-forming event. Unlike in small eruptions, where the chamber pressure decreases exponentially through time (Hreinsdóttir et al., 2014), we envisage several scenarios that, individually or in combination with each other, could cause chamber pressure to oscillate during a large caldera-forming event: (1) Decompression of the upper parts of a magmatic system could destabilize deeper magma and/or deeper magma bodies (Sparks and Cashman, 2017), triggering recharge events that partially restore the chamber pressure (Blake, 1981); (2) caldera roof block(s) could undergo stick-slip subsidence (Kumagai et al., 2001; Stix and Kobayashi, 2008), repeatedly restoring lithostatic pressure to the top of the chamber (Folch and Martí, 2009; Martí et al., 2000); and/or (3) chamber pressure could be partially restored during temporary blockage of vents, possibly during early eruptive phases prior to the establishment of an open and interconnected conduit system (Myers et al., 2016), followed by pressure release during opening of new vents.

Several geometric relationships within the shatter zone hint toward a possible trapdoor style of caldera collapse during the shatter zone–forming eruption. Pinching out of the shatter zone to the west in the southern segment may reflect a trapdoor collapse with a hinge on the western side of the complex. This possibility is consistent with the more dynamic nature of shatter zone facies, particularly shatter zone B, and evidence for upwelling of hotter and more mafic type III matrix magma in the eastern segment, which may reflect more intense venting and a more open conduit/vent system on that side of the complex, opposite to the inferred hinge. However, differences in shatter zone facies and geometry between the eastern and southern segments also reflect differences in depth of exposure, and disentangling the effects of depth and collapse geometry is not possible with present-day exposure.

The four caps of Cadillac Mountain granite resting structurally on top of the shatter zone (Fig. 2) must have solidified there after eruptions from the ring vents ceased. We attribute the presence of Cadillac Mountain granite above the shatter zone as evidence for upward and lateral flow of weak Cadillac Mountain granite crystal mush that was squeezed toward and into the inactive vents along the ring fault by the collapsing caldera roof block. A similar process of collapse-driven flow of crystal mush was recently described at the Altenberg-Teplice caldera in the Bohemian Massif (Tomek et al., 2018). Together, these examples suggest that the downward pressure exerted by subsiding caldera blocks can play a major role in mobilizing otherwise sluggish and noneruptible crystal mush.

Model for Caldera Development Associated with the Cadillac Mountain Granite Complex

The evidence presented here allowed us to reconstruct the following sequence of events that occurred within the Cadillac Mountain granite magmatic system and led to the formation of the shatter zone as it is now preserved (Fig. 15).

• (1) Shatter zone A records the initial explosion and shattering of the country rock beneath erupting vents along the ring fault of a caldera system. This interpretation is supported by the fragment size distributions, which imply explosive shattering (Roy et al., 2012). Similar explosive brecciation linked with a major eruption was previously described at the Ossipee ring dike by Kennedy and Stix (2007). The pervasive, fine-grained matrix infilling the intricate fracture networks in the country rock and clasts within shatter zone A likely consisted of fragmental material that was forced into the fractures in a gas-particle jet at high temperatures and retained sufficient heat to weld and crystallize within the fractures.

• (2) Shatter zone B records dynamic processes within erupting ring fault vents at the highest exposed structural levels. The blocks hosted in this facies were incorporated from the conduit walls and milled in a gas-particle jet within the conduit, before collapsing down the conduit onto crystal mush in the underlying chamber and forming a lithic debris pile that was buttressed against the fractured country rock comprising shatter zone A. Variable degrees of rounding and fracturing of the blocks reflect varying residence times (and hence varying degrees of attrition and rounding) within the conduit, as well as differing degrees of reaction with their host matrix. The deeper exposures of the ring fault along the south coast lack an “explosive” facies (i.e., facies shatter zone A) at the fault, and breccias near the fault (shatter zone D) lack rounded and reacted diabase blocks and partially melted Bar Harbor Formation blocks, indicating the blocks close to the fault were not transported up into the vents, but collapsed downward.

• (3) Shatter zone C and shatter zone E record the collapse and fracturing of the chamber roof, and subsequent settling of large blocks derived from the roof onto Cadillac Mountain granite crystal mush.

• (4) Mineralogical variations in matrix across the shallow eastern segment of the shatter zone document an inward increase of temperature from shatter zone A to shatter zone C. The two-feldspar biotite granite matrix (type I) characteristic of shatter zone A and much of shatter zone B consists of cool hydrous felsic magma that was drawn up through the shatter zone. The more mafic hypersolvus granitic matrix (type III) characteristic of shatter zone C and inner shatter zone B reflects syneruptive upwelling of hot hybrid magma derived from below the Cadillac Mountain granite crystal mush and above the gabbro-diorite unit.

• (5) Strong positive Eu anomalies in the type III matrix typical of shatter zone C and inner shatter zone B of the eastern segment suggest that much of the felsic liquid component of this hybrid magma was lost to eruption. We interpret oscillatory Na-K zoning in euhedral hypersolvus feldspars in this matrix to record a sequence of pressure fluctuations related to eruptive activity.

• (6) Mineralogical variations in matrix across the deeper southern segment of the shatter zone similarly document an increase in temperature of the shatter zone matrix from the ring fault to the Cadillac Mountain granite. Shatter zone D has a matrix of two-feldspar biotite granite (type I) that grades to hypersolvus granite in shatter zone E (type IV). The great extent of the type I matrix at these deep structural levels of the shatter zone indicates that this matrix must have been sourced from a deeper silicic body or bodies and was drawn up syneruptively through the permeable shatter zone.

• (7) The width of the shatter zone narrows westward along the southern margin. This may suggest a trapdoor-type collapse geometry with a hinge in the west. The absence of the type III hybrid matrix along the deeper shatter zone exposures along the southern segment implies that upwelling of this hybrid magma was localized to the eastern part of the complex, possibly also reflecting a trapdoor-style collapse.

The structure, facies, and matrix chemistry of the shatter zone provide direct evidence for formation of the shatter zone during a major caldera-forming eruption, thus implying that the Cadillac Mountain granite reflects the remnants of a large subcaldera crystal mush zone that fed explosive eruptions and was subsequently preserved as a large granitic pluton. Excellent preservation of the shatter zone and exposure across a range of structural depths allowed us to establish clear links between the plutonic record and explosive volcanic processes through field relationships and feldspar zoning, which are independent of the limitations of the isotopic dating methods generally used to study the plutonic-volcanic connection. While generally supportive of this reconstruction, the available radiometric dates proved insufficient to constrain the intimate connection between the granites on Mount Desert Island and the derived Cranberry Island volcanic series that is evident in the outstanding field relations preserved in the shatter zone. This direct link between plutonism and volcanism demonstrated by the shatter zone supports an origin for some granite plutons as the cumulate roots of large volcanic systems, and it demonstrates that the energetic and dynamic conditions experienced in the shallow vent and conduit system in major eruptions may also be felt and recorded along the deeper margins of large magmatic systems.

Funding was provided in part by National Science Foundation award EAR-0536655, to R.A. Wiebe. S. Kolzenburg acknowledges support from H2020 Marie Skłodowska-Curie Fellowship DYNAVOLC no. 795044. Wiebe acknowledges the Keck Foundation, through the Keck Geology Consortium, for funding undergraduate field projects in 1993, including two to the shatter zone by Michelle Coombs and Jonathan Holden. Over several years, Wiebe greatly benefited from conversations in the field with David Hawkins, Don Snyder, John Eichelberger, Sam Roy, and Duane Braun. We benefitted from helpful reviews by Calvin Miller, Eric Christiansen and two anonymous reviewers. We appreciated the editorial help of Shanaka de Silva and Guilherme Gualda.