Making sense of brittle deformation in rhyolitic lavas: Insights from Obsidian Dome, California, USA

Extrusive silicic rocks are common throughout the geologic record from a range of tectonic environments and are important hosts of epithermal mineralization. Whereas lava domes are ubiquitous at intermediate and silicic volcanoes (e.g., Mount St. Helens, Washington, USA, 1980–1986, 2004–2008) and have been the subject of numerous studies (e.g., Fink et al., 1990; Sparks et al., 2000; Pallister et al., 2013), direct recorded observations of active rhyolitic lava flows did not occur until two Chilean eruptions, Chaitén and Cordón Caulle, in 2008 and 2011–2012, respectively (Lara, 2009; Schipper et al., 2013). Lava flows occur on the flanks of volcanoes (e.g., Medicine Lake volcano, California, USA) where the lava is not restricted to a deep crater, and when sufficient magma supply is available, the flow develops into a unidirectional lava flow (a “coulee”; e.g., the Chao dacite, Chile [de Silva et al., 1994]; Big Obsidian Flow, Newberry volcano, Oregon, USA [Donnelly-Nolan et al., 2011]). They also form on low-relief caldera floors where they can spread radially (e.g., Deadman Dome, California, USA; Sampson and Cameron, 1987).

Unlike intermediate-composition lava domes, where there are several active worldwide at any one time, most of what is known about the dynamics of rhyolitic and rhyodacitic lava flows must be inferred from prehistoric examples. This paper documents a field-based study at Obsidian Dome, California, USA (Fig. 1), to investigate the emplacement of silicic lavas through the lens of brittle deformation of the upper surface and margins. We attempted to constrain the relative timings, mechanisms, and conditions under which deformation occurred through a combination of textural and structural observations. Specifically, this study examined how different fracture types and sizes relate to one another, and whether they formed through a single continuous deformation process, or if they represent different processes operating at different times. In so doing, we documented evidence of synchronous localized explosive processes intimately linked to the deformation.

Herein, we use the term “lava” for prehistoric examples and reserve “lava flows” for active flows. We do this (1) to avoid confusion between active and prehistoric extrusions, and (2) because it can be difficult to distinguish lava domes and lava flows in the geologic record, but both are coherent lavas.

The 2011–2012 eruption of Cordón Caulle, Chile, was the first large-volume, crystal-poor (i.e., obsidian), rhyolite lava emplacement observed throughout the entire eruptive event. It is important because (1) it featured a wide range of emplacement processes, including endogenous and exogenous (Farquharson et al., 2015; Magnall et al., 2017) and advance-limiting modes (Castruccio et al., 2013) acting contemporaneously in different parts of the lava flow, and (2) the advance continued for ~8 mo after new magma ceased being erupted from the vent (Tuffen et al., 2013). Additionally, contemporaneous explosive eruptions from the advancing lava accompanied effusion (“hybrid activity”; Schipper et al., 2013; Castro et al., 2014; Castro and Walter, 2021). Hybrid activity is attributed to outgassing through fracture networks (tuffisites) within and above the shallow conduit during the lava’s effusion (Castro et al., 2014; Wadsworth et al., 2020). These observations challenge the theory (“permeable foam model”) that rhyolitic lavas are degassed upon effusion, and the transition into a dense glassy lava occurs by collapse of the bubble and pore space (permeable network; Eichelberger et al., 1986). Experimental studies have demonstrated that even high-porosity foams cannot support effective permeability without fracturing (Ryan et al., 2019). The outgassing pathways are sustainable until viscous relaxation closes the fracture pathways (Heap et al., 2019; Unwin et al., 2021).

Holoceneage obsidian lavas in California and Oregon, USA (e.g., Fink, 1983; Fink and Anderson, 2017), are among the most comprehensively studied lavas worldwide; additional important late Pleistocene and Holocene examples exist in the Aeolian Islands (Italy), the Chilean Andes, Iceland, Japan, and New Zealand (e.g., Stevenson et al., 1994; Maeno and Taniguchi, 2006; Lara, 2009; Tuffen and Castro, 2009; Pallister et al., 2013; Tuffen et al., 2013; Bullock et al., 2018). Investigations of these lavas have yielded important insights into silicic lava emplacement mechanisms through detailed studies of (1) morphology (e.g., Fink, 1980a; Ramsey and Fink, 1999; Deardorff et al., 2019; Leggett et al., 2020), (2) lithology and structure (e.g., Manley and Fink, 1987; Smith and Houston, 1994; Smith, 2002), and (3) microstructure and petrofabrics (e.g., Castro et al., 2002; Cañón-Tapia and Castro, 2004; Rust et al., 2003; Manga et al., 2018). Observations and quantitative data provide inspiration and constraints for many informative numerical and analog simulations (e.g., Fink and Griffiths, 1992, 1998; Merle, 1998; Lescinsky and Merle, 2005; Farrell et al., 2018; Kenderes, 2021).

Despite the fact that most silicic lavas have only a simple, single chemical composition, they typically display two or more distinct lithofacies, that is, physically and texturally distinctive rock types. Although the lithofacies naming schemes vary between studies (Fink, 1980a, 1983; Manley and Fink, 1987; Stevenson et al., 1994; Maeno and Taniguchi, 2006), the typical lithostratigraphic units identified are: (1) microcrystalline rhyolite (can be devitrified or spherulitic; bulk density ~2100 kg/m³); (2) finely vesicular pumice; (3) coarsely vesicular pumice; and (4) avesicular obsidian (bulk density >2000 kg/m³).

The seminal study of the Big and Little Glass Mountain lavas (ca. 1.2 ka) at Medicine Lake volcano, California, USA, by Fink (1983) described three lithofacies (finely vesicular pumice, coarsely vesicular pumice, and avesicular obsidian) on the upper surfaces and at the distal margins. Furthermore, the extensive talus at the margins was inferred to obscure a laterally continuous layer of avesicular obsidian and coarsely vesicular pumice, underlain by a basal breccia. Scientific drilling at Obsidian Dome (Eichelberger et al., 1984, 1985) and the ca. 130 ka Banco Bonito lava at Valles Caldera, New Mexico, USA, revealed the same general lithostratigraphy (Fink and Manley, 1987; Manley and Fink, 1987; Fig. 2). This stratigraphy was inferred to be laterally continuous within Obsidian Dome, where it was correlated between two boreholes near the vent and the margin (Fink and Manley, 1987; Manley and Fink, 1987; Fig. 2). Several drill cores from the ca. 51 ka Takanoobane rhyolite lava of SW Japan (Furukawa and Uno, 2015) revealed alternating layers of coarsely vesicular pumice and obsidian in the uppermost 20 m, with the AVL2 core resembling the observed stratigraphy of the RDO-2A core at Obsidian Dome (Fig. 2).

Similar lithofacies have been identified at silicic lavas in New Zealand (Ben Lomond rhyolite, ca. 100 ka; Stevenson et al., 1994) and at Showa Iwo-jima, Japan (88 yr old; Maeno and Taniguchi, 2006), where the presence of a basal breccia overlain by a rhyolite core grading upward into obsidian is ubiquitous. Importantly, although Maeno and Taniguchi (2006) described extensive coarsely vesicular pumice at the upper surface of Showa Iwo-jima, nowhere has laterally continuous coarsely vesicular pumice at depth been demonstrated conclusively (cf. Fink, 1983), and coarsely vesicular pumice is absent from many lavas, including the Ben Lomond lava (Fig. 2; Stevenson et al., 1994) and lavas in the Landmannalaugar area, Iceland (e.g., Wilson et al., 2007).

Features indicative of hybrid explosive-effusive activity identified at Chaitén and Cordón Caulle, Chile, such as expansive pyroclastic ridges, tuffisite veins, and glassy to pumiceous ballistic bombs, have also been described at Big Glass Mountain, California, USA (Castro and Walter, 2021). There, obsidian pyroclasts display similar hydrous geochemical signatures to pyroclasts from the hybrid Chaitén eruption.

The upper surfaces of silicic lavas are dominated by a carapace of superficially erratically placed, angular boulders of avesicular obsidian and finely vesicular pumice (e.g., Anderson et al., 1998; Plaut et al., 2004; Leggett et al., 2020). Immediately beneath the boulders, the surface of the coherent lava is composed of in situ finely vesicular pumice and avesicular obsidian that are ubiquitously disrupted by tensile fractures, the largest of which provide windows into the lava’s interior. Crease structures (Anderson and Fink, 1992) are large, splayed-open fractures penetrating deep into the lava, interpreted to form during lateral spreading. Although better exposed elsewhere (e.g., Medicine Lake volcano), several crease structures are present at Obsidian Dome, including ones that have been partially quarried.

Superimposed on this foundation, there are ~15–20-m-amplitude ridges and troughs that are regularly spaced, laterally continuous, arcuate, and usually parallel to the nearest flow margin and perpendicular to the inferred flow direction (Fink, 1980b). Ridges are antiformal and expose “deeper” lithofacies like avesicular obsidian under a thin or nonexistent carapace of finely vesicular pumice boulders. Troughs are synformal and are filled by the thick finely vesicular pumice boulder carapace. The ridges and troughs are termed ogives and are commonly interpreted as buckle-style folds (Fink, 1980a, 1980b, 1983; Anderson and Fink, 1992; Fink and Anderson, 2017); however, Andrews et al. (2021) suggested that ogives form by dilation of decimeter-scale circumferential fracture sets that form horst (ridge) and graben (trough) morphologies.

Herein, we use “brittle-ductile” to refer to deformation processes that initiated in the brittle regime (i.e., exceeding the tensile strength of the rock rapidly enough to induce fracture) and then experienced ductile flow (i.e., plastic flow where melt viscosity is the dominant physical control on rheology). The brittle-ductile transition is a dynamic boundary separating the brittle and ductile regimes, and in crystal-poor obsidian flows, it is closely related to the glass transition (T_g), which is sensitive to temperature, composition, and strain rate (Dingwell, 1996). Ductile behavior in lava is favored by higher temperatures, higher dissolved water contents, and lower strain rates. Near T_g, lava may initially deform in a brittle fashion but then undergo relaxation and viscous flow over a longer time scale. Understanding this spectrum of processes from ductile through to brittle deformation regimes requires close observation of the results of the deformation (e.g., fracture surfaces).

Extensive research has been conducted on the origins and implications of flow banding (changes in crystallinity, vesicularity, or fragmentation and annealing of tuffisite) in silicic lavas, including those at Obsidian Dome (e.g., Castro et al., 2005; Gonnermann and Manga, 2005; Tuffen et al., 2013). However, for the purpose of this research, and in the absence of evidence to the contrary, we assumed that all flow bands were developed before or during effusion at the vent and were passive layers (i.e., weak) that faithfully recorded ductile deformation during lateral emplacement. These flow bands are typically steep at the upper surface and subhorizontal elsewhere (Smith, 1996). Reconstruction of cross sections through several rhyolite lavas indicated that originally subhorizontal flow banding became steepened during fracturing and tilting (Andrews et al., 2021). Subhorizontal flow banding from outcrops (e.g., Smith, 1996) and analog experiments (e.g., Merle, 1998) is interpreted to result from ductile, gravitational loading (where the maximum principal stress [σ₁] is vertical) and subhorizontal (i.e., parallel to ground surface) radial spreading (σ₂ and σ₃) where σ₁ > σ₂ = σ₃. This is coaxial or “pure” shear strain. Coaxial, ductile deformation in obsidian is also recorded in the preferred orientations of microlite crystals (Castro et al., 2002; Castro and Mercer, 2004; Manga et al., 2018), vesicles (Rust et al., 2003), and magnetic fabrics (Cañón-Tapia and Castro, 2004).

Taken together, the structures observed at silicic lavas record a prolonged history of extensional, coaxial deformation: brittle and brittle-ductile at the upper surface, and ductile in the interior. However, the body of existing literature is heavily weighted toward inferences about the entire flow’s behavior based on its large-scale morphology (e.g., Fink and Griffiths, 1998; Cioni and Funedda, 2005; Bullock et al., 2018; Farrell et al., 2018; Leggett et al., 2020), with few studies directed at the upper surfaces. Therefore, this study focused exclusively on the architecture and structural geology of the upper surface at Obsidian Dome, California, USA. We examined the possible continuum of deformation recorded in the tensile fractures ubiquitous in the silicic lavas and fracture-facilitated outgassing throughout the entirety of lava emplacement.

Obsidian Dome is part of the Holocene Inyo volcanic chain nestled on the eastern side of the Sierra Nevada Mountains, eastern California, USA, in the Long Valley volcanic region. The dome sits ~1.5 km north of the Pleistocene Long Valley caldera margin (Fig. 1). The Inyo chain includes silicic lavas and phreatic craters above a N-S–trending dike zone (Bailey et al., 1976; Fink, 1985; Fink and Pollard, 1985; Miller, 1985; Reches and Fink, 1988). Obsidian Dome is one of the three youngest domes (~650–550 yr old; Wood, 1977; Miller, 1985), along with Glass Creek dome and Deadman Dome, and all three are coeval with extensive pumice and ash tephra deposits that blanket the surrounding land-scape. The ~600-yr-old lavas and their feeder dike share different proportions of two geochemically distinct rhyolitic magmas (Sampson and Cameron, 1987). At Obsidian Dome, the low-silica (~70 wt%), microcrystalline end member is restricted to the central peak region over the inferred vent, whereas the high-silica (73.5 wt%) end member is ubiquitous at the margins (Vogel et al., 1989). Whether or not these geochemical differences are sufficient to have influenced the emplacement of the lava is uncertain.

Obsidian Dome is a pancake-shaped silicic lava, covering an area of ~2 km², and it has an estimated eruptive volume of 0.1 km³ (Eichelberger et al., 1986). The dome slopes down very gradually to the east and south and has a central peak reaching ~2608 m above sea level. The thickness of the flow margins is variable between a maximum thickness of ~60 m in the northeast, where the lava flowed into a paleovalley, and ~20 m along the southeastern margin.

The exceptionally rough surfaces of silicic lavas (Plaut et al., 2004) make them slow and hazardous to traverse, and they lack enough relief or distinctive features to support navigation using the available 1:24,000 scale topographic maps or 1-m-resolution light detection and ranging (LiDAR)– derived digital terrain models (DTMs). Therefore, we developed a workflow to produce color, orthorectified air photos and accompanying DTMs of sites of interest using a small unoccupied aerial vehicle (sUAV) and then digitally mapped structural features onto those base images.

We collected global positioning system (GPS)– located, overlapping, planview photographs along raster flight paths with a DJI Phantom 4 Advanced sUAS, following the approach of Leggett et al. (2020). This is a consumer-grade sUAV with a 20 megapixel, color, gimbal-mounted camera, and it is controlled from an Apple iPad 4 mini. Flight paths were planned and optimized for photography using Dronedeploy (Drone Deploy, 2020) and Drones Made Easy’s Map Pilot applications for Apple iOS, usually with fixed flight altitudes between 30 m and 50 m above the ground and flight paths designed to provide 85% overlap between adjacent images along and across flight lines. Flights were typically conducted in the late morning to take advantage of lower wind speeds and to maintain similar bright, high-angle lighting between missions.

DTMs and orthorectified air photos were constructed by structure-from-motion photogrammetry (e.g., Westoby et al., 2012) in Agisoft Metashape v.1.5 (Agisoft LLC, 2019) following the approach of Leggett et al. (2020). The dense point clouds for each site returned root-mean-square errors of ±0.1 m and generated DTMs with ~5 cm resolution. The resulting orthorectified air photos had resolutions of ~2–5 cm²/pixel.

We mapped data onto ruggedized iPad 4 mini tablets with Midland Valley Exploration’s FieldMOVE (Petroleum Experts, 2020) installed. Additional structural measurements were collected using cell phones (Apple iPhone 6 and 7; Google Android-powered LG X) with FieldMOVE Clino installed. Most software systems using GPS location produce horizontal errors of 5–10 m, but this was mitigated by manually correcting the measurement location against the high-resolution orthorectified base image (e.g., Allmendinger et al., 2017). Typically, iPhone smartphones running FieldMOVE Clino produce better data than Android-powered units (e.g., Allmendinger et al., 2017; Novakova and Pavlis, 2017, 2019; Trede et al., 2019). Comparisons of FieldMOVE (iPad 2) and FieldMOVE Clino (iPhones 7 and 8) against hand-held compass-clinometers identified a mean error of ±0.25° in dip angle (Hama et al., 2014) and <10°–23° in strike azimuth (Hama et al., 2014; Allmendinger et al., 2017); the azimuth value is known to be very susceptible to disturbance of the smartphone’s built- in magnetometer (e.g., by power lines, batteries, etc.). We recalibrated the iPad 4 minis and iPhones regularly, occasionally confirmed strike and dip measurements with handheld compass-clinometers, and manually adjusted measurement locations to the base image.

Every structural measurement, sample, and image were recorded with spatial metadata. Using Google Earth, all the .kmz files from each FieldMOVE project were initially assessed for accuracy of locations and large-scale structural patterns. The .kmz files contained all structural information, locations, and orientations of all field photos taken, making them a powerful tool in Google Earth. Original plane.csv files from FieldMOVE with all planar structural measurements and locations were imported into QGIS 3.10 for further analyses using the orthorectified base images. Data and orthorectified images were combined to create local structural maps across Obsidian Dome. Each structural map was accompanied by several stereonets displaying the attitude of the different planar features and lineations measured. All strike and dip information was compiled and analyzed using Stereonet 11 (Allmendinger et al., 2011).

We use the term “lithofacies” to refer to different textural forms and emphasize that different lithofacies readily grade into one another. Three volumetrically significant rhyolite lithofacies are recognized at Obsidian Dome. By different proportions and sizes of vesicles, they are: avesicular obsidian, finely vesicular pumice, and coarsely vesicular pumice (Fink, 1983). A fourth microcrystalline rhyolite lithofacies is present directly over the vent and at depth recorded in the drill cores. It appears to be restricted to discrete flow bands within obsidian and is associated with neither finely vesicular pumice nor coarsely vesicular pumice.

The avesicular obsidian is defined as variably flow-banded, black, 95% glass groundmass with small (≤2-mm-diameter) albite and sanidine phenocrysts, sparse clinopyroxene, amphibole, and Fe-Ti–oxide phenocrysts (Vogel et al., 1989), and sparse quartz spherulites. It is the same as the “dense obsidian” described by Manley and Fink (1987) from the Obsidian Dome drill cores as well as outcrops at many other silicic lavas (e.g., Fig. 2). The obsidian is typically avesicular with measured porosities of ~1%–4%, with ~0.15–0.22 wt% H₂O and bulk density of ~2300 kg/m³ (Andrews et al., 2021). Where it is transitional with coarsely vesicular pumice, or in low-strain zones (e.g., fold hinges, dilational shear bands), large (≥1-mm-diameter), isolated vesicles form a scoriaceous texture (Fig. 3A). Scoriaceous obsidian is typically very strongly flow banded with an abundance of ductile and brittle deformation structures, including: intrafolial isoclinal recumbent folds, boudins of avesicular obsidian and microcrystalline rhyolite (e.g., Fig. 3C), rotated crystals, faults, and tension gashes.

Obsidian is best exposed at Obsidian Dome around the margins of the lava, in surrounding talus, and in small, discreet masses (≤10 m across) on the upper surface. It is typically massive at the margins and forms semicontinuous horizons separated by lenses of autobreccia (Figs. 4A–4B). Obsidian on the upper surface grades laterally and upward into the extensive carapace of finely vesicular pumice, where faint flow banding becomes more pronounced, and the rock becomes lighter gray in color.

The upper surface of Obsidian Dome is dominated by decameter-scale angular blocks, boulders, pinnacles, and randomly rotated slabs of finely vesicular pumice (Fig. 3D). Finely vesicular pumice is light beige to dark gray in color, is flow banded, and contains ubiquitous small (≤0.5-mm-diameter), spherical vesicles (Fig. 3E). The measured total porosities of finely vesicular pumice are ~30%–40%, with ~0.23 wt% H₂O, and its bulk density is ~1670–1750 kg/m³ (Andrews et al., 2021).

Coarsely vesicular pumice has >40% vesicles and is a dark olive-green color with large (0.1–1-cm-diameter), often stretched and interconnected, vesicles (Fig. 3B; Fink, 1983; Manley and Fink, 1987; Fink et al., 1992). Coarsely vesicular pumice has a total porosity of 20%–80%, with ~0.14 wt% H₂O, and a bulk density of ≤1000 kg/m³ (Andrews et al., 2021). The coarsely vesicular pumice makes up ~1% of the surface of Obsidian Dome and is concentrated around the WSW and NW margins of the dome as positive topographic features around crease structures (Fig. 3D).

We did not identify laterally continuous coarsely vesicular pumice layers at the flow margins of Obsidian Dome as proposed in Fink (1983) and related models, regardless of the lava thickness. Instead, the observed lithostratigraphy at the flow margins of Obsidian Dome is, from the base upward, talus obscuring the bottom ~10–30 m of the front, followed by prominent, massive layers of obsidian (~2–10 m thick; Fig. 4A), which are often interbedded with breccia, coarsely vesicular pumice, or micro-crystalline rhyolite (Fig. 4B), and capped by a veneer of blocky finely vesicular pumice blocks (~5 m thick).

The different lithofacies described at Obsidian Dome are disrupted by a plethora of different-sized fractures associated with distinctive outcrop morphologies. Herein, we use the word “fracture” purely descriptively and not genetically. A fracture is a planar or curviplanar dislocation of measurable surface area (i.e., >10 cm²) and orientation (i.e., dip and strike) that is superimposed upon the original flow banding or surface. A fracture may be concordant with flow banding, or it may be discordant, and it may have measurable displacement across it (i.e., dilation; mode 1) or along it (i.e., a fault; modes 2 or 3), or it may have neither and represent an unknown mode. A fracture terminates at a tip point (in two dimensions) or a tip line (in three dimensions). Finally, a fracture may be “healed” such that it is closed and no longer represents a mechanical weakness. We describe the fractures in order of increasing size (i.e., length, width, and depth). We describe four classes of planar features based upon scale, morphology, and their relationships to the lithofacies and other features (Table 1). The classes are based on a qualitative assessment and form a continuum.

The smallest and most ubiquitous fractures on the surface of the lava are cracks. They crosscut all the lithofacies but are most prevalent in the finely vesicular pumice, where they are irregular or very weakly curviplanar. Cracks in coarsely vesicular pumice are irregular planes, and those in avesicular obsidian can be irregular planar or strongly curviplanar (e.g., Fig. 4C). Everywhere they are ≤1 m in length and ≤1 m deep (Fig. 5A) and are usually open and dilated (≤20 cm across), widening upward. Cracks do not show plane-parallel displacement. Cracks typically occur in orthogonal sets of intersecting vertical fractures, mutually perpendicular to subhorizontal flow banding (Fig. 5B), lithological layering, and the upper surface (Fig. 5C). Cracks can incorporate different surface ornamentations (see Fig. 5) and are themselves crosscut by larger fractures (Fig. 4C). Only cracks are observed to be closed and healed, and only when formed in obsidian.

Clefts are larger than cracks with lengths of 2–10 m, depths of 1–5 m, and widths of 0.5–3 m, large enough to allow a person to stand within them (Fig. 6A). Cracks crosscut and are cut by clefts. Clefts occur in all lithofacies but are best developed in avesicular obsidian (Fig. 6B), where they typically have very irregular, although generally planar, sides.

Crevasses are typically larger still but only occur in the finely vesicular pumice and avesicular obsidian lithofacies, where they are characterized by very smooth planar margins symmetrical about a near-vertical axis (Fig. 7). Fracture depths of ≥5 m, lengths of 10–20 m, and widths of 2–5 m define crevasses; they are notable for the upwelling of cool air as you stand atop or within them, suggesting that they penetrate deep into the lava. Crevasses are the primary fracture class that define the margins of ogive structures, and therefore they play a larger role in defining the surface physiography than clefts or cracks. Their surfaces are cut by later perpendicular cracks and are rarely, if ever, ornamented (Figs. 7A–7B). Crevasses narrow and terminate toward tips buried under the finely vesicular pumice carapace, but unlike clefts, the crevasses are empty of finely vesicular pumice boulders (e.g., Fig. 7C).

The largest and least common type of fractures is the crease structures, which only occur in coarsely vesicular pumice and in five locations at Obsidian Dome (Fig. 8A) where the lava is estimated to be thickest (50–60 m). Aerial images display the stark color and textural contrast between the smooth, expansive crease structure surface in coarsely vesicular pumice and the surrounding finely vesicular pumice blocks (Fig. 8B). Although pumice mining operations have altered several crease structures, their general size and orientation are preserved, and these aspects show that the main fracture is not systematically orientated relative to the nearest flow fronts (Fig. 8A), nor do the crease structures define the ogives; rather, where they occur, they overprint the local ogive pattern.

The crease structures are characterized by (1) long fracture lengths (>10 m; Fig. 8B), (2) large fracture widths (>2 m), and (3) a flared shape, where their very smooth curviplanar surfaces routinely rotate from near vertical at the base to subhorizontal (Figs. 8C–8D). Vertical to subhorizontal flow banding is often intersected by curviplanar surfaces (Fig. 8E) with identical patterns, for example, intrafolial folds, exposed on both sides. The curviplanar surfaces of crease structures at other lavas display striations (e.g., Fink, 1980b; Anderson and Fink, 1992), but no striations were found on any of the crease structures at Obsidian Dome. The crease structure fracture surfaces are penetrated by cracks but not clefts or crevasses and are not ornamented.

Cracks and clefts are often associated with fracture surface ornamentations (Table 1). Four very different types of ornamentation were identified: red oxidized surfaces, red leathery surfaces, pink tessellated surfaces, and welded breccias. To the best of our knowledge, this is the first description of fracture ornamentations on the surface of silicic lavas.

Rust-red surfaces generally occur as thin (≤0.5 mm) veneers coating crack and cleft surfaces in any of the different lithofacies (e.g., Fig. 4C), and they are very common on finely vesicular pumice and obsidian blocks in the carapace, where they typically cover one side only. The deepest red colors occur at the flow margins and are associated with avesicular obsidian interlayered with breccia (Fig. 4B). Many red surfaces are cut by later, smaller cracks that expose the original lithology color beneath the veneer (Fig. 4C).

Red “leathery” surfaces on cracks and clefts are also thin veneers cut by later, smaller cracks but are distinctly wrinkled and contorted, and they occur on all lithofacies (Fig. 9A). Different red leathery surfaces can exhibit a variety of small-scale morphologies. In some cases, there are two perpendicular sets of wrinkles, where the larger set is parallel to the underlying flow banding, and the smaller and later wrinkles are discordant and subhorizontal (e.g., Fig. 9A). Some wrinkles are demonstrably differential extrusions of certain flow bands across an initially planar fracture.

Pink tessellated surfaces are thin (≤0.1–5-mm-thick) veneers of pink tuff annealed to a fracture surface (Fig. 9B). They occur in all lithofacies but are most readily identified in coarsely vesicular pumice, where they can be concordant or discordant with flow banding. The thin layers are nearly always stretched to form polygonal wafers (tesserae) with separation of 1–5 mm (Fig. 9C).

Oxidized lapilli- to block-sized, angular to sub-rounded, obsidian clasts fill many otherwise open cracks and, especially, clefts at the upper surface of Obsidian Dome. The breccias are typically tack-welded together and to the fracture walls (e.g., Fig. 4C) as well as strongly flattened and welded in narrow fractures within obsidian. The lapilli and blocks are occasionally nearly in situ and jigsaw-fit (e.g., Fig. 10A) where they are not discolored and are very angular; however, in most cases, the clasts are somewhat rounded and stained with an orange veneer (Fig. 10B) and appear to be allochthonous. Where tack-welded, a finer-grained orange matrix is composed of ash-sized clasts that are sub-rounded. Similar tack-welded breccias form tall pillars at the margins of the lava’s upper surface (Fig. 10C).

The upper surfaces of Obsidian Dome and all other silicic lavas and domes are characterized by rough, uneven, block-strewn, irregular masses of pumice and obsidian (e.g., Anderson et al., 1998; Plaut et al., 2004; Leggett et al., 2020). Although the carapace buries much of the coherent lava, understanding how the carapace formed is important to understanding the entire lava emplacement process. This is because, unlike typical sedimentary breccias or even the marginal talus slopes at Obsidian Dome, the clasts in the upper carapace cannot have been transported very far, if at all, from where they originated. Moreover, the clasts have not been liberated from the coherent lava by many of the typical forms of erosion and mechanical weathering that interfere with rocks, for example, erosion by water or ice, or disruption by root-wedging. We are confident of this because (1) Obsidian Dome is postglacial, (2) it has no surface drainage network and allows for no standing water, and (3) it is almost completely devoid of vegetation. Thermal expansion and contraction coupled with ice-wedging are plausible mechanisms to generate some fractures postemplacement; however, they are much more likely to accentuate preexisting (i.e., synemplacement) fractures that rainwater and snowmelt can exploit. With these caveats in mind, we discuss our interpretation of how the fractures, and their ornamentations, record lava emplacement processes.

Cracks and clefts are mode 1 tensile fractures formed by wholly brittle (finely vesicular pumice, obsidian, and coarsely vesicular pumice; Fig. 6) or brittle-ductile (obsidian only; Fig. 5B) deformation. They form perpendicular to a preexisting surface (i.e., parallel to σ₁), propagate in the σ₂ direction, and dilate in the σ₃ direction (i.e., plane strain). Where orthogonal fracture sets σ₂ and σ₃ are approximately equal (i.e., coaxial shear strain), the original surface that is fractured may be the top of the coherent lava (Fig. 5A), or it may be part of a larger fracture class. Stresses produced during the vesiculation of silicate melt exceed its tensile strength to form finely vesicular pumice. Therefore, many of the smallest cracks, especially orthogonal sets and those superimposed on crevasse and crease structure faces, were probably initiated in part by volume increases associated with the formation of finely vesicular pumice at the upper surface.

We infer that the irregular nature of these fractures was caused by the linkage of many small cracks into fewer, larger cracks and clefts. Therefore, clefts developed from growth and linkage of parallel-striking cracks under the same stress state. Linkage of smaller costriking fractures is usually achieved when opposing tips propagate past each other and then “hook” through the intervening rock barrier (e.g., Lamarche et al., 2018); this is plausible at Obsidian Dome, where the fractured material is assumed to be mechanically isotropic.

The presence of weakly curviplanar cracks and healed cracks in obsidian, and the formation of leathery surfaces on cracks and clefts in obsidian strongly suggest that the deformation in obsidian layers was brittleductile. In this scenario, cracks initiated in the overlying finely vesicular pumice by brittle failure during stretching and propagated downward into melt that was above T_g (i.e., capable of viscous flow). We are confident that fractures initiated in the finely vesicular pumice because (1) they are the most dilated (i.e., furthest from the tip) at the surface and (2) the tensile strength of finely vesicular pumice is expected to be less than that of the lowporosity obsidian (Heap et al., 2021). There, the fracture-tip propagation is fast enough to create a high shear strain rate, which is briefly faster than the viscous relaxation rate. This locally and temporarily causes the melt around the propagating crack tip to behave rigidly, causing continued brittle failure of the melt (e.g., Cordonnier et al., 2012; Huang and Hassager, 2017). Upon release of the tensile stress, the strain rate becomes zero, and the volume returns across T_g to the ductile regime, allowing the margins of the fracture to relax viscously and close together (e.g., Dingwell and Webb, 1990; Wadsworth et al., 2020). The presence of leathery surfaces on cracks and clefts in obsidian strongly suggests that, despite being at elevated temperature (i.e., above T_g), some flow bands experienced more viscous relaxation than others, and, moreover, these different bands alternated frequently. However, we found no evidence that they fractured differently. Minor but significant differences in dissolved water content probably best explain the different responses under the same temperature and stress conditions. DeGroat-Nelson et al. (2001) and Andrews et al. (2021) measured dissolved water contents in obsidian ranging from 0.13 to 0.22 wt%. Even though this range is small (~0.1 wt%), dissolved water has the greatest effect on melt viscosity when at the lowest concentrations (Romine and Whittington, 2015), and, therefore, very small differences in water content between adjacent flow bands may yield very different responses. If this is correct, it implies that the flow banding in the otherwise homogeneous obsidian formed from the mixing of a more hydrous component and a less hydrous component (e.g., Seaman et al., 2009).

Crevasses and crease structures are different from the smaller fracture classes because they are characteristically smooth and are not ornamented. Although they are both often penetrated by cracks and clefts, the surfaces of crevasses and crease structures are notably continuous and simple. Crevasses and crease structures share some important features: (1) They both appear to “push” through the finely vesicular pumice boulder carapace (e.g., Fig 7A) and do not have significant numbers of boulders resting inside them, and (2) both have vertical σ₁ and subhorizontal σ₂ and σ₃, where σ₂ > σ₃ (i.e., plane strain). However, the key difference between crevasses and crease structures is whether or not they are curviplanar. Crevasses are planar and exhibit a simple V-shaped cross-sectional profile (e.g., Fig. 7B); in contrast, crease structures are very curviplanar (Fig. 8C; Anderson and Fink, 1992). We interpret the crevasses and crease structures to be end members on a continuum when the rheology allowed for viscous relaxation (Wadsworth et al., 2018).

In this model, a single, large-magnitude, tensile failure event, possibly initiated in the finely vesicular pumice, forms a single, large crevasse fracture and a maximum of 2–5 m of dilation. We infer a single fracturing event because of the absence of striations (e.g., Anderson and Fink, 1992). If the fracture penetrates to a depth that is below T_g during the same or a subsequent stretching event, then there is the potential for viscous relaxation through horizontal flow (where σ₃ switches to become σ₁) (1) to partially close the fracture, forming a curviplanar surface, and (2) to rotate the competent lid upward, forming a crease structure (Fig. 11). The degree to which this process may depressurize the >T_g lava interior and induce spontaneous vesiculation, possibly forming coarsely vesicular pumice, will be the topic of a future study.

Cracks formed throughout the emplacement process, probably for as long as finely vesicular pumice formed and for the duration of flow. The tensile strength of the finely vesicular pumice was so low that it would have readily fractured at even very low differential stresses (<5 MPa; Heap et al., 2021). Once formed, the cracks were permanent mechanical weaknesses (unless healed by welding) that were exploited by continued stretching (Byerlee’s rule; Byerlee, 1978). As cracks continued to grow and new cracks initiated, they linked together to form clefts that penetrated deeper into the lava. We suggest that this is the basic mode by which the upper surface, finely vesicular pumice, and top of the avesicular obsidian deformed during extension.

Superimposed upon this foundation, crevasses and crease structures formed locally where the stress and rheological conditions permitted and displaced earlier-formed cracks and clefts. Crease structures, and to a lesser degree crevasses, are located at topographic highs, which may represent the thickest, and presumably slowest cooling, parts of the lava (Kenderes, 2021). Taken together, this implies that these largest fractures formed relatively late in the emplacement of any given area of the upper surface. Both penetrated thick sections of lava into the silicate melt as single, smooth planes, suggesting that a relatively thick rigid lid had to have formed before they initiated. The different sizes of crevasses and crease structures can be explained by their age relative to the gradual cooling and thickening of the lid (e.g., Park and Iversen, 1984; Stasiuk et al., 1993). In this scenario, ogives bordered by crevasses formed relatively late when the lid had reached a thickness 0.8–1.2 times the ogive spacing. At Obsidian Dome, this would equate to crevasses forming when the lid was ~8–12 m thick. The nonsystematic orientations of crease structures at Obsidian Dome (Fig. 8A) support an interpretation where tensile stresses were independent of stretching at the margins (i.e., the gravitational spreading of the flow) and instead represent local stress heterogeneities (Fink, 1983).

In summary, cracks initiated as finely vesicular pumice formed at the upper surface, and then cracks developed into clefts, and new cracks formed as long as the finely vesicular pumice remained under tension. Cracks were probably the first and last fracture types to form. Crevasses and crease structures only formed late in the emplacement of any given area of the upper surface when a relatively competent layer had formed and grown to ~10 m thickness. Under these conditions, older cracks and clefts were probably either healed or too shallow to prevent the buildup of stress in the competent layer. The rigid upper thermal boundary layer failed in a single event when a high enough stress overcame the tensile strength of the competent layer; in crevasses, the stress was due to continued spreading of the lava, whereas in crease structures, it may have been due to local heterogeneities, including the thickness and buoyancy of the coarsely vesicular pumice layer (Fig. 11).

Andrews et al. (2021) determined that T_g determined by viscometry ranges around 680 °C, 650 °C, and 614 °C for finely vesicular pumice, coarsely vesicular pumice, and avesicular obsidian, respectively. Both finely vesicular pumice and coarsely vesicular pumice form by vesiculation of silicate melt and initiate under ductile conditions and above T_g but then evolve very differently (Fink, 1983). In the case of finely vesicular pumice, the thermal gradient is very high, and it would cool below T_g in ~16–10 h (Andrews et al., 2021). However, the tensile stresses from initial volume expansion readily exceed the tensile strength of the increasingly dry and fragile finely vesicular pumice carapace, causing the formation of cracks in this time window, and probably increasing the cooling rate further by increasing the surface area. We infer that the dissolved magmatic water contents in finely vesicular pumice (≤0.23 wt%; DeGroat-Nelson et al., 2001) are preserved by fracture-enhanced cooling in a positive feedback loop where exsolution of water makes the melt more fragile (i.e., increasing porosity; Heap et al., 2021) and drives volume expansion and increasing hydrostatic pressure (i.e., stress) that exceeds the decreasing tensile strength as the carapace dries and cools (Ryan et al., 2019). Repeated generation of new cracks rapidly increases the surface area and cooling rate (tens to hundreds of minutes), eventually leading to quenching of the finely vesicular pumice carapace and terminating volatile exsolution.

Dissolved water contents in the obsidian are variable, from 0.22 wt% (the minimum for finely vesicular pumice) down to 0.13 wt% (similar to the minimum for coarsely vesicular pumice) (Andrews et al., 2021). This suggests that some parts of the obsidian were quenched rapidly and experienced minimal volatile loss, like finely vesicular pumice, whereas others remained above T_g long enough (hours) to lose some water before quenching (Wadsworth et al., 2018). We did not analyze the textural evidence in the obsidian to determine whether or not it had vesiculated and then become welded (e.g., Wadsworth et al., 2020). Where coarsely vesicular pumice formed, tensile stresses from initial vesiculation did not exceed the tensile strength, and brittle deformation did not interfere. Instead, a volatile-rich zone inflated through local vapor pressure, and vesicles could continue to grow and (probably) nucleate until stopped by cooling-induced quenching (Manley and Fink, 1987).

Taking the lowest value for T_g (614 °C in obsidian; Andrews et al., 2021) as a minimum, ductile deformation must have occurred above that. Brittle deformation can occur above T_g if the strain rate is high enough and the stress is great enough to exceed the strength of the glass (Dingwell, 1996). Therefore, the evidence allows for a range of processes, including explosive fragmentation, that may only have occurred when the rheological and stress-strain conditions were appropriate (Gonnermann, 2015), as when crevasses evolved into crease structures. Upon formation of a rigid upper thermal boundary layer, continued or renewed tensile deformation would have been concentrated upon single, large, planar fractures (crevasses). If the fracture did not penetrate through and beyond the contemporaneous T_g isosurface (Fig. 11), then the crease structure was not modified further. On the other hand, if the fracture penetrated into the melt, (1) horizontal viscous relaxation across the fracture would have closed any void space formed by dilation and plastically deformed the fracture margins, forming a crease structure, and (2) upward flow of melt would have been possible into the wrenched-open space in the lid. We envisage this process being localized to the volume immediately around the fracture and generating a rotational strain in the lid where melt was channeled up into the gap in the lid while simultaneously being drained from beneath the flanks of the crease structure. This might explain the correlation between the changing attitude of flow banding and the curvature of the crease structure margins (Fig. 8D).

The existing literature on silicic lavas arguably understates the pervasiveness and complexity of fractures at the upper surface. This is important for understanding lava cooling and lava flow dynamics, especially in analog and numerical models, similar to the ways in which ’a’ā lavas are modeled (e.g., Applegarth et al., 2010). Models by necessity simplify the rheological and mechanical properties and profiles of lavas (e.g., Griffiths and Fink, 1997; Merle, 1998). However, more sophisticated models will benefit from incorporating temperature and strainrate dependence, and temporal evolution. For example, improved understanding of the ways in which the strength and thickness of the upper thermal boundary layer evolve in response to fracturing and outgassing will yield more reasonable simulations. Similarly, cooling rate can be accelerated by the fracturing process and is probably underestimated by simple conductive cooling models.

We interpret the presence or absence of different ornamentations on fracture surfaces as snapshots of outgassing processes and different rheological responses to stress during emplacement. Based on lithofacies, ornamentation, and fracture morphology, we can consider a further sub-division of processes occurring within “cold” lava (i.e., wholly below T_g) and those that incorporate “hot” lava as well (i.e., crossing the contemporaneous T_g isosurface; Fig. 11). Hot features exhibit evidence of ductile deformation of silicate melt either as postfracture viscous flow or vesiculation and explosive fragmentation.

Fracturing occurred during and after plastic flow of the lava. Contemporaneous fracturing and ductile flow processes were recorded by welding and healing of cracks and clefts, and the evolution of crevasses to crease structures. Moreover, these features were typically accompanied by red surfaces and leathery surface ornamentations. These features were subsequently cut by planar, nonornamented cracks. Rustred surfaces are inferred to have been oxidized, probably as magnetite microlites to hematite, during exsolution of water across the fracture surfaces (e.g., Manley, 1996) at elevated temperature (>500 °C; e.g., Watkins and Haggerty, 1967; Saito et al., 2004), but this has not been tested. Therefore, red surfaces and red leathery surfaces must have formed immediately after a fracture formed and dilated before the rock could cool significantly. Leathery surfaces are plastically deformed, so they must have formed above T_g (614 °C for obsidian). Fractures therefore represent important outgassing pathways through enhanced permeability (e.g., Kushnir et al., 2017; Heap et al., 2019).

Fracture surfaces coated with tuff (pink tessellated surfaces) and lapilli-sized juvenile volcaniclastic grains are widespread but not volumetrically significant. The pink color probably resulted from oxidation by outgassing water vapor, but in these cases, the fluid also entrained ash-sized particles. The most plausible explanation for forming ash-sized tephra is the explosive fragmentation of frothy low-porosity melt, indicating rapid decompression of volatile-rich melt at the base of a fracture (Heap et al., 2021). Similar welded ash- to lapilli-size clasts are recorded within the products of effusive and hybrid lavas (e.g., Tuffen and Castro, 2009; Castro et al., 2012; Castro and Walter, 2021).

The textural and structural observations recorded in this study demonstrate and catalogue the pervasive tensile fracturing of the upper surface of silicic lavas during emplacement. Fractures characterized by size and shape record two separate continua of brittle and brittle-ductile deformation occurring in the late stages and throughout the emplacement of Obsidian Dome. We show that small-scale cracks grew and linked to form larger cracks and clefts in all lithofacies during the lava’s effusion. We hypothesize that the largest fractures, crevasses, and crease structures lie as end members along a fracture continuum occurring in the final stages of lava emplacement. A single, large-scale fracture event generated a crevasse, which may have developed into a crease structure if the fracture penetrated deep enough to reach lava above its glass transition and capable of further vesiculation to form coarsely vesicular pumice. Furthermore, ornamentation recorded on these fracture surfaces illuminates areas across the lava where fractures penetrated the brittle-ductile transition, resulting in explosive fragmentation. The identified annealed to welded ash- to lapilli-sized tuff across the surface and at the margins highlights the ongoing outgassing during lava emplacement.

We sincerely appreciate the extensive, insightful review provided by Jonathan Fink and an anonymous reviewer, which improved and expanded this manuscript. This research was supported by the National Science Foundation through awards EAR-1725131 and EAR-1935764 to G. Andrews and EAR-1724581 to A. Whittington, and the John C. and Mildred Ludlum Geology Endowment to S. Isom. The authors benefited from discussions with Kenneth Befus, Tyler Leggett, Sarah Brown, and field assistants Holly Pettus and Levi Fath.