The exposed Mule Creek vent deposits record the structure of a volcanic conduit during a hybrid explosive–effusive eruption

Silicic volcanic eruptions commonly begin with the explosive ejection of pyroclastic material, before transitioning to gentler effusion-dominated activity. Well-exposed dissected silicic systems are scarce and poorly studied, hindering the advances in our understanding of the explosive–effusive transition needed to improve interpretations of volcanic unrest and hazard forecasting. The Mule Creek vent (New Mexico, USA) is a dissected silicic conduit that records the processes controlling conduit formation and evolution, and the role tuffisites (fractures filled with variably welded pyroclasts) play in conduit dynamics. Here, we use decimeter-scale photo-mapping of lithostratigraphic units and thin section analysis to differentiate and interpret three dominant emplacement styles during vent evolution. First, there was repeated deposition and erosion of pyroclastic material at the conduit walls, recorded by erosive surfaces in pyroclastic breccia and agglomerates at the conduit margins. Second, sub-vertical domains of dense melt-dominated magma were emplaced and preserved as glass-dominated vitrophyre and brecciated vitrophyre, with the textural hallmarks of assembly from welding of pyroclasts. Finally, the sub-horizontal fracturing of previously deposited lithologies produced laterally cross-cutting tuffisites. The vent deposits track the widening and then narrowing of the conduit through time and reflect progressive insulation and generally higher temperatures towards the conduit center as pyroclasts accumulate. Welding of pyroclastic fill and the formation of dense vitrophyres towards the conduit center lowers deposit porosity and effective wall permeability. This drives localized gas pressure increases and results in gas-driven fracturing, generating tuffisites, which act as transient outgassing pathways. The structure of the Mule Creek vent records an explosive–effusive transition, constraining the processes controlling conduit evolution and aiding our interpretation of volcanic unrest.


Introduction
Silicic volcanic eruptions are preceded by the opening of fractures driven by high-pressure gas-ash mixtures (McGowan 2016), often in fissures (Lara 2009). These fissures then localize to single vents (Schipper et al. 2013). The first pyroclastic deposits recording this opening phase of explosive eruptions typically involve a high proportion of lithics, diagnostic of country rock fracturing, and ventclearing and widening (e.g., Stasiuk et al. 1996;Campbell et al. 2013;Schipper et al. 2013). The volcanic conduit is the ascent pathway along which material is transported towards the surface, and the vent is the uppermost part of this conduit system, which often flares as it approaches within ~300-500 m of the surface (Wilson & Head 1981;Cataldo et al. 2013). In silicic systems, the eruption onset is often characterized Editorial Responsibility: N. Pardo by high-energy explosive activity (Cassidy et al. 2018) involving the eruption of gas and pyroclastic material, before transitioning to effusive activity dominated by the comparatively gentle eruption of lava (Eichelberger et al. 1986). The switch between explosive and effusive activity is thought to broadly represent a transition from closed system degassing to more open-system outgassing, where gas can escape from the magma and the conduit system, releasing gas pressure (Eichelberger et al. 1986;Jaupart & Allègre 1991).
Open-system outgassing may occur vertically or laterally, with lateral outgassing allowing gas to escape from the magma and flow through permeable conduit margins, either through permeable host rock (Jaupart & Allègre 1991;Rust et al. 2004;Lavallée et al. 2013;Farquharson et al. 2015;Kolzenburg et al. 2019), permeable tephra-lined conduit margins (Rust et al. 2004) or permeable particle filled fractures-known as external tuffisites-created by the injection of ash-gas mixtures into fractures within the conduit walls (Cloos 1941;Heiken et al. 1988;Goto et al. 2008;Unwin et al. 2021). In contrast, open-system vertical outgassing requires either long-range magma permeability (e.g., an extensive magmatic foam), long-range fractures (such as internal tuffisites), a compacting column of fragmental material within the vent (Kolzenburg & Russell 2014), or dynamic outgassing during fragmentation and explosive eruption, before welding (Wadsworth et al. 2020). Internal tuffisites have been proposed to form in hot magma within a volcanic conduit when shear stresses reach or exceed the strength of the magma, forming interconnected fracture networks (Kendrick et al. 2016;Saubin et al. 2016). Tuffisites in magma have been touted as permeable outgassing pathways, potentially allowing sufficient outgassing to moderate eruption explosivity (Jaupart 1998;Castro et al. 2012Castro et al. , 2014Berlo et al. 2013;Farquharson et al. 2017;Heap et al. 2019;Kolzenburg et al. 2019). However, tuffisite permeability is only transient as welding progressively destroys pore space between hot fracture-propping particles (Stasiuk et al. 1996;Tuffen et al. 2003;Heap et al. 2019;Kolzenburg et al. 2019;Wadsworth et al. 2021).
Direct observations of silicic eruptive activity at Volcán Chaitén (2008-2009) and Cordón Caulle (2011, both in Chile, show that the effusive eruption of silicic lava is interrupted by intermittent explosions (hybrid activity), with gas and ash erupted through fractures in the vent-filling lava itself (Castro & Dingwell 2009;Schipper et al. 2013). The recognition that fragmented material trapped in lava-hosted fractures can weld to produce a continuous melt by viscous sintering (Gardner et al. 2018(Gardner et al. , 2019Wadsworth et al. 2019Wadsworth et al. , 2021Farquharson et al. 2022), and the discovery of obsidian clasts and bombs formed of welded fragments (Cabrera et al. 2015;Gardner et al. 2017;Giachetti et al. 2021;Schipper et al. 2021), has led to models suggesting that the accretion of clastic material on to the conduit walls could form both obsidian pyroclasts (Gardner et al. 2017(Gardner et al. , 2019 and silicic lava (Wadsworth et al. 2020. Simultaneous explosive and effusive activity has been explained by continued fragmentation at depth, with effusive lava formed by either material accreting and welding at the conduit walls (Wadsworth et al. 2020), or a rising column of intact and degassing magma that does not fill the entire conduit (Schipper et al. 2021).
Dissected mafic systems, such as diatremes, are more common and better studied than silicic vents (e.g., White & Ross 2011;Valentine & White 2012;Ross et al. 2017). Diatremes and silicic vents share many similarities, with both formed of an inverse cone-shaped body of pyroclastic material created by blasting and then deposition, and a complex internal structure formed by the continued overturning of previously deposited material (e.g., Ross & White 2006;Schipper et al. 2021;Valentine & Cole 2021;Wadsworth et al. 2022). Comparing and contrasting these systems would help to improve our understanding of both groups. Dissected silicic vents provide an opportunity to examine structures and textures of an upper conduit system and test the diverse hypotheses for vent processes described above. The Mule Creek vent, New Mexico (USA), is a classic example of a dissected silicic vent and is a type locality for examining the relationships between tuffisites and the vent that they intersect ( Fig. 1a and b; Stasiuk et al. 1996). Many features of the Mule Creek vent are described in detail by Stasiuk et al. (1996). Here, we build on their observations, focusing on a large outcrop of pyroclastic breccia at the vent margin that provides new insights into the early stages of vent evolution.

Geological setting of the Mule Creek vent
The Mule Creek vent (MCV) is located ~8 km north of Mule Creek in the Gila National Forest, southwestern New Mexico, USA (Fig. 1a), and is the best exposed of at least six genetically linked rhyolitic plugs and vents distributed along the northwest trending Potholes Country Graben (Stasiuk et al. 1996;Ratté 2004), together with voluminous pyroclastic deposits. The MCV was formed during an eruption emplacing the 20-21 Ma Potholes Country rhyolite, a lava flow which is mostly confined within the Potholes Country Graben (Stasiuk et al. 1996). The Mule Creek vent is exposed in a tall canyon 350 m deep, cutting through the ~23-25 Ma Bearwallow Mountain andesite host rock (Fig. 1b). The flared shape of the vent towards the canyon rim suggests that the canyon exposes the uppermost ~350 m of the vent. The upper half of the canyon is encased in bedded rhyolitic tephra up to 150 m thick, emplaced during vent formation, indicating the position of the palaeosurface before the formation of the MCV ( Fig. 2a and b; Ratté 2004). The Bearwallow Mountain andesite at the canyon base is composed of vesicular brecciated andesite lavas that are red brown to grey in color. The pyroclastic deposits are sub-horizontally bedded > 100 m from the vent, and their dip into the vent increases closer to the contact with the Potholes Country rhyolite lava (Stasiuk et al. 1996).

Field and analytical methods
A large outcrop of pyroclastic breccia at the vent margin was identified during fieldwork at the MCV in 2010 (HMC09, Figs. 1b and 2), with a greater thickness and more complex internal structure than that described by Stasiuk et al. (1996).
Samples and photographs of the MCV units and intersecting tuffisites were collected, enabling the multiscale (macro and micro) analysis of the four different vent units (pyroclastic breccia, vitrophyre breccia, vitrophyre, and central rhyolite lava) and their intersecting tuffisites. Thin sections were prepared from a sub-set of the hand specimens and have been used as part of a detailed microstructural study. In addition, polished thin sections were examined using a Hitachi SU-70 high-resolution scanning electron microscope with a 15 kV beam voltage and 15 mm working distance. Backscattered electron (BSE) images of the tuffisite matrix  Stasiuk et al. (1996) and Ratté et al. (2004).  The same view as in part a but with a simplified geological interpretation overlain. c A zoomed-in photograph showing the structure of the vent more clearly. d The same view as in part c but with a simplified geological interpretation overlain. Outcrop names and key features are marked were taken of all samples at an appropriate scale to capture ~200 grains in each image for particle size distributions, with the collected images each 1280 × 960 pixels. The area of the pyroclastic breccia imaged to produce the particle size distribution was 0.016 mm 2 at × 800 magnification with a scale of 17.5 pixels/μm. The area of vitrophyre breccia imaged was 0.1 mm 2 at × 300 magnification with a scale of 3.0 pixels/μm. Particle size distributions for particles within tuffisites were determined by measuring the area of particles in backscattered electron images using ImageJ (Schneider et al. 2012). The low contrast between particles and their touching margins prevented effective image thresholding for particle detection (Lormand et al. 2018). Therefore, the margins of each particle were traced manually. The traced particles were then separated out and each particle measured individually using ImageJ. The particle area data collected reflects 2D slices through 3D samples in the thin section but will here be referred to as particle size for simplicity.
The particle area was converted to an equivalent diameter for each particle using d = √ 4A∕ , where A is the measured particle area and d is the equivalent spherical particle diameter. The circularity of the average particle in each image ranged from ~0.74 to 0.85, calculated with the in-built algorithm within ImageJ (imagej. nih. gov/ ij; Schneider et al. 2012). Any particle with an equivalent diameter less than 10 pixels across or an area greater than 1% of the measured image area were considered not to be sufficiently well represented in the data and were filtered out (Shea et al. 2010). The particles were sorted into a series of geometric bins, with the smallest bin equal to the minimum size threshold (10 pixels in diameter) and each subsequent bin 10 0.1 larger than the previous. The histogram was then normalized so that the sum of the areas is equal to one, converting to particle area fraction.

Field observations
The MCV is composed of a series of four major vent units, with the outermost unit comprising a pyroclastic breccia in contact with the andesite country rock (Figs. 2c and d and 3;Stasiuk et al. 1996). The pyroclastic breccia is bordered on the inside edge by two dark-colored units with a glassy appearance (i.e., nominally aphyric albeit with some microlite-dominated flow bands), termed by Stasiuk et al. (1996) as a vitrophyre breccia and a vitrophyre. The vitrophyre breccia is composed of heavily fractured clasts of glassy material and borders the pyroclastic breccia. The vitrophyre breccia becomes less heavily fractured and more sheared towards the center to form a vitrophyre unit formed of more coherent glass (Figs. 3a and b). The vitrophyre unit grades into a flow-banded rhyolite lava at the vent center (Stasiuk et al. 1996). The vent units are intersected by particle-filled fractures known as tuffisite veins, which form either steeply dipping and aligned features or sub-horizontal structures (Figs. 3a and b). Tuffisites injected into the walls of the MCV are defined as external tuffisites and may be constrained to an individual vent unit or in some cases cross-cut multiple vent units. These external tuffisite veins are filled with fragments of their host rock, as well as ashsized particles that preserve sedimentary structures such as laminations.
The following sections describe field and sample observations that augment and supplement those described by Stasiuk et al. (1996). We report observations on a locality down-slope (~250 m deeper) from those investigated by Stasiuk et al. (1996), focusing on the pyroclastic breccia, vitrophyre breccia and vitrophyre units, and tuffisites (HMC09; Figs. 2c and d).

Pyroclastic breccia and apron
Pyroclastic deposits around the vent form a ≤ 150 m thick apron (Fig. 1b;Stasiuk et al. 1996), with subunits closer to the rhyolite lava contact dipping more steeply (≤ 20°) into the vent than those further away (Fig. 3b) and commonly showing evidence of gravitational slumping. At distances > 100 m from the rhyolite lava contact, the pyroclastic deposits are approximately sub-horizontally stratified or dip away from the vent (≤ 5°). A transect through this pyroclastic apron at the canyon rim reveals that they become relatively less lithic-rich (fewer non-juvenile clasts) and increasingly pumice-dominated moving stratigraphically up-section ( Fig. 4a; Stasiuk et al. 1996).
The outermost unit of the MCV is a pyroclastic breccia that is in contact with either the pyroclastic apron or the Bearwallow Mountain andesite country rock (Stasiuk et al. 1996). Clasts in the pyroclastic breccia are composed of rounded pumice and angular lithic fragments derived from the Bearwallow Mountain andesite (Fig. 5a). Close to the canyon rim, Stasiuk et al. (1996) recorded the pyroclastic breccia as being 0.2-2 m thick and infilling irregularities at the vent margin. This study focuses on the characteristics of a much thicker (> 14 m) pyroclastic breccia sequence exposed closer to the base of the canyon on the eastern side of the vent (HMC09, Figs. 2a and d and 6). This outcrop is ~100 m deeper than the base of pyroclastic apron mapped by Stasiuk et al. (1996), which can be seen in satellite imagery and field photographs to be higher up the canyon wall (Figs. 1b and 2b). We therefore interpret this outcrop of pyroclastic material as representing part of the pyroclastic breccia emplaced within the evolving vent during its formation, rather than as part of the pyroclastic apron, though these two deposits may be related to one another (Valentine & Cole 2021).
The structure of the pyroclastic breccia is complex, formed of multiple stratified subunits distinguished by their overall grain size and componentry. The outer subunits are exposed at the base of the outcrop where they form sub-horizontal layers that grade into one another (Fig. 6). Closer to the vent center, the pyroclastic breccia subunits dip steeply towards the vent center before becoming subvertical, with a greater number of steep erosive surfaces. Many subunits contain slump features and steeply dipping extensional faults that offset subunit contacts by up to tens of centimeters (Fig. 7a). Faults are differentiated from slump or erosive features as more planar features, a Vent structure towards the canyon rim as identified by Stasiuk et al. (1996). b Conduit structure as identified from the outcrop HMC09, see Fig. 1b for location. The outermost part of the vent consists of pyroclastic breccia, which is texturally very similar to the surrounding pyroclastic apron. Adjacent to the pyroclastic breccia is a unit of vitrophyre breccia, with glassy, often jigsaw-fit clasts that become more sheared and compacted towards the vent centre. The vitrophyre breccia grades inwards to become a more coherent vitrophyre unit. The vitrophyre grades into the rhyolite lava unit that forms the vent centre. The rhyolite lava unit is flow banded with flow bands that become more widely spaced towards the vent centre. Every unit of the conduit is cross-cut by tuffisite veins, with subvertical tuffisites in the rhyolite lava and vitrophyre units. In the vitrophyre breccia and pyroclastic breccia, tuffisites dip less steeply. Some tuffisites are laterally continuous, up to 13 m in length, while other tuffisites can only be followed for up to 50 cm confined to the pyroclastic breccia that offset sections of an identifiable subunit.
The different subunits of the pyroclastic breccia contain sub-angular clasts of the andesite host rock as well as rounded pumice clasts. There are varying clast sizes, with coarser subunits containing ≤ 30 cm lithic (non-juvenile) clasts and ≤ 80 cm pumice clasts, while finer grained deposits contain ≤ 2 cm clasts (Fig. 6). Sedimentary structures such as graded beds and coarser lenses of material are common in the finer grained subunits (Figs. 7a and b). A transect through the outcrop indicates highest lithic content towards its base (Fig. 4b). Some jigsaw-fit clasts are separated by thin ribbons of pyroclastic breccia (Fig. 7c). In some layers of pyroclastic breccia, the intergranular space between the larger clasts is filled with a matrix of densely welded finegrained < 5 μm particles (Fig. 7b).
The nature of the contacts between the various layers of pyroclastic material are highly variable, ranging from sharp, steeply inclined sub-vertical margins marked by a distinct, narrow glassy 'rind,' defined by its low vesicularity and absence of visible clasts at outcrop scale, to more complex and highly irregular morphologies with 'feather-like' interdigitating contacts. An example of a highly complex feather-like contact is shown in Fig. 8 where a laminated pyroclastic deposit composed of fine lapilli-and ash-sized grains is cross-cut by a relatively younger lithic breccia. Within a few centimeters of a sub-vertical contact running down the center of the photograph (Fig. 8a), the lamination is deflected upwards to become parallel to this glassy boundary (Fig. 8b).

Vitrophyre breccia and vitrophyre
The vitrophyre breccia and vitrophyre (Fig. 3) are lithologically very similar, forming subvertical dark-colored units with a glassy texture (Figs. 5b and c). These units are both non-vesicular. The vitrophyre breccia is a 0-3-m thick unit composed of mostly angular-subangular vitrophyre clasts (ranging from 1 to 10 cm across) and a few rare andesite clasts ~3 cm across ( Fig. 5b; Stasiuk et al. 1996). The vitrophyre clasts often show a jigsaw-type fit, separated by red-colored welded clastic material, suggesting in situ fracturing of the material. However, Stasiuk et al. (1996) recorded some clast imbrication, suggesting at least some localized transport of clasts. Close to the contact with the vitrophyre (< 50 cm from contact) clasts become more elongate than those further from Fig. 4 Logs of the maximum clast size and % lithics in the different pyroclastic breccia subunits in a Stasiuk et al. (1996), measured close to the canyon rim and b this study in outcrop HMC09. Componentry was completed at outcrop-scale (clasts > 2 cm across). Locations of both logs are marked on Fig. 2b. The trace of the log from outcrop HMC09 is drawn on Fig. 6b ▸ the contact, indicative of shear deformation parallel to the vent margins, and clasts are also more strongly welded together (Stasiuk et al. 1996).
The vitrophyre unit is typically ~1 m thick but locally reaches 7 m. Elsewhere within the vent, this vitrophyre unit can be absent (Stasiuk et al. 1996). Flow bands 1-100 μm thick in the vitrophyre are seen to be parallel to the vent margins in oriented samples but can be contorted and discordant where the contact is irregular ( Fig. 5c; Stasiuk et al. 1996). The vitrophyre grades into the central rhyolite lava unit.

Flow-banded rhyolite lava
The innermost 25 m (measured from an approximate conduit center-line close to the canyon base) of the MCV is composed of grey to pink flow-banded rhyolite lava (Figs. 3 and 5d). The flow banding is more narrowly spaced (< 0.5 mm thick) towards the margins of the rhyolite lava compared with the unit centre (Fig. 3b). Towards the center of the MCV the flow banding becomes relatively wider spaced and increasingly deformed by open to tight, steeply inclined isoclinal folds (Stasiuk et al. 1996). The flow bands consist of layers mostly defined by variations in the degree of vesicularity of the rhyolite lava, with some layers containing clearly visible, open subrounded vesicles, while other layers are less vesicular in appearance and display deformed coalesced vesicles (Stasiuk et al. 1996). Some flow bands are instead defined by alternating bands of aligned microlites and cryptocrystalline rhyolite. The coarsely flow banded rhyolite lava at the very center of the conduit is the most vesicular compared with the outer parts of this unit at outcrop scale. This unit shows welldeveloped sub-vertical platy fracturing (Fig. 5d). Small sub-millimeter spherulites in the center of the rhyolite lava increase in size and abundance towards the rhyolite lava margins, before decreasing once again where the rhyolite lava grades into the adjacent vitrophyre unit over a zone approximately 1 m thick (Stasiuk et al. 1996).

Pyroclast-filled fractures
Pyroclast-filled fractures have been recognized in each different unit of the MCV (Fig. 3) and typically form crosscutting veins that record fluid transport and injection into the host rock. The size of these veins varies considerably from laterally extensive features which are > 30 m in length to much shorter veins that can only be followed for ~2 cm. The morphology of the pyroclast-filled veins varies depending on their host lithology. Consequently, we have divided the veins into the following categories-those hosted in (1) country rock, (2) pyroclastic breccia, (3) vitrophyre or vitrophyre breccia, (4) cross-cutting multiple units, and (5) flow-banded rhyolite lava. The characteristics of the pyroclast-filled veins within each host rock are summarized in Table 1. Many of these observations are similar to those found closer to the top of the canyon by Stasiuk et al. (1996).

Pyroclast-filled veins hosted in the country rock
The mostly sub-horizontal pyroclast-filled veins (~2 cm in width) within the country rock extend into the surrounding Bearwallow Mountain andesite or pyroclastic apron ( Fig. 3; Stasiuk et al. 1996). The veins divide into multiple narrower veins (< 1 cm in width) at their tips to produce a branching network (Figs. 3 and 9a and b; Stasiuk et al. 1996). The veins are filled with a matrix of red, oxidized, ash-sized particles with fragments of quartz and feldspar phenocrysts and larger (< 1 cm across), subangular clasts of rhyolite lava and andesite (Stasiuk et al. 1996). Laminations and cross-laminations are common in the finer ash-sized particles found at the fracture margins.

Pyroclast-filled veins hosted in the pyroclastic breccia
Pyroclast-filled veins in the pyroclastic breccia are typically ~5 cm in width with diffuse margins, giving veins a wispy appearance (Figs. 9c and d). The fill is entirely composed of ash-sized particles that are typically welded (<5 μm across; Figs. 10a and 10b). This material has a similar appearance to the matrix of the pyroclastic breccia and the material forming tuffisites in the country rock, suggesting a similar composition. The finest material is at the fracture margins where it forms sub-horizontal to irregular, finely laminated features that infill originally open pore spaces and/or voids within the host breccia. Clasts of host material also sometimes protrude across vein margins ( Fig. 11a) and clasts from the host breccia that are substantially larger than those of the surrounding vein material can also be found at vein centers ( Fig. 11a). This is in contrast to tuffisites seen elsewhere that occur within fractures. Within the pyroclastic breccia, the clast-filled veins were observed forming isolated areas of fine-grained material that are not apparently connected to the margin of a pyroclastic breccia subunit, though could be connected in 3D.
The pyroclastic breccia-hosted veins typically contain complex internal structures, including laminations, climbing ripples, and internal self-injections (Fig. 11b). The host pyroclastic breccia adjacent to the veins has experienced softsediment deformation, with localized vein-adjacent fluidization producing complex structures including disharmonic folds, cuspate to irregular boundaries between the laminated vein and breccia (Fig. 11a), and the disruption of the matrix to the breccia to form 'ball-like' structures enclosed within the vein (Figs. 11b and c). Soft-sediment deformation of the host pyroclastic breccia was also observed and associated with the fragmentation of these fine-grained pyroclast-filled veins themselves, which are broken up into jigsaw-fit clasts of densely welded material.

Pyroclast-filled veins hosted in the vitrophyre and vitrophyre breccia
Pyroclast-filled veins within the vitrophyre are mostly oriented sub-vertically, parallel to the banding within the vitrophyre (Fig. 5c). Most of the veins are too narrow (< 1 mm in width) to contain visible internal structure and are filled with brown-colored altered devitrified material, with fuzzy Many pyroclast-filled veins within the vitrophyre-breccia have the same appearance as those in the vitrophyre. However, some fractures within the vitrophyre breccia are larger, up to 1 cm in width, and contain either discontinuous laminae or irregular patches or clasts of clastic material resulting in complex internal structures (Figs. 5b and 11d). The fracture margins are often lined with vitrophyre fragments up to 20 μm across (Fig. 10c) that are welded-locally to such an extent that the individual clasts are no longer visible (Fig. 10d). Horizons of ash-sized particles, angular vitrophyre clasts, and crystal fragments weave between those vitrophyre blocks. The clast-filled veins can also contain irregularly shaped blobs of melt with microlites that can be aligned in sheared clasts (Fig. 10d). The microlite texture of each of the melt blobs is qualitatively similar and the melt blobs are much more microlite-rich than the surrounding vitrophyre. Narrower veins within the vitrophyre breccia have structures that are less complex, but often still show cross-cutting relationships. The erosive contacts between the different injected domains are sharp, with no indication of soft-sediment deformation. These veins do not show other internal structures such as cross-bedding or lamination (Fig. 11d).

Cross-conduit pyroclast-filled veins
Pyroclast-filled veins are not restricted to spanning the width of one vent unit, and Stasiuk et al. (1996) reported laterally extensive veins originating in the vitrophyre breccia and continuing outwards through the pyroclastic breccia and into the surrounding andesite country rock. In this study, one laterally extensive vein with a length > 30 m was traced through the pyroclastic breccia and into the country rock (HMC07, Figs. 2a and 3b). These veins have a similar appearance to the matrix of the pyroclastic breccia and the pyroclastic brecciahosted tuffisites, suggesting a similar composition.

Pyroclast-filled veins hosted in flow-banded rhyolite lava
The pyroclast-filled veins within the rhyolite lava are typically ~2 cm wide and tens of centimeters long and steeply dipping, aligned parallel to flow banding within the host rock ( Fig. 9e; Stasiuk et al. 1996). The internal structure of these veins is complex, with sharp well-defined erosive to gradational contacts separating clastic domains with different grain sizes. Soft-sediment deformation is clearly seen, with curved walls to injections and deformed laminations. The pyroclast-filled veins can form an angular network filled with welded material.

Interpretations
The units of the MCV are interpreted to have all been emplaced during a waning explosive eruption, forming in order from the pyroclastic breccia at its outer margins deposited first, then cross-cut by the accretion of the vitrophyre breccia and vitrophyre, before finally the emplacement of the rhyolite lava within the centre of the evolving vent ( Fig. 12 and animation (Online resource 2)). Repeated fracturing and the injection of material into the vent units would have been facilitated by pulsations within this overall waning phase.

Formation of the pyroclastic breccia
The different subunits of the pyroclastic breccia are often separated by erosive surfaces, suggesting repeated cycles of erosion and deposition. The extensional faults seen within the pyroclastic breccia ( Fig. 7a) reflect material collapsing back into the vent at the end of a phase of deposition. Subhorizontal subunits away from the vent center are interpreted to have formed first, cross-cut by later subunits deposited closer to the vent center ( Fig. 12 and animation (Online Resource 2)). Welding of the pyroclastic breccia allowed it to transition from a loose pile of fragmented rock and pyroclasts that would have been frequently overturned to a more coherent rock. Later sub-vertical pyroclastic breccia would have been injected through this more coherent material, allowing steeper and better-defined erosive contacts to form. Fractured jigsaw-fit clasts separated by ribbons of clastic material (Fig. 7c) record a smaller-scale version of this process, with sudden pressure pulses breaking clasts during deposition. The pyroclastic breccia contains a substantial volumetric proportion of lithic clasts, suggesting that the breccia was emplaced during an early stage of vent formation when explosive processes were quarrying, transporting and depositing clasts of the surrounding country rock (Fig. 4; Stasiuk et al. 1996). The erosive contacts suggest that significant erosion was still occurring, interspersed with periods of deposition. The lower lithic clast proportion (< 10 vol.%) in the later sub-vertical pyroclastic breccia subunits could indicate that earlier breccia subunits acted as a barrier preventing further quarrying of the country rock (Fig. 4). Through time, subunits became less lithic-rich as the conduit began to stabilize in both size and shape, reducing the overturning of the pyroclastic  breccia already deposited and breaking less material from the conduit walls (Valentine & Cole, 2021). The steep glassy contact (Fig. 8) is an example of one of the erosive margins between different breccia subunits. Its glassy margin formed via accretion of ash-sized particles to the sub-vertical surface, where they could weld to form dense glass, through the same processes thought to form obsidian or vitrophyre at conduit margins (Gardner et al. 2017). The formation of the glassy margin suggests that this position in the vent was sufficiently hot for welding to occur, and if deposited rapidly enough, this may have helped to strengthen the margin, allowing it to exist at such a steep angle.

Formation of the vitrophyre breccia and vitrophyre
The similarities between the vitrophyre breccia and vitrophyre units, both dark units with a glassy appearance and low vesicularity, suggest that these units are formed of the same material. These units are therefore interpreted to have formed through the same process, with the vitrophyre breccia a brecciated derivative of the vitrophyre. The vitrophyre is formed by the hydration of obsidian due to post-emplacement alteration. Obsidian can form by the accretion of pyroclasts at the conduit margins (Online Resource 3; c.f. Gardner et al., 2017), and if continued this process could also produce a thick unit coating the conduit walls (Wadsworth et al. 2020). This is the same process that we suggest formed the thin, localized glassy contact within the pyroclastic breccia ( Fig. 8; see "Formation of the pyroclastic breccia"), but here occurring on a far larger scale in terms of both the thickness and vertical continuity of the dense glassy layer formed. The lack of lithics in the vitrophyre breccia and vitrophyre suggest that quarrying of the country rock had decreased by the time these units were emplaced, perhaps due to the source of lithics becoming cut off as the margins of the conduit were no longer country rock, or because explosivity waned.
There are two potential mechanisms by which the vitrophyre could brecciate to form the vitrophyre breccia: either the latter formed through in situ fracture of a coherent vitrophyre, supported by the pseudo-jigsaw-fit appearance of many clasts and the presence of tuffisites, or clasts were eroded and transported from greater depth, before being deposited to form the vitrophyre breccia (Gardner et al. 2017;Wadsworth et al. 2020). Compaction and shearing later altered the shape of these clasts towards the vitrophyre breccia-vitrophyre contact (Fig. 3), causing them to become more elongate, masking the evidence for these processes across much of the unit.
The persistence of lithics suggests some clasts within the vitrophyre breccia were sourced from elsewhere, perhaps sourced from a greater depth and transported up the conduit to their place of deposition (Stasiuk et al. 1996;Saubin et al. 2016). Imbricated clasts must have been transported at least a few metres before being deposited. Most clasts however have a jigsaw-fit appearance, suggesting that while some clasts may have been transported from elsewhere, the local clast production by in situ fracturing was more prevalent (Online Resource 3). Shear traction on the assembling vitrophyre walls due to the bypassing high velocity gas and pyroclasts could produce a sufficiently high shear stress, relative to the elastic shear modulus, to brecciate the vent margin ). If the brecciation of the vitrophyre walls is indicative of high velocity bypassing flow, then the transition to non-brecciated vitrophyre could represent eruption waning and a further reduction in explosivity, as dense conduit-plugging material experienced a lessening degree of damage by pulses of pressurized magmatic fluids.

Formation of tuffisites
The pyroclast-filled veins found cross-cutting the different vent units are interpreted as tuffisites, with all veins except those in the pyroclastic breccia occurring within fractures. We interpret the fracture-hosted tuffisites as hydrofractures driven by pressurized magmatic fluids (gas-pyroclast mixtures) above the fragmentation level in the eruption-feeding conduit (Wadsworth et al. 2020;Unwin et al. 2021). Most of the tuffisites seen within the MCV have complex internal structures that must reflect either variations in the fluid velocity along the fracture or changes in fracture width, or both. Variations in fluid velocity can be driven either externally, by gas pressure fluctuations within the volcanic conduit, or internally by local variations in the fracture width, created by pre-existing narrower fracture sections or by clogging the fracture with deposited material (Perkins & Kern 1961;Cosgrove 2001;Phillips et al. 2013;Unwin et al., 2021).
Within the MCV we identify four different phases of tuffisite formation: (1) the formation of tuffisites within the country rock, associated with the opening of the conduit pathway ( Figs. 9a and b), (2) diffuse tuffisites within the pore space of the pyroclastic breccia (Figs. 9c and d), (3) Fig. 10 Grain size distributions of particles seen in SEM images. The particles measured are colored in red in the SEM image, and n is the number of particles measured. a, b Particles forming a diffuse tuffisite within the pore space of the pyroclastic breccia. Image location is given in Fig. 11a. c, d Particles forming a tuffisite within the vitrophyre breccia. The location of the SEM image is given in Fig. 11d. The original untraced images can be found in Online Resource 1 n = 639 mean = 7 µm 50 μm 20 μm c b a d n = 244 mean = 2 µm tuffisites hosted within the vitrophyre and vitrophyre breccia (Figs. 5b, 5d, and 11d), and (4) tuffisites within the central rhyolite lava (Fig. 9e).

Tuffisites in the country rock
Tuffisite veins within the andesite country rock contain clasts of the host rock at their margins that have been fractured away from the vein wall (Figs. 9a and b). This fracturing process records the propagation and widening of the fracture network, with country rock tuffisites recording the first stages of vent evolution (Cloos 1941). These tuffisites record some of the first fractures that were opened as part of the fracture network that developed into the MCV. Parts of the fracture network that have been abandoned and not destroyed during the quarrying of the vent can be filled with pyroclastic material and preserved as tuffisites within the country rock ( Fig. 12 animation (Online Resource 2)).
The diffuse margins of tuffisites hosted by the pyroclastic apron reflect the difficulties of opening and propagating a pathway through the highly permeable host rock. Instead, gas escape exploited, and potentially widened, pre-existing open voids and intergranular pore spaces within the breccia, which progressively filled with laminated/stratified ash. Deposition of these tuffisite cavity fills would have led to a locally significant reduction in the permeability of the pyroclastic apron, acting to seal areas of the evolving vent margins.

'Early' tuffisites in the pyroclastic breccia
The tuffisites within the pyroclastic breccia are not fracture hosted, but instead comprise fine-grained materials likely swept through the pore space of the surrounding pyroclastic breccia and driven by lateral pressure gradients within the volcanic conduit ( Fig. 11a and b). Particles carried into the pore space would likely be deposited either where the fluid pressure decreased as it entered a larger void, or where the Fig. 11 Photomicrographs and SEM images tuffisites within the pyroclastic breccia and vitrophyre breccia. White labelled boxes indicate the position of SEM images within the larger scale image. a Photomicrograph of a tuffisite within the pyroclastic breccia, with finely laminated margins. The injected material grades into coarser particles towards the tuffisite center. b Photomicrograph of tuffisite in the pyroclastic breccia, with irregular margins, formed of fine-grained laminated material, and a coarser center. c Backscattered electron image of fine-grained injection within the pyroclastic breccia. Small ash-sized particles fill the pore space of the host breccia. d Sharp-edged tuffisite within the vitrophyre breccia containing different phases of injected material separated by erosive surfaces fluid flux was lowest. These tuffisites leave behind a record of this process in the areas of lower permeability or where fluid flow pathways become blocked. Unlike the tuffisites found in other vent units, these tuffisites are not formed by hydrofracturing, instead representing outgassing through dilation of existing permeable pathways that did not require fracturing to further increase permeability.
The soft sediment deformation surrounding these features reflects the passage of pressurized gas through the pore spaces of the unlithified breccia, which reduced the effective confining pressure, lowered its cohesive strength, and allowed for shifting and slumping. More extensive tuffisites (< 13 m in length) may represent areas where the fluid flow within the pyroclastic breccia was more constrained to a single pathway, allowing for more focused deposition of material into a single vein. These more laterally extensive tuffisite veins are consistent with the infilling of the pre-existing intergranular pore spaces and cavities, not only lowering the permeability of the breccia but also reducing the friability of these pyroclastic deposits, facilitating subsequent fracture development.

Tuffisites in the vitrophyre and vitrophyre breccia
The morphology of the tuffisites within the vitrophyre and vitrophyre breccia reflects the greater cohesion and lower permeability of the vitrophyre compared to the pyroclastic breccia. The wide variety of tuffisite orientations seen in the vitrophyre breccia (Fig. 5b) suggest that tuffisite formation is not necessarily controlled by pre-existing fabrics, although the branches at the margins of the tuffisite in Fig. 11d do exploit pre-existing fractures in the surrounding vitrophyre breccia. The more poorly developed fabric of the outermost parts of the assembling vitrophyre breccia would act as less  2). a Fractures are opened by a high-pressure gas-pyroclast mixture. If preserved, material deposited within these fractures forms tuffisites within the country rock (e.g., Fig. 9a). b Continued fracturing forms a network of fractures that can grow in size (thickness and length) due to the quarrying of pyroclastic material. c Propagation and widening of fractures allows for the formation of a proto-conduit. d Repeated deposition and erosion as fluid pressure within the developing conduit fluctuations allows for the emplacement of multiple subunits of pyroclastic breccia. Pyroclastic breccia-hosted tuffisites can form (e.g., Fig. 9bd). e Once the conduit reaches a great enough temperature, welding of ash-sized particles at the conduit margins produces the vitrophyre. Repeated fracturing and healing forms the vitrophyre breccia with jigsaw-fit clasts and vitrophyre breccia-hosted tuffisites. Cross-conduit tuffisites that intersect multiple vent units form. f Once the explosivity drops sufficiently, the accreting vitrophyre will not be as greatly fractured, producing a more intact vitrophyre unit. Fracturing of the vitrophyre by high-pressure gas-ash mixtures produces tuffisites hosted within the vitrophyre of a control on the morphology of intersecting tuffisites than the stronger fabric seen within the vitrophyre. The tuffisites within the vitrophyre and vitrophyre breccia were emplaced after the development of the fabric, once the vitrophyre had cooled sufficiently to have fractured by shearing, producing pathways that tuffisites could easily exploit.
The tuffisites in the vitrophyre breccia are interpreted as fracturing and injecting material into the stationary conduit wall, leading us to consider them to be external tuffisites, rather than internal tuffisites formed by in situ shear failure of highly viscous magma (c.f. Stasiuk et al., 1996, Tuffen et al., 2003. The interpretation of the vitrophyre forming by accretion at the conduit margins suggests that these tuffisites formed instead as external tuffisites by tensile failure, with some clasts broken from the fracture walls as the fracture opened, and other clasts injected. Some obsidian-hosted tuffisites described elsewhere are angular particle-filled fractures that have developed tractional bedforms (Tuffen et al. 2003;Tuffen & Dingwell 2005), and display very similar morphologies and internal structures to the Mule Creek vitrophyre breccia and vitrophyre-hosted tuffisites. These were originally documented at Torfajökull, Iceland, and interpreted as internal tuffisites. In that interpretation, the clastic fracture fill was thought to be locally derived clasts from fracture walls (Tuffen et al. 2003). However, the observations made herein suggest that these similar features may actually have been external tuffisites with clastic fill sourced from an erupting and bypassing dispersion travelling up a volcanic conduit prior to conduit sealing by continued welding. This reinterpretation would provide a source for the trace metal concentrations in the Torfajökull veins, which Berlo et al. (2013) attributed to a genesis that was 'deep' relative to the tuffisite position. Similarly, the observations made here suggest that the pyroclasts filling tuffisites found in vitrophyre-like bombs, such as the bomb documented by Saubin et al. (2016), may be sourced from magmatic fragmentation, contrary to models that suggest that the fill is formed by disaggregation of a foamy magma directly beneath the vitrophyre. Taken together, these comparisons between the MCV vitrophyre-hosted tuffisites and similar tuffisites found elsewhere lend support to models that envisage the shallow conduit as progressively plugging by sintering of the products of deeper fragmentation (Wadsworth et al. 2020Farquharson et al. 2022). We propose that tuffisites can be considered to be predominantly external rather than internal, as a large proportion of their fill is distally derived, and carried through tensile fracture networks within largely passive, conduit-lining dense magma. The particle source is the deeper gas-ash-pyroclast mixture feeding the ongoing eruption, consistent with models for combined explosive-effusive eruption behavior (Wadsworth et al. 2020), and failure is driven by gas overpressure (Unwin et al. 2021). This model provides a distinct fluid and particle source able to fluidize and transport particles within the tuffisite and emphasizes the passive role of the dense conduit lining over the timescale of overpressuredriven brittle failure events, even if it remained capable of some viscous deformation.

Tuffisites hosted in flow-banded rhyolite lava
Tuffisites within the central flow-banded rhyolite lava (internal tuffisites) have previously been proposed to form by the same mechanisms as those in vitrophyre or obsidian and are often considered to form when stresses within magma meet or exceed the magma shear strength, opening networks of magma-hosted fractures (Tuffen et al., 2003;Tuffen & Dingwell 2005;Kendrick et al. 2016;Paisley et al. 2019). The tuffisites within the rhyolite lava of the MCV are steeply dipping, reflecting either failure parallel to the plane of maximum shear stress, rotation of veins by subsequent viscous flow and shear (Tuffen et al. 2003;Gonnermann & Manga, 2003), or exploitation of weaker pore-rich bands associated with devitrification (Schipper et al. 2015).
The clasts in internal tuffisites have previously been thought to derive entirely locally from fracture walls (Tuffen et al., 2003;Tuffen & Dingwell 2005), or from a foamy magma immediately beneath, and in contact with, a dense magma plug (Saubin et al. 2016). Neither of these models is consistent with the observed lithic content of internal tuffisites (Heap et al. 2019). Flow banding in the central rhyolite lava at the MCV and the granulation at the margins of lava plugs elsewhere demonstrate that the lava plug is pushed upwards and out of the vent (Pallister et al. 2013;Gaunt et al. 2014). Tuffisites in the vent center will therefore also likely be deformed (Tuffen et al. 2003), changing their planar geometry, and obscuring their relationship with particle sources. For this reason, as suggested in "Formation of tuffisites," we propose that it is consistent with our genetic model to assume that what appear to be internal tuffisites may be consistent with flow-deformed external tuffisites. These fractures within conduit-blocking magma, assembled by the welding of pyroclastic material, provided pathways for pyroclastic material produced by continued fragmentation at depth to reach the surface during hybrid explosive-effusive activity (Wadsworth et al. 2020Schipper et al. 2021).

A model for the formation of the Mule Creek vent
It follows from our interpretation that the material at the conduit center is relatively younger than the material forming the conduit margins, and as such, each conduit unit records a different period of conduit evolution. Tuffisites are found within every unit of the MCV, formed at different stages of conduit evolution and at a variety of scales, with very few cross-cutting the entire conduit width. This indicates that there was a continuous but pulsating supply of pyroclastic material to feed tuffisite formation during vent evolution.

Conduit opening and the erosive formation of the vent structure
The propagation of fractures by pressurized gas-pyroclast mixtures through the overlying andesite created pathways for fluid flow (Fig. 12a and animation -Online Resource 2). A branched network of fractures formed, exploiting the weakest pathways for fracture opening (e.g., bed/unit contacts, joints, and faults). Where this propagating fracture network separated blocks from the surrounding country rock, these could become entrained into the ascending fluid, providing a source of the lithic clasts. Variations in fluid velocity, driven by fluctuating fluid pressure, would have caused pyroclastic material and country rock fragments to be deposited within the opening fractures (Fig. 12b). Through time, the network of fractures was refined into a major pathway with a simpler shape, forming an evolving conduit. Some abandoned pyroclast-filled fractures may be preserved, forming tuffisites that cut through the country rock adjacent to the major pathway.

The onset of vent-filling emplacement of the pyroclastic breccia
Once a vent is opened the initial activity is likely to be purely erosive and would therefore not be recorded in any vent-filling units. The pyroclastic breccia unit was the first within-vent unit emplaced during conduit formation. A relatively high proportion of country rock fragments records the quarrying and filling of space with pyroclastic material and fragments of country rock, together with pyroclastic material (Fig. 4). The first subunits deposited are subhorizontal, as the quarried-out area is filled with pyroclastic material, but repeated erosion and deposition produces later subunits that dip more steeply towards the conduit center (Fig. 12c). The permeable pyroclastic breccia filling the evolving conduit would allow for lateral outgassing from the conduit, with erosion, deposition and slumping causing the fluid pathway to migrate and seal through time. Ash-sized particles transported into the pore space of the surrounding pyroclastic breccia by outgassing can be deposited to form what we term 'diffuse tuffisites' (Fig. 12d).
At Cordón Caulle, Chile in 2011 and Chaitén, Chile in 2008, the initial purely explosive phase of activity lasted about ten days (Lara 2009;SERNAGEOMIN-OVDAS 2011). If the Mule Creek eruption had a similar timescale, the vent-filling breccia would be deposited over a similar period as the purely explosive phase waned and conduit occlusion prevailed, before the dense lava was produced and erupted.

Vent wall steepening and localization: a transition from unwelded pyroclastic breccia to welded vitrophyre breccia
The sustained input of fragmented material from depth would supply material to become trapped within the conduit. The continued deposition of hot material within the conduit and the increasing volume of the accreted deposits would have inhibited cooling of the conduit fill, enabling ever-greater welding of the surrounding pyroclastic breccia subunits (Kolzenburg et al. 2019).
Welding increased the cohesion of the pyroclastic breccia, stabilizing the conduit walls and preventing the overturning of the deposited material. The increased stability would also allow tuffisites hosted within the pyroclastic breccia to become larger and more extensive. As the permeability of the pyroclastic breccia decreased, outgassing would have been constrained to within the central portion of the conduit and promoted overpressure development, especially if the conduit diameter were to constrict (Castro et al. 2016) due to pyroclastic breccia emplacement at its walls. The gradual heating of the conduit walls during pyroclastic breccia emplacement would have eventually allowed fine-grained ash particles to stick to the conduit margins, allowing a transition to the formation of vitrophyre breccia (Fig. 12e).
The presence of the glassy contact within the pyroclastic breccia (Fig. 8) suggests that a sufficient temperature for welding at the conduit margins was reached before the start of vitrophyre breccia and vitrophyre formation. Whether the transition from pyroclastic breccia to vitrophyre breccia can occur will depend on multiple factors, such as the eruption explosivity and the grain size of material, where higher explosion intensity results in finer grain sizes (Kueppers et al. 2006;Fowler & Scheu 2016), thus facilitating more efficient welding. The shielding of the surrounding country rock by the pyroclastic breccia would reduce the number of lithics available, and the increased cohesion of the pyroclastic breccia would prevent overturning and reworking (Schipper et al. 2021). A drop in explosivity would also reduce the amount of quarrying that could occur, leaving a supply of ash-sized particles for vitrophyre breccia and vitrophyre formation that contained fewer lithic clasts (Fig. 12f). The vitrophyre breccia and vitrophyre were nearly entirely formed by the welding of ash-sized particles, with a few rare lithic clasts captured by the accreting material. Coherent vitrophyre, rather than vitrophyre breccia, would have been formed once the explosivity had dropped sufficiently to prevent the repeated fracturing of the accreting material.
Increased overpressure in the constricting conduit opened fractures that would become tuffisites within the vitrophyre and vitrophyre breccia, now acting as a passive and effectively static 'country rock'. These tuffisites may have crosscut only the vitrophyre or vitrophyre breccia, but given greater overpressure the fractures may have propagated further to intersect the pyroclastic breccia or surrounding country rock (Fig. 12e).

Conduit sealing and the formation of the innermost vent units
Fine-grained, crystal-poor material that sticks to conduit margins can weld to form a dense glass (Gardner et al. 2017), preserved here as vitrophyre (Figs. 5c and 12e). The pressure-driven fracturing of this dense glassy vitrophyre produced the in situ jigsaw-fit clasts that form the vitrophyre breccia, separated by tuffisites. As the eruption waned further and erosive quarrying decreased, the accreting vitrophyre at the conduit margin was no longer fragmented and ejected, and merely damaged, until eventually that damage ceased. If the filling of the conduit were promoted by the welding of pyroclastic material and the entire silicic magma column were expected to be wholly fragmented at depth, the central conduit-plugging rhyolite lava would share a similar origin (Wadsworth et al. 2020). The continued accretion and welding of pyroclastic material would have eventually sealed part of or all of the vent, leading to the onset of welded-lava effusion to produce the flow-banded rhyolite lava at the vent center. A body of magma assembled at shallow depth could then be fractured to produce a network of permeable pathways along which pyroclastic material from below could be transported and deposited to form tuffisites (Wadsworth et al. 2020).

Similarities between conduits and tuffisites
The initiation of an eruption can be considered as the continued propagation, expansion, and connection of tuffisites that evolve into a conduit. In the shallow subsurface (< 2-km depth) the silicic magma column is likely to be fragmented (Wadsworth et al. 2020). The deposition of clastic material within the conduit during the waning phase of an eruption means that preserved, dissected conduits will qualitatively resemble a tuffisite, as each is a pathway for pyroclastic discharge. The characteristics and overall structure of the MCV are remarkably similar to those of the tuffisites intersecting the vent units, albeit on a larger scale. Processes such as repeated erosion and deposition of material, injections, and welding, occur not only on a small scale within tuffisites but also on a much larger scale within the vent units (Schipper et al. 2021).
Once repeated injections of material into a tuffisite allow it to reach a sufficient thickness, the tuffisite centre becomes hot enough for welding and viscous deformation to occur (Unwin et al. 2021). The same process at the conduit-scale will allow the pyroclastic breccia to weld and strengthen, changing the conduit pathway through time. Conduits are much larger than tuffisites and so will generally reach greater temperatures and cool more slowly, allowing more complete welding to occur (Kolzenburg et al. 2019). In a conduit where the margins have become sufficiently hot, ash particles can stick to and coat the conduit margins to form a vitrophyre unit (Gardner et al. 2017;Wadsworth et al. 2020). Similarly, sticking and coating of tuffisite vein walls by fine-grained ash can create dense veneers (Farquharson et al. 2022), which will have low permeability relative to the surrounding country rock .
However, there are many important differences between tuffisites and conduits that control their evolution through time. While tuffisites are the initial stage of conduit evolution, the greater dimensions of a conduit allow for more mixing and overturn of material, preventing the formation and preservation of the complex internal structures seen within tuffisites (Valentine & Cole 2021). The shape of the vent is likely to create deposits that are more prone to slumping than within tuffisites, controlling the structure of the vent produced. The pressure gradients driving gas and particle flow within the conduit will likely be lower than within a tuffisite, due to the greater dimensions of the conduit, especially during waning activity, controlling the ability for pyroclasts to be re-entrained into the flow and for internal fracturing to occur (Unwin et al. 2021). The greater permeability of the conduit margin than tuffisite walls will also help to reduce the pressure gradient within the conduit by allowing for lateral gas escape (Heiken et al. 1988), lowering the flow velocity within the conduit. To inject material into a preexisting host fracture to form a tuffisite, the fluid pressure must be great enough to displace the surrounding material (Unwin et al. 2021), while in a conduit continuous mixing, overturning and soft sediment deformation allows for the emplacement of new material at lower fluid pressures.

Flux through tuffisites during vent evolution
Tuffisites have been suggested to act ubiquitously as permeable pathways along which outgassing can occur, perhaps defusing gas pressure from a conduit (Farquharson et al. 2017;Kolzenburg et al. 2019). Therefore, tuffisite formation is thought to be key to understanding explosive-effusive transitions. By contrast, here, we show that the initial hostrock for 'early' tuffisites is a highly permeable vent-lining breccia (Online Resource 3) that would allow highly efficient outgassing (Eichelberger et al. 1986). 'Early' tuffisites in this host rock are not necessarily only hydrofracture driven and instead can also involve invasion of the pyroclastic debris with ash-and clast-laden fluids, which deposit as they percolate (White & Ross 2011). Such invasion could create deposits that appear superficially tuffisitic, but actually sealed up outgassing routes in pre-existing deposits, without any initial high-permeability opening phase.
Once the conduit walls have been heated sufficiently for welding of fine material at the margins to form the vitrophyre and vitrophyre breccia, these lower permeability units (relative to the pyroclastic breccia) will seal the margins of the conduit, constraining fluid flow and increasing the fluid pressure within the active portion of the conduit. If a sufficient fluid pressure can be achieved to fracture the vitrophyre breccia, tuffisites can form a pathway through the less permeable material and into the more permeable pyroclastic breccia beyond, where outgassing can occur more effectively (Online Resource 3). Tuffisites therefore become necessary for lateral outgassing to occur in this partially or wholly sealed phase of the conduit evolution. Fracturing the dense vitrophyre breccia will require a greater fluid pressure than invading a pathway through the weaker, less compacted pyroclastic breccia, suggesting that once the vitrophyre unit begins to coat the conduit margins a higher fluid pressure can be achieved within the conduit. We propose that this corresponds to the formation of a dense lava plug, accompanied by repetitive Vulcanian explosions. If the dense glassy plug extends to the surface then hybrid activity could occur, as observed at Chaitén in 2008 (transitional phase 2; Pallister et al. 2013) and Cordón Caulle from June 15 th 2011 (Schipper et al. 2013(Schipper et al. , 2021Wadsworth et al. 2022).
Fine ash-sized particles, whether injected and deposited within a fracture or deposited within the pore space of coarser-grained material, will rapidly weld, sealing fluid flow pathways and reducing further outgassing. The permeability of these fluid pathways will therefore only be transient. The time-dependent porosity of a welding matrix with time, ϕ(t), can be estimated using the 'vented bubble model'  where ϕ i is the initial porosity, ΔP is the difference in pressure of the gas phase and stress acting on the particles involved in the welding, μ is the viscosity of the hot particles, Γ is the interfacial tension between the gas and the particle, and 〈a i 〉 is the mean size of the pore spaces between the welding particles (related to the particle size; Wadsworth et al. 2016). When ΔP → 0, Eq. (1) can be approximated by ϕ ≈ ϕ i exp(−3t/(2λ)) where λ = ζμ〈a i 〉/Γ is the welding timescale in the ΔP = 0 regime (Wadsworth et al. 2014(Wadsworth et al. , 2016. Here, ζ is a correction factor that accounts for the polydispersivity of the particle size distribution .
Referring to Eq. (1), ΔP ≈ 0 is justified for the case where the gas and particles are in pressure equilibrium (i.e., the effective pressure is zero). At Mule Creek, this is likely to be the case in the conduit itself (i.e., during vitrophyre formation by welding) or in the early stages of tuffisite formation and welding. As tuffisites seal and gas pressure decays, there may be an increase in the pressure acting to compress the assemblage of particles relative to the gas pressure between the particles, causing ΔP to rise to ΔP > 0, invalidating this simple approximation. Nevertheless, if we use the ΔP ≈ 0 assumption, we can give an upper-bound on λ. In order to compute λ, we must convert R i toa i and ζ. To do this, we use the model provided by Wadsworth et al. (2017), which, using the two distributions given in Fig. 10, yields 〈a i 〉 = 2.7 ×10 −7 m with ζ = 17.59 (Fig. 10a) and 〈a i 〉 = 7.33 × 10 −6 m with ζ = 7.33 (Fig. 10c), respectively. Using Γ = 0.22 N/m (Gardner & Ketcham, 2011), 5 × 10 7 < μ < 1.5 × 10 9 (using Hess & Dingwell, 1996; with a water content of 0.8 wt.% and for 700 and 800 °C from Stasiuk et al., 1996), we find that 18 <λ < 540 min for the 2 μm particles in the tuffisites within the pyroclastic breccia, or 37 <λ < 1107 min for the 7 μm particles in the tuffisites within the vitrophyre breccia. These timescales are far shorter than conduit cooling timescales (Kolzenburg et al. 2019), especially when thermally buffered by an ongoing bypassing eruption, with hot material flowing past the conduit walls, and well within the timescales associated with ongoing silicic eruptions (e.g. Schipper et al. 2013), lava assembly, and effusion (Wadsworth et al. 2020). This process will act more efficiently in early, high energy fragmentation events due to the correlation of grain size and fragmentation energy (Kueppers et al. 2006;Fowler & Scheu 2016) and slow down during the later waning stages of an eruption-further enhancing the discrepancy in the need for pressure release during the high energy stages and a more efficient outgassing and depressurization during later, waning phases. The welding timescales in both cases explored here suggest that the welding material could efficiently lower the permeability of any outgassing pathways, potentially sealing them completely. In the pyroclastic breccia, welding of fine-grained intruded material could locally reduce the permeability, but with fluid flow pathways through the unconsolidated and permeable breccia relatively unconstrained, this reduction in permeability would likely not have a great effect on the bulk permeability of the unit. The accretion of the vitrophyre and vitrophyre breccia could quickly seal the conduit margins, rapidly decreasing their permeability. With gas unable to escape laterally through the impermeable vitrophyre and vitrophyre breccia, regular and repeated fracturing would be required to continually produce new lateral outgassing pathways to replace those sealed by welding.
As tuffisites close, gas pressure drops and ΔP will rise, and therefore, our estimate of the rate of porosity change is a conservative estimate (c.f. Eq. (1)). Similarly, if the gas pressure rises in tuffisites or between the welding vitrophyre material, Eq. (1) predicts that ΔP will drop and can become negative, causing dϕ/dt to be positive, and therefore porosity to increase with time as the gas pushes the sintering system apart again. Fractured masses of sintered particles within tuffisites in the vitrophyre breccia (Fig. 10d) suggest that this process does occur, perhaps due to the sealing or nearby permeable pathways locally increasing gas pressure (Unwin et al. 2021).

Future challenges
Tuffisite veins are pervasive throughout the width of the MCV, from the vitrophyre into the country rock, suggesting that they might play a role in influencing conduit dynamics throughout vent evolution. Characterization of the grain sizes, porosity, and permeability of tuffisites would determine how significant their role in outgassing might be, and whether tuffisites could carry a sufficient flux to defuse pressure within the evolving conduit. Further investigation of conduit walls would help constrain how their permeability and erodibility might vary through time, controlling the lateral outgassing flux through conduit walls and the conduit morphology throughout its evolution. This evolution of the permeability and erodibility of the conduit walls may also influence flow within the conduit, perhaps acting as a control on the steadiness of conduit flow through time. Detailed analysis of textures and H 2 O contents across the margins of vents similar to Mule Creek may help to constrain degassing mechanisms. More closely integrating the spatial context provided by analysis at ancient dissected systems with the fragmental but fresh record preserved in pyroclasts from recent silicic eruptions is then needed to further improve our understanding of the processes controlling the explosive-effusive transition. Detailed textural analysis of the different clasts forming the vitrophyre breccia at Mule Creek and similar vents may also help to confirm whether this unit was brecciated in situ or formed of vitrophyre clasts fractured away from the vent margin and redeposited at a shallower depth. Similar structures to the those seen within the Mule Creek vent are seen in basaltic diatremes, and applying and adapting our understanding of vent formation to less silica-rich systems poses an additional challenge.

Conclusions
The structure of the Mule Creek vent is complex, with the characteristics of the different vent-filling units controlling its morphology during vent evolution. The pyroclastic breccia contains many erosive surfaces, reflecting the repeated deposition, erosion, and overturning of pyroclastic material until sintering and compaction increased its cohesion and more stable conduit walls could form. Heating of the conduit walls and waning explosivity allowed the accretion of ash-sized particles to form the vitrophyre breccia unit, with glassy material repeatedly fractured and re-healed until explosivity waned sufficiently for a more coherent vitrophyre to accrete. The accretion and sintering of pyroclastic material eventually blocked the vent, leading to the eventual effusion of sintered lava and production of flow-banded rhyolite lava at the vent center.
The tuffisites that intersect every vent unit have complex internal structures, reflecting repeated periods of erosion and deposition within the fractures. Diffuse tuffisites in the pyroclastic breccia record lateral outgassing through the high permeability breccia lining the conduit margins. Repeated periods of erosion and deposition, driven by fluctuations in the conduit fluid pressure, caused the conduit width to vary through time. The accretion of ash-sized particles to form the vitrophyre and vitrophyre breccia would have effectively narrowed the active conduit, sealing the conduit margins and restricting lateral outgassing to the permeable pathways created by tuffisites.