Relations between thermal history and secondary structures of ignimbrites exclusive of rheomorphism

Columnar joints, devitrification spots, lithophysal cavities, and horizontal joints are characteristic secondary structures of ignimbrites. Columnar joints and formation of cavities by gas have been modeled in fluid basalt lavas but not in tuffs. Devitrification of glass has been replicated under laboratory conditions, which are only approximately similar to those in cooling ash deposits. It is difficult to explicitly model the formation of horizontal joints in tuffs because neither initial nor boundary conditions at their initiation can be precisely specified. The stratigraphic distribution of these features is described herein for some tuffs whose density profiles are modeled by the compaction program of Riehle et al. (1995, 2010). Once the density profile has been replicated by this model, histories of temperature, compaction, and gas pressure can be hindcast. The distribution of features can then be discussed on a phenomenological basis with reference to the model histories. Profiles have been selected for their illustration of times to devitrification, to vertical columnar joints, to development of lithophysal cavities, and to horizontal jointing, and are presented in that order.

The compaction model of Riehle et al. (1995, 2010) is used to calculate temperatures, gas pressures, and progression of compaction of rhyolitic ash-flow deposits; readers are referred to the earlier papers for details. In brief, the method calculates a model density profile for assumed values of initial thickness, temperature at emplacement, and, for complex cooling units, the cooling interval between ash-flow deposits. After comparing the model profile to the measured profile, the user adjusts input variables and calculates a second run. A close fit of model to measured densities typically can be achieved with five iterations. Sample model runs (Fig. 1) illustrate the model’s workings. Model profiles are sensitive, if not accurate, to ±2 °C and ±0.2 m thickness. Correlations among nearby profiles of the same deposit having different total thickness suggest that cooling intervals are reproducible to about ±50%, e.g., 10 ± 5 days; this seemingly large inaccuracy of cooling intervals is perhaps not surprising, considering that a small amount of rainfall can have a large effect on the hindcast cooling interval. There is no independent basis for inferring rainfall, so for consistency, solely radiative plus convective cooling of the upper surface is assumed for all model deposits.

Gas pressure is an important variable in this report; therefore, the method for calculating pressure is explained further. A lithostatic pressure gradient is assumed at deposition by an ash flow (Miller, 1990; Riehle et al., 1995). If greater than lithostatic, gas escapes in the overriding cloud, and if less, a plug of ash settles to the ground. Immediately following deposition, gas outfluxes mainly by diffusion upwards to atmospheric pressure. A small amount fluxes downward owing to pressure reduction by conductive cooling at the base. As pressure falls below load pressure, compaction under lithostatic load begins; residual gas pressure increases as pore space decreases.

Results for two 10-m-thick model deposits emplaced 20 days apart (Fig. 1A) show that gas pressures decay to one atmosphere in less than a month throughout the entire deposit. The lower deposit has already completed ∼80% of its final compaction by the time the upper deposit is emplaced 20 days later; the upper deposit completes much of its compaction within a month after that. Subsequently, compaction mainly “fills in” the density low at the boundary of the deposits, where temperatures remain highest for the longest period.

Air or gas permeability of tuff samples measured at different porosities is shown in Figure 1C. Researchers were aware of complications introduced by fractures and attempted to measure only matrix permeability. Nonetheless, the range of values is striking. The program uses the log-linear curve for permeability as a function of porosity shown by the green line in Figure 1C. To constrain the effect of the large variation of permeability, there is an option to calculate pressure and compaction values using a value of permeability that is 1/10 that of the green line (the dashed line in Fig. 1C). This lower function is well below the likely range of permeability for naturally occurring tuffs; it serves to demonstrate the potential effect of lower permeability.

Compaction results for 50-m model deposits (Fig. 1B) are generally similar to the 10-m deposits except that compaction requires two to four years for completion, by which time the density reversal at the boundary is nearly obliterated due to the greater lithostatic load. Even for these thicker deposits, outfluxing reduces gas pressures to <1.04 atmospheres within two months including at the bottom of the deposit. If 1/10 lower permeability is assumed, however, then residual gas pressure of as much as 2 atmospheres remains at the bottom of the deposit after two months. An unanticipated result of lower permeability is that final densities are as much as 10% higher in the zone of residual pressures at the bottom. Although pore pressure reduces lithostatic load, net load is not reduced to zero; instead, the slightly higher water pressure lowers the effective viscosity of the glass shards, which increases the rate of compaction.

In summary, the model should adequately account for the behavior of initial gas pressures in deposits up to 50 m thick at emplacement. Thicker deposits can also be modeled provided there are density reversals as evidence for cooling breaks when episodic degassing occurred from the tops of succeeding deposits. Any errors in calculated density at the base of thick deposits should be small, because thick deposits compact to zero porosity regardless of the permeability function that is employed. There is no attempt to model behavior of gas generated by devitrification two to three years after deposition.

Profiles herein are labeled by their figure number. Where more than one profile is shown in a figure, a letter is included. Individual sample labels follow a hyphen after the profile label—for example, 2B-D is the profile shown in Figure 2B and D is a particular sample.

The Rattlesnake Tuff at Highway 74, John Day, Oregon (site locations are in  Appendix I) is partly devitrified in its upper third (Riehle et al., 2010, fig. 7), while a second occurrence of Rattlesnake Tuff in John Day is completely vitric throughout (photomicrograph jpeg files available from the author). Measured and model profiles at each of these sites (Fig. 2) show that the Highway 74 deposit remained above 600 °C for three years in its interior, whereas the second, thinner outcrop has no zone that remained >600 °C for two or more years.

Columnar joints penetrate from top to bottom of the Aravaipa Tuff at the north headland of Whitewash Canyon in SE Arizona (Fig. 3). The density profile of the Aravaipa Tuff at the nearby south headland of Whitewash Canyon, where a profile was easily measured, shows one major and several minor reversals (Fig. 4A). A density profile measured by Krieger (1979) is included. Her sample spacing is too great to show details of density reversals, but a major cooling break (∼115 ± 10 days) occurs just below the center of each profile, supporting the interpretation that these density reversals represent correlatable events. Model compaction results indicate that all deposits were emplaced below 750 °C and some below 700 °C.

Compaction and thermal histories of the Aravaipa Tuff at Whitewash Canyon (Figs. 4B and 4C) show that density maxima developed quickly. The bottom four deposits attained their maximum density at 15 m by the end of the major cooling break at 22 m. Similarly, in the top half of the deposit, density maxima at centers of individual deposits developed within six months after the last deposit. Over the next 18 months, compaction essentially “filled in” density minima but did not much change density maxima. This means that low-permeability horizons were present early after deposition.

A profile of the Rattlesnake Tuff at Highway 395 near Burns, Oregon, shows well-developed cavities and spherulites, and it is discussed in detail in the Lithophysal Cavities section. The profile has been modeled as comprising six preserved ash flows ranging from 731 to 786 °C emplacement temperatures (Fig. 5); two additional deposits, covered and/or now partly eroded, are inferred by correlation to nearby sections by Riehle et al. (2010, fig. 11). The least devitrified portions at the top are missing or covered. The model temperature history for this section (Fig. 5B) shows that the deposit remained above 600 °C for two to three years between 7 and 18 m above the base.

Many of the deposits examined for this study show complex horizontal jointing or are locally densely fractured. Such features are well displayed by the Peach Springs Tuff (Young and Brennan, 1974; Buesch and Valentine, 1986) near the intersection of Route 66 and I40, one km S of Kingman, Arizona. A composite of four closely adjacent sections is described in detail in the Horizontal Fractures section. The Peach Springs Tuff has multiple density reversals above a maximum density about one-quarter of the thickness above its base (Fig. 6A). A separate profile, 3 km away, has been eroded in its upper part. The preserved part is closely similar to the lower part of the Route 66 profile; it is included here to support the inference that density reversals are amenable to stratigraphic correlation rather than random variation.

Density modeling shows a complex cooling unit comprising 14 individual flows, ranging from 8 to 27 m in initial thickness and 664 to 780 °C emplacement temperature. Cooling intervals range from 13 to 140 days. Several density reversals show significant density lows relative to immediately adjacent density highs (asterisks, Fig. 6). Such extreme reversals are unusual; density lows typically make more gentle reversals and differ by only 0.1–0.2 g/cm³ relative to adjacent density highs (for example, Fig. 5). This aspect of the Peach Springs Tuff profile is discussed in the Horizontal Fractures section. The temperature history (Fig. 6B) shows two intervals, at 25–50 m height and 70–90 m height, that remained well above 600 °C for >4 years, separated by a slightly lower temperature interval reflective of the 140-day cooling break.

The Rattlesnake Tuff shows wide variation in development of secondary structures over short horizontal distances (Streck and Grunder, 1995). One such locality near Paulina, Oregon, is on federal land 300 m W of Camp Creek Road and is included here to evaluate the cause of lateral variability. The base of the Rattlesnake Tuff is not exposed under talus at the foot of the section. The top of the talus apron descends 15–20 m over a 150-m interval from the thinner profile to the thicker one. A density profile was measured on foot in the thinner part of the deposit, and a professional climber collected samples while rappelling down the face of the cliff in the thicker part of the deposit. The density decreases to ∼1.60 g/cm³ at the lowest exposure of basal vitrophyre (Fig. 7) in each profile. By comparison with the Burns profile (Fig. 5), where the base is exposed beneath vitrophyre, each profile at Camp Creek Road must be exposed nearly to its base.

The two profiles (Fig. 7) match in detail, supporting the concept that detailed density reversals are most likely boundaries between individual ash-flow deposits. Each model profile comprises an early deposit (736 °C), a 769 °C deposit, a 749 °C deposit, and then three hotter deposits (780–793 °C). Three cooler deposits follow (710–750 °C), and then a major cooling break of 94 ± 16 days. Two cooler, uppermost deposits are partially exposed in the thicker section. Only one sample remains to suggest that these deposits were originally present in the thinner section as well. A third profile from Nicoll Ridge Road, colinear with the Camp Creek Road site and the inferred source SE of Burns, confirms that two uppermost deposits followed a major cooling break. Erosion has apparently removed these younger, more porous deposits from the thinner Camp Creek profile; their former presence is assumed in its model, which thereby yields a better match to the thicker profile. Despite the variation in thickness between the two sites, model results show that initial emplacement temperatures and cooling intervals are approximately similar.

Compaction under lithostatic load and devitrification—crystallization below the solidus (e.g., Lofgren, 1971)—are two of the most important processes governing structures and textures of ignimbrites. Devitrification converts ductile, vitric ash shards to finely granular, brittle rock, and in turn is associated with, or even responsible for, spots, lithophysal cavities, and dense fractures. This important feature is the first of the secondary structures to be discussed.

Devitrification textures typically overprint compaction foliation (Smith, 1960), and therefore compaction must have ended before the onset of devitrification. Indeed, this is required by the fact that compaction occurs while shards and pumice are still ductile. Once devitrification proceeds to some stage as yet unquantified, the consequential increase of ash rigidity prematurely arrests compaction relative to that of compaction arrested solely by conductive cooling. There may be some compaction after devitrification by rigid-particle packing; however, Sheridan (1970), Wilson and Hildreth (2003), and Riehle et al. (2010) have all shown that there is no measurable discontinuity in the density profile of a compacted tuff across the contact between vitric and devitrified tuff. This confirms that compaction precedes devitrification and also means that the slight increase in density from glass (2.35 g/cm³) to quartz-feldspar devitrification assemblages (2.65 g/cm³) must be accompanied by 10% increase in porosity. Textural evidence for such porosity increase is shown later.

Temperature, cooling history, and water vapor are known from experimental studies to be the main factors controlling devitrification of rhyolitic glass (Marshall, 1961; Yagi, 1962; Lofgren, 1970). Dissolved species (Lofgren, 1971) and presence of nucleation sites (Prakash, 1968; Lofgren, 2006) are other factors. A key question about devitrification of ignimbrites is: how long after deposition does devitrification begin? Lofgren (1970) reports complete devitrification of 5-mm glass cylinders in only 4–6 days at 400 °C, but 1 kb PH₂O was required; this is an unrealistically high water pressure for ash-flow deposits <100 m thick at the Earth’s surface. Friedman and Long (1984; Fig. 2) estimate that a million years are needed for the devitrification of rhyolitic glass having a few tenths of a percent of H₂O at 600 °C, typical of rhyolitic glass at surface conditions. Devitrification can occur below 500 °C, but progressively longer times are required at lower temperatures and below some temperature, nucleation and growth cease.

The two Rattlesnake Tuff profiles from John Day (Fig. 2) help to estimate the minimum time required for the onset of devitrification in tuff deposits. As a rough rule of thumb, a minimum of two years above 600 °C is adopted here for the onset of devitrification in deposits <50 m thick. Devitrification is a rate process; thus, a longer period at slightly lower temperatures is theoretically equivalent.

The Aravaipa Tuff profile has a complex distribution of devitrification textures. Three thin-section samples are entirely vitric; the other three (Fig. 4A, 4A-B, D, and F) show incipient devitrification. Samples 4A-D and F are below the major, 115-day cooling break and had already been in place for nearly a year at the emplacement of the top deposit; their relevant temperature plot is more nearly that for “six months after the last deposit,” which is just above 600 °C. The seemingly random distribution of incipient devitrification reflects the complicating effects of temperature reversals at multiple ash-flow boundaries that persisted for the first year of cooling.

The occurrences of devitrification in the profiles of the Peach Springs Tuff and the Rattlesnake Tuff at Burns and on Camp Creek Road are shown on their respective figures and are discussed in sections on “Lithophysal Cavities” and “Horizontal Fractures.”

Vapor-phase mineralization in open spaces is not explicitly modeled here but is related to devitrification and so is mentioned briefly in this section.

Vapor-phase mineralization is spatially related to zones of devitrification. Ross and Smith (1961, p. 28) state deposits “… may have 2 zones of vapor-phase minerals, 1 above and 1 below the densely welded zone which may be either completely or partially devitrified. … the lower zone is typically thinner than the upper … vapors … tended to move upward but were trapped below the densely welded zone … .” Sheridan (1970) finds that fossil fumaroles in the Bishop Tuff overlie paleotopographic basins where the tuff is thicker, are more indurated there by vapor-phase mineralization than in other areas, and regardless of degree of compaction of the underlying tuff, are absent where the underlying tuff remains vitric. Sheridan’s observations imply a second pulse of vapors, after the initial vapors responsible for fluidization of the ash flow have outfluxed, and clearly indicate the origin of this second pulse in devitrification. Agarwal (1989) studied devitrification of industrial glasses and reported vesicles that formed above 1100 °C by “exsolution of gas” consequent to devitrification. Swanson et al. (1989, p. 170) studied subtle textural differences between extrusive and intrusive portions of rhyolitic Obsidian Dome, California, and conclude that “cavities [in devitrified obsidian] are formed by some combination of vapor exsolution during crystallization [of anhydrous mineral assemblages] … .”

Modeling here shows that gas present at deposition quickly outfluxes due to a high initial permeability in deposits <50 m thick; if this were not so, compaction could not proceed to zero porosity before arrest by devitrification. For the rest of this report, it is assumed that a second pulse of gas pressure ensues from devitrification of hydrous glass. Even though only tenths of a percent of vapor may remain dissolved in the glass after deposition, tens of meters of devitrifying deposit can yield a significant flux of vapor, which locally can cause fumarolic alteration, expansion of gas vesicles (lithophysae), or vapor-phase deposition in the overlying deposit.

The Highway 74 Rattlesnake Tuff deposit shows essentially no vapor-phase deposition (Riehle et al., 2010, Fig. 7), probably due to a combination of proximity of the zone of devitrification to the upper surface of the deposit—so vapors readily escaped—and its small thickness. The Rattlesnake Tuff on Nicoll Ridge Road shows some vapor-phase deposition (Fig. 8) throughout most of the deposit; the greatest deposition appears in a low-density zone below the uppermost two ash flows (Fig. 7). The Nicoll Ridge deposit is at least 15 m thicker than the Highway 74 deposit and has a devitrified zone that is more than twice as thick as the Highway 74 deposit. Peach Springs Tuff near Kingman, Arizona, has variable vapor-phase deposition above the zone of devitrification, which is ∼60 m thick and extends well below the zone of vapor-phase deposition (Fig. 6).

Three general origins of vertical joints in tuffs have been previously identified. Jointing in response to underlying topography or to regional stress depends on variables particular to individual sites and is beyond the scope of this paper. However, columnar jointing in response to cooling contraction is intrinsic to all tuff deposits that were deposited hot enough to compact, and it is considered here. A brief review of the three types of vertical joints is merited in order to distinguish among them.

In the Bandelier Tuff of New Mexico, Walters (1996) finds two orthogonal joint sets that he interprets as tectonic; he interprets three other joint sets intersecting at 120° as cooling. Wohletz (2006) finds that intensity and to some extent orientation of fractures in the Bandelier Tuff vary in response to their proximity to buried fault zones. These fractures typically lack evidence for shear movement such as slickensides, indicating that they are Mode 1 joints. The absence of colored margins caused by hot gas is a good indicator of their late formation, although even early-formed rimmed joints may be preferentially oriented by regional stresses.

Sheridan (1970) shows that vertically extensive joints in the Bishop Tuff were the locus of fumarolic activity early during cooling. Some such joints form conjugate pairs oriented at Mohr angles to the local ground-surface slope, and Sheridan infers that these joints formed as tension fractures accompanying down-dip creep into a basin after the deposit had developed relief by differential compaction. That is, where the deposit was thicker in a paleobasin, its compaction was also greater and a surface basin developed that mirrors the underlying paleobasin. It is this surface slope that provides a lateral free face toward which the deposit slowly moves.

Rheomorphic deformation of a Miocene ignimbrite on Gran Canaria occurred on slopes of 6°–8° following compaction (Kobberger and Schminke, 1999). The tuff deposit in the Valley of Ten Thousand Smokes (Hildreth, 1983, 1987) has topographic benches and cross-valley fractures that formed shortly after its emplacement. Present ground-surface slopes in the Valley of Ten Thousand Smokes are only 2°–3°. A pull-apart of the Huckleberry Ridge Tuff in eastern Idaho was triggered by diapiric doming of the underlying, water-saturated lake and fluvial deposits (Embree and Hoggan, 1999); the regional ground slope in the headwall area (“Hog Hollow”) is uncertain, but inspection of the Newdale 7.5-min quadrangle map clearly indicates that it is low. In short, the mobility of tuffs after compaction but before significant cooling is remarkable, and secondary deformation ranging from joints to rheomorphic folds or gravity slides is feasible even on low slopes. Embree and Hoggan (1999) describe vertical zonation of deformation features in the Huckleberry Ridge Tuff. Cool brittle tuff near the surface rode along, whereas deformation was concentrated below a brittle-ductile transition at depth.

Vertical columnar jointing like that in basalt flows is typical in welded or indurated portions of tuffs (Ross and Smith, 1961). Columns may be square or rectangular in cross section, unlike the hexagonal columns in basalt lavas. Departures from vertical such as rosettes (“fan jointing”) are probably related to departures from horizontal isotherms such as adjacent to fumaroles or near lateral contacts in paleogullies. Sheridan (1970) interprets shallow vertical joints in the Bishop Tuff that resemble orthogonally intersecting joints in permafrost as having formed by cooling contraction. Engineers have long termed such joints “thermoelastic contraction features” consequent to shrinkage upon cooling. Study of rectangular patterned ground in permafrost ground (Lachenbruch, 1960a, 1960b, 1961) shows that fracture sets tend to have orthogonal intersections because the stress field is modified near an early-formed fracture, such that an approaching fracture changes its orientation to perpendicular. Thus orthogonal joints in cooled media must have slightly different ages and formed over a protracted period.

A combination of origins is reported for vertical joints in the Topopah Springs Tuff at Yucca Mountain, Nevada, where complex interaction among cooling tuff deposits, concurrent faulting, and local relief in the evolving Basin and Range Province occurred. Throckmorton and Verbeek (1995) identify three sets of joints that formed during cooling of the Topopah Springs Tuff and the overlying Tiva Canyon Tuff, based on relative ages of joints, their smoothness, and high-temperature alteration rims. Columnar cooling joints comprise mainly two sets of mutually perpendicular joints that, because they show a tendency to cluster in orientation, may have formed in stress fields reflecting underlying paleotopography similar to the Bishop Tuff. Tectonic joints formed later based on absence of high-temperature features and relative ages to the cooling joints; these joints have rough surfaces and show a shift in orientation with age that correlates with a known shift in the regional stress field. Throckmorton and Verbeek (1995) note that the cooling joints dominate in the upper part of the Tiva Canyon Tuff, while tectonic joints dominate lower in the formation, presumably reflecting proximity to tectonic stresses in underlying bedrock.

Columns in basalt flows are typically shown in textbooks as hexagonal, supposedly owing to parent cracks that bifurcated at 120° during propagation. Saliba and Jagla’s (2003) finite-element modeling of stress-strain in a cooling medium yielded columns having footprints that evolved during downward growth into ones that were not hexagonal but instead comprised polygons with different numbers of sides. Regardless, some columnar joints are hexagonal (Throckmorton and Verbeek, 1995), so thermoelastic joints in tuffs are not exclusively either orthogonal or hexagonal in plan. The best evidence for their origin is a concentration in the upper and lower parts of a deposit, planar to smoothly sinuous faces, rims, and their regular spacing (which likely changes with vertical position).

Lachenbruch (1960a, 1960b) shows that, for some conditions, a joint formed at the surface of a cooling body penetrates well below the layer under thermal stress (“overshoots”). This condition is owing to stress concentration at the sharp tip of a growing joint, possibly aided by kinetic energy of outward-moving joint faces (Lawn, 1993). DeGraff and Aydin (1993) elegantly model the growth of columns in basalt lava flows and show that growth increments are controlled by tip blunting in more ductile lava above the solidus temperature (980 °C for basalt). Fracture strength in brittle material decreases at higher temperatures, but fracture resistance—“toughness”—in ductile material increases at higher temperatures. A fracture begins when cooling contractile stress exceeds the fracture strength; stress concentration at the sharp tip may lead to overshooting, but growth is shortly halted in much hotter, more ductile lava. Upon cessation of growth, tip blunting by flow in ductile material in turn reduces stress concentration. Each new joint increment is initially delayed by blunting, but eventually there is enough cooling to reinitiate fracturing. Because fracture growth is rapid relative to progression of isotherms, each growth increment occurs at higher temperature and higher fracture toughness. This means that growth increments are progressively larger.

where σ_x is stress across a plane normal to an isothermal surface (here, horizontal). Material properties for tuffs vary with porosity, amount of devitrification, and temperature and are only approximately known. Young’s modulus E of compacted tuff at room temperature is 10⁴ megapascals (MPa), and Poisson ratio ν is 0.10 (Robertson, 1959; Keller, 1960; Lawn, 1993). Thermal expansion coefficient α at 300 °C is ∼2 × 10^–5 deg^–1; α and E increase by factors of 5–10 as porosity decreases from 40% to 5% (Ciancia and Heiken, 2006). Coefficients α and ν increase by a factor of 2–3 from 20 to 900 °C, while E decreases slightly (Peck and Minakami, 1968; Ryan and Sammis, 1981). The tensile strength σ_x of Topapah Springs Tuff (density 2.10–2.30 g/cm³) ranges from 1 to 12 MPa at room temperature (Ciancia and Heiken, 2006); it decreases with increasing porosity and is lower for vitric samples than for devitrified samples at the same porosity. By analogy with basalt (DeGraff and Aydin, 1993), the tensile strength of tuff likely decreases up to T_G and then increases. New measurements of physical properties at elevated temperatures and at various porosities, particularly of σ_x and E, are needed before precise values of ΔT can be calculated.

For compacted vitric tuff >T_G, σ_x is estimated to be ∼15 MPa, α ∼6 × 10^–5, E ∼8 × 10³, and ν ∼0.2 (see previous paragraph). Vertical fracturing should then occur for cooling ΔT of 25 °C (Equation 1). All deposits studied here cool by >25 °C before two years, so material properties of vitric tuff are appropriate. Granite at 315 °C (<<T_G) theoretically fractures when cooled by 20 °C (Smith et al., 1973), and basalt near its solidus temperature is estimated to fracture when cooled by 80 °C (Peck and Minakami, 1968); therefore, 25 °C seems approximately correct for low-porosity, vitric tuff. The T_G for rhyolitic glass having a few tenths of a percent H₂O is as high as 775–975 °C for timescales of minutes but decreases to as low as 500–600 °C for timescales of days (Zhang et al., 2003; Romine et al., 2012; see discussion of glass rheology in following section “Lithophysal Cavities…”). A single-value cooling threshold for initiation of fracturing is perhaps simplistic. However, as ΔT increases by cooling and σ_x increases, the tuff also approaches T_G and the fracture toughness decreases, aiding in initiation of fracturing.

The Aravaipa Tuff at the north headland of Whitewash Canyon is flat lying; some columnar joints extend upwards from the base and then terminate or merge (Fig. 3, arrows), while others terminate downwards or appear to extend continuously through the entire deposit. At the south headland, the Aravaipa Tuff has a structural dip of 12° (Fig. 9A). Other evidence of structural adjustment includes zones of brecciation and fractures partly filled by quartz veins (Fig. 9B). Fine, pervasive vertical striations are common (Fig. 9C). Faint intermediate-dipping fractures (Figs. 9C and 9D, arrows) are suggestive of splays accompanying shear on column faces (A. Aydin, 2013, personal commun.), which implies that the vertical striations are a type of slickenline. These structures likely formed during tilting at the south headland well after cooling.

One columnar joint shows two bands (Fig. 9D, brackets) that resemble hackle described by DeGraff and Aydin (1987). The vertical ribs in profile all show the same sense of curvature (Fig. 9E), a characteristic of fracture hackle. Each band is ∼30 cm high, at the upper end of the range of growth increments reported for basalt lava flows (DeGraff and Aydin, 1987). This indicates that joint growth was incremental like that of basalt flows. Aravaipa Tuff deposits were likely >T_G during early cooling over days prior to jointing (deposit ∼680 °C, T_G ∼500–600 °C); however, for stresses acting for ∼1 second during joint propagation, they would have been <T_G (T_G >750 °C), and the tuff was presumably weaker. Thus, there would have been overshooting until growth was halted by intersecting hotter tuff (lower ΔT and σ_x).

This means that, to a first approximation, fractures followed the inward progression of a 25-degree cooling isotherm. The progression of intervals that cool by 25 °C are marked on the temperature plots (Fig. 4C); the presence of so many cooling breaks means that columnar joints developed nearly simultaneously with progressive deposition and were all but complete within a month after the last deposit, at which time compaction was ∼90% complete as well. Jointing occurred inward from both the top and bottom of each individual deposit and accompanied upward accumulation of the deposit. Temperature gradients—which would have governed fracture increments—in the deposit interior at the passing of the 25-degree cooling isotherm were nearly identical, except at temperature rollovers (Fig. 4C), where they were zero. However, hackle is rare on Aravaipa Tuff columns, and the possibility of thicker growth increments at temperature rollovers in the deposit interior cannot be confirmed.

Lithophysal (Greek, “rock bubble”) cavity was first applied by von Richtofen (1860) to shells of volcanic glass having hollow interiors. Some were surrounded by radial crystal growths (spherulites) and others by concentric crystalline shells (orbicules; Johannsen, 1931). The origin was identified as due to gas expansion, the assumption being that expansion was triggered by pressure relief at eruption. Gas expansion in tuffs probably occurs instead by incremental free gas from devitrification aided by temperature rise consequent to latent heat. Streck and Grunder (1995) use lithophysa to refer to a devitrified patch; they distinguish a lithophysal cavity from a gas cavity (i.e, a cavity having no encircling lithophysa). All cavities examined here have at least an incipiently devitrified border, so the distinction between gas cavity and lithophysal cavity is not made. Spot was used extensively at Yucca Mountain to refer to a visible patch of crystallized tuff without a central cavity (Case and Buesch, 2004) and is equivalent to lithophysa. Spot is retained here for its ease of use as a nongeneric term; all spots examined microscopically are devitrified tuff.

Breitkreuz (2013) thoroughly reviews the history of thought about the origins of spherulites and lithophysal cavities in silicic lavas and tuffs and provides details of their petrographic features. The present paper takes a more phenomenological approach; for example, the relation between thermal history and stratigraphic distribution of cavities is emphasized. Devitrification textures are not described in detail, which could lead to further insights into their origins.

Near-spherical lithophysal cavities are abundant in the Prater Creek Ash-Flow Tuff (Walker, 1969) at Highway 395 near Burns, Oregon (Fig. 10). Compaction foliation is deformed around some cavities (Fig. 10B), clearly indicating their growth by ductile inflation after compaction foliation had developed. Emplacement of the Prater Creek Ash-Flow Tuff above T_G is suggested by the paucity of phenocrysts. Most cavities are surrounded by a concentric zone of slightly higher degree of devitrification than the groundmass farther from the cavity (Fig. 10C). Short horizontal fractures extend from some cavities (Fig. 10A).

Some lithophysal cavities in the Jackpot Member of the Rogerson Formation (Andrews et al., 2008), near Jackpot, Nevada, have short, angular arms extending from the main body of the cavity (Fig. 11) and resemble the star-shaped cavity of Breitkreuz (2013). Arms extend both parallel and perpendicular to compaction foliation. Euhedral crystals, including cristobalite and feldspar, line the cavity walls, and the tips of the angular arms are typically lined or bordered by opaque oxide grains, indicating former vapor. Again, cavities are bordered by incipient devitrification.

A section of Rattlesnake Tuff on Highway 395 north of Burns, Oregon (Fig. 5) has abundant spots and cavities. Pink spots appear at the top of a basal vitrophyre and are prominently developed within 2 m of the base (Fig. 12Q). In thin section the spots appear dark rather than pink, owing to cryptocrystalline iron oxides in the groundmass. The spots are incipient devitrification; the adjacent groundmass is wholly vitric and shows delicate, curling perlitic fractures (Figs. 13R1 and 13R2). Spots occur below the lowest cavity, but by 3 m above the base, lithophysal cavities occur along with spots, and the groundmass adjacent to spots shows incipient devitrification (Figs. 12P and 13O). The lowest cavities are mostly small, arcuate rims on spots. By 8 m above the base, the groundmass shows greater devitrification, vapor-phase deposition occurs, and cavities include larger, spherical openings (Fig. 12J). By 12 m above the base, vitroclastic texture in spots is obliterated, and some spots form rings of devitrified material with a core of less devitrified groundmass. Both solid spots and ring spots typically have a partial bounding fracture; other large spots have an arcuate internal fracture that evidently marks an early boundary of the spot, which later continued growth outside the fracture (Fig. 14). Fractures show both vapor-phase deposition and banded opal fill.

Small, Mode-1 fractures are common and show zigzag traces, bifurcations, or high-angle intersections (Fig. 15F). True spherulitic habit is best developed in individual pumice clasts (Fig. 15J-2), although it is incipiently developed in spots or at random places in the groundmass. Its occurrence seems more to reflect continuity of glass framework in precursor pumice clasts, rather than to have any genetic significance. Vapor-phase deposition in vesicles is common near the top of the section but is rare below 10 m height. The uppermost sample is slightly less intensely devitrified (Fig. 15A) and has fewer spots and cavities than samples beneath it; zigzag fractures occur in this sample as well. Cavities in the uppermost samples (Figs. 12B and 12D) are the most star-shaped and also the most elongate in the plane of foliation.

The stratigraphic distribution of these features is summarized as follows. Spots appear at the top of the basal vitrophyre and occur thereafter to the top of the preserved section. The spots in otherwise pristine, perlitic vitrophyre are clearly nucleation sites for initial devitrification. Lithophysal cavities first appear ∼2 m above the lowest spots and also occur nearly to the top. Both spots and lithophysal cavities tend to be more abundant where density is lower and permeability is higher (Fig. 5). Groundmass devitrification first occurs above the spots at approximately the level of the lowest lithophysal cavities. The intensity of devitrification of the center of cored spots is always at least equal to that of the adjacent groundmass or is greater (Fig. 14); this relationship was noted as well by Streck and Grunder (1995, Fig. 11B). Vapor-phase deposition is restricted to the upper half of the section.

Compaction as well as formation of lithophysal cavities involves strain of a glassy medium at or near its transition temperature T_G. T_G demarcates the transition from solid-like (glass) to liquid-like (melt) behavior (Wong and Angell, 1976). Complications arise with the introduction of time. Glass may behave brittly over periods of seconds or minutes yet still behave ductilely under stress applied for days. This phenomenon is reviewed by Dingwell and Webb (1990), and the following summary is excerpted from their study. Glass behavior near T_G is idealized by a Maxwell viscoelastic model of a spring (elastic) in series with a dashpot (viscous). The instantaneous response to a stress is elastic deformation (recoverable), followed over time by viscous relaxation (permanent and equilibrium); such a setup has a characteristic timescale τ defined as (η/M), where η is viscosity (Pa s) and M is the instantaneous elastic modulus (Pa). Plotted in time-temperature space, τ separates glassy behavior at lower temperatures or shorter times from liquid (equilibrium) behavior at higher temperatures or longer times.

The M of real silicate melts is nearly constant with melt composition or temperature (log₁₀ M ∼10); thus τ varies linearly with viscosity η (Dingwell and Webb, 1990, p. 431). In Friedman et al.’s (1963) classic compaction experiment on which the compaction model of Riehle et al. (1995) is based, rhyolitic ash at zero water pressure and 635 °C compacted from 50% to 40% porosity in four to five months. The same compaction increment at 1.72 bars water pressure required only a few hours. Viscosities of the ash used in this experiment are reported by Friedman et al. (1963) to range from 10¹² Pa s (785 °C, 0.2% H₂O) to 10^14.5 Pa s (535 °C, 0.7% H₂O). Dividing by 10¹⁰, it follows that τ of the ash ranges from 0.5 day at high viscosity, to 100 s at low viscosity. Thus, experimental compaction times—as well as compaction times in naturally occurring tuffs—are well within the field of liquid (equilibrium) behavior.

Real silicate melts exhibit more complex behavior than the Maxwell ideal. Melt fibers fail brittly at high strain rates that are still well below the calculated equilibrium strain rate; this occurs without necking and possibly reflects structural changes in the melt (Dingwell and Webb, 1990, p. 435). Eruption of ash flows involves brittle failure of the melt to form shards, possibly a result of increased viscosity at constant strain rate due to loss of water during rise in the conduit. Similar changes in material behavior may be involved during formation of lithophysal cavities. Preserved structures in tuff deposits are the net result of an integrated deformation history over a range of temperatures and periods of days to months; therefore, it is difficult to uniquely hindcast a single variable such as temperature history or PH₂O solely from observed deformation. Here, only relative differences in style of deformation are discussed. Future experiments on formation of cavities during crystallization may provide insights applicable to naturally occurring tuffs.

Devitrification is exothermic. Agarwal (1989) measured exotherms by differential thermal analysis in industrial glasses, which he identified as formation of wollastonite, diopside, and anorthite at ∼900 °C, 950 °C, and 1000 °C, respectively. From 1000 °C to melting at 1150 °C, bubbles formed in the residual glass by exsolution of gas. Recent studies of naturally occurring spherulites report water concentration profiles in the surrounding rhyolitic obsidian that document diffusion of water away from the anhydrous mineral assemblage formed during crystallization (Castro et al., 2008; Watkins et al., 2009). Water concentration is highest immediately adjacent to the spherulite. Modeling of the profiles implies growth of cm-scale spherulites on a timescale of days to months (Castro et al., 2008). Watkins et al. (2009) conclude that thermal diffusivity of obsidian is too high to allow for temperature rise due to release of latent heat; this conclusion is examined in the following section.

Is exothermic devitrification a possible source of heat to inflate lithophysal cavities? Watkins et al. (2009) argue that spherulite growth is limited by diffusivity of OH^–, which is 10^–6 mm²/s in obsidian, and because thermal diffusivity is orders of magnitude higher, latent heat is conducted away well before temperature can rise. However, theoretical spherulite growth times reported by Castro et al. (2008) are shorter than those of Watkins et al. (2009) and are verified by experimentally grown spherulites (Castro et al., 2009). Second, even if OH^– diffusivity is a limiting growth factor in nonporous obsidian, it is less likely to constrain growth in partly permeable tuffs. Lastly, Watkins et al.’s (2009) analysis ignores scale effects, as is shown next.

For spheres of radius a = 1–25 cm, diffusivity D = 0.0047 cm²/s, and time t of 4.3 × 10⁴ s (0.5 days, used below), the term for n = 2 in the summed series is effectively zero because of the large negative exponent of e. Thus, a single term (n = 1) suffices.

For conduction across a boundary of solids having the same thermal conductivity and diffusivity—a reasonable assumption here—the initial boundary heat content C₀ is equal to the average of the two solids (Hsu, 1963, p. 72). Suppose a sphere of radius a = 1 cm (a single spherulite) begins devitrification and releases 0.28 cal/g the first day (C₁). Ambient heat begins at zero; thus, C₀ = 0.14 (average of 0.28 and 0). After 15 min, C = C₀ to the third decimal place, C₀ begins to decrease, and Watkins et al.’s (2009) conclusion of rapid heat loss is verified.

Now consider a sphere of radius 25 cm, which represents not a single spherulite but instead a volume of rock in which multiple, closely packed spots have nucleated (for example, Fig. 12Q). The thermal history can be modeled by Equation (2) with an important modification: Heat lost from the central sphere is added to the enclosing rock, which results in higher values of C₀for each time increment. The amount of increase in C₀ is uncertain, but the goal here is only to demonstrate that heat builds up—not the exact value of buildup. The amount lost to the environment is the average amount in the sphere after any time increment subtracted from the total amount added for all increments. The average amount in the sphere is that at radius 20 cm, which is 50% of the sphere volume. Crank (1957, fig. 6.1) shows that after a time increment of 0.5 days, the radial distribution of heat in the sphere is nearly linear; therefore, heat at r = 20 cm is simply C₀ + 0.2(C – C₀). By keeping calculation increments short (0.5 days) relative to the overall period of crystallization (50 days), average values afford a reasonable approximation to continuous variation. After the first half-day increment, heat content at the sphere center is 0.0757 cal/g, that at the outer surface is 0.07 cal/g (the prescribed starting value), and the average heat content of the sphere (r = 20 cm) is 0.0711 cal/g. Subtracting 0.0711 from 0.14, the heat lost to the environment was 0.0689 cal/g; it is highest next to the sphere and drops to zero farther away, defining a heated shell. The next half-day increment begins by adding 0.14 cal/g to 0.0711 cal/g, or C₁ = 0.2111. Assume that now the effective ambient heat content is (0.0689)/2, that is, the central sphere “sees” an ambient heat content that is the average of the surrounding shell. C₀ for the second interval is then (0.2111 + 0.03445)/2 or 0.1228.

Continuing with this method, the heat content of the 25-cm sphere at the start of the last time increment is 4.8 cal/g, which is a temperature increase of 16 °C. The difference from the 1-cm sphere is due to the larger volume. For isolated small spherulites (the Watkins et al. [2009] case), heat is lost rapidly. But when numerous heat centers interact, heat is not so quickly lost, and the temperature rises. These figures are based on the assumption that by the end of the 50-day period, the entire volume of sphere has recrystallized. For lesser intensity of devitrification, lower temperatures result. Alternatively, the effective ambient heat content “seen” by the sphere is likely to be that of the shell nearest the sphere, a little higher than 50% of the lost heat, and the temperature will rise a little more than 16 °C. Conversely, heating by latent heat is reduced by ongoing conductive heat loss of the total deposit during the 50-day period. Detailed modeling of the Rattlesnake Tuff profile at Burns (Fig. 5) shows that temperatures over 50 days beginning two years after deposition (onset of devitrification) fall by 4 °C at 3.6 m height, by 6 °C at 7.4 m, and by 7 °C at 13.1 m; these temperatures must be subtracted from the 16-degree temperature rise calculated for a sphere. Thus the heating effect of latent heat is concentrated in the bottom half of the deposit.

With the foregoing as background, the crescentic voids that partly encircle devitrified spots are here interpreted to indicate ductile expansion of compacted tuff—in agreement with Breitkreuz (2013)—by vapor released by devitrification. The voids form after some minimum amount of devitrification (i.e., spot size), and they form at the boundary of the spot where water concentration is highest. Cavity abundance tends to be greatest in horizons that are slightly less than maximum density (Figs. 5 and 6), suggesting that residual porosity is important to their formation. This implies that growth of lithophysal cavities is driven by the combination of increased free gas above the previous lithostatic pressure and its adiabatic expansion by a modest temperature increase. If porosity is too high, the ratio (gas released/pore volume) drops below a critical minimum; if zero porosity, released gas remains in glass solution. Expansion that appears to begin at a small amount of devitrification (e.g., Figs. 10 and 11) may reflect effects of gas at higher pressure and temperature that has diffused upwards a short distance.

Deflection of compaction foliation (Fig. 10) requires growth of cavities above T_G, similar to deflected microlites around some spherulites (Castro et al., 2008). For short times of one day, the corresponding viscosity is 10¹⁵ Pa s, which for rhyolitic glass having 0.4% water content, translates to ∼535 °C. Model temperatures at the base of the Rattlesnake Tuff at Burns, Oregon (Fig. 5B) were >500 °C after two years at and above 4 m height, which is approximately where the first lithophysal cavities appear.

Fractures extending from cavities (Fig. 10A) are interpreted to indicate brittle failure by stress concentration at cavities after ductile cavity formation, and star-shaped cavities apparently reflect brittle failure of cavity walls. Bubbles in supersaturated obsidian grow by diffusion of vapor into the bubble (Stevenson et al., 1997; Blower et al., 2001) with consequential increase of bubble-wall viscosity. Conversely, water concentration around growing spherulites in impermeable obsidian is high at the spherulite boundary, so viscosity is lower and cavities begin as ductile deformation. The shift to brittle behavior (fractures) at a single cavity in permeable tuff probably reflects viscosity increase of the cavity wall consequent to devitrification, possibly aided by more rapid cavity growth (non-Newtonian behavior; Dingwell and Webb, 1990) as pressure rises due to heating and vapor increase.

Model temperatures based on conductive cooling of the Burns profile remained >600 °C for >2 years only between 7 and 18 m height (Fig. 5). Spots and lithophysal cavities extend to just below 20 m, close to the model 18-m limit, but spots occur in the basal vitrophyre beginning at 3-m height, well below the 7-m model limit. Devitrification started in the hottest part of the deposit, between 10 and 18 m; latent heat raised the temperature in this central portion and then conducted up and down. Heat loss by conduction at the base is less than by radiation and convection at the upper surface; thus, the secondary temperature pulse reached further toward the base than toward the top. Temperatures need not have exceeded 600 °C at 3-m height; they only needed to have been maintained >500 °C for some time.

Devitrified annuli having a core cavity represent radial expansion that was eventually halted by increasing rigidity of the crystalline annulus. Their origin may be as a variation of crescentic fractures. Devitrification in an annulus that surrounds a central core of less devitrified ash shards (Fig. 14) is more difficult to explain. Degree of undercooling is decreased by both increased water pressure (lower solidus temperature) as well as increased temperature. However, the increase of water pressure is minimal, at most perhaps a bar over load pressure. So one possible explanation is that initial devitrification of the cores was slowed by decreased undercooling, but increased water pressure as vapor migrated outward from the core enhanced nucleation by a viscosity decrease more than it decreased undercooling. Resolution of this interesting feature requires experimental devitrification studies.

The interpretation here of devitrified annuli differs from that of Streck and Grunder (1995), who interpret more devitrified annuli around a less devitrified core as recrystallization of initially homogeneously devitrified groundmass. However, the mechanism for a just-devitrified patch of tuff to then recrystallize is doubtful. Alternatively, spots are clearly more highly devitrified than non-spot groundmass. Their occurrence throughout the section means that they were likely the first devitrification in pervasively devitrified parts of the section as well. Uniformly devitrified groundmass occurs only above the lowest occurrence of spots and in the company of spots. The conjunction of these facts implies that spots are nucleation sites of initial crystallization; wholesale groundmass devitrification occurs later, in response to greater time and increased water pressure and temperature rise.

Unlike the Burns section of Rattlesnake Tuff, the Peach Springs Tuff at Kingman (Fig. 16) does not show a basal vitrophyre passing up into a zone of spots, although obscure spots do occur in pervasively devitrified groundmass in the center of the deposit. Vertical joints at the base are dm to 3 m apart and irregularly developed, but by 3 m above the base, the joints have organized into regular spacings of 1–3 m and smooth faces. The joints comprise two sets that intersect at high angles and are almost certainly columnar cooling joints. Polyhedral fracture blocks of dm scale occur along some vertical fractures (Fig. 17D) more than 1 m above the base; some blocks appear to have long axes oriented perpendicular to the columnar joints—that is, they resemble horizontal columnar joints.

Some sites have a 0.5-m-thick swarm of planar to anastomosing horizontal fractures between 3 and 6 m above the base, distinguishable from single, planar horizontal fractures. Other such swarms occur near the top of the section (Fig. 18C). Some vertical joints truncate against such horizontals, indicating formation of the horizontals early in the cooling history. At 16 m above the base, where the tuff passes from partly compacted to densely compacted (Fig. 16), vertical joints show upward-bifurcating swarms (Fig. 17B). All have well-developed rims indicative of formation in the presence of hot gas. Vertical swarms are a few meters below the bottom of the lithophysal-cavity zone; in the intervening horizon, both verticals and horizontals are short and weakly developed.

At ∼22 m above the base, density reaches its maximum of 2.35–2.40 g/cm³ (Fig. 16; porosity <1%). This is also the base of an 8-m-thick zone of lithophysal cavities (Fig. 18A), which occur mainly in a narrow zone within and immediately above the horizon of maximum density. Cavities are largest and most spherical at the bottom of the zone and irregularly become smaller and more elliptical upwards (horizontal long axes). In the cavity zone, horizontal fractures are longer and more abundant than below the zone and have hackly surfaces. Horizontals may pass around cavities, or form the upper or lower boundary of cavities, or pass through cavities. Vertical fractures also intersect some lithophysal cavities, indicating that the cavities formed first; other verticals curve around a large cavity, indicating deflection of the fracture during inflation. Verticals tend to be smooth, but the rock adjacent to verticals may be a zone of dense fracturing; the horizontal tops of such densely fractured zones differ in elevation across bounding verticals (Fig. 17A).

Most horizontals do not carry across the nearest long vertical; that is, they do not align with horizontals in the adjacent column. Thus, columnar joints bound domains that developed horizontal fractures independent of the adjacent domain. Most of such columnar joints—master verticals—formed before most horizontals. Short secondary verticals and horizontals truncate one another about equally. Hackly horizontals may merge laterally into microbreccia at the bounding vertical. In the zone of lithophysal cavities, the spacing of horizontals within a domain tends to increase upwards (Figs. 16 and 18D); above the zone of cavities, horizontals occur with greater, irregular spacing or in small swarms. Terminations of horizontals at verticals tend to be bifurcations (Fig. 17C).

The highest horizontal fractures are 12–14 m below the ground surface, where most truncate minor verticals but are truncated by master verticals. One vertical was observed to curve into a horizontal fracture ∼5 m below the surface; other verticals have decimeter apertures filled with angular clasts of tuff. Beneath a long horizontal that extends from one master vertical to the next master vertical, small swarms of horizontals occur with upward-decreasing spacing.

To summarize Peach Springs Tuff structures: (1) Long “master verticals” are early-formed columnar joints. (2) Most horizontals truncate at master verticals and formed later. These secondary horizontals have shorter lengths than master verticals, and in different horizons, truncate or are truncated by shorter verticals. Secondary fractures developed independently in each column. (3) The zone of lithophysal cavities is bounded on its bottom by the highest density and lowest permeability of the entire section. (4) In the zone of cavities, horizontals have hackly surfaces and dominate short verticals. Most verticals in this zone are masters (columnar joints) and are bounded by densely fractured borders. Horizontals and secondary verticals developed approximately contemporaneously with cavities. (5) Lack of slickenlines on horizontals, their independent development in vertical columns, splaying by some at a truncating vertical, and their hackly surfaces rule out shear as their origin; these are Mode 1 fractures.

An important distinction between columnar joints in basalt flows and the Peach Springs Tuff joints is the protracted period of joint formation in the Peach Springs Tuff, beginning with vertical columnar joints that are similar to those in basalt flows, but following with a second period of complexly interacting jointing in the Peach Springs Tuff associated with cavity formation. This second period apparently reflects reheating and gas pressurization consequent to devitrification.

A key concept of jointing is that the longer the joint, the greater the release of stored elastic energy (Lachenbruch, 1960a, 1960b). A corollary is that any early-formed joint at the upper or lower margin of a cooling deposit reduces the stress field for shorter neighbors (DeGraff and Aydin, 1993); that is, some joints “starve” their neighbors and become “master joints.” This is observed at the base of the Peach Springs Tuff, where the lowest vertical joints have the smallest spacings but within 2–3 m above the base, joint spacings have stabilized at 1–3 m. Some vertical joints near the base are bounded by dm-scale polyhedral fracture blocks. There is no evidence of chilling of the base of the Peach Springs Tuff at Kingman by substrate water, such as rosette jointing or palagonitized tuff. Similar polyhedral blocks adjacent to columnar joints in the vitric Rattlesnake Tuff in John Day are bounded by pure mode-1 fractures. Polyhedral blocks in vitric tuff likely formed by thermoelastic contraction in response to isotherms that were parallel to the vertical joints. Other authors (e.g., Ross and Smith, 1961; Sheridan, 1970) have interpreted closely spaced, nonvertical joints to be the result of isotherms deflected from the horizontal due to convective heating by gas escaping along columnar joints, and this explanation likely applies to these features; the absence of rims means that degassing at the base was completed in a short period.

Columnar joints are vertical because a cooling body is free to contract in the vertical by lowering its upper surface but is constrained in the horizontal by effectively infinite extent. Thus the question arises: once a vertical column has formed, how is it that any further jointing can occur, either vertical or horizontal? Subsequent jointing obviously does occur, as evidenced by later subdivision of early-formed crustal plates on the surface of lava lakes (e.g., Peck and Minakami, 1968), which must reflect differential stress between the edges and the interior of a plate. Stress release is greatest nearest to a just-formed joint and perpendicular to the joint (Lachenbruch, 1960a, 1960b). The component of horizontal stress parallel to a new joint is less reduced, and thus new joints will eventually form in the interiors of early-formed columns as cooling proceeds, typically normal to the nearest bounding joint.

The author is unaware of any study that deterministically models horizontal thermoelastic joints; for example, DeGraff and Aydin (1993) simply acknowledge the existence of horizontal joints and discuss their effect on further growth of vertical joints. This is due at least in part to the difficulty in specifying the initial and boundary conditions at the formation of horizontal joints—the material properties over a range of temperatures, the three principal stress components, and the precise geometry of newly formed joint blocks. Holmes’ (1945) re-publication of Tomkeieff’s obscure 1940 paper explains horizontal joints in basalt columns of the Giant’s Causeway as due to “efficient cooling” by gases escaping along previously formed vertical joints. Cooling by gases is doubtful; in contrast, tuff researchers typically refer to convective heating adjacent to fissures that feed fossil fumaroles.

Alternatively, Nemcok et al. (2004) infer development of horizontal fractures in the geothermal system of a Javanese volcano by cyclic overpressures together with support of caprock by asperities. It is suggested here that horizontal fractures in Peach Springs Tuff columns formed by combined effects of gas pressurization and thermoelastic contraction but where some fraction of the column weight was born by asperities—deviations from the vertical—of bounding columnar joints. Indeed, horizontal thermoelastic fractures could not have formed without the prior formation of vertical fractures. This conclusion is implicit in DeGraff and Aydin’s (1993) discussion of horizontal joints formed after passage of the vertical joint. Horizontal joints near the base of the deposit, for example, may have formed shortly after an interval of vertical joint growth; unfractured tuff above the vertical joint tip partially suspended the column, which then fractured horizontally. Without a vertical joint, all vertical thermoelastic contraction simply results in uniform settling of the deposit surface. The addition of vertical joints subdivides the deposit into domains each capable of independent response to the evolving stress field.

A digression from the Peach Springs Tuff is made here to discuss other evidence for gas pressure in tuffs in excess of load pressure. Dunne et al. (2003) describe jointing in the interior of the 100-m-thick Tiva Canyon Tuff at Yucca Mountain, Nevada. The earliest formed fractures are orthogonal vertical joints that have mm- to cm-diameter tubes in their faces. Tube-bearing joints formed prior to lithophysal cavities, because none intersect cavities and some in close proximity to cavities are deflected by the inflation. The tubes that formed after the joints occur only on joint faces and not within rock between joints (where instead, lithophysal cavities occur), and expanded joint faces up to 15% vertically. Vapor-phase grains occur on tube walls, and devitrification extends a tube diameter into the matrix, but joint faces show little of either. Vertical expansion by both lithophysal cavities and tubes is inferred by Dunne et al. (2003) to indicate gas pressure in excess of load pressure.

Tube-bearing joints occur in the majority of uppermost lithostratigraphic units of the Tiva Canyon Tuff, and some occur on early-formed horizontal joints as well (Throckmorton and Verbeek, 1995). Although overpressured gas seeped along the joints, joints were still sufficiently confined to inflate the tubes. Morgan’s (1984) thin sections showing recrystallized joints are interpreted by Throckmorton and Verbeek (1995) to indicate joint sealing due to expansion of the tuff mass during the period of overpressure. Sealing may have been aided by mineral deposition. Daily et al. (1987) found an order of magnitude decrease in permeability of fractured tuff in the first 83 days of an experiment at 89 °C; they attributed this decrease to silica deposition. Thus, early-formed joints were generally not highly permeable conduits for gas later in the cooling of the Tiva Canyon Tuff, which is why vapor-phase grains and devitrification occur mainly in tubes and cavities and not joints.

Bifurcating fractures in the Peach Springs Tuff at Kingman occur as vertical branches at the transition from partly compacted tuff up to densely compacted tuff and at the termini of horizontal fractures truncated by columnar joints. Densely clustered, bifurcating fractures are characteristic of “dynamic fracturing,” which is reviewed by Sagy et al. (2001) and is caused either by high strain rates or by near-constant strain rates imposed by regional tectonism. A key geometric feature of dynamic fracturing is branching in the direction of growth (Fig. 17B); an associated feature is surface roughness (surface roughness, then, may not be a unique indicator of joint formation at low temperature). Branch fractures are linear or nearly so (Fig. 17B), a feature interpreted by Sagy et al. (2001) to indicate limited interaction amongst growing fractures due to their high velocity of propagation. Sagy et al. (2001) experimentally replicated outcrop-scale, bifurcating fractures. The experiments revealed “fracture clustering”: initial fractures formed a single branch, but within one second, numerous overlapping branches formed, generating a fragmentation zone.

The vertical bifurcating fractures low in the Peach Springs Tuff (Fig. 17B) are rimmed and formed in the presence of gas. Initial gas outfluxing occurred within a few days; thus, the gas and related thermoelastic stress must have been part of a second pulse during devitrification. Dynamic upward fracturing was likely caused by the same pressure that formed lithophysal cavities slightly higher in the deposit. Horizontal bifurcating fractures are more common in the upper half of the Peach Springs Tuff (Fig. 17C). Their growth was toward a bounding columnar joint. After formation of the vertical joint, the interior of the column retained the highest residual contractile stress, and the horizontal fracture originated in the column interior. Upon approaching the vertical joint, the velocity of growth accelerated not because of a higher temperature gradient but because the horizontal component of stress had been recently reduced at the vertical joint, and the ratio of vertical/horizontal residual stress components was increased.

The origin of the anastomosing horizontals near the base and top of the section is uncertain. Similar, opal-lined horizontal fractures are described by Dunne et al. (2003) in the Tiva Canyon Tuff; they attribute these fractures to faulting. However, these anastomosing fractures could also owe their origin to dynamic fracturing. Another example of dynamic fracturing is provided by Riley and Tikoff (2010) in the roof of a Sierra Nevada batholith and called by them tabular fracture clusters (tfc). Differences between tfc and the anastomosing clusters in the Peach Springs Tuff are (1) scale and (2) orientation. Tabular fracture clusters in the Sierra Nevada batholith occur along vertical fractures formed by doming and are inferred to reflect volatile overpressures developed late in the crystallization of the batholith. Tabular fracture clusters, however, are opportunistic and could form equally well along a horizontal fracture, such as in the Peach Springs Tuff. Tabular fracture clusters are 4–40 cm wide; anastomosing clusters in the Peach Springs Tuff are slightly wider (50 cm), which could be due to lower tensile strength of the porous tuff.

An intriguing aspect of horizontal fractures in the Peach Springs Tuff is the regular upward increase of their spacing in the center of the deposit (Figs. 16 and 18D). Horizontals are prominent in the zone of lithophysal cavities and continue upwards to within a few meters of the inferred paleosurface; however, their spacing in the upper third of the deposit is irregular. Horizontal fractures are superficially similar to exfoliation sheeting in granitic rocks, but the spacing of which typically decreases upwards. Another possible explanation is that the horizontal fractures formed simultaneously with vertical crack progression and thus mimic increments of vertical joint growth. However, at least some horizontal fractures formed simultaneously with lithophysal cavities in the zone of maximum density, and these cavities did not form until devitrification, well after initial formation of columnar joints. A third possible mechanism is that upward-increasing spacing of horizontals coincides with a zone of upward-decreasing density and, therefore, tensile strength, but upward-decreasing strength should have resulted in decreasing spacing (at constant stress), not increasing.

A fourth factor that may account for the spacing of the horizontal fractures is gas pressure in the zone of lithophysal cavities. As discussed, overpressures apparently persisted for some time despite the presence of vertical joints. Residual thermoelastic stress may have been involved in horizontal fracturing, but downward-increasing gas pressure may have governed spacing of the horizontal fractures in the lithophysal-cavity zone. Higher pressures just above the point of maximum density may have resulted in closer-spaced fracture failures.

Densely clustered, dm-scale fractures occur in the Rattlesnake Tuff as well as in the Peach Springs Tuff at Kingman (Figs. 17A and 18B). There is no evidence at these sites for chilling by either rainwater or substrate water. Short, mode-1 fractures are clearly associated with spots of incipient devitrification in the Rattlesnake Tuff at Burns (Figs. 13 and 15F). Some fractures extend from a spot into surrounding perlite, while others border spots. In more intensely devitrified zones, mode-1 fractures occur with bifurcating or “zigzag” traces, with blunt terminations, or with pinch-and-swell walls (Fig. 19). These features resemble fractures in combusted Miocene mudstone described by Eichhubl (2004), who attributes them to ductile fracturing consequent to high-temperature sintering, in contrast to brittle fracturing such as is responsible for columnar joints. The key difference is the mechanism for fracture growth: in brittle fractures, stress concentrates at a sharp joint tip, which extends the joint. In ductile fracturing, growth is by coalescence of pre-existing pores. Coalescence is favored along planes of maximum shear stress ahead of the fracture, and these planes extend at intermediate angles rather than along the fracture extension (Fig. 20A).

As previously discussed, higher-density devitrification assemblages must be accompanied by a slight increase in porosity in order that bulk density remains constant. It is suggested here that the preexisting pores required for ductile fracturing in tuff are residual porosity after compaction, enhanced by dilation accompanying devitrification. Ductile fracturing, then, may result by porosity increase in the middle to late stages of devitrification, before complete rigidity. At Nicoll Ridge Road, Burns, and Rattlesnake Canyon sites of the Rattlesnake Tuff, dense fracturing occurs below the highest occurrence of devitrification (Fig. 20D), unlike polyhedral fracture blocks bordering vertical joints at the base of the Peach Springs Tuff (Fig. 17D) and in the John Day II profile, where the adjacent tuff is vitric. At some sites, dense fracturing appears influenced by, or concentrated along, master verticals such as at the base of the Peach Springs Tuff (Fig. 17A), which may reflect enhanced heat transfer and consequential devitrification along the verticals.

Some of the tuff sections studied for this report do not appear to have through-going columnar joints. The Rattlesnake Tuff in Rattlesnake Canyon, east of Burns, Oregon, shows an upper zone of columnar joints that penetrate only to a central zone, where columnar joints are obscured or terminated by dm-scale, chaotic fractures (Fig. 20D), which are late-formed, ductile fractures. Ross and Smith (1961) noted that many tuffs show horizontal platy jointing near the point of maximum density; in some, vertical joints do not penetrate this zone of horizontal jointing, whereas in others, they do. This implies that the horizontal jointing formed before growth of the columnar (vertical) joints by cooling had penetrated the interior of the deposit. The author is not aware of the cases on which Ross and Smith base their description, but by the analysis here, master columnar joints form earlier than any horizontal joints in tuff deposits <50 m thick. Secondary vertical joints are certainly truncated by horizontal joints, but in any case, dense ductile fracturing can obscure older, vertical columnar joints to the extent that they seem to disappear.

Model results show that compaction at density reversals in the Peach Springs Tuff section continued for as long as four years (Fig. 6C), yet two years above 600 °C is adopted here as the threshold for onset of devitrification. This raises the possibility that the marked density reversals in the Kingman section of Peach Springs Tuff reflect premature arrest of compaction by devitrification.

Density highs are attained quickly after deposition (e.g., Fig. 1) and are not likely to have been affected by devitrification. If compaction at the lows was prematurely arrested, the model cooling intervals in the central half of the section are too large (Fig. 21A). There is no way to evaluate by how much density lows may have been affected. A model profile was recalculated for cooling intervals that were reduced by an arbitrary 30% in the central part of the section, which required that emplacement temperatures also be reduced so as to match the density highs between the cooling breaks (Fig. 21B). Emplacement temperatures had to be reduced by as much as 24 °C (Fig. 21C). These results mean that the model results for the central half of the Peach Springs Tuff section at Kingman are probably reliable only to within ∼15 °C and a factor of 2 for cooling intervals. (Note: Devitrification did not create density reversals—it preserved them where they might otherwise have become unrecognizable by compaction.)

The issue of premature arrest arises because of emplacement temperatures >720 °C and thickness >100 m of the Peach Springs Tuff section. High lithostatic pressure drove compaction at density lows in the interior of the deposit long after deposition. A resolution would be to model another section of Peach Springs Tuff that is only 50 m or so thick and compare those model results with the Kingman section. However, the goal here is not an exhaustive study of the Peach Springs Tuff but a discussion of its secondary features. Toward that goal, note that temperature histories of the initial profiles and the revised model profiles are nearly identical (Fig. 21D). This seemingly impossible result is because cooling history depends technically on heat content, not just initial temperature, and enough heat was lost during the longer cooling intervals of the initial, hotter model profile that the long-term cooling histories are nearly identical. In any case, the discussion of the relationship of the secondary structures to the thermal history of the section is not materially affected.

In future studies, for any density reversal requiring >2 years for completion of compaction, a larger uncertainty will be assigned to the cooling interval. However, the effect is not likely to be major. Two years is the minimum to onset of devitrification, which begins at isolated spots (see section “Lithophysal Cavities ...”) and which raises the temperature of the surrounding deposit by latent heat. This in turn reduces the viscosity and increases the rate of compaction before complete devitrification, all of which minimizes the effect of premature arrest. Details of how devitrification slows compaction require laboratory study.

Streck and Grunder (1995, p. 164) state that “... strong variations in welding and crystallization facies [in the Rattlesnake Tuff] occur over a distance of a few hundred meters with little or no change in thickness ....” Streck and Grunder suggest that such variations are likely the result of subtle variations in thickness, temperature, and cooling history. The Rattlesnake Tuff at Camp Creek Road has abundant lithophysal cavities in one profile and very few cavities in the profile within 150 m laterally. This site provides an opportunity to investigate the cause of lateral variability in development of secondary features in the Rattlesnake Tuff.

From afar, the continuous mesa top lends the appearance of layer-cake stratigraphy and suggests near-constant thickness (Fig. 22). The mesa top slopes gently up away from the cliff edge, rising as much as 8 m above the cliff; exposures are sparse on the top. Secondary features at the Camp Creek site (Fig. 23) were closely observed along a foot traverse of one profile and in the bottom 8 m of the adjacent profile and its samples. Geometry and relative ages of horizontal and vertical joints are generally similar to those in the Peach Springs Tuff at Kingman and are not discussed in detail. The important difference between the two profiles is the near-absence of lithophysal cavities in the first profile and their abundance near the base of the second profile.

A vitrophyre is the lowest exposed rock in each profile, which by correlation with the Rattlesnake Tuff at Burns suggests that the depositional base is within 1 or 2 m. In the second profile (inset, Fig. 23), the vitrophyre has spots at its top and is overlain by dense, devitrified tuff. A zone of lithophysal cavities extends from the base of the devitrified zone 2 m above the exposed base to ∼8 m; the cavities occur together with elliptical spots whose long axes are horizontal.

A cross section shows the buildup of the model deposits at the adjacent profiles (Fig. 24C). The section emphasizes that what appears to be uniform thickness across the 150 m separating the two profiles is actually a 15-m difference in the elevation of the base of each deposit and a difference of ∼18 m in the pre-compacted, aggregate thickness (Fig. 7) of each deposit. Detailed density profiles reveal the internal architecture of these deposits, which enables recognition of the otherwise occult difference in their thicknesses. Correlation of the tops of the deposits at different stages in their emplacement shows that although the pre-eruption topographic low at the thicker profile was progressively filled, nonetheless, even after the last deposit there still remained a slight low due to greater compaction of the thicker profile.

Temperature and compaction histories (Figs. 7 and 24) are similar in the two profiles but show the following differences: (1) the thicker profile has a zone of likely devitrification (>600 °C for two years) that is twice as thick as in the thinner profile; and (2) the bottom of this zone is 42 m deep in the thicker profile but only 30 m deep in the thinner profile. Thus, more heat and gas were released by devitrification in the thicker profile, but upward gas diffusion must have been more hindered by the greater thickness. Moreover, conductive heat loss would have been less effective in reducing temperature rise from latent heat (see “Genetic Interpretations of Spots…” section) low in the thicker profile than in the thinner profile. The 18 m of greater thickness equates to only 0.18 MPa of greater lithostatic pressure, which serves to emphasize the slight differences in initial and boundary conditions that result in lithophysal cavities.

The Camp Creek profiles afford an opportunity to revisit one of the more enigmatic features of these deposits—horizontal cooling joints. Horizontal joints occur at several localities of the Rattlesnake Tuff: Camp Creek (Fig. 22, inset), Tamarack Creek, John Day, and Rattlesnake Canyon. These sites have in common well-developed devitrification and at least incipient lithophysal cavities. In contrast, subhorizontal joints are mostly lacking from the Rattlesnake Tuff at Nugget Street in John Day (entirely vitric) and from the Aravaipa Tuff at the N headland of Whitewash Canyon (incipiently devitrified). The origin of horizontal joints, as inferred for the Peach Springs Tuff, must be related to gas pressurization accompanying devitrification and consequential inflation of the rock mass. A complete understanding of their genesis requires a detailed mapping study in combination with quantitative stress-strain modeling.

• A period of two years above 600 °C is an empirical threshold for the onset of devitrification of the Rattlesnake Tuff, Peach Springs Tuff, and Aravaipa Tuff.

• Compaction proceeds rapidly enough that in ignimbrites <40 m thick, compaction is arrested by the combination of conductive cooling and approach to zero porosity before devitrification can occur. In thicker deposits emplaced >720 °C, however, compaction at flow margins may continue for >2 years and, as a result, is subject to premature arrest. Details of how devitrification affects compaction merit laboratory study.

• Columnar joints formed by thermoelastic contraction are present within a few weeks after deposition of an ignimbrite, and in compound cooling units, may grow about as fast as accumulation of the succeeding ash flows because of multiple cooling breaks.

• Castro et al. (2008) show that cm-scale spherulites grow in days to a few months. Using that rate of heat production in Crank’s (1957) equation for heat conduction, a spherical volume of pervasively devitrifying tuff >25 cm radius could heat by as much as 16 °C after 50 days (12 °C allowing for ongoing conductive heat loss).

• Lithophysal cavities may have a variety of origins (e.g., Breitkreuz, 2013), but at least some originate in expansion of a ductile medium as shown by deflection of compaction foliation or of adjacent vertical joints. A mechanism for expansion is increasing free vapor and its adiabatic expansion consequential to exothermic devitrification.

• Jointing of tuff deposits that undergo devitrification is far more complex than jointing in basalt flows because of the postdeposition pulse of heat and water vapor accompanying devitrification. Early-formed vertical columns create structural domains, which then respond independently to evolving stress fields. Asperities aided by gas overpressure may provide a mechanical basis for development of horizontal fractures that form after the vertical columns.

• The occurrence of cavities and tubes deep in devitrified deposits shows that gas pressures in excess of load pressure persisted during devitrification, despite the presence of early-formed columnar joints. As suggested by Throckmorton and Verbeek (1995), the joints are likely sealed by a combination of inflation under gas overpressure and mineral deposition.

• Bifurcations of both vertical and horizontal joints indicate dynamic fracturing at high velocities. Bifurcations are associated with zones of lithophysal cavities or occur where one joint approaches an older orthogonal joint. Each occurrence means high residual stresses even after early columnar jointing; the occurrence associated with cavities is probably related to the same pulse of gas expansion that led to cavity formation.

• Short fractures concentrated in devitrified horizons show features characteristic of ductile fracturing—zigzag traces, blunt terminations, and pinch-and-swell walls. The fractures may have originated by a slight porosity increase accompanying devitrification.

• A difference of only 18 m in initial noncompacted thickness (18%) was sufficient to cause formation of lithophysal cavities in one section of the Rattlesnake Tuff and not a closely adjacent section, confirming Streck and Grunder’s (1995) suggestion that subtle variations in thickness can lead to lateral variability of secondary features such as cavities.

Geology departments at Oregon State University, Washington State University, and North Idaho College (NIC) are gratefully acknowledged for granting access to polarizing microscopes for the photomicrographs in this report; the majority were made at NIC (thanks, Bill). Professor Atilla Aydin pointed out the dipping fractures in Figure 9D and interpreted their significance as shear deformation. Wes Hildreth, David Brown, and an anonymous reviewer made comments that led to material improvements in the manuscript. Another anonymous reviewer made two especially insightful comments: (1) that the average heat content of a spherical heat source is best represented by a volume mean, not a radial mean; and (2) that conductive cooling of the entire deposit acts in opposition to the temperature rise by latent heat of devitrification.

Jim Ablao rappelled the section at Camp Creek Road and obtained precisely located samples under difficult conditions; John and Denise aided in collecting this profile. David Buesch assisted in sampling several Rattlesnake Tuff profiles and the Peach Springs Tuff profiles at Kingman, as well as making observations and taking photographs, and he called the author’s attention to the Breitkreuz paper. Unfortunately, Buesch was unable to participate in interpretations or writing; therefore, any errors in the conclusions reached in this report are exclusively those of the author. Lorraine Flint kindly provided an updated list of tuff permeability studies. Don J. Thorstenson assisted in collecting the Aravaipa density profile at the south headland of Whitewash Canyon and took photographs.

Rattlesnake Tuff

John Day, Oregon; Highway 74: 44°24′30.2″N, 118°59′15.1″W

John Day II: On private property; owner requests anonymity.

John Day, Nugget Street: 44°23′51.4″N, 118°57′03.6″W

Rattlesnake Canyon, Burns, Oregon: 43°40′51.1″N, 118°47′48.0″W

Highway 395, Burns, Oregon: 43°39′33.4″N, 118°59′56.4″W

Nicoll Ridge Road, Riley, Oregon: 43°41′48.6″N, 119°44′02.6″W

Camp Creek Road, Paulina, Oregon: 44°04′23.0″N, 120°08′12.5″W

Tamarack Creek, Paulina, Oregon: 44°14′24.0″N, 119°44′35.1″W

Prater Creek Tuff

Highway 395, Burns, Oregon: 43°42′08.4″N, 119°00′56.4″W

Aravaipa Tuff

North headland of Whitewash Canyon, Arizona: 32°54′10.0″N, 110°33′15.4″W

South headland of Whitewash Canyon, Arizona: 32°53′23.4″N, 110°33′07.1″W

Peach Springs Tuff

Kingman, Arizona, at Interstate 40: 35°10′37.9″N, 114°04′15.1″W

At Route 66: 35°10′43.3″N, 114°03′50.9″W

At trailer park on Route 66: 35°09′56.7″N, 114°04′05.5″W

Sewer treatment ponds: 35°10′43.7″N, 114°03′44.4″W

Jackpot Member, Rogerson Formation, Jackpot, Nevada: 41°57′13.5″N, 114°40′42.9″W