Field and laboratory studies of the 1959 Kīlauea Iki lava lake have provided insight into differentiation processes in mafic magma chambers. This paper explores how partially molten basaltic mushes responded to unloading as a consequence of drilling. Most holes drilled from 1967 to 1979 terminated in a melt-rich internal differentiate with a sharp crust-melt interface. These interfaces were not stable, so the boreholes were backfilled by melt-rich (<5% crystal) ooze. This process, with melt ascent rates of 1.3–4.2 m/s, occurred within minutes of intersecting the bodies, mimicking volcanic eruptions, albeit on a small scale.

One borehole (KI79-1), which did not encounter such a discontinuity, was backfilled over a period of 16 days by upward flow of crystal-rich mushes rather than melt-rich ooze. The first interval of ooze recovered had undergone extensive internal differentiation. Its most conspicuous feature was production of melt-rich layers by lateral migration of interstitial melt from the wallrock into the rising crystal-rich mush. In addition, two smaller-scale processes occurred within the rising mush: segregation of melt into discrete blebs within the rising mush column and aggregation of groundmass crystals into crystal-rich clumps formed adjacent to coarser olivine crystals. The upper parts of the ooze are enriched in melt relative to deeper samples, which suggests that the melt blebs rose relative to their olivine-rich matrix. Similar melt blebs and crystal-rich clumps are observed in naturally occurring diapiric bodies within the lava lake. These processes appear to be intrinsic to the upwelling of narrow cylindrical mush bodies whether constrained within a borehole (like the oozes) or unconstrained (as were the diapirs in the lava lake).

The most striking behavior observed during repeated reentry of KI79-1 was a sharp change in rheology during the second and third re-entries of the borehole. The shift in behavior observed was that the oozes rose up the borehole, with ascent rates of 1.0–1.7 m/s, which are comparable to the rates of the crystal-poor oozes from melt-rich internal differentiates. These oozes contain more melt than the original core at equivalent depths, presumably because melt moved relative to crystals down the pressure gradient created by the open borehole. Groundmass textures in these inflated mushes show erosion of crystal outlines, especially of grain-to-grain contacts between different phases, so that the tenuous crystalline network observed in the original core samples was replaced by rounded crystals in continuous melt at crystallinities of 55–65 vol%. The transition from stable coherent mush to inflatable mush occurred at 25–28 vol% melt. This behavior appears similar to certain types of reactive transport observed in other studies.

Field and laboratory studies of historic lava lakes (1959 Kīlauea Iki, 1963 Alae, and 1965 Makaopuhi lava lakes; Fig. 1) have provided insight into differentiation processes in mafic magma chambers. Previous papers have emphasized the general processes of cooling, crystallization, and chemical differentiation of basaltic magmas. The purpose of this paper is to describe naturally occurring discontinuities in melt content within the partially molten parts of Kīlauea Iki lava lake and how they and the rest of the partially molten zone responded to and were changed by drilling.

Three of the historic lava lakes at Kīlauea have been studied: Kīlauea Iki (formed in 1959), Alae (formed in 1963), and Makaopuhi (formed in 1965). Of these, the 15-m-thick Alae lava lake crystallized completely in 13 months (Peck et al., 1966). It and the 1965 Makaopuhi lava lake were covered up by new lava in February 1969, rendering both inaccessible for further study. Observations and interpretations of the Alae and Makaopuhi lava lakes were summarized in Wright et al. (1976), Wright and Peck (1978), and Wright and Okamura (1977).

Of the three bodies, only Kīlauea Iki lava lake remained accessible for study for an extended period. Accordingly, it has been the most extensively investigated: a total of 27 boreholes (locations are shown in Fig. 2) were drilled from 1967 to 1988, and ~1400 m of core was recovered from them. The 1967, 1975, and 1988 drilling programs were conducted by the U.S. Geological Survey Hawaiian Volcano Observatory (HVO) to document the progressive cooling and crystallization of the lava lake. Drilling in 1976, 1979, and 1981 was conducted by Sandia National Laboratory as an investigation into extracting heat directly from a magma body (Hardee, 1980; Hardee et al., 1981). The 1979 drilling greatly expanded the number of locations drilled in the lava lake (Colp, 1979), and observations from this episode form the basis of the present paper.

Most work on the Kīlauea Iki drill core has focused on physical and chemical differentiation within the lava lake with particular emphasis on the physical mechanisms of processes. The results are summarized in Table 1, which gives the depths, temperature range, and timing of the various processes identified and key references for particular processes. In addition, a short overall summary was given in Helz (1987). The first set of three processes took place at high temperatures within the interior of the lake, and two of them were triggered by instabilities in the density profile in the melt column. The second set of processes, seen within the upper and lower coherent mush zones surrounding that innermost core, reflect disruption of the mush, which was initiated by the upward migration of vesicle plumes; the mush was sufficiently crystalline to retain the textural changes imposed by their passage. The intervening section (depths of 58–77 m), although traversed by two different diapiric processes (Helz et al., 1989), was too melt-rich to retain textural evidence of their passage. All of these studies assumed implicitly that the phase compositions and textures seen in the partially molten core had been successfully quenched and reflected those intrinsic to the lava lake. In contrast, this paper focuses on where and how drilling disturbed partially molten material in the lava lake.

In the historic lava lakes, drilling and core recovery in the partially molten zone were relatively easy, given an adequate supply of water to cool the bit. In Alae and Makaopuhi, most holes terminated where the interstitial melt fraction within the crust was 40–45 vol%, which occurred at a temperature of 1070 °C. The drillable crust in those bodies was fairly thin (~1.5 m, see figs. 2–3 in Wright et al., 1976), so the intervals of partially molten core recovered were short.

Initially, the thickness of the drillable partially molten zone in Kīlauea Iki was similar to that in Alae and Makaopuhi in thickness (Richter and Moore, 1966), but it broadened later, so that the total amount of partially molten mush recovered was 260 m from all holes (see core logs in Helz et al., 1984; Helz and Wright, 1983; Helz, 1993). For drill cores near the center of the lake (locations A and B in Fig. 2), Figure 3 shows how the upper crust grew over time and also how the partially molten intervals became wider. In 1967–1979, the part of the mush zone that could be drilled and sampled (melt < 45% by volume) was fairly narrow (4–6 m) at those locations. In contrast to the Alae and Makaopuhi lava lakes, the temperature at these interfaces was higher (T = 1105–1115 °C), reflecting the higher crystallinities at a given temperature observed in this olivine-rich lava lake. Drilling and core recovery terminated abruptly at the interface between the overlying mush and melt. The earliest exception was hole KI79-1 (location C in Fig. 2), where the termination was arbitrary and the partially molten interval recovered was 12 m thick. Most of the later boreholes, drilled in 1981 and 1988, ended either arbitrarily (Helz and Wright, 1983) or because it was physically impossible to continue drilling with the available equipment (Helz, 1993).

The configuration with a sharp crust-melt interface was never stable, and partially molten material backfilled the boreholes almost immediately. So, in addition to the intended recovery of partially molten core during first-pass drilling, a total of 32 m of backfilled material, referred to as “ooze” in this paper, was recovered by re-entering boreholes from 1967 through 1979. These re-drilling episodes, summarized in Helz and Wright (1983), provide information on how partially molten material, of varying melt content, responded to sudden and repeated unloading.

This paper focuses on (1) the nature of the crust-melt interfaces encountered, (2) the response of the 1979 lava lake to unloading of those interfaces with sharp discontinuities, and (3) the response of the lava lake to unloading in the absence of a discontinuity (in borehole KI79-1). The paper includes descriptions of the field behavior of the backfilling ooze, plus the petrographic and chemical nature of the backfilled material. Lastly, it documents the differentiation processes observed that resulted from these episodes of induced small-scale magma segregation and mush mobilization.

The melt fraction in the partially molten Kīlauea Iki core varies from near zero to 40%–45% by volume (Barth et al., 1994), with still higher fractions in material from just below the crust/melt interface. In most cases, that interface has been found to be a pre-existing segregation vein.

The locations and thickness of the ferrodiabasic segregation veins in cores recovered from holes drilled from 1967 to 1981 are shown in Figure 4 (modified from fig. 15 of Helz et al., 2014). Here the figure has been annotated to indicate the nature of the melt-rich body responsible for blocking various holes. Of 12 holes that terminated with an abrupt increase in melt below a sharp crust-melt interface, seven were blocked by pre-existing segregation veins, and two probably were, based on the presence of segregation veins in nearby cores recovered in later drilling projects. One (KI75-1, marked by “v”) was blocked by a vertical, olivine-rich body (or “vorb,” see Helz et al., 1989); this was confirmed when the hole was extended (as KI75-1R) in late 1978 (Helz et al., 1984). The nature of the crust-melt interface in two holes (marked by “?” in Fig. 3) is less obvious.

Four examples of crust-melt interfaces produced by segregation veins are shown in Figures 5A5D; in three cases, the interface is a few millimeters thick. Distinctly coarser crystals of clinopyroxene and plagioclase are visible in the photomicrographs below the crust-melt interface. These confirm that the melt bodies encountered were preexisting segregation veins, which had begun to crystallize before the drill bit reached them. Quenching temperatures of these melts varied from 1105 °C to 1115 °C (Helz, 2020). In contrast, the crust-melt interface in KI75-1 (Fig. 6A) is marked by large vesicles, large olivine crystals, and a sharp boundary between lighter (and hotter) melt invading the overlying interstitial melt within the mush, which are expected characteristics of a vorb. The range of quenching temperatures within this section is ~10 °C. This range is larger than that in the sections in Figures 5A5D, and the temperatures (1120–1129 °C) are higher.

The crust-melt interface sample from hole KI79-6 (Fig. 6B) shows subangular fragments of coherent mush in a glass-rich matrix, but there are no textural clues as to the nature of the melt-rich body. In this core there is no difference in quenching temperature between the interstitial melt and the large areas of glass in the sample shown here (Helz, 2020). Helz and Wright (1983) concluded that this hole was most likely blocked by a shallow-dipping, olivine- and melt-rich body, as it was terminated at depths below those where segregation veins and vorbs are found (Table 1). KI81-3, the hole closest to KI79-6 (locations D in Figs. 2 and 4), contains several such olivine-glass bodies (or “ogbs”), one of which is shown in Figure 6C. These contain none of the coarse pyroxene and plagioclase crystals characteristic of segregation veins (cf. Figs. 6B6C versus Figs. 5A5D). Other indications for a different kind of body as the melt-rich source are (1) the quenching temperature of the glass in KI79-6-190.4 (1131 °C) is higher than that of any other bottom-hole sample and (2) the melt-rich volume produced more ooze than any other source, and it surged 14.4 m back up the borehole (Helz and Wright, 1983).

The most enigmatic termination of a drill hole is that of KI67-1 (location E in Figs. 2 and 4), which ended abruptly but with no textural indication as to why. In the 1965 Makaopuhi lava lake, the core recovered from some holes showed a gradual change from more to less crystalline mush, with all core samples—including very melt-rich samples—having similar bulk compositions throughout the interval (Wright and Okamura, 1977). Thus, the crust-melt transition in that lava lake was not a chemical boundary but merely a gradual change in crystallinity.

At first it was thought that the deepest material recovered from KI67-1 might be similar. However, the pattern observed in Kīlauea Iki is that all melts recovered from below the crust-melt interface have bulk compositions (Helz et al., 1994) like those of the interstitial glasses in the overlying mush (Helz, 2020). To evaluate the termination of KI67-1, a later hole (KI75-2) was drilled 0.38 m away from KI67-1 to check the appearance of core from the 21.2–26.5 m depth interval, which had been partially molten in 1967. No textural anomalies were found: it appears that the crystal mush found at 26.5 m in 1967 collapsed when drilled, perhaps from the vibration of drilling itself, producing a small pocket of almost 100% melt, as was suggested in Helz and Wright (1983). But the melt in that pocket was differentiated, not equivalent in composition to the overlying mush, as shown in Table 2, which compares compositions of melt and lower crust from KI67-1 with those from KI79-3 (shown in Fig. 5C).

As discussed above (Fig. 4), most of the holes drilled between 1967 and 1979 terminated in pre-existing, melt-rich internal differentiates that defined the local crust-melt interface, as indicated in the field logs (Helz et al., 1984). When a melt-rich body was reached and the string backed off to retrieve the last interval of core, black material (designated “ooze” in the field reports) moved up the borehole within minutes of the string being raised. The history of this backfilling was reviewed in Helz and Wright (1983).

The commonest type was black, crystal-poor ooze that was finely but variably crystallized but opaque in thin section except for scattered plagioclase crystals. A photomicrograph of such ooze, adjacent to partially molten wallrock, was shown in figure 13B of Helz et al. (2014). The ooze cores from KI79-3 are shown in Figure 7, and bulk compositions of four ooze samples taken from KI79-3 are given in Table 3. They have segregation-vein compositions similar to those of the glass below the various crust-melt interfaces. In those samples, however, the glass is brown and translucent (see Figs. 5A5D and 6A6B), while the ooze is black and opaque. The crystallization of the ooze must have occurred as it rose within the (water-cooled) borehole wall; the temperature at which this occurred is uncertain, but the time available was less than an hour in most cases.

In 1979, repeated attempts were made to drill through the small, isolated melt bodies encountered in the hope of reaching the larger melt zone below the effective crust-melt interface (see, e.g., Colp, 1979), as it was thought that they were too thin to pose a serious obstacle to drilling. The 1979 drilling episode thus offers the most information on rheological instabilities at the base of the upper crust in Kīlauea Iki.

A series of efforts was made to extend KI75-1R, KI79-3, and KI79-4, three closely spaced drill holes (location A, Figs. 2 and 4), beyond the ~6-cm-thick segregation vein in which they all terminated, with the following results:

  1. KI75-1R was drilled to 52.76 m on 11 December 1978 and re-entered on 12 December 1978, with 1.76 m of ooze recovered from between 50.20 m and 51.96 m (see Fig. 4C; Helz and Wright, 1983). The hole backfilled immediately with ooze to 50.6 m and was abandoned.

  2. KI79-3 was drilled to 52.67 m on 10 January 1979. When it was re-entered an hour later, drilling began at 47.24 m, and 3.76 m of ooze was recovered from between 47.24 m and 51.0 m (the R1 ooze). The hole was re-entered 15 min later, with drilling begun at 48.70 m, and 2.13 m of coherent ooze was recovered with much fragmental material (the R2 ooze). Both are shown in Figure 7. After that, the hole backfilled to 48.74 m and was abandoned.

  3. KI79-4 was drilled to 52.46 m using standard coring techniques; on 18 January 1978, the hole was extended with a special non-coring bit to a depth of ~56.4 m (dashed box in Fig. 4), but the hole failed, backfilled, and was abandoned (Helz et al., 1984).

To what extent did repeated drilling drain this segregation vein? We can calculate the size of the depleted source using simple geometry, assuming the ooze and the depleted regions are both cylinders, as follows:


The diameter of the hole was ~8.5 cm, so the radius of the ooze is 4.25 cm. The total height of the ooze recovered was 610 cm, giving a volume of 0.035 m3 for the ooze. The segregation vein was ~6 cm thick, so the radius of the drained volume of the vein would have been ~43 cm, assuming 100% drainage. Thus, the two surges of ooze recovered from KI79-3 were derived from a pancake-like volume less than 1 m in diameter, and the vein immediately backfilled the hole a third time. The closely spaced drill holes in location A (Figs. 2 and 4) were 3–5 m apart, so draining the observed amount of ooze from boreholes KI75-1R and KI79-3 would not have significantly depleted their common source. It proved impossible to drain or chill this thin segregation vein sufficiently to drill past it.

This picture of local draining of a single segregation vein is supported by the chemical data on the KI79-3 ooze samples, as the ooze is uniform in composition throughout (Table 3). This suggests that the sole source was the small segregation vein, and there was no input from the deeper mush zone that existed below the vein in 1979. The thickness of this deeper layer can be estimated by looking at the segregation veins found subsequently in nearby borehole KI81-1: there is a vein at 52.82 m (the melt that blocked drilling in 1979) and another at 54.44 m (still glassy in 1981; see Helz and Wright, 1983), as shown at location A in Figure 4. The core from this later hole shows that there was perhaps 1.6 m of mush below the 1979 crust-melt interface, but access to it was effectively blocked by the 6-cm-thick segregation vein at 52.76 m. The observation that the deeper melts were not tapped has implications for the extraction of melts from within larger mush systems.

The lithostatic pressure on the segregation vein at 52.76 m in the 1979 drill holes, estimated using the density data of Johnson (1980), was calculated to be 1.5–1.6 MPa. This modest pressure drop was sufficient, given the low crystal content of the core of the segregation vein, to forestall further drilling, even though the source was only centimeters thick. Repeated re-entry of the boreholes, with coring and removal of ooze, never exhausted these pre-existing, melt-rich sources.

KI79-1 (location C, Fig. 4) was the first deep borehole in Kīlauea Iki that was not blocked by a discrete melt-rich body. As in all of the 1979 holes, the core recovered was ~6 cm in diameter. The hole was drilled without incident to 56.24 m by 20 December 1978, and when drilling resumed the next day, the hole required only modest reaming. The presence of a thin, melt-rich layer in the core, found 0.3 m below the overnight stop (Fig. 8), provides evidence for incipient instability of the mush, giving context for the behavior observed later. On 21 December 1978, the hole was drilled to 62.2 m, where it was stopped without having hit any significant melt pocket. The hole was again stable overnight, with only minor collapse at the bottom, so it was possible to take a temperature profile in the completed (but uncased) borehole to a depth of 61.4 m on the following day (22 December 1978; see Colp, 1979).

Over the next 16 days, KI79-1 backfilled not with black glassy ooze but with partly crystalline, olivine-rich ooze, which had reached a depth of 50.8 m when a thermocouple profile was taken down the hole. The hole was re-drilled three times on 25 January 1979, and internally differentiated and/or massive, olivine-rich ooze was recovered, as shown in Figure 9. The re-entry history is in Helz et al. (1984), and petrographic logs of the ooze are presented in Helz (1993). The character of this ooze versus that of the initial core recovered, the complex textures seen in the first ooze recovered, and the major shift in rheological behavior of the ooze between the initial drilling, the first backfilling, and the subsequent very rapid backfilling prior to the second and third re-entries, are the focus of the rest of this paper.

Field Observations of Re-drilling and Ooze Recovery

The first redrilled core, from a depth of between 50.8 m and 53.9 m (the second column from left in Fig. 9), moved up over a period of 16 days after termination of the original hole. After this interval was retrieved and the string lowered again, the bottom of the hole was found to be at 52.9 m; drilling recovered uniform, olivine-rich mush from 52.9 m to 54.25 m (column 3 in Fig. 9). After wire-lining this second interval of core, drilling resumed yet again at 52.9 m and an additional 3 m of ooze was recovered (column 4 in Fig. 9). This time, the bit had advanced to a depth of 55.4 m when the driller announced that the core barrel was blocked. When the 10 ft core barrel was pulled up, it was found to be completely full, suggesting that the crystal-rich mush was rising into the core barrel spontaneously during drilling. Individual pieces of ooze were heavily corrugated and twisted (Fig. 10), and their appearance is consistent with the idea that ooze was intruding into the core barrel as the bit advanced. At this point, the hole was abandoned; a thermocouple lowered down it on 17 February 1979 determined the bottom of the hole to be at a depth of 53.9 m (Helz and Wright, 1983).

Observations on the Differentiated First (R1) Ooze

The appearance of the first ooze (R1) is strikingly different from anything seen before from Kīlauea Iki, including in the original KI79-1 drill core. The upper part of the column of ooze contains a series of melt-rich layers of variable thickness that are bounded above and below by more crystalline mush (column 2, Fig. 9; details are in Helz, 1993). Most of the olivine-rich parts of the ooze contain small bodies of melt, as indicated in the sketch in Figure 9, that extend almost to the bottom of R1. The melt-rich spots gave the R1 core a superficial resemblance to the partially molten, foundered crust seen in the 1967 cores, which was designated as “leopard rock” by R.S. Fiske (see 1967 logs in Helz et al., 1984). This contributed to the early impression that hole KI79-1 had bottomed in a block of foundered crust (Hermance and Colp, 1982). Subsequent drilling in 1981 and 1988 showed that, although there are blocks of foundered crust with such textures deep in Kīlauea Iki, there are none at the 50–60 m level where this ooze was recovered (Helz and Wright, 1983; Helz, 1993).

The four melt-rich layers (two thick and two thinner layers all in the upper part of R1) have a combined thickness of 0.96 m. The R1 ooze is 3.05 m long, so the melt-rich layers make up 31% of its length. The melt layers contain very fine crystals in a matrix of brown glass, so they were cooled below their liquidus prior to drilling. They do not contain any olivine phenocrysts but do include rare crystals resembling those in segregation veins. Contacts between the melt-rich layers and the olivine-rich matrix are sharp and appear to be intrusive, and there is no indication of exchange of material between the layers.

Textures in the olivine-rich R1 ooze are shown in Figure 11. Each sample contains abundant olivine phenocrysts in a matrix of finer-grained mush. Crystalline phases in the mush are olivine + clinopyroxene + plagioclase, with rare Fe-Ti oxides, and the melt is continuous throughout the sections. Two sections (Figs. 11A11B) are from the same piece of core (one cut horizontally and the other vertically); this piece is just below the deeper section of the thick, melt-rich layers. Both sections show a range of sizes and shapes for the melt-rich bodies. In the two deeper samples (Figs. 11C11D), the melt is more mafic and hotter (Table 4). The blebs in Figure 11C are smaller, and their shapes simpler, while they are essentially absent in the deepest sample (Fig. 11D). Vesicles in the melt blebs and elsewhere are sparse in all sections. Together the four slides suggest that the blebs were moving slowly upward relative to the enclosing mush and that some coalescence of blebs has occurred. Also, the bleb shapes and the locations of vesicles found in some blebs suggest that the blebs were still moving upward just before the core was quenched.

Another process observed is the clumping of groundmass crystals into tight aggregates near or interstitial to the olivine phenocrysts in samples from 52.36 m to 53.89 m, as shown in Figure 12. These two disturbed mush textures (the large blebs of melt and the tight crystal clumps) resemble textures observed in the vorbs, the cross-cutting, cylindrical diapir tracks found in all holes with core between 40 m and 58 m (see Table 1).

The crystal clumps in R1, plus details of the vorbs’ disrupted textures (anastomosing melt channels coupled with unusually coarse olivine and dense clumps of groundmass crystals), are shown in Figures 12A12C. Larger fields of view of the same vorb can be seen in fig. 2B of Helz et al. (1989) and fig. 12A of Helz et al. (2014) The crystal clumping in R1 (Fig. 12A) is locally very tight, but contacts between the clump and more melt-rich areas are not as sharp as in the vorbs. The R1 ooze differed somewhat from a vorb, as it moved up within a borehole and past undercooled walls, but the coupling of melt segregation (blebs) and the clumping of groundmass crystals is similar.

Melt-rich bodies and segregations have been found in the lower crust of the lava lake, and they also appear to be associated with the vesicle streaming process that formed the vorbs, as summarized in Table 1. Figure 13A shows two small blebs, and Figure 13B shows a closer view of a large bleb. Both are from R1 for comparison with “speckled rock” samples (Figs. 13C13D) from deep in the lake. The blebs are similar in thin section, where both the R1 blebs (Fig. 13A) and the blebs in the speckled rock (Figs. 13C13D) appear to displace (at least slightly) the crystals in the surrounding mush. This texture gives the impression that these blebs moved up within the mush, presumably because they were less dense than the mush. In contrast, the larger bleb from R1 (Fig. 13B) has irregular, wavy boundaries within the enclosing mush and contains some groundmass crystals in its center. These may have been derived from the walls of smaller blebs as they coalesced into the larger one.

Compositional and thermal constraints on the development of the R1 ooze are provided by the compositions of the melt-rich layers near the top of the ooze, which are appreciably more differentiated than normal segregation veins. With bulk MgO contents of 3.42–3.48 wt% (Helz et al., 1994), they fall in a gap in the array of whole-rock compositions found at Kīlauea Iki (see fig. 8 of Helz, 1987) and correspond to melts with liquidus temperatures of ~1082–1085 °C (Helz and Thornber, 1987). Such melt is segregated at the hand-specimen scale nowhere else in the recovered core.

The source of the melt that created these layers was most likely interstitial melt from the adjacent wall rock that moved sideways into the olivine-rich mush as it crept up the borehole. The interstitial melt present in KI79-1 between 52.7 m and 53.3 m has the right composition, but there is only 20 vol% melt in the matrix (see Table 5). However, this interval includes a segregation vein (at 52.97 m) and a vorb (at 53.3 m). The vein, correlative with the one that blocked hole KI79-3 (see locations A and C in Fig. 4), contained appreciable melt, although it was cooler (1088 °C versus 1110 °C) and somewhat thinner than the vein at location A. It seems likely that the melt-rich layers in the ooze were derived largely from this vein, as they contain rare crystals with vein habits. The vein did not flow directly into the open borehole, but moved in only after the olivine-rich mush had risen to its level (see column 2 in Fig. 9). The presence of two crystal-poor layers suggests that there were two distinct pulses of lateral intrusion, separated by a day or so, as the ooze crept upward at a rate of roughly 0.3 m/day. There is little mingling between the olivine-poor melt layers and the enclosing olivine-rich mush, which suggests that the vein intrusions were somewhat cooler than the olivine-rich mush.

The relatively high bulk iron content of the crystal-poor layers (Helz et al., 1994) is consistent with their being derived from a relatively olivine-poor source like the segregation vein (Helz, 1987). Because this melt migration produced almost 1 m of ooze (~0.006 m3) from a very thin vein (~3 cm thick), there must have been some lateral migration of melt within the segregation vein over a total lateral distance of perhaps 24 cm. The amount of melt production from the vein next to R1 was minor, however, compared with the 6.1 m (~0.035 m3) of ooze derived from the vein that blocked hole KI79-3 (discussed above), which was thicker as well as 20 °C hotter.

To summarize, the appearance of the R1 ooze is consistent with its having formed during slow upward creep over the 16 days that elapsed between completing KI79-1 and the re-entry on 6 January 1979. The melt blebs are present in the first-run core recovered, diminishing in size and abundance downward, but nothing like them is present either in the original core (column 1 in Fig. 9) or in the other sections of ooze recovered (columns 3 and 4 in Fig. 9). The variation in bleb development with depth in R1, plus the complete absence of this texture in the R2 and R3 oozes, argue for its having developed during the (unique) slow rise of ooze in the borehole. The abundance of blebs in the upper part of R1 suggests that the blebs rose relative to their enclosing mushes, so that the upper part of R1 is enriched in the differentiated melts they contain relative to the lower sections. There is no textural evidence for them joining the discrete, melt-rich layers.

Comparison of the melt compositions and quenching temperatures of the R1 samples (Table 4) with the compositions and melt quenching temperatures of the segregation-derived oozes in Tables 23 shows that the flow differentiation seen in the R1 ooze occurred at distinctly lower temperatures (and hence higher melt viscosities) than temperatures seen in segregation veins and other structures relating to their formation (T = 1100–1140 °C, Table 1). The internal differentiation seen in the R1 ooze, though resembling textures developed elsewhere in the lava lake, is unique to this setting.

Observations of Oozes Recovered in the Second and Third Re-entries of KI79-1

The oozes recovered during the second and third re-entries of KI79-1 (Fig. 9) were relatively featureless olivine-rich crystal mushes with MgO contents ranging from 16.41 wt% to 19.03 wt% (Helz et al., 1994). The second pass (R2, column 3 in Fig. 9, recovered between 52.9 m and 54.35 m) was uniform in appearance. The third pass (R3, column 4 in Figure 9, recovered between 52.9 m and 55.4 m, but with a full 3 m of ooze in the core barrel) contains a patch of vorb, from the middle or deepest of the three vorbs seen in the original core, and also (at a recovery depth of 54.4 m) a block of aphanitic material incorporated in a more plastic matrix.

In plotting data from the R2 and R3 passes, their recovery depths must be corrected, as the oozes in the two later passes clearly rose from deeper than their recovery depths (see arrows in Fig. 9). For the second pass, the simplest adjustment is to assume that the top of R2 was just below 53.9 m and to add 1 m to the depths for all samples from R2. The top of R3 was just below 54.35, which suggests that samples from R3 should have 1.4 m added to their recovery depths. This correction assumes that the R2 and R3 oozes rose from directly below the surviving part of the borehole.

Temperature information for hole KI79-1 (Fig. 14) includes two thermocouple profiles (Colp, 1979) and glass quenching data for all glassy samples (Helz, 2020), plus estimated temperatures for two partly devitrified oozes where the mesostasis was fine-grained enough to obtain analyses that appear close to expected glass compositions for MgO and CaO, though it contains too many very fine-grained Fe-Ti oxides for the FeO and TiO2 contents to be reliable. Glass and devitrified-mesostasis compositions for oozes from the three re-entry passes are presented in Table 4 with apparent quenching temperatures (uncertainty ± 8 °C; Helz and Thornber, 1987).

The thermal data for the oozes fall between the glass quenching temperatures of the original core (on the right) and the thermocouple profile taken the day after the drill hole was completed (on the left). The 22 December 1978 profile shows the maximum effect of the cooling water on the wall rock, and the fact that the profile extends to 61.4 m shows that the mush in the wall rock did not collapse immediately after drilling was completed. All R1 ooze quenching temperatures fall smoothly on the extension of the temperature profile obtained on 6 January 1979 and fall slightly below the glass quenching temperatures of the original KI79-1 core, which suggests that they had equilibrated thermally prior to being drilled on 25 January 1979.

For the R1 samples, the initial recovery depths are appropriate, but for samples from R2 and R3, the depths were adjusted as discussed above. Using the adjusted depths, temperatures from the top pieces in R2 and R3 fall slightly below the January thermocouple profile, perhaps reflecting the effects of the cooling water used during renewed drilling, but the bottom pieces are close to temperatures in the original drill core at the corresponding depths. Interestingly, the estimated temperature for the aphanitic block in R3 falls close to the borehole temperatures observed one day after drilling. Its corrected depth places it near the position of the overnight stop in drilling KI79-1, which was at 56.24 m on 20 December 1978. This suggests that the block is a chunk of the chilled margin of the original borehole. If so, its interstitial melt was quenched to a glass and then devitrified prior to recovery on 25 January, which meets the definition of devitrification used by Burkhard (2001) and Deardorff and Cashman (2017). Its unique and conspicuous devitrification suggests that the rest of the R2 and R3 oozes are not derived from the original chilled margin.

Nature and Behavior of the R2 and R3 Oozes

The central question posed by the behavior of these later oozes is why were these mushes so very mobile when compared with original core? Figure 15 shows groundmass views of four samples: two from the KI79-1 core, both from below the depth of the overnight stop on 20 December 1978, with sections of the two deepest samples from R2 (KI79-1R2-177) and R3 (KI79-1R3-181.3).

The photomicrographs of the original core samples (Figs. 15A15B) show that their groundmass consists of olivine + augite + plagioclase in abundant brown glass. The crystal network is very fragile, but apparently continuous, as the walls of the borehole did not collapse overnight. The groundmass appears to have a cellular structure with networks of crystals surrounding pools of melt; this is most visible in sample KI79-1-189.0 (Fig. 15B). Acicular plagioclase may be a crucial part of these fragile networks, as was found by Philpotts and Carroll (1996) and Philpotts et al. (1998). However, in these samples, groundmass augite appears to be present with plagioclase in the crystal chains.

The samples from the bottom of R2 and R3 (Figs. 15C15D) look superficially like the original core in that they consist of olivine + augite + plagioclase in a matrix of abundant brown glass. The textural differences are subtle, but the grains in the ooze samples are more rounded, and grain-to-grain contacts, especially between different phases, are largely gone. These textural changes are clearer in the backscattered electron images (Figs. 15E15F). In Figure 15E, there appear to be plagioclase chains (per Philpotts and Carroll, 1996; Philpotts et al.,1998). But the erosion of plagioclase crystals is progressive and more extensive in the R3 sample, even though the quenching temperature of the melt in that sample is lower than in the original core.

These oozes differ from the original core in more than groundmass texture. Figure 16 shows how the bulk compositions of ooze and original core samples compare for Na2O, TiO2, K2O, and P2O5 (all are plotted against bulk MgO content). The Na2O levels in the oozes, like most other major oxides, are the same as those in the original core, at the same MgO content. However, the somewhat more incompatible elements (here including TiO2, as some of these samples were quenched above the temperature at which the Fe-Ti oxides begin to crystallize) are enriched in all of the R1, R2, and R3 ooze samples relative to the original core samples. The extent of enrichment is similar for all except the shallowest olivine-bearing R1 ooze (sample KI79-1R1-170.9), which is notably enriched in K2O and P2O5 relative to the others. The simplest explanation is that the oozes contain more melt than the original core did.

Data on the melt content (in volume percent) of samples from several boreholes in Kīlauea Iki are given in Table 5, and the results are plotted against depth in Figure 17. These data were obtained by making point counts on polished thin sections in reflected light to minimize the masking of melt by crystals during the count. Table 5 also includes data on vesicle contents, which were collected as part of the same effort. For the 1975 and 1976 data in Figure 17, the range of melt contents (14.3%–44.4%) is similar. The original KI79-1 core samples show melt content increasing with depth, steeply at first and then (below ~58 m) more gradually, similar in form to the shape of the temperature profile seen in Figure 14. The maximum amount of melt observed is 40.4 vol%, somewhat lower than the 41–44 vol% observed in the deepest samples from the 1975 and 1976 drill cores.

Oozes from R1, R2, and R3 are enriched in melt relative to the original core from the same depths. The deepest R3 sample has 43 vol% melt, at but not beyond the normal limit of core recovery. Nevertheless, this and the R2 material rose up the borehole in minutes, possibly just behind the drill string as it was backed off while wire-lining the core. It appears that there was movement of excess melt (that is, melt moving relative to crystals) into the mush near the base of the borehole, presumably in response to the drop in pressure (1.5–1.6 MPa) that occurred near the base of the open, uncased borehole. This melt migration inflated the mush and resulted in the preferential erosion of grain-to-grain contacts, as can be seen in Figures 15C15F, which destroyed the coherence of the original crystal network. The quenching temperatures of the glasses in these inflated mush samples are similar to those in the original core or slightly lower (Fig. 14). The observed resorption of grain boundaries thus appears to be an example of reactive transport triggered by a pressure gradient, which is similar to that observed in the experiments of Pec et al. (2015).

A wide range of phenomena was observed during re-entry of the Kīlauea Iki drill holes and the recovery of ooze. Before further discussion of their causes, however, it may be useful to review the physical parameters controlling the ooze surges seen in the 1979 drill holes. These include pressure, temperature, and volatile content, plus constraints on the depth of origin of the KI79-1 oozes.

Constraints on Oozing Behavior

Two basic controls on ooze behavior are the lithostatic pressure at the base of the open hole and the temperature of the ooze source. Pressure varied only slightly (1.5–1.8 MPa) from borehole to borehole in 1979, while temperatures varied from 1108 °C in KI79-3 to 1120–1131 °C in KI79-1 and KI79-6. As for the volatile content of the mush, during the 1979 drilling emission of sulfur from the boreholes during drilling was negligible, in contrast to the heavy sulfur emissions observed in 1967, 1975, and 1976 (Helz and Wright, 1983). Thus, the melt retained very little of its original volatile content, as can be seen in Figure 18, which shows the vesicle contents for samples in Table 5 plus data for two more cores (KI79-3, from Mangan and Helz, 1986; KI81-1, from Barth et al., 1994). Almost all samples from 40 m to 70 m have vesicle contents of 1%–3% by volume, perhaps decreasing slightly with depth. This sparse through-going population would have been filled by dilute steam, mostly of meteoric origin, prior to drilling in 1979.

Four samples from KI79-1, which were closely exposed to unloading just prior to quenching, have higher vesicle contents. One is the sample seen in Figure 8, from 0.3 m below the overnight stop, another is the next sample below it (shown in Fig. 15A), and the others are the lowermost pieces of ooze in R1 and R3 (Figs. 15C15D). These samples contain 3.6%–4.7% vesicles, not frothy but nevertheless perceptibly more vesicular than the general population (Fig. 18). In contrast, the only original core sample (Fig. 18) quenched without prior unloading but that has comparable vesicularity is an ogb (KI81-1-225). Its vesicularity is consistent with the inference that these bodies are, like the vorbs and speckled rock (Table 1), somehow related to vesicle streaming in the lake.

Ascent Rates and Viscosity of Oozes in Kīlauea Iki Boreholes

Observations on the timing of redrilling and extent of backfilling are presented in Table 6. These data allow calculation of the rate of rise of ooze for three of the 1979 boreholes. The most conspicuous result is that almost all oozes rose at rates between 1.0 × 10−3 m/s and 4.2 × 10−3 m/s. For comparison, the rate of propagation of the intrusive body that moved down Kīlauea’s east rift before the 2018 eruption was 0.7 × 10−2 m/s (Neal et al., 2019), that is, one or two orders of magnitude faster than the ooze surges up open boreholes. The difference in velocity reflects the facts that the cross-section of the 2018 body was larger and its flow aided by gravity, whereas the borehole oozes were rising against gravity.

The Table 6 results also allow calculation of the viscosity of the oozes, using the equation for simple pipe (Poiseuille) flow:


where µ is viscosity, ΔP is the pressure drop, r is the radius of the borehole, V is the ascent rate of the ooze, and L is the height of the ooze. Estimated viscosities for the crystal-poor oozes and for the R2 and R3 oozes from KI79-1 range from 0.4 × 104 Pa s to 5 × 104 Pa s. Estimated viscosities for the slow-rising, internally differentiated R1 ooze are 4–9 × 106 Pa s, two to three orders of magnitude higher.

The limited variation in pressure, temperature, and dissolved volatile content in the ooze sources are reflected in the limited range of height to which the oozes rose, which ranges from 47.3 m to 52.7 m in most cases. The one exception is the final surge in KI79-6, which came from the deepest and hottest source: it rose up >12 m while the drill string was out of the hole so that the bit could be replaced. Even there, however, the overall rate of rise is similar to most of the other surges, where the time elapsed before re-entry is more closely constrained.

A final question involves constraints on the depth of origin of the three runs of recovered ooze from borehole KI79-1. The overall redrilling history shows that the original hole walls were preserved to a depth of 53.9 m (Fig. 9), so that the top of the R1 ooze must lie immediately below that depth. The presence of an isolated block of strongly devitrified material at a reconstructed depth of 55.8 m (near the depth of the penultimate overnight stop at 56.2 m, both shown in Fig. 14) shows that at least some material in the R3 ooze came from that depth. These bracketing depths are used in Table 6 as the maximum range of starting points for the R1 ooze. Even hypothesizing the deeper source, however, the rate of rise of the R1 ooze is slower than any of the other surges in Table 5, and its estimated viscosity is higher.

Processes and Possible Causes of the Behavior Exhibited by the 1979 Oozes

Physical, petrologic, and chemical phenomena observed in the various oozes include:

  1. Melt surging up a borehole as a consequence of unloading melt-rich bodies: This behavior may pose significant problems for drilling into magma (as proposed in e.g., Eichelberger, 2019) unless melt distribution within the body being drilled is uniform. It appears that a sharp change in melt content combined with even a small pressure drop will be unstable at temperatures above 1000 °C.

  2. Melt in crystal-poor oozes is derived from the melt-rich body only, without contributions from adjacent interstitial liquids. This has implications for the origin of magmas from bodies with heterogeneous melt distributions, as suggested by Sinton and Detrick (1992) and Barth et al. (1994), who hypothesized that thin sills at the top of or within mush zones were the effective immediate source of mid-oceanic-ridge basalts.

  3. Production of melt-rich layers within previously uniform mush. The two processes involved are unloading coupled with the lateral migration of melt from wallrock into the mush. The simplest example is seen in sample KI79-1-185.5 (Fig. 8), where the melt-rich layer was generated in response to unloading of the hot, melt-rich mush under a pressure drop of 1.6–1.7 MPa. The melt in the layer is “extra,” raising the total melt content of sample KI79-1-185.5 from 32.5 vol% to 39.8 vol% (Table 5), well above that of adjacent samples (Table 5). The higher vesicle content of this sample also reflects the overnight unloading. In spite of these small-scale responses, however, the sample did not rise up into the open borehole. In contrast, the melt-rich layers in the R1 ooze were produced by the lateral injection of melt from a segregation vein in the wall rock and occurred as the olivine-rich mush crept past the vein. The timing of the injection is unequivocal, as the two main melt-rich layers were recovered 1 m and 2 m above the level of the source vein.

  4. Segregation of melt into blebs within the mush. Perhaps the most widespread process of melt segregation within partially molten mushes, both in R1 and in Kīlauea Iki, has been the production of melt spots or blebs. This may be a slower process than the formation of melt layers, as it was not observed in sample KI79-1-185.5 (Fig. 8). The process is pervasive in the R1 ooze, increasing from the base of R1 to just below the uppermost melt layer (Figs. 11A11D), and is widespread in the lower mush zone (Table 1 and Figs. 13C13D) of the lava lake. The textures of blebs in R1 suggest they have risen relative to the enclosing mush and that coalescence into larger but highly irregular blebs has occurred. This enrichment in melt in the shallower R1 samples is supported by the observation that the shallowest ooze in R1 is most strongly enriched in K2O and P2O5 relative to the original core (Fig. 16).

    The rise of R1 took place over 16 days, whereas in the lava lake, melt blebs had longer lifetimes prior to quenching. Thus, once segregated, they tended to organize into vertical arrays (Fig. 13C), even coalescing into melt chimneys similar to those seen in experiments by Tait and Jaupart (1992) and Tait et al. (1992). Their formation appears to be related to the local upwelling of melt and vesicles deep within the lava lake. However, the time and temperature of their formation is not well constrained: in 1981, temperatures ranged from 1130 °C to 1090 °C at depths of 77–94 m, but many of the actual samples were not recovered until 1988, when they had cooled to temperatures of 1100 °C or less (Helz, 2020).

    How did the parallel R1 and lava lake textures develop? Two possibly related phenomena from the literature are the generation of segregation vesicles (Anderson et al., 1984; Goff, 1996) and the gas filter pressing triggered by second boiling of Sisson and Bacon (1999). More recently, Fowler et al. (2015) modeled three phase (gas, liquid, and solid) upwelling within a vapor-saturated mush pile. In their modeling, the buoyancy that triggers the formation of vertical cylindrical structures is provided by steam bubbles exsolving from the (water-saturated) interstitial melt, a condition that existed in Kīlauea Iki.

    This later model is more like the Kīlauea Iki situation, in which the phenomena develop deep within the lava lake, in a melt-rich environment, at temperatures near or above 1100 °C, where the melts and mushes involved flow plastically (as was clearly true for ooze R1; see especially Fig. 13B). The simpler gas filter-pressing mechanism (Anderson et al., 1984; Goff, 1996) depends on there being a difference in pressure on the gas phase relative to the interstitial melt, which is difficult to achieve when the matrix is plastic rather than rigid. It seems unlikely that this precise mechanism could have operated within the lower Kīlauea Iki mush zone at the temperatures observed, as the lake was not rigid, but subsided continually from 1960 to 1988 (Pallon et al., 1994), as degassing and crystallization proceeded. Furthermore, it is difficult to define a “second boiling” in Kīlauea Iki: the lava lake always had a sparse population of vesicles migrating through it (Fig. 18), though they were maintained from 1979 on by access of meteoric water.

    Admittedly, how the melt blebs (as seen in the R1 ooze and the lava lake) formed is not completely clear. For example, Figure 13D shows a swarm of tiny melt blebs with many separated from each other by very thin septa of groundmass crystals. If one hypothesizes that the blebs were formerly vesicles filled with dilute steam, which have been subsequently filled by the brown silicate melt now present, the question arises as to how the fragile walls now separating the blebs survived the replacement of dilute steam by silicate melt. It seems more reasonable to suggest that the melt blebs (with or without small vesicles) are primary and were not formerly gas-only voids. In any case, once the blebs form, they move up on their own, as implied by the Fowler et al. (2015) model, whether coalescing (as in R1, Figs. 11A11B) or not (Fig. 11C). They do not disperse back into the matrix mush.

  5. Segregation of crystals from melt into crystal-rich clumps. This process appears to lag the formation of melt spots, having been observed mostly in the lower parts of ooze R1. In Kīlauea Iki itself, crystal clumping is characteristic of the vorbs, which are found between 40 m and 58 m, but is absent from the various disrupted textures observed in the lower mush zone (Table 1; also, Helz, 1993). The pattern suggests that crystal clumping requires more extended transport than does melt segregation into blebs and that it perhaps facilitates this transport by minimizing internal friction within the moving mush.

    All textures are exhibited either within oozes that rose up with the KI79-1 borehole or in vertical pipe-like bodies in otherwise unconstrained, plastic lava lake mushes. These observations appear to be consistent with other predictions from the experimental and theoretical literature. Spiegelman et al. (2001) modeled flow organization during melt transport in a deformable medium and discovered that the process created pipe-like, melt-rich channels separated by impermeable (crystal-rich) regions. The scale on which melt and crystal segregation is operating in the vorbs and oozes is much smaller than that of the mantle processes envisioned by Spiegelman et al. (2001), and the mushes in the lava lake are perhaps more easily deformed than mantle rocks. Nevertheless, the resemblances suggest that the coupling of these processes will occur over a wide range of conditions. Also, the present triggers (1) rising, self-organizing vesicle plumes, as proposed in Helz et al. (1989) for the vorbs, which are similar to the model results of Parmigiani et al. (2016) for other systems, or (2) ooze rising in boreholes in response to unloading, which is unique to the low-pressure environment of the lava lake, but the textural results are broadly similar. The processes seen here are governed by vertical unloading with minor enhanced vesiculation and some shearing intrinsic to diapiric upwelling, but without the significant lateral shear frequently used in laboratory experiments investigating melt segregation (see, e.g., Kohlstedt and Holtzman, 2009).

  6. Inflation of mushes as melts move relative to crystals down a pressure gradient. This process explains the behavior of the R2 and R3 oozes, which differs so markedly from the original KI79-1 core. The original material held its shape against the open borehole over the nights of 20–21 December 1978, with minor relaxation and additional vesiculation 0.3–1.0 m ahead of the bottom of the hole (Fig. 18). After being unloaded again, on 25 January 1979, the olivine-rich mush rose rapidly and repeatedly, such that the effective base of the crust was 54 m as opposed to the 62 m observed in December.

The geometry of the former drill hole is shown schematically in Figure 19. The melt content of the original wall rock at the new base of the crust was ~25 vol% (Fig. 17). The arrows in the sketch show the melt moving into the previous volume of the hole sideways from all directions as opposed to flowing up a stable hole. This is because the R2 and R3 ooze, though enriched in melt, appear to contain very little of the original chilled margin of the hole, which suggests that the walls were effectively gone.

The melt content at the new base of the crust is similar to the >26 vol% melt considered by Sawyer (1994) to be the upper limit of melt that could be accommodated within a stable mush zone. Beyond that point, his analysis showed that the continuity of the matrix breaks down and spontaneous diapiric upwelling occurs. Sawyer was modeling melt segregation in crustal compositions, with rising temperatures, so it is remarkable that the results observed in the picritic mushes found in Kīlauea Iki (with falling temperatures) show the boundary between stable crust and instability at nearly the same melt fraction.

The R2 and R3 oozes are enriched in melt relative to the original core recovered at the same depths. That and observed textural changes suggest that they exhibit a type of reactive transport by which inflated mushes, because of their corroded grain-to-grain contacts, behave as fluids. Their calculated viscosities (Table 6) are similar to those of oozes with low crystal contents, which is consistent with this interpretation. This conspicuous change in behavior occurred at 35–45 vol% melt fraction (55–65 vol% crystallinity), near the boundary between magma and rigid partially molten mush, which is usually set at 40% melt and 60% crystals (see, e.g., Cashman et al., 2017).

This behavior is similar to the reactive transport in response to a pressure gradient, as seen in the experiments of Pec et al. (2015), so it differs somewhat from the style of reactive liquid flow that mainly involves reintrusion of a mush by hotter melt (e.g., Leuthold et al., 2014). It also resembles the “yield surface model” for mush behavior during an eruption, in which the magma near the conduit is hypothesized to inflate and behave as a liquid, whereas magma farther away from the pressure drop is unaffected and remains nearly rigid (Karlstrom et al., 2012; Parmigiani et al., 2014). As in those models, there is no evidence that the change in mush behavior in R2 and R3 involved the injection of hotter melts from deeper in the lava lake.

How widespread are the phenomena observed and processes inferred in Kīlauea Iki? Sill-like structures like the segregation veins have been widely recognized in mafic intrusives, but reports of most of the other structures observed here are rare. If the presence of segregation veins indicates that diapiric processes like those observed in Kīlauea Iki were active, why do we not see their traces? One obvious explanation is that the vorbs, melt chimneys, and speckled rock diapirs are easy to recognize in partially molten drill core but much more difficult to recognize in subsolidus material and in natural outcrops (other than the very cleanest of surfaces).

Comparison of Kīlauea Iki’s internal structures with those seen in the cliff face of the prehistoric Makaopuhi lava lake illustrates the problem. Both bodies contain a trellis of large, laterally continuous segregations that are present from 14 m to 23 m in the prehistoric lake (fig. 14 of Helz et al., 2014) and 18–56 m in Kīlauea Iki (Table 1). Moore and Evans (1967) also found short, vuggy, olivine-rich pipes in the lower part of the Makaopuhi lake, and the one analysis shows compositional affinities with the vorbs, though their physical position and limited vertical extent (a few centimeters) suggests they are more like the melt chimneys in Kīlauea Iki. No vertical structures were recognized at higher levels. However, the prehistoric Makaopuhi lava lake has a bulk MgO = 9.7 wt% and olivine phenocrysts are scarce in the upper parts of the lake, so any diapiric structures would be difficult to recognize without the olivine signature, in subsolidus rock, and in the natural outcrop. Another body, the 1965 Makaopuhi lava lake (bulk MgO = 8.2 wt%), showed some development of segregation veins at 13–17 m, but as we have no core deeper than 18–20 m, its deeper structures and later crystallization history are unknown.

One occurrence of well-developed vertical segregations is found in the Jurassic North Mountain Basalt of Nova Scotia as described by Kontak and Dostal (2010). They refer to the bodies found there as segregation pipes, and they occur in the lower part of the uppermost (Brier Island Basalt) flow, the thickness of which (~151 m) is slightly greater than that of Kīlauea Iki (Helz, 1993). Though these segregations are crystallized and altered, they are clearly vuggy, vertical pipes containing coarser crystals and more vesicles than the host basalt. As they are restricted to the lower part of the flow, they appear to be analogous to the melt chimneys and speckled-rock diapirs seen between 77 m and 94 m in Kīlauea Iki. Like those bodies, the segregation pipes include some coarser and possibly entrained crystals within a finer matrix. The pipes are one to a few meters high, and their diameter ranges from 2 cm to 3 cm (both dimensions are similar to those of the vorbs in Kīlauea Iki) to 50–60 cm (broader than the structures in Kīlauea Iki). Some outcrops afford a plan view of the pipes, where they very strongly resemble the melt chimneys of Tait et al. (1992). The areal density of the pipes at one locality averages five pipes/m2. This is intriguing, as it is consistent with observations from locations A and B in Kīlauea Iki, where each drill hole within these closely spaced clusters contains one or more vorbs.

It may be that awareness of the possible existence of these structures will lead to their wider recognition. Another example of a possibly unrecognized phenomenon may be the very tight crystal clusters generated during upward diapiric flow: these might be expected to survive in larger magma chambers whether they are volcanic or intrusive. If their internal textures are unique, like the various types of crystalline enclaves characterized by Holness et al. (2019), they may be recognizable in other bodies and therefore could extend recognition of that process.

This paper reviews how melts and crystalline mushes of varying crystal contents behaved in response to unloading, as observed during the 1979 drilling of Kīlauea Iki lava lake.

Most 1979 boreholes terminated in a melt-rich internal differentiate. In this situation, the holes were stable for less than an hour, with crystal-poor (<5 vol% crystals) melt surging up the open boreholes at 1.0–4.2 × 10−3 m/s. The ooze was derived exclusively from the melt-rich body with no contribution from the more crystalline mush.

One hole (KI79-1) was not blocked by a discrete melt-rich body. In this hole, the crystal-rich mush first showed signs of instability at ~56 m, though the hole was drilled to a depth of 62 m, and the hole remained open overnight. Over a period of 16 days, KI79-1 slowly backfilled with ooze to a depth of 50.8 m. When this hole was re-entered, crystal-rich oozes of varying character recovered. The first surge of ooze recovered in KI79-1 (R1, which had crept up the borehole over 16 days) displayed internal differentiation, including apparently coupled formation of melt-rich blebs within the rising, more crystalline mush, and incipient formation of crystal-rich areas. Similar textures are observed in naturally occurring diapiric bodies within the lava lake and appear to be intrinsic to the upwelling of narrow, cylindrical mush bodies.

Two oozes recovered immediately after the R1 ooze were homogeneous, olivine-rich mushes with crystal contents of 55%–65%. These rose at rates of 1.0–1.7 × 103 m/s and as quickly as the crystal-poor oozes. This conspicuous shift in rheology is attributed to melt moving relative to crystals down the pressure gradient created by the open borehole: groundmass textures in the resulting inflated mushes show the erosion of crystal outlines, with crystals floating in continuous melt. The effective base of the crust during the repeated re-entries of KI79-1 was 54 m (vs. the 62 m depth of the original hole), with the original hole apparently almost completely effaced below that depth. The transition from stable, quasi-rigid crust to unstable, inflatable, highly mobile mush occurred at 25%–28% melt by volume.

The results presented here depend on the work of the many persons from multiple institutions involved in drilling Kīlauea Iki lava lake. All available drill core from Kīlauea Iki, from 1960 through 1988, is now in the collections of the National Museum of Natural History, Smithsonian Institution, Washington, D.C., USA, and is curated by that institution. This paper was improved by reviews from C.R. Bacon (U.S. Geological Survey, Menlo Park, California, USA), B.J. Andrews (National Museum of Natural History, Smithsonian Institution, Washington, D.C.), an anonymous reviewer, and K.P. Cashman (University of Bristol, Bristol, UK), to whom special thanks are given for detailed and very helpful comments in her review of this paper.

Science Editor: Brad S. Singer
Associate Editor: Michael Ort
Gold Open Access: This paper is published under the terms of the CC-BY license.