Rapid Deformation of the South Flank of Kilauea Volcano, Hawaii

The south flank of Kilauea volcano has experienced two large [magnitude (M) 7.2 and M 6.1] earthquakes in the past two decades. Global Positioning System measurements conducted between 1990 and 1993 reveal seaward displacements of Kilauea's central south flank at rates of up to about 10 centimeters per year. In contrast, the northern side of the volcano and the distal ends of the south flank did not displace significantly. The observations can be explained by slip on a low-angle fault beneath the south flank combined with dilation deep within Kilauea's rift system, both at rates of at least 15 centimeters per year.

per cubic centimeter. We determined the aerosol size distribution and the surface-area concentration by measuring the aerosol number density and the extinction of ultraviolet light (200 to 400 nm) by the aerosol. The aerosol is characterized by a log-normal radius distribution (10) given by dN = V log s exp { 2(1og S)2 d log r where N is the number of particles, rpk is the peak radius, and s is the geometric standard deviation. The H2SO4 aerosol was generated by passing a portion of the carrier gas over hot concentrated H2S04 and then into a cool region. The aerosol size was varied by changing the gas flow rate and the temperature of the liquid H2S04. We adjusted the composition of the aerosol by passing the aerosol through a conditioner containing a 59.9% H2SO4 solution (by weight) held at the same temperature as the flow tube reactor. The conditioner was either a 35-cmlong cylindrical tube with an inside diameter of 3 cm half filled with the solution or a coiled (-0.5 cm inside diameter by 100 cm long) glass tube coated with the solution. On the basis of model calculations of the liquid-to-aerosol vapor transfer rate within the conditioners, the composition of the aerosol exiting the conditioners was within 10% (by weight) of the conditioner liquid composition. For the experiments with HCI, the carrier gas was bubbled through a 59.9% H2SO4 solution (by weight) doped with 7.8 x 10-4 M HCI, before mixing with the aerosol. The aerosol stream was then passed through the coiled conditioner, which was charged with the same solution as the bubbler. The conditioner and bubbler transferred both H20 and HCI vapor to the aerosol and were maintained at the same temperature (± 1 K) as the reactor by circulating chilled methanol through their jackets. 12 14. In our experiments, [HCI] was 6.6 x 10-4 M and the activity of H20 was 0.12 (that is, [H20] -6.4 M). Therefore, the ratio of the second-order rate coefficient for CIONO2 + HCI to that of CIONO2 + H20 is approximately 1.5 x 105, which is comparable to the ratio extracted from measurements at 202 K (-1 x 105) (6). We did not account for a possible surface reaction between HCI-CIand CIONO2 (6). It is expected that the surface reaction will contribute <150% to the reactivity in these experiments. By using the measured e value and an estimate of 1 X 10-6 cm2 s-1 for the CIONO2 diffusion constant, we derive estimates for the hydrolysis and HCI rate coefficients of 1 x 104 and 1.5 x 109 M-1 s-, respectively, for 60% H2S04 at 250 K. The HCI rate coefficient is in the diffusion-limited regime. 15. J. I. Gmitro and T. Vermeulen, Am. Inst. Chem. Eng. J. 10, 740 (1964). 16. The results from the present study suggest that the HCI reaction is diffusion-limited at 250 K in 60% H2S04 solution [see (14)]. In this case, the rate coefficient is proportional to the product of the capture radius and the diffusion constant. If we assume that the capture radius for the Cl-+ CIONO2 reaction is independent of temperature, we calculate that e is 0.03 plm for CIONO2 hydrolysis in 60% H2S04 at 202 K. 17. D. J. Hofmann, Science 248, 996 (1990 (4,5) were caused by seaward motion on a subhorizontal fault that may coincide with the base of the volcanic edifice (6). Swarms of shallow earthquakes accompany intrusions of magma into either of Kilauea's two rift zones (Fig. 1), but these earthquakes are typically small (M ' 4) and nondestructive (7). Earthquakes south of the rift zones have focal depths of 5 to 13 km (7), leading some to suggest that slip occurs on multiple parallel faults (5). However, precise relative relocations place many of the earthquakes on a welldefined northward dipping fault (8 angle fault zone 7 to 9 km beneath Kilauea's south flank. The structure of the rift system at depths of more than a few kilometers is uncertain because few earthquakes occur below 5 km. This lack of seismicity might indicate that deformation is concentrated in the shallow rift system, or simply that elevated temperatures at depth inhibit brittle failure. Delaney and colleagues (9) suggested that subsidence observed on leveling lines across the rift zones was caused by steady dilation at depths of -4 to 9 km. Dvorak and colleagues ( 10) argued that deep dilation is not required and that the data can be explained by episodic shallow intrusions in combination with slip on a low-angle fault at 6 km. Horizontal deformation measurements on the north flank could distinguish between competing models of south-flank and rift zone kinematics. Electronic distance measurements (EDM) required line-of-site visibility, so that it was previously not possible to measure baselines on the heavily forested north flank of Kilauea (11,12). Furthermore, it was not possible to determine displacements relative to sites removed from active volcanism and seismicity. EDM data for the period 1983   south flank (13); however, the network was geometrically weak and the data do not uniquely determine the velocity field (14). With the Global Positioning System (GPS), it is possible to determine threedimensional displacements on both sides of the rift system with respect to a stable remote site (15). GPS surveys were conducted on Kilauea in August and September 1990, February and March 1992, and August 1993 (16). In order to constrain the GPS satellite orbits, the 1990 and 1992 data were processed in 3to 4-day orbital arcs (17), including data from between 3 and 10 permanent GPS sites in Hawaii, North America, and around the Pacific Rim. We fixed the coordinates of three fiducial sites to values determined from very long baseline interferometry (VLBI) (18) and then estimated satellite orbital parameters, the remaining station coordinates, tropospheric delay parameters, and integer cycle ambiguities (19). The 1993 ambiguity-fixed solutions used singleday orbital arcs and applied orbits determined by the International GPS Service (IGS) (20).
For each year, the individual multiday or daily solutions were combined with the use of a least-squares network adjustment. The resulting covariance matrices were scaled so that the residual sum of squares divided by the number of degrees of freedom equaled 1.0 (21). The residuals to the network adjustment provided a measure of data precision (Table 1). Average velocities for each station were calculated from the three annual solutions, with the assumption that the displacement rates are constant in time (22). Velocities were computed with respect to HP-6, a station on the island's stable northwest side (Fig. 2).
Our results (Fig. 2) show that the central south flank displaced horizontally at -10 cm/year (11.7 ± 1.0 cm/year at M801 and 9.1 ± 0.3 cm/year at KAEN). Stations south of the east rift zone moved roughly perpendicular to the rift axis. The oblique motion of KALA may have been affected by the eruption at Pu'u O'o, only a few kilometers to the west. Stations near the southwest rift zone (SAND, KOAE, and KAMK) moved approximately perpendicular to the southwest rift axis. In contrast to the rapid velocities south of the rift zones, velocities north of the rifts were small or insignificant. Even UWEK, on the northwest rim of the caldera, did not move horizontally. There are also steep gradients in velocity parallel to the rift system; velocities northeast of MOAN and southwest of KAMK are negligible. These gradients in velocity (strain rates) are dramatic. The extension rate across the caldera between UWEK and SAND, in a direction normal to the east rift zone, is -15 microstrains per year. The average extension rate across the east rift zone between FERN and KMOA is -5 microstrains per year. Locally, strain rates across the rift zone must be even higher. Extension rates across the rift system (for instance, HULU-GOAT and UWEK-SAND) are an order of magnitude greater than extension rates within the central south flank (for instance, PALI-APUA and GOAT-KAPU). One baseline, M801-KAEN, appears to have contracted, although the velocity of M801 is poorly determined. Velocity gradients parallel to the rift system (shear strain rates) are also large, averaging about 4 microstrains per year for the baselines HEIH-66YY and MOAN-66YY. In contrast, the maximum strain rates on the San Andreas fault are on the order of 0.3 microstrains per year (23).
Only four stations exhibited vertical displacement rates that were significant at two standard deviations: KALA, UWEK, HULU, and HEIH (Fig. 2) The fault beneath the south flank was modeled by two horizontal planes of shear dislocation in a homogeneous elastic halfspace (24), with slip normal to the middle east rift zone. We modeled the rifts with three opening dislocations positioned along the rift zones (Fig. 3). We set the depth of the fault plane (D) at 9 km, consistent with the seismic data discussed above, and adjusted the other parameters to fit the GPS data. The resulting model (Fig. 3) shows that the dilating rift zones extend from a depth (d) of 1 to 3 km down to the fault plane at 9 km. The fault slip rate and rift dilation rate are both 20 cm/year, except along the upper east rift zone where the rift dilation rate is 10 cm/year. In comparison, the average slip rate on the San Andreas fault is only 3.5 cm/year. The steep velocity gradients across the rift zones make it impossible to fit the data if only fault slip at 9 km depth is assumed (Fig. 4). Rift dilation alone also cannot fit the data. A combination of fault slip and dilation of the rift zone at depth effectively decouples the south flank at the rift axes and permits a reasonable fit to the data (Fig.  3). It is possible to fit the observed gradient across the rift zones without rift dilation, but only if the fault is very shallow (-3 kin). Such a model is inconsistent with the depths of major south flank earthquakes. Other model parameters are reasonably well constrained by the GPS observations. The dip of the fault must be subhorizontal in order to fit the modest coastal uplift. On the other hand, increasing the offshore extent of the fault and concurrently decreasing the slip rate to 15 cm/year does not seriously degrade the fit to the data. There is also a trade off between the height of the rift and the rate of rift opening. For example, it is possible to fit the data with rift dilation extending from 4 to 9 km if the rate of rift dilation is - 30 (Fig. 1) are the headwall scarps of a giant landslide (26). There is evidence from wave deposits and from sidescan sonar data that older Hawaiian volcanoes have collapsed in catastrophic landslides (27,28). A series of closed basins offshore of the Hilina Palis (29) (Fig. 1) must be young, because the high influx of volcanic sediments from Kilauea would quickly fill any depression (30). These observations, together with the high rates of south flank seismicity (Fig. 1) and deformation, have led some to suggest that Kilauea has the potential to fail in a catastrophic landslide (27). We observed the highest rates of deformation where the Hilina Pali faults are well expressed (Fig. 2). Papa' Seamount, the prominent linear feature in the bathymetry, has been suggested to form the westemn boundary of a Hilina megalandslide system (27). However, we find that the westemn edge of the modeled fault is east of this feature (Fig. 1). The eastemn edge is located near a small right step in the east rift zone at HEIH but does not correspond to a prominent bathymetric feature. The seaward edge of the basal fault plane is poorly defined by the GPS data and could be extended to the postulated landslide toe without seriously degrading the data fit. We are unaware of any evidence for lateral "ttear" faults at either the eastemn or westemn edges of the rapidly deforming zones, which suggests that the deformation pattemn has varied over time.
The GPS data allow us to distinguish between the two competing models for rift zone subsidence discussed above. A consequence of Delaney and colleagues' model of deep rift dilation (9) is a decoupling of the south flank, which we have shown clearly exists in the GPS results (Fig. 2). Dilation of the rift zones also has implications for the volcano's magma supply budget. The dislocation model presented here implies a 0.06 km3/year volume increase due to rift dilatation. This is roughly 60% of the long-term magma supply rate as determined from the accumulation of lavas and estimated changes in summit magma chamber volume (31), which demonstrates that rift zone dilation must be accounted for when computing magma supply rate.
The dislocation model is merely kinematic, and although it puts constraints on mechanical processes at work in Kilauea, it does not directly address the forces driving deformation (32). It has been postulated that gravitational spreading of dense residual melts (33) or dunite cumulates (34) may drive south flank motion. It has also been suggested that high fluid pressures in abyssal sediments trapped beneath the volcano promote slip on the basal fault (35).  The absence of a tear fault near HEIH suggests that, over sufficiently long periods of time, the south flank adjacent to the lower east rift also displaces seaward, which may occur seismically. The greatest potential hazard associated with the south flank of Kilauea is probably a tsunami, which could effect the more populated areas of the Big Island or indeed other islands. These could range in size from the tsunami that accompanied the 1975 earthquake (36) to a much larger event if the south flank fails in a catastrophic landslide. Analysis of baseline length changes determined by geodimeters between 1970 and 1990 suggests that the present rate of south flank motion is not anomalously fast. The south flank was probably moving faster in the years after the 1975 earthquake. It is notable that as south flank seismicity decayed to background values (10), the rate of south flank motion presumably decreased to 10 cm/year. More frequent monitoring is needed to evaluate changes in deformation rates on Kilauea's unstable south flank. ve inland and uplifts the rift flanks, which is also inconsistent with the data. Electroluminescent (EL) devices based on organic thin layers are one of the most promising next-generation flat panel display systems. These devices can be made into large-area, extremely thin full color displays that can be operated by batteries. The structure of the devices is simple, having organic thin layers sandwiched between two electrodes. Because organic layers can be formed by vacuum evaporation or solution casting, the fabrication cost could be less than that of the liquid crystal displays, which are today the most widely used flat panel displays. In organic EL devices, the generation of light is the consequence of the recombination of holes and electrons injected from the electrodes. Such carrier recombination in the organic emitter layer excites the emitting centers. In a chemical sense, the reaction of radical cations (holes) and radical anions (electrons) provides excited 1332 molecules that emit light as one of the decay processes. A variety of materials have been investigated as active materials (1)(2)(3)(4)(5)(6)(7)(8)(9), and a number of device structures (10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) have been proposed because it is important to find an appropriate device structure to maximize the carrier recombination efficiency.
The use of multilayer structures was first demonstrated by Tang and VanSlyke (1). They used a hole-transport layer for hole injection from the electrode into the electron-transporting emitter layer, which significantly improved the EL efficiency to 1.5 lm/W. Bright electroluminescence of over 1000 cd/mi2, which is high enough for practical applications, with a voltage below 10 V was achieved first. Their device consisted of two active layers with an organic holetransport layer and an electron-transporting luminescent metal complex layer. The hole-transport layer plays an important role not only in transporting holes but also in blocking electrons, thus preventing electrons from moving into the opposite elec-SCIENCE * VOL. 267 * 3 MARCH 1995 trode without recombining with holes. The recombination, therefore, mainly takes place in the emitter layer. Since this first device, various types of EL devices with a multilayer structure have been reported. For instance, a luminescent hole-transporting material can also be an emitter layer when combined with an appropriate electrontransport layer (4,9,13).
In the above examples, various emitting colors have been obtained, including the three primary colors of blue, green, and red. However, there are few white light-emitting devices owing to the lack of the organic fluorescent dyes with white fluorescence. We have therefore focused on developing white light-emitting EL devices and have recently developed one made from a polymer emitter layer doped with three kinds of fluorescence dyes, each emitting blue, green, or red to produce white light (17).
In this study, we developed a device structure with three emitter layers, each emitting in the different region of the visible spectrum to generate white light. In general, organic EL devices are composed of two organic layers, each transporting either holes or electrons. In such a device, one of the two layers functions as an emitter layer and the other as a carrier transport layer. Therefore, the emission color is determined by the fluorescence properties of the emitter material, and the light is of the particular color of the emitting material.
One way to add another emission color to such a device is to dope the emitter layer with a fluorescent dye of a different color. For example, it was demonstrated that, in a device with an electron-transporting greenemitting aluminum chelate emitter and a hole-transport layer, when the chelate layer is partly doped with a red-emitting fluorescent dye, the resulting light becomes a mixture of the lights from the two emitter materials, thus producing yellow light (12).