Skip Navigation


Journal of Petrology Advance Access originally published online on April 22, 2005
Journal of Petrology 2005 46(9):1747-1768; doi:10.1093/petrology/egi032
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Supplementary Data
Right arrow All Versions of this Article:
46/9/1747    most recent
egi032v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (3)
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by DREHER, S. T.
Right arrow Articles by LARSEN, J. F.
Right arrow Search for Related Content
GeoRef
Right arrow GeoRef Citation
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author 2005. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oupjournals.org

The Petrology and Geochemistry of the Aniakchak Caldera-forming Ignimbrite, Aleutian Arc, Alaska

S. T. DREHER1,*, J. C. EICHELBERGER2 and J. F. LARSEN2

1 DEPARTMENT OF EARTH SCIENCES, UNIVERSITY OF DURHAM, DURHAM DH1 3LE, UK
2 ALASKA VOLCANO OBSERVATORY, UNIVERSITY OF ALASKA FAIRBANKS, FAIRBANKS, AK 99775, USA

RECEIVED JANUARY 22, 2004; ACCEPTED MARCH 8, 2005


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 GEOLOGICAL BACKGROUND
 METHODS
 RESULTS
 DISCUSSION
 CONCLUSION
 SUPPLEMENTARY DATA
 REFERENCES
 
Aniakchak caldera, Alaska, produced a compositionally heterogeneous ignimbrite ~3400 years ago, which changes from rhyodacitic at the base to andesitic at the top of the eruptive sequence. Interpretations of compositionally heterogeneous ignimbrites typically include either in situ fractional crystallization of mafic magma and generation of a stratified magma body or replenishment of a silicic magma chamber by mafic inputs. Another possibility, silicic replenishment of a more mafic chamber, exists. Geochemical characteristics of the caldera-forming rhyodacite and several late pre-caldera rhyodacites indicate independent origins for each, within a maximum of ~5000 years prior to caldera formation. Isotopic considerations preclude derivation of the caldera-forming rhyodacite from the caldera-forming andesite. However, the caldera-forming rhyodacite can be explained as the residual liquid of a mostly crystallized basalt, with addition of crustal material. The Aniakchak andesite probably formed in a shallow chamber by successive mixing events involving small volumes of basalt and rhyodacite, together with contamination. The pre-caldera rhyodacites represent erupted portions of intruding silicic magma, whereas another portion homogenized with the resident mafic magma. The caldera-forming event reflects a large influx of rhyodacite, which erupted before significant mixing occurred and also triggered draining of much of the andesitic magma from the chamber.

KEY WORDS: Aniakchak; caldera-forming eruption; geochemistry; ignimbrite; silicic replenishment


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
GEOLOGICAL BACKGROUND
 METHODS
 RESULTS
 DISCUSSION
 CONCLUSION
 SUPPLEMENTARY DATA
 REFERENCES
 
Chemically heterogeneous ignimbrites are often attributed to eruption from a zoned magma chamber (Smith, 1979Go; Wolff et al., 1990Go), wherein low-density, high-silica residual melt forms by in situ fractional crystallization of mafic magma. The lighter liquid collects at the roof of the chamber, forming a stratified magma body. In other cases, ignimbrite heterogeneity is thought to represent intrusion of mafic magma into a silicic magma body (e.g. Bacon & Druitt, 1988Go). Again, the result is a stratified magma body. When the volcano erupts, the silicic cap is evacuated first and joined later by the underlying mafic magma (Spera, 1984Go; Blake & Ivey, 1986Go). This view has recently been challenged by Eichelberger et al. (2000)Go, who cited unrealistically high silicic magma production rates (1 km3 per 1000 years, in the case of Katmai–Novarupta) necessary to produce the observed silicic magma volumes in the time following the last previous mafic eruption. Eichelberger et al. (2000)Go suggested instead that the silicic and mafic components of bimodal ignimbrites have no direct genetic linkage, but are derived independently of one another. According to this model, the two magmas encounter one another just prior to the eruption. Furthermore, Eichelberger & Izbekov (2000)Go and Eichelberger et al. (2000)Go proposed that some of these caldera-forming eruptions arise when silicic magma is injected into a chamber of more mafic magma, rather than the opposite but more commonly accepted notion of mafic replenishment. These researchers hypothesized that low-density silicic magma ascends to the top of the chamber, destabilizing the roof, and initiating a catastrophic eruption. A full evaluation of this model depends in part on understanding the origins of the magmatic components in bimodal ignimbrites.

Of proposed mechanisms for silicic magma genesis, the widely accepted end-member processes are fractional crystallization of a mafic magma and partial melting of crustal rocks. Either process presents a problem in explaining large volumes of silicic magma. Examination of incompatible trace elements in magmatic suites often indicates that to produce a rhyolite from a basalt, 80–90% of the parental magma must be removed in solid phases. In other words, to distill 1 km3 of rhyolite, 10 km3 of parental basalt are necessary. Consequently, a large mass of mafic to ultramafic cumulate material would be expected, which in many cases is not observed (Knesel & Davidson, 1997Go; Hannah et al., 2002Go). On the other hand, partial melting requires similarly large volumes of source rock, unless the source is itself silicic in composition. Substantial involvement of much older basement granitoids can, however, often be ruled out by isotopic criteria. For more mafic source rocks, progressive partial melting will eventually produce melts diverging from the ‘granite minimum’. In either case, large volumes of silicic magmas pose a problem.

The origin of andesites is less well constrained geochemically. These may form by magma mixing between basaltic and silicic magmas (Eichelberger, 1975Go) or by fractional crystallization of mafic parent magmas (e.g. Brophy & Dreher, 2000Go). Contamination of basaltic magma by crustal material may also generate andesites (Price et al., 2005Go). Mixing of magmas to produce andesite requires subequal volumes of the end-members. Because smaller degrees of crystallization are necessary to generate andesitic liquids from basalt, typically 50% or less, cumulate volume is less of an issue for distillation of andesite than for more silicic magmas.

With regard to bimodal andesite–rhyolite ignimbrites, the question of when the andesite formed is as important as how it formed. In situ fractional crystallization requires that the mafic and silicic components have similar residence times in the subvolcanic system, and that the silicic component has chemical characteristics consistent with derivation from the andesite (e.g. unfractionated radiogenic isotopic ratios). Mafic injection models require that the andesite be formed prior to or during the intrusion event, such that its residence time in the subvolcanic system is shorter than that of its silicic counterpart. Finally, silicic replenishment models imply shorter residence time for the silicic magma and potentially independent origins of silicic and mafic components.

Aniakchak Caldera, Alaska, erupted a minimum of 14 km3 dense rock equivalent (DRE) of rhyodacite (67–70 wt % SiO2) and 13 km3 DRE of andesite (57–60 wt % SiO2) during the caldera-forming eruption, which occurred ~3400 years ago (Miller & Smith, 1987Go). This study examines the geochemical characteristics of the Aniakchak caldera-forming rhyodacite and the co-erupted caldera-forming andesite. Data obtained from several late pre-caldera rhyodacite units are also included for comparison with the caldera-forming rhyodacite.

The caldera-forming rhyodacite at Aniakchak is the most silicic magma erupted from the volcano. It is crystal poor, and the crystals that are present appear to be in equilibrium with the surrounding glass. Petrographic evidence suggests a mixing origin for the contemporaneous andesite. Trace quantities of both probably xenocrystic olivine and quartz are present, as well as a bimodal plagioclase population, comprising low-Ca plagioclase derived from a rhyodacite parent, and high-Ca plagioclase derived from basalt parent. In addition to these petrographic observations, we examine major and trace element whole-rock compositions, Sr, Nd, and U-series isotopic ratios (George et al., 2004Go), and mineral chemistry.


   GEOLOGICAL BACKGROUND
TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL BACKGROUND
METHODS
 RESULTS
 DISCUSSION
 CONCLUSION
 SUPPLEMENTARY DATA
 REFERENCES
 
Aniakchak Volcano is a broad andesite shield located on the Alaska Peninsula (56°54'36''N, 158°7'48''E; see Fig. 1). A large, nearly circular caldera 10 km in diameter was formed ~3400 years BP (Miller & Smith, 1987Go). A US Geological Survey party discovered the crater in August 1922 (Smith, 1925Go). Further detailed exploration by Bernard Hubbard revealed the active nature of the volcano (Hubbard, 1931Go). Outcrops of a pre-caldera andesite ignimbrite, estimated to be between 10000 and 4400 years old (Miller & Smith, 1987Go), exist on the northern and southwestern flanks of the volcano, as well as in the eastern rim on the caldera. There is, so far, no evidence available to determine if an earlier caldera-forming event was associated with this ignimbrite.



View larger version (36K):
[in this window]
[in a new window]
  		Fig. 1. Map of the Alaska Peninsula showing the location of Aniakchak caldera.

 
The caldera-forming eruption resulted in the deposition of a layered ignimbrite in which successive zones of rhyodacite, rhyodacite–andesite, and andesite are in sharp contact (Fig. 2). This event erupted the most silicic material associated with the volcano along with ‘normal’ andesite. Some mingling between the two magmas produced banded pumice, but no hybrid (i.e. thoroughly mixed) pumice intermediate to the andesitic and rhyodacitic compositions has been discovered within samples from the caldera-forming eruption sequence.



View larger version (80K):
[in this window]
[in a new window]
  		Fig. 2. Field photographs of contacts between rhyodacite and mixed ignimbrites (bottom) and between mixed and andesite ignimbrites (top). The generalized stratigraphic column on the right of the photographs represents the sequence of the caldera-forming eruption; this is a composite section, however, and no single outcrops have been found in which this entire sequence is exposed. Sample names refer to data in Table 1.

 
Post-caldera activity has been extensive, with numerous maars, cinder cones, and composite cones including Vent Mountain, which rises nearly 500 m above the caldera floor. The most recent eruption took place in 1931 from a vent near the base of the west wall inside the caldera. Post-caldera pumices and lavas range continuously but unsystematically from 53 wt % SiO2 to 68 wt % SiO2 (Bacon et al., 1997Go), spanning the chemical gap that is so pronounced in the caldera-forming eruption.

Underlying the earliest cone-building lavas at Aniakchak are Upper Jurassic and Lower Cretaceous sedimentary rocks (Detterman et al., 1981Go). Two stratigraphic units, the Naknek Formation and the Staniukovich Formation, are exposed in the inner caldera walls, with thicknesses equal to the thickness of the volcanic edifice in some places (Detterman et al., 1981Go). Jurassic plutonism, the earliest recognized widespread episode of magmatic activity on the Alaska Peninsula (Reed & Lanphere, 1973Go), is represented in the vicinity of Aniakchak by large granodioritic cobbles in Naknek Fm. conglomerate. Large boulders of Naknek conglomerate were included in lithic breccias deposited on the caldera rim, presumably during caldera formation. The large size of the included plutonic clasts suggests a short transport distance from the sediment source, making it reasonable to assume that the Alaska–Aleutian Range Batholith described by Reed & Lanphere (1970Go, 1973Go) is present in the basement beneath Aniakchak. Naknek-derived arkosic sandstones have been found along most of the length of the Alaska Peninsula (McLean, 1979Go), suggesting the presence of granitic rocks in the subsurface. Volcanic rocks from the Eocene to Miocene Meshik arc surround the SE half of the present Aniakchak caldera (Wilson, 1985Go). A relatively flat gentle slope occupies the area to the NW between Aniakchak and the Bering Sea. The NE-trending Aleutian Range stands between the SW flank of the caldera and the Pacific Ocean.


   METHODS
TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL BACKGROUND
 METHODS
RESULTS
 DISCUSSION
 CONCLUSION
 SUPPLEMENTARY DATA
 REFERENCES
 
Whole-rock analyses were determined at the Geoanalytical Laboratory at Washington State University. Major elements were analyzed by X-ray fluorescence (XRF), trace elements by inductively coupled plasma mass spectrometry (ICP-MS). Analytical details are available from the Geoanalyical Lab website at: http://www.wsu.edu/~geology/Pages/Services/Geolab.html. Briefly, for major elements, 3·5 g of rock powder were combined with 7 g of lithium tetraborate and fused. The beads were then reground and re-fused. The glass beads were then loaded into a Rigaku 3370 XRF spectrometer for analysis.

For trace elements, 2 g each of sample powder and lithium tetraborate were mixed and fused. The fused beads were reground, and 250 mg of the resulting powder was dissolved in a solution of 6 ml HF, 2 ml HNO3, and 2 ml HClO4 and evaporated. After a second evaporation of 2 ml HClO4, 3 ml of HNO3, 8 drops of H2O2, 5 drops of HF, and internal standards of In, Re and Ru were added to the sample. This solution was then diluted to a total volume of 60 ml and analyzed by ICP-MS using a Sciex Elan 250 system.

Mineral compositions were measured on a Cameca SX-50 four-spectrometer electron microprobe at the University of Alaska Fairbanks. Standard operating conditions include 15 keV accelerating voltage, 10 nA beam current, and 5–10 µm spot size. Peak and background counting times were 10 s each for all elements. Sodium was analyzed first in all cases to minimize effects of migration.

Isotopic compositions discussed in this paper were originally reported by George et al. (2004)Go. The reader is referred to that work for details of the analytical techniques.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL BACKGROUND
 METHODS
 RESULTS
DISCUSSION
 CONCLUSION
 SUPPLEMENTARY DATA
 REFERENCES
 
Petrology and geochemistry of caldera and pre-caldera rhyodacites
The caldera-forming rhyodacite at Aniakchak is present as a pure rhyodacitic ignimbrite and as a mixed ignimbrite that comprises clasts of both rhyodacite and andesite pumice (Dreher, 2002Go). Pumice blocks are all crystal-poor (~2% crystals). Phenocrysts are dominated by plagioclase (~80%) with subordinate orthopyroxene (~10%), magnetite (~10%) and clinopyroxene (~2%), and trace amounts of hornblende, ilmenite, and apatite. Because of the crystal-poor nature of the samples, the indicated visually estimated mineral proportions are subject to substantial uncertainty. Of 15 analyzed samples of the caldera-forming rhyodacite, the mean SiO2 content is 68·98 ± 0·51 wt %. Table 1 shows representative analyses (the complete dataset is available as an Electronic Appendix from the Journal of Petrology website at http://www.petrology.oupjournals.org). In addition to the caldera-forming rhyodacite, several other rhyodacite pumice and lava units are known, all of which are stratigraphically constrained to have erupted within 5000 years prior to the caldera-forming event (Dreher, 2002Go). Mineral assemblages in all are similar to that in the caldera-forming rhyodacite, although the pre-caldera samples also contain clinopyroxene in addition to orthopyroxene. Pre-caldera lavas (as opposed to pumice units) also contain trace amounts of quartz. Analyses of 10 pre-caldera rhyodacite samples are contained in the Electronic Appendix and summarized in Table 2.


View this table:
[in this window]
[in a new window]
  		Table 1: Representative whole-rock major and trace element data for Aniakchak caldera rhyodacite (RD), andesite (And), and high-alumina basalt (HAB)

 

View this table:
[in this window]
[in a new window]
  		Table 2: Representative whole-rock major and trace element data for Aniakchak pre-caldera rhyodacites

 
Figure 3a shows chondrite-normalized trace element patterns for caldera and pre-caldera rhyodacite samples. All samples show similar arc-type patterns, with relative enrichments of light rare earth elements (LREE) over heavy REE, low Nb and Ta concentrations, and enrichments in large ion lithophile elements (LILE) relative to high field strength elements (HFSE). These patterns also show depletions of Sr and Ti. Chondrite-normalized REE patterns (Fig. 3b) for all rhyodacite samples show negative Eu anomalies.



View larger version (29K):
[in this window]
[in a new window]
  		Fig. 3. (a) Chondrite-normalized incompatible multi-element diagram for Aniakchak caldera-forming (squares) and pre-caldera rhyodacites (triangles). Chondrite normalization values from Sun (1980)Go. (b) Rare earth element patterns for caldera and pre-caldera rhyodacites. Two samples with slightly lower REE abundances (97ANB45 and 97ANB46, Table 2 and online supplementary data) are from lava flows on the caldera flanks. Chondrite normalization values from Sun & McDonough (1989)Go.

 
Plagioclase compositions in all rhyodacite samples are similar. Anorthite content exhibits unimodal distributions, with compositions between An40 and An45 (Fig. 4). Orthopyroxene compositions for all samples are also similar, in the compositional range En63–65Fs32–35Wo3–4.



View larger version (24K):
[in this window]
[in a new window]
  		Fig. 4. Histograms of plagioclase core compositions in Aniakchak caldera-forming andesite and rhyodacite, pre-caldera rhyodacite and high-alumina basalt (HAB).

 
Dreher (2002)Go determined temperatures in the caldera-forming rhyodacite to be between 888° and 898°C based on Fe–Ti oxide compositions, using the equations of Andersen & Lindsley (1988)Go. The pre-caldera rhyodacite lavas have calculated temperatures between 900° and 1182°C based on Fe–Ti oxide compositions. Two-pyroxene geothermometry (Wood & Banno, 1973Go) gives temperatures between 1018° and 1082°C. Pre-caldera hyodacite pumice samples give Fe–Ti oxide temperatures between 933° and 948°C (Dreher, 2002Go).

There are some crucial differences among the different rhyodacite units. The caldera-forming rhyodacite is more evolved in terms of SiO2 content; however, incompatible trace elements are often lower in the caldera samples relative to the pre-caldera rhyodacites. Several SiO2 variation diagrams are shown in Fig. 5. Rb, Zr, and Ba concentrations in the pre-caldera samples are similar to or higher than those in the caldera samples with higher SiO2 concentrations.



View larger version (20K):
[in this window]
[in a new window]
  		Fig. 5. Variation of Zr, Rb, and Ba vs wt % SiO2 for caldera and pre-caldera rhyodacites. Upper and lower pre-caldera pumice refers to two pre-caldera rhyodacite pumice units on the caldera rim that are situated such that one directly overlies the other. The distinction between these two units is made primarily on the darker color and slightly higher crystallinity in the upper unit.

 
Geochemistry of caldera-forming andesite
The caldera-forming andesite samples from Aniakchak are sparsely phyric, although they tend to contain slightly higher concentrations of crystals than the rhyodacite (~8 vol. % crystals). Phenocrysts are dominated by plagioclase, visually estimated to account for ~80–85% of the phenocryst population. The most common ferromagnesian phase is clinopyroxene (~5–10%); trace abundances of resorbed olivine (<1%) are found in some samples. Magnetite (~5–10%) is also present, along with trace quantities of ilmenite and resorbed quartz (each <1%). As with the rhyodacites, the crystal-poor nature of these samples leads to large uncertainties in the visually estimated crystal proportions.

The andesite that erupted during the caldera-forming event is geochemically and petrographically more complex than its silicic counterpart. The most significant aspect of this complexity is a bimodal population of plagioclase crystals. A histogram of plagioclase core compositions (Fig. 4) shows a correspondence between the low-calcium plagioclase in the andesite and the plagioclase in the caldera-forming rhyodacite as well as in pre-caldera rhyodacites. Histograms of plagioclase in other pre-caldera rhyodacite units also corroborate this correspondence. High-calcium plagioclase in the caldera-forming andesite is compositionally similar to plagioclase in Aniakchak high-alumina basalts (HAB).

The only major ferromagnesian silicate mineral in the caldera-forming andesite is augite, with a typical composition of En44–48Fs12–14Wo40–43 and molar Mg-number [100 x MgO/(MgO + FeO)] between 79 and 85. Augite crystals in the caldera-forming andesite typically have molar Kd values [(FeO/MgO)cpx/(FeO/MgO)melt] between 0·17 and 0·22. FeO and Fe2O3 concentrations were calculated from total FeO using the algorithm of Sack et al. (1980)Go. These values are low in comparison with equilibrium crystallization Kd values between 0·23 and 0·26 (Grove & Bryan, 1983Go). Although these Kd values are based on equilibrium with mid-ocean ridge basalts, if applied to the Aniakchak andesite they indicate equilibrium crystallization of clinopyroxene from a more magnesium-rich melt. Aniakchak samples with appropriate FeO/MgO ratios to crystallize such clinopyroxenes are basalts containing around 52 wt % SiO2.

The caldera-forming andesite samples have major element compositions broadly similar to pre- and post-caldera data (Nye et al., 1993Go; Bacon et al., 1997Go; Bacon, 2000Go; Nicholson, 2003Go). The caldera-forming andesite data coincide with maxima in pre- and post-caldera trends in TiO2 and P2O5 vs SiO2. Decreasing concentrations of Al2O3, MgO, FeO, and CaO are observed with increasing SiO2 in the caldera-forming andesite samples. Na2O and K2O increase with increasing SiO2, whereas TiO2 and P2O5 show no clear correlation with SiO2. Incompatible trace element concentrations in the caldera-forming andesite are also consistent with trends defined by pre- and post-caldera data. Ba and Rb, for example, in Fig. 7 increase with increasing SiO2, but the degree of scatter in the Th data makes any distinct correlation with SiO2 unrecognizable in the caldera-forming andesite data. There are decreasing concentrations of Sr with increasing SiO2 in the Aniakchak suite in general, and the caldera-forming andesites in particular. Notably, however, Sr as well as LREE and middle REE (MREE) are enriched in the caldera-forming andesites relative to post-caldera-forming andesites at similar SiO2 and MgO concentrations (Figs 7 and 8).



View larger version (35K):
[in this window]
[in a new window]
  		Fig. 6. SiO2 (wt %) vs major element oxides (wt %) for Aniakchak pre-caldera, caldera, and post-caldera samples.

 


View larger version (26K):
[in this window]
[in a new window]
  		Fig. 7. SiO2 (wt %) vs Rb, Ba, Th, and Sr for Aniakchak pre-caldera, caldera, and post-caldera data. Symbols and grey field are as in Fig. 6.

 


View larger version (31K):
[in this window]
[in a new window]
  		Fig. 8. Chondrite-normalized rare earth element diagrams comparing caldera-forming andesite with post-caldera-forming andesites of similar major element composition. MgO and P2O5 concentrations in individual samples are also indicated, to illustrate the correlation between elevated phosphorus and enrichment in LREE and MREE in the Aniakchak andesites.

 
Temperature estimates based on Fe–Ti distribution in two magnetite–ilmenite pairs (Stormer, 1983Go; Andersen & Lindsley, 1988Go) indicate temperatures of 889°C and 907°C. Ilmenite is rare in caldera products, magnetite–ilmenite pairs in contact with each other even more so. These pairs pass the equilibrium criterion proposed by Bacon & Hirschmann (1988)Go based on Mg–Mn partitioning between the two phases. Oxygen fugacities associated with these two pairs are –11·9 and –11·5 log units, respectively, corresponding approximately to an NNO +1 buffer curve (where NNO is nickel–nickel oxide). Given the relatively mafic composition of the andesite, compared with the rhyodacite, it seems unlikely that they would exist in equilibrium at the same temperature. One other pair from the caldera-forming andesite, which also passes the Bacon–Hirschmann test, gives an anomalously high temperature of 1174°C and log (fO2) of –7·3. This temperature is higher than expected for an intermediate magma in equilibrium with the observed mineral assemblage.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL BACKGROUND
 METHODS
 RESULTS
 DISCUSSION
CONCLUSION
 SUPPLEMENTARY DATA
 REFERENCES
 
Caldera-forming andesite petrogenesis
Bimodal populations of plagioclase, as observed in the Aniakchak caldera-forming andesite, are commonly interpreted as indicators of magma mixing (Sakuyama, 1981Go; Brophy, 1987Go). Couch et al. (2001)Go, however, have proposed that some petrological indicators of disequilibrium, such as reversely zoned crystals and phase compositions not in equilibrium with the host rock, may actually result from convective stirring of a thermally zoned magma body. Although this is certainly a viable scenario at some volcanoes, for Aniakchak the close correspondence between the two plagioclase populations in the andesite and the single plagioclase compositional modes in the basalt and rhyodacite groups, respectively, makes basalt–rhyodacite mixing the more plausible explanation for the disequilibrium mineral assemblage. The model of Couch et al. (2001)Go might be expected to generate a more continuous range of plagioclase compositions, corresponding to a smooth thermal gradient from bottom to top. Thermal convection resulting from heating a magma from below would then juxtapose crystals precipitated originally in equilibrium with magma at a range of temperatures. The decidedly bimodal nature of the caldera-forming andesite plagioclase population contrasts with this expectation. Furthermore, the relative abundance of disequilibrium phases suggests that mixing has been a major process in the formation of the caldera-forming andesite, implying that the bulk composition of the andesite derives largely from basaltic and rhyodacitic sources, not an andesite that has had minor contact with basalt and rhyodacite. Many geochemical trends corroborate an interpretation of magma mixing for the caldera-forming andesite, but some do not. The FeO–SiO2, CaO–SiO2, and MgO–SiO2 data in Fig. 6 form tight linear arrays between basaltic and rhyodacitic end-members. The K2O–SiO2 data may be interpreted as a mixing line, although the data are more scattered. Many incompatible trace elements also support a mixing origin for the andesite. Although there is some scatter in the trace element data (Fig. 7), they define fairly restricted, linear trends between basalt and rhyodacitic end-member compositions.

In spite of the petrographic and geochemical indicators of magma mixing, some significant deviations from simple binary mixing exist in the geochemical dataset. Sr and Nd isotopic compositions determined by George et al. (2004)Go demonstrate that the andesite is not related to the Aniakchak basalts and rhyodacite in any simple way (Fig. 9 and Table 3). Sr and Nd isotopic ratios of the one analyzed caldera-forming rhyodacite sample are 0·70340 and 0·51308, respectively. The caldera-forming andesite sample has a Sr isotopic ratio of 0·70333 and a Nd isotopic ratio of 0·51313 (George et al., 2004Go). Isotopic analyses of four high-alumina basalts from Aniakchak give 87Sr/86Sr values of 0·70324–0·70331. Nd isotopic ratios determined for three of the basalts are between 0·51298 and 0·51311(George et al., 2004Go). Isotopic ratios of pre-caldera rhyodacites have not been determined. Binary mixing calculations indicate that the isotopic compositions of the caldera-forming andesite reflect more than mixing alone. Figure 9 shows that 143Nd/144Nd is elevated relative to that expected from basalt–rhyodacite mixing. Nd concentrations in the caldera-forming andesites (~25 ppm Nd) are elevated relative to other andesites from Aniakchak of similar composition (~19 ppm Nd). The Sr isotopic composition of the caldera-forming andesite is consistent with binary basalt–rhyodacite mixing; however, the Sr concentration in the caldera-forming andesite is almost 200 ppm greater than predicted by the mixing calculations in Fig. 9. The caldera-forming andesites contain around 100 ppm more Sr than other Aniakchak andesites of similar composition. Subsequent fractionation of plagioclase would tend to diminish rather than increase the Sr concentration. Although there are some high-Sr basalts shown in Fig. 7 that might be candidates for the mafic mixing end-member responsible for the Sr concentration of the caldera-forming andesite, these are all from the early pre-caldera phase of the volcano, and are unlikely to have had much influence on the relatively recent caldera magmatism.



View larger version (16K):
[in this window]
[in a new window]
  		Fig. 9. 87Sr/86Sr vs Sr (a) and 143Nd/144Nd vs Nd (b) for basalts and caldera-forming rhyodacite and andesite; (c) 87Sr/86Sr vs 143Nd/144Nd. Curves represent bulk mixing lines between basalt and rhyodacite end-members. Ticks along mixing curves represent 10% increments.

 

View this table:
[in this window]
[in a new window]
  		Table 3: Isotopic compositions of caldera andesite, caldera rhyodacite, and high-alumina basalts from Aniakchak

 
The largest deviation from linear mixing trends is for P2O5, plotted versus SiO2 in Fig. 6. The caldera-forming andesite has the highest P2O5 content in the entire dataset, which, as discussed below, is probably controlled by apatite. Figure 10a shows a positive correlation between (Sm/Yb)N and P concentration. A similar plot of (La/Yb)N vs P (Fig. 10b) shows constant values of (La/Yb)N over the range of P concentrations. Whatever process is enriching the caldera-forming andesite in P apparently also enriches the MREE.



View larger version (28K):
[in this window]
[in a new window]
  		Fig. 10. Chondrite-normalized Sm/Yb (a) and La/Yb (b) vs phosphorus concentrations in Aniakchak dataset. (c) P/Rb vs Rb. Curves represent models of P/Rb evolution by fractional crystallization from a basaltic parent. Ticks on curves represent 10% increments of crystallization. DRb calculated from the mass balance model phase assemblage, and assuming an identical DP. Symbols and grey field are as in Fig. 6.

 
Phosphorus enrichment could be the result of fractional crystallization of a basaltic magma, prior to apatite saturation. The caldera-forming andesite would then represent the point on the liquid line of descent at which apatite began to crystallize. This is, however, inconsistent with the trace element data. Figure 10c shows a plot of P/Rb vs Rb. Using a bulk D for Rb of 0·18, calculated from major element mass balance fractionation models, and setting the partition coefficient of P at zero, closed-system fractional crystallization of basalt with 0·2 wt % P2O5 and 19 ppm Rb cannot generate the P2O5 levels observed in the caldera-forming andesites. As the bulk partition coefficient of phosphorus cannot be less than zero, this diagram indicates that phosphorus must have been added to the system, rather than accumulating as a result of exclusion from crystallizing solid phases. Furthermore, at temperatures of 900°C apatite saturation in Aniakchak andesites occurs with ~0·3 wt % P2O5 in the melt (Harrison & Watson, 1984Go). Because basalt–rhyodacite mixtures are expected to have around 0·2 wt % P2O5, only limited phosphorus enrichment could occur by diffusion. To reach P2O5 concentrations of 0·5–0·7 observed in the caldera-forming andesite, more apatite than is observed would have to crystallize, and be retained in the magma. At 1000°C, the P2O5 concentration required for apatite saturation in a 1:1 mixture of Aniakchak basalts and rhyodacite is around 0·7 wt % (Harrison & Watson, 1984Go). At this higher temperature, P2O5 enrichment could occur by dissolution of apatite from an external source, reaching the observed levels in the caldera-forming andesite. Dissolution of apatite into the hybrid andesite, rather than admixture of solid apatite crystals into the andesite, is a more likely contamination mechanism, because very few apatite crystals have been observed in the andesite. The small amount of apatite that has been observed in the caldera-forming andesite is typically found as inclusions in other crystals, usually magnetite. Microprobe analyses of the groundmass glasses in the andesite samples reveal similarly high concentrations of P2O5.

Based on the enrichments in both P and MREE in the caldera-forming andesite relative to post-caldera andesites from Aniakchak (Fig. 8), we infer that an ‘apatite component’ has been added to the hybrid magma. As there is no potential mixing end-member with high enough P2O5 concentrations to explain the caldera-forming andesites by mixing alone, and as some apparently hybrid andesites from the post-caldera sample suite lack P2O5 enrichment (Bacon et al., 1997Go; Nye et al., 1997Go; Bacon, 2000Go), the basalt–rhyodacite mixing must then be decoupled from the apatite contamination (unless affected end-members simply never erupted). One possibility for the contamination process is dissolution of apatite-bearing feldspar crystals in granitoid host rocks in the shallow crust.

Nonetheless, raising the P2O5 concentration from 0·2 (as observed in the non-caldera-forming andesites) to the values around 0·6 wt % found in the caldera-forming andesite requires addition of ~1 wt % pure apatite. Such addition would increase the CaO concentration by only about 0·5 wt %, less than the range in CaO observed in the caldera-forming andesite data. The apatite contaminant would need around 525 ppm Nd to account for the discrepancy between caldera and non-caldera-forming andesites. If this 1 wt % apatite is the sole source of excess Nd, it must have a 143Nd/144Nd ratio between 0·51330 and 0·51345 to generate the observed Nd isotopic signature of the caldera-forming andesite. Sr concentrations around 1 wt % are necessary to explain the Sr enrichment of the caldera-forming andesites, however—an unrealistic composition for natural apatite (e.g. Dempster et al., 2003Go).

Models for Aniakchak rhyodacite petrogenesis
The trace element data for the caldera and pre-caldera rhyodacites imply that the rhyodacite units do not represent compositional steps along a single liquid line of descent. Incremental evolution from a single parental magma would generally generate a systematic increase in incompatible trace elements with increasing SiO2. Combined assimilation plus fractional crystallization processes (AFC) processes could potentially produce the observed rhyodacites from a single starting composition, but would require different assimilants and/or r values to generate the range of compositions. Thus, it is unlikely that all the rhyodacites evolved continuously from a single parental melt in a single magma chamber. As discussed below, the data for the late pre-caldera rhyodacites, including equal or higher concentrations of incompatible elements at lower SiO2 concentrations relative to the caldera-forming rhyodacite, indicate that each rhyodacite sampled was produced independently from the others, each from sources with slightly different compositions and/or resulting from varying degrees of differentiation. Some major element trends, e.g. P2O5–SiO2 (Fig. 6), within the rhyodacite samples indicate mixing between caldera-forming andesite and caldera-forming rhyodacite. Multiple mixing relationships probably exist between the caldera-forming andesite and rhyodacites; however, it is unlikely that this mixing occurred in a single event between a single pair of end-members, resulting in a compositionally zoned boundary layer. In such a boundary layer model, the magma would become progressively more mafic with depth in the chamber. In addition, the magma in the chamber would become increasingly homogenized overall with time. The early eruption of rhyodacites more affected by interaction with andesite, followed by the least affected rhyodacite in the caldera event, is opposite to these expectations. Finally, the presence of quartz—usually one of the latest phases to crystallize—in some pre-caldera samples but not in the caldera-forming rhyodacite further supports an independent origin for each of the various rhyodacite units. The similarities in major element composition and mineral assemblages suggest that all of the rhyodacites followed largely similar evolutionary paths, each mixing with andesite to varying degrees prior to eruption.

George et al. (2004)Go showed that 87Sr/86Sr in the prehistoric eruption products is positively correlated with SiO2. Samples from the 1931 eruption exhibit a less pronounced correlation between 87Sr/86Sr and SiO2, and a higher 87Sr/86Sr ratio for a given SiO2 concentration (George et al., 2004Go). The 1931 samples also show a decrease in 226Ra/230Th with increasing differentiation, which was inferred by George et al. (2004)Go to reflect mixing between an andesite with a 226Ra excess and a rhyodacitic component in secular equilibrium (226Ra/230Th = 1). George et al. (2004)Go further reasoned that, because it is unlikely that a rhyodacitic magma would remain liquid in the shallow crust for the ~8000 years necessary to reach 226Ra/230Th equilibrium, the evolved mixing component is more likely to be partially melted wall-rocks in the crust (see also Smith et al., 2004Go; Price et al., 2005Go).

Generation of rhyodacite from caldera-forming andesite
Perhaps the simplest scenario for the relationship between the caldera-forming andesite and rhyodacite is a genetic one, in which in situ fractional crystallization of the andesite leads to production of low-density evolved liquids, which segregate into a silicic cap at the top of the magma chamber. Such a scenario has been advocated previously (e.g. Hildreth, 1979Go, 1987Go; McBirney et al., 1985Go; de Silva & Wolff, 1995Go; de Silva, 2001Go) for generating chemically zoned ignimbrites. However, this model is difficult to reconcile with the compositional gap, spanning ~10 wt % SiO2, in the Aniakchak ignimbrite. Prolonged contact between the parent and evolved magmas would eventually lead to compositional gradation, rather than a sharp boundary, by chemical diffusion (Koyaguchi, 1989Go). Convection within the chamber would further smear out the compositional boundary (Cardoso & Woods, 1996Go, 1999Go). The idea of double-diffusive convection advocated by Sparks et al. (1984)Go leads to an expectation of multiple independently convecting layers, separated by small compositional differences. Marsh (1996Go, 2002Go) and Brophy & Dreher (2000)Go have, on the other hand, proposed that magmatic differentiation occurs within solidification fronts along the roof and floor of a magma chamber. In this model, strongly differentiated liquids are trapped interstitially within the mostly crystallized portion of the solidification front, whereas the interior of the parent magma remains largely undifferentiated. Evolved melts produced in this way are volumetrically small compared with the volume of parent magma, unless they can be segregated from the crystallized rind, collected and stored.

The caldera-forming andesite has higher 143Nd/144Nd than the caldera-forming rhyodacite, which is evidence against closed-system fractionation to generate the caldera-forming rhyodacite from the andesite. Major element mass balance models generally require around 40 wt % crystallization to generate rhyodacite from the caldera-forming andesite. The predicted mineral assemblage in these models typically includes around 60–65 % plagioclase. Under these conditions, the bulk partition coefficient for Sr would need to be around two to generate the Sr abundances observed in the rhyodacites. Given that reported Sr partition coefficients are usually less than two for plagioclase alone, and much less than one for clinopyroxene, orthopyroxene, and magnetite, a bulk partition coefficient of two is unrealistically high. The positive correlation of Sr isotopic ratio with SiO2 (George et al., 2004Go) is indicative of open-system magmatic evolution, either by AFC or magma mixing, to generate the entire spectrum of compositions from basalt to rhyodacite. The caldera-forming andesite cannot be both parental to the caldera-forming rhyodacite and at the same time derived from the rhyodacite through mixing with basalt. Based on petrographic evidence for magma mixing in many intermediate-composition rocks at Aniakchak, we infer that magma mixing between basaltic and rhyodacitic end-members is the dominant process responsible for the Aniakchak compositional spectrum, with other processes such as crustal contamination, crystal fractionation, and crystal accumulation exerting a subordinate (but important) control. Although this argument does not rule out fractional crystallization as the dominant process for generating rhyodacites from mafic parents elsewhere in the subvolcanic system, it seems unlikely that the caldera-forming andesite is directly parental to the rhyodacites at Aniakchak.

Simple fractional crystallization from basalt
Whereas the range of compositions at Aniakchak is probably controlled by various degrees of basalt–rhyodacite mixing, the origin of the silicic end-member is still unexplained. Both fractional crystallization and partial melting of crustal rocks can potentially generate rhyodacitic liquids similar to those erupted at Aniakchak (e.g. Bacon & Druitt, 1988Go; Beard & Lofgren, 1991Go). Major element mass balance calculations allow the possibility of fractional crystallization from basalt to generate rhyodacite. Table 4 shows the most successful mass balance calculation. The sum of squared residuals ({Sigma}r2) for this model is 0·09. The results of this model show that rhyodacite sample 97AC37B can be derived from high-alumina basalt (HAB) sample 92CNA22 by removal of plagioclase, clinopyroxene, magnetite, and olivine. Other calculations using different rhyodacite and basalt samples give similar results. The basalt samples contain phenocryst assemblages that are consistent with these results.


View this table:
[in this window]
[in a new window]
  		Table 4: Fractional crystallization and assimilation–fractional crystallization mass balance model

 
Fractional crystallization of basalt combined with crustal assimilation (AFC)
Whereas fractional crystallization of basalt sufficiently explains the major element composition of the caldera-forming rhyodacite (and by analogy, the compositions of the pre-caldera rhyodacites), as noted above, Sr–Nd isotopic data obtained by George et al. (2004)Go require a contribution from a source with more radiogenic Sr than typical Aniakchak basalts. Furthermore, the trace element data are not well explained by fractional crystallization alone, as shown in Fig. 11. The fractional crystallization scheme outlined above along with incorporation of crustal material (AFC) can potentially account for the trace element characteristics of the caldera-forming andesites, although little is known about possible contaminants in the crust beneath Aniakchak caldera. Large granitic cobbles, presumably derived from the Alaska–Aleutian batholith, are abundantly present in the lithic breccia present on the caldera rim, both as solitary blocks and as clasts in larger boulders of conglomerate (Dreher, 2002Go). A sample of this granite was used to represent the composition of a potential assimilant. A mass balance model testing possible scenarios is given in Table 4. The calculation uses an HAB as the parent magma, the granite sample as the assimilant, and the caldera-forming rhyodacite as the derivative magma. The model gives good results ({Sigma}r2 = 0·33) with a fractionating mineral assemblage similar to that calculated in the fractional crystallization model described above.



View larger version (28K):
[in this window]
[in a new window]
  		Fig. 11. Rb vs K/Rb (a), Sr (b), Zr (c), and Nb (d) for the Aniakchak dataset. Curves represent modeled fractional crystallization (FC) and fractional crystallization combined with crustal contamination (AFC) with ticks at 10% increments of crystallization. Numbers on the three AFC trajectories indicate r values used in each of three models. Details of the models are given in the text and Table 4. Partition coefficients are listed in Table 5. Symbols and grey field are as in Fig. 6.

 
Figure 11 shows the variation of K/Rb, Zr, Sr, and Nb with Rb content compared with fractional crystallization and AFC trends. The AFC models use r values of 0·3, 0·5, and 0·9. The r value is the ratio of the mass added to the magma by addition of assimilant to the mass removed from the magma by crystallization. Partition coefficients used in the models are shown in Table 5. The ratio K/Rb (Fig. 11a) can be taken as an indication of crustal influence on magmatic evolution. Because bulk partition coefficients of K and Rb are similar, crystallization does not fractionate K from Rb. This is borne out by the fractional crystallization model in Fig. 11a, which produces a nearly horizontal line. The assimilant chosen for this model has a lower K/Rb ratio than most of the volcanic samples from Aniakchak. As Fig. 11 illustrates, the AFC model from an HAB parent can adequately explain the rhyodacite compositions. Figure 11b shows that fractional crystallization alone would deplete the evolving melt in Sr beyond what is observed in the rhyodacites. Addition of the assimilant, however, compensates for this to give predicted rhyodacitic compositions similar to those observed in the caldera-forming rhyodacite. The HFSE Zr (Fig. 11c) and Nb (Fig. 11d) in the rhyodacites are lower than predicted by fractional crystallization models, and addition of the assimilant again corrects the deficiency. The AFC model is consistent in terms of mass balance in each of the four diagrams of Fig. 11: each shows that the remaining liquid mass necessary to produce a rhyodacitic composition is between 50 and 55%, relative to the initial mass of basalt.


View this table:
[in this window]
[in a new window]
  		Table 5: Trace element partition coefficients used in fractional crystallization and AFC models

 
The REE are also consistent with an AFC model for the petrogenesis of the rhyodacites. Figure 12 shows chondrite-normalized REE patterns for HAB, caldera-forming rhyodacite and the granitic assimilant used in the model. Fractional crystallization of an HAB alone does not reproduce the rhyodacite composition. Starting with a melt produced by 70% fractional crystallization, however, and mixing three parts residual melt with two parts assimilant reproduces the rhyodacite composition remarkably well. Furthermore, this fractionation and mixing scenario results in a final liquid volume of 50% of the starting liquid volume, in good agreement with the final volume estimated by the previously discussed AFC models.



View larger version (17K):
[in this window]
[in a new window]
  		Fig. 12. Chondrite-normalized rare earth element diagram showing composition of high-alumina basalt, caldera-forming rhyodacite and granitic contaminant. Also shown are modeled residual liquids resulting from 70% crystallization of basalt and 70% basalt crystallization plus 40% mixing with granite. Crystallization modeling uses a mode of 60:23:10:7 plagioclase:clinopyroxene:magnetite:olivine, and partition coefficients listed in Table 5. Mixing model uses the residual liquid from the fractional crystallization model and Alaska–Aleutian batholith granite in proportions of 60:40 liquid:granite.

 
The isotopic composition of the granite used as the model assimilant is unknown. However, the slightly more radiogenic Sr isotopic ratio in the rhyodacite is consistent with assimilation of older granitic crustal rocks.

Models involving a mostly solid mafic source
Although rhyodacites from Aniakchak are generally crystal-poor (~2% crystals in the caldera-forming rhyodacite), they include single populations of plagioclase, with the mode of plagioclase core compositions around An40 (Fig. 4). Plagioclase crystals are generally free from disequilibrium textures, and either lack zonation or display mild to moderate oscillatory zonation. Melt inclusions analyzed in rhyodacite-hosted plagioclase are rhyodacitic themselves, much like the host rock (Table 6). The phenocryst population in general appears to be in equilibrium with the host magma.


View this table:
[in this window]
[in a new window]
  		Table 6: Average compositions of melt inclusions in rhyodacite plagioclase, rhyodacite matrix glass and rhyodacite whole rocks

 
Based on these observations, we infer that the crystals found in the caldera-forming rhyodacite have formed since the time of separation of the rhyodacitic magma from its source or parent. Crystals retained from a more mafic parental magma would be likely to have more calcic cores, grading into more sodic compositions towards the rim. Melt inclusions would also probably preserve compositions of a less-evolved precursor magma.

If the rhyodacite-hosted phenocrysts were all formed within the rhyodacitic melt, the rhyodacite magma must have been isolated from its parent or source as nearly 100% liquid. In other words, the caldera-forming rhyodacite does not represent the culmination of a gradual progression down a liquid line of descent, wherein a single magma evolves homogeneously in minute increments as crystallization proceeds. Rather, a more likely scenario is one in which melt is efficiently extracted from a mostly solid source. Once isolated, the extracted evolved melt then begins to crystallize independently. Given the low crystallinity and multiple saturation of plagioclase, orthopyroxene, and magnetite in the caldera-forming rhyodacite, the time between separation and eruption must have been short or the rate of crystallization low.

We propose a model for rhyodacite formation at Aniakchak in which basaltic magmas are emplaced within the crust and begin to crystallize in a solidification front around the margins of the magma chamber (Marsh, 1996Go). As the magma cools from the margin inward, the liquidus isothermal surface will advance inward. At any given time, the state of the magma will range from completely liquid at the liquidus surface through progressively greater proportions of crystals towards the chamber margin.

During the cooling process, the temperature at the chamber–wall-rock interface remains roughly constant (Turcotte & Schubert, 1982Go) until the entire volume of magma cools below its emplacement temperature. If the interface temperature is at or below the solidus temperature of the magma, the range in crystallinity in the solidification front will be from 0% crystals at the liquidus surface to 100% crystals at the chamber margin. If the initial interface temperature is above the solidus of the magma, the range in crystallinity will not extend to 100%. Within this crystallinity range, any interstitial melts between the point of critical crystallinity, typically around 50% (Marsh, 1988Go, 1996Go), and the chamber margins are separable from the crystalline material. Instability and foundering of portions of the mostly solid shell around the chamber margins may allow trapped melt to pool separately from the parent magma body, as discussed by Marsh (1996Go, 2002Go) and Brophy & Dreher (2000)Go.

In addition, crystallization of basalt will provide not only its own sensible heat, but also latent heat produced by crystallization. If this heat is sufficient to induce partial melting in surrounding crustal rocks, a ready source of a crustal contaminant is generated. To summarize, we propose that the Aniakchak caldera-forming rhyodacite represents the interstitial liquid in a crystal-rich mafic mush. Once extracted by solidification front instability or tectonic squeezing and filter pressing, the resulting rhyodacitic liquid is free to ascend to shallower levels.

Relationships between rhyodacites and caldera-forming andesite
One possible model for the formation of the caldera ignimbrite is that after formation of the rhyodacitic magma it was intruded by basalt, which partially mixed with the rhyodacite to produce the caldera-forming andesite. Subsequent eruption of the remaining unmixed rhyodacite and hybrid andesite produced the observed bimodal ignimbrite. The major shortcoming of this model is that it does not explain the compositional gap between the caldera-forming andesite and rhyodacite. Mixing resulting from intrusion of basalt into rhyodacitic magma might be expected to produce a compositional gradation from unmixed basalt at the base to hybrid andesite in the middle and unmixed rhyodacite at the top. Applying this model to the Aniakchak ignimbrite requires that a portion of the rhyodacite was physically isolated from the mixing region of the chamber, or that any hybrid magma compositionally intermediate between the andesite and rhyodacite was prevented from erupting. The former possibility means that the caldera-forming rhyodacite was a separate magma body, but genetically related to the rhyodacite end-member involved in producing the caldera-forming andesite. The latter requires some improbable physical mechanism for excluding the hybrid transition zone from erupting. Additionally, the rhyodacite is petrographically simpler than the andesite, which is more crystal-rich and has undergone both mixing and subsequent contamination processes, suggesting a shorter—not longer—history for the caldera-forming rhyodacite than the andesite as a discrete magma. The extremely low crystallinity of the caldera-forming rhyodacite provides a severe time constraint for storage unless the magma was heated by direct contact with mafic magma, but this is unlikely given the absence of any mixed zone.

Eichelberger et al. (2000)Go suggested that bimodal andesite–silicic ignimbrites can arise from intrusion of new silicic magma into a shallow-level established pod of andesitic magma. The low density of the silicic magma allows it to ascend through the andesitic pod to the roof. A sufficiently large intrusion will subsequently lead to instability of the chamber roof, and a catastrophic eruption. This hypothesis is intended to account for the characteristic crystal-poor nature and mineralogical simplicity of the silicic component in these eruptions, together with their separation from the andesite by a chemical gap. It is also consistent with the distinct isotopic compositions of Aniakchak caldera-forming andesite and rhyodacite reported by George et al. (2004)Go.

That traces of quartz are found in the caldera-forming andesite and pre-caldera rhyodacite, but not in the caldera-forming rhyodacite implicates pre-caldera rhyodacites in the origins of the caldera-forming andesite. The caldera-forming andesite probably existed as a hybrid magma prior to the appearance of the caldera-forming rhyodacite in the shallow magmatic system.

The pre-existence of a hybrid caldera-forming andesite is further supported by mixing trends on major element variation diagrams between the pre-caldera rhyodacites and the caldera-forming andesite. This is especially clear on the P2O5–SiO2 variation diagram (Fig. 6), where the caldera-forming andesite is easily distinguished from other Aniakchak samples. If the various pre-caldera rhyodacite samples represent independently evolved magmas as argued above, these observations can suggest that the andesite was formed by several successive mixing events, prior to the caldera-forming eruption. Each event involved input of rhyodacitic magmas into a more mafic chamber, some of which erupted whereas the remainder homogenized with the resident andesite. The erupted portions of the pre-caldera rhyodacites also incorporated variable amounts of the more mafic host magma.

A model for the Aniakchak caldera ignimbrite
The geochemical data so far presented can be summarized as follows.

(1) The caldera-forming rhyodacite and pre-caldera rhyodacites were formed and evolved independently from one another.

(2) The rhyodacites are best modeled as residual liquids from a mostly crystallized basaltic mush, with contamination by granitic country rocks, independently of the caldera-forming andesite.

(3) The caldera-forming andesite formed prior to the caldera-forming event by mixing between basaltic and pre-caldera rhyodacitic magmas; binary magma mixing was subsequently overprinted by addition of an apatite-rich crustally derived component.

In light of the necessity of a crustal contaminant, and textural and geochemical evidence for efficient extraction of rhyodacitic melts from their source residue, we find any model involving slow, incremental differentiation of basalt to rhyodacite untenable. Rather, we suggest that the Aniakchak rhyodacites represent the residual liquid squeezed out of a nearly congealed gabbroic pluton (e.g. Marsh, 1996Go, 2002Go; Brophy & Dreher, 2000Go). Melts formed in the surrounding granitic wall-rocks—a likely consequence of basalt intrusion and crystallization—are similarly extracted and mixed with the rhyodacitic liquid liberated from the gabbroic mush. At what level in the crust this occurs is unclear; however, for the rhyodacitic melt to interact with the caldera-forming andesite prior to eruption, it must have been generated at greater depths than that of andesite storage.

Based on these conclusions, we favor a model in which successive and independently derived batches of rhyodacite, generated in the lower to middle crust, ascend through the crust and intersect a shallow mafic magma pod (Fig. 13a and b). Successive small inputs of rhyodacitic magma leave an accumulating influence on the originally mafic magma in the shallow chamber. Small portions of rhyodacitic magma may have ‘leaked’ to the surface before complete homogenization with the host magma, leading to the apparent mixing trend displayed by the pre-caldera rhyodacites (Fig. 13c). Upon intrusion of an unusually large batch of rhyodacitic magma, the caldera-forming eruption occurred immediately with little time for mixing with the host andesite (Fig. 13d).



View larger version (27K):
[in this window]
[in a new window]
  		Fig. 13. Summary diagram of the proposed model for the Aniakchak ignimbrite petrogenesis. (a) Initial emplacement of basalt into a shallow crustal reservoir. This basalt undergoes some post-emplacement differentiation, as indicated by the slight positive slope in the SiO2–time diagram in the bottom of the panel. (b) Rhyodacitic residual liquids from crystallizing mafic magmas in the lower to middle crust, combined with anatectic melts of surrounding crustal rocks, ascend to shallow depths, and mix with the previously emplaced basalt. Mixing generates a basaltic andesite magma, which continues to evolve after the mixing event. (c) Small inputs of rhyodacitic melts from deeper in the crust may lead to small eruptions of silicic lavas, after minor interactions with the host basaltic andesite. The erupted lava is then slightly less silicic than the original injected rhyodacitic magma. The portion of injected magma that remains in the chamber continues to homogenize with the resident mafic magmas, driving the hybrid composition closer to andesite. (d) If an unusually large influx of crustally derived rhyodacite intrudes into the subvolcanic chamber, this could trigger an eruption immediately with little time for mixing between the injected and host magmas. In the SiO2–time diagrams: •, composition of the resident magmas; {circ}, composition of crustally derived rhyodacitic inputs.

 
Not all magma intrusion events in the shallow crust lead to eruption. In addition, seismic swarms surrounding active volcanoes are commonly interpreted as rock fracturing associated with magmatic intrusion at relatively shallow depths. Such an event occurred at Akutan volcano in the Aleutians in 1996 without leading to an eruption (Keith et al., 1996Go). If the intrusion is relatively small, fracturing the wall-rocks may relieve the overpressure resulting from the increased volume of magma before the fractures reach the surface. In this case, the intruded and host magmas remain in contact with one another and more intimate mixing may occur. Injecting large volumes of magma, such as the caldera-forming rhyodacite, on the other hand, might sufficiently overpressurize the chamber to initiate an immediate eruption, before significant mixing can occur (e.g. Jellinek & DePaolo, 2002Go). Post-caldera eruptions span the entire chemical range within a single eruption, filling in the caldera eruption chemical gap. We can infer that this relates to the change in the magmatic plumbing system attendant on caldera collapse. After the caldera-forming eruption, the magmatic plumbing system will probably occupy a more highly fractured volume of rock. As such mixing between various batches of melt is more easily facilitated. This is yet another style of m