- Rumble III D (ed) (1976) Oxide minerals. Rev Mineral vol 3, Mineral Soc America, Chelsea, MichiganGoogle Scholar
- Lindsley DH (ed) (1991) Oxide minerals: Petrologic and magnetic significance. Rev Mineral, vol 25, Mineral Soc America, Chelsea, MichiganGoogle Scholar
- Deer WA, Howie RA, Zussman J (1962) Rock-forming minerals. Vol 5: Non-silicates. Longmans, LondonGoogle Scholar
About Cookies, including instructions on how to turn off cookies if you wish to do so. By continuing to browse this site you agree to us using cookies as described in About Cookies. Remove maintenance message. On the basis of oxide mineral assemblages, we now discuss oxidation state and process of our rock samples. Oxides in type A samples are characterized by homogeneous grains, while those in type B are exsolved and composed of two or three phases. According to the oxide classification by Haggerty (1991), type A oxides are classified as C1 and R1 stage. Oxides in type B samples are classified as more than C3 and R6 stage because titanomagnetite is separated by trellis lamellas and titanohematite.
Zitierte Literatur
- Lindsley DL (1976) The crystal chemistry and structure of oxide minerals as exemplified by the Fe-Ti oxides. In Rumble III D, ed (1976) Oxide minerals. Rev Mineral vol 3, L1–L88. Mineral Soc America, Chelsea, MichiganGoogle Scholar
Abstract
Analyses of Fe–Ti oxides help constrain models of magma storage region processes for the Soufrière Hills Volcano, Montserrat (W.I.), and provide clear evidence of the nature of transient heating events in the magma storage region. To constrain timescales of magma heating and remobilization, the TiO2 zoning patterns in a time series of natural titanomagnetites were compared with those produced in controlled phase equilibrium experiments on the andesite. Most samples of andesite erupted from 1995 to 2002 contain titanomagnetite crystals with uniform core compositions (TiO2∼7·8 wt %). Many crystals are characterized by rimward increases in TiO2, interpreted to be Ti diffusion gradients caused by heating of the andesite by invading basaltic magma. Some andesites erupted during periods of the highest observed mass eruption rate, however, contain titanomagnetite with uniformly low TiO2 contents from core to rim. The observation that no Ti diffusion gradients, and no elevated core TiO2 contents, occur in the vast majority of titanomagnetite grains in magma batches that were erupted more than 2 years after the onset of the present eruption strongly suggests, first, that heating of the batches of andesite occurred just before eruption, and, second, that injection of basaltic magma has continued throughout the eruption. Heat, but little mass, may be transferred from the invading basalt to the andesite in the magma storage region by injection of dikes or formation of sills. Ponding of basaltic magma at the base of a pre-existing andesitic magma storage region is the simplest explanation consistent with observations. The Fe–Ti oxide data strongly suggest that an internal conduit within the andesitic magma storage region carries magma from the zone of heating to the overlying conduit, which carries the magma through the upper arc crust. In this model, the magma chamber is being emptied from the bottom, at the contact between pre-existing andesite and newly intruded basalt.
INTRODUCTION
The 1995-to-present (December 2002) eruption of the Soufrière Hills Volcano, Montserrat (W.I.), is the latest in a series of andesitic, dome-forming eruptions in a history spanning >175 kyr (Roobol & Smith, 1998; Harford et al., 2002). The current eruption is thought to have been triggered by injection of magma of basaltic or basaltic andesite composition into a pre-existing andesitic magma storage region (SiO2∼57–61 wt %; Devine et al., 1998a), judging from the presence of ubiquitous ‘mafic inclusions’ (<1 vol. %) with diktytaxitic textures found in the andesite (Murphy et al., 1998, 2000). The most recent eruptions before the new activity, which generated magma of generally similar composition, occurred at ∼400 and ∼3950 years bp (Young et al., 1996; Roobol & Smith, 1998), and there were seismic crises in the 1890 s, 1930 s, and 1960 s that are thought to have been caused by magma movement (Shepherd et al., 1971). It seems likely that magma mixing processes have been important throughout the recent history of the volcano.
The mineral assemblage of the newly erupted andesite includes ∼45–55 wt % phenocrysts of plagioclase (An48–93), amphibole, orthopyroxene, titanomagnetite, and minor quartz and ilmenite. Clinopyroxene occurs as small grains in the groundmass and as fine-grained reaction rims on quartz and some orthopyroxene phenocrysts. Apatite and sulfide are accessory phases.
Following an initial phreatic stage, the eruptive style has been largely extrusive, except for ash venting episodes (Norton et al., 2002) and several episodes of vulcanian explosive eruptions (Druitt et al., 2002). The first explosive eruption occurred on 17 September 1996, following a partial dome collapse, and there were two cyclic series of explosions in early August and then in September and October 1997 (Miller et al., 1998; Voight et al., 1998). Occasional dome collapse episodes have resulted in pyroclastic flows that have, in many cases, reached the coast. These flows, and tephra fall deposits, were sampled when conditions permitted.
One component of the volcanic hazard monitoring program of the Montserrat Volcano Observatory has involved petrographic examination and chemical analysis of dome lava and tephra fall samples, carried out at Brown University (USA) and the University of Bristol (UK). The aim has been to complement monitoring that has included earthquake seismology, ground deformation studies (global positioning system, electronic distance measuring, gravity), dome volume measurement, and gas emissions monitoring [correlation spectrometry (COSPEC) and Fourier transform infrared spectroscopy (FTIR); see references in special issues of Geophysical Research Letters, vol. 25, nos 18 and 19 (S. R. Young et al., eds), and Geological Society of London Memoir21 (T. H. Druitt & B. P. Kokelaar, eds), for descriptions of methods and results].
This paper uses Fe–Ti oxides to constrain models of magma storage region processes. The rationale was to monitor Fe–Ti oxide compositions to try to detect whether or not ‘global’ heating of the andesite was occurring, i.e. gradual heating of the entire magma body. Some important questions relating to global heating are: What are the relative volumes of (1) invading basalt, (2) andesite resident in the magma storage region, and (3) erupted andesite? How is the energy required to remobilize the andesite transferred from basalt to andesite? How soon does remobilized andesite erupt after being heated? If the cumulative volume of injected basalt becomes large (e.g. comparable with the andesite resident in the magma storage region), how much heat can be transferred from basalt to overlying andesite in the absence of extensive magma hybridization, and what would be the timescale of such heat transfer? What is the likelihood of eventual extensive hybridization? What would be the potential effect of open-system behavior, with respect to volatile species in basalt such as H2O, CO2, and SO2, on the degree of vapor saturation of remobilized andesite? In our view, the occurrence of global heating would require a fundamental reassessment of volcanic hazard zonation maps (e.g. Wadge & Isaacs, 1989), because it might indicate that a relatively large volume of andesite has been remobilized, which could potentially be ejected in an explosive eruption larger than any observed to date. Therefore, we monitored Fe–Ti oxide chemistry of eruptive products to assess the nature of magma heating and remobilization.
Fe–Ti oxides generally change composition much faster than do silicates in response to magma system temperature rises that may be caused by magma mixing (e.g. Venezky & Rutherford, 1999). They are therefore sensitive probes of magma storage region heating episodes that take place on timescales of a few days to months before eruption of affected magma batches. Silicate phenocrysts, with the possible exception of amphibole, generally may not change composition (or become resorbed) fast enough to record changes in magma chamber temperatures that take place on such short timescales. So the Fe–Ti oxides may provide the only clear evidence of the nature of transient heating events in the inaccessible magma storage region. For example, Nakamura (1995) used Fe–Ti oxide analyses to show that the 1991 eruption of Unzen Volcano (Japan) was caused by a series of injections of andesitic magma into an overlying dacitic magma chamber, emphasizing that the injections continued after the onset of the eruption.
In the present study, the TiO2 concentration zoning patterns in natural mineral grains are compared with those produced in titanomagnetite grains in controlled experiments designed to determine the phase relations of the natural andesite (Rutherford & Devine, 2003). The comparisons are then used to constrain estimates of timescales of dynamic magma heating and remobilization.
PREVIOUS RESULTS
The bulk composition of erupted andesite has remained within a relatively narrow range (SiO2 ∼57–61 wt %) over the course of the eruption, with no apparent systematic change in composition with time (Devine et al., 1998a). Analyses of the cores of coexisting titanomagnetite and ilmenite crystals in recent Soufrière Hills Volcano products were used to infer pre-eruptive temperature and oxygen fugacity of the andesitic magma in the storage region [T∼840°C; logfO2∼NNO + 1 (where NNO is nickel–nickel oxide); Devine et al., 1998a] using the algorithm of Andersen & Lindsley (1988) as amended by Andersen et al. (1991). Pre-eruptive melt water contents of ∼4·7 wt % (Barclay et al., 1998; Devine et al., 1998a) suggested that the top of the magma chamber was located at depths in excess of 5–6 km at a pressure of ∼130 MPa [crustal density distribution model assumed to be similar to that determined for Mount St. Helens by Williams et al. (1987)], a depth estimate that is in accord with estimates derived from the lower bound of the vast majority of earthquake hypocenter locations (∼5–6 km; Aspinall et al., 1998).
One observation that is consistent with recent reheating of the andesite is that quartz phenocrysts are either partially resorbed or are mantled by fine-grained reaction rims of clinopyroxene. Quartz crystals, although rare (1–2 grains per thin section), occur in nearly all samples over the >7 years of the eruption.
In addition, the complex compositional zoning of plagioclase phenocrysts supports the inference that there have been numerous episodes of heating and cooling of the andesite (Devine et al., 1998a). Reversely zoned plagioclase microphenocrysts in andesites (Murphy et al., 1998, 2000) also provide evidence for recent reheating. The presence of andesite-derived xenocrysts in mafic inclusions and the presence of mafic magma-derived pargasitic amphibole crystals in the andesite (Rutherford & Devine, 2003) indicate that some magma mingling and hybridization have occurred, even though they have not yet been manifested as a monotonic mixing trend in whole-rock major element analyses. Variations in magma bulk chemistry have been observed, however, that are larger than analytical error, as has been confirmed by independent analyses of the same samples by the Brown and Bristol research groups.
RESULTS
Petrographic observations and analytical results
Two-oxide disequilibrium
Electron microprobe analyses of Fe–Ti oxides in a time series (1995–2002) of andesites were used in two-oxide geothermometry (Andersen & Lindsley, 1988). Rim-to-rim transects of titanomagnetite grains were also obtained to detect the potential development of near-rim Ti diffusion gradients. Our sample selection favored tephra fall deposits. Magma ascent rate influences the nature of eruption products as well as eruptive style. For oxide minerals, slow magma ascent provides opportunities for near-surface oxidation and slow cooling, with formation of exsolution lamellae of ilmenite or titanohematite within titanomagnetite crystals. For example, the first-out magma of the current eruption had a reddish-buff color as a result of such oxidation. Tephra fall deposits are less affected by conduit processes such as plagioclase microlite crystallization and oxidation, so they provide the clearest insight into the pre-eruptive magmatic conditions in the storage region.
The cores of large ilmenite and titanomagnetite grains in oxide pairs in Soufrière Hills Volcano andesites are relatively homogeneous, and the relatively low TiO2 contents of the titanomagnetites have suggested temperatures within the range 835–850°C (logfO2 ∼−11·8, or NNO + 1·1; Devine et al., 1998a). Analyses of rare ilmenite–titanomagnetite in-contact pairs, however, reveal Ti zoning in the titanomagnetites (Fig. 1). In contrast, the titanomagnetites of in-contact oxide pairs from many other well-studied eruption products (e.g. the Minoan eruption of Santorini; the 18 May 1980, eruption of Mount St. Helens; the 1902 eruption of Mt. Pelée) are homogeneous (Venezky & Rutherford, 1999; Pichavant et al., 2002).
In-contact titanomagnetite–ilmenite pairs, sample MVO291 (29 September 1997, explosive eruption). (a) TiO2 vs distance from titanomagnetite rim, transect in direction of ilmenite crystal core. [Note the gradient in TiO2 concentration in titanomagnetite near contact with the ilmenite, but essential lack of TiO2 gradient at the titanomagnetite crystal rim, which is in contact with melt (now glass-bearing groundmass). Note also the slightly lower TiO2 concentration in ilmenite near the contact with titanomagnetite; in most cases, however, the composition of the ilmenite is effectively uniform across the grain.] ○, ILM–MT pair 3 (see below). (b) Apparent temperature (Celsius) vs distance from titanomagnetite rim for ILM–MT pair 3 above, estimated by five methods. Geothermometer algorithm of Andersen & Lindsley (1988), as amended by Andersen et al. (1991): ○, solution model of Stormer (1983), with error bars calculated by the program (see text); □, solution model of Anderson (1968); ⋄, solution model of Carmichael (1967); ▵, solution model of Lindsley & Spencer (1982). ▿, geothermometer of Ghiorso & Sack (1991). Average two-oxide temperature estimated on the basis of all measured crystal core compositions and averages of the four estimation methods of Andersen & Lindsley (1988) is equal to ∼830 ± 10°C. For reasons discussed in the text, temperature estimates plotted in other figures throughout the rest of this paper are those based on the Stormer solution model alone.
In-contact titanomagnetite–ilmenite pairs, sample MVO291 (29 September 1997, explosive eruption). (a) TiO2 vs distance from titanomagnetite rim, transect in direction of ilmenite crystal core. [Note the gradient in TiO2 concentration in titanomagnetite near contact with the ilmenite, but essential lack of TiO2 gradient at the titanomagnetite crystal rim, which is in contact with melt (now glass-bearing groundmass). Note also the slightly lower TiO2 concentration in ilmenite near the contact with titanomagnetite; in most cases, however, the composition of the ilmenite is effectively uniform across the grain.] ○, ILM–MT pair 3 (see below). (b) Apparent temperature (Celsius) vs distance from titanomagnetite rim for ILM–MT pair 3 above, estimated by five methods. Geothermometer algorithm of Andersen & Lindsley (1988), as amended by Andersen et al. (1991): ○, solution model of Stormer (1983), with error bars calculated by the program (see text); □, solution model of Anderson (1968); ⋄, solution model of Carmichael (1967); ▵, solution model of Lindsley & Spencer (1982). ▿, geothermometer of Ghiorso & Sack (1991). Average two-oxide temperature estimated on the basis of all measured crystal core compositions and averages of the four estimation methods of Andersen & Lindsley (1988) is equal to ∼830 ± 10°C. For reasons discussed in the text, temperature estimates plotted in other figures throughout the rest of this paper are those based on the Stormer solution model alone.
Typical Ti concentration gradients in titanomagnetite phenocrysts in contact with ilmenite phenocrysts are illustrated in Figs 1a and b, and 2a and b. Two-oxide geothermometry calculations (examples of which are given in Table 1) initially suggested that the higher concentrations of Ti in titanomagnetite lying closest to the interface with the ilmenite were recording transient temperatures of up to ∼900°C. Experimental investigation of amphibole stability in the Soufrière Hills Volcano andesite has shown, however, that amphibole crystals thermally decompose to a mixture of melt and anhydrous crystals (Cpx, Plag, Fe–Ti oxide) in experiments held at 880°C for only 48 h (P = 130 MPa, water saturated; Rutherford & Devine, 2003). Such thermal decomposition of amphibole is observed only rarely in the natural samples, indicating that heating of the andesite by the invading basalt has not exceeded the thermal stability limit of amphibole, which is ∼855°C at 130 MPa, under water-saturated conditions (Rutherford & Devine, 2003). We conclude that what initially appeared to be diffusion gradients in titanomagnetite crystals abutting ilmenite grains are actually reaction fronts: as a result of the andesite being heated by invading basalt, the ilmenite is in fact being replaced by titanomagnetite along an advancing reaction front.
Analytical transects of titanomagnetite crystals in Soufrière Hills Volcano andesite sample MONT 153 (vesiculated pumice, 17 September 1996, explosive eruption). (a) Ilmenite–titanomagnetite in-contact oxide pair; TiO2 content (connected •) and ‘apparent’ temperatures (connected ○) vs distance (from the grain interface) across the titanomagnetite grain; ‘apparent’ temperatures calculated using the method of Andersen & Lindsley (1988) (see text for caveat about temperature calculations; also for interpretation of TiO2 concentration gradients). (b) Ilmenite–titanomagnetite in-contact oxide pair [see caption for (a)]. (c)–(f) rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass).
Analytical transects of titanomagnetite crystals in Soufrière Hills Volcano andesite sample MONT 153 (vesiculated pumice, 17 September 1996, explosive eruption). (a) Ilmenite–titanomagnetite in-contact oxide pair; TiO2 content (connected •) and ‘apparent’ temperatures (connected ○) vs distance (from the grain interface) across the titanomagnetite grain; ‘apparent’ temperatures calculated using the method of Andersen & Lindsley (1988) (see text for caveat about temperature calculations; also for interpretation of TiO2 concentration gradients). (b) Ilmenite–titanomagnetite in-contact oxide pair [see caption for (a)]. (c)–(f) rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass).
Example geothermometry calculations forFig. 1(MVO291, ILM–MT pair 3)
Oxide* | Wt % | Element | Cation frac. | Formula/3 sites |
---|---|---|---|---|
Spinel phase | ||||
TiO2 | 7·65 | Ti | 0·0717 | 0·22 |
Al2O3 | 1·97 | Al | 0·0289 | 0·09 |
Cr2O3 | 0·00 | Cr | 0·0000 | 0·00 |
Fe2O3 | 52·70 | Fe3+ | 0·4943 | 1·48 |
FeO | 36·20 | Fe2+ | 0·3774 | 1·13 |
MnO | 0·56 | Mn | 0·0059 | 0·02 |
MgO | 1·17 | Mg | 0·0217 | 0·07 |
Total | 100·25 | |||
X(Usp) spinel | 0·223 | Stormer model | ||
X(Usp) spinel | 0·203 | Anderson model | ||
X(Usp) spinel | 0·215 | Carmichael model | ||
X(Usp) spinel | 0·218 | Lindsley model | ||
Rhombohedral phase† | ||||
TiO2 | 42·86 | Ti | 0·4036 | 0·81 |
Al2O3 | 0·22 | Al | 0·0032 | 0·01 |
Cr2O3 | 0·02 | Cr | 0·0002 | 0·00 |
Fe2O3 | 20·09 | Fe3+ | 0·1893 | 0·38 |
FeO | 33·86 | Fe2+ | 0·3546 | 0·71 |
MnO | 0·95 | Mn | 0·0101 | 0·02 |
MgO | 2·09 | Mg | 0·0390 | 0·08 |
Total | 100·08 | |||
X(Ilm) rhomb | 0·800 | Stormer model | ||
X(Ilm) rhomb | 0·789 | Anderson model | ||
X(Ilm) rhomb | 0·807 | Carmichael model | ||
X(Ilm) rhomb | 0·807 | Lindsley model | ||
Pair passes Bacon & Hirschmann (1988) equilibrium test |
Oxide* | Wt % | Element | Cation frac. | Formula/3 sites |
---|---|---|---|---|
Spinel phase | ||||
TiO2 | 7·65 | Ti | 0·0717 | 0·22 |
Al2O3 | 1·97 | Al | 0·0289 | 0·09 |
Cr2O3 | 0·00 | Cr | 0·0000 | 0·00 |
Fe2O3 | 52·70 | Fe3+ | 0·4943 | 1·48 |
FeO | 36·20 | Fe2+ | 0·3774 | 1·13 |
MnO | 0·56 | Mn | 0·0059 | 0·02 |
MgO | 1·17 | Mg | 0·0217 | 0·07 |
Total | 100·25 | |||
X(Usp) spinel | 0·223 | Stormer model | ||
X(Usp) spinel | 0·203 | Anderson model | ||
X(Usp) spinel | 0·215 | Carmichael model | ||
X(Usp) spinel | 0·218 | Lindsley model | ||
Rhombohedral phase† | ||||
TiO2 | 42·86 | Ti | 0·4036 | 0·81 |
Al2O3 | 0·22 | Al | 0·0032 | 0·01 |
Cr2O3 | 0·02 | Cr | 0·0002 | 0·00 |
Fe2O3 | 20·09 | Fe3+ | 0·1893 | 0·38 |
FeO | 33·86 | Fe2+ | 0·3546 | 0·71 |
MnO | 0·95 | Mn | 0·0101 | 0·02 |
MgO | 2·09 | Mg | 0·0390 | 0·08 |
Total | 100·08 | |||
X(Ilm) rhomb | 0·800 | Stormer model | ||
X(Ilm) rhomb | 0·789 | Anderson model | ||
X(Ilm) rhomb | 0·807 | Carmichael model | ||
X(Ilm) rhomb | 0·807 | Lindsley model | ||
Pair passes Bacon & Hirschmann (1988) equilibrium test |
0·800 | 832 ± 19 | −11·88 ± 0·14 | Stormer | |
0·203 | 0·789 | 823 ± 18 | −11·92 ± 0·14 | Anderson |
0·215 | 0·807 | 823 ± 19 | −12·05 ± 0·15 | Carmichael |
0·218 | 0·807 | 825 ± 19 | −12·02 ± 0·15 | Lindsley |
919 | −10·51 | Ghiorso & Sack |
0·800 | 832 ± 19 | −11·88 ± 0·14 | Stormer | |
0·203 | 0·789 | 823 ± 18 | −11·92 ± 0·14 | Anderson |
0·215 | 0·807 | 823 ± 19 | −12·05 ± 0·15 | Carmichael |
0·218 | 0·807 | 825 ± 19 | −12·02 ± 0·15 | Lindsley |
919 | −10·51 | Ghiorso & Sack |
Oxide | Wt % | Element | Cation frac. | Formula/3 sites |
Spinel phase | ||||
TiO2 | 9·85 | Ti | 0·0919 | 0·28 |
Al2O3 | 1·84 | Al | 0·0269 | 0·08 |
Cr2O3 | 0·08 | Cr | 0·0008 | 0·00 |
Fe2O3 | 48·75 | Fe3+ | 0·4552 | 1·37 |
FeO | 37·98 | Fe2+ | 0·3941 | 1·18 |
Oxide | Wt % | Element | Cation frac. | Formula/3 sites |
Spinel phase | ||||
TiO2 | 9·85 | Ti | 0·0919 | 0·28 |
Al2O3 | 1·84 | Al | 0·0269 | 0·08 |
Cr2O3 | 0·08 | Cr | 0·0008 | 0·00 |
Fe2O3 | 48·75 | Fe3+ | 0·4552 | 1·37 |
FeO | 37·98 | Fe2+ | 0·3941 | 1·18 |
Oxide* | Wt % | Element | Cation frac. | Formula/3 sites |
---|---|---|---|---|
MnO | 0·85 | Mn | 0·0089 | 0·03 |
MgO | 1·20 | Mg | 0·0222 | 0·07 |
Total | 100·56 | |||
X(Usp) spinel | 0·284 | Stormer model | ||
X(Usp) spinel | 0·263 | Anderson model | ||
X(Usp) spinel | 0·276 | Carmichael model | ||
X(Usp) spinel | 0·278 | Lindsley model | ||
Rhombohedral phase‡ | ||||
TiO2 | 42·35 | Ti | 0·4004 | 0·80 |
Al2O3 | 0·21 | Al | 0·0031 | 0·01 |
Cr2O3 | 0·01 | Cr | 0·0001 | 0·00 |
Fe2O3 | 20·72 | Fe3+ | 0·1960 | 0·39 |
FeO | 33·36 | Fe2+ | 0·3508 | 0·70 |
MnO | 1·00 | Mn | 0·0106 | 0·02 |
MgO | 2·08 | Mg | 0·0390 | 0·08 |
Total | 99·74 | |||
X(Ilm) rhomb | 0·793 | Stormer model | ||
X(Ilm) rhomb | 0·782 | Anderson model | ||
X(Ilm) rhomb | 0·801 | Carmichael model | ||
X(Ilm) rhomb | 0·800 | Lindsley model |
Oxide* | Wt % | Element | Cation frac. | Formula/3 sites |
---|---|---|---|---|
MnO | 0·85 | Mn | 0·0089 | 0·03 |
MgO | 1·20 | Mg | 0·0222 | 0·07 |
Total | 100·56 | |||
X(Usp) spinel | 0·284 | Stormer model | ||
X(Usp) spinel | 0·263 | Anderson model | ||
X(Usp) spinel | 0·276 | Carmichael model | ||
X(Usp) spinel | 0·278 | Lindsley model | ||
Rhombohedral phase‡ | ||||
TiO2 | 42·35 | Ti | 0·4004 | 0·80 |
Al2O3 | 0·21 | Al | 0·0031 | 0·01 |
Cr2O3 | 0·01 | Cr | 0·0001 | 0·00 |
Fe2O3 | 20·72 | Fe3+ | 0·1960 | 0·39 |
FeO | 33·36 | Fe2+ | 0·3508 | 0·70 |
MnO | 1·00 | Mn | 0·0106 | 0·02 |
MgO | 2·08 | Mg | 0·0390 | 0·08 |
Total | 99·74 | |||
X(Ilm) rhomb | 0·793 | Stormer model | ||
X(Ilm) rhomb | 0·782 | Anderson model | ||
X(Ilm) rhomb | 0·801 | Carmichael model | ||
X(Ilm) rhomb | 0·800 | Lindsley model |
X(Usp) sp | X(Ilm) rh | T (°C) | log fO2 | Solution model |
0·284 | 0·793 | 879 ± 19 | −11·24 ± 0·14 | Stormer |
0·263 | 0·782 | 870 ± 19 | −11·28 ± 0·13 | Anderson |
0·276 | 0·801 | 869 ± 19 | −11·42 ± 0·14 | Carmichael |
0·278 | 0·800 | 871 ± 19 | −11·39 ± 0·14 | Lindsley |
923 | −10·45 | Ghiorso & Sack |
X(Usp) sp | X(Ilm) rh | T (°C) | log fO2 | Solution model |
0·284 | 0·793 | 879 ± 19 | −11·24 ± 0·14 | Stormer |
0·263 | 0·782 | 870 ± 19 | −11·28 ± 0·13 | Anderson |
0·276 | 0·801 | 869 ± 19 | −11·42 ± 0·14 | Carmichael |
0·278 | 0·800 | 871 ± 19 | −11·39 ± 0·14 | Lindsley |
923 | −10·45 | Ghiorso & Sack |
Calculated using the algorithms of Andersen & Lindsley (1988; as amended by Andersen et al., 1991), and Ghiorso & Sack (1991); solution models from Carmichael (1967), Anderson (1968), Lindsley & Spencer (1982), and Stormer (1983).
Composition of ilmenite in crystal core.
Composition of ilmenite close to interface with titanomagnetite.
Example geothermometry calculations forFig. 1(MVO291, ILM–MT pair 3)
Oxide* | Wt % | Element | Cation frac. | Formula/3 sites |
---|---|---|---|---|
Spinel phase | ||||
TiO2 | 7·65 | Ti | 0·0717 | 0·22 |
Al2O3 | 1·97 | Al | 0·0289 | 0·09 |
Cr2O3 | 0·00 | Cr | 0·0000 | 0·00 |
Fe2O3 | 52·70 | Fe3+ | 0·4943 | 1·48 |
FeO | 36·20 | Fe2+ | 0·3774 | 1·13 |
MnO | 0·56 | Mn | 0·0059 | 0·02 |
MgO | 1·17 | Mg | 0·0217 | 0·07 |
Total | 100·25 | |||
X(Usp) spinel | 0·223 | Stormer model | ||
X(Usp) spinel | 0·203 | Anderson model | ||
X(Usp) spinel | 0·215 | Carmichael model | ||
X(Usp) spinel | 0·218 | Lindsley model | ||
Rhombohedral phase† | ||||
TiO2 | 42·86 | Ti | 0·4036 | 0·81 |
Al2O3 | 0·22 | Al | 0·0032 | 0·01 |
Cr2O3 | 0·02 | Cr | 0·0002 | 0·00 |
Fe2O3 | 20·09 | Fe3+ | 0·1893 | 0·38 |
FeO | 33·86 | Fe2+ | 0·3546 | 0·71 |
MnO | 0·95 | Mn | 0·0101 | 0·02 |
MgO | 2·09 | Mg | 0·0390 | 0·08 |
Total | 100·08 | |||
X(Ilm) rhomb | 0·800 | Stormer model | ||
X(Ilm) rhomb | 0·789 | Anderson model | ||
X(Ilm) rhomb | 0·807 | Carmichael model | ||
X(Ilm) rhomb | 0·807 | Lindsley model | ||
Pair passes Bacon & Hirschmann (1988) equilibrium test |
Oxide* | Wt % | Element | Cation frac. | Formula/3 sites |
---|---|---|---|---|
Spinel phase | ||||
TiO2 | 7·65 | Ti | 0·0717 | 0·22 |
Al2O3 | 1·97 | Al | 0·0289 | 0·09 |
Cr2O3 | 0·00 | Cr | 0·0000 | 0·00 |
Fe2O3 | 52·70 | Fe3+ | 0·4943 | 1·48 |
FeO | 36·20 | Fe2+ | 0·3774 | 1·13 |
MnO | 0·56 | Mn | 0·0059 | 0·02 |
MgO | 1·17 | Mg | 0·0217 | 0·07 |
Total | 100·25 | |||
X(Usp) spinel | 0·223 | Stormer model | ||
X(Usp) spinel | 0·203 | Anderson model | ||
X(Usp) spinel | 0·215 | Carmichael model | ||
X(Usp) spinel | 0·218 | Lindsley model | ||
Rhombohedral phase† | ||||
TiO2 | 42·86 | Ti | 0·4036 | 0·81 |
Al2O3 | 0·22 | Al | 0·0032 | 0·01 |
Cr2O3 | 0·02 | Cr | 0·0002 | 0·00 |
Fe2O3 | 20·09 | Fe3+ | 0·1893 | 0·38 |
FeO | 33·86 | Fe2+ | 0·3546 | 0·71 |
MnO | 0·95 | Mn | 0·0101 | 0·02 |
MgO | 2·09 | Mg | 0·0390 | 0·08 |
Total | 100·08 | |||
X(Ilm) rhomb | 0·800 | Stormer model | ||
X(Ilm) rhomb | 0·789 | Anderson model | ||
X(Ilm) rhomb | 0·807 | Carmichael model | ||
X(Ilm) rhomb | 0·807 | Lindsley model | ||
Pair passes Bacon & Hirschmann (1988) equilibrium test |
0·800 | 832 ± 19 | −11·88 ± 0·14 | Stormer | |
0·203 | 0·789 | 823 ± 18 | −11·92 ± 0·14 | Anderson |
0·215 | 0·807 | 823 ± 19 | −12·05 ± 0·15 | Carmichael |
0·218 | 0·807 | 825 ± 19 | −12·02 ± 0·15 | Lindsley |
919 | −10·51 | Ghiorso & Sack |
0·800 | 832 ± 19 | −11·88 ± 0·14 | Stormer | |
0·203 | 0·789 | 823 ± 18 | −11·92 ± 0·14 | Anderson |
0·215 | 0·807 | 823 ± 19 | −12·05 ± 0·15 | Carmichael |
0·218 | 0·807 | 825 ± 19 | −12·02 ± 0·15 | Lindsley |
919 | −10·51 | Ghiorso & Sack |
Oxide | Wt % | Element | Cation frac. | Formula/3 sites |
Spinel phase | ||||
TiO2 | 9·85 | Ti | 0·0919 | 0·28 |
Al2O3 | 1·84 | Al | 0·0269 | 0·08 |
Cr2O3 | 0·08 | Cr | 0·0008 | 0·00 |
Fe2O3 | 48·75 | Fe3+ | 0·4552 | 1·37 |
FeO | 37·98 | Fe2+ | 0·3941 | 1·18 |
Oxide | Wt % | Element | Cation frac. | Formula/3 sites |
Spinel phase | ||||
TiO2 | 9·85 | Ti | 0·0919 | 0·28 |
Al2O3 | 1·84 | Al | 0·0269 | 0·08 |
Cr2O3 | 0·08 | Cr | 0·0008 | 0·00 |
Fe2O3 | 48·75 | Fe3+ | 0·4552 | 1·37 |
FeO | 37·98 | Fe2+ | 0·3941 | 1·18 |
Oxide* | Wt % | Element | Cation frac. | Formula/3 sites |
---|---|---|---|---|
MnO | 0·85 | Mn | 0·0089 | 0·03 |
MgO | 1·20 | Mg | 0·0222 | 0·07 |
Total | 100·56 | |||
X(Usp) spinel | 0·284 | Stormer model | ||
X(Usp) spinel | 0·263 | Anderson model | ||
X(Usp) spinel | 0·276 | Carmichael model | ||
X(Usp) spinel | 0·278 | Lindsley model | ||
Rhombohedral phase‡ | ||||
TiO2 | 42·35 | Ti | 0·4004 | 0·80 |
Al2O3 | 0·21 | Al | 0·0031 | 0·01 |
Cr2O3 | 0·01 | Cr | 0·0001 | 0·00 |
Fe2O3 | 20·72 | Fe3+ | 0·1960 | 0·39 |
FeO | 33·36 | Fe2+ | 0·3508 | 0·70 |
MnO | 1·00 | Mn | 0·0106 | 0·02 |
MgO | 2·08 | Mg | 0·0390 | 0·08 |
Total | 99·74 | |||
X(Ilm) rhomb | 0·793 | Stormer model | ||
X(Ilm) rhomb | 0·782 | Anderson model | ||
X(Ilm) rhomb | 0·801 | Carmichael model | ||
X(Ilm) rhomb | 0·800 | Lindsley model |
Oxide* | Wt % | Element | Cation frac. | Formula/3 sites |
---|---|---|---|---|
MnO | 0·85 | Mn | 0·0089 | 0·03 |
MgO | 1·20 | Mg | 0·0222 | 0·07 |
Total | 100·56 | |||
X(Usp) spinel | 0·284 | Stormer model | ||
X(Usp) spinel | 0·263 | Anderson model | ||
X(Usp) spinel | 0·276 | Carmichael model | ||
X(Usp) spinel | 0·278 | Lindsley model | ||
Rhombohedral phase‡ | ||||
TiO2 | 42·35 | Ti | 0·4004 | 0·80 |
Al2O3 | 0·21 | Al | 0·0031 | 0·01 |
Cr2O3 | 0·01 | Cr | 0·0001 | 0·00 |
Fe2O3 | 20·72 | Fe3+ | 0·1960 | 0·39 |
FeO | 33·36 | Fe2+ | 0·3508 | 0·70 |
MnO | 1·00 | Mn | 0·0106 | 0·02 |
MgO | 2·08 | Mg | 0·0390 | 0·08 |
Total | 99·74 | |||
X(Ilm) rhomb | 0·793 | Stormer model | ||
X(Ilm) rhomb | 0·782 | Anderson model | ||
X(Ilm) rhomb | 0·801 | Carmichael model | ||
X(Ilm) rhomb | 0·800 | Lindsley model |
X(Usp) sp | X(Ilm) rh | T (°C) | log fO2 | Solution model |
0·284 | 0·793 | 879 ± 19 | −11·24 ± 0·14 | Stormer |
0·263 | 0·782 | 870 ± 19 | −11·28 ± 0·13 | Anderson |
0·276 | 0·801 | 869 ± 19 | −11·42 ± 0·14 | Carmichael |
0·278 | 0·800 | 871 ± 19 | −11·39 ± 0·14 | Lindsley |
923 | −10·45 | Ghiorso & Sack |
X(Usp) sp | X(Ilm) rh | T (°C) | log fO2 | Solution model |
0·284 | 0·793 | 879 ± 19 | −11·24 ± 0·14 | Stormer |
0·263 | 0·782 | 870 ± 19 | −11·28 ± 0·13 | Anderson |
0·276 | 0·801 | 869 ± 19 | −11·42 ± 0·14 | Carmichael |
0·278 | 0·800 | 871 ± 19 | −11·39 ± 0·14 | Lindsley |
923 | −10·45 | Ghiorso & Sack |
Calculated using the algorithms of Andersen & Lindsley (1988; as amended by Andersen et al., 1991), and Ghiorso & Sack (1991); solution models from Carmichael (1967), Anderson (1968), Lindsley & Spencer (1982), and Stormer (1983).
Composition of ilmenite in crystal core.
Composition of ilmenite close to interface with titanomagnetite.
The replacement reaction of ilmenite by titanomagnetite is best illustrated in Fig. 3, which is a colorized TiKα X-ray dot map of an ilmenite–titanomagnetite contact zone. The ilmenite crystal (red) has a reaction rim (green) of ∼10 µm thickness of high-Ti titanomagnetite. The abutting titanomagnetite phenocryst has a low-Ti core (dark blue) and higher-Ti rim compositions (light blue), the latter characterized by true diffusion gradients. The high-Ti reaction rim surrounding the ilmenite phenocryst is interpreted to be a mixture of sub-micron scale domains of titanomagnetite and relict ilmenite (which could not be resolved using the electron microprobe).
Colorized Ti Kα X-ray dot map of ilmenite–magnetite in-contact oxide pair from dome lava sample MVO1208A (September 2000); ilmenite is being replaced by titanomagnetite along an advancing reaction front; the ilmenite crystal (red tones) has a reaction rim of ∼10 µm thickness (green tones) composed mainly of high-Ti titanomagnetite; the abutting titanomagnetite phenocryst has a low-Ti core (dark blue) with higher-Ti rim compositions (light blue), the latter characterized by true diffusion gradients; the high-Ti reaction rim surrounding the ilmenite phenocryst is interpreted to be composed of a mixture of sub-micron-scale domains of titanomagnetite and relict ilmenite; results of analytical transects A–A′ and B–B′ are given in Fig. 4.
Colorized Ti Kα X-ray dot map of ilmenite–magnetite in-contact oxide pair from dome lava sample MVO1208A (September 2000); ilmenite is being replaced by titanomagnetite along an advancing reaction front; the ilmenite crystal (red tones) has a reaction rim of ∼10 µm thickness (green tones) composed mainly of high-Ti titanomagnetite; the abutting titanomagnetite phenocryst has a low-Ti core (dark blue) with higher-Ti rim compositions (light blue), the latter characterized by true diffusion gradients; the high-Ti reaction rim surrounding the ilmenite phenocryst is interpreted to be composed of a mixture of sub-micron-scale domains of titanomagnetite and relict ilmenite; results of analytical transects A–A′ and B–B′ are given in Fig. 4.
Transects of the titanomagnetite illustrated in Fig. 3 reveal relatively high TiO2 concentrations at the crystal rim, where in contact with groundmass (Fig. 4). Two-oxide geothermometry calculations, using the titanomagnetite rim composition for the spinel phase, and the average ilmenite composition for the rhombohedral phase, yield apparent temperature estimates that are again in excess of the thermal stability limit of amphibole in the andesite, and so are considered unrealistic. Therefore, although Ti diffusion gradients at the rims of titanomagnetite grains are produced by, and are therefore evidence of, magma reheating, it is not possible to obtain meaningful temperature estimates from them. On the other hand, two-oxide temperature and oxygen fugacity estimates based on phenocryst core compositions are believed to be valid.
Analytical transects of titanomagnetite crystal shown in Fig. 3; dome lava sample MVO1208A (September 2000). (a) Transect A–A′ in Fig. 3, which is roughly perpendicular to the ilmenite–titanomagnetite interface (see text for explanation of spuriously high ‘apparent’ temperature estimates derived from two-oxide geothermometry calculations). (b) Transect B–B′ in Fig. 3, which is roughly parallel to the ilmenite–titanomagnetite interface, 130 µm away (note asymmetry of rim-to-core diffusion gradients on opposite sides of the crystal; also caveat about spuriously high ‘apparent’ temperature estimates derived from two-oxide geothermometry calculations).
Analytical transects of titanomagnetite crystal shown in Fig. 3; dome lava sample MVO1208A (September 2000). (a) Transect A–A′ in Fig. 3, which is roughly perpendicular to the ilmenite–titanomagnetite interface (see text for explanation of spuriously high ‘apparent’ temperature estimates derived from two-oxide geothermometry calculations). (b) Transect B–B′ in Fig. 3, which is roughly parallel to the ilmenite–titanomagnetite interface, 130 µm away (note asymmetry of rim-to-core diffusion gradients on opposite sides of the crystal; also caveat about spuriously high ‘apparent’ temperature estimates derived from two-oxide geothermometry calculations).
The accuracy and precision of two-oxide temperature estimates have been investigated by several research groups (e.g. Frost & Lindsley, 1991; Ghiorso & Sack, 1991). We have previously determined, however, that the closest match between two-oxide temperature estimates for experimental charges and thermocouple-measured temperatures in our hydrothermal experiments (±5°C accuracy) is obtained when the solution model of Stormer (1983) is used in conjunction with the temperature–fO2 algorithm of Andersen & Lindsley (1988) (see Geschwind & Rutherford, 1992; Gardner et al., 1995; Cottrell et al., 1999). Use of other solution models can result in either higher or lower temperature estimates relative to the Stormer–Andersen formulation (Fig. 1a). In some cases, the estimated temperatures using different solution models are more than 10°C lower than the estimates obtained by our preferred method [e.g. the solution model of Anderson (1968)]. The potentially systematic error in our two-oxide temperature estimates for natural sample phenocryst core pairs could possibly be greater than 10°C. The precision of the estimates is generally ±20°C, as calculated by the Andersen & Lindsley (1988) algorithm (Fig. 1b).
Time series of Fe–Ti oxide analytical transects
Petrologic monitoring of the newly erupted andesite by our group included estimation of magma ascent rates from amphibole breakdown and estimation of magma heating from Fe–Ti oxide geochemistry on a near real-time basis. The results of representative analytical transects of the oxides are presented in Fig. 2 and in Figs 5–11, which are arranged in order of eruption date. The scales of the vertical axes are the same in each figure (unless otherwise noted), which allows one to easily assess subtle changes in zoning patterns that have occurred as the eruption has progressed. In our view, there is no other ‘index’ that would provide an adequate description of the significant variability of zoning patterns within a single sample (e.g. Fig. 8). Finally, a caveat applies to the figure axes labeled ‘Apparent T (C)’; as indicated above, near-rim titanomagnetite compositions may give spuriously high temperature estimates, because there is no ilmenite in equilibrium with this magnetite.
Analytical transects of titanomagnetite crystals in sample MVO573 (vesiculated pumice from one of the September–October 1997 explosive eruptions). (a)–(f) Six of 12 rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass); only two grains, illustrated in (b) and (f), contained slight Ti diffusion gradients at the phenocryst rims, the other 10 crystals being essentially free of measurable Ti diffusion gradients (see text for discussion).
Analytical transects of titanomagnetite crystals in sample MVO573 (vesiculated pumice from one of the September–October 1997 explosive eruptions). (a)–(f) Six of 12 rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass); only two grains, illustrated in (b) and (f), contained slight Ti diffusion gradients at the phenocryst rims, the other 10 crystals being essentially free of measurable Ti diffusion gradients (see text for discussion).
Analytical transects of titanomagnetite crystals in combined samples MVO1125 and MVO1126 (lithic-rich ashes from the 9 November 1999 explosive eruption). (a)–(e) Rim-to-rim and rim-to-core transects with variable Ti diffusion gradients and absolute concentrations (see text for discussion). (f) Titanomagnetite grain with incipient unmixing, probably caused by slow cooling within the conduit or dome (note change of scales with respect to other figures).
Analytical transects of titanomagnetite crystals in combined samples MVO1125 and MVO1126 (lithic-rich ashes from the 9 November 1999 explosive eruption). (a)–(e) Rim-to-rim and rim-to-core transects with variable Ti diffusion gradients and absolute concentrations (see text for discussion). (f) Titanomagnetite grain with incipient unmixing, probably caused by slow cooling within the conduit or dome (note change of scales with respect to other figures).
Analytical transects of titanomagnetite crystals in sample MVO1175 (dome lava erupted in March 2000). (a)–(d) Ilmenite–titanomagnetite in-contact oxide pairs (see caption for Fig. 2a); (e) and (f) rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass).
Analytical transects of titanomagnetite crystals in sample MVO1175 (dome lava erupted in March 2000). (a)–(d) Ilmenite–titanomagnetite in-contact oxide pairs (see caption for Fig. 2a); (e) and (f) rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass).
Analytical transects of titanomagnetite crystals in sample MVO1208C (dome lava erupted in September 2000). All grains occur in andesite that is within 7 mm of the contact with a large mafic inclusion; wide variation in Ti diffusion gradient patterns and absolute concentrations should be noted (see text for discussion).
Analytical transects of titanomagnetite crystals in sample MVO1208C (dome lava erupted in September 2000). All grains occur in andesite that is within 7 mm of the contact with a large mafic inclusion; wide variation in Ti diffusion gradient patterns and absolute concentrations should be noted (see text for discussion).
Analytical transects of titanomagnetite crystals in sample MVO1228B (dome lava erupted in September 2001). (a) Ilmenite–titanomagnetite in-contact oxide pair; TiO2 content (connected •) and ‘apparent’ temperatures (connected ○) vs distance (rim-to-rim) across the titanomagnetite grain; ‘apparent’ temperatures calculated using the method of Andersen & Lindsley (1988). (b)–(f) Rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass) [note more pronounced Ti diffusion gradients, and only slightly higher core TiO2 concentrations of some grains, relative to earlier, explosively erupted samples (Figs 2 and 5)].
Analytical transects of titanomagnetite crystals in sample MVO1228B (dome lava erupted in September 2001). (a) Ilmenite–titanomagnetite in-contact oxide pair; TiO2 content (connected •) and ‘apparent’ temperatures (connected ○) vs distance (rim-to-rim) across the titanomagnetite grain; ‘apparent’ temperatures calculated using the method of Andersen & Lindsley (1988). (b)–(f) Rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass) [note more pronounced Ti diffusion gradients, and only slightly higher core TiO2 concentrations of some grains, relative to earlier, explosively erupted samples (Figs 2 and 5)].
Analytical transects of titanomagnetite crystals in sample MVO1229B (dome lava erupted in September 2001). (a)–(f) Rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass) (note more pronounced Ti diffusion gradients, and only slightly higher core TiO2 concentrations of some grains, relative to earlier-erupted samples).
Analytical transects of titanomagnetite crystals in sample MVO1229B (dome lava erupted in September 2001). (a)–(f) Rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass) (note more pronounced Ti diffusion gradients, and only slightly higher core TiO2 concentrations of some grains, relative to earlier-erupted samples).
Analytical transects of titanomagnetite and ilmenite crystals in samples MVO1234A, MVO1234B, and MVO1234C (dome lavas erupted in September 2002). (a) Ilmenite-titanomagnetite in-contact oxide pair; TiO2 content (connected •) and ‘apparent’ temperatures (connected ○) vs distance (rim-to-rim) across the titanomagnetite grain; ‘apparent’ temperatures calculated using the method of Andersen & Lindsley (1988). (b)–(j) Rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass) (note less pronounced Ti diffusion gradients, and slightly lower core TiO2 concentrations of some grains, relative to samples MVO1228B and MVO1229B).
Analytical transects of titanomagnetite and ilmenite crystals in samples MVO1234A, MVO1234B, and MVO1234C (dome lavas erupted in September 2002). (a) Ilmenite-titanomagnetite in-contact oxide pair; TiO2 content (connected •) and ‘apparent’ temperatures (connected ○) vs distance (rim-to-rim) across the titanomagnetite grain; ‘apparent’ temperatures calculated using the method of Andersen & Lindsley (1988). (b)–(j) Rim-to-rim analytical transects of titanomagnetite phenocrysts that are in contact with melt (now glass-bearing groundmass) (note less pronounced Ti diffusion gradients, and slightly lower core TiO2 concentrations of some grains, relative to samples MVO1228B and MVO1229B).
September 1996
Tephra from the first explosive eruption on 17 September 1996, which was triggered by a major dome collapse, contain a diversity of titanomagnetite grains, some of which lack Ti diffusion gradients at the crystal rim (MONT153; Fig. 2a and f), some of which had hints of rim Ti diffusion gradients (Fig. 2c and d), and some of which had slight rim Ti diffusion gradients (Fig. 2b and e).
September–October 1997
Pumice fall sample MVO573, a single clast of which was used as an experimental starting material (Rutherford & Devine, 2003), is typical of tephra produced by the cyclic series of eruptions in September and October 1997. Of 12 randomly analyzed titanomagnetite phenocrysts (Fig. 5), 10 had no Ti diffusion gradients and two had only slight gradients at the crystal rims (Fig. 5b and f). Sample MVO291, a pumice fall deposit produced by an explosive eruption on 29 September 1997, (i.e. during this same period), also has titanomagnetites that lack Ti rim diffusion gradients or are characterized by weak gradients.
November 1999
The volcano was in an eruptive hiatus from about March 1998 until April 1999, when the onset of ash venting episodes and minor explosions signalled that the eruption was not over (Norton et al., 2002). Samples of ash produced by the explosions were examined for a juvenile component. Although glassy clasts and vitreous amphibole fragments were present, it was not immediately clear whether they were newly erupted material, or just young ash from earlier eruptions that had been carried aloft by phreatic explosions. On 9 November 1999, an explosion occurred that was clearly too large to have been caused merely by phreatic processes. Fe–Ti oxides from the ash produced by that explosion (samples MVO1125 and MVO1126) were analyzed, and transects of titanomagnetites (Fig. 6) suggested diverse origins for the respective grains. Some grains contained elevated TiO2 contents (Fig. 6b and e), relative to the low TiO2 contents (<8 wt %) of the cores of phenocrysts in explosively erupted tephra from the 1995–1998 phase of the eruption (e.g. Figs 2 and 5), whereas others had lower values similar to those of earlier-erupted grains (Fig. 6a and d). Some grains had begun to unmix, resulting in compositional banding (Fig. 6f), and still others had become variably altered to hematite (not shown). The Fe–Ti oxide data suggested that the ash was a mixture of old and new tephra and/or lithic fragments.
March 2000
Sample MVO1175 (Fig. 7) is a fragment of dome lava collected in March 2000, after dome extrusion resumed at the beginning of the second eruptive phase. All ilmenite–titanomagnetite pairs are characterized by reaction fronts at their contact that yield spuriously high temperature estimates (Fig. 7a–d). Many titanomagnetite phenocrysts, however, have low-TiO2 (<8 wt %) cores, and some have modest Ti diffusion gradients at the rim (Fig. 7e and f).
September 2000
Samples 1208A, 1208B, and 1208C are fragments of dome lava erupted in September 2000. Sample 1208C contained a large (15 cm × 8 cm × 5 cm), flattened, ellipsoidal, diktytaxitic mafic inclusion that was enclosed by andesite. Oriented thin sections were made such that the distance from titanomagnetite grains in the andesite to the andesite–mafic inclusion interface could be measured. The objective was to see if there was a systematic variation in the Ti zoning patterns in the titanomagnetite grains as a function of distance from the mafic inclusion. The reasoning was that quenching of the mafic blob caused by injection into the andesite might cause transient heating of the andesite near the blob, recorded by compositional changes in the titanomagnetite crystals. The results of the analytical transects are presented together in Fig. 8.
There is considerable diversity of Ti diffusion profiles. Some grains lack diffusion gradients and also have low-TiO2 cores, whereas others contain long, short, or even abrupt Ti diffusion gradients at the crystal rims, but have ‘normal’, low-TiO2 cores. At least one relatively small grain has an elevated core TiO2 content and no diffusion gradient, although this may only be a sectioning artifact. High TiO2 contents are also commonly observed in titanomagnetite microlites in the groundmass of all andesites described above. These high TiO2 contents cannot be used for temperature estimates, for reasons outlined above.
The observed diversity of compositional zoning in titanomagnetite phenocrysts in sample MVO1208C all occurs within 7 mm of the interface with the mafic blob, and there is no systematic variation of zoning pattern or core Ti content as a function of distance from the andesite–mafic inclusion interface. Titanomagnetite grains that have undergone different thermal histories have been brought together before, or during, eruption. This inference also applies to the silicate phases in the andesite.
September 2001
Dome lava samples MVO1228B (Fig. 9) and MVO1229B (Fig. 10), which erupted in September 2001, contain titanomagnetite phenocrysts with the most pronounced rim-to-core Ti diffusion gradients. The rim-to-rim profiles of the titanomagnetite phenocrysts are generally symmetric, yet the TiO2 concentrations in the cores of the grains generally remain like those that are characteristic of the andesite before the most recent heating event. Most amphibole phenocrysts in these samples have thin or no decompression-induced breakdown rims.
September 2002
Several textural types of dome lava were sampled from dome collapse deposits that were extruded in September 2002. Samples MVO1234A, MVO1234B, and MVO1234C include moderately friable, slightly oxidized, brownish blocks, and denser, light gray varieties. All contain a mixture of different types of amphibole phenocrysts that are dominated by euhedral crystals with thin or no decompression breakdown rims, with lesser amounts of other amphibole grains that have decompression or thermal breakdown features of variable extent and thickness. Small fragments and/or microphenocrysts of amphibole also occur in the groundmass (see Rutherford & Devine, 2003). Mafic inclusions appear to be slightly more abundant in these samples than in earlier-erupted samples. Fe–Ti oxide phenocrysts are free of exsolution lamellae (Fig. 11). Rare ilmenite–titanomagnetite in-contact pairs are observed (Fig. 11a). Although TiO2 gradients are observed in titanomagnetite crystals near the contact with ilmenite, suggestive of replacement of the former by the latter, ilmenite crystals are still present in eruption products more than 7 years after its onset. Partially resorbed quartz phenocrysts are also still present. Although Ti diffusion gradients of variable extent occur at the rims of titanomagnetite phenocrysts, the TiO2 contents of the cores of the grains remain at the low values obtained since the beginning of the eruption.
DISCUSSION
Geothermometry
Our results indicate that the average TiO2 content of titanomagnetite phenocryst cores (≥30 µm from the rim) in tephra fall deposits is 7·78 ± 0·28 wt % (see Figs 2 and 5). Two-oxide geothermometry calculations, based on average core compositions of titanomagnetite and ilmenite phenocrysts in tephra fall deposits, suggest that, before the recent heating event, the andesite was at a temperature of 830 ± 10°C and a log fO2 of −11·8, about 1·1 log units above the NNO synthetic buffer; this is essentially the same result as that reported by Devine et al. (1998a). Experiments show (1) that, at 130 MPa, quartz does not crystallize until the temperature falls to ∼825°C, and (2) that the general stability of amphibole in the natural magma indicates that the temperature in the magma is unlikely to have been raised above ∼850–860°C for more than a few hours (Rutherford & Devine, 2003). The rise in temperature of the magma remobilized just before eruption is therefore likely to be from ∼825 ± 10°C to ≤855, i.e. ≤30°C.
Causes of reheating of the andesiticmagma
The essential lack of Ti diffusion gradients in most titanomagnetite grains in magma batches such as samples MVO573 and MVO291, erupted during the August–October 1997 period of highest volume extrusion rate (∼10 m3/s; Druitt et al., 2002), has implications for the nature of magma heating episodes. Magma erupted during that period travelled from the magma storage region at ∼5–6 km depth to the surface in <2 days (Devine et al., 1998b; Rutherford & Devine, 2003); therefore, it was not affected by conduit processes, such as plagioclase microlite crystallization, that might overprint phase relations in the storage region just before eruption. Furthermore, the inferred heating that caused remobilization of these magma batches did not last long enough to produce TiO2 diffusion gradients in the titanomagnetite phenocrysts.
The pronounced Ti diffusion gradients in titanomagnetite crystals in samples such as MVO1228 and MVO1229 (erupted in September 2001) are equally significant. In contrast to the explosively erupted samples from the September–October 1997 period, the processes that resulted in eruption of these lava batches did permit formation of Ti diffusion gradients. But the question arises whether the zoning observed in such titanomagnetite crystals is (1) produced in the magma storage region, as a result of heating of the host andesite by invading basalt or perhaps changes in oxygen fugacity, or (2) produced during magma ascent, as a result of release of latent heat caused by microlite crystallization, which is brought on by degassing of H2O, or perhaps a combination of both processes (R. S. J. Sparks & M. Pichavant, personal communication, 2002).
Potential magma heating as a result of latentheat effect
The efficacy of the latent heat effect depends on the nature of heat flow in the conduit walls, and whether the magma in the conduit is an open system with respect to volatile species derived from magma at depth, which may also carry heat out of the system (Devine & Rutherford, in prep.). Modal analysis, microprobe analyses of minerals and glasses, and mass balance calculations suggest that the maximum temperature rise as a result of release of latent heat would be ∼30–40°C (J. D. Devine, unpublished data, 2002), assuming the unlikely boundary condition of zero heat loss from the magma in the conduit to the wall rocks or the atmosphere.
Regarding the question of the magnitude of heat flow from the magma in the conduit into the wall rocks, the Soufrière Hills Volcano has previously been studied for its extensive and long-lived hydrothermal system, which included a hot water pond with magmatic H2O-influenced isotopic characteristics located >5·5 km from the summit domes (Chiodini et al., 1996; Hammouya et al., 1998). This indicates that hydrothermal fluids are probably capable of carrying heat and volatiles away from the conduit. In addition, the D/H ratios of amphibole phenocrysts in some early-erupted dome samples indicate that magma has at times interacted with meteoric water (Harford & Sparks, 2001), again suggesting the potential for heat loss from the conduit as a result of hydrothermal activity. Furthermore, hundreds of thousands of earthquakes have been recorded in a laterally extensive, >5 km2 area since the onset of the present eruption, most of them located in the 0–3 km depth range, many characterized by the sharply impulsive onsets resulting from rock fracture (Aspinall et al., 1998). It is inferred that the country rock in the immediate vicinity of the conduit is highly fractured and therefore accessible to hydrothermal fluids.
On the other hand, the eruptive style of the volcano appears to depend at times on catastrophic release of pressurization of the lava dome or upper part of the conduit (Woods et al., 2002), suggesting that the wall rocks of the conduit, or the sides of the growing lava dome, may at times become impermeable, perhaps as a result of precipitation of silica from volcanic exhalates. At a time of frequently occurring vulcanian eruptions, in August–October 1997, however, the eruptive dynamics suggested that leakage of volcanic gas from the conduit through the surrounding rocks was occurring (Clarke et al., 2002a, 2002b).
Thus the rise in upper conduit magma temperature potentially caused by release of latent heat during microlite crystallization may well be considerably less than the maximum ∼30–40°C calculated assuming negligible loss of heat from the conduit. In addition, experimental evidence described below suggests that there is insufficient time for the chemical zoning observed in natural titanomagnetites to be produced by the latent heat effect.
Potential effect of changes in oxygen fugacity
Experimental petrologists have shown that compositions of titanomagnetites in hydrous andesitic and dacitic magmas vary strongly with changes in oxygen fugacity (e.g. Rutherford & Devine, 1988; Frost & Lindsley, 1991; Martel et al., 1999). The question therefore arises whether the compositional zoning observed in the titanomagnetite phenocrysts in the newly erupted andesite is due to changes in fO2 as well as temperature (M. Pichavant, personal communication, 2002).
For changing fO2 to have a significant effect on titanomagnetite composition, the change must be in a direction essentially normal to the common fO2 buffers in fO2–T space (e.g. Frost & Lindsley, 1991, p. 443). The observed titanomagnetite zoning would require perhaps an order of magnitude change in fO2, which we consider to be unlikely, at least at relatively constant T. In fact, the variation of fO2 in island arc magmas in general, and in the Lesser Antilles in particular, is more or less parallel to the common fO2 buffers (e.g. Arculus & Wills, 1980; Gill, 1981; Devine, 1987).
Investigations of the effects of temperature and fO2 changes on the compositions of coexisting Fe–Ti oxides in simple systems indicate (1) that rising temperature with fO2 varying parallel to the common fO2 buffers results in an increase in the TiO2 content (i.e. mole fraction of ulvöspinel) of coexisting titanomagnetite, and a smaller decrease in the TiO2 content of coexisting ilmenite; (2) that decreasing fO2 at constant T results in increases in the TiO2 contents of both phases (e.g. Frost & Lindsley, 1991). The cores of ilmenite grains are relatively invariant in composition and ilmenite in contact with titanomagnetite appears to be being replaced by the latter. The TiO2 content of ilmenite near such contacts is essentially the same as that in the crystal cores (e.g. Fig. 11a), and in some cases slightly lower, rather than higher, than that in the crystal core (e.g. Fig. 1). This is the opposite of the effect one would expect if the increase in titanomagnetite rim TiO2 contents were due to a significant lowering of fO2. We infer that limited compositional variations in ilmenite do not support the idea that the fO2 of the andesite has been substantially decreased, relative to the trends of the common buffers, as a result of the recent magma mixing episode, an effect that could potentially produce higher TiO2 contents in coexisting titanomagnetite phenocryst rims. Small changes in fO2 must occur, but we conclude that they are mainly due to the magma being heated, rather than to reduction by the invading basalt. We therefore interpret the elevated TiO2 contents of titanomagnetite crystal rims to be due largely to a rise in the temperature of the host magma (see also below).
Constraints on conduit processes from petrographic observation and experiments
We have conducted melting, crystallization, and isothermal decompression experiments to estimate the effects of magma ascent processes on the compositional zoning of titanomagnetite crystals (Rutherford & Devine, 2003). The starting material for the experiments was coarsely crushed sample MVO573, which is a partially glassy tephra. As indicated above, the natural titanomagnetite phenocrysts are essentially unzoned with respect to TiO2 (Fig. 5), so any zoning in run products was probably produced during the experiments.
First, isobaric heating experiments show that Ti enrichment of titanomagnetite phenocryst rims can be produced in a variety of ways. For example, samples that are heated to a high temperature (say, 880°C; P = 130 MPa) for a short period of time (e.g. 2 days; experiment M32) produce TiO2 zoning patterns in titanomagnetite phenocrysts (Fig. 12a) similar to those in crystals from samples heated at lower temperatures (e.g. 850°C) for longer periods of time (e.g. 2 weeks; run M56; Fig. 12b–d; compare Figs 9 and 10).
(a–f) Analytical transects of titanomagnetite crystals in experimental samples. In (b)–(d), (h), and (j) the similarity of zoning patterns in long-duration experiments compared with those in natural samples MVO1228B and MVO1229B should be noted. (See text for discussion.)
(a–f) Analytical transects of titanomagnetite crystals in experimental samples. In (b)–(d), (h), and (j) the similarity of zoning patterns in long-duration experiments compared with those in natural samples MVO1228B and MVO1229B should be noted. (See text for discussion.)
The effects of the experimental temperature rises, such as those cited above, on the other phases can be drastically different in the respective experiments. Petrographic observation of the natural samples can be used to rule out some heating scenarios. Specifically, in the first case mentioned above (heating for 2 days to 880°C at P = 130 MPa), the amphibole phenocrysts in the experimental charge undergo extensive thermal decomposition. Such decomposition is observed in only a few grains per thin section in natural samples, most other grains showing only decompression breakdown rims of variable thickness or opacitization caused by near-surface oxidation. This strongly suggests that heating for short periods at temperatures above the amphibole stability limit is unlikely to be the main cause of the Ti diffusion gradients observed in most natural titanomagnetite grains. Transient heating of the andesite undoubtedly occurs during episodes of basalt injection but the experiments suggest that the high temperatures are likely to be dissipated in a matter of hours.
Second, the thicknesses of decompression breakdown rims on amphibole phenocrysts may be used to estimate magma ascent rates and ascent times from the magma storage region at 5–6 km depth, because an experimentally calibrated speedometer has been determined by Rutherford & Devine (2003, fig. 9). For example, in the most recently examined samples (MVO1228, MVO1229, MVO1234), many amphibole phenocrysts lack decompression breakdown rims, indicating magma ascent rates of ≥0·017 m/s and ascent times of ≤4 days from the storage region to near-surface depths (Rutherford & Devine, 2003). Assuming that most degassing of ascending magma occurs in the upper 1–2 km of the 5–6 km conduit (e.g. Melnik & Sparks, 1999, 2002), that would indicate that any effects on the TiO2 zoning in titanomagnetite phenocrysts, which would potentially be caused by release of latent heat as a result of microlite crystallization, would probably have to occur in less than 1–2 days. That is, within 1 or 2 days of the onset of potential microlite crystallization in the upper conduit, the magma will be expelled from the conduit and therefore cooled considerably, effectively retarding any currently occurring reactions (other than oxidation) as a result of rapidly decreasing, temperature-dependent Ti diffusion rates. Experiments discussed in a later section also indicate that the pronounced Ti gradients observed in titanomagnetites in some of the most recently erupted samples cannot be produced in a period of only 1–2 days.
Constraints on magma conduit processes from decompression experiments
Decompression experiments were conducted to investigate the effects of gradually decreasing pressure, such as may occur during slow magma ascent, on amphibole stability and Ti zoning in titanomagnetite crystals (Rutherford & Devine, 2003). The oxygen fugacity was held near NNO + 1. Experiment M29 involved gradual isothermal (860°C) decompression of MVO573 starting material from 130 to 4 MPa over 10 days. Experiment M36 involved the same isothermal decompression conditions, but the duration was 23 days. Transects of titanomagnetite phenocryst fragments in the experimental charges show only modest near-rim enrichments in TiO2 (Fig. 12e and f), even though the experiments were artificially maintained at the relatively high temperature of 860°C, requiring external input of heat to the furnace apparatus. The longer-duration experiment has better-developed Ti diffusion gradients than those in the shorter experiments, but neither profile is as pronounced as those in some natural samples. It is concluded, because of the weak Ti diffusion gradients observed in experiments M29 and M36, that the pronounced gradients observed in titanomagnetite crystals in natural samples such as MVO1228 and MVO1229 cannot be produced by slow decompression P, T trajectories, even at the relatively high temperature of 860°C. Although lowering of melt water content owing to decompression drastically lowers Ti diffusion rates in the melt adjacent to the phenocryst, the most likely explanation for lack of pronounced gradients in the experimental titanomagnetites is that the diffusion rate of Ti through the crystal is the most important rate-limiting factor. Using available data for diffusion of Ti in magnetite, Venezky & Rutherford (1999) calculated that it would take 30 days to develop a Ti-rich rim of 20 µm thickness on titanomagnetite at 850°C.
Some experiments were designed to simulate rapid ascent of magma from the storage region (P = 130 MPa), followed by staging at a shallower level (50 or 90 MPa; Rutherford & Devine, 2003). None of the shorter-duration experiments of this sort produced the pronounced Ti zoning observed in titanomagnetite phenocrysts in natural samples MVO1228 and MVO1229. Examples of the zoning profiles produced in the 50 MPa experiments are illustrated in Fig. 12g (experiment M25; 10 day run) and Fig. 12h (experiment M25 + 1; 18 day run), and in the 90 MPa experiments in Fig. 12i (experiment M23; 5 day run) and Fig. 12j (experiment M40; 27 day run) We conclude that magma ascent processes cannot account for the Ti diffusion gradients observed in natural samples MVO1228 and MVO1229, because these magma samples took <4 days to ascend from the storage region to the surface, judging by the lack of decompression-induced breakdown rims on amphibole phenocrysts. The diffusion gradients in the natural samples were therefore caused by magma storage region processes.
Experimental constraints on timescales of heating of andesite in the storage region
Fe–Ti oxide minerals can re-equilibrate (i.e. become homogeneous with respect to Ti distribution) within a few days to months of a rise in system temperature, depending primarily on temperature and grain size (Gardner et al., 1995; Nakamura, 1995; Venezky & Rutherford, 1999). Therefore, the existence of Ti diffusion gradients in Montserrat samples (e.g. MVO1228B, MVO1229B) indicates a recent heating event shortly before extrusion. Equally importantly, the lack of Ti diffusion gradients in the crystal rims of most titanomagnetite phenocrysts in explosively erupted sample MVO573, and their preserved, low, ‘original’ grain TiO2 contents, indicate that heating of that batch of magma probably happened within a few days or weeks of eruption. The time of eruption (September–October 1997), however, may have been a period of the most rapid reheating and remobilization of the andesite, because a few hundred thousand cubic meters of lava were being erupted explosively every nine hours or so (Sparks et al., 1998; Druitt et al., 2002). The observation that no Ti diffusion gradients, and no elevated core TiO2 contents, occur in the vast majority of titanomagnetite grains in that magma batch (which was erupted more than 2 years after the onset of the present crisis), strongly suggests that (1) heating of the MVO573 batch of andesite occurred just before its eruption, and (2) injection of basaltic magma has continued throughout the eruption.
Several experiments help constrain likely timescales of magma heating in the storage region and subsequent eruption. Experiment M56 (MVO573 starting material; Rutherford & Devine, 2003) was run for 2 days at 870°C, then for 14 more days at 850°C at 130 MPa, thus simulating heating by basalt injection into the andesitic magma storage region. This experiment produced rim-to-core Ti diffusion gradients in titanomagnetite phenocryst fragments in the charge (Fig. 12b–d) that are essentially similar to those in natural oxide crystals in samples MVO1228 and MVO1229 (Figs 9 and 10). Amphibole persisted as a stable phase. The experiment was rapidly quenched, thereby freezing in the compositional gradients that existed in the simulated magma chamber. The natural samples MVO1228 and MVO1229 may have been similarly heated before rapid transport to the surface. Experimental results suggest, however, that these natural samples could not have been briefly heated to temperatures as high as 880°C for 2 days, because this would have resulted in extensive thermal decomposition of amphibole grains, which is not observed in these samples.
In contrast, experiment M21 (MVO573 starting material), run at 860°C for only 4 days at 130 MPa, produced no obvious Ti diffusion gradients in the rims of titanomagnetite phenocryst fragments (Fig. 12k), but continued heating under those conditions for a total of 10 days (experiment M34) resulted in ‘rounding’ and embayment of amphibole phenocryst fragments, production of very slight Ti diffusion gradients in the rims of the largest titanomagnetite phenocryst fragments (Fig. 12l), and higher TiO2 concentrations, relative to the MVO573 starting material, in the smallest titanomagnetite phenocryst fragments. The Ti gradients in the largest phenocryst fragments in M34 (Fig. 12l), however, do not extend to the high rim TiO2 contents in some titanomagnetite phenocrysts in natural samples MVO1228 and MVO1229 (Figs 9 and 10). It is also possible that the smallest titanomagnetite grains formed by recrystallization, rather than by diffusive processes, as there are other reactions occurring in the experiments.
It is concluded that heating periods of two or more weeks may be recorded by some natural samples with pronounced Ti diffusion gradients and high rim TiO2 contents in titanomagnetite phenocryst rims. It seems highly unlikely that extensive Ti diffusion gradients observed in samples MVO1228 and MVO1229 can be produced at realistic temperatures in <4 days, even at the 130 MPa pressure inferred for the magma storage region, where diffusion rates for Ti in the melt will probably be high relative to those prevailing in the upper conduit, but diffusion rates in titanomagnetite will remain relatively slow. The residence time of the magma within the conduit is also generally too short for such Ti diffusion gradients to have been produced there, whether by release of latent heat or any other process. It is therefore also concluded that reheating and changes in Fe–Ti oxide compositions of the andesite are caused by injection of basaltic magma into an andesitic magma storage region.
Titanomagnetite phenocrysts in the most recently erupted lavas (MVO1234; September 2002; Fig. 11) lack the pronounced Ti diffusion gradients observed in crystals from earlier-erupted samples (MVO1228, MVO1229; September 2001; Figs 9 and 10). This suggests that there has been no monotonic increase in the TiO2 contents of crystals in the andesite during the year that elapsed between eruptions of these samples. Thus, there continues to be a lack of evidence for global heating of the andesite. The TiO2 contents of most crystal cores essentially remain as low as those in the earliest eruption products.
Constraints on the magma system
Presence of magma storage regions within the arc crust
The major conclusions of our previous studies of the andesite, and the work reported here and by Rutherford & Devine (2003), are that, before heating and remobilization, the erupted andesite was at a pressure of ∼130 MPa, and that injection of mafic magma caused its temperature to be raised ∼30°C, from ∼825°C to ∼855°C before eruption (Devine et al., 1998a; compare Couch et al., 2001). The energy required to remobilize the andesite is probably supplied by transfer of heat across some sort of interface, because only minor magma hybridization has as yet occurred. There are a number of ways in which such heat transfer may occur, depending on the geometry of the magma plumbing system, and the relative fluidity of the hydrous, mafic magma with respect to the colder, more viscous, crystal-rich andesite. Geophysical observations that would help constrain models of the volcanic plumbing system are, however, few.
Monitoring of the volcano has indicated that most of the earthquakes and ground deformation are caused by near-surface phenomena occurring within and around the conduit that connects the growing volcanic dome with the storage region. Therefore, this monitoring cannot be used to infer the size or shape of the magma storage region or the geometry of the conduits. In contrast to volcanic centers where tectonic and volcanic earthquakes have allowed delineation of the lateral extent of magma chambers, such as Mount St. Helens (Scandone & Malone, 1985) and Mount Pinatubo (Mori et al., 1996; Pallister et al., 1996), the relative scarcity of Montserrat volcanic earthquakes with hypocentral depths greater than 5–6 km leaves open questions about the size and shape of the Soufrière Hills Volcano andesitic magma storage region. Similarly, ground deformation studies since the eruption began have indicated that any ground motions caused by expansion or deflation of the magma storage region would be masked by the much larger near-surface motions caused by deformation of the volcanic edifice and upper conduit (Jackson et al., 1998; Shepherd et al., 1998; Voight et al., 1998; compare Mattioli et al., 1998).
The dimensions of the uppermost conduit can be inferred from a time series of measurements of extruded magma volume, from the distribution of ballistics from explosive eruptions, and from calculations of magma extrusion rates and ascent rates. The upper part of the conduit between the growing dome and the top of the underlying magma storage region was estimated by Robertson et al. (1998) to have a cross-sectional area of ∼700 m2, based on ballistics analysis and calculated discharge rates of the 17 September 1996 explosive eruption. This equates to a circle with a diameter of ∼30 m. An independent estimate of the conduit diameter can be obtained using magma ascent rate data (Devine et al., 1998b) and the eruption rate calculated from dome volume changes (Sparks et al., 1998). The eruption rate appears to have been relatively constant at about 1·9 m3/s in the period 110–200 days into the eruption (Sparks et al., 1998). Samples of the 12 May 1996 eruption contain hornblendes with a well-defined population of thin decompression breakdown rims (∼2 µm thick), which, using the Montserrat-specific experimental calibration of the amphibole breakdown magma ascent rate speedometer of Rutherford & Devine (2003), yield estimated average ascent rates between 0·017 and 0·022 m/s. Assuming conduit transport in a region where magma volume is constant (i.e. before significant bubble formation), one can calculate the approximate cross-sectional area of the conduit by dividing the volume eruption rate by the ascent rate. This calculation yields a conduit diameter of ∼11–12 m for a cylindrical geometry (tacitly assuming a uniform flow velocity; R. S. J. Sparks, personal communication, 2002), with an error of ±7 m, assuming errors of ±0·5 m3/s in eruption rate and ±0·001 m/s in the magma ascent rate. It is possible that conduit diameter varies with depth, being ∼30 m diameter near the surface and ∼11–12 m at greater depths. The point is that the conduit is narrow.
Using the larger estimated conduit diameter of 30 m, and assuming that the conduit has a uniform diameter with depth down to the top of the andesitic magma storage region (5–6 km), <1 vol. % of the total erupted volume of >440 × 106 m3 of andesite (Montserrat Volcano Observatory open file reports; R. Herd, personal communication, 2002) could have been contained at any one time in the discharge conduit before, or during, the eruption. Similarly, if one presumes that the mafic magma that has triggered the present eruption of andesite (from its ‘upper-crustal’ reservoir at 5–6 km depth) is itself derived from an underlying, ‘mid-crustal’ basaltic magma reservoir (say, 10 km deep; depth calculation from Devine & Rutherford, in prep.), by flow up a lower conduit of similar cross-section (i.e. 30 m diameter), one may also conclude that such a basalt-filled conduit could not contain a sufficient volume of magma to have pushed out the >440 × 106 m3 of andesite from its upper-crustal reservoir on a volume-for-volume basis.
These observations support the inference that there is a pre-existing, extensive reservoir of andesite in the upper crust. Additional support for the idea is that the eruption products of the Soufrière Hills Volcano have been in a relatively narrow compositional range for the last 31 kyr (Roobol & Smith, 1998), implying that magmatic processes have produced some sort of quasi-steady state within the andesite range. The relatively low Al contents of amphiboles in the andesites indicate a relatively narrow range of water pressures averaging ∼130 MPa (Rutherford & Devine, 2003), an estimate consistent with H2O contents of melt inclusions in plagioclase phenocrysts (Barclay et al., 1998; Devine et al., 1998a). This suggests that the andesitic magma storage region, although of unknown dimensions, does not extend over a significant pressure range, and possibly has the general shape of an oblate ellipsoid. It is probably intermittently connected by a narrow conduit to an underlying magma storage region filled with hydrous, high-Al basalt, as illustrated schematically in Fig. 13.
Schematic cross-section of the Lesser Antilles arc crust beneath Montserrat showing inferred volcanic plumbing. The proposed layering and density structure of the arc crust are an extrapolation of the seismic refraction study of Boynton et al. (1979). It is proposed that ascending mantle-derived hydrous basaltic melts stagnate at the interface between the intermediate- and lower-crustal layers as a result of the effect of the density structure on melt buoyancy (Devine, 1995). An andesitic magma storage region has been developed at the interface between the upper- and intermediate-crustal layers (see text for discussion).
Schematic cross-section of the Lesser Antilles arc crust beneath Montserrat showing inferred volcanic plumbing. The proposed layering and density structure of the arc crust are an extrapolation of the seismic refraction study of Boynton et al. (1979). It is proposed that ascending mantle-derived hydrous basaltic melts stagnate at the interface between the intermediate- and lower-crustal layers as a result of the effect of the density structure on melt buoyancy (Devine, 1995). An andesitic magma storage region has been developed at the interface between the upper- and intermediate-crustal layers (see text for discussion).
From time to time, the andesitic reservoir is invaded by injections from the basaltic magma storage region, either because of density instability in the latter, or because the basaltic reservoir has been injected with a new aliquot of mantle-derived, primitive basaltic melt. Petrologic observations (i.e. relatively narrow range of Al contents in natural amphibole crystals; narrow range of melt inclusion estimated water contents; pronounced oscillatory zoning of plagioclase crystals; Devine et al., 1998a) and experiments (e.g. variation of Al contents in experimental amphiboles with pressure; Rutherford & Devine, 2003) rule out models of the Soufrière Hills Volcano plumbing system that involve polybaric, decompression crystallizaton of ascending andesitic mush columns of the type proposed by Blundy & Cashman (2001). Extensive petrographic evidence for magma mixing in all eruptive products of the Soufrière Hills Volcano, i.e. the ubiquitous ∼1 vol. % mafic inclusions and mineral disequilibrium features, strongly suggests that the andesite is derived by fractional crystallization of a hydrous basaltic parent magma (see below); that is, the andesite is not a primary hydrous partial melt of sub-arc mantle (see Carmichael, 2002). Magma mixing and crustal assimilation processes do occur, but the most important differentiation process is fractional crystallization (e.g. Devine, 1995).
One possible explanation for the predominance of andesite in past eruptive products is that the inferred andesitic reservoir is large enough to assimilate, by intimate hybridization, inputs of potentially hundreds of millions of cubic meters of basaltic magma, without producing measurably large changes in the average composition of andesitic magma in the reservoir. There is as yet no geophysical evidence for or against the conjecture that such a large magma chamber actually exists (e.g. attenuation of shear waves; J. B. Shepherd, personal communication, 2002). Another possibility is that only a small portion of the total basaltic magma that is input into the upper-crustal reservoir actually mixes with the andesite (i.e. it is only the ∼1 vol. % comprising the mafic inclusions observed in the erupted andesite); the rest of the basalt mainly transfers the heat required to mobilize the andesite.
Heat transfer models for the two-magma system
There are at least two viable models of processes in which heat, but not mass, may be transferred from the invading basalt to the andesite in the inferred magma storage region. One model postulates that dike-like injections of basaltic magma into the andesite reservoir provide ample surface area for heat transfer; but viscosity contrast prevents extensive mechanical mixing of the two magmas (see, e.g. Sparks & Marshall, 1986; Blundy & Sparks, 1992). Alternatively, basalt ponds at the base of the andesitic magma storage region, with mostly heat being transferred across a horizontal interface (e.g. Snyder, 2000). Couch et al. (2001) have proposed a model of the latter type for the Soufrière Hills andesite magma chamber, in which a heated boundary layer at the base of the andesitic magma storage region buoyantly rises, causing ‘self-mixing’ of the andesite and then eruption (Fig. 14a).
Schematic illustration of possible evolution of the andesitic magma reservoir. (a) Onset of the present eruption, with small batches of injected basalt ‘ponding out’ (i.e. stagnating as a result of rheological or density contrast) at the base of the andesitic magma storage region, causing remobilization of a thin boundary layer of andesite; ‘ponded basalt’ refers to the most recent injection increment; ‘remobilized andesite’ refers to the andesite layer heated by the most recent basalt injection; shape and dimensions (∼60 km3) of the storage region are hypothetical (note that eruptions of several volcanoes in the Lesser Antilles have exceeded 30 km3 dense rock equivalent). (b) Potential result of hypothetical continuation of the present eruption, if injection of basaltic magma continues and begins to collect at the base of the andesitic reservoir, rather than become hybridized with the andesite; dimensions and layer thicknesses are schematic (note that the inferred layering in the magma storage region mirrors the larger-scale inferred layering of the arc crust).
Schematic illustration of possible evolution of the andesitic magma reservoir. (a) Onset of the present eruption, with small batches of injected basalt ‘ponding out’ (i.e. stagnating as a result of rheological or density contrast) at the base of the andesitic magma storage region, causing remobilization of a thin boundary layer of andesite; ‘ponded basalt’ refers to the most recent injection increment; ‘remobilized andesite’ refers to the andesite layer heated by the most recent basalt injection; shape and dimensions (∼60 km3) of the storage region are hypothetical (note that eruptions of several volcanoes in the Lesser Antilles have exceeded 30 km3 dense rock equivalent). (b) Potential result of hypothetical continuation of the present eruption, if injection of basaltic magma continues and begins to collect at the base of the andesitic reservoir, rather than become hybridized with the andesite; dimensions and layer thicknesses are schematic (note that the inferred layering in the magma storage region mirrors the larger-scale inferred layering of the arc crust).
We consider first the general case of heating and remobilization of the andesitic magma by dike-like or sill-like injections of basalt. If the remobilized magma were to be erupted essentially immediately, the Fe–Ti oxide minerals in the erupted magma might not have time to complete compositional changes caused by that recent heating (i.e. increases in titanomagnetite Ti contents by diffusion of Ti from the melt, and eventual rehomogenization of grains with diffusion gradients). If a time interval of, say, 2–4 weeks occurred between the time of heating and the time of eruption, experimental evidence suggests that measurable Ti diffusion gradients in titanomagnetite phenocryst rims should be developed (Venezky & Rutherford, 1999; Rutherford & Devine, 2003).
We consider next the case of the andesite in the magma reservoir being heated from below by an increasingly larger volume of basaltic magma, which has ponded at the base of the reservoir (Fig. 14b). As the volume of basalt increases, the cumulative amount of heat transferred to the andesite should increase. This heating might be recorded by a general increase in the TiO2 contents of the titanomagnetite phenocrysts in the andesite.
Nature of mafic magma injections
We now examine potential geometries of mafic magma injection into the andesite, in light of petrologic and geophysical constraints. Basaltic magma may pond at the base of silicic magma chambers in sill-like layers as a result of density contrast (e.g. Snyder, 2000; Couch et al., 2001), but this may not necessarily be the case if the basalt has a sufficiently high H2O content (Ochs & Lange, 1999; Lange, 2002). On the other hand, there is no evidence to indicate whether or not dike-like injections of basalt have played an important role in the current eruption. To date, only minor amounts of basalt appear to have mingled with the andesite, and there have been no episodes of basalt eruption, so if dike-like plumbing is important, the basalt has not yet stoped through the andesite.
One incidence of nearly simultaneous eruption of intermediate and basaltic magma occurred in nearby St Kitts, where basaltic cinders (SiO2 ∼ 49–50 wt %) are interbedded with andesitic tephra (SiO2 ∼ 59–62 wt %) in the Mansion Series of Mt. Misery (now called ‘Liamuiga’; Baker, 1968, 1969, 1980; Baker & Holland, 1972). This example, and the presence of the South Soufrière Hills Volcano basaltic center so close to the Soufrière Hills Volcano, points to the possible eruption of basaltic magma from the latter. Roobol & Smith (1998) reported banded blocks in marker bed ‘M’, subunit III (∼17 ka BP) of the Soufrière Hills Volcano, which are composed of intimately mingled andesite and dark basaltic andesite, but they did not report the presence of any discrete basaltic layers. There is as yet no evidence to support a model of dike-like injection of basaltic magma into, or through, the andesitic magma storage region beneath the Soufrière Hills Volcano.
One possibility for control of the geometry of basalt–andesite interaction at the Soufrière Hills Volcano is the relatively high viscosity of the pre-eruptive andesite (>107 Pa s at T≤840°C; Devine & Rutherford, in prep.). The relatively stiff andesite, before invasion by the more fluid basalt (viscosity ∼102 Pa s; Devine & Rutherford, in prep.), may prevent the basalt from ascending any higher than the transition at 5–6 km depth from brittle failure to ductile deformation in the country rocks that is suggested by the vertical distribution of earthquake focal depths (Aspinall et al., 1998; J. B. Shepherd, personal communication, 2001). In oceanic crust, such a brittle–ductile transition appears to coincide with the 750°C isotherm (Phipps Morgan & Chen, 1993a, 1993b; Perfit & Chadwick, 1998; see also Fournier, 1999). In the case of the Soufrière Hills Volcano, the high contrast between the viscosities of the highly fluid, hydrous basalt and the highly crystallized (∼50 wt %) andesite might prevent commingling of the two magmas (see also Sparks & Marshall, 1986). Ponding of basaltic magma at the base of a pre-existing andesitic magma storage region is the simplest explanation consistent with observations.
Nature of heating of the remobilized andesite
The strict time constraints placed on andesitic magma remobilization and eruption by the Fe–Ti oxide data strongly suggest that the remobilized batches of andesite are relatively small (Fig. 15). This in turn suggests that the basaltic injections are small, perhaps sill-like bodies with thicknesses of the order of 1–2 m (Fig. 16; e.g. Snyder & Tait, 1995). The transfer of heat from a sill-like intrusion of basaltic melt into the base of a silicic magma chamber has been analyzed by Snyder (2000), so it is possible to calculate the likely thicknesses of remobilized layers of Soufrière Hills Volcano andesite using thermal constraints developed above. In short, a basaltic sill of 1 m thickness (say, at 1000°C) will remobilize a layer of 830°C andesite ∼1·6 m thick, the heat transfer probably taking just a couple of days (Devine & Rutherford, in prep.).
Schematic illustration of basaltic magma injection into the base of the andesitic magma storage region. The 760 m scale bar is intended to illustrate the volume of magma erupted as of September 2002, of ∼440 × 106 m3; 760 m is the cube root of the volume estimate. Scale of diagram is consistent with Fig. 14. Area enclosed by box is enlarged in Fig. 16.
Schematic illustration of basaltic magma injection into the base of the andesitic magma storage region. The 760 m scale bar is intended to illustrate the volume of magma erupted as of September 2002, of ∼440 × 106 m3; 760 m is the cube root of the volume estimate. Scale of diagram is consistent with Fig. 14. Area enclosed by box is enlarged in Fig. 16.
Schematic enlargement of the zone of basalt injection, showing the possible geometry of sill-like intrusions of basalt (see also Snyder & Tait, 1995). Basaltic sill-like injections (dark pattern) are shown at the base of the reservoir of andesite (lightly stippled pattern). A thin boundary layer of remobilized andesite (heavily stippled pattern) is extracted by a vertical conduit that drains the layer and feeds the overlying conduit that extends through the upper crust. Break in scale between schematic bottom and top of andesitic reservoir should be noted. The lower part of the diagram is equivalent to the area of the box in Fig. 15. Basalt injections may be intermingled with partially hybridized andesite at the bottom of the remobilized andesite layer. The thickness of the remobilized layer and the extent of any potential magma hybridization may vary with the size of the basaltic magma injections.
Schematic enlargement of the zone of basalt injection, showing the possible geometry of sill-like intrusions of basalt (see also Snyder & Tait, 1995). Basaltic sill-like injections (dark pattern) are shown at the base of the reservoir of andesite (lightly stippled pattern). A thin boundary layer of remobilized andesite (heavily stippled pattern) is extracted by a vertical conduit that drains the layer and feeds the overlying conduit that extends through the upper crust. Break in scale between schematic bottom and top of andesitic reservoir should be noted. The lower part of the diagram is equivalent to the area of the box in Fig. 15. Basalt injections may be intermingled with partially hybridized andesite at the bottom of the remobilized andesite layer. The thickness of the remobilized layer and the extent of any potential magma hybridization may vary with the size of the basaltic magma injections.
Couch et al. (2001) made a similar analysis of the heat transfer from basalt to andesite, using different boundary conditions, which were based on the assumption that the andesite could be heated to much higher temperatures than those inferred here. They concluded that the heat transfer was sufficient to cause ‘self-mixing’ of the andesite, although they stated that a boundary layer in the andesite only 2–4 m thick would be mobilized. In our view, this rather limited heating may be insufficient to cause large-scale mixing of the andesite; first, because the storage region must be sufficiently large to have contained all the magma erupted to date; second, because the andesite outside the 1–2 m, or 2–4 m, boundary layer will not ‘see’ the heating event (Snyder, 2000); third, because the remobilized andesite in the boundary layer is apparently erupted within a few weeks of the onset of reheating. We agree with the Bristol group that continued injection of basaltic magma into the base of the plumbing system over time periods as long as decades is likely to result in global heating of all the andesite in the magma reservoir.
The volume heating problem is illustrated with reference to Fig. 15, which is a schematic representation of the basaltic magma injection zone at the base of the andesitic reservoir. The 760 m scale bar in the figure is used to indicate the volume of magma that has been erupted from the storage region since 1995 (∼440 × 106 m3). It is unlikely that andesite away from the zone of basalt injection could be appreciably heated by an individual injection of basalt only 1–2 m thick. A more appropriate question is how to move small aliquots of andesite from the heated injection zone to the surface.
Magma ascent pathways are used over and over again in the Lesser Antilles, with most volcanoes producing multiple eruptions from the same conduit (e.g. Boynton et al., 1979). Certainly the previous two eruptions of the Soufrière Hills Volcano (∼400 and ∼3950 years ago) used the current conduit. Whether the conduit is a tectonic feature (e.g. controlled by conjugate faults), or a pre-warmed pathway, is unknown. However, the Fe–Ti oxide data suggest that an internal conduit within the andesitic magma storage region exists, which carries magma from the zone of heating to the overlying conduit that in turn carries the magma through the upper arc crust. In this model, the magma chamber is being emptied from the bottom, not the top. The possibility that the zone of basaltic magma injection is at the top of the andesitic magma storage region, rather than the bottom (e.g. because the hydrous basaltic melt is possibly less dense than the colder, partially crystallized andesite), seems unlikely, because there is no evidence of extensive magma hybridization.
The model of melt extraction from the bottom of the andesitic magma storage region satisfies the thermal constraints imposed by the Fe–Ti oxide data, especially the observation that many of the titanomagnetite crystals erupted more than 2 years into the eruption retain their ‘original’ low-TiO2 (i.e. low-temperature) core compositions. It is possible that at times of high volume eruption rate (August–October 1997), there was a higher than average basaltic magma injection rate, with potentially thicker sills, and relatively rapid accumulation and extraction of layers of heated andesite. It is also possible that basaltic magma is displacing andesite on a volume-for-volume basis (Fig. 14b), and that crystallization of the accumulated basalt will, over time, result in global heating of the remaining andesite. Continuous monitoring of Fe–Ti oxide compositions may reveal if that is the case.
Finally, we conclude that petrographic evidence for limited magma hybridization can be reconciled with a lack of evidence for a monotonic change in magma composition, which one might expect from a process of continuing, thorough hybridization, by postulating that hybridization takes place only within the heated boundary layer at the base of the andesitic reservoir. The Fe–Ti oxides indicate that any such mixtures are quickly harvested. Slightly more or less hybridization with time of perhaps up to 10 or 20 wt % basaltic magma with andesite in the boundary layer could account for the variability and range of SiO2 contents observed during the course of the eruption.
CONCLUSIONS
Fe–Ti oxide compositions in 1995–2002 Soufrière Hills Volcano eruption products have been used in conjunction with experimentally determined phase relations of the andesitic magma (Rutherford & Devine, 2003) to infer the nature of heating and remobilization of the andesite by invading mafic magma. Comparison of zoning observed in natural titanomagnetite crystals with mineral zoning produced in controlled experiments constrains the timescales of magma heating and remobilization. Most batches of magma are heated within ∼4 weeks of eruption; in some cases, perhaps just days before eruption. The cores of most zoned titanomagnetite crystals have retained their pre-eruption TiO2 contents, suggesting that: (1) the magma remained at the pre-eruptive temperature of ≤830°C until heated and remobilized; (2) insufficient time had elapsed between heating and eruption to allow complete re-equilibration; (3) global heating of all the andesite in the mid-crustal storage region at about 5–6 km depth (P ∼ 130 MPa) has not yet occurred.
The andesite is probably heated and remobilized by relatively small, sill-like injections of basaltic magma into the base of the andesitic magma storage region. The heated boundary layer in the andesite adjacent to the basaltic injections rises from the bottom to the top of the andesitic magma storage region, and then up through the mid-crustal conduit that connects the storage region with the surface. The same ascent pathway appears to have been used by the magmas erupted at ∼400 years and ∼3950 years bp.
The conclusion that global heating of the andesite in the storage region has not occurred indicates that there is no evidence suggesting that the probability of large explosive eruptions is increasing (or decreasing). Therefore, there is no compelling evidence that a change in the present volcanic hazard zonation map for Montserrat is required.
The present eruption is inferred to have been triggered by injections of hydrous, high-Al basalt, probably derived from its own storage region underlying the andesitic reservoir (>10 km?). The driving force of the eruption is a function of processes occurring in the basaltic reservoir. Basaltic melt may potentially be ejected from its storage region as a result of several factors. It is probably hydrous, so fractional crystallization will drive up the H2O content of residual liquid, making it increasingly less dense, more buoyant and gravitationally unstable. Crustal fracturing might then allow the basalt to rise and either intersect an overlying more evolved reservoir (such as Soufrière Hills), or erupt directly at the surface (e.g. South Soufrière Hills). This idea is consistent with the observation that volcanic eruptions in the Caribbean region may follow large earthquakes (Carr, 1977). Alternatively, the ejection of evolved, hydrous, high-Al basalt from its storage region could be caused by injections of parental basaltic melts derived directly from the mantle wedge. In either case, the real cause of the 1995–present eruption of the Soufrière Hills Volcano may be once or twice removed from processes occurring within the andesitic magma storage region.
Present address: British Geological Survey, Keyworth NG125 GG, UK.
Present address: 1506 Cordova Street, Coral Gables, FL 33134, USA.
This work was funded by the UK Department for International Development (formerly Overseas Development Administration) and administered by the British Geological Survey and the Montserrat Volcano Observatory. This work benefited from helpful discussions with Steve Sparks, John Shepherd, Don Snyder, and Yan Liang. Thorough reviews by Steve Sparks, Michel Pichavant, and Peter Kokelaar helped improve the manuscript. The assistance of MVO staff in obtaining the samples used in this and previous studies is gratefully acknowledged. Charlie Mandeville, of the American Museum of Natural History, is thanked for assistance in obtaining the X-ray dot map. Published by permission of Director, British Geological Survey (NERC).
REFERENCES
Andersen, D. J. & Lindsley, D. H. (). Internally consistent solution models for Fe–Mg–Mn–Ti oxides.American Mineralogist
, 714
–726.Andersen, D. J., Bishop, F. C. & Lindsley, D. H. (). Internally consistent solution models for the Fe–Mg–Mn–Ti oxides: Fe–Mg–Ti oxides and olivine. American Mineralogist
, 427
–444.Anderson, A. T. (). Oxidation of the LaBlache Lake titaniferous magnetite deposit, Quebec. Journal of Geology
, 528
–547.Arculus, R. J. & Wills, K. J. A. (). The petrology of plutonic blocks and inclusions from the Lesser Antilles island arc. Journal of Petrology
, 743
–799.Aspinall, W. P., Miller, A. D., Lynch, L. L., Latchman, J. L., Stewart, R. C., White, R. A. & Power, J. A. (). Soufrière Hills eruption, Montserrat, 1995–1997: volcanic earthquake locations and fault plane solutions. Geophysical Research Letters
, 3397
–3400.Bacon, C. R. & Hirschmann, M. M. (). Mg/Mn partitioning as a test for equilibrium between Fe–Ti oxides. American Mineralogist
, 57
–61.Baker, P. E. (). Petrology of Mount Misery Volcano, Saint Kitts, West Indies. Lithos
, 124
–150.Baker, P. E. (). The geological history of Mount Misery volcano, St. Kitts, West Indies. Overseas Geology and Mineral Resources (Great Britain, Institute of Geological Sciences)
, 207
–230.Baker, P. E. (). Geology and geochemistry of the Mansion pyroclastic fall succession, St. Kitts. Bulletin Volcanologique
, 303
–310.Baker, P. E. & Holland, J. G. (). Geochemical variations in a pyroclastic succession on St. Kitts. Bulletin Volcanologique
, 472
–490.Barclay, J., Rutherford, M. J., Carroll, M. R., et al. (). Experimental phase equilibria constraints on pre-eruptive storage conditions of the Soufrière Hills magma. Geophysical Research Letters
, 3437
–3440.Blundy, J. & Cashman, K. (). Ascent-driven crystallisation of dacite magmas at Mount St. Helens, 1980–1986. Contributions to Mineralogy and Petrology
, 631
–650.Journal of Petrology
, 1039
–1104.Boynton, C. H., Westbrook, G. K., Bott, M. H. P. & Long, R. E. (). A seismic refraction investigation of crustal structure beneath the Lesser Antilles island arc. Geophysical Journal of the Royal Astronomical Society
, 371
–393.Carmichael, I. S. E. (). The iron–titanium oxides of salic volcanic rocks and their associated ferromagnesian silicates. Contributions to Mineralogy and Petrology
, 36
–64.Carmichael, I. S. E. (). The andesite aqueduct: perspectives on the evolution of intermediate magmatism in west–central (105–99°W) Mexico. Contributions to Mineralogy and Petrology
, 641
–663.Carr, M. J. (). Volcanic activity and great earthquakes at convergent plate boundaries. Science
, 655
–657.Chiodini, G., Cioni, R., Frullani, A., Guidi, M., Marini, L., Prati, F. & Raco, B. (). Fluid geochemistry of Montserrat Island, West Indies. Bulletin Volcanologique
, 380
–392.Clarke, A. B., Voight, B., Neri, A. & Macedonio, G. (). Transient dynamics of vulcanian explosions and column collapse. Nature
, 897
–901.Clarke, A. B., Neri, A., Voight, B., Macedonio, G. & Druitt, T. H. (). Computational modelling of the transient dynamics of the August 1997 Vulcanian explosions at Soufriere Hills Volcano, Montserrat: influence of initial conduit conditions on near-vent pyroclastic dispersal. In: Druitt, T. H. & Kokelaar, B. P. (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir21, 319–348.
Cottrell, E., Gardner, J. E. & Rutherford, M. J. (). Petrologic and experimental evidence for the movement and heating of the pre-eruptive Minoan rhyodacite (Santorini, Greece). Contributions to Mineralogy and Petrology
, 315
–331.Couch, S., Sparks, R. S. J. & Carroll, M. R. (). Mineral disequilibrium in lavas explained by convective self-mixing in open magma chambers. Nature
, 1037
–1039.Devine, J. D. (). Role of volatiles in Lesser Antilles island arc magmas. Ph.D. thesis, University of Rhode Island, 612 pp.
Devine, J. D. (). Petrogenesis of the basalt–andesite–dacite association of Grenada, Lesser Antilles island arc, revisited. Journal of Volcanology and Geothermal Research
, 1
–33.Devine, J. D., Murphy, M. D., Rutherford, M. J., et al. (). Petrologic evidence for pre-eruptive pressure–temperature conditions, and recent reheating, of andesitic magma erupting at the Soufrière Hills Volcano, Montserrat, W. I. Geophysical Research Letters
, 3669
–3672.Devine, J. D., Rutherford, M. J. & Gardner, J. E. (). Petrologic determination of ascent rates for the 1995–1997 Soufrière Hills Volcano andesitic magma. Geophysical Research Letters
, 3673
–3676.Druitt, T. H., Young, S. R., Baptie, B., Bonadonna, C., Calder, E. S., Clarke, A. B., Cole, P. D., Harford, C. L., Herd, R. A., Luckett, R., Ryan, G. & Voight, B. (). Episodes of repetitive vulcanian explosions and fountain collapse at Soufrière Hills Volcano, Montserrat. In: Druitt, T. H. & Kokelaar, B. P. (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir21, 281–306.
Fournier, R. O. (). Hydrothermal processes related to movement of fluid from plastic into brittle rock in the magmatic–epithermal environment. Economic Geology
, 1193
–1212.Frost, B. R. & Lindsley, D. H. (). Occurrence of iron–titanium oxides in igneous rocks. In: Lindsley, D. H. (ed.) Oxide Minerals: Petrologic and Magnetic Significance. Mineralogical Society of America, Reviews in Mineralogy25, 433–468.
Gardner, J. E., Carey, S., Rutherford, M. J. & Sigurdsson, H. (). Petrologic diversity in Mount St. Helens dacites during the last 4,000 years: implications for magma mixing. Contributions to Mineralogy and Petrology
, 224
–238.Geschwind, C.-H. & Rutherford, M. J. (). Cummingtonite and the evolution of the Mount St. Helens magma system: an experimental study. Geology
, 1011
–1014.Ghiorso, M. S. & Sack, R. O. (). Fe–Ti oxide geothermometry: thermodynamic formulation and estimation of intensive variables in silicic magmas. Contributions to Mineralogy and Petrology
, 485
–510.Gill, J. B. (). Orogenic Andesites and Plate Tectonics. Berlin: Springer, 390 pp.
Hammouya, G., Allard, P., Jean-Baptiste, P., Parello, F., Semet, M. P. & Young, S. R. (). Pre- and syn-eruptive geochemistry of volcanic gases from Soufrière of Montserrat, West Indies. Geophysical Research Letters
, 3685
–3688.Harford, C. L. & Sparks, R. S. J. (). Recent mobilisation of shallow-level intrusions on Montserrat revealed by hydrogen isotope composition of amphiboles. Earth and Planetary Science Letters
, 285
–297.Harford, C. L., Pringle, M. S., Sparks, R. S. J. & Young, S. R. (). The volcanic evolution of Montserrat using 40Ar/39Ar geochronology. In: Druitt, T. H. & Kokelaar, B. P. (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir21, 93–113.
Jackson, P., Shepherd, J. B., Robertson, R. E. A. & Skerritt, G. (). Ground deformation studies at Soufrière Hills Volcano, Montserrat I: electronic distance meter studies. Geophysical Research Letters
, 3409
–3412.Lange, R. A. (). Constraints on the preeruptive volatile concentrations in the Columbia River flood basalts. Geology
, 179
–182.Lindsley, D. H. & Spencer, K. J. (). Fe–Ti oxide geothermometry: reducing analyses of coexisting Ti-magnetite and ilmenite. Transactions, American Geophysical Union
, 471
.Martel, C., Pichavant, M., Holtz, F., Scaillet, B., Bourdier, J.-L. & Traineau, H. (). Effects of fO2 and H2O on andesite phase relations between 2 and 4 kbar. Journal of Geophysical Research
, 29453
–29470.Mattioli, G. S., Dixon, T. H., Farina, F., Howell, E. S., Jansma, P. E. & Smith, A. L. (). GPS measurement of surface deformation around Soufrière Hills volcano, Montserrat from October 1995 to July 1996. Geophysical Research Letters
, 3417
–3420.Melnik, O. & Sparks, R. S. J. (). Nonlinear dynamics of lava dome extrusion. Nature
, 37
–41.Melnik, O. & Sparks, R. S. J. (). Dynamics of magma ascent and lava extrusion at Soufrière Hills Volcano, Montserrat. In: Druitt, T. H. & Kokelaar, B. P. (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir21, 307–317.
Miller, A. D., Stewart, R. C., White, R. A., Luckett, R., Baptie, B. J., Aspinall, W. P., Latchman, J. L., Lynch, L. L. & Voight, B. (). Seismicity associated with dome growth and collapse at the Soufrière Hills Volcano, Montserrat. Geophysical Research Letters
, 3401
–3404.Mori, J., Eberhart-Phillips, D. & Harlow, D. H. (). Three-dimensional velocity structure at Mount Pinatubo: resolving magma bodies and earthquake hypocenters. In: Newhall, C. G. & Punongbayan, R. S. (eds) Fire and Mud: Eruptions and Lahars of Mount Pinatubo. Philippine Institute of Volcanology. Seattle, WA: University of Washington Press, pp. 371–382.
Murphy, M. D., Sparks, R. S. J., Barclay, J., et al. (). The role of magma mixing in triggering the current eruption at the Soufrière Hills volcano, Montserrat, West Indies. Geophysical Research Letters
, 3433
–3436.Murphy, M. D., Sparks, R. S. J., Barclay, J., Carroll, M. R. & Brewer, T. S. (). Remobilization of andesite magma by intrusion of mafic magma at the Soufrière Hills Volcano, Montserrat, West Indies. Journal of Petrology
, 21
–42.Nakamura, M. (). Continuous mixing of crystal mush and replenished magma in the ongoing Unzen eruption. Geology
, 807
–810.Norton, G. E., Watts, R. B., Voight, B., Mattioli, G., Herd, R. A., Young, S. R., et al. (). Pyroclastic flow and explosive activity at Soufrière Hills Volcano, Montserrat, during a period of virtually no magma extrusion (March 1998 to November 1999). In: Druitt, T. H. & Kokelaar, B. P. (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir21, 467–481.
Ochs, F. A. & Lange, R. A. (). The density of hydrous magmatic liquids. Science
, 1314
–1317.Pallister, J. S., Hoblitt, R. P., Meeker, G. P., Knight, R. J. & Seims, D. F. (). Magma mixing at Mount Pinatubo: petrographic and chemical evidence from the 1991 deposits. In: Newhall, C. G. & Punongbayan, R. S. (eds) Fire and Mud: Eruptions and Lahars of Mount Pinatubo. Philippine Institute of Volcanology. Seattle, WA: University of Washington Press, pp. 687–731.
Perfit, M. R. & Chadwick, W. W., Jr (). Magmatism at mid-ocean ridges: constraints from volcanological and geochemical investigations. In: Buck, W. R., Delaney, P. T., Karson, J. A. & Lagabrielle, Y. (eds) Faulting and Magmatism at Mid-Ocean Ridges. Geophysical Monograph, American Geophysical Union106, 59–115.
Phipps Morgan, J. & Chen, Y. J. (). Dependence of ridge-axis morphology on magma supply and spreading rate. Nature
, 706
–708.Phipps Morgan, J. & Chen, Y. J. (). The genesis of oceanic crust: magma injection, hydrothermal circulation, and crustal flow. Journal of Geophysical Research
, 6283
–6297.Pichavant, M., Martel, C., Bourdier, J.-L. & Scaillet, B. (). Physical conditions, structure, and dynamics of a zoned magma chamber: Mount Pelée (Martinique, Lesser Antilles Arc). Journal of Geophysical Research
, ECV 2–1
to 2–30.Robertson, R. E. A., Cole, P., Sparks, R. S. J., et al. (). The explosive eruption of Soufrière Hills Volcano, Montserrat, West Indies, 17 September, 1996. Geophysical Research Letters
, 3429
–3432.Roobol, M. J. & Smith, A. L. (). Pyroclastic stratigraphy of the Soufrière Hills volcano, Montserrat—implications for the present eruption. Geophysical Research Letters
, 3393
–3396.Rutherford, M. J. & Devine, J. D. (). The May 18, 1980, eruption of Mount St. Helens, III. Stability and chemistry of amphibole in the magma chamber. Journal of Geophysical Research
, 11949
–11959.Rutherford, M. J. & Devine, J. D. (). Magmatic conditions and magma ascent as indicated by hornblende phase equilibria and reactions in the 1995–2002 Soufrière Hills magma. Journal of Petrology
, 1433
–1453.Scandone, R. & Malone, S. D. (). Magma supply rate, magma discharge and readjustment of the feeding system of Mount St. Helens during 1980. Journal of Volcanology and Geothermal Research
, 239
–262.Shepherd, J. B., Tomblin, J. F. & Woo, D. (). Volcano-seismic crisis in Montserrat, West Indies. Bulletin Volcanologique
, 143
–163.Shepherd, J. B., Herd, R. A., Jackson, P. & Watts, R. (). Ground deformation measurements at the Soufrière Hills Volcano, Montserrat II: rapid static GPS measurements June 1996–June 1997. Geophysical Research Letters
, 3413
–3416.Snyder, D. (). Thermal effects of the intrusion of basaltic magma into a more silicic magma chamber and implications for eruption triggering. Earth and Planetary Science Letters
, 257
–273.Snyder, D. & Tait, S. (). Replenishment of magma chambers: comparison of fluid-mechanic experiments with field relations. Contributions to Mineralogy and Petrology
, 230
–240.Sparks, R. S. J. & Marshall, L. A. (). Thermal and mechanical constraints on mixing between mafic and silicic magmas. Journal of Volcanology and Geothermal Research
, 99
–124.Sparks, R. S. J., Young, S. R., Barclay, J., et al. (). Magma production and growth of the lava dome of the Soufrière Hills Volcano, Montserrat, West Indies: November 1995 to December 1997. Geophysical Research Letters
, 3421
–3424.Stormer, J. C. (). The effects of recalculation on estimates of temperature and oxygen fugacity from analyses of multi-component iron–titanium oxides. American Mineralogist
, 586
–594.Venezky, D.Y. & Rutherford, M. J. (). Petrology and Fe–Ti oxide reequilibration of the 1991 Mount Unzen mixed magma. Journal of Volcanology and Geothermal Research
, 213
–230.Voight, B., Hoblitt, R. P., Clarke, A. B., Lockhart, A. B., Miller, A. D., Lynch, L. L. & McMahon, J. (). Remarkable cyclic ground deformation monitored in real-time on Montserrat, and its use in eruption forecasting. Geophysical Research Letters
, 3405
–3408.Wadge, G. & Isaacs, M. C. (). Mapping the volcanic hazards from Soufrière Hills Volcano, Montserrat, West Indies, using an image processor. Journal of the Geological Society, London
, 541
–551.Williams, D. L., Abrams, G., Finn, C., Dzurisin, D., Johnson, D. J. & Denlinger, R. (). Evidence from gravity data for an intrusive complex beneath Mount St. Helens. Journal of Geophysical Research
, 10207
–10222.Woods, A. W., Sparks, R. S. J., Ritchie, L. J., Batey, J., Gladstone, C. & Bursik, M. I. (). The explosive decompression of a pressurised lava dome: 26 December 1997 at Soufrière Hills Volcano, Montserrat. In: Druitt, T. H. & Kokelaar, B. P. (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir21, 457–465.
Young, S. R., Hoblitt, R. P., Smith, A. L., Devine, J. D., Wadge, G. & Shepherd, J. B. (). Dating of explosive volcanic eruptions associated with dome growth at the Soufrière Hills volcano, Montserrat, West Indies. In: Second Caribbean Conference on Natural Hazards and Hazard Management. MVO Open File Report 96/22, Kingston, Jamaica.