watt_supporting_material

advertisement

1

2

3

4

5

35

36

37

38

39

40

31

32

33

34

41

42

43

44

45

46

47

48

49

24

25

26

27

28

29

30

11

12

13

14

15

16

17

6

7

8

9

10

18

19

20

21

22

23

Supporting Material

Arc magma compositions controlled by linked thermal and chemical gradients above the subducting slab

Sebastian F.L. Watt, David M. Pyle, Tamsin A. Mather, José A. Naranjo

Sample descriptions

This study analysed melt inclusions from four separate volcanic eruptions, at the monogenetic cones of Apagado, South Minchinmávida, Palena (a volcanic field comprising several cones), and at the basaltic volcano of Hornopirén. Although Hornopirén is not a monogenetic cone, its only known products are basalts of very similar composition [ Watt et al.

, 2011b], and the conical morphology and relatively small size of the edifice suggests a young volcano with a simple history.

Each of the four studied eruptions produced well-bedded scoria fall deposits (hereafter referred to as eruptive units). These deposits typically comprised multiple fining-upward beds of ash or lapilli, on a cm-scale. For each deposit, samples were taken from individual beds within proximal exposures (1–10 km from source). All site locations are given in Table S1.

Each of the four erupted units studied here were found to contain unzoned olivine as a major phenocryst phase. For Ap1, this is the only phenocryst phase, while in the other units plagioclase is also present. The macroscopic appearance of each deposit is broadly similar, with no changes in any deposit that suggest significant co-eruptive compositional variations.

For each unit, free olivine crystals were picked from selected samples for melt inclusion analysis, bulk scoria compositions were measured by X-ray fluorescence spectrometry

(XRF), and additional trace element compositions on multiple samples, from separate beds, we measured by solution ICP-MS (Tables S1 to S5). For any one eruption, samples had very similar major and trace element compositions, suggesting homogeneous magmatic compositions throughout the scoria-producing eruptive phase. The exception to this was variation in MgO content in Ap1, which we suggest (see below) was the result of excess olivine accumulation in some Ap1 samples. Given the homogeneous nature of each individual eruptive unit, we assume that our measured XRF compositions (Table S4) are representative, even if they were not from the same bed from which the melt inclusions were sampled (although most, in fact, are from the same samples (Supplementary Figure 1)).

Ap1 eruption, Apagado

The Ap1 fall deposit has a volume of 1.0 km

3

, and has been correlated up to 50 km from source [ Watt et al.

, 2011b]. The magnitude of Ap1 is notably large for a pulsatory explosive eruption of mafic magma. The bulk composition of the dispersed scoria deposit is also unusually primitive, reaching in excess of 15 wt% MgO in some samples. Samples from the top surface of the cone are less primitive (7 wt% MgO) [ Watt et al.

, 2011b], and we suggest that magmas affected by greater degrees of crystal fractionation and crustal assimilation were erupted in the later stages of the eruption. This is also indicated by the presence of augite phenocrysts in the cone samples, which are absent in the regional scoria fall deposit. For this reason, we focus simply on the earliest erupted material, within the coarse, lower part of the scoria fall deposit [cf. Watt et al.

, 2011b].

Olivine, ubiquitously associated with Cr-spinel, is the single phenocryst phase in Ap1.

All olivines have a very thin, late-stage rim, which is less forsteritic. In all samples the groundmass is dominated by acicular plagioclase (An

~75

), with augite, olivine (Fo

79

), rarer pigeonite and Ti-magnetite also present. Three Ap1 samples are studied here, taken from two sites: 21-5 (3 km west of the cone) and 15-6 (10 km east of the cone). The samples are from three separate beds and are, from stratigraphically lowest to highest positions, 21-5B, 15-6B,

85

86

87

88

89

90

81

82

83

84

91

92

93

94

95

96

97

98

99

65

66

67

68

69

70

61

62

63

64

54

55

56

57

50

51

52

53

58

59

60

71

72

73

74

75

76

77

78

79

80

15-6A (Supplementary Figure 1). Bulk MgO contents of each of these beds, measured by

XRF, range from 12.0 to 15.1 wt%, and silica contents from 48.2 to 50.7 wt% [ Watt et al.

,

2011b]. Olivine phenocryst core compositions in each bed are indistinguishable, with an average of Fo

88.7

in 15-6A and 21-5B, and Fo

88.9

in 15-6B. Bulk rock Mg# (atomic

Mg/(Mg+Fe)) in the same rocks ranges from 76 in 15-6B to 74 in 15-6A and 72 in 21-5B.

Given the lack of observable differences between olivine phenocryst compositions, we suggest that the slight variation in bulk-rock MgO content is due to variable amounts of olivine accumulation, up to an estimated maximum of 15% phenocrysts, by volume, in 15-6B

(bulk rock, 15.1 wt% MgO). This is confirmed by the expected relationship between bulkrock MgO and olivine forsterite content at K d

Fe-Mg

(ol-liq)

values of 0.3–0.35, where sample

15-6B falls slightly below the line, indicative of olivine accumulation, in contrast to the other two samples. Given these observations, and the fact that all olivines are unzoned, euhedral, and of near-identical composition in each Ap1 sample, we conclude that our sampled olivines were in equilibrium with their host magma, and that none of our samples are xenocrysts.

Ho1 eruption, Hornopirén

The Ho1 fall deposit has a volume of approximately 0.3 km

3

[ Watt et al.

, 2011b], and was a less energetic eruption than Ap1. Here, we analyse samples from one site (801-1), 5 km east of the summit of Hornopirén. Three individual beds were sampled, all from the lower, coarser two-thirds of the deposit, with the lowest being 801-1D and the highest 801-1B

(Supplementary Figure 1). Samples from 801-1D were found to contain more forsteritic olivine phenocrysts (Fo

84-87

) than the upper two samples (Fo

82

), and we suggest the earlier stages of the eruption sourced the most mafic magma. We thus divide the Ho1 samples into

Ho1a (early, more mafic stages of eruption) and Ho1b (later, more evolved stages of eruption), and analyse each of these as separate sub-units. We only have bulk XRF analyses for Ho1b, which is more mafic (6.2 wt% MgO; 50.7 wt% SiO

2

) than Hornopirén lavas (4.7–

5.5 wt% MgO; 52.7–54.3 wt% SiO

2

) (Table S4). Ho1 samples contain plagioclase phenocrysts (An

75-80

), and have a groundmass dominated by plagioclase, with orthopyroxene, pigeonite and rare olivine.

Palena eruptions

The Palena volcanic group consists of at least five monogenetic cones, along a 37 km axis which splays west from the main lineament of the Liquiñe-Ofqui fault zone [ Cembrano et al.

, 1996], at a strike of 026°. The cones are densely vegetated and inaccessible, but road cuts expose scoria fall deposits. These deposits exceed 3m thickness in places, and may be the product of multiple separate eruptions. At their thickest, individual beds contain up to 20 cm of coarse lapilli (grain size up to 25 mm).

Here, we analyse two samples from beds in separate parts of the deposit. Although we cannot be certain of the source cone, we suggest from the bed coarseness and the presence of more than one nearby potential cone, that these samples are <5 km from source. The lower sample (826-3B) may be from a separate eruption to the upper sample (826-4A), since they are separated by a relatively fine, structureless section in the fall deposit (Supplementary

Figure 1), although there is no intervening soil or clear unconformity. XRF data suggest similar compositions for separate parts of the Palena deposit, with a silica content of 49 wt% and MgO content of 8 wt%. The lower sample (826-3B) contains slightly more forsteritic olivine phenocrysts (Fo

85.2

) than the upper sample (Fo

84.3

), although it is unclear if this difference is significant. Olivine is the dominant phenocryst phase in the Palena deposit, and plagioclase is also present.

A 15-cm-thick bedded scoria documented by Heusser et al.

[1992] 25 km north of our sampling site, and dated at ~11.5 kyr BP, may correlate with the Palena deposits on the basis of its description

124

125

126

127

128

129

116

117

118

119

120

121

122

123

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

107

108

109

110

111

112

113

114

115

100

101

102

103

104

105

106 and stratigraphic position. The age of the Palena deposits is also constrained by the overlying Yan1 tephra [ Naranjo and Stern , 2004], dated at ~9.5 kyr BP, and we therefore suggest that it erupted in the earliest Holocene.

South Minchinmávida

The South Minchinmávida cone is an isolated single cone, 17 km south of the summit of Minchinmávida volcano and on the main lineament of the Liquiñe-Ofqui fault zone. It is unclear whether this cone has a close genetic relationship to the Minchinmávida stratovolcano. Aerial imagery suggests that similar cones may occur in the vicinity, but this has not been verified. A track across the cone exposes scoria, bedded on a cm-scale. A single bed was sampled (C16-G7B), and has a bulk silica content of 50.6 wt% and bulk MgO content of 8.1 wt%. Olivine and plagioclase are present as phenocryst phases. In contrast to all of the other units studied here, the scoria consists of homogeneous black grains with a very open structure, whereas all other units comprise grains of two colours (dark-grey and yellow-brown or orange). The cone represents a relatively small eruption, and has a volume of ~0.1 km

3

. It is densely vegetated, and of unknown age (although presumed, from its preservation, to be Holocene).

Methods

Sample preparation and analysis

Following collection, selected scoria samples were analysed for major elements and selected trace elements using XRF at the Open University, Milton Keynes, U.K., on coarse sieved fractions. Powder pellets (for trace elements) and glass discs (for major elements) were produced for the samples and a range of internal and external rock standards. Standard recoveries were all within 1% relative for major elements (3% for Ca and Mn). Picked scoria grains from selected samples were analysed for trace element compositions by solution inductively-coupled-plasma mass spectrometry (ICP-MS) at the University of Oxford, U.K.

Solutions were produced by a high-pressure acid digestion process [cf. Watt et al.

, 2011b], using HF and HNO

3

. Final measurements were made on indium-spiked solutions with an approximate dilution factor of 10

4

, on a Thermo-Finnegan Element 2 ICP-MS. Measurements were calibrated using a range of synthetic standards, corrected against digestion blanks, and calibrated for signal drift using the USGS basalt BHVO-4. A few elements were rejected from the final results on the basis of poor duplication on repeat measurements or poor recovery against reference standards.

Free olivine crystals from selected samples were picked from 250-μm to 1-mm sieved fractions, for the analysis of olivine-hosted melt inclusions. These crystal grains were mounted in epoxy resin and polished with three grades of silicon carbide paper, followed by diamond polishing to a 1 μm grade. The polished and mounted grains were then examined under an optical microscope to select melt inclusions suitable for micro-analysis. Inclusions that had irregular edges, were microcrystalline, which appeared to be potentially linked to the crystal exterior via embayments below the polished surface, or which were elongate towards the crystal edges or linked to cracks in the host crystal, were all avoided. For each eruptive unit, several melt inclusions, in several separate olivine grains, were thus chosen for further analysis.

Selected trace element, CO

2

and H

2

O compositions of melt inclusions were measured of gold coated samples by ion microprobe, using secondary ion mass spectrometry (Camecaims-4f at the University of Edinburgh; Hinton [1990]). All analyses used an O

-

ion source (15 kV) and primary beam current of 5 nA, with a beam spot of <20 μm, on spot sites away from the edges of melt inclusions. Sites were pre-cleaned using a rastered 50 μm beam for 2

173

174

175

176

177

178

179

167

168

169

170

171

172

180

181

182

183

184

185

186

187

188

189

190

191

157

158

159

160

161

162

163

164

165

166

149

150

151

152

153

154

155

156

192

193

194

195

196

197

198 minutes before each analysis. Anomalous compositions, resulting from crystal interference in inclusions that may have partially crystallised, were rejected.

1 H, 26 Mg, 30 Si, 47 Ti, 88 Sr, 89 Y, 90 Zr, 93 Nb, 138 Ba, 139 La and 140 Ce were analysed in ten cycle runs, with all ten used to calculate trace element concentrations, and the final five cycles for H. All counts were normalised to

30

Si.

12

C was measured on separate 20 cycle runs, using results from the final ten cycles, on the same spots, since a higher mass resolution was required in order to distinguish the peak from

24

Mg

2+

ions. Calibration was against basaltic and rhyolitic glass standards for C and H, and NIST SRM610 for trace elements, using the

Jochum et al.

[2011] reference values. We estimate an accuracy of 5–15% relative for all elements, and relative precision of better than 10% for all elements at the observed concentrations.

The same inclusions were later analysed for major element glass compositions (using these to correct ion microprobe trace element analyses against the measured silica content), as well as adjacent host olivine compositions, by electron microprobe (JEOL8600 at the

University of Oxford). Samples were cleaned and carbon coated prior to analysis, and analysed with an accelerating voltage of 15 kV, a 6 nA beam current and a 10 μm spot size.

Analyses of basaltic glass secondary standards indicated accuracy within 2% relative for all major element oxides. Glass major element oxide totals were all >95% (generally 97-98%), except for the most water-rich Ap1 melt inclusions (>92 wt%).

A further examination of our ion-probe data was made by comparing the trace element concentrations of our post-entrapment corrected, undegassed olivine-hosted melt inclusions with whole rock trace element data, obtained by solution ICP-MS and XRF [ Watt et al.

,

2011b, and authors’ data] (Table S5). We would not expect an exact match between these datasets, because compositions preserved within melt inclusions, postulated to be trapped during the earliest stages of crystallisation, are likely to differ from the whole rock, which may have been affected by later modifying processes. Nevertheless, given the primitive nature of these rocks, we would not expect extensive shallow-level assimilation or mixing, and thus expect the two datasets to broadly correspond. For example, systematic differences between eruptive units should be observable within both the whole-rock and melt-inclusion datasets, even if absolute values have been affected by greater degrees of crystal fractionation in the whole rock data. The ICP-MS data show broadly matching incompatible trace-element patterns for each eruptive unit, but clear differences between units. For the elements analysed within olivine-hosted melt inclusions, the mean melt-inclusion concentrations are systematically slightly lower in the melt inclusion data versus the whole rock data (Table S5), for all trace elements except Nb. The deviation is generally larger for the more evolved units

(Hornopirén and South Minchinmávida), as expected, because these units are likely to have undergone more extensive fractionation following entrapment of the olivine-hosted melt. We note that the measured Ti contents are consistent for the melt inclusions in both ion probe and electron probe data (Table S2). We also note that, in spite of variation in absolute values, systematic differences between eruptive units, such as the order of increasing Nb/Y values from Ap1 to South Minchinmávida, is recorded in both olivine-hosted melt inclusions and the

ICP-MS whole rock data.

Corrections, calculations and modeling

Post-entrapment crystallisation and melt inclusion re-equilibration

All measured compositions were corrected for post-entrapment crystallisation of olivine on the inclusion walls, by determining the olivine composition that the measured inclusion is in equilibrium with, and then incrementally adding increasingly forsteritic olivine until the revised melt inclusion composition is in equilibrium with the measured phenocryst core composition. We applied such a correction using the PETROLOG software [ Danyushevsky ,

207

208

209

210

211

212

213

214

215

216

199

200

201

202

203

204

205

206

223

224

225

226

227

228

229

217

218

219

220

221

222

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

2011]. The correction is based on glass major element compositions (including H

2

O), with equilibrium calculations applied at a nominal 1 kbar pressure and oxygen fugacity set at the

QFM buffer, using the olivine-melt equilibrium model of Ford et al.

[1983], with trace element measurements adjusted proportionally. The amount of olivine added was notably lower for CO

2

-rich samples. This suggests that post-entrapment volatile loss is associated with higher levels of crystallisation; there is a positive correlation between measured H

2

O and CO

2

contents and the forsterite percentage that the original glass analyses are in equilibrium with. We therefore reject samples which we assume to have undergone extensive degassing, which particularly reduced the size of our Ap1 data set. We set this degassing criterion conservatively, rejecting any inclusions with <120 ppm CO

2

, along with four Ap1 samples that showed very low water contents. All of the rejected samples had required large post-entrapment corrections, up to 28% (Table S2). In contrast, the samples defined as undegassed required <9% olivine addition in all cases, except for one Ho1 a analysis (11%).

Additionally, the corrected, undegassed samples had tightly clustered compositions for each eruptive unit, whereas the degassed samples, requiring large corrections, showed wide scatter

(Supplementary Figure 2), which we suggest results from errors propagating through an excessively large correction. One outlier remains in the Ap1 samples, at anomalously low

MgO (8.4 wt%, hosted in an Fo

89.3

olivine). The reasons for this are unclear, but the anomalous bulk composition also results in low equilibration pressure and temperature estimates for this outlier (Figure 3b).

Associated with post-entrapment crystallisation, re-equilibration of the melt inclusion with its mineral host may occur, resulting in diffusion of Fe out of the inclusion

[ Danyushevsky et al.

, 2000]. This loss of Fe is manifested as lower FeO* values in the melt inclusion relative to the whole rock trend, and it may be particularly significant in highly forsteritic inclusions. However, the degree of re-equilibration is dependent on the degree of pre-eruptive cooling of the magma following melt-inclusion entrapment [ Danyushevsky et al.

,

2000]. Thus, large time periods between entrapment and eruption, and related magmatic evolution and cooling, may lead to significant Fe-loss in melt inclusions. We have assessed the degree of Fe-loss in all our samples by comparing melt inclusion FeO* and host-olivine forsterite content with bulk-rock FeO* contents, following the method of Danyushevsky et al.

[2000, 2002]. This assessment illustrates full overlap of our melt inclusion compositions with the whole rock trends. Although some melt inclusions extend to slightly lower FeO* compositions, this is matched by scatter above the whole rock trend to higher FeO* compositions, with the whole scatter approximately ±1 wt% FeO* relative to whole rock values. The coincidence of the two trends suggests that Fe-loss has not significantly affected our samples, and we therefore make no correction for melt-inclusion re-equilibration. The absence of Fe-loss may be related to the nature of our samples, which were erupted explosively and may have risen rapidly from depth, with little time between phenocryst growth (and melt inclusion entrapment) and eruption.

Parental magma compositions

In order to estimate compositions of parental magmas, defined as being in equilibrium with mantle olivine (Fo

90

), we add olivine incrementally to our (undegassed) melt inclusion compositions. The method again uses PETROLOG, calculated at 0.01% increments and 1 kbar, with the Ford et al.

[1983] olivine-melt model, and a QFM+1 buffer, selected as suitable for arc magmas [cf. Kelley and Cottrell , 2009]. Ap1 required only a small addition of olivine (0–2 wt%), as expected from its already highly forsteritic phenocrysts, in contrast with the other units, which required 10–16 wt% olivine addition. In this calculation of parental magma compositions, we assume that olivine is the only liquidus phase to have been significantly fractionated from our samples. To first assess this, we examined major element

257

258

259

260

261

262

263

264

265

266

249

250

251

252

253

254

255

256

273

274

275

276

277

278

279

267

268

269

270

271

272

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297 patterns in the samples from each unit. CaO increases with decreasing MgO, and variation in

Al and Mg also suggest that plagioclase or clinopyroxene have not begun to crystallise significantly. For this reason, we do not apply corrections along the plagioclase-olivine cotectic [e.g., Kelley et al.

, 2010], which would be suitable for dry (back-arc) melts, but are likely to be less suitable for these wet melts. Water is likely to suppress plagioclase crystallisation until melts have reached bulk MgO contents lower than those observed in our melt inclusions. For example, all the arc samples identified as having been affected by plagioclase crystallisation in Kelley et al.

[2010] have bulk MgO contents of <6 wt%. We tested the application of a correction along the plagioclase-olivine cotectic (to bulk MgO contents of 8 wt%, following Kelley et al.

[2010]) for the Ho1 samples, and found it to produce unrealistic compositions after this first correction stage, with melts in equilibrium with Fo

>90

. We consider the assumption of olivine-only fractionation to therefore be supported by petrological and chemical observations in each unit, and to produce reasonable results. However, it remains likely that small amounts of clinopyroxene or Fe-oxide fractionation have affected our magmas, particularly those hosting less forsteritic olivines

(e.g. Ho1b), without being clearly evident in major element trends. Failing to account for such fractionation would lead to slightly elevated SiO

2

concentrations in our calculated parental magmas for these less primitive units, and this would consequently feed through our calculations to give lower mantle equilibration pressures. On the latter point, we note that our most primitive unit, Ap1, actually gives the lowest equilibration pressures (see below) of our samples.

The maximum amount of olivine addition required, of 16 wt%, compares to an average of 22 wt% for the Marianas samples in Kelley et al.

[2010]. We also note that the majorelement parental magma compositions of our relatively drier units are all closely similar, as expected, in contrast to the wetter Ap1, in agreement with predictions of the effect of water on the MgO and silica content of primary lherzolite melts [ Wood and Turner , 2009].

Melt fraction and source water contents

Following Kelley et al.

[2006, 2010], TiO

2

can be used as a proxy for melt fraction, since its mass flux from the slab is likely to be relatively minor. This approach assumes TiO

2 content is simply a function of melt fraction and of the Ti-content of the mantle source. Trace element patterns suggest that our mantle source is slightly enriched relative to an NMORB source, with slight relative depletion of the source beneath Ap1 compared to the other units.

We use the Kelley et al.

[2006] batch melting relationship and our calculated primary melt composition to calculate melt fraction, and then use similar batch melting relationships to calculate source water contents. To estimate the source Ti concentration we use the Kelley et al.

[2006] approach of using Ti/Y systematics. We also repeat the batch melting concentrations with an assumed constant TiO

2

concentration of 0.133 wt% (MORB-source value of Salters and Stracke [2004]), but note that this is relatively low, given that trace element signatures suggest a slightly enriched source, and that the TiO

2

contents estimated from Ti/Y relationships are higher (0.162-0.196 wt%). Given that our trace element data suggest slightly variable source fertility, and overall enrichment relative to NMORB, we prefer the approach of using Ti/Y systematics, although note that this provides maximum likely estimates of melt fraction and source water contents (particularly since Ti may not be entirely immobile, and our assumption of immobility may slightly over-estimate source Ti content). We therefore suggest that the most likely values of melt fraction and source water content lie somewhere between our two sets of estimates, which are based on high- and lowendmember source Ti contents. In any case, the relative variation between the different eruptive units is preserved in both approaches, and it is simply the absolute values that vary

306

307

308

309

310

311

312

313

314

315

298

299

300

301

302

303

304

305

322

323

324

325

326

327

328

316

317

318

319

320

321

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

(Table S3). The Ti/Y relationships suggest that the greatest mantle enrichment is for the

Palena and South Minchinmávida sources.

Models of melting relationships

To estimate the pressures and temperatures of parental magma equilibration, we use the thermobarometer of Lee et al.

[2009], based on Si and Mg contents. The results of this thermobarometer are not open to simple interpretation, since they reflect an approximate mean equilibration condition for a potential mixture of multiple pooled melts, forming across a range of conditions. They are nevertheless useful at providing insights into relative differences in melt generation across the arc. To make these calculations we manually apply the method of Lee et al.

[2009]. Results correspond closely with those obtained by using the spreadsheet macro provided by Lee et al.

[2009]. Calculated values fall along clear, approximately solidus-parallel lines for each volcano, with a distinct lower temperature offset for Ap1. In general, all of the pressures indicated are relatively low (see, for example, the higher pressures for the Marianas arc in Kelley et al.

[2010]), and suggest pooling of melts towards the base of the crust. There is not a strong relationship between source water content and pressure. The solidus relevant to each melt differs, and the plotted values cannot be interpreted in light of a single solidus. The melting function will vary due to source fertility

(not likely to be a major effect here) and due to variable water contents. All the data plot above the dry solidus, in contrast to what might initially be expected for arc rocks.

We follow the method of Kelley et al.

[2010] to further interpret melting relationships, and to provide a dataset that can be compared to their results from the Marianas. We use their hydrous melting parameterisation, which predicts mantle melt fraction against source water content for different PT conditions. We use mean PT values, representative of each eruptive unit, obtained from the Lee et al.

[2009] thermobarometry, and assume a fertile mantle source. The results show the amount of melting expected over a range of water contents, and are generally consistent with the melt fraction and source water contents estimated for each sample based on Ti:Y (see section above), suggesting that wet melting of fertile mantle at the calculated PT conditions can explain the formation of, and differences between, our parental melts. For example, the wettest melt (Ap1) formed at lower PT conditions, and has the lowest melt productivity (a low melt fraction for a given water content), but because of its high water content its melt fraction is higher than the other units. Our results are consistent with a strong influence of water flux on melting.

The hydrous melting curves are slightly higher in their suggested productivity (the melt fraction produced at a given source water content) than the values from Ti:Y, but the variation between the different units is captured. Moreover, if an enriched mantle source was considered, as may be suitable for the units other than Ap1, the melting curves would shift to be closer to the data estimated from Ti:Y.

The results from thermobarometry suggest relative hot temperatures of equilibration, and a role for dry melting (i.e. a position above the dry mantle solidus) for units other than

Ap1. To test this interpretation with our batch melting relationships, and thereby check the consistency of our batch-melting and thermobarometric calculations, we use the melting model developed by Portnyagin et al. [2007]. By modelling the relationship between temperature, melt fraction and the parental-magma water content at a constant pressure, the curves illustrate the equilibrium melt temperature relative to the dry peridotite solidus. This confirms that Ap1 lies on the dry solidus (and therefore requires water-fluxed melting for its generation), in contrast to the other units. The deviation between the two sets of data derived using different source Ti-contents indicates the uncertainty within our calculations.

Nevertheless, the two are broadly consistent relative to the dry solidus.

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

356

357

358

359

360

361

362

363

364

365

348

349

350

351

352

353

354

355

372

373

374

375

376

377

378

379

366

367

368

369

370

371

397

Slab-surface temperatures

Slab-surface temperatures were estimated by Ce and K geothermometry [ Plank et al.

,

2009; Cooper et al.

, 2012]. Initial temperature estimates were made by simply applying the relationships in Plank et al.

[2009] to the relevant species concentrations in our undegassed melt inclusions (corrected for post-entrapment crystallisation). However, Cooper et al.

[2012] showed the need to correct this thermometer for both mantle contributions (Ce and H

2

O) and to variable pressure (the original thermometer is only valid at 4 GPa). To correct for mantle contributions, we follow the method of Cooper et al.

[2012], using the most water-rich melt inclusions for each unit, and using Nb/Ce ratios. The key assumption in this correction is the selected Nb/Ce ratio of the source. Given the uncertainties, we use a single NMORB source value for each unit [cf. Cooper et al.

, 2012], of 0.311 (note this is slightly enriched compared to the values in Salters and Stracke , [2004]), and an Nb/Ce ratio of the slab fluid of 0.04. This correction leads to an increased estimate of the slab fluid H

2

O/Ce ratio, and thereby reduces the estimated temperatures by between 20 °C (Ho1a) and 85 °C (South Minchinmávida), a significant reduction. If a slightly more enriched source was assumed, the change would be reduced slightly, but results would still be significantly lower than the original, uncorrected values. Following this mantle-source correction, we apply a pressure correction, again following the Cooper et al.

[2012] method, and using the slab surface depths provided in

Table 1. This correction increases the temperature estimate for Palena, and reduces it for Ho1 and Ap1, the latter by >100 °C. The resultant slab-surface temperature estimate for Ap1 is approximately 150 °C lower than that at Ho1. We consider such a large temperature change, over this relatively short distance, to be unrealistic, and suggest that the applied corrections are, at present, poorly parameterised. Nevertheless, the estimates fit well with estimates from thermal models [e.g., van Keken et al.

, 2011], and importantly, support a relatively low temperature for Apagado (below, or perhaps more realistically, close to the wet-sediment solidus) in comparison with all the other units, which exceed this solidus. For comparison with the results of the Ce thermometer, we apply a K/H

2

O thermometer, by simply using the equation in Plank et al.

[2009] (note the misplaced negative sign in the published equation.

The correct version is: K

2

O/H

2

O = 6×10

5

exp(0.0098T)). All temperature estimates are similar and the same overall pattern occurs. However, the K-thermometer estimate for Ap1

(797 °C) is substantially higher than that from the Ce thermometer (722 °C) (Table S3).

S

upplementary Figures

Supplementary Figure 1

Stratigraphic columns for the Ap1, Ho1 and Palena deposits, at representative proximal sites. Sample names and positions are indicated in italics, for all analysed samples (melt inclusions and bulk compositions). Site coordinates are given in Table S1. A column is not provided for South Minchinmávida, since this unit was sampled from a single exposure in the scoria cone itself, rather than from a locally dispersed fall deposit.

Supplementary Figure 2

Data plotted as in Figure 2a, but for a wider range of MgO contents, and discriminating between individual arc stratovolcanoes (Osorno to Maca) [ L ó pez-Escobar et al.

, 1992, 1993,

1995;

D’Orazio et al.

, 2003; Naranjo and Stern , 2004; Gutierrez et al.

, 2005; Watt , 2010;

Watt et al.

, 2011a]. Also shown are compositions of degassed melt inclusions, following correction for post-entrapment olivine crystallisation. Degassed inclusions required large corrections; note their wide scatter, in contrast to the clustered data of their undegassed counterparts.

S

upplementary Tables

409

410

411

412

413

414

415

416

417

418

419

420

421

398

399

400

401

402

403

404

405

406

407

408

429

430

431

432

433

434

435

436

437

438

422

423

424

425

426

427

428

439

440

441

442

443

444

445

446

Tables 1 to 5 are provided as separate PDF documents. In each table, Unit refers to a single scoria fall deposit (i.e. Ap1), while sample refers to scoria sampled from this deposit at a particular site. Selected scoria samples were then used for a variety of different analyses

(i.e. whole rock chemistry using XRF and ICP-MS; melt inclusion glass compositions on picked free olivine crystals).

Supplementary References

Cembrano, J., F. Hervé, and A. Lavenu (1996), The Liquiñe-Ofqui fault zone: a long-lived intra-arc fault system in southern Chile. Tectonophysics, 259, 55-66.

Cooper, L. B., D. M. Ruscitto, T. Plank, P.J. Wallace, E. M. Syracuse, and C. E. Manning

(2012), Global variations in H

2

O/Ce: 1. Slab surface temperatures beneath volcanic arcs, Geochem., Geophys., Geosyst., 13, Q03024.

Danyushevsky, L. V., F. N. Della-Pasqua, and S. Sokolov (2000), Re-equilibration of melt inclusions trapped by magnesian olivine phenocrysts from subduction-related magmas: petrological implications, Cont. Mineral. Petrol., 138, 68-83.

Danyushevsky, L. V., A. W. McNeill, and A. V. Sobolev (2002), Experimental and petrological studies of melt inclusions in phenocrysts from mantle-derived magmas: an overview of techniques, advantages and complications. Chem. Geol., 183, 5-24.

Danyushevsky, L. V., and P. Plechov (2011), Petrolog3: Integrated software for modeling crystallization processes, Geochem., Geophys., Geosyst., 12, Q07021, doi:10.1029/2011GC003516.

D'Orazio, M., F. Innoceti, P. Manetti, M. Tamponi, S. Tonarini, O. González Ferran, A.

Lahsen, and R. Omarini (2003), The Quaternary calc-alkaline volcanism of the

Patagonian Andes close to the Chile triple junction: geochemistry and petrogenesis of volcanic rocks from the Cay and Maca volcanoes (45° S, Chile), J. S. Am. Earth Sci.,

16, 219-242.

Ford, C. E., D. G. Russell, J. A. Craven, and M. R. Fisk (1983), Olivine-Liquid equilibria: temperature, pressure and composition dependence of the crystal/liquid cation partition coefficients for Mg, Fe

2+

, Ca and Mn, J. Petrol., 24, 256-266.

Gutiérrez, F., A. Gioncada, O. González Ferran, A. Lahsen, and R. Mazzuoli (2005), The

Hudson Volcano and surrounding monogenetic centres (Chilean Patagonia): An example of volcanism associated with ridge-trench collision environment, J. Volcanol.

Geotherm. Res., 145, 207-233.

Heusser, C. J., L. E. Heusser, and A. Hauser (1992), Paleoecology of late Quaternary deposits in Chiloé continental, Chile, Rev. Chil. Hist. Nat., 65, 235-245.

Hinton, R. W. (1990), Ion microprobe trace-element analysis of silicates: Measurement of multi-element glasses, Chem. Geol., 83, 11-25.

Jochum, K. P., U. Weis, B. Stoll, D. Kuzmin, Q. Yang, I. Raczek, D. E. Jacob, A. Stracke, K.

Birbaum, D. A. Frick, D. Günther, and J. Enzweiler (2011), Determination of reference values for NIST SRM 610-617 glasses following ISO guidelines, Geostand. Geoanal.

Res., 35, 397-429.

Kelley, K. A., and E. Cottrell (2009), Water and the oxidation state of subduction zone magmas, Science, 325, 605-607.

Kelley, K. A., T. Plank, T. L. Grove, E. M. Stolper, S. Newman, and E. Hauri (2006), Mantle melting as a function of water content beneath back-arc basins, J. Geophys. Res., 111,

B09208.

Kelley, K. A., T. Plank, S. Newman, E. M. Stolper, T. L. Grove, S. Parman, and E. H. Hauri

(2010), Mantle melting as a function of water content beneath the Mariana arc, J.

Petrol., 51, 1711-1738.

455

456

457

458

459

460

461

462

463

464

447

448

449

450

451

452

453

454

471

472

473

474

475

476

477

478

479

480

481

482

483

465

466

467

468

469

470

Lee, C. A., P. Luffi, T. Plank, H. Dalton, and W. P. Leeman (2009), Constraints on the depths and temperatures of basaltic magma generation on Earth and other terrestrial planets using new thermobarometers for mafic magmas, Earth Planet. Sci. Lett., 279, 20-33.

López-Escobar, L., M. A. Parada, H. Moreno, F. A. Frey, and R. L. Hickey-Vargas (1992), A contribution to the petrogenesis of Osorno and Calbuco volcanoes, Southern Andes

(41° 00'-41° 30'S): a comparative study, Rev. Geol. Chile, 19, 211-226.

López-Escobar, L., R. Kilian, P. D. Kempton, and M. Tagiri (1993), Petrography and geochemistry of Quaternary rocks from the Southern Volcanic Zone of the Andes between 41° 30'and 46° 00'S, Chile, Rev. Geol. Chile, 20, 33-55.

López-Escobar, L., M. A. Parada, R. Hickey-Vargas, F. A. Frey, P. D. Kempton, and H.

Moreno (1995), Calbuco Volcano and minor eruptive centers distributed along the

Liquiñe-Ofqui Fault Zone, Chile (41°-42° S): contrasting origin of andesitic and basaltic magma in the Southern Volcanic Zone of the Andes, Contrib. Mineral. Petrol.,

119, 345-361.

Naranjo, J. A., and C. R. Stern (2004), Holocene tephrochronology of the southernmost part

(42° 30'-45° S) of the Andean Southern Volcanic Zone, Rev. Geol. Chile, 31, 224-240.

Plank, T., L. B. Cooper, and C. E. Manning (2009), Emerging geothermometers for estimating slab surface temperatures, Nature Geosci., 2, 611-615.

Portnyagin, M., K. Hoernle, P. Plechov, N. Mironov, and S. Khubunaya (2007), Constraints on mantle melting and composition and nature of slab components in volcanic arcs from volatiles (H

2

O, S, Cl, F) and trace elements in melt inclusions from the

Kamchatka Arc, Earth Planet. Sci. Lett., 255, 53-69.

Salters, V. J. M., and A. Stracke (2004), Composition of the depleted mantle, Geochem.,

Geophys., Geosyst., 5, Q05B07. van Keken, P. E., B. R. Hacker, E. M. Syracuse, and G. A. Abers (2011), Subduction factory:

4. Depth-dependent flux of H

2

O from subducting slabs worldwide, J. Geophys. Res.,

116, B01401.

Watt, S. F. L. (2010), Records of volcanism and controls on volcanic processes in southern

Chile. D.Phil thesis, Department of Earth Sciences, University of Oxford, U.K., 367 pp., (http://ora.ouls.ox.ac.uk/).

Watt, S. F. L., D. M. Pyle, and T. A. Mather (2011a), Geology, petrology and geochemistry of the dome complex of Huequi volcano, southern Chile, Andean Geol., 38, 335-348.

Watt, S. F. L., D. M. Pyle, J. A. Naranjo, G. Rosqvist, M. Mella, T. A. Mather, and H.

Moreno (2011b), Holocene tephrochronology of the Hualaihue region (Andean southern volcanic zone, ~42° S), southern Chile, Quat. Int., 246, 324-343.

Wood, B. J., and S. P. Turner (2009), Origin of primitive high-Mg andesite: Constraints from natural examples and experiments, Earth Planet. Sci. Lett., 283, 59-66.

Download