Mid-Holocene change in types of degassing volcanoes, using indium in Antarctic ice as a tracer of volcanic source type
 Proportions of trace metals in Antarctic ice samples indicate that the type of volcanoes that dominated atmospheric emissions changed at about the middle of the Holocene from relatively mafic, deep source volcanoes to more silicic, shallower-source volcanoes. We base this inference on the strong contrast in the abundances of the trace metal indium (In), relative to other trace metals present in ice, deposited at different times in the past, and on contrasting In abundances in modern emissions of volcanoes of different types. Indium is more abundant in the emissions of deep-source mafic volcanoes than in more felsic, shallower-source volcanoes. Earlier workers have shown, on the basis of petrologic and some meteoritic evidence, that In may be partitioned to the interiors (stony mantles) of differentiated planets, or enriched in the liquids of partly crystallized mafic melts.
 Between climatic regimes, changes in load pressures on the Earth's surface from the accumulation and removal of ice sheets, and subsequent sea level and tectonic changes [Lowry, 2006, and references therein], may influence volcanic activity, including both energetic eruptions and continuous degassing at vents. Other workers have pointed to chemical changes in extrusive rocks or explosion frequency on local or regional scales following deglaciation (e.g., Iceland [Maclennan et al., 2002] and Mediterranean [McGuire et al., 1997]); to increased broad-scale explosion frequency in response to cooling events [Bay et al., 2006]; to varying response times of types of volcanism following changes in ice volume (e.g., California [Jellinek et al., 2004]); and to apparent peaks in volcanic activity in interglacial times (California [Glazner et al., 1999]). Evolution of magmas as they ascend through shallow depths has been shown to occur over times of only a few thousand years [Turner et al., 2000], and cosmogenic isotope and sediment accumulation studies at subduction margins may apply to times scales of glacial cycles [Morris et al., 2002]. But there has previously been no evidence for broad-scale compositional change in volcanism in response to major changes in ice loading or sea level change.
 Content of the rare, volatile chalcophile trace metal indium (In) varies in emissions of volcanoes of different type, being relatively abundant in emissions of basaltic hot-spot volcanoes, and lower from more silicic, shallow source volcanoes [Hinkley et al., 1994]. There is meteoritic evidence that In is partitioned to the interiors (stony mantles) of differentiated planets, and depleted in their crusts [Onuma et al., 1968; Yi et al., 2000], and therefore this observed difference between volcano types is expected.
 In ice from the Antarctic Taylor Dome site, masses of a suite of volatile trace metals are best accounted for by emission to the atmosphere of quiescently degassing volcanoes, not by mineral dust or other sources [Matsumoto and Hinkley, 2001]. The proportions of indium in the suite of Taylor Dome ice samples change distinctively in the middle Holocene, indicating a change to more silicic volcanism.
 Metals in the Taylor Dome Antarctic ice samples were measured by isotope dilution followed by thermal ionization mass spectrometry (ID-TIMS). Ice cores were prepared, and the samples treated, according to contamination-controlled methods described by Matsumoto and Hinkley . Samples containing In (commonly 10–100 × 10−9 g of In) were collected as metal-bearing particles by filtering the physical mixtures that make up the plumes of degassing volcanoes [Hinkley et al., 1994, and references therein). Some of the volcano emission samples were measured by the isotope dilution method, some by inductively-coupled plasma mass spectrometry (ICP-MS). Indium is present in ice at concentrations that are very low, both absolutely (typically 10–100 × 10−15 g In/g ice) and in relation to other rare trace metals. It is doubtful that the low concentrations of indium in ice samples could have been reliably measured by other presently-available methods. Due to its efficient ionization from a filament, it can be measured with superior precision and accuracy by isotope dilution and mass spectrometry. Analysis of In is relatively free from problems of contamination. In Tables 1 and 2 all available analytical data are presented (not selected from any larger body of data).
Table 1. Proportions of Pb, Cd, and In in Quiescent Emission Plumes of a Deep-Source Basaltic (Kilauea) and a Shallower-Source Andesitic Volcano (Etna)a
| ||“skylight”b|| || || |
| || ||10||0.24||0.86|
| || ||10||0.12||0.71|
| ||Bocca Nuova||10||n.d.||0.003|
|(all 1988)c||Bocca Nuova||10||n.d.||0.018|
| ||S.E. Crater||10||n.d.||0.21|
| ||diabase W-1e||10||0.19||0.1|
Table 2. Proportions of Trace Metals in Antarctic Ice From Taylor Dome, From Different Time Periods During Holocene and Last Glaciala
|Four Later Holocene Samples|
|Three Earlier Holocene Samples|
|Two Samples From the Last Glacial Period|
 Suites of ordinarily-rare, low melting- and boiling-point metals (e.g., Pb, Cd, In, Tl, Au) preferentially cross the melt-vapor (i.e., lava-air) interface in quiescently degassing volcanic systems, to such an extent that their concentrations in the plumes are raised to the status of major constituents. Major constituents of the lava (e.g., Si, Al, Fe, Mg) are left behind. Particles in the plumes that bear the rare metals may be widely transported as components of the atmospheric load [Hinkley et al., 1994, 1999, and references therein]. There is also evidence that in explosive eruptions the same rare metals may be more distantly transported in the atmosphere than grains of tephra that have the low trace element concentrations of ordinary volcanic rock (J. McConnell, personal communication). However, although volcanoes of andesitic composition (more silicic than basaltic volcanoes) can and do explode, it is probable that quiescent (non-explosive) degassing of andesitic volcanoes accounts for the majority of their trace metal emissions to the atmosphere [Nho et al., 1996], and therefore most of the emissions of both andesitic and basaltic volcanoes are into the troposphere, having the same starting point for atmospheric transport and deposition. However, we acknowledge that atmospheric patterns may have differed, and that efficiency or geographic representation of transport of volcanic products to Antarctica may be biased by time period, to some unknown degree. In Antarctica, especially at low-dust times, volcanic emission of these rare trace metals is commonly the dominant source of their total measured concentrations in the ice, dwarfing the contributions from crystals of minerals in dust (or from ocean solute). This has been shown to be true at the Taylor Dome site in Antarctica for metals including In, for ice deposited over at least the last 25,000 years [Matsumoto and Hinkley, 2001]. A dominant volcanic source for trace metals is also apparent in Antarctica at the Law Dome and Coats Land sites [Vallelonga et al., 2002, 2005; Planchon et al., 2003] for some (pre-industrial) time periods. Most samples analyzed from those sites are younger than the Taylor Dome samples.
 The trace metal indium has somewhat complicated and ambiguous geochemistry in the Earth and in meteorites. In different cases it apparently displays chalcophile or lithophile behavior, and its volatility appears to have varying importance under different circumstances. Petrologic evidence suggests that it may be partitioned into the stony mantles of differentiated planets [Yi et al., 2000], and that it may be partitioned into the liquid fractions of partially crystallized basaltic magmas (due to exclusion of In from olivine, and to a neutral (unit) partition with respect to pyroxene [Matsui et al., 1977; Onuma et al., 1968]). A review of equilibrium constants [Wood and Samson, 2006] and other parameters for various species of In (and other trace metals) in aqueous systems constrains its behavior, but firm data for high temperature conditions are few. Details of proportions of various trace metals are different in the plumes of volcanoes of different types: Table 1 shows that indium is consistently much more abundant in present-day plume samples from the deep-source basaltic volcano Kilauea than in the more silicic, shallower-source volcano Etna.
 The differences in efficiency of transport of the different members of the metals suite through the water cycle of the atmospheric are not quantitatively known, and fractionation may occur between emission and deposition [Banic et al., 1992]. Exact proportions of metals in volcano plumes at emission sites may be substantially different than at sites of deposition. However, Table 2 shows that the abundance of In (normalized to lead, Pb) is much lower in the four samples of Taylor Dome ice representing times at or later than about the middle of the Holocene (i.e., 7000–1,300 y.b.p.; In/Pb ∼0.02–0.06) than in the three samples representing the early Holocene (10,000, 10,400, and 13,400 y.b.p; In/Pb ∼0.2–0.3). According to calculations of Matsumoto and Hinkley , more than 95 percent of the Pb and In in these seven ice samples are attributable to volcanic emissions (except for Pb in the 10,000 y.b.p. sample, in which only about 85 percent was attributable to the volcanic source, and 15 percent to dust minerals). The strong change in relative abundance of In through time in the ice record indicates that, during these two time periods there is a difference in the average compositions of the emissions of quiescent volcanoes that are injected into and transported by the atmosphere: larger amounts of In consistent with deeper source, relatively ferromagnesian volcanoes are seen in the earlier period, and smaller amounts of In from more silicic, shallower source volcanoes in the later period. Also, both of the two ice samples from the earlier, full glacial period (27,200 and 72,900 y.b.p.; In/Pb ∼0.1) have proportions of Pb and In that are intermediate between the two groups of samples from the early and late Holocene, respectively.
 There may be three distinct modes of volcanism dominant during the three time periods: (1) mixed types of volcanoes active during the full glacial time, when some regions of land were heavily loaded with and depressed by ice sheets, and sea level was low; (2) relatively ferromagnesian, deep-source volcanism dominant for several millenia after the rapid deglacierization and rapid sea level rise with which the Holocene began; and (3) more silicic, shallower-source volcanism dominant during the subsequent millenia, after a certain amount of crustal adjustment to the changed conditions.
 Measurements of the concentration of indium in Antarctic ice samples have been made only at the Taylor Dome site. At that site the climate and atmospheric depositional regime are intermediate between coastal and continental. At some other Antarctic sites (coastal) the contents in ice of suites of rare, volatile trace metals are dominated by volcano emissions rather than by dusts (data of Vallelonga et al.  and of Planchon et al. ); also, calculations indicate that volcano emissions are important at low-dust times at the interior site Vostok (data of Hong et al. ), and Pb isotopic proportions indicate that the same is true at the interior Dome C site (volcanic-type radiogenic Pb in low-dust times, continental dust-type, less-radiogenic Pb in dusty times, data of Vallelonga et al. ). As to the question of whether indium and other volatile trace metals in the ice in Antarctica originate specifically from Antarctic volcanoes, or whether they represent an integration of volcanoes from a wider area (viz. ocean island and other volcanoes over the southern hemisphere), Matsumoto and Hinkley  state that the proportions and amounts of trace metals, as well as the isotopic proportions of lead (Pb), correspond well between the deposition rates of metals to the ice and the estimated worldwide source strength of degassing volcanoes.
 Values for the trace elements Cd and Tl are available for only some samples (no Cd data for Mt. Etna emissions, no Tl for either Kilauea or Etna), and the patterns of concentration for these metals appear to be less distinctive than for In. However, Pb and Cd emission data for the Indonesian arc volcanoes (andesitic; more silicic than Kilauea and similar to Etna) and for Kilauea by Nho et al.  and Hinkley et al.  show that the Pb:Cd ratios are about 10 : 3 for Kilauea and 10 : 0.7 for Indonesia, corresponding with the contrast in the ratios in the early and late Holocene ice, respectively. This supports the indication from the indium data of this present paper that emissions from basaltic volcanoes dominated the early Holocene, and more silicic volcanoes dominated the late Holocene.
 To extend the preliminary findings of the present work, plume particles from additional quiescently degassing volcanoes, and ice cores from varied locations and ages, should be sampled and analyzed. Also, lava production volumes at representative basaltic volcanoes, and time and mass frequency of andesitic tephras in ice, should be tabulated to assess overall magnitude of volcanic activity by time period.
4. Summary and Conclusions
 Several studies have proposed links between glacial and sea level cycles, and intensities of volcanic activity. The present study indicates that the chemical type of volcanism, or at least of volcanic emissions to the atmosphere, changed from relatively ferromagnesian to relatively silicic, about the middle of the Holocene, after the Earth's crust had had several thousand years to adjust to the rapid changes in ice cover and in sea level that occurred at the end of the last ice age. The data also indicate that volcanism during the last ice age was of a type intermediate between the types dominant during the early and late parts of the Holocene. The basis of the indications is that different amounts of the trace metal indium are preserved in strata of the Antarctic ice sheet from these different times, and that different amounts of indium are present in the atmospheric emissions of modern volcanoes of different chemical types, or that tap different depths within the Earth.
 Thanks to K. Misawa, E. Castellano, and R. Bay for discussions and to J. S. Pallister, G. P. Meeker, and an anonymous reviewer for comments that improved the ms.