0016-7037/94/S6.00
Geochimica d Cosmochimica Acta Vol. 58, pp. 75-83 Copyright 0 1994 Ekvier Science Ltd. Printed in USA
+ .oO
Severe leaching of trachytic glass without devitrification, Terceira, Azores JAMESE. MUNGALLand ROBERTF. MARTIN Department of Earth and Planetary Sciences, McGill University, 3450 University St., Montreal, Quebec H3A 2A7, Canada (Received December 23, 1992; accepted in
revisedform July 8, 1993)
Abstract-Pumice produced during the subplinian phase of eruption of a peralkaline trachyte lava on the island of Terceira, Azores, has undergone extensive leaching at ambient conditions by groundwater, while remaining in a glassy state. Over 25% of F, Na, K, Si, Fe, Ti, and Mn originally present has been removed, whereas U, Al, Nb, Ca, Y, and Rb show smaller but significant losses. Thorium, Zr, Hf, Ta, and Lu have remained immobile. Water content (measured as LOI) has increased up to tenfold, whereas the Sr content has risen by up to 80%. The absence of devitrification requires that leached elements escaped by diffusing through the glass; a coefficient of chemical diffusion L&. between 1.8 X lo-l9 cm’ S -’ and 6.1 X lo-l9 cm2 s-’ has been calculated, in excellent agreement with previously published experimentally derived values for comparably low temperatures. Water has not replaced the leached cations on a charge-equivalent basis, a finding that requires an explanation other than simple binary interdiffusion. We suggest that a simultaneous influx of molecular water and cation exchange between Na+ in the glass and H+ in the groundwater opens the glass structure and frees cations, which can ditfuse outward by hydrolysis of bridging oxygen atoms. Reported mobilities have important implications for such diverse fields as igneous petrology, nuclear waste management, and agriculture. INTRODLJCI-ION
and higher, alkalis and alkaline earths were selectively leached from basaltic glass undergoing hydrothermal alteration. FYFE et al. ( 1978) found that K was leached, whereas P was conserved in rhyolitic and basaltic glasses to which several wt% K3P04 had been added. Hydration experiments with synthetic glasses in simple systems have demonstrated the high mobility of Na within an essentially immobile framework of Si and Al tetrahedra (SCHOLZE, 1988; HOUSER et al., 1980; LANFORD et al., 1979; DOREMUS, 1975; BUNKER et al., 1983; SMETSand LOMMEN, 1983; DUNKEN and DOREMUS, 1987; DRAN et al., 1988). VERNAZ and DUSSOSSOY(1992) reviewed the mechanisms of alteration of synthetic nuclearwaste glass in some detail, concluding that an initial phase of hydration is succeeded by alkali-water exchange and eventual “congruent hydration” of all cations in the glass. The resulting gel is then thought to evolve by diffusion of mobile species within a fixed framework. Their results indicate complete immobility for actinides, REEs, Fe, Zr, and Al. This contribution reports on chemical changes in pumice inferred to have been leached by groundwater at ambient conditions. We use the approach of GRESENS( 1967) to estimate the bulk mobility of a wide range of elements not normally considered in similar studies. The advantage of such an approach is that it permits precise quantitative analysis of samples for minor and trace elements not accessible to conventional depth-profiling microanalytical methods; the disadvantage is that we cannot detect chemical zonation within the sample, but report instead a net chemical change in the bulk sample.
LOW-TEMPERATURE INTERACTIONS between water and glass, leading to its leaching and hydration, are of scientific interest in several fields. For example, such interactions bear on the routine dating of obsidian by anthropologists ( MICHELSet al., 1983); on disposal of nuclear waste (PETIT, 1992); on the selection of glass compositions for industrial, medical, and household use (SCHOLZE, 1988); on an understanding of chemical mass-balances that exist between volcanic glass and either seawater (STAUDIGEL and HART, 1983) or groundwater (WHITE et al., 1980); and on the development of single-application glass-based fertilizers for agriculture ( FYFE et al., 1978). We report here an extreme example of selective leaching of elements during low-temperature (“secondary”) hydration of silicic pumiceous glass, which forms an extensive tephra deposit on the western flanks of Santa Barbara volcano on Terceira island, in the Azores. Previous studies of natural glasses that have undergone hydration without devitrification have documented more limited compositional variations (on the order of several percent) accompanying hydration and have generally reported selective mobility of a limited number of elements. For example, MAGONTHIERet al. ( 1992) reported losses of alkalis and silica but stressed the immobility of Fe, Ti, and Al. WHITE ( 1983) reported selective partial removal of Na, K, and Al into solution in experiments at 25’C on felsic glasses but suggested that Si release into solution results from congruent dissolution. TSONG et al. ( 1978) showed that natural hydration of obsidian is accompanied by significant losses of Na, K, Ca, Mg, and Li, whereas Si and Al are conserved. ZIELINSKI et al. ( 1977) and LIPMANet al. ( 1969) documented limited preferential leaching of alkalis and alkaline earths from calcalkaline obsidians during perlite formation. NOBLE ( 1967) and JEZEK and NOBLE( 1978) reported losses of Na, K, and Si with hydration. BERGERet al. (1988) found that at 50°C
OCCURRENCE OF THE SILICIC GLASS The glassy ejecta described here were produced in an eruption estimated to have occurred between 2000 and 600 years BP, from the Pica do Camheiros vent on the western flank of Santa Barbara volcano ( SELF, 1976 ) . An initial subplinian 15
76
J. E. Mungall and R. F. Martin
phase generated an estimated 0.09 km3 (0.02 km3 dense-rock equivalent) of pumiceous material dispersed over 80 km2, following which 0.3 km3 of peralkaline trachyte was extruded as a series of coalescing domes and flows covering 4.1 km*. The trachyte flows consist of a felsite interior and a glassy exterior and are surrounded by vent-area aprons of obsidian clasts and pumice. Phenocrysts of anorthoclase, sodian augite, and magnetite are common, whereas fayalite and apatite are minor constituents. The assemblage and compositions of phenocrysts are the same in both the subplinian pumice-fall deposit and the lava, allowing identification of the pumice at considerable distances from the vent. In the flow rocks, a
complete range of groundmass textures exists, from holohyaline obsidian to holocrystalline felsite. We restricted our sampling to glassy rocks because we suspected felsite to have lost alkalis upon crystallization or devi~fi~tion (cf. WEAVER et al., 1990). We collected pumice samples about I km from the vent, where the deposit is about 1 m thick (SELF, 1976); its top lies less than 1 m below the soil surface. The sampling location, situated on the western side of the island, is subjected to heavy rainfall over most of the year; the presence of lush vegetation rooted in a well-develop soil immediately above the sampled pumice suggests that the sampled pumice layer remains partially saturated with groundwater all year. A few
FIG. 1. Electron micrographs of altered pumice. (a) Backscatteredelectron image of polished thin section of pumice demonstrates extremely high porosity and specific area of glass. Septa range from ~0.5 to >20 pm thick, with majority of material residing in 1-4 em thick bubble walls. There is no evidence for compositional zoning inward from giassair interfaces; such zoning should appear as variations in intensity of the backscattered electrons as a function of the average atomic number of the material. (b) Tensile cracks in thin bubble wall, consistent with volume reduction during leaching. This texture appears in contrast to the very finely pointed te~inations of incomplete septa in la that ap~n~y resulted from arrested coalescence of bubbles during cooling. (c) Mildly weathered surface adjacent to large fracture in pumice fragment showing incipient development of etched surfaces and possible amorphous material or clay mineral. Such weathered surfaces make up a very small proportion of the exposed surface area of the glass. (d) Closeup of etched surface in area free of weathering overgrowths; EDS spectra of rough areas are indistinguishable from those of adjacent smooth areas, indicating that roughness is produced by small degrees of bulk dissolution, not by deposition of secondary minerals. Uniformity of the EDS spectra extends to freshly broken surfaces of bubble walls, suggesting a minimal amount of chemical zoning between the inner &ass and its altered surface. The usefulness of this observation is tempered by the fact that the activation volume of the EDS analysis probably extends at least 2 pm into the sample.
Severe leaching of trachytic glass
roots penetrate the sampled layer, but it remains largely free of decomposed organic material; external surfaces of pumice clasts are visibly weathered to clay, but fresh internal surfaces are firm and glassy. The pumice appears homogeneous in examinations using both optical and electron microscopy (Fig. 1); no evidence for chemical zonation is apparent in backscattered electron images. All samples of pumice remain in a glassy state, as evidenced by isotropic character in transmitted light and the absence of peaks in X-ray diffmctograms other than those attributable to observed phenocryst phases (cf. PALMER et al., 1988). COMPOSITIONAL
DATA
Sample compositions, two of vent-area obsidian clasts, three of glassy trachyte from the flow-top, and three of pumice from the airfall deposit, all produced by the Pica do Camheiros eruption, are given in Table 1. Samples were crushed in an iron jaw crusher and tungsten carbide puck mill. SiOz, Ti02 , A1203, Fe20:, MnO, CaO, Na20, KzO, Pz05, and Ba were measured by X-ray fluorescence (XRF)(BENNETand OLIVER,1992) on glass beads fluxed with lithium tetraborate; Nb, Zr, Y, Sr, and Rb were determined by XRF of pressed powder pellets at McGill University. Concentrations of the REEs, Ta, and Hf were determined at the University of Waterloo by Instrumental Neutron Activation Analysis (INAA) (GIBSONand JAGAM, 1980) of rock powders. Powders were activated for four hours in a flux of I X 10 I2 neutron cm-* s-’ at McMaster University and counted at Waterloo using a Li(Ge) detector with a resolution of 0.65 keV at 122 keV. Concentrations of U, Th, and F were determined
77
at Activation Laboratories Ltd., Ancaster, Ontario; F by specific ion electrode after digestion; and U and Th by INAA (see above). Precisions based on replicate analyses are reported where available in Table 1;the relative precisions of the REE, U, Th, Hf, and Ta results are based on counting statistics and are generally better than 1%. Density was measured on dried rock powders using a gas pycnometer under N2. Pumice samples have loss on ignition (LOI, which refers to weight loss upon ignition after drying at 100°C) values as high as 7.48 wt%, in marked contrast to the nearly volatile-free nature of the co-erupted flow-rocks. The agpaiticindex [(Na + K)/AI (molar)] ofthelavas ranges from 1.026to 1.185,whereasthe samplesofpumiceare markedly peraluminous, despite the presence of phenocrysts of peralkaline clinopyroxene. These phenocrystscould not have grown in equilibrium with a melt having the composition of the pumiceous glass in whichtheyarefound. Allofthe dense lavas of Santa Barbara can be related by simple fractional crystallization of the phenocryst phases observed in them using least-squares mixing models (BRYAN et al., 1968). A plot of Nb vs. Zr ( ppm) (Fig. 2) shows a covariation that might be expected of incompatible elements in a magma undergoing crystal fractionation, with pumice samples appearing much more evolved than the samples of lava. However, a plot of total alkalis vs. silica (Fig. 3) shows a wholly contradictory result, with the pumice appearing to be less fractionated. Furthermore, a graph of Al (cation units) vs. Zr (Fig. 4) for all the glassy felsic extrusive rocks from Santa Barbara for which we have data shows that the trend relating the Pica do Camheiro pumice samples to their coeval lavas requires conservation of both Al and Zr, whereas the differentiation trend for the suite as a wholeshowsa gradual decline in Al,reflecting fractionation of alkalifeldspar,thedominantphenocrystphase.Conservativebeha of Al in a lava undergoing fractional crystallization of feldspar is impossible; attempts to derive the composition of pumice 89-39a
Table 1. Compositions of Pumice and Lava
Sample SiO, TiO,
A&O,
Fe,O,'MnO
89-39 61.80 0.31 89-42 59.04 0.32 89-43 62.75 0.34
16.77 4.66 17.74 5.00 16.81 4.84
89-72 89-73 89-74 89-75 91-43
14.69 14.76 14.69 14.54 14.07
Maior Elements -Na,O W PA
MgO
CaO
0 17 0 19 0 17
0.16 0.13 0.15
0 74 0 80 0 73
5.56 5.22 5.70
4.11 3.77 4.37
0.03 0.04 0.03
5.55 7.47 4.28
0 0 0 0 0
22 22 22 22 22
0.01 0.04 0.11 nd 0.09
0 0 0 0 0
6.56 6.31 6.55 6.59 6.81
4.34 4.40 4.53 4.42 4.49
0.03 0.03 0.03 0.03 0.03
Mean 67.42 0.40 14.55 5.42 0 22 C.V.% 0.139 0.511 0.292 0.093 1 86
0.05 2.09
0 69 6.56 0 244 1.07
4.44 0.12
0.03 2.08
67.49 67.52 67.60 67.85 66.64
0.04 0.04 0.04 0.04 0.39
5.38 5.48 5.40 5.50 5.35
Sample Ba
Nb
Zr
Y
89-39 89-42 89-43
343 330 313
205 225 222
1237 1348 1332
91 14 98 25 102 25
89-72 614 89-73 597 89-74 577 89-75 498 91-43 1290
183 185 184 189 177
984 984 999 1021 1077
715 C.V.0 1.67
172 0.73
1048 0.57
Mean
Sr
85 85 87 89 95 88 .57
Rb
69 69 68 71 66
Trace elements Th U Ce
107 18 98 20 115 21
7.0 7.8 8.1
234 -
11 11 9 10 10
101 101 102 104 104
12 15 12 15 15
5.6 6.7 6.4 5.9 6.0
209 211 -
10 .43
102 1.1
16 1.0
6.1 1.0
210 0.3
LO1
P.I.
S.G.
99.62 99.50 99.91
0.811 0.714 0.839
2.484 2.584 2.365
0.18 0.44 0.10 0.06 0.04
99.67 100.10 99.94 99.95 98.37
1.054 1.026 1.067 1.075 1.185
2.565 2.581 2.518 2.516
0.16
99.94
1.081 0.037
2.545
Eu
Total
Nd
Sm
105
20.4 2.21 2.95 9.13 1.52 14.1 27.9 _ _
_ _
Tb
Yb
Lu
Ta
Hf
_ _
95.4 96.0
19.1 19.7 __
3.56 3.47
2.78 2.82
8.38 8.38
1.27 1.28
11.6 11.8
22.7 23.0
95.7 0.6
19.4 0.14
3.52 0.26
2.80 1.36
8.38 0.54
1.28 0.51
11.7 1.15
22.8 0.33
Samples 89-39athrough 89-43 are pumice; 89-72 through 91-43 are dense lava. Mean refers to lavas Major elements reported as wt%, trace elements as ppm. Analytical methods listed in text. only. S.G. is density in g cm-'. C.V.% is the coefficient of variation (relative precision), equal to the standard deviation divided by the mean of ten replicate analyses, expressed as a percentage, except for REE, Hf, Ta where it is based on counting statistics.
J. E. Mungall and R. F. Martin
78
200
0
-
‘3”“““““’ 500
1000
1509
Zr @pm, dry) FIG. 2. Concentrations of Nb and Zr. All data recalculated on dry basis, in ppm. Solid squares represent pumice from Pica do Camheiros: open squares, dense lava from Pica do Camheiros; open triangles, other trachyte lava from Santa Barbara volcano. Pumice and lava are related by process that has conserved both Nb and Zr; if this process was magmatic fractionation, it would require that the pumice be much more evolved than the dense lava.
FIG. 4. Concentrations of Al and Zr. All data recalculated on dry basis, Al in cation units, Zr in ppm; symbols as in Fig. 2. Pumice is related to dense lavas by process that has conserved both Al and Zr, whereas differentiation trend for other Santa Barbara lavas reflects extensive fractionation of feldspar.
LEACHING from either the co-erupted trachyte, or from a more primitive trachyte that might be. considered a possible parent to both, using the leastsquares mixing models, were uniformly unsuccessful, resulting in unacceptably large residual errors and absurd proportions of the crystallizing phases. In particular, they failed to reproduce the observed variation in peralkalinity between the lavas and the pumice. This failure is demonstrated graphically in Fig. 5, a plot of total alkalis vs. Al (both as cation units). Although feldspar and clinopyroxene are both slightly peralkaline owing to NaFe substitutions involving Fe’+, they and all other phenocryst phases present in both the pumice and the dense lavas plot very close to the line Na + K = Al (all phenocrysts other than feldspar lie essentially at the origin of Fig. 5 ) Visual inspection of Fig. 5 shows that no combination of phenocrysts can be removed from the Pica do Camheiros lavas to produce the compositions of the pumice samples. Addition of very large quantities of the phenocrysts (corresponding to crystal accumulation) could produce the desired effect; however, the Nb vs. Zr diagram (Fig. 2) shows that the pumice cannot be derived from the lavas by a process of crystal accumulation, since accumulation would dilute these incompatible elements in the bulk composition and place the pumice closer to the origin in Fig. 2. These arguments work equally well in reverse; i.e., the pumice cannot be taken to represent a more primitive melt from which the lavas are derived by either fractional crystallization or accumulation. It is clear that whereas pre-eruptive differentiation may have had small effects on sample compositions, a major component of the variation must be ascribed to another process.
We contend that the evident deficiency of alkalis in the composition of the pumiceous glass compared with the peralkaline compositions of the lava samples can best be explained by leaching during secondary (low-temperature) hydration of glass, a process vastly enhanced in the pumice relative to the massive obsidian by the extremely high specific area and permeability of the pumice. It is very unlikely that high-temperature orthomagmatic or hydrothermal fluids could be responsible for the alteration, because in the numerous examples of such high-temperature alteration present on Terceira, the phenomenon has invariably led to very rapid devitrification and deposition of recognizable hydrothermal mineralization. There is, furthermore, no plausible source for such fluids in a one meter thick blanket of tephra deposited a kilometer from the active vent. If we assume that the composition of the pumice and that of the average lava were once identical, a view supported by
l-
Na+K=AI
.
AA
I’ ,’
A
A
. .
,’
a .
I.‘#
0.14
,
,
0.16
,
,
0.10
,
,
0.2
,
,
J
0.22
Al (cation units)
62
64
66 SQ
66
70
(dry)
FIG. 3. Concentrations of Alkalis vs. Silica. All data recalculated on dry basis, in wt%; symbols as in Fig. 2. Pumice appears much less affected by crystal fractionation than dense lava produced by same eruption. Compare with Fig. 2, where pumice appears more evolved.
FIG. 5. Concentration of (Na + K) vs. Al. All data as cation units. Dashed line (agpaitic index = 1.O) includes all phenocrysts found in both lavas and pumice; composition of feldspar is shown; those of others lie essentially at origin. Removal of no combination of phenocrysts can drive Pica do Camheiros lavas (open squares) across dashed line towards pumice (closed squares). Phenocryst accumulation has already been ruled out by Fig. 2 above; hence, no igneous process could have generated the pumice samples from the suite of lavas.
Severe leaching of trachytic glass the uniformity in composition of the lavas and the common population of phenocrysts among the samples, we can perform simple mass-balance calculations to determine the exact extent of the chemical modifications that have taken place. GRESENS ( 1967) developed equations relating volume changes and densities of fresh and altered rocks to the net gains and losses of individual elements. Although the approach of GRESENS ( 1967) is cumbersome and difficult to apply to large data-sets drawn from ancient altered rocks (cf. BARRETTand MACLEAN, 199 1)) it is ideally suited for situations in which the initial and final compositions are directly measurable and small numbers of samples are concerned. In the general equation,
79
3 2.5
*
2
,
9’.5-
I, 1
0.5
I ‘I’,
inferred
1
factor
I.
.!.I..!!
v0lume
I I I I I I I I I !..!..!..Y!.., ....*_-.. ._ 1’1, I .
(0.824)
’ I
Element
FIG. 6. Composition-volume relations, sample 89-39a. Volume factor intercepts of Gresens’s equation plotted for each element; bars indicate uncertainty (two standard deviations). Dashed lines (Fe*+, AX, is the net gain or loss of element n, A is the initial amount of material, Vr is the ratio between initial and final volumes of material, X $B are the amounts of element n in initial a! and final /3rocks, and pasBare densities of the initial and final rocks. Appropriate units of measure for X,, are moles of the element per liter of rock, with A being a standard volume such as one liter; this yields values of AX, in moles. In the absence of a precise knowledge of volume changes, we can estimate volume changes based on reasonable assumptions about element mobilities ( APPLEYARD and WOOLLEY, 1979). The mass-balance equation for each element is treated as one in the form y = mx + b where y = AX,, m = Vf multiplied by a measurable p8/p”, and b = X ,“. This treatment results in the generation of a family of linear equations that can be plotted in terms of volume change vs. composition ( Vf vs. AX,,) for each pair of initial and final compositions. The amount of each element added or subtracted is calculated from an appropriate value of Vf. This value of Vf is selected either by assuming one element to have been immobile, or by finding a group of lines that pass through the zero mass-change axis all at the same Vf, possibly indicating that all have remained immobile together. We have recalculated the mean composition of the average dense glassy lava and each of the three pumice samples on a molar basis, including moles of H. The values for H were estimated by assuming that the entire volatile component released upon ignition consisted of water, and further that oxygen was added during ignition such that all Fe’+ in the original sample was oxidized to Fe’+. This approach ignores the presence of halogens, known to be present at levels of hundreds of ppm, and incorrectly dismisses the persistence of small amounts of ferrous iron during ignition. We feel justified in making these approximations in light of the variation in H content attendant on changing atmospheric conditions, as documented by SCHATTNERet al. ( 1988) and indirectly observed by us as mass changes of individual samples from day to day. In light of this variability, we do not feel that attempts to measure water contents more accurately are warranted. The uncertainty in water content will affect densities and, consequently, values of Vr; however, the uncertainty in Vf in no way affects our assessment of the mobilities of the various elements other than hydrogen. Composition-volume changes for one sample of pumice are presented in Fig. 6. Elements whose Vf intercepts lie
P, Mg) indicate extremely large standard errors resulting from abundances near detection limits. Error bars were not calculated for 0 and H because both are inferred either from stoichiometry or LOI; they are included for completeness. Dashed horizontal line indicates a choice of volume factor at which no high-field-strength elements have been added, and two least mobile elements (Zr and Th) have
been conserved.
below the chosen volume-factor are inferred to have been added; those with higher intercepts have been preferentially removed. The value of AX,, at the true volume-factor (i.e., the real change in volume of the sample) is the actual change in abundance of an element n. It is immediately clear that there is a range in behaviors, from extreme enrichment to extreme depletion of elements. Not surprisingly, H is added in great abundance for any reasonable choice of Vf, elements such as Nb, Zr, and Al show zero AX, at intermediate values of Vf; and the cations Si, K, and Na, which are known to be mobile (see references above), show net losses at those values of V, that would indicate immobility of Nb, Zr, and Al. The choice of Vf for each pumice sample was dictated by the assumption that appreciable amounts of high-field-strength cations such as Zr and Nb are not present in groundwaters of meteoric origin, and therefore could not be added. Furthermore, these elements are difficult to remove in aqueous solution, and are likely to be conserved in the residue after leaching. Mg and P are both present near their detection limits and consequently have very large coefficients of variation; they are included only for completeness. The search for a value of Vf for which all the traditionally immobile elements are conserved within the 95% confidence interval (error bars), failed in all three cases. We have instead chosen the average between Zr and Th, considered the two least mobile elements, since they consistently overlap one another and show lower relative mobility than all the others. Net additions and subtractions of elements following this assumption are shown in Table 2 and Fig. 7. The estimated reductions in volume, although subject to large errors because of uncertainty in measured densities, are consistent with results of experimental leaching of K-rich glass (BUNKER et al., 1983 ) and with textures observed in pumice (Fig. 1). The pattern of addition and removal of cations from the glass is quite consistent from sample to sample. We can group the elements into four sets:
80
J.E.Mungalland R. F.Martin Table 2. Mass changes in altered pumice. 89-39a
89-42
89-43
(3.7) -37.089 -31.394 Si -28.047 (3.6) (2.1) (3.6) -42.239 Ti -38.860 (3.7) (2.6) -37.030 (4.7) -12.409 Al -9.527 (4.5) (2.8) -14.839 (4.5) -33.751 Fe -32.536 (4.4) (3.6) -34.336 (3.2) -37.956 Mn -39.344 (3.1) (2.0) -43.041 (252.0) 112.631 (274.0) (264.0) 79.602 Mg 141.526 -21.560 (6.5) -16.221 Ca -15.325 (6.2) (5.7) (6.4) -42.869 Na -33.510 (6.1) (5.9) -35.990 (6.1) -38.945 -27.385 K -27.273 (5.9) (5.4) P 2.045 (745.0) -4.212 (775.3) -26.288 (713.0) 894.877 - 1592.30 H 1273.810 F -26.981 (14.0) -41.525 (14:3) -35.716 (14.0) Ba -62.358 (20.0) -66.855 (20.6) -67.744 (19.0) (5.4) -12.807 Nb -12.478 (5.5) (4.1) -10.963 (6.7) -3.070 Zr -2.854 (6.7) (5.5) -1.319 (7.2) -20.177 Y -19.012 (7.1) (6.5) -14.755 Sr 7.739 (26.0) 76.080 (27.3) 80.667 (26.3) (5.5) -31.246 Rb -17.978 (5.5) (4.3) -17.218 Th 3.026 (10.0) 3.271 (9.6) (9.3) 1.354 (4.9) -8.438 u -10.217 (5.3) (3.2) -2.440 (4.3) Ce -12.534 (4.6) Nd -13.877 (4.5) Sm -17.458 (8.9) EU -50.717 Tb -17.299 (16.7) (6.7) Yb -14.479 Lu -6.053 (13.5) (7.7) Ta -5.403 (5.2) Hf -4.366 -24.237 -23.900 0 -18.239 0.854 0.697 V, 0.824
DISCUSSION
The main species affecting the rate of bulk diffusion of water through obsidian, if leaching of cations into aqueous solution is not involved, is molecular HzO. Hydrolysis of bridging oxygen atoms by molecular Hz0 produces OH groups whose diffusivity is orders of magnitude lower than that of the Hz0 itself ( ZHANG et al., 1991). Equilibrium
concentrations of OH groups in silicate glasses containing 2 and 10 wt% Hz0 are on the order of 0.5 wt% at 200°C ( STOLPER, 1989 ). With an assumption
-60 -lOO-
H
Sr P Hf Al Nb Nd Eu Yb Ca 0 F Si No Mn MS Th Zr Ta U Ce Sm Tb Lu Rb Y K Fe Ti Ea
8 * ’ ’
c 3 8 ’ ’ ’ c I ’ ’ t s
All changes given as percentages of amount originally present; figures in parentheses are Elements near two standard deviations. detection limits have very large standard errors and are included only for subjective comparison. Error analysis was not attempted for H and 0 because they are inferred, 0 from stoichiometry and H from LOI. Volume factors are all less than one, indicating volume reduction during alteration.
1) H and Sr have been added; 2) Zr and Th have been assumed to be immobile, thus con-
89-42
’
-60 : -80 -
straining Hf, Ta and Lu also to have been immobile; 3) Al, U, Nb, Ca, Rb, Y, and most of the REEs have undergone minor removal; 4) F, K, Si, Fe, Na, Ti, Mn, and Eu have been removed in quantities ranging from 15 to more than 50% of the amounts originally present. Although it is not evident from the figures, it is noteworthy that H has not been added in sufficient quantity to match the net loss of cations on a charge-equivalent basis. Therefore, the mechanism of leaching did not depend on a direct, chargeequivalent replacement of cations by H+. The reported mobility of several elements whose equilibrium solubilities in low-ionic-strength solutions are extremely low (cf. HOGFELDT, 1982) is surprising but not irreconcilable with the observations. The fluid-rock ratio in this system is extremely high owing to high rainfall and the permeability of the pumice; flushing of “immobile” metals is common in lateritization ( KRONBERG et al., 1979) and so might occur on Terceira to the limited extent we have proposed. Furthermore, the metal ions in question do not sit in thermodynamically stable sites in the glass and consequently can be expected to show metastable solubilities much higher than the equilibrium solubilities of their minerals would suggest.
that a similar ratio applies
I H
Sr
I
.I
I P
Hf
Al Nb N
Eu Yb Ca tn
0
I F
Si Na Yn
, , , ( , , ,
I
89-43
-100
1
I I 1 1
-40 :
-60 -80 -
_I
H
Sr P Hf Al Nb Nd Eu Yb Ca 0 F Si Na Hn I Mg Th Zr Ta U Ce Sm Tb Lu Fib Y K Fe Ti Ba
c 8 8 3 ’ 8 8 m0 3 ’ 8 ’ 8 8 8
Element FIG.7.Changesin mass expressed as percentage
of amount otiginally present, at selected values of volume change (see Table 2). Symbols as in Fig. 6. H does not appear because additions were several hundred percent in all cases. Al, U, and Nb have been significantly reduced in each sample, with the possible exception of U in 89-43; Ca, Rb, Y, F, K. Si, Fe, Na, Ti, Mn, and Ba have all been removed in large amounts. All elements show differing degrees of mobility, indicating that the immobile elements Zr and Th are not simply residual from process of bulk dissolution of the glass, but that incongruent dissolution has dominated.
Severe leaching of trachytic glass at ambient conditions, most water dissolved in the Pica do Carnheiro pumice must be present as molecular H20. This assumption is consistent with the spectroscopic data of PANDYA et al. ( 1992), who found abundant water to be present exclusively as molecular Hz0 in synthetic glass hydrated in air under ambient conditions for periods of several years. Allowance of longer times for equilibration in natural conditions might be expected to result in somewhat higher (but still small) abundances of hydrolyzed oxygen atoms. On the other hand, hydration of alkali silicate glasses in aqueous solution is accompanied by Na+-HsO+ (or H+) interdiffusion (LANFORD et al., 1979; HOUSER et al., 1980; TSONGet al., 1980; MAGONTHIERet al., 1992; VERNAZand DUSSOSSOY,1992). The process is greatly enhanced in solutions with constant low pH, where an abundance of H30+ ions is available for exchange with Na+ at the glass-water interface (BUNKER et al., 1983). This view of hydration as an ion-exchange process is fundamentally different from that of ZHANG et al. ( 199 1) and is inconsistent with our observation that there has not been a charge-equivalent replacement of cations by H+. DRAN et al. ( 1988) investigated this apparent contradiction and concluded that glass reacting with an aqueous solution at low temperatures is subjected to both ion exchange between Na+ and H+ and to inward diffusion of molecular water, aided by the porous structure of the Naexchanged glass. VERNAZand DUSSOSSOY( 1992) extended this model by suggesting that once the structure of the glass has been opened and saturated with water to form a gel, Si is free to diffuse outward to attain equilibrium with the surrounding aqueous phase. In the case of the Pica do Camheiros glass, we suggest that the models of DRAN et al. ( 1988) and VERNAZ and DusSOSSOY( 1992) should be extended still further to allow for diffusion of other high-field-strength elements (HFSE) besides Si. Although H,O+-cation interdiffusion has probably dominated the leaching of alkalis and assisted inward diffusion of molecular water, we envisage a catalytic role for water in the diffusion of more highly charged cations, including Si4+ ( DONNAYet al., 1959). The presence of water speeds diffusion of cations in silicate melts by up to two orders of magnitude (BAKER, 199 1). When the pumice contains 7% Hz0 by weight, approximately half of the cations in the glass are hydrogen ions; at this concentration, water can be expected to play a very major role in diffusive processes. Successive hydrolysis and dehydrolysis reactions might allow HFSE to migrate without requiring charge-equivalent replacement of cations by H +; instead, each water molecule could participate in the leaching of several cations in succession. This mechanism should operate equally well for elements such as Zr as for Si, so that the controlling factor governing diffusive loss of HFSE might be their solubility in the ambient aqueous solution rather than their mobility within the hydrated glass itself. The effective diffusion coefficient D of sodium in pumice from Pica do Carnheiro may be estimated if several simplifying assumptions are made. Assuming that the bulk of the pumice resides in the thin curviplanar walls of the bubbles (Fig. I ), we can use an approximate solution to Fick’s law to estimate ditfusivity of Na ( TALBOTand K~TCHENER,1956; CRANK, 1975), as follows:
81
where M, is the amount of Na that has left the septum at time t, M, is the corresponding amount at infinite time, D is the effective diffusion coefficient of Na, and I is one half the thickness of the septum. If we assume that the steadystate concentration of Na at the glass-water interface is close to zero ( SCHOLZE,1988), the ratio M,/M, becomes simply the fraction of Na lost from the sample. For sample 89-43, this ratio is estimated as 0.40. For a mean bubble wall thickness 21 of 4 pm and a period of time t = 2000 years, the corresponding value of D is 1.8 X lo-l9 cm2 s-’ . Using 600 years as the youngest possible age of the deposit increases the estimated value of D to 6.1 X lo-l9 cm2 s-’ . These estimates supersede our preliminary results ( MUNGALLand MARTIN, 1992). Low-temperature self-diffusion coefficients for Na in silicate glasses are available from a number of sources. HOUSER et al. (1980) found DNa = 1.48 X lo-” cm2 s-’ at 30°C in soda-silicate glass. LANFORDet al. ( 1979) reported a value of DNa equal to lo-l3 cm2 s-’ at 90°C in soda-lime silicate glass. WHITE (1983) estimated DNa = 5 X lo-l9 cm2 s-’ based on experiments on obsidian at 25°C. The values of DNa listed above were all derived assuming binary interdiffusion of Na and H, and reflect a strong dependence of DNa on Na content or peralkalinity of the glass. Our approach allows us to calculate an effective diffusion coefficient D of Na which can be correlated with the self-diffusion coefficients according to the following relation: D=
DtiD~a
C’,D,
+
(1 - CH)DNa ’
(3)
where DNa.~ are the self-diffusion coefficients of Na and H, and CNa.u are their respective concentrations (DOREMUS, 1975 ) . The inferred process of leaching at Pica do Camheiro took place near the ground surface in a maritime temperate climate at a probable average soil temperature between 10” and 25’C. Using the temperature dependence offi. reported by WHITE ( 1983), we estimate that ha lies between 1 X 10m2’ and 5 X lo-l9 cm2 s-’ at these temperatures. As noted by WHITE ( 1983), values of DNa are slightly higher than the effective diffusion coefficients as calculated here because values of DH are somewhat lower. One thus expects an effective binary diffusion coefficient for Na to lie near the low end of this range. Our slightly elevated D probably reflects the mildly peralkaline character of the starting composition of lavas and pumice erupted from Pica do Camheiro. However, the close agreement between our inferred D and the values measured experimentally by WHITE ( 1983) indicates that the observed variation in Na contents in pumice and lava from Pica do Camheiro is fully consistent with the known effects of leaching of glass by water. The apparent addition of Sr is perhaps surprising, but COUSENSet al. ( 1993 ) documented a similar uptake of Sr in glassy pyroclastic rocks from the island of Gran Canaria, and argued convincingly, on the basis of a very complete isotopic data-set, that the effect there must have been due to lowtemperature leaching, involving addition of Sr from meteoric water with a possible seawater component.
82
J. E. Mungall and R. F. Martin
IMPLICATIONS AND CONCLUSIONS The results outlined above have several interesting consequences. We have shown that leaching of glass can induce drastic changes in composition without inception of devitrification. In this instance, a moderately peralkaline trachyte glass has been altered to a strongly peraluminous composition with anomalously high abundances of incompatible trace elements while remaining in an entirely glassy state. Preservation of glass free of devitrification is clearly not proof that a rock has escaped chemical alteration; even elements generally considered to be immobile may have been lost. The presence of up to 80% additional Sr indicates that measurement of isotope ratios for petrogenetic studies could be seriously compromised. If Sr has actually diffused into the glass, in a manner exactly analogous to that in which other elements have left the system, then a brief leaching of the glass surface to remove contaminants, as DUPR~ et al. ( 1982) did on actual samples from Terceira, would not detect the added Sr. If glassy rocks with high specific areas are selected for isotopic analysis, one must ensure that samples have not undergone the type of alteration described here. The release of K, Mn, and other organically important nutrients by the low-temperature leaching of glass documented here accounts in part for the very fertile nature of soils developed on glassy pyroclastic materials ( QUANTIN et al., 1991). Soft, friable pumice and ignimbrite show great potential as single-application fertilizers, whereas the synthetic glass preparations and mixtures investigated by FYFE et al. ( 1978) would seem to be prohibitively expensive. The mobility observed here for U, Y, REEs, and Nb has important ramifications for projected storage of nuclear wastes in the form of glass. Previous studies have shown no evidence for mobility of U or Th in naturally weathered glass ( ROSHOLTand NOBLE, 1971; MAGONTHIERet al., 1992), but it is clear from our results that U and some other highfield-strength heavy elements are not locked into the structure of felsic glass and can be released in significant amounts in times as short as 1000 years. It is therefore incumbent on proponents of glass-based storage of nuclear waste to demonstrate that radionuclides will not be liberated by processes like those documented to have occurred in our natural example. Acknowledgments-This study comprises part of a doctoral dissertation by J. E. Mungall. The work was supported by NSERC operating grants to R. F. Martin. The generous assistance of I. L. Gibson at Waterloo in acquiring REE data is particularly appreciated. Numerous colleagues offered helpful criticism along the way, particularly R. Wares and T. Skulski; the comments of A. White and two anonymous reviewers have helped greatly to correct errors and clarify our discussion. Editorial handling: M. F. Hochella Jr.
REFERENCES APPLEYARDE. C. and WOOLLEYA. R. ( 1979) Fenitization: An example ofthe problems of characterizing mass transfer and volume changes. Chem. Geol. 26, 1-15. BAKERD. R. ( 199 I ) Interdiffusion of hydrous dacitic and rhyolitic melts and the efficacy of rhyolite contamination of dacite enclaves. Contrib. Mineral. Petrol. 106, 462-473.
BARRETTT. J. and MACLEANW. H. ( 1991) Chemical, mass, and oxygen isotope changes during extreme hydrothermal alteration of an Archean rhyolite, Noranda, Quebec. Econ. Geol. 86, 406414.
BENNETH. and OLIVERC. R. ( 1992) XRF Analysis oJCerumics, Minerals, and Allied Materials. J Wiley & Sons. BERGERG., SCHOTT J., and GUY C. (1988) Behaviour of Li, Rb, and Cs during basalt glass and olivine dissolution and chlorite, smectite, and zeolite precipitation from seawater: Experimental investigations and modelization between 50 and 300°C. Chem. tieol. 11, 297-312. BRYANW. B., FINGER L. W., and CHAYESF. (1968) Estimating proportions in petrographic mixing equations by least-squares approximation. Science 163, 926-927. BUNKERB. C., ARNOLDG. W., BEAUCHAMPE. K., and DAY D. E. ( 1983) Mechanisms for alkali leaching in mixed Na-K silicate glasses. J. Non-Cryst. Solids 58, 295-322, COUSENSB. L., SPERAF. J., and DOBSONP. F. ( 1993) Post-eruptive alteration of silicic ignimbrites and lavas, Gran Canaria, Canary Islands: Strontium, neodymium, lead, and oxygen isotopic evidence. Geochim. Cosmochim. Acta 57,63 I-640. CRANKJ. ( I975 ) The Mathematics of Diffusion, 2nd ed. Clarendon Press. DONNAYG., WYAR~J., and SABATIERG. ( 1959) Structural mechanism of thermal and compositional transformations in silicates. Z. Kristallogr. 112, I6 I - 168. DOREMUSR. H. ( 1975) Interdiffusion of hydrogen and alkali ions in a glass surface. J. Non-Cryst. Solids 19, 137-144. DRAN J.-C., DELLA MEA G., PACCAGNELLA A., PETITJ.-C.. and TROTIGNONL. ( 1988) The aqueous dissolution of alkali silicate glasses: Reappraisal of mechanisms by H and Na depth profiling with high energy ion beams. Phys. Chem. Glasses 29, 249-255. DUNKENH. and D~REMUSR. H. ( 1987) Short time reactions of a NazO-CaO-SiOz glass with water and salt solutions. J. Non-Cryst. Solids 92, 6 I-72.
DIJPR~B., LAMBRETB., and ALL~GREC. J. ( 1982) Isotopic variations within a single oceanic island: The Terceira case. Nature 299,620FYFEW. S., KRONBERGB. I., and LEONARDOS 0. H. (1978) Phosphate release from synthetic glasses and inhibition of phosphate fixation on ferric hydroxide. Soil Sci. Plant Nutr. 24, 455-464. GIBSONI. L. and JAGAMP. ( 1980) Instrumental neutron activation analysis of rocks and minerals. In Neutron Activation Analysis in the Geosciences (ed. G. K. MUECKE); Mineral. Assoc. Canada Short-Cour.se Handbook, Vol. 5. vv. 109- 131. GRESENSR. L. ( 1967) Composition-volume relationships of metasomatism. Chem. Geol. 2, 47-65. HOCFELDTE. ( 1982) Stability constants of metal-ion complexes. Part A: Inorganic ligands. IUPAC Chemical Data Series, Vol. 2 1, Pergamon. HOUSERC. A., HERMANJ. S., TSONG 1. S. T., WHITE W. B., and LANFORDW. A. ( 1980) Sodium-hydrogen interdiffision in sodium silicate glasses. J. Non-Cryst. Solids 4i, 89-98. JEZEK P. A. and NOBLED. C. ( 1978) Natural hvdration and ion exchange of obsidian: An electron microprobe study. Amer. Mineral. 63, 266-273.
KRONBERGB. l., FYFEW. S., LEONARDOS 0. H., JR., and SANTOS A. M. ( 1979) The chemistry of some Brazilian soils: Element mobility during intense weathering. Chem. Geol. 24, 2 I 1-2 19. LANFORDW. A., DAVISK., LAMARCHEP., LAURSENT., GROLEAU R., and DOREMUSR. H. ( 1979) Hydration of soda-lime glass. J. Non-Cryst. Solids 33, 249-266.
LIPMANP. W., CHRISTIANSEN R. L., and VAN ALSTINER. E.(1969) Retention of alkalis by talc-alkalic rhyolites during crystallization and hydration. Amer. Mineral. 54, 286-29 1. MAGONTHIERM.-C., PETITJ.-C., and DRANJ.-C. ( 1992) Rhyolitic glasses as natural analogues of nuclear waste glasses: Behaviour of an Icelandic glass upon natural aqueous corrosion. Appl. Geochem. 1, 83-93 (suppl.). MICHELSJ. W., TSONG1. S. T., and NELSON C.M. ( 1983) Obsidian dating and East African archeology. Science 219, 36 l-366. MUNGALLJ. E. and MARTINR. F. ( 1992) Selective leaching of cations accompanying secondary hydration of pumiceous silicic glass,
Severe leaching of trachytic glass Terceira, Azores. Geol. Assoc. Canada-Mineral.
Assoc. Canada
Prog. Abstr. 17, A8 I. NOBLE D.
C. ( 1967) Sodium, potassium, and ferrous iron contents of some secondarily hydrated natural silicic glasses. Amer. Mineral.
52,280-287.
PALMERH. C., TAZAKIK., FYFE W. S., and ZHOU Z. ( 1988) Precambrian glass. Geology 16,22 l-224. PANDYAN., SHARMAS. K., MUENOWD. W., and SHERRIFFB. L. ( 1992) Hydration of alkali silicate glasses at ambient conditions. EOS; Trans. AGU 73, 36 I (abstr.). PETIT J.-C. ( 1992) Reasoning by analogy: Rational foundations of natural analogue studies. Appl. Geochem. 1, 9- 11 ( suppl.). QUANTINP., BALFSDENT J., B~ULEAUA., DEL~UNEM., and FELLER C. ( I99 I ) Premiers stades d’alteration de ponces volcaniques en climat tropical humide, Montagne PelCe, Martinique. Geodenna 50, 125-148. ROSHOLTJ. N. and NOBLED. C. ( 197 1) Mobility of uranium and thorium in glassy and crystallized silicic volcanic rocks. Econ. Geol. 66, 1061-1069. SCHA-ITNERK. H., DOREMUSR. H., and LANFORDW. A. (1988) Hydrogen analysis of soda-lime silicate glass. J. Non-Cryst. Solids 102, 1 l-18.
H. ( 1988) Glass-water interactions. J. Non-Cryst. Solids 102, I-10. SELFS. ( 1976 ) The Recent volcanology of Terceira, Azores. .I. Geol. QHOLZE
Sot. London 132,645-666.
SMETSB. M. J. and LOMMENT. P. A. ( 1983) The role of molecular water in the leaching of glass. Phys. Chem. Glasses 24, 35-36. STAUDIGELH. and HART S. ( 1983) Alteration of basaltic glass: Mechanisms and significance for the oceanic crust-seawater budget. Geochim. Cosmochim. Acta 47,337-350.
83
STOLPERE. ( 1989) Temperature dependence of the speciation of water in rhyolitic melts and glasses. Amer. Mineral. 74, 12471257.
TALBOTA. and KITCHENERJ. A. ( 1956) Diffusion (or conduction) along a slightly tapering tube, and its application to the determination of diffusion coefficients. Brit. J. Appl. Phys. 7,96. TSONGI. S. T., HOUSERC. A., YUSEFN. A., MESSIERR. F., WHITE W. B., and MICHEL~J. W. ( 1978) Obsidian hydration profiles measured by sputter-induced optical emission. Science 201, 339341. TSONG 1. S. T., HOUSER C. A., and TWANG S. S. C. ( 1980) Depth profiles of interdiffusing species in hydrated glasses. Phys. Chem. Glasses 21, 197-198. VERNAZE. Y. and Dussosso~ J. L. ( 1992) Current state of knowledge of nuclear waste glass corrosion mechanisms: The case of R7T7 alass. ADDI.Geochem. 1. 13-22 (suuul.). WEAVER-S. D.,- GIBSON 1. L., ‘HOUGHTON B. F., and WILSON C. J. N. ( 1990) Mobility of rare earth and other elements during crystallization of peralkaline silicic lavas. J. Volcano/ Geotherm. Rex 43, 57-70.
WHITE A. F. ( 1983) Surface chemistry and dissolution kinetics of glassy rocks at 25°C. Geochim. Cosmochim. Acta 47,805-8 15. WHITEA. F., CLAASSENH. C., and BENSONL. V. ( 1980) The effect of dissolution of volcanic glass on the water chemistry in a tuffaceous aquifer, Rainier Mesa, Nevada. USGS Water Supp. Pap. 1535-Q.
ZHANGY., STOLPERE. M., and WASSERBURG G. J. (1991) Diffusion of water in rhyolitic glasses. Geochim. Cosmochim. Acta 55,441456.
ZIELINSKIR. A., LIPMANP. W., and MILLARDH. T., JR. (1977) Minor-element abundances in obsidian, perlite, and felsite of calcalkalic rhyolites. Amer. Mineral. 62,426-437.