GEOLOGICAL SETTING AND SAMPLE DESCRIPTION

Tenerife is the largest and tallest (Teide: 3718 m) of the Canary Islands, and is located in the central part of the island chain (Fig. 1). The oldest exposed units are the `Basaltic Shields' which form three deeply eroded massifs, Roque del Conde in the SW, Teno in the NW, and Anaga in the NE. These shields are dominated by alkali basaltic lavas, although their upper parts include evolved lavas and pyroclastics (Fúster et al., 1968; Ancochea et al., 1990). K-Ar analyses suggest age ranges of 8·5-3·8 Ma for Roque del Conde, 6·7-4·5 Ma for Teno, and 6·5-3·3 Ma for the Anaga Massif (Ancochea et al., 1990). The spatial relationships between the three shields at depth are unknown.

Figure 1. Map of Tenerife, Canary Islands, showing sample locations ([circles]). Based on map by Ancochea et al., (1990). SdT, Santiago del Teide; BdN, Buenavista del Norte; Ch, Chico; SI, San Isidro; V, Vilaflor.

A large volcanic complex, the Las Cañadas edifice, consisting of basanitic to phonolitic lavas and pyroclastics subsequently formed in the central part of Tenerife (Ancochea et al., 1990). K-Ar dating suggests an age range of >3·3-0·17 Ma (Martí et al., 1990, , 1994; Fúster et al., 1994). The top of the Las Cañadas edifice was destroyed through several caldera collapses, which formed the composite Las Cañadas caldera (Martí et al., 1994, , 1997). Subsequent extrusion of basanites to phonolites formed the Teide-Pico Viejo volcanic complex inside the Las Cañadas caldera (e.g. Ablay et al., 1998). A series of lava flows (mostly basaltic; 0·87-0·56 Ma) form the dorsal ridge between the Las Cañadas edifice and the Anaga massif (Ancochea et al., 1990). Both the Teide-Pico Viejo complex and many flank vent systems on the Las Cañadas edifice have been active into historical time [Ancochea et al., (1990) and references therein]. Sample localities are shown in Fig. 1 and listed in the Appendix. The data presented here include basaltic to phonolitic lavas and dykes of different ages and from different parts of the island, and thus give a general overview of chemical variations among the mafic to intermediate rocks and their minerals in different parts of Tenerife.

ANALYTICAL METHODS

Major elements were analysed on fused pellets using 9:1dilution with sodium tetraborate, and alpha factor corrections. The analyses were performed on a Philips PW 2400 X-ray fluorescence spectrograph with X47 software at the Department of Geology, University of Oslo.

Trace element concentrations were determined by different methods. All rock samples were analysed for selected trace elements by X-ray fluorescence (XRF) spectrometry on pressed powder pellets (cemented by Paraloid), using matrix corrections based on Compton Top measurements. Some samples were analysed for trace elements by epithermal instrumental neutron activation analysis (INAA) at the Mineralogisk-Geologisk Museum, University of Oslo. Powdered rock samples were wrapped in Al foil, and placed in a cadmium box, and were irradiated for 48 h with a thermal neutron flux of 1·5 * 10¹³ n/cm² per s in the JEEP-II reactor at the Institute of Energy Technology, Kjeller, Norway. Gamma spectrometry measurements were performed using a coaxial Ge(Li) detector with Canberra electronics. The detector was calibrated using pure Eu and Co sources. Each sample was counted twice, after 3-9 days for 1800 or 3600 s, and after 25-30 days for 7200 or 14 400 s. The international rock standards BCR-1, BHVO-1 and G-2 were used for calibration [using standard values recommended by Govindaraju, (1989)] and included as unknowns in each run. The method has been described by Brunfelt & Steinnes, (1969). Standard deviations are based on eight repeated countings of one sample for 50 000 and 14 400 s. Another group of samples was analysed for trace elements by inductively coupled plasma emission spectrometry (ICP-MS) at ACTLABS, Ancaster, Ontario, Canada. In general, there is excellent agreement between results obtained by different methods on the same samples.

Minerals were analysed for major elements on an automatic wavelength-dispersive CAMECA CAMEBAX electron microprobe fitted with a LINK energy dispersive system at the Mineralogisk-Geologisk Museum, University of Oslo. An acceleration voltage of 15 keV, sample currents of 20 nA (mafic silicates and oxides) or 10 nA (plagioclase and glass), and counting times of 10 s were used. Oxides and natural and synthetic minerals were used as standards. Matrix corrections were performed by the PAP procedure in the CAMECA software. Analytical precision (2[sgr]) evaluated by repeated analyses of individual mineral grains, is better than ±1% for elements with concentrations of >= 20 wt % oxide, better than ±2% for elements in the range 10-20 wt % oxide, better than 5% for elements in the range 2-10 wt % oxide, and better than 10% for elements in the range 0·5-2 wt % oxide.

Trace element compositions in minerals were determined by laser ablation ICP-MS on polished thick sections (110 µm) at the Geochemical Analyses Unit, GEMOC Key Centre, Macquarie University, Sydney. A detailed description of the laser microprobe and ICP-MS instrumentation and operating procedures has been given by Norman et al., (1996). The laser is a Q-switched Nd-YAG laser, operating at 266 nm (UV). Analyses were performed with repetition rate of 2-4 Hz and an energy of ~0·5-2 mJ/pulse. These conditions produced a sampling area 30-50 µm in diameter and a maximum drill rate of ~0·5 µm/s. The ablated material is transported in a stream of high-purity Ar directly into the ICP-MS system. Plasma operating conditions for the Perkin Elmer ELAN 5100 ICP-MS system included a forward power of 1040 W and nebulizer gas flow of 0·96 l/min, which gave a ²⁴⁸Th/²³²Th ratio of 0·5-1%. Dwell times of 50-100 ms were used and counts were collected in peak-hopping mode with one sweep per reading and one reading per replicate. A typical analysis included 30-40 replicates on background and 50-100 replicates on the sample signal. Data collection, reduction procedures, precision and accuracy have been described by Norman et al., (1996). They demonstrate an instrumental precision of <5% at ppm levels.

Microscopic observations and microthermometric measurements of fluid inclusions were performed on doubly polished thick sections, ~180 m thick, on a Chaixmeca heating-cooling stage with a cooling medium of pre-cooled nitrogen. The stage was calibrated with a series of pure synthetic substances with known melting points in the range -142°C (methylcyclopentane) to +30°C (gallium). The calibration curve of the instrument is flat in the entire interval, with a deviation from the true temperature of maximum 1·0°C. All CO₂ melting and homogenization temperatures have been corrected accordingly. The reproducibility of CO₂ melting point measurements is estimated to be ±1·0°C.

MINERALS

Petrography and major element chemistry

A number of the analysed lavas and dykes are aphyric, but the majority of the samples carry phenocrysts which may make up a significant proportion of the rocks (up to 55 vol. %). From basanites to phonolites, the phenocryst assemblages are: ol, ol + cpx, ol + cpx + mt + plag ± ap, cpx ± plag ± mt, ± ol + cpx + plag + krs ± ap ± il (ol, olivine; cpx, clinopyroxene; mt, magnetite; plag, plagioclase; krs, kaersutite; ap, apatite; il, ilmenite; Fig. 2). Modal compositions, determined on the basis of point counting 1500-3000 points per thin section, are listed in the Appendix. Samples with <5 vol. % phenocrysts have been defined as aphyric and near-aphyric. A few dykes and cinder cones carry ultramafic xenoliths (spinel harzburgites, spinel lherzolites and spinel dunites), indicating fast ascent from mantle levels to the upper crust. A study of the petrographic relations in Tenerife lavas was first presented by Scott, (1976).

Figure 2. (a) Relative proportions of olivine ([squares]), clinopyroxene ([triangles]) and plagioclase ([diamonds]) of the phenocryst assemblages of Tenerife volcanics, plotted against the forsterite content in cores of olivine phenocrysts. (b) Generalized overview of changes in phenocryst assemblages with decreasing Fo content. Continuous lines, major phenocryst phase in most rocks; dashed line, major phenocryst phase in some rocks, or accessory phase present in most rocks; dotted line, accessory phase in some rocks; grey field, transition from spinel to Ti-magnetite; [circles], rocks with kaersutite xenocrysts. (Note that Cr-Al-spinel-titano-magnetite occurs only as inclusions in the most mafic rocks.)

Olivine (5 to 1 mm long) occurs as euhedral to subhedral phenocrysts, broken fragments, strongly corroded grains, and as cores enclosed by clinopyroxene. Both normal and reverse zoning are observed (Table 1). Olivine is the dominant phenocryst phase in the most mafic samples (Fo_88·7-78·0; Fig. 2) and is an accessory phenocryst phase in the somewhat less magnesian rocks (Fo_78-60). The most evolved samples generally do not contain olivine (Fig. 2), but phonolite TF126 [mg-number = 35·2; mg-number = atomic proportion Mg * 100/(Mg + Fe_total)] contains olivine (Fo₇₄) with reaction rims consisting of clinopyroxene + oxides (Fig. 3b). This olivine is clearly not in equilibrium with the host lava. A few olivine-clinopyroxene clusters are believed to represent cumulate xenoliths (Fig. 3a).

Table 1. Representative olivine compositions

Figure 3. Photomicrographs of samples. (a) Xenolith consisting of a fine-grained intergrowth of pale olivine and yellowish clinopyroxene (TF60). The overgrowth of zoned, subhedral brown clinopyroxene rim in the upper right corner of the xenolith (field of view width 1·7 mm) should be noted. (b) Corroded olivine crystal with reaction rim consisting of oxides, pale brown glass and pale Na-salite (TF126; field of view width 1·7 mm). Black blobs are ink markings for microprobe analyses. (c) Clinopyroxene consisting of green Na-salite mantled by light brown Al-salite (TF103; field of view width 0·85 mm). Arrows mark the position of cross-section presented in Fig. 5. (d) Zoned clinopyroxene consisting of a very dark green, rounded Na-salite core mantled by an inner zone of pale brown Al-salite, and an outer zone of colourless Al-salite (TF86, grain I in Table 2; field of view width 3·4 mm). Arrows mark the position of cross-section presented in Fig. 6. (e) Fragment of zoned clinopyroxene with broken side, showing evidence of incipient melting (TF99; field of view width 1·7 mm). (f) Kaersutite crystal with reaction rim consisting of Na-salite, oxides and glass (TF126; field of view width 0·85 mm).

Clinopyroxene phenocrysts (7 to 1 mm long) are present in most rocks (Fig. 2). The most common type of clinopyroxene is relatively Al-Ti-rich diopside to salite (mg-number = 87-66; referred to below as Al-salite) which occurs as euhedral to anhedral, zoned, pinkish, beige or olive grains, as zoned overgrowths on corroded cores of Na-rich salite (mg-number = 79-49; referred to below as Na-salite; Fig. 3c and d), and as fragments, some of which show evidence of incipient dissolution (Fig. 3e). Al-salite contains 0·9-6·1 wt % TiO₂, 3-11 wt % Al₂O₃, up to 0·87 wt % Cr₂O₃, and 0·3-0·8 wt % Na₂O (Table 2), and defines a trend of increasing TiO₂ and Al₂O₃, and relatively constant Na₂O with decreasing mg-number (Fig. 4). The zoning pattern of Al-salites is often complex, including repeated zones of normal and reverse zoning (Table 2; Fig. 6, below); sector zoning is visible in some grains. In some lavas homogeneous, corroded cores of Al-salite of one composition are coated by Al-salite of another composition. Na-salite generally forms rounded or corroded, green to greenish brown cores that lack visible zoning; these cores are mantled by Al-salite, although in a few evolved lavas (e.g. phonolite TF126) Na-salite occurs as euhedral to subhedral microphenocrysts. The Na-salite is characterized by low TiO₂, Al₂O₃ and Cr₂O₃, and high Na₂O and MnO contents (0·5-3·9 wt % TiO₂, 1·3-7·5 wt % Al₂O₃, <0·1 wt % Cr₂O₃ and 0·8-2·4 wt % Na₂O). Na₂O increases with decreasing mg-number (Fig. 4). The mantled grains generally show a sharp transition in composition from Na-salite cores to Al-salite overgrowths (Figs 5 and 6), although gradual transitions also occur. The mg-number normally increases from Na-salite cores to Al-salite overgrowths, but decreasing mg-number has also been observed (Table 2, Figs 4- 6). It is to be noticed that Al-salites are, on average, more magnesian than Na-salites. The occurrence of corroded green, sodic ferrosalite as cores in clinopyroxene phenocrysts in mafic to intermediate lavas from Tenerife was first described by Scott, (1976).

Table 2. Representative clinopyroxene compositions

Figure 4. TiO₂, Al₂O₃, Cr₂O₃ and Na₂O in Tenerife clinopyroxenes plotted against mg-number. [squares], Na-salite; [diamonds], Al-salite. Zones of different compositions within single crystals are connected by lines. The Na-salites and Al-salites define separate compositional trends. (See text for further explanation.)

Figure 5. Zoning profile for mantled clinopyroxene in sample TF103, shown in Fig. 3c. Z1, rounded Na-salite core; Z2, Al-salite overgrowth. The homogeneous composition of the Na-salite core should be noted.

Figure 6. Zoning profile for mantled clinopyroxene in sample TF86 (grain I in (Table 2; Fig. 3d). The complex zoning pattern in the Al-salite overgrowth, with alternating normal and reverse zoning, should be noted. Z1, corroded Na-salite core; Z2a, inner zone of Al-salite overgrowth (brownish in Fig. 3d); Z2b, outer zone of Al-salite overgrowth (pale in Fig. 3d). The transition between zones Z2a and Z2b is marked by reverse zoning.

Oxides are present in all the rocks. In the most mafic lavas the oxide phase is restricted to small inclusions of Al-Cr spinel (sp) in olivine and clinopyroxene. There is a general tendency for Al + Cr content in spinel inclusions to decrease, and Ti + Fe³⁺ to increase, with decreasing Fo content in the host olivine (Table 3). More evolved lavas contain titanomagnetite both as inclusions in silicate phenocrysts and as separate phenocrysts (<1 mm in diameter). An abundance of titanomagnetite phenocrysts is only present in rocks with Fo_<77 (Fig. 2). Exsolution lamellae of ilmenite are common in thetitanomagnetite, and a few of the most evolved lavas also contain ilmenite as separate grains.

Table 3. Representative oxide compositions

Plagioclase ( <= 10 mm long) is present in moderate proportions in some of the more mafic rocks (Fo_86-83), is the dominant phenocryst phase in some lavas in the range Fo_82-78, and the dominant phenocryst phase in all the more evolved lavas (Fo_<78; Fig. 2). Like other phenocryst phases, plagioclase occurs both as subhedral grains, as rounded grains, and as broken fragments. In some rocks the plagioclase is `sieve-textured', suggesting partial melting. Plagioclase phenocrysts are generally zoned; the total compositional range is An_83·6Ab_15·9Or_0·5 to An_23·3Ab_68·7Or_8·0. K₂O contents are relatively low (0·5-0·7 wt %) in the most calcic plagioclase (An_70-84), but increase with decreasing CaO from about 1·8 wt % K₂O at An_60-65, through 4·2 wt % at An₄₂, to 8·0 wt % K₂O at An₂₃. One tephrite (TF41-2, Fo_83-72) contains rounded alkali feldspar grains (An₇Ab₅₉Or₃₄) with reaction rims.

Kaersutite is present in a few samples ( Fig. 2) as euhedral to ovoid grains (<0·5 mm in diameter) with reaction rims of Na-salite, oxides and glass (Fig. 3f). In one sample (TF99: Fo₇₇) kaersutite forms inclusions in clinopyroxene phenocrysts. The kaersutite contains 5·6-7·0 wt % TiO₂, 10·7-13·4 wt % Al₂O₃, and covers a restricted range in mg-number (58-69; Table 4).

Table 4. Representative kaersutite compositions, together with glass in reaction rim on the kaersutite in sample TF126

Apatite is frequently present as minute inclusions in clinopyroxene, plagioclase and kaersutite phenocrysts in relatively evolved lavas (e.g. Fig. 3c), and forms an accessory phenocryst phase in a few of the most evolved lavas.

Inclusions

Olivine phenocrysts frequently contain numerous small, equant inclusions of spinel or titanomagnetite, and sometimes also inclusions of devitrified glass or glass-bearing polyphase inclusions. Devitrified glass and polyphase inclusions are also common in clinopyroxene and plagioclase phenocrysts; in some samples they are concentrated along growth zones. Globular sulphide inclusions also are common whereas fluid inclusions are rare. However, secondary trails of minute fluid inclusions are occasionally found in olivine, clinopyroxene or plagioclase phenocrysts.

Trace element chemistry

Six lavas of different ages and from different parts of the island (TF67, TF68, TF86 in the Teno part of the `Basaltic Shield'; TF114 in the Anaga part of the `Basaltic Shield'; TF58, TF126 in the younger Las Cañadas edifice; Fig. 1) were selected for trace element analyses of clinopyroxenes (Table 2). In addition, olivine was analysed in four samples (TF58, TF67, TF68 and TF114; Table 1) and amphibole in one (TF126, Table 4). Samples TF58 (near-aphyric) and TF86 (strongly porphyritic) carry phenocrysts of olivine (Fo_83-81 and Fo_86-81, respectively), mantled clinopyroxene and titanomagnetite. Samples TF67, TF68 and TF114 carry phenocrysts of olivine (Fo_86-81, Fo_80-76 and Fo₇₉, respectively), and Al-salite; TF114 also carries titanomagnetite phenocrysts. The phonolite TF126 is nearly aphyric, but contains microphenocrysts of Na-salite, Al-salite and plagioclase, plus corroded xenocrysts of olivine (Fo_76-74) with reaction rims of Na-salite + titanomagnetite, and of kaersutite surrounded by a reaction zone of transitional salite, titanomagnetite and glass (Fig. 3b, f, Appendix).

Clinopyroxene

The differences in major element compositions exhibitedby Al-salites and Na-salites are reflected in their trace element chemistry. Compared with Na-salites, the Al-salites show lower concentrations in REE and Y, and are higher in compatible trace elements (e.g. Al-salites: 24-335 ppm Zr, 1·8-15·3 ppm Sm, 3-201 ppm Ni, and 23-135 ppm Sc; Na-salites: 144-777 ppm Zr, 8·9-34·2 ppm Sm, 2-15 ppm Ni, and 19-59 ppm Sc; Table 2). Furthermore, the Al-salites show mildly convex primitive-mantle-normalized trace element patterns, including weak negative anomalies for Nb and Sr, and higher primitive-mantle-normalized ratios for Sc and V than for the heavy REE (HREE). With increasing concentrations of lithophile elements, the differences between primitive-mantle-normalized ratios for the HREE, Sc and V decrease. These ratios are similar for the most enriched Al-salites, which also tend towards weak negative Ti anomalies (sample TF58). The Na-salites exhibit flat to slightly concave trace element patterns with pronounced negative anomalies for Sr and Ti, and are depleted in Sc and V relative to HREE (Fig. 7). It is to be noticed, however, that the most pronounced negative Sr and Ti anomalies are found in relatively magnesian Na-salites (mg-number = 75-76 in samples TF58 and TF126; Fig. 7). The trace element patterns thus indicate a general tendency for negative correlations between REE and non-lithophile elements such as Sc and V and progressive depletion of Sr and Ti with increasing concentrations of REE.

Figure 7. Trace element patterns for clinopyroxenes in Tenerife lavas of mafic to intermediate composition (Table 2), normalized to primitive mantle (PM; Sun & McDonough, 1989). Na-salites are indicated by data-points (squares and diamonds) connected by lines; Al-salites are shown only as lines and fields. (See text for further information.)

Typical trace element zoning patterns are presented in Fig. 8. Mantled clinopyroxene grains in samples TF58 and TF86 show large compositional contrasts between Na-salite cores (Z1; relatively high contents of Na₂O, MnO, Y, REE and Zr, and low contents of TiO₂, Al₂O₃, Sr, Ni and Sc), to Al-salite overgrowths (Z2; lower Na₂O, MnO, Y, REE and Zr, and a tendency for higher TiO₂, Al₂O₃, Sr, Ni and Sc). In sample TF86 these contrasts (exhibited by three separate grains) are accompanied by a strong increase in mg-number, whereas samples TF58 shows a mild decrease in mg-number from Na-salite core to the Al-salite rim. It is to be noticed that different phenocrysts from the same sample show similar compositional variations. The spacing of points in laser ICP-MS analyses is too large to study the transitions between zones in any detail. However, the major element profile presented in Fig. 6 implies that the transition from Na-salite core to Al-salite rim in clinopyroxenes in sample TF86 is abrupt. Negative Sr and Ti anomalies are measured as the ratios between primitive-mantle-normalized Sr and Ti (Sr_N and Ti_N), and hypothetical values falling on the straight lines Pr_N-Nd_N and Eu_N-Gd_N, respectively, in Fig. 7 [Sr_N/Sr_N* = Sr_N * 2/(Pr_N + Nd_N), Ti_N/Ti_N* = Ti_N * 2/(Eu_N + Gd_N)]. Al-salite grains and overgrowths commonly show alternating normal zoning (decreasing mg-number and Ni, constant or decreasing Ti_N/Ti_N* and Sr_N/Sr_N*, and increasing TiO₂, Al₂O₃, Y and REE within each of the zones Z2a, Z2b and Z2c) and reverse zoning (increasing mg-number and Ni, decreasing TiO₂, Al₂O₃, Y and REE at the transitions from zone Z2a to Z2b, and from Z2b to Z2c).

Figure 8. Trace element zoning profiles for representative clinopyroxenes in Tenerife lavas (Table 2). It should be noted that in each lava different grains show similar chemical variations. Z1, rounded, corroded Na-salite cores; Z2, Al-salite overgrowths; Z2a, Z2b and Z2c, different Al-salite zones within which the zoning appears to be normal (decreasing mg-number and Ni, increasing TiO₂, Al₂O₃, REE and Y), separated by stages of reverse zoning (increasing mg-number and/or increasing Ni). Grain I in TF86 (five data points) is also shown in Figs 3d and 6, but the trace element analyses (and corresponding major element analyses) are taken along a profile about 90° to that exhibited in Fig. 6. It should be noted that major elements show the same concentrations in the profiles in Figs 6 and 8. The ratios Sr_N/Sr_N* and Ti_N/Ti_N* measure degree of Sr and Ti anomalies, and are defined in the text. c, core; r, rim.

Kaersutite

Kaersutite in sample TF126 is highly enriched in lithophile elements [e.g. enrichment factors >100 for Nb and light REE (LREE)], has marked negative anomalies for Sr and Ti, and higher primitive-mantle-normalized ratios for Lu, Sc and V than for all other REE (Table 4, Fig. 9).

Figure 9. Trace element patterns for kaersutite with reaction rims in the near-aphyric lava TF126 (filled signs) normalized to primitive mantle (PM; Sun & McDonough, 1989). The trace element patterns of kaersutite in basaltic glass in mantle xenoliths from La Palma are shown for comparison (E. Wulff-Pedersen & E.-R. Neumann, unpublished data, 1996). It should be noted that whereas kaersutite in basaltic glass in mantle xenoliths shows no Sr anomaly, kaersutite in lava TF126 exhibits marked negative Sr anomalies, suggesting crystallization from melts with a history of removal of plagioclase.

Olivine

The olivines show a range in trace element concentrations that appears to be unrelated to Fo contents (Table 1; Fig. 10). The highest concentrations of Al, Ti, Ni and Co are exhibited by the most magnesian olivines (TF114: Fo₈₇; Table 1, Fig. 10). Different olivine grains within the same sample may have significantly different trace element concentrations. The analysed olivines show a general positive correlation between Al and Ti, whereas different samples define separate trends in a Co-Ni plot (Fig. 10).

Figure 10. Trace element relations in olivine phenocrysts in Tenerife basalts showing trace element differences both between samples and between different crystal in the same sample.

WHOLE-ROCK CHEMISTRY

The rocks included in this study show total ranges in MgO and mg-number of 15·2-0·5 wt % and 69-23, respectively (Table 5; Fig. 11). Rocks with MgO > 9 wt %, which are highly porphyritic and contain large phenocrysts of olivine and/or clinopyroxene, show no correlation between (mg-number)_{whole rock} and Fo content of olivine. This indicates that their mafic character reflects accumulation of Mg-rich phases rather than a primitive nature. The most mafic aphyric to near-aphyric sample has 8·8 wt % MgO and mg-number = 55 (Fo₈₃ in olivine). Using the classification system of Le Bas et al., (1986), aphyric to near-aphyric rocks classify as basanites to phonolites. Estimated CIPW norms (assuming an Fe²⁺/Fe_total ratio of 0·75) show a total range of 3·0-19·4 % normative nepheline, with the strongest degree of silica undersaturation among the most evolved samples.

Table 5. Whole-rock compositions; major and trace element compositions of aphyric and near-aphyric lavas and dykes, and porphyritic rocks in which minerals have been analysed for trace elements

Figure 11. (a)Fig. 11.(a) Major element concentrations and agpaitic index [atomic ratio (K + Na)/Al], and (b) selected trace elements vs wt % MgO in Tenerife lavas of mafic to intermediate composition. Sample localities are shown in Fig. 1. The `Basaltic Shields' (Roque del Conde in SW, Teno in NW, and Anaga in NE) are shown as closed symbols, whereas younger lava series are indicated by open symbols. (See text for discussion.)

The analysed volcanics contain up to 5·0 wt % TiO₂, 0·5-1·3 wt % P₂O₅, and cover a wide range in incompatible trace elements (e.g. 49-244 ppm Nb, 49-142 ppm La, and 232-823 ppm Zr; Table 5). Lithophile elements such as K, Na, Rb, Th, U, Nb, Ta and REE increase, and elements compatible in olivine, clinopyroxene and Fe-Ti-oxides (e.g. Mg, Ca, Fe, Ni, Sc and Cr) decrease with decreasing MgO (Fig. 11). TiO₂, P₂O₅ and Sr define maxima at about 6, 4·5 and 4·5 wt % MgO, respectively (Fig. 11). Also the agpaitic index [atomic ratio (Na + K)/Al] increases strongly with decreasing MgO, reaching values of >= 1·0 in the most evolved rocks; these rocks are therefore peralkaline. It should be noticed, however, that mafic aphyric to near-aphyric rocks with similar MgO contents show considerable differences in the concentrations of minor and trace elements (Fig. 11). The majority of the aphyric and near-aphyric rocks define relatively smooth, parallel trace element patterns with the highest primitive-mantle-normalized ratios for Ba, Ta, Nb and La, and decreasing degree of enrichment from La to Lu (Fig. 12). The rocks most enriched in Cs, Rb, Th-La, Zr and Hf exhibit negative anomalies for Sr, P, Ti and Ba, and near-constant concentrations of intermediate and heavy REE. However, it is noteworthy that among the aphyric and near-aphyric samples, pronounced negative Sr, P and Ti anomalies are found among mafic as well as among evolved samples (Fig. 12).

Figure 12. Representative trace element patterns for aphyric and near-aphyric lavas (<5% phenocrysts) of mafic to intermediate composition, normalized to primitive mantle (PM; Sun & McDonough, 1989). (See text for comments.)

TEMPERATURES AND PRESSURES OF EQUILIBRATION

Temperature estimates

Temperature estimates are based on clinopyroxene-liquid equilibria (Putirka, 1997), olivine-liquid equilibria (Leeman, 1978; Putirka, 1997) and multivariate statistical analyses of clinopyroxene compositions (Soesoo, 1997). For aphyric and near-aphyric samples whole-rock compositions have been assumed to be equivalent to `liquid' compositions. For porphyritic samples the groundmass compositions were estimated by correcting whole-rock compositions for their phenocryst assemblages (using point counting and microprobe data). Samples for which the estimated groundmass compositions did not give an (Mg/Fe²⁺)_groundmass/(Mg/Fe²⁺)_ol ratio of 0·30 ± 0·02 (assuming Fe²⁺/Fe_total = 0·75) were not used for temperature estimates.

The cpx-liq geothermometer of Putirka, (1997) is based on the jadeite-diopside/hedenbergite exchange between clinopyroxene and liquid. It has long been known that the solubility of the jadeite component in clinopyroxenes increases with increasing pressure and temperature (e.g. Kushiro, 1962; Thompson, 1974). However, the concentration of Na₂O in clinopyroxene is not a P-T indicator in itself. Sodium also combines with Fe³⁺ to form the acmite component which is common in many clinopyroxenes formed or equilibrated at low pressures. The Na-salite cores of clinopyroxene phenocrysts in Tenerife lavas contain up to 1·9 wt % Na₂O. In the Na-salites the atomic proportion of Fe³⁺ is generally higher than that of Na, implying that all, or most, sodium goes to form acmite. Consequently, the high sodium content exhibited by the green cores cannot be taken as an indication of formation at high temperature or pressure. As the geothermometer of Putirka, (1997) assigns all Na to the jadeite component, samples rich in Fe³⁺ were not used for P-T determinations.

Estimated temperatures lie in the range 1040-1260°C (Table 6). In general there is good agreement between the Putirka, (1997) cpx-liq and ol-liq temperatures, whereas temperatures estimated by the Leeman, (1978) and Soesoo, (1997) geothermometers give systematically lower values (Table 6). Also the high Ti contents of the kaersutites (0·63-0·77 cations Ti per 23 oxygens; Table 4) indicate crystallization at high temperatures (>1050°C; Helz, 1982).

Table 6. Estimated temperatures and pressures of crystallization, obtained by geothermometers and geobarometers by Leeman (1978), Grove et al. (1989), Putirka (1997) and Soesoo (1997)

Pressure estimates based on mineral chemistry

Pressures (Table 6) were estimated on the basis of ol-cpx-liquid equilibria (Grove et al., 1989), cpx-liquid equilibria [Putirka, (1997) and references therein], and multivariate analyses of clinopyroxene (Soesoo, 1997). The geobarometer of Nimis, (1995), based on clinopyroxene crystal-structure modelling, gave mostly negative values, and therefore the results are not presented in Table 6. We suspect that this barometer is not adequately calibrated for low pressures, although this is not mentioned in the paper by Nimis.

The geobarometer of Grove et al., (1989) gave a pressure range of 0·4-1·0 GPa, with an average of 0·7 GPa (Table 6). Similar results (0·3-1·1 GPa, with an average of 0·9 GPa) were obtained by the geobarometer of Putirka, (1997), whereas that of Soesoo, (1997) indicates pressures below 0·5 GPa for most of the samples. Single samples rarely show agreement between the results obtained by different geobarometers. We have also used the computer program MELTS (Ghiorso & Sack, 1995) to test the liquidus assemblages of Tenerife lavas at different pressures, assuming temperature-oxygen fugacity relations between those defined by the quartz-fayalite-magnetite (QFM) and the nickel-nickel oxide (NNO) buffers, and 0·3 wt % H₂O. The choice of H₂O concentration was based on ion probe analyses of basaltic glass inclusions in phenocrysts. The results indicate that melts of compositions corresponding to aphyric and near-aphyric Tenerife lavas are only in equilibrium with their observed phenocryst assemblages within a pressure range of about 0·2-0·4 GPa. This supports the low pressures indicated by the Soesoo, (1997) geobarometer. To obtain more reliable estimates, pressures of crystallization were also estimated by combining fluid inclusion densities with temperature, as described below.

Fluid inclusion compositions and densities

Fluid inclusions in phenocrysts in the lavas proved to be too rare and small for reliable density measurements. However, some gabbroic and nepheline syenite xenoliths are rather rich in CO₂ inclusions. These xenoliths have mineral and trace element characteristics which indicate close association with the Tenerife volcanism. Assuming that the gabbroic and syenitic aggregates may give general information about the depths of magma chambers formed during the Tenerife magmatism, we chose one gabbroic and two syenitic xenoliths for fluid inclusion studies.

Upon cooling to temperatures below about -100°C the fluid inclusions re-equilibrated to a solid CO₂ + vapour assemblage. On heating, the solid CO₂ melted within experimental error of the triple point of pure CO₂ (-56·6°C). On further heating, the liquid + vapour phase assemblage of the inclusions homogenized to liquid below the critical point of pure CO₂ (+31·4°C). Homogenization temperatures (T_h) for CO₂ fluid inclusions in the xenoliths are given in Fig. 13a. There are no significant differences between homogenization temperatures of inclusions in clinopyroxene and plagioclase. The xenoliths gave significantly different homogenization temperatures (Fig. 13a) of +20·7 to +22°C for the gabbro, and +26·1 to +28°C for the syenites, corresponding to molar volumes of 59·1 cm³/mol and 68·6 cm³/mol (or densities of 0·74 g/cm³ and 0·64 g/cm³, respectively). Isochores calculated on the basis of the modified Redlich-Kwong equation of state of Bottinga & Richet, (1981) are shown in Fig. 13b. Using the obtained temperature range of 1040-1260°C (Table 6), the gabbro isochore gives a pressure range of about 0·4-0·5 GPa. The lower densities of CO₂ inclusions in the syenite xenoliths, which have clearly also formed at lower temperatures than the gabbros, suggest pressures of 0·2-0·3 GPa. The fluid inclusion data thus support the estimates by the Soesoo, (1997) geobarometer and the computer program MELTS (Ghiorso & Sack, 1995) of crystallization at low pressures. A pressure range of roughly 0·2-0·5 GPa, or depths of 5-14 km, seems most likely.

Figure 13. (a) Homogenization temperatures for CO₂ inclusions in olivine in one gabbro and two syenite xenoliths from Tenerife. (b) Isochores for gabbro and syenites calculated on the basis of the modified Redlich-Kwong equation of state of Bottinga & Richet, (1981). The gabbro isochore combined with the total temperature range of 1080-1240°C obtained by geothermometry on the basalts (Table 6) suggests a pressure range of about 0·4-0·5 GPa.

DISCUSSION

Fractional crystallization

Evidence from major element chemistry

The geochemical data presented above clearly imply that the magmas which gave rise to the Tenerife lavas were subjected to a complex combination of processes, one of which was fractional crystallization. The lavas and dykes show the following features consistent with removal of ol and ol + cpx from the most mafic magmas, followed by ol + cpx + mt, and ± ol + cpx + mt + plag + ap from the more evolved magmas (olivine appears not to have been on the liquidus in the most evolved melts; Figs 11 and 12): (1) general trends of increasing concentrations of lithophile elements such as K₂O, Rb, Th, Nb Ta and REE, and decreasing concentrations of compatible elements such as Ni, Sc and Cr, with decreasing MgO (removal of ol ± cpx; (Fig. 11); (2) increasing agpaitic index with decreasing MgO (removal of plag); (3) transition from incompatible to compatible behaviour of TiO₂, P₂O₅ and Sr as a result of eventual removal of magnetite, apatite and plagioclase, respectively, with decreasing MgO and increasing concentrations of strongly incompatible elements (e.g. Th, Nb, Ta and Zr); (4) transition from strongly incompatible to mildly incompatible behaviour of REE with increasing enrichment in strongly incompatible elements (e.g. Th, Nb, Ta and Zr). This is most probably caused by compatible behaviour of REE in apatite (e.g. mineral-melt partition coefficients of 10-100; e.g. Wörner et al., 1983; Fujimaki, 1986); removal of apatite will therefore significantly decrease the enrichment of REE relative to Cs, Rb, Th-La, Zr and Hf in the residual melt.

Evidence from clinopyroxenes and kaersutite

Fractional crystallization is also indicated by clinopyroxene chemistry. The analysed clinopyroxenes show wide ranges in concentrations of lithophile elements, strong positive correlations between La, Y and Sm, strong negative correlations for Sr_N/Sr_N*-Sm and Ti_N/Ti_N*-Sm, and a rough trend of decreasing Sc and V with increasing Sm (Table 2, Figs 7, 8 and 14).

Figure 14. Concentrations in selected lithophile elements, Sr_N/Sr_N* and Ti_N/Ti_N* (defined in the text) in clinopyroxenes in Tenerife lavas, plotted against ppm Sm and mg-number. Na-sal, Na-salite; Al-sal, Al-salite. The figure shows strong covariations between REE, Y, Sr_N/Sr_N* and Ti_N/Ti_N*, but only very weak covariations with mg-number. Different evolutionary trends in different samples are suggested by dotted lines.

The trace element composition of a magmatic clinopyroxene is the product of the trace element content of its host melt, and cpx-melt partitioning. Partition coefficients clinopyroxene-silicate melt (D^cpx-melt) for REE, Sr and Ti have been found to increase with increasing (SiO₂)_melt or degree of polymerization (e.g. Ryerson & Hess, 1978; Mahood & Hildreth, 1983; Wörner et al., 1983; Green & Pearson, 1985; Nash & Crecraft, 1985; Sisson, 1991; Green, 1994), increasing mole fraction of Al in the melt (Gallahan & Nielsen, 1992), falling temperature (e.g. Green & Pearson, 1985; Gallahan & Nielsen, 1992), and decreasing pressure within the range 0·5-3 GPa (Adam & Green, 1994). A typical relationship between D values and degree of polymerization of the melt [R = O/(Si + Ti + P + Al*), where Al* is aluminium stabilized in tetrahedral position by combination with Na and K] is shown in Fig. 15. The Al-salites in samples TF58, TF67, TF68, TF86, and TF114 crystallized at similar temperatures (1150-1200°C) and low pressures (Table 6). Judging from the groundmass chemistry of these samples (whole-rock compositions corrected for phenocrysts), the melts which gave rise to these rocks were depolymerized, with R values within the limited range 2·80-3·06. Addition of H₂O or CO₂ causes depolymerization or polymerization, respectively, of the melt. We do not know the H₂O and CO₂ contents of the different melts. However, judging from the presence of CO₂ inclusions in mineral in basalts and gabbroic xenoliths and the presence of hydrous minerals in the gabbroic rocks from Tenerife, it seems likely that the magmas contained both H₂O and CO₂, so that some of the effects of the two species cancelled each other. Small deviations (<0·2) from the estimated R values will not siginificantly affect the partition coefficients (Fig. 15). On this basis it seems reasonable to assume that for most elements Al-salite-melt partition coefficients were similar for these samples. Similar cpx-melt partition coefficients for the Al-salites imply that their different trace element concentrations and Sr_N/Sr_N* and Ti_N/Ti_N* ratios reflect corresponding differences among the melts from which they formed.

Figure 15. Clinopyroxene-melt partition coefficients for Sm plotted against degree of polymerization of the host melt (R; see text for definition). Based on data from Onuma et al., (1968, , 1981), Grutzeck et al., (1974), Matsui et al., (1977), Nicholls & Harris, (1980), Dostal et al., (1983), Mahood & Hildreth, (1983), Wörner et al., (1983), Fujimaki et al., (1984), Green & Pearson, (1985), Nash & Crecraft, (1985), McKay et al., (1986), Francalanci, (1989), Sisson, (1991) and Foley et al., (1996). Grey fields show R values for samples discussed in the text.

The Na-salites, in contrast, appear to have formed from evolved, highly polymerized melts (the aphyric lava TF126 gives R = 2·12), and to have crystallized at lower temperatures than the Al-salites. This implies significantly higher partition coefficients and large variations with small changes in degree of polymerization (R) for most trace elements in the Na-salite-melt systems (Fig. 15). The higher concentrations of La, Sm, etc. in Na-salites than in Al-salites therefore do not necessarily reflect more extensive crystallization, but may be the result of higher partition coefficients. However, the well-defined trends exhibited by the Sr_N/Sr_N*-Sm and Ti_N/Ti_N*-Sm plots strongly suggest that partition coefficients cpx-melt for REE, Sr and Ti have increased at roughly similar rates from the Al-salite to the Na-salite systems, implying that the Na-salites formed from evolved magmas which have been subjected to extensive removal of plagioclase and Fe-Ti-oxides and other phases. There is also a tendency for the pyroxenes in different samples to define separate pyroxene trends (see the La-Sm and Y-Sm plots in Fig. 14) common to both Al-salites and Na-salites in that sample. This suggests a genetic relationship between Na-salites and Al-salites in the same samples.

The kaersutite analysed for trace elements (sample TF126) shows a pronounced negative Sr anomaly (Table 4, Fig. 9). Kaersutite in basaltic melts in mantle xenoliths from La Palma (Wulff-Pedersen et al., 1996), in contrast, shows flat trace element patterns through Sr (Fig. 9), suggesting similar kaersutite-melt partition coefficients for Sr and light to intermediate REE. The marked negative Sr anomalies exhibited by kaersutite in sample TF126 are therefore interpreted as the result of crystallization from a melt depleted in Sr relative to REE; such a depletion is probably a consequence of removal of a plagioclase-bearing mineral assemblage.

However, rocks and minerals also exhibit compositional relations which cannot be accounted for by fractional crystallization alone. Aphyric and near-aphyric Tenerife rocks with similar MgO contents show significant differences in concentrations of strongly incompatible elements (Fig. 11) which are not related to age or geographic position. Similarly, the clinopyroxenes show only very weak tendencies for covariation between trace element concentrations and ratios (e.g. REE, Sr_N/Sr_N* and Ti_N/Ti_N*), and mg-number (Fig. 14). These relations imply the operation of processes which cause decoupling between major and trace elements.

Magma mixing

The rocks of this study present considerable evidence of mineral-melt disequilibria and dynamic processes which are interpreted as the results of magma mixing. The evidence includes: (1) corrosion and fragmentation of many phenocrysts (Fig. 3e); (2) corroded crystals of alkali feldspar with reaction rims in some of the rocks; (3) corroded crystals of olivine with reaction rims in highly evolved lavas (Fig. 3b); (4) K-poor calcic plagioclase rimmed by K-rich, sodic feldspar; (5) significant ranges in concentrations of Al and Ti in olivine phenocrysts with similar Fo contents at the scale of a thin section (Fig. 10); (6) inverse zoning in the clinopyroxenes (increasing mg-number, Ni, Sc, V, Sr_N/Sr_N* and/or Ti_N/Ti_N*, and decreasing TiO₂, Al₂O₃, REE and Zr, Table 2, Figs 5, 6 and 8), which clearly results from mixing between mafic melts relatively depleted in lithophile elements and evolved melts highly enriched in lithophile elements; (7) the presence of euhedral microphenocrysts of both Na-salite and Al-salite in phonolite TF126.

The corroded cores of Na-salite (Figs 3c, d, 5 and 6) are clearly out of equilibrium with their host magmas and represent crystals `accidentally' introduced into the magmas before formation of the Al-salite overgrowths. The Na-salite cores may have been added through magma mixing or represent xenocrysts extracted from the wall-rocks during magma ascent. A mantle origin seems out of the question. As shown above, all Na in the Na-salites is balanced by Fe³⁺ in the acmite molecule (Table 2), not in the jadeite molecule as would be expected by clinopyroxenes formed at high pressures. Furthermore, the Na-salites tend towards relatively low mg-number values (79-40; Fig. 4) and have trace element compositions typical of pyroxenes formed from highly evolved melts (e.g. high REE and Y concentrations, strong depletion in Sr and Ti relative to REE, low concentrations of Cr₂O₃, Sc, V and Co; Table 2, Figs 4, 7 and 14). Their LREE-HREE enriched trace element patterns (Fig. 7) also preclude affinity to the N-MORB rocks which make up the old ocean crust on which the Canary Islands are built. It therefore seems highly unlikely that the Na-salites represent xenocrysts in the sense of being derived from a pre-existing solid rock. Formation from highly evolved melts is supported by Fig. 16, which shows that the Na-salites occupy an intermediate position between the trends of Al-salites and the trend defined by Na-rich, Ti-Al-poor clinopyroxenes typical of phonolitic and nepheline syenitic rocks in Tenerife (Scott, 1976; Wolff, 1987; J. A. Wolff, personal communication, 1997). Finally, the tendency for Al-salites and Na-salites in the same rock to fall on a common trace element trend (Fig. 14) suggests some kind of genetic relationship, such as formation in the same magma chamber. Altogether it seems most likely that the Na-salites were introduced into the mafic magmas through magma mixing, with the Al-salites forming from the mixed magmas. The common presence of Na-salite cores in clinopyroxenes in basaltic Tenerife lavas thus emphasizes the importance of this mechanism in the plumbing system.

Figure 16. The compositional ranges of Al-salites and Na-salites of this study compared with that of Na-rich clinopyroxenes in phonolitic and nepheline syenitic rocks in Tenerife [data from Scott, (1976), Wolff, (1987) and J. A. Wolff, personal communication (1997)].

In addition to the evidence presented here, mixed mafic and evolved magmas, including pieces containing both mafic and evolved magmas, are seen in many scoria deposits in Tenerife (e.g. Wolff, 1985; Araña et al., 1994; Ablay & Martí, 1995; Ablay et al., 1995a, 1995b; Bryan, 1995; Martí et al., 1995a). Furthermore, the stratigraphy of Tenerife shows continuous alternation between mafic and evolved magmas from the oldest to the youngest parts of the island, implying that both evolved and mafic magmas have been more or less continuously available (e.g. Wolff, 1985; Ancochea et al., 1990; Martí et al., 1990, , 1994, 1995a, 1995b; Ablay & Martí, 1995).

Fractional crystallization of periodically refilled magma chambers (FCM)

If we accept that the Na-salites formed from highly evolved Tenerife melts, some of the zoned pyroxenes (e.g. cpx in sample TF86; Figs 6 and 8) have recorded (at least) four cycles of fractional crystallization, interrupted by episodes of mixing between evolved melts and batches of mafic melts injected into the magma chamber. Such processes are referred to below as FCM processes (fractional crystallization, mixing). Evidence of FCM processes is apparent in clinopyroxenes in basaltic lavas belonging to the `Basaltic Shields' (TF67, TF68, TF86, TF114), as well as in younger lavas (TF58, TF126). O'Hara, (1977) showed that fractional crystallization in periodically refilled magma chambers (FCM processes) leads to decoupling between major and trace elements; for example, strong enrichment in incompatible elements with only minor decreases in mg-number, as exhibited by the clinopyroxenes of this study (Fig. 14). Further evidence is presented in Fig. 17, which shows concentrations of Nb in Tenerife lavas (corrected for phenocryst contents to simulate melt concentrations) plotted against Fo_ol (rim compositions), compared with expected Nb-Fo_ol relations in melts evolving through fractional crystallization (FC), equilibrium crystallization (EC), and FCM processes. Nb was used as an example of a trace element which may be assumed to be strongly incompatible (bulk distribution coefficient 1) in magmas with ol ± cpx ± plag ± mt ± ap on the liquidus. The Mg/Fe ratio in a melt is expressed as Fo content in rims of olivine in equilibrium with the groundmass. For aphyric and near-aphyric lavas without olivine, a hypothetical olivine composition is estimated on the basis of (Fe₂O₃)_total and MgO concentrations in the lava, assuming Fe³⁺/Fe_total = 0·25, and (Mg/Fe²⁺)_melt/(Mg/Fe²⁺)_ol = 0·3. The procedure for FCM calculations was outlined by O'Hara, (1977). In our calculations, the first FCM cycle starts with a batch of injected, mafic initial melt, l_o (35 ppm Nb, Fo_ol = 0·91), which is subjected to 55% fractional crystallization (F = 0·45), forming an evolved, residual melt, l_e¹. The next cycle starts with mixing between a new batch of injected melt (l_o) and evolved melt (l_e^l) in the proportion 0·6/0·4, to form mixed melt l_m², which is subsequently subjected to 55% fractional crystallization, giving rise to evolved melt l_e², etc. At the start of each new cycle (cycle i) the evolved melt (l_e^{i - 1}) formed at the end of the preceding cycle (cycle i - 1) mixes with an injected batch of l_o melt in the proportion 0·6/0·4, and subsequently undergoes 55% fractional crystallization. We have used bulk distribution coefficients of 1·1 for Fe²⁺, 3·1 for Mg and 0·01 for Nb.

Figure 17. (a) Estimated Nb_melt/Nb_o (where Nb_o is concentration in initial melt) plotted against the forsterite contents in olivine in equilibrium with the melts, in melts evolving through fractional crystallization (FC), equilibrium crystallization (EC), and mixing between primitive (l_o) and evolved melt in different proportions. The grey field shows the range covered by the analysed lavas. Nb_melt is analysed Nb corrected for phenocryst content. We assume an initial melt (l_o) with 11·0 wt % MgO, 7·2 wt % Fe₂O₃, in equilibrium with olivine Fo_91·0, and Nb_o = 35 ppm. Numbers indicate mass fraction of residual melt (F). We have used mineral-melt bulk distribution coefficients of 1·1, 3·1 and 0·01 for FeO, MgO and Nb, respectively. (b) Nb_melt/Nb_o (corrected for phenocryst content) vs Fo content in rims of olivine phenocrysts in volcanics in Tenerife, compared with the geochemical evolution expected as the result of fractional crystallization in a periodically refilled magma chamber (FCM processes), using l_o as the composition of eachbatch of mafic magma injected into the magma chamber, and 0·55% fractional crystallization (F = 0·45) in each cycle. Consecutive crystallization cycles are indicated by dotted lines and cycle numbers. At the beginning of each cycle (starting with cycle 2) mafic melt (l_o) mixes with evolved melt in the ratio 60:40. (c) Changes in Ba/Th ratio induced by FCM processes, using the model described for (b), assuming 220 ppm Ba and 2 ppm Th in melt l_o. Changes in melt composition during each cycle were calculated from the MELTS program of Ghiorso & Sack, (1995). Data on mineral-melt partitioning (D) from Onuma et al., (1968, , 1981), Shimizu, (1974), Drake & Weill, (1975), Matsui et al., (1977), Dostal et al., (1983), Mahood & Hildreth, (1983), Wörner et al., (1983), Fujimaki et al., (1984), Nash & Crecraft, (1985), Green & Pearson, (1985), Yurimoto & Sueno, (1987), Francalanci, (1989), Sisson, (1991), Hart & Dunn, (1993), Beattie, (1994), Horn et al., (1994), Nielsen et al., (1994), Terakado & Fujitane, (1995) and Foley et al., (1996) were used to draw regression curves for changes in partition coefficients increasing degree of polymerization, as shown in Fig. 15. Using these regression curves, the mineral-melt partition coefficient for each mineral was changed in several steps during each cycle, following changes in the degree of polymerization of the melt (Ba: 0·000002-0·0001 for ol, 0·0012-0·005 for cpx, 0·2-1·5 for plag, 0·01-0·03 for mt; Th: 0·00015 for ol, 0·01-0·04 for cpx, 0·001-0·01 for plag, 0·01-0·05 for mt). Average Ba/Th ratios given by Weaver, (1991) for N-MORB, primordial mantle (PM), and ocean islands with HIMU and EM I signatures, are shown for comparison. [circles], aphyric lavas; [diamonds], porphyric lavas. (See text for further explanation and discussion.)

Figure 17 shows that after the first few cycles, each new mixed melt (l_mⁱ) will have approximately the same mg-number (and be in equilibrium with olivine of the same composition) as the mixed melt of the preceding cycle (l_m^{i - 1}), whereas the concentrations in incompatible elements will become progressively higher. Melts extruded at different stages of the FCM process will therefore show a complete lack of correlation between Nb and mg-number in the melt (or between Nb_melt and the Fo content in olivine phenocrysts). Figure 17b shows a wide range in Nb concentrations with no relation to the Fo content in olivine phenocrysts (or mg-number in the melt) among lavas and dykes in Tenerife, consistent with FCM processes. FCM processes will lead to different degrees of enrichment for incompatible elements with slightly different bulk distribution coefficients, resulting in progressive fractionation between such elements. Such changes may occur for trace elements whose ratios in mafic lavas are generally taken to be inherited from the mantle source (e.g. Ba/Th, Nb/Zr). An important consequence of fractional crystallization in periodically refilled magma chambers is that incompatible element ratios significantly different from those of the primary magmas may be encountered in lavas with high mg-number values (containing highly magnesian olivine) formed through mixing at the onset of each FCM cycle (Fig. 17c). Such processes may thus induce significant changes in trace element ratios without changing the radiogenic isotope ratios. Furthermore, mixed melts without plagioclase or titanomagnetite phenocrysts may show negative Sr and/or Ti anomalies, inherited from residual melts formed during previous cycles of fractional crystallization involving removal of these phases. The trace element characteristics of mafic lavas should therefore be used with caution as a source of information about their mantle source unless their evolutionary history is well known.

FCM processes may account for the range in compositions exhibited by the Na-salites (Figs 4 and 16). Sodium and potassium are incompatible elements which will be progressively enriched in the residual magma, whereas aluminium and titanium will be removed with crystallizing plagioclase and titanomagnetite. The mixed melt, l_mⁱ, at the beginning of cycle i is therefore expected to have a higher agpaitic index and be depleted in Ti relative to the mixed melt l_m^{i - 1} at the start of cycle i - 1. Clinopyroxene forming from melt l_mⁱ will consequently be richer in Na, and poorer in Al and Ti than clinopyroxene formed from melt l_m^{i - 1}. This is expected to give rise to a series of trends between those of Al-salite and phonolite-nepheline syenite in Fig. 16.

Figure 17 predicts only the chemical evolution in a single magma chamber where the FCM processes follow the strict pattern outlined above, whereas the lavas and dykes of this study were emplaced in different parts of Tenerife over a long period of time, and clearly formed from a large number of independent magma chambers. FCM processes governed by different factors in separate magma chambers (e.g. different degrees of fractional crystallization, different mixing proportions, fractionation processes at different depths) will give rise to additional compositional complexity. Chemical diversity may also arise from chemical differences among the primitive magmas injected into the magma chambers. Such differences are indicated by the fact that mafic Tenerife lavas show a larger range in incompatible element ratios (e.g. Ba/Th; Fig. 17c) than predicted by the FCM model. Small but significant variations in Sr-Nd-Pb isotope ratios have also been found to exist among mafic Tenerife lavas of different age and geographic location (Seim, 1996; Simonsen et al., 1998). Sr-Nb-Pb isotopic data and their relations to trace elements in the Tenerife lavas are the topics of a separate paper.

Implications for crustal structure

Estimates suggest that the FCM processes took place at pressures of about 0·2-0·5 GPa, or depths of 5-14 km. Crystallization at elevated pressures is supported by the presence of kaersutite and K-poor plagioclase. A general feature of amphibole stability in mafic to intermediate magmas is that at pressures below ~0·5 GPa the amphibole-in curve has a positive P-T slope, and crosses the solidus at low pressure (e.g. Helz, 1982). The liquidus assemblage will thus change with decreasing pressure. Experimental studies by Huckenholz & Gilbert, (1984) and Kunzmann, (1998) imply that melts of basanitic to phonolitic compositions with P_H2O < P_total have clinopyroxene, TiO₂-rich amphibole, phlogopite, olivine and apatite on, or near, the liquidus at pressures of 0·7-2·5 GPa, whereas liquidus and near-liquidus phases at pressures <0·3 GPa are clinopyroxene, plagioclase, olivine, titanomagnetite and apatite. Amphibole crystallizing from an H₂O-undersaturated basanitic to phonolitic magma at higher pressures becomes unstable during ascent, and typically melts incongruently to a mixture of rhönite + cpx + spin-mt + ol (Kunzmann et al., 1986; Huckenholz & Kunzmann, 1988). However, under suitable H₂O-CO₂ and f_O2 conditions amphibole may coexist with an alkali basaltic magma at pressures to <0·2 GPa (Holloway & Burnham, 1972). The reaction rims on kaersutites are believed to result from decompression during ascent. The stability of kaersutite in basaltic to intermediate Tenerife magmas at elevated pressures is supported by the common presence of kaersutite in plagioclase-free pyroxenite cumulate xenoliths from Tenerife (Borley et al., 1971; E.-R. Neumann, unpublished data, 1997). Kaersutite is also present in basalt veins in gabbro and peridotite xenoliths from La Palma (Wulff-Pedersen et al., 1996; Sørensen, 1998). Furthermore, Ablay et al., (1998) found kaersutite to be common in phono-tephritic to tephri-phonolitic lavas (Fo_<65) from Pico Viejo, but rare or absent in Pico Teide series. This difference was interpreted as the result of evolution of Pico Viejo magmas under conditions of higher P_H2O than Pico Teide magmas. The low Or contents of calcic plagioclase also suggest crystallization at somewhat elevated P_H2O. Calcic, low-Or plagioclase rimmed by relatively Or-rich, sodic plagioclase probably results from ascent of mafic magmas from deeper magma chambers to shallower ones where the mafic magma mixed with evolved ones.

The conclusion that magma chambers associated with the formation of Tenerife have been subjected to FCM processes rather than simple fractional crystallization has important implications with respect to the crustal structure. The proportion of a given volume of primary melt deposited in the crust as the result of crystallization processes is generally estimated on the basis of simple fractional crystallization. If we assume a primary magma in equilibrium with Fo₉₁, a cumulate making up 50% of the original volume will be reached in a residual melt in equilibrium with olivine of about Fo₇₁ (Fig. 17); that is, a rather evolved melt. However, an important result of FCM processes is that they drastically change the relation between proportion of cumulates formed and the mg-number in the residual melt. This is demonstrated in Fig. 18. At the beginning of each cycle the total mass of cumulate crystals (M_xtl) relative to the total mass of melt intruded into the magma chamber (M_melt) is

where we assume that identical batches of l_o melt are intruded into the magma chamber at the beginning of each crystallization cycle, and l_mⁱ is the mass of mixed melt present in the magma chamber at the beginning of each crystallization cycle.

where A is the ratio of primitive melt (l_o) relative to evolved melt (l_e) mixed at the beginning of each cycle. Figure 18 shows the result using the same parameters as in Fig. 17 (A = 0·6/0·4, F = 0·45).

Figure 18. Estimated mass ratio M_xtl/M_melt plotted against Fo content in olivine in equilibrium with the melt, for a magma undergoing fractional crystallization in a periodically refilled magma chamber (FCM processes, parameters as in Fig. 16). M_xtl, total mass of cumulates formed at a given stage; M_melt, total mass of magma emplaced into the magma chamber at that stage. FC, simple fractional crystallization. (See text for further explanation and discussion.)

The discussion presented above implies that most of the exposed lavas in Tenerife (including lavas belonging to the `Basaltic Shields') have a history of evolution in a complex magma plumbing system with magma chambers located at different depths between about 5 and 14 km, where large volumes of mafic to intermediate cumulates were deposited. A recent seismic study combined with gravity modelling presented by Watts et al., (1997) places the top and bottom of the old oceanic crust at about 8 and 15 km below sea-level, respectively. This means that magma chambers and intrusions associated with the Tenerife volcanism are located both in the downflexed old oceanic crust, and in the younger volcanic load that makes up the island. Figure 18 shows that already at the onset of the third cycle of fractional crystallization, the M_xtl/M_melt ratio may be near 0·5 in residual melts in equilibrium with olivine Fo₈₆. Such large relative masses of cumulates are in general only expected to be associated with rather Fe-rich residual melts (Fig. 18). The deposition of cumulates is expected to have caused thickening of the old oceanic crust. The crustal structure model of Watts et al., (1997) did not restrict the position of the Moho under Tenerife well enough to shed light on the degree of thickening. Watts et al., (1997) found evidence that underplating is significant under Hawaii and the Marquesas, but not under Tenerife. There is thus an apparent disagreement between the data of Watts et al., (1997) and the conclusions of this paper. However, our results suggest that the ocean island cumulates deposited under Tenerife are scattered throughout the crust, rather than forming massive underplating. Furthermore, many of these cumulates are expected to contain large proportions of plagioclase and have geophysical properties similar to those of MORB gabbros. It may therefore be difficult or impossible to differentiate ocean island cumulates and intrusives from MORB gabbros by geophysical methods.

Conclusions

(1) Clinopyroxene phenocrysts in Tenerife lavas of mafic to intermediate compositions typically exhibit alternating normal and reverse zoning. Up to four cycles of normal zoning, interrupted by reverse zoning, have been observed within a single crystal.

(2) The lavas show general trends of increasing concentrations in lithophile elements, and decreasing contents of elements compatible with olivine and clinopyroxene with decreasing MgO, with TiO₂, P₂O₅ and Sr defining maxima at about 6, 4·5 and 4·5 wt % MgO. However, for any given MgO content the rocks exhibit a wide range in concentrations of trace and minor elements.

(3) Both clinopyroxenes and lavas show a tendency for decoupling between major and trace elements.

(4) The observed compositional variations in clinopyroxenes and lavas are interpreted as the results of fractional crystallization (removal of ol + cpx -> ol + cpx + mt -> ± ol + cpx + mt + ap + plag), repeatedly interrupted by mixing between evolved melts and batches of more mafic melts (FCM processes); that is, fractional crystallization in periodically refilled magma chambers.

(5) The FCM processes took place in crustal magma chambers at temperatures of 1040-1260°C and pressures of about 0·2-0·5 GPa or depths of 5-14 km; that is, partly within the old oceanic crust and partly within the overlying sequence of Canary Islands lavas.

(6) Evidence of FCM processes is found in lavas from the `Basaltic Shields', as well as from the younger series, indicating the common occurrence of FCM processes in Tenerife.

(7) FCM processes will lead to progressively higher ranges in incompatible element concentrations for each cycle, whereas after the first cycle, the range in mg-number will stay essentially constant.

(8) An important aspect of FCM processes is that they will lead to progressive fractionation between incompatible elements with slightly different bulk distribution coefficients. Such changes may occur for trace element ratios, which, in mafic lavas, are generally taken to be inherited from the mantle source (e.g. Ba/Th, Nb/Zr).

(9) Another important result of FCM processes is that melts in equilibrium with Fo-rich olivine may have undergone repeated cycles of fractional crystallization and magma mixing, and may thus be associated with much larger masses of cumulates in the crust than expected from a simple fractional crystallization model.

ACKNOWLEDGEMENTS

This project was supported by the Commission of the European Communities, DGXII, Environment Programme, Climatology and Natural Hazards Unit, under Contract EV5V-CT-9283, and by grants from the Norwegian Research Council (NFR) and Comision Interministerial de Ciencia y Technología (CICYT AMB96-0498-C04). We thank the Instituto para la Conservación de la Naturaleza (ICONA) for permission to undertake this research. We are also grateful to Director Suzanne Y. O'Reilly for giving us liberal access to, and technical assistance with, the laser ICP-MS at GEMOC, School of Earth Sciences, Macquarie University, Sydney, Australia. Fieldwork in Tenerife was made very pleasant because of the accommodation at the Parador de Cañadas del Teide, and efforts of the friendly staff there. The paper has been improved through discussions with Drs John Wolff and Giray J. Ablay, and constructive criticism by Giray J. Ablay, Wendy A. Bohrson, Michael J. O'Hara and John A. Wolff of earlier versions of the manuscript.

REFERENCES

APPENDIX

Listed below are the sample locations and volume percent of phenocrysts (based on point counting) for basanitic to phonolitic lavas and dykes in Tenerife included in this study. The locations are grouped according to age and geographic position. As far as it has been possible to determine stratigraphic positions, the lavas within each group are listed after decreasing age. Phen, phenocrysts; ol, olivine; cpx, clinopyroxene; mt, titanomagnetite; plag, plagioclase; kaers, kaersutite; ap, apatite; il, ilmenite. Small amounts of phenocrysts are indicated by parentheses.

Roque del Conde (11.6-6.44 Ma; Ancochea et al., 1990)
TF102-1	N28°7·226[prime] W16°43·345[prime]. Lowermost exposed lava in lower part of Barranco del Infierno (right below power station). Very small ultramafic xenoliths. Phen: ol13·7
TF102-2	N28°7·226[prime] W16°43·345[prime]. Lava about 6 m above TF102-1, separated by flow breccias and two thin lava flows. Phen: ol.
TF99	N28°7·804[prime] W16°43·138[prime], 420 m a.s.l. (above sea-level). One of several thin flows, about 10 m below pyroclastic sequence. A few flows from top. Phen: ol0·2 cpx4·9 mt0·6 plag3·0 kaers0·1 ap0·1.
TF101	N28°7·693[prime] W16°43·286[prime], 300 m a.s.l. Lava (0·5-3 m thick) immediately below pyroclastics. Phen: ol0·1 cpx1·4 mt1·0 plag5·5 ap0·1.
TF97	N28°7·85[prime] W16°42·960[prime], 420 m a.s.l. Lava (12-20 m thick) overlying about 15-20 m of thick light pyroclastic units. Aphyric.
TF98	N28°7·755[prime] W16°43·266[prime]. Same as TF97, closer to Adeje. Aphyric.
TF100	N28°7·814[prime] W16°43·116[prime]. Dyke cutting lower basalt sequence and pyroclastics, but ending at the base of lava TF97. Phen: cpx + plag + mt ( + ol).
Teno Massif Road to Punta del Teno (~7·4-4·5 Ma; Ancochea et al., 1990)
TF85	Lowermost lava along western part of the road eastwards from Faro de Teno, in barranco where road crosses small bridge. Some pahoehoe structures. The grey basalt is overlain by a flow that looks like a poorly welded basaltic tuff with orange alteration surface. Phen: ol + plag ( + cpx).
TF91	Lowermost lava in the Teno massif along the eastern part of the road Casa Blanca-Faro de Teno. Difficult to ascertain stratigraphic position relative to sample TF85, but TF91 is probably younger. Phen: ol0·4 cpx14·0 mt0·4 plag2·2.
TF90	Rather massive lava flow near parking place below Punta de la Monja, stratigraphic position below TF89. Phen: ol + (cpx + mt).
TF89	Lava below TF88, west of Punta de la Monja. Phen: ol8·4 mt0·2.
TF88	Lava in small opening between tunnels. Difficult to see if this flow lies below or above TF87. Phen: ol4·7.
TF87	Uppermost lava flow along the road to Faro de Teno, about two bends west of westernmost tunnel entrance. Phen: plag.
TF86	Large dyke (~7 m wide) between localities TF85 and TF87. Contains small ultramafic xenoliths concentrated along the margins of the dyke. Phen: ol2·3 cpx9·8 mt0·2.
TF84	Faro de Teno, cinder cone at lighthouse. Probably younger than Teno massif sensu stricto. Collected large cinder pieces. Phen: ol2·3 cpx5·1 mt0·5 plag3·7.
Road from Buenavista to Santiago del Teide
TF92	Lava just below Las Palmas. Phen: ol2·4 cpx7·6mt1·3.
TF93	Lava flow a good distance further up the road. Near aphyric.
TF94	Lava flow slightly above TF93.
TF95	Lava flow at pass. Difficult to know the stratigraphic position relative to the lavas in the Masca area. Near aphyric.
TF96-2	Lava block in fossil breccia along road above Carrizal Alto. Phen: ol9·9 cpx6·6 mt0·1.
TF96-1	Lowermost lava flow along road above Carrizal Alto. This must represent a younger lava than TF96-2. Phen: ol20·7 cpx13·3 mt2·2 plag0·5.
TF70	N28°18[prime]28·3[prime][prime] W16°50[prime]22·4[prime][prime], 570 m a.s.l. Lava at lowest point along road past Masca, just before parking lot in gully, columnar jointing. Phen: cpx + plag.
TF69	N28°18[prime]12·6[prime][prime] W16°49[prime]58·9[prime][prime], 720 m a.s.l. About 2 m thick lava flow close to place with picnic arrangement outside road. Phen: cpx2·0 plag12·2.
TF68	N28°18[prime]16·3[prime][prime] W16°49[prime]47·8[prime][prime], 900 m a.s.l. Lava in sharp left turn. Phen: ol9·9 cpx6·6 mt0·1.
TF67	N28°18[prime]1·1[prime][prime] W16°49[prime]36·5[prime][prime], 990 m a.s.l. Phen: ol10·3 cpx38·0 mt0·1.
TF66	N28°18[prime]1·6[prime][prime] W16°49[prime]30·4[prime][prime]. Lava along road to Masca, at first passing place on the way down. Phen: ol + cpx.
TF1	Dyke with ultramafic xenoliths along road above Masca, between localitites TF68 and TF69.
Anaga Massif (5·8-3·6 Ma; Ancochea et al., 1990)
TF109	N28°30·645[prime] W16°11·365[prime]. Lava along road from Barrio de San Andres towards Puertito de Igueste. Phen: (ol + cpx).
TF111	N28°30·779[prime] W16°10·765[prime]. Lava or dyke. Near aphyric.
TF112	N28°31·063[prime] W16°10·555[prime], 90 m a.s.l. Lava in gully
TF113	N28°21·009[prime] W16°10·314[prime], 30 m a.s.l. Aphyric lava with xenoliths.
TF114	28°31·284[prime] W16°10·030[prime], 60 m a.s.l. Rounded boulder in breccia, very common lava type with many phenocrysts. Clearly relatively high stratigraphic position. Phen: ol12·1 cpx41·5 mt1·3.
TF115	N28°32·606[prime] W16°11·877[prime], 480 m a.s.l. Lava along road from Barrio de San Andres to El Bailadero, close to the latter. Phen: ol1·8 cpx2·5 mt0·2 plag0·1.
TF116	N28°32·673[prime] W16°13·109[prime], 690 m a.s.l. Lava stratigraphically above TF115 along road from El Bailadero to Las Casa de la Cumbre. Phen: ol4·7 cpx3·4 mt0·1.
TF4	N28°31·159[prime] W16°10·051[prime], 240 m a.s.l. Dyke with lots of pyroxenite and amphibole-rich xenoliths. Aphyric.
TF110	N28°30·666[prime] W16°11·174[prime], 60 m a.s.l. 8-10 m wide dyke. Phen: ol0·4 cpx14·0 mt0·4 plag2·2. Phen: cpx + mt ( + ol).
TF9-2	N28°31·472[prime] W16°17·275[prime], 660 m a.s.l. Mirador del Valle de Aguera. Dyke with minute xenoliths. Phen: ol1·0 mt0·3 plag2·1.
The Las Cañadas edifice Southeastern part of Cañadas caldera wall (about 3-2 Ma; Martí et al., 1994)
TF126	Topo de la Grieta; caldera floor below Valle Blanco. Aphyric.
TF125	Montón de Trigo, caldera wall, lower group, 2 m thick lava (age 2·0-3·4 Ma; Martí et al., 1994). Phen: ol0·4 cpx1·5 mt1·0 plag23·1 ap0·1.
TF127	La Angosturas, caldera wall, about 22 m above caldera floor. Sequence with white ignimbrite alternating with basaltic lavas and scoria. Fifth lava flow above ignimbrite (<1·5 Ma; Martí et al., 1994). Phen: cpx.
Western part of Las Cañadas edifice, along road to Chio (2·2- <1 million years; Fúster et al., 1994)
TF65	N28°13[prime]0·7[prime][prime] W16°47[prime]5·7[prime][prime], 510 m a.s.l. Lava NW of Guia de Isora. Probably same general age group as TF63, TF64. Phen: cpx0·3 mt0·6 plag0·7 kaers1·2.
TF64	N28°13[prime]22·2[prime][prime] W16°45[prime]21·7[prime][prime], 1020 m a.s.l. Lower of two lava flows right above km 23 mark. Upper one rather altered. Same general unit as TF63. Phen: ol1·0 mt0·3 plag2·1 kaerstr.
TF63	N28°14[prime]30·7[prime][prime] W16°45[prime]50·8[prime][prime], 1170 m a.s.l. Lava 100 m above km 20 mark. Near aphyric (cpx + mt + plag + kaers).
TF62	N28°16[prime]51·3[prime][prime] W16°46[prime]3·1[prime][prime], 1440 m a.s.l. Young-looking lava probably erupted from Mta Boca Cangrejo. Aphyric.
TF59-1	N28°15[prime]54·8[prime][prime] W16°44[prime]16·5[prime][prime], 1740 m a.s.l. Older of two (or more) different flows in 18-20 m high roadcut. Phen: cpx0·2 mt0·1 plag2·7 kaers0·7 aptr.
TF59-2	N28°15[prime]54·8[prime][prime] W16°44[prime]16·5[prime][prime], 1740 m a.s.l. Lava overlying TF59-1. Phen: cpx0·4 mt0·4 plag3·3 kaers0·5 aptr.
TF60	N28°18[prime]27·1[prime][prime] W16°45[prime]18·8[prime][prime], 1560 m a.s.l. Lava overlying TF27-2. Phen: ol3·2 cpx6·3 mt1·7 plag1·2.
Southern part of Las Cañadas edifice, along road between San Isidro and Vilaflor (1·9-0·1 Ma; Ancochea et al., 1990)
TF57	N28°5[prime]14·6[prime][prime] W16°36[prime]32·8[prime][prime], 450 m a.s.l. Road from Las Socas to San Miguel, 5-6 m thick lava flow between pyroclastic deposits.Probably younger than 1 Ma (500-600 ka??).Aphyric.
TF76	N28°5[prime]5·5[prime][prime] W16°34[prime]0·7[prime][prime], 330 m a.s.l. Ol-rich basalt just N of San Isidro. Phen: ol15·2 cpx0·9 mt 0·1 plag0·5.
TF58	N28°5[prime]33·1[prime][prime] W16°37[prime]56·9[prime][prime], 630 m a.s.l. Crossing between road C-822 (W of San Miguel) and road to Vilaflor. Phen: ol1·1 cpx0·1 mt0·1.
TF77	N28°5[prime]53·1[prime][prime] W16°34[prime]19·2[prime][prime], 390 m a.s.l. Higher stratigraphic position than TF76. Phen: ol (+ cpx + mt).
TF78	N28°6[prime]57·1[prime][prime] W16°35[prime]4·8[prime][prime], 720 m a.s.l. Above Granadilla. Phen: ol1·7 cpx0·1.
TF79	N28°8[prime]0·5[prime][prime] W16°36[prime]4·1[prime][prime], 810 m a.s.l. Near aphyric lava (plag + mt).
TF80	N28°7[prime]59·8[prime][prime] W16°36[prime]37·4[prime][prime], 1050 m a.s.l. Higher stratigraphic position than TF79. Aphyric lava.
TF81	N28°8[prime]7·1[prime][prime] W16°36[prime]58·3[prime][prime], 1140 m a.s.l. Aphyric lava.
TF82	N28°8[prime]11·4[prime][prime] W16°37[prime]3·6[prime][prime]. Aphyric lava.
TF83	N28°9[prime]7·3[prime][prime] W16°37[prime]57·8[prime][prime], 1290 m a.s.l. Just S of Vilaflor, below stone wall around fields. Near aphyric lava. Phen: (plag).
TF31	N28°12[prime]28·4[prime][prime] W16°40[prime]48·1[prime][prime], 2190 m a.s.l. Lava in roadcut along road to Vilaflor (near km 54 mark), in upper part of Boca Tauce Fm, just outside the Canadas caldera. Age >2·5 Ma. Phen: ol + cpx + plag + mt + il + ap.
Diego Hernandez Fm outside caldera (~540-170 ka; Martí et al., 1994)
TF48	N28°18[prime]22·4[prime][prime] W16°33[prime]5·8[prime][prime], 2070 m a.s.l. Lava in roadcut below the cone Cerrillal. Phen: ol12·2 cpx4·2 mt0·5.
TF50	N28°18[prime]17·7[prime] W16°30[prime]25·2[prime], 2100 m a.s.l. Very fresh lava flow overlying cinder and ash layers. Roadcut outside caldera to the E. Phen: ol0·4 cpx2·2 mt0·8 plag6·6 aptr.
TF51-2	N28°18[prime]35·0[prime] W16°30[prime]9·6[prime][prime], 2310 m a.s.l. 4-6 m thick lava. Phen: ol3·5 cpx7·2 mt1·3.
TF72-2	N28°15[prime]34·6[prime] W16°31[prime]19·4[prime][prime], 2040 m a.s.l. Lava flow at the top of the general sequence, E flank, outside caldera, probably belongs to Diego Hernandez Fm. Phen: ol0·1 cpx0·6 mt0·3 plag1·2.
TF49	N28°18[prime]20·3[prime][prime] W16°31[prime]54·3[prime][prime], 2340 m a.s.l. Rather oxidized bombs in cinder layer by the road below Mta El Cerrillal. Probably post-caldera. Phen: cpx + mt.
TF51-1	N28°18[prime]35·0[prime][prime] W16°30[prime]9·6[prime][prime], 2310 m a.s.l. Dyke 0·5 m thick. Phen: ol0·7 cpx0·8 mt0·4.
Diego Hernandez Fm, in caldera wall (542-170 ka; Martí et al., 1994)
TF124	Lava flow on the caldera floor (Diego Hernandez). Phen: ol11·5 cpx2·4 mt0·1.
TF018	N28°16[prime]15·1[prime][prime] W16°32[prime]55·7[prime][prime]. Basalt flow in W part of Diego Hernandez Fm, E of the Diego Hernandez cave. Older than TF41-2. Phen: ol2·3 cpx16·6 mt1·2 plag2·5.
TF38-1	N28°16[prime]41·1[prime][prime] W16°33[prime]7·2[prime][prime], 2190 m a.s.l. Lowest of three basaltic flows at base of Diego Hernandez Fm. Phen: ol4·5 cpx3·6 mt0·7 plag2·1.
TF38-2	N28°16[prime]41·1[prime][prime] W16°33[prime]7·2[prime][prime]. Middle flow. Phen: ol3·0 cpx0·7 mt0·6 plag0·3 kaerstr.
TF38-3	N28°16[prime]41·1[prime][prime] W16°33[prime]7·2[prime][prime]. Uppermost flow. Phen: ol0·6 cpx1·3 mt0·2 plag0·3.
TF41-2	N28°16[prime]24·7[prime][prime] W16°32[prime]56·6[prime][prime]. 2070 m a.s.l. Lowest part of thick lava sequence in middle part of Diego Hernandez Fm. Phen: ol0·5 cpx0·9 mt0·3 il0·1·
TF33-1	N28°16[prime]50·7[prime][prime] W16°33[prime]6·8[prime][prime], 2220 m a.s.l. Lowermost of two lavas in eastern caldera wall, W of fault. Very vesicular, mugearitic composition. Phen: ol0·6 cpx1·3 mt0·2 plag0·3.
TF33-2	N28°16[prime]50·7[prime][prime] W16°33[prime]6·8[prime][prime], 2222 m a.s.l. 2 m above TF33-1. Phen: cpx (+ mt + ap).
TF75-3	N28°16[prime]14·5[prime][prime] W16°32[prime]45·6[prime][prime], 2280 m a.s.l. Lower of three lava flows. Phen: ol1·9 cpx6·5.
TF75-2	N28°16[prime]14·5[prime][prime] W16°32[prime]45·6[prime][prime]. Flow above TF75-3. Phen: ol3·5 cpx0·3.
TF75-1	N28°16[prime]14·5[prime][prime] W16°32[prime]45·6[prime][prime]. Uppermost of the three lava flows, sampled 10-12 m above TF75-3. Phen: ol3·4 cpx0·3 mt0·1.
TF74	N28°16[prime]30·6[prime][prime] W16°32[prime]45·1[prime][prime], 2370 m a.s.l. Probably below TF73-1, slightly further S. Phen: ol0·3 cpx1·1 mt0·3 plag3·1.
TF73-1	N28°16[prime]39·1[prime][prime] W16°32[prime]48·1[prime][prime]. 2250 m a.s.l. Uppermost basaltic lava in the Diego Hernandez Fm. Phen: cpx + mt + plag.
TF47	N28°16[prime]55·6[prime][prime] W16°34[prime]23·3[prime][prime], 2220 m a.s.l. Basaltic lava near crater rim of Mta Mostaza. Phen: ol0·3 cpx1·1 mt0·3 plag3·1.
TF32	N28°16[prime]55·4[prime] W16°33[prime]10·4[prime][prime]. Dyke parallel to caldera wall in E part of Diego Hernandez Fm. Phen: ol6·3 cpx3·7.
TF36	N28°16[prime]48·3[prime][prime] W16°33[prime]8·2[prime][prime], 2100 m a.s.l. Small dyke, about 1 m wide. Phen: cpx0·9 mt0·2.
TF37	N28°16[prime]49·4[prime][prime] W16°33[prime]7·5[prime][prime]. Dyke cutting ignimbrites, about 2 m thick, but variable, possibly same as TF36. Phen: cpx1·8 mt0·5 ap0·1 kaerstr.
TF41-1	N28°16[prime]24·7[prime] W16°32[prime]56·6[prime][prime]. 2070 m a.s.l. Thick dyke cutting a thick sequence of lavas in the central part of Diego Hernandez Fm.
TF43	N28°16[prime]2·0[prime][prime] W16°33[prime]14·1[prime][prime], 2070 m a.s.l. Dyke close to red pyroclastic deposit W of Diego Hernandez Fm. Phen: ol + cpx (+ plag).
TF20	N28°16[prime]5·8[prime][prime] W16°32[prime]55·7[prime][prime]. Dyke standing up in a remarkable manner. Phen: ol2·3 cpx1·4.
TF22	Cinder cone along N part of caldera wall with Diego Hernandez Fm, Mta Colorada (or possibly Mta de las Arenas Negras).
Dorsal ridge Road from La Laguna towards Las Canadas (1·67-0·43 Ma; Ancochea et al., 1990)
TF117	N28°26·328[prime] W16°22·469[prime], 960 m a.s.l. Lava close to La Laguna. Phen: ol20·0 cpx2·1 mt0·4 plag0·2.
TF118	N28°25·874[prime] W16°22·885[prime], 1380 m a.s.l. Lava. Phen: ol4·0 cpx13·1 mt1·8.
TF119	N28°24·327[prime] W16°25·425[prime], 1620 m a.s.l. Lava. Phen: ol4·3.
TF120	N28°23·577[prime] W16°26·181[prime], 1740 m a.s.l. Lava. Phen: ol22·5 cpx18·2 mt0·6.
TF121	N28°23·058[prime] W16°27·356[prime], 1740 m a.s.l. Lava. Phen: ol2·6 mt0·3.
TF122	N28°21·831[prime] W16°27·915[prime], 1770 m a.s.l. Lava. Phen: ol17·2 cpx1·4 mt0·4.
TF123	N28°19·083[prime] W16°29·703[prime], 2190 m a.s.l. Lava. Phen: ol0·3 cpx1·1 mt0·3 plag3·1. Near aphyric (cpx + mt).
TF53-2	N28°20[prime]47·4[prime][prime] W16°28[prime]50·0[prime][prime], 2100 m a.s.l. Lava flow, 8-10 m thick. Phen: ol3·9 cpx9·3 mt3·8 plag13·8.
TF54	N28°21[prime]17·4[prime][prime] W16°28[prime]14·8[prime][prime], 2070 m a.s.l. Lava in a series of relatively thin lavas above TF53-2. Phen: ol1·8 cpx1·5 mt1·1 plag22·8.
TF53-1	N28°20[prime]47·4[prime][prime] W16°28[prime]50·0[prime][prime], 2100 m a.s.l. Dyke. Phen: ol1·8 cpx1·6 mt1·5 plag19·7.
Road NW from Güimar
TF103	N28°18·283[prime] W16°23·009[prime], 210 m a.s.l. Lava west of high-way at Güimar exit. Phen: ol1·6 cpx0·4.
TF105	N28°20·600[prime] W16°25·172[prime], 450 m a.s.l. Northern part of Arafo, widespread lava, massive part about 2 m thick. Phen: ol7·5 cpx11·1 mt0·3.
TF106	N28°21·444[prime] W16°24·890[prime]. 600 m a.s.l. Lava. Phen: ol6·3 cpx1·8 mt0·1.
TF107	N28°21·898[prime] W16°25·962[prime], 960 m a.s.l.Relatively thin flow (0·5-1·0 m), of unclear stratigraphic position. Phen: cpx0·2 mt0·7 kaers0·1.
TF108	N28°23·008[prime] W16°25·704[prime], 1260 m a.s.l. Lava. Phen: ol12·4 cpx3·1 mt0·1.
TF104	N28°19·589[prime] W16°23·961[prime]. 240 m a.s.l. Lava along road NNE of Güimar. Probably <150 ka. Phen: ol (+ cpx).
TF55	N28°23[prime]5·2[prime][prime] W16°25[prime]51·1[prime][prime], 1320 m a.s.l. Lava along road to Güimar. Phen: ol13·5 cpx7·9 mt0·2.
Historical lavas Outside Cañadas caldera (<300 years; J. Martí & J. Mitjavila, personal communication, 1994)
TF72-1	N28°15[prime]34·6[prime][prime] W16°31[prime]19·4[prime][prime], 2040 m a.s.l. Lava flow from Mta de Siete Fuentes (year 1704) sampled in a gulley. Phen: ol5·3 cpx3·9 mt0·4.
TF56	N28°19[prime]42·4[prime][prime] W16°25[prime]36·5[prime][prime], 570 m a.s.l. February 1705 flow from Volcan de Güimar or Mta de las Arenas. Sampled aboveGüimar. Phen: ol5·7 cpx8·1 mt1·7 plag0·2.
TF23-1	N28°17[prime]1·7[prime][prime] W16°30[prime]54·0[prime][prime], 2220 m a.s.l. Volcan de Fasnia (or Mta Roquillo), erupted in year 1705. Northernmost cone.
TF23-2	N28°17[prime]1·7[prime][prime] W16°30[prime]54·0[prime][prime], 2220 m a.s.l. Southernmost cone. Phen: ol + cpx (+ mt + plag).
TF61	N28°17[prime]20·3[prime][prime] W16°45[prime]36·6[prime][prime], 1500 m a.s.l. Lava flowing from Mta de Chinyero, 1909 eruption. Phen: cpx (+ ol + mt + plag).
TF027	Believed to be flow from Mta Chinyero, further from eruption centre.
Inside Cañadas caldera (<200 years; J. Martí & J. Mitjavila, personal communication, 1994)
TF45	Pico Viejo, east side, scoria cone. Identical composition to TF24. Aphyric.
TF24	Las Cañadas SW along road. Probably tongue of lava from Chahorra (the 1798 cone on the side of Pico Viejo. Minute gabbro xenoliths. Aphyric.
TF71	N28°13[prime]49·7[prime][prime] W16°41[prime]30·3[prime][prime]. 2310 m a.s.l. 1798 flow from Chahorra (cone on side of Pico Viejo), same flow as TF46, sampled along road. Aphyric.
TF46	Bomb from the Chahorra crater (1798 eruption). Near aphyric (cpx + mt + plag + ap + kaers).