Fluid metasomatism
MTIB
Subduction components
NTVZ SBS
KBS MHY
PCY KYSTTVG
N
TAIWAN TLS
Nb/La
Fig. 12. Nb/La vs Ba/Rb for the NTVZ volcanic rocks. Values of N-MORB (normal MORB), E-MORB and OIB are from Sun & McDonough (1989). Other data sources are the same as in Figs 8 and 9.
involvement of both subduction components in the meta-somatism (Table 4; Fig. 13).
The average composition of the continental sediments on Taiwan was chosen to represent that of the subducted terrigenous sediments (Lan et al., 1990) because of the active NW subduction of the Philippine Sea plate along the eastern margin of the Taiwan orogen. Compositions for the slab-derived fluid were taken from Tatsumi &
Kogiso (1997) and Ayers (1998). Using P-type (enriched) ETO as a starting material representing local enriched MORB mantle source unmodified by recent subduction metasomatism, mixing calculations show that all the
NTVZ volcanic rocks plot along a mixing trend between the P-ETO and TLS (Fig. 13). A fluid-dominated meta-somatism coupled with a 20---30% contribution from average terrigenous sediment is also shown in the TLS mantle source region that lies close to the mixing line between the average terrigenous sediment and the slab-derived fluid. The fluid-dominated metasomatism may have resulted in a phlogopite-bearing harzburgitic mantle (Wyllie & Sekine, 1982). Such a mantle has a substantially lower solidus temperature than refractory SCLM, and could be preferentially partially melted during the initial stages of thermal perturbation owing to extension.
0.5134 0.5129
0.5124 0.5119
0.1 1 10 100
143Nd/144Nd
0.5134 0.5129
0.5124 0.5119
143Nd/144Nd
20 40
60 80
E-MORB (P-ETO)
Slab-derived fluid
20 40
60 80
E-MORB (P-ETO) (a)
(b)
Ce/PbNb/U
NTVZ SBS
KBS MHY
PCY KYSTTVG
N
TAIWAN TLS
0.1 1 10 100
Terrigenous sediments
Slab-derived fluid
Terrigenous sediments
Fig. 13. Nb/U (a) and Ce/Pb (b) vs143Nd/144Nd for the NTVZ volcanic rocks. Data for enriched-type ETO (P-ETO) are from Sun et al. (1979), Sun (1980) and Jahn (1986). Small ticks with numbers on the mixing curve between the slab-derived fluid (Tatsumi & Kogiso, 1997; Ayers, 1998) and terrigenous sediment (Lan et al., 1990) indicate percentages of sedimentary components. Compositions used in mixing calculation are shown in Table 4. In comparison with the TLS lavas, which plot close to the mixing curve, most NTVZ rocks plot between the mixing curve and an E-MORB mantle source.
Other evidence for fluid-dominated metasomatism of the mantle source might come from generation of high-Al basalts in the NTVZ. High-Al basalts have been consid-ered to be primary melts from large degrees (50---60%) of partial melting of eclogites in subducting plates (e.g.
Brophy & Marsh, 1986; Myers et al., 1986). However, considering the difficulty of deriving large-degree partial melts from the old and cold subducting Philippine plate (45 Ma) in the Ryukyu subduction system, high-Al basalts in the NTVZ are most likely to be the result of fractional crystallization rather than representing primary melts. Alternatively, Gaetani et al. (1993) pointed out that the presence of water in the mantle wedge above a subduction zone can significantly alter the composi-tional path followed by residual liquids, and modify the types and compositions of crystals that accumulate at the site of cooling. Accordingly, under hydrous conditions, early crystallization of plagioclase is suppressed so that aluminium contents in residual liquids increase to yield high-Al basalts (Fig. 14a). Figure 14b shows that most of the NTVZ volcanic rocks lie along the olivine---clinopyroxene cotectic for hydrous crystallization. This suggests that their mantle source region might be hydrous. The SBS lavas, all high-Al basalts, plot closest to the olivine---clinopyroxene---plagioclase ternary eutec-tic, indicating that high-Al basalts in the NTVZ are not primary melts but have undergone some fractional crystallization.
In comparison with the MTIB in northwestern Taiwan erupted prior to the Taiwan arc---continent collision (Chen, 1990; Chung et al., 1995b), the NTVZ magmas with subduction-modified geochemical characteristics
suggest that the upper mantle domain underwent meta-somatism related to subduction zone processes during the buildup of the NTMB. Considering the tectonic evolu-tion of Taiwan region since arc---continent collision at
10 Ma (e.g. Teng, 1990), the Ryukyu subduction system is the only one that approached the northern Taiwan region and was most likely to cause the required metasomatism. Most NTVZ volcanic rocks having geologically meaningless negative Nd model ages may also result from this recent metasomatism.
The mixing of mantle source components and partial melting conditions
The mixing trend in Fig. 13 also suggests that mixing of partial melts of enriched asthenospheric mantle and sub-duction-metasomatized SCLM could explain the geo-chemical characteristics of the NTVZ volcanic rocks.
Although the MHY magmas have lower values than the P-ETO, they still plot at the other end of the mixing trend (Fig. 13a). The MHY mantle source might have been fluxed by slab-derived fluid but not sediment melts, so their Ce/Pb ratios are much lower (Fig. 13b). The effect from the slab-derived fluid is not clear for Nb/U prob-ably because the MHY magmas have higher Nb contents similar to E-MORB. A similar mixing trend is observed in the Nd---Sr isotope diagram (Fig. 8), in which the other NTVZ volcanic rocks lie along a mixing trend between the depleted MHY and enriched TLS mantle sources despite the elevated Sr isotopic ratios of the MHY mag-mas. A simple source sediment contamination model is not preferred because variable sediment flux in the
CPX60
OL60
SBS KBS
PCY TTVG KYS TLS
CPX
Pl Ol
CPX60
PL
Basalt parent Dry crystallization
Wet crystallization
Projected from silica (oxygen units)
(a) (b) MHY
Fig. 14. (a) Schematic liquid lines of descent for a primitive basalt under hydrous and anhydrous conditions. (b) Compositions of the NTVZ volcanic rocks. It should be noted that there are different eutectic points and paths of fractional crystallization for parental basaltic magmas under anhydrous and hydrous conditions. Diagrams are modified from Gaetani et al. (1993).
NTVZ mantle sources unrelated to NTVZ distribution would be highly unlikely to lead to systematic isotopic variation among the NTVZ. Alternatively, we propose that variable mixing of melts from the depleted astheno-spheric mantle at the site of formation of the MHY magmas, and the subduction-metasomatized SCLM where the TLS magmas formed could satisfy most of the geochemical characteristics of the NTVZ magmas.
The proportions of the melts from the metasomatized SCLM mixing with asthenospheric melts can be calcu-lated based on the NTVZ Nd---Sr isotopic ratios:
12---17% for the SBS, 20% for the KBS, 13% for the PCY, 30% for the TTVG and 20% for the KYS, respectively (Fig. 8), using compositions of the MHY and TLS magmas as the two end-members. In the calcula-tion, the most enriched Sr isotope value of E-MORB (87Sr/86Sr¼ 070308; Sun et al., 1979) instead of that of the MHY magmas was used because the MHY magmas have elevated Sr isotope ratios as a result of subduction-fluid flux. The degree of partial melting also plays an important role in controlling the spatial geochemical variation among the NTVZ, which the simple source sediment contamination model can not decipher.
The mantle xenoliths entrained within the MTIB are dominantly refractory harzburgites (Yang et al., 1987), so this part of the SCLM is not a productive magma source.
Whereas the fluid-dominated metasomatism substantially lowers the solidus temperature of the refractory SCLM, the TLS magmas, indicative of the most fluid-dominated metasomatized SCLM source, still show extremely low degrees of partial melting. Using the Batan sub-arc phlogopite-bearing harzburgites from the Philippines (Maury et al., 1992) (which are the analogue of the TLS mantle source and generated similarly high-K magmas) as the starting source material in the graphical concen-tration ratio (CR) method of Maaloe (1994), 52% of partial melting is estimated for the TLS magmas (Table 3; Fig. 15). Thus, except for the MHY and TLS magmas, which are separately derived from the astheno-spheric and lithoastheno-spheric mantle, most of the NTVZ magmas are the products of variable degrees of partial melting of asthenospheric mantle, variously ‘contaminated’
by melts from the metasomatized SCLM during ascent.
The degrees of partial melting for the NTVZ can be evaluated by the CR method using an enriched MORB mantle source as the starting source material, which is estimated from basalts in the MOT with a subduction flux similar to that for the MHY magmas (Table 3;
Fig. 15). We suggest that the parental magmas of the offshore volcanic fields in the NTVZ were generated by larger degrees of partial melting (8---20%) than the onshore fields (2---5%). A southwestward decrease in degree of partial melting is also revealed, which is evident in the plot using dominantly mantle-derived elements or conservative elements (e.g. Nb and Yb; Pearce & Parkinson,
1993; Pearce & Peate, 1995) without significant sub-duction contribution (Fig. 15b). The degrees of partial melting of the asthenospheric source may depend on both the overlying lithospheric thickness (the minimum depth at which the melting occurred) and the specific mantle domain composition. The latter cause is not pre-ferred here as the isotopic characteristics of the NTVZ magmas are decoupled from the inferred degree of partial melting. In addition, the overlying lithospheric thickness also determines the extent to which the metasomatized SCLM contaminated the asthenospheric melts. In the NTVZ, the spatial and temporal changes in magma composition, reflecting lithosphere architecture and history within a coherent tectonic terrane, allow some important inferences to be made about the geodynamic evolution of this complex region (see next section), and also about the variable nature of the lithosphere traversed by the mantle-derived mafic magmas in time and space.
Petrogenetic model
To test our hypothesis, modelling of trace-element concentrations in the mantle sources and melts was carried out. Non-modal batch melting was assumed and the compositions of the mantle sources for the MHY and TLS magmas were used as starting points. As Wang et al. (2002) suggested that asthenospheric mantle ascended to 60---70 km depth to generate the MHY magmas at 26 Ma, we assume that all melts, including the TLS melts from the SCLM, occur in the garnet stabilizing field of the mantle. Using the trace-element partition coefficients of Halliday et al. (1995), Johnson (1998) and Schmidt et al. (1999), the trace-element com-positions of the MHY and TLS mantle sources were calculated from the average lava compositions using the estimated degrees of partial melting shown in Fig. 15 (8%
for the MHY and 2% for the TLS), and appropriate source and melt modes (garnet lherzolites for the MHY and phlogopite---garnet harzburgites for the TLS; see Table 5 for details). Having thus established the trace-element compositions of the enriched asthenospheric and metasomatized SCLM source components, the trace-element composition of the asthenospheric melts can be derived for the individual NTVZ volcanic fields. Finally, a simple mixing calculation based on the estimated com-positions of the asthenospheric melts and those from the metasomatized SCLM, using the proportions shown in Fig. 8 for the individual NTVZ volcanic fields, provides their model trace-element compositions. The model trace-element compositions are plotted in Fig. 16 compared with the actual compositions of the NTVZ magmas (except for the MHY and TLS magmas used as end-members). Although the model compositions for the NTVZ do not perfectly coincide with their actual values in Fig. 16, the geochemical variation trends of
the NTVZ magmas are successfully modelled by this method.
GEODYNAMIC MODEL
Wang et al. (1999) proposed that the magmatism of the northern Taiwan volcanic zone resulted from post-collisional extension in the NTMB since Plio-Pleistocene times. Initiation of the NTVZ thus serves as a straight-forward constraint for the onset of the collapse of the NTMB. Whereas structural and seismological data (Suppe, 1984; Lee & Wang, 1988; Yeh et al., 1991)
provide lines of evidence for post-collisional extension in the Quaternary, a geodynamic model for the tectonic evolution accommodating seismological and geophysical data is still unavailable. This is especially the case when the NTVZ is considered not to be part of the Ryukyu Arc. Using analogies from other localities of Tertiary post-collisional magmatism, e.g. Tibet, China (e.g.
Turner et al., 1996; Maheo et al., 2002) and the Betic---Alboran domain of SE Spain (Turner et al., 1999), we suggest a geodynamic model for the post-colli-sional extension based on our interpretation of the NTVZ geochemical data.
20 10
0 0 2 4 6 8
Nb (ppm)
Yb (ppm)
20 15
10
5 4
Garnet facies Spinel facies
Primitive mantle Estimated enriched MORB mantle
(b)
20 10
0 0.0 0.2 0.4 0.6 0.8 1.0
La/Yb
Tb/Yb
20 15
10
5
4
2 1 0.5 0.1
Garnet faci es
Spinel facies Primitive
mantle
Estimated enriched MORB mantle
(a)
NTVZ SBS
KBS MHY
PCY KYSTTVG
TAIWANTLS
Spinel facies Garnet facies 15 10
5 4
2
10
Batan phlogopite harzburgite
Batan phlogopite harzburgite 2015 10
5 4 2
N
Fig. 15. Variation of (a) Tb/Yb vs La/Yb and (b) Nb vs Yb for the NTVZ volcanic rocks. Bold continuous lines are model curves for partial melting of the enriched MORB mantle, estimated from basalts in MOT (Wang, 1998), with spinel-facies and garnet-facies mineralogy. Dashed lines connect the same degree of partial melting between the two facies. Fine continuous lines are model curves for partial melting of Batan phlogopite harzburgite (Maury et al., 1992), an analogue of the TLS mantle source, with spinel-facies and garnet-facies mineralogy. Small ticks with numbers represent the degrees of partial melting. Values used in modelling (Table 3), the melting model and distribution coefficients are the same as in Fig. 10b. The primitive mantle (Sun & McDonough, 1989) is shown for comparison.
Models involving a retreating subduction zone (e.g.
Lonergan & White, 1996), slab detachment or break-off (e.g. Davies & von Blankenburg, 1995), delamination (e.g. Bird, 1979) and convective removal of the litho-sphere (e.g. Houseman et al., 1981) have been used to explain post-collisional extension. All these processes cause upwelling of the asthenosphere and this will perturb the original thermal gradient. Consequences that follow are magma generation and sequential exten-sion. Combining the magmatic history with other constraints based on the tectonic evolution of an area may lead to a more robust dynamic model (e.g. Turner et al., 1999).
Relative to the present-day location of Taiwan, the initial arc---continent collision started in the NE at
10 Ma and propagated southwestward corresponding to the relative motion between the Philippine and Eurasian plates (Suppe, 1984). According to the sedimen-tary record (Teng, 1990), the collision activity was most intensive at 6---5 Ma, resulting in an uplift of nearly 4000 m above sea level (Fig. 1). Meanwhile, both the arc and associated back-arc volcanism in the middle part of Ryukyu subduction experienced a hiatus in the period of 6---2 Ma (Letouzey & Kimura, 1986; Kamata &
Kodama, 1994; Park, 1996). This indicates a feedback relationship between the Taiwan collision and the
development of the part of the Ryukyu subduction system near northern Taiwan.
Severe compressional forces near the northern Taiwan region probably stalled the Philippine Sea plate subduc-tion so this slab could not encroach beneath the region.
Consequently, the Ryukyu subduction re-established and re-generated arc and back-arc volcanism at a later stage.
Two phases of the Okinawa Trough opening at 10---6 and
2 Ma (Miki, 1995; Sibuet et al, 1998) may respond to the re-establishment of the Ryukyu subduction system.
The thickened lithosphere caused by arc---continent colli-sion would have also prevented western propagation of the Ryukyu subduction system (Fig. 17a). This model differs from previous ones suggesting that the Ryukyu subduction zone was able to continue developing in the northern Taiwan region (Chen, 1990; Teng et al., 1992;
Teng, 1996).
Several lines of evidence also support the interpretation that the NTVZ is unlikely to be part of the Ryukyu Arc.
They include a horizontal displacement (of about 150 km) from the western end of the actual Ryukyu volcanic arc (Iriomotejima; Fig. 1) to the NTVZ, and the fact that the north-dipping Benioff zone of the sub-ducting Philippine Sea plate is now sitting200---250 km beneath the NTVZ (Eguchi & Uyeda, 1983; Kao et al., 1998) compared with 100 km for the normal Ryukyu Arc. There are apparent differences in the duration of volcanism between the NTVZ and the Ryukyu Arc, as the latter became dormant in the earliest Pliocene (Shinjo, 1998). In terms of their geochemical character-istics, if the NTVZ magmas represent arc volcanism, their systematic along-arc (spatial) geochemical variations (reflecting systematic variable degrees of partial melting in their mantle source) should represent a typical cross-arc variation for the cross-arc volcanism. This is not supported by the actual spatial distribution of the NTVZ paralleling the Ryukyu Trench. Moreover, the along-arc variation is also difficult to ascribe because there is no apparent spatial variation in the subducting plate, i.e. sediment flux etc., if the NTVZ is treated as arc-related volcanism.
As there was no active subduction near northern Taiwan during that time, tectonic models involving a retreating subduction zone and slab detachment or break-off models cannot be realistic for the NTVZ. However, metasomatism by the Ryukyu subduction zone could still affect upper-mantle domains beneath northern Taiwan because of the lateral migration of subduction components (Fig. 17a).
The geochemical characteristics of the MHY asthenospheric melts indicate that significant upwelling of asthenosphere to 60---70 km depth has occurred since
26 Ma (Wang et al., 2002) at the time of initiation of the NTVZ activity (Fig. 17b). Taking account of the continental crustal thickness of 30 km in northern Taiwan (Yeh et al., 1989), emplacement of the MHY magmas Table 5: Values used in trace-element modelling
Garnet peridotite Phlogopite---garnet harzburgite
Source mode (X)
Melting mode (Pi)
Source mode (X)
Melting mode (Pi)
Ol 0.6 0.1 0.6 0.05
Opx 0.2 0.18 0.2 0.32
Cpx 0.1 0.3
Garnet 0.1 0.42 0.1 0.4
Phl 0.1 0.23
Ref. 1 2
Bulk D Melt mode (P)
Bulk D Melt mode (P)
MHY (av.)
TLS (av.)
Nb 0.0025 0.0074 0.0081 0.0199 5.98 16.12
La 0.0062 0.0171 0.0011 0.0020 5.58 24.68
Sm 0.0503 0.1742 0.0239 0.0906 2.96 4.59
Zr 0.0490 0.1643 0.0360 0.1138 78.86 117.51
Y 0.2549 0.9845 0.2109 0.8121 19.71 14.61
Yb 0.7050 2.8262 0.6650 2.5803 1.52 1.32
Ref. 3 1, 3, 4
1, Johnson (1998); 2, Kelemen et al. (1993); 3, Halliday et al.
(1995); 4, Schmidt et al. (1999).
suggests that part of the SCLM was removed or thinned to let the underlying asthenosphere upwell to sufficiently shallow depths for decompressional melting to occur.
Because the collision started to the NE of present-day Taiwan close to the MHY, this is inferred to have been the region of thickest lithosphere, and most likely to be removed subsequently, similar to the scenario described for Tibetan post-collisional magmatism by Turner et al.
(1996). Actually the NTVZ distribution along the under-lying northern Taiwan metamorphic basement, which represents parts of the Taiwan orogen and is character-ized by folded and tilted Tertiary strata resulting from the collision (Fig. 1; Wageman et al., 1970; also named the southern Taiwan---Sinzi Folded Zone), strongly suggests a link between generation of the NTVZ and collapse of the Taiwan orogen. Seismic profiles show that offshore
northeastern Taiwan has been characterized by high-angle normal faults, which have been reactivated from pre-existing, collision-induced reverse faults, since late Pliocene times (Hsiao et al., 1998; Kong et al., 2000;
Fig. 1). On this basis, Teng (1996) proposed that exten-sional collapse of the NTMB took place in Plio-Pleistocene times. The uplift history revealed by the Taiwan base-ment rocks and sedibase-mentary accumulation in the fore-land basins suggests that there was a major acceleration in the rate of uplift of the arc---continent collision zone and an increase in the sediment accumulation rate at
3---25 Ma, following steady-state uplift since 8 Ma (Teng, 1990, fig. 8). This supports a model of rapid uplift after parts of the SCLM were removed or attenuated.
The SBS, TTVG and later PCY (21 Ma) magmas show HFSE depletion (Fig. 7a), indicating a likely greater
0.3 0.2
0.1 0.0
0.0 0.1 0.2
Sm/Y
Nb/Zr
SBS KBS PCY TTVG KYS Actual values
Model values
NE
SW (b)
1.0 0.8
0.6 0.4
0.2 0.0
0 10 20
Nb/Y
La/Yb
NE
SW
(a)
SBS KBS PCY TTVG KYS Actual values
Model values
Fig. 16. La/Yb vs Nb/Y (a) and Nb/Zr vs Sm/Y (b) diagrams showing similarities between model and actual ratios in the NTVZ volcanic rocks.
involvement of subduction components in the metasoma-tized asthenospheric mantle and/or the SCLM. The SBS and PCY magmas, both offshore and close to the MHY field, are also low-K and calc-alkalic magmas produced by higher degrees of partial melting and with higher
143Nd/144Nd ( 051284---051289 and 051289---051297, respectively). The TTVG magmas are calc-alkaline and have lower 143Nd/144Nd ( 051268---051280). The correlation might suggest that upwelling
143Nd/144Nd ( 051284---051289 and 051289---051297, respectively). The TTVG magmas are calc-alkaline and have lower 143Nd/144Nd ( 051268---051280). The correlation might suggest that upwelling