Average I-type granite of the LFB contains 31 ppm La, 66 ppm Ce and 31 ppm Y, whereas average S-type granite has 27 ppm La, 61 ppm Ce and 32 ppm Y (Chappell and White, 1992), indicating that there is very little difference in the whole-rock REE abundance between the two types of granite, except for some very fractionated I- and S- type granites which show marked LREE depletion and relative HREE enrichment. Thus, the difference in REE distribution patterns between apatites from S-type granites and those from mafic I-type granites does not result from the difference in the whole-rock REE distribution patterns of their host granites, but must have originated from the
redistribution of REE among different REE-bearing phases. In I-type granites, except very felsic ones, apatite, zircon and allanite are the major REE repositories, whereas in S-type granites monazite and xenotime occur in addition to apatite and zircon. Hence, the absolute abundance of REE in an individual phase, such as apatite, depends on its ability to compete with other phases for these elements. Therefore, the assemblage of REE-bearing minerals, and the partitioning of REE among these minerals and between minerals and melts in I- and S-type magmas are the dominant factors controlling the REE abundance in a phase for a given bulk REE abundance.
S-type magmas are more reduced and peraluminous. The solubility of monazite is very low in peraluminous melts (Rapp et al., 1987; Montel, 1993; Wolf and London, 1995). The redox ratio Ce3+/Ce4+ may be enhanced through a similar mechanism to that affecting the Fe2+/Fe3+ ratio, because of relatively lower oxygen fugacities and stronger peraluminosity of S-type magmas compared with I-type magmas, For these reasons, in S-type granitic magmas, which are always peraluminous, monazite is much more stable and becomes saturated earlier than apatite. As a result, monazite can occur as a liquidus phase, or crystallise from the melt prior to, or simultaneously with, apatite. Because of the early occurrence of monazite, most of the LREE goes into this mineral, so that the silicate melt and apatite will become depleted in LREE. Thus, an LREE-depleted V- shaped distribution pattern results for apatites from felsic I- and S-type granites.
By contrast, most I-type magmas are oxidised and metaluminous. The solubility of monazite increases with decreasing ASI value in the metaluminous region according to the experimental results of Montel (1986). Hence, monazite becomes saturated less easily in metaluminous I-type magmas than in S-type and peraluminous felsic I-type magmas. Furthermore, high Ca contents of mafic I-type magmas compared to their S- type equivalents lead to the relatively early saturation of apatite, which depletes the melt in REE, hence further delaying the saturation of monazite. Thus, the precipitation of monazite is suppressed in oxidised, metaluminous high-Ca magmas, until very late stages when their residual melts become peraluminous and very felsic.
Ilmenite (FeTi0 3) is stable in reduced peraluminous S-type and felsic I-type magmas because its formation requires a high Fe2+/Fe3+ ratio and high Fe2+ concentration in the melt. In oxidised metaluminous magmas, titanite (CaTiSiOs) occurs as the main Ti accessory mineral because of higher oxygen fugacity and high Ca concentration in the melt. Another consequence of high-Ca and oxidising melts is the occurrence of allanite, Ca(LREE, Ca)Al (Al, Fe3+)(Fe2+, Al)(Si0 4)3(0H). Allanite contains large amounts of Ca and Fe3+ in its structure, which requires that melts contain higher Fe3+ and Ca concentrations. Although Ce3+/Ce4+ ratios in allanite have not been systematically determined, it is expected that this ratio would be lower in allanite than in monazite. Therefore, apatite + titanite + allanite is the major mineral assemblage controlling the behaviour of light REE in I-type granitic magmas. Allanite and titanite are known to
have a strong ability to concentrate LREE in their structures and have high mineral/melt partition coefficients for REE. Therefore, the crystallisation of these two minerals will also deplete the melt in LREE.
However, the degree of LREE depletion in an I-type melt caused by the precipitation of allanite and titanite is significantly smaller than the degree of depletion caused by the occurrence of monazite, either as a liquidus phase or as a precipitate, in S-type and felsic I-type magmas, as supported by the following evidence. Firstly, chemical analyses of accessory minerals from the LFB granites show that monazite contains 10- 12 wt% La, 20-25 wt% Ce, 3-5 wt% Pr, 7.5-12 wt% Nd and 0.9-2 wt% Sm, with a total LREE of 52-55 wt%, which is much higher than those of allanite and titanite. Allanite contains about 1-5 wt% La, 2-12 wt% Ce, 0.3-2 wt% Nd, 0.1-0.5 wt% Sm, whereas titanite contains 0.1-0.2 wt%, 0.2-0.4 wt% Ce, 0.1-0.3 wt% Nd, and 0.01-0.09 wt% Sm. Thus, the precipitation of monazite will cause much stronger depletion in LREE than allanite and titanite.
Secondly, the partition coefficients of LREE between monazite and melt can be calculated based on the solubility data of monazite in silicate melts. The partition coefficients range from 1,307 to 2,106 for Ce, and from 2,036 to 2,825 for Sm, which are calculated from the solubility data of pure CePCT* and SmPÜ4 monazite in
peraluminous melts with ASI values of 1.07 to 1.11 at 800 °C and 0.2 GPa (Montel, 1986, 1993). La, Pr and Nd should have a similar range of partition coefficients. The silicate melts covered by the above experiments are only slightly peraluminous, but over 65 to 70% of S-type granites in the LFB are more peraluminous than 1.11, as shown in the statistics of Chappell and White (1992). As the solubility of monazite decreases with increasing peraluminosity and silica content, as well as with decreasing temperature (Montel, 1986, 1993), the concentrations of individual LREE in monazite-saturated melts that are more peraluminous than ASI=1.11 and more felsic than 70 wt% SiCH at temperatures lower than 800 °C could be as low as 10 to 100 ppm, and REE monazite/melt partition coefficients, thus, could be as high as 4,000 to 12,000.
Although we do not have experimental data to test this prediction, the monazite/whole-rock ratio in monazite-bearing granites will give an estimate of the minimum values for "apparent REE partition coefficients". A granite (VB94) from the S-type Burrandana pluton (71.46 wt% SiC>2 and ASI=1.17) contains 28 ppm La, 61 ppm Ce, 22 ppm Nd and 6.1 ppm Sm (Chappell, 1994, unpublished data), and monazite from this granite contains 110,664 ppm La, 229,264 ppm Ce, 92,236 ppm Nd and 14,900 ppm Sm. Therefore, the "apparent partition coefficients" are 3952 for La, 3,758 for Ce, 4,193 for Nd and 2,443 for Sm. A granite (VB98) from the S-type Mt. Flakney pluton, which is more felsic and more peraluminous (73.49 wt% SiÜ2, ASI=1.25) than the Burrandana granite, contains 8 ppm La, 19 ppm Ce, 9 ppm Nd and 2 ppm Sm (Chappell, 1994, unpublished data), and the monazite contains 113,000 ppm
La, 234,500 ppm Ce, 95,400 ppm Nd, and 15,500 ppm Sm, yielding "apparent partition coefficients" of 13,452 for La, 12,088 for Ce, 10,258 forNd and 6,458 for Sm. These two examples clearly demonstrate that REE partition coefficients between monazite and melt should increase significantly with increasing peraluminosity, and silica content, especially in the high silica region (e.g., >70 wt% SiCb).
REE partition coefficients between titanite and melt range from 10 to -200 (Noyers et al., 1983; Luhr et al., 1984; Green and Pearson, 1986). REE partition coefficients between apatite and melt vary from about 2 to 120, but are mainly between 10 and 90 in melts of basaltic to rhyolitic composition (Nagasawa, 1970; Nagasawa and Schnetzler, 1971; Watson and Green, 1981; Sawka et al., 1984). The "apparent partition coefficients" of Ce, Nd and Sm calculated from allanite/whole-rock ratios for I-type granites from the LFB range from 404 to 2,775 for La, 453 to 2,206 for Ce, 505 to 1,799 for Nd, and 532 to 972 for Sm, the upper limits of which are in agreement with the partition coefficients between allanite phenocryst separates and glasses from high-silica rhyolites (Mahood and Hildreth, 1983).
Therefore, it can be concluded that the REE partition coefficients for monazite are about 10 times higher than those for allanite, and 100 to 1000 times higher than those for titanite and apatite. For a given silicate melt containing the same REE concentrations, the crystallisation of monazite is, thus, about 10 times more efficient than allanite and
100 to 1000 times more efficient than titanite in depleting LREE in the melt.
Moreover, extensive mineral separation and SEM imaging of LFB granites show that allanite is less abundant than monazite in granites and often crystallises later than apatite. LREE in the melts enter apatite first because it occurs as a liquidus phase or early precipitate from the melt, resulting in an LREE-enriched pattern in apatites from I- type granites. However, in cases where allanite is very abundant and crystallises earlier than apatite, we should see LREE depletion. However, such depletion in apatite is characterised by a decrease mainly in La and less commonly in Ce only (e.g., sample KB22); other LREEs such as Pr, Nd and Sm, are usually not very much affected. This is because allanite shows a strong preference for La and Ce. In contrast, the precipitation of monazite causes depletion in all LREE in apatite and coexisting silicate melt, because the monazite structure can accommodate more LREE. SEM imaging shows that titanite usually crystallises later than apatite in LFB granites, although if the melt contains extremely high contents of TiC>2 and CaO, titanite may crystallise earlier than apatite.
Because the distribution pattern and magnitude of the REE partition coefficients for titanite are very similar to those of apatite, the precipitation of titanite from the melt would mainly cause a decrease in the absolute REE abundances in apatite, but not significantly change the REE distribution patterns of the apatite, although the enrichment in LREE in apatite may be less striking.
monazite can occur. For example, monazite and xenotime are present in the felsic I-type Bodalla pluton, Moruya batholith. Thus, apatites from felsic I-type granites display REE patterns similar to those of apatites from S-type granites (Figs 5-1 OB and 5-10C).
In summary, the depletion of LREE in apatites from S-type and felsic I-type granites results mainly from the occurrence of monazite, which causes much stronger depletion in LREE concentrations in silicate melts than allanite and titanite. Therefore, apatites formed from S-type and felsic I-type peraluminous magmas that contain monazite show stronger LREE depletion than apatites formed from mafic I-type metaluminous magmas that contain no monazite. Thus, the difference in accessory mineral assemblages between I-type and S-type granites, which are primarily controlled by the magma and source rock compositions, as well as intensive variables, is a direct cause for the fractionation and contrasting patterns for REE in apatites from different types of granite.