Thesis for The Degree of Ph D MTCHTGAN STATE UNIVERSITY . :1; * STEPHEN T. cam" - 1973 LI B RA It. 3 Michigan State University ‘3. Ute-Mm. mam.“ my... - This is to certify that the thesis entitled DISTRIBUTION OF TITANIUM AND PHOSPHORUS IN OCEANIC BASALTS AS A TEST OF ORIGIN presented by Stephen I. Chazen has been accepted towards fulfillment of the requirements for Ph.D. degree m Geology ()me Major profes%n [ '_ Date June 21, 1973 0-7 639 ABSTRACT DISTRIBUTION OF TITANIUM AND PHOSPHORUS IN OCEANIC BASALTS AS A TEST OF ORIGIN BY Stephen I. Chazen A model is proposed to explain the distribution of titanium and phosphorus in oceanic basalts. This model is based on, the existence of titanium and phosphorus as primary components of specific mineral phases, the use of volumetric considerations to predict the stabilities of these phases and the existence of water as OH_ in the mantle. If this model is correct then the distribution of titanium and phosphorus and their ratio may be used to test models for the origin of oceanic basalts. The amount of titanium and/ or phosphorus is dependent on (in order of importance), the percent water in the original melt, amount of fractionation of silicate phases, and the depth. The ratio is largely dependent only on depth of partial melting provided titanium and phosphorus are not depleted in the mantle by partial melting. A x2 analysis of the frequency distri- butions of these elements and their ratio is preformed in Stephen I. Chazen order to test several models for the origin of basalts. From this analysis it is concluded that partial melting is the primary process for the generation of basalts but the degree of partial melting is relatively small and that the alkalinity of basalts increase with depth of partial melting. DISTRIBUTION OF TITANIUM AND PHOSPHORUS IN OCEANIC BASALTS AS A TEST OF ORIGIN BY Stephen I. Chazen A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Geology 1973 Copyright by STEPHEN l. CHAZEN I973 ACKNOWLEDGMENTS I wish to thank Dr. Thomas Vogel for valuable advise and assistance in preparing this thesis as well as many years of moral support. I also thank Dr. Sam Upchurch, Dr. Charles Spooner, Dr. Max Mortland and Dr. Robert Ehrlich for their suggestions. I am grateful to the faculty and most especially the graduate students of the Geology Department of Michigan State University for providing the atmosphere that is conducive to this type of study. Thoughtful discussions of Gary Byerly, Leland Younker, Bruce Walker and Graham Ryder are appreciated. I am grateful to Dr. Felix Chayes of the Geophysi— cal Laboratory for providing the data reductions for this study. The understanding nature of Patricia Chazen is always appreciated. ii TABLE OF CONTENTS Page ACKNOWLEDGMENTS . . . . . . . . . . . . . ii LIST OF TABLES . . . . . . . . . . . . iv LIST OF FIGURES . . . . . . . . . . . . . V INTRODUCTION . . . . . . . . . . . . . . l PREVIOUS WORK . . . . . . . . . . . . . 3 MODEL FOR THE DISTRIBUTION OF TITANIUM AND PHOSPHORUS . 6 EFFECTS OF FRACTIONAL CRYSTALLIZATION . . . . . . 9 A TEST OF GREEN'S PETROGENETIC GRID . . . . . . . 11 TEST OF OTHER MODELS FOR BASALT PETROGENESIS . . . . 18 CONCLUSION . . . . . . . . . . . . . . . 28 REFERENCES CITED . . . . . . . . . . . . . 30 APPENDIX . . . . . . . . . . . . . .' . . 33 iii 2. 4. LIST OF TABLES X2 Analysis Breakdown of Frequency Distributions of Rock Types Shown in Figure l . . . . X2 Analysis of Frequency Distributions for 10 Rock Types Shown in Figure l for the Parameters TiOZ/P2O 5 . . . . . X2 Analysis of Frequency Distributions Comparing Quartz Tholeiite, Tholeiite (0-5), and Olivine Tholeiite (5-15) between Oceanic and Island Arc Regions, for the Parameter TiO2/P205 . Significance of Correlations (Product-Moment) Between TiO and P205 for Rock Types from Oceanic Regions . . . . iv Page l6 19 20 27 LIST OF FIGURES Figures 1. Petrogenetic grid for mantle derived basalt based on partial melting of pyrolite containing . % water . . . . . . 2. Plot of TiO2 vs P205 for 52 analysis from Kilaavea\kflran3(Wright and Fiske, 1971) 3. Plot of Ti02 vs P205 for 10 analyses from Thingmuli Volcanoe, Iceland (Carmichael, 1964) . . . . . . . . . . . . Page 12 22 24 INTRODUCTION Several hypotheses exist for the generation of basaltic magmas (Kushiro, 1973a; Green, 1971; O'Hara, 1968). These schemes have been based largely on experimental evidence. Since these hypotheses yield differing results in terms of percent partial melting, depth, and mode of origin for each basaltic composition it is apparent that some method of testing these hypotheses using evidence other than laboratory experiments would be useful. Of the major elements only the "incompatible"l ones bear any promise of differentiating between various origins for each rock composition. The nomenclature of igneous rock classification depends on the amount of major elements such as calcium, silicon, aluminum. Those elements that are independent of this classification scheme and that have predictable trends in distribution are the only ones that may be used for testing hypotheses concerning the origin of basalts. Two of these "incompatible" elements, titanium and phosphorus, are unique in that they form minerals that lElements K, P, Ti, U, and Th are defined as incompatible by Green and Ringwood (1967). contain these elements as primary components rather than existing only as traces in silicates. Stabilities of minerals such as rutile, ilmenite, sphene, perovskite, and apatite will control the distribution of titanium and phosphorus in melts since the bulk of the whole rock titanium and phosphorus is in these minerals rather than 'silicates. PREVIOUS WORK Several hypotheses exist that are used to explain the distribution of titanium and phosphorus in basaltic systems. The central problem that needs to be explained is the enrichment factors of titanium and phosphorus. That is, how does, for example, titanium increase from .71% in estimated mantle compositions to 3—4% in some alkali basalts. This is especially puzzling since rocks thought to be most deeply formed often have the greatest enrichment. It has been suggested that the earth is zoned with respect to these elements (Green and Ringwood, 1967). In Green and Ringwood's model this zonation has been caused by repeated passing of heat through the earth which drives the ”incompatible" elements out of silicate mineral struc- tures in which they do not fit very well (Harris, 1957). This process of zone refining is not applicable to titanium and phosphorus because of their tendency to form their own mineral phases (e.g., rutile, ilmenite, apatite) rather than being included as traces in silicates. Experimental attempts to duplicate this process envisioned by Harris (1957) for rock systems have not been very successful for titanium and phosphorus (Vinogradov, Yaroshevsky, and Ilyin, 1971). O'Hara (1968) suggests that titanium and phosphorus are incompatible with eclogites. If as O'Hara suggests eclogites are an intermediate product of partial melting between mantle compositions and basalts it may be possible to obtain the appropriate enrichment of titanium and phos- phorus. This argument, while it may be correct does not explain the reason for the enrichment but only reports that it occurs. Experimental studies in the system MgO-SiOZ-TiO2 (MacGregor, 1968) show that with increasing pressure more titanium is forced into the partial melt. The shift, while present, is insufficient to account for the enrichment of titanium observed in rock systems. The choice of this particular system is unfortunate since magnesium titanates are extremely rare in nature. Flower (1971) analyzed a specimen of phlogopite thought to be derived from the mantle. He found about 6% TiO2 in this specimen. He postulated that the distribution of titanium was controlled by the breakdown of phlogopite in the mantle. There are few phlogopite crystals found with as much TiO2 as this specimen (Deer, Howie, and Zussman, 1962), but even if this were common the percent phlogopite in the mantle is certainly less than 1%. There- fore only a maximum of .06% TiO in the mantle could be 2 accounted for with phlogopite. Green and Ringwood (1967) have suggested that rising plutons absorb titanium and phosphorus from the wall rock to yieldtjmzappropriate enrichment. The criticism of O'Hara (1968) seems proper in that it is unlikely that this mechanism would work as well the second time a magma passed through a conduit to the surface as it did the first time. The bulk of the current notions concerning the distribution of these two elements either use a trace element approach to their distribution or some theory that is untestable. Kushiro (1973b) has done some work in basaltic systems with TiO2 and P205 added. He reports that additions of small amounts of either element will shift the invariant point to the alkalic compositions. This work does not provide a source for the TiO2 and P205. Some other mecha— nism is needed to provide the titanium and phosphorus to the melts. MODEL FOR THE DISTRIBUTION OF TITANIUM AND PHOSPHORUS Titanium and phosphorus occur dominantly in mineral phases that contain either element as a major component and for that reason cannot be considered as trace elements. Due to the relatively low geothermal gradient in the mantle volumetric (AV) relationships should be the most important factor controlling the trends in the stability of phases (Fyfe, 1970). These trends are therefore predictable based on AV relationships of model reactions. If we assume that water exists primarily as OH_ in the mantle (Hamilton, Burnham and Osborn, 1964) then the stability of iron and calcium titanates can be predicted. For example: + TiO 2 1. FeTiO + ZOH— = Fe(OH)2 3 3 has a negative AV for all pressures. Thus larger amounts of TiO3: would be found in melts formed under increasingly higher pressures. Similar relationships can be described for calcium titanates. Therefore with increasing pressure more titanium would be forced into partial melts. The stability of ilmenite and sphene (or perovskite) with respect to rutile decreases with increasing pressure. Ilmenite (FeTiO3) will be increasingly replaced by rutile (TiO with depth and thus the source for reaction (1) 2) decreases in abundance. Rutile is relatively insoluble in OH_ and therefore there will be smaller rate of increase of titanium with increasing depth of partial melting. The stability of apatite, the primary phosphate bearing phase, can be written: C OH + 5 a( )2 b H! 2. Ca5(PO4)3OH + 9 OH = 3PO Using estimates of the compressibility of apatite and portlandite (Ca(OH)2) (Clark, 1966) the AV for this reaction becomes increasingly negative with depth. No known, OH- insoluble, phosphate phases, replace apatite with depth and the phosphorus content of the melt will increase un- impeded with depth of partial melting. Thus if OH_ is present in the mantle, both elements will increase, in the melt, with depth of partial melting. Due to the limiting effects on the availability of ilmenite (reaction [1]) by increasing rutile stability with pressure and the lack of limits on the availability of apatite the proportion of titanium to phosphorus will decrease with depth. This ratio will depend primarily on the interrela— tionships of titanium bearing mineral stabilities and apatite stability at a particular pressure. This proportion will be largely independent of reasonable amounts of partial melting since limited melting affects the absolute amount but not their proportion. The presence of titanium and phosphorus in all basalts would be expected only if the degree of partial melting was not enough in almost all cases to use up either element. If these elements were commonly used up in partial melting processes many basalts would be expected that had very little of either element. In all rocks thought to be residual products of partial melting (e.g., lherzolites) discernable amounts of both titanium and phosphorus are found (Wyllie, 1967). On these bases it is concluded that the degree of partial melting for most rocks cannot be sufficient to use up all of titanium and phosphorus in the source. Since the titanium and phosphorus are not depleted by partial melting then the ratio in the melt will be responsive to the pressure at which the melting occurred. In the source rock some titanium may be tied up as a trace element in silicates. The effect of pressure on these silicates is to force the titanium out of the lattice into melts (Verhoogen, 1962). This effect approximates the trend due to reaction [1]- EFFECTS OF FRACTIONAL CRYSTALLIZATION The effect of fractional crystallization on the titanium to phosphorus ratio must be considered. The effect on the ratio TiOZ/PZOS of fairly large amounts of fraction- ation, in sills and other igneous complexes is negligible since only trace amounts of titanium or phosphorus are incorporated into solid phases until at least 40% of the liquid is solidified (Wager and Brown, 1967; Walker, 1969; Anderson and Greenland, 1969). The remaining liquid, after some fractionation, will have essentially the same ratio as the initial ratio. In the analysis discussed below, since a large number of samples are used, errors due to liquid fractionation if present will be minimal. Of the primary silicate phases in basaltic rocks only clinopyroxene can contain more than a trace amount of titanium. The amount of clinopyroxene that would be formed by fractional-crystallization of basaltic magmas is in the range of 5-10% of the total magma with a maximum of 20%. A maximum estimate of the titanium content of the pyroxene from basaltic rocks is 2%. Using this maximum estimate, combined with the maximum estimate of pyroxene fractionated from basaltic magmas, not more than .4% of the whole rock titanium would be in the fractionated pyroxene (a more realistic estimate would be .l-.2%). If one does not con- sider individual samples but rather element distributions with large numbers of samples the effect of .4% titanium on the trends of titanium and TiOZ/P would be quite 205 small. The net effect may be quite close, to the analytical error associated with the analyses used. The ratio TiOZ/PZO5 is not correlated with olivine (normative) content (product-moment correlation in oceanic regions of about .05). The independence from olivine content is a further indication that fractionation of primary sili- cate phases does not effect the ratio. A TEST OF GREEN'S PETROGENETIC GRID Green (1971) has presented a comprehensive model for the generation of basaltic magmas. His petrogenetic grid (Figure 1), based on reconnaissance experiments in water deficient basaltic and pyrolitic systems, relates magma types to depth of partial melting of a pyrolite source and amount of water present. Although Green's model needs further experimental verification, it can be tested by comparing the theoretical magma types of his petrogenetic grid with the real patterns of distribution of certain elements in the equivalent basaltic rock types that occur in oceanic areas. In order to test Green's model, ten rock types, defined in the same manner as Green and Ringwood (1967) were selected from his petrogenetic grid (Figure 1). Sample frequency distributions for 1870 non-continental Cenozoic basalts, grouped by rock type, were obtained for the para— meters TiO and TiOZ/P2O5 from RKNFSYS (Chayes, 1970). 2' p205 In any suite of rocks, the trends in frequency distributions of titanium and phosphorus (in chemical analyses as TiO and P205) should be sensitive to depth 2 of origin and amount of water in the melt. The amount of water effects the amount of each element due to the 11 l2 .pHmE CH mcH>HHo o>HumEnoc pcmoumd ow Howey memogpcmumm CH mHoQEsz .HopOE m.c®muw co pmmMQ mmmwp xoou mo mcoHpHmOQEoo paw moumcflpmooo .Hmpmz Ma. OQHCHMDGOO mwflaouhm we mcflpHmE HMHDHMQ co pmmMQ wammwb pm>flump maucme new pane oapmcmmonuomTT.H onsmflm l3 :2: 7: 2:2: ES 2:: m. m. 2 q _ _ _ 4 lom-omc 2mm-omi mwflwmv 32:83 2:56:82 T 2. 2 222-0225 2:25 iom-omc 2mm-m:: m :m oxmm-ww_>_ :33 32:33 T 2.. . E . ._o 2:25 2:22 2222225 28.21 T 2:22: 2:25 22-2 Te T 2:22: 2:25 2:22: 2:22: Ego _ _ — — mm 2 2 m oszE LEE/E Emu Em mm 3 (EDI) HHHSSBHd 14 shifting of reactions [1] and [2] to the right with increas- ing OH-. The ratio of the two elements will be relatively constant at a particular depth since the AV which controls the equilibrium of the reactions is pressure dependent and the source of titanium and phosphorus is "unlimited" for reasonable degrees of partial melting. For any given amount of water in the melt the amount of titanium and phosphorus in the melt should increase with depth, and similarly, at any given depth the amount of water increases. If Green's model is viable, the ratio (TiOZ/PZOS) should shift downward with increasing depth and most of the variation should be between various depths. In the case of the individual elements, the most pronounced effect should be an increase in the elements with percent water in the proposed melt and a lesser increase in the elemental abundance with increasing depth. A X2 contingency table analysis was performed on the frequency distributions (Cochran, 1954) (see Appendix), of the rock types in order to test the predicted distribution of TiOZ/P2O5 and TiO2 and P205 with depth and amount of water in the suggested melt. For rock types that have the same proposed depth of origin, their TiOZ/PZO frequency 5 distributions were added together and compared with other depth groups. Similarly, for rock types that have the same suggested percent water (or degree of partial melting), their frequency distributions for the parameters TiO2 and 15 P205 (separately) were added together and compared with other percent water groups. Trends in the ratio table (Table 1, part a) that parallel the proposed pressure were tested while trends in the individual elements' table (Table 1, parts b and c) paralleling both the degree of partial melting and the percent water, were tested. A significant negative correlation was obtained between pressure and the ratio, TiOZ/PZOS’ indicating as the ratio decreases the pressure increases. A significant positive correlation was obtained between the amount of water in the proposed melt and each of TiO and P O 2 2 5’ The xz/df column in each part of Table 1 gives a measure of relative importance of each source of variation. Table 1, part a indicates that the grouping of TiO2/P205 based on depth of origin yields relatively more important variation than does the within depth variation. The within depth variation may be, in part, explained by several possible depths of origin for each of the rock types. The linear correlation with pressure is clearly the most important factor in explaining the variation between rock groupings. In Table 1, parts b and 0 both elements show insig- nificant correlation with degree of partial melting but highly significant and important correlation with percent water in the melt. The grouping based on percent water in the melt explained the bulk of variation between rock types. The within percent water also yielded significant variation l6 .9HmE CH Hm9m3 HCmonm .uHmE CH 929m3 quonmm >9 pmmsonm >9 meCOHm mCoH959HHHme >0Cm5qmnm NOHB on mCOH959HH9me >oCmsqum mOmm 99v Tmum memm\m0He Amy M mm>Hm CEDHoo wp\mx m9: Co omEHOHHmm mmz mHm>HMCm pCmHB .CHmHHo mo 99mmp >9 pmmsonm mCoH959HH9mH© >oCme "CoHpmHHm> mo ooHCOm pmpmHH mo ooCM9HOQEH m>HueHmH mo mnsmmma .Hm>mH mo. .H demHm CH C309m mmHMCHpHOOO mCHms mmH9MH >0CmmCH9Coo o99 9m pmpmmH mums muHsmom an no.m .mHm en.mmm VHH mCHmsoum Hopes quonm CHCHHB mm.m .mHm no.mom mm COHHMHmHHoo HM®CHHH>HSU nm.mv .mHm nm.me H Hmumz pCmonm CHHB COHHMHmHHoo mo. .mHm poz mo. H mCHpHmE HMkumm mo mmnmmp CHHB COHHMHoHHOU Hv.e .mHm nv.Hmm mm mCHmCOHm Hm9m3 quoHom 09 mso o pHmm Ho.v .mHm mm.mmm em mCHmsoum HmHMB quonm CH99H3 Hh.v .mHm :H.mmH ow COHHMHmHHoo HmoCHHH>HCU mn.me .mHm m:.m:H H Hopes quonm CHHB CoHumHmHHou MH.N .mHm uoz mH.N H mCHuHmE HMHunmm mo wmummp 99H3 CoHumHmHHOU mh.m .mHm mo.mmm Ne mCHQCOHm Hm9m3 pCmoumm 09 mso 9 99mm o:o.m .mHm oe.meH m: enema cHesz mm.H .me 902 mo.ov mm CoHumHoHCoo HmmCHHH>HSO mh.mw .mHm on.m: H musmmmum 99H3 COHHMHoHHoo HmmCHH me.m .mHm Ha.mmH mm mcHdsoum enema 02 as: m 920m wp\mx moCmonHCmHm mx mp oousom CH CBOCm mmm>e 900m mo mCOH939H99mHQ >oCm5qum mo C30pxmmum mHm>Hm9¢ M.H mHCmHm m XTT.H mamCe 17 which to a large extent is due to depth of origin, but it is not as relatively important as the correlation with water content. Therefore, in each of the three x2 parts of the table the method chosen to group the rock types (either by depth or percent water) yielded the relatively most important variation. By using the distribution of these elements in oceanic basaltic rocks, it is clear that the TiOZ/P205 ratios are consistent with Green's depth of partial melting, and the absolute abundances are compatible with his proposed distri- bution of water in the melt. In other words, we would have had to reject his model if these comparisons were not signifi- cant. The lack of correlation with percent water in the melt may indicate that Green's melting curves require too large percentage of partial melting (Wyllie, 1971). More appro- priate values would show a smaller range in percent partial melting than Green currently shows. TEST OF OTHER MODELS FOR BASALT PETROGENESIS Other models for the origin of basalts may be tested using the distribution of titanium and phosphorous. Kushiro (1973a) has suggested a model which relates alkalinity with depth. In a general sense, Green (1971) has done this in his grid but some rock types are out of place (e.g., basanite [15-25] with a model based on parallel alkalinity and depth). If we group TiOZ/P2OS for the 10 rock types shown in Figure 1 in a manner such that there is decrease in silica saturation the following rock types would be grouped: l. Quartz Tholeiite 2. Olivine Tholeiite and Tholeiite 3. Alkaline Olivine Basalt 4. Olivine Basanite and Olivine Nephelinite The first group is therefore quartz normative, the second is olivine normative but not nepheline normative, the third has less than 5% nepheline in the norm and the fourth has more than 5%. Picrites were placed in the appropriate groups based on their nepheline content. The results of the x2 analysis on the frequency distributions of TiOZ/PZO ratio for the rock types grouped 5 as above is shown in Table 2. As can be seen by comparing this table with Table 1, part a this method of grouping is 18 19 clearly superior to Green's grouping. This can be seen by comparing the size of the "Due to Depth" x2 values and by noting the lack of significant (at the 5% level) "Within Depth" factors in Table 2. TABLE 2.--x2 Analysis of Frequency Distributions for 10 Rock Types Shown in Figure l for the Parameters TiOZ/P2O5.* 2 Sig. 2 Source df x (5% level) x /df Due to Depth 36 203.1345 Sig. 5.64 Within Depth 72 71.7349 Not Sig. 1.00 Rock types were grouped according to silica saturation. The within depth portion is largely caused by partial melting affects on the ratio. It might be argued that the ratio TiOz/PZO is cor- 5 related in some heretofore unknown way with alkalinity and not with depth per se. Two observations strongly suggest that this is not the case. First, the frequency distributions for the ratio are clearly polymodal in all rock types examined. This would be eXpected only if the ratio were correlated with some variable, such as depth, that had several possible values for a given rock type. Due to the tight restraints on the rock type compositions it is unlikely that compositional control, by the major elements, would give patterns such as those shown for these rocks. 20 Second, the oceanic regions have different depth relationships for the origin of rock types than circumoceanic regions (Dickinson, 1970), and if we compare three rock types found in large numbers in both island arcs and oceanic regions we should see this depth difference if the ratio is a measure of depth. If it is strictly a compositional indicator there should be no differences between regions only possible differences between rock types. The three rock types chosen were quartz tholeiite, tholeiite (0-5) and olivine tholeiite (5-15). These three types are extremely common in both areas and are the only basaltic rock types with sufficient number of analyses in the island arcs to work with. Table 3 shows the result of this analysis. There is no significant difference (at the 5% level) between rock types. There is a large, highly important, very sig- nificant difference due to region indicated. If the ratio truly measures only composition there should have been only differences between rock types. The ratio must therefore be a measure of some genetic factors. TABLE 3.--x2 Analysis of Frequency Distributions Comparing Quartz Tholeiite, Tholeiite (0-5), and Olivine Tholeiite (5-15) between Oceanic and Island Arc Regions, for the Parameter TiOz/PZO 5. 2 Sig. 2 Source df x (5% level) x /df Due to rock type 28 41.133 Not Sig. 1.47 Due to region 42 525.842 Sig. 12.52 21 The melting relationships proposed by Wyllie (1971) for the fusion of a garnet peridotite source is a scheme that might be tested. Relatively small (compared to Green, 1971) amounts of melt are produced in Wyllie's system. This model is likely to give results comparable to the degree of partial melting that best fits the distribution of TiO2 and P205. That is, his system would give a smaller range of partial melting than Green has proposed. Unfortu- nately sufficient data does not exist to rigorously test this model. The model of O'Hara (1968) (O'Hara and Yoder, 1967) is more difficult to evaluate than the other schemes because of its reliance on fractional crystallization for many compositional differences. In a general sense, this type of model may also be evaluated using the distribution of titanium and phosphorus. If fractional crystallization is an important process then plots of TiO2 vs P205 will have good correlations. The amount of titanium and phosphorus should increase with fractionation but the relative amount of each (the ratio) should be the same. Figure 2 illustrates this for Hawaiian volcanics while Figure 3 is for Icelandic volcanics. Both are tholeiites. Both clearly have important fractionation components in their origin. The Icelandic series has a deeper origin than the Hawaiian (as indicated by the slope). This is supported by geophysical evidence independent of 22 .Av.n mumEHHmm oHumuv mmMH. mH onHm ®CE .moHHom OHHHHmHOCH UwHMHquHwHMHp mHCH 90w mOmm pCm NOHB Coo39o9 mpmeo Amnmm. n HV CoHpmHmHHoo mCouum < .prxo m99 HO quOme HCmHoB CH mum CBOCm mwsz> .AHan .mmem pCm HCmHHBV OCm0H0> mesmHHM Eoum mom>HmCm mm MOM mOmm m> NOHB mo uonTT.N mHCmHm 23 IF.“ T 2 A _ x «him» .5» «95$; 2 2 232.1 + # W23 2 an 24 .Am.m mumEHHmw OHHMHV mom. mH mmon wCB .meHmm UHHHHwHOCu pmuwHquHomep mH99 HOw mONm pr NOHB wa39m9 AmmHm. H My mumem CoHumHmHHoo p000 .opon o99 Ho pCmoCmm HCmHmB CH mum CBOCm mosHm> .AvomH rHoMCOHEHmoV pCmHmOH roOCm0H0> HHsEmCHCB EOHM mom>HMCm 0H How mONm m> NOHB mo ponTT.m mHCmHm 25 ow mm on ma cu m: . _ _ _ _ _ .00 X XLN. X XX x 1*. X . 3d 42% X X vb 26 Chemical reasoning for both regions (Carmichael, 1964; Wright and Fiske, 1971). The correlation between titanium and phosphorus in both cases is very high. In an area where partial melting is more important than fractionation, the correlation between titanium and phosphorus will be low. That is all the rocks will cluster around a particular value yielding a low correlation coef- ficient but a small variance of each of TiO2 and P205. In order to evaluate the entire oceanic region each area within it would have to be considered separately. An approximation to this more involved process can be made by looking at a product moment correlation between titanium and phosphorus for some of the previously defined rock types. Table 4 shows some values for oceanic regions. This type of approach is only an approximation since it does not take into account regional differences but it does give some indication of the relative importance of processes. Not surprisingly the more undersaturated rocks are derived by partial melting (or some complex combination of partial melting and fractionation and partial melting again) while the more silica rich rock types have important evidence of fractionation in their origin. It is clear that the distribution of titanium and phosphorus can be useful in determining whether a series of rocks had been fractionated or had a simple partial melting origin. Other major elements would not be as useful as this 27 TABLE 4.——Significance of Correlations (Product-Moment) Between Ti02 and P205 for Rock Types from Oceanic Regions. Significance Level Rock Type of Correlation Quartz Tholeiite gtr. than .001 Tholeitte (0-5) gtr. than .001 Olivine Tholeiite (5-15) gtr. than .001 Olivine Tholeiite (15-20) gtr. than .001 Olivine Tholeiite (20-25) Not Sig. Olivine Rich Basanite (15-25) Not Sig. Picrite Not Sig. Olivine Rich Basanite (20-30) Not Sig. Alkaline Olivine Basalt Not Sig. 1': Minimum significance level was .10. pair for several possible reasons. First, often pairs of elements are correlated because of numerical necessity (Chayes, 1949, 1971, etc.). Secondly, titanium and phos- phorus travel together and respond to the same factors except under the most extreme conditions of fractionation. CONCLUSION One of the clearest observations concerning the distribution of titanium and phosphorus is their regularity within regions and rock types. The trends in frequency distributions were as predicted by the model for the occur- rence of these elements and is generally consistent with accepted models for the origin of basaltic magmas. The 1 change in the ratio, in the case of Green's model was closely parallel with the change in pressure. Since alkalinity increase with pressure underlies all models for oceanic regions it would be expected that most other models would also linearly correlate the TiOZ/P ratio with pressure. 205 The absolute amount of each element is dependent on the pressure and percent water in the melt. Green's percent water direction is nearly an increasing normative olivine direction with decreasing water. The apparent negative logarithmic-type correlation (similar to percent water in melt) between percent olivine and TiO2 and P205 distributions could easily explain the apparent enrichment of TiO2 and P205 in basalts. The amount of titanium (or phosphorus) in picrites is not much greater than the amount of titanium assumed to be in the mantle. With a log-type increase in water in melts the titanium and phosphorus would increase logarithmically also. 28 29 A model that would be consistent with the distri- bution of TiO2 and P205 and their ratio should contain each of the following characteristics: 1. The model should be based primarily on partial melting but locally important fractionation at shallow depths may be used to generate more siliceous rock types. 2. The degree of partial melting should be rela- tively small (a smaller range than Green has suggested), perhaps as small as 10% maximum. 3. The trends and water contents in basaltic magmas must be similar to those suggested by Green. 4. The model should relate increasing alkalinity with depth. Most models currently suggested meet the above criteria, at least generally. Consideration of the distri- bution of these two "incompatible" elements should be useful as a guide to further experimental work in basaltic systems. The distribution of these elements clearly puts limits on what is and is not possible for generation of magma compo- sitions similar to basalts that exist in oceanic regions. REFERENCES CITED Anderson, A. T., and Greenland, C. P., 1969. Phosphorus Fractionation Diagrams as a Quantitative Indicator of Crystal Differentiation of Basaltic Liquids, Geochemica Et Cosmo. Acta, 33, 493-506. Carmichael, I.S.E., 1964. Petrology of Thingmuli Volcanoe, Iceland, J. Petrology, 5, 435-460. Chayes, F., 1949. On Correlation in Petrography, J. Geology, 57, 239-254. , 1970. Electronic Storage, retrieval, and reduc- tion of data about the chemical composition of common rocks, Carnegie Inst. of Wash. Yearbook, 70, 1970-71, 197—201. , 1971. Ratio Correlation. Chicago, I11.: Univ. of Chicago Press, p. 99. Clark, S. 0., Jr., 1966. Handbook of Physical Constants, GSA Memoir 97, Geol. Soc. Am., New York, p. 587. Cochran, W. G., 1954. Some methods of strengthening the common X2 test, Biometrics, 10, 417-451. Deer, W. A.; Howie, R. A.; and Zussman, J., 1962. Rock Forming Minerals, Vol. 3, London, Longmans, Green and Co., p. 270. Dickinson, W. R., 1970. Global Tectonics, Science, 168, 1250-1259. Flower, M. F. J., 1971. Evidence for the role of phlogopite in the genesis of alklai basalts, Cont. Min. Pet., 32, 126-137. Fyfe, W. S., 1970. Lattice Energies, Phase transformations and volatiles in the mantle, Phys. Earth and Planetary Interiors, 3, 196—200. Green, D. H., 1971. Compositions of basaltic magmas as indicators of conditions of origin: Application to Oceanic volcanism, Phil. Trans. Roy. Soc. London, Ser. A., 268, 707-725. 3O 31 Green, D. H., 1973. Contrasted melting relationships in a pyrolite upper mantle under mid-oceanic ridge, stable crust, and island arc environments, Tectonophysics, 17, 285-297. Green, D. H., and Ringwood, A. E., 1967. The genesis of basaltic magmas, Cont. Min. Pet., 15, 103-190. Hamilton, D. L.; Burnham, C. W.; and Osburn, E. F., 1964. The solubility of water and effects of oxygen fugacity and water content on crystalization in mafic magmas, J. Petrology, 5, 21-39. Harris, P. L., 1957. Zone refining and the origin of pltassic basalts, Geochem. et Cosmo. Acta, 12, 195-208. Kushiro, I., 1973a. Origin of some magmas in oceanic and circumoceanic regions, Tectonophysics, 17, 211-222. , 1973b. Regularity on Shift of Liquidus Boundaries between silicate minerals and its significance in Imxflm1genesis,Carnegie Inst. Wash. Yearbook 72, in press. MacGregor, Ian, 1969. The system MgO-SiOZ-TiO and its bearing on the distribution of Ti02 in basalts, Am. J. Science, 267A, 342-363. Maxwell, A. E., 1961. Analysing Qualitative Data. London: Methuen and Co. Ltd., p. 163. O'Hara, M. J., 1968. The bearing of phase equilibria studies in synthetic and natural systems on the origin and evolution of basic and ultrabasic rocks. Earth Sci. Rev., 4, 69-133. O'Hara, M. J., and Yoder, H. S., 1967. Formation and fractionation of basic magmas at high pressure, Scottish J. Geol., 3, 67-117. Verhoogen, J., 1962. Distribution of titanium between silicates and oxides in igneous rocks, Am. J. Science, 260, 211-220. Vinogradov, A. P.; Yaroshevsky, A. A.; and Olyin, N. P., 1971. Phil. Trans. Roy: Soc. London, Ser. A., 268, 409-421. Wager, L. R., and Brown, G. M., 1967. Layered Igneous Rocks, San Francisco: W. H. Freeman & Co., p. 588. Walker, Wright, Wyllie, 32 K. R., 1969. The Palisades Sill, New Jersey. A reinvestigation, Geol. Soc. Am. Sp. Paper 111, Geol. Soc. Am.: Boulder, Colorado, p. 178. T. L., and Fiske, R. S., 1971. Origin of the dif- ferentiated and hybrid lavas of Kilauea Volcano, Hawaii, J. Petrology, 12, 1-65. P. J., 1967. Ultramafic and Related Rocks. New York: John Wiley and Sons Inc., p. 464. , 1971. The role of water in magma generation and interpretation of diapiric uprise in the mantle, J. Geophy. Res., 76, 1328-1338. APPENDIX A description of the x2 analysis used in this work may be obtained from Maxwell (1961). Deviations from chance for frequency distributions between groups may be evaluated by using X2 analysis. The expected value of the x2 is equal to the number of degrees of freedom. This number is calcu- lated by creating a contingency table and multiplying the number of rows minus one times the number of columns minus one. Deviations from this expected value are evaluated using a X2 table. A significant result means that there are differences between categories, within the contingency table, that deviate significantly from chance. In this analysis (see Table 1, part a), for example, a contingency table was made that had a TiO2/P205 frequency distribution for each rock type (10 rock types and 12 divisions within each frequency distribution). A X2 value was obtained for this "Total" table. Another table was made that grouped and added the frequency distributions of the rock types that were thought to have the same depth (4 depth groups and 12 divisions of the distributions). The x2 value obtained for this table was the "due to depth" value. The difference between this value and the value for the "Total" table is ascribed to "within depth." 33 34 Trends in these tables may also be tested. Using a table that has both the vertical and horizontal axis in a natural order we can test to see if a continuous variable underlies the distribution. That is, is the shift in the frequency distribution continuous and is the shift related to the assigned coordinates for each cell of the table. In this analysis, the contingency table with the 4 depth groups and the 12 divisions of the distribution was examined to see if the frequency distributions shifted to one direction continuously and if this shift parallels the change in pressure in magnitude and direction. The x2 value gives the significance of the regression coefficient with one degree of freedom.