American Mineralogist, Volume 64, pages 501-513, 1979

Statistical analysis of clinopyroxenes from deep-sea basalts

Elntur L..ScHwrnznn, Jeurs J. Pepxn eNo A. EnwnRn BnNcr

Department of Earth and Space Sciences, State Uniuersity of New YorkStony Brook, New York I1794

Abstract

Pyroxene analyses of superior quality, from Dspp Legs I 1, 15, 17, 34, and 37, have beenused to identify the various basalt groups from pyroxene chemistry. Cluster analysis, factoranalysis, and variation plots give insight into the crystal chemistry of the pyroxenes. Thetholeiitic and alkalic pyroxene groups differ primarily in CrrOs, TiOr, and Na2O contents.Tholeiitic pyroxenes are high in CrrOs and low in NazO and TiOr, relative to an averagepyroxene. For the same oxides, pyroxenes from alkalic basalts are high in TiOz and NarO, andlow in Cr2Or. Two pairs of cations, Ti-IvAl and Fe3+-IvAl, have high correlation coefficientsin all ofthe groups. It appears that common substitutions which occur in deep-sea pyroxenesare: Al for Si in a tetrahedral site combined with Ti or Fe8+ substituting for Mg or Fe2+(respectively) in the octahedral Ml site.

Introduction

Pyroxenes are important recorders of basalt pet-rogenesis. Detailed investigations of the crystal chem-istry of mare basalt pyroxenes (Bence and Papike,1972) in conjunction with experimental studies(Grove and Bence, 1977) reveal that the pyroxenesare recorders of bulk chemistry,./Cr, mineral para-genesis, and cooling rate. In ocean-ridge basalts,clinopyroxene follows olivine and plagioclase in thelow-pressure crystallization sequence, and its chemis-try reflects conditions of late-stage crystallization.However, at high pressures (: l0 kbar), clinopy-roxene can be a liquidus phase (Kushiro, 1973;Bender et al., 1978), and its chemistry will reflectthese conditions of crystallization.

We have studied pyroxenes from 8 drill sites of 5Dsur legs, in order to get a broad view of deep-seabasalt pyroxene chemistry. Cluster and factor anal-yses were used to simplify the large data base (1583pyroxene analyses). The major objectives are to usethe pyroxene chemistry (l) to identify the variousbasalt groups and the parageneses within thosegroups, and (2) to define systematics for pyroxenecrystal chemistry. Preliminary results have been pub-lished by Schweitzer et al. (1978).

Geologic background

The pyroxenes are from a variety of deep-sea basaltlocalities (Fig. 1). Basalts from four of the five legs

0003404x /7 9 /0506-050 I $02.00

are oceanic tholeiites (Dsnr Legs I l, 15, 34, and 37).Alkalic basalts recovered from Leg l7 are also con-sidered. Basalts from Legs 11,34, and 37 appear tohave erupted at spreading centers, while the alkalicbasalts from Leg 17 are products of off-ridge vol-canism. Leg 15 basalts may have been the product offlood-type volcanism (Donnelly, 1973), but this inter-pretation is disputed by Christofersson (1976). Ba-salts from these legs were chosen for statistical analy-sis for the following reasons: (l) to represent the twomajor oceanic basalt types, while maintaining, atleast nominally, the relative abundance of the twotypes; (2) to use tholeiites which varied in age andlorphysical characteristics; and (3) to eliminate inter-laboratory analytical biases by using only data col-lected at Stony Brook.

The basalts range from extensively altered (Leg I l)to very fresh (Leg 37). Many of the extensively al-tered ones contain smectite and zeolites, and showreplacement of olivine and plagioclase (Myers er a/.,1975; Ayuso er al.,1976). Sites l@ and 105 ofLeg llyielded the oldest basalts used in this study, approxi-mately 130 m.y. old (Hollister et a1.,1972). Sites 146,151, and 153 of Leg 15 in the central Caribbeanyielded Cretaceous tholeiites (Edgar et al.,1973). Al-though the Leg l5 samples are from a tectonic settingdifferent from that of oceanic tholeiites, they havesimilar major and trace element chemistries (Bence eral., 1975). Basalts from Sites 3l9A and 321(Leg3a)underlie sediments which are 15 and 40 m.y. old,

501

502 SCHWEITZER ET AL..' CLINOPYROXENES

Fig. l. Map showing the locations of the sites sampled.

respectively (Yeats et al., 1976). The youngest sitesampled is Site 332 of Leg 37. The basalts from thissite are approximately 3.5 m.y. old (Dmitriev,1977),and some of the samples have been characterized asfresh, chemically primitive tholeiitic basalts (Hodgesand Papike, 1977).

The alkalic basalts from Leg l7 are from two sites:165A and 169. Site 169 also yielded tholeiitic rocks,but the pyroxene analyses used are only from thealkalic sill (Myers et al.,1975). These alkalic sampleswere selected because the pyroxenes are fresh andwere analyzed in one electron microprobe laboratory.

MethodsData selection

The pyroxene analyses were collected on an auto-mated Anl Errrx-srvt microprobe. Analyses used inthis paper represent individual point analyses takenin most cases on zoned crystals; they were selected torepresent the maximum composition ranges for thepyroxenes for the specific rocks studied. Data reduc-tion procedures, carried out on-line, are those ofBence and Albee (1968). Nine elements (Si, Al, Ti,Cr, Fe, Mn, Mg, Ca, Na) were analyzed. Each analy-sis also had ferric iron estimated by the method of

Papike et al. (1974). In order to be selected for statis-tical treatment, each analysis had to have an oxidesum : 100+2 weight percent, and it had to pass all ofthe following tests: (l) the sum of Si + rvAl : 2+.0.02atoms/6 oxygens; (2) octahedral cations (Mn, Fe2+,Fe3+, Mg, Ti, Cr, vIAl) sum to greater than 0.98; (3)the M2 site occupancy equals 1.010.02; and (4) thecharge balance equation {vIAl + vlFes+ + vrcrs+ +2Ttn+ : rvAl + MzNa) balances within 0.02 charge.

The number of cations per formula unit was usedin the statistical programs because it gives more in-sight into the crystal chemistry of the pyroxenes thandoes similar treatment of the oxides, and providesinformation equally ,suitable to discriminate amongthe basalt groups.

The statistical programs could handle no morethan 800 analyses at a time; consequently, the origi-nal data set was cut to 750 analyses (150 for each leg).Exactly 150 analyses were selected from the dataavailable for each leg by using the crystal-chemicalcriteria stated above and more stringent constraintson the cation sums. The sum of the tetrahedral cat-ions must equal 2.000, and the total cation sum de-viates at most only slightly from 4.000 (3.985 ( .r (

4 .010) .

SCHWEITZER ET AL.:

S tatistical techniques

Cluster and factor analysis are both multivariatestatistical techniques which condense and simplifydata. Cluster analysis calculates distances betweenpoints in n-space (n : number of variables) and thenforms clusters of points based on the distances be-tween them. Factor analysis also considers casesplotted in n-space and simplifies the variables so thatthe cases can be described in a more economicalmanner by a smaller number of factor axes.

Each set of 150 analyses was clustered using Bio-medical Computer Program P2M (Dixon, 1975). Thecations IVAI, Fe2+, Fe8+, Mg, Mn, Ti, Cr, Ca, Na,and vIAl were the variables, and each analysis in theset was compared with the other 149 analyses so thatmore similar ones were grouped together in the re-sulting histogram. After the cluster analysis was per-formgd, each set was separated into recognizablegroups based on the dendritic histogram of the com-puter output. The groups and their mean cation andoxide compositions were then used as "summaries"of the total data set studied.

A factor analysis program (Bune4M; Dixon, 1975)was run for each set of 150 analyses, and also for eachgroup which was recognized from the cluster analy-ses. The variables used for the factor analysis werethe same cations as above. with the addition of Si. Inaddition to the factors determined by this analysis,correlation coefficients were calculated for every pairof cations.

A cautionary note is appropriate at this point.Multivariate analysis is a means of increasing per-ception of a complex system by compressing manyvariable dimensions into a few. The power of thisstatistic is that the reduction can be performed andchecked by computer. However, one must be awarethat some of the procedures are highly dependentupon pattern recognition techniques (Naney et a/.,1976). Thus, the reader must be careful not to over-interpret the pyroxene groups presented in this paper.Different techniques and different data sets couldproduce different clusters. In fact, in a compositionalcontinuum the cluster analysis might define "groups."Nevertheless, the approach is extremely useful insimplifying the data and in identifying chemical sys-tematics that might be otherwise overlooked.

Results

The clustering procedure separates the data into 9groups of unequal numbers of analyses. Analyses

CLINOPYROXENES

Table l. Site distribution for the cluster groups

Group " Site (l of analyses)

503

l l - l

tr-2

l 5

t 7 - l

t7 -2

t't -3

34-r

34-2

37-332

r00 (87)105 (47)r00(3)

146 (r 04)tst Q4)ts2 (22)

t6e (47)165A (3 l )

l65A (s0)

169 (16 )

3 le (64)32t Q7)

32r (r 5)3 le ( r0 )

332 (132)

" Groups are labelled as leg number-group number.

from Legs I I and 34 formed 2 groups each, thosefrom Leg I 7 formed 3 groups, and those from Legs l5and 37 formed one group each (Site 332 is the onlysite analyzed for Leg 37). The nine groups, brokendown by site and the number of analyses from eachsite, are shown in Table l. The cluster groups areidentified by leg number and group number (i.e., Leg17, group I is l7-l). Analyses which did not fall intoan identifiable group were not used in the groupaverages. The actual number of analyses in eachgroup is listed in Table l.

The mean oxide and cation compositions for eachpyroxene group are shown in Table 2. As with thewhole-rock compositions, the NarO contents of thepyroxenes from alkalic basalts differ significantlyfrom corresponding values for the tholeiites, but, forthe other oxides, there is more variation within thethree cluster groups of the alkalic basalts than be-tween cluster groups for the alkalic and tholeiiticbasalts. The cluster groups l7-1, l7-2, and l7-3 arefrom the alkalic basalts, and the 6 remaining clustergroups are from tholeiitic basalts.

Average "alkalic" and "tholeiitic" pyroxenes (Fig.2a) were calculated using the mean oxide values fromthe alkalic and tholeiitic cluster groups. These twoaverage pyroxene compositions were combined withthe mean compositions of the 9 cluster groups incalculating the average pyroxene composition in the

504 SCHWEITZER ET AL.: CLINOPYROXENES

''Ar cl Mg Mn l{ol l

'"Alt t l

Fc!' l* ' t t lCo Xp.

t lTiALICLIC -.-

nouenrc --.- -.--.r,lr.---x-r...r.....--.-oo

f Z -ffZ'-- +'-t-O.A*--tr-*===*------OO

o.ot .o

z @ o 5

oo

r o

r 'o

o.5

t o

o.5

oo

r o

-o.5

o or o

-o.5

o o

t .o

r o

o.o

t7 -2 o 5

o.o7 2

t o

o 5t 7 - 3

a Sl0" frzOr TOz ftO CoOL si vht cP' M9 Mn NoD t t r t t r r r r t t l

''Ar Fc3' Ti Fc* co xF.AlrO! CraQ MgO |,ho ilodo

Fig. 2. (a) Oxide variations of pyroxene group averages, normalized to the average pyroxene of Table 2. See text for details of thenormalization procedure. (b) Cation variations ofpyroxenegroup averages, normalized to the average pyroxene ofTable 2. Xr.: S:""* I(Fe'z+ + Mg). See text for details of the normalization procedure.

SCHWEITZER ET AL.: CLINOPYROXENES

Table 2. Mean pyroxene cluster group compositions

505

t't -3l't-2t7 -ll 5Il-2l l - t 34-l 34-2 37-332 Avg. Pyx.

sio,AlrosTio,CrzOtFeO"FezOa"MnOMgoCaONarO

siI V A I

) te t rahedra l

TivrAl

Cr.. Fe3+"

Fe2+MnMg

)oc tahedra l

CaNa2aor"t

"raror"

51.32 48.953.93 4.810.34 l . l90.63 0.124.29 10. 131.87 1.950 .11 0 .23

17.97 14.0t19.19 18.290 .14 0 .18

99.79 99.86

1.876 1.8400.124 0.1632.000 2.003

0.009 0.0340.046 0.0500.018 0.0030.051 0.0550 . 131 0 .3180.003 0.0070.979 0.783

t.z3',t r.250

0.752 0.',t360.010 0.0133.999 4.N2

51.50 49.292.75 3.750.62 1.500.25 0.028.49 8.760.98 2.160.07 0.24

16.81 r3.2r18.20 20.620.05 0.30

99.72 99.85

1.908 1.8540.w2 0.1452.000 t.999

0.017 0.0420.028 0.0210.007 0.0010.02't 0.0610.265 0.2'160.002 0.0080.927 0.'t40

t.2't3 1.t49

4.722 0.8310.003 0.0223.998 4.001

100.06 99.66

l 655 I .93l0.345 0.0672.000 1.998

0.120 0.0200.021 0.0160.000 0.0000.1 l5 0.0380.179 0.6950.005 0.0190.654 0.355

I .094 1.143

0.874 0.8330.032 0.0284.000 4.N2

100.11 99.79

t .869 1.8910.130 0.107l.999 1.998

0.035 0.0400.015 0.0030.005 0.0010.058 0.0380.262 0.4750.008 0.0130.860 0.751

1.243 1.32r

0.741 0.6670.018 0.0r 54.000 4.001

51.21 49.363.10 3.860.54 1.360.23 0.168.54 9.7 |t .4r | .960.20 0.24

t6.24 | 3.9818.40 19.030.10 0.23

99.97 99.89

1.898 1.8580.102 0.1422.000 2.000

0.015 0.0370.033 0.0260.007 0.0050.039 0.0530.27 | 0 .3190.006 0.0080.892 0.771

|.263 | .219

0.729 0.7650.007 0.0163.999 4.000

43.948.244.240.005.684.050. l6

1 1 . 6 521.660.44

48.56l . 8 00.690.01

20.851.260.556.04

19.540.36

50.263.301 .250 . t 88.412.080.26

t5.52l8.600.25

49.552.471.,100.02

14.87I . J J

0.41t3.2116.320.21

Note: Ferric iron was calculated for each analysis by the method of Papike et al. (1974\. Groups l7-1, 17-2, and l7-3 are from alkalicbasalts, all other groups are from tholeiitic basalts. Avg. Pyx. is an average pyroxene calculated using the mean compositions from thecluster groups, see text for details.

column labelled "Avg. Pyx." in Table 2. However,the mean cation composition in the Avg. Pyx. columnwas calculated using only the mean cation valuesfrom the 9 cluster groups.

The average whole-rock analyses listed in Table 3indicate by the low sums, relatively high KzO andhigh FerOs (where analyzed), that alteration has oc-curred. A major objective of this study is to attemptto see through the alteration of the whole rocks bychemically characterizing the essentially pristine py-roxenes. The whole-rock compositions listed in Table3 are arranged to correspond to the 9 pyroxene clus-ter groups. Whole-rock analyses for sites and inter-vals corresponding to those of the pyroxene clustergroups were taken from the literature and averagedtogether so that they would correspond to the aver-age pyroxene compositions of the cluster groups. Thewhole-rock alkalic basalt compositions ("cluster"groups l7-1, 2,3) are high in TiO2, Na2O, and KzOrelative to the tholeiites. [n the pyroxene groups, onlyNa2O is consistently higher in the pyroxenes fromalkalic basalts.

Chemical characte rist ics

The differences among the pyroxene cluster groupsare easily observed on variation diagrams (Fig. 2a,b).The 9 cluster groups are arranged in Figure 2 so thatsimilar patterns are in close proximity. The oxides orcations are ordered so that the ionic radii of thecations increase from left to right. The actual numeri-cal values are derived from the ratios oxide/(oxide *average) and cation/(cation + average), where"oxide" and "cation" are the group mean values(Table 2) and "average" is the value for Avg. Pyx.(Table 2). The ratios are normalized so that Avg.Pyx. plots on the 0.5 line. If any oxide or cation valueis higher than that of Avg. Pyx. it plots above the 0.5line, and if it is lower it plots below the line. Theresulting variation diagrams show differences andserve. like chondrite-normalized REE and transition-metal plots, to distinguish the basalt groups. Two ofthe groups, 34-l and ll-2,have mean compositionsthat are very similar to the mean of the whole dataset; consequently, their patterns appear relatively flat.

506 SCHI4/EITZER ET AL.: CLINOPYROX ENES

Table 3. Whole rock average compositions

I l-l n-2 15 l7-t l7-2 l7-3 34_l 34_2 37_332

sio,Al,osTio,C12O,FeOFerO,MnOMeoCaONaroKrO

t

49.716.40.870.038.78n.a.0 . l38 .51

12.42.420 . 1 I

99.24

49.015 .31.050.069.70n,a .0 . 1 91 . 6 3

13 .62 . t80.29

99.20

50.415.40.920.009.93n.a .0.058.67

12.42.380.04

100.19

44.261 4 . 3 12.95n.a .7.544.750.226.869.883.6',1t .25

95.69

42.M13.614.89n.a.7.'t25.260.246.228.s74.U2 . 1 2

94.71

46.49t 5 .011.02n.a .7.374.250. r91.50

11 .203.300.39

96.72

50.2t4.N1.880.02

lt.2n.a.0 . 1 67.03

l 1 . 32.790.31

99.29

49.913.92 . t 90.02

12.2n .a .0 . l96.60

10.82.700.24

98.74

48.46t7.950.740.033 . r53.860 .127.05

t4. t82.020 .18

97.74

n.a. = not analyzed.

l7-3 Site 169 (Bass et al., 1973); XRF and wet chemical.?tl qitr 319 (Mazzullo and Bence, 1976); electron microprobe.11-?-!iq 319 and 321 (Mazzullo and Bence, 19?6); etection microprobe.37-332 Sites 332,4 and 3328 (Blanchard et al., t976); XRF and wei chemical.

The behavior of Cr2O, changes progressively fromthe top group to the bottom group in Figure 2a; i.e.,CrrOr is high relative to "average pyroxene" in thetop group and low in the bottom groups.

The average compositions of pyroxenes from al-kalic and tholeiitic basalts, grouped separately, areshown at the top of Figure 2a and reflect clearly thechemical differences between the two basalt types.The major differences between tholeiitic and alkalicbasalts, as reflected in the pyroxene chemistry, are

\ . .r t t r .

, l

o o o r 0 2 0 3 0 4 0 5 0 6 0 ? 0 8 0 9{ rcz*z rczf us}

Fig. 3a. Ti us. Fe,+/(Fe2+ + Mg) (atomic). Three cluster groupsfor pyroxenes from alkalic basalts. Ti units are atoms per formulauni t .

Fig. 3. Ti us. Fd+/(Fsr+ + Mg) (atomic). The three groupsplotted are: closed circles = alkalic basalt pyroxenes, open circles: tholeiitic basalt pyroxenes, closed triangles (in shaded region) =pyroxenes from basalts collected at Site 100 of Leg I l. Units arecations per formula unit.

SCHWEITZER ET AL.: CLINOPYROXENES 507

(Feztl lez'rwl

Fig. 4. Na us. Fe'1rl(Fd+ * Mg) (atomic). Symbols for groupsand units are as in Fig. 3.

best seen in the components other than CaSiOg-MgSiOr-FeSiOs. The pyroxenes from alkalic rocksare relatively low and those from tholeiites are rela-tively high in CrrOr. Alkalic basalt pyroxenes arehigh in TiOz and NazO. To a lesser degree AlzOg andFezOg are low in tholeiitic basalt pyroxenes and highin those from alkali basalts. The patterns for groupsI l - l and I 5 closely follow the pattern of the tholeiiticgroup. Group l7-3 is very different from both thetholeiitic groups and the other two alkalic groups.Among the tholeiitic groups, 34-2 is unique.

Groups 34-2 (tholeiitic) and l7-3 (alkalic) do notappear to follow the patterns of the other groups oftheir types. For example, the Fe2+/(Fe'+ * Mg) val-ues for these two groups are higher than the averagepyroxene value, and are higher than the Fez+/(Fe'z+1Mg) value in any other group. Also, the overall ap-pearance of the patterns for 34-2 and l7-3 are moresimilar to each other than they are to the othergroups. Both of these pyroxene groups are fromhighly fractionated basalts, and this fact probablyaccounts for their similar patterns.

On a series of plots with various cations plottedagainst the atomic ratio Fe'z+/(Fd+ + Mg) : Xr,(Figs. 3-10) we observe two features: (l) the charac-teristics of the two major basalt types (tholeiitic andalkalic); and (2) an indication of the crystal-chemical

(ftz+/Fez+ +Mg)

Fig. 5. Ca us. Fe2+/(Fez+ + Mg) (atomic). Symbols for groupsand units are as in Fig. 3.

1612+7p62+ +Mg)

Fig. 6. Cr us. Fe'z+/(Fe'z+ + Mg) (atomic). Symbols for groups

and units are as in Fig. 3.

behavior of the various cations. For simplicity wedistinguish only three groups of pyroxenes on theplot, namely those collected from alkalic basalts,those from tholeiitic basalts (exclusive of clustergroup ll-l), and those from Site 100 of Leg ll. Thepyroxenes from the Site 100 basalts constitute all ofcluster group ll-1. It is significantly different fromthe other tholeiitic pyroxene cluster groups, con-sisting largely of analyses of xenocrysts andlorphenocrysts.

With the large numbers of analyses used, charac-teristics of the two basalt types are readily observed.Pyroxenes from alkalic basalts have a wide range ofTi concentrations with high values at low Xr" and lowvalues at high X"" (Fig. 3). In contrast Ti is positivelycorrelated with Xr" in tholeiitic basalt pyroxenes.Pyroxenes from Site 100 basalts are very low in Tiand very low in Xp".

The trend of the alkalic basalt pyroxenes (Fig. 3)can be explained with the help of Figure 3a. This

1 6"e+7 5g2++ Mg )

Fig. 7. Mn ut. p4+ /(Fez+ * Mg) (atomic). Symbols for groupsand units are as in Fig. 3.

.r.J:-ro€ 09

o l o? 03 04 05 06 07 08 09 02 03 04 05 06 0? 08 09

508

figure is the same as Figure 3 except that it shows onlythe 3 alkalic cluster groups, l7-1, 17-2, and l7-3.These 3 groups represent 2 alkalic basalt drilling sites,and as shown in Table l, groups l7-2 and lj-3 eachrepresent only one site (165,4, and 169, respectively)while group l7-l consists of analyses from bolft sites.

r " l

tclu rfl.ugl

Fig. 9. rvAl DS. Fel+ /(Fe+ * Mg) (atomic). Symbols for groupsand units are as in Fig. 3.

SCHWEITZER ET AL: CLINOPYROXENES

o o . l '

The pyroxenes of groups l7-2 and l7-3 differ in Ticontent, those of l7-2being relatively high in Ti andlow in X"", and those of l7-3 being low in Ti andrelatively high in X"" (Fig. 3a). Therefore, groups l7-2 and l7-3, reflecting the two sites, form 2 separatecurves which appear as branches of the single alkalicpyroxene curve in Figure 3. Group l7-l falls in themiddle, where the slope of the curve in Figure 3changes, at a Ti content of approximately 0.04 cat-ions, which is where the two groups of Figure 3ameet. The difference in Ti content of the groups re-flects the difference in bulk chemistry of the hostbasalts (Table 3).

The sodium contents of all the pyroxenes are ratherlow (Fig. 4), but a difference is noticeable betweenpyroxenes of alkalic basalts and those of tholeiiticbasalts. Those from alkalii basalts are generallyricher in Na than those from tholeiites. Pyroxenesfrom alkalic basalts are also slightly richer in calciumthan pyroxenes from tholeiitic basalts (Fig. 5). So-dium and calcium contents are essentially independ-ent of X"".

Tholeiitic basalt pyroxenes show an inverse rela-tion between Cr and X"" (Fig. 6). Pyroxenes from alltholeiitic basalts which appear to be phenocrysts/

o3

o o t

j o - .

- o o i

f o o ^

tb'

( Fez*/ Fe z*+Mg )

Fig. 8. Si us. Fe2+ /(Fe2+ + Mg) (atomic). Symbols for groups and unirs are as in Fig. 3.

,:,-:. \ cI .a t d

. l 'O 2

' l

t.

. te D

t .

t.

, t . o

t . '

:nI c '

: ' .

o.9o.807o'6o4o3 o5o.2o.loo

SCHWEITZER ET AL: CLINOPYROXENES 509

xenocrysts are characterized by high Cr and very lowX"" (Fig. 6). Groundmass pyroxenes from these samesites are low in Cr and tend to have high X"". Thealkalic basalt pyroxenes are low in Cr.

The pyroxenes of all three plotted groups show astrong positive correlation between Mn content and&" (Fig. 7). This trend is consistent with the com-mon assumption that Mn behaves similarly to Fe2+ inpyroxenes.

Figures 8 and 9 give some insight into the crystalchemistry of tetrahedral sites in clinopyroxenes. Avalue of 2.00 Si (Fig. 8) would indicate that thetetrahedral sites are completely filled by silicon. Inthe alkalic basalt pyroxenes Si commonly falls wellbelow the value of 2.00, with the deficiency com-pensated by IvAl (Fig. 9). As X"" increases in alkalicbasalt pyroxenes, Si also increases and tvAl de-creases. The distributions of the three cluster groupsrepresenting alkalic basalts in Figures 8 and 9 aresimilar to their distributions in Figure 3, with groupsl7-2 and l7-3 forming the low and high Xp" branches,respectively. In contrast to the behavior of pyroxenesfrom alkalic basalts, those from tholeiitic basaltsshow a much narrower range of IVAI. However, as agroup, the tholeiitic basalt pyroxenes resemble clustergroup l7-3 as regards IvAl.

Aluminum substitutes more extensively in tetrahe-dral than in octahedral sites in these pyroxenes. Fig-ure l0 provides documentation for this statement inthat the trend for IVAI (as seen in Fig. 9) pre-dominates even when combined with vIAl.

, a.z+ 7 pgz+ + Mg )

Fig. 10. (tuAl + vrAl)/Si us. Fe'z+/(Fe'z+ + MC) (atomic).Symbols for groups and units are as in Fig. 3.

vht *zTi*ct

Fig. I I . Charge deficiencies ( )/ axis) us. charge excesses (X axis).Symbols for groups and units are as in Fig. 3.

The ferric iron contents of the pyroxenes may beinferred from a plot (Fig. ll) of charge deficienciesus. charge excesses. These charge imbalances are cal-culated relative to a pyroxene of charge-balancedcomposition, or "Quad" pyroxenes. For example,substitution of Na for Ca in the pyroxene M2 site orAl for Si in a tetrahedral site creates a charge defi-ciency relative to "Quad," whereas Al, Cr3+, Fe3+, orTia+ in an octahedral site creates a charge excess. InFigure I l, lateral displacement of points to the left ofthe line Fe8+ = 0 indicates the relative amount offerric iron present. The large number of analysesfalling to the left of the Fe8+ = 0 line provides com-pelling evidence for significant ferric iron in thesepyroxenes.

C ry s t al - chemi cal charac te rist ics

The general formula for a pyroxene can be writtenXtYZ2Oa, where X represents Ca, Na, Mn2+, Fe2+,Mg, and Li in the 6-8-coordinated M2 site; Y repre-sents Mn2+, Fe2+, Mg, Fe8+, Cre+, Al, and Tia+ in the6-coordinated Ml site; and Z represents Si and Al inthe 4-coordinated T site. Thus, the pyroxene struc-ture accommodates most of the major cations thatoccur in basaltic systems. Papike et al. (1974) suggesta subdivision of pyroxenes into two major groups ofstructural chemical components which they call"Quad", and "Others." "Quad" pyroxene com-ponents are those which plot in the quadrilateralsubsystem composed of the end-members diopside

t{

+oz

0 2 0 3 0 4 0 5 0 6 0 7

5 1 0 SCHIYEITZER ET AL.: CLINOPYROXENES

Table 4. Selected "Other-Other" correlations

Cluster Group vIAl-IVAI vIFesr-IvAl vICf+-IvAl vrTir+-IvAl Na-vrAl Na-vrFes+ Na-vrcf+ Na-vrTi.+

i l - lll-2l 51 7 - lt7 -2I / - J

34-l34-237-332

Tholeiit icAlkal ic

0.460.5780.190

- 0 .318-0.406

0.6870.057

- 0.31 I0.3690.320

- 0.041

0.489-0.026

0.6050.6010.3390.2340.4040.8690.6350.5740.718

0.1480.0430.145

- 0.0860.095

- 0. t6 l0.2790.4580.24s0.056

-0 . t 85

0.5740.7530.5990.8220.9350.6770.5050.8860.57 l0.5690.965

0 . t680.4090.062

-0 .135-0 .01 I- 0.078

0 . 1 t 7-0 . r l 9

0.227- 0.032- 0.051

0.2350.0460.2690.2870.0690.0100.0530.6250.2220.3350.s2

0 . t680.081

- 0.083- 0.307

0.1960.150

- 0.0150.@40.239

-0.098-0 .2 t8

0.2860.2230.1250.268

-0.M20.3280.3800.7330.2590.4600.302

CaMgSirOu, hedenbergite CaFeSi2Ou, enstatiteMgSiOs, and ferrosilite FeSiOr. Deviations from thisreference state result in the "Other" pyroxene com-ponents. Pyroxene components in the ..Others"group contain the cations Al, Fes+, Cr8+, Ti{+. andNa. A site charge-balance equation indicates the gimportant coupled substitutions which are possible inthe "Others" group:

charge excess charge deficiencyvIAl + vlFes+ + vrcrs+ + 2vrTi4+ _ rvAl * M2Na

o 1 8

0 . r 6

When vIAl, vIFes+, or vlcrs+ substitutes in the pyrox-ene Ml site, a charge excess of *l results relatiue tothe (Mg,Fe'+) occupancy of this site in "Quad" py-roxene components. By similar reasoning, a Tia+ sub-stitution in Ml causes a charge excess of *2 relativeto "Quad." Obviously, for charge balance to bemaintained in the pyroxene structure, these sitecharge excesses must be counterbalanced by sitecharge deficiencies. The substitution of Al for Si inthe tetrahedral site causes a deficiency of - l. Simi-larly, the substitution of Na in the M2 site for(Ca,Fe2+,Mg) also results in a site charge deficiencyof - I relative to "Quad." Although the classificationof pyroxenes suggested by Papike et al. was specifi-cally designed to minimize any arbitrary sequentialcalculation of components, our new results from fac-tor analysis of pyroxenes from deep-sea basalts en-ables us to identify relative importance of the differ-ent components and thus gives us a basis for acalculation sequence.

Correlation coefficients have been calculated fromthe correlation matrices of the factor analyses forevery pair of cations studied. For simplicity, the cor-relations are separated into two groups. Table 4 pre-sents the "Other-Other" cation correlations whichcorrespond to the eight important coupled "Other"substitutions mentioned above. Table 5 presents the"Other-Quad" correlations. Silicon is not included.since it is redundant with rvAl (tuAl : 2 - Si).

Among the "Other-Other" pairs, Ti-IvAl and Fes+-IvAl are among the most strongly correlated, in-dicating that vrTia+-2lvAl and vIFes+-IvAl are thetwo most important "Others" substitutional couplesin pyroxenes from deep-sea basalts. They are ofroughly equal importance in pyroxenes from tho-leiitic basalts, but in pyroxenes from alkalic basaltsvrTir+-2IvAl is more important. Thus, important sitesubstitutions are Tia+ or Fe8+ for (Mg,Fer+) in theMl site, coupled with Al for Si in the tetrahedral site.

o-t2 . l t

ir... ..

.ogl d < b o o

o r b

02 0.3 0.4vlAl * lvAl

Fig. 12. Ti us. total Al. Units are atoms per formula unit. Seetext for discussion.

o.6o.5o. l

. f ; t . .

''f3'o

" -':.t;

-i... .' '7 .F..;1... o o r

| -

SCHWEITZER ET AL.

Table 4 summarizes important information con-cerning the most important "Others" substitutionalcouples. In pyroxenes from tholeiitic basalts the mostto least important are: vrFet+-IvAl, vrTi4+-IvAl, Na-vITi4+, Na-vIFet+, vlAl-IvAl. In pyroxenes from al-kalic basalts the sequence is: vrTi4+-IvAl, vlFes+-IvAl, Na-vIaat*, Na-utTi4+. Additional correlationcoefficients (not shown) and the results presented inTable 4 indicate that NaTi(SiAl)O. is a significantcomponent in pyroxenes from both tholeiitic andalkalic basalts. The substitutional couples NavlCrs+(ureyite) and NavrAl fiadeite) are not important inpyroxenes from either group. However, vrAl-IVAI is asignificant substitution in pyroxenes from tholeiiticbasalts but is insignificant for pyroxenes from alkalicbasalts. The high positive correlation between Ti andAl (especially for alkalic pyroxenes, correlation coef-ficient : 0.965) is il lustrated in Figure 12.

Table 5 enables us to see if these substitutionalcouples correlate preferentially with any of the"Quad" cations (Ca,Fe2+,Mg). In pyroxenes fromtholeiitic basalts Ti is positively correlated with Fe'+and negatively correlated with Mg; this indicates a Tibuildup in the pyroxenes with fractionation. Thus, inthese pyroxenes the vlTil+-2lvAl couple is more im-portant in late-stage pyroxenes. In contrast, since Tiis negatively correlated with Fd+ in the pyroxenesfrom alkalic basalts, the vlTin+-21vAl couple is moreimportant in the early-crystallizing pyroxenes fromalkalic basalts and falls off with fractionation.

Figure l3 gives additional important insights intopyroxene chemical systematics for the three majorgroups, alkalic, tholeiitic, and site 100. Pyroxenesfrom alkalic basalts plot along a line connecting vITiIvAlz and vrFes+rvAl, which are the two most impor-tant "Others" substitutional couples. In early-crystal-lizing pyroxenes from these basalts vlTirvAlz is thedominant couple, but in late-stage pyroxenes vlFes+IvAl dominates. Pyroxenes from tholeiitic basaltsplot off the TiAlz-Fe3+Al line toward the rvAl apex ofthe data display. This reflects the additional impor-tant "Others" couples in these pyroxenes, namely,Cr3+Al (especially for Site 100) and vlAlrvAl. Thus,for Site 100 pyroxenes the crystallization trend wouldbe initiated at the Ti-rvAl sideline near rvAl. Thesepyroxenes would be enriched in the Cr8+ Al and vrAlIvAl components. With increased crystallization thetrend is toward vrFes+IvAl.

Factor analysis, correlation coefficients, and sequentialcalculations of pyroxene components

Papike et al. (1974) introduced a method for char-acterizingpyroxenes that specifically avoided any ar-

CLINOPYROXENES

€

o

F

C A ,

> d

l ot>

l +

Z t ;

e . q> v

- ! +< 'i,: I !

-! ^", < g

I

< i i

i { ,v E

I

36

' r d

l ai :>

,f .q

L {zz

z Q

. t {

a>

J ..qo v

, t +r "F

I+ 4

l o Er

I

I&

- ! +-<p

b o

5i

- r € 9 d O q O O F o

$ $ e s s x G F S F Sr i : : : i a a i ^ ' A

- - o 6 9 € h q - N o6 n r 6 V d + - N O 6- - O r Q O d N € S r

o o o d o o o o o c i c jt t l l l l l l

R B K S : F ; X € 3 FS h N d h h 6 - E - d: : : A : : A A A A :

t t t l l l l l l l

= = e S s g F R g S GN N - € - € N S - - 9l : : : : i ^ ' A d ^ :

t t l l r l l l l l l- < r * 6 @ O S r 6 9

i l e K $ = ; = X a T B: : : ^ a : : ^ A d ^ '

t t t t l l l iN - F h 6 @ n A N 6 -

! 8 R n 5 € = R S ? 5^ l : : : : : : i A A

I

B:EE5X8HF=Ec t c i o c i o o o o o c i ot l l l l

F s - 6 F h - O € O Qg s s S F : + 8 8 s R: : : A : : A ; A A

t r l l l r l l l l

EESHHHHERgXo o o o o o o o o o o

l l l lN N @ r r 9 6 - 9 - -

3 e € 6 s R € 5 $ 4 8o o o o o o o o o o ot l l l l l l l l

N - d 9 r d - 6 6 - 6

8 5 9 = ; I 5 8 F F RJ : : : A A A A A A ;

l l l l l

g8:REEIEf ; :E: : : : : A A ; A A A

l r l l l

R3X€A8Fi lE=;o o o o o o o o o o ot t l t l l l l l l l

3E:BEI*5!EEo o o o o o o o o o o

l l l l l lN - N 6 6 N - $ d r O

8 8 9 8 S R e 5 9 e !c i o o o o o o o d o ot t l

r N d h @ - o @ o 9 r

3 e 9 F 3 g B K 3 F 8o o o o o o o o o o o

t t l l l l l

$R9gEEBHRSEo o o o o o o o o o dl t l l l

+ r € 9 6 r - F - o o6 - r € h O E h 6 N oO d - * n 9 6 d h N -

d c i o d o o o o o 6 ot l l l l

g = ' e- N - d - - d i ' ; d

- l - 1 6 * * * + + * = r_ _ _ _ _ E <

5 l I

512

Fe- v'Fei- ,YAl tvAl

Fig. 13. Ti -Fee+-IVAI (atomic) data display. Symbols as inFig. 3.

bitrary sequential calculation of pyroxene com-ponents. By using the methods of factor analysis andcation-cation correlation coefficients we may returnto sequential calculation of pyroxene components.

For an example of a calculation of the pyroxenecomponents, we refer to Tables 2, 4, and,5 and willapply the calculation specifically to pyroxene clustergroup l7-2. The correlation coefficients presented inTables 4 and 5 indicate that CaTiAlrO. is the most

the Wo, En, and Fs proportions of euad as 43.1,12.5, and 44.4 percent, respectively.

Figure 14 illustrates the pyroxene data plotted on aCa-Mg-Fe'z+ ternary. Cluster group l7-2 (Fig. lab)plots in a somewhat curious position on this plotabove the join CaMg-CaFe2+. At first glance onemight assume that this implies that some Ca is enter-ing the pyroxene Ml site, but this is not the case. Thecorrelation coefficients indicate that Ca is highly cor-related with vrTi and vlFes+ and thus some of itshould be included with the "Others,', not the"Quad" components. When one makes this adjust-ment as illustrated in the calculation above. we seethat cluster group l7-2 plots properly within the py-roxene quadrilateral.

Based on the results of this study the correct se-quential calculation for pyroxenes from alkalic ba-

AL.: CLINOPYROXENES

salts is caTiAlzor, cavlFes+SiAlOr, NaFee+Si2Ou,NaTia+SiAlO.. In contrast the correct sequential cal-culation for pyroxenes from tholeiitic basalts isCaFes+SiAlO., Fe2+TiAlrO., NaTiSiAlOr, NaFes+SirOu, CavrAlSiIv,4,106, MgvICrs+SiAlOu.

Conclusions

Differences in pyroxene chemistry reflect chemicaldifferences in the types of basalt in which they occur.Pyroxenes from alkalic and tholeiitic rocks differ pri-marily in the abundances of the components CrrOr,TiOr, and NarO. With increased fractionation [asmeasured by an increase in Fe'z+/(Fe,+ * Mg) ratio]several trends in these components are observed.Thus, in the pyroxenes from tholeiitic rocks there isan inverse relation between fractionation and Cr con-

CoFe.

F"i

Fig. 14. (a) Pyroxene data for tholeiitic basalts plotted on Ca-Mg-Fg'z+ (atomic) ternary diagrams. The upper diagram is forpyroxenes from Leg I 1. Triangles represent cluster group I l-l andopen circles I 1-2. Tbe lower diagram illustrates combined data forcluster groups 15,34-1,34-2,37-332. (b) Pyroiene data for alkalicbasalts plotted on a Ca-Mg-Fer+ (atomic) ternary diagram. Threecluster groups are indicated.

SCHWEITZER ET

Mgz

M 9 .

vlFel+ lvAl

SCHWEITZER ET AL.:

tent. In pyroxenes from alkalic rocks Ti content isinversely related to fractionation.

Using correlation coefficients and crystal-chemicalreasoning, we infer from our data that substitutionswhich commonly occur in these deep-sea pyroxenesare: Al for Si in the tetrahedral site, balanced byeither Ti or Fe3+ for Fe2+ or Mg in an octahedral Mlsite.

Fresh pyroxenes, even in highly altered basalts,provide clues about the original chemical character oftheir host rocks.

AcknowledgmentsK. Baldwin is acknowledged for his advice on the use of the

computer programs. Thorough reviews by M. Bass, E. Dowty, andF. Chayes markedly improved the manuscript. This research wassupported by NSF grant OC87622193A01 (AEB) and NASANSG-9M4 (JJP), which we gratefully acknowledge.

References

Ayuso, R. A., A. E. Bence and S. R. Taylor (1976) Upper Jurassictholeiitic basalts from DSDP Leg ll.J. Geophys. Res.,81,4305-4335.

Bass, M. N., R. Moberly, J. M. Rhodes, C. Shih and S. E. Church(1973) Volcanic rocks cored in the Central Pacific, Leg 17, DeepSea Drif ling Project. Inilial Rep. Deep Sea Drilling Project, 17,429-503.

Bence, A. E. and A. L. Albee (1968) Empirical correction factorsfor electron microanalysis of silicates and oxides. J. Geol.,76,382-403.

- x1d J. J. Papike (1972) Pyroxenes as recorders of lunarbasalt petrogenesis: Chemical trends due to crystal-liquid inter-actions. Proc. Lunar Sci. Conf. 3rd, 431469.

and R. A. Ayuso (1975) Petrology of submarinebasalts from the Central Caribbean: DSDP Leg 15. J. Geophys.Res.,80, 47'154804.

Bender, J. F., F. N. Hodges and A. E. Bence (1978) Petrogenesis ofbasalts from the Project Fnnaous area: experimental study fromI atm. to 15 kb. Earth Planet. Sci. Lett. ,41,277-302.

Blanchard, D. P., J. M. Rhodes, M. A. Dungan, K. V. Rodgers,C. M. Donaldson, J. C. Brannon, J. W. Jacobs and E. K. Gibson(1976) The chemistry and petrology of basalts from Leg 37 oftheDeep Sea Drilling Project. J. Geophys. Res., 81, 4231-4246.

CLINOPYROXENES

Christofersson, E. (1976) Colombian Basin magnetism and Carib-bean plate tectonics. Geol. Soc. Am. Bull.,87, 1255-1258.

Dixon, W. J., Ed. (1975) BMDP Biomedical Computer Programs'University of California Press, Berkeley, California.

Dmitriev, L. (197'l) Petrochemistry of basalts and plutonic rocks,Leg 37 , Deep Sea Drilling Project. Inilial Rep. Deep Sea DrillingProject, 37 , 681-693.

Donnelly, T. W. (1973) Late Cretaceous basalts from the Carib-bean: a possible flood basalt of vast size (abstr.). EOS, Trans.Am. Geophys. Union, 54, 1004-1006'

Edgar, N. T., J. B. Saunders et al. (1973) Initial Reports of the DeepSea Drilting Project, vol. 15. U. S. Government Printing Office,Washington, D. C.

Grove, T. L. and A. E. Bence (1977) Experimental study ofpyrox-ene-liquid interaction in quartz-normative basalt 15597. ProcLunar Sci. Cod. 8th, 1549-1579.

Hodges, F. N. and J. J. Papike (1977) Petrology of basalts, gab-

bros, and peridotites from DSDP Leg 37. Initial Rep' Deep SeaDrilling Project, 37, 711-723.

Hollister, C. D., J. I. Ewing, et al. (1912) Initial Reports of the DeepSea Drilling Project, vol. ll. U. S. Government Printing Office,Washington, D. C.

Kushiro, I. (1973) Origin of some magmas in oceanic and circum-oceanic regions. Tectonophysics, I 7, 2l l-222.

Mazzullo. L. J. and A. E. Bence (1976) Abyssal tholeiites fromDSDP Leg 34: the N azca Plate. "/. G eophy s. Res., 8 l' 4327 -435 l.

Myers, C. W., A. E. Bence, J. J. Papike and R. A. Ayuso (1975)Perrology of an alkali-olivine basalt sill from Site 169 of DSDPLeg 17: the Central Pacific Basin. ,/. Geophys. Res.,80,807-822.

Naney, M. T., D. M. Crowl and J. J. Papike (1976) The Apollo l6drill core: Statistical analysis of glass chemistry and the charac-terization of a fligh llumina-'Silica Poor (HASP) glass' Proc.Lunar Sci. Conf.,7, 155-184.

Papike, J. J., K. L. Cameron and K. Baldwin (1974) Amphibolesand pyroxenes: characterization of othe r than quadrilateral com-ponents and estimates of ferric iron from microprobe data(abstr.). Geol. Soc. Am. Abslracts wilh Programs,6, 1053-1054.

Schweitzer, E. L., J. J. Papike and A. E. Bence (1978) Clinopyrox-enes from deep sea basalts: a statistical analysis. Geophys. Res.Leu.. 5. 573-576.

Yeats, R. S., S. R. Hart et al. (1976) Initial Reports of the Deep SeaDritling Project, vol. 34. U. S. Government Printing Office,Washington, D. C.

Manusuipt receiued, JulY 3, 1978;accepted for publication, December 9, 1978.

5 1 3