American Mineralogist, Volume 66, pages 526-546, l98l
Ion microprobetechniques and analysesof olivine and low-ca pyroxene
I. M. SreBrE, R. L. HnRvIc, I. D. HurcHEoN AND J. V. Snanu Department of Geophysical Sciences, (Iniversity of Chicago Chicago, Illinois 60637
Abstract
Experimentalconditions for major, minor, and trace elementanalysis of olivine and low- Ca pyroxene are describedand analytical accuracy tested using suites of natural samples spanninga wide range of Mg/Fe. Special attention was given to sample cleanlinessto avoid contamination, instrumental vacuum to eliminate hydrides in the secondaryion spectrum,and samplepreparation to give good precisionin measuredintensities. Especially bothersome were molecular interferences in the secondaryion spectrumwhich were separatedfrom analytical peaksusing massresolu- tion (M/AM) over 3000.Careful analysisof the secondaryion spectrumallowed choiceof ei- ther high or low massresolution depending on the presenceof interferences.Lithium (0.005), F (l?), Na (0.01),Mg, Al (0.1),Si, P (5), K (0.05),Mn (0.1),Fe, and Co (l) were analyzedat low resolutionwith detectionlimits (ppm) in parentheses.Elements requiring high resolution includeCa (l), Sc (l), Ti (l), V (l), Cr (l) and Ni (5). The secondary-ionintensities for Mg and Si do not correlate linearly with composition whereasFe is nearly linear. The simple relation of the Mgl(Mg + Fe) ratio to the measured secondaryion ratio Mg*/(Mg* + Fe*) enabled major element determination to within +l mol% of forsterite or enstatitecontent. The yield of Mg and Fe secondaryions is a complex function of composition,and Mg*/Si* and Fe*/Si* are not simply related to atomic ratios in the target. To test accuracyof minor elementdetermination as a function of major elementvariation, secondary-ioniatensities were compared with compositionsbased on electron probe mea- surements.Some elements (Al, Cr, Ti, Mn) give a linear relationshipwith no obvious matrix eflect,but Ni, and possiblyCa and Co, definitely dependon the matrix. The linear relation- ship allows compositiondetermination to within tl\Vo of the amount presentby referenceto known standards.A major efort is required to calibrateLi, K, Na, F, v and Sc for which few reliable standardsexist.
Introduction analysis)are available for analysisof trace elements, but most of them require mechanical separation of The development of the electron nicroprobe in the material to be analyzed.The ion microprobe was 1960-1970brought about a revolutionin the studyof recognizedaround 1970to be a potential "chemical rocks and minerals: indeed, it could be called a microscope"for many trace elements,and qualitative "chemical microscope" becauserapid non-destruc- analysesrapidly became routine especially in the tive analyses possible became at spotsselected by op- study of metals and doped semiconductors.Quan- tical study. It was no longer necessaryto separate tification of analysesof materials has posed severe minerals for analysesof major and minor elements practical and theoretical problems. Although these with atomic number (Z) greaterthan 10, and chem- problems are not fully solved,we now describeana- ical zoning could be studiedwith a spatial resolution lytical techniquesfor.reliable ion microprobe analy- near I ;.rm.However, light elementscould not be ana- sesof various trace elementsin olivine and low-Ca lyzed, and the continuous X-ray spectrum placed a pyroxene.Particular emphasisis placedon the useof detectionlimit near 40 ppm (2o) for elementswith Z high massresolution for identification and separation > 10. Many techniques(e.g., spark-sourceand iso- of massinterferences. tope-dilution mass spectrometry;neutron activation The ion microprobe utilizes a focusedprimary ion oo03-004x/ 8 | /0506-0526$02.00 526 STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE 521 beam to sputter a small volume of material from a Sampleselection and preparation targeUsome sputtered atoms are ionized and these secondaryions (SI) are analyzedwith a mass spec- Natural olivines lie very close to the binary join trometer. The extremely low background in a SI (Mg,Fe)rSiOo,but severalelements including Ni, Ca massspectrum allows analysisof trace elements,and and Mn are at high enough levels for comparisonof the focusedbeam can be positioned carefully on the ion and electron probe analyses.Low-Ca pyroxenes samplesurface. Although the potential of this analyt- lie fairly close to the binary join (Mg,Fe)SiO' and ical tech-niquehas been discussed(e.9., Evans, 1972; severalelements, especially Al, Ca, Mn, Cr and Ti Liebl, 1975;Werner, 1975),relatively little has ap- are readily measuredwith an electron microprobe. pearedregarding its potential for analysisof geologi- Chemically-analyzedlow-Ca pyroxeneswere usedby cal materials. To provide a fundamental test of the Howie and Smith (1966)to test correctionprocedures technique with typical applications, we have em- for the electron microprobe, and a selectedsuite of barked on a systematic study of important rock- thesepyroxenes was mounted for ion probe analysis. forming minerals using an AEI IM-20 ion probe with Olivines were selectedfrom those analyzedby Sim- high mass-resolution(Banner and Stimpson, 1975). kin and Smith (1970) and Smith (1966). Several Thesespecific studies follow severalyears devoted to other homogeneousolivines were used as well as an instrument developmentand to practical experience extensive suite of Mg-rich olivines and ortho- in analysisof geologicsamples. pyroxenes from mantle-derived peridotites which By far the greatesthindrance to measurementof had been carefully analyzed with the electron probe raw SI intensity dzta are molecular interferences (Hervig, 1980).Choice of sampleswas basedmostly which often are far more intensethan an analytical on grain size and absenceof visual inclusions. Be- peak. For example,in mafic silicates24Mgt6O+ occurs cause only six one-inch diameter samples can be at the samenominal mass(: 40) as the most abun- placed in the sample chamber of the ion probe and dant Ca isotope.Thus measurementwith a massres- becauserapid sampleexchange is not feasibleif vac- olution sufficient only to separateadjacent nominal uum is to be maintainedat <10-'torr, grain mounts masseswill give incorrect intensity data. with up to 36 samplesin each sample holder were Two contrastingtechniques are currently available prepared. A second important consideration in to reducethe effectsof molecular interferences.Her- choosing this mounting method is that the relative zog et al. (1973)and Shimizu et al. (1978) described ion intensitiesappear to be sensitiveto the physical application of energy filtering to increasethe signal environment(holes, cracks, etc.) rn the vicinity of the to interferenceratio while Bakale et al. (1975) and analysispoint; thus a rather large insulating surface Steeleet a/. (1980)described use of high massresolu- with few defects is required. Sample mounts were tion to separateanalytical peaks from interferences. preparedby scribing a grid of 2 mm squareson one Becausethe AEI Itvt-20ion microprobe allows only side of a l-inch diameter fused silica disk in a 6 x 6 limited energyfiltering (Steeleet al.,1977a), we pre- array. A thin layer of epoxy was placed on the non- fer to use high mass resolution when interferences scribedside of the silica disk, and grains were placed are present. within individual squaresduring observation with The precision and accuracy of measuremenJsis transmitted light under a stereomicroscope.Each crucial to use of the ion microprobe. Central thereto grain was checked for visual defects while in the is the influence of the mineral matrix on the sput- epoxy, and removed if necessary.With practice, 36 tered ion intensitiesand any systematicinstrumental sampleseach of -5 grains were mounted in a two effects.To attempt to answerthis questionwe assem- hour period. Each mount was ground to thin-section bled a suite of natural samplesof olivine and low-Ca thickness and polished. To avoid possible surface pyroxene to span nearly the entire range of Mg/Fe. contamination (seebelow), diamond polishing com- After careful electron-probeanalysis for all detect- pound was made by mixing gradeddiamond powder 'Aaselins". able elements we systematically examined the SI with commercial fill commercial dia- spectrumto determineanalytical conditions for each mond pasteswere found to have high levelsof K, Na, element as well as other elementsnot detectedwith CL Si, F, Ca and other elements.Commercial vase- the electron probe. We then compared ion-probe line was found to be contaminant-free at detection count rates against electron-probemeasurements to levelsof the electronmicroprobe. Because of possible test the quantitative ability of the ion-probe tech- hydrocarbon contamination and degassingunder nique. high vacuum, thick epoxy mounts were not used; STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE
standardthin sectionsdo not noticeably degradethe cies derived from the duoplasmatronitse$ and vacuum. A reflected light photograph was taken of from possibleleaks, but at the expenseof a re- each sample and any particular featuressuch as in- duced intensity. clusionsand exsolutionlamellae were noted. (2) The primary beam has an impact energy of Before analysisby the ion probe, selectedgrains of 20kV; higher potentialsmight increasethe sput- each sample were analyzed with an ARL-EMx-sM tering and ion yields but are more susceptibleto electronprobe. Operatingat 25 kY, with a high beam electricalbreakdown. current (2p"A) and,a spot diameter (-25pm) com- (3) The condenserlens is adjusted to give a l0nA parable to that of the ion probe, analysesfor major primary beam current at a Faraday cup in the elementsas well as Na, Al, Ca, Ti, Cr, Mn and Ni sample position from a maximum observed were obtained for both olivines (Table l) and pyrox- value of 90nA. This adjustmentprovides greater enes (Table 2) with detection levels (2o) below 100 stability in the primary beam at the sample be- ppmw. Standards used were CarPrO, (66.9 wt.Vo causethe beam becomesinsensitive to the vari- PrOr), Corning V glass (0.80 wt.VoTiOr, 016 wt.Vo ous voltage fluctuations prior to the final aper- CrrO.), diopside.r-jadeite,, glass (22.0 wt.Vo CaO, ture. Also by reducing the primary beam, the 3.78trt.VoAl2O3, 2.30wt.Vo NarO), P-140 olivine (7.41 overall ion-optical propertiesallow better focus- wt.VoFeO), Corning W glass (0.64 wt.VoMnO) and ing of the primary beam to a routine diameter CorningX glass(0.71wt.Eo NiO). Analysesof low-Ca of - 15pm. The l0nA beam current was chosen pyroxenes agree satisfactorily with those given by to allow good SI intensities for a reasonably Howie and Smith (1963).Analyzed sporswere noted small spot. Becausevariations in primary beam on photographsfor later ion-probe analysisbecause intensity cannot be monitored during analysis, the pyroxene grains differed slightly in composition. the current is measuredbefore and after analy- ln all, 7 olivine and 16 low-Ca pyroxene samples sis by moving the sample from the beam path were selectedfor ion probe analysis in addition to and collecting the beam in a Faraday cup. Ob- many samplesfrom peridotites. The preparation of serveddrifts of primary intensity are less than the thin section and electron microprobe analyses 2Voof the total over 20 minutes. added considerablesurface contamination and a (4) Although rasteringof the prinary beam in con- cleaning procedure was required before ion probe junction with an electronicaperture should pro- analysis. After final polishing with diamond com- vide a cleaner target surface without edge ef- pound, surfaceswere cleaned with Q-tips and ace- fects, we prefer to use a stationary focused tone followed by carbon tetrachloride and immedi- beam. This gives a faster stabilization of the SI ately coated with a conducting gold layer of -25A signal and doesnot consumeas much material. and storedin a dessicator.Gold is preferredover car- Uncertaintiesor changeswith time of edge ef- bon becausethe Au layer can be sputteredfaster, re- fects in the periphery of the focusedbeam can ducesthe possibility of forming hydrocarbonsduring be correctedby periodic referenceto standards. sputtering, and does not contribute to the SI spec- (5) The sampleis observedwith both reflectedand trum. transmitted light. Specialcare in focusing with high power (150x) reflectedlight is necessaryto Analytical conditions ensurethat the analysispoint remains constant with respectto the SI extraction system. Possiblythe most difrcult aspectin comparing ion (6) Extraction of SI is provided by a number of fo- probe results between different laboratories is the cusing and deflecting electrodesbetween tl.re wide variety of instrumental conditions. Below are sampleand sourceslit of the massspectrometer. given our routine operating parameterswith some justifications. The observedintensity ratios of S[ and the posi- tion of the primary ion beam are greatly influ- (l) The primary beamis negativelycharged oxygen encedby the potentialson theseelectrodes. As a which was shown (Aadersen,1969) to generally set procedure all controls are adjuritedto give give high yield of positive SI for insulating sam- maximum Si ion yield. The settingsare nearly ples. Primary beam intensities in excessof constant from day to day. Adjustments in the l00nA are routinely obtained from the duoplas- repeller and extractor potentials afect the posi- matron source.The beam is mass analyzed to tion of the primary beam and are fixed during ''O- provide a pure beam by exclusionof spe- routine measurements.Deflection/centering ad- STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE s29
Table l. Compositions(w.Vo) and secondaryion intensities for (7) The sourceslit is adjustablefrom fully closedto olivine 0.025cm and, in conjunction with an adjustable collector slit, controlsthe massresolution of the massspectrometer. For low resolution,the col- sio 2 41.3 38.8 36.8 35.5 30.8 29-1 29.6 '1 lector slit is openedfully and the sourceslit ad- FeO 4 19.0 3A.t 38 3 60 1 65.9 61 6
MnO o 10 a.25 0 39 0.56 I10 0.91 2 -30 justed to provide no more than about 106cts/sec M9o 50.9 41-.9 32.5 26 4 6 -43 2 -51 0.08 for the most intense peak to avoid damage to NiO o 33 0 t2 0.06 0.o1 0.00 0 00 0.00 the first dynode of the electronmultiplier and to CaO o-ao4 0.19 a.oA 0 03 0 16 0-06 o.10 avoid serious dead-time corrections. Under Tio nd O A29 0.014 nd .O63 nd i,010 2 peakshave a top that is flat to ol:0 nd 0.040 0.007 nd ,r.OOg ,i.AA1 theseconditions 3 Na2U nd 0,0I 1 nd nd nd nd n(l +l%o of peak height. The extreme range in cr0 nd O.Ol4 nd nd nd nd nd 23 count rates betweenmajor and minor elements
I 100.034 100.356 100.511 100.ao 99 2r:I 99 -\41 99.a.96 precludescollection of data under identical slit settings for both; however, some major ele- Catlons ,/ 12 OxYg.'ns ments, such as Si, have minor isotopeswhich Si 3.003 2-913 2-9A() 2 981 1.001 2.99I 1.011 can provide a SI signal of acceptablecount rate 0.450 l 218 2 -Af9 2 .6)0 1 946 5. 550 5 15I with slits set for maxirrum sensitivity. Mn o 006 o 016 o-021 o-040 o o9l 0.083 r).I98 Mg t -5I1 4 la6 3.923 I 304 A.934 0.)11 t).nr2 (8) Although the mass spectrometerallows photo-
NI o. 019 o.00? o.o04 al-001 0 000 0-000 0 00/) graphic recording of SI spectra,exclusive use is made of electrical detectionwhich incorporates o 000 0 016 a.oo4 0 003 0.017 0-007 o 011 20-stageelectron multiplier, preamplifier and Ti 0 000 o a)02 0.oot 0.o00 0 005 0.ooo 0-001 a
AI 0 .000 o.oo4 0.001 0-000 0 001 0.oo1 0 000 scalerwith a 100 megaHertzcounting capabil- o.000 0 002 0.oo0 0 000 0.000 0 000 0.000 ity. Cr 0 .000 0.0o1 o 000 o-oo0 0 000 0 o00 lt.0(x) (9) Magnetic peak switchingat low massresolution
MS./(Mg+Fc ) 92 16 t-2 is automatic using an on-line Data General comput€r and an analytical relation between field intensity sensedby a Hall-effectprobe and si 44O lA4 211 246 362 192 39(, secondarymagnet current. Peak switching over FE t45 461 94A I t 60 25Ia) 2994 lO40 Mn I 10 9 88 tf).5 23-3 57.8 \1 A )-32 a 10 amu (atomic massunit) range and stabili- Mg 4320 414t) 423t) 3624 10 20 434 I l Ni 1 56 t 06 0.a,9 0.15 010 0.O1 nd zation of the magnetic field requires about 5 0 0l 9.38 I 65 t.t3 l -60 | 19 4.2A seconds.Because the presentelectronic configu- Co 0-26 0 51 1.10 0.99 4.62 o-28 0 0I Cr I 1I0 nd nd nd nd nd ration does not allow precisemagnet scanning nd2I nd nd nd nd high mass resolution, this is done man- Ti a 260 60 4 lro 50 1l under
SC 318 10 3 ually with manual recording of data. Instru- AI f3 2ItA 400 55 700 40 fa K 895 13 22 15 t3 mental modification for scanningat high resolu- NA 45 6100 72\ 106 265 118 105 tion is in progress,using the designdeveloped at L1 t00 190 1120 2750 1080 11 50 166A P 251 13 1l 63 Cambridge,England (Long et al.,1980). (10) To allow for long-term drift and to provide an * Samples: 1. P-I40 Balsam Cap N.C dunite; 2 TS.4 = SCl.l 3 (Srmkln and Smith, 1970); 3. TS-8 = KI-2O99 (Sinkin and Snith, 197.1) t indication of precision, standard samples are 4. rS-9 = YS-Il (Smith, 1956); 5. YS-5 (Smith, 7966);6- rS-tO = 1282 (simkin and smith, l97O; 7 R3517 = ys-I (smtth, 1965). analyzed at regular intervals. Such samples are a* Counts/scc x IO-3 except foJ Cr, V, Ti, Sc, A1, K, Na, Li,'p whicl, chosenfor homogeneity,grain size, and chem- are counts/sec. Cornt rates nornalized to 100% isotopic abundancc. hw mass resoluticn: Si, Ee, Mn, !19 Session 1- Mn. Co, AI, K, Na, Li, ical proxirrity to the sample type being ana- P Sessior, 2 - normalized to Mn of Scssion l. High nass resolution: Ni, Ca, Cr. V. Ti, Sc lyzed.A sequenceof analysesis ignored if count nd - not detected. rateson the standardschange by more than 5Vo. These standards also allow comparisonsfrom day to day for which the relative count rates justmentsjust before the sourceslit are sensitive may differ appreciably(+25Vo) even though an- to the initial energy of the particular SI as well alwical conditions are nominallv identical. as to any misalignmentof the extraction optics with respect to the analyzed point and source Aside from the basic principle, important differ- slit. Theseare changedfor eachmeasurement to ences which are not always apparent exist between give a maximum SI signal for each ion type. ion and electron probe tech:riques.(l) the SI spec- 530 STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE
Table 2. Compositions(wt.Vo) allld secondary ion intensitiesfor low-Ca pyroxenes
*l 11 I2 13 I5 16
sio2 56. B 57.8 54.7 55.1 55.8 53.5 52.4 53.5 5r.2 52.8 49.9 50.6 50.7 49.0 49.r 41,3 T102 0 .10 0.00 0.08 o.r2 0.11 0.06 0. 14 0.04 0.08 0.06 0.04 0.12 0.09 0.07 0.r3 0.03 o1203 0.95 0. 31 4,76 1.27 r.68 3.86 r .80 1.07 3.11 0.82 3.00 1.51 0.62 o,99 0.60 1.00 Cr2Oj 0. 18 0 .12 0 .08 0.50 0.16 o .22 0.00 0.01 0.00 0.00 0.03 0.01 0.00 0.00 0.00 0.00 Feo 4.40 6.60 r0.0 10.7 r3.1 t4.4 22.O 24.1 24.2 24.7 2a.o 29.9 31.0 34.8 36.2 4r.5 MnO 0.09 0 .09 0 .28 0,23 0.24 0.30 0 .62 0.73 0.16 0.87 0.57 0.53 r.O2 0.66 0.90 0.91 Mgo 35.0 15.2 3L6 30.6 30.0 21.4 23,2 22.9 2I -2 2l-.6 18.6 17.8 16.3 ll.7 12.6 8.8 N10 0 .11 0.07 0.06 0.07 0.02 0.09 0.00 0 .0r 0 .01 0.01 0.01 0.01 0.00 0.01 0.00 0.00 CaO O.47 0.07 0.41 1.20 0.46 0.66 0.63 o .29 0. 30 0.60 o .29 0 .80 0 .-12 0.50 0.93 0.9r Naro 0. L5 0.00 0.00 0.01 0.c0 0.01 0.01 0.00 0.o2 0.02 0.00 0.01 0.01 0 .01 0 .o2 0 .01
I 98.45 100.26 r01. 39 r00.40 101.57 100.90 100.80 r02.65 100.88 101.48 100.44 101.29 100.46 99.74 100.48 100.46
Cal1ons / 12 Oxygens sl 3.950 3.975 3.790 3.920 3.909 1.806 3 8i6 3.910 ).a20 3.926 3 811 3.869 3.918 3.901 3.92t 3.888 Tl 0.005 0.0 0.004 0.006 0.006 0.00:i 0.008 0.0o2 0.004 0.003 0.002 0.007 0.006 0.004 0.008 0.002 A1 0.078 0.025 0.140 0.105 0.139 0.324 0.157 0.092 0.214 0.012 0.270 0.136 0.057 0.093 0.057 0.091 cr 0.021 0.006 0.004 0.028 0.009 0.012 0.000 0.001 0.000 0.000 0.002 0.001 0.000 0.000 0.000 0.000 Fe 0.256 0.380 0.580 0.630 0.767 0.857 1.360 r.41j 1.510 r.516 1.788 1.9t2 2.0I4 2.32r 2.4ra 2.A53 Mn 0.005 0.005 0.016 0.014 0.014 0.018 0.039 0.045 0.048 0.0s5 0.037 0.034 O.067 0.045 0.061 0.064 Ms 3.628 3.608 3.264 l.2rO 1.7)2 2.948 2.556 2.495 2.357 2.394 2.1r7 2.029 1.A87 1.628 1.500 1.078 Ni 0.006 0.004 0.003 0.004 0.00r 0.005 0.000 0.00r 0.001 0.001 0.001 0.001 0.000 0.00r 0.000 0.000 ca 0.035 0.005 0.o32 0.090 0.014 0.050 0.050 0.023 0.o24 0.048 o.o24 0.066 0.060 0.043 0.080 0.080 Na 0.020 0.000 0.000 0.001 0.000 0.001 0.001 0.000 0 003 0.003 0.000 0.002 0.002 0.002 0.003 0.002 En 92.6 9o-4 84.2 81.7 79.6 16.5 64.5 62.5 60.6 60 2 53.9 50.6 4t.6 40.8 3t.5 26.9 Fs 6.6 9.5 15.0 16.0 19.5 22.2 34.2 36.9 18.8 38.6 45.5 47.7 50.8 58.1 60.5 77.1
Sccondary ion intensicy** si 285 3t4 2a5 29t 312 214 240 214 238 25? 246 246 255 241 262 267 Ii 990 120 1010 1560 880 570 1570 550 850 950 450 1050 920 960 ' 1180 2aO A1 24.3 9.62 108 4r.0 59.2 r29 61.1 l3.r 107 31.6 r09 40.1 18 5 32.5 20.5 38.4 Cr 1650 990 150 5800 1880 2ilo 2 240 25 25 320 78 8 567nd l'(, 59 84 r24 159 2tO 2I5 328 316 180 404 447 472 5r9 545 605 684 Mn t.80 ).71 6.1 5.11 6.r 6 . Br Il.2 t1 I 15.5 20.6 11.5 13.t 23.6 15.2 22.2 24.O 1550 1510 rl50 1530 r610 1390 1160 1270 1050 1160 1000 918 873 109 109 518 Ni 725 387 187 404 r04 544 217 126 98 23 66 88 38 109 41 50 Ca 1.55 0.21 1.15 5.01 1.56 r .97 2.52 1.28 1.08 2.05 0.93 | .94 2.44 1.67 3.42 2.75 Na l].2 0.48 (i-26 7.19 2.00 2.93 2.04 0.99 1.62 2.40 1.55 3.82 2.15 o.42 2.9I 2.O1, Co 28 t9 46 tJO 68 100 10B 1l0 r19 19 108 122 r29 116 92 71 58 :]9 99 r10 I 20 7r 43 91 r00 15 170 61 30 110 21 nd Sc 61644 89 68 40 78 a6 170 53 61 48 160 90 180 34 K 240 960 410 6100 580 430 450 1200 25(\ 340 440 38000 260 400 310 580 Li 740 2900 6800 1200 7200 5800 1300 3900 9900 2600 7900 1400 4600 2200 6400 410 P 1l 12 22 /18 21 6 916 510 224 464
*SamP1('s: 1.7)-165;2. SAc; l. R62; 4. Yl8; 5. EzL;6. H66; 7. Y9; 8. Rl3b; 9. SP18; 70. ?.2270i 1L.294I,72.7286; lf. 137; 14. 4642L: 15. 115: 16. 1002. ;l't(iounEs/scc x l0-4 excepE tor K, Ti, Cr, Ni, Co, \,, Sc whiclt arc counts/sec. A11 count rates normlized to l00Z isotopic abundance, l,ow mass rcs()lution: Si, At, rr', Mn, Mg, Na, K, Li, P from one analysis session. High mass resolution: Ti, Cr, Ni, Ca, Co, V, Sc fronr sccond analysis scssjon. nd: noL dctcctcd trum containsno equivalent of the intensewhite ra- ing statistics.(2) Whereasthe ratio of X-ray intensity diation background which statistically limits preci- to concentration varies little for different elements sion with the electron probe. Ion probe precision is with I | < Z < 26, SI signalsrange over 5 orders of only limited by instrument stability and peak count- magnitude resulting in a wide range of sensitivities STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE 531 for different elements.(3) An X-ray signal stabilizes quickly after the electron beam impinges on the Olivrne sample,whereas the ion beam requires-5 minutes to 73-76 producea stableSI signal.Thus points cannot be an- alyzedat a rate comparableto that possiblewith the electron probe. Rastering of the primary beam re- quires considerabletime to give a stable SI signal from a large area, whereasimaging with an electron probe is rapid. (4) Sample change requires several hours for the ion probe in order to maintain a clean vacuumof l0-8 torr. In practice,approximately 30 mineral analysesfor -10 elementsare possiblein a l0 hour day. Because many elementscan be analyzedunder low massreso- lution, whereasothers require high mass resolution to separateinterferences, two sessionsare required to 45 obtain a complete analysis; reproducible, rapid 50 49 4B 47 46 switchingfrom low to high massresolution is not fea- Fig. l. Low resolutionmass scan oftitanium isotopesin73-76 sible. olivine (Fo 94,TiO2 = 30 ppm, Na2O : 120 ppm) from garnet lherzolite.Isotopes at n/e : 50, 49, 48, and 46 clearly have large Mass spectrumof Mg,Fe silicates interferencesbecause m/e: 47 is very weak. Insets show high mass resolution scans illustrating that even m/e : 47 has an The first step in analysisof a mineral matrix is an interferencethought to be NaMg*. examination of the SI massspectrum. As a guide to possible interferencesa computer-generatedlist is zero (Fig. 2). The simplestexplanation was that some first made for a specific set of matrix elements.For sampleshave an interferenceat m/e : 47, and in- olivine and low-Ca pyroxene a matrix was prepared deed high-resolution scans on sampleswith excess for Mg, Fe, Si, Al, Ca and O. Although the most counts at m/e : 47 on Figure 2 showed a doublet abundantisotope is the obviouschoice for analysis,it (Fig. l). The only plausible interferencewas might be preferable to collect intensity data at low "Na2oMg* and indeed the Na content of olivine cor- massresolution for a weaker isotope if the attenua- relates qualitatively with the excesslow-resolution tion of transmissionof the abundant isotopeat high signal at m/e : 47 on Figure 2, as indicated by the massresolution reversesthe advantagein sensitivity. dashedline for sampleswith Na > 25 ppm. This in- The absenceof an interferencefor a weaker isotope terferencecould have been overlooked quite easily is more difficult to confirm at high mass resolution becauseNa is absent at the 40 ppm (2o) detection becausethe transmissionis lower, and a wide range level of the electron microprobe in most olivines of compositionsmust be examinedso that an inter- crystallizedat low pressure(<-10 kbar), and only ferenceis not overlookedwhen it occursonly in some olivines from the deep-seatedperidotites display Na specimens. substitutionconsistently. The "Na2oMg* interference Theseparticular problems are illustrated by mass is significant becauseof the very high yield of sec- scansfor the five Ti isotopesin an olivine (Fig. l). ondary ions involving alkali metals.Titanium analy- From the known isotopic ratios for n/e (: chargeto sesmust be obtained at high massresolution for Na- massratio) of 46-50(8.0, 7.8, 73.4, 5.5, 5.3Vo),larye bearing olivines.For Na-free olivines,low resolution interferencesat 46, 48, 49, and 50 must have raised analysiszt m/e: 47 gives a higher sensitivity than the intensity relative to that of 47. From the com- high resolution at m/e : 48. puter-generatedlist of mass defects(Table 3), most Whether the analytical technique useshigh mass interferencescan be assignedeasily: e.g.,'oMgi at m/ resolution or energy filtering to suppress inter- e : 48. Although m/e : 47 appearedto be inter- ferences,it is necessaryto first identify the magnitude ference-freefrom both the absenceof computer-gen- of the interference.Table 4 summarizesthe system- erated interferencesand high-resolution scans, in- atic examination of interestingelements represented tensity data obtained at high resolution for m/e: 48 in the spectra of (Mg,Fe) silicates.Undoubtedly and low resolution for m/e : 47 for a range of oli- other elementscan be detected.Some general fea- vines did not fall on a single line passing through tures are: (l) for elementswith Z < 12, interferences STEELE ET AL: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE
Table 3. Interferencesfor Ti isotopesfrom mafic silicate tonic but non-linear decreasefrom Fe-rich to Fe- poor compositions,for Mg there is a maximum SI C Major-elementsystematics Olrvine l 200 o No( 25ppm The variation of major-element SI signals with a 0) e 25(No(75 composition should provide tests of postulated cor- E o No ) BOppm rection proceduresand theoriesof interactionswhich or, occur during sputteringand ionization. Although the -c_ a same information can be obtained from minor ele- q) ro0 ments,major-element data are much easierto usebe- a cause of high a count rates, negtgible interferences O and well-known composition either from stoichio- L_ metry or electron-probeanalysis. co $ "/" The simplest presentationis a plot of raw count //e rate againstthe atomic concentrationas in Figure 3a, nlr "o ro0 200 b where Mg*, Fe* and Si* are plotted against mole%o 47-.' forsterite or enstatite(mg: Mg/(Me + Fe) for oli- I r (cts/sec)low resolutron vine and pyroxene,respectively. Although the atomic Fig. 2. Comparisonofsecondary ion count ratesfor aETiat high : fraction of silicon is near-constantfor all olivines. the resolution and m/e 47 at low resolution for a suite of olivines from peridotitesra.ging from Fo66to Foro. The dashedline was SI yield changesdrastically with a minimum near mg : fitted by eye betweenthe data for sanples with Na > 25 ppm. 0.65and a greateryield for the Mg than for the Fe Because it does not pass through the origin, an interference is end member.Whereas Fe count rates show a mono- implied at m/e = 47. STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE compositionaldependence of IMlIsi is affectedby the mentsby electronand ion microprobesare now com- complex form of the Si yield vs. composition curve. pared. Although every effort was made to atalyze Although the Fe SI data appearnear-linear with the identical areaswith the two techniques,it must be atom fraction of Fe for both olivines and low-Ca py- emphasizedthat the ion probe usesSI sputteredfrom roxenes, I../I* would be distinctly non-linear. (2) the surface whereasthe electron probe uses X-rays LIseof"relative sensitivityfactor" data for derivation excited fron a light-bulb-shapedregion with mean of atomic ratios (Meyer, 1978; Storms et al., 1977) depth about 7 to 2 Fm below the surface Tiny in- has severelimitations becauseof compositional de- clusions and exsolution lamellae, especially of Ca- pendence.For the data of Figure 3a, the sensitivity rich pyroxene, are particularly troublesomein low- factor of Mg in olivine ranges(Fig. a) from about 3 Ca pyroxene, and the pyroxene specimensrepre- for forsteriteto I I at Fo 60; for Fe, the sensitivityfac- sented in the figures are separatedinto groups for tor rangesfrom 8 to 4.Data in Meyer (1978)show at which electron probe analysesshowed variable or least 2-fold variations for Li, Be, F, Fe, Zr and Rb, near-constant analyses of Ca (Howie and Smith, dependingon the standard.Although such diagrams 1966). Data from Tables I and 2 are now shown are useful for illustration of wide sensitivity ranges, graphically and discussed. errors in deriving atomic ratios therefrom are large. A simplerelation is obtainedbetween the observed Mg*/(Mg* * Fe*) SI ratio and the Mg/(Mg + Fe) Ca atomic ratio (Fig. 5, center), but not for the corre- Figures 6a and 6b compare ion and electron mi- spondingMg/(Mg + Si) plot (Fig. 5, upper left). croprobe analysesfor low-Ca pyroxene and otvine, Smoothtrends occur for the Mgl(Mg * Fe) plot, but respectively.If the inhomogeneoussamples 4, 12 and with slightly different trends for olivine and low-Ca 16 are ignored, there is a linear correlation within pyroxene.Three referencecurves with K : 1.0, 0.6 1107oabsolute of the electron-probevalues (dashed and 0.4 are drawn for the following analytical form: lines), especially for samples with constant Ca (squaresymbols). Furthermore there is no systematic K: (t - mS*)l/IQ - mg) mg*l Img bias with Mg/Fe about the visually-drawn central where mB: atomicMg/(Mg + Fe) and mg* : SI ra- lines. The Ca level in olivine is much lower than in tio Mg*/(Mg* + Fe*). The data for low-Ca pyroxene pyroxene, and multiple analyseswere made with a (squares)lie closeto a curve with K : 0.52, and the narrow electron beam in the electron probe to test scatterof about +l atomic Voirt mg is comparableto sample homogeneity in the area analyzed with the that for routine electron microprobe analyses.The ion probe (Fig. 6b). Analyseswere also made on the data for six olivines have stayed consistentduring samespots on two separatedays (open and filled cir- severaltests over two years, and yield K: 0.42. A cles). Samples5 and 6 yielded a range in electron larger suite of olivines with specimensof inter- probe analysesconsiderably greater than the statisti- mediatezg is neededto provide a more rigoroustest. cal error ofthe electronprobe. There is considerable The advantageof using Figure 5 for mg determina- scatter about the visually-fitted linear averagesand tion is that only one standard is neededto confirm deviationstend to be the same for the two days, as the value of K, and the analytical function is easy to particularly displayed by specimens2 and 6 which handle. have the largest deviations. BecauseMg-rich speci- The shapeof the referencecurves in Figure 5 in- mens 2 and 3 lie abovethe lines and Mg-poor speci- dicates a constant distribution between Mg and Fe mens 6 and 7 (marginally) lie below the lines, while secondaryions to Mg and Fe atomsin the target, and specimens4 and 5 are near the line, it appearsthat is the sameas that used for ideal solution models of the SI yield for Ca may be higher for Mg-rich than site populations (e.9., Saxena, 1973). Perhaps this Mg-poor olivines. Such a matrix effect must be con- regularity results from the similarity of chemical firmed by measurementof more specirnensbefore it properties of Mg'z* and Fe2* ions, and can be ex- can be justified. A scatterof similar size was found ploited in the theory of sputtering. for Mg determinationsin plagioclase(Steele el a/., 1977b),but absenceof a correlation with Na/Ca ra- tio indicated that there was no matrix effect. These Minor snd trace elements two studiesindicate the needfor caution in assuming In order to evaluatethe experimentalaccuracy and a linear dependencefor ion probe analysesof minor the matrix influence,analyses of minor and trace ele- elementsin a mineral group with large compositional STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE C ono6'0@E a a o .iq o . N .d c o VP.ioC o Hl gF .3. .: .i FO o 6 p 1 | - Bi 3 o 6 q x o o o.i o .ioEiiu - j @ - o L : I t ! ! q .d a D p i o ot: . ' d .d i a > fl a !E q @ F o o o! f, o o ! oF ! ts F@ i H a a oi o ts o u N q 2! ! d x ' .i >+ q c !! Q . O ? P N dod Eo ! otr A O ots! ! > po O ! i >do 'i> A OX a@ o o qi u o tr !o q c F a o ts o oN ! ! o a 2 a o > a tr Ftr o o | .il o * C L o ol ! 3 o E o o_o o o .io o.sqotsopqiFQ ts i ! od o I ! o o oi ! ! ! 3 x ! Eo ou c oo o , s 09" ag d ci @ oo o o o Eq c E o o a la o o 'n .i! OO€ P C a [email protected].;! tr o ! .i d o > EC o ! o p c.i ci i ! o e A 5 J.i O o d i ! F E Fq ! o ! O p C o h o! oo oo a p 6 tr 6! O ! c i u o 'o . o co i ! o.i odtr @ o !6 o - u s t o trz !o,ald o d il o ll i o @c i ot G ts i :o ! ip d .d o I ov o d q.i E \o \ > 6 o .i3 .- E O Eq o c c i oi oQ @c o ! ! os tr .d o a * t3,9 '' t'! IJr , * F. '- ! Z -.a ) ! 6 ! o S bo o Qli o o .ii E >! >do q:i ! o @ ts i tsco !d o ots >: o a D.4g o o o O ts o d o E< ! + J c o !.i oo c o oo . oJ o! .aq ! ! > c. o oh o t 9f.i ;38 ?, 5 t- j > 6 b >:o o o! _ o a O L L J ! O p6 Z 4 N ] .- F ! ! i o LC,:oOO.-JF!O.!+O.f!!de a o oo! ,q o! ocd Cko €o No .iod aa o .ll i p !q ( ! ! , @ c o! ! c o i .i .i o o o o u9 0 €o ! !o E! aatr ooc aa .i.i oo! oo t @q o! .io! ! !o lod !6 q @ ! r i: 2HS 3.: *_3 5E.; 2.1t 8g R e 9.9.3 ,:3 A E3 o o E o J6 cd ^ | tO E o J EO vO J^ rl c. vO i c j 60 | N ov 6 0 .6 0 N . @ Jo6 . o 9i o .o c11 id O \+ NN Oh ?tlltl??? $ o oooaooooooo6 F 3 ooo @io-@-@ojoos l6N6 NOOO@N@OS$6O 66{l t$N@o-oi@i6NN QN {oi ro s.{oo@rf@s6o o@@ 6io@@rNirN6- rtttl rlll +t Nt iN ATtudtuAoLAAAAAtuO ooooooooooooa NN N N +o + c++r+E N @ ff-oN.iNsoi +- + + * + @ E O :E O6 O h O N oo i.io@@d.ad@@@.nd z@4oizz6@Z>1 EEa2 i@rOjgoori ss6- INNNIIO6N-ONNNNN c 6 s Ntr 9{.oooo ) o.NO;Oao ooNo dr6i@i;i d66iN o o oo 6 IIJ€€TZ< do i@N6io-r i606 diNN oos$ S?:EELE ET AL,: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE 535 o o OO aooooooooooooooooooo OO OT oijoiQjNn6rr@r@iioNO si6@r6o@66onrs@@@s@{ ON NNiNNnoNiNioNi-ii-i{ x o! o xo N ss6 '': q '':,';: d dq 6. i: T ": 1 i 1 1 I I i i a ? {N dioao6ro@N69NONOOiO6 I o! JNNN NNNNNiNOOiN{d?JIii 3 o Ho i oo H! o -E oo 'i dts @i .iii.iNdso d E< a4<@ooZZz @r @NNOO€No-06S q. NNN N NN .i N i6 N i i :I O N N N O NN O O O d@@@ d@@ o .i o .a da o .d.i d @.il @ i @.i ! .d \6 EEE@fu@O o oc o \! 6@ t E! O6 N r6 NQ ! o! !.i o! q oo o >! o F> tr @i u N oo I o oq (! 536 STEELE ET AL: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE (n a C C f l o _o c + a a C C + a) C C '"2"'"4FA \ilI IOOMgl(Mg.Fe) MgrSiOo FeSiO. IOOMql(M9-Fe) MqSiO. Fig. 3(a) Variation of secondaryion count ratesfor Si* , Mg+ and Fe+ in olivine as a function of forsteritecontent. (b) Variation of secondaryion count rates for Si+, Mg+ and Fe* in low-Ca pyroxeneas a function of enstatitecontent. Curves for Mg* and Fe* are possiblylinear while Si+ is not. Lines fitted by eye. variation of major elements.For a caunca ion mi- Ni croprobe, Shimizu et al. (1978) found that the Ca A less favorable peaklbackground ratio for the yield for ferromagnesianminerals was en-hancedin electronprobe, lower concentrationsof Ni, and posi- Fe-rich samples.Detailed measurementsof the same tive correlation of Ni and Mg in natural samples sampleswith carrlsce and AEI ion microprobesare make comparisonsof Ni determinationsmore diffi- neededto testwhether the apparentdifference in ma- cult than for Ca. Becauseion and electron probe trix effectsresults from instrumentalparameters (e.9., analysesof Ni in low-Ca pyroxene (Table 2; filled degreeof energypelection of the secondaryions) or circles,Fig. 9a) correlatedpoorly, repeat determina- from choiceof normalization (Shimizu et al. normal- tions were made with both the electronprobe and ion ized SI intensitiesto those of Si). probe (open circles) for four out of the five samples Figure 7 demoristratesthat there is a nearly linear showing highest Ni. The relative position of each correlation betweenelectron and ion probe analyses of Ca in olivines for which the range of zg is small enough (0.85-0.94)for matrix effectsto be low. All the samplesare from peridotite nodules in kimber- lites. The electron microprobe analyses were ob- + Mq-.' tlro ^-.,o--olo\ tained on thin sectionswhereas the ion probe analy- a ^).-' \ *\ /o- o seswere made on grain mounts. Much of the scatter cD -.---t---. - -o: \ -"--t:._o/ot-.--o---'---.--.. -.- lies within the 20 counting precision of the electron . \ -t- t. o \ probe analyses,and a small amount can be ascribed ,,' +._ J '\ to mineral inhomogeneity detected in the electron a re/ probe analyses.Even if all the scatterwere attributed ,c) .o to the ion probe (for which the counting precision is L trivial), most analyseswould lie within a +l0%o range. 20 40 60 80 roo Data taken in one analyical session(Fig. 8) dem- IOOMql(Mq.Fe) onstrate that the yield for low-Ca pyroxenes SI is Fig. 4. Variation of Mg and Fe secondary ion "sensitivity about one-third higher than for olivines, as also was factor" 1: Mg+,/Si+ or Fe+/Si+) for olivine as a function of found for Mn (seebelow). forsterite cont€nt. STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE the same as that at Chicago, and a diferent energy band of the SI may be used.It appearsfrom the dif- ferenceof slopesof lines in Figure lOb that an empir- ical matrix correction must be deterrrined for each set of analytical conditions. In view of the non-lin- earity of data in Figure 3a, it is possible that the trends in Figures lOa and lOb would be found to be slightly non-linear if further measurementsof higher precisionwere made. Mn This element showsa strong correlation with Fe, as shown by the progressionof sample numbers . Olivine . Low-CoPyroxene along the band of data in Figures lla and llb for low-Ca pyroxene and olivine. Becausethe data points are randomly scatteredabout a straight line, there is no evidence for a change of ionization effi- ciency with Fe/Mg ratio, as was demonstratedfor Ni. However Figure 12 showsthat the SI yield for Fig. 5. Variation of mg (: 100Mgl(Mg + Fe) atomic) in targ€t with secondaryion mg+ (: 100 Mg*/(Mg+ + Fe+)) for olivine Mn in low-Ca pyroxeneis about 25Vogreater than for and low-Ca pyroxene.Dashed lines show variation of 100 Mg/ olivine. It should not be assumedthat this difference (Mg + Si) in target as a function 100 Mg+/(Mg* + Si+) of of "relative sensitivity factor" is constant from one secondaryion spectra.See text for discussion. ion microprobeto another and from one set of oper- 31ingconditions to another. point is qualitatively the samefor the two determina- tions although the ion probe count rate is about 407o AI lower for the second determination. The scatter There is a good linear correlation (Figure 13) for about any correlation line is large even if analytical the four low-Ca pyroxenes(squares) listed as homo- errors or zoning are considered.Following the obser- geneousby Howie and Smith (1966),and all but two vation by Reed et al. (1979) that the Ni yield is a of the in-homogeneousspecimens (filled circles) lie function of the Fe contentof olivine. we testedthe ef- closeto the line. There is no evidencefor a matrix ef- fect of the FelMg ratio of low-Ca pyroxeneon the Ni fect dependenton the FelMg ratio. SI yield. Figure lOa showsx 11s41-lins4rcorrelation Becausethe Al content of olivine is rarely much betweenthe Fe content expressedas mole%oferrosi- greaterthan the detectionlevel with the electronmi- lite and the Ni yield (e): croprobe(70 ppm AlrO, at 2o), it was not possibleto make a rigorous calibration of ion probe analysis. e : (atomic FelNi) . (ion probe cts Ni/Fe) However. all olivines with Al detectedwith the elec- If the Ni SI yield were independentof the FelMg ra- tron microprobe show high count rates with the ion tio, the points on Figure lOa would lie on a horizon- probe. tal line. The positive slope indicates that with in- creasingFe (thus decreasingMg), the same atomic Cr concentrationof Ni givesgreater Ni SI yield. Becauseof the rapid decreaseof Cr with increase Similar datl arc given on Figures 9b and lOb for of Fe in low-Ca pyroxene,only a few samplescould the four olivines for which Ni is detectablewith the be usedto test ion probe analysis.For thesesamples, electronprobe. Again, an approximatelinear depen- there is a linear relation (Fig. la) between electron dence of e with Fe expressedas moleTo fayalite is and ion probe analyses, with the scatter mostly present.Comparison with the results of Reed el a/. within the 2o counting error of the electron micro- (1979) is complicated becauseReed et a/. obtained probe analyses.Only one otvine containedCr at the both the Ni and Fe analyses at high resolution detection level of the electron microprobe, and it had whereasthe present data for Fe were taken at low a substantialion probe signal. Chromium occurs at resolution.Furthermore the extractionoptical system detectablelevels in olivines from deep-seatedperido- in the ant ion microprobe used by Reed et al. is rrot tites, and there is a good correlationbetween electron 538 STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE Low-CoPyroxene v t^ I -Constont Co -/ 1--\ + o-lrregulor Co -- -.' X tl/ // ./O rc -- O- (n \ 7: va T o I Co/ 12Ox. (electron probe) Olivine ro oI X O Q) .+-a O o O ool o02 Co/12Ox (electronprobe) Fig. 6(a) Ca* secondaryion couot rate vs.Ca cations/12 oxygensin low-Ca pyroxene.Numbers adjacentto data points correspondto analysesin Table 2 and higher sample numbers repres€nt samples richer in Fe. Square symbols are used for samples known to have constant Ca and circles for those with variable Ca. Dashed lines represent l0% absolute errors from eye-estimatedlinear fit. (b) Ca+ secondaryion count rate vs. Ca cations/I2 oxyg€ili in olivine. Open and solid symbols repres€ntanalyses made on different days. Dashed lines represent tl0% absolute errors from eye-estimatedlinear fit. Sample numbers are shown. STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE 539 (t) _o electronprobe o\- counis o_ 6,OOO \-56 + ,/' ./ C) to"n'u Q) -*rdS''"= J*l*4* Q) -4 3.37 3 36 3.35 o E wovelength(A) o_ ttt' o_ a"tt-t' ,a t o2 ."- o C) 4.'/ .t4" ' 1 2o error or!gf''1 J 4a. o /a o l;OOO 2,OOO 3.OOO ' 4r\'Co- count/second Fig.1. Ca+ secondaryion count rate vs. Ca ppm lbr olivines from peridotites. At these low concentrations,mole fraction is proportional to weight fraction. Upper left inset showselectron probe wavelengthscans for 390 and 45 ppmw CaO and lower right inset showsion probe massscan at high resolutionto s€parateMgO+ molecularinterference from aoCa*.Error is basedon counting statistics. and ion probe analyses;however, the small range of the concentrationsare too low for accurateelectron FelMg in these samplesprecludes testing for a ma- probe analysis,there is a linear correlation with ion trix efect. probe analyses within the 2o counting error of the electronprobe. Becauseof the small rangeof Fe/Mg, Ti thesesamples do not allow testing of a possiblema- The ion probe signal at o'Ti has an unresolvedo'Ca trix effect. interferenceeven at high resolution, but the inter- For olivines of Table I only sample2 from a wehr- ferenceis small for low-Ca pyroxene(Fig. 15: circle, uncorrected;dot, corrected) and trivial for olivine. The correction was made by dividing the count rate *Ca a for by the acceptedisotopic ratio of 97.0:0.18 ,/, for *Ca to n'Ca.Some of the scatterabout the line on fo Figure 15 can be assignedto counting error in the I Low-Copvroxene. electron probe analyses,and some may result from \ .-. inhomogeneity caused by exsolution of Ti-bearing tr OI -'t o) oxide. There is no evidencefor a systematicbias with tt .a" FelMB ratio. For the olivine suite, Ti was detected a -'/o - I by electron probe analysis for only four out of the 4 \nr,u,n. :"o.1^) -/o v sevensamples, and the electronprobe analysescorre- O ^...-: nb.-XU, l lated linearly with the ion probe analyseswithin the "o oo2 0.o4 2o counting precision. Co/l? Ox (electronprobe) Na Fig. 8. Secondaryion yield for Ca+ in low-Ca pyroxene com- pared to olivine. Data obtainedduring sameanalysis session. Ver- Thorough analyseswere made of Na in olivines tical bar shows reproducibility of secondaryion count rate for 73- from mantle-derivedperidotites (Fig. 16). Although 165pyroxene during scssion. 5.IO STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE Low- Co Pyroxene 8 Y. Ao 6 t >< o, c) x.4 (n 4tz + ; Z 2 I -a ,^ s B! li 2o- error r5 .ll nlj l3 "oo 0002 0004 0.006 Ni/ t2 Ox (electronprobe) r5 Olivi ne fo aI r0 tr x O q) { a tr 3 05 z. 4 2s-error o6 0.01 o02 Ni /tZ Ox.(electron probe) Fig.9(a). Ni* secondaryion count rat€ vs.Ni cations/12 oxygensfor low-Ca pyroxene.Open and solid symbolsrepr€sent analyses made on different days. (b) Ni+ secondary ion count rate vs. Ni cations /12 oxygens for olivine. STEELE ET AL: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE 54I analyses of carefully selected homogeneousstan- dards; particularly difficult will be testing for matrix Low-CoPyroxene uo- lP effects.An approximatecalibration can be obtained z using the relative SI yields for other materials for which the concentrationsof two elementsare known. C l three adjacent transition elements o For example,the il Cr, V and Ti give approximatelyequal SI yields for z pure metal targets(Storms et al., 1977),and Cr in a o? :-- 0) pyroxene standard could be used as a semi-quan- LL -2/' /-/a O titative referencestandard for V in pyroxenes.This J E -'-' assumesa similar relative yield in an oxide matrix. 6 ./ -/l Relative yield factors are a function of instrumen- ll tal parametersas well as matrix. Becausewe have not determinedsensitivity factors for our instrument, we rely on data of Meyer (1978), Storms et al. (1977), Fs(mol %) and Gittins et al. (1972). The relative SI yields of trace elementsare now discussedand comparedwith .(D pyroxenes L observedtrends, especiallyfor several for Olivine which emission optical spectrographic analyses zo 4 a rReedetol (t979) (Table 5a) are available(Howie, 1955). c f Assumingthat the SI yields for Ti and V are equal o 4 --d- on an atomic basis,and not matrix dependent,1550 7A ,-t 3 ppm (Fig. 15) =v 3 cps for Ti* correspondsto 840 Ti and .c) LL hence 860 ppmv. Thus I cps should correspondto o 0.53 ppm. These estimatesfor five samples(Table E trr o -"' . 5b) are systematicallylow by -2.2. Yanadium was '-O YE02 lo 20 30 40 detectedin only two olivines in Table l, 4,ndthe 2 Fo (mol%) cps for sample2 would correspondto -l ppm if the Fig. l0(a). Secondaryion yield factor for nickel (e) vs.ferrosilite pyroxenecalibration is applicableto olivine. (Fs) contentfor Ni-rich, low-Ca pyroxenes,Symbols as in Fig. 9a. Cobalt ion yields can be expectedto be close to The cause of the anomalous position of the dot for specimen 2 is both Fe and Ni, but becausehigh-resolution Fe data unknown. (b) Secondaryion yield factor for nickel (e) vs.fayalite were not obtained,reference will be to Ni. Using the (Fa) content for Ni-rich olivines. Solid line is from Reed et a/. relative ion yields of Co to Ni as -1.75 for silicate (1979)adjusted to sameinterc€pt. glass (Meyer, 1978) and Ni data for sample 6 of Table 2 (544 cps Ni* equals710 ppm Ni), I cps cor- lite inclusion in basalt shows high (130 ppm) Na. respondsto 0.73 ppm Co. The resulting estimates Other olivines,all from rocks formed under near-sur- (Table 5b) are in the correct range but agreepoorly face conditions,have Na levelsof near I ppm with a in detail with Howie's (1955) data in Table 5a. The restrictedrange. With the exception of sample I of calculated Co values for sampleswith high mg are Table 2, electronprobe analysesof low-Ca pyroxenes low while those for sampleswith low mg Lte too have insufficient precision to test for matrix efects. hig[ differentialmatrix effectsmight be responsible. Assumingno matrix effectsthe ion probe count rates Scandium ion yields are about 1.5 times those of in Table 2 show no clear trend with zg although the Ti (Gittins et al.,1972), and I cps should correspond highestvalues occur at intermediatemg as for several with 3 ppm Sc. Calculated data in Table 5b appear other elements. twice as high as values in Table 5a, but these latter ones are all close to detection.If the pyroxene cali- Otherelements (Li, K, Sc, V, Co) bration is carried over to olivine, the count rates in Although theseelements can be measuredreadily Table I correspondto l.G-6.0 ppm. Scandiumcorre- in most olivines and low-Ca pyroxenesby ion probe lateswith Fe in low-Ca pyroxenes(Deer et al., 1978), analysis,they are too low in concentration for the and such a correlation is also implied for olivine. electronprobe. A major effort will be neededto cali- Potassiumhas a very high yield of SI, and Meyer brate the ion-probe data by comparison with bulk (1978)gives a yield about twice that of Na in feldspar 542 STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE r t6oz those of Howie (1955) but not by a constant factor. D',rnwana Becausethe Li determinationsof Howie are all near Low-Co ,uto_ ,tt l5r. his detection limit of 5 ppm, further comparison is r"',/ al/"" probably not warranted. Carrying over the calibra- /c/ v I '//'/ tion of 600 cps : I ppm to olivine yields 0.1-4 ppm a- .,' ,z o/t 9 t// X 12 / ,.t4 (Table l). ..4 /a a.. ' r/.7 O tz/ "' Discussionand prospectsfor ion microprobeanalysis a / " g 9'.a ' Whereasthe instrumentation for the ion probe is + 5 ..t' c aa/' much more complexthan for the electronprobe, and ..,17 samplepreparation and analytical techniquesare far more stringent, the ability to measuremany trace ele- points ""oo 002 004 006 ments at chosen encouragesfurther develop- Mn/ )? Ox (electronprobe) ment of the ion probe. Even now, detectionlevels are quite satisfactory for many geochemicalproblems, and theselevels can be improved at least l0-fold, and Olivrne 7 probably 100-fold,by improved design. Unlike ele- mentswith Z > l0 in the electronmicroprobe, detec- t2 tion levelswith the ion microprobevary greatly from I elementto elementdepending mainly on: (a) the SI o X yield which variesby about 105;(b) the abundanceof the measuredisotope, and (c) whetherfiigh massres- (-) a) olution is needed. The detection level can be esti- Ea mated from the count rate in col. 6 of Table 4. and -+o^ rangesfrom -0.01 ppm for alkali metals to -l ppm c "// for most transition metals. Even for Ni a detection "/z level better than l0 ppm can be obtained,in spite of a the weak yield of SI, and the needfor high-resolution nE "o ot 02 analysiswith a low-abundanceisotope. Mn/l? Ox (electronprobe) Lower count rates result at high mass resolution becauseof attenuation of the transmitted beam by Fig. ll(a). Mn* secondaryion count rate ys.Mn cations/12 ox- narrow slits. In the present AEI instrument, the in- '-Ca ygensfor lo\ pyroxenes.Most points lie within tl07o absolute tensity drops about 20-fold in adjusting from low to dashcderror limits about visual linear fit. (b) Mn+ secondaryion high resolution,mainly becauseof poor designof the count rate ys. Mn cations/12 oxygensfor olivine. matrix. For low-Ca pyroxene number l, Table 2, 0.15 vtrt.VoNarO correspondsto 3.3 x l0' cps, and I ppm K should correspondto 300 cps. The resulting values of l-2 ppm for the five low-Ca samplesin rO Table 5 are -103 tirneslower than the emissionspec- I Low-Copyr(ene trographic analyses.This very large discrepancyis x ,r " best explained by impurities in the (J V./ spectrographic a) analyses.Carrying over this caHbration a ,"/\ - to olivine u) \oliuine gives values of 30 to 100 ppb for K count rates of ,."r) ./r' Table 1. c t4 Lithium has a high SI yield about one-quarterthat a for Na (Meyer, 1978).For a plagioclasewith known oE Li our instrument gives approximately600 cps/ppm "o oo5 0ro 0r5 o20 Li which is somewhat higher than that calculated Mn/12 0x (electronprobe) from Meyer (1979).{Jsing our Li calibration, data for Fig. 12. Mn+ Secondary ion yield for low-Ca pyroxene 5 low-Ca pyroxenes(Table 5) are high relative to compared to olivine. Data obtained during sameanalysis session. STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE 543 r6 Low-CoPyroxene z/a - r ConstonlAl t"O t2 o - lrregulorAl r.c) I X va (, (n + + -f4-{ NL U o.l o2 o3 Al/ 12Ox. (electron probe) Fig. 13.Al+ secondaryion count rate vs.Al cations/12 oxygensfor low-Ca pyroxene.Linear fit includesmost sampleswithin *107o absoluteerror limits. Squaresor circles,respectively, show samplesfound to be homogeneousor inhomogeneousby electron probe analysis. 6 42 /t / ./ Low-CcPyroxene ,. ./ A rOa I - X q) E 9p O_ +\- rl Bj,! rr o|j.-o o.ol o.o2 003 Cr/ 12Ox. (electron probe) Fig. 14. Cr* secondaryion count rate ys.Cr cations/12 oxygensfor low-Ca pyroxene. 54 STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE or,, r5 Low- Co pyroxene N I I x to ro () I q) a a + O 5 + '/ lu/ tr Zf 2r ,u # aE "o 0002 0004 0006 oo08 Ti /12Ox (electronprobe) Fig. 15.Secondary ion count rate vr. cations/12 oxygensfor titanium in low-Ca pyroxenes.Filled symbol is correctedfor unresolvable aECainterference based on 4Ca count rate. SI extraction optics. After modification by a design dards. Currently there is no reliable quantitative the- which should involve only a 3-fold decrease(De- ory for secondary-ionanalysis, though a generalthe- Ereveet al., 1979),the detection level for elements ory (Andersenand Hinthorne, 1973)is valuable for measured at high resolution should improve. semi-quantitativepurposes (within a factor of 2?). Roughly speaking,the detection levels of most ele- Unfortunately it is at besttedious to assemblea set of mentswith Z > lO listed in Tables I and2 are over a reference standards for ion microprobe analysis, and 100-fold better with the ion microprobe than the at worst very difficult or even impossible for some electronprobe, especiallyfor the alkali metals. elements.Ion probe usersshould exchangematerial Now that the sensitivityproblem has been solved, to comparethe responseof the various instruments. the major problem is to obtain accuratecalibrations. In spite of the problems, ion microprobe analysis The demonstrationof a linear relation between SI of trace elementshas an important role in future re- signal and atomic concentrationfor some elements searchin geochemistry,mineralogy, and petrology. within one mineral type (e.g., for Mn in ortho- Calibration problens are least difficult for minerals pyroxene)is encouragingbecause it greatly simplifes with near-identicalor similar major-elementchemis- quantitativeprocedures. However, the dependenceof try, such as thosefrom upper-mantleperidotites, and the SI yield of Ni on the FelMg ratio of olivine and many important problems can be tackled. Minerals pyroxene provides a warning that a linear response belonging to binary series can also be handled if should not be assumedunless necessary because of standardscan be assembledto testpossible matrix ef- lack of appropriate standards.Furthermore the dif- fects, and again there are many important problems ference of catbration curves for olivine and pyrox- involving feldspars,olivines, and low-Ca pyroxenes. ene, even for an elementwhose responseappears to Minerals with complex compositional ranges, e.8., be independentof Fe/Mg ratio (e.g., Mn, Fig. l2), garnets and amphiboles, will be difficult to handle provides a secondwarning about casualuse of stan- unlessa quantitativetheory of SI emissioncan be de- STEELE ET AL.: ION MICROPROBE ANALYSES OF OLIVINE AND PYROXENE 545 repeotedonolyses on 7',, two doysoveroged to- .. ,/ somemeon ro" aI x Clivine ?rn \J lt,| tv (l) a repeotedonolyses a + () on differentspots on somedoy + o z 50 too t50 No2Oppmw (electron probe) Fig. 16. Na+ secondaryion count rate ys. Na ppm for olivines from mantle-derivedperidotites. At these low concentrations,mole fraction is proportional to weight fraction. Round symbols and crossesare from different analysis sessionsnormalized to give the same m€an coutrt rate for Na-rich olivine 73-105. veloped.Development of ion microprobe techniques lead to major new results in geochemistry,mineral- for trace elementswill be more diffi.cult than of elec- ogy and petrology. tron microprobe techniques for major and minor ele- sinultaneous useof both methodsshould ments,but Acknowledgments JVS wishesto point out that the analytical procedureswere de- veloped and the analysesobtained by the other authors. He and R. Table 5. Analysesoftrace elementsin pyroxenes N. Clayton conceivedthe project ofdeveloping an ion microprobe with high mass-resolution capability for trace elements and iso- topic ratios,in order to extendlow mass-resolutionstudies. Recog- a) Emissionspectrograph analyses (ppmw)of low-Ca with instrumental dcvelop- pyroxenes(Howie, I955). nizing that there would be difficulties ment and analytical procedures,the Earth SciencesSection ofthe 'lt Sample *.l0 t3 l4 t5 National ScienceFoundation funded the project to test whether trac€-element and isotope analyseswould be feasible at high mass Li 3 52 43 resolution. Only the AEI Company of Manchester, England' was Co B0 ]00 100 100 40 willing to attempt construction of a high-resolution ion mi- v40 then 200 50 125 l0 instrument was modified considerably' Sc <10 steady and skillful development of ion microprobe techniques Congresson X-ray Opticsand Microanalysis,p. 316-321.Pend- over a long period when publishable results were rare. ell Publishing Co., Midland, Michigan. Meyer, C. (1978) Ion microprobe analysesof aluminous lunar glasses:a test of the "rock type" hypothesis.Proceedings Lunar and PlanetaryScience Conference 9th, 155l-1570. References Recd, S. J. B., Scott,E. R. D. and Long, J. V. P. (1979)Ion micro- probe analysisofolivine in pallasitemeteorit€s for nickel. Earth Andersen,C. A. (1969)Progress in analytical merhodsfor the ion and PlanetaryScience Letters, 43, 5-12. microprobe mass analyzer. International Journal of Mass Spec- Saxena,S. K. (19?3)Thermodynamics of rock-forming crystalline trometry and lon Physics,2,6l-74. solutions.Springer-Verlag. Andersen,C. A. and Hinthorne, J. R. (1973)Thermodynamic ap- Shimizu, N. (1978)Analysis of zoned plagioclasefrom different proach to the quantitative interpretation of sputtered ion mass magmatic environments:a preliminary ion-microprobe study. spectra.Analytical Chemistry, 45, 142l-1488. Earth and PlanetaryScience Letters, 39, 398-406. Bakale, D. K., Colby, B. N. and Evans, C. A. (1975)High mass Shimizu, N., Semet,M. P. and Allegr6, C. J. (1978)Geochemical resolution ion microprobe massspectrometry to complex matri- applications of quantitative ion-microprobe analysis. Geochem- ces. Analytical Chemistry 47, tS32-1536. ica et Cosmochimica Acta, 42, 132l-1334. Banner,A. E. and Stimpson,B. P. (19?5)A combinedion probe/ Simkin T. and Smith, J. V. (1970)Minor-elemenr distribution in spark sourceanalysis system. Vacuum, 24, sllril1.. olivine. Journal of Geology, 78,3U-325. Deer, W. A., Howie, R. A. and Zussma&J. (1978)Rock Forming Smith, J. V. (1966)X-ray emissionmicroanalysis of rock-forming Minerals, Vol. 2A, Single-Chaia Silicates. Longman, London. minerals.IL Olivines, Journal of Geology,74, l-16. Degr6ve,F., Figaret,R. and Laty, P. (1979)Depth profiling by ion Steele,I. M., Hutcheon, I. D., Solberg,T. N., Smith, J. V. and microprobe with high mass resolution. International Journal of Clayton, R. N. (1977a)Effect of energyselection on quantitative Mass Spectrometryand Ion Physics,29, 351-361. analysis in secondary ion microanalysis. International Journal Evans,C. A. (1972)Secondary ion massanalysis: a tcchnique for ofMass Spectrometryand Ion Physics,23,293-305. three-dimensional characterization. Analytical Chemistry, zl4, Steele,I. M., Hutcheon, I. D., Solberg,T. N., Clayton, R. N. and 674-80A. Smith, J. V. (1977b)Ion microprobeanalysis ofplagioclase feld- Gittins, R. P., Morgaq D. V. and Dearnaley,G. (t972) The appli- spars (Ca,Nat-*All**Si3-*O8)for major and minor elements. cation of the ion microprobe analyser to the measurementof the ProceedingsEighth International Conference on X-ray Optics distribution of boron ions implanted into silicon crystals. Jour- and Microanalysis,180A-180F. nal of PhysicsD: Applied Physics,5, 165+1663. Steele,I., and Hutcheon, I. (1979) Ion probe analysisof natural Hervig, R. L. (1980) Minor and trace element composition of olivine: secondary-ionintensity variation and systematicsfor a mantle minerals: Ca-Mg exchange between olivine and ortho- simple binary silicate. In Microbeam Analysis 1979, Pto- pyroxene as a geobarometer: origin of harzburgites. ph.D. ceedingsof the l4th Annual Conferenceof the Microbeam Thesis,University of Chicago. Analysis Society,338-340. p. Herzog, R. F. K., Poschenrieder,W. and Satkiewicz,F. G. Steele,I. M., Hutcheon, I. D. and Smith, J. V. (1980)Ion micro- (1973)Observation ofclusters in a sputteringion source.Radia- probe analysis of plagioclasefeldspar (Ca1-,Na*Al2-*Si2**Os) tion Effects, 18, 199-205. for major, minor and trace elements.Eighth International Con- Howie, R. A. (1955)The geochemistryof tle charnockiteseries of gresson X-ray Optics and Microanalysis,p. 515-525.Pendell Madras, India. Royal Society of Edinburgh Transactions,62, Publishing Co., Midland, Michigan. 725-768. Storms,H. A., Brown, K. F. and St€in,J. D. (1977)Evaluation of Howie, R. A. and Smith, J. V. (1966) X-ray emission micro- a cesium positive ion source for secondary ion mass spectrome- analysis of rock-forming minerals. V. Orthopyroxenes, Journal try. Analytical Chemistry, 49, 2023-2030. of Geology, 74,4/'3462. Werner,H. W. (1975)Theoretical and experimentalaspects of sec- Liebl, H. (1975) Ion probe microanalysis.Journal of physics E: ondary ion massspectrometry. Vacuum, 24,493-5M. ScientificInstruments, 8, 797-808. Long, J. V. P., Astill, D. M., Coles, J. N., Reed, S. J. B. and Charnley,N. R. (1980)A computer-basedrecording system for Manuscript received, Augttst 5, 1980; high mass-resolution ion-probe analysis. Eighth International acceptedfor publication, January 5, 1981.