American Mineralogist, Volume 66, pages 547-552, 1981 Quantitative methodsfor electron microprobeanalysis of sodiumin natural and synthetic glasses CHaRI-Bs H. NInI-sBN'aND HAnaLDUR SIGURDSSoN GraduateSchool of OceanograPhY University of RhodeIsland Narragansett,R. I. 02882 Abstract Two methodshave been developedfor the microprobe analysisof Sodiumin natural and synthetic glassesthat exhibit time-dependentelement migration during electron bombard- ment. A cryogenicmethod is basedon the cooling of the sampleto -90oC at which temper- ature sodium diffusion rate from the excitation volume is reducedto z€ro. An empirical cor- rection method determinesthe shapeof the sodium decaycurve as a function of time during electronbombardment, which under normal operating conditions,gives initial sodium con- centration by extrapolation. Introduction et al., 1978).We describehere two analytical meth- that addressthis problem. One approsgl slimi- The precise determination of sodium concentra- ods natesthe lossof sodium and potassiumduring analy- tion in hydrous and rhyolitic synthetic and natural sis by a cryogenic lsshniqu€, while the other more glassesis one of the principal problems in quan- practical approachprovides a quantitative correction titative electronmicroprobe analysis due to the time- procedure. dependentloss in intensity of the Na K line during continuous electron bombardment.Concurrent with Previouswork the alkali loss one may observea correspondingin- creasein counts and, hence,apparent concentrations Typically 0.02 pA at l5kv representsreasonable of the non-mobile elementsin the phase.This phe- working conditions for quantitative analysisof geo- nomenon has particularly presentedproblems in ex- logic materials. For many natural and synthetic perimental petrology in the analysis of hydrous glassesconsiderable structural damage is incurred glassycharges (Helz, 1976;Kushiro, 1972).A loss of even at theselevels due to localizedheating from the over 50Voof the sodium counts and someloss in po- focusedbeam after the first few secondsof exposure. tassiumwas observed,for example,during the analy- The most notable of the beam-induced composi- sis of glassy basaltic quench products with a beam tional changesis a reduction in concentrationof mo- current of as low as 0.01p.Aat l5kV and a 5 g,mspot bile alkali metal ions. The effect is most prominent size(Mysen and Kushiro,1975). The increasedappli- for sodium and to a lesserextent for potassium,cal- cation of the electronprobe microanalyzer(nrue) to cium and barium (Borom and Hanneman,1967). Re- the study of terrestrial and deep-seatephra deposits ports of alkali losscan be found extensivelythrough- has further emphasizedthe problem of sodium loss. out the literature, for example,see Varshneya et al. For example,the sodium loss during analysisof the (1966), Vassamilletand Caldwell, (1969), Goodhew Minoan tephra from Santorini volcano was mea- and Gulley, (1974),and Watkins et al. (1978)' sured as 5OVoof the original concentrationduring 30 Depletion of alkali ions from glassduring micro- secondsanalysis time at a beam current of 0.0125pA probe analysisis most likely due to the processfirst at 15kV and a beam diameterof 5 to l0 pm (Watkins describedby Lineweaver (1962). A fraction of in- cident electrons separatesthe alkali atoms from the bridging oxygenatoms forming two ions. The oxygen as a gas I Presentaddress: JEOL (USA) INC., ll Dearborn Road, Pea- ions migrate to the surfaceand may appear body, Massachusetts01960. bubble trapped under the carbon coating.The alkali 00n3-00/'X/ 8r / 0506-0547$02.00 NIELSEN AND SIGURDSSON: METHODS FOR MICR2PR1BE ANALYSIS oF S1DIT]M ions remain in place until a particular temperatureis Methods reachedat which time they begin to ditruse through Rhyolitic glass fragments (KN-18) containing the sample.A spacecharge layer, formed severalmi- 5.687oNarO and4.39VoKrO and ranging in size from crons beneath the surface where the incident elec- 20 to 200 pm in diameter were mounted in a small trons cometo rest within the sampleacts as a sink for brass disc with epoxy. The mount was polished and thesemobile alkali ions. The mobile ions diffuse out coatedwith approximately200 angstromsof carbon. of the excitedvolume toward the spacecharge layer, The loss of sodium from the rhyolitic glass during resultingin a decreasedalkali X-ray production. The electron bombardmentwas monitored both at 25oC thermal characterof the processaccounts for the ex- under vacuum, and under cryogenic conditions. A istenceof what Vassamilletand Caldwell (1969)have cryogenic cold-finger cooled the sample through a termedthe incubation period, i.e., anearly flat region copperbraid to a final temperatureof -110'C in a in the beginning of the decay curve where little, if specimenchamber under vacuum of 5 x 10-6Torr. any, depletion takes place. The incubation period, During cooling the rate of sodium loss was mon- therefore,represents the time required for the sample itored every lOoC for a 60 second period. Counts to attain the temperaturenecessary for noticeable al- were automaticallyaccumulated in 2 kali diffusion. secondintervals from samplesexposed to a focused,0.02p.A, 15 kV Severalattempts have been made to eliminate al- electron beam. The EIMA (JEoLJxA-5OA) was stan- kali loss by altering the analytical conditions. Mov- dardized using basaltic glass VG-2 and a synthetic ing the sampleduring the analysisallows collection mineral composition of 85Vodiopside-l5Zo jadeite. of X-rays beforethe material has a chanceto heat up. The standardswere mounted in a brassdisc with the Rastering or defocusingthe beam over a large area unknowns so that standardizationcould be checked also has the efect of reducingthe heating rate. These without having to warm the sample. Quantitative methodsare effectiveonly when large (-100 pm) ho- nrue analyseswere performed on unexposedsam- mogeneoussamples are available which is generally ples of known composition at a temperature of not the casefor silicic glassytephra or experinental -l10oC, with countingtimes of 30 secondsper ele- charges.Goodhew (1975)has shown that by increas- ment. Correctionswere made on-line with a Bence- ing the acceleratingvoltage the spacecharge layer is Albee routine supplied by Krisel Control. formed at a greaterdepth in the sample.This reduces the electric field to which the alkali ions in the ex- cited volume are subjected and hence slows the decay rate. The combination of increasedaccelera- Results ting voltage, reduced beam current and energy dis- The decaycurves of Na and K X-ray intensitiesat persive spectrometersprovides reasonablequan- room temperatureare presentedin Figure l. The er- titative results(Goodhew, 1975).Unfortunately, this ror bars indicate plus or minus one standard devia- method requires acceleratingvoltages much higher tion uncertainty in the peak rate from simple count- than those compatiblewith the Bence-Albeecorrec- ing statistics.The drarratic slope of the Na curve tion schemeand relies on energy-dispersivedetectors illustrates the severity of the problem of alkali-loss which limit the accuracyof the analysis. which reducesthe initial count rate by 50Vowithin In order to utilize wavelengthdispersive spectrom- the first l0 secondsof exposureto the beam. The eters(WDS) for quantitative analysisof sodium and shapeofthe sodium curve is very nearly exponential potassiumin natural and synthetic glasses,we have and seemsto reach a steadystate condition after ap- found it necessaryto developa samplecondition that proxinately 45 seconds.Observed loss of potassium eliminatesor at least minimizes alkali mobility dur- counts is much lessand nearly linear. ing exposureto the beam. Vassamilletand Caldwell The decay rate, as measuredby the steepnessof (1969) have shown that for potassium the diffusion the curve, variesas a function ofseveral physical pa- rate increaseswith increasinginitial specimentem- rameters.For instance,the sample size is important perature.It should follow, therefore,that by reducing becausethe smallerthe samplethe fasterthe temper- the specimentemperature sufficiently the diffusion ature will rise for similar materials.Other factors in- processshould cease.If so, the initial concentration clude the embedding material, the thicknessof the can then be measuredat a temperaturewhere diffu- conductive coating, the thermal conductivity of the sion is negligible. sampleand the alkali and water contentsof the glass. NIELSEN AND SIGURDSSON:METHODS FOR MICROPROBE ANALYilS OF SODIUM -d too Y90 X o \80 o KinKN- F70 a \ z. o H60 z o E. ><I 40 o l'l -5U t\ o or No in KN-lB, O O2pA ot 15kV = oo qq-" (r .g o o:o qo z to " "o " ".*{_ . 5 rO 15 20 25 30 35 40 45 50 55 60 TIME(SECONDS) Fig. L sodium and potassium loss from natural rhyolitic glass during electron microprobe analysis under normal operating conditions. A. Empiricalcorrection method for the decay curve, plot the log of intensity versus time, and perform the statistical analysis. Sodium As noted above, the shape of the alkali metal standardsthat are stableunder the beam, such as a decay curves appear to obey some functional rela- solid solution. of 85Vodiopside-l5%o jadeite, wete tionship with time. The function which most closely usedto calibrate the system.To test the routine, sev- fits the decline of sodium intensity with time was em- materials of known composition which experi- pirically determined to be a simple exponential. eral encedsodium depletion were analyzed.Table I lists Hence, a straight-line relationship should be ob- the "true" sodium