Effective L-Series Mass Absorption Coef®cients for EDSDavid G. Rickerby and Norbert WachterIHCP, European Commission Joint Research Centre, 21020 Ispra (Varese), ItalyAbstract. Energy dispersive X-ray spectroscopy hasinsuf®cient resolution to separate the individual linesof low energy L-series peaks. However, the massabsorption coef®cients for L and L radiation differsigni®cantly for elements with atomic numbersbetween 21 and 32. Effective mass absorption coef®-cients for the entire L-shell emission were determinedby measuring the variation of the X-ray intensitiesemitted from pure element standards, as a functionof the accelerating voltage, and ®tting the experimentaldata with a theoretical curve using the XMAC soft-ware. These experimentally determined mass absorp-tion coef®cients were compared with average valuescalculated on the basis of the theoretical line intensi-ties, taking into account the primary vacancy genera-tion and the radiationless Coster-Kronig transitions.Key words: Mass absorption coef®cients; energy dispersive ana-lysis; L-series.The component lines of the L-series are not completelyresolvable by energy dispersive spectrometry for X-rayenergies less than approximately 3 keV. This results inthe L and L peaks appearing as a single peak withan energy shift  10 eV with respect to the exact Lline position. Use of the undeconvoluted L peak forquantitative analysis may therefore lead to large errorsin cases where mass absorption differs signi®cantly forthe individual lines [1±3]. The elements principallyaffected are those lying between atomic numbers 21(scandium) and 32 (germanium), for which the energyof the L line is slightly greater than the ionizationenergy of the L3subshell. Due to the proximity of theabsorption edge the mass absorption coef®cients for Lradiation in these elements are up to a factor of sixhigher than for L.An additional source of error in analysis constitutesthe uncertainty in the reported values of the massabsorption coef®cients for the L lines of the elementsfrom Ti to Zn [4 ± 6]. The self absorption coef®cientfor atoms bonded with atoms of other elements canfurthermore vary noticeably from that in the pureelement because the electron transition probabilitiesand X-ray absorption properties are in¯uenced bymodi®cations in the structure of the valence bandcaused by alloying. The 3d transition elements, inwhich the valence band is incompletely ®lled, aremost strongly affected and the self absorption of NiL in alloys such as Ni-Al or Ni-Zn, for example, isconsequently weaker than in the pure metal [7, 8].Because of the uncertainties in analysis using softX-rays, a method has been developed to determineprecise values for mass absorption coef®cients bymeasuring the variation of the emitted line intensitywith accelerating voltage. The experimentally mea-sured values of the X-ray emission rate per unit beamcurrent are compared to a theoretical curve computedby the XMAC software [8, 9], which is based on theXPP model of X-ray generation. Using an iterativeprocedure to optimize the ®t a value for the massabsorption coef®cient can be obtained.Two alternative approaches may be used to performquantitative energy dispersive analysis using com-pound L-series peaks. In the ®rst, the L intensity isestimated by multiplying the total intensity of the L-shell radiation by a relative intensity factor, aL,corresponding to the ratio of the L intensity to thatemitted by the entire L-shell. In the second, the entireL-shell emission is considered [10] and the use of anMikrochim. Acta 132, 157±161 (2000) To whom correspondence should be addressedeffective L-shell mass absorption coef®cient isnecessary [2]. Values of aLhave been determinedexperimentally [11, 12] and, provided that the massabsorption coef®cients of L and L radiation aresimilar, should be relatively independent of theaccelerating voltage. Where this is not the case, aLwill depend strongly on the electron energy [3].The present work discusses the experimentaldetermination of effective mass absorption coef®cientsfor L-series radiation for the elements between atomicnumbers 22 and 33. These data will be compared withtheoretical values derived from the probabilities of X-ray generation calculated from the radiative decay ratesand effective vacancy distributions.Experimental ProcedureEnergy dispersive X-ray spectra were acquired with an OxfordInstruments Link Pentafet (Si-Li) detector, operated in thewindowless mode, using an eXL II microanalysis system. Thedetector was installed on the column of a JEOL 6400F ®eldemission scanning electron microscope. The energy resolution ofthe detector was 133 eV at Mn K and the take off angle was 35degrees. The pumping system of the microscope was equippedwith a liquid nitrogen foreline trap and an activated alumina ®lter,allowing a ®nal pressure of 2  10ÿ5Pa to be achieved in thespecimen chamber. The beam was rastered in the TV scanningmode, using a magni®cation of 100 x, in an attempt to minimizethe in¯uence of hydrocarbon contamination [13]. Conditioning ofthe detector was carried out immediately before the experiments toremove the ice layer on the surface of the crystal. Spectra wereobtained at eight different accelerating voltages between 5±30 kVfor each of the elements of interest.The standards employed were uncoated polished metal discs of> 99.9% purity mounted in a Geller UHV-compatible, typeNM537F circular holder and a polished specimen of GaAs in anASTIMEX METM25-44 metal mount. These standards were storedunder vacuum before use to reduce oxidation of their surfaces. Theelectron beam current was monitored before and after spectrumacquisition by a pneumatically operated JEOL Probe CurrentDetector (PCD) connected to a Keithley Model 485 autorangingpicoameter. Beam currents were typically between 0.2 and 1.0 nAand were found to be stable to within about  1% over a period ofseveral minutes, with the microscope operated in the constant probecurrent mode. An acquisition live time of 200 s was used, while thedead time was generally in the range 30±40%. The total net X-raycounts above the linearly interpolated background were determinedwithin an energy window defined between the channels at 1.5 timesthe FWHM on the low energy side of the Li line and 1.5 times theFWHM on the high energy side of the L line.TheoryAn effective mass absorption coef®cient for the total L-series emission