high-resolution eds in semHigh spatial resolution energydispersive X-ray spectrometry inthe SEM and the detection of lightelements including lithiumSimon Burgess, Xiaobing Li, and James HollandOxford Instruments NanoAnalysis, High Wycombe, UKIntroductionEnergy-dispersive spectrometry (EDS) is theestablished technique for the determination ofconstituent elements and elemental compositionon the microscale in the scanning electronmicroscope (SEM). The past 20 years has seensignificant advances in detector technologywhich has improved the capability for thedetection of low-energy X-rays. This hasopened up new applications in the analysis ofnanomaterials, and also materials containing thelight elements from boron to fluorine.Here we consider how progress in three areas,low-energy resolution, sensor size and X-raydetection efficiency, are together producing leapsforward in the sensitivity of EDS detectors to lowenergy X-rays, and what current developmentsmay offer in terms of nano- and light elementanalysis.Energy ResolutionDetection of light elements by EDS becamecommon in the 1990s with the introductionof vacuum-tight polymer windows. Thesewere robust to normal SEM operation, but alsorelatively transparent to low-energy X-rays,and they extended the elemental range of EDSto include the elements boron, carbon, oxygen,nitrogen and fluorine. In combination withfield-emission SEM, light-element detectors alsopromised analysis of smaller structures throughthe use of lower accelerating voltages whichreduce the volume of X-ray generation within thesample [1] (Figure 1).Resolution of EDS detectors has beenimproving since their commercialization in the1970s. Resolution is measured using a radioactive55Fe source, or Mn metal in a SEM, to produce aMnKa peak whose resolution is determined as thefull width at half the height (FWHM). Detectorresolution is of vital importance for low-energyanalysis as spacing of the X-ray lines decreaseswith decreasing energy. Mn resolution was foundto be a poor indicator of resolution performanceat low energy due to its insensitivity to detectornoise and incomplete charge collection. In2002 a new ISO specification 15632:2002 [2] waspublished, stressing the need to measure EDSresolution in the SEM using carbon X-raysto guarantee low-energy performance. Thisrequirement for low-energy X-ray performancemeasures encouraged the development ofdetectors with excellent low-energy resolutionS8that could easily separate common low-energylines such as BKa and CKa.Improved resolution was promised with thesub-130 eV headline Mn resolution promisedby the new technology of silicon drift detectors(SDD) in the mid-2000s. However, early SDDtechnology suffered from incomplete chargecollection, and was incompatible with low-energyanalysis. The charge collection properties of SDDsensors have improved to match their excellentlow noise to provide the promised improvementsin low-energy resolution in the best sensors.One way to test low-energy resolution is to lookat the separation of Til lines around 0.4 keV [3].To separate these lines requires both excellentnoise and charge collection for silicon-basedsensors. The improvement in performance can beseen in Figure 2. The best Si(Li) detectors with Mnresolution around 125 eV could almost begin toseparate these lines (green) but not if they showedany incomplete charge collection (yellow). Thebest SDD sensors with excellent charge collectionshow significant separation of these lines (red).This can aid identification of the very closelyspaced lines typical of low energies.The lightest element detectable withFigure 1X-ray generation volumes calculated using Monte Carlo simulation for pure iron at different acceleratingvoltages [1]. By reducing kV the area analysed is reduced and the spatial resolution of analysis increased.Figure 2Spectra collected from Ti metal at 5 kVusing a 10 mm2 Si(Li) detector which showsincomplete charge collection (yellow), a 30mm2 Si(Li) detector with excellent chargecollection and the latest 50 mm2 silicondrift detector (SDD). Note the carbon peakfrom the 10 mm2 Si(Li) detector is shifted tolower energy due to the incomplete chargecollection, and this loss of resolution meansno separation of TiLn/TiLa is possible. The 30mm2 has the same noise and Mn resolutionbut with much better charge collectionshows better carbon peak resolution andposition and separates OKa from TiLa better,the TiLn/TiLa shape is beginning to bevisible. The 50 mm2 SDD, has excellent noiseand charge collection. Although it is onlyabout 3 eV better at MnKa, its resolution atCKa is over 10 eV better, and the TiLn/TiLaseparation is now clear.Compositional Analysis Supplement May 2013 MicroscopyandAnalysis

high-resolution eds in semFigure 3 (left)Spectrum collected fromBe metal at 5 kV usinga X-MaxN 50 mm2 SDDdetector. The BeKa peakis completely separatedfrom the noise and has aFWHM of 35 eV.Figure 4 (right)EDS spectrumsimulations frompure iron at differentaccelerating voltages butusing the same beamcurrent and acquisitiontime. Reducingaccelerating voltagereduces line choicefor analysis to the lowenergy L line series andalso causes a dramaticreduction in the X-raycounts emitted by thesample.commercial EDS is beryllium which was firstseparated from detector noise in the 1980s [4].When first reported the resolution of the BeKapeak was over 100 eV [4, 5]. However, in the time ithas taken to improve Mn resolution by less than15 eV, the resolution of the BeKa has improvedby over 65 eV, with the best SDD detectors nowshowing peaks with resolutions around 35 eV(Figure 3).Sensor SizeLight element containing samples are oftenhighly beam sensitive, limiting the beam currentthat can be used. Reducing the acceleratingvoltage for improved spatial resolution meanslower X-ray yields even assuming beam currentcan be maintained (Figure 4). For practicalanalysis of low-energy lines therefore traditional10mm2 Si(Li) detectors were too small. Makinglarger sensors was possible, but only at the costof resolution performance, the largest Si(Li)detectors made for low-energy analysis in SEMbeing 30 mm2 [6].Developments in silicon drift sensors (SDD),have changed this in the past 5 years. Early SDDdetectors were small, with very poor low-energyresolution. However, they have the capability ofresolution performance which is independent ofsize, using designs with fixed anode dimensionswhere capacitance is fixed. This performanceis now being realized with the release of largearea SDD detectors. Currently sensors up to 150mm2 in size are now available with low-energyresolution similar or even better than thatproduced by Si(Li). In combination with parallelimprovements in FEG-SEM performance, thishas revolutionized nanoanalysis, making X-raymapping at 3 kV on structures less than 50 nmin size now practical (Figure 5). The gap betweenSEM and TEM for chemical analysis is closing.X-ray detection efficiencyImproving sensitivity for low-energy X-rays byincreasing sensor size while maintaining or evenimproving peak resolution, has revolutionisedEDS creating a true nanoanalysis capability,however, it is unlikely to be the route for futureprogress. There are two barriers to progress in thisABCDFigure 5X-ray maps for FeLa (A), NiLa (B) and CuLa (C) collected at 3 kV from a memory alloy using a X-MaxN 150mm2 SDD detector. (D) is a zoomed-in image where the three maps have been overlaid on the secondaryelectron image, showing structures as small as 21 nm are clearly x1.5x1.4ImprovementTable 1Improvements in detector efficiency for a windowless configuration compared to a polymervacuum window (Moxtek AP3.3).MicroscopyandAnalysis May 2013 Compositional Analysis SupplementS9

high-resolution eds in semAarea: firstly, ballistic deficit caused by the distancecharge carriers travel within the sensor, starts toreduce resolution performance at high count rateas sensor size increases [7]. Secondly, and moreimportantly, count rate is not only proportionalto area; it is also inversely proportional to thesquare of the distance between sample and sensor.Enlarging sensor size is only useful for large areasensors while the tube of the detector can be madesmall enough to be positioned close the sample.Lens geometry and tube diameter therefore limitmaximum sensor size.Another area to improve sensitivity is to lookagain at the polymer thin window. Removing thewindow reduces X-ray attenuation, increasingintensity for low-energy X-rays, with increasinggains at very low energy (Table 1). Windowlessdetectors are gaining acceptance on the TEMwhere vacuum environments are tightlycontrolled. These gains have caused investigationsof potential benefits for SEM also.Detection of Very Low EnergyX-rays (SiLl, AlLl and MgLl )One of the less heralded areas is the progress inthe efficiency of the detection of very low-energyX-rays. While the detection of BeKa using Si(Li)technology in SEM was first reported almost 30years ago [4], it was only a few years ago whenSDDs had progressed sufficiently to detect BeKa.However, progress has continued and we are nowable to report detection results that surpass anyprevious result with a solid-state EDS detector.Low noise sensor electronics and enhancedcharge collection efficiency have improvedsensitivity to SiLl at 92 eV in the latest SDDsensors. The improvements achieved haveresulted in clear SiLl and AlLl lines (79 eV)being commonly observed in materials such asAl-Si-Sr alloys using large-area detectors. Withthe combination of low-energy sensitivity andhigh collection efficiency, count rates sufficientfor X-ray mapping can now be achieved for SiLlusing the largest sensors. However, the poortransmission of very low-energy X-rays throughthe window remains a significant barrier, withlittle usable transmission below AlLl .Therefore, we have built a special detectorusing a windowless design to maximize lowenergy sensitivity. For this work, we selecteda large (80 mm2) detector with excellent noiseperformance (C resolution 45 eV). Results havebeen extremely promising with significantincreases in intensity for SiLl (Figure 6A) and AlLlS10BCFigure 6X-ray spectra collected from Si (A), Al (B) and Mg (C) at 5 kV in a FEG-SEM using a special version of 80mm2 X-MaxN detector with windowless configuration. Spectra show enhanced sensitivity for very lowX-ray lines SiLl and AlLl, and detection of MgLl line at 49 eV.ABCDEFFigure 7X-ray maps collected from a semiconductor structure at 3 kV in a FEG-SEM using a special version of an80 mm2 X-MaxN detector with windowless configuration.(A-D) Background and overlap corrected maps for NKa, TaMa, AlKa and SiKa, showing structures as smallas 10 nm can be identified.(E-F) Conventional X-ray maps of AlLl and SiLl .Compositional Analysis Supplement May 2013 MicroscopyandAnalysis

high-resolution eds in semlines (Figure 6B), compared to detectors fittedwith polymer windows.The excellent low-energy sensitivity suggestedthis configuration had the potential to detectand measure even lower energy X-ray lines. In afurther series of tests we have seen clear evidenceof this potential. The lowest energy X-ray that hasbeen detected with this set-up is MgLl at 49 eV.The peak is separated from noise sufficiently forelement identification (Figure 6C).X-ray Mapping of very lowenergy X-ray linesDetection of these very low-energy X-ray linesoffers the potential to use lower acceleratingvoltage, further improving spatial resolutionof analysis. In applications where Si and Al areimportant constituents, accelerating voltage canbe reduced below the 3 kV needed to excite the Klines of these elements.Figure 7 shows an example of the X-raymapping of structures in a semiconductorwith widths 10-200 nm. 10-20 nm structurescontaining nitrogen (silicon and titanium nitride-A) and tantalum (B) are clearly visible, withexcellent NKa intensity shown (Table 1). Datacollection at 3 kV allows the comparison of Kaand Ll lines for Al and Si to be made (C-F).Detection of Lithium X-raysusing EDSLithium is an increasingly important elementwith uses in growing areas such as Li-ionbatteries and Li-Al alloys used in aerospacematerials. It is the only metallic element whichcannot currently be detected by commerciallyavailable EDS detectors, although advancesin diffraction-grating spectrometry promisepotential analysis in combination with electronprobe microanalysis (EPMA) [8]. The energyof LiKa X-rays (54 eV) is higher than that ofMgLl detected using the new windowless SDDtechnology. Therefore the detection of X-raysfrom lithium (LiKa 54 eV), has been investigatedusing samples of pure lithium. Initial attemptsused a sample cut from a Li metal rod in a bagpressurised with pure N2 to minimise surfacecontamination through reaction with air, beforerapid transfer to the load lock of the SEM. Aspectrum collected from the sample is shown inFigure 8A and shows a pronounced peak at around54 eV, well above the X-ray background andclearly separated from the triggered noise peak.Other peaks, particularly OKa, but also CKa,NaKa and in some areas NKa were also detected.The OKa is always higher than the Li peak, butthis is to be expected because the efficiency of thedetector at 500 eV is many times that at 50 eV,even with this special detector configuration.To make sure that the observed Li peak wasnot some artifact of the excitation or detectionsystem, spectra were collected from a number ofother materials with no Li content using exactlythe same microscope, detector and electronicconfiguration. No spurious signal was seen in therange 40–70 eV in any spectrum other than thatfrom the Li metal, providing strong evidence thatthe peak near 54 eV detected in the spectrum fromLi metal is indeed due to the LiKa characteristicX-ray emission. Figure 8B shows the spectrumobtained from Cr metal where there is clearly nolow-energy peak at the LiK energy.ABFigure 8(A): X-ray spectrum collected from a Li metal sample in a FEG-SEM using a special version of 80 mm2X-MaxN detector with windowless configuration. The spectrum shows the presence of a peak due tocharacteristic LiKa X-rays, with the peak enlarged in the inset. (B) Spectrum collected with the samehardware under identical conditions from Cr-metal shows no peak at the LiKa energy, confirming thepeak from Li metal is genuine.MicroscopyandAnalysis May 2013 Compositional Analysis SupplementFigure 9X-ray spectrum collected froma clean lithium foil sample at2.5 kV. Despite the much loweraccelerating voltage for thisanalysis making the signals fromany surface oxidation muchgreater, the signal from oxygenhas been greatly reduced by usinga vacuum for sample preparationand handling.S11

high-resolution eds in semABCDbiographySimon Burgess obtainedhis BSc in geochemistyfrom the University ofSt. Andrews and his PhDin geology from theUniversity of Edinburgh.Simon worked inthe EPMA unit at theUniversity of Edinburghbefore moving to OxfordInstruments as an expert in WDS. For thepast 15 years Simon has worked on thedevelopment of EDS and WDS hardware,software and applications.abstractCurrent technology for energy-dispersiveX-ray spectrometry using silicon-driftdetectors in combination with thelatest field-emission scanning electronmicroscopes delivers routine nanoscalecharacterisation of structures as smallas 50 nm. This performance relies on acombination of high spectral resolution andlarge sensor area. Results from the latestdevelopments in low-noise electronics andwindowless detectors show the possibilityof analysing structures down to 10 nm andthe sensitive detection of light elements,including, for the first time, lithium.Figure 10(A) Secondary electron image collected on a lithium foil sample after deliberate exposure to air for 10minutes showing characteristic hexagonal areas of reaction. Area of X-ray map collection is shown by thered box. The X-ray maps were collected at 2.5 kV and are shown for LiKa (B), CKa (C) and OKa (D). TheX-ray maps show in the areas of reaction the formation of Li carbonates and oxides/hydroxides.lithium compounds formed onlithium metalLi metal is reactive to a number of gases thatthe sample was exposed to during preparationincluding N2 to form Li3N and atmosphericwater vapour to form LiOH [9]. Therefore furthertests have been conducted on a Li-foil samplethat was prepared and loaded through a specialmicroscope loading process under vacuum andargon. Spectra collected from this sample (Figure9 at 2.5 kV) show significantly lower levels ofOKa, allowing lower kV to be used. The benefit ofthis method was shown by exposing this sampleto air for 10 minutes. This exposure results in arapid alteration to oxygen and carbon containingspecies, forming characteristic hexagonalstructures (Figure 10A). X-ray maps collected forLiKa, CKaand OKa after exposure (Figure 10B-D), show the formation of lithium carbonatesand oxides/hydroxides in these regions duringexposure to air.ConclusionsBy combining large sensor size and excellent lowenergy resolution, modern silicon drift detectorbased energy-dispersive X-ray spectrometrysystems in combination with the latest fieldemission scanning electron microscopes provideever more detailed views of nanoscale structures.With current technologies structures in the rangeless than 50 nm can be investigated at 3 kV.To achieve even greater detector sensitivity inthe future, to further improve spatial resolution,may require the use of windowless detectors toreduce the loss of X-rays due to attenuation. Thisstudy has shown that by combining large sensorarea, low noise electronics, and windowless SDDtechnology, detection and X-ray mapping of verylow-energy X-rays is feasible. In addition X-raysfrom lithium have been detected and analysed byEDS in SEM for first time.Clearly, further work is required to determinethe applicability of this new technology to realapplication problems, particularly in the case oflithium. More information is also required tounderstand the design requirements necessary forsuccessful commercialization of this technology.References1. Barkshire, I. et al. High-spatial-resolutionlow-energy electron beam X-ray microanalysis.Mikrochim. Acta 132:113-128, 2000.2. ISO15632:2002, ISO, Switzerland, 2002.3. Statham, P. Measuring Performance of EnergyDispersive X-ray Systems. Microsc. Microanal.4:605-615, 1999.4. Statham, P. Accuracy, reproducibility and scopefor X-ray microanalysis with Si(Li) detectors.Journal de Physique 45:C2-175 - C2-180, 1984.5. Statham, P. Instrumental considerations forEDX Microanalysis at Intermediate Voltages. In:Intermediate voltage electron microscopy andits application to materials science, Ed: K Rajan,Philips Electronic Instruments Inc, New Jersey,pp 3-10, 1987.6. Burgess, S. et al. X-ray Analysis of Nanostructures Using the INCAPentaFET-x3 Si(Li)detector. Microsc. Microanal. 13(S02):1438-1439,2007.7. Bell, D. et al. Low Voltage Electron Microscopy,MicroscopyandAnalysis May 2013 Compositional Analysis SupplementacknowledgementsThe authors would like to thank Fibics Incfor supplying the semiconductor device,Hitachi High Technology Corp. for the use oftheir vacuum sample preparation and SEMloading facility, and Peter Statham of OxfordInstruments NanoAnalysis for help andadvice preparing this paper.Corresponding author detailsDr Simon Burgess,Oxford Instruments NanoAnalysis,Halifax Road,High Wycombe, Bucks, HP12 3SE, UKTel: 44 (0) 1494 442255Email: [email protected] and Analysis 27(4):S8-S13(EU), 2013 2013 John Wiley & Sons, LtdPrinciples and Applications, Wiley, pp 185-200,20138. Terauchi, M. et al. Ultrasoft-X-ray emissionspectroscopy using a newly designedwavelength-dispersive spectrometer attached toa transmission electron microscope. Journal ofElectron Microscopy 61(1):1-8, 2012.9. Wietelmann, U., Bauer, R. J. Lithium andLithium Compounds. Ullmann’s Encyclopedia ofIndustrial Chemistry, Wiley, 2000.S13

Oxford Instruments NanoAnalysis, High Wycombe, UK Introduction Energy-dispersive spectrometry (EDS) is the established technique for the determination of constituent elements and elemental composition on the microscale in the scanning electron microscope (SEM). The past 20 years has seen significant advances in detector technology