Energy DispersiveX-ray Microanalysis Hardware

Oxford Instruments Analytical – technical briefing

Oxford Instruments Analytical – technical briefing

Introduction

Ease of use has become a major focus in the selection of EDS analyzers. The hardware

that acquires the data is often taken for granted.

The detection and measurement of X-rays in an electron microscope requires a complexmeasurement chain, which, if functioning correctly, can provide the accurate and stable datarequired to complement the software and allow reliable automatic peak identification andstandardless quantitative analysis. The test of a good design is whether results on a given sampleremain accurate irrespective of changing count rate or environment.

This guide introduces the operation of the EDS hardware, and highlights the important aspectsthat make a difference in accurate and efficient X-ray analysis.

X-ray Detector Pulse Processor2

Analyzer

Principal System ComponentsAn EDS system is comprised of three basic components that must be designed to work together toachieve optimum results (Fig. 1).

X-ray DetectorDetects and convertsX-rays into electronic signals

Pulse ProcessorMeasures the electronic signalsto determine the energy ofeach X-ray detected

AnalyzerDisplays and interpretsthe X-ray data

Fig. 1.

Al Kα

3

Oxford Instruments Analytical – technical briefing

Components of an EDSDetector

1. Collimator assemblyThe collimator provides a limiting aperturethrough which X-rays must pass to reach thedetector. This ensures that only X-rays fromthe area being excited by the electron beamare detected, and stray X-rays from otherparts of the microscope chamber are notincluded in the analysis.

2. Electron trapElectrons that penetrate the detector causebackground artefacts and also overload themeasurement chain. The electron trap is apair of permanent magnets that stronglydeflect any passing electrons. This assemblyis only required on detectors with thinpolymer windows, as thicker berylliumwindows efficiently absorb electronsbelow 20keV in energy.

3. WindowThe window provides a barrier to maintainvacuum within the detector whilst being astransparent as possible to low energy X-rays.There are two main types of windowmaterials. Beryllium (Be) is highly robust, butstrongly absorbs low energy X-rays meaningthat only elements from sodium (Na) can bedetected. Polymer-based thin windows canbe made much thinner than Be windowsand therefore are transparent to much lowerenergy X-rays, many allowing detection ofX-rays down to 100eV. Although thesewindow materials are far less robust, byplacing them on a supporting grid they canwithstand the pressure difference betweenthe detector vacuum and a ventedmicroscope chamber at atmospheric pressure.

The greater transmission of the polymer-based windows means that they have largelyreplaced Be as the material used for detectorwindows.

4. CrystalThe crystal is a semiconductor device thatthrough the process of ionization convertsan X-ray of particular energy into electriccharge of proportional size. To achieve thisa charge-free region within the device iscreated. Two main materials are used forthe detecting crystal. The most common issilicon (Si), into which is drifted lithium (Li)to compensate for small levels of impurity.High purity germanium crystals (HpGe) arealso used. Si(Li) was the first material usedin EDS detectors and remains the mostcommon choice today. HpGe offersperformance advantages when measuringhigher energy X-rays.

5. FET The field effect transistor, normally referredto as the FET, is positioned just behind thedetecting crystal. It is the first stage of theamplification process that measures thecharge liberated in the crystal by an incidentX-ray and converts it to a voltage output.

Section 1:

4

6. CryostatThe charge signals generated by the detectorare small and can only be separated from theelectronic noise of the detector if the noiseis reduced by cooling the crystal and FET.Most EDS detectors work at close to liquidnitrogen temperatures (90K), and are cooledusing a reservoir of liquid nitrogen held ina dewar. The low temperatures required canalso be achieved using mechanical cooling

devices. However, these are more expensiveto build and maintain, particularly if lowvibration is essential, and so are normallyused only where liquid nitrogen is notavailable. The ‘cold finger’ that cools thecrystal is insulated from the wall of thedetector snout by a vacuum. The vacuum ismaintained at a low enough level to preventthe condensation of molecules on the crystal.

Fig. 2. Cut-away diagram showingthe construction of a typical EDS detector.

The EDS Detector

6. Cryostat

5. FET 4. Crystal

3. Window

2. Electron trap

1. Collimatorassembly

5

Oxford Instruments Analytical – technical briefing

How the EDS DetectorWorks

The EDS detector converts the energy ofeach individual X-ray into a voltage signalof proportional size. This is achieved througha three stage process. Firstly the X-ray isconverted into a charge by the ionizationof atoms in the semiconductor crystal.Secondly this charge is converted into thevoltage signal by the FET preamplifier.Finally the voltage signal is input into thepulse processor for measurement. Theoutput from the preamplifier is a voltage‘ramp’ where each X-ray appears as avoltage step on the ramp.

EDS detectors are designed to convert theX-ray energy into the voltage signal asaccurately as possible. At the same timeelectronic noise must be minimized to allowdetection of the lowest X-ray energies.

Electrons

Holes

X-ray FET contact

3b

K

L

N

M

Electron beam

Secondary electron

X-ray

Time

Vo

ltag

e

Charge restore

X-ray induced voltage step

Charge signal C Voltage ramp

Pre-amplifierFETCrystal

h

e

X-ray

HV

3c

3d

Fig. 3. Conversion of X-ray signals into a voltage ‘ramp’ by the EDSdetector. (a) generation of a characteristic X-ray in a sample by electronbombardment. (b) generation and measurement of electron-hole pairs

in the crystal. (c ) circuit diagram of the EDS detector. (d) typical outputvoltage ‘ramp’ showing events induced by MnKα X-rays.

3a

6

How the crystal converts X-rayenergy into charge

When an incident X-ray strikes the detectorcrystal its energy is absorbed by a series ofionizations within the semiconductor tocreate a number of electron-hole pairs(Fig. 3b). The electrons are raised into theconduction band of the semiconductor andare free to move within the crystal lattice.When an electron is raised into theconduction band it leaves behind a ‘hole’,which behaves like a free positive chargewithin the crystal. A high bias voltage,applied between electrical contacts onthe front face and back of the crystal,then sweeps the electrons and holesto these opposite electrodes, producinga charge signal, the size of which isdirectly proportional to the energyof the incident X-ray.

The role of the FET

The charge is converted to a voltage signalby the FET preamplifier (Fig. 3c). Duringoperation, charge is built up on the feedbackcapacitor. There are two sources of thischarge, current leakage from the crystalcaused by the bias voltage applied betweenits faces, and the X-ray induced charge thatis to be measured. The output from the FETcaused by this charge build-up is a steadilyincreasing voltage ‘ramp’ due to leakagecurrent, onto which is superimposed sharpsteps due to the charge created by eachX-ray event (Figs. 3d, 4). This accumulatingcharge has to be periodically restored toprevent saturation of the preamplifier.Therefore at a pre-determined charge levelthe capacitor is discharged, a process calledrestoration. Restoration can be achievedeither by pulsed optical restore where lightfrom an LED is shone onto the FET, or byusing direct injection of charge into aspecially designed FET.

Time

b

a

Vo

ltag

eV

olt

age

Fig. 4. Output voltage ramp (a) when no X-rays are detected, the ramp results from leakage current from the detector.(b) when X-rays are detected, steps representing each X-ray are superimposed onto the ramp. At high count ratesexcessive rate of restores can affect the ability of the detector to measure X-rays accurately.

4a

4b

7

Oxford Instruments Analytical – technical briefing

The noise is strongly influenced by the FET,and noise determines the resolution of adetector particularly at low energies. Lownoise is also required to distinguish lowenergy X-rays such as beryllium fromnoise fluctuations (Fig. 5) . Direct chargerestoration via the FET introduces less noisethan optical restore. At high count rates,the restoration periods limit the maximumoutput rate and any after-effects of therestoration (Fig. 4) will affect pulsemeasurement. Direct charge restoration viathe FET is considerably faster and avoids theafter-effects associated with optical restoreso that noise and resolution are less likely todegrade with increasing count rate.

What Makes a GoodX-ray Detector

The following section introduces some of themost important issues that can be used toevaluate the ability of an X-ray detector toaccurately and efficiently detect X-rays.

Manganese resolution, MnKαFWHM

Resolution is quoted as the width of thepeak at half its maximum height (FWHM).The lower the number the better theresolution a detector has and the better it will be at resolving peaks due toclosely spaced X-ray lines.

Energy (keV)

0

500

1000

1500

2000

2500

3000

3500

4000

4500

Co

un

ts

-0.1

FeS2

Zr

Si

BeNoise

Be Kα

Si LI

S LI

Zr Mz

0 0.1 0.2

Fig. 5. Spectra collected from Be, Si, FeS2, and Zr, showing the detection and separation from noise of low energy K, L, and M lines.

8

Manganese resolution is measured byplacing a piece of pure manganese underthe electron beam. On a microscope theresolution is often quoted at 1000cps.This count rate is much lower than is used inpractice for most microanalysis experiments.It is important to determine whether thisresolution is maintained at more realisticand productive count rates.

The importance of low energyresolution

The identification and quantification ofclosely spaced X-ray peaks becomes easierand more accurate as the separationbetween them increases. X-ray lines getcloser together at low energy and this isapparent in Fig. 6 where the energyof the main alpha line for K, L and M series is plotted for all elements. At the MnKαenergy most commonly used to specifyEDS resolution, peaks are well separated.However, it is clear that much more seriousoverlaps occur below 3keV and theresolution performance at low energies iscritical to good performance for all elements.

0 1 2 3 4 5 6 7 8 9 10

Kα linesMα lines

Lα lines Detector resolutionLa U

Be Ge

F 65eV FWHM Mn 129eV FWHM

Ca Hg

Energy (keV)

Fig. 6. Energy of Kα, Lα and Mα lines. Note that the separation between lines decreases and density of lines increases at low energy.

9

Fluorine resolution FKα FWHM

The width of the peaks in a spectrum willvary depending on the energy of the X-rayline. In Fig. 8 this variation with energy hasbeen calculated to show how the resolutionshould change with X-ray energy for Si(Li)and HpGe detectors with different Mnresolution specifications. The curves arecalculated from the equation

FWHM2 = k . E + FWHMnoise2

where k is a constant for the detectormaterial, and E is the energy.

These curves demonstrate that as the energydecreases, the resolution of the X-ray peaks

improves. The variation is alsoclearly different for Si(Li) andHpGe crystals. At low energythe electronic noisecontribution (FWHMnoise) has agreater effect on resolution(Fig. 9). Mn FWHM is a veryinsensitive measure tocharacterize the noise of adetector and predict theresolution at low energy.Therefore to characterize theresolution of a detector at lowenergy, the resolution is alsoquoted for another line, typically fluorine Kα.

Oxford Instruments Analytical – technical briefing

7a

7b

Fig. 7. Spectra collected from a nickelalloy at different accelerating voltages.(a) 20kV.(b) 5kV.

When small features <1µm in size are beinganalyzed, the beam voltage needs to bereduced to avoid electron scattering outsidethe feature. However, at low kV only lowenergy lines are available for analysis.Spectra collected from a nickel alloy at 20kVand 5kV (Fig. 7) illustrate the importance ofresolution at low energy. When working at20kV, the separation of widely spaced K linesof Cr, Fe and Ni, will not be affected much bya few eV variation in resolution. Whenworking at 5kV however, whereidentification relies on L lines, a detectorwith a few eV better resolution will allowthe L lines of Cr, Fe and Ni to be separatedenough for confident identification.

10

results in counts appearing in the spectrum atlower energies than the energy of theX-ray which they represent, typically as a tailon the low energy side of the peak. Alldetectors suffer from incomplete chargecollection to some extent. Low energy X-rayshave a very shallow depth of penetration andICC is usually poor near the front contact. Thepeak for a low energy X-ray will therefore bebroader and have a mean energy lower thanexpected, due to varying levels of chargecollection as each X-ray is measured (Fig. 9).

Thus, incomplete charge collection results indetectors with resolutions measured at lowenergy that are worse than those predicted by

theory. In extreme cases it can be the dominantfactor controlling resolution at these very lowenergies (Fig. 9).

Carbon resolution CKα FWHM

It is more important to know what the resolutionspecification of a detector is at low energy than it isat MnKα. A detector with good energy resolution atlow energy will have good resolution throughoutthe X-ray spectrum. The same is not true in reverse.A good manganese specification guarantees nothing

about performance at low energy. This isbecause a measurement of resolution atlow energy is affected by both noise andincomplete charge collection.

The resolution of the CKα peak measuredusing pure carbon is very useful due to itssensitivity to incomplete chargecollection. A low value guaranteed herereally does mean excellent detectorresolution for all energies.

Energy (keV)0

0

20

40

60

80

100

120

140

160

180

200

1 2 3 4 5 6 7 8 9 10

Res

olu

tio

n F

WH

M (

eV)

Resolution 129eV Si(Li) FWHM at Mn Kα

Resolution 133eV Si(Li) FWHM at Mn Kα

Resolution 115eV HpGe FWHM at Mn Kα

Resolution 129eV HpGe FWHM at Mn Kα

Energy (keV)0

01 2 3 4 5 6

Res

olu

tio

n F

WH

M (

eV)

DetectorICC Noise

20

40

60

80

100

120

140

160

Fig. 9. The variation in detector resolution with energy iscontrolled by the contribution of three factors: a constantdispersion based on the crystal material used in a detector,the level of electronic noise, and low energy peakbroadening due to incomplete charge collection. The curveshere are based on a detector with a resolution of 140eV atmanganese that has severe incomplete charge collection.

Fig. 8. Curves showing how resolution changes with energy fordifferent detectors. The calculation ignores the effects of incompletecharge collection that can make some detectors give worseperformance than these curves predict, particularly at low energies.

Incomplete charge collection

If all the electron-hole pairs generated by anX-ray are not swept to the electrical contacts,the charge signal measured by the FET willbe lower than expected, and the energymeasured lower than the energy of theincident X-ray. This phenomenon is known asincomplete charge collection (ICC), and

11

Oxford Instruments Analytical – technical briefing

Detector specifications based ontests using an Fe55 radioactivesource

Testing the performance of an EDS detectorusing a radioactive source is convenient for adetector manufacturer because it can bedone without having to mount a detector onan electron microscope column. However,Fe55 source specifications don’t necessarilyguarantee the performance when mountedon a column and collecting X-rays emittedduring electron bombardment. They are alsolimited in scope and do not reveal someimportant aspects of detector performance,in the real situation of electron microscope-based X-ray microanalysis.

The reason that the resolution of a detectoris traditionally specified for Manganese KαX-rays at 5.895 keV, is because this is theenergy of the most intense X-ray line emittedby the Fe55 source (Fig. 10).

One useful feature of the Fe55 source is theabsence of the continuum or bremsstrahlungX-rays that would be generated in anelectron microscope. Therefore, in an Fe55

spectrum the very low intensity backgroundat lower energies than the characteristic MnX-ray peaks is due to incomplete chargecollection. The most common measurementof ICC is the peak: background ratiocomparing the height of the MnKα peak tothe average background between 0.9-1.1keV(Fig. 10). Events appearing at 1keVcorrespond to MnKα photons where 83% ofthe charge has not been collected, whereasevents appearing at say 4keV have lost 32%of the charge. Although the original IEEEstandard suggested measurement at severalenergies, manufacturers have mostcommonly used 1keV. As shown in Fig. 10,1keV is typically the lowest part of the ICCbackground and a low value at this energy(high peak:background ratio) does not

Fig 10. A schematic of a spectrum collected from an Fe55 radioactive source showing how peak:background ratio is measured on the bench.

12

Energy (keV)

Co

un

ts

Mn Kα

Mn Kα esc

Fe55 P/B backgroundmeasured here

LinearLinear x100

0 1 2 3 4 5 6 7

guarantee that charge collection will begood for the smaller charge losses thatcontribute to the tail or plateau on the lowenergy side of a peak. Fe55 peak:backgroundonly gives a general indication of chargecollection efficiency, but does not ensuregood peak shapes or give any indication ofICC for low energy X-rays. Incomplete chargecollection is most important where X-rays arestrongly absorbed and only penetrate a shortdistance into the crystal, for example justabove the absorption edge of the elementsmaking up the crystal (1.84keV for Si(Li)detectors), and for very low energy X-rays(less than 0.5keV). The CKα resolutionmeasurement on the microscope representsa more useful and sensitive measurementof incomplete charge collection thanpeak:background ratios measured with anFe55 source. Low values for CKα FWHMguarantee excellent charge collection, andgood peak shapes at all energies.

Detector performance changeswith time

A major cause of detector degradation overtime is the build up of contaminants thatabsorb X-rays before they can be detected bythe crystal. Common examples include thecondensation of oil on the collimator ordetector window and ice forming on the faceof the crystal. These contaminants will causepreferential absorption and a drop-off insensitivity for low energy X-rays.

Ice forms on the cold crystal due to themigration and condensation of any watermolecules in the detector vacuum.

Two sources of water vapor exist: impuritiespresent during manufacture, and moleculesthat migrate through the window when thedetector is exposed to high vapor pressures.Modern manufacturing techniques meanthat when installed, a detector vacuumshould be free of water molecules. In SEMswhere variable vacuum or environmentalmodes are used, the detector spends time inconditions where water is present in themicroscope chamber. Some types of polymerthin window, which are predominant inmodern EDS detectors, have been shownto degrade and become porous underconditions where water molecules arepresent.

A gradual decrease in low energy sensitivityover time will result in a decrease in theheight of peaks at low energy. This can bechecked by monitoring the relative height ofK and L lines from a transition metalelement. The ratio of L to K line heights frompure nickel measured at 20kV is a commontest used. A more sensitive test for thepresence of ice on a crystal is to look at theL spectrum from pure Cr. The L line spectrumconsists of the Ll line at 0.5keV, and the Lαline at 0.571keV. The Lα line is on the highenergy side of the oxygen absorption edge(energy 0.531keV) whilst the Ll line is on thelow energy side. Therefore CrLα X-rays aremuch more efficiently absorbed by ice thanCrLl X-rays. On a detector with little or no iceon the crystal face the Lα line should behigher than the Ll line (Fig. 11b). On adetector which has ice built up on the crystalthe Ll line will be higher (Fig. 11a).

13

Oxford Instruments Analytical – technical briefing

The ice must be removed to regain the lightelement sensitivity of a detector. This can bedone in two ways. If the detector is thermallycyclable it can be allowed to warm up. Whenthe crystal starts to warm the ice will sublimeand the vapor will disperse into the vacuum.When the detector is cooled down again thewater vapor will normally condense withinthe dewar because this area cools down first.

This technique is time-consuming, requiringthe liquid nitrogen reservoir in the dewar to

be exhausted which can take a number ofdays, or to speed up the process, the detectorcan be removed from the column and thenitrogen poured out. Some detectors have abuilt-in heating circuit called a conditioner.This circuit warms up the crystal enough tosublime off any ice. Conditioning can be donewhilst the detector is cooled down on themicroscope column, and can remove any icein as little as 2 hours.

Geometry

To a first approximation, X-rays areemitted equally in all directions,therefore the collection efficiency isgoverned by the proportion ofspace intercepted by the detectoractive area (A). This in turn isproportional to the ‘solid angle’which would be 2π steradians if allthe X-rays in a hemisphere abovethe specimen were collected. If dis the sample-crystal distance, thensolid angle = A/d2 is a usefulapproximation for solid angles lessthan 0.2 steradians.

EDS detectors are available withdifferent sizes of crystals. Thecrystal size is often measured inarea, 5mm2, 10mm2, 30mm2, 50mm2

etc. There is a trade off inperformance; normally the largerthe crystal, the worse will be itsresolution, particularlyat low energy.

11a

11b

Fig. 11. CrL spectra collected from pure Cr at 5kV. (a) A spectrum collected from a detectorused on a variable vacuum microscope showing CrLα absorption equivalent to 100nm of ice.(b) A spectrum collected from the same detector after the ice layer has been sublimed awayby a two hour conditioning cycle.Note partially resolved CrLl and Lα lines and their relative intensities.

14

The collection efficiency is more sensitive tothe distance between crystal and the sample,and for maximum efficiency the detectormust be positioned as close to the sample aspossible. A 10mm2 crystal at a distance of5cm will have the same solid angle as a30mm2 crystal at 8.7cm. The distance thecrystal will be from the source of X-rays onthe sample will depend on the dimensionsand design of the detector and also thegeometry of the microscope (Fig. 12a). Thetype of collimator and window assembliesused also affect the solid angle.

The requirement for an electron trap placesthe crystal further from the sample (Fig. 12a),and the need for a support grid for thewindow reduces the active area A bytypically 20% compared to self-supportingthin windows (Fig. 12b).

Therefore if count rate is an issue, whenconsidering crystal size it is much more usefulto know what the maximum solid angle willbe on the microscope to be used, rather thanjust the crystal area. For most microscopessufficient solid angle can be achieved with10mm2. In some situations, for example whenusing a transmission electron microscope, theX-ray signal is very low, and a bigger crystalis used to improve signal collection.

Fig. 12. Calculation of the solid angle of an EDS detector. (a) The detector size and the geometry of the microscope control the distancebetween X-ray source and crystal. (b) The solid angle is proportional to the active area of the crystal, rather than its actual area, and inverselyproportional to the square of the distance.

12a 12b

15

Active areaPole piece

d

d

w

w(d) Crystal to X-ray source distance

(w) Solid angleElectron trap

Crystal

Grid

Oxford Instruments Analytical – technical briefing

Summary• The EDS detector converts the energy

of a single X-ray photon into a step ofproportional size on a voltage rampusing a semiconductor crystal andFET-preamplifier

• Good analytical performance requiresgood resolution at low energies

• Mn resolution (5.9keV) does not give areliable indication of resolution at lowenergies because it is insensitive to ICC and does not separate noise and detectormaterial components

• F resolution (0.7keV) is more sensitive tolow energy noise and is a better guide tolow energy performance

• C resolution accounts for noise and is verysensitive to incomplete charge collection.An excellent resolution at carbon impliesgood resolution at all energies

• The collection efficiency of X-rays isdetermined by the collection solid angle,not the area of the crystal

16

Section 2: The Pulse ProcessorThe Role of the PulseProcessorThe charge liberated by an individualX-ray photon appears at the output ofthe preamplifier as a voltage step on alinearly increasing voltage ramp (Fig. 13a).The fundamental job of the pulse processoris to accurately measure the energy of theincoming X-ray, and give it a digitalnumber that is used to add a count to thecorresponding channel in the computer(Fig. 13c). It must also optimize the removalof noise present on the original X-ray signal.It needs to recognize quickly and accuratelya wide range of energies of X-ray eventsfrom 110eV up to 80keV. It also needs todifferentiate between events arriving inthe detector very close together in time,otherwise the combination produces thespectrum artefact called pulse pile-up.

Signal measurement

There are a number of ways of measuringthe size of the steps on the voltage ramp,which depend on the type of pulse shapingbeing used: digital or analog.

Analog pulse shaping

In analog shaping the signal from thepreamplifier is converted to a pulse by wayof analog shaping electronics (Fig. 14). Theheight of the pulse is then measured andconverted into a digital signal by an analogto digital converter (ADC). A longer peakingtime (TP) is used to reduce noise and improveresolution.

Fig. 13. Accurate measurement of the voltage ramp is the role of the pulseprocessor.(a) typical output voltage ramp showing events induced by MnKα X-rays.(b) layout of a typical pulse processor showing the measurement and pulsepile-up inspector channels.(c) Output of the measured steps gives an X-ray spectrum showing numberof counts vs. energy.

Time

Vo

ltag

e

Charge restore

X-ray induced voltage step

Output

Pile-up inspection channels

Pulse measurement channel

Fast/high energy events

Detector/Preamp

Acceptor reject

Slow/low energy events

Medium/medium energy events

13b

13c

13a

17

Oxford Instruments Analytical – technical briefing

Accurate height measurement of each pulserequires the shaping circuitry to be reset tothe baseline level before another event canbe accepted for measurement. The decaytime (TD) gives the effective minimum timebefore another step can be measuredaccurately (Fig. 14).

The challenge with analog shaping is tocompensate for changes in the baseline. Tomeasure a 10keV photon to within 1eVrequires the baseline to be known within0.01% of the pulse height. The time requiredfor the pulse to decay to below 0.01% of thepulse height will be many times the peakingtime. As the count rate increases the residualson the baseline left over from each shapedpulse begin to build up, resulting in ameasurable shift in the apparent baseline. Inaddition, the baseline will move at high inputrates due to the cumulative after-effects of allthe large restore signals. Thus, as the inputcount rate increases, peaks will shift inposition.

Even at low count rates, the slope on thevoltage ramp will affect the position of thebaseline. This slope, caused by leakage

current from the detector (see Fig. 4), willmean that shaped pulses will not returnexactly to zero, and therefore the baselinealways has some offset.

When calibrating systems with analogshaping two key adjustments are required.One to compensate for detector leakagecurrent, another to compensate for countrate-induced peak shift. Since the degree ofshift depends on the shaping time, theseadjustments have to be performed for everyshaping time constant and gain setting.Count rate effects may also vary withspectrum content and even if exactcompensation is achieved for an ‘average’sample spectrum, peak shifts of a few eV maystill be expected when different samples areanalyzed. Furthermore, if detector leakagechanges, the calibration process for baselineoffset has to be repeated.

Some pulse processors are described asdigital, because they use digital circuits andsoftware instead of control knobs to controlthe circuitry. These processors will have thesame problems with baseline measurementand calibration shown by any pulse processorthat uses analog shaping.

Time

Vo

ltag

e

TP TD

Shaping electronics

Pulse heightanalysis

Time

Vo

ltag

e

Fig. 14. Measurement of a step on the voltage ramp by analog shaping electronics. Each pulse is shaped for a peaking time TP to remove noise, and toform a pulse that can be measured by a pulse height analyzer. TD is the time taken for the pulse to decay sufficiently close to zero to allow anotherpulse to be measured accurately.

18

Time variant shaping

Time variant shaping can be used toovercome the calibration problems and ratesensitivity of analog shaping (Kandiah et al.1975). By switching to a very short timeconstant right after the measurement iscomplete, the pulse returns very rapidly tothe baseline. The shaping circuit can then beswitched back to long time constants readyto measure the next voltage step. Thismethod removes any count rate-induced shiftbut requires complex circuitry because eachevent has to be recognized as being presentbefore the sequence of shaping can beinitiated. In the absence of any real events,the electronics can also measure the voltageramp, and therefore monitor the baselineautomatically, so that changes in the slope ofthe ramp caused by variable detector leakageare compensated and peak shift is negligible.

Digital pulse shaping

In a processor using digital shaping, thesignal from the preamplifier is digitized atthe input of the pulse processor, and shapingand noise reduction are achieved by digitalcomputation.

The preamplifier output is sampledcontinuously by an analog to digitalconverter (ADC) and X-ray pulse heights aremeasured by subtracting the average of oneset of values, measured before an X-rayevent, from that for another set, measuredafter the event. The resultant value of thestep measurement is then sent directly to thecomputer multichannel analyzer.

The noise on the voltage ramp from thedetector is effectively filtered out byaveraging the signal (Fig. 15). The time overwhich the waveform is averaged (oftencalled the process time) is equivalent to thepeaking time for an analog shaper (TP).

Energy (keV)

Co

un

ts

0.5 0.6 0.7 0.8 0.9

TP

Time

Vo

ltag

e

Time

Vo

ltag

e

Energy (keV)

Co

un

ts

0.5 0.6 0.7 0.8 0.9

TP

Fig. 15. Measurement of steps on a voltage ramp by averaging differing numbers of measurements of the signal. (a) Short TP permits all steps to bemeasured, but the variation of each measured step is large, so the X-ray energy is not measured accurately and peaks show poor resolution. (b) Long TPmeans that some steps arrive too close together to be measured. However, noise averaging is better and therefore peaks show better resolution.

15a 15b

19

Oxford Instruments Analytical – technical briefing

By measuring the voltage ramp in this waythe benefits of perfect baseline recovery areachieved without the complexities requiredfor time variant shaping circuits. Thereforepeaks should not shift with count rate. Thereis also the potential for measuring the zerobaseline (‘strobing’) by averaging the digitaloutput of the ramp when no events arepresent (Fig. 18a). This will provide not onlyautomatic correction for changes in the slopeof the voltage ramp caused by changes indetector leakage current, but will alsomeasure the effective noise resolution.Calibration procedures for processors usingdigital pulse shaping should bestraightforward and reliable. One energycalibration should be sufficient to guaranteeaccurate energy calibration at all count rates.

Fig. 16. Spectra collected atshort (red) and long (blue)process times.

Processors with fixed process time

The longer the process time (TP), the lowerthe noise. If noise is minimized, theresolution of the peak displayed in thespectrum is improved (see Fig. 15), and itbecomes easier to separate or resolve, fromanother peak that is close in energy (Fig. 16). However, there is a trade-off between theprocess time that is used, and the speedat which data can be measured. The longerthe process time, the more time is spentmeasuring each X-ray, and the fewer eventsthat can be measured. The longest processtime used by a processor gives the bestresolution possible while the shortestprocess time gives the maximum throughputinto the spectrum, but with the worstresolution (Fig. 16).

Energy (keV)

0

10000

20000

30000

40000

50000

60000

70000

80000

90000

100000

Co

un

ts

1 1.2 1.4 1.6 1.8 2

Mg Kα Al Kα Si Kα

20

Productivity depends on the rate of countsmeasured, called the acquisition rate, ratherthan the input rate (into the detector).As the input rate increases so will theacquisition rate, but an increasing number ofevents are rejected because they arrive in ashorter time period than TP (Fig. 15). If inputrates increase sufficiently, the proportionrejected will exceed the increase in measuredevents and the acquisition rate will start todecrease with further increases in input rate(Fig. 17).

Input count rate into the detector (counts per second)

20,000 40,000 60,000 80,000 100,000 120,000

Acq

uis

itio

n r

ate

into

th

e sp

ectr

um

(co

un

ts p

er s

eco

nd

)

Short process time high acquisition rate

Long process time best resolution

maximum acquisition

rate

maximum acquisition

rate

Fig. 17. Curves showing the variation of acquisition rate with input rate for two fixed process times.The longer process time gives good resolution but limited maximum acquisition rate, the shortprocess time allows much higher acquisition rates but resolution is worse.

Therefore for each process time there is amaximum acquisition rate (Fig. 17) whichcorresponds to the maximum speed possiblefor a chosen resolution. The maximumacquisition rate for each process time ischaracteristic of the pulse processor used.By determining, for each processor setting,the maximum acquisition rate and theresolution at this rate, the productivity andperformance of a processor can be evaluated.

In a processor where the process time isfixed, the trade off between resolution andacquisition rate can be controlled and theresolution accurately defined.

21

Oxford Instruments Analytical – technical briefing

Processors which useadaptive pulse shaping

It is possible to use variableaveraging times to measurethe size of each step. Eventsthat arrive far apart can bemeasured with more noiseaveraging than events arrivingclose together (Fig. 18b). Thisproduces a spectrum whereeach peak is built up of manyGaussian shapes with adistribution of resolutionsdetermined by the distributionof the timing of the arrival ofevents on the voltage ramp.

At low input rates this type ofprocessor will deliver similarresolution to a fixed processorusing a long process time.However, as the input rate isincreased the adaptive shaperbegins to use shorteraveraging times and theresolution becomes worse,although the acquisition rateis better than with a fixedprocess time. The adaptivepulse shaper thus ensuresefficient measurement of asmany X-rays as possible. Themaximum acquisition rate thatthis type of processor canachieve will still be controlledby the shortest time allowedto measure a voltage step. Thiswill be similar to the maximumcount rate achieved by a fixedprocess time processor usingits shortest process time.

Fig. 18. Voltage ramps measured by:(a) a digital pulse processor with a fixed process time, and(b) a digital pulse processor using adaptive pulse shaping.The blue areas show the averages used to measure each step. With a fixed processtime, events like (2) and (3) which arrive closer together than TP are rejected. The fixedprocess time electronics continuously ‘strobes’ the zero level (red areas) when notmeasuring the voltage steps. In the adaptive pulse shaper the averaging time isallowed to vary before and after the step. Therefore some events like (1) are measuredwith longer averaging time than with the fixed shaper but others like (2) and (3) aremeasured with short averaging times and therefore worse resolution. The effectivenoise contribution is a complex mixture of Gaussians of different width and there is noway of measuring equivalent noise resolution by ‘strobing’. Throughput is thereforebetter than with a fixed shaper but resolution and peak shape change with count rateand resolution cannot be predicted by strobe measurement.

18a

18b

Time

Vo

ltag

e TP

(1)

(2)

(3)

Time

Vo

ltag

e

(1)

(2)

(3)

22

Using this type of shaping the analystloses control over resolution of thespectrum. The resolution varies with countrate, for example when moving from onephase to another, or with any change inbeam current. ‘Strobing’ cannot be used tomeasure resolution because the averagingtime is variable. As a consequence resolutionand peak shape are poorly defined andthis compromizes the accuracy of spectrumprocessing.

Does resolution change withcount rate?

Even when using a fixed process timeresolution may change with count rate whenusing analog or digital shaping. For example,this may be due to changes in the slope ofthe ramp caused by variable leakage current,or where the contribution of after-effects oframp restore increases with restore rate.Furthermore, instabilities in the detectorand variations in electronic baseline maybe accentuated by increasing count rate.

Although most systems should be ableto maintain resolution sufficiently forX-ray mapping, only the best designswill achieve stability (within 1%)required for accurate peakdeconvolution.

Pulse pile-up inspection

Pulse pile-up inspection channels areused to ensure that only one photon ismeasured at a time.

Inspection circuits (Fig. 13b) sense thearrival of an incoming event. Eachcircuit has a time constant thatdetermines the smallest voltage step andtherefore the lowest energy event that can

be inspected. If the time constant is tooshort, noise levels will be high and lowenergy events will not be efficientlydetected, but if too long it will be unable todistinguish between closely arriving eventsand the pile-up protection efficiency of thecircuit will be compromized. Pulse processorsshould have more than one of these circuits,each with a different time constant to ensureefficient pulse pile-up protection over thefull range of detectable energies. This meansthe processor will be suitable for use with athin polymer window detector for lightelement detection, as well as for measuringhigher energy lines.

The result of efficient pulse inspection is toprovide a spectrum with no pile-up artefactssuch as sum peaks. These peaks, causedwhen two X-rays arriving close together arecounted as one, may cause incorrect peakidentification, and inaccurate results (Fig. 19).If a photon is missed by all the inspectionchannels, it can contribute to a tail or ‘pile-up continuum’ on the high side ofevery peak.

Fig. 19. A spectrum collected from kyanite (Al3SiO5) at high input rate, showssum peaks that could suggest the presence of elements not present in thesample. The peak at 1keV is the result of oxygen X-rays arriving together

(O + O sum peak), but may be confused for a NaKα peak. The Al + Osum peak at about 2keV may be incorrectly identified as PKα.

23

Oxford Instruments Analytical – technical briefing

Comparing DifferentPulse ProcessorsFor accurate and efficient EDS analysis theperformance of the pulse processor is asimportant as the detector.

EDS detector specifications typically revealthe best possible performance that thedetector can achieve. Any pulse processorwill perform its best at very low count rates,when the voltage steps on the ramp arewidely spaced and easy to measure.Therefore detector specifications are oftenquoted at 1000cps. However, this count rateis well below what is required for efficientanalysis and only the best designs of pulseprocessor will maintain good and stabledetector performance as the input count rateis varied.

How does performance change when thecount rate is increased? At more commonlyused input rates between 2000 and 10000cps,processors may not be able to maintain thebest resolution if resolution degrades withcount rate, or process time is shortened toachieve a useful acquisition rate. Thereforeone useful measure of how an EDS hardwaresystem will perform is the resolutionachieved when the input rate is at least2500cps, and the acquisition rate is belowthe maximum for the process time chosen.

Modern EDS software is designed to givereliable automatic identification of X-raypeaks and accurate standardless analysis.This is a relatively straightforward task whenpeaks are well separated, but for overlappedpeaks where there is no clear valley between

the peaks, accurate energy calibration isvital. Some systems may appear to havestable performance with count rate becausepeaks do not move more than 5eV and X-rayline markers are always in the correctchannel at 10eV/channel. However, whenquantifying peaks about 35eV apart (e.g.SiKα and WMα) only a 4eV shift in energycalibration can introduce a 10 weight% error(Statham 2002). If the resolution also variesand the width of peaks is wrongly predictedby the software then larger errors may occur.

Analog pulse shaping designs have difficultymaintaining a stable baseline so the energycalibration may vary as count rate increases.Time variant shapers or digital pulse shapersshould show much less variation, andprovide more reliable results, without theneed to keep input rate constant. Moreover,processors which are able to constantlymonitor the zero level will be able tomeasure shifts in the baseline and, inaddition, some systems can also monitorresolution changes for correction bysoftware.

The best method to test how energycalibrations and resolutions change over auseful operating count rate range, and howwell these changes are compensated, is toanalyze real samples at a useful range ofinput rates, for example 1000, 2500, 5000and 10000cps. By doing this the effect ofany variation on the ability of a system toperform reliable analysis can be tested.The ability to resolve a severe overlap canbe tested by analyzing a pure material forelements known not to be present. Forexample, by collecting spectra from a pureelement such as Si, at different count rates.

24

If the spectra are quantified assuming Si, andTa and W are present (TaM and WM are veryclose to SiK), a system that will provideaccurate analysis in this count rate rangeshould find levels of Ta and W belowstatistical significance at all count rates. Byoverlaying spectra collected at the differentrates it may also be possible to see peakshifts or resolution changes with count rate.

For example Fig. 20 shows four spectracollected from pure silicon at different inputrates using a digital pulse processor wherecount rate stability is good. Quantitativeanalysis of these spectra, assuming Si, Taand W are present (Table 1) shows thatwithin statistical significance only Si ispresent, whatever count rate is used tocollect the data.

Energy (keV)

0

Co

un

ts

1.6 1.65 1.7 1.75 1.8 1.85 1.9

5000cps

10000cps

1000cps

2500cps

10000

20000

30000

40000

50000

60000

70000

80000

90000

100000

Fig. 20. Spectra collected from pure silicon at 1000, 2500, 5000 and 10000 cps input rate using a digital pulse processor measuringat the same process time. Even when displayed at 5eV per channel no change in resolution or position can be seen for these four spectra.

1000cps 2500cps 15000cps 10,000cps

Si 99.68 � 1.78 99.43 � 1.41 99.90 � 0.97 99.23 � 0.90

Ta -0.82 � 1.56 0.31 � 1.22 -0.99 � 0.85 0.04 � 0.78

W 1.14 � 0.89 0.26 � 0.70 1.09 � 0.49 0.74 � 0.46

Table 1Quantitative analysis of pure silicon calculated from the spectra shown in Fig 20 assuming Si, Ta and W are present.At all input rates Ta and W are below statistical significance (3 sigma). Note that negative numbers should be reported,because negative values greater than statistical significance also indicate incorrect spectrum deconvolution.

25

Oxford Instruments Analytical – technical briefing

Summary• Both detector and pulse processor

are equally important parts of themeasurement chain

• Pulse processor performance ischaracterized by the maximum acquisitionrate and resolution achievable at eachprocess time

• Measuring the change in energycalibration and resolution as the countrate changes shows how reliably themeasurement chain will provide accuratedata for automatic peak identificationand standardless quantitative analysis

• Even small changes in resolution orenergy calibration can lead to largeerrors when analyzing severely overlappedX-ray lines

ReferencesK. Kandiah, A. J. Smith and G. White,IEEE Trans. Nucl. Sci., NS-22, 2058 (1975)

P.J. Statham, In Proceedings NIST-MASSpecial Topics Workshop, “Understandingthe Accuracy Barrier in Quantitative ElectronProbe Microanalysis and the Role ofStandards” NIST, Gaithersburg, MD, USA,April 8-11, (2002)

26

27

Oxford Instruments Analytical Limited at High Wycombe, UK operates Quality Management Systems approved to the requirements of BS EN ISO 9001. This publication is the copyright of Oxford Instruments AnalyticalLimited and provides outline information only which (unless agreed by the company in writing) may not be used, applied or reproduced for any purpose or form part of any order or contract or be regarded as a representation relating to the products or services concerned. Oxford Instruments’ policy is one of continued improvement. The company reserves the right to alter without notice the specification, design or conditions of supply of any product or service. INCA® is the registered trademark of Oxford Instruments Analytical Limited. Oxford Instruments acknowledges all trade marks and registrations.

Oxford Instruments Analytical Limited, 2002. All rights reserved.Printed in England.

Ref: OIA/074/A/0702

PATENTSEP 0325383EP 0302716US 4931650GB 2192091US 5170229US 5357110JP 2557692JP 2581597EU 0568351

CERTIFICATE NUMBER FM29142

Oxford InstrumentsAnalytical

UKHalifax Road, High WycombeBucks, HP12 3SE EnglandTel: +44 (0) 1494 442255Fax: +44 (0) 1494 461033Email: analytical@oxinst.co.uk

AustraliaSydney, N.S.W. 1715Tel: +61 2 9484 6108Fax: +61 2 9484 1667Email: sales@oxinst.com.au

AustriaA – 1030 WienTel: +43 (0) 1 710 61 98Fax: +43 (0) 1 710 61 98Email: rob.wills@oxinst.at

ChinaBeijingTel: +86 (0) 10 6833 0336/5/4Fax: +86 (0) 10 6833 0337Email: oiadmin@oxford-instruments.com.cn

FranceSaclay, CedexTel: +33 (0) 1 69 85 25 24Fax: +33 (0) 1 69 41 86 80Email: analytical-info@oxford-instruments.fr

GermanyWiesbadenTel: +49 (0) 6122 937 176Fax: +49 (0) 6122 937 178Email: analytical@oxford.de

JapanTokyoTel: +81 (0) 3 5245 3591Fax: +81 (0) 3 5245 4466/4477Email: yasunori.yoda@oxinst.co.jp

Latin AmericaClearwater FLTel: +1 727 538 7702Fax +1 727 538 4205Email: oxford@gate.net

ScandinaviaLink Nordiska ABLidingö, SwedenTel: +46 8 590 725 50Fax: +46 8 590 725 58Email: info@linknord.seWeb: www.linknord.se

SingaporeTel: +65 6337 6848Fax: +65 6337 6286Email:analytical.sales@oxford-instruments.com.sg

USAConcord MATel: +1 978 369 9933Toll Free: +1 800 447 4717Fax: +1 978 369 8287Email: info@ma.oxinst.com

www.oxford-instruments.com

c