Energy-Dispersive X-ray Spectrometry in the AEM

1
Energy-Dispersive X-ray Spectrometry in the AEM
Charles Lyman
Based on presentations developed for Lehigh
University semester courses and for the Lehigh
Microscopy School
2
Why Do EDS X-ray Analysis in TEM/STEM?

• Spatial resolution
• 2-20 nm (103 times better than SEM/EPMA)
• Elemental detectability
• 0.1 wt - 1 wt, depending on the specimen (same
  as SEM/EDS)
• Can use typical TEM specimens (t 50-500 nm)
• EELS requires specimens lt 20-30 nm
• Straightforward microanalysis
• Qualitative analysis gt Which element is present?
• Quantitative analysis gt How much of the element
  is present?
• Easy x-ray mapping

3
Example of an X-ray Spectrum

• 2 Types of X-rays
• Characteristic x-rays
• elemental identification
• quantitative analysis
• Continuum x-rays
• background radiation
• must be subtracted for quantitative analysis

Example of EDS x-ray spectrum
from Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer,
2003.
4
Continuum X-rays

• Interactions of beam electrons with nuclei of
  specimen atoms
• Accelerating electric charge emits
  electromagnetic radiation
• Here the acceleration is a change in direction
• The good
• The shape of the continuum is a valuable check on
  correct operation
• The not-so-good
• I bkg increases as ib increases
• I bkg is proportional to Zmean of specimen
• I max bkg rises as beam energy rises
• Peak-to-background ratio
• Ratio of Ichar / Ibkg sets limit on elemental
  detectability

Continuum x-rays
Absorption of continuum
from Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer,
2003.
5
Generation of Characteristic X-rays

• Mechanism
• Fast beam electron has enough energy to excite
  all atoms in periodic table
• Ionization of electron from the K-, L-, or
  M-shell
• X-ray is a product of de-excitation
• Example
• Vacancy in K-shell
• Vacancy filled from L-shell
• Emission of a Ka x-ray
  (or a KLL Auger electron)
• Important uses
• Qualitative use x-ray energy to identify elements
• Quantitative use integrated peak intensity to
  determine amounts of elements

6
Compute Energy of Sodium Ka Line
Beam electron
If beam E gt EK, then a K-electron may be
ionized
?
?
X-ray energy is the difference between two energy
levels
L
M
K
Energy levels EK and EL3 are in Bearden's "Tables
of X-ray Wavelengths and X-ray Atomic Energy
Levels" in older editions of the CRC Handbook of
Chemistry and Physics
For sodium (Z11)
EK 1072 eV EL3 31 eV
Beam electron loses EK
For Na only see one peak since the Kb is only 26
eV from the Ka line
7
Families of Lines
If K-series excited, will also have L-series
Note this is a simplified version of Goldstein
Figure 6.9 showing only lines seen in EDS
8
Fluorescence Yield ?

• ? fraction of ionization events producing
  characteristic x-rays
• the rest produce Auger e
• ???increases with Z
• ?K typical values are
• 0.03 for carbon (12) K-series _at_ 0.3 keV
• 0.54 for germanium (32) K-series _at_ 9.9 keV
• 0.96 for gold (79) K-series _at_ 67 keV
• X-ray production is inefficient for low Z lines
  (e.g., O, N, C) since mostly Augers produced
• ?? for each shell ?K???L???M
• X-ray production is inefficient for L-shell and
  M-shell ionizations since?
• ?Land??M always lt 0.5
• ?L 0.36 for Au (79)
• ?M 0.002 for Au (79)

from Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer,
2003.
9
X-ray Absorption and Fluorescence

• X-rays can be absorbed in the specimen and in
  parts of the detector
• Certain x-rays fluoresce x-rays of other elements
• X-rays of element A can excite x-rays from
  element B
• Energy of A photon must be close to but above
  absorption edge energy of element B
• Example Fe Ka (6.40 keV) can fluoresce the Cr
  K-series (absorption edge at 5.99 keV)

Greater absorption when -- x-ray energy is just
above absorber absorption edge -- path length t
is large
10
EDS Dewar, FET, Crystal

• LN dewar is most recognizable part
• To cool FET and crystal
• Actual detector is at end of the tube
• Separated from microscope by x-ray window
• Crystal and FET fitted as close to specimen as
  possible
• Limited by geometry inside specimen chamber

Schematic courtesy of Oxford Instruments
11
Electron-Hole Pair Creation

• Absorption of x-ray energy excites electrons
• From filled valence band or states within energy
  gap
• Energy to create an electron-hole pair
• ? 3.86 eV _at_ 77K
• (value is temperature dependent)
• Within the intrinsic region
• Li compensates for impurity holes
• Ideally electrons holes
• electron-hole pairs is proportional to energy
  of detected x-ray

Conduction band
Excited e-
Donor
Energy
Energy gap
Acceptor
Hole
Valence band
For Cu Ka 8048 eV/3.8 eV 2118 e-h pairs
(after drawing by J. H. Scott)
12
Details of Si(Li) Crystal
Ice?
Anti-reflective Al coating 30 nm
Gold electrode
Gold electrode 20 nm
Silicon inactive layer (p-type) 100 nm
X-ray
Electrons
Holes
()
()
Window Be, BN, diamond, polymer 0.1 mm 7 mm
Active silicon (intrinsic) 3 mm
(after drawing by J. H. Scott)
13
X-ray Pulses to Spectrum
spectrum
charge staircase
slow amplifier
analog-to-digital converter
energy bins
collects e-h as charge
fast amplifier
from Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer,
2003.
14
Slow EDS Pulse Processing

• EDS can process only one photon at a time
• A second photon entering, while the first photon
  pulse is being processed, will be combined with
  the first photon
• Photons will be recorded as the sum of their
  energies
• X-rays entering too close in time are thrown away
  to prevent recording photons at incorrect
  energies
• Time used to measure photons that are thrown away
  is dead time
• Lower dead time -gt fewer artifacts
• Higher dead time -gt more counts/sec into spectrum
• Processor extends the live time to compensate

fast amplifier
slow amplifier
slower amplifier
Counting is linear up to 3000 cps (20-30 dead
time)
from Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer,
2003.
15
Things for Operator to Check

• Detector Performance
• Energy resolution (stamped on detector)
• Incomplete charge collection (low energy tails)
• Detector window (thin window allows low-energy
  x-ray detection)
• Detector contamination (ice and hydrocarbon)
• Count rate linearity (counts vs. beam current)
• Energy calibration (usually auto routine)
• Maximum throughput (set pulse processor time
  constant to collect the most x-rays in a given
  clock time with some decrease in energy
  resolution)

Topics in red explained in next few slides
16
Energy Resolution

• Natural line width 2.3 eV (Mn Ka)
• measured full width at half maximum (FWHM)
• Peak width increases with statistical
  distribution of e-h pairs created and electronic
  noise
• Measured with 1000 cps at 5.9 keV
• Mn Ka line

E x-ray energy N electronic noise F Fano
factor (0.1 for Si) E 3.8 eV/electron-hole pair
Mn Ka line
from Williams and Carter, Transmission Electron
Microscopy, Springer, 1996.
17
X-ray Windows

• Transmission curve for a windowless detector
• Note absorption in Si
• Transmission curves for several commercially
  available windows
• Specific windows are better for certain elements

from Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer,
2003.
18
Ice Build Up on Detector Surface

• All detectors acquire an ice layer over time
• Windowless detector in UHV acquires 3µm / year
• Test specimens
• NiO thin film (Ni La / Ni Ka)
• Cr thin film (Cr La / Cr Ka)
• Check L-to-K intensity ratio for Ni or Cr
• L/K will decrease with time as ice builds up
• Warm detector to restore (see manufacturer)

Windowless detector
Immediately after warmup
After operating 1 year
Courtesy of J.R. Michael
19
Spectrometer Calibration

• Calibrate spectrum using two known peaks, one
  high E and one low E
• NiO test specimen (commercial)
• Ni Ka (high energy line) at 7.478 keV
• Ni La (low energy line) at 0.852 keV
• Cu specimen
• Cu Ka (high energy line) at 8.046 keV
• Cu La (low energy line) at 0.930 keV
• Calibration is OK if peaks are within 10 eV of
  the correct value
• Calibration is important for all EDS software
  functions

0.930 keV
8.046 keV
from Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer,
2003.
20
Artifacts in EDS Spectra

• Si "escape peaks
• Si Ka escapes the detector
• Carrying 1.74 keV
• Small peak 1 of parent
• Independent of count rate
• Sum peaks
• Two photons of same energy enter detector
  simultaneously
• Count of twice the energy
• Only for high count rates
• Si internal fluorescence peak
• Photon generated in dead layer
• Detected in active region

from Williams and Carter, Transmission Electron
Microscopy, Springer, 1996
21
Expand Vertically to See EDS Artifacts
from Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer,
2003.
22
EDS-TEM Interface

• We want x-rays to come from just under the
  electron probe, BUT
• TEM stage area is a harsh environment
• Spurious x-rays, generated from high energy
  x-rays originating from the microscope
  illumination system bathe entire specimen
• High-energy electrons scattered by specimen
  generate x-rays
• Characteristic and continuum x-rays generated by
  the beam electrons can reach all parts of stage
  area causing fluorescence
• Detector can't tell if an x-ray came from
  analysis region or from elsewhere

23
The Physical Setup

• Want large collection angle, W
• Need to collect as many counts as possible
• Want large take-off angle, a
• But W reduced as a is increased
• Compromise by maxmizing W with a 20 at 0 tilt
  angle
• can always increase a by tilting specimen toward
  detector -- but this increases specimen
  interaction with continuum from specimen

from Williams and Carter, Transmission Electron
Microscopy, Springer, 1996
24
Orientation of Detector to Specimen

• Detector should have clear view of incident beam
  hitting specimen
• specimen tilting eucentric
• specimen at 0 tilt
• Identify direction to detector within the image
• Analyze side of hole "opposite the detector
• Keep detector shutter closed until ready to do
  analysis

EDS detector
Rim
Edge of hole furthest from detector
Thinned
Detector
from Williams and Carter, Transmission Electron
Microscopy, Springer, 1996
Top view of disc
25
Spurious X-rays in the Microscope

• Pre-Specimen Effects
• spurious x-rays gt hole count due to column
  x-rays and stray electrons
• spurious x-rays gt poor beam shape from large C2
  aperture
• Post-Specimen Scatter
• system x-rays gt elements in specimen stage, cold
  finger, apertures, etc.
• spurious x-rays gt excited by electrons and
  x-rays generated in specimen
• Coherent Bremsstrahlung
• extra peaks from specimen effects on
  beam-generated continuous radiation

The analyst must understand these effects to
achieve acceptable qualitative and quantitative
results
26
Test for Spurious X-rays Generated in TEM

• Detector for x-rays from illumination system
• thick, high-Z metal acts as hard x-ray sensor"
• Uniform NiO thin film used to normalize the
  spurious "in hole" counts, thus
• NiO film on Mo grid

Electron beam
Beam-generated Ni K x-ray (good)
Spurious (bad) Mo K-series
NiO film on C
Hole

Mo grid bar
- on film results in hole results -
Usually take inverse hole count ratio
Hard x-ray from illumination system
see Egerton and Cheng, Ultramicroscopy 55
(1994) 43-54
27
Spectrum from NiO/Mo
Measure in center of grid square on NiO/Mo
specimen
Maximize P/B ratio for Ni Ka

• Spurious x-rays
• Inverse hole count (Ni Ka/ Mo Ka)
• Want high inverse hole count
• Fiori P/B ratio
• Ni Ka/B(10 eV)
• Increases with kV
• Want high to improve element detectability

Minimize spurious Mo x-rays by using thick C2
aperture
Egerton and Cheng, Ultramicroscopy 55 (1994) 43-54
28
Figures of Merit for an AEM

• Fiori PBR full width of Ni Ka divided by 10 eV
  of background
• (Ni Ka) / ( Mo Ka) is inverse hole count

Obviously, we want to use the highest kV
from Williams and Carter, Transmission Electron
Microscopy, Springer, 1996
29
Beam Shape and X-ray Analysis

• Calculated probes (from Mory, 1985)
• Effect on x-ray maps (from Michael, 1990)

C2 aperture too large
correct C2 aperture size
witchs hat beam tail excites x-rays
Properly limited Spherically
aberrated
30
Qualitative Analysis
Which elements are present?

• Collect as many x-ray counts as possible
• Use large beam current regardless of poor spatial
  resolution with large beam
• Analyze thicker foil region, except if light
  elements x-rays might be absorbed
• Scan over large area of single phase gt avoid
  spot mode
• Use more than one peak to confirm each element

31
Qualitative Analysis Setup 1
Specimen

• Use thin foils, flakes, or films rather than
  self-supporting disks to reduce spurious x-rays
  (not always possible)
• Orient specimen so that EDS detector is on the
  side of the specimen hole opposite where you take
  your analysis
• Collect x-rays from a large area of a single
  phase
• Choose thicker area of specimen to collect more
  counts
• Tilt away from strong diffracting conditions
• (no strong bend contours)
• Operate as close to 0 tilt as possible (say, 5
  tilt toward det.)

32
Qualitative Analysis Setup 2

• Microscope Column
• Use highest kV of microscope
• Use clean, top-hat Pt aperture in C2 to minimize
  hole count effect
• Minimize beam tails
• (C2 aperture or VOA should properly limit beam
  angle)
• Use 1 nA probe current to maximize count rate
• This may enlarge the electron beam (analyze
  smaller regions later)
• Remove the objective aperture

33
Qualitative Analysis Setup 3

• X-ray Spectrometer
• Ensure that detector is cranked into position
• Keep detector shutter closed until you are ready
  to analyze
• Use widest energy range available (0-20 keV is
  normal)
• 040 keV for Si(Li) detector
• 080 keV for intrinsic Ge detector
• Choose short detector time constant (for maximum
  countrate)
• Count for a long time  100-500 live sec

34
Peak Identification

• Start with a large, well-separated, high-energy
  peak
• Try the K-family
• Try the L-family
• Try the M-family
• Remember -- these families are related
• Check for EDS artifacts
• Repeat for the next largest peak
• Important
• Use more than one peak for identification
• If peak too small to "see", collect more counts
  or forget about identifying that peak peak
  should be greater than 3B1/2

35
Chart of X-ray Energies (0-20 keV)
from Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer, 2003
36
Chart of X-ray Energies (0-5 keV)
from Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer, 2003
M-series
L-series
K-series
37
Know X-ray Family Fingerprints
38
Some Peaks will Look Similar
At low energies each series collapses to a single
line From 1 keV to 3 keV, the K, L, or M lines
all look similar At 2.0 keV Z 15 (P) Ka
_at_ 2.013 keV Z 40 (Zr) La _at_ 2.042 Z 77
(Ir) Ma _at_ 1.977
M-series
Z
L-series
K-series
after Goldstein et al., Scanning Electron
Microscopy and X-ray Microanalysis, Springer, 2003
39
Unknown 1
Energy (keV)
40
Data Analysis for Unknown 1
26
6.4 Fe(26) Ka
4.5
7.0 Fe(26) Kb
4.9
41
Unknown 1
from xray.optics.rochester.edu/.../spr04/pavel/
42
Unknown 2
after www.pentrace.com/ nib030601003.html
43
Analysis of Unknown 2
Start
44
Qualitative Analysis
45
Automatic Qualitative Analysis?
Check the results of every automatic qualitative
analysis

• Are suggested elements reasonable? Tc and
  Pm are unusual, Cl and S are not
• 2. Do not use peak energy alone to identify
• Lines of other elements may have the same energy
• Consider logic of x-ray excitation
• All lines of an element are excited in TEM/STEM
  (100-300 kV)
• If L-series indicated, K-series must be present
• If M-series indicated, L-series must be present

46
Automatic Qualitative Analysis Blunders
Identification by peak energy alone without
considering x-ray families or peak shape
Identification without considering other lines of
same element
from Newbury, Microsc. Microanal. 11 (2005)
545-561
47
Summary

• EDS in the TEM has more pitfalls than in SEM
• Use the highest kV available
• Understand the effects of
• detector-specimen geometry
• spurious x-rays from the illumination system
• post-specimen scatter
• beam shape and spatial resolution gt the witchs
  hat
• Identify every peak in the spectrum
• Even artifact peaks
• Forget peaks of intensity lt 3 x (background)1/2
• Collect as many counts as possible
• Use large enough beam size to obtain about 1 nA
  current
• Qualitative analysis use
• use long counting times or
• thicker electron-transparent regions with a short
  pulse processor time constant, if appropriate
• Assume data might be used for later quantitative
  analysis (determine t if possible)