Scanning Electron Microscopy

Volume 1986 Article 7

Number 1 Part I

4-9-1986

Principles of X-Ray Spectroscopy

Edwin R. Levin
David Sarnoff Research Center

Follow this and additional works at: https://digitalcommons.usu.edu/electron

Part of the Biology Commons

Recommended Citation
Levin, Edwin R. (1986) "Principles of X-Ray Spectroscopy," Scanning Electron Microscopy: Vol. 1986 : No. 1
, Article 7.
Available at: https://digitalcommons.usu.edu/electron/vol1986/iss1/7

This Article is brought to you for free and open access by

the Western Dairy Center at DigitalCommons@USU. It
has been accepted for inclusion in Scanning Electron
Microscopy by an authorized administrator of
DigitalCommons@USU. For more information, please
contact digitalcommons@usu.edu.
SCANNING ELECTRON MICROSCOPY /1986/I (P ages 65-76) 0586-5581/86$1 .0 0+0S
SEM Inc., AMF O'H are (Chi cago ) , IL 60666 - 0507 USA

PRINCIPLES OF X-RAY SPECTROSCOPY

Edwin R. Levin

RCA Laboratories
David Sarnoff Research Center
P.O. Box 432
Princeton, NJ 08543 - 0432
Phone no.: (609) 734 - 3233

(R ece ived for publication March 31, 1984, and in revised form April 09, 1986)

Abstract Introduction

X-ray spectroscopy has become an essential The remarkable growth that has taken place in
tool for obtaining local che mi cal constit uent imaging capabilities and applications in both
data, particularly for identifying features scanning electron microscopy (SEM) and trans-
observed in imaging syste ms . This tutorial paper mi ssion electro n microscopy (TEM) during the past
is intended to introduce the basic ideas of X-ray 25 years has broug ht a dramatically increased need
spectroscopy, es pecially to the newcomer to to identify features observed. This need has been
scanning electron micro scopy (SEM) and trans- met largely by use of x - ray spectroscopy (XRS)
mission electron microscopy (TEM). The primary techniques in SEM and (later) TEM in strumen ts.
focus is on ene rgy-dispersi ve analysis (EDS); The bl endi ng of XRS and e l ectro n-opti cal imaging
however, the close parallels with wavelength - systems presents a powerful combination for
dispersive analysis (WDS) are s hown. Topics co mpre hensive analysis of finely detailed
inc lud e elec tron beam-sp eci men interactio ns, X- ray struct ur es . This paper is intended as a basic
generation and detection, spectral analysis, introduction to X-ray spectroscopy as applied in
approac hes to quantitation, and limits of electron - optical imaging systems. The e mphasis is
detection and resolution. on energy -di spersive analysis in SEM; however,
many comparisons with wavelength - dispersive
analysis are made, a nd numerous aspects of the
discussion are common to both. While the attempt
is made to give a comprehensive overview, a treat-
ment of this sort can only touch on many topics.
A broad general literature is available for fuller
details of a ny aspect of interest (See, e.g.,
references 1, 4, 6, 14, 16, 18, 19, 24, 27).

X-Ray Generation and Detection

Generation of charac teri stic X-rays

When a beam of energetic electrons strikes a
solid specimen, many different effects are
produced (Fig. 1). Each of these may be used as a
signal to provide information about the spec imen.
I
1° Electron Beam
Back seat tared
;Electrons
Characteristic X - Rays

KEY WORDS: X-ray Spectroscopy, X- ray Micro- Secondary

Electrons
' ""
- --..__:~
·"--,
/
Ca thodol um i nascence
11

____
.,.:re 9c~~on s
analysis, Energy Dispersive, Wavelength Di s per-
sive, Scanning Electron Micr oscope, Transmission
Electron Microscope, EDS, WDS.
Specimen
current Transmitted
Electrons

i
Fig . 1. Effects produced by e le ctro n beam
bombardment.

65
 Edwin R. Levin

Some signals are functions principally of the Wavelengt h-d ispersive spectroscopy (WDS)
physical configuration of the specimen; th e most Equation (1) il lu s trates that the X-ray
important of these are the seco nd ary electrons, photon may b e characterized by either its wave-
which are used to form the fa miliar SEM "emissive- length or its energy. X-ray detection may be
mode" i mages. Other signals are mor e closely acco mpl is hed by utilization of eit her of these
associated with the specific properties of the properties. In wavele ngth -di spersive spectro-
speci men, and may b e employed to obtain informa- ~ (WDS), a diffracting crystal is used to
tion about the material itself. Principal separate X-rays according to wavelength. The
examples of such signals include back - scattered crystal spectrometer depends on Bragg's law of
electrons, X- rays, Auger elec t rons and cathodo - X- ray diffraction, which states that a strong
luminescence. The analysis of characteristic diffracted beam is obtained only when
X-r ays --i. e., X-ray spec troscopy--is the most
widely used of these techniques; it has becom e nA = 2d si n 0 (2)
indispensable in many SEM laboratories, and its
application in TEM is growing rapidly. where A is the X- ray wavelength, d the spacing
Characteristic X- rays are produced when between adjacent planes of t he crystal, 0 the
elec trons make transitions from higher to low er angle between the crystal planes and the incident
energy states within an atom, as illustrated in (parallel) X- ray beam, and n is an integer. The
Fig . 2. An external excitation causes ejection of geometry involved in this relationship is shown in
an inner - shell (K) e lectron from the atom. The Fig. 3 . The diffracted X-rays are detected by a
vaca ncy in the (K) she ll is filled by an electron proportional counter, which is arranged to rotate
with the diffracted bea m as the crystal is scanned
through a range of angles.

INCIDENT
Incid en t X-RAY BEAM
Electron

CRYSTAL
d PLANES

Fig. 3. Geometry of wavelength-dispersive

detection (Bragg diffraction condition).
Fig. 2. Generation of characteristic X-rays.

Energy-dispersive spectroscopy (EDS)

In energy-dispersive spectroscopy (EDS),
from an outer (L) shell. The energy lost by the
X- rays are dete cte d by a silicon diode, formed by
electron in this transition appears as an emitted
diffusing a thin layer of lithium into th e surface
X- ray photon.
of a silicon crystal and then allowing the Li to
Several important points derive from this
drift uniformly through the 3mm thick crystal. A
picture. First, the energy of the emitted photon
schematic drawing of a lithiu m-drifted silicon
correspo nds exactly to the difference in th e
[Si(Li)] detector i s shown in Fig. 4. When an
energies of the two levels involved in the
X-ray is absorbed in the Si(Li) crystal, a number
electron transition. Since the atomic energy
of electron-hole pairs are created. Under the
levels--and consequently the differences between
action of an electric field in the crystal, set up
levels --ha ve specific values, the X- ray photons
by an ex t ernally applied voltage, th e excited
have well - defined energies completely character -
carriers are collected, giving rise t o a curr ent
istic of the ato m from which they are emitted:
pulse whose magnitude is proportional to the X-ray
E E E he photon energy. The induced curre nt pulse is con-
photon= L - K = Y:- (1)
verted to a voltage, amplified, an d scaled to
correlate with the photon energy.
Here EK and E1 are the energies of the electrons Figure 5 is a photograph of a t ypi cal
in the r especEive shells, a nd E h t on and A are energy -d is p ersive spectro meter. The Si(Li)
the energy and wavele ngth, respgc~ively, of th e detector mus t be maintained at low temperature to
emitted X-ray . Electron transitions b etween prevent d iffusion of the lithium. Also, since the
numerous atomic levels are possible; thus each actual current pulse from the detector is very
excited atom emits a series of X- rays, small, the field-effect transistor, which provides
char acterized by a specific distribution of the first stage of ampl ification, must be mounted
energies and intensities. This spectrum of X- rays close to the detector and opera t ed at low
identifies the ato mi c species uniquely. temperature to reduce noise. The detector a nd FET

66
 Principles of X-ray Spectroscopy
elements in the periodic tabl e. On the high
energy si de, the main limitation is that imposed
by the maximum available excitation energy.
Most SEM's have upper electron beam energy

r
capabilities of 20-30 keV; this limits efficient
n-TYPE Si X- ray production to the range be low 20 keV. The
X-RAY low - energy limit is determined by the detector
PHOTON METAL CONTACT
efficiency, which drops off sharply below about 1
~
kV, because the softer (i .e. , lower-energy) X-rays
GOLD are strongly absorbed in the Be window of the
CONTACT ~- -- +-H-- Li - DRIFTED detector. Fluorine (Z = 9) is usually the
REGION lightest element detectable with the conventional
Si(Li) detector. Recently, "windowless" Si(Li)
Si
DEAD detectors, which have provi sio n for removing the
LAYER Be window in suitable vacuum conditions, have
become practical and popular . With windowless
detection, the range of analysis is extended down
750 V to carbon (Z = 6) or even boron (Z = 5). For WDS,
a variety of analyzing crystals is available,
allowing detection of all elements down to Be
Fig. 4. Si(Li) energy-dispersive detector. (Z = 4) .
A block diagram of the main co mponents of an
EDS system is shown in Fig. 6. X-r ays excited by
electron bombardment of the specimen impinge on
the Si(Li) spectro meter (Figs. 4, 5), which
out puts a series of voltage pulses proportional to
th e X- ray energies. The pulses are amplified,
converted from analog to digital signa l s and input
to a mult i - channel analyzer (MCA). The MCA sorts
the incoming pulses, assigning eac h e nergy to a
particular chan nel and ke epi ng a cou nt of the
number of pulses (X-ray counts) in eac h cha nnel.
The data accumula t ed by the MCA are placed in
memory, and may be stored, displayed, or computer
processed as desired. Figure 7 is a typical
display of an X-ray spectrum from a mineral,
showing the number of counts (X- ray intensity) as

Liquid
Nitrogen
Reservoir Si (Li) L _J
Detector
Electron
FET Beam
Preamp

Amplifier H.V.
Fig. 5. Energy-dispersive spectrometer, wi th Supply
preamplifier and liquid nitrogen dewar
(Court esy of Princeton Gamma Tech). A/D Converter
are placed in contact with a "cold finger" from CRT
the liquid nitrogen reservoir and the whole
Multichannel Display
arrangement is enclosed in a metal tube that is
evacuated to reduce heat exchange through the Analyzer
walls. The end of the cylinder, directly in front
Printer/
of the Si(Li) detector, contains a very thin
Plotter
(typically 7 µm) Be window, which transmits
without appreciable loss all but the lowest-energy Com
rrnter Chart
X-rays, yet maintains the vacuum integrity of the Recorder
structure.
Energy-dispersive detectors are capable of
detecting X-rays from almost the entire range of Fig. 6. Block diagram of EDS system.

67
 Edwin R. Lev in

Qualitative Analysis

Characteristic X- ray series

The c haracteristic peaks in a spectrum
V,
identify the elements present in the specimen .
This ele mental identification is aided by the fact
that the X-ray peaks appear in series with known
energy and intensity relationships. K, L, and M
series are usually used. Fig. 8 shows Kand L
V, spec tra for several ele ments. In the K series,
I-
z: the Ka and K~ peaks always appear together. As Z
::::,
0 increases, the Ka and K~ energies -- and the
u
~~ separation between them--increase, and the peaks
LL.
0
"" nJ nJ become broader. For lower-Z ele ment s (e.g., Na,
::,: uu
""
w
0:,
:>L
:a: o>
::::, ::;:: ··········1··.. ········•·······...........................
z: '6 <!'..
'"6
s- :>L :>L l ·i
"" "" <l) <l)
u.. '-6<:n.. i
"" u..
[
:
I
l
I
'
CHANNEL -· ·•·•.......
26Fe
,..., "'r..1.1..,..,!f:,,
J . l
....n.........
111Wpm .. f

'
73To
'" .......

.
, .. .. .. . . ......

·i
.

Fig. 7. Energy-dispersive spectrum of a mineral. :

I

:'
a function of the MCA channel number (X-ray
energy),
peaks.
with identification of the e l emental i
I

..... . .................... -. '!.' ............. ~ ........... ...... ... , .. , , _,,,.,, ,,,l.il ,llill.,l!l,~il liillhiill,u .. .,

Comparisons of WDS and EDS 27CO -

•l
p
'
!
i 74
w ·I
·
·
This description of XRS systems has !
emphasized EDS. However, the brief discussion II !
allows so me immediate comparisons of WDS and EDS , I'

to be made. Table 1 lists the principal factors

,
f
'
f
relating to the two techniques. In many ways,
WDS and EDS incorporate similar ideas, considera-
tions and procedures. The principal differences
are that WDS gives higher resolution and lower
background levels, thus can produce better
quantitative results. However, since WDS spectrum
collection is a sequential process while the EDS
detector receives data from the entire spectrum
simultaneously, EDS spectrum collection generally
is much faster. Also, EDS commonly employs
appreciably smaller beam current, therefore may
produce less speci men damage than WDS. (a) ( b)

TABLE 1. COMPARISONSOF EDS AND WDS

Fig. 8. Wavelength -di spersive spectra of pure
metals, showing shifts in peak positions with
ms wns
l . 1\na ! ys is mode
atomic number. a. K-series, b. L-series.
Si mu I Laneous Sequentia l
2. Spectru m collection Li me H in uu•s Te ns of minu t es
3. ErlC'rgy reso lut ion 150 eV 5 cV Mg), the Ka and K~ peaks may not be resolved, but
4. ElemP nl de t ection l imit Be 1,dn d o-..•: Z > 9 appear together in a single narrow envelope.
Win do ,..,Jess: Z-~ 6(5?) Z ~ 4
For the higher-Z elements, the K ph oton
5. Backgro und Hay be App reciable Low
energy may be too high to allow the K X-rays to
6. Background corrcct1on Dif fi c u lt for low S i mple
energies be excited efficien tly (or at all) with the
7. Beam current (Typicdl)
accelerating voltages available. Then the L
( Min) peaks, which hav e lower energies than th e K X-rays
8. Probe dia meter (Ty p ical) soo R
from the same atom, becom e more useful for
(Mi n ) 50 ~ a nalysis. The L spec tra are usually identified by
9. Det Pc t or geomet ry Variable ; col l ection characteristic groupings of peaks comprising the
may be opt i mized La, L~ and Ly series. Most EDS L spectra appear
10. Spectral artifac ts Peak overlaps , sum and Rare
as four principal peaks: La, L~, 1~ and Ly; for
escape pea k s , pul se 2
pi le -u p lower Z elements (e.g., Ag), the four L peaks are

68
 Principles of X-ray Spectroscopy

distinguishable, but may not be compl etely speci me n . As an X-ray beam travels through a
resolved. With the higher resolution of WDS material, the intensity decreases expone ntially:
spectra , individ ual component peaks within the La,
L~, Ly series are clearly resolved (Fig. 8b). (3)
In s trong er spectra, other, l ess intense, L peaks
may be de te cted as well. Again, as Z increases, where xis the distance traversed and p the
the energies of the L lines and th e separatio n density of the medium. The mass absorption co-
between them inc rease, and the peaks become efficient, µ , is a function of the effective
broader. When the L-p ea k energies are too high to
atomic numbe~ of medium a nd also varies with
be excited efficie ntly, the M series may b e photon e nerg y, as shown in Figure 10. The absorp-
employ e d. Fig. 9 shows the relation between tion coeffic~ nt increases wit h decreasing photon
atomic number and energies of characteristic K, L, energy (as E 3) until a sharp discontinuity is
and M X-rays. The efficiency of X- ray genera-
reached at the absorption edge KAb' which occurs
tion -- and, therefore, the choice of lines used in at an energy just above that of Lfie K X-ray
analysis--is a function of the excitation emission . The strong absorption at KAb is related
conditions (electron beam energy, e t c.) employed, to efficie nt excitation of electrons Irom the K
as will be discussed later. shell of the Fe atoms, and is accompanied by cor-
respondingly s trong emission of Fe K photons.
X-rays with ene rgi es below KAb can not exc ite K
electrons in the Fe atoms ana, therefore,
experience relatively littl e absorption . However ,
as the X- ray e nergy is decreased below KAb' µ
again begins to increase as another absorptioW
edge (L) is approac hed. Absorption effects can
70
a:
UJ
influ e nce both the shape of the background
CD
:, 60 contin uum and the distribution of intensities of
::,
z 50
emitted characteristic peaks: regions of the
u spectr um for which th e speci men is highly
~
0
"' absorbing will ex hibit decreased intensities,
while the intensities of the c haracteris tic peak s
'"' JO of the absorbing atoms are enhanced. These
20
effects must be caref ully acco unt ed for in
10 quantification of the spectral data.
50 ,o 0 ,, 0 200 25 O

Energy (keV)

Fig. 9. Energy of chara cteris t ic p eaks as 500

function of atomic number (Courtesy of Fe
Princeton Gamma Tech). 400
Modern X-ray spectro meter systems have E
computer files of stored ele mental X-ra y photon ::I.. 300
energies a nd intensities. The peak positions and
relative intensities within series of any selected
ele ments may be displayed wit h the accumu l ated
200
spectra. This facility provides appreciable aid
in element identification of unkn own spectra. 100
Background radiation - Bremsstrahlung
The characteristic X-ray p ea ks appear on a 00 2 4 6 8 10 12 14 16
background continuum of non-characteristic X-RAY ENERGY (keV)
radiation . This continuum --o r Bremsstrahlung- -
radiation is produced as beam e lectrons are Fig. 10 . Variatio n of mass absorption
scattered by the atomic nuclei in the speci men. coefficient of Fe with photon energy.
The photon energies correspond to the e nergies
lost by individual electrons in the sca ttering
Spectrum analysis
proc ess, and can vary from zero up to th e full The X-ray spectrum from an unknown thus
energy of the incident electron. The continuum consists of a collection of well-defined peaks
radiation intensity is low at th e high er energies superimposed on a continuum background which may
(decreasing to Oat the incident e l ectro n energy), contain discontinuities at absorption e dges.
and increases with decreasing photon e nergy. At Complete qualitative analysis co nsist s of
the low e nerg y side, the observed continuum determining the elements present in the specimen
intensity drops off sharply as a result of rapidly by identification of all the c hara c teristic peak s,
decreasing (EDS) detector efficiency. making certain that t~corre c t intensity
relationships between peaks in th e same series are
Absorption observed. In this connection, it is important to
The appearance of a spectrum may also be note th at there are numerous situations in which a
markedly affected by absorption of X-rays in the peak of one element overlaps a peak of a different

69
 Edwin R. Levin

element. Fig. 11 shows the case of overlap of the

MnK~ and FeKa peaks . These situations must be Cu
recognized to assure proper identification of all
the constituents, and proper assignment of the Ni

Fe

MnKCX: FeKCX: Cr
Al
MnK13

Fig. 12. SEM image and X-ray dot maps of

stack of metal s heets .

5 6 7
keV
Fig. 11. Mn-Fe spectrum (solid line) with
computer-generated MnK~ peak (dashed line),
showing overlap of FeKa and MnK~ peak s.

peak intensities. Comprehensive procedures are

available for deconvolution of overlapped peaks
[21,22].
In some situations, it is suffic ient si mply
to identify the elements present without regard to
relative abundance. More frequently, it is
important to differ entiate between major and min or
constituents. This may be done qualitatively by Fig. 13. X-ray line profiles for specimen
observing the relative intensities of peaks of of Fig 12.
different elements in the same region of the
spectr um, wjth the min or co nstituent peak s
appearing much weaker than the major constituents.
Where differences between specimens are of on the specimen. Fig . 13 shows line profiles of
interest, it is often sufficient to compare the major constituents for the metals sandwich of
spectra directly, noting differences in the Fig. 12. The quality of the line profile can be
relative intensities of different e l ement peaks substa ntially improved by repeating the scan a
between spectra collected under the same number of times, acc umulating the counts in
conditions. s uccessive pas ses . Many SEM- EDS systems hav e pro-
Spatial distribution of elements visions for superimposing the line scan profiles
It is sometimes of interest to have (or dot maps) on the SEM image; this is obviously
information on th e spatial distribution of one or helpful in correlating the X- ray data with image
more elements. For these purposes, two alterna- features .
tive modes of analyzer operation and display are
employed: ele ment al (or "dot") maps and lin e Limits and Complications
profile analysis. In both of these modes, the
analyzer is set to count X-rays only within a Minimum detectability lim it
selected narrow band of energies , or "window", The range of ele ment s that may be detected
including only the peak of i nt eres t. Variations has been discussed previously, in connection with
in the intensity of X-rays within the window are conside ration of detectors. The minimum quantity
then observed as the electron beam is scanned over of a given element that may be detected is a
the specimen. question of considerable interest and importance.
In elemental mapping, the electron beam is The detectability limit is a function of many
scanned over a complete raster. A bright dot is factors, including the accelerating voltage
displayed in the image whenever a photon within employed, the X- ray peak analyzed, the X-ray co unt
the selected window is detected. The density of rate and integrating time, specimen composition
dots in any area then is related to the local and density, and detector efficiency and geometry.
concentration of the element of interest. This is Because of the number and range of these factors,
illustrated in Fig. 12, which shows the SEM image it is not possible to give a simple, universal
of a stack of sheets of common metals, along with value for the detectability limit. Fig. 14 shows
sections of dot maps for principal elements the variation in detectability limit with atomic
represented. number for a typical set of conditions. Under the
In line profile analysis, the electron beam most favorable conditions, constituents present
is scanned along a single line, and the number of at levels a little less than 0.1% may be detected.
X-ray counts for the selected peak is recorded as Doubling the analysis time would be expected to
a function of the beam position. The data are i mprove the detectability limit by a factor of
displayed as a plot of X-ray intensity vs position about ../2.

70
 Principles of X-ray Spectroscopy

,~wn-
10 ~------------------------, LOW Z HIGH Z

¥
9

1
.9
.8
.7
.6
"'""
?5~ (a)

ii .5
.=-
~ .4
:J
i'.: .3 THIN SECTIONN \ /
:J INTERACTIO~
iii
<(
t; .2
'
~Cl I
/
\
BULK SPECIMEN--\ I
.1
INTERACTION "-- ,,/ ( b)
.09
.OS
.07
.06 Fig. 15. a. Effect of beam voltage and speci men
.05 atomic number on beam-specimen interaction region .
.04 b. Comparison of bulk specimen and thin section
.03
resolution .

.02 interpreting EDS spectra is the appearance of the

so-called "artifact" peaks: sum peaks and escape
peaks.
.01 ~-----,----~----r------,------.---.---~ Sum peaks occur when two X-ray photons of
70
10 30 40 60 60
energy£ are incident on the detector almost
ATOMIC NUMBER
simultaneously. In that case, the system
circuitry cannot distinguish two separate pulses
Fig. 14. Typical detectability limits for EDS of energy E, but counts a single pulse of energy
analysis obtained at 25 kV for 200 sec (Courtesy 2E. Sum peaks appear only at high count rates and
of Princeton Gamma Tech). only for major constituents in the specimen. The
most probable situation is, e.g ., in a steel
Spatial resolution specimen, for two FeKa peaks (6.4 kV) to combine
The spatial resolution, i.e., the s mallest in a sum peak at twice the energy of the FeKa
area or volume that may be analyzed, is a function (12.8 kV). It is also possible for FeKa and FeK~
principally of the excitation voltage and the (7.06 kV) photons to combine, producing a peak at
effective density of the speci men. Although th e the correspo nding sum energy (13.46 kV); however,
electron beam in a modern SEM may be focussed to since the K~ is less intense, the 13.46 kV su m
a probe diameter of 100 i or less, the interaction peak would be significantly smaller than the 12.8
between the beam and the solid specimen -- and kV sum. Sum peaks disappear rapidly as the X-ray
hence, the excitation of characteristic X-rays-- count rate is reduced.
extends over a considerably larger region. Fig. Escape peaks have energies 1 .7 4 kV(=SiKa)
15(a) illustrates the differences in interaction below prominent characteristic peaks of the matrix
volume for different electron beam energies and elements. They result from interactions within
specimen atomic numbers. The electron beam- the Si(Li) detector. When an X-ra y photon (say
speci men interaction region may be several orders FeKa) is absorbed in the detector, many ioniza-
of magnitude greater than the beam diameter. For tions occur (at the expenditure of 3.8 eV per
exa mpl e , in iron excited by 20 kV electrons , electron - hole pair created); also inner-shell
X-rays are emitted from a region about 2.4 µm e lectrons are ejected from some Si atoms, leading
in diameter; with 15 kV excitation, the resolution to the emission of Si X-rays. If the Si X- rays
improves to about 1.3 µm. In thin sections , the are re - absorbed within the detector, no energy is
electron beam - specimen interaction is restricted lost, and the full energy of the FeKa is detected.
to a much narrower region, and the spatia l If, however, a Si X- ray escapes from the detector,
resolution for X-ray analysis is muc h closer to the net energy detected is reduced by the amount
the probe size, as illustrated in Fig . 15(b). of the SiKa energy (1.74 kV), and a count at the
escape peak energy, 6.40-1.74 = 4.66 kV, is
Artifact peaks registered. No escape peak can be produced for
The effects of X-ray absorption within the elements below Z = 15 (P), be cause the
specimen have been discussed; and the occurrence characteristic X- rays do not have sufficient
of overlapping peaks has been noted . Another energy to excite the SiKa . Higher energy X-rays
effec t that can lead to errors or confusion in yield weaker escape peaks, because the Si X-rays

71
 Edwin R. Levin
are produced deeper in th e detector and have less Specimen - dete ctor geometry
probability of escapi ng. The intensity of the The specimen-detector geometry involves two
escape peak relative to the parent peak can be principal factors which are important in influenc -
predicted as a function of energy : the P Ka escape ing the intensity of the dete c ted X-rays: these
peak ha s 1.4% the intensity of the parent, while are the X-ray take-off angle,~' and the solid
the ZnKa escape peak is only 0.13% as intense as angle, 0, subtended at the specimen by the
th e parent peak. detector. Fig. 16(a) illustrates the relationship
At extremely high count rates, an EDS system of these factors.
may be unable to properly register dead time or The detection solid angle, O, is given
photon energy. When this occurs, fictitious approximately by
values of count rate and dead time may be shown,
and severe peak distortions exte nding to the
0 '\, na2 sin a (4)
low - energy side of the true peak position s may
appear. These distortions disappear rapidly as
'\,
s2
the count rate is reduced to more moderat e
operating ranges. where a is the radius of the ac tive detector area,
S the distance from the point of X- ray emission to
the center of the detector, and a the angle
between Sand the detector plane. Since X-rays
SEM Parameters and Detection Geometry are emitted from the specimen in all directions,
0 determines th e fraction of emitted X-rays that
Electron beam voltage reach the dete ctor . Because of the inverse
The choice of excitation voltage i s important proportionality between O and S2 , moving the
in determining the intensity of X- rays produced, det ec tor closer to the specimen can be useful in
the spatial resolution and the region of the improving peak intensity in low count - rate
specimen excited. Characteristic X-rays are situa tions .
produced only when the incident electron e nergy The take-off angle,~. is the angle between
exceeds th e excitatio n potential for the X-ray the specimen surface and the line from th e
emission in question. Generally, overvoltages of speci men to the detector. In the geometry of Fig.
2- 3 times the p ea k e ner gy are desirable for 16( a)
efficient X-ray excita tion. Smaller overvoltages 2
re duc e the intensity of the peak. Higher over- . -1 D
~ = e + sin s (5)
voltages degrade the spatial resolution, as
discussed above, and increase th e depth in the
speci men at which X-rays are produced. It is not where 0 is the spec imen tilt angle and ZD is the
always possible to work within this optimum range vertical displa ce ment betwe e n the specimen and the
of overvoltage because of the limited available detector. The take-off a ngle is related to the
mi croscope voltage or a wide range of critical distance an X-ray generated beneath the surface
excita tion potentials in the speci men analyzed .
For materials co ntaining principally light
ele ments, it is ofte n sufficient to use 10 kV or
DETECTOR
less for analysis. This p roduces adequate X-ray (AREA 1ro 2 )
intensities while allowing excellent lateral and
depth resolution, factors of importance when
analyzing small features or r ela tivel y thin
s urface layers. For medium-Z e lements, 20-25 kV
electrons are co mmonly us ed to excite the K- series
peaks. For higher-Z elements, Lor M peaks are
usually used for analysis, and the electron beam
voltage must be chosen according to their
energies. In qualitative analysis of a completely (a )
unknown material, it i s sometimes important to
carry through analyses at several different
voltages to insure proper identification
elements present .
of the
j
Io

TO
Electron beam current e DETECTOR
The electron beam current is usually selected
to produce a suitable X-ray count rate. In most
SEM's, the beam diameter is very much smaller than d
the excitation region in the specimen, as
discussed above, so the variation of beam diameter
with beam current does not affect spatial resolu-
tion. However, it is i mportant to restrict the ( b}
beam current- - and the corresponding count rat e --to Fig. 16. EDS detector collection geometry.
moderate ranges, where artifact peaks are a. factors involving relati ve position of spec -
minimized and spectrum distortion effects imen and dete c tor . b. Relat io nship betw ee n
prevented.
tak e -off angle and X-ray path length in specimen.

72
 Principles of X-ray Spectroscopy
must travel through the specimen in its path X- ray intensity. These rapid shifts in intensity
toward the detector, thus to the amount of can be very misleading, falsely suggesting abrupt
absorption that takes place in the specimen. The changes in ele mental concentration.
relevant features of the geometry are shown in Fig. 16(a) shows the special case where the
Fig. 16(b). Here, an X-ray (hv) is generated at detector (cylinder) axis is horizontal and
a distanced below the surface of a specime n perpendicular to the specimen tilt axis. More
tilted at angle 6 relative to normal incidence often in electron-optical columns, neither of
of the electron beam. In its path to the these conditions is met. Rather the detector
detector, the X- ray traverses a distance x in axis is offset by a horizontal azimuth ang l e,
the specimen, and emerges at take-off angle~- typically 45°, fro m the perpendicular to the tilt
Substituting x = dcos6cos~ for t he X- ray path axis; also the detector axis is usually inclined
length, the absorption equation (3) becomes to the horizontal. In these cases, the equations
for O (5) and~ (6) must be modified to take
account of the additional geometric factors.
I (6)
0 The take-off angle and the detection solid
angle may be varied by adjusting the vertical
Because of the csc~ factor in the exponential, position of the specimen. However, the presence
the emitted (detected) X-ray intensity may vary of a collimator or other physical structure at the
appreciably with take - off angle. Fig. 17 shows front end of the detector may restrict the range
the ~- dependence obtained from equation (6) for of specimen positions from which X-rays may reach
X-rays generated 1 µm below the surface of an the detector.
untilted specimen (6 = 0). Absorption losses
increase at an increasing rate as the take-off Approaches to Quantitation
angle is decreased . When the absorption
coefficient is large, as in the case of AlKa The intensities of the peaks in a spectrum
radiation in a Cu matrix, the intensity varies can be used to obtain the concentrations of the
rapidly with take-off angle, even as~ approaches elements in the specimen. However, this is not a
90°. In mor e moderate absorption situations, the straightforward process. The intensity of a
emitted intensity is relatively insensitive to spectral line increases with the concentration of
changes in tak e -off angle until~ becomes much the element, but usually not in a linear manner.
smaller, e.g., less than about 40° for CuKa Also, since the efficiency for production of
X-rays in iron. To minimize the strong dependence X-rays varies from element to element, even with
of detected X-ray intensity on take-off ang le, it the same excitation conditions, the concentrations
is advisable to make~ as large as possible. of constituents in a compound specimen are not
An associated problem in low take-off angle obtained simply from ratios of the elemental
situations is that relativ ely small lateral intensities.
displacements of the electro n beam on the speci men
may produce effective changes in~ which, because Empirical methods
of the steep slopes in some portions of the curves An empirical approach (3,32] provides the
in Fig. 17, are accompanied by large shifts in most direct way to obtain actual concentrations
from the elemental peak intensities . If standard
materials with known compositions similar to the
unknown are available, it may be possible to
construct calibration curves of measured intensity
as a function of concentration for each of the
elements concerned. The composition of the
unknown may then be determined simply by referenc -
ing the observed elemental intensities to the
calibration curves. Alternatively, a set of
simultaneous equations derived from the calibra-
0 0.4 t io n data might be used to correct for inter-
11>
H element effects. In practice, however, the
'.,..0.3
H
empirical
suitable
methods have restricted
standards generally
use, because
are not available.

k-ratio
The first estimate of the concentration of an
0.2 element is the ratio of the intensity of a
characteristic peak for the element in the unknown
to the intensity of the same peak obtained under
the same analytical conditions in the pure
material. This fraction is called the k-ratio for
the element. In the absence of absorption and
other matrix effects, the k-ratio would be c l ose
30 60 90 to the actual concentration of an element , and the
1/t(deg) tota l concentration of the specimen could be
obtained simply from the k-ratios of all the
Fig . 17. Effect of varying take - off angle on constituents. Because X- ray interactions within
peak intensity. the specimen are appreciable, the k-ratios

73
 Edwin R. Levin
generally are not good approximations for the In the "conventional" method, considered by
elemental concentrations, and corrections must be most workers to be the most accurate, pure element
made. intensities are obtained by direct measurement of
standard materials. These may be pure elements
ZAF corrections (e.g., common metals: Al, Cu) or compounds of known
The theoretical approach to quantitation of stoichiometry or composition (e.g., FeS ) that_can
the X-ray spectroscopic data is to calculate 2
be prepared in the proper form. It is imperative
corrections for the matrix effects from that the standards data be collected under
fundamental considerations, and use these to conditions as nearly identical as possible to
obtain the true specimen composition [18, 20]. those under which the unknown is measured. These
Three factors are involved in modifying the conditions include the electron beam current and
k-ratio data to account for matrix effects. voltage, specimen tilt, and specimen-detector
These are: geometry. In addition, the specimen (standard or
Z, the atomic number correction. This factor unknown) should be flat- - at least over the area
adjusts for the "stopping power" of the scanned--in order to keep the take-off angle
material, which is a function of the constant and well defined. For the best results,
effective atomic number. It also corrects it is often recommended that the standards be
for the fraction of ele ctrons that are measured immediately before and/or after the data
backscattered from the specimen and do not are collected for unknowns. It is possible,
contribute to X-ray excitation. and widely practiced, to store libraries
A, the aLsorption correction. This facto r of standards data to be called up when required;
corrects for absorption of X-rays by other the accuracy of the analysis in such cases is
elements in the matrix. strongly dependent on how closely the conditions
F, the fluorescence correction. This factor employed in obtaining the unknown spectrum
corrects for enhancement of the peak correspond to those which apply to the standards.
intensity as the result of excitation by The most reliable analyses generally employ
X-rays from other elements in the matrix. standards derived from materials similar to the
Application of these factors to obtain the unknowns. This is especially true in studies of
corrected concentrations of the constituents in ceramics and minerals, where oxygen is generally
the matrix is commonly referred to as the ZAF determined by difference, and in low - Z element
method [31]. All three of the ZAF factors can be analyses with windowless detectors, where matrix
calculated [9,10,25,26]; however, the values are correction factors are less well known.
themselves dependent on the composition to be An alternative approach, employed very widely
determined. Thus, an iterative technique[8,22] is and successfully, is to use measured values for a
used, utilizing the relation selected group of elements to derive a curve that
C = k·Z·A·F (7) gives pure-element intensities for all elements
[2]. Such a curve is specific to the series of
The k-ratios determined from the unknown spectrum X- ray lines (e . g . , Ka) employed, the excitation
are used as estimated concentrations, and from conditions (beam voltage and current) and the
these the ZAF factors are determined. These specimen/detector geometry (tilt, take-off angle,
values are applied to equation (7), and an etc.). Fig. 18 shows a pure-element intensity
improved estimate of concentration, C, is ob- curve for Ka peaks obtained at 16 kV. Separate
tained. A second application of equation (7) is curves are required for K-, L- and M-series peaks,
then made, using new values of the ZAF factors and for each electron beam energy of interest.
obtained from the revised values of the C's . The The utility of the curves derives from the fact
process is repeated until no further changes in that ED detector efficiencies remain relatively
the concentration values occur. Programs current- constant over considerable periods of time. It is
ly in use obtain convergence in 3 to 4 iterations . usual to store the pure-element intensity curves
In order to use the ZAF method, k-ratios must
be supplied for all elements of significant 300 ..-------------------,
concentration in the specimen, even though some (/)
t-
may not be of interest. However, the k- ratio of z
one (and only one) element may be obtained by :::> Ka
0
difference, if all other elements are represented u 16 kV
by measurements. A prominent example is that of ::.:: 20 0
analysis of minerals, where the oxygen k-ratio is >-
often obtained by difference. t-
(/)
z
w 100
Use of elemental standards t-
The success of application of the ZAF method z
H
depends on the quality of the k-ratio data input.
::.::
The k- r atio itself is i mportantly affected by the <!
appropriateness of the value used for the pure w
a... 0 ~-,L2-1L4 _ 1L6-1~8-2~0-2~2---'2~4---'
2~6---'2~8-=3o
element i ntensity. Several different methods for
deriving the k-ratios for ZAF analyses are in
ATOM IC NUMBER
common use; these methods differ principally in
their use of standards for the pure-element Fig. 18. Pure-element Ka peak intensity as
intensities. function of element atomic number.

74
 Principles of X-r ay Spectroscopy
(or th e data from which they are derived), and 2. Barbi NC, Skinner DP, Blinder S. (1976) The
then call them up when specific stan dard intensity calculation of pure element X-ray intensi-
data are required for an analysis. One mode of ties from empir ically derived expressions,
employment, preferable to relying co mpl etely on and its application to SEM/EDS analysis,
prerecorded intensity data, is to measure one in: Proc. 11th Annual Conf. of the Micro-
standard along with the unknown spectrum, and use beam Analy sis Soc. , Sa n Francisco Press,
that intensity to scale the pure-element standards San Francisco, 8.
curve. An alternative is to scale the standards 3. Bence AE, Albee A. (1968) Empirical correc-
curve in direct proportion to the beam current tion factors for the electron microanalysis
employed. of silicates and oxides. J. Geology 76,
Pure - element intensity curves also can be 382 -40 3. -
derived from first principles (13,29]. This 4. Bertin EP. (1978) Introduction to X-Ray
calculation is based on a knowledge of the Spectrometric Analysis, Plenum Press, New
fluorescent yield of the X-ra ys, the detector York, Chaps. 1-4, 6, 7, 9.
characteristics, and th e beam energy employed 5. Blum F, Brandt MP. (1973) Evaluation of the
(2,5,11]. With such theoretical pure-element use of a scanning electron microscope
intensity data, it is possible to obtain co mbin ed with an energy dispersive X- ray
quantitative analyses without reference to any analyzer for quant i tative analysis. X-r ay
measured pure-element standard. "No standards" Spectrom. ~. 121-124.
analyses are fairly successful for major 6. Boekestein A, Stadhouders AM, Stols ALH,
constituents of medium-Z materials (e .g ., Roomans GM. (1983) Quantitative biological
stainless steels), but are not applicable to low-Z X-ray microanalysis of bulk speci mens, Scan-
element analysis. ning Electron Microsc. 198 3; II: 725-736.
The methods discussed to this point apply to 7. Cliff G, Lorimer GW. (1972) Quantitative
specimens which are "infinitely thick" with analysis of thin metal foils using EMMA -4 :
respect to electron beam penetration and X-ray The ratio technique, in: Proc 5th Annual
exci tation. Analysis of thin sections requires a European Conf. on Electron Microscopy, In-
so mewhat different approach (2 3,29 ,30]. In thin stit ute of Physics, London, 140-141.
sections (see Fig. 15), X-ray absorption and 8. Criss JW, Bir ks LS. (1966) in: The Electron
fluorescence effects often are negligible, so ZAF Microprob e , TD McKinley, KFJ Heinrich and DB
corrections are not necessary (15], and the Wittry (eds), Wil ey, New York, 217.
intensity of an X-ray peak may be assumed to be 9. Duncumb P, Reed SJB. (1967) in: Quantitative
directly proportional to the elemental intensity. Electron Probe Microanalysis, KFJ Heinrich
Variations or uncertainties in speci men thickness ( ed) , US Nat'l. Bur. Stds. Publ . 298,
may complicate development of reliable reference Washington, DC.
intensity standards. However, the ratios of pure 10. Eckert R. (1982) A post-correction met hod to
e lement intensities obtained from sta ndard s curv es improve the results of a ut omatic ZAF correc -
suc h as Fig. 18, corrected for thin section tion in quantitative analysis by X-rays,
parameters, may provide an excellent basis for Scanning Electron Microsc. 1982; IV:
quantitation. Several methods incorpo rating thes e 1429-1433.
ideas for analysis of thin sections hav e been 11. Gedcke DA, Byars LG, Hardy WH. (1982) ZAP-A
described (7,1 2, 17]. standardless X-ray microanalysis co mputer
program, Scanning Electron Microsc. 1982;
Counting statistics III: 981-993.
Successful quantitative analysis with any of 12 . Geiss R, Huang T. (1975) Quantitative X-ray
the methods discussed requires that the spectra e nergy dispersive analysis with the trans-
contain sufficient X-ray counts for statistical mission elec tron mi croscope. X-ray Spectrom.
accuracy. The standard deviation of a peak ~. 196- 20 1.
intensity I is ✓ I. Statistically, this mea ns 13 . Goldstein JI, Costley JL, Lorimer GW, Reed
there is only a 65% probability that the tru e SJB. (1977) Quantitative X-ray analysis
intensity is in the range I± ✓ I. For this reason, in the electron mi croscope , Scanning Elec-
it is unlikely that a single intensity measure - tron Microsc. 1977; I: 315-323.
ment will yield the true intensity. Clearly the 14. Goldstein JI, Newbury DE, Echlin P, Joy DC,
statistical error improves as th e measured Fiori C, Lifshin E. (1981) Scanning Electron
intensity in creases. Since the precision of Microscopy and X-ra y Microanalysis, Plenum
quantitative analysis cannot exceed that of the Press, New York, Chaps. 5 - 8.
measured intensities, the analyst must assure that 15. Goldstein JI, Lorimer GW, Cliff G. (1976)
the number of counts in the analyzed peaks are Quantitative X-r ay analysis of thin films
sufficient to keep the statistical errors within in the electron microscope, in: Pro c .
desired limits. 11th Annual Conf. of Microbeam Analys is
Soc., San Francisco Press, San Francisco,
25A-25C .
References 16. Gooley R. (1981) Electron probe mi croa na lysis
in geosciences, Scanning Electron Microsc.
1. Barbi NC, Skinner DP. (1976) Tec hni ques for 1981; I: 493-502.
elec tron microscop ic identification of small 17. Hale TA. (1968) in: Quantitative Electron
particles, Scanning Electron Microsc. 1976; Probe Analysis, KFJ Hei nri ch (ed), Nat'l Bur.
I: 394 -400 . Stds. Publ. 298, Washington, DC, 269.

75
 Edwin R. Levin

18. Heinrich KFJ. (1981) El e ctron Beam Micro- 26. Reed SJB . (1975) Electron Microprob e
analysis, VanNostrand Re inhold, New York, Analysis, Cambridge University Press, Chap.
Chaps 5, 7-9, 12. 9, 10, 17 .
19. Heinrich KFJ, Newbury DE, Myklebust RL 27. Russ JC (ed). (1971) Energy Dispersion X-ray
(eds) . (1981) Energy Dispersive X- ray Analysis: X-ray and Electron Probe Analysis,
Spectrometry, Nat'l. Bur. Stds. Special Special Tech. Publ. 485, American Society
Publication 604, Washington, D.C., 1-4, for Testing and Materials, Philadelphia,
71-96, 177-192, 381 - 390. PA 19103.
20. Jenkins R, Gould RW, Gedcke D. (1981) 28. Russ JC. (1974) Quantitative microanalysis
Quantitative X-ra y Spectrometry, Marcel with min i mum pure element standards, in:
Dekker, New York , Chaps. 1-4, 8-10. Proc. 9th Annual Conf. of Microbeam Analysis
21. Lifshin E, Ciccarelli MF, Bolon RB. (1975) Soc., San Francisco Press, San Francisco, 22.
X-ra y spectral measurements and interpreta - 29. Russ JC. (1984) Fundamentals of Energy
tion, in: Practical Scanning Electron Dispersive X-ray Analysis, Butterworths,
Microscopy, JI Goldstein, H Yakowitz (eds), London, Chap. 13.
Plenum Press, New York, Chap 5. 30. Williams DB, Goldstein JI, Michael JR .
22 . Maggiore C, Foster M. (1975) Least squares (1982) Quantification of energy-dispersive
peak fitting within the ZAF iteration loop, spectra from thin-foil specimens, in:
in: Proc. 10th Ann. Conf. of Microbeam Microbeam Analysis 1982, KFJ Heinrich (ed),
Analysis Soc., San Francisco Press, San San Francisco Press, San Francisco, 21-36.
Francis co, 24. 31 . Yakowitz H. (1975) Methods of quantitative
23. Nockolds C, Cliff G, Lorimer GW. (1980) X-ray analysis used in electron probe micro-
Characteristic X-ray fluorescence correction analysis and scanning electro n microscopy,
in thin-foil analysis. Micron 11, 325-326. in: Practical Scanning Electron Microscopy,
24. Panessa-Warren BJ. (1983) Basicbiological JI Goldstein, H Yakowitz (eds), Plenum
X-ra y microanalysis, Scanning Electron Press, New York, Chap. 9.
Microsc. 198 3; II: 713-723. 32. Ziebold TO, Ogilvie RE. (1964) An empirical
25. Philibert J. (1962) Method for calculating method for electron microanalysis.
the absorption correction in electron-probe Analytical Chemistry 36, 322 - 327.
microanalysis. Proc. 3rd Int'] Symposium on
X-ray Optics and X-Ray Microanalysis, HJ{
Pattee, VE Cosslett, A. Engstrom (eds),
Academic Press, New York (1963), 379 - 392. Editor's Note: All of the reviewers' concerns were appro priate-
ly addressed by text changes , hence there is no Discussion with
Reviewers.

76