Appli cat i on N ote 5 2 7 6 9

SEM/EDS Techniques for

Analyzing NdFeB Magnet Material
in the Automotive Industry
John Konopka, Thermo Fisher Scientific, Madison, WI, USA

Key Words
NORAN System 7 X-ray Microanalysis System, NORAN Analytical
Software (NSS), COMPASS Phase Mapping, Heterogeneous Materials,
Neodymium Magnet, Permanent Magnets, Point Analysis, SEM/EDS

Introduction
Permanent magnets are very important materials for the
automotive industry. Typical automobiles today use a large
number of magnets in motors and sensors throughout
a car. These are used for drive motors, power steering,
power windows, starters, generators, power seats, fuel
sensors and more.

For the most part, these magnets are comprised of Nd,

Fe and B. However, additions of small amounts of other
For most other technologies the sample is first put into
elements change the magnetic properties of the material
solution by digestion in an acid or other solvents. These
as well as the corrosion properties and improve utility at
techniques produce a very high precision analytical result
higher temperature.
describing the average composition of the material but
As part of incoming inspection, automobile companies lacking in any structure information. This is critically
need to confirm that the correct grade of material was important as it is the size, distribution and composition of
supplied or whether the supplied material satisfied the the phases which, in part, determine the magnetic properties
requirements of the grade. In some cases it may be of the material.
important to investigate failures.
Although heterogeneous materials cannot easily be
The nature of the material makes this a difficult analysis. quantified they may be characterized. It is straightforward
NdFeB magnets are formed by sintering a mixture of the to inspect the material and analyze anomalies. Also, it may
powdered constituents at certain temperatures and be possible to know that two materials are either very
pressures. The result is a dominant matrix populated by at similar or different even if the materials cannot be
least two other phases. The exact composition depends on quantitatively described. Finally, it is possible to measure
the grade of the material. Heterogeneous structures like the microstructure in a reproducible manner. Following
this are difficult to analyze. are descriptions of the possible measurements that can
be employed.
To obtain the average composition with high precision,
the typical procedure would be to use X-Ray Fluorescence A tool well suited for this kind of analysis is a Scanning
(XRF) technology, Inductively Coupled Plasma Mass Electron Microscope (SEM) equipped with an Energy
Spectroscopy (ICP-MS) or similar technique. In order to Dispersive X-ray Spectrometer (EDS). The SEM uses a
use these analytical methods one must first homogenize fine electron beam to probe the sample. The EDS is able
the sample. For XRF the sample is milled and blended and to identify the elements present wherever the electron
pressed into a pellet or dissolved in a flux to form a disk. beam impinges on the sample. The electron beam can be
2

focused very tightly such that features much less than one
micron across can be inspected. Moreover, little sample
preparation is required beyond polishing the sample to
a flat surface. For magnets, one additional requirement
is that they be demagnetized.

In cases where a fracture has occurred, the fracture face

may be inspected in the SEM directly, with no polishing or
grinding, to attempt to identify any foreign inclusions that
may have caused the fracture.

Point Analysis
One of the simplest EDS techniques is to visually inspect
a sample and direct the electron beam to conduct an
analysis at a specific point. In this technique an image is
Figure 1: Point and Shoot regions selected for analysis captured from the electron microscope then the analyst
points at regions in the image to be analyzed. The
computer directs the electron beam to those areas and
conducts the analysis. The result is one or more elemental
analyses as well as an image documenting their locations
and the appearance of the sample.

Figure 1 shows a typical field of view. The scale marker

indicates the length of 5 microns in the image. This is a
backscattered electron image (BSE). In this image the
brighter areas indicate higher average atomic number.
The visible inclusions range from about five microns
across to sub-micron.

Note in the figure that three areas are highlighted. The

spectra for these three areas are shown in Figure 2. In this
example one measurement was made of the matrix, one of
a dark grey inclusion and one of a bright white particle.
(See reference1 for more information about X-ray spectra
and how to read them.)
• The matrix consists largely of Nd and Fe
• The bright particles consist of Nd, Pr, and O
• The dark particles also consist of Nd, Pr, and O but
with more O than the bright particles
Figure 3 shows the visual comparison of the spectra from
the two kinds of inclusions seen in the sample.

The approximate composition of each area may be

calculated. This data is presented in Table 1. In a simple
analysis the accuracy is about two or three percent. With
care this can be greatly improved. It is important to be aware
that these analyses represent just the tiny points indicated
in the image, and do not represent the whole sample.

Table 1: Quantitative estimates of compositions at the four analyzed regions

indicated in Figure 1

O-K Fe-K Pr-L Nd-L

Figure 2: Spectra from each of the three analysis regions indicated in Figure 1 NdFeB(4)_pt1 15.5 2.1 17.4 65.0
NdFeB(4)_pt2 22.0 3.4 15.5 59.1
NdFeB(4)_pt3 75.2 4.5 20.3
 3

Figure 3. Comparison of spectra from the bright (pt 1)

and grey (pt 2) inclusions seen in Figure 1. Even
without numerically analyzing them, comparisons
provide a way for contrasting the compositions of
multiple features.

In special cases where the peaks are difficult to separate

by EDS a different spectrometer can be mounted on the
SEM and used together with the EDS detector. The
Wavelength Dispersive Spectrometer (WDS) is an excellent
complement to EDS as it provides much higher peak
resolution which in turn provides for much greater
certainty and precision when performing quantitative
analysis of trace elements.2

Point analysis is well suited to quickly characterizing

features which can be selected by eye. This can easily be
used to check the composition of the matrix and other
known features as well as the composition of anomalies
that appear in the field of view.

Mapping Figure 4: Greyscale image of area mapped with Spectral Imaging

A more sophisticated analysis technique is called Spectral
Imaging (SI). In this case the field of view is selected by
the analyst then the computer rasters the electron beam in
a grid pattern across the sample. At each node in the grid
a spectrum is measured and stored. The result is a two
dimensional array of spectra that describe the elemental
composition of this area.

For a fuller discussion of this technology and what it

provides see the following reference.3 For the purposes of
this discussion only two features of SI will be used.

The main purpose of mapping is to reveal the distribution

of an element in the field of view. Because the technique
analyzes a large area it does not depend on the analyst
locating a feature by eye for inspection. Every part of the
sample is analyzed. Figure 4 shows an image of the area
mapped and Figure 5 shows the spectrum acquired from
the entire field of view. This shows the elements that may
be found in this region.

This spectrum reveals the presence of Nd, Pr, Fe, Nb, Co, Figure 5: Gross spectrum from full field of view of Figure 4
O, C, and a trace of Dy.

Figure 6 shows maps of the discovered elements except

for Dy which is present at too low a level to map easily in
this particular sample and C which is likely merely a
contaminant. These are processed maps in which back-
ground information and peak interference overlaps have
been removed.
4

Maps reveal the size and shape of the inclusions containing

each element. These maps clearly show that the Nb and
Co are located in just a few areas, while Fe and Nd are
abundant throughout the map.

Because a full spectrum is stored at each point in the map

it is easy to extract a spectrum from a feature to confirm
the composition. Figure 7 shows the spectra extracted
from Nb and Co rich regions confirming the correctness
of those identifications.

Displays of Net Counts values for specific elements requires

knowing that one should look for those elements. Another
technique parses the Spectral Imaging data set automatically
and can find elements that weren’t expected to be present.
Figure 6: Maps showing elemental distributions in a region of the sample
Available exclusively on the Thermo Scientific™ NORAN™
System 7 X-ray microanalysis system, COMPASS™ uses
principal component analysis (PCA), a powerful algorithm4
that finds element groupings in the field of view then
produces a map of the abundance of that set of elements
across the field of view.

For a simple example consider a sample that has three kinds

of magnesium silicates – olivine, pyroxene and serpentine. All
contain Mg and Si but in different ratios. Simple mapping
would show maps for Mg and Si. COMPASS PCA would
identify each ratio as a different component and produce
a map showing the distribution for each component.

COMPASS phase mapping groups together nodes in the

map with similar values and flags it as a unique phase.
The software then calculates the percent area of that
phase in the image and collects all of the individual
spectra from each point in the phase and sums them into
Figure 7: Extracted spectra from Nb and Co rich regions. Note that Co is just at the trace level. one spectrum. This reveals a map that shows the
composition of the phase.

Figure 8 shows the COMPASS PCA results calculated from

the same map data shown above. The components look
like spectra but actually mathematically derived shapes.
 5

Figure 8: COMPASS PCA components and distributions

6

Figure 9 shows the phases calculated from the results. The tool for screening incoming material. Note that in the
spectrum for each phase is shown as well. results COMPASS PCA automatically isolated the Nb and
Co rich areas even though they were minor constituents.
The proprietary method is very useful as no prior knowledge
COMPASS PCA is very good at finding features that are
about the sample is required before applying it to the map.
different from the matrix, even if they comprise a very
No judgements are required from the analyst so every
small fraction of the field of view.
operator gets the same results. This makes it an excellent

Figure 9: Phase distributions

 7

A second sample was examined with Spectral Imaging.

The image is shown in Figure 10 and the maps are shown
in Figure 11. In this case it was suspected that P and S might
be present. They appear to exist at three points in this
field of view.

Figure 12 shows the spectrum extracted from the SI data

set. Because the particle was so small in this image the
spectrum was quite weak. To improve the confidence of
the identification Point and Shoot was used. Figure 12
also shows this image and the spectrum acquired from a
point indicated by the map as containing P and S. As the
spectrum shows the identification was correct. In this way
mapping and point analysis complement each other. SI
helps in the rapid identification of where something might
be located while Point and Shoot excels at a confident Figure 10: Image of area mapped in second example
identification of the sample at a specific area.

To quickly evaluate a relatively large area SI can be

extended with the Analysis Automation option which
integrates NORAN analytical software (NSS) software
with the stage motors of the electron microscope. In this
way an analyst can setup an analysis to collect maps from
a grid of areas, move that data to another workstation
and then process it offline to search for anomalies. In this
way product can be screened in a routine manner and when
anomalies are discovered they can be brought to the
attention of a senior engineer.

Figure 11: Elemental maps of area shown in Figure 11. Note the presence of P and S.
8

Figure 12: The spectrum (top left) was extracted from the map at one of the
spots in Figure 11 indicating the presence of P and S. To improve the precision
of the identification Point and Shoot was used at the location indicated by
the map. This enlarged image is shown below. The spectrum (bottom left)
was acquired from Point 1 in the image. It shows with greater confidence the
identification of P and S in this sample at this point.

Match This requires some small effort to setup the methodology

Another technology that can be brought to bear is the and sample preparation for particular materials as improper
Match algorithm. Most analytical techniques attempt to sampling would bias the result. Once the method has been
characterize a sample by measuring the abundance of established for a particular material it can be employed
each element in the sample. This works well for simple, with little operator input.
homogeneous samples. However, there is no standard way To setup a test several spectra were taken from a NdFeB
of analyzing a heterogeneous sample. sample – some from the matrix and some from the bright
However, a remarkable feature of EDS is that if the same areas with slightly different composition.
material is measured under the same instrument conditions In the first example, a spectrum was taken from a different
the same spectrum will be obtained (except for small, area of the matrix. The results are shown in Figure 13. A
statistical fluctuations). Due to the physics of X-ray better match is indicated by a smaller Chi-squared value.
generation nothing else matters. This is true even for As expected, the two other matrix spectra are show very
complex, heterogeneous samples. low chi-square values.
Match takes advantage of this repeatability by attempting
to identify materials by numerically comparing a spectrum
from a material with a set of spectra taken from known
samples. Rather than trying to identify the elements in each
spectrum and measure their abundance it simply looks at
the shapes of the spectra and finds the most similar shapes
among the candidate spectra. This makes it well suited to
answering the question “Is material A reasonably similar
to material B?” even when the materials are complex.

Figure 13. Match results for unknown from the matrix

 9

A second spectrum was acquired from a bright particle.

The result is shown in Figure 14. This is a good match for
the first type of bright.

Once it has been setup Match is a very fast way to identify

materials, even those with complex compositions.

Match also can be extended with Analysis Automation

such that a large number of areas across a sample can be
atomically measured and processed. With complex samples
a distribution of results from a large number of measure-
ments is often more helpful to characterize a sample.

Feature Sizing and Chemical Typing (FSCT) Figure 14: Match results for unknown from a bright particle
Finally, Feature Sizing and Chemical Typing can be used
to count the different visible features in a field of view and
measure their sizes, shapes and elemental compositions.
Most often this is used with Analysis Automation to
measure features over a large number of fields. The result
is a table of numbers characterizing hundreds, if not
thousands of features in a material. Histograms are
calculated for measurements such as area, diameter and
perimeter, enabling the analyst to quantitatively view the
nature of the sample.

With modern EDS detectors and SEMS the measurement

of each feature may require from about 0.5s to 5s depending
on the nature of the feature and the desired precision.
Setting up a method for a particular sample type may require
some investment of effort after which the application of
the method is simple and routine and mostly automated.
Figure 15: Field of view used to generate features sizing data
As an example the region shown in Figure 15 was analyzed
by FSCT. The discriminators were set to select the bright
features. Fifty one particles were detected. Figure 16 shows
the area measurements plotted as a histogram. Many other Statistics
kinds of measurements are possible. Minimum: 0.81 µm2
Maximum: 21.04 µm2
An X-ray spectrum was taken from each particle and Mean: 5.10 µm2
Median: 2.70 µm2
stored in the analyzer. The software automatically sorted StdDev: 5.39 µm2
these spectra into five groups. Table 2 lists these categories. Number of Frames: 1
The analyst may create their own categories or they may Average Field Area: 3333.18 µm2
Average Field Area%: 7.80%
modify the generated categories and apply them to the
stored data and resort the data.

Figure 16: Histogram plot of the measured areas of 51 particles in a sample of NdFeB magnet

Table 2: Categories automatically determined by software when spectra were acquired from particles. These mixed element categories consist of a center value
and a range for each element. It is also possible to define categories as a single element and a minimum value.

Fe ± Pr ± Nd ± O ±
Learn 0 20 10 21 10 53 10
Learn 1 19 10 66 10 9 10
Learn 2 34 10 18 10 46 10
Learn 3 55 10 10 10 31 10
Learn 4 8 10 21 10 55 10 14 10
 Appli cat i on N ote 5 2 7 6 9
Tables 3 and 4 are summaries of the particle results. Table 3: Report summarizing the distribution of found particles among the
Table 3 lists the area fractions of particles in each composition categories
category and Table 4 summarizes the size ranges of each Chem Type # Particles % Particles % Total Area
composition category.
All Types 51 100 7.80
Figure 17 shows part of the result table. The full table of
Learn 0 12 24 0.83
results for all particles is available to the analyst as a
spread sheet. Learn 1 26 51 5.87
Learn 2 6 12 0.64
In case of complex samples it is possible to differentiate
particles such that even particles within particles can Learn 3 3 6 0.16
be measured. Learn 4 4 8 0.31
The value of this kind of measurement arises from the Table 4: Report summarizing the size distributions of found particles of the
context of comparing it to similar analyses of known good composition categories
samples. Then it is possible to see if particle sizes and shapes Area
have changed. This is a powerful technique for discovering 0.00–10.00 10.00–20.00 20.00–30.00 30.00–Max
Ranges
if incoming material has changed in the size, shape, and All Types 44 5 2 0
composition of the constituent features. Enormous
Learn 0 12 0 0 0
numbers of particles may be sized and analyzed and the
results summarized and categorized by the software. Learn 1 20 4 2 0
Learn 2 5 1 0 0
Site License
An additional characteristic of this platform is that it Learn 3 3 0 0 0
includes a site license. With this the identical application Learn 4 4 0 0 0
software used to collect the data can be installed at an
unlimited number of computers around the laboratory.
This makes it possible to review data and generate reports
while the SEM/EDS is busy collecting fresh data.

Summary
• NdFeB is a difficult material to analyze because it
is a composite consisting of micron to submicron
sized grains
• SEM/EDS provides multiple ways to analyze individual
features down to dimensions less than one micron
• Point and Shoot provides for the analysis of individual
features identified by the analyst
• SI Net Counts mapping shows the spatial distribution
of elements chosen by the analyst in a field of view
• COMPASS phase mapping automatically parses a map
into its constituent phases with little input from the
operator beyond setting up the analytical conditions Figure 17: Sample of the full table of results of particle measurements

• Match provides a simple, reliable method for identifying

materials based on the similarity of their measured References
spectra to spectra from known materials 1. E
 nergy Dispersive X-ray Microanalysis: An Introduction
(TM50627_E 06/13M).
• Feature Sizing is a powerful technique that measures
the sizes and shapes and compositions of the visible 2. W
 hite Paper: Principles and Applications of Parallel Beam WDS
(WP52608_E).
particles in a field of view
3. E
 DS Spectral Maps with XPhase (AN52456).
• Analysis Automation is an extension of the other
4. K
 otula P. G., Keenan M. R., and Michael J. R. 2003. Automated
features which allows these analyses to be automatically
analysis of SEM X-ray spectral images: A powerful new microanalysis
conducted at a large number of points across a sample tool. Microscopy and Microanalysis 9:1–17.

www.thermoscientific.com
©2015 Thermo Fisher Scientific Inc. All rights reserved. All trademarks are the property of Thermo Fisher Scientific and its subsidiaries.
This information is presented as an example of the capabilities of Thermo Fisher Scientific products. It is not intended to encourage use of these
products in any manners that might infringe the intellectual property rights of others. Specifications, terms and pricing are subject to change.
Not all products are available in all countries. Please consult your local sales representative for details.

Africa +43 1 333 50 34 0 Denmark +45 70 23 62 60 India +91 22 6742 9494 New Zealand +64 9 980 6700
Australia +61 3 9757 4300 Europe-Other +43 1 333 50 34 0 Italy +39 02 950 591 Russia/CIS +43 1 333 50 34 0
Austria +43 810 282 206 Finland/Norway/Sweden Japan +81 45 453 9100 Spain +34 914 845 965
Belgium +32 53 73 42 41 +46 8 556 468 00 Latin America +1 561 688 8700 Switzerland +41 61 716 77 00
Canada +1 800 530 8447 France +33 1 60 92 48 00 Middle East +43 1 333 50 34 0 UK +44 1442 233555
China +86 21 6865 4588 Germany +49 6103 408 1014 Netherlands +31 76 579 55 55 USA +1 800 532 4752
AN52769_E 11/15M