A step-by-step analysis of the polishing
process for tungsten specimens
Armin Manhard*, Gabriele Matern and Martin Balden
Max-Planck-Institut für Plasmaphysik, EURATOM Association, Boltzmannstraße 2, 85748
Garching, Germany
*corresponding author: armin.manhard@ipp.mpg.de
Keywords:
tungsten, polishing, cross-section, focused ion beam (FIB), scanning electron microscopy
(SEM), deformation layer
Abstract
The different stages of the polishing process of polycrystalline tungsten samples were
investigated by scanning electron microscopy of both the sample surface and of cross-sections
prepared with the help of a focused ion beam. It is shown that a distortion layer is present at
the sample surface after mechanical fine grinding and even after polishing with diamond
suspension, although the sample has a mirror-like finish. A sufficiently long chemo-
mechanical polishing step using an alkaline colloidal silica suspension was able to remove
this distortion layer. Although electropolishing produced an even smoother surface, the
microstructure quality after chemo-mechanical polishing is comparable to that of an
electropolished sample.
Introduction
Tungsten is a promising material for the plasma-facing wall in future nuclear fusion reactors.
In the environment of such a fusion reactor, the wall is subjected to large fluxes of deuterium
and tritium ions with typical energies between several eV and several 100 eV. For safety
reasons, it is important to estimate the amount of radioactive tritium that can be stored in
plasma-facing components due to ion implantation [1]. Therefore, tungsten specimens are
exposed to deuterium plasmas or deuterium ion beams under laboratory conditions to study
the interaction of hydrogen isotopes with tungsten surfaces. In order to perform such
experiments under controlled conditions, it is important to use specimens with well-defined
surfaces, in particular since the stopping range of hydrogen isotope ions at energies relevant
for fusion reactor applications (see above) is short, between a few nm and several 10 nm.
It is generally known to metallographers that mechanical polishing of metals can lead to
deformation layers at the surface (see, e.g. [2]), in particular for soft metals such as copper.
Nevertheless, this is also observed for tungsten [3]. Apart from obscuring the true bulk
microstructure during, e.g., scanning electron microscopy (SEM) analysis, such a deformation
layer could have a large influence on the deuterium retention, which can depend strongly on
the microstructure of the tungsten samples [4]. There is also experimental evidence that a
deformed surface layer can influence, e.g., the formation of blisters on deuterium-plasma-
exposed tungsten [3]. There are numerous methods for removing this deformation layer, e.g.,
electropolishing [2]. For the work presented here, a final chemo-mechanical polishing step
using an alkaline colloidal silica suspension was applied.
The aim of the work presented here was to optimise the polishing procedure for tungsten
specimens. Therefore, a set of specimens was first ground with increasingly fine SiC abrasive
paper (P400 up to P4000). The polishing was then continued with 1µm diamond suspension,
followed by the final chemo-mechanical step. At several intermediate steps of the process
individual specimens were removed in order to study the effect of the polishing procedure up
to that point. In the following, the specimens were analysed with a FEI HELIOS NanoLab
600 dual-beam scanning microscope, which is equipped both with an electron beam for
imaging and a focussed ion beam (FIB) for in situ cross-section preparation. To complete the
picture, roughness measurements were performed with an Olympus LEXT OLS4000 confocal
laser scanning microscope.
Experimental
The specimens used in the experiments presented here were hot-rolled tungsten with a
nominal purity of 99.97 wt.% and manufactured by PLANSEE. They had a thickness of
0.8 mm and were cut into rectangular pieces of 12x15 mm
2
. All specimens were from one
single manufacturing batch and can therefore be considered to be identical.
The first specimen was removed after the final grinding step with P4000 SiC paper. This step
was performed with a force of 20 N for 1 minute. Subsequently, polishing with diamond
suspension was performed with a force of 35 N on a textile pad for 10 minutes. After this
treatment, the second specimen was removed. The remaining specimens were chemo-
mechanically polished with alkaline colloidal silica suspension (Logitech "SF1") for 10, 20
and 30 minutes, respectively. For this polishing step a synthetic felt pad (Logitech "SUBA-
X") was used, the applied force was 30 N. The specimens were then all cleaned with
isopropanol in an ultrasonic bath in order to remove any residual polishing agent or debris.
For comparison, an additional specimen was electropolished for 2-3 minutes with 2.5%
NaOH solution in water after grinding. The polishing voltage was 25 V, the electrolyte
temperature was stabilised at 20 °C. The specimen was then heated to 1200 K in vacuum to
remove any residual oxide layer that formed during electropolishing. The grain structure of
the sample did not change during this heat treatment.
On each specimen, plan view images of the surface as well as cross-section images of a FIB
cut parallel to the rolling direction were acquired. The images were typically recorded with a
secondary electron detector. It should be noted at this point that the orientation contrast visible
in SEM images does not allow discerning between small-angle and large-angle grain
boundaries. Taking into account the fabrication method of the specimen (i.e., hot rolling), it
can be expected that most of the grains visible in the images are separated by small-angle
grain boundaries and are, accordingly, actually sub-grains, as it was reported, e.g., for hot-
rolled molybdenum [5].
All surface images presented here were acquired with the specimen surface normal parallel to
the electron beam. Prior to the FIB cut, a protective layer of an amorphous Pt-C compound
was deposited in situ on the specimen surface by locally decomposing a precursor gas with
the help of the electron beam. The electron beam was chosen for the deposition instead of the
ion beam in order to avoid any artefacts produced by ions impinging on the unprotected
surface during the initial phase of the layer deposition. This layer is visible as a dark, slightly
structured area at the top of all cross-section images presented here. For cutting and imaging
of the cross-section, the specimen surface normal was tilted by 52° with respect to the
electron beam. This corresponds to normal incidence of the ion beam. All cross-section
images are presented without a geometric tilt correction for the 38° inclination of the cross-
sectional plane towards the electron beam.
For each of the samples, a roughness analysis was performed on a 256×256 µm
2
scan
acquired by the confocal laser scanning microscope. This image size corresponds roughly to
the lower end magnifications typically used for SEM analysis of the grain structure. For
example, the surface SEM images presented here have a full width of 25.6 µm. RMS
roughness values R
q
were calculated for 3 horizontal, 3 vertical and 2 diagonal profiles of the
3-D surface maps. The values given for each sample are averaged over the values for these 8
profiles. The error margins correspond to the standard deviations of each such set of
roughness measurements.
Results
The specimen that was removed after grinding (see Figure 1) shows numerous scratches, pits
and cracks, and an apparently very fine grain structure on the surface (a). The cross-section
image (b) reveals that this is due to a top layer of about 0.5 µm where the grains have
fragmented. Deeper into the bulk, the grains are much larger. After diamond polishing, the
specimen surface appears already mirror-like to the naked eye. During this step, the near-
surface layer of strongly fragmented grains is mostly removed (see Figure 2). However, a
clear grain structure is not yet discernible in the surface SEM image (a). The grains at the
surface accordingly still have to be considered as heavily deformed. The deformation
becomes especially obvious when comparing the top layer of grains with bulk grains in the
cross-section images (b). Also, both the surface and the cross-section view reveal that the
specimen surface still contains many cracks, pores and small craters. It can therefore already
be concluded that diamond polishing alone is insufficient for producing specimens where the
near-surface grains have the same properties as the bulk grains.
The following chemo-mechanical polishing step then gradually removes the damage that
remains after polishing with diamond suspension. However, at least 30 minutes of this
treatment are necessary to fully remove all traces of a deformation layer. A clear grain
structure is then visible on the specimen surface (see Figure 3a), with only a small number of
scratch traces. The cross-section image (Figure 3b) shows that the top layer of grains is now
indeed representative for the bulk of the specimen. The slight deformations visible in some
surface grains occur in bulk grains as well.
Surface (Figure 4a) and cross-section (Figure 4b) images of the electropolished specimen look
remarkably similar to those of the specimen that was chemo-mechanically polished for 30
minutes. Only some grain boundaries show slightly enhanced contrast in the surface image of
the surface (Figure 4). This can be attributed to slight preferential etching of grain boundaries
intersecting with the surface. As for the chemo-mechanically polished specimen, there is also
no discernible difference between bulk grains and grains adjacent to the specimen surface.
This provides further evidence that the remaining distortions, which are visible in some grains
as different grayscale shadings, are not preparation artefacts. They are rather caused by the
hot-rolling process and are therefore an intrinsic property of the specimens. Accordingly, the
chemo-mechanical polishing process presented here is indeed able to produce specimens with
a surface quality that is comparable to the result of electrochemical polishing.
Roughness analyses of the samples after the different polishing treatments show that after
removing the large initial surface roughness by grinding with SiC paper up to P4000, the
RMS roughness R
q
is reduced from R
q
= 22 5 nm to R
q
= 17 3 nm by polishing with
diamond suspension. 10 minutes of chemo-mechanical polishing reduces the roughness even
further to R
q
= 10 2 nm. For further chemo-mechanical polishing, the roughness barely
changes anymore and stagnates at about R
q
= 9 2 nm for 20 minutes respectively R
q
= 9 1
nm for 30 minutes. During this final chemo-mechanical polishing treatment, only the internal
microstructure of the surface is refined by removing the distortion layer caused by the
previous polishing steps, as it was already discussed above. Only by electropolishing, even
smoother surfaces could be produced. Here an RMS value of R
q
= 3 1 nm was achieved.
This value may still be overestimated since it is close to the resolution limit of confocal
microscopy and some measurement artefacts were already visible on the image.
Compared to electropolishing, the chemo-mechanical polishing method presented here has the
advantage of being highly reproducible and can be applied to multiple samples at the same
time with a suitable polishing machine. With the equipment used for this study, up to 6
samples could be polished at the same time, which allows for a quick production of a large
number of identical samples. Electropolishing, on the other hand, was found to produce
reliable results only if the samples were embedded in epoxy resin in order to prevent any
contact between the electrolyte and the anode current supply. The embedding of the samples
in resin and particularly the removal of the resin after the electropolishing are tedious
processes. They strongly increase the total time required to prepare one sample, thus making
electropolishing unsuitable producing a large number (i.e., hundreds) of identical samples for
systematic parameter studies. Yet, electropolishing may still be of merit if extremely smooth
surfaces are required.
Summary
It was investigated by scanning electron microscopy on surfaces and cross-sections how
mechanical and chemo-mechanical polishing affects the grain structure of tungsten at the
polished surface, even if the visual appearance of the specimen is already mirror-like. It was
demonstrated that grinding and diamond polishing produce both microstructures at the surface
that are not equivalent to the bulk structure. These preparation artefacts can be removed by 30
minutes of chemo-mechanical polishing using an alkaline colloidal silica suspension.
Although electropolishing can produce even smoother surfaces, the final near-surface
microstructure then closely resembles that of an electropolished surface, where the grains
adjacent to the surface are virtually indistinguishable from bulk grains.
Literature
[1] Roth, J. et al.: J. Nucl. Mater. 390-391 (2009), 1-9
[2] Petzow, G.: Metallographisches Ätzen, 5. Auflage, Gebrüder Borntraeger, Berlin-
Stuttgart, Germany, 1976, S. 8ff
[3] Alimov, V. Kh. et al.: J. Nucl. Mater 420 (2012) 1-3, 519-524
[4] Manhard, A. et al.: J. Nucl. Mater. 415 (2011) 1S, S632-S635
[5] Primig, S. et al.: Prakt. Metallogr. 48 (2011) 7, 345-355
About the Authors
Armin Manhard has been working on his Ph.D. thesis from 2008-2011 and is
now working as a postdoc at the Max-Planck-Institute for Plasma Physics in
Garching. The focus of his research is the retention of deuterium in tungsten
and its correlation with the microstructure and defect density of the tungsten.
Gabriele Matern works as a metallographer at the Max-Planck-Institute for
Plasma Physics in Garching. She performs sample preparation as well as
optical and scanning electron microscopy analysis of material samples and
various components, e.g., from the fusion experiment ASDEX Upgrade.
