J Mater Sci (2011) 46:34053414

DOI 10.1007/s10853-010-5229-2

Microstructural characterization of AISI 431 martensitic stainless

steel laser-deposited coatings
I. Hemmati V. Ocelk J. Th. M. De Hosson

Received: 24 October 2010 / Accepted: 28 December 2010 / Published online: 12 January 2011
The Author(s) 2011. This article is published with open access at Springerlink.com

Abstract High cooling rates during laser cladding

of stainless steels may alter the microstructure and phase
constitution of the claddings and consequently change their
functional properties. In this research, solidification structures and solid state phase transformation products in single and multi layer AISI 431 martensitic stainless steel
coatings deposited by laser cladding at different processing
speeds are investigated by optical microscopy, Scanning
electron microscopy (SEM), energy dispersive spectroscopy (EDS), orientation imaging microscopy (OIM), ternary phase diagram, Schaeffler and TTT diagrams. The
results of this study show how partitionless solidification
and higher solidification rates alter the microstructure and
phase constitution of martensitic stainless steel laser
deposited coatings. In addition, it is shown that while
different cladding speeds have no effect on austenite
martensite orientation relationship in the coatings,
increasing the cladding speed has resulted in a reduction of
hardness in deposited coatings which is in contrast to the
common idea about obtaining higher hardness values at
higher cladding speeds.

Introduction
Austenitic stainless steels are being extensively used for
structural components to be employed in corrosive environments. However, increasing number of applications
requires superior friction and wear behavior in addition to

I. Hemmati
V. Ocelk (&)
J. Th. M. De Hosson

Department of Applied Physics, Materials innovation institute
(M2i), University of Groningen, Nijenborgh 4,
9474 AG Groningen, The Netherlands
e-mail: v.ocelik@rug.nl

corrosion resistance [1, 2]. Laser cladding by powder

injection is an attractive surface engineering method to
improve the wear resistance of austenitic stainless steels by
producing high quality, dense, metallurgically bonded and
thick coatings with minimal heat input into the work piece.
This improvement can be obtained by coating of new parts
or refurbishment of damaged components with materials
having excellent combination of corrosion resistance and
hardness such as martensitic stainless steels.
As the heat input during laser cladding is very localized,
cooling rates during solidification and subsequent cooling
in solid state are usually very high. It has been shown
before that high cooling rates during solidification have
altered the microstructures in stainless steel weld deposits
[3, 4]. In other words, rapid solidification of stainless steels
may produce microstructures different from those observed
during conventional solidification with differences in mode
of solidification, scale of microstructural features and solute segregation patterns. These discrepancies can strongly
affect the final mechanical, chemical and technological
properties of the coatings [3, 4].
In the case of laser-deposited coatings, it is generally
believed that faster cooling rates obtained at higher cladding speeds produce finer microstructures with increased
hardness values [5]. While this idea is well known for alloy
systems including precipitation hardening, it is not studied
for alloys in which the main strengthening mechanism is
martensitic transformation. In this investigation, single and
multi-layer coatings of AISI 431 martensitic stainless steel
was deposited on AISI 304 austenitic stainless steel by
laser cladding at various speeds. The objective of this study
is to explain in detail the microstructural features and phase
constitutions resulting from rapid solidification and the
possible changes in the orientation relationship between
solid state transformation products during subsequent solid

123

3406

J Mater Sci (2011) 46:34053414

state cooling as the starting point to understand and control

the functional properties of this type of coating. In addition,
the implications of microstructural changes on hardness of
clad layers are discussed.

Experimental procedures
Commercial 42C powder from Sulzer Metco (equivalent to
AISI 431) with particle size in the range of -140 to ?325
mesh was deposited on AISI 304 rods with 40 mm diameter using a continuous wave IPG fiber laser with wavelength of 1.07 lm and variable power up to 3.3 kW. The
laser beam was defocused to obtain a laser spot size of
3.3 mm on the surface in all experiments. The overlapping
between subsequent tracks was 30% and thickness of single
and five layer deposits were almost 1 and 4 mm, respectively. Chemical compositions of the cladding and substrate materials are presented in Table 1. Powder feeding
system consisted of Metco Twin 10C powder feeder, argon
as carrier and shielding gas and ALOtec Dresden GmbH
Cu-based side cladding nozzle with cyclone and 2-mm
nozzle opening. Single layer and five layer samples
were deposited at scanning speeds (V) of 5, 25, 58, and
117 mm/s. In multi-layer deposits, the time between
deposition of subsequent layers was 5 s. Solidification
structure was studied by optical microscopy on samples
prepared by standard mechanical grinding/polishing and
etched with Kalling 1 reagent for 510 s.
Solid state transformation structures were studied by
optical microscopy and Philips XL30 FEG scanning electron microscope (SEM) equipped with orientation imaging
microscopy (OIM). As the distorted lattice of untempered
martensite makes it difficult to get good quality electron
back scatter diffraction (EBSD) patterns, using a suitable
sample preparation method which produces minimum
surface relief was essential. In this research, surface preparation procedure recommended by Struers Co. for EBSD
analysis of ferrous alloys was employed [6]. This procedure consists of mechanical grinding with SiC-paper
number 320 and 9 lm diamond suspension, polishing with
3, 1, and 0.25 lm diamond suspensions and final polishing
with OP-AA suspension. Using this preparation method,
percentage of indexed data was higher than 95% in all of
the OIM measurements. EBSD data were obtained by TSL

Table 1 Nominal composition of the coating and substrate materials

(wt%) [8]
Alloy

Mn

Si

Cr

Ni

AISI 431

0.2

1517

1.252.5

AISI 304

0.08

1820

123

810.5

OIM Data Collection 5 and analyzed by TSL OIM Analysis

5.31 software. Previous research [7] showed that choosing
other crystal symmetries than bcc for indexing martensite
EBSD patterns will produce poor results and consequently,
gamma (c) and alpha (a) Iron were included as the constituent phases in all of the OIM analyses in this research.
Energy dispersive spectroscopy (EDS) elemental mapping was employed to analyze the distribution of alloying
elements and their possible partitioning. EDS maps of Cr,
Ni, Si, and Mn were obtained at 20 kV and map of C was
collected at 4 kV accelerating voltage. Ternary phase diagram, Schaeffler and TTT diagrams were employed to
analyze the experimental findings and determine the effects
of rapid solidification on deviations from equilibrium state.

Results and discussion

Solidification structures
Figure 1 shows the three distinct microstructures which
could be identified in the deposits. Observation of interfacial regions (Fig. 1a) showed that a thin layer of planar
growth was present at all interfaces with an extension of
around 20 lm. This plane growth zone became unstable as
solidification proceeded and broke down into cellular
dendrites within short distances from the interface. As a
result, cellular dendrites formed the majority of solidification microstructures. In multi layer deposits, a columnar
to equiaxed transition (CET) occurred at the top of deposits
resulting in a thin layer of equiaxed grains (Fig. 1b).
To explain the observed microstructural changes, it is
necessary to know the parameters which control the
solidification structure. These parameters are mainly solid
liquid interface growth rate R and temperature gradient in
the melt G. The G/R ratio determines the solidification
mode and the G
R product controls scale of the solidified
structure [9]. CET takes place when nucleation of equiaxed
grains occurs in the liquid ahead of the columnar structure.
In laser cladding, this will not happen unless the ratio of
G/R is lower than some critical value [10]. As can be seen
in Fig. 2, by lowering the G/R ratio, solidification structure
changes from planar to cellular and then columnar and
equiaxed dendrites. In single layer coatings, only the
transition from planar to cellular and columnar dendrites
happened but in five layer deposits, all of the transitions
were visible. These changes imply that during solidification, G/R ratio was not constant. Previous research [4]
shows that in laser solidification treatments the G/R is
infinite at the bottom of the melt pool which translates into
a planar growth layer but rapidly decreases and as a result,
structure of laser claddings mostly consists of columnar
growth of dendrites [4].

J Mater Sci (2011) 46:34053414

3407

Fig. 1 Solidification structures of the coatings revealed by optical microscopy: a planar growth at substrate-coating interface, b cellular dendritic
structure with columnar to equiaxed transition at top of the deposit ((a) transversal and (b) longitudinal cross sections)

Fig. 2 Effect of G and R on morphology and scale of solidification

structures [9]

Recent studies [11, 12] confirmed that in multi layer

laser claddings with constant power, temperature, and size
of the melt pool increase from each layer to the next
because of heat accumulation. This was also observed in
this study in which size of the melt pool in multi layer
deposits (measured by optical microscopy after sectioning)
was almost doubled from the first to the fifth layer. In fact,
this heat accumulation reduces the high temperature gradients typically associated with columnar microstructures
and provides the opportunity for a transition from columnar
to equiaxed structure [11].
Increasing the cladding speed (i.e., increasing the
solidification rate) did not affect the microstructural features such as epitaxial growth across subsequent layers or
transitions in the solidification mode but scale of the dendritic structures was significantly refined as is obvious by
the reduction of cell spacing at higher cladding speeds as
shown in Fig. 3. Higher cladding speeds produced smaller
cell spacing with values down to around 9 lm in five-layer

Fig. 3 Effect of cladding speed on cell spacing of single and multi

layer deposits. Dashed lines show the fits for V-0.5 dependence

and 3 lm in single layer deposits. This can be explained by

the fact that growth rate is a direct function of cladding
speed and higher growth rates are obtained at higher
deposition velocities. A variation of 15% is included in the
data of Fig. 3 to account for changes in cell spacing values
from bottom to top of the coating because of local variations in solidification conditions [13]. The difference in the
cell spacing values of single and multi-layer deposits in
spite of their similar cladding speeds can be attributed to
the heat accumulation in multi-layer claddings as explained
above but in both cases, cell spacing followed the empirical
power law of cell spacing V-0.5 reported previously [14].
Solidification phases
The connection between chemical composition of stainless
steel laser depositions and their equilibrium phase constitution can be obtained by equilibrium diagrams such as

123

3408

Schaeffler and ternary phase diagrams. Figure 4 shows the

composition of AISI 431 on Schaeffler diagram [15] and on a
section of ternary FeCrNi phase diagram at the latest stage
of solidification (1,400 C) calculated by 3D Demo system
for Ternary Phase Diagrams from Japan Science and Technology Corporation. A dilution of around 10% from substrate was taken into account in calculating the position of
AISI 431 on the above-mentioned diagrams but because of
big differences in the amounts of C, Cr and Ni between
substrate and coating, Creq. and Nieq. values with and without
dilution were approximately the same and this allowed
neglecting dilution from substrate and having the same
position on diagrams for both single and multi layer deposits.
Based on these diagrams, equilibrium solidification of
AISI 431 is a mixed austenitic-ferritic solidification which
finally results in around 10% delta ferrite (d) along with
martensite and retained austenite at room temperature.
While these diagrams are reliable for predicting structures
at cooling rates close to equilibrium, numerous previous
researches [3, 1619] showed that solidification and solid
state cooling rates have a significant effect on the final

J Mater Sci (2011) 46:34053414

microstructures, thus making it unreliable and even

impossible to predict the phase constitution of rapidly
cooled stainless steels from the conventional diagrams. This
unreliability was shown in the current research in terms of
the amount and distribution of d in claddings deposited at
different speeds. Figure 5 shows the optical micrographs
of 5-layer samples deposited with cladding speeds of 5
(Fig. 5a) and 117 mm/s (Fig. 5b). It can be seen, especially
at track overlapping areas, that by increasing the cladding
speed less d (visible as white islands) was formed. In
addition, distribution of d is not homogenous and as shown
in Fig. 6, while little amounts of d are present at interdendritic areas; huge quantities are formed at the beginning of
each solidification step, i.e., at substrate-coating interface or
in remelted parts of the previous track. A point which
should be emphasized here is that high cooling rates after
solidification tends to suppress the solid state transformation of d to c. Consequently, concentration of d in some
areas should be a result of solidification step.
Studying the solute redistribution in the overlapping
areas (which contain considerable quantities of both c and

Fig. 4 a Schaeffler diagram

[15] and b section of ternary
FeCrNi phase diagram.
Composition of AISI 431 is
shown with a dot in each
diagram

Fig. 5 Reduction in the amount of d with increasing cladding speed: a 5 and b 117 mm/s. d phase appears as bright islands mostly at track
interfaces

123

J Mater Sci (2011) 46:34053414

3409

Fig. 6 Distribution of d (bright islands shown by arrows): a at substrate-cladding interface and b at track overlapping areas

Fig. 7 SEM micrograph of a dual phase area in AISI 431 laser deposit and the EDS spectral maps showing the distribution of Ni, Cr, and C
in that area

d) by EDS spectral mapping showed that even in the lowest

cladding speed, there was no detectable partitioning of
alloying elements between these two phases. Figure 7
shows the EDS maps for Cr, Ni (obtained at 20 kV), and C
(obtained at 4 kV) as the main alloying elements in AISI
431 stainless steel. Uniformity of chemical composition in
dual phase areas suggests that solidification of these coatings occurred in a partitionless manner probably because of
the high cooling rate and its associated undercooling during
the laser deposition process.

Taking the high cooling rate and partitionless solidification into account, the liquidus line in phase diagram will
not be the start of solidification anymore and instead, To
line should be considered as the beginning of solidification
as shown in Fig. 8. To line is the loci of To temperatures at
which the new phase can appear with a net decrease in free
energy at the same composition as the parent phase. In fact,
To line defines the boundary below which diffusionless
transformation is possible. Below To, free energy of the
single phase c or d is lower than the free energy of

123

3410

J Mater Sci (2011) 46:34053414

Figure 6 also shows that after a short time, c outgrew d

and became the primary solidification phase. This can be
attributed to the faster growth kinetics of c in comparison
to d which was observed in several previous research works
on welding of stainless steels [3, 16, 20]. It seems that
while partitionless solidification provides the opportunity
for both c and d to nucleate at initial stages of solidification, more favorable growth kinetics of c (especially at
higher solidification rates) dominates the rest of solidification resulting in a mostly austenitic structure with isolated block of d at the interfaces and occasional stringers at
interdendritic areas.
Characteristics of the solid state transformation product
Fig. 8 Position of AISI 431 composition in ternary phase diagram
during partitionless solidification [3, 16]

undercooled liquid and partitionless solidification will be

theoretically possible [3]. To curve is located between the
liquidus and solidus lines of equilibrium phase diagram
(Fig. 8) in which the data from references [3] and [18] are
combined to show the composition of AISI 431 in the
phase diagram with To lines. In this case, solidification
structure depends on whether the To line (or its extension)
of c, d or both of them is crossed by the undercooled melt
and relative quantity of c and d is governed by the difference in To temperatures of the two phases [16]. As can be
seen in Fig. 8, To temperatures of c and d for AISI 431
composition are very close and this means that if the
undercooled liquid crosses the To lines, both phases will
form in similar quantities. This idea is confirmed by
micrographs of Fig. 6 which show that at the beginning of
each solidification stage, c and d phases were formed in
similar proportions. It can be concluded that the undercooling achieved during rapid solidification was sufficient to
bring the liquid metal below both To lines making it possible for both phase to nucleate.

It is expected that by changing the cladding speed, not only

the solidification rate changes, but also cooling rate in solid
state is affected and this can have implications on solid
state transformation products. According to TTT diagram
of the AISI 431 stainless steel [21], this type of steel has a
very high hardenability and to form any transformation
product other than martensite, such slow cooling rates are
needed that take 103104 s to reach around 600 C from
austenitizing temperature. Under our processing conditions, having these slow cooling rates is virtually impossible and only the transformation of austenite to martensite
can occur. Optical microscopy and SEM microstructural
investigation of the samples deposited at different cladding
speeds revealed that this was the general case in which
typical lath martensite was the dominant microstructural
feature. Figure 9 shows examples of martensitic microstructure of the coatings.
Orientation imaging microscopy (OIM) was employed
to analyze the characteristics of solid state transformation
products with special interest in austenite/martensite orientation relationship (OR) and characteristics of the austenite/martensite interphase boundaries. Figure 10 shows
examples of grain map, phase map and [001] inverse pole

Fig. 9 Lath martensitic microstructure of the deposits observed by (a) optical microscope and (b) SEM

123

J Mater Sci (2011) 46:34053414

figure (IPF) map of the studied coatings which are superimposed on image quality (IQ) map (The IQ parameter
describes the quality of an electron backscatter diffraction
pattern [22]). In these figures, martensite lathes and traces
of retained austenite are clearly visible. The data of the
images presented in Fig. 10 are used in the subsequent
calculations of OR and boundary characteristics.
As the formation of martensite involves coordinated
movement of atoms, parent austenite and product martensite always have some form of orientation relationship
(OR). Over the years, several models such as Kurdjumov
Sachs (KS) [23], NishiyamamWassermann (NW)
[24, 25], and GreningerTroiano (GT) [26] have been
proposed to describe the fcc/bcc OR formed during martensitic transformation. The basic idea in all of these
models is that closed-packed planes of the two phases

Fig. 10 Examples of OIM scan results obtained from the studied

coatings a grain, b phase (martensite: green, austenite: orange), and
c [001] inverse pole figure (IPF) maps superimposed on image quality
(IQ) map (Color figure online)

3411

(i.e., {111} for austenite and {110} for ferrite) are parallel
or nearly parallel. By confirming the validity of the concept
of parallel close-packed planes, it will be reliable to use
these models to study the OR in our coatings. To explore
this task, pole figure (PF) texture plots for close-packed
planes of austenite and martensite were obtained and as
shown in Fig. 11, it is obvious that the criterion of parallel
close-packed planes is valid in our case because the projection of all {111} plane poles for austenite coincides with
the projection of some {110} planes for martensite.
Having this fundamental concept confirmed, discrete PF
plots for different planes of austenite and martensite were
obtained and their combined PF plot was compared with
the theoretical PFs for each OR model. The combined PF
plot of the two phases shows close resemblance to the PF
of GT model as shown in Fig. 12.
Usually different austenitemartensite ORs are associated with different values of misorientation across the
boundaries between these two phases [27]. Consequently
the misorientation across the austenitemartensite boundaries was also studies by OIM to check the existence of
GT orientation relationship. Figure 13 shows a close-up
of an area of the images in Fig. 10 containing both martensite and retained austenite.
This section was selected in such a way that it contained
only one originally austenite grain (evident from the
inverse pole figure map in Fig. 13b) to avoid considering
the cc interfaces in the boundary measurements. By
measuring the misorientation angle across interphase
boundary, it was known that the two phases are separated
by high angle boundaries with misorientation of around
45. More detailed analysis of the misorientation values
showed that there was in fact a range of boundary misorientations from 42 to 46 degrees as highlighted in
Fig. 13d. Highlighting the 4246 boundaries in the whole
scan area revealed that this is the general case and almost
all of the interphase boundaries are high-angle boundaries

Fig. 11 PF texture plots for

close-packed planes of c and a.
Identical location all {111}
plane poles for c and some
{110} planes for a confirms that
theses two sets of planes are
parallel

123

3412

J Mater Sci (2011) 46:34053414

Fig. 12 (100) PF plots obtained

for a c ? a (c poles are shown
with red dots) and b fcc/bcc OR
in G-T model. Similarity of
a and b shows the existence
of GT orientation relationship
between c and a in the studied
coating (Color figure online)

Fig. 13 a section of the phase

map containing martensite and
only one austenite grain as
shown by the single orientation
(single color) grain of austenite
in [001] IPF of
b. c misorientation angle across
the line drawn in a. d 4246
boundaries are highlighted

with misorientations in the range of 4246 (Fig. 14).

Average misorientation angle between austenite and martensite was 44.27 which is very close to 44.3 misorientation calculated for GT orientation relationship from
Eulerian angles typical of this OR [27]. By repeating the
above-mentioned analyses for single and multi layer coatings deposited at the lowest (5 mm/s) and the highest

123

(117 mm/s) cladding speeds, it became obvious that different cooling rate associated with various cladding speed
had no noticeable effect on boundary characteristics or OR
in the final structure of the coatings.
On the other hand, EBSD results showed that while in
multi layer deposits the amount of retained austenite was
around 4% for all cladding speeds, in single layer coatings

J Mater Sci (2011) 46:34053414

3413

Fig. 14 4246 boundaries

highlighted in the whole scan
area showing that
misorientations of nearly all
interphase boundaries are in this
range (Color figure online)

Fig. 15 Effect of cladding

speed on hardness values in
a multi layer and b single layer
deposits

more austenite (up to 11%) was retained at the higher cladding speed. This results in a variation of hardness values of
single layer deposits produced at different scanning speeds as
shown in Fig. 15a but in multi layer coatings, hardness values
were the same in all cases (Fig. 15b). The observed differences can be attributed to the austenite stabilizing effect
of finer dendrites produced at higher cladding speeds [13].

Conclusions
AISI 431 martensitic stainless steel was deposited by fiber
laser at four different cladding speeds up to 117 mm/s.
Solidification structure of the deposits mostly consisted of
dendritic structure, but CET was observed at the top of
multi layer claddings. Higher cladding speeds resulted in
finer dendritic structures with cell spacing values down to
around 9 lm in multi-layer and 3 lm in single layer
deposits. Formation of d phase which was distributed heterogeneously with higher concentration at track boundaries
was suppressed by higher cladding speeds as a result of
faster growth kinetics of c. In addition, homogenous distribution of alloying elements across dual phase areas
consisting of former c plus d showed that solidification of
these deposits occurred in a partitionless manner. Finally,
cooling in the solid phase produced lath martensitic
structure having 4246o high angle boundaries and GT
orientation relationship with retained austenite. Morphology, orientation relationship and boundary misorientation
angle of martensite produced in samples with different
cladding speeds did not show big differences, but the
amount of retained austenite was increased by increasing

the cladding speed with a maximum of around 11 percent

in single layer coating deposited at 117 mm/s. This austenite stabilizing effect of high speed cladding reduced the
hardness values of the samples with the most refined
microstructures. This finding is in contrast to the behavior
reported for precipitation-hardening alloy systems.
Acknowledgements The work is part of the research program of
M2i (Materials innovation institute), The Netherlands, Project number: MC7.06259. M. Dutka is acknowledged for his assistance in laser
experiments.
Open Access This article is distributed under the terms of the
Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any
medium, provided the original author(s) and source are credited.

References
1. Dearnley PA, Smith GA (2004) Wear 256:491
2. Blawert C, Weisheit A, Mordike BL, Knoop RM (1996) Surf
Coat Technol 85:15
3. David SA, Vitek JM, Hebble TL (1987) Weld J 66:289
4. Kurz W, Trivedi R (1994) Mater Sci Eng A 179180:46
5. Kathuria YP (2000) Surf Coat Technol 132:262
6. Ellemann R, Guesnier A Preparation of ferrous metals for electron backscatter diffraction (EBSD) analysis. http://www.struers.
com/resources/elements/12/108620/Application_Note_EBSD_
English.pdf. Accessed Sept 2010
7. Ryde L (2006) Mater Sci Technol 22:1297
8. Metals handbook, vol 1, 10 edn, 4th printing. ASM International,
Materials Park, p 843
9. Kou S (2003) Welding metallurgy, 2nd edn. Wiley, New Jersey
10. Kurz W, Bezencon C, Gaumann M (2001) Sci Technol Adv
Mater 2:185

123

3414
11. Bontha S, Klingbeil N, Kobryn P, Fraser H (2009) Mater Sci Eng
A 513514:311
12. Hofman J (2009) Development of an observation and control
system for industrial laser cladding. PhD thesis, University of
Twente, The Netherlands, pp 121126
13. Colaco R, Vilar R (2004) Mater Sci Eng A 385:123
14. Kurz W, Fisher DJ (1992) Fundamentals of solidification, 3rd
edn. Trans Tech Publications, Switzerland
15. Metals handbook, vol 6, 10 edn, 4th printing. ASM International,
Materials Park, p 678
16. Elmer JW, Allen SM, Eagar TW (1989) Metal Trans A
20:2117
17. Zambon A, Bonollo F (1994) Mater Sci Eng A 178:203

123

J Mater Sci (2011) 46:34053414

18. Vitek JM, David S (1994) In: Mazumder J, Mukerjee K, Mordike
BL (eds) Laser mater process IV. AIME, Warrendale, pp 153167
19. Fukumoto S, Kurz W (1998) ISIJ Int 38:71
20. Kelly TF, Cohen M, VanderSande JB (1989) Metal Trans A
15:819
21. (1995) Heat treaters guide: practices and procedures for irons
and steels, 2nd edn. ASM International, Materials Park
22. Wright SI, Nowel MM (2006) Microsc Microanal 12:72
23. Kurdjumow G, Sachs G (1930) Z Phys 64:325
24. Nishiyama Z (1934) Sci Rep Tohoku Imp Univ Tokyo 26:637
25. Wassermann G (1935) Mitt K-Wilh-Inst Eisenforsch 17:149
26. Greninger AB, Troiano AR (1949) J Met Trans 185
27. Nolze G (2006) Cryst Res Technol 41:72