ON THE PRECIPITATION BEHAVIOUR IN MARAGING
STEELS
B. Jamnig and R. Ebner
Materials Center Leoben (MCL)
Franz-Josef Straße 13
A-8700 Leoben
Austria
P. Staron and H. Clemens
GKSS Research Center, Institute for Materials Research
Max-Planck-Straße
D-21502 Geesthacht
Germany
H. Leitner and F. Jeglitsch
Department of Physical Metallurgy and Materials Testing
University of Leoben
Franz-Josef Straße 18
A-8700 Leoben
Austria
P. Warbichler
Research Institute for Electron Microscopy, Center for Electron Microscopy
University of Graz
Steyrergasse 17
A-8010 Graz
Austria
549
550
6TH INTERNATIONAL TOOLING CONFERENCE
Abstract
The evolution of precipitates in a cobalt-free chromium containing maraging
steel was studied using the complementary techniques of small angle neu-
tron scattering (SANS) and energy-filtering transmission electron microscopy
(EFTEM). EFTEM measurements were conducted to determine the shape of
the precipitates and to gain information on the chemical composition which
are two important input parameters for analyzing the SANS data. With SANS
the size, distribution, particle number density and volume fraction of the pre-
cipitates were determined. EFTEM measurements exhibited homogeneously
distributed precipitates and their size correlate well with SANS results. Also
the chemical composition investigated by EFTEM and verified by the liter-
ature are in good agreement with the chemical composition calculated from
the ratio of magnetic to nuclear scattering intensities obtained by SANS.
Keywords:
maraging steel, SANS, EFTEM, precipitation behaviour
INTRODUCTION
Maraging steels are widely used in many technological sectors such as
aerospace, military as well as for tools and dies. For most applications
these steels require high strength in combination with good toughness. Also
good weldability, high strength to weight ratio and dimensional stability
during aging are attractive features for applications. Occasionally corrosion
resistance is needed.
The almost carbon-free maraging steels transform to a soft nickel marten-
site after cooling from solution treatment. In this condition they can easily
be machined and if desired cold worked. Subsequent ageing at temperatures
between 723 and 823 K leads to the strengthening which is caused by the
formation of intermetallic phases. A number of studies have been carried out
on the precipitation behaviour, applying a variety of techniques. These in-
clude direct methods like transmission electron microscopy (TEM) [1] with
electron energy-loss spectroscopy (EELS), atom probe field ion microscopy
(APFIM) [2] and indirect methods like small angle X-ray scattering (SAXS)
[3], small angle neutron scattering (SANS) [4] and Mössbauer spectroscopy
[5]. The reason of this variety of examination methods is the difficulty to
make statements on chemistry and structure of the mostly nanometer fine
precipitates and their quantitative properties. Yet there are still no complete
elucidations about intermetallic phases.
In the present study SANS measurements were carried out on a cobalt-free
corrosion resistant maraging steel which was aged at various temperatures
On the Precipitation Behaviour in Maraging Steels
551
and times. Small-angle scattering (SAS) for the study of maraging steels
was first used by Servant et al [3, 4]. SANS is one of the most powerful
methods for probing three-dimensional chemical heterogeneities and density
fluctuations, the size of which may vary between 1 and 100 nm. Employing
SANS, size, size distribution, particle number density and volume fraction
of the precipitates can be determined. In addition, it is possible to study
the nature of the interface between the inhomogeneities and the surrounding
matrix [6]. The advantage of SANS is that these quantitative values can be
measured within a rather large sample volume which is hardly possible with
other methods. Neutrons interact not only with atomic nuclei but also with
the magnetic field of the electrons. This gives the chance of checking an
assumed chemical composition of non-magnetic precipitates embedded in a
ferromagnetic matrix.
The metallurgical interpretation of the SANS results is discussed making
reference to results obtained by energy filtered TEM (EFTEM) and previous
investigations reported in literature.
EXPERIMENTAL
A model alloy based on the maraging steel studied by Gemperle et al. [7]
was used. The chemical composition is given in Table 1. All samples were
annealed for 1 hour at 1273 K and then air cooled. Subsequently, different
age hardening treatments were conducted as summarized in Table 2.
Sample 1 was only solution annealed at 1273 K and air quenched without a
subsequent age hardening treatment. In the following, this sample is referred
to as reference sample. Samples 2, 3 and 4 were aged at 748 K for 0.25 h,
12 h and 100 h, respectively, to study the precipitation kinetics. The heat
treatment of sample 3 was chosen in a way that peak hardness is achieved.
Sample 5 was slightly overaged at 798K for 16 h.
Table 1.
Chemical composition of the investigated maraging steel
C
Si
Mn
Cr
Ni
Mo
Ti
Al
Fe
wt%
0.01
0.55
0.11
12.3
8.89
0.98
0.83
0.57
75.8
552
6TH INTERNATIONAL TOOLING CONFERENCE
Table 2.
Summary of age hardening treatments; prior to these heat treatments all samples
were solution annealed at 1273 K and then quenched in air. Sample 1 is referred to as
as-quenched reference sample
Sample
1
2
3
4
5
Heat
Treatment
as-quenched
0.25 h/748 K
12 h/748 K
100 h/748 K
16 h/798 K
EFTEM
EFTEM was used to gain information on composition of the matrix and
shape of the precipitates. A comprehensive description of the EFTEM tech-
nique can be found in [8, 9]. The instrument used was a Philips CM20/STEM
equipped with a Gatan imaging filter (GIF). The preparation of the TEM
specimen (∅ 3 mm) are described elsewhere [10].
SANS
The experimental technique is comprehensively described in [11]. The
used cylindrical samples have a circular cross section with a diameter of
25 mm and a thickness of 1 mm. Measurements were carried out with the
instrument SANS-2 at the Geesthacht Neutron Facility (GeNF). Selector-
monochromated neutrons with a mean wavelength of
λ = 0.57 nm and a
wavelength spread of
∆λ/λ = 10% were used. The neutron beam had a
diameter of 8 mm. The samples were magnetized to saturation in a field
of 2 T. The measured intensities were corrected for sample transmission,
background intensity as well as detector response. The macroscopic differ-
ential scattering cross sections have been obtained by calibration against a
Vanadium standard.
The scattering curves of the samples aged at 748 K for 0.25 h and 12 h
have shown an interference maximum which obviously is caused by a high
particle number density. In order to treat this interparticle interference the
so-called local monodisperse approximation as described by Pedersen was
adopted [12]. In this approximation, the SANS scattering cross section of a
dispersion of precipitate particles can be described in the framework of the
On the Precipitation Behaviour in Maraging Steels
553
two-phase model according to:
dΣ
dΩ
(q) = (∆η)
2
Z
∞
0
n(R)V (R)
2
F (q, R)
2
S(q, R
HS
, f
HS
) dR
(1)
where
dΣ/dΩ is the macroscopic differential scattering cross section, ∆η
is the difference in the scattering length densities of particle and matrix,
n(R) dR is the number density of particles with sizes between R and R+dR
and
V (R) stands for the particle volume. q is the scattering vector, q =
4π sin(θ)/λ, where 2θ is the scattering angle. The majority of precipitates
were expected to have a spherical shape, therefore the particle form factor
F (q, R) for spherical particles [13] was used. S(q, R
HS
, f
HS
) is the structure
factor of the monodisperse hard-sphere model describing the interparticle
interference effect.
R
HS
is the hard-sphere radius of a precipitated particle
that includes a depleted zone around the particle.
R
HS
is given by
R
HS
=
C
HS
R, where C
HS
is a constant.
f
HS
is the volume fraction of hard spheres
and is related to the actual particle volume fraction by
f
HS
= f C
3
HS
.
A lognormal distribution was used for the particle size distribution
n(R).
The three free parameters of this distribution describe the position of the
maximum (
R
0
), the width of the distribution (
β) and the number density of
particles contained in the distribution (
n
0
). These parameters, together with
the additional parameter
C
HS
, are determined by fitting a scattering curve
calculated by equation (1) to a measured scattering curve by means of a
least-squares procedure.
In case of magnetic scattering there is a magnetic scattering contrast ,
∆η
mag
in addition to the nuclear one,
∆η
nuc
. The magnetic scattering
cross section depends on the angle
α between the scattering vector and
the magnetic field. In the maraging steel under investigation the precipi-
tates are assumed to be non-magnetic and thus can be considered as "mag-
netic holes" in a ferromagnetic matrix. When the matrix is magnetized
to saturation and the chemical size of the precipitates is assumed to be
the same as their "magnetic" size,
∆η in equation (1) can be replaced by
∆η = (∆η
nuc
)
2
+ (∆η
mag
)
2
sin
2
α [14]. Thus, the nuclear cross section
can be measured at
α = 0, while the sum of nuclear and magnetic cross
section is measured at
α = π/2 . Then the magnetic cross section can be
calculated by subtracting
dΣ/dΩ (q, α = 0) from dΣ/dΩ(q, α = π/2).
The magnetic scattering length density of the matrix was calculated as
η
mag
,
m
= 3.96 · 10
10
cm
−
2
, which is identical with the difference in scatter-
554
6TH INTERNATIONAL TOOLING CONFERENCE
ing length density between matrix and particle,
∆η
mag
, since the particles
are considered to be non-magnetic, i.e.
η
mag
,
p
= 0. The ratio A of magnetic
to nuclear scattering intensity
A =
dΣ
dΩ
d,
π
2
dΣ
dΩ
(q, 0)
= 1 +
(∆η
mag
)
2
(∆η
nuc
)
2
(2)
depends on the chemical composition of the precipitates [15]. In the case
of non-magnetic precipitates, equation (2) can be used for estimating the
nuclear scattering length density of the precipitates from measured scattering
curves. Therefore, an assumed chemical composition of the precipitates can
be tested by comparing experimental nuclear scattering length densities to
calculated nuclear scattering length densities. The nominal composition
of the alloy was used for the matrix, neglecting possible deviations of the
matrix composition by the formation of precipitates. The precipitates were
assumed to be coherent and the same mean volume per atom as in the matrix
was used.
RESULTS
EFTEM
Figure 1 shows various EFTEM mapping images of sample 5 which was
aged at 798 K for 16h. In the bright field TEM image (Fig. 1(a)), the
precipitates can hardly be separated from the matrix, however, in the mapping
images they can be seen (Fig. 1(b) – 1(f)). Here, dark areas indicate an
depletion of the scanned element whereas bright areas show an enrichment.
The mapping images shown in Fig. 1 reveal that the precipitates are depleted
in iron and chromium and enriched in nickel, titanium and silicon.
Figure 2 shows the evolution of the precipitates with ageing time at 748
K. After aging for 0.25 h the precipitates are extremely fine and can not be
resolved with EFTEM (Fig. 2(a)). Prolongation of aging to 12 h (Fig. 2(b))
has the effect that the precipitates can clearly be seen. Figure 2(b) indicates
that the precipitates are spherical and uniformly distributed. Aging for 100 h
at 748 K leads to a slight coarsening of the spherical precipitates. In addition,
formation of rod-like precipitates is observed (Fig. 2(c)).
From the iron mapping images shown in Fig. 2(a) and 2(b) an attempt
was made to estimate the size of the spherical precipitates. Due to the
On the Precipitation Behaviour in Maraging Steels
555
Figure 1.
EFTEM analysis conducted on sample 5 (ageing treatment: 16 h at 748 K). (a)
bright field image. Mapping images: (b) Fe; (c) Ti; (d) Cr; (e) Ni; (f) Si. The arrow in (a)
points at a grain boundary (GB).
weak contrast an exact evaluation was difficult. Taking into account all
uncertainties diameters ranging from 2 to 4 nm were estimated for both
conditions. The size of the rod like particles (Fig. 2(c)) was estimated to be
30 nm in length and 6 nm in diameter.
Figure 2.
Iron mapping images (EFTEM). Heat treatment: all samples solution annealed
at 1273 K for 1 h, air cooled and subsequently aged at 748 K for (a) 0.25 h, (b) 12 h, and (c)
100 h.
556
6TH INTERNATIONAL TOOLING CONFERENCE
SANS
Figures 3 and 4 show the scattering curves of samples 1 to 5. It can be seen
that the curves of the reference sample 1 are flat at large scattering vectors
q.
This indicates that a good homogenization was achieved by annealing at 1273
K for 1 h. The flat cross section of sample 1 at large
q results from incoherent
scattering and other diffuse scattering contributions. This constant value is
added to the calculated cross sections, assuming that it does not change
significantly during ageing. After ageing an increase in scattering intensity
at large q can be observed due to the formation of precipitates. In order to
calculate the particle number density (
n
0
) the magnetic cross sections were
fitted using equation (1). From Fig. 3 it can be seen that calculated curves
fit well with the measured data.
Figure 3.
Magnetic cross sections as obtained by SANS. Symbols: measured data; solid
lines: calculated curves according to equation (1).
On the Precipitation Behaviour in Maraging Steels
557
Because the precipitates are assumed to be non-magnetic they can be
treated as "magnetic holes" in a magnetic matrix. Therefore, the volume
fraction of the precipitates can be obtained from the magnetic cross section.
For the evaluation the magnetic scattering length density of the matrix must
be known. Knowledge of the chemical composition of the precipitates is not
needed. The resulting particle size distributions are shown in Fig. 5. The
volume fraction of particles contained in the particle size distribution
n(R)
is given by
f =
Z
∞
0
n(R)V (R) dR =
Z
∞
0
f (R) dR
(3)
The fitting parameter
n
0
of the particle size distribution
n(R), the vol-
ume fraction f and the parameters of the distribution of the particle volume
fractions (Fig. 5) are summarized in Table 3.
Table 3.
Evaluation of the SANS data according to equations (1) and (3);
n
0
is the particle
number density contained in the particle size distribution,
f is the particle volume fraction;
R
max
is the position of the maximum and FWHM is the full width at half maximum. Numbers
in brackets give the error in the last digits as calculated from the statistical error of the data.
Sample
R
max
[nm]
FWHM [nm]
n
0
[cm
−
3
]
f
0.25h/748 K
0.69
0.38
4.3(3) · 10
19
0.054(4)
12h/748 K
1.35
0.71
7.0(1) · 10
18
0.066(1)
100h/748 K
2.05
1.65
2.52(4)·10
18
0.078(1)
16h/798 K
2.45
2.59
1.3(1) · 10
18
0.065(4)
At 748 K, after the shortest aging time of 0.25 h, the particle volume
fraction is 5.4% and increases up to 7.8% after 100 h. The particle radius
increases from 0.7 nm up to 2.05 nm (position of the maximum) while the
size distribution becomes significantly broader (Fig. 5). The particle number
density decreases from
4.3 · 10
19
cm
−
3
to
2.5 · 10
18
cm
−
3
, indicating that
coarsening has taken place. At 798 K the reaction is faster and after 16 h
the precipitates have grown to a radius of 2.45 nm with a volume fraction of
6.5% and a number density of
1.3 · 10
18
cm
−
3
.
The ratio
A (equation 2) is determined by multiplying the calculated
magnetic cross section by
1/A such that it fits the nuclear cross section .
From Fig. 4 it can be seen that the resulting curves for the nuclear cross
558
6TH INTERNATIONAL TOOLING CONFERENCE
Figure 4.
Nuclear cross section as obtained by SANS. The solid lines are obtained by
multiplying the fit of the corresponding magnetic cross section by 1/A.
Figure 5.
Particle size distributions obtained from evaluation of the SANS curves.
On the Precipitation Behaviour in Maraging Steels
559
Table 4.
Measured and calculated A-ratio of some intermetallic precipitates.
Measured
(this study)
Ni
3
Ti
Ni
2
Ti
NiTi
NiAl
Ni
4
Mo
Ni
3
Mo
Ti
6
Si
7
Ni
16
A
4.6. . . 6.4
9.2
3.8
1.8
9.3
18.8
26.4
4.4
section do not fit the data equally well as in case of the magnetic cross
section.
One reason for this deviation could be that the chemical size of the precip-
itates might be slightly smaller than the magnetic hole in the matrix because
the magnetic field does not drop sharply to zero at the chemical precipitate
boundary. This effect might be relatively large for very small precipitates.
However, the resulting values of
A can be used to check an assumed chemical
composition of the precipitate.
The measured A-ratios range from 4.6 to 6.4. A comparison of
A with
calculated values for different potential intermetallic compounds is given in
Table 4.
DISCUSSION
Nanometer-sized precipitates are the most relevant microstructural con-
stituents in maraging steels to determine their properties. The precipitates
can be characterised with regard to shape, size, composition, particle number
density, volume fraction and distribution.
In the early stage of the precipitation process and at lower temperatures
spherical precipitates (samples aged for 0.25 h and 12 h at 748 K) were
exclusively observed by EFTEM. This is the reason why the form factor
for spheres was used for the analysis of the SANS data. In addition, larger
rod-like particles have been observed in samples aged for 100 h at 748 K
as well as 16 h at 798 K. To avoid an over-interpretation of the data these
rod-like particles were not included in the SANS data analysis. However,
these larger particles could possibly be responsible for the small deviation
between the measured and calculated scattering curves (Fig. 3 and 4) of the
samples aged for 100 h at 748 K and 16 h at 798 K, but the corresponding
volume fraction is considered to be small in comparison with the volume
fraction of the spherical particles. Therefore, the interest was focussed in
560
6TH INTERNATIONAL TOOLING CONFERENCE
the analysis of the very small precipitates originating from the early stage of
the ageing process which are believed to control the strength properties of
this steel. The best fits of the SANS data were achieved for samples aged for
0.25 h and 12 h at 748 K. The mean radii are 0.69 and 1.35 nm, respectively
(Table 3). In case of the sample aged for 0.25 h the mean radius could not
be measured by EFTEM due to the insufficient resolution. However, the
calculated size of the precipitates by SANS of the other samples aged for 12
h and 100 h at 748 K correlates well with the size estimated by EFTEM.
Generally, the spherical particles are much smaller than the rod-like par-
ticles. After ageing at 798 K for 16 h the rod-like precipitates have lengths
up to 30 nm and a diameter of 6 nm, whereas the spherical ones are only 4
nm in diameter. In reference [16] it was speculated that the rod-like pre-
cipitates might be aggregates of fine spheroidal particles, which have been
precipitated along dislocation lines.
EFTEM investigations have shown that both spherical and rod-like pre-
cipitates consist of a considerable amount of titanium, nickel und silicon
(Fig. 1). From Table 4 it is obvious that the G-phase (Ti
6
Si
7
Ni
16
) might be
one of the potential candidates. In fact, the existence of G-phase in a steel of
similar composition was reported by Gemperle et al. [7] and Sha et al. [17].
In addition, the assumed chemical composition of the precipitates was
checked by SANS. A comparison of the A-ratio with calculated values for
different intermetallic compounds (Table 4) leads to the conclusion that the
precipitates in the samples aged for 0.25 h and 12 h at 748 K (A value 4.6
and 6.4, respectively) are most likely Ti
6
Si
7
Ni
16
, because the corresponding
A = 4.4 for Ti
6
Si
7
Ni
16
matches well with the observed values. The phases
NiAl, Ni
3
Mo, Ni
4
Mo and Fe
7
Mo
6
, that were found in similar steels [18, 19]
are not compatible with the observed A values. Theoretically, also a mixture
composed of Ni
3
Ti and Ni
2
Ti (Table 4) might be possible. However, all
EFTEM investigations have shown that the precipitates contain a significant
amount of silicon. This observation excludes Ni
3
Ti as a possible phase,
because it shows no pronounced solubility for Si [7]. Thus, it is tempting to
speculate that the rod-like precipitates in the samples aged for 100 h at 748
K and 16 h at 798 K also consist of G-phase Ti
6
Si
7
Ni
16
. This result is in
contrast with the investigations by Gemperle et al. [7] who suggested that
the rod-like particles are Ni
3
Ti phase. In our study, however, the rod-like
precipitates also incorporate silicon (as measured with EFTEM).
On the Precipitation Behaviour in Maraging Steels
561
Up to now it is unclear if the observed range in
A (Table 4) reflects an error
width or true changes in particle composition during the ageing treatment.
One might speculate that in the early stage of precipitate formation, some
of the constituting elements of the G phase are replaced by other elements
present in the matrix. These elements will influence the average scattering
length density of the precipitate and thus alter the
A-ratio. Therefore, further
investigations with atom probe field ion microscopy or electron diffraction
will be conducted to gain additional understanding on the chemical compo-
sition of the precipitates in this particular alloy system.
The particle number density and the distribution of fine precipitates in
bulk specimen can be determined at higher accuracy by means of SANS than
it is possible by means of TEM. Especially for particles in the nanometer
range the TEM resolution was found to be insufficient. A second problem
is that the precipitates are smaller than the thickness of the thin TEM foil
and, therefore, they can superpose each other. As a result, it was impossible
to calculate the exact particle number density by analysing the TEM data.
Applying SANS a larger sample volume can be investigated and this leads
to a good statistic within a short measurement time which is impossible to
achieve with direct methods like TEM.
SUMMARY
SANS and EFTEM investigations were carried out to study the precipi-
tation behaviour in a cobalt-free chromium containing maraging steel. By
combining these methods quantitative results on size, distribution, volume
fraction as well as chemical composition of the precipitates were achieved
for the early stages of the ageing process. After aging at 748 K for 0.25 h,
12 h and 100 h, the particle volume fractions are 5.4 %, 6.6 % and 7.8 %,
respectively. The particle radius increases from 0.7 nm up to 2.05 nm and
the particle number density decreases at the same time from
4.3 · 10
19
cm
−
3
to
2.5·10
18
cm
−
3
. In the early stage of the precipitation process and at lower
temperatures spherical precipitates were exclusively observed by EFTEM
whereas additional larger rod-like particles appeared in samples aged longer
and/or at higher temperatures. The
A-value, which is the ratio of mag-
netic to nuclear scattering intensities obtained by SANS, corresponds with
the intermetallic G-phase Ti
6
Si
7
Ni
16
. The phase plays a major role in the
precipitation behaviour and thus is considered to dominate the mechanical
properties of this particular maraging steel.
562
6TH INTERNATIONAL TOOLING CONFERENCE
ACKNOWLEDGMENTS
The authors would like to thank Böhler Edelstahl GmbH & Co KG for
supply of the model alloy. Financial support from the Austrian Kplus Com-
petence Center Programme is gratefully acknowledged.
REFERENCES
[1] V. K. VASUDEVAN, S. J. KIM and C. M. WAYMAN, Metall. Trans. 21A (1990)
2655.
[2] W. SHA, A. CEREZO and G. D. W. SMITH, Metall. Trans. 24A (1993) 1221.
[3] N. BOUZID, C. SERVANT and O. LYON, Phil. Mag. B 57 (1988) 343.
[4] C. SERVANT and N. BOUZID, Phil. Mag. B 60 (1989) 659.
[5] X. D. LI, Z. D. YIN, H. B. LI, T. C. LEI, M. L. LIU, X. W. LIU AND M. Z. JIN,
Mater. Chem. Phys. 33 (1993) 277.
[6] R. Tewari, S. Mazumder, I. S. Batra, G. K. Dey and S. Banerjee, Acta mater. 48 (2000)
1187.
[7] A. GEMPERLE, J. GEMPERLOV ´
A, W. SHA and G. D. W. SMITH, Mat. Sci. Techn.
8 (1992) 546.
[8] F. HOFER, P. WARBICHLER, W. GROGGER and O. LANG, Micron 26 No.5 (1995)
377.
[9] P. WARBICHLER, F. HOFER, P. HOFER and E. LETOFSKY, Micron 29 No.1 (1998)
63.
[10] F. HOFER and P. WARBICHLER, Prakt. Metallgr., Sonderband 26, Carl Hanser,
M¨unchen (1995) 393.
[11] G. KOSTORZ, in "Treatise on Materials Science and Technology, Vol. 15: Neutron
Scattering" (Academic Press, 1979).
[12] J. S. PEDERSEN, J. Applied Crystallography 27 (1994) 595.
[13] A. GUINIER and G. FOURNET, in "Small-Angle Scattering of X-Rays" (John Wi-
ley&Sons, Chapman &Hall, 1955).
[14] G. E. BACON, in "Neutron Diffraction" (Clarendon Press, Oxford, 1975).
[15] M. GROSSE, A. GOKHMAN and J. B ¨
OHMERT, Nucl. Instr. Meth. Phys. Res. B 160
(2000) 515.
[16] B. R. BANERJEE and J. J. HAUSER, in "Symposium of the Transformation and
Hardenability in Steels" February 1967 (University of Michigan) p. 133.
[17] W. SHA, A. CEREZO and G. D. W. SMITH, Metall. Trans. A 24 (1993) 1241.
[18] V. SEETHARAMAN, M. SUNDARARAMAN and R. KRISHNAN, Mat. Sci. Eng.
47 (1981) 1.
On the Precipitation Behaviour in Maraging Steels
563
[19] C.V. ROBINO, P.W. HOCHANADEL, G.R. EDWARDS and M.J. CIESLAK, Metall.
Trans. 25A (1994) 697.