Professional Documents
Culture Documents
Charitidis2012 PDF
Charitidis2012 PDF
Charitidis2012 PDF
a r t i c l e i n f o a b s t r a c t
Article history: Residual stress during Friction Stir Welding (FSW) of lightweight alloys is of major concern, due to their
Received 19 November 2011 functionality and applications in transport and industry elds. Several conventional techniques which
Received in revised form 30 January 2012 are used to measure and characterize welded aluminum alloys are destructive. This drawback has raised
Accepted 31 January 2012
nanoindentation as the non-destructive alternative technique with many advantages, such as easy prepa-
Available online 8 February 2012
ration and high spatial resolution. In this study a methodology was brought forward to investigate the
applicability of nanoindentation method in order to overcome limitations of this technique and measure
Keywords:
residual stress in two of most commonly used aluminum alloys in transport and industrial applications.
Nanoindentation
Aluminum alloys 2012 Elsevier B.V. All rights reserved.
Friction stir welding
Residual stress
1. Introduction alloys, but since then FSW has also been utilized on joining a large
range of materials. In FSW the interaction of a non-consumable and
In transport and industrial applications the lightweight alloys rotating tool with the work pieces being welded creates a welded
undergo signicant mechanical loads under different applied loads joint through frictional heating and plastic deformation at temper-
and time dependent conditions. More specically, high-demanding atures below the melting temperature of the alloys being joined.
mechanical properties and high accuracy to strength-to-weight The contacting shoulder applies frictional heat to the weld region
ratio of such materials have led to an increased research. Aluminum and prevents highly plasticized material from being expelled dur-
alloys offer a high potential for weight reduction in automotive ing the welding operation. The combined frictional heat from the
and other transportation vehicle construction. Aluminum alloy pin and the shoulder creates a plasticized condition around the
AA6082-T6 (AA: provided by the Aluminum Association) is a immersed pin and the contacting surface of the shouldered region
high-strength AlMgSi alloy containing manganese to increase of the work piece top surface. Material ows around the tool and
ductility and toughness. The T6 condition is obtained through coalesces behind it while relative traverse between substrate and
articial ageing at a temperature of 170200 C, mainly for the rotating tool occurs. Notwithstanding the widespread interest
welding applications [1]. Aluminum alloy AA5083-H111 is an in the possibilities offered by FSW, data concerning the mechan-
AlMg alloy which can be classied as wrought alloy prod- ical behavior of joints obtained using this process is still scarce.
uct with excellent corrosion resistance. The H111 condition is The heat input (most of it generated by the friction between
obtained through both cold hardening and partial annealing. In tool shoulder and materials to be joined) increases by increas-
particular, both alloys are often applied in shipbuilding industry ing tool rotational speed and by decreasing welding speed. Low
[2]. heat input causes the intermittent material ow and improper stir-
The increasing relevance of aluminum alloys in transportation ring action around the tool pin due to insufcient plasticization
requires research on more efcient and reliable joining processes. of the materials under the tool shoulder. On the other side, high
Friction Stir Welding (FSW) is a solid-state joining process, which heat input causes turbulent material ow around the tool pin due
emerged as an alternative technique to be used in high strength to excess plasticization of material under the tool shoulder. Fur-
alloys that were difcult to join with conventional joining tech- ther details and analysis of the process are reported elsewhere
niques. This welding process was initially developed for aluminum [3].
The frequent failure of structural components by time depen-
dent degradation in severe operating environments has recently
Corresponding author. Tel.: +30 2107724046; fax: +30 2107722339. become a concern. In particular, pipeline operating condi-
E-mail address: charitidis@chemeng.ntua.gr (C.A. Charitidis). tions are made more severe by cryogenic contents and many
0921-5093/$ see front matter 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2012.01.129
C.A. Charitidis et al. / Materials Science and Engineering A 540 (2012) 226234 227
1000
(a) Advancing side (-) Retreating side (+)
(a)
Al<200>
Al<111>
2000
800
1 23
2 4 5 7
3 5 6 1
AA 6082-T6
1750
base metal
1500 2 cm
600
3 (-3) cm (1)
Intensity
Load ()
-Al12(Fe,Mn)3Si
-1 cm (3)
400 1000 -0.8 cm (4)
0 cm (5)
0.5 cm (6)
-Mg2Si
750
1.3 cm (7)
200
AlFeSi
Al6Mn
500
250
0
40 50 60 70 80 90 100 110 0
0 20 40 60 80 100 120 140 160 180 200
2 Theta / degree
Displacement (nm)
1000
(b)
Al<200>
Al<111>
800
AA 6082-T6
2500
1 23 4 5 6 1
AlFeSi
2250
welding zone
600 2000
20 mm
Intensity
1750
-Al12(Fe,Mn)3Si
-15.5 mm (1)
Load (N)
1500 -10 mm (2)
400 - 8mm (3)
1250 0 mm (4)
-Mg2Si
4 mm (5)
1000
Al6Mn
200 11 mm (6)
750
500
0
250
40 50 60 70 80 90 100 110
0
2 Theta / degree 0 20 40 60 80 100 120 140 160 180 200
Displacement (nm)
Fig. 2. XRD analysis in diffractogram form of base metal and welding zone for
AA6082-T6. Fig. 4. Nanoindentation load-displacement curves at 200 nm indentation depth for
(a) the AA 6082-T6 aluminum alloy and for (b) the AA 5083-H111 aluminum alloy
(transverse). Hatched area indicates the welding zone.
3. Analytical modeling
3.1. Nanoindentation analysis on the half-space elastic deformation theory, H and E values can be
extracted from the experimental data (load displacement curves)
Nanoindentation testing was performed with Hysitron Tribo- using the O&P method [7], where derived expressions for calculat-
Lab Nanomechanical Test Instrument. Details about the instrument ing the elastic modulus from indentation experiments are based on
and the experimental setup have been presented elsewhere [2]. Sneddons elastic contact theory [22]:
The surface of the samples was prepared by grinding and polishing
(with 1 and 0.1 m agglomerated -alumina suspension). Based S
Er = (1)
2 Ac
Compressive Stress free state where S is the unloading stiffness (initial slope of the unloading
stress state load-displacement curve at the maximum displacement of pene-
LC tration (or peak load)), Ac is the projected contact area between the
tip and the substrate and is a constant that depends on the geom-
etry of the indenter ( = 1.167 for Berkovich tip) [3,4]. Conventional
nanoindentation hardness refers to the mean contact pressure; this
Tensile
Load ()
LT where
2 1/2 1/16
A(hc ) = 24.5hc + a1 hc + a1/2 hc + + a1/16 hc (3)
and
Displacement (nm)
Pm
hc = hm (4)
Fig. 3. Typical nanoindentation load-displacement curves for different stress states. Sm
C.A. Charitidis et al. / Materials Science and Engineering A 540 (2012) 226234 229
2250
2000
1 2 3 45 6
1750
2 cm
1500 -2 cm (1)
Load (N)
-1 cm (2)
1250 0 cm (3)
1 cm (4)
1000 1.2 cm (5)
3 cm (6)
750
500
250
0
0 20 40 60 80 100 120 140 160 180 200
Displacement (nm)
1750
1 2 3 4 5 6 1
1500
20 mm
1250 -15.5 mm (1)
Load ()
-10.5 mm (2)
1000 -8.5 mm (3)
0 mm (4)
4.5 mm (5)
750
8.5 mm (6)
500
250
0
0 20 40 60 80 100 120 140 160 180 200
Displacement (nm)
Fig. 5. Nanoindentation load-displacement curves at 200 nm indentation depth for (a) the AA 6082-T6 aluminum alloy and for (b) the AA 5083-H111 aluminum alloy (cross
section). Hatched area indicates the welding zone.
where hm is the total penetration displacement of the indenter at shear deviator stress D . The surface-normal deviator stress in z-
peak load, Pm is the peak load at the indenter displacement hm , direction (normal to the surface of indentation) zD is 2res /3 by
and is an indenter geometry constant, equal to 0.75 for Berkovich removing the hydrostatic stress 2res /3 from the surface resid-
indenter [23]. ual stress res and is added to the contact pressure [9]. LT L0 is
dened as a product of the selected deviator stress component and
3.2. Residual stress measured by nanoindentation its corresponding contact area ATC . Thus, an equation for the equibi-
axial residual stress is derived in terms of the indentation load and
A surface residual stress is assumed to be in an equibiaxial ten- contact area as:
sile state and uniform in the near-surface region (taken as about
(L0 LT )
three times the indentation depth) [8,9]. If an arbitrary indentation res = 3 (5)
state (ht , L0 ) is attained in an unstressed state and if the tensile 2ATC
inplane stress res is applied to the loading state at a xed pen-
etration depth ht , the indentation load L0 is reduced to a load LT Here, ATC in the tensile stress state is calculated from LT A0C /L0
due to the decrease in surface penetration resistance (in Fig. 3 typ- because the contact hardness Hc or L0 /A0C = LT /ATC is independent
ical nanoindentation load-displacement curves for different stress of the elastic residual stress. This measurement is based on the ini-
states are presented). The stress tensor is separated into mean tial elastic unloading part of the load-displacement curve which is
stress M (hydrostatic stress) and plastic-deformation-sensitive unaffected by the residual stresses for materials with isotropic or
230 C.A. Charitidis et al. / Materials Science and Engineering A 540 (2012) 226234
1000 600
(a) AA6082-T6 Welding zone (a) AA6082-T6 Welding zone
Advancing side Retreating side Advancing side Retreating side
800 400
600
200
400
0
200
-200
0
-400
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 -4 -3 -2 -1 0 1 2 3 4
Distance from weld center (cm) Distance from weld center (cm)
3000
(b) AA5083-H111 Welding zone (b) 1000
AA5083-H111 Welding zone
Advancing side Retreating side Advancing side Retreating side
2500
500
1500
0
1000
500
-500
-500 -1000
-24 -20 -16 -12 -8 -4 0 4 8 12 16 20 24 -24 -20 -16 -12 -8 -4 0 4 8 12 16 20 24
Distance from weld center (mm) Distance from weld center (mm)
Fig. 6. Residual stress distribution for (a) the AA 6082-T6 aluminum alloy and for (b) Fig. 7. Residual stress distribution for (a) the AA 6082-T6 aluminum alloy and for
the AA 5083-H111 aluminum alloy (transverse). Hatched area indicates the welding (b) the AA 5083-H111 aluminum alloy (cross section). Hatched area indicates the
zone. welding zone.
120
(a) AA6082-T6 Welding zone 110
AA5083-H111
AA6082-T6
100
y=55.25x-6.71
Hmicro (HV0.3)
100
80
60 90
Hmicro (HV0.3)
40 80
2.0
1.8 70
Hnano (GPa)
1.6
60
1.4
1.2
50
1.0
-5 -4 -3 -2 -1 0 1 2 3 4 5
1.0 1.2 1.4 1.6 1.8 2.0 2.2
Distance from weld center (cm)
Hnano (GPa)
(b) AA5083-H111 Welding zone
90
Fig. 9. Distribution of micro- and nano-hardness for both alloys [43].
Hmicro (HV0.3)
80
propagate downward and toward the surface as the loading con-
70 tinues and then the crack becomes unstable [32].
In the measured residual stresses shown in Fig. 6(a) for AA6082-
60 T6 the stresses appear to rise to a tensile stress of 800 MPa at
1.8 1 cm from the center, before reversing in trend to form a ten-
1.6
1.4
sile region near the weld line of about 600 MPa. It can be seen
that this tensile region peaks at around 1 cm from the weld line,
Hnano (GPa)
1.2
1.0 a distance corresponding to the edge of the tool shoulder, before
0.8
dropping slightly at the weld centre as it is shown in Fig. 6(a). Also,
0.6
0.4 in the measured residual stresses shown in Fig. 6(b) for AA5083-
0.2 H111 the stresses from the parent material to the welding zone
0.0 appear to decrease linearly to a compressive stress steady state of
-16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16
Distance from weld center (mm)
Fig. 8. Transverse distribution (comparison) of Hmicro and Hnano for both alloys (a)
AA 6082-T6 and (b) AA 5083-H111. Hatched area indicates the welding zone [43].
(a)
hc/hm
AA6082-T6
AA5083-H111
10 100 1000
Displacement, hm (nm)
10
(b)
hc/hm
AA6082-T6
AA5083-H111
0.1
0.015 0.02 0.025 0.03 0.035
H/E*
about 500 MPa near the weld line. This form of prole has been 4.2. Nano- and micro-hardness distribution across the weld for
observed previously for aluminum AA5083-H111 by Peel et al. [29]. both alloys
The dip around the pin position may be attributed to the high
temperatures associated with this region, which would limit the Transverse distribution of Hmicro and Hnano presented the same
capacity of the material to support the generated load. Alterna- shape [3,33] and the same decrease of values between base metal
tively, it could be a consequence of stress relief that occurred when and welding zone (20% for AA5083 and 33% for AA6082), as
the plate was cut up. For both alloys, a small plateau region appears shown in Fig. 8.
near the welding line; yet, the peak stresses are much higher in the In Fig. 9 the distribution of micro- and nano-hardness for both
case of AA5083-H111. alloys is presented. The linear regression equation for this relation-
In the measured residual stresses shown in Fig. 7(a) for AA6082- ship is:
T6 the stresses appear to rise to a tensile stress of 250 MPa at 1 cm
Hmicro (HV ) = 55.26 Hnano (GPa) 7.28 (8)
from the center, before reversing in trend to form a compressive
region near the weld line. It can be seen that this tensile region where Hmicro represents the microhardness and Hnano represents
peaks at around 1 cm from the weld line, a distance correspond- the nanohardness.
ing to the edge of the tool shoulder, before dropping slightly at The square of the correlation coefcient R2 for (8) is 90, indi-
the weld centre. Also, in the measured residual stresses shown in cating that approximately 90% of the total variation is explained by
Fig. 7(b) for AA5083-H111 the stresses appear to decrease linearly this linear regression equation. The Vickers and Berkovich inden-
to a compressive stress steady state of about 500 MPa near the ters have similar geometric effects on indentation geometry, so it
weld line. is reasonable that Vickers testing and nanoindentation Hc values
C.A. Charitidis et al. / Materials Science and Engineering A 540 (2012) 226234 233
exhibit reasonable statistical relationship. The indentation shape a both interesting and complicated problem for the estimation of
produced by the Berkovich indenter (used for the Hnano tests in this residual stress by nanoindentation.
study), is designed to be similar to the indentation created by the
Vickers indenter. The experimental results show that Hnano results 5. Conclusions
from a nanomechanical test instrument with a Berkovich indenter
can be reliably compared with Hmicro (Vickers) test results. Thus, Friction stir welding, contrary to classical welding techniques,
the selection of an appropriate H test should be based on the prob- has the advantage that most of the welding parameters can be
lem being investigated. The results of this investigation show that controlled in a precise manner, thus controlling the energy input
Hnano test results from a nanomechanical test instrument with a into the system. However, the effect of different welding param-
Berkovich indenter can be used to reliably analyze Hnano distribu- eters on the weld properties remains an area of uncertainty. In
tion on a friction stir welded alloy. The enhancement of elasticity this work, the results of microstructural, mechanical integrity and
for weaker loads may be a real physical effect of superelastic behav- residual stress investigations of two friction stir welds (AA6082-T6
ior of materials under mN scale forces [34], due to the inactivation and AA5083-H111 aluminum alloys (in both the longitudinal (par-
of dislocations. It may also be an artefact due to the piling-up of the allel to tool travel) and transverse (perpendicular to tool travel)
surface during indentation. directions) produced under varying conditions), are reported. It
was found that the weld properties were dominated by the ther-
mal input rather than the mechanical deformation by the tool.
4.3. Pile-up/sink-in deformation It is clearly from this work that nanoindentation could arise
like an alternative technique to estimate and describe residual
The contact area is inuenced by the formation of pile-ups stresses distribution. It is found that residual stress distribution
and sink-ins during the indentation process. To accurately mea- in the longitudinal direction of AA5083-H111 is in agreement
sure the indentation contact area, pile-ups/sinks-ins should be in trend with classical methods such as synchrotron residual
appropriately accounted for. The presence of creep during nanoin- stress analysis. Increasing the traverse speed (and hence reduc-
dentation has an effect on pile-up, which results in incorrect ing the heat input) narrows this weld zone. Further investigation
measurement of the material properties. FischerCripps observed includes extended pile-up investigation and synchrotron residual
this behaviour in aluminum where the measured elastic modu- stress analysis, which is reported to indicate whether the weld
lus was much less than expected [35]. Rar et al. [36] observed zone is in tension in both the longitudinal and transverse direc-
that the same material when allowed to creep for a long dura- tions.
tion produced a higher value of pile-up/sink-in indicating a switch
from an initial elastic sink-in to a plastic pile-up [36]. Creep- Acknowledgements
ing time is reported to have no impact on the pile-up/sink-in of
rate sensitive aluminum alloy [37]. In Fig. 10, SPM imaging of AA Authors would like to thank D.I. Pantelis and N.M. Daniolos for
6082-T6 [3] and AA 5083 alloys for 5000 N of applied load is pre- providing the FSW aluminum alloys.
sented; AA 6082-T6 exhibits signicant pile-up compared to AA
5083. References
In Fig. 11, the normalised pile-up/sink-in height hc /hm is plot-
ted vs. displacement hm and the normalized hardness H/E*. Higher [1] A. Scialpi, L.A.C. De Filippis, P. Cavaliere, J. Mater. Des. 28 (2007) 11241129.
stresses are expected in high H/E*, hard materials, and high stress [2] R. Nandan, T. DebRoy, H.K.D.H. Bhadeshia, Prog. Mater. Sci. 53 (August (6))
(2008) 9801023.
concentrations develop towards the indenter tip, whereas in case [3] E.P. Koumoulos, C.A. Charitidis, N.M. Daniolos, D.I. Pantelis, Mater. Sci. Eng. B
of low H/E*, soft materials, the stresses are lower and are dis- 176 (2011) 15851589.
tributed more evenly across the cross section of the material [35]. [4] J.Y. Kim, K.W. Lee, J.S. Lee, D. Kwon, Surf. Coat. Technol. 201 (2006) 42784283.
[5] J. Jang, D. Son, Y.H. Lee, Y. Choi, D. Kwon, Scr. Mater. 48 (2003) 743748.
Rate sensitive materials experience less pile-up compared to rate
[6] L. Yajiang, W. Juan, C. Maoai, S. Xiaoqin, Indian Acad. Sci. 27 (2004) 127132.
insensitive materials due strain hardening. Cheng and Cheng [38] [7] W.C. Oliver, G.M. Pharr, J. Mater. Res. 7 (1992) 15641583.
reported a 22% pile-up for a work hardening exponent. This is [8] S. Suresh, A.E. Giannakopoulos, Acta Mater. 46 (1998) 57555767.
[9] Y.-H. Lee, D. Kwon, Scr. Mater. 49 (2003) 459465.
consistent with the fact that when hc /hm approaches 1 for small
[10] N. Huber, J. Heerens, Acta Mater. 56 (2008) 62056213.
H/E*, deformation is intimately dominated by pile-up [39,40]. On [11] S.K. Chionopoulos, C.H.I. Sarafoglou, D.I. Pantelis, V.J. Papazoglou, Proc. Int. Conf.
the other hand when hc /hm approaches 0 for large H/E* it corre- Manuf. Eng. (ICMEN), Chalkidiki, 2008.
sponds to purely elastic deformation and is apparently dominated [12] K. Nakata, S. Inoki, Y. Nagano, T. Hashimoto, S. Johgan, M. Ushio, J. Jpn. Inst.
Light Met. 51 (2001) 528533.
by sink-in in a manner prescribed by Hertzian contact mechanics [13] W.B. Lee, J.W. Kim, Y.M. Yeon, S.B. Jung, Mater. Trans. 44 (2003) 917923.
[41]. [14] W.B. Lee, Y.M. Yeon, S.B. Jung, Mater. Sci. Technol. 19 (2003) 785790.
An advantage of extracting residual stress by nanoindentation [15] S.H.C. Park, Y.S. Sato, H. Kokawa, T. Tsukeda, in: S.A. David, T. DebRoy, J.C.
Lippold, H.B. Smartt, J.M. Vitek (Eds.), Proceedings of the 6th International
is that there is no prerequisite of the knowledge of the stress state Conference on Trends in Welding Research, ASM International, USA, 2003, pp.
of a reference sample or any particular mechanical properties of 267272.
the test materials; nanoindentation gives the opportunity of con- [16] Y. Li, L.E. Murr, J.C. McClure, Mater. Sci. Eng. A 271 (1999) 213223.
[17] Z.Y. Ma, R.S. Mishra, M.W. Mahoney, Acta Mater. 50 (2002) 44194430.
tinuous measurement of residual stress and strain based on the [18] Y.J. Kwon, N. Saito, I. Shigematsu, Mater. Sci. Lett. 21 (2002) 14731476.
continuous measurement of stiffness [42]. However, as the method [19] Y.S. Sato, M. Urata, H. Kokawa, Metall. Mater. Trans. A 33 (2002) 625635.
relies on an accurate determination of the hc /hm ratio and as a [20] Y.J. Kwon, I. Shigematsu, N. Saito, Mater. Trans. 44 (2003) 13431350.
[21] Y.J. Kwon, I. Shigematsu, N. Saito, Scr. Mater. 49 (2003) 785789.
consequence of the contact area, experimental factors such as sur-
[22] I.N. Sneddon, Proc. Camb. Philos. Soc. 44 (1948) 492507.
face roughness, creep and pile-up may result in large error in the [23] R.B. King, Int. J. Solids Struct. 23 (1987) 16571664.
experimental determination of the hc /hm ratio from the unload- [24] A.E. Giannakopoulos, J. Appl. Mech. 70 (2003) 638643.
[25] Y.H. Lee, D. Kwon, Acta Mater. 52 (2004) 15551563.
ing curve of indentation. In nanoindentation measured areas that
[26] J.H. Underwood, Exp. Mech. 13 (1973) 373380.
exhibit pile-up the contact area calculated by the O&P method will [27] Y.H. Lee, K. Takashima, Y. Higo, D. Kwon, Scr. Mater. 51 (2004) 887891.
be overestimated, which leads to an underestimation of hardness [28] Y.H. Lee, J.Y. Kim, J.S. Lee, K.H. Kim, J.Y. Koo, D. Kwon, Philos. Mag. 86 (2006)
of both free and residual stress state region. Also, many uncertain- 54975504.
[29] M. Peel, A. Steuwer, M. Preuss, P.J. Withers, Acta Mater. 51 (2003) 47914801.
ties are connected with creep, since O&P method is unable to take it [30] X.D. Li, B. Bhushan, Thin Solid Films 315 (1998) 214221.
into account time dependent deformation. These drawbacks pose [31] X.D. Li, B. Bhushan, Thin Solid Films 355 (1999) 330336.
234 C.A. Charitidis et al. / Materials Science and Engineering A 540 (2012) 226234
[32] P. Venkateswaran, Z.-H. Xu, X. Li, A.P. Reynolds, J. Mater. Sci. 44 (2009) [39] R. Hill, B. Storakers, A.B. Zdunek, Proc. R. Soc. Lond. A 423 (1989) 301330.
41404147. [40] S. Biwa, B. Storakers, J. Mech. Phys. Solids 43 (1995) 13031333.
[33] D.P.P. Booth, M.J. Starink, I. Sinclair, Mater. Sci. Technol. 23 (2007) 276284. [41] H. Hertz, Miscellaneous Papers by H. Hertz, Macmillan, London, 1896.
[34] G. Cross, A. Schirmeisen, A. Stalder, P. Grtter, M. Tschudy, U. Drig, Phys. Rev. [42] Z.-H. Xu, X. Li, in: F. Yang, J.C.M. Li (Eds.), Micro and Nano Mechanical Testing
Lett. 80 (1998) 46854688. of Materials and Devices, Springer Verlag, New York, 2008, pp. 139154.
[35] A.C. Fischer-Cripps, Mater. Sci. Eng. A 385 (2004) 7482. [43] E.P. Koumoulos, C.A. Charitidis, N.M. Daniolos, D.I. Pantelis, Determination of
[36] A. Rar, S. Sohn, W.C. Oliver, D.L. Goldsby, T.E. Tullis, G.M. Pharr, Fund. Nanoin- onset of plasticity (yielding) and comparison of local mechanical properties of
dentation and Nanotribology III. Symp, 2005, pp. 119124. friction stir welded aluminum alloys using the micro- and nano- indentation
[37] S. Mandal, S. Kose, A. Frank, A.A. Elmustafa, Int. J. Surf. Sci. Eng. 2 (2008) 4151. techniques, Int. J. Struct. Integrity (2011) in press.
[38] Y.-T. Cheng, C.M. Cheng, Philos. Mag. Lett. 78 (1998) 115120.