": ~ ; ' ,, :\

journal of
MOLECULAR
LIQUIDS
ELSEVIER Journal of Molecular Liquids 86 (2000) 273-277
www.elsevier.nl/locate/molliq

The Snoek relaxation

Grzegorz Haneczok

Institute of Physics and Chemistry of Metals, Silesian University, 40-007 Katowice, .

Bankowal2, Poland, e-mail haneczok@us.edu.pi.

Theory of the Snoek relaxation is presented and discussed by taking into account the
interaction of interstitial solute atoms (i.e. random cooperative strain interaction model).
© 2000 Elsevier Science B.V. All rights reserved.
1. I N T R O D U C T I O N

Interstitial solute atoms (ISAs) such as oxygen, nitrogen and carbon in bcc metals occupy
octabedral sites causing local strain distortions of tetragonal symmetry. Thus an ISA in the
bcc lattice forms an elastic dipole which can lie along one of the three cubic axes (denoted as
/_,=x, y or z) as presented in Fig.1. The strain ellipsoid of the elastic dipole is completely
characterised by the principal values (gu) of the co-called lambda tensor, defined as the strain
tensor per unit mole fraction of ISAs having the same orientation. For a defect of tetragonal
symmetry we have obviously Lteg2=g3 [1, 2].
Application of external stress tr leads to differences in free energy for different dipole
orientations. The solid approaches a new thermodynamic equilibrium by stress-induced
transitions between sites of different energy corresponding to reorientation jumps of the
elastic dipoles. This gives rise to an anelastic behaviour known as the Snoek relaxation, which
may be observed in static experiments (o---const.) as a time-dependent increase of strain -
creep (or if tr is released as an elastic after-effect). In dynamic experiments (o=tr0exp(icot), co
- the angular frequency) usually a loss tangent maximum (Q-I) is measured versus
temperature or frequency [ 1-6].
For uniaxial external stress o-=tr= the strain operator of the anelastic deformation
associated with the defect at site i is defined as [2, 6]:

_ l
e, ,u 1 _,J2 l
(i)
/t

where P / ' is the operator whose eingenvalue is equal to one if the i-th octabedral site is
occupied by an elastic dipole in orientation/z and is zero otherwise, N is the number of host
atoms and AA=A1-22. In the absence of cr the three directions x, y and z are occupied with the
same probability. For try0 the time evolution of the expectation value of P,fft) denoted as
<P,~(t)> is described by the kinetic equations of the form [I, 2, 7]:

a <e/(t)>_ <g(t)>-4e v < ,e'(t)>] (2)

art ~,

0167-7322/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.
PII SO167-7322(99) 00149-X
274

where W~u are frequencies oflSAs jump from a site of type vto/a. The kinetic equations must
satisfy a detailed balance condition in order to assure a thermal equilibrium with the canonical
d tsmouuon
.......
( ,~,T,
~ CO>= ur, for t ~ ) . In other words, the energy levels of elastic dipoles are split
by stress and they will subsequently be repopulated in accordance with the canonical
distribution. So the detailed balance condition is:

W,~ = exp[fl(gf - g~ )]Wu, (3)

where # = l / k T (with k the Boltzmann constant and T the temperature), g~a represents the
change of the energy level of site i (p orientation) and in general consists of two parts - let
say - internal h,a describing possible interaction of elastic dipoles and external due to the
presence of the external stress (elastic energy is Vco;, with V the specimen volume). So one
can write:

g," = h: + VoZ~,t,~(a~, - ~) (4)

where 6~ is the Kronecker delta and vo=V/N is the specific volume of the host lattice. Usually
one assumes that the unperturbed frequencies of jumps are: W~y=Wy~=Wy~=W~= W0exp(-flH),
with Wo the pre-exponential factor and H the activation enthalpy of ISAs diffusion.
Based on the model described above one can calculate the static as well as the dynamic
response functions i.e. the mechanical compliance J(t) and J(ro) which are defined by [1, 2]:

)-~, < 8, (t) > ®

d(t) = Jv + - - and d(co) = - i c o f d ( t ) exp(icot)dt (5)
G o

where Jv is the unrelaxed compliance given by the Hooke's law. The loos tangent curve {24
defined as ImJ(oJ)/Re J(ro) usually can be approximated by ImJ(ro)/Ju.

2. N O N I N T E R A C T I N G SYSTEM

Let us assume that the concentration of elastic dipoles in the bcc lattice is so small that
interactions between them can be precluded. Thus h~a---0 and equation (4) takes the form:
g,a=voAAo(6~,-1/3). In this case, as it can be expected, the response functions J(t) and J(ro)
describe a single relaxation process of the Debye type:

A/
J ( t ) = J u + &/(1 - exp~-t/r)) and J(w) = Jv 4--- (6)
1 - iror

where fi/=2,&:v0(A2)2/9 is proportional to the ISAs concentration c. The relaxation time ~,

defined as ~l/3W0)exp(flH), has an Arrhenius temperature dependence and can be related to
the diffusion of ISAs in bcc metals. For the model with the elementary jumps between
neighbouring octahedral site only, the diffusion coefficient D is expressed as: D---a2/36r (a -
the lattice parameter). This fact has been used quite extensively for determining the activation
enthalpies of interstitial migration [1, 2]. It may be emphasised that the noninteracting system
 275

of elastic dipoles is characterised by a single relaxation time r, which depends on temperature

via the Arrhenius relation and cot/oc1/T.

3. INTERACTING SYSTEM

Now the concentration of ISAs in the bcc lattice is assumed to be high enough that long-
range interactions between them should be taken into account. The interaction matrix of the
system of elastic dipoles can be defined as [2, 7, 9]:

= ±(FII - E $ ) (7)

where E~ and E~ are the energy of pairs of elastic dipoles in parallel and perpendicular
orientations, respectively. Let notice that for bcc crystals only these two cases are possible, so for
a noninteracting system ~)=0. The essential idea of the so-called random cooperative strain
interaction (RCSI) model consists of the assumption that the elements of the o¢~ matrix are
distributed randomly with the Gaussian probability and zero mean value. It is obvious that
interaction energies of pairs of elastic dipoles depend on the distances between them and if
the distribution oflSAs is random the ~,) elements should also be random. Thus the matrix ~¢,)
plays a role of a random matrix Whose eigenvalue spectrum extends from -25" to +2S (where S is
the intensive parameter proportional to c u2) and the eigenvaiue density p(E) obeys a
semicircular law [10]:

(8)

with -2S <_E<_ +25. It is proper to add that the idea of random bonds was originally proposed
for explaining magnetic properties of spin glasses (i.e. a magnetic impurity dissolved in a
non-magnetic host [11]). The model predicts the existence of a certain critical temperature Tc
defined as Tc=S/k below which a local order out of randomness may be expected. As it has
been shown using linearised mean field approximation [2, 8, 12] the expression for h/' valid
for T>Tc is given by:

h~i=3'}-"&[<PT>-Ct~j(<P~>-½)] (9)
/

with a random distribution of ISAs in the absence of external stress. In the flame of this
model the response functions J(t) and ,/(co) are described by a continuous spectrum of
relaxation times [2, 8]~

+28 ÷~
J(t) = J~ + r~l j ~ ~ and J(co) J~ i~l I~s~_a~ e~ 0o)
-2S -28

where r corresponds to a noninteracting system. From (10) for d(co) one obtains:
 276

J(co) = Ju + 2T2Js: [1 + fl ~$2 -icor-,4(1 + fl2s~ -icar )2 _ 4fl~s~ ] (11)

which leads to the following expression for the loss tangent curve:

e,= ''' al l (12)

where A=(1-ffSX)Lco2z2 and B=2(l+ffS2)cor. Equation (12) in In(co0 scale (or in "I/T if an
Arrhenius relation for r is assumed) represents a peak positioned approximately at cor=l.
From equations (10) to (12) the non-Debye features of the Shock relaxation can be easily
recoguised. The following variations of the maximum Q'l(cor) occur, if the interaction effect
is enhanced (i.e. by an increase of 176): (i) the peak height decreases, and this decrease is
accompanied by (ii) an increasing broadening and (iii) a shift of the peak position towards
smaller col-values (i.e. higher temperatures). Obviously, for/76'---~ the loss tangent curve
tends to a single Debye process.

4. DISCUSSION

The Snoek relaxation was examined experimentally by applying the static as well as
dynamic measurements In earlier papers it was recognised that J(t) determined for concentrated
alloys (c>0.1 at %) does not correspond to a single exponential process and the loss peak
(Q-t(/)) is much broader than a single Debye curve. These effects were interpreted as the result
of interaction of ISAs. The experimental curves were described in terms of a clustering model
by introducing additional processes arising from small complexes of ISAs (pairs, triplets etc.).
Later, Weller questioned the clustering model and proposed a continuous distribution of
relaxation times which should correspond to long-range interaction models [13, 14].
In order to explain the problem of interaction of ISAs the Snoek relaxation in systems such
as Nb-O [12, 13, 15], Ta-O [16] and Fe-C [17] was carefully examined by dynamic
measurement and the following experimental facts were established:
1. The mean activation enthalpy H does not depend on ISAs content (H values calculated
from the frequency shift of the Snoek maxima between 1 Hz and 1 kHz experiments).
2. The peak temperature of the Shock maximum shifts towards higher temperatures with
increasing ISAs content (maximum about 5 K ).
3. The Shock maxima are broadened with increasing ISAs content.
4. The broadening depends on temperature, since it is less pronounced for the maxima
measured at higher temperatures (kHz range).
It has been shown by making use of a nonlinear fit method [15] that the RCSI model is
able to describe quantitatively all experimental facts listed above. Fig.2 presents the plot In(S)
against In(c) for the data obtained in [12-17]. From this figure it can be recognised that
independent of the examined material the points correlate on one straight line with the slope
0.47 ± 0.02, which is in excellent agreement with the value 0.5 predicted by RCSI model.
Based on these results a hypothesis that, in the first approximation the interaction of ISAs in
bee metals does not depend on the solvent lattice nor, possibly, on the kind of solute atoms
can be formulated [8]. In addition, it also has been shown that for a dilute case (e<0.01 at%)
the experimental data are well described by a single Debye process.
 277

Finally it can be concluded that the RCSI model offers a consistent picture of the
interaction of ISAs in a broad range of concentrations (0<c<l at %) by taking into account the
essential features of many-body interactions. The model explains the fact that the Snoek
relaxation examined for relatively low concentration of ISAs (or high temperatures) in the
fn'st approximation can be treated as a Debye one. The shift of the peak temperature as well as
the broadening may be considered as the secondary characteristics of this relaxation.

3.2

Z8
•• Fe-C f
Ta-O
• Nb-O
slope0.47+0.02

.j
24

t'- ZO

1.6

4.0 -3.5 -RO -Z5 -ZO -1.5 -1.0 -0.5 0.0

In(c[%at])
x Z X

Fig. 10ctahedral sites in bcc lattice Fig.2 Plot of In(S) against In(c) for the data
(x, y, z), and schematically the obtained by applying the RCSI model.
activation enthalpy barrier and the
deformation elipsoid oflSA.

REFERENCES

1. A.S. Nowik, B.S. Berry, Anelastic Relaxation in Crystalline Solids, Academic Press, New
York 1972.
2. G. Haneczok, Interaction of interstitial foreign atoms dissolved in bcc metals, Silesian
University Press, Katowice 1996 (in Polish).
3. M. Weller, J. Phys., Pads, 46 (1985) C10-7.
4. M. Weller, J. Phys., Pads IV, 5 (1995) C7-199;
5. M. Weller, J. Phys., Pads IV, 6 (1996) C8-63.
6. V. Balakrishnan, S. Dattagupta, G. Venkataraman, Phil. Mag. A37 (1978) 65.
7. S. Dattagupta, R. Balakfischnan, R. Ranganathan, J. Phys. F Met. Phys., 12 (1982) 1345.
8. G. Haneczok, Phil. Mag., 78 (1998) 845.
9. S. Dattagupta, J. Phys. F: Met. Phys., 12 (1982) 1363.
10. S.F. Edwards, R.C. Jones, J. Phys. A, 9 (1976) 1595.
11. S. Kirk'patrick, D. Shorington, Phys. Rex,. B, 17 (1978) 4384.
12. G. Haneczok, M. Waller, J. Diehl, Phys. Star. Sol. Co), 172 (1992) 557.
13. M. Waller, G. Haneczok, J. Diehl, Phys. Star. Sol. Co), 172 (1992) 145.
14. M. Weller, J.X. Zhang, G.Y. Li, T.S. KS, J. Diehl, Acta Metal1,, 29 (1981) 1057.
15. G. Haneczok, M. Weller, J. of Less Common Metals, 159 (1990) 269.
16. G. Haneczok, M. Weller, J. Diehl, J. Alloys and Compounds, 211 (1994) 71.
17. G. Haneczok, M. Weller, J. Diehl, Mater. Sci. Forum, 119-121 (1993) 101.