Diffusion Solids Fundamentals Diffusion Controlled Solid State Episode 1 Part 7 pot

23.8.2.2 Activation Volumes of Solute Diffusion For diffusion of interstitial solutes no defect formation term is required.. 8.3 Correlations between Diffusion and Bulk Properties 141∆Vσ:1.

8.2 Pressure Dependence 139vation volumes are considered as evidence (among others) for self-interstitialmediated diffusion in silicon (see [6, 15] and Chap 23).

8.2.2 Activation Volumes of Solute Diffusion

For diffusion of interstitial solutes no defect formation term is required.

Then, from Eqs (8.11) and (8.24) the activation volume is

where V M is the migration volume of the interstitial solute As already tioned, ‘small’ atoms such as C, N, and O in metals diffuse by this mechanism

men-The effect of pressure was studied for C and N in α-Fe, for C in Co and for

C in Ni and for N and O diffusion in V (for references see Chap 10 in [1]).Interstitial diffusion is characterised by small values of the activation volume

For example, for C and N diffusion in α-iron small values between -0.08 and

+0.05 Ω were reported (see Table 8.1) This implies that interstitial diffusion

is only very weakly pressure dependent

Diffusion of substitutional impurities is mediated by vacancies

Ac-cording to Sect 8.1 the diffusivity can be written as

different from the migration volume V M

1V of the vacancy in the pure matrix

Finally, the term C2 arises from the pressure dependence of the solute

cor-relation factor V2M + C2 can be interpreted as the migration volume of thesolute-vacancy complex

The activation volumes for various solutes in aluminium listed in ble 8.1 show a considerable variation As we shall see in Chap 19, transitionmetal solutes are slow diffusers, whereas non-transition elements are normaldiffusers in Al Self-diffusion in Al has been attributed to the simultaneousaction of mono- and divacancies and a similar interpretation is tenable for the

Ta-diffusion of non-transition elements such as Zn and Ge On the other hand,transition elements in Al have high activation enthalpies and entropies ofdiffusion, which can be attributed to a repulsive interaction between vacancyand solute (see Chapt 19) The high activation volumes for the transition met-als diffusers Mn and Co indicate large formation and/or migration volumes

of the solute-vacancy complex [6]

8.2.3 Activation Volumes of Ionic Crystals

The pressure dependence of the ionic conductivity has been studied in eral alkali halide crystals (KCl, NaCl, NaBr, KBr) with Schottky disorder

sev-by Yoon and Lazarus [22] These crystals consist of sublattices of cations

(index C) and anions (index A) In the intrinsic region, i.e at high atures, cation and anion vacancies, VC and VA, are simultaneously present

temper-in equal numbers (Schottky pairs) In the extrtemper-insic region of crystals, dopedwith divalent cations, additional vacancies in the cation sublattice are formed

to maintain charge neutrality (see Chaps 5 and 26):

(i) In the intrinsic region the conductivity is due to Schottky pairs The

formation volume of Schottky pairs is

V SP F : 61± 9 for KCl, 55 ± 9 for NaCl, 54 ± 9 for KBr, 44 for NaBr.

(ii) In the extrinsic region the conductivity is dominated by the motion

of cation vacancies because anion vacancies are less mobile Thus, fromthe pressure dependence of the conductivity one obtains the migration

volume of the cation vacancy, V M

V C The following values have been ported [22]:

re-V M

V C:8± 1 for KCl, 11 ± 1 for NaCl, 11 ± 1 for KBr, 8 ± 1 for NaBr.

Due to the higher mobility of cation vacancies the activation volume of

the ionic conductivity in the intrinsic region, ∆Vσ, is practically given by

∆Vσ = V SP F /2 + V V M C (8.39)

In principle, anion vacancies also contribute to the conductivity (seeChap 26) However, as the anion component of the total conductivity

is usually small this contribution has been neglected in Eq (8.39)

A comparison between activation volumes in metals and ionic crystals withSchottky disorder may be useful In units of the molar volumes of the crystals,

V m, the activation volumes of the ionic conductivity in the intrinsic regionare:

8.3 Correlations between Diffusion and Bulk Properties 141

∆Vσ:1.03 Vm for KCl, 1.28 Vm for NaCl, 1.23 Vm for KBr, 1.37 Vm for NaBr.The activation volumes for intrinsic ionic conduction, which is due to themotion of vacancies, are of the order of one molar volume These values aresimilar to activation volumes of self-diffusion in close-packed metals, wherethe activation volume is also an appreciable fraction of the atomic volume

of the material (see above) In contrast, the migration volumes of cationvacancies are smaller:

V V M C :0.21 Vm for KCl, 0.26 Vm for NaCl, 0.25 Vm for KBr, 0.25 Vmfor NaBr.These values indicate a further similarity between metals and ionic crystals

In both cases, the migration volumes of vacancies are only a small fraction

of the atomic (molar) volume

The α-phase of silver iodide is a typical example of a fast ion conductor

(see Chap 27) The immobile I−ions form a body-centered cubic sublattice,

while no definite sites can be assigned to the Ag+ions In the cubic unit cell 42sites are available for only two Ag+ions Because of these structural features,

Ag+ions are easily mobile and no intrinsic defect is needed to promote their

migration The pressure dependence of the dc conductivity in α-AgI was

studied up to 0.9 GPa by Mellander [23] The activation volume is 0.8 to0.9 cm3mol−1 This very low value can be attributed to the migration of Ag+ions, confirming the view that migration volumes are small

8.3 Correlations between Diffusion and Bulk Properties

Thermodynamic properties of solids such as melting points, heats of melting,and elastic moduli reflect different aspects of the lattice stability It is thusnot surprising that the diffusion behaviour correlates with thermodynamicproperties Despite these correlations, diffusion remains a kinetic propertyand cannot be based solely on thermodynamic considerations In this sec-tion, we survey some correlations between self-diffusion parameters and bulkproperties of the material These relationships, which can be qualified as ‘en-lightened empirical guesses’, have contributed significantly to the growth ofthe field of solid-state diffusion The most important developments in thisarea were: (i) the establishment of correlations between diffusion and meltingparameters and (ii) Zener’s hypothesis to relate the diffusion entropy withthe temperature dependence of elastic constants These old and useful corre-lations have been re-examined by Brown and Ashby [24] and by Tiwari

et al.[25]

8.3.1 Melting Properties and Diffusion

Diffusivities at the Melting Point: The observation that the

self-diffusi-vity of solids at the melting point, D(T ), roughly equals a constant is an old

one, dating back to the work of van Liempt from 1935 [27] But it was notuntil the mid 1950s that enough data of sufficient precision were available

to recognise that D(Tm) is only a constant for a given structure and for

a given type of bonding: the bcc structure, the close-packed structures fccand hdp, and the diamond structured elements all differ significantly Asdata became better, additional refinements were added: the bcc metals were

subdivided into two groups each with characteristic values of D(T m) [26];

also alkali halides were seen to have a characteristic value of D(T m) [31].Figure 8.7 shows a comparison of self-diffusion coefficients extrapolated to themelting point for various classes of crystalline solids according to Brown andAshby[24] The width of the bar is either twice the standard deviation of thegeometric mean, or a factor of four, whichever is greater Data for the solidusdiffusivities of bcc and fcc alloys coincide with the range shown for pure

metals It is remarkable that D(Tm) varies over about 6 orders of magnitude,

being very small for semiconductors and fairly large for bcc metals

At the melting temperature Tm according to Eq (8.6) the self-diffusivity

The constancy of the diffusivity at the melting point reflects the fact that for

a given crystal structure and bond type the quantities D0 and ∆H/(kBTm)

are roughly constant:

Fig 8.7 Self-diffusivities at the melting point, D(T m), for various classes of talline solids according to Brown and Ashby [24]

crys-8.3 Correlations between Diffusion and Bulk Properties 143

The pre-exponential factor D0 is indeed almost a constant According

to Eq (8.5) it contains the attempt frequency ν0, the lattice parameter a, geometric and correlation factors, and the diffusion entropy ∆S Attempt

frequencies are typically of the order of the Debye frequency, which lies inthe range of 1012 to 1013s−1 for practically all solids The diffusion entropy

is typically of the order of a few kB Correlation factors and geometric termsare not grossly different from unity

The physical arguments for a constancy of the ratio ∆H/(kBT m) areless clearcut One helpful line of reasoning is to note that the formation of

a vacancy, like the process of sublimation, involves breaking half the bondsthat link an atom in the interior of the crystal to its neighbours; the enthalpy

required to do so should scale as the heat of sublimation, Hs The migration

of a vacancy involves a temporary loss of positional order – it is somehowlike local melting – and involves an energy that scales as the heat of melting

(fusion), Hm One therefore may expect

that ∆H/(kBTm) should be approximately constant, too.

Activation Enthalpy and Melting Properties: From practical

consider-ations, correlations between melting and activation enthalpy are particularly

useful Figure 8.8 shows the ratio ∆H/(kBTm) for various classes of

crys-talline solids It is approximately a constant for a given structure and bondtype The constants defined in this way vary over a factor of about 3.5 Theactivation enthalpy was related to the melting point many years ago [27–29].These correlations have been reconsidered for metals and alloys by Brownand Ashby [24] and for pure metals recently by Tiwari et al [25] Theactivation enthalpy of diffusion is related via

to the melting temperature (expressed in Kelvin) of the host crystal This

relation is called the van Liempt rule or sometimes also the Bugakov – van Liempt rule [30].

One may go further by invoking the thermochemical rule of Trouton,

which relates the melting point of materials to their (nearly) constant entropy

of melting, Sm Trouton’s rule, Sm = Hm/Tm ≈ 2.3 cal/mol = 9.63 J/mol,

allows one to replace the melting temperatuire in Eq (8.42) by the enthalpy

of melting, Hm Then, the van Liempt rule may be also expressed as

∆H ≈ K1

Fig 8.8 Normalised activation enthalpies of self-diffusion, ∆H/(kB m), for classes

of crystalline solids according to Brown and Ashby [24]

Fig 8.9 Activation enthalpies of self-diffusion in metals, ∆H, versus melting

temperatures, T m, according to Tiwari et al [25]

K1 and K2 are constants for a given class of solids Plots of Eqs (8.42) and(8.43) for metals are shown in Figs 8.9 and 8.10 Values of the slopes for

metals are: K1= 146 J mol −1 K −1 and K2= 14.8 [25].

The validity of Eq (8.42) has been demonstrated for alkali halides by

[31] For inert gas solids and molecular organic solids,

8.3 Correlations between Diffusion and Bulk Properties 145

Fig 8.10 Activation enthalpies of self-diffusion in metals, ∆H, versus melting

enthalpies, H m, according to Tiwari et al [25]

the validity of Eqs (8.42) and (8.43) has been established by Chadwickand Sherwood[32]

The correlations above are based on self-diffusion, which is indeed themost basic diffusion process Diffusion of foreign elements introduces addi-tional complexities such as the interaction between foreign atom and vacancyand temperature-dependent correlation factors (see Chaps 7 and 19) Corre-lations between the activation enthalpies of self-diffusion and substitutionalimpurity diffusion have been proposed by Beke et al [33]

Activation Volume and Melting Point: The diffusion coefficient is

pres-sure dependent due to the term p∆V in the Gibbs free energy of activation The activation volume of diffusion, ∆V , has been discussed in Sect 8.2.

Nachtrieb et al.[34, 35] observed that the diffusivity at the melting point

is practically independent of pressure For example, in Pb and Sn the tice diffusivity, as for most metals, decreases with increasing pressure in such

lat-a wlat-ay thlat-at the increlat-ased melting point resulted in lat-a constlat-ant rlat-ate of

diffu-sion at the same homologous temperature If one postulates that D(Tm) is

independent of pressure, we have

d [ln D(T m)]

Then, we get from Eq (8.6)

∆V = ∆H(p = 0) Tm(p = 0)

dTm

if the small pressure dependence of the pre-exponential factor is neglected.

This equation predicts that ∆V is controlled by the sign and magnitude

of dTm/dp In fact, Brown and Asby report reasonable agreement of

Eq (8.45) with experimental data [24] In general, dT m /dp is positive for

most metals and indeed their activation volumes are positive as well For

plutonium dT m /dp is negative and, as expected from Eq (8.45), the

activa-tion volume of Pu is negative [36]

Later, however, also remarkable exceptions have been reported, which

violate Eq (8.45) For example, dT m /dp is negative for Ge [37], but the

acti-vation volume of Ge self-diffusion is positive [38] (see also Chap 23) Neitherthe variation of the activation volume with temperature due to varying contri-butions of different point defects to self-diffusion nor the differences betweenthe activation volumes of various solute diffusers are reflected by this rule

8.3.2 Activation Parameters and Elastic Constants

A correlation between the elastic constants and diffusion parameters wasalready proposed in the pioneering work of Wert and Zener [39, 40] They

suggested that the Gibbs free energy for migration (of interstitials), G M,represents the elastic work to deform the lattice during an atomic jump

Thus, the temperature variation of G M should be the same as that of an

appropriate elastic modulus µ:

The subscript 0 refers to values at absolute zero The migration entropy S M

is obtained from the thermodynamic relation

S M =− ∂G M

and H M = G M + T S M yields the migration enthalpy In the Wert-Zener

picture both H M and S M are independent of T if µ varies linearly with perature If this is not the case, both S M and H M are temperature dependent

tem-Substituting the thermodynamic relation and G M

0 ≈ H M in Eq (8.46), weget:

At temperatures well above the Debye temperature, elastic constants

usu-ally vary indeed linearly with temperature The derivative Θ ≡ −∂(µ/µ0)/

∂(T /Tm) is then a constant Its values lie between −0.25 to −0.45 for most metals Then H M and S M are proportional to each other and the model ofZener predicts a positive migration entropy

For a vacancy mechanism, Zener’s idea is strictly applicable only to themigration and not to the formation property of the defect One can, however,

References 147

always deduce a diffusion entropy via ∆S = kBln[D0/(gf a2ν0)] from the

measured value of D0 Experimental observations led Zener to extend hisrelation to the activation properties of atoms on substitutional sites:



λ is a constant that depends on the structure and on the diffusion mechanism For self-diffusion in fcc metals λ ≈ 0.55 and for bcc metals λ ≈ 1 The relation

Eq (8.49) is often surprisingly well fulfilled We note that this relation also

suggests that the diffusion entropy ∆S = S M +S Fis positive This conclusion

is supported by the well-known fact that the formation entropy for vacancies,

S F, is positive (see Chap 5)

8.3.3 Use of Correlations

The value of the correlations discussed above is that they allow ties to be estimated for solids for which little or no data are available Forexample, when diffusion experiments are planned for a new material, theserules may help in choosing the experimental technique and adequate thermaltreatments The correlations should be used with clear appreciation of thepossible errors involved; in some instances, the error is small

diffusivi-We emphasise that the correlations have been formulated for self-diffusion.Solute diffusion of substitutional solutes in most metals differs by not morethan a factor of 100 for many solvent metals and the activation enthalpies byless than 25 % from that of the host metal (see Chap 19)

There are, however, remarkable exceptions: examples are the very slowdiffusion of transition metals solutes in Al and the very fast diffusion ofnoble metals in lead and other ‘open metals’ (see Chap 19) Also diffusion

of interstitial solutes (see Sect 18.1), hydrogen diffusion (see Sect 18.2), andfast diffusion of hybrid foreign elements in Si and Ge (see Chap 25) do notfollow these rules

References

1 H Mehrer (Vol Ed.), Diffusion in Solid Metals and Alloys, Landolt-B¨ornstein,Numerical Data and Functional Relationships in Science and Technology, NewSeries, Group III: Condensed Matter, Vol 26, Springer-Verlag, 1990

2 D.L Beke (Vol Ed.), Diffusion in Semiconductors and Non-Metallic Solids,

Landolt-B¨ornstein, Numerical Data and Functional Relationships in Scienceand Technology, New Series, Group III: Condensed Matter; Vol 33,

Subvolume A: Diffusion in Semiconductors, Springer-Verlag, 1998;

Subvolume B1: Diffusion in Non-Metallic Solids, Springer-Verlag, 1999

3 N.L Peterson, J of Nucl Materials 69–70, 3 (1978)

4 H Mehrer, J of Nucl Materials 69–70, 38 (1978)

5 R.N Jeffrey, D Lazarus, J Appl Phys 41, 3186 (1970)

6 H Mehrer, Defect and Diffusion Forum 129–130, 57 (1996)

7 H Mehrer, A Seeger, Cryst Lattice Defects 3, 1 (1972)

8 M Beyeler, Y Adda, J Phys (Paris) 29, 345 (1968)

9 G Rein, H Mehrer, Philos Mag A 45, 767 (1982)

10 R.H Dickerson, R.C Lowell, C.T Tomizuka, Phys Rev 137, A 613 (1965)

11 M Werner, H Mehrer, in: DIMETA-82, Diffusion in Metals and Alloys, F.J.

Kedves, D.L Beke (Eds.), Trans Tech Publications, Diffusion and Defect graph Series No 7, 392 (1983)

Mono-12 J.N Mundy, Phys Rev B3, 2431 (1971)

13 A.J Bosman, P.E Brommer, G.W Rathenau, Physica 23, 1001 (1957)

14 A.J Bosman, P.E Brommer, L.C.H Eickelboom, C J Schinkel, G.W

Ra-thenau, Physica 26, 533 (1960)

15 U S¨odervall, A Lodding, W Gust, Defect and Diffusion Forum 66–59, 415

(1989)

16 E Nygren, M.J Aziz, D Turnbull, J.F Hays, J.M Poate, D.C Jacobson, R

Hull, Mat Res Soc Symp Proc 86, 77 (19?8)

17 A Th¨urer, G Rummel, Th Zumkley, K Freitag, H Mehrer, Phys Stat Sol

20 H Wollenberger, Point Defects, in: Physical Metallurgy, R.W Cahn, P Haasen

(Eds.), North-Holland Physics Publishing, 1983, p.1139

21 R.M Emrick, Phys Rev 122, 1720–1733 (1961)

22 D.N Yoon, D Lazarus, Phys Rev B 5, 4935 (1972)

23 B.E Mellander, Phys Rev B 26, 5886 (1982)

24 A.M Brown, M.F Ashby, Acta Metall 28, 1085 (1980)

25 G.P Tiwari, R.S Mehtrota, Y Iijima, Solid-State Diffusion and Bulk

Prop-erties, in: Diffusion Processes in Advanced Technological Materials, D Gupta

(Ed.), William Andrew, Inc., 2005, p 69

26 N.L Peterson, Solid State Physics 22, 481 (1968)

27 J.A.M van Liempt, Z Physik 96, 534 (1935)

28 O.D Sherby, M.T Simnad, Trans A.S.M 54, 227 (1961)

29 A.D Le Claire, in: Diffusion in Body Centered Cubic Metals, J.A Wheeler,

F.R Winslow (Eds.), A.S.M., Metals Parks, 1965, p.3

30 B.S Bokstein, S.Z Bokstein, A.A Zhukhvitskii, Thermodynamics and Kinetics

of Diffusion in Solids, Oxonian Press, New Dehli, 1985

31 L.W Barr, A.B Lidiard, Defects in Ionic Crystals, in: Physical Chemistry –

an Advanced Treatise, Academic Press, New York, Vol X, 1970

32 A.V Chadwick, J.N Sherwood, Self-diffusion in Molecular Solids, in: Diffusion

Processes, Vol 2, Proceedings of the Thomas Graham Memorial Symposium,

University of Strathclyde, J.N Sherwood, A.V.Chadwick, W.M Muir, F.L.Swinton (Eds.), Gordon and Breach Science Publishers, 1971, p 472

33 D Beke, T Geszti, G Erdelyi, Z Metallkd 68, 444 (1977)

34 N.H Nachtrieb, H.A Resing, S A Rice, J Chem Phys 31, 135 (1959)

References 149

35 N.H Nachtrieb, C Coston, in: Physics of Solids at High Pressure, C.T.

Tomizuka, R.M Emrick (Eds.), Academic Press, 1965, p 336

36 J.A Cornet, J Phys Chem Sol 32, 1489 (1971)

37 D.A Young, Phase Diagrams of the Elements, University of California Press,

1991

38 M Werner, H Mehrer, H.D Hochheimer, Phys Rev B 32, 3930 (1985)

39 C Wert, C Zener, Phys Rev 76, 1169 (1949)

40 C Zener, J Appl Phys 22, 372 (1951)

41 C Zener, in: Imperfections in Nearly Perfect Crystals, W Shockley et al (Eds.),

John Wiley, 1952, p 289

In this chapter, we consider the diffusion of two chemically identical atomsthat differ in their atomic masses Their diffusivities are different and this dif-

ference is denoted as isotope effect The isotope effect, sometimes also called

the mass effect, is of considerable interest It provides an important mental means of gaining access to correlation effects Correlation factors ofself- and solute diffusion are treated in Chap.7 and values for correlation fac-tors of self-diffusion in several lattices and for various diffusion mechanismsare listed in Table 7.2 Correlation factors of solute diffusion are the subject

experi-of Sect 7.5 We shall see below that the isotope effect is closely related to thecorrelation factor Since correlation factors of self-diffusion often take valuescharacteristic for the diffusion mechanism, isotope effects experiments canthrow light on the mechanism

9.1 Single-jump Mechanisms

Let us consider two isotopes α and β of the same element labelled by their isotopic masses m α and m β Because of their different masses, the two iso-topes have different diffusion coefficients in the same host lattice For self-and impurity-diffusion in coordination lattices the tracer diffusivities can bewritten as:

D ∗

α = Aω α f α , and D ∗

The quantity A contains a geometrical factor, the lattice parameter squared,

and for a defect mechanism also the equilibrium fraction of defects or the

defect availability next to the solute The atom-defect exchange rates ωα

or ωβ are factors in Eq (9.1) The correlation factors for vacancy-mediateddiffusion in fcc, bcc, and diamond lattices according to Eq (7.46) have the

same mathematical form, sometimes called the ‘impurity form’ :