You are on page 1of 19

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259649176

Hot Topics in Thermal Analysis and Calorimetry

Chapter · January 2013


DOI: 10.1007/978-90-481-3150-1_3

CITATIONS READS

6 148

2 authors:

Pavel Holba David Sedmidubský


University of West Bohemia University of Chemistry and Technology, Prague
80 PUBLICATIONS   606 CITATIONS    287 PUBLICATIONS   3,761 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Thermodynamics of partly open systems View project

Optically active centers in ZnO and zirconia crystals View project

All content following this page was uploaded by David Sedmidubský on 30 July 2015.

The user has requested enhancement of the downloaded file.


Chapter 3
Crystal defects and nonstoichiometry
contributions to heat capacity of solids

P. Holba1, D. Sedmidubský
1
Independent scientist, member of working group on Thermal Analysis of the Czech Chemical Society, K Lipám 293, CZ-19000
Praha, e-mail: holbap@gmail.com
1
New Technology — Research Centre in the Westbohemian Region, West Bohemian University, Universitní 8, CZ-30114
Pilsen, Czech Republic

3.1 Introduction

Heat capacity is one of the essential thermophysical characteristics determining the thermal behavior of chemical
substances. Its temperature dependence reflects all the excitations that the substance undergoes when heated. As
such it relates to essential thermodynamic quantities such as (standard) enthalpy and entropy which are used in
thermodynamic descriptions of both stoichiometric compounds and their solutions. Considering the third law of
thermodynamics defining the absolute entropy, the heat capacity as a function of temperature represents, along
with the enthalpy of formation, the only information necessary to specify the thermodynamic behavior of a stoi-
chiometric compound at constant pressure by constructing the Gibbs energy function. It can be also used to eval-
uate the enthalpy of formation and entropy from high temperature equilibrium data using the second law of
thermodynamics.
At low temperatures, the heat capacity measurements provide a valuable tool to analyze various physical phe-
nomena such as electronic specific heat coefficient in strongly correlated systems, magnetic ordering transitions
and magnon dynamics, spin transitions, charge and orbital ordering effects as well as transitions between local-
ized electronic states (as it is explained e.g. in [1]).
In the high temperature range, the knowledge of heat capacity is essential for modeling the thermal behavior
of numerous construction materials applied at elevated temperatures, refractories and nuclear fuel materials.
Note that the second essential material characteristics relevant in this kind of studies, the heat conductivity, is
frequently derived from the measured thermal diffusivity using heat capacity data obtained in an independent
calorimetry measurement.
The high temperature heat capacity involves some additional contributions being absent at low temperatures
such as anharmonic effects giving rise to difference in heat capacity measured at constant pressure and volume
and a contribution due to formation of various crystal defects (vacancies, interstitials, Schottky defects, Frenkel
pairs, anti-site defects etc.). The heat capacity associated with crystal defect formation represents in fact a partic-
ular case of a more general contribution due to homogeneous equilibria evolving with temperature and being ac-
companied by individual endothermic effects. The energetics connected with crystal defect formation and its in-
fluence on heat capacity has been discussed in a number of textbooks and papers, particularly focused on metals
and semiconductors. However, relatively little attention has been paid to systems where the defect formation in-
volves a simultaneous change of the overall stoichiometry of one or more components. As a typical example let
us mention the mixed valence oxides that exchange oxygen with the surrounding atmosphere upon heating or
cooling in a dynamic atmosphere (with a fixed activity of the shared component). Although the effect of
nonstoichiometry of a shared component on the heat capacity enhancement has been experimentally documented
for several oxides (see e.g. [2] and nitrides, a detailed thermodynamic analysis of this phenomenon is, to our
knowledge, still missing. The relevance of this analysis stems, among others, from the fact that it is experimen-
tally very difficult to ensure the fixed content of shared component during high temperature calorimetric meas-
urements of those partly open systems. Just as much important as the analysis of nonstoichiometry contribution
to heat capacity itself is the proper separation of other contributions, in particular the phonon and electronic part.
These are examined in the next section.
2

3.2 Concepts and models of heat capacity

Heat capacity is in general defined as the quantity of heat needed to enhance the temperature of a unit amount
(mass unit, mole) by a unit degree of temperature. Depending on the character of energy exchange between the
system under study and its surroundings we distinguish the heat capacity at constant volume

 ∂U 
CV [isolated system ] =   (3.1)
 ∂T V ,{Ni }= const

applicable to an isolated system maintained under isochoric conditions ensuring that all supplied heat is trans-
formed into an increase in internal energy U, and the heat capacity at constant pressure

 ∂H 
CP [ closed system ] =   (3.2)
 ∂T  P ,{Ni }= const

relevant for closed systems in which the supplied heat is spent at expense of an external work done against the
constant pressure and an increase of internal energy, where both these contributions constitute the change of en-
thalpy H. The difference between CP and CV can be shown to assume the value of universal gas constant R for
ideal gases, while for real gases as well as for liquids and solids a general relation holds

T ( ∂V ∂T ) P
2

CP − CV =
− TV α 2 β
= (3.3)
( ∂V ∂P )T

where α = V–1(∂V/∂T)P is the isobaric coefficient of thermal expansion and β = –V–1(∂V/∂P)T is the isothermal
coefficient of pressure compressibility (the volume V is related to the same unit amount as the heat capacity).
Except for very low temperatures, the heat capacity of vast majority of substances is dominated by a contribu-
tion of lattice vibrations (phonons). The most widely used approach to describe the phonon spectrum of a crys-
talline solid is the harmonic crystal approximation assuming a quadratic dependence of the lattice potential on
the individual atomic displacements [3]. In such a case the crystal lattice reveals an infinite thermal conductivity
and no thermal expansion. Hence, the latter effect implies that the heat capacities at constant volume can be fair-
ly well described within the harmonic crystal model,

νmax 2
 hν  e hν / kBT
CV = Char = R ∫
0
  hν / kBT
 kBT  [e − 1]2
g (ν )dν (3.4)

where R, h, and kB are, respectively, the universal gas, the Planck and the Boltzmann constants, ν is a specific
phonon frequency and g(ν) is the pertinent phonon density of states (PDOS). Since the x2ex/(ex – 1)2 part of the
integrand goes to unity for x → 0 and the integral ∫ g (ν )dν = 3 N , where N is the number of atoms per primitive
unit cell, the heat capacity Char approaches the value of 3NR in the limit of high temperatures. This is referred to
as Dulong–Petit limit (DPL) discovered firstly in 1819 [4]. In fact the real temperature range where the heat ca-
pacity approaches DPL also depends on the respective phonon spectrum. Heavy and weakly bonded atoms give
rise to PDOS situated in low frequency region and Char thus quickly saturates towards DPL. By contrast, strong
bonds between light atoms result in large force constants and large dynamical matrix elements and, consequent-
ly, PDOS extends to high frequencies and the heat capacity grows slowly. The PDOS can be directly derived
(performing the summation over the first Brillouin zone [5]) from the phonon band structure, which can be either
experimentally determined (inelastic neutron scattering) or calculated from the first principles (direct method
from Hellmann–Feynman forces [6, 7]). If PDOS is not explicitly available, various models and approximations
are to be applied.
3

Every phonon spectrum consists of three acoustic branches and (3N – 3) optical bands. In the long-
wavelength limit the acoustic branches exhibit a linear dependence of frequency on the wave vector ν = C |q| (as-
suming a constant value of velocity C and applicable for low values of |q|) corresponding to a quadratic form of
g(ν) = 6Vcν2/C3. This is the basis of Debye model [8] which, in its simplest form, results from cutting-off the
quadratic PDOS function g(ν) at the Debye frequency νD implied by the normalization condition ∫ν D g (ν )dν = 3
and substituting into Eq. 3.4. On the other hand, the PDOS of narrow optical bands can be approximated a delta
function δ(ν – νE) which is essentially the basis of the Einstein model assuming the atoms to vibrate as independ-
ent harmonic oscillators at a characteristic frequency νE . Optionally, if the bands are broader, the respective part
of g(ν) can be replaced by a bar-shaped function. Combining the Debye and Einstein model to describe the
acoustic and optical branches, respectively, we get

3 xD 3 N −3
 T  x4ex xE2i e xEi
Char = 9R   ∫ ( e − 1) dx + R ∑ (3.5)
 ΘD  (e )
2 2
0
x
i =1
xEi
−1

where xD = θD/T and xEi = θEi/T with θD = hνD/kB and θEi = hνEi/kB being the characteristic Debye and Einstein
temperatures which can be considered as free parameters. To reduce the number of parameters to be fitted, the
Einstein modes can be grouped so that the essential features of the phonon spectrum and their weights are repro-
duced. The hybrid Debye–Einstein model [9] is not only suitable for the analysis of experimental data. Since the
characteristic temperatures reveal clear trends within the isostructural series of compounds reflecting the de-
pendence of lattice dynamics on atomic masses and strengths of individual bonds, it can be used to estimate the
heat capacities in cases the experimental data are not available.
As mentioned, if we want to involve the effects of anharmonicity we need to consider higher than quadratic
powers in the expansion of lattice potential with respect to atomic displacements. This gives rise to thermal ex-
pansion of the lattice and CP – CV difference. Note that for most solids CP – CV becomes relevant for tempera-
tures higher than 700–800 K. The problem to calculate CP is usually treated in terms of quasiharmonic approxi-
mation which retains the harmonic model scenario for the calculation of PDOS, however, PDOS is varied with
volume and the vibrational free energy

νmax
  hν 
Fvib (V , T ) = RT ∫
0
g (ν ,V ) ln  2sinh 
  2kBT
  dν

(3.6)

obtained from PDOS is minimized with respect to V for each temperature. Hence, from V(T) we get the coeffi-
cient of thermal expansion α(T) and from Fvib(V,T) the bulk modulus B(T) (equal to reciprocal value of isother-
mal compressibility β for isotropic crystals)

 ∂ 2 (U + Fvib ) 
B (=
T ) 1=
β V  (3.7)
 ∂V 2 
 T

which can be substituted into Eq. 3.3.


Aside from the phonon part discussed above some additional contributions to heat capacity are to be consid-
ered in specific cases. These involve, among others, electronic specific heat of conduction electrons, magnetic
contributions and Schottky type anomalies [10] due to nuclear or electron magnetic moments.
The electronic heat capacity of metals and alloys, i.e. systems with itinerant electrons, is commonly described
by the well known Sommerfeld formula (see e.g. [11])

2π 2
C=
el γ=
T N ( EF )kB2 T (3.8)
3
4

where N(EF) is the density of electronic states at Fermi level EF and γ is the electronic specific heat coefficient.
However, this expression neglects some important features. First of all, since the energy interval around EF
probed by the electronic heat capacity is of the order ~4kBT, the assumption of constant density of states appears
to be a poor approximation at high temperatures for systems with N(E) rapidly varying with E. Second, EF in Eq.
3.8 should be rather replaced by a temperature dependent chemical potential μ = μ(T) determined by the condi-
tion that the total number of electrons is conserved. Once again, the variation of μ with T depends on the slope of
N(E), since, to lowest order, μ(T) – EF = –(π2/6)(kBT)2N’/N. Hence, a more general formula reads

( E − µ )2  ∂f 
Cel = 2kBT ∫ kBT
N (E )  −  dE
 ∂E 
(3.9)

with f(E) = [exp((E – μ)/kBT) + 1]–1 being the Fermi–Dirac distribution function. Furthermore, the density of
states appearing in Eq. 3.8 may be renormalized due to electron-phonon (and other) many-body effects, which
generally contribute to Cel enhancement at low temperatures while in the high temperature limit Cel tends to the
normal “free-electron” model.
In magnetic materials the magnetic moments are excited into spin waves (magnons) analogous to lattice vi-
brations, since both magnons and phonons obey the same Bose–Einstein statistics. Using the corresponding spin
wave frequency dispersion relations the magnon contributions to heat capacity can be derived being proportion-
al, in the low temperature limit, to CM ∝ T3/2 and CM ∝ T3 for ferromagnetic and anti-ferromagnetic arrange-
ments, respectively. Unfortunately, the spin wave model based on collective excitations is not applicable up to
the critical temperature TC as it apparently underestimates the total magnetic entropy. Order–disorder models
predicting a logarithmic dependence CM ∝ log(TC – T) have thus to be applied.
Last, consider a system in which the particles can exist in a group of discrete energy levels (with degeneracies
gi) separated from the ground state by specific energies εi. The corresponding contribution to heat capacity is
sometimes termed the Schottky effect (or Schottky anomaly) and for a two level system it forms a characteristic
hump superimposed on the lattice and other contributions:

2
 ε  g exp( ε1 kBT )
CSh = R  1  0 (3.10)
 B  1 [1+ ( g 0 g1 ) exp( ε1 kBT ) ]
2
k T g

If more levels are involved,

d2 z
CSh = R T 2 (3.11)
d(1/T ) 2

=
where z ∑ gi exp(− ε i kBT ) represents the partition function. The Schottky effects are observed in
paramagnetism of crystalline salts, localized f-electron transitions in rare-earths compounds and for nuclear
magnetic state splitting (hyperfine structures). Moreover, formally the same behavior of heat capacity is ob-
served as a result of various crystal defects formation. This issue will be addressed in more detail and applying a
more elaborated formalism in the following sections.

3.3 Isoplethal and isodynamic heat capacities (at constant pressure)

The quantity called isobaric heat capacity CP is known from phenomenological thermo-dynamics, where it is de-
fined for closed systems (systems of constant composition) as a partial derivative of enthalpy H with respect to
temperature T at a constant pressure P (and constant composition). For an n-component system comprising
N1, ..., Ni, ..., Nn moles of components 1, ... i, ... n (arithmetic vector of chemical composition could be represent-
ed as {Ni}) the quantity CP is therefore defined by Eq. (3.2).
5

However, under the conditions of TA experiments the composition of a sample is generally not conserved at
constant value and the decomposition processes can be studied under the so-called „dynamic (controlled) atmos-
phere“ – in a flow of gaseous mixture with defined activities a f1 , a f2 , ... of volatile (free) components f1 , f 2 , ...
(the presence of at least one volatile component in the sample is a condition for obtaining a non-uniform curve of
thermogravimetry). The heat capacity measured under controlled atmosphere (assuming the sample with only
one free/volatile component) should be defined respecting the fact, that not the content Nf but the activity af of
the volatile component f is kept constant and called “isodynamic” [12] heat capacity (the arithmetic vector {Nc}
means a vector of mole amounts of (n – 1) components other than the free component f)

 ∂H 
CP , a f =   (3.12)
 ∂T  P ,{Nc }, a f = const

in contrast to “isoplethal” (“isocompositional”) heat capacity defined by Eq. (3.2).


The equation for the “isodynamic” (“isoactivital”) heat capacity could be obtained as a partial derivative of a
composite function:

(
 ∂H T , N f (T )
CP , a f = 
) 
 ∂T 
 
(3.13)
 ∂H   ∂H   ∂N f 
=   +  ⋅ 
 ∂T  P ,{Nc }, N f = const  ∂N f 
 P ,T , Nc≠ f  ∂T  P , Nc≠ f , a f

which can be re-written as

 ∂N f 
C= CP , N f + h f   (3.14)
 ∂T  P , a f
P,a f

where hf means partial molar enthalpy of a component f consisting of two terms: H 0f denoting the molar en-
thalpy of pure component f (at the standard state) and Δhf denoting the relative partial molar enthalpy of com-
ponent f

 ∂H 
hf ≡   = H 0f + ∆h f (3.15)
 ∂N f 
  P ,T , Nc≠ f

Let us note that a relation analogous to (3.15) is valid for any partial molar quantity including the partial molar
Gibbs free energy gf which is identical with the chemical potential μf :

 ∂G 
gf ≡   = G 0f + ∆g f ≡ µ= µ 0f + RT ln a f (3.16)
 ∂N f  f
  P ,T , Ni≠ f

Applying the Gibbs–Helmholtz equation

∂ (G T )
=H (3.17)
∂ (1 T )
6

to equation (3.16):

(
∂ gf T
= h=
)H 0f + R
∂ ln a f
(3.18)
∂ (1 T ) ∂ (1 T )
f

the relation for Δhf is obtained:

∂ ln a f
∆h f = h f − H 0f = R (3.19)
∂ (1 T )

Substituting (3.19) and (3.15) into (3.14) the following relation between “isodynamic” CP , a f and “isoplethal”
CP , N f isobaric heat capacities is obtained:

 ∂ ln a f   ∂N f 
CP , a f = CP , N f +  H 0f + R ⋅ 
 ∂ (1 T )   ∂T  P , a
f
(3.20)
 ∂N f 
(
= CP , N f + H 0f (T ) + ∆h f T , N f ( )) ⋅ 
 ∂T  P , a f

If we denote the second term in (3.20) as a “saturation contribution” to heat capacity

 ∂N f   ∂N 

= sat C P hf  =   H 0f (T ) + ∆h f (T , N f )  ⋅  f  (3.21)
 
 ∂T  P , a f  ∂T  P , a f

then the relation between isodynamic and isoplethal heat capacity simply reads

C=
P,a f CP , N f + ∆ sat CP (3.22)

3.4 Heat capacity as a function of deviation from stoichiometry

Assume the isoplethal heat capacity CP , N 0 of n-component condensed phase φ containing N1... Nn–1 moles of
f

conservative components (composition vector {Nc}) and a single free component (f) whose content is allowed to
vary (at a temperature T) from N Lf to N Hf around the stoichiometric composition N 0f such that
N Lf ≤ N 0f ≤ N Hf . The question now arises how to estimate the difference ΔdevCP, between the isoplethal heat ca-
pacity of a nonstochiometric composition, CP , N ≠ N 0f
, and that of the stoichiometric composition, CP , N = N 0f
, of a
f f

phase φ

∆ dev CP = CP , N f − CP , N 0 (3.23)
f
7

It is clear that the difference should be an integral (from N 0f to N f ) of some quantity expressing the dependence
of heat capacity on the free component content (where the derivative ∂CP ∂N f ≡ (∂CP ∂N f )T , P ,{ Nc } is known
also as the partial molar heat capacity of a component f)

 ∂CP 
∆ dev CP =  ∫
 ∂N f
 dN f

(3.24)
N 0f  

If CP is defined as a temperature derivative of enthalpy (at constant pressure) then using an analogy to the Max-
well relations the following relation is valid:

 ∂CP  ∂2 H ∂2 H ∂h f
=  = = (3.25)
 ∂N f  ∂T ∂N f ∂N f ∂T ∂T
 

which can be modified using (3.19) into

 ∂CP  ∂H 0f ∂∆h f ∂∆h f


 = + =CP0 , f + (3.26)
 ∂N f  ∂T ∂T ∂T
 

where CP0 , f means the molar heat capacity of pure component f in its standard state.
Equation (3.24) is then transformed into:

∂h f ∂∆h f
∆ dev CP = ∫ ∂T
( )
dN f = N f − N 0f CP0 , f + ∫ ∂T
dN f (3.27)
N 0f N 0f

where the first term on the right hand side refers merely to the heat CP0 , f of free component in gaseous state,
while the other term is connected with changes in mechanism providing the deviation from stoichiometry and
with the change in number of vibrational modes associated with atoms released from or incorporated into the
condensed phase.

3.5 Heat capacity as a function of (intrinsic) equilibrium dissociation

The Gibbs free energy G of an E-component homogeneous phase φ consisting of several species (ions, mole-
cules, crystal defects etc.) under conditions of closed system is given as a sum

=G ∑
= n µ
j
∑ n (G
j j
j
j
0
j + RT ln a j ) (3.28)

where the M-component vector n1 , ..., n j , ..., nM or {nj} represents the amounts of individual species giving mi-
croscopic composition of the homogeneous phase φ and other M-component vector a1 , ..., a j , ..., aM or {aj} rep-
resents activities of the mentioned species.
The amounts of species should satisfy the following balances: a) constant amounts of chemical elements (in-
variable macroscopic composition – E balances); b) electroneutrality condition (one balance) and c) constant ra-
tio of sites belonging to different crystal sub-lattices (S – 1 balances for crystal phase with S different sub-
lattices) [13].
8

Applying the above-mentioned balances the amounts nj of (E + S) species can be expressed in terms of the
amounts of the remaining R (R = M – E – S) species n1 ≡ λ1, ..., nr ≡ λr, ..., nR ≡ λR, which are also denoted as the
degrees of conversion (vector {λr}) and in terms of nonstoichiometry γf

γ=
f N f − N 0f (3.29)

as their linear function

(
n j λ1 ... λr ... λR , γ f ≡ n j
nj = ) ν j 0 + ∑ν jr λr + ν jf γ f
({λr } , γ f ) = (3.30)
r

(it should be noted that for j = r the identity relation is valid: nj = λr) and the amounts of these species are deter-
mined from the conditions of homogeneous equilibrium:

 ∂Gϕ 
  =
∆Gr =
∆Gr0 + RT ln K r =
0 (3.31)
∂λ
 r λs≠r ,γ f

where ΔGr is the Gibbs free energy change corresponding to homogeneous reaction (Rr) in the phase φ, ∆Gr0
stands for the standard Gibbs free energy of the reaction (Rr)

 ∂n j 
=∆Gr0 ∑ G=
j
 
 ∂λ 
∑ν
0
j
r λs ≠ r j
0
jr G j (3.32)

where νjr is obtained from (3.30), and Kr is the equilibrium constant of reaction (Rr)

∂
 ∑(n ln a j 
j

)   ∂ ln a j  
=ln K r 

j
=
∂λr


∑ ν jr ln a j + n j 

 
 ∂λr λs≠r 
(3.33)
  j 
 λs≠r

Assuming the activity of jth species being proportional to its amount nj

aj = k jnj (3.34)

the following relation is found:

 ∂ ln a j   ∂ ln n j   ∂ ln k j 
ajnj   = nj   + nj   =
 ∂λr λs≠r  ∂λr λs≠r  ∂λr λs≠r
(3.35)
n j  ∂n j  n j  ∂k j  n j  ∂k j 
=   +   ν jr + 
= 
n j  ∂λr λ k j  ∂λr λ k j  ∂λr λ
s≠r s≠r s≠r

so that

  ∂ ln k j  
=
ln K r ∑ 
(
ν jr ln n j + ln k j + 1 + n j  )  
 ∂λr λs≠r 
(3.36)
j 
9

which could be simplified assuming the coefficients kj are nearly independent of λr

∂ ln k j
ν jr ( ln n j + ln k j + 1)  n j (3.37)
∂λr

into a form

=
ln K r ∑ν j
jr ln n j + ln K 0 r (3.38)

where the values of coefficients νjr (in equations (3.32) and (3.34)) are either negative (species consumed by re-
action Rr) or positive (species produced in reaction Rr) or zero (species not participating in reaction Rr) and

=
ln K 0 r ∑ν ( ln k + 1)
j
jr j (3.39)

Substituting (3.30) into (3.38) the following relation for logarithm of equilibrium constant is found:

 
ln K r =ln K 0 r + ∑ν
j
jr ln ν j 0 +


∑ν r
jr r λ + ν jf γ f 

(3.40)

The sum in equation (3.28) can be then rearranged into:

G = G id + ∑ ∆G λ
r
r r = G id + ∑ λ (∆G
r
r
0
r + RT ln K r ) (3.41)

where Gid represents the Gibbs energy of the phase φ consisting of minimum number of distinguishable species –
it contains only one species of molecules or a species corresponding to an ideally occupied crystal lattice (with-
out vacancies, interstitials and substituent ions). It is surprising the Gibbs free energy in homogeneous equilibri-
um is independent on the degree of conversion of any internal reaction as it is due to the validity of equation
(3.31) – zero value of any ΔGR at equilibrium.
However, this invariance with respect to the degrees of conversion does not apply for the enthalpy Hφ of the
phase φ. Dividing equation (3.41) by temperature T and applying the Gibbs–Helmholtz equation (see (3.17)), the
following relation is obtained:

∂ (G T )  ∂ ln K r 
H=
∂ (1 T )
= H id + ∑ λ ∆H
r
r r = H id + ∑ λ  ∆H
r
r
0
r +R 
∂ (1 T ) 
(3.42)

and for the heat capacity the following equation is found

 ∂H   ∂λr   ∂∆H r 
CP =   = CP +
 ∂T  P
id
∑ ∆H
r
0
r  ∂T  +
 P
∑ λ 
r
r
∂T 
(3.43)

where the last sum can be neglected assuming

 ∂∆H r   ∂λ 
λr    ∆H r0  r  (3.44)
 ∂T   ∂T  P
10

so that

 ∂λr 
CP = CPid + ∑ ∆H
r
0
r  ∂T 
 P
(3.45)

For such a homogeneous reaction the degree of conversion (expressed as the mole number of one defect species)
is given as

 ∆Gr0   ∆Sr0   ∆H r0   ∆H r0 
λr ≅ exp  − =
 Lr exp   ⋅ exp  − =
 Ar exp  −  (3.46)
 RT   R   RT   RT 

where Lr is the multiplying constant (including K0r (see (3.39)) and νj0 in (3.40) and other constants following
from chosen form of balances) and

 ∆S 0 
Ar ≡ Lr exp  r  (3.47)
 R 

so that for a temperature derivative (∂λr/∂T) the following relation is obtained:

∂λr −∆H r0 −1  ∆H r0  ∆H r0  ∆H r0 
= Ar ⋅ 2 ⋅ exp  − =  A exp  −  (3.48)
∂T
r
R T  RT  RT 2  RT 

and the contribution of one dissociation (or defect formation) reaction (Rr) to heat capacity of crystalline phase is
given by equation

∆ r CP =CP − CPid
2
 ∂λ  Ar  ∆H r   ∆H r0  (3.49)
0
=
∆H r0  r =   exp  − 
 ∂T  R T   RT 

similar to equation (8.51) in [14].


This contribution should be considered at any (stoichiometric or nonstoichiometric) crystalline solid phases
due to point defects formation (see [15]) e.g.
∆H r0
null ⇒ VaMg (periclas) + VaO (periclas) [Schottky] 7.7 eV (740 kJ/mol)
TiTi (rutil) ⇒ Tii (rutile)+ VaTi (rutile) [Frenkel] 12.0 eV (1160 kJ/mol)
OO (rutil) ⇒ Oi (rutile) + VaO (rutile) [Frenkel] 8.7 eV (840 kJ/mol)
null ⇒ e– (periklas) + h+ (periklas) [electron-hole] 6.3 eV (606 kJ/mol)
at 1400 °C
and also in fluids (gases and liquids) e.g.
∆H r0
CO2(g) ⇒ CO(g) + ½ O2(g) 2.9 eV (283 kJ/mol)
N2O4(g) ⇒ 2 NO2(g) 0.6 eV (57 kJ/mol)
H2(g) ⇒ 2 H(g) 4.5 eV (430 kJ/mol)
2 H2O (L) ⇒ H3O+(L) + OH–(L) 0.8 eV (80 kJ/mol)
2 NH3 (L) ⇒ NH4+(L) + NH2–(L) 1.2 eV (115 kJ/mol)
11

Let us note that the equation (3.49) is valid only in cases of one dominating intrinsic reaction, i.e. when no
other parallel intrinsic reaction occurs (in a given temperature range) and for ∆H r0  RT .

3.6 Equilibrium nonstoichiometry (deviation from stoichiometry)

The evaluation of the saturation contribution, ΔsatCP (Eq. 3.21), and the contribution due to deviation from stoi-
chiometric content of a free component f, ΔdevCP (Eq. 3.27), requires the knowledge of functional dependences
for the partial molar enthalpy or the relative partial molar enthalpy of free component

hf =h f (T , N f ) =h f (T , γ f ), ∆h f =∆h f (T , N f ) =∆h f (T , γ f ) (3.50)

and eventually

= γ f γ f (T , a f )
N f N f (T , a f ) ⇒= (3.51)

The equilibrium content of free component f in the system under isodynamic conditions (af = const) can be de-
termined from the condition of minimum of hyperfree energy Zf which has been shown to be effectively applied
as a thermodynamic potential for partially open systems [16]. For a system with a single free component f it is
defined as the Legendre transformation of the Gibbs free energy with respect to free component content

G − Nf µf =
Zf = G − N f (G 0f + RT ln a f ) (3.52)

The minimum condition (at constant T, P, Ni≠f ,{λr}) results in

 ∂Z f   ∂Z f 
  ≡ 
 ∂N f   
 T , P , Ni≠ f ,λr  ∂γ f T , P , Ni≠ f ,λr
(3.53)
 ∂G 
=   − G=
f − RT ln a f
0
0
 ∂γ f 
 T , P , Ni≠ f ,{λr }

Starting from the expressions for G (3.28) and for nj (3.30) the following equation is found:

 
∑ν 0
fj Gj − G 0f + RT 
 ∑ν fj ln a j − ln a f  =

0 (3.54)
j  j 

which can be understood as the equilibrium condition for the incorporation reaction of the free component from
surroundings into the phase φ (reaction If)

∆GI f =
∆GI0f + RT ln K I f =
0 (3.55)

where

∆=
GI0f ∑νj
0
fj Gj − G 0f (3.56)
12

and

ln=
KI f ∑ν j
fj ln a j − ln a f ≈ ∑ν j
fj ln n j − ln a f + ln K 0 I f (3.57)

The equation (3.55) expresses implicitly the relation between the activity af and the equilibrium
nonstoichiometry γf (or free component content Nf) since upon substituting (3.40) we obtain

∆GI0f  
Φ (T , γ f ) ≡
RT
+ ln K 0 I f + ∑ν
j
fj ln ν 0 j +


∑νr
rj λr + ν f j γ f  − ln a f

(3.58)
=0

and for the partial derivative we get

 ∂N f   ∂γ f  ∂Φ ∂T ∆GI0f RT 2
  ≡   =
− = (3.59)
 ∂T a f  ∂T a f ∂Φ ∂γ f ν 2f j n j ∑ j

3.7 Dependence of the relative partial molar enthalpy on the free component content

It would seem that the enthalpy of incorporation reaction entirely determines the partial molar enthalpy of the
free component f. However, the choice of species selected as degrees of conversion λr is arbitrary, so that the in-
corporation reaction can include quite different species. Then, the partial molar Gibbs free energy gf (see equa-
tion (3.16)) should be derived from (3.28) by differentiating G as a composite function:

gf ≡ 
(
 ∂G γ f ,{λr (γ f )} )   ∂G
≠


 ∂γ f   ∂γ f 
 T , P , Ni≠ f  T , P , Ni≠ f ,{λr }

(
 ∂G γ f ,{λr } )   ∂G   ∂G   ∂λr 
gf ≡ 
 ∂γ f 
=
 
 ∂γ f 
 T , P , Ni≠ f ,{λr }
+ ∑  ∂λ  
r   ∂γ f


λs≠r
(3.60)
 T , P , Ni≠ f r

 ∂λr 
=
∆GI f + G 0f + RT ln a f + ∑ ∆G  ∂γ
r
r 

f λ
s≠r

Respecting that g=
f G 0f + RT ln a f (see (3.16)) and substituting (3.55) and (3.57) into (3.60) the following
equation is obtained

   ∂λr 
∆GI0f + RT 
0=
 ∑ν fj ln n j + ln K 0 I f  − RT ln a f +
 ∑ ∆G  ∂γ r 

f λ

 j  r
s≠r
(3.61)
   ∂λ 
⇒ RT ln a f =
∆GI0f + RT 
 ∑ν fj ln n j + ln K 0 I f +
 ∑ ∆Gr  r

 ∂γ f


λs≠r
 j  r
13

Dividing (3.61) by temperature T, differentiating with respect to (1/T) and using the Gibbs–Helmholtz equation
(see (3.17)), we obtain

 
∂ ln a f
∂
 ∑ν fj ln n j + ln K 0 I f 
  ∂λ 
+R  +
∑ ∆H
j
R =
∆H I0f  r  (3.62)
∂ (1 T ) ∂ (1 T )
r  ∂γ f 
r  λs≠r

Neglecting the second term on the right side provided that

 
∂
 ∑ν fj ln n j + ln K 0 I f 
  ∂λr 
R    ∆H 0 +
∑ ∆H
j
  (3.63)
∂ (1 T )
If r  ∂γ f 
r  λs≠r

and assuming

∆H r ≅ ∆H r0 (3.64)

equation (3.62) is reduced (including (3.19)) into

∂ ln a f  ∂λ 
∆h f ≡ R
∂ (1 T )
≅ ∆H I0f + ∑ ∆H
r
0
r  r
 ∂γ f



λs≠r
(3.65)

where the partial derivatives (∂λr ∂γ f )λs≠r can be estimated from a relation for equilibrium constant Kr of reac-
tion Rr:

= (λr , γ f ) const
K r f=

 ∂λr  ∂K r ∂ γ f
∑ν ν rj fj nj
(3.66)
=− ≈−
j
 
 ∂γ f


λs≠r ∂K r ∂λr ∑ν j
2
rj nj

so that

∂ ln a f
∑ν ν rj fj nj
∆h f ≡ R
∂ (1 T )
≈ ∆H I0f − ∑
r
∆H r0 ⋅
∑ν
j
2
rj nj
(3.67)

3.8 Example: Generic case of an oxygen deficient binary oxide

Let us first consider a generic case of a nonstoichiometric binary oxide MOn–δ with an oxygen deficiency δ re-
sulting from vacancy formation on the oxygen anion sublattice. For the sake of simplicity we only consider the
14

configuration entropy due to random distribution of vacancies on anion sub-lattice (with n sites) while disregard-
ing the mixing of different valence cations being formed due to charge balance on the cation sub-lattice. Since
the system is partly open for oxygen the corresponding hyper-free energy reads, according to Eq. 3.52,


Z MOn−δ = GMO n
− δµO +
 δ   n − δ  n − δ (3.68)
+ RT δ ln   + (n − δ ) ln   − ⋅ (GO 2 + RT ln PO2 )
 n  n  2

and the condition of minimum for this thermodynamic potential

∂Z MOn−δ  δ PO1/ 2  GO


= −∆g O + RT ln  2
 = 0, ∆g O = µO − 2 (3.69)
∂δ  n −δ  2
 

results in the equilibrium oxygen non-stoichiometry

nK O  ∆g O   ∆hO ∆sO 
=δ =
, where K O exp=
 RT  exp  RT − R  (3.70)
PO1/22 + K O    

The standard chemical potential of oxygen µO involves not only the enthalpy hIO for oxygen incorporation into
the anion sub-lattice but also the vibrational enthalpy hvO and the corresponding entropy term T · svO. Note that
unlike the formation of most kinds of defects such as the Schottky, Frenkel, anti-site or substitutional defects
where the changes in lattice dynamics are nearly compensated, the oxygen uptake or release is accompanied, re-
spectively, by a population of new vibrational modes or an extinction of the existing ones. Within the simplest
Einstein approximation

 1 1
3 RΘ E O 
hν O = +  (3.71)
 ( O
)
 exp Θ E T − 1 2 

and

 ΘE  ΘE    Θ EO  
=sν O 3R  O coth  O  − ln  2sinh     (3.72)
 2T  2T    2T   

which can be substituted into the relations for hO = hIO + hvO and µO = hIO + hvO – T · svO as well as into the re-
spective relative partial molar quantities ∆hO = hO − 1 2 H O 2 and ∆g O = µO − 1 2 GO 2 .
As mentioned, the difference between isoplethal and isodynamic heat capacity of a nonstoichiometric oxide
exchanging oxygen with the surrounding atmosphere involves the saturation contribution (Eq. 3.21) and the con-
tribution due to deviation from stoichiometry (Eq. 3.27) which, applied to the present case and assuming ΔhO
and ΔsO being independent of temperature and hO invariable with respect to δ, adopt the forms

1/ 2
 ∂δ  hO ∆hO nPO2 K O
∆ sat CP =
−hO   = (3.73)
 ∂T  RT 2 ( PO1/22 + K O ) 2

 ∂h   ∂h  nK
∆ dev CP =
0

−  O  dδ =
 ∂T 
−  O  1/ 2 r
 ∂T  ( PO2 + K O )
(3.74)
15

As it is seen from Fig. 1, the first contribution has a form of the Schottky anomaly (a peak with a steep start-up
and a slow decay) and it is predominantly related to the enthalpy – hIO required to create oxygen vacancies.
The second part, on the other hand, scales the variation of δ with T and the term (∂ hO/∂T) = (∂hvO/∂T), which
represents the Einstein heat capacity connected with the oxygen vibration modes and limits to Dulong–Petit val-
ue 3R at high temperatures. ΔdevCP thus approaches the value –3δR as δ simultaneously saturates at constant val-
ue in the high temperature limit (δ → n = 1 for the model parameters used to calculate ΔCP in Fig. 1, i.e. a com-
plete reduction to a hypothetical metal with the same structure as has the cation sub-lattice in MOn). In fact, the
assumptions applied to derive the relations on the right hand sides of Eqs. (3.73, 3.74) seem to be rather stringent
since both ΔhO and ΔsO should vary with temperature due to different vibrational characteristics of oxygen in
crystal lattice and oxygen molecule. In the calculation shown in Fig. 1, hO, ΔhO and ΔsO were thus assumed as
temperature dependent (Eqs. (3.71, 3.72) and the tabular data for gaseous oxygen) and substituted directly into
the general relations on the left hand sides of Eqs. (3.73, 3.74). This only slightly affects the shape of the result-
ing curves, but the general features of the simplified model are retained.

20
∆Cp
15 ∆satCp
∆devCp
∆Cp (R)

10

500 1000 1500 2000 2500 3000


T (K)
Fig. 3.1 Difference between isoplethal (n = 1, δ = 0) and isodynamic (pO2 = 1) heat capacity ΔCP (in R = 8.31447 J mol–1 K–1) of an
oxygen deficient oxide MOn–δ and its contributions ΔsatCP and ΔdevCP (Eq. 3.27). Calculated for hIO = –1.5 × 104 R, ΘEO = 1000 K.

It should be noted, however, that in real cases only the growing part of the ΔCP curve can be mostly observed
experimentally, since at higher temperature the system undergoes either a melting or a phase transformation to
an oxide with lower valence of M, both accompanied by an abrupt release of oxygen.
Let us now examine a more complex case assuming an additional formation of Frenkel pairs of oxygen in the
MOn–δ oxide discussed above. This corresponds to a crystal chemical formula [M][On–δ–γδ+γ][m–γOγ] with m-
site sublattice partly occupied by O2– interstitials (the squares stand for vacancies). We thus have two equilibri-
um conditions, (∂Z MOn−δ −γ ∂δ ) = 0 and (∂GMOn−δ −γ ∂γ ) = 0 , resulting in two equations for δ and γ

mPO1/22 K r K O nK O
=γ = , δ −γ (3.75)
1 + PO2 K r K O
1/ 2
PO2 + K O
1/ 2

where

(δ + γ ) PO1/22  ∆g 
=KO = exp  O  (3.76)
n −δ −γ  RT 
16

γ (δ + γ )  ∆g   ∆h 
=Kr = exp  r  ≅ exp  r  (3.77)
(m − γ )(n − δ − γ )  RT   RT 

and Δhr < 0 represents the enthalpy of the Frenkel pair annihilation. Hence, according to Eqs. (3.67) and (3.66),
the partial molar enthalpy of oxygen

 ∂γ   ∂K r ∂δ 
h= hIO + hν O + ∆hr  =  hIO + hν O + ∆hr  −  (3.78)
 ∂δ   ∂K r ∂γ 
O

involves, aside from the previously defined incorporation and vibrational enthalpy, the effect of intrinsic defect
formation.

0.001

0.000

-0.001
∆CP(R)

-0.002
∆satCp''
∆rCp
-0.003

-0.004

-0.005
500 1000 1500 2000 2500 3000
T (K)
Fig. 3.2 The effect of oxygen Frenkel pair formation in an oxygen deficient oxide MOn–δ. Direct effect, ΔrCP (Eq. 3.80), and the in-
fluence on the saturation contribution to heat capacity (second term in Eq. (3.79), ΔsatCP’’). Calculated for Δhr = –1.5 × 104 R, m = 1
and the parameters specified in Fig. 3.1.

The Frenkel pairs thus indirectly affect the saturation contribution to the heat capacity

 ∂δ   ∂K r ∂δ   ∂δ 
∆ sat CP = −(hIO + hν O )   − ∆hr  −   = ∆ sat CP′ + ∆ sat CP′′ (3.79)
 ∂T   ∂K r ∂γ   ∂T 

by influencing δ and hO, but they also give rise to a new excess term

 ∂γ 
∆ r CP = −∆hr   (3.80)
 ∂T 

due to crystal defect formation itself. Interestingly, if we model the effect of Frenkel pairs (second term in Eq.
(3.79) and Eq. (3.80)) by applying the previously used parameters and m = 1, Δhr = –1.5 × 104 R, the saturation
part ΔsatCP’’ turns out to be mainly negative and by one order in magnitude larger than ΔrCP (Fig. 3.2). Neverthe-
less, it remains yet negligible compared to the prevalent ΔsatCP’ term (Fig. 3.1) for the current choice of parame-
ters.
17

3.9 Conclusion

It follows from what was stated above that the model of high temperature heat capacity of a disordered non-
stoichiometric phase exchanging one or more components with surrounding atmosphere includes three additional
contributions compared to the parent stoichiometric phase:
a) contribution ΔrCP due to change of microscopic composition based on a temperature dependent intrinsic
equilibrium of dissociation (or defect formation) reaction Rr (Eq. 3.49).
b) contribution ΔdevCP due to deviation from stoichiometric content of a free component (f) (Eq. 3.27) reflecting
the temperature dependence of partial molar enthalpy. This is caused by the change in number of normal vi-
brational modes as a result of free component release or uptake and eventually by the variation of incorpora-
tion mechanism of free component as a consequence of mutual interplay of different kinds of defects.
c) contribution ΔsatCP connected with isodynamic conditions (instead of isoplethal conditions) at measuring the
high temperature heat capacity of nonstoichiometric phases (Eq. 3.21). It scales the variation of free compo-
nent content with temperature and the absolute value of partial molar enthalpy at given conditions. Similarly,
the latter quantity can involve, in addition to the main contribution of free component incorporation enthalpy,
the effects of other intrinsic equilibria for dissociation reactions Rr or defect formation (cf. a) and Eq. 3.67).

Acknowledgement

One of the authors, D. Sedmidubský, acknowledges a support of the Ministry of Education of the Czech Republic (Research Project
N° MSM6046137302)

References

1. Gopal E.S.R (1966): Specific Heats at Low Temperatures, Heywood Books London, 1966.
2. Inaba H., Naito K. (1977): Heat Capacity of Nonstoichiometric Compounds, Netsu Sokutei (J. Calor. Therm. Anal., Ja-
pan) 4 [1] 10-18.
3. Einstein, A. (1911), Beziehung zwischen dem elastischen Verhalten und der Spezifischen Wärme mit einatomigem
Molekül, Ann. der Physik, vol. 34, p. 170 (1911); Elementare Betrachtungen über die thermische Molekularbewegung in
festen Körpern, Ann. der Physik, vol. 35, p. 679.
4. Petit A.-T., Dulong P.-L. (1819): Recherches sur quelques points importants de la Théorie de la Chaleur. In: Annales de
Chimie et de Physique 10, 395–413 (1819) [English translation in Annals of Philosophy 14, 189-198
http://web.lemoyne.edu/~giunta/PETIT.html.
5. Brillouin, L. (1930). "Les électrons dans les métaux et le classement des ondes de de Broglie correspondantes". Comptes
Rendus Hebdomadaires des Séances de l'Académie des Sciences 191, 292-294.
http://gallica.bnf.fr/ark:/12148/bpt6k31445.
6. Hellmann, H (1937). Einführung in die Quantenchemie. Leipzig: Franz Deuticke. p. 285.
7. Feynman, R. P. (1939). Forces in Molecules. Phys. Rev. 56 (4): 340-343.
8. Debye P. (1912). "Zur Theorie der spezifischen Wärme". Ann. Der Physik 39: 789-839.
9. Sedmidubský, D., Leitner, D., Svoboda, P., Sofer, Z., and Macháček, J. (2009): Heat capacity and phonon spectra of A
IIIN – Experiment and calculation, Journal of Thermal Analysis and Calorimetry 95 [2] 403-407.
10. Schottky, W. (1922): Über die Drehung der Atomachsen in festen Körpen (Mit magnetischen, thermischen und
chemischen Beziehungen). Physikalische Zeitschrift, 23, 448-455.
11. Kittel, C. (1976): Introduction to Solid State Physics, J.Wiley&Sons.
12. Holba P. (2009) „New Thermodynamic Potentials and Clapeyron Equations for Condensed Partly Open Systems" In:
„Some thermodynamic, structural and behavioral aspects of solids accentuating amorphous materials" Ed. J. Šesták,
Publ. Westbohemian Univ. Plzeň 2009.
13. Holba P. (1994): Thermodynamics and Ceramic Systems, In: Structure and Properties of Ceramic Materials (A. Koller,
Editor), Elsevier, Amsterdam – London – New York – Tokyo 1994, pp.17-113.
14. Stolen S. & Grande T. (2004): „Chemical Thermodynamics of Materials: Macroscopic and Microscopic Aspects“, Wiley
2004, see p. 260.
18
15. Chiang, Y.-M., Birnie, D.P., III, Kingery, W.D., (1997) Physical Ceramics, Principles for Ceramic Sciences and Engi-
neering, J.Wiley, pp. 109, 126.
16. Holba P. (1992): Thermodynamics of Partially Open Systems, Czech J. Phys. B42, No.6, 549-575.

View publication stats

You might also like