Universal equation of state for elastic solids; MgSiO3 perovskite is an example

József Garai

Department of Earth Sciences, Florida International University, Miami, FL 33199, USA

Abstract

Universal (P-V-T) equation of state is derived from theoretical considerations. Correlation

coefficients, root-main-square-deviations, and Akaike Information Criterias are used to evaluate

the fitting to the experiments of perovskite 0-109 GPa and 293-2000 K. The proposed equation

remains valid through the entire pressure and temperature range and has superior fitting

parameters in comparison to the Birch-Murnaghan, Vinet, and Roy & Roy equation of states.
Contents

1. Introduction

2. The equation of state

2.1 Thermal EoS

2.2 The temperature effect on the bulk modulus

2.3 Isothermal EoS

2.3.1 Finite-strain EoS

2.3.2 Inter-atomic potential EoSs

2.3.3 Empirical EoSs

3. Fundamental components of the volume in solid phase

4. New description for the pressure-temperature-volume relationship

4.1 The affect of pressure and temperature

5 Testing the EoS to experiments of perovskite

5.1 Fitting criterias

5.2 Fitting parameters

6. Conclusions

Acknowledgement

References

-2-
1. Introduction

The relationships among the pressure, the volume, and the temperature are described by the

Equation of State (EoS). The volume-temperature relationship is described by the definition of

the volume coefficient of expansion [ α V ]

1  ∂V 
αVp ≡   (1)
V  ∂T  p

The relationship between the pressure and the volume is given by the isothermal bulk

modulus [B T ]

 ∂p 
B T ≡ − V  . (2)
 ∂V  T

For the validity of equation (2) it is assumed that the solid is homogeneous, isotropic, non-

viscous and has linear elasticity. It is also assumed that the stresses are isotropic; therefore, the

principal stresses can be identified as the pressure p = σ1 = σ 2 = σ 3 . The schematic

relationships between the thermodynamic quantities are shown on Fig. 1-a.

Experiments show that both the volume coefficient of expansion and the isothermal bulk

modulus are pressure and temperature dependent; therefore, it is necessary to know the

derivatives of these parameters.

 ∂α V   ∂α   ∂B   ∂B 
  ;  V  ;  T  ;  T  (3)
 ∂T  p  ∂p  T  ∂T  p  ∂p  T

A universal EoS must cover the entire pressure and temperature range; therefore, it is necessary

to incorporate all of the derivatives of the volume coefficient of expansion and the isothermal

bulk modulus. There is no single expression known for universal (P-V-T) EoS (MacDonald,

-3-
1965; Baonza et al., 1996). An attempt is made here to derive and test the first universal EoS for

elastic solids.

2. The equation of state

In order to overcome the complexity of the EoS, the common practice is that the temperature

of the substance is raised first and then the substance is compressed along the isotherm of interest

(Duffy and Wang, 1998; Angel, 2000). The relevant equations are called the thermal and the

isothermal equation of state (EoS) respectively.

The thermal EoS is used to calculate the volume at atmospheric pressure and temperature T

[V ] .
0 ,T It is also necessary to know the temperature affect on the bulk modulus [B 0 (T )] . Using

the values of the volume and the bulk modulus at the corresponding temperature the isothermal

EoS calculates the affect of pressure by using the first and the second derivates of the bulk

 ∂B   ∂ 2B 
modulus,   and  2  at the given temperature.
 ∂p  T  ∂p  T

The simplest complete thermodynamic description of a single component solid then requires a

minimum of four parameters:

 ∂B   ∂B 
αV ; B 0 (T0 ) ;   ;   (4)
 ∂T  p=0  ∂p  T

a more precise description would require six parameters

 ∂α V   ∂B   ∂B   ∂ 2B 
αV ;   ; B 0 (T0 ) ;   ;   ; and  2  . (5)
 ∂T  p=0  ∂T  p=0  ∂p  T  ∂p  T

-4-
2.1 Thermal EoS

The simplest thermal equation of state is derived by the integration of the thermodynamic

definition of the volume coefficient of thermal expansion Eq.(1)

T

∫ αVp (T )dT (6)

V0 (T ) = V0 (T0 )e T0
.

If a wider temperature range is considered then the temperature dependence of the volume

coefficient of thermal expansion should be known. Knowing the first derivative of this

parameter allows one to calculate the high temperature values:

 ∂α 
α V p (T ) = α V p (T0 ) +  V  (T − T0 ) . (7)
 ∂T  p

The thermodynamic Gruneisen-Anderson parameter [δ T ] is defined as (Anderson, 1987):

 ∂ ln K T  1  ∂ ln K T 
δ T =   = −   (8)
 ∂ ln ρ p α V  ∂T  p

Assuming that the solid at higher temperatures follows classical behavior, then the product of

α Vp K T is constant and the Gruneisen-Anderson parameter is independent of temperature,

Anderson et al. (1992) and Shanker (1993) proposed the following isobaric EoS:

[
V = V0 1 − α V0 δ 0 (T − T0 ) ]
1
−
δ0 (9)

where the subscript zero values of the parameters refers to the initial temperature of T0 .

Assuming that the product of α Vp K T is constant and the Gruneisen-Anderson parameter

changes linearly with the volume, the following EoS has been proposed by Kumar (2002) and

Kushwah et al. (1996)

-5-
  1
[ 
V = V0 1 − ln 1 − α V0 A(T − T0 )  ] (10)
 A 

where A = δ 0 + 1 .

Thermal EoS have been suggested by Akaogi and Navrotsky (1984; 1985), assuming that the

thermal expansion is quadratic in the temperature, and independent of pressure

[
Vp = V0 1 + α V0 (T − T0 ) + α 'V0 (T − T0 ) ,
2
] (11)

where α 'V0 is the temperature derivative of α V at temperature T = T0 . Taking into consideration

the affect of the pressure the equation can be written as:

−1

 B ' (p − p 0 )  K '0
Vp ,T = V0 1 + 0
B0
'
[
 1 + α V0 (T − T0 ) + α V0 (T − T0 ) ,
2
] (12)
 

Fei and Saxena (1986) revised the quadratic relationship of Eq. (12) and proposed the

following empirical expression:

 1 −1 
Vp = V0 1 + α V0 (T − T0 ) + α 'V0 (T − T0 ) − α V0 (T − T0 )  .
2
(13)
 2 

Assuming linear change as a function of temperature in the volume coefficient of thermal

expansion leads to the following expression:

1
α V0 ( T − T0 )+ α 'V0 ( T − T0 )2
Vp = V0 e 2 (14)

The proposed general expression of Eq. (14) for bcc iron is:
−1

   ∂B   p − p 0  K '0 α V0 (T −T0 )+ 12 α'V0 (T −T0 )2

Vp = V0 1 + B '0 B 0 −   (T − T0 )  e (15)
 ∂T  0
2
   B0 

An empirical expression has been given by Plymate and Stout (1989)

-6-
  −
1  ∂B
 α +  T 
1 

∂B
 ( T − T0 )+  α 'V +  T 
2
1  ( T − T0 )2 


   ∂B   T − T   B'0  V0  ∂T 
B B '   0  ∂T  2 ' 
B B 2 
V = V0 1 +  T    e 
0 0 0 0  0 0 0 
 (16)
  ∂T  0  B 0  
 

2.2 The temperature effect on the bulk modulus

The temperature has an effect not only on the volume and the volume coefficient of thermal

expansion but on the bulk modulus as well. In order to use the isothermal EoS it is necessary to

know the value of the bulk modulus at the temperature of interest, which can be obtained from

 ∂B 
B T 0 (T ) = B T 0 (T0 ) +  T  (T − T0 ) . (17)
 ∂T  p

Theoretically the temperature dependence of the elastic constants can be determined as the

sum of the anharmonic terms (Kittel, 1968; Levy, 1986). At sufficiently low temperatures the

elastic constant should vary as T4 (Born & Huag, 1956, p 437). Contrary to this suggestion some

metallic substances have been found to show a T2 rather than a T4 dependence at low

temperatures (Alers, 1961; Chang. & Graham, 1966). There is no general prediction for higher

temperatures. Experiments on refractory oxides, conducted at higher than room temperature,

show a linear relationship between the bulk modulus and the temperature (Wachtman, 1959).

The third law of thermodynamics requires that the derivative of any elastic constant with

respect to the temperature must approach zero as the temperature approaches absolute zero.

Combining this criterion with the observed linear relationship at higher temperatures, Wachtman

et al. (1961) suggested an equation in the form of

 T0 
− 
B = B 0 − b1Te  T 
. (18)

-7-
where K 0 is the bulk modulus at absolute zero, and b1 and T0 are arbitrary constants.

Theoretical justification for the Wachtman’s Equation was suggested by Anderson (1966).

Based on shock-wave and static-compression measurements on metals, a linear relationship

between the logarithm of the bulk modulus and the specific volume has been detected for metals

(Grover et al., 1973):

 ∆V 
ln B T = ln B 0 + α , (19)
 V 

where α is a constant depending on the material. The linear correlation is valid up to 40%

volume change. Using this linear correlation Jacobs and Oonk (2000) proposed a new equation

of state. They rewrite equation (19) as

 B 0 (T ) 
Vm0 (T ) = Vm0 (T0 ) + b ln 0  , (20)
 B (T0 ) 

where Vm0 denotes molar volume, T0 the reference temperature and the superscript “0” refers to

standard pressure (1 bar). Equation (20) successfully reproduces the available experimental data

for MgO, Mg 2 SiO 4 , and Fe 2 SiO 4 (Jacobs and Oonk, 2000; Jacobs et al. 2001; Jacobs and

Oonk, 2001).

Assuming that the product of the volume coefficient thermal expansion and the bulk modulus

is constant at temperatures higher than the Debye temperature analytical solution for the

temperature dependence of the bulk modulus was derived (Garai and Laugier, 2006).
T
−
∫T =0 δTα VdT (21)
BT = e K T =0 .

where δ T is the isothermal Anderson-Grüneisen parameter given by:

-8-
 1  ∂B T 
δT = −   . (22)
α Vp B T  ∂T  p

Equation (21) was able to mimic experiments with high accuracy for the investigated substances.

2.3 Isothermal EoS

The determined values of the volume and the bulk modulus at temperature T can be used as

initial parameters for an isothermal EoS. The isothermal equation of states follow finite strain,

interatomic potential, or empirical approach.

2.3.1 Finite-strain EoS The Birch-Murnaghan EoS (Birch, 1947; Murnaghen, 1937, 1944)

assumes that the strain energy of a solid can be expressed as a Taylor series in the finite Eulerian

strain, f E . Expansion to fourth order in the strain yields an EoS:

5 3 3 35  
p = 3B 0 f E (1 + 2f E )2 1 + (B'−4)f E + B 0 B"+(B'−4)(B'−3) +  f E2  (23)
 2 2 9 

where f E is

 V0  3
  −1 (24)
V
fE =   .
2

The Birch-Murnaghan equation (23) is the most widely used isothermal EoS.

Quite recently Sushil et al. (2004) used n=1 instead of the n=2 in the Eulerian strain measure
n

 V0  3
  −1 1
(25)
 V  V0  3
fE = =   −1
n V

and using the method of Stacey (2001) proposed a modified three-parameter Eulerian strain EoS,

-9-
 9  −
4
−
5
− 
7

p = B 0  − A1 x 3 + A 2 x 3 − A 3 x −2 + A 4 x 3  . (26)
2  

where

V 26
A1 = B 0 B"0 + (B '0 − 3) +
2
x= ,
V0 9
66
A 2 = 3B 0 B"0 + (B '0 − 3)(3B '0 − 8) +
9
60
A 3 = 3B 0 B"0 + (B '0 − 3)(3B '0 − 7 ) +
9
20
and A 4 = B 0 B"0 + (B '0 − 3)(B '0 − 2 ) +
9

The authors claimed that their modified Eulerian strain EoS is more rapidly convergent than the

Birch-Murnaghan EoS.

2.3.2 Inter-atomic potential EoSs The theoretical base for the interatomic potential EoS lays in

the thermodynamic relationship

 ∂p   ∂U 
p = T  −   (27)
 ∂T  V  ∂V  T ,m

where m stands for a mol quantity. Assuming constant temperature, as the case for isothermal

EoSs, the first term can be neglected. Approaching the second term, the so-called internal

pressure, with the volume derivative of the biding energy allows determining the pressure-

volume relationship. The resulting EoS contains three parameters, the zero pressure values of

the molar volume, the isothermal bulk modulus, and the pressure derivative of the bulk modulus.

Using the potential function proposed by Mie and extended by Grunesisen (Poirier, 1991 p.

37; Partington, 1957)

- 10 -
 A B A B
U(r ) = − − n = − m + n . (28)
r m
r    
V 3  V 3 

where r is the interatomic spacing and A, B, m, and n are constants (not necessarily integers) the

P-V equation of state can be written as (Anderson, 1989, p. 83):

  m +3 

 n +3 
3K T (0)  Vo  3   Vo  3  
  

p=   −   . (29)
m − n  V  V 
 

The so-called universal EoS derived by Rose from a general inter-atomic potential (Rose,

1984) which was promoted by (Vinet, 1987-a, -b) is also commonly used:

1 − f V  2 ( B ' −1)(1−f V )

3

p = 3K 0 e (30)
f V2

where

 V 3
f V =   . (31)
 V0 

The Vinet EoS gives very accurate results for simple solids at very high pressure.

Some authors (Parsafar and Mason, 1994; Campbell and Heintz, 1991) pointed out that there

is a restriction on Eq. (30) when it is applied to high-pressure phase solids under low pressure

conditions. The use of p = 0 and V = V0 is sometimes arbitrary since the high pressure phase

might not exist under this condition. In order to overcome on this problem Fang, 1998 suggested

modifying the original Vinet equation (30) by introducing an additional parameter. In this

modified equation it was assumed that the isothermal bulk modulus varies linearly with the

pressure.

- 11 -
 Precise knowledge of the interatomic forces in the stress-free state and their variation with

pressure and temperature would allow calculating all the thermodynamic properties. The lack of

such knowledge has resulted in many two and three-parameters empirical EoSs.

2.3.3 Empirical EoSs Empirical EoSs can be divided into two major groups. One uses the

original Eulerian strain or Interatomic potential EoSs and refines their parameters in order to find

a better fit to experiments (e.g. Keane, 1953; Davis and Gordon, 1967: Freud and Ingalls, 1989 ;

Kumari and Dass, 1990). The other approach is to find a mathematical function which gives the

best fit to the experiments (e.g. Mao, 1970; Huang and Chow, 1974; Roy and Roy, 2003; Saxena,

2004).

Roy and Roy (2005) give a good review and evaluate the fittings of the currently used EoSs.

Their proposed (Roy and Roy, 1999) three parameter empirical EoS is

 ln (1 + ap ) 
V = V0 1 − , (32)
 b + cp 

where

1 3(B ' + 1) + (25B ' 2 + 18B ' − 32B B '' − 7 )12 

a=  0 0 0 0 0 
8B 0

1
b = 3(B '0 + 1) + (25B '02 + 18B '0 − 32B 0 B '0' − 7 )2 
1

8 

1  '  1 ' 
( ) ( ) 2 (B ' + 1) − ( ) ( )
1 1
c= 3 B 0
+ 1 + 25 B '2
0
+ 18 B '
0
− 32 B 0
B ''
0
− 7  0 3 B 0
+ 1 + 25 B '2
0
+ 18 B '
0
− 32 B 0
B ''
0
− 7 2

16   8

They used shock compression data of different metals (Nellis 1988) and the calculated EoS of

halite (Decker, 1971) to evaluate the proposed equation up to ultra high pressures.

- 12 -
 The empirical nature of these equations usually leads to a lack of generality and careful

inspection reveals that a particular equation is typically gives excellent fitting only for special

substances or a specially selected pressure and/or temperature range.

Many of the parameters in the EoS are inter-related, which adds to the complexity of

calculations. The optimum values of each of the interrelated parameters have to be determined

by confidence ellipses (Angel, 2000, Mattern et al., 2005). The thermodynamic description of

solids is complicated, time consuming, labor intensive, and expensive.

3. Fundamental components of the volume in solid phase

Contrarily to gasses Avogadro’s principle does not apply to solids. Matter in solid phase

occupies an initial volume [Vo ] at zero pressure and temperature.

Vo = nVom , (33)

In Eq. (33) n is the number of moles and Vom is the molar volume of the substance at zero

pressure and temperature. The pressure modifies this initial volume by inducing elastic

deformation while the temperature by causing thermal deformation. Using equations (1) and (2)

the actual volume at given pressure and temperature can be calculated (Cemic, 2005) by

allowing one of the variables to change while the other one held constant

T p 1
∫p=0 BT =0 dp
∫T =0 α V p =0dT
[V ]
−
[V ]
T p =0
= V0 e or p T =0 = V0 e (34)

and then

T p 1
∫p=0 BT dp
[V ] = [V ] ∫T =0 α Vp dT
[V ] = [V ]
−
e or e (35)
T p p T =0 p T T p =0

- 13 -
These two steps might be combined into one and the volume at a given p, and T can be

calculated:
T p 1 T p 1
∫T =0 α Vp=0 dT − ∫p=0 BT dp ∫T =0 αVp dT − ∫p=0 BT =0 dp (36)
Vp ,T = V0 e = V0 e

The total volume change related to the temperature will be called thermal volume [V th ] while

the total volume change resulted from elastic deformation will be called elastic volume [V el ] .

The thermal volume at zero pressure is

 T αVdT 
 ∫ 
[VTth ]p=0 = Vo  e T=0 − 1 , (37)
 
 

while the elastic volume at zero temperature is

 − p 1 dp 
 ∫ BT 
[Vpel ]T =0 = Vo  e p=0 − 1 . (38)
 
 

The thermal volume at pressure p is

p p

∫
1
∫
1  T αV dT 
−
BT
dp −
BT
dp
 ∫ 
[VTth ]p = [VTth ]p=0 e p =0
= Vo e p =0
 e T =0 − 1 , (39)
 
 

and the elastic volume at temperature T is

T T
 p 1 
∫ α VdT ∫ αV dT  −p∫=0 BT dp 
[Vpel ]T = [Vpel ]T =0 e T =0 = Vo e T =0  e − 1 . (40)
 
 

The actual volume is the sum of the volume components:

[V ]
T p
= Vo + [Vpel ]
T =0
+ [VTth ] p or [V ] p T
= Vo + [Vpel ] + [VTth ] p =0
T
(41)

Since

- 14 -
 [V ] = [V ]
T p p T (42)

from Eq (41) follows that

[VTth ]p − [VTth ]p=0 = [Vpel ]T − [Vpel ]T=0 . (43)

The compressed part of the thermal volume is the same as the expanded part of the elastic

volume. Since the volume difference in Eq. (43) both temperature and pressure dependent I will

call this volume difference to thermo-elastic volume [∆Vpth −el ]T

[Vpth −el ]T = [VTth ]p − [VTth ]p=0 = [Vpel ]T − [Vpel ]T=0 . (44)

The thermoelastic volume can be calculated as:

 T α VdT   − p 1 dp 
 ∫   ∫ BT 
[Vpth −el ]T = Vo  e T =0 − 1  e p=0 − 1 . (45)
   
   

It can be concluded that the actual volume comprises from four distinct volume parts, initial

volume, thermal volume at zero pressure [VTth ] p=0 , elastic volume at zero temperature [V ]
el
p
T =0
,

and thermo-elastic volume [Vpth −el ] (Fig. 2).

T

V = Vo + [VTth ] p=0 + [Vpel ] T=0 + [Vpth −el ] . (46)

T

These fundamental volume components are related to the thermo-physical variables as:

Vo = f (n ) , (47)

[VTth ]p=0 = f (n, T ) , (48)

[Vpel ]T=0 = f (n, p ) , (49)

- 15 -
and

[Vpth −el ]T = f (n , T, p ) . (50)

4. New description for the pressure-temperature-volume relationship

Recent study (Garai, 2007) suggested that the mechanical equivalency of heat or the first law

of thermodynamics is correct only if the energy from the mechanical work is conserved by the

same physical process as the heat. This condition is not satisfied in solid phase; therefore, heat

and work is not interchangeable and they must be treated separately. The lack of

‘communication’ between the heat and work results that thermoelastic volume should not exist

Eq. (50). Thus the volume should comprise only from the initial, thermal, and elastic volumes.

Deducting the thermal-elastic volume from the actual volume gives

 T α dT p
1 
 ∫ V −
∫ BT dp 
V − [Vpth −el ]T = Vo  e T =0 + e p=0 − 1 . (51)
 
 

The elimination of the thermoelastic volume requires the transformation of the V − [Vpth −el ]T

volume to the actual volume V:

 T α dT p
1 
 ∫ V −
∫ BT dp 
V − [Vpth −el ]T = Vo  e T =0 + e p=0 − 1 ⇒ V . (52)
 
 

This transformation can be achieved by redefining the volume coefficient of thermal expansion

and the bulk modulus

α Vp ⇒ α o BT ⇒ Bo . (53)

- 16 -
Superscript o is used for the new parameters which are defined as:

1 ∂V th
αo ≡ (54)
Vp =0 ∂T

and

∂p
B o ≡ −VT =0 . (55)
∂V el

Using the new definition of the volume coefficient of thermal expansion the thermal volume is

 T αodT 
 ∫ 
Voth = Vo  e T =0 − 1 (56)
 
 

while the elastic volume can be calculated as:

 − p 1 dp 
 ∫ Bo 
Voel = Vo  e p=0 − 1 . (57)
 
 

Subscript o is used to indicate that these fundamental volume parts were determined by using B o

and α o . The actual volume is the sum of the initial, thermal and elastic volumes (Fig. 1)

V = Vo + Voel + Voth . (58)

Substituting the fundamental volume components gives the actual volume

 − p 1 dp T

 ∫ Bo ∫ α o dT 
V = Vo  e p=0 + e T =0 − 1 . (59)
 
 

Eq. (59) is identical with the required expression given in Eq. (52) except the conventional

volume coefficient of expansion and bulk modulus has been replaced with the newly defined

ones. Thus the transformation of the volume is completed by the introduction of the new

definitions [Eq. (54) and (55)].

- 17 -
4.1 The affect of pressure and temperature

Equation (59) assumes a constant value for volume coefficient of thermal expansion and the

bulk modulus. In order to take into consideration the pressure and temperature affect on these

parameters linear factors are introduced. The bulk modulus is approximated as

B o (p, T ) = ap + bT + B o . (60)

where a and b are linear factors relating to the pressure and temperature respectively. Equation

(59) can be rewritten then as:

 − p 1 dp T

 ∫ ap+ Bo ∫ α o dT 
V = Vo  e p=0 + e T =0 − 1 (Universal 4) (61)
 
 

and

 − p 1 dp T

 ∫ ap+ bT + Bo ∫ αodT 
V = Vo  e p=0
+ e T =0 − 1 (Universal 5). (62)
 
 

The numbers after Universal refers to the number of parameters in the equation.

Investigating highly symmetrical atomic arrangements linear correlation between the volume

coefficient of thermal expansion and the thermal heat capacity was detected (Garai, 2006).

Based on this correlation the integral

∫ α dT
T =0
o (63)

is approximated by an area of trapezoid (Fig. 3) The integral below the Debye temperature [Tθ ]

- 18 -
is then
T ≤ Tθ
α T2 
∫ α o (T )dT ≈  o  (64)
T =0  2Tθ  T≤Tθ

while at temperatures higher than the Debye temperature is

∫ α (T )dT ≈ [α (T − T )]
T > Tθ
o o θ Tθ
. (65)
Tθ

Combining the two parts Eqs. (64) and (65) gives the general formula

T
αoT 2  T 
∫ α o (T ) dT ≈ I A (T ) + [1 − I A (T )] α o  T − θ  , (66)
T =0
2Tθ  2

where

1 if T ≤ Tθ
I A (T ) =  . (67)
0 if T > Tθ

Substituting Eq. (66) into Eq. (61) gives

 − p 1 dp 
 ∫ ap+ Bo α T2  T 
I A ( T ) o + [1− I A ( T )] α o  T − θ  
V = Vo  e p =0
+e 2 Tθ  2 
− 1 (Debye 4). (68)
 
 

Assuming linear pressure dependence for the Debye temperature requires the introduction of an

additional multiplier c

Tθ (p ) = cp + Tθ (p = 0) . (69)

Equation (68) can be written then as:

 − p 1 dp 
 ∫ ap+ Bo I A (T )
αoT 2 
+ [1− I A ( T )] α o  T −
cp + Tθ 
 
V = Vo  e p=0 +e 2 ( cp + Tθ )  2 
− 1 (Debye + pressure 5). (70)
 
 

- 19 -
The validity of Eqs. (61), (62), (68), and (70) will be tested to experiments and I will call these

equations to Universal 4 parameter, 5 parameter, Debye, and Debye plus pressure respectively.

Equation (61) has an analytical solution for the pressure

V 
ln − e αT + 1
p = Bo  Vo  .
(71)
V 
− a ln − e αT + 1 − 1
 Vo 

and for the temperature

V −p

ln  −e ap b + Bo
+ 1
 Vo  (72)
T=  .
αo

Equation (62) has analytical solution only for the pressure

V 
ln − e αT + 1
p = (B o + cT )  Vo  . (73)
V 
− a ln − e αT + 1 − 1
 Vo 

Iteration was used to calculate the temperature from Eq. (62).

5. Testing the EoS to experiments of perovskite

Perovskite, the most abundant mineral of the mantle, has been extensively investigated at high

pressures and temperatures. The availability of a wide range of pressure and temperature

experiments makes this mineral ideal for thermodynamic studies. Experiments up to 25-30 GPa

pressure usually use multi-anvil apparatus while at higher pressures diamond anvil cells (DAC)

are used. The experimental results of multi anvil press (Funamori, et al., 1996; Wang et al.,

- 20 -
1994; Morishima et al., 1994; Utsumi et al., 1995) and diamond anvil (Fiquet et al., 1998; 2000;

Saxena et al., 1999) are used in this study (Fig. 4).

5.1 Fitting criteria

The fitting accuracy of empirical EoSs with the same number of parameters is evaluated by

correlation coefficients and/or root-mean-square deviations (RMSDs). The number of wiggles of

the data deviation curve allows detecting the possible standard error in the fitting equation (Roy

and Roy, 2005). The fit quality of models using different numbers of parameters can not be

evaluated by their correlation coefficients only (Lindsey, 2004; Burnham and Anderson, 2002;

2004). The test devised assessing the right level of complexity is the Akaike Information

Criteria AIC (Akaike, 1973; 1974). Assuming normally distributed errors, the criterion is

calculated as:

 RSS 
AIC = 2k + n ln , (74)
 n 

where n is the number of observations, RSS is the residual sum of squares, and k is the number

of parameters. The preferred model is the one which has the lowest AIC value.

5.2 Fitting parameters

The calculated fitting parameters, correlation coefficient, RMSD, and AIC are given in Table

1 for Eqs. (61), (62), (68), and (70). The best fit is achieved by the universal 5 parameter

equation [Eq. (62)]. The approximations used for the volume coefficient of thermal expansion

in Eqs. (68) and (70) did not increase the fitting and better fit was achieved by assuming constant

- 21 -
value for the volume coefficient of thermal expansion. Based on visual inspection the residuals

seem to be random (Fig. 5). The RMSD or uncertainty of the universal 5 parameter equation is

0.05 cm3, 0.8 GPa, and 123 K for the volume, pressure, and temperature respectively. These

values are in the range of the uncertainties of the experiments, since the laser heated DAC data

have an order of magnitude uncertainty in the temperature measurement (Shim and Duffy, 2000).

The uncertainty is significantly smaller if an electrical heater is used in the DAC. Having the

same uncertainty from the fitting of the universal as from the experiments is a clear indicative

that the proposed EoS correctly describes the P-V-T relationship of perovskite.

Starting from 300K the experiments were separated into 200 K wide temperature groups.

Using the three most widely used isothermal EoS, Birch-Murnaghan, Vinet, and Roy & Roy,

equations (23), (30), and (32) respectively the fitting parameters were determined for each of

these temperature range. Using averages determined from the overall fitting the RMSD, and

AIC was calculated for the universal 4 and 5 parameter equations in of the temperature range.

The fitting parameters were also determined for the conventional bulk modulus [ K T conv.] as:

−p

V = V0 T e ap+ K T (Conventional). (75)

The calculated values of the five parameter universal EoS have equal or better fitting parameters

than any of the investigated isothermal EoSs (Table 2). The four parameters universal EoS has

better fitting parameters in seven temperature ranges than the conventional bulk modulus

equation. The only exception is the highest temperature range (1700-1900K). The better fitting

indicates that the proposed new definition of the bulk modulus describe the volume pressure

relationship more accurately than the conventional one.

- 22 -
6. Conclusions

Assuming that adiabatic conditions do not exist in elastic solid phase universal P-V-T EoS has

been derived by using the newly defined expressions of the volume coefficient of thermal

expansion and the bulk modulus. Using the high pressure and temperature experiments of

perovskite, the fitting parameters, correlation coefficients, RMSD, and AIC were calculated.

The calculated fitting parameters of the new universal EoS are superior to the Birch-Murnaghan,

Vinet, and Roy & Roy equations. Additional advantage of the proposed universal EoS is its

simplicity and the fact that it allows the back and forward calculation of any of its quantities.

Acknowledgement

I would like to thank Alexandre Laugier for his encouragement and Mike Sukop for reading and

commenting the manuscript. This research was supported by Florida International University

Dissertation Year Fellowship.

- 23 -
References

Akaike H 1973 Second International Symposium on Information Theory (Budapest, Akademia Kiado) p. 267
Akaike H 1974 IEEE Transactions on Automatic Control 19 716
Akaogi M and Navrotsky A 1984 Phys., Earth Planet. Interior. 36 124
Akaogi M and Navrotsky A 1985 Phys. Chem. Mineral. 12 317
Alers G A and Waldorf DL 1961 Phys. Rev. Lett. 6 677
Anderson O L 1966 Phys. Rev. 144 553
Anderson D L 1989 Theory of the Earth (Blackwell Scientific Publications) p. 366
Anderson D L 1987 Phys. Earth Planets. Interiors. 45 307
Anderson O L Isaak D and Oda H 1992 Rev. Geophys. and Space Phys. 30 57
Angel R J 2000 Rev. in Mineral. and Geochem. 41 35
Baonza V G Taravillo M Caceres M and Nunez J 1996 Phys. Rev. B 53 5252
Birch F 1947 Phys. Rev. 71 809
Born M and Huang K 1956 Dynamical Theory of Crystal Lattice (Oxford University Press, New York)
Burnham K P and Anderson D R 2002 Model Selection and Multimodel Inference: A Practical-Theoretic Approach
(Springer, New York)
Burnham K P and Anderson D R 2004 Multimodel Inference: understanding AIC and BIC in Model Selection
(Amsterdam Workshop on Model Selection)
Campbell A J and Heinz D L 1991 J. Phys. Chem. Solids 52 495
Cemic L 2005 Thermodynamics in Mineral Sciences (Springer) p. 43.
Chang R and Graham L J 1966 J. Appl. Phys. 37 3778
Davis l A and Gordon R B 1967 J. Chem. Phys. 46 2650
Decker D L 1971 J. Appl. Phys. 42 3239
Duffy T S and Wang Y 1998 Rev. in Mineral. and Geochem. 37 425
Fang ZH 1998 Phys. Rev. B 58 20
Fei Y and Saxena SK 1986 Phys. Chem. Minerals 13 311
Fiquet G. et al 1998 Phys. Earth and Planet. Int. 105 21
Fiquet G Dewaele A Andrault D Kuntz M and Le Bihan T 2000 Geophys. Res. Lett. 27 21
Freud J and Ingalls R J 1989 Phys. Chem. Solids 50 263
Funamori N Yagi T and Utsumi W 1996 J. Geophys. Res. 101 8257
Garai J 2007 arXiv:0705.1484v1
Garai J 2006 CALPHAD 30 354
Garai, J., and Laugier, A 2007 J. Appl. Phys. 101 023514
Grover R Getting I C and Kennedy G C 1973 Phys. Rev. B 7 567
Huang Y K and Chow C Y 1974 J. Phys. D: Appl. Phys. 7 2021
Jacobs M H G de Jong B H W S and Oonk H A J 2001 Geochim. Cosmochim. Ac. 65 4231
Jacobs M H G and Oonk H A J 2001 Phys. Chem. of Min. 28 572
Jacobs M H G and Oonk H A J 2000 CALPHAD 24 133
Keane A 1953 Nature 172 117
Kittel C 1986 Introduction to Solid State Physics 6th edition (New York, Wiley) p. 114.
Kumari M and Dass N 1990 J. Phys.: Condens. Matter 2 3219
Kushwah S S Kumar P and Shanker J 1996 Physica B 229 85
Kumar M 2002 Physica B 311 340
Levy R A 1968 Principles of Solid State Physics (New York, Academic) p. 141.
Lindsey J K 2004 Introduction to applied statistics a modeling approach, Second Ed. (Oxford University Press Inc.,
New York)
MacDonald J R 1969 Rev. Mod. Phys. 41 316
Mao N H 1970 J. Geophysical Res. 75 7508
Mattern E Matas J Ricard Y and Bass J 2005 Geophys. J. Int. 160 973
Morishima H Ohtani E and Kato T 1994 Geophys. Res. Lett. 21 899
Murnaghan F D 1937 Am. J. Math. 49 235
Murnaghan, F D 1944 Proceedings of the National Academy of Sciences 30 244
Nellis WJ et al 1988 Phys. Rev. Lett. 60 1414

- 24 -
Parsafar G and Mason E A 1994 Phys. Rev. B 49 3049
Partington J R 1957 An Advanced Treatise on Physical Chemistry, Vol. 3, (Longmans, London) p. 345
Poirier J-P 1991 Introduction to the physics of the Earth’s interior (Cambridge University Press, Cambridge)
Rose JH Smith JR Guinea F and Ferrante J 1984 Phys. Rev. B 29 2963
Roy S B and Roy P B 1999 J. Phys. Condens. Matter. 11 10375
Roy S B. and Roy P B 2005 J. Phys. Condens. Matter. 17 6193
Roy S B and Roy P B 2003 J. Phys. Condens. Matter. 15 1643
Saxena S K 2004 J. Phys. Chem. Solids 65 1561
Saxena S K Dubrovinsky L S Tutti F and Le Bihan T 1999 Amer. Mineral. 84 226
Shanker J and Kumar M 1993 Phys. Stat. Sol. (b) 179 351
Shim Sang-Heon and Duffy TS 2000 Amer. Miner. 85 354
Stacey F D 2001 Phys. Earth Planet Inter. 128 179
Sushil K Arunesh K Singh P K and Sharma B S 2004 Physica B 352 134
Utsumi W Funamori N and Yagi T 1995 Geophs. Res. Letts. 22 1005
Vinet P Ferrante J Rose J H and Smith J R 1987-a J. Geophys. Res. 92 9319
Vinet P Smith J R Ferrante J and Rose J H1987-b Phys. Rev. B 35 1945
Wachtman J B Jr and DG Lam Jr. 1959 J. Am. Ceram. Soc. 42 (5) 254
Wachtman J B Jr Tefft W E Jr Lam D J Jr and Apstein C S 1961 Phys. Rev. 122 1754
Wang Y Weidner DJ Liebermann R C and Zhao Y 1994 Phys. of the Earth and Planet. Int. 83 13

- 25 -
Figure 1. Thermo-physical relationships (a) solid phase conventional description (b) proposed

new description (c) new description at constant volume. (The arrow ↔ represent a reversible

while → represents an irreversible relationship or process.)

- 26 -
Figure 2. The fundamental volume components of the actual volume, in accordance to the

conventional thermo-physical description of solids.

Figure 3. Approximation used for the volume coefficient of thermal expansion in Eqs. (68), and

(70).

- 27 -
Figure 4. Pressure-temperature range covered by the data.

- 28 -
Figure 5. The residuals plotted against (a) volume (b) pressure (c) temperature.

- 29 -
 0-109 GPa [N=257] Ko Vo[cm3] αo [10-5 K-1] a b c R RMSD AIC
[GPa]
V(p,T) (Univ. 4) 274.2 24.345 1.451 1.291 0.99934298 0.055 -1486.1
V(p,T) (Univ. 5) 280.2 24.232 2.362 1.405 -0.0277 0.99957336 0.044 -1595.0
p(V,T) ( Univ. 4) 276.3 24.370 1.303 1.277 0.99932559 1.038 27.3
p(V,T) ( Univ. 5) 282.9 24.210 2.499 1.406 -0.0319 0.99960000 0.797 -106.7
T(V,p) ( Univ. 4) 269.8 24.326 1.620 1.323 0.95180023 144.2 2563.1
T(V,p) ( Univ. 5) 123.1 2474.0
V(p,T)(Debye; 4) 273.8 24.483 1.774 1.305 0.9991373 0.063 -1416.1
V(p,T)(Debye+pres; 5) 273.5 24.433 3.857 1.467 51.6 0.9992888 0.057 -1463.7
Average ( Univ. 4 ) 273.4 24.347 1.458 1.297
Average ( Univ. 5) 281.5 24.221 2.431 1.405 -0.0298

Table 1 P-V-T fitting parameters and results.

- 31 -
- 33 -
Table 2 P-V fitting parameters and results

- 34 -