You are on page 1of 19

Fundamentals of Chemical Processes summary

L. Spinelli

1 Chemical equilibrium
In order to study chemical reactions, we must first introduce some fundamental quantities. The centrepiece of chemistry is the
stoichiometry of reaction, which we can express as follows:

4NH3 + 4NO + O2 → 4N2 + 6H2 O (1.1)

representing the standard SCR reaction, utilised in NOx abatement. In general we can write

νA A + νB B
νP P + νQ Q (1.2)

where stoichiometric coefficients νi represent how many moles of a given species are consumed/produced per unit extent of
reaction ε = dnA/νA = dnB/νB = dnP/νP = dnQ/νQ = dni/νi . By convention, we consider positive νi for products, and negative for
reactants.
We can also introduce some important quantities such as:

• conversion of species A
nA,converted n◦ − nA
χA = = A ◦ (1.3)
nA,fed nA

• selectivity of A to product P
nP,produced/|νP | nP − n◦P |νA |
SP,A = = (1.4)
nA,converted/|νA | n◦A − nA |νP |

• yield of P from A
nP,produced/|νP | nP − n◦P |νA |
ηP,A = = = χA · SP,A (1.5)
nA,fed/|νA | n◦A |νP |

1.1 Equilibrium criteria


Let’s now consider a closed system containing an arbitrary number of species, which is in equilibrium with its surroundings, that is
T = Tsurr and p = psurr . The II law of thermodynamics states that

dS + dSsurr ≥ 0 (1.6)
δQsurr
where the equality holds for reversible transformations between equilibrium states: in such an instance, we have dSsurr = Tsurr =
− δQ
T ⇒ dS ≥ δQ
T : we can plug this result into the I law of thermodynamics to write

dU = δQ − pdV ⇒ dU − T dS + pdV ≤ 0 (1.7)

which we can further elaborate to get

d (U + P V − T S)p,T = dU + pdV +   − T dS − 
V dp ≤ 0
SdT


dG|p,T ≤ 0 (1.8)

we can interpret this result as follows: any spontaneous transformation will decrease its Gibbs free energy until equilibrium is
reached when dG|p,T = 0.

1
Considering G = G (T, p, ni ) = G (T, p, n◦i + νi ε), we can write its total differential as
     
∂G ∂G ∂G
dG = dT + dp + dε (1.9)
∂T p,ni ∂p T,ni ∂ε T,p

that becomes, for a given temperature and pressure,


 
∂G
dG|T,p = dε , ∆G · dε
∂ε T,p

we can thus reformulate our equilibrium criterion as ∆G ≤ 0.

1.1.1 Chemical potential

Let’s now consider a homogeneous system with ns species: we can express its differential Gibbs free energy at fixed temperature
and pressure as
ns   ns ns
X ∂G X X
dG|T,p = dni ≡ µi dni = µi νi dε (1.10)
i=1
∂ni T,p,nj6=i i=1 i=1

where we defined
P µi ≡ Ḡi or chemical potential. We can therefore summarise the equilibrium condition for homogeneous systems
as ∆G = i νi µi = 0.
If we consider a system undergoing multiple reactions, we’ll have
ns
X
dni = νi,j dεj
j=1
P
which replaced in the definition of the equilibrium condition give ∆Gj = i µi νi,j = 0.
We can now examine heterogeneous (multiphase) systems; in this case the Gibbs potential can be expressed as G = α Gα
P
where the suffix α indicates the considered phase. In this case we have
np X
X ns np X
X ns
dG|T,p = µα α
i dni = µα α α
i νi dε
α=1 i=1 α=1 i=1

let’s first consider physical transformations, their stoichiometric coefficients are always unitary, therefore the phase equilibrium
β γ
condition α µα α α
P
i νi = 0 translates to µi = µi = µi = µi . We’re hence left with

ns
X np
X ns
X
dG|T,p = µi dnα
i = µi νi dε
i=1 α=1 i=1

of phase transitions cancel out, and α νiα dεα = νi dε. The chemical equilibrium condition can
P
since the signed sum of extents
P
thus be written as ∆G = i νi µi = 0, like in the homogenous case.
Expanding on the concept of chemical potential, we can derive its expression by exploiting the relation
       
∂µi ∂ ∂G ∂ ∂G ∂V
= = = ≡ V̄i (1.11)
∂p T,ni ∂p T,ni ∂ni T,p ∂ni T,p ∂p T,ni ∂ni T,p

where the Schwarz theorem on mixed derivatives was employed. Under the assumption of ideal gas, we may now employ the I.G.
law, which states pV = nRT ⇒ V̄i = RT
p for a pure species. Rearranging, we get

dp
dµi |T = RT = RT · d ln p|T
p T

from which, by means of integration we obtain

I.G.,◦ ◦ p
µI.G.
i,pure (T, p) = µi,pure (T, p ) + RT ln (1.12)
p◦

2
If we now consider a mixture of ideal gases, by definition the properties of each component are equal to those it would have if it
occupied the system’s volume alone, therefore
pi
µ̂i (T, p) ≡ Ḡi (T, p) = G̃i (T, pi ) = µ◦i (T, p◦ ) + RT ln (1.13)
p◦

from which we can differentiate to obtain dµi |T = RT ln pi .


Lewis proposed to extend this to any substance in any physical state by replacing partial pressures with fˆi , or fugacity of the
i-species in the mixture and writing
fˆi
µ̂i = µ◦i + RT ln ≡ µ◦i + RT ln âi (1.14)
fi◦
Departments from the ideal fugacity for both pure species and mixtures can be accounted for with the introduction of a fugacity
coefficient

fi = ϕi (T, p) p (1.15)
fˆi = ϕ̂i (T, p, yi ) pi = ϕ̂i yi p (1.16)

moreover, since the reference state is set at pure, ideal gas at the temperature of the system and p◦ = 1bar (or 1atm), numerically
we have âi = fˆi/fi◦ = fˆi .

Poynting factor If we consider liquid-vapour equilibria (as well as ones between solids and vapours), physical equilibrium imposes
µli = µvi , which can be written as
fil
µli − µvi = RT ln =0
fiv
that is fil = fiv = ϕi,sat (T, psat ) psat . On the other hand, we must provide a correction in case the pressure of the system is not
the one found in saturation conditions at the given temperature: we can employ 1.11 to integrate between the saturation pressure
and the actual pressure:  
∂µi ∂ ln fi
= RT = V̄i
∂p T,ni ∂p T,ni
 l 
vi (p − pi,sat )
fil = ϕi,sat (T, psat ) psat · exp (1.17)
RT
where
n the exponential
o is also known as Poynting correction factor. In our reference state, we can calculate the Poynting factor as
vil (p◦ −pi,sat )
exp RT , so that in general the activity of a pure condensed species becomes

vil (p − p◦ )
 
ali = exp ∼
=1
RT

negligible for all but very high pressures. The same is valid for solid-liquid equilibria.

Liquid mixtures are described by means of an activity coefficient γi = γ (T, p, xi ) such that
 
γi xi fi fi ∼
âi = = γi xi = γi xi (1.18)
fi◦ fi◦

which, by definition, is equal to 1 for ideal mixtures. In case of real mixtures, we may have two limiting behaviours:

• symmetric, when the i-species exists in condensed state at the given T, p

lim γi = 1
xi →1

where the system abides by the Lewis-Randall rule, stating fˆi = xi fi ; assuming ideal gas behaviour for the vapour phase,
we have the Raoult’s law
yi p = xi pi,sat (T ) (1.19)

3
• antisymmetric, in presence of incondensables or solids in absence of solvent

lim γi = 1
xi →0
 
dfˆi
in this region of concentrations we get a slope dxi ≡ Hi or Henry’s constant, so that high diluted dissolved gas
xi =0
respond to Henry’s law
pi = yi p = Hi xi (1.20)

Solutions, on the other hand, are better treated by means of molality mi [molsolute/kgsolvent ]

1.2 Equilibrium of reacting systems


We can now apply
P our understanding of chemical equilibrium to the evolution of reactions: we know that for a multicomponent
system ∆G = i µi νi , and through 1.14 we can evaluate the chemical potential of each species; plugging in, we get
ns
X ns
X ns
Y
∆GR = νi µ◦i + νi RT ln âi = ∆G◦R + RT ln âiνi (1.21)
i=1 i=1 i=1

from this expression, we grouped up the term ∆G◦R = ◦


P
i νi µi function of temperature only since the reference state is set at
fixed pressure
 and composition; moreover, it is equivalent to the variation of Gibbs free energy per unit extent of reaction, since
∆G◦R = ∂G
∂ni = G (1) − G (0), and since G is a function of state we may employ the same method used for the enthalpy of
T,p◦
reaction to write
ns
X
∆G◦R (T ) = νi ∆G◦f,i (T ) (1.22)
i=1

where ∆G◦f,i = ∆Hf,i ◦ ◦


− T ∆Sf,i may be found in empirical tables.
At equilibrium, we have ∆GR = 0, therefore we can write

∆G◦R = −RT ln Keq

where we introduced
∆G◦R
  Y
Keq = exp − = âνi i (1.23)
RT i

or equilibrium constant: this quantity provides a tool to correlate T, p and composition in a reaction problem through the calculation
of ∆G◦R and activity coefficients.
We’re interested in evaluating how the standard Gibbs free energy varies with temperature, and to do so, we employ the following
relation: ! ! !
dG
T d U +pVT −T S d H
T−S H
= = =− (1.24)
dT dT dT T2
p p p

to write
∆G◦
!

 
d RT
R
d ln Keq ∆HR (T )
=− =− (1.25)
dT dT p RT 2
p

known as Van’t Hoff equation. We can integrate this assuming ∆HR 6= f (T ) , which yields

 
ln Keq (T ) −∆HR (298K) 1 1
= −
ln Keq (298K) R T 298K
 ◦

= ∂H = i νi H̄i◦ must be taken into account: to do so, we must
P
More rigorously, the temperature variability of ∆HR ∂ε
T,p
integrate the quantity
◦ ns ◦ ns ns
∂H ◦
     
∂∆HR X ∂ H̄i X ∂ X
= νi = νi = νi c◦p,i (1.26)
∂T p i=1
∂T p i=1
∂ni

T,p ∂T p i=1

4

RT
νi c◦p,i dT + ∆HR

P
obtaining ∆HR (T ) = 298K i (298K). We must integrate as follows:
T T
∆G◦R (T ) ◦
Z   Z
∆HR (T )
d =− dT
298K RT 298K RT 2

T
RT P ◦
∆G◦R ∆G◦R (298K) − ◦ ◦
i νi cp,i dT
Z
(T ) ∆HR (298K) ∆HR (298K) 298K
= + − dT (1.27)
RT R · 298K RT RT 2
 298K

2 Phase equilibria
Gibbs’s phase theorem. Let’s consider a system comprising np phases and ns species: in orderP to define its intensive state, we
need to know the temperature, pressure and composition of each phase, under the constraint that i xα i = 1; this is the same as
solving a system with nv = np · (2 + ns − 1) variables. When thermal (uniform T α = T β = . . . ), mechanical (uniform pα = pβ =
β
. . . ) and chemical (µα
i = µi = . . . i = 1, . . . , ns) equilibria conditions are satisfied, we can write ne = (np − 1) · (2 + ns)
equations, which lead to nfint = nv − ne = 2 + ns − np degrees of freedom, known as Gibbs’s phase rule.

Duhem’s theorem. If we need to define the extensive state of a system, we need to know the total number of moles in each phase,
so that the number of variables becomes nv = np · (2 + ns); on the other hand, we can now write the molar/mass balance for each
component in the form α nα
P
i = ni so that the number of constraints becomes ne = (np − 1) · (2 + ns) + ns. The degrees of
freedom can now be computed as nftot = 2, which is the statement of the Duhem theorem, that states that for any closed system
with a given number of moles of each species, the thermodynamic state is completely determined once two variables are fixed, with
the number of extensive variables given by 2 − nfint .

2.1 Vapour-liquid equilibrium


As previously stated, in order to have equilibrium
between two separate phases, their chemical potential must be equal, that is
µ̂li = µ̂vi ; moreover, since dµ̂i |T ≡ RT · d ln fˆi , by means of integration we obtain that we must have

T

fˆil = fˆiv (2.1)

which, assuming ideal gas behaviour for the vapour phase (ϕ̂i = 1) and ideal mixture behaviour for the liquid phase (γi = 1,) as
well as negligible Poynting factor correction, we get
xi psat,i (T ) = yi p (2.2)

known as Raoult’s law.


We can employ this relationship to plot the bubblepoint and dewpoint curves, being the boundaries after which respectively the
first bubble of vapour and the first droplet of liquid phase occur. This is done by
P
• imposing i yi = 1, we get the expression of the bubble-line
ns
X
xi psat,i (T ) = p (2.3)
i=1

P yi
• imposing i xi = 1, expressing xi = psat,i p and plugging this back in, we have

1
p = Pns yi (2.4)
i=1 psat,i

Given an i-species, we can describe its tendency to partition itself in vapour space by defining the equilibrium ratio Ki = yi/xi ,

which, in the case of ideal gases and ideal mixtures, follows the Raoult’s law and is equal to

psat,i (T )
Ki = (2.5)
p

5
We can employ this relation in flash drums problems in order to evaluate the outlet composition: from the total molar (mass)
balance on the system we get F = L + V , while from the balances on the single species we get zi F = xi L + yi V ; we can define
a new quantity α = V /F so that we can write
zi = xi (1 − α) + yi α
we can now plug in the relation xi = yi/Ki and rearrange into
ns ns
X X Ki zi
yi = =1
i=1 i=1
1 + α (Ki − 1)

the same holds for the liquid phase, where


ns ns
X X zi
xi = =1
i=1 i=1
1 + α (Ki − 1)
both these equations may be used in order to solve a flash calculation, but they present trivial solutions for α = 1 and α = 0
respectively. A more mathematically sound and stable description is obtained by subtracting one from the other and writing
ns ns
X X zi (Ki − 1)
(yi − xi ) = =0 (2.6)
i=1 i=1
1 + α (Ki − 1)

known as Rachford-Rice equation: this formulation decreases monotonically, and will therefore give rise to a single solution.
In case of components that are uncondensable at the system pressure, the Raoult’s law fails to describe their transfer into liquid
phase, and we must resort to a different model. Assuming ideal gas, ideal mixture and negligible Poynting factor, this behaviour is
described by the Henry’s law, stating
yi p = Hi xi (2.7)
nonetheless, a good degree of approximation is achieved by neglecting xi of the uncondensables. Henry constants may be
calculated through Van’t Hoff equation
d ln (Hi (T )) ∆Hides
1
 =−
d T R
with ∆Hides which can be assumed independent on temperature

2.2 Calculation of fugacity/activity coefficients


Deviations from the ideal model become apparent when considering states close to the critical point. In the p-T diagram, this is
most apparent near the cricondenbar and the cricondentherm, respectively maximum pressure and temperature at which we could
still encounter liquid phase: these fall at values higher than the critical point, and we may thus have retrograde condensation, in
which isothermal expansion causes the fluid to condensate and re-evaporate.
In order to evaluate these deviations, we must first define the ideal states. Considering an ideal gas, we have ϕ̂i = 1 by
definition, so that
fˆi yi p
µ̂I.G.
i (T, p, yi ) − µI.G.
i (T, p) = RT ln = RT ln = RT ln yi
fi p
which can be written in terms of Gibbs free energy as

ḠI.G.
i = G̃I.G.
i + RT ln yi

We can use this result to define the ideal mixture condition, where

Ḡid
i = G̃i + RT ln xi

from which we derive


fˆiid
µ̂id
i − µi = RT ln = RT ln xi
fi
from which fˆiid (T, p, xi ) = xi fi (T, p) known as Lewis-Randall law: we may conclude γi (T, p, xi ) = 1 and, in terms of real gases
mixtures, fˆiid = ϕ̂id id
i yi p = yi fi = yi ϕi p from which ϕ̂i (T, p, yi ) = ϕi (T, p).
Moreover, we can ascribe two important features to ideal mixtures:

6
• isochoric mixing, remembering that (∂ Ḡi/∂p)T,ni = V̄i , we have
!
∂ Ḡid
 
i ∂ G̃i
V̄iid = = = Ṽi (2.8)
∂p T,ni ∂p
T,ni
P P
that is, V = i ni V̄i = i ni Ṽi ;
• isoenthaplic mixing, from (∂ Ḡi/∂T )p,ni = −S̄i , we can write
!
∂ Ḡid
 
i ∂ G̃i
S̄iid =− =− − R ln xi = S̃i − R ln xi
∂T p,ni ∂T
p,ni

by definition, we can now write the enthalpy of the species in the mix as:

H̄iid = Ḡid id
i + T S̄i = G̃i + 
RTln
x + T S̃ − RTln
i i 
x) = H̃
i i (2.9)

2.2.1 Deviation from ideality

We can now introduce residual properties, which measure the departure from ideal gas

M R (T, p, yi ) = M (T, p, yi ) − M I.G. (T, p, yi )

and excess properties, which measure the departure from ideal mixture

M E (T, p, xi ) = M (T, p, xi ) − M id (T, p, xi )

Considering the residual Gibbs free energy of the i-species, we may write

fˆi
ḠR I.G.
i = Ḡi (T, p, yi ) − Ḡi (T, p, yi ) = RT ln = RT ln (ϕ̂i (T, p, yi ))
yi p
moreover, if we differenciate this with respect to pressure at constant T, ni , we get

∂ ḠR ∂ ḠI.G.
     
i ∂ Ḡi i
= − = V̄i − ṼiI.G. (2.10)
∂p T,ni ∂p T,ni ∂p T,ni

where we simplified V̄iI.G. (T, p, yi ) = ṼiI.G. (T, p) by definition. We can now employ the generalised ideal gas equation of state
in the form pV = nZRT , where Z is the compressibility factor, unitary for ideal gases. Dividing 2.10 by ṼiI.G. = RTp , we get

∂ ḠR
 
1 i p
= V̄i − 1
RT ∂ ln p T,ni RT

we can express specific volume as a funzion of Z by substituting the generalised equation of state into its definition:
   
∂V RT ∂nZ RT
V̄i = = = Z̄i
∂ni T,p p ∂ni T,p p

which we can plug back into the differential equation to obtain


 
∂ ln (ϕ̂i (T, p, yi ))
= Z̄i (T, p, yi ) − 1
∂ ln p T,ni

which we can integrate from 0 to p (remembering that for pressures tending to zero, the gas behaves ideally, hence ϕ̂i = 1)
Z p 
ln ϕ̂i = Z̄i − 1 d ln p (2.11)
0

The same can be obtained for the activity coefficient, provided excess properties are considered:

ḠE id
i = Ḡi (T, p, xi ) − Ḡi (T, p) = RT ln γi

7
2.2.2 Equations of State

We can approach the evaluation of the compressibility factor by evaluating the equations of state for real systems, which can be
written in the form
pṼ
Z= (2.12)
RT
the most common representation is given by the virial expansions, which can be expressed equivalently as

B C D
Z =1+ + + + ...
Ṽ Ṽ 2 Ṽ 3
Z = 1 + B 0 p + C 0 p2 + D0 p3 + . . .
where we can switch from one to the other by applying the definition in 2.12. The physical representation of the terms in these
expressions is the statistical mechanical interaction between a number of bodies equal to the exponent of the term.
More accurate equations, such as the cubic equations of state, take into account the finite volume of these bodies, known as
covolume b:
RT a
p= − (2.13)
Ṽ − b Ṽ 2
is known as Van der Waals equations, and truncates the virial expansion at two body interactions only. The cubic nature of this
equation becomes apparent when it is expressed as a function of Ṽ or Z
 
3 RT a ab
Ṽ − b + Ṽ 2 + Ṽ − =0
p p p
abp2
 
bp ap
Z3 − + 1 Z2 + 2 Z − 3 =0
RT (RT ) (RT )
If we plot these equations in the p-Ṽ diagram, we get a monotonic decreasing behaviour for T ≥ TC , with an inflexion point for
the isotherm passing through the critical point TC , pC , while for T < TC a particular behaviour is observed, with a minimum and
a maximum point within the saturation dome: although this does not correspond to the real evolution of a system, the points within
the bell can be ascribed to metastable states.
The presence of three coincident solutions at the critical point allows us to determine the values of the a, b coefficients by
comparing the Van der Waals cubic equation evaluated at TC , pC to
 3
Ṽ − ṼC =0

Ṽ 3 − 3Ṽ 2 ṼC + 3Ṽ ṼC2 − ṼC3 = 0


From Van der Waals equations we can derive more complex equations of state, such as the Soave-Redlich-Kwong or the
Peng-Robinson, which provide better fitting of the experimental data; we can write these as

RT a (T )
p= −  
Ṽ − b Ṽ + εb Ṽ − σb

α(T ,ω)R2 T 2
with b=Ω RT
pC , a (T ) = Ψ
C R
pC
C
, where TR = T/TC is the reduced temperature, and ω = −1 − log (psat,R (TR = 0.7))
known as Pitzer acentric factor.
The existence of tree different acceptable values for Z in the VLE region, the lowest one corresponding to liquid phase and the
highest one to vapour, allows us to introduce a new approach for solving equilibria, known as the ϕ-ϕ method, as opposed to the
conventional γ -ϕ one.

Theorem of corresponding states. Moreover, we notice that a, b only depend on reduced properties, this means that fluids at
same TC and pC will approximately have the same volumetric behavior, thus deviating from ideal behavior by the same degree.

When we consider mixtures, we need to modify our correlation by applying mixing rules: the simplest of these are Van der
P P 1/2
2
Waals rules, which apply a linear average for b = i yi bi and a quadratic average for a = i yi ai , and calculates fugacity
   
∂na
coefficients for mixtures by applying āi = ∂ni and b̄i = ∂nb
∂ni .
T,nj6=i T,nj6=i

8
2.2.3 Activity coefficients

When intermolecular interactions between liquid are too complex to be described by EoS, the ϕ-ϕ method fails, and we must
evaluate the activity coefficients of the mixture. Recalling the definition of excess properties, we may write
ḠE id
i (T, p, xi ) = Ḡ (T, p, xi ) − Ḡi (T, p) = RT ln (γi (T, p, xi ))
E
on the other hand, since excess Gibbs free energy is a state function, we may compute the differential of the quantity G /RT as
1

GE dGE dGE GE
 
d RT
d = + GE · dT = − dT
RT RT dT RT RT 2
where dGE = −S E dT + V E dp + ḠE E E E
P
i i dni from classical thermodynamics: plugging in both results, being G = H − T S
we get
ns
GE S E dT
 V E dp X HE S E dT
 

d =−  + + ln γi dni − 2
dT + 
RT RT RT i=1
RT RT
We can thus write the following:
E E E
HE
     
∂ G /RT ∂ G /RT ∂ G /RT
= V E; =− ; = ln γi
∂p T,ni ∂T p,ni RT 2 ∂ni T,p,nj6=i

from this, we can apply the Schwarz theorem of cross derivatives, and write
H̄iE
   
∂ ln γi ∂ ln γi
= V̄iE ; =−
∂p T,ni ∂T p,ni RT 2
these allow us to evaluate pressure and temperature dependence of activity coefficients from the experimentally determined effects
of mixing on partial volume and enthalpy.
Gibbs-Duhem equation. Since GE is an equation of state, we can apply additivity of partial molar properties GE =
P
i ni Ḡi
E
and, dividing by n, G̃E = G /n = i xi Ḡi to write
P

ns
G̃E X
= xi ln γi
RT i=1

at the same time, we have !


ns
X ns
X
E

dG̃ = d xi Ḡi = dḠi xi + Ḡi dxi (2.14)
i=1 i=1

and, as previously shown, dG̃E = −S̃ E dT + Ṽ E dp +


P
i Ḡi dxi : plugging the first into the second, we get
X ns
E
dG̃ = −S̃ E dT + Ṽ E dp +  E
dG̃ − dḠE
i xi = 0
 

i=1

which, at constant T, p, yields


ns
X ns
X
dḠE
i = RT xi d ln γi = 0 (2.15)
i=1 i=1
P
or i xi d ln γi = 0.
An interesting quantity from the experimental point of view is G̃E/RT x1 x2 in binary mixtures, in particular the limits for x1 → 0
and x1 → 1:
 
  :0
 E   E   E 
d ln γ2 x2 ) + (ln γ1 − ln γ2 ) 

G̃/RT G̃ /RT dG̃/RT  (d ln γ1
x1
+ dx
1
lim = lim = lim = lim   =
x1 →0 x1 x2 x1 →0 x1 x1 →0 dx1 x1 →0  dx
 1

γ1
= ln γ1∞ = A12
= lim ln
γ2 x1 →0
 E 
G̃ /RT γ2
lim = lim ln = ln γ2∞ = A21
x1 →1 x1 x2 x1 →1 γ1

9
where we applied both 2.15 and 2.14. We may use these limits to correlate experimental data through a linear trend, expressed by

G̃E/RT
= A21 x1 + A12 x2
x1 x2
if we multiply this by the total number of moles, we get

GE nG̃E n1 n2
= = (A21 n1 + A12 n2 )
RT RT n1 + n2
from this, if we derive with respect to n1 and n2 we obtain respectively
  E 
∂ G
RT
ln γ1 =   = x22 [A12 + 2x1 (A21 − A12 )] (2.16)
∂n1
p,T,n2

and ln γ2 = x21 [A21 + 2x2 (A12 − A21 )], known as Margules equation, from which activity coefficients calculation can be made.
An alternative representation, which actually draws on the physical phenomenon and can be extended to wider temperature ranges
and multi-component mixtures, is given by the Wilson model, which can be summarised by
 
ns ns
x Λ
P k ki
X X
ln γi = 1 − ln  xj Λij  −
j=1 j xj Λkj
k=1

Ṽj aij 
where Λii = 1 and Λij = exp RT with aij = f (T, xi , xj ).
Ṽi

Azeotropes Some mixtures show a particular behaviour, where we observe evaporation and condensation with no change in
composition. This is equivalent to a stationary point (minimum/maximum) in the p-x diagram. This can be demonstrated by
applying the Gibbs-Duhem equation, which is valid for any function of state: if we consider Ḡi/RT = ln fˆi , we have i xi ln fˆi = 0
P
and thus, dividing by dx1 ,
d ln fˆ1 d ln fˆ2
x1 + x2 =0
dx1 dx1
where fˆil (T, p, xi ) = fˆiv (T, p, yi ) = ϕ̂vi (T, p, yi ) yi p; if we plug this into the former equation, we get

v v
   
1 dp 1 dy1 d ln ϕ̂ 1 dy2 d ln ϕ̂
+  1 + x2 +  2 =0
 
+ x1
p dx1 y1 dx1 dx1 y2 dx1 dx1
P
where we simplified xi ln ϕ̂i = 0: rearranging we get
i
     
1 dp dy1 x1 x2 dy1 x1 (1 − 1 ) − y1 (1 − 
y x
1) dy1 x1 − y1
= − = =
p dx1 dx1 y1 y2 dx1 y1 y2 dx1 y1 y2

that implies that when we have a stationary point dp/dx1 = 0 and x1 = y1 . In order to evaluate the composition of an azeotropic
equilibrium, we must consider the equilibrium equation yi = Ki xi i = 1, 2: from this we can define a relative volatility α12 =
K1/K2 where K = γi ϕsat,i psat,i . In order to have an azeotrope state, we must have either lim
i ϕ̂i p x→0 α12 > 1 ∧ limx→1 α12 < 1 or
vice-versa, since for yi = xi we must have α = 1.

2.2.4 Property changes of mixing

As we previously stated through the definition of ideal mixtures, Ḡid id id


i (T, p, xi ) = G̃i (T, p)+RT ln xi but V̄i = Ṽi and H̄i = H̃i .
E id mix
Mixing
P property changes, though are not the same as excess properties G̃ = G̃ − G̃ : they are defined as M̃ = M̃ −
x M̃
i i i , so in the case of specific Gibbs free energy
ns
X ns
X ns
X
G̃mix = G̃ − xi G̃i = G̃E + G̃id − xi G̃i = G̃E + RT xi ln xi (2.17)
i=1 i=1 i=1

10
on the other hand, for enthalpy of mixing, we have H̃ mix = H̃ − E id  id
P P
− 

i xi H̃i = H̃ +H̃ i xi H̄i , which allows us to

experimentally evaluate the temperature variability of activity coefficients in a mix through equation
H̄ E
 
∂ ln γi
=−
∂T p,ni RT 2
In order to have a spontaneous reaction, we must have dGT,p ≤ 0; if we consider a mixing process, we may assume the
evolution of Gibbs free energy to be a linear function of the transformation coordinate:
ns
X
G̃ (ε) = εG̃mix + xi G̃i
i=1
 
∂ G̃
= G̃mix = G̃E + RT xi ln xi ≤ 0 or H̃ E − T S̃ E ≤ RT
P P
from which ∆G̃ = ∂ε i i xi ln xi . Moreover, experimental
T,p,ni
data show the need for an equilibrium stability condition in order to avoid concavity inversion with consequent phase separation
!
∂ 2 ∆G̃
>0
∂x21
T,p

which we can evaluate by calculating the partial derivatives of ∆G̃/RT = G̃mix/RT = G̃E/RT + x1 ln x1 + x2 ln x2 with respect to
x1 :    E 
∂ ∆G̃/RT ∂ G̃ /RTx1 x2
= + ln x1 +  − ln x2 − 
∂x1 ∂x1 x1 x2
and again  2   2 E 
∂ ∆G̃/RT ∂ G̃ /RT 1 1
= + + >0
∂x21 ∂x21 x1 x2
E
moreover, G̃ /RT = x1 ln γ1 + x2 ln γ2 , so
 E 
∂ G̃ /RT d ln γ
1
 d ln γ
2

= x1 + x2 + ln γ1 − ln γ2
∂x1  dx1  dx1
 2 E 
∂ G̃ /RT d ln γ1 d ln γ2 1 d ln γ1
2 = − =
∂x1 dx1 dx1 x2 dx1
where we applied the Gibbs-Duhem law in both cases; we can plug this back into the disequation to get

1 d ln γ1 x2+ x:1
1
 > − (2.18)
x
 2 dx 1 x
2 x 1
d ln γi
which can be generalised as dxi > − x1i .

2.3 Liquid-liquid equilibrium


The instability effects are very unlikely to occur in vapour-liquid equilibria, but may be significant in liquid solution, where intermolec-
ular forces are stronger andP cause immiscibility regions, where two distinct liquid phases are present, to be in place. In this region
we have G̃mix = G̃E − RT i xi ln xi ≥ 0: we can neglect the influence of pressure as we’re considering liquid phases, therefore
we shall consider the variation of the excess Gibbs free energy, which we conveniently already know being equal to
E
H̃ E
 
∂ G̃ /RT
=−
∂T p,ni RT
since immiscibility is more likely the larger G̃E/RT
becomes, upon increasing temperature we have a larger region for exothermic
mixing and a smaller one for endothermic mixing.
As usual, equilibrium for split phases must respect the equality of chemical potential, yielding
 
fˆiα (T, p, xα ˆβ T, p, xβ
i ) = fi i

β β β
where fˆiα = γiα xα l l α α α
i fi where fi = fi = fi ; plugging this in, we get γi xi = γi xi from which, rearranging, we get

γiα xβi
ln = ln (2.19)
γiβ xα
i

11
3 Kinetics
Kinetics of chemical reactions studies the rate of transformation of reacting systems towards chemical equilibrium, through the
evaluation of the reaction rate
1 dnA 1 dnP 1 dni
r= = = ··· = (3.1)
Z νA dt Z νP dt Z νi dt
where, given a generic reaction stoichiometry like 1.2, Z is an appropriately chosen extensive quantity such as reactor volume,
catalyst surface area, number of active sites, etc...

3.1 Ideal reactors


This quantity cannot be directly measured, but it can be inferred from mass balances in the form
 
DCi ∂Ci ∂Ci ∂ ∂Ci
= + wi = Di + νi r (3.2)
Dt ∂t ∂xi ∂xi ∂xi
from which we can formalize the definition of the following ideal reactors:
• Continuously Stirred Tank Reactor, where the presence of a mixer allows us to consider a steady state condition with no
diffusive flows, thus writing, for an outlet flow FA [mol/s] of species A

0 = FA◦ − FA + νA rV
FA − FA◦ V̇ ◦ C ◦ χA
rνA = = (CA − CA )= A (3.3)
V V τ
where τ = V /V̇ is known as the average residence time;

• Perfectly Mixed Batch Reactor, a closed system in which mixing makes diffusion negligible: we can write
dnA
= νA rV
dt
from which we obtain the definition of reaction rate for a constant volume system
1 dnA dCA ◦ dχA
rνA = = = CA (3.4)
V dt dt dt
• Plug Flow Reactor, where we may consider a differential portion of reactor of length dz and volume dV = S · dz
0 = −dFA + νA rdV
dFA dχA dχA
rνA = = FA◦ =  
dV dV d FV◦
A

where V /F ◦
A is known as space time; if the molar flow is constant along the axial direction (equimolar stoichiometry), then

also the velocity wz will be constant, and we can write


dCA 
0 = wz Sdz + νA rAdV

dz
dCA ◦ dχA
νA r = = CA (3.5)
dτ dτ

3.1.1 Arrhenius equation

Experience tells us that the temperature dependence of r = r (T, p, yi ) can be grouped up into another parameter to write
r = k (T ) · f (p, yi ): this is known as rate constant, and it can be determined through the Arrhenius equation
 
Eact
k = A · exp − (3.6)
RT
where A is known as the frequency factor and Eact is known as activation energy, and can be evaluated by plotting ln k against
1/T .

12
3.1.2 Power law expressions

Some experimental correlations were developed in order to fit the composition and pressure dependence of reaction rates, known
as power law expressions. They may appear in the form
a b r p s
r = k (T ) CA CB CR CP CS

r = k (T ) paA pbB prR ppP psS


where exponents (which may be R 0) represent the reaction order of the corresponding species and their sum is the global reaction
order. Only in very few cases these quantities are equal to the stoichiometric coefficients of the reaction, like in that of elementary
reactions.
Elementary reactions are mono-/bi-molecular reaction events, which can be described by means of stoichiometry and have
reaction rates

− |ν | |ν |
r = k + CA A CB B ; →
− |ν | |ν |
r = k − CP P CQ Q
where, in general, thermodynamic equilibrium conditions must still be valid


r

∆G

k+

∆G◦


− = exp ; = exp
r RT k− RT

3.2 Study of reactions


Rate constants of elementary reactions may be theoretically estimated. Collision theory provides a description of the frequency
factor A, which strongly overestimates the actual number of collisions which actually result in a reaction. A more accurate model
involves quantum mechanics, and it is provided by the transition state theory (developed by Eyring): if we plot the behaviour of
Gibbs free energy as a function of the extent of reaction, we notice that reactants must overcome a maximum, called transition state
∆‡ G◦ , in order to achieve a structure known as activated complex, which is in quasi-equilibrium with both reactants and products,
and has equal probability to be converted into either.
This way, we compute
‡ ◦
 

− ∆ G
r = k ‡ exp − pA pB (3.7)
RT
where k ‡ =
kT
n ‡ ◦
oh with kn Boltzmann’s
o constant and h Planck’s constant; from which we can obtain Arrhenius law by expanding
n ‡ ◦o
∆ G ∆‡ H ◦
exp − RT = exp − RT exp ∆ RS thus obtaining

∆‡ S ◦ ∆‡ H ◦
   

− kT
r = exp · exp − pA pB
h R RT
n o
∆‡ S ◦
where A = kT
h exp R and Eact = ∆‡ H ◦ .

3.2.1 Radical chain reactions

Often, even bi-molecular stoichiometry fails to be described as elementary reactions; an example of such reaction is the one for the
production of hydrogen bromide, which was extensively studied by Bodenstein and Lind

H2 + Br2 → 2HBr

the observations on the reaction rate suggested that the reaction followed steps that were much different than the ones suggested
by the stoichiometry: the actual elementary steps involve one initiation, three propagations and one termination reaction, and the
process is known as chain reaction:  k


 M? + Br2 →1 M + 2Br·
 k2
Br · +H2 → HBr + H·



k
Br2 + H· →3 HBr + Br· (3.8)

 k4
H · +HBr →

 H2 + Br·

 k5
M + 2Br· → M? + Br2

13
where H·, Br· are radical species, which act as chain carriers for the reaction and are consumed and produced very quickly; and
M? is a third body which provides the energy for initiation and absorbs the energy released during termination. We can thus express
the rate of reaction by accounting for each of these elementary step reactions as

dCHBr
= k2 CBr CH2 + k3 CBr2 CH − k4 CH CHBr
dt
dCH
= k2 CBr CH2 − k3 CBr2 CH − k4 CH CHBr ≈ 0
dt
(
dCBr (( (((( 2
= 2k1 CBr2 −k2 CBr CH2(+(k3(
C( (C
Br2 H + k 4 H HBr − 2k5 CBr ≈ 0
C C
dt (( ( ( (
where we approximate the radical species to be in quasi steady state due to their high reactivity, so that no accumulation occurs.
We can thus eliminate the dependance on the concentration of radical species by plugging
 1/2
k1 1/2
CBr = CBr2
k5
 1/2 1/2
k1
k2 k5 CBr2 CH2
CH =
(k3 CBr2 + k4 CHBr )
into the first reaction rate equation, to get
 1/2  
dCHBr k1 1/2 k3 CBr2 − k4 CHBr
= k2 CBr2 CH2 1 + =
dt k5 k3 CBr2 + k4 CHBr
 1/2 1
/2
k2 kk15 CBr2 CH2 (2 k3
CBr
+ k 
2 4 CHBr − 
 k
 4 CHBr )

=   =
k3  1 + k4 CHBr
CBr

 2 k3 CBr2
 1/2 1
/2
2k2 kk15 CBr2 CH2
=
1 + kk43 C HBr
CBr 2

which corresponds to the formulation found empirically by Bodenstein and Lind.

Quasi-steady state We can adopt this approximation when we know that the rate of reaction of a given species is much larger
than the other species involved in a chain reaction. Given the general reaction, with k2  k1

k k
A →1 B →2 C

we may express the reaction rates by means of elementary step power laws

dCA ◦ −k1 t
 dt = −k1 CA
 CA = CA e
dCB ◦ −k1 t
dt = k1 CA − k2 CB = k1 CA e − k2 CB
 dCC

dt = k C
2 B

k1 ◦
e−k1 t − e−k2 t ; since the second reaction occurs much

solving the integral by variable separation, we get CB = k2 −k1 CA
k1 ◦ −k1 t
faster than the first, we can simplify CB ≈ k2 CA e = kk12 CA : plugging this back into the differential equation, we get dC
dt ≈
B

k1 CA − k2 kk21 CA = 0.

3.2.2 Combustion reactions

In combustion reactions, power laws are not suitable to describe the evolution in time of the system, as we have an initial slow
induction phase (the length of which depends on the inverse of temperature), then a very steep increase of reaction rate with the

14
formation of numerous intermediates, such as CO, which are slowly transformed into CO2 and H2 O. These reactions may be
described by slightly modifying the chain reaction model:
 k
 A → C?
1

 initiation

? k2 ?
A + C → A + αC branching



k
A + C? →3 P + C? propagation (3.9)

? k4

C → W wall termination



 ? k5

C →G gas phase termination

we can thus evaluate the rate of reaction of chain carriers under the assumption of quasi-steady state
dCC?
= k1 CA + (α − 1) k2 CA CC? − k4 CC? − k5 CC? ≈ 0
dt
that is
k1 C A
CC? =
k4 + k5 − (α − 1) k2 CA
from which the rate of product generation becomes
2
dCP k3 k1 C A fr
= k3 CA CC? = = (3.10)
dt k4 + k5 − (α − 1) k2 CA fw + fg − (α − 1) fb
it’s evident how for a sufficiently large α this rate may tend to infinity, leading to explosive combustion. We can plot this behaviour
on the p-T diagram to identify two explosion limits:

• for high p, we have fw  fg , so if fw = (α − 1) fb explosions occur; fw increases on increasing the surface to volume
ratio, and consequently decreases for bigger reactor sizes;

• for low p, fg  fw and we have explosions when fg = (α − 1) fb .

3.2.3 Catalytic reactions

Catalysis is a kinetic phenomenon in which reactants interact with another substance that changes the reaction steps followed by
the reactants, which has the twofold benefit on yield of increasing the conversion rate of reactants, and improving the selectivity
towards the desired products. In order to study this reactions, it’s best to consider an alternative definition of the reaction rate which
either takes Wcat weight of the catalyst or the number of active sites into account.
Catalysts may be homogeneous (in the same phase of the reactants), but in most industrial processes we consider solids
(heterogeneous reaction) such as pellets or monoliths. The process may be divided into seven steps:

1. diffusion out of the fluid bulk and onto the catalytic surface (external transport)

2. diffusion within the pores in the surface (intraparticle transport)

3. adsorption of reactants onto the catalyst surface

4. surface reaction

5. desorption of products from the surface

6. intraphase transfer out of the porous structure

7. interphase transfer to the fluid bulk;

we’re mainly interested in steps 3 to 5, as they involve chemical processes.


In order to study these, we may adopt the empirical approach of power laws, which fares well in correlating experimental data
with simple mathematics, but has feeble fundamental background and its results may not be extrapolated. If we need a more
solid theoretical description, we must resort to the Langmuir-Hinshelwood-Hougen-Watson (LHHW) model. This model studies a
sequence of three steps, comprising adsorption, surface reaction (which is assumed to be the Rate Determining Reaction) and
desorption, under the assumption of ideal uniform catalytic surface (described by Langmuir isotherms) and quasi-steady state of
adsorbate intermediaries.

15
Langmuir theory of adsorption This model is based under the following assumptions:

• the nature and number of total sites does not change with time;
• the enthalpy of adsorption is not a function of coverage;
• attractive/repulsive forces between adjacent molecules are neglected.
We can express the rates of adsorption and desorption as

rads = kads pA Cσ? ; rdes = kdes CA?

= exp − ∆G

from which we can define Kads = kads/kdes
RT
ads
: in equilibrium, we have rads = rdes , and thus
?
CA = Kads pA Cσ?
? ?
we may express this in terms of Ctot = CA + Cσ? total number of sites, we get
?
? Kads Ctot pA
CA = (3.11)
1 + Kads pA
?
known as Langmuir isotherm. The behaviour of surface coverage ϑA = CA /C ?tot with temperature is dictated by the Van’t Hoff
equation
d ln Kads d (−∆Gads/RT ) ∆Hads
= =
dT dT RT 2
and, since adsorption is an exothermal phenomenon, ϑA will decrease at higher temperatures, in contrast to what would be
predicted by Arrhenius equation.

Rate determining step We can consider the following elementary steps for catalytic reactions:



 A + σ → Aσ A adsorption
B + σ → Bσ B adsorption



Aσ + Bσ → P σ + Qσ surface reaction (3.12)

Pσ → P + σ P desorption





Qσ → Q + σ Q desorption

as previously stated, adsorbate intermediates are much more reactive than species in the fluid phase, therefore we may employ a
steady state approximation where

r = r1+ − r1− = r2+ − r2− = r3+ − r3− = r3+ − r3− = r4+ − r4− = r5+ − r5−

nonetheless, there is one step which consumes all of the driving force (∆G) available for the reaction: that is the RDS, and we have
|ri+ |/|ri− | ∼
= 1 ∀i 6= RDS .
This means that while adsorption and desorption steps may be considered at quasi-equilibrium and described by Langmuir
isotherms
? A Q
CA = Kads pA Cσ? ; ?
CB B
= Kads pB Cσ? ; CP? = Kads
P
pP Cσ? ; ?
CQ = Kads pQ Cσ?
as far as the RDS is concerned, we may calculate its reaction rate as

CP? CQ
?
 
r= k3+ CA
? ?
CB − k3− CP? CQ
?
= k3+ ? ?
CA CB − (3.13)
KSR
we can solve this in an easier way by manipulating the expressions for the concentration of active sites:
 
? ? ? Q
Ctot = CA + CB + CP? + CQ
?
+ Cσ? = Cσ? 1 + Kads
A B
pA + Kads P
pB + Kads pP + Kads pQ

from which by expliciting Cσ? and plugging it into a general Langmuir isotherm, we get
i ?
Kads pi Ctot
Ci? =
1+ A p
Kads B p + KP p + KQ p
+ Kads
A B ads P ads Q

16
which we can now use to solve 3.13:
Q
K P Kads
   
? 2 ? 2 pP pQ
k3+ Ctot A
Kads B
Kads pA pB − K A adsB p P p Q k +
3 Ctot K A
ads K B
ads p A p B − K
ads Kads KSR eq
r=  2 =  2 (3.14)
A B P Q A B P Q
1 + Kads pA + Kads pB + Kads pP + Kads pQ 1 + Kads pA + Kads pB + Kads pP + Kads pQ

known as the Langmuir-Hirshelwood-Hougen-Watson rate equation. While the assumptions in this model may not be the most
stringent, the general structure of this equation holds for every set of assumptions, in particular

[kinetic factor] · [driving force group]


r=
[adsorption group]

In general, an increase of partial pressure of reactant species A may result in an initial increase of reaction rate due to better
adsorption and driving force, but it soon causes a reduction due to the lower adsorption of the reactant species B . Increasing total
pressure may lead to adsorption of spectator species.
As far as temperature is concerned, if we consider Arrhenius law we see an increase of kinetic coefficients for greater temper-
atures, but since Kads is negatively affected by temperature, we might have a complex trend of the reaction rate as a function of
temperature.

4 Fuel cells
Fuel cells are electrochemical devices that are able to directly transform chemical energy into electrical energy, which allows to
achieve higher theoretical efficiency than thermal machinery, which is bound by ideal thermodynamic cycles as an upper limit.
The components of a fuel cell system include:

• a reformer upstream the fuel cell which converts hydrocarbon fuel to syngas or pure hydrogen;
• electrodes, anode (-) where the oxidation B → B+ + e− reaction takes place, and cathode (+) where reduction A + e− → A−
occurs; these must be porous systems, as they have to act as catalysts for the aforementioned reactions1 ;

• an electrolyte, a medium that is very permeable to ions, and allows their flow between the electrodes, while barring the flow
of fuel or oxygen, and non conductive to electricity;

• an external circuit that allows the flow of charge from anode to cathode: we can extract useful work by connecting a load;
• inlets and ducts to allow the feed and exhaust flows to enter or exit the system;

• electronic devices such as transformers and inverters, to convert the low voltage direct current to more useful forms of
electricity.

Different fuel cell technologies are available, differentiated on chemical reaction, electrodes and electrolyte: some examples of the
most prominent are shown in the following box.
Type Anodic reaction Cathodic reaction
Polymer Electrolyte FC H2 → 2H+ + 2e− 1 + −
2 O2 + 2H + 2e → H2 O
− − 1 − −
Alkalyne FC H2 + 2 (OH) → 2H2 O + 2e 2 O2 + H2 O + 2e → 2 (OH)

H2 + CO= 3 → H2 O + CO2 + 2e 1 − =
Molten Carbonate FC =
CO + CO3 → 2CO2 + 2e − 2 O2 + CO2 + 2e → CO3

Phosphoric Acid FC H2 → 2H+ + 2e− 1 + −


2 O2 + 2H + 2e → H2 O
= −
H2 + O → H2 O + 2e 1 = −
Solid Oxide FC
CO + O= → CO2 + 2e− 2 O2 → O + 2e

The global stoichiometry is, in general, either H2 + 21 O2 → H2 O or CO + 21 O2 → CO2 .


1
The link between stoichiometry and charge is represented by the quantity F = NA · e = 96485.33C known as Faraday constant.

17
4.1 Fuel cells thermodynamics
Since useful work is being extracted from the system, in the form Wel = nF E where n is the number of electrons produced at the
anode per mole of fuel and E is the reversible electric potential difference between the electrodes.
Equilibrium condition tells us that
ns
Y
∆G + Wu = 0 ⇒ ∆G = −nF E = ∆G◦ + RT ln aiνi (4.1)
i=1

which, for an ideal gas, can be rearranged into Nernst’s law


ns ns
∆G◦ RT Y νi RT Y νi
E=− − ln pi = E ◦ − ln p
nF nF i=1 nF i=1 i

giving the ideal maximum potential2 , which is never achieved since the cell never reaches equilibrium during operation.
From this, we can define a set of efficiencies that may help us characterize fuel cells:
Wu ∆G
• reversible efficiency, ηid = −∆Hf◦ = ∆Hf◦ which is the ratio of the ideal work derived by Nernst’s law and the heating value
of the fuel;
V
• voltage efficiency, ηv = E comparing the reversible voltage to the real one, which is higher due to kinetic irreversibilities;
I
• current efficiency, ηI = = U , or fuel utilization factor (χf ) due to the non complete conversion of fuel in the reaction
Imax
due to thermodynamics and kinetics;

• global efficiency
η = ηref ηid ηv ηI ηinv (4.2)

which includes all of the former, the efficiency of the upstream reformer system and the efficiency of inverters/transformers.

Studying the effects of the thermodynamics state on fuel cells efficiency, we know that hydrogen combustion decreases the number
of moles of the system, making the variation of entropy negative: we have
   
∂E ∂ ∆G/nF ∆S
=− = <0
∂T p ∂T p nF

therefore an increase in temperature is met with a reduction of reversible potential; nonetheless, losses due to kinetic irreversibilities
are reduced with temperature, and higher temperatures allow for a more efficient recovery of enthalpy losses. Therefore, in general,
high temperature fuel cells are preferred.
As far as fuel utilization is concerned, Nernst’s law applied on H2 combustion tells us

RT pH2 O
E = E◦ − ln 1/2
nF pH2 pO2

therefore increasing the partial pressure of H2 O will have a negative effect on potential, which is also magnified by temperature.
Pressures have an all around positive effect on fuel cell performance; not only do they improve kinetic factors, but we also have
    P
∂E ∂ ∆G/nF ∆V iνi V̄i
=− =− =− >0
∂p T ∂p T nF nF

since the number of moles decreases, with a consequent contraction of volume. Nevertheless, fuel cell operation (sealing and
manufacturing) proves very complicated at high pressure, and usually moderate values are preferred.
2
for H2 systems, E ◦ = 1.229V for liquid phase at 25◦ C

18
4.2 Polarization curve
The polarization curve shows the real behaviour of fuel cell potential when increasing the intensity of current produced; the deviation
from ideality shows a predictable behaviour, that is a greater reduction in voltage efficiency the farthest we are from thermodynamic
equilibrium (greater reaction rate causes larger current).
Many factors affect the real behaviour of cell voltage, such as kinetics of electrochemical reactions (activation polarization),
transfer through the electrolyte (ohmic polarization), mass transfer from the bulk of the gaseous phase to the electrode surface
(concentration polarization); moreover, we may have further pressure losses due to reactant crossover and internal short-circuit
currents, responsible for losses up to 0.2V.

4.2.1 Activation losses

Unlike normal catalytic reactions, electrodic half-reactions are able to exploit electrical energy in order to overcome the ∆‡ G◦ energy
barrier: this is not desirable, as it depletes some of the available electrical potential. This polarization potential, or overpotential,
can be calculated through Tafel’s equation
RT i
∆Vc = − ln (4.3)
αnF i0
where α depends on the nature of the reaction and the materials of the electrodes. It may appear as though temperature affects
this quantity negatively, but it causes a much larger increase in i0 = A exp − ERT

act
f (C).
Activation losses cause major concern, expecially at the cathode of low temperature fuel cells; they may be counteracted upon
by using high porosity, higher activity electrodes.

4.2.2 Ohmic polarization

These losses are due to resistance to the flow of ions through the electrolyte, and the flow of electrons through electrodes and
junctions, following the Ohm’s law ∆V = R · I . They may be reduced by using electrolytes that are highly conductive to ions, as
well as reducing the distance between electrodes.

4.2.3 Concentration polarization

Mass transfer losses occur when the bottleneck for the fuel cell operation is the insufficient transport of reactants (and thus charge)
to the reacting surface by diffusion, which can be expressed as

dq nF D
i= = (Cb − Cs )
dt δ
when mass transfer control is complete, reactants are consumed as soon as they reach the surface, and Cs = 0: we may wrtite a
limiting condition
nF D
imax = Cb (4.4)
δ

19

You might also like