Notes
Notes
INTERMEDIATE THERMODYNAMICS
Joseph M. Powers
updated
12 March 2024, 9:51pm
2
Preface 7
1 Review 9
2 Cycle analysis 25
2.1 Carnot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1.1 Analysis for a calorically perfect ideal gas . . . . . . . . . . . . . . . 25
2.1.2 Exergy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2 Rankine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.2.1 Classical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.2.2 Reheat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.3 Regeneration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.4 Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.2.5 Cogeneration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.3 Air standard cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.4 Brayton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.4.1 Classical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.4.2 Regeneration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.4.3 Ericsson cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.4.4 Jet propulsion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.5 Reciprocating engine power cycles . . . . . . . . . . . . . . . . . . . . . . . . 62
2.6 Otto . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.7 Diesel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.8 Stirling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.9 Refrigeration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.9.1 Vapor-compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.9.2 Air standard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.10 Rejected thermal energy on a national scale . . . . . . . . . . . . . . . . . . 78
3 Gas mixtures 83
3.1 Some general issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.2 Ideal and non-ideal mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.3 Ideal mixtures of ideal gases . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3
4 CONTENTS
Appendix 295
Bibliography 315
These are lecture notes for AME 50531 Intermediate Thermodynamics (AME 54531 for
students in our London and Rome programs), the second of two undergraduate courses in
thermodynamics taught in the Department of Aerospace and Mechanical Engineering at
the University of Notre Dame. Most of the students in this course are juniors or seniors
majoring in aerospace or mechanical engineering. The objective of the course is to survey
both practical and theoretical problems in classical thermodynamics.
The notes draw heavily on the text specified for the course, Borgnakke and Sonntag’s
(BS) Fundamentals of Thermodynamics, Ninth Edition, John Wiley, New York, 2017, es-
pecially Chapters 8-14. In general the nomenclature of BS is used, and much of the notes
follow a similar structure as that text. In addition, Abbott and van Ness’s Thermodynamics,
McGraw-Hill, New York, 1972, has been used to guide some of the mathematical develop-
ments. Many example problems have been directly taken from BS and other texts; specific
citations are given where they arise.
These notes emphasize both problem-solving as well as some rigorous undergraduate-level
development of the underlying classical theory. It should also be remembered that practice is
essential to the learning process; the student would do well to apply the techniques presented
here by working as many problems as possible.
The notes, along with information on the course, can be found on the world wide web at
https://www3.nd.edu/∼powers/ame.50531. At this stage, anyone is free to make copies
for their own use. I would be happy to hear from you about suggestions for improvement.
Joseph M. Powers
powers@nd.edu
https://www3.nd.edu/∼powers
Notre Dame, Indiana; USA
CC BY: $ = 12 March 2024
\
The content of this book is licensed under Creative Commons Attribution-Noncommercial-No Derivative
Works 3.0.
7
8 CONTENTS
Review
Here we give a brief review of some standard concepts from a first course in thermodynam-
ics. In this review, we confine attention to a pure substance, defined as a material that
has homogeneous and invariable composition. An example would be pure helium. Later,
beginning in Ch. 3, we shall study mixtures, for example hydrogen-oxygen-nitrogen mixtures
that could react exothermically and convert chemical energy into thermal energy.
9
10 CHAPTER 1. REVIEW
• density:
1
ρ= , (1.1)
v
mass per unit volume, SI-based units are kg/m3 .
P v = RT, u − uo = cv (T − To ). (1.2)
RT
P (T, v) = . (1.5)
v
SI-based units are kPa.
– e.g. the sound speed for CPIG:
√
c(P, v) = kP v. (1.6)
the ratio of specific heats. A more common alternate notation for the ratio of
specific heats of a CPIG is cP /cv = γ.
• energy:
1
E
|{z} U + m(v · v) + mgz .
= |{z} (1.8)
|{z}
|2 {z } potential
total energy internal
kinetic
• first law:
dE = δQ − δW. (1.9)
If kinetic and potential energies are ignored,
dU = δQ − δW. (1.10)
dU = T dS − P dV. (1.11)
Sometimes this version of the Gibbs equation is called the “fundamental thermody-
namic equation.”
• second law:
δQ
. dS ≥ (1.12)
T
Here the heat transfer δQ can be reversible, in which case the equality holds, or irre-
versible, in which case the inequality holds.
• process: moving from one state to another, in general with accompanying heat trans-
fer and work.
• cycle: process which returns to initial state.
• specific reversible work:
Z 2
1 w2 = P dv, δw = P dv. (1.13)
1
Figure 1.1 gives an example of an isothermal thermodynamic process going from state 1
to state 2 in various thermodynamic planes. Figure 1.2 gives a sketch of a thermodynamic
cycle.
Example 1.1
Consider an ideal gas in the T − s plane. Compare the slope of an isochore to that of an isobar at
a given point.
Recall the Gibbs equation for a simple compressible substance, which is obtained by scaling the
extensive Eq. (1.11) by mass m to get the more common intensive form:
T ds = du + P dv. (1.15)
P T
v1
1 v2
P1
1
2
T1=T2 2
P2 T1=T2
v1 v2 s1 s2 s
v
P v
v1
P2
P1 v2
1
2
v2
1
P1
P2 2 v1
1
T1=T2 T1 = T2 T
T
Here we have used ds = dS/m, du = dU/m, and dv = dV /m. We have for the ideal gas
This holds for all ideal gases, be they calorically perfect or imperfect. Thus, the Gibbs equation can be
rewritten as
T ds = cv dT +P dv. (1.17)
| {z }
=du
∂T T
= . (1.19)
∂s v cv
h = u + P v. (1.20)
P T
v s
∂T T
= . (1.26)
∂s P cP
Example 1.2
Consider the following isobaric process for air, modeled as a CPIG, from state 1 to state 2. P1 =
100 kPa, T1 = 300 K, T2 = 400 K. Show the second law is satisfied.
Because the process is isobaric, P = 100 kPa describes a straight line in the P − v and P − T planes
and P2 = P1 = 100 kPa. Because we have an ideal gas, we have for the v − T plane:
R
v = T, straight lines! (1.28)
P
kJ
RT1 0.287 kg K (300 K) m3
v1 = = = 0.861 , (1.29)
P1 100 kPa kg
RT2 0.287 kgkJK (400 K) m3
v2 = = = 1.148 . (1.30)
P2 100 kPa kg
du = cv dT. (1.31)
Because our ideal gas is also calorically perfect, cv is a constant, and we get
Z u2 Z T2
du = cv dT , (1.32)
u1 T1
u2 − u1 = cv (T2 − T1 ), (1.33)
kJ
= 0.7175 (400 K − 300 K), (1.34)
kg K
kJ
= 71.750 . (1.35)
kg
Also we have
T ds = du + P dv, (1.36)
T ds = cv dT + P dv, (1.37)
RT R RT
from ideal gas : v = : dv = dT − 2 dP, (1.38)
P P P
RT
T ds = cv dT + R dT − dP, (1.39)
P
dT dP
ds = (cv + R) −R , (1.40)
T P
dT dP
= (cv + cP − cv ) −R , (1.41)
T P
dT dP
= cP −R , (1.42)
T P
Z s2 Z T2 Z P2
dT dP
ds = cP −R , (1.43)
s1 T1 T P1 P
T2 P2
s2 − s1 = cP ln − R ln . (1.44)
T1 P1
(1.45)
du = δq − δw, (1.53)
δq = du + δw, (1.54)
Z 2 Z 2 Z 2
δq = du + δw, (1.55)
1 1 1
1 q2 = (u2 − u1 ) + 1 w2 , (1.56)
kJ kJ
= 71.750 + 28.700 , (1.57)
kg kg
kJ
= 100.450 . (1.58)
kg
Now in this process the gas is heated from 300 K to 400 K. One would expect at a minimum that the
surroundings were at 400 K. Check for second law satisfaction.
1 q2
s2 − s1 ≥ ? (1.59)
Tsurr
kJ 100.450 kJ/kg
0.2890 ≥ ? (1.60)
kg K 400 K
kJ kJ
0.2890 ≥ 0.2511 , yes. (1.61)
kg K kg K
Let us review non-constant specific heat in some more detail. For calorically imperfect
ideal gases (CIIG), e.g. O2 at moderate to high temperatures (300 K < T < 6000 K):
• u = u(T ),
P T
T1 = 300 K v1
T2 = 400 K
1 2 P =P =100 kPa v2
1 2 T2
2
1
T1
v1 v2 s1 s2 s
v
P v
v1
v1 1
T1 T2 T1 T2 T
T
• cv = cv (T ),
• h = h(T ),
• cP = cP (T ).
For such temperatures, our assumption of constant cv is not as valid. But for ideal gases,
we can still take cv = cv (T ), so
du
= cv (T ). (1.62)
dT
We can integrate via separation of variables to get
du = cv (T ) dT, (1.63)
Z 2 Z 2
du = cv (T ) dT, (1.64)
1 1
Z 2
u2 − u1 = cv (T ) dT. (1.65)
1
We can interpret the difference in u as the area under the curve in a plot of cv (T ) versus T
as plotted in Fig. 1.4. More generally, we could say
T1 T2 T
Figure 1.4: Relation between u2 − u1 and area under curve in a plot of cv (T ) for calorically
imperfect ideal gas.
Z T
u(T ) = uo + cv (T̂ ) dT̂ , (1.66)
To
Now, cv , cP and R all have units of kJ/kg/K. Let us consider the ratio
cv
cv cv M M cv
= = R
= . (1.69)
R RM M R
The ratio is now in terms of molar specific properties with cv and R having units of
kJ/kmole/K. Note that R is the universal gas constant. A plot of cv /R versus T for a
variety of simple molecules is given in Fig. 1.5. We note some remarkable facts:
• For monatomic gases, such as Ar, O, and H, cv /R = 3/2 for a wide variety of temper-
atures.
• For diatomic gases, such as O2 and H2 for T < 600 K, cv /R ∼ 5/2, and for T > 600 K,
cv /R → 7/2.
6 CO2
H2O
4 O2
7/2
H2
5/2
2 O
3/2
Ar, H
0
0 2000 4000 6000 T (K)
What we are seeing actually reflects some fundamental physics. We first note that sta-
tistical thermodynamics proves
• For diatomic molecules, we summarize the behavior in the sketch given in Fig. 1.6.
7/2
5/2
Figure 1.6: Sketch of cv /R as a function of T for a model diatomic gas. Sketch adapted
from J. D. Anderson, 1989, Hypersonic and High Temperature Gas Dynamics, McGraw-Hill,
New York, p. 442. Note that a real gas would liquefy in the low temperature region of the
plot! So this model is really for a non-existent gas that has no liquid-inducing intermolecular
forces.
• For triatomic molecules such as H2 O or CO2 , there are more modes of motion that can
absorb energy, so the specific heat is higher still.
The dissociation and ionization behavior of the components of air is clearly displayed in
Fig. 1.7. Here we see at low temperatures, T < 1000 K, diatomic N2 and O2 are dominant
in air. Both of these major components begin to dissociate at higher temperatures. For
T > 6000 K, we no longer find diatomic N2 and O2 , but instead find their monatomic
components N and O. At higher temperatures still, the molecule loses electrons, and positive
ions remain.
Figure 1.7: Equilibrium composition of air at low density and various temperatures. Figure
from W. E. Moeckel and K. C. Weston, 1958, Composition and thermodynamic properties
of air in chemical equilibrium, NACA Technical Note 4265.
Feynman1 summarizes the argument that this preference for one type of energy over
another (translation, rotational, vibrational) depending on temperature is surprising to those
not versed in quantum mechanics and violates standard assumptions of classical statistical
mechanics. In fact, he notes that Maxwell had a hint of the problem as early as 1859, and
stated this concern more directly in 1869. Maxwell summarized those concerns in an 1875
lecture, transcribed in Nature.2 Feynman argues that the reason for the energy partition
observed in diatomic gases is a “failure of classical physics” and instead is a pure effect
of quantum mechanics; that is to say k = cP (T )/cv (T ) = k(T ) is a non-classical result!
Employment of the theories of quantum and statistical mechanics allows an accounting for
the observation that there is a preference of molecules to exist in lower energy states, and at
those states, the discrete quantization is important. High energy vibrational states are less
likely than translational states at low temperature. At higher temperature, there is a higher
1
R. P. Feynman, R. B. Leighton, and M. Sands, 1963, The Feynman Lectures on Physics, Volume 1,
Addison-Wesley, Reading, Massachusetts, pp. 40-7–40-10.
2
J. C. Maxwell, 1875, “On the dynamical evidence of the molecular constitution of bodies,” Nature,
11(280): 374-377.
probability that all states are populated, and one recovers results consistent with classical
physics.
Let us also recall that cP (T ) − cv (T ) = R; thus, cP (T ) − cv (T ) = R. Let us summarize
• for monatomic gases,
3
cv = R, (1.70)
2
5
cP = cv + R = R, (1.71)
2
5
cP R 5
= k = 23 = = 1.6667. (1.72)
cv 2
R 3
To summarize, usually the most problematic case is whether or not specific heats vary
with temperature in ideal gases. For low temperatures, the specific heat is well modeled as a
constant; here the internal energy change is strictly proportional to the temperature change.
For moderate to high temperatures, a temperature-variation of the specific heat is observed.
Changes in internal energy are no longer strictly proportional to changes in temperature.
The behavior is analogous to solid mechanics. At low strain ǫ, stress σ is proportional to
strain, and the constant of proportionality is the modulus of elasticity E. For high strains,
the linearity is lost; we could say the elastic modulus becomes a function of strain. We give
a sketch in Fig. 1.8 of the comparison to solid mechanics
There are four main ways to calculate changes in enthalpy for ideal gases:
• assumption of constant cP evaluated at 298 K,
• assumption of constant cP evaluated at an intermediate temperature,
R2
• using a known analytic form of cP (T ) in the direct integration of 1 cP (T ) dT , or
• estimation using the ideal gas tables.
Example 1.3
Calculate the heat transferred per unit mass to N2 in an isobaric process that starts at T1 = 300 K
and finishes at T2 = 1000 K. Use the four different means of calculating enthalpy changes to estimate
the heat transfer.
u
calorically
imperfect
region
E higher cv
lower
plastic
region
elastic calorically
region perfect
E region cv
1 1
Figure 1.8: Sketch of comparison of stress-strain behavior in solids with ideal gas internal
energy-temperature behavior.
u2 − u1 = 1 q2 − 1 w2 . (1.76)
R2
For an isobaric process, 1 w2 = 1
P dv = P (v2 − v1 ) = P2 v2 − P1 v1 . So
Z 2
1 q2 = u2 − u1 + P2 v2 − P1 v1 = h2 − h1 = ∆h = cP (T ) dT. (1.77)
1
• Constant cP at 298 K. From Table A.5 in BS, we find cP = 1.042 kJ/kg/K. Thus, we estimate
kJ kJ
∆h = h2 − h1 = cP (T2 − T1 ) = 1.042 ((1000 K) − (300 K)) = 729.4 . (1.78)
kg K kg
T1 + T2 300 K + 1000 K
Tave = = = 650 K. (1.79)
2 2
Table A.6 in BS has polynomial curve fits for cP . For N2 , we find from Table A.6 that
T
cP = C0 + C1 θ + C2 θ2 + C3 θ3 , θ≡ . (1.80)
1000 K
The numbers Co , . . . , C3 actually have units and are designed to yield a value for cP in kJ/kg/K. We
get
2
kJ kJ T kJ T
cP = 1.11 − 0.48 + 0.96
kg K kg K 1000 K kg K 1000 K
3
kJ T
− 0.42 , (1.81)
kg K 1000 K
2
kJ kJ 650 K kJ 650 K
= 1.11 − 0.48 + 0.96
kg K kg K 1000 K kg K 1000 K
3
kJ 650 K
− 0.42 , (1.82)
kg K 1000 K
kJ
= 1.08826 . (1.83)
kg K
Let us change variables from T to θ. We have θ = T /(1000 K), so dθ = dT /(1000 K), and dT =
(1000 K) dθ. Thus,
Z θ2
∆h = C0 + C1 θ + C2 θ2 + C3 θ3 ((1000 K) dθ), (1.86)
θ1
θ
θ2 θ3 θ4 2
= (1000 K) C0 θ + C1 + C2 + C3 , (1.87)
2 3 4 θ1
2
kJ kJ θ
= (1000 K) 1.11 θ − 0.48
kg K kg K 2
3 4 1000
1000 K
kJ θ kJ θ K
+ 0.96 − 0.42 . (1.88)
kg K 3 kg K 4 300 K
1000 K
• Use of ideal gas tables. Lastly, we can use the ideal gas tables. For us, Table A.8 of BS is best. We
find h(300 K) = 311.67 kJ/kg and h(1000 K) = 1075.91 kJ/kg. So
kJ kJ kJ
∆h = h2 − h1 = 1075.91 − 311.67 = 764.24 . (1.90)
kg kg kg
Example 1.4
A calorically imperfect ideal gas is known to have
cP (T ) = cP o + aT. (1.91)
Because internal energy changes are path-independent, we need not worry about any details of
the process, including its path. That is because we know the end states. So we can use a version of
Eq. (1.65) to get
Z T3
u3 − u1 = cv (T ) dT. (1.92)
T1
We are given cP , but need cv . We can use Mayer’s relation, Eq. (??) to get
So
Z T3
u3 − u1 = ((cP o − R) + aT ) dT. (1.94)
T1 | {z }
cv (T )
Integrating, we find
a
u3 − u1 = (cP o − R)(T3 − T1 ) + (T32 − T12 ). (1.95)
2
Had we been asked to find the heat transfer, 1Q3 , we would have had to calculate the work 1W3 , and
the details of the path would have been required. Because internal energy is a state function, only the
end states are required to evaluate the change in internal energy.
Cycle analysis
2.1 Carnot
The Carnot cycle is the most well-known thermodynamic cycle. It is a useful idealization,
but is difficult to realize in practice. Its real value lies in serving as a standard to which
other cycles can be compared. These are usually fully covered in introductory courses. The
cycle can be considered as follows
• 1 → 2: isentropic compression,
• 2 → 3: isothermal expansion,
• 3 → 4: isentropic expansion, and
• 4 → 1: isothermal compression.
This forms a rectangle in the T − s plane The P − v plane is more complicated. Both are
shown in Figure 2.1.
25
26 CHAPTER 2. CYCLE ANALYSIS
P T
2
isot
isotherm 3
her 2
m
isentrope
isentrope
isent
ro p
3
e
iso
th e
ise
1 rm ro 1 isotherm 4
nt
p e
v s
The slope of the isentrope is found in the following way. Consider first the Gibbs equation,
Eq. (1.15):
T ds = du + P dv. (2.4)
du = cv dT, (2.5)
T ds = cv dT + P dv. (2.6)
P v = RT, (2.7)
P dv + v dP = R dT, (2.8)
P dv + v dP
= dT. (2.9)
R
So then
P dv + v dP
T ds = cv +P dv, (2.10)
R
| {z }
=dT
c cv
v
T ds = + 1 P dv + v dP, Take ds = 0, (2.11)
R R
c cv ∂P
v
0 = +1 P + v , (2.12)
R R ∂v s
cv
∂P +1 P
= − R cv , (2.13)
∂v s R
v
cv
c −c
+1 P
= − P cvv , (2.14)
cP −cv
v
cv + cP − cv P
= − , (2.15)
cv v
cP P
= − , (2.16)
cv v
P
= −k . (2.17)
v
Because k > 1, the magnitude of the slope of the isentrope is greater than the magnitude of
the slope of the isotherm:
∂P ∂P
> . (2.18)
∂v s ∂v T
Example 2.1
(Adapted from BS) Consider an ideal gas Carnot cycle with air in a piston cylinder with a high
temperature of 1200 K and a heat rejection at 400 K. During the heat addition, the volume triples.
The gas is at 1.00 m3 /kg before the isentropic compression. Analyze.
m3
T1 = 400 K, v1 = 1.00 . (2.19)
kg
Now isentropically compress to state 2. By the standard relations for a CPIG, one finds
k−1 k−1
T2 v1 P2 k
= = . (2.21)
T1 v2 P1
So k−1
1 1
T1 m3 400 K 1.4−1 m3
v2 = v1 = 1.00 = 0.06415 . (2.22)
T2 kg 1200 K kg
Note that v2 < v1 as is typical in a compression. The isentropic relation between pressure and volume
can be rearranged to give the standard
P2 v2k = P1 v1k . (2.23)
Thus, one finds that
k m3
!1.4
v1 1.00
kg
P2 = P1 = (1.148 × 102 kPa) 3 = 5.36866 × 103 kPa. (2.24)
v2 0.06415 mkg
The pressure has increased in the isentropic compression. Check to see if the ideal gas law is satisfied
at state 2:
kJ
RT2 0.287 kg K (1200 K)
P2 = = 3 = 5.36866 × 103 kPa. (2.25)
v2 0.06415 mkg
This matches. Now the expansion from state 2 to 3 is isothermal. This is the heat addition step in
which the volume triples. So one gets
m3 m3
v3 = 3v2 = 3 0.06415 = 0.19245 , (2.26)
kg kg
T3 = T2 = 1200 K. (2.27)
The ideal gas law then gives
kJ
RT3 0.287 kg K (1200 K)
P3 = = m3
= 1.78955 × 103 kPa. (2.28)
v3 0.19245 kg
Process 3 to 4 is an isentropic expansion back to 400 K. Using the isentropic relations for the CPIG,
one gets
k−1 1 1
T3 m3 1200 K 1.4−1 m3
v4 = v3 = 0.19245 = 3.000 , (2.29)
T4 kg 400 K kg
k 3 !1.4
v3 3
0.19245 mkg
P4 = P3 = (1.78955 × 10 kPa) m3
= 3.82667 × 101 kPa. (2.30)
v4 3.000 kg
Check:
kJ
RT4 0.287 kg K (400 K)
P4 = = m3
= 3.82667 × 101 kPa. (2.31)
v4 3.000 kg
A summary of the states is given in Table 2.1.
Now calculate the work, heat transfer and efficiency. Take the adiabatic exponent for air to be
k = 1.4. Now because of Eq. (1.7) k = cP /cv , and Meyer’s relation, Eq. (1.27), cP − cv = R, one gets
R + cv
k = , (2.32)
cv
kcv = R + cv , (2.33)
cv (k − 1) = R, (2.34)
kJ
R 0.287 kg K kJ
cv = = = 0.7175 . (2.35)
k−1 1.4 − 1 kg K
3
T (K) P (kPa) v ( mkg )
1 400 1.148 × 102 1.000
2 1200 5.36886 × 103 0.06415
3 1200 1.78955 × 103 0.19245
4 400 3.82667 × 102 3.000
Recall the first law, u2 − u1 = 1 q2 − 1 w2 . Recall also the caloric equation of state for a CPIG: u2 − u1 =
cv (T2 − T1 ). Now process 1 → 2 is isentropic, so it is also adiabatic, hence 1 q2 = 0, so one has
u2 − u1 = 1 q2 −1 w2 , (2.36)
|{z}
=0
cv (T2 − T1 ) = −1 w2 , (2.37)
kJ
0.7175 (1200 K − 400 K) = −1 w2 , (2.38)
kg K
kJ
1 w2 = −5.7400 × 102 . (2.39)
kg
The work is negative as work is being done on the system in the compression process.
Process 2 → 3 is isothermal, so there is no internal energy change. The first law gives
u3 − u2 = 2 q3 − 2 w3 , (2.40)
cv (T3 − T2 ) = 2 q3 − 2 w3 , (2.41)
| {z }
=0
0 = 2 q3− 2 w3 , (2.42)
Z v3
2 q3 = w
2 3 = P dv, (2.43)
v2
Z v3
RT
= dv, (2.44)
v2 v
Z v3
dv
= RT2 , (2.45)
v2 v
v3
= RT2 ln , (2.46)
v2
m3
kJ 0.19245 kg
= 0.287 (1200 K) ln m3
, (2.47)
kg K 0.06415 kg
kJ
= 3.78362 × 102 . (2.48)
kg
kJ kJ kJ
Process ∆u kg
q kg
w kg
1→2 5.74000 × 102 0 −5.74000 × 102
2→3 0 3.78362 × 102 3.78362 × 102
3 → 4 −5.74000 × 102 0 5.74000 × 102
4→1 0 −1.26121 × 102 −1.26121 × 102
Total 0 2.52241 × 102 2.52241 × 102
Note that
wcycle = qcycle . (2.65)
Check now for the cycle efficiency. Recall that the thermal efficiency is
Tlow
η =1− . (2.68)
Thigh
400 K 2
η =1− = ∼ 0.6667. (2.69)
1200 K 3
Recall further that for a Carnot cycle, one has
qlow Tlow
= . (2.70)
qhigh Thigh
−4 q1 Tlow
= , (2.71)
2 q3 Thigh
kJ
1.26121 × 102 kg 400 K
kJ
= , (2.72)
3.78362 × 102 kg 1200 K
1 1
= . (2.73)
3 3
Indeed, the appropriate relation holds.
2.1.2 Exergy
Here we will introduce the concept of exergy, which relies on a Carnot cycle for its motivation.
This concept is widely used in some industrial design applications; its use in the fundamental
physics literature is not extensive, likely because it is not a thermodynamic property of a
material, but also includes mechanical properties. This system quantity is one measure of
how much useful work can be extracted from a system which is brought into equilibrium
with a so-called reference rest state.
control
volume
Carnot
engine
First let us imagine that the surroundings are at a reference temperature of To and a
reference pressure of Po . This yields a reference enthalpy of ho and a reference entropy of so .
We also take the surroundings to be at rest with a velocity of v = 0, and a reference height
of zo .
Consider the sketch of Fig. 2.2. Let us consider a steady flow into and out of a control
volume. The conservation of mass equation yields
dmcv
= ṁin − ṁout . (2.74)
dt
For steady flow conditions, we have d/dt = 0, so we recover
We will soon relate Q̇cv /ṁ to qH,Carnot as sketched in Fig. 2.2; this will introduce a sign
convention problem. For now, we will maintain the formal sign convention associated with
Q̇cv connoting heat transfer into the control volume. For steady flow conditions, we get
1
0 = Q̇cv − Ẇcv + ṁ hin − hout + (vin · vin − vout · vout ) + g(zin − zout ) . (2.77)
2
Now take the “in” state to be simply a generic state with no subscript, and the “out”
state to be the reference state, so
1
0 = Q̇cv − Ẇcv + ṁ h − ho + (v · v) + g(z − zo ) . (2.78)
2
Let us next scale by ṁ so as to get
1
0 = qcv − wcv + h − ho + v · v + g(z − zo ). (2.79)
2
Here we have taken qcv ≡ Q̇cv /ṁ and wcv = Ẇcv /ṁ.
Now, let us insist that wcv = 0, and solve for qcv to get
1
qcv = − h − ho + v · v + g(z − zo ) . (2.80)
2
Now, we imagine the working fluid to be at an elevated enthalpy, velocity, and height
relative to its rest state. Thus in the process of bringing it to its rest state, we will induce
qcv < 0. By our standard sign convention, this means that thermal energy is leaving the
system. This energy which leaves the system can be harnessed, in the best of all possible
worlds, by a Carnot engine in contact with a thermal reservoir at temperature To to generate
useful work. It is this work which represents the available energy, also known as the exergy.
For the Carnot engine, we have
The standard sign convention for work and heat transfer is abandoned for the Carnot
analysis! It is wCarnot which gives the availability or exergy, which we define as ψ:
Now let us require that the heat input to the Carnot engine be the heat associated with the
heat transfer from the control volume. Because of the inconsistency in sign conventions, this
requires that
qH,Carnot = T (s − so ), (2.84)
qL,Carnot = To (s − so ). (2.85)
So, by substituting Eqs. (2.83) and (2.85) into Eq. (2.82), the availability ψ is
ψ = −qcv − To (s − so ). (2.86)
Now use Eq. (2.80) to eliminate qcv from Eq. (2.86) to get
1
ψ = h − ho + v · v + g(z − zo ) − To (s − so ), (2.87)
2
1
ψ = h − To s + v · v + gz − (ho − To so + gzo) . (2.88)
2
• high velocity v,
Example 2.2
Find the exergy for a CPIG.
h − ho = cP (T − To ), (2.89)
T P
s − so = cP ln − R ln . (2.90)
To Po
So we get
T P 1
ψ = cP (T − To ) − To cP ln − R ln + v · v + g(z − zo ). (2.91)
To Po 2
For T ∼ To P ∼ Po , we can use Taylor series to simplify this somewhat. First, we recall the general
Taylor series for a log function near unity. Consider
y(x) = ln x, (2.92)
dy 1 d2 y
y(x) ∼ y(1) + (x − 1) + (x − 1)2 + . . . . (2.93)
dx x=1 2 dx2 x=1
dy 1 d2 y 1
= , 2
= − 2, (2.94)
dx x dx x
and thus
dy d2 y
y(1) = ln(1) = 0, = 1, = −1. (2.95)
dx x=1 dx2 x=1
So
1
y(x) = ln x ∼ 0 + (x − 1) − (x − 1)2 + . . . (2.96)
2
We then expand ψ via the following steps:
T P 1
ψ = cP (T − To ) − cP To ln + RTo ln + v · v + g(z − zo ), (2.97)
To Po 2
T T k−1 P 1
= cP T o − 1 − ln + ln + v · v + g(z − zo ), (2.98)
To To k Po 2
2 ! !
T T 1 T k−1 P
= cP T o −1 − −1 − − 1 + ... + − 1 + ...
To To 2 To k Po
1
+ v · v + g(z − zo ), (2.99)
2 !
2 !
1 T k−1 P 1
= cP T o − 1 + ... + − 1+ ... + v · v + g(z − zo ). (2.100)
2 To k Po 2
In the neighborhood of the ambient state, relative pressure differences are more effective than relative
temperature differences at inducing high exergy.
2.2 Rankine
2.2.1 Classical
The Rankine cycle forms the foundation for the bulk of power generating devices which
utilize steam as a working fluid. The ideal cycle is described by
T
3
3
turbine
boiler
4 2
2
1 4
pump condenser
1
s
Figure 2.3: Schematic for Rankine cycle and the associated T − s plane.
A schematic for the Rankine cycle and the associated path in the T − s plane is shown
in Figure 2.3.
Example 2.3
(adopted from Moran and Shapiro) Consider steam in an ideal Rankine cycle. Saturated vapor
enters the turbine at 8.0 MPa. Saturated liquid exits the condenser at P = 0.008 MPa. The net power
output of the cycle is 100 MW. Find
• thermal efficiency
• back work ratio
Use the steam tables to fix the state. At the turbine inlet, one has P3 = 8.0 MPa, and x3 = 1
(saturated steam). This gives two properties to fix the state, so that
kJ kJ
h3 = 2758 , s3 = 5.7432 . (2.101)
kg kg K
State 4 has P4 = 0.008 MPa and s4 = s3 = 5.7432 kJ/kg/K, so the state is fixed. From the saturation
tables, it is found then that
kJ kJ
s4 − sf 5.7432 kg K − 0.5926 kg K
x4 = = = 0.6745. (2.102)
sg − sf 7.6361 kgkJK
State 1 is saturated liquid at 0.008 MPa, so x1 = 0, P1 = 0.008 MPa. One then gets h1 = hf =
173.88 kJ/kg, v1 = vf = 0.0010084 m3 /kg.
Now state 2 is fixed by the boiler pressure and s2 = s1 . But this requires use of the sparse
compressed liquid tables. Alternatively, the pump work is easily approximated by assuming an incom-
pressible fluid so that
Ẇ
h2 = h1 + = h1 + v1 (P2 − P1 ), (2.104)
ṁ
kJ m3 kJ
= 173.88 + 0.0010084 (8000 kPa − 8 kPa) = 181.94 . (2.105)
kg kg kg
Now the first law for the turbine and pump give
Ẇt Ẇp
= h3 − h4 , = h1 − h2 . (2.107)
ṁ ṁ
= 0.371. (2.111)
By definition the back work ratio bwr is the ratio of pump work to turbine work:
Ẇp
bwr = , (2.112)
Ẇt
h1 − h2
= , (2.113)
h3 − h4
kJ
173.88 kg − 181.94 kJ
kg
= , (2.114)
kJ kJ
2758 kg − 1794.8 kg
= 0.00837. (2.115)
The desired mass flow can be determined because we know the desired net power. Thus
Ẇcycle
ṁ = , (2.116)
(h3 − h4 ) + (h1 − h2 )
100 × 103 kW
= , (2.117)
kJ
2758 kg − 1794.8 kJ
kg + 173.88
kJ
kg − 181.94 kJ
kg
kg
= 104.697 , (2.118)
s
kg s kg
= 104.697 3600 = 3.769 × 105 . (2.119)
s hr hr
Ẇcycle = Q̇in + Q̇out = (269.7 MW) − (169.7 MW) = 100 MW. (2.128)
For the condenser mass flow rate now perform a mass balance:
dEcv
= Q̇cv − Ẇcv +ṁc (hin − hout ) + ṁ(h4 − h1 ), (2.129)
|{z} |{z}
| dt
{z } =0 =0
=0
0 = ṁc (hin − hout ) + ṁ(h4 − h1 ), (2.130)
ṁ(h4 − h1 )
ṁc = − , (2.131)
h − hout
in
104.697 kg
s 1794.8 kJ
kg − 173.88
kJ
kg
= − , (2.132)
62.99 kJkg − 146.68 kJ
kg
kg
= 2027.79 , (2.133)
s
kg 3600 s
= 2027.79 , (2.134)
s hr
kg
= 7.3 × 106 . (2.135)
hr
The enthalpy for the cooling water was found by assuming values at the saturated state at the respective
temperatures of 15 ◦ C and 35 ◦ C.
Example 2.4
Compute the exergy at various points in the flow of a Rankine cycle as considered in the previous
example problem.
So the exergy has gone up. Here ψ2 − ψ1 = h2 − h1 because the process is isentropic. Here ψ2 − ψ1 =
h2 = h1 = 181.94 − 173.88 = 8.06 kJ/kg.
After the boiler, just before the turbine, we have
Relative to the pump, the boiler has added much more exergy to the fluid. After the turbine, just
before the condenser, we have
Note that ψ4 − ψ3 = h4 − h3 because the process is isentropic. The actual exergy (or available work)
at the exit of the turbine is relative low, even though the enthalpy state at the turbine exit remains at
an elevated value.
2.2.2 Reheat
In a Rankine cycle with reheat, the steam is extracted from an intermediate stage of the
turbine and reheated in the boiler. It is then expanded through the turbine again to the
condenser pressure. One also avoids liquid in the turbine with this strategy. This generally
results in a gain in cycle efficiency. Geometrically, the behavior on a T − s diagram looks
more like a Carnot cycle. This is often covered in an introductory thermodynamics class, so
no formal example will be given here. A schematic for the Rankine cycle with reheat and
the associated T − s diagram is shown in Figure 2.4.
2.2.3 Regeneration
In a Rankine cycle with regeneration, some steam is extracted from the turbine and used
to pre-heat the liquid which is exiting the pump. This can lead to an increased thermal
efficiency, all else being equal. The analysis is complicated by the need to take care of more
T
3
3 5
turbine
boiler
4
5
2 4
2 6
1
6
pump condenser
1
s
Figure 2.4: Schematic for Rankine cycle with reheat along with the relevant T − s diagram.
complex mass and energy balances in some components. A schematic for the Rankine cycle
with regeneration and open feedwater heating is shown in Figure 2.5.
Example 2.5
(adopted from BS) Steam leaves a boiler and enters a turbine at 4 MPa, 400 ◦ C. After expansion
to 400 kPa, some steam is extracted for heating feedwater in an open feedwater heater. Pressure in
feedwater heater is 400 kPa, and water leaves it at a saturated state at 400 kPa. The rest of the steam
expands through the turbine to 10 kPa. Find the cycle efficiency.
turbine
boiler
7
6
4
feedwater
heater condenser
2
pump
P2
3
pump
1
P1
Figure 2.5: Schematic for Rankine cycle with regeneration and open feedwater heating.
wP 1 = v1 (P1 − P2 ), (2.157)
m3
= 0.00101 ((10 kPa) − (400 kPa)) , (2.158)
kg
kJ
= −0.3939 . (2.159)
kg
A sign convention consistent with work done by the fluid is used here. At this point the text abandons
this sign convention instead.
Now consider the turbine
dmcv
= ṁ5 − ṁ6 − ṁ7 , (2.160)
dt }
| {z
=0
ṁ5 = ṁ6 + ṁ7 , (2.161)
ṁ6 ṁ7
1 = + , (2.162)
ṁ5 ṁ5
dEcv
= Q̇cv −Ẇcv + ṁ5 h5 − ṁ6 h6 − ṁ7 h7 , (2.163)
| dt
{z } |{z}
=0
=0
Now consider the feedwater heater. The first law for this device gives
dEcv
= Q̇cv − Ẇcv +ṁ2 h2 + ṁ6 h6 − ṁ3 h3 , (2.170)
| dt
{z } |{z} |{z}
=0 =0
=0
ṁ2 ṁ6
h3 h2 +
= h6 , (2.171)
ṁ3 ṁ3
ṁ7 ṁ6
= h2 + h6 , (2.172)
ṁ5 ṁ5
ṁ6 ṁ6
= 1− h2 + h6 , (2.173)
ṁ5 ṁ5
kJ ṁ6 kJ ṁ6 kJ
604.73 = 1− 192.194 + 2685.6 , (2.174)
kg ṁ5 kg ṁ5 kg
ṁ6
= 0.165451. (2.175)
ṁ5
wP 2 = v3 (P3 − P4 ), (2.179)
m3
= 0.001084 ((400 kPa) − (4000 kPa)) , (2.180)
kg
kJ
= −3.9024 . (2.181)
kg
Now
h4 = h3 + v3 (P4 − P3 ), (2.182)
kJ m3
= 604.73 + 0.001084 ((4000 kPa) − (400 kPa)) , (2.183)
kg kg
kJ
= 608.6 . (2.184)
kg
Now get the net work
qh = h5 − h4 , (2.191)
kJ kJ
= 3213.6 − 608.6 , (2.192)
kg kg
kJ
= 2605.0 . (2.193)
kg
Thus, the thermal efficiency is
wnet 975.677 kJ
kg
η= = kJ
= 0.375. (2.194)
qh 2605.0 kg
This does represent an increase of efficiency over a comparable Rankine cycle without regeneration,
which happens to be 0.369.
2.2.4 Losses
• Turbine: These are typically the largest losses in the system. The turbine efficiency is
defined by
wt h3 − h4
η= = . (2.195)
wts h3 − h4s
Here h4s and wts are the enthalpy and work the working fluid would have achieved had
the process been isentropic. This is for a control volume.
• Pump: Pump losses are usually much smaller in magnitude than those for turbines.
The pump efficiency for a control volume is defined by
wps h2s − h1
η= = . (2.196)
wp h2 − h1
• Piping: pressure drops via viscous and turbulent flow effects induce entropy gains
in fluid flowing through pipes. There can also be heat transfer from pipes to the
surroundings and vice versa.
Losses will always degrade the overall thermal efficiency of the cycle.
2.2.5 Cogeneration
Often steam is extracted after the boiler for alternative uses. A good example is the Notre
Dame power plant, where steam at high pressure and temperature is siphoned from the
turbines to heat the campus in winter. The analysis for such a system is similar to that for
a system with regeneration.
A schematic for cogeneration cycle is shown in Figure 2.6.
• CPIG,
In some cycles, it is common to model the working fluid as a fixed mass. In others, it is
common to model the system as a control volume.
6
7
thermal
process
4 steam load
(e.g. heating the condenser
ND campus)
3 2
pump pump
mixer
P2 P1
2.4 Brayton
2.4.1 Classical
The Brayton cycle is the air standard model for gas turbine engines. It is most commonly
modeled on a control volume basis. It has the following components:
• 1 → 2: isentropic compression,
A schematic for the Brayton cycle is shown in Figure 2.7. Diagrams for P − v and T − s for
the Brayton cycle are shown in Figure 2.8.
The efficiency of the air standard Brayton cycle is found as
wnet
η= . (2.197)
qH
fuel
combustion
chamber 3
2
compressor turbine w
1 air products 4
P T
2 3 3
isentrope
isobar
2
isentrope 4
4 1 isobar
1
v s
dEcv
= Q̇cv − Ẇcv +ṁhin − ṁhout . (2.199)
dt } |{z}
| {z =0
=0
So
One might wonder about the work going from 2 to 3, which has area under the curve in the
P − v diagram of Fig. 2.8. In the combustor there is no shaft work, so Ẇcv = 0. But the
work of the expanding fluid is recognized by the enthalpy difference. Our Eq. (2.203) for
the combustor predicts q = h3 − h2 = cP (T3 − T2 ). Using the definition of h, it also says
q = u3 + P3 v3 − (u2 + P2 v2 ). And because P3 = P2 , we get q = u3 − u2 + P2 (v3 − v2 ). So for
the isobaric combustor, we could cast the first law as ∆u = q − P ∆v.
So the thermal efficiency for the Brayton cycle is
T4
wnet qH − qL qL cP (T4 − T1 ) T 1 T1
− 1
η= = =1− =1− =1− . (2.204)
qH qH qH cP (T3 − T2 ) T T3 − 1
2 T2
T1 1 1
η =1− = 1 − T2 = 1 − (k−1)/k . (2.210)
T2 T1 P2
P1
If a Carnot cycle were operating between the same temperature bounds, its efficiency,
ηCarnot = 1 − T1 /T3 , would be greater than that for the Brayton cycle. Relative to the
Rankine cycle, the Brayton cycle has a large fraction of compressor work. So the backwork
ratio bwr is larger.
One must also account for deviations from ideality. These effects are summarized in
component efficiencies. The relevant efficiencies here, assuming a control volume approach,
are
h2s − h1
ηc = , (2.211)
h2 − h1
h3 − h4
ηt = . (2.212)
h3 − h4s
For a CPIG, one gets then
T2s − T1
ηc = , (2.213)
T2 − T1
T3 − T4
ηt = . (2.214)
T3 − T4s
Example 2.6
(adopted from Moran and Shapiro). Air enters the compressor of an air-standard Brayton cycle
at 100 kPa, 300 K with a volumetric flow rate is 5 m3 /s. The pressure ratio in the compressor is 10.
The turbine inlet temperature is 1400 K. Find the thermal efficiency, the back work ratio and the net
power. Both the compressor and turbine have efficiencies of 0.8.
First calculate the state after the compressor if the compressor were isentropic.
k/(k−1)
P2 T2s
= = 10, (2.215)
P1 T1
T2s = T1 (10)(k−1)/k , (2.216)
(1.4−1)/1.4
= (300 K)(10) , (2.217)
= 579.209 K. (2.218)
Now
T2s − T1
ηc = , (2.219)
T2 − T1
T2s − T1
T2 = T1 + , (2.220)
ηc
(579.209 K) − (300 K)
= (300 K) + , (2.221)
0.8
= 649.012 K. (2.222)
Now state three has T3 = 1400 K. Recall that 2 → 3 involves heat addition in a combustion chamber.
Now calculate T4s for the ideal turbine. Recall that the expansion is to the same pressure as the
wnet qH − qL qL cP (T4 − T1 ) T4 − T1
η = = =1− =1− =1− , (2.231)
qH qH qH cP (T3 − T2 ) T3 − T2
860.101 K − 300 K
= 1− , (2.232)
1400 K − 649.012 K
= 0.254181. (2.233)
If the process were ideal, the back work ratio would have been
wcomp
bwrideal = , (2.243)
wturb
cP (T2s − T1 )
= , (2.244)
cP (T3 − T4s )
T2s − T1
= , (2.245)
T3 − T4s
(579.209 K) − (300 K)
= , (2.246)
(1400 K) − (725.126 K)
= 0.413721. (2.247)
Now get the net work. First get the mass flow rate from the volume flow rate. This requires the specific
volume.
kJ
RT1 0.287 kg K (300 K) m3
v1 = = = 0.861 . (2.248)
P1 100 kPa kg
Now 3
V̇ 5 ms kg
ṁ = ρ1 V̇ = = m 3 = 5.8072 . (2.249)
v1 0.861 kg s
The inefficiencies had a significant effect on both the back work ratio and the net work of the engine.
2.4.2 Regeneration
One can improve cycle efficiency by regeneration. The hot gas in the turbine is used to
preheat the gas exiting the compressor before it enters the combustion chamber. In this
cycle, one takes
• 1 → 2: compression,
• x → 3: combustion chamber,
y regenerator
x 4
combustion
2 chamber 3
compressor turbine w
For this to work, the gas at the turbine exit must have a higher temperature than the gas
at the compressor exit. If the compression ratio is high, the compressor exit temperature is
high, and there is little benefit to regeneration.
Now consider the regenerator, which is really a heat exchanger. The first law holds that
dEcv
= Q̇cv − Ẇcv +ṁh2 − ṁhx + ṁh4 − ṁhy , (2.256)
dt }
| {z |{z} |{z}
=0 =0
=0
0 = h2 − hx + h4 − hy , (2.257)
= cP (T2 − Tx ) + cP (T4 − Ty ), (2.258)
= T2 − Tx + T4 − Ty . (2.259)
Now consider the inlet temperatures to be known. Then the first law constrains the outlet
temperatures such that
T2 + T4 = Tx + Ty . (2.260)
If the heat exchanger were an ideal co-flow heat exchanger, one might expect the outlet
temperatures to be the same, that is, Tx = Ty , and one would have Tx = Ty = (1/2)(T2 + T4 ).
But in fact one can do better. If a counter-flow heat exchanger is used, one could ideally
expect to find
Tx = T4 , Ty = T2 . (2.261)
This is the Carnot efficiency moderated by the pressure ratio. As the pressure ratio rises,
the thermal efficiency declines.
The regenerator itself will not be perfect. To achieve the performance postulated here
would require infinite time or an infinite area heat exchanger. As both increase, viscous
losses increase, and so there is a trade-off. One can summarize the behavior of an actual
regenerator by an efficiency defined as
hx − h2 actual
ηreg = = . (2.273)
hx − h2
′ ideal
Example 2.7
(adapted from Moran and Shapiro) How would the addition of a regenerator affect the thermal
efficiency of the isentropic version of the previous example problem?
One may take the rash step of trusting the analysis to give a prediction from Eq. (2.272) of
(k−1)/k
T1 P2
η = 1− , (2.275)
T3 P1
300 K
= 1− (10)(1.4−1)/1.4 = 0.586279. (2.276)
1400 K
Without regeneration, the thermal efficiency of the ideal Brayton cycle had a value of 0.482053. Had
the engine used a Carnot cycle between the same temperature limits, the efficiency would have been
0.785714.
P T
2 3 3 4
isotherm
isobar
isobar
isotherm
1 4 2 1
v s
• 1 → 2: isothermal compression,
Note that P2 /P1 > 1, so wc < 0; work is done on the fluid in the compressor.
For the turbine, again for the control volume, the key is the enthalpy difference and
dh = T ds + v dP = δqt − δwt , with δqt = T ds and δwt = −v dP . The turbine work would
be
Z 4 Z 4
dP P4
wt = − v dP = −RT3 = −RT3 ln . (2.278)
3 3 P P3
But because P2 = P3 and P1 = P4 , the turbine work is also
P1 P2
wt = −RT3 ln = RT3 ln . (2.279)
P2 P1
qC = h3 − h2 = cP (T3 − T2 ). (2.280)
P2
qt = wt = RT3 ln . (2.281)
P1
P2
qH = qC + qt = cP (T3 − T1 ) + RT3 ln . (2.282)
P1
T1 1
= 1− T , (2.287)
T3 1− T1
1+ 3
(k−1)/k
P
ln P2
1
T1
T1 1 − T3
∼ 1− 1 − (k−1)/k + .... (2.288)
T3
| {z } ln PP12
Carnot | {z }
correction
This is expressed as a Carnot efficiency modified by a correction which degrades the efficiency.
The efficiency is less than that of a Carnot cycle. For high temperature ratios and high
pressure ratios, the efficiency approaches that of a Carnot engine.
Now if one used multiple staged intercooling on the compressors and multiple staged
expansions with reheat on the turbines, one can come closer to the isothermal limit, and
better approximate the Carnot cycle.
burner 3
2
1 a 4
diffuser compressor turbine nozzle
isentrope 5
a
a
1 isobar
1 5
v s
The goal of the jet propulsion cycle is to produce a thrust force which is necessary to
balance a fluid-induced aerodynamic drag force. When such a balance exists, the system is
in steady state. One can analyze such a system in the reference frame in which the engine
Solving for the thrust force, neglecting the small differences in pressure, one gets,
F = ṁ (v5 − v1 ) . (2.290)
The work rate done by the thrust force is the product of this force and the air speed, v1 .
This gives
|Ẇp | = |F · v1 | = |ṁ(v5 − v1 ) · v1 | . (2.291)
Now the efficiency of the cycle is a bit different. The net work of the turbine and compressor
is zero. Instead, the overall efficiency is defined as
Note the following unusual behavior for flight in an ideal inviscid atmosphere in which
the flow always remains attached. In such a flow D’Alembert’s paradox holds: there is
no aerodynamic drag. Consequently there is no need for thrust generation in steady state
operation. Thrust would only be needed to accelerate to a particular velocity. For such
an engine then the exit velocity would equal the entrance velocity: v5 = v1 and F = 0.
Moreover, the overall efficiency is ηo = 0.
Other efficiencies are often defined in jet propulsion. The propulsive efficiency, ηp is often
defined as the propulsive power scaled by the net change in kinetic energy per unit time.
Using our notation, and assuming negligible mass flow rate of the fuel relative to that of the
air, we have
|Ẇp |
ηp = , (2.293)
v52 v12
ṁ 2
− 2
ṁv1 (v5 − v1 )
= 2 , (2.294)
v5 v12
ṁ 2 − 2
v1 (v5 − v1 )
= 2 , (2.295)
v5 v12
2
− 2
2v1 (v5 − v1 )
= , (2.296)
(v5 − v1 ) (v5 + v1 )
2v1
= , (2.297)
v5 + v1
2
= . (2.298)
1 + vv15
|Ẇp | |Ẇp |
ηp = , ηo = . (2.299)
˙
|∆KE| |Q̇H |
˙
|∆KE|
ηth = . (2.300)
|Q̇H |
With these definitions, we get the overall efficiency to be the product of the thermal and
propulsive efficiencies.
ηo = ηth ηp . (2.301)
Example 2.8
(adopted from Çengal and Boles). A turbojet flies with a velocity of 850 ft/s at an altitude where
the air is at 5 psia and −40 F. The compressor has a pressure ratio of 10, and the temperature at
the turbine inlet is 2000 F. Air enters the compressor at a rate of ṁ = 100 lbm/s. Assuming an air
standard with CPIG air, find the
• temperature and pressure at the turbine exit,
• velocity of gas at nozzle exit,
• overall efficiency, and
• propulsive efficiency.
It is usually sufficient to neglect the kinetic energy of the fluid inside the engine. While it still has
a non-zero velocity, it has been slowed enough so that enthalpy dominates kinetic energy inside the
v1 · v1
ha = h1 + , (2.305)
2
v1 · v1
0 = h1 − ha + , (2.306)
2
v1 · v1
= cP (T1 − Ta ) + , (2.307)
2
v1 · v1
Ta = T1 + . (2.308)
2cP
Now get the pressure at the compressor inlet via the isentropic relations:
k/(k−1) 1.4/(1.4−1)
Ta 480 ◦ R
Pa = P1 = (5 psia) = 8.0 psia. (2.311)
T1 420 ◦ R
The temperature at the entrance of the turbine is known to be 2000 F = 2460 ◦ R. Now the turbine
work is equal to the compressor work, so
wc = wt , (2.314)
h2 − ha = h3 − h4 , (2.315)
cP (T2 − Ta ) = cP (T3 − T4 ), (2.316)
T2 − Ta = T3 − T4 , (2.317)
T4 = T3 − T2 + Ta , (2.318)
= 2460 ◦ R − 927 ◦ R + 480 ◦ R, (2.319)
= 2013 ◦ R. (2.320)
Expansion through the nozzle completes the cycle. Assume an isentropic nozzle, therefore
(k−1)/k
P5
T5 = T4 , (2.322)
P4
(k−1)/k
P1
= T4 , (2.323)
P4
(1.4−1)/1.4
5 psia
= = (2013 ◦ R) , (2.324)
39.7 psia
= 1114 ◦ R. (2.325)
v5 · v5 v4 · v4
h5 + = h4 + , (2.326)
2 2 }
| {z
small
v52
0 = h5 − h4 + , (2.327)
2
v5 · v5
= cP (T5 − T4 ) + , (2.328)
p 2
v5 = 2cP (T4 − T5 ), (2.329)
v !
u
u Btu 25037
2
ft
= t 2 0.240 (2013 ◦ R − 1114 ◦ R) s2
, (2.330)
lbm ◦ R 1 Btu
lbm
ft
= 3288 . (2.331)
s
lbm ft ft ft lbm
|F| = ṁ(v5 − v1 ) = 100 3288 − 850 = 243800 , (2.332)
s s s s2
ft lbm
243800 s2
= lbm ft
, (2.333)
32.2 lbf s2
= 7571.4 lbf. (2.334)
The power is the product of the thrust force and the air speed, so
ft Btu
|ẆP | = |F · v1 | = (7571.4 lbf) 850 , (2.335)
s 777.5 ft lbf
Btu
= 8277 , (2.336)
s !
Btu 1.415 hp
= 8277 Btu
, (2.337)
s s
= 11711 hp. (2.338)
The remainder of the energy is excess kinetic energy and excess thermal energy, both of which ultimately
dissipate so as to heat the atmosphere.
The propulsive efficiency is
2 2
ηp = = = 0.410826. (2.344)
1 + vv51 1 + 3288 ft/s
850 ft/s
Other variants of the turbojet engine include the very important turbofan engine in which
a large cowling is added to the engine and a an additional fan in front of the compressor
forces a large fraction of the air to bypass the engine. The turbofan engine is used in most
large passenger jets. Analysis reveals a significant increase in overall efficiency as well as
a reduction in jet noise. Other important variants include the turboprop, propfan, ramjet,
scramjet, jet with afterburners, and rocket.
S = 2Rcrank . (2.346)
See Figure 2.13 for an illustration of this geometry for a piston-cylinder arrangement in two
configurations. The total displacement for all the cylinders is
2Rcrank
Note that
B2
Acyl = π . (2.348)
4
The compression ratio is not the ratio of pressures; it is the ratio of volumes:
Vmax
rv = CR = . (2.349)
Vmin
For these engines the volumes are closed in expansion and compression, so the net work is
I
wnet = P dv = Pmef f (vmax − vmin ). (2.350)
Here one has defined a mean effective pressure: Pmef f . The net work per cylinder per cycle
is
Wnet = mwnet = Pmef f m(vmax − vmin ) = Pmef f (Vmax − Vmin ). (2.351)
Now consider the total power developed for all the cylinders. Assume the piston operates at
a frequency of ν cycles/s:
Ẇnet = Ncyl Pmef f (Vmax − Vmin )ν. (2.352)
s
It is more common to express ν in revolutions per minute: RPM = ν(60 min
), so
RPM
Ẇnet = Pmef f Ncyl (Vmax − Vmin ) s , (2.353)
| {z } 60 min
=Vdispl
RPM
= Pmef f Vdispl s . (2.354)
60 min
This applies for a two-stroke engine. If the engine is a four stroke engine, the net power is
reduced by a factor of 1/2.
2.6 Otto
The air-standard Otto cycle approximates the gasoline engine. It employs a fixed mass
approach. Diagrams for P − v and T − s for the Otto cycle are shown in Figure 2.14.
P T
3
3
isentrope isochore
2
4
2 isentrope 4
1 isochore
1
v s
For isochoric heating, such as 2 → 3, in a fixed mass environment, the first law gives
u3 − u2 = − 2 w3 ,
2 q3 (2.355)
Z v3
= 2 q3 − P dv, but v2 = v3 , (2.356)
v2
Z v2
= 2 q3 − P dv , (2.357)
v2
| {z }
=0
2 q3 = u3 − u2 , (2.358)
= cv (T3 − T2 ), if CPIG. (2.359)
1
η = 1 − rv1−k = 1 − . (2.371)
rvk−1
If the compression ratio increases, the thermal efficiency increases. High compression ratios
introduce detonation in the fuel air mixture. This induces strong pressure waves in the
cylinder and subsequent engine knock. It can cause degradation of piston walls.
Some deviations of actual performance from that of the air-standard Otto cycle are as
follows:
Example 2.9
(adopted from Moran and Shapiro). The temperature at the beginning of the compression process
of an air-standard Otto cycle with a compression ratio of 8 is 540 ◦ R, the pressure is 1 atm, and the
cylinder volume is 0.02 ft3 . The maximum temperature is 3600 ◦ R. Find
• temperature and pressure at each stage of the process,
• thermal efficiency, and
• Pmef f in atm.
One can use the ideal gas law to get the pressure at state 2:
P2 V2 P1 V1
= , (2.375)
T2 T1
V1 T2
P2 = P1 , (2.376)
V2 T1
1240.69 ◦ R
= (1 atm)(8) , (2.377)
540 ◦ R
= 18.3792 atm. (2.378)
Now V3 = V2 because the combustion is isochoric. And the maximum temperature is T3 = 3600 ◦ R.
This allows use of the ideal gas law to get P3 :
P3 V3 P2 V2
= , (2.379)
T3 T2
V2 T3
P3 = P2 , (2.380)
V3 T2
|{z}
=1
3600 ◦ R
= (18.3792 atm)(1) , (2.381)
1240.59 ◦ R
= 53.3333 atm. (2.382)
k−1
T3 V4
= , (2.383)
T4 V3
k−1
T4 V3
= , (2.384)
T3 V4
k−1
V2
T4 = T3 , (2.385)
V1
1−k
V1
= T3 , (2.386)
V2
= (3600 ◦ R)(8)1−1.4 , (2.387)
◦
= 1566.99 R. (2.388)
P4 V4 P3 V3
= , (2.389)
T4 T3
V3 T4
P4 = P3 , (2.390)
V4 T3
V2 T4
= P3 , (2.391)
V1 T3
1 1566.99 ◦ R
= (53.3333 atm) , (2.392)
8 3600 ◦ R
= 2.90184 atm. (2.393)
1
η = 1− , (2.394)
rvk−1
1
= 1 − 1.4−1 , (2.395)
8
= 0.564725. (2.396)
P1 cv T3 − T4 + T1 − T2
= , (2.402)
T1 R 1 − VV2 1
P1 1 T3 − T4 + T1 − T2
= , (2.403)
T1 k − 1 1 − VV2 1
2.7 Diesel
The air standard Diesel cycle approximates the behavior of a Diesel engine. It is modeled
as a fixed mass system. Here the compression is done before injection, so there is no danger
of premature ignition due to detonation. No spark plugs are used. Diagrams for P − v and
T − s for the Diesel cycle are shown in Figure 2.15. One can outline the Diesel cycle as
follows:
P T
2 3
3
isobar
isentrope 2
4
isentrope
4 1 isochore
v s
All else being equal, the Otto cycle will have higher efficiency than the Diesel cycle.
However, the Diesel can operate at higher compression ratios because detonation is not as
serious a problem in the compression ignition engine as it is in the spark ignition.
Example 2.10
(adopted from BS). An air standard Diesel cycle has a compression ratio of 20, and the heat
transferred to the working fluid has 1800 kJ/kg. Take air to be an ideal gas with variable specific heat.
At the beginning of the compression process, P1 = 0.1 MPa, and T1 = 15 ◦ C. Find
• Pressure and temperature at each point in the process,
• Thermal efficiency,
• Pmef f , the mean effective pressure.
The enthalpies and entropies at the reference pressure for variable specific heat air are tabulated
as functions of temperature in Table A7.1. Recall that
Z T
h(T ) = href + cP (T̂ ) dT̂ . (2.412)
Tref
P
= soT (T ) − R ln . (2.417)
Pref
Here the reference pressure is Pref = 0.1 MPa. This happens to be the inlet state, which is a coincidence;
so at the inlet there is no correction to the entropy for pressure deviation from its reference value. At
the inlet, one has T1 = 15 + 273.15 = 288.15 K. One then gets from interpolating Table A7.1 that
kJ kJ kJ
h1 = 288.422 , soT1 = s1 = 6.82816 , u1 = 205.756 . (2.418)
kg kg K kg
For the isentropic compression, one has
kJ 1
s2 = s1 = 6.82816 , V2 = V1 . (2.419)
kg K 20
Now from the ideal gas law, one has
P2 V2 P1 V1
= , (2.420)
T2 T1
P2 T2 V1
= , (2.421)
P1 T1 V2
P2 T2
= 20 . (2.422)
Pref 288 K
Now
P2
s2 (T2 , P2 ) = soT2 − R ln , (2.423)
Pref
kJ kJ T2
6.82816 = soT2 − 0.287 ln 20 , (2.424)
kg K kg K 288 K
kJ T2 kJ
0 = soT2 − 0.287 ln 20 − 6.82816 ≡ F (T2 ). (2.425)
kg K 288 K kg K
kJ kJ
T2 (K) soT2 kg K
F (T2 ) kg K
−4
900 8.01581 8.57198 × 10
1000 8.13493 8.97387 × 10−2
800 7.88514 −9.60091 × 10−2
850 7.95207 −4.64783 × 10−2
899.347 8.01237 −2.37449 × 10−3
This then becomes a trial and error search in Table A7.1 for the T2 which satisfies the previous equa-
tion. If one guesses T2 ∼ 900 K, one gets soT2 = 8.01581 kJ/kg/K, and the right side evaluates to
0.000857198 kJ/kg/K. Performing a trial and error procedure, one finds the results summarized in
Table 2.3. Take T2 = 900 K as close enough!
Then
T2 V1
P2 = P1 , (2.426)
T1 V2
900 K
= (0.1 MPa) (20), (2.427)
288 K
= 6.25 MPa. (2.428)
qH = h3 − h2 , (2.430)
h3 = h2 + qH , (2.431)
kJ kJ
= 933.15 + 1800 , (2.432)
kg kg
kJ
= 2733.15 . (2.433)
kg
One can then interpolate Table A7.1 to find T3 and soT3
2733.15 kJ
kg − 2692.31 kJ
kg
T3 = (2350 K) + ((2400 K) − (2350 K)), (2.434)
kJ kJ
2755 kg − 2692.31 kg
= 2382.17 K, (2.435)
kJ
soT3 = 9.16913
kg K
kJ
2733.15 kJ
kg − 2692.31 kg kJ kJ
+ 9.19586 − 9.16913 , (2.436)
kJ
2755 kg − 2692.31 kJ kg K kg K
kg
kJ
= 9.18633 . (2.437)
kg K
kJ kJ
T4 (K) soT4 kg K
F (T4 ) kg K
−1
400 7.15926 −9.34561 × 10
800 7.88514 −4.07614 × 10−1
1400 8.52891 7.55464 × 10−2
1200 8.34596 −6.31624 × 10−2
1300 8.44046 8.36535 × 10−3
1250 8.39402 −2.68183 × 10−2
1288 8.42931 −1.23117 × 10−4
One also has P3 = P2 = 6.25 MPa. Now get the actual entropy at state 3:
P3
s3 (T3 , P3 ) = soT3 − R ln , (2.438)
Pref
kJ kJ 6.25 MPa
= 9.18633 − 0.287 ln , (2.439)
kg K kg K 0.1 MPa
kJ
= 7.99954 . (2.440)
kg K
Now 3 → 4 is an isentropic expansion to state 4, which has the same volume as state 1; V1 = V4 . So
the ideal gas law gives
P4 V4 P1 V1
= , (2.441)
T4 T1
P4 T4 V1
= , (2.442)
P1 T1 V4
P4 T4 V1
= . (2.443)
Pref T1 V4
Now consider the entropy at state 4, which must be the same as that at state 3:
P4
s4 (T4 , P4 ) = soT4 − R ln , (2.444)
Pref
T4 V1
= soT4 − R ln
s3
T1 V4 , (2.445)
|{z}
=1
kJ o kJ T4
7.99954 = sT4 − 0.287 ln , (2.446)
kg K kg K 288 K
o kJ T4 kJ
0 = sT4 − 0.287 ln − 7.99954 = F (T4 ). (2.447)
kg K 288 K kg K
Using Table A7.1, this equation can be iterated until T4 is found. So T4 = 1288 K. At this temperature,
one has
kJ kJ
h4 = 1381.68 , u4 = 1011.98 . (2.448)
kg kg
= 0.552098. (2.451)
Now one has
kJ kJ kJ
qL = u1 − u4 = 205.756 − 1011.98 = −806.224 . (2.452)
kg kg kg
So
kJ kJ kJ
wnet = qH + qL = 1800 − 806.224 = 993.776 . (2.453)
kg kg kg
So
Wnet
Pmef f = , (2.454)
Vmax − Vmin
wnet
= , (2.455)
vmax − vmin
wnet
= , (2.456)
v1 − v2
wnet
= RT1 RT2
, (2.457)
P1 − P2
kJ
993.776 kg
= , (2.458)
kJ 288 K 900 K
0.287 kg K 100 kPa − 6250 kPa
2.8 Stirling
Another often-studied air-standard engine is given by the Stirling cycle. This is similar to
the Otto cycle except the adiabatic processes are replaced by isothermal ones. The efficiency
can be shown to be equal to that of a Carnot engine. Stirling engines are difficult to build.
Diagrams for P − v and T − s for the Stirling cycle are shown in Figure 2.16.
One can outline the Stirling cycle as follows:
• 1 → 2: isothermal compression in the compression stroke,
• 2 → 3: isochoric heat transfer in the combustion stroke,
• 3 → 4: isothermal expansion in power stroke, and
• 4 → 1: isochoric rejection of heat to the surroundings.
P T
3
3 4
isotherm isochore
isochore
2 isotherm 4
2 1
1
v s
2.9 Refrigeration
A simple way to think of a refrigerator is a cyclic heat engine operating in reverse. Rather
than extracting work from heat transfer from a high temperature source and rejecting heat
to a low temperature source, the refrigerator takes a work input to move heat from a low
temperature source to a high temperature source.
2.9.1 Vapor-compression
A common refrigerator is based on a vapor-compression cycle. This is a Rankine cycle in
reverse. While one could employ a turbine to extract some work, it is often impractical.
Instead the high pressure gas is simply irreversibly throttled down to low low pressure.
One can outline the vapor-compression refrigeration cycle as follows:
• 1 → 2: isentropic compression,
3 condenser 2
expansion compressor
valve
4 evaporator
1 T
2
compressor
3 condenser
temperature of
surroundings
valv
evaporator
e
temperature inside
refrigerator 4 1 refrigerator
Figure 2.17: Schematic and T − s diagrams for the vapor-compression refrigeration cycle.
of performance, β, is defined as
what one wants
β = , (2.460)
what one pays for
qL
= . (2.461)
wc
A heat pump is effectively the same as a refrigerator, except one desires qH rather than
qL . So for a heat pump, the coefficient of performance, β ′, is defined as
qH
β′ = . (2.462)
wc
Example 2.11
(from Moran and Shapiro) R − 12 is the working fluid in an ideal vapor-compression refrigeration
cycle that communicates thermally with a cold region at 20 ◦ C and a warm region at 40 ◦ C. Saturated
vapor enters the compressor at 20 ◦ C and saturated liquid leaves the condenser at 40 ◦ C. The mass
flow rate of the refrigerant is 0.008 kg/s. Find
• compressor power in kW,
• refrigeration capacity in ton,
At the compressor inlet for the ideal cycle, one has a saturated vapor.
T1 = 20 ◦ C, x1 = 1. (2.463)
The tables then give
kJ kJ
h1 = 195.78 , s1 = 0.6884 , P1 = 5.6729 bar. (2.464)
kg kg K
Now at the end of the condenser (state 3), one has T3 = 40 ◦ C. The condenser is on an isobar, so the
pressure is the saturation pressure at the temperature, which is
P3 = P2 = 9.6065 bar. (2.465)
So, two properties at the end of the compression are known: pressure and entropy. The allows deter-
mination of the enthalpy at the end of compression via the tables:
kJ
h2 = 205.1 . (2.466)
kg
State 3 is at the end of the condenser, so x3 = 0, and one finds the enthalpy from the tables to be
kJ
h3 = 74.59 . (2.467)
kg
Now the throttling device has constant enthalpy, so
kJ
h4 = h3 = 74.59 . (2.468)
kg
However, the fluid has been throttled to a lower pressure: that of the evaporator, which is the same as
state 1, the compressor inlet, so
P4 = P1 = 5.6729 bar. (2.469)
The compressor work is
Ẇc = ṁ(h2 − h1 ), (2.470)
kg kJ kJ
= 0.008 205.1 − 195.78 , (2.471)
s kg kg
= 0.075 kW. (2.472)
Now one desires the heat which leaves the cold region to be high for a good refrigerator. This is the
heat transfer at the low temperature part of the T − s diagram, which here gives
Q̇L = ṁ(h1 − h4 ), (2.473)
kg kJ kJ
= 0.008 195.78 − 74.59 , (2.474)
s kg kg
= 0.9695 kW, (2.475)
!
kJ 60 s 1 ton
= 0.9695 kJ
, (2.476)
s min 211 min
= 0.276 ton. (2.477)
The unit of ton is used for power and is common in the refrigeration industry. It is the power required
to freeze one “short ton” of water at 0 ◦ C in 24 hours. It is 12000 Btu/hr, 3.516853 kW, or 4.7162 hp.
A short ton is a U.S ton, which is 2000 lbm. A long ton is British and is 2240 lbm or 1016.047 kg. A
metric ton, also called a tonne, is 1000 kg and is 2204 lbm.
The coefficient of performance is
Q̇L
β = , (2.478)
Ẇc
0.9695 kW
= , (2.479)
0.075 kW
= 12.93. (2.480)
Q̇L
βmax = , (2.481)
Ẇc
Q̇L
= , (2.482)
Q̇H − Q̇L
1
= , (2.483)
Q̇H
Q̇
−1
L
1
= TH
, (2.484)
TL −1
TL
= , (2.485)
TH − TL
20 + 273.15
= , (2.486)
40 + 273.15 − (20 + 273.15)
= 14.6575. (2.487)
3
2
w
expander compressor
33.0 quad
ηU S = = 0.329. (2.488)
(33.0 quad) + (67.3 quad)
Figure 2.21 shows a comparable plot for 2011. One can see evolution of the energy
budget.
Example 2.12
If all the waste heat in the US in 2022 were directed into Lake Michigan, find its temperature rise.
T
2
isobar
4 isobar
Now, Lake Michigan has a volume of 4900 km3 . Therefore the mass of water in Lake Michigan is
roughly
kg 3
103 m 3
m = ρV = 997 3 4900 km = 4.88 × 1015 kg. (2.490)
m km
If all the waste energy were dumped into Lake Michigan, we could expect from a first law analysis to
find a temperature rise of
Q 7.10 × 1019 J
∆T = = = 3.5 K. (2.491)
mcP (4.88 × 1015 kg) 4180 J
kg K
Locally on the University of Notre Dame campus, both St. Mary’s and St. Joseph’s Lakes would be
vaporized many times over.
Figure 2.20: Chart of distribution of energy sources and usage in the US in 2022. Data from
Lawrence Livermore National Laboratory, https://flowcharts.llnl.gov/.
Figure 2.21: Chart of distribution of energy sources and usage in the US in 2011. Data from
Lawrence Livermore National Laboratory, https://flowcharts.llnl.gov/.
Gas mixtures
One is often faced with mixtures of simple compressible substances, and it the thermody-
namics of such mixtures upon which attention is now fixed. Here a discussion of some of
the fundamentals of mixture theory will be given. In general, thermodynamics of mixtures
can be a challenging topic about which much remains to be learned. In particular, these
notes will focus on ideal mixtures of ideal gases, for which results are often consistent with
intuition. The chemical engineering literature contains a full discussion of the many nuances
associated with non-ideal mixtures of non-ideal materials.
N
X
m = m1 + m2 + m3 + · · · + mN = mi , mass, units=kg, (3.1)
i=1
N
X
n = n1 + n2 + n3 + · · · + nN = ni , moles, units=kmole. (3.2)
i=1
83
84 CHAPTER 3. GAS MIXTURES
Recall 1 mole = 6.02214179 × 1023 molecule. The mass fraction of component i is defined as
ci 1 :
mi
ci ≡ , mass fraction, dimensionless. (3.3)
m
ni
yi ≡ , mole fraction, dimensionless. (3.4)
n
Now the molecular mass of species i is the mass of a mole of species i. It units are typically
g/mole. This is an identical unit to kg/kmole. Molecular mass is sometimes called “molec-
ular weight,” but this is formally incorrect, as it is a mass measure, not a force measure.
Mathematically the definition of Mi corresponds to
mi kg g
Mi ≡ , = . (3.5)
ni kmole mole
mi
ci = , (3.6)
m
ni Mi
= , (3.7)
m
ni Mi
= PN , (3.8)
j=1 mj
ni Mi
= PN , (3.9)
j=1 nj M j
ni Mi
= 1
PN n , (3.10)
n j=1 nj Mj
ni Mi
n
= P N nj Mj , (3.11)
j=1 n
yi Mi
= PN . (3.12)
j=1 y j Mj
1
Note that ci is the unusual notation for mass fraction used by BS, to which we adhere. A more common
notation for mass fraction from the combustion and physical chemistry literature is Yi .
2
The notation yi for mole fraction is also the notation of BS and much of the combustion and physical
chemistry literature.
Similarly, one finds mole fraction in terms of mass fraction by the following:
ni
yi = , (3.13)
n
mi
Mi
= PN mj , (3.14)
j=1 Mj
mi
Mi m
= PN mj , (3.15)
j=1 Mj m
ci
Mi
= P N cj . (3.16)
j=1 Mj
Example 3.1
Air is often modeled as a mixture in the following molar ratios:
O2 + 3.76N2 . (3.21)
Find the mole fractions, the mass fractions, and the mean molecular mass of the mixture.
n1 = 1 kmole, (3.22)
and N2 to be
n2 = 3.76 kmole. (3.23)
The molecular mass of O2 is M1 = 32 kg/kmole. The molecular mass of N2 is M2 = 28 kg/kmole. The
total number of moles is
Note that
N
X
yi = 1. (3.27)
i=1
That is to say y1 + y2 = 0.2101 + 0.7899 = 1. Now for the masses, one has
kg
m1 = n1 M1 = (1 kmole) 32 = 32 kg, (3.28)
kmole
kg
m2 = n2 M2 = (3.76 kmole) 28 = 105.28 kg. (3.29)
kmole
So one has
m = m1 + m2 = 32 kg + 105.28 kg = 137.28 kg. (3.30)
The mass fractions then are
m1 32 kg
c1 = = = 0.2331, (3.31)
m 137.28 kg
m2 105.28 kg
c2 = = = 0.7669. (3.32)
m 137.28 kg
Note that
N
X
ci = 1. (3.33)
i=1
That is c1 + c2 = 0.2331 + 0.7669 = 1. Now for the mixture molecular mass, one has
m 137.28 kg kg
M= = = 28.84 . (3.34)
n 4.76 kmole kmole
Now postulates for mixtures are not as well established as those for pure substances.
The literature has much controversial discussion of the subject. A strong advocate of the
axiomatic approach, Truesdell (1984), proposed the following “metaphysical principles” for
mixtures, which are worth considering.
3. The motion of the mixture is governed by the same equations as is a single body.
Most important for the present discussion is the first principle. When coupled with fluid
mechanics, the second two take on additional importance. The approach of mixture theory
is to divide and conquer. One typically treats each of the constituents as a single material
and then devises appropriate average or mixture properties from those of the constituents.
The best example of this is air, which is not a single material, but is often treated as
such.
U = U(T, P, n1 , n2 , . . . , nN ). (3.36)
A partial molar property is a generalization of an intensive property, and is defined such that
it is the partial derivative of an extensive property with respect to number of moles, with T
and P held constant. For internal energy, then the partial molar internal energy is
∂U
ui ≡ . (3.37)
∂ni T,P,nj ,i6=j
Pressure and temperature are held constant because those are convenient variables to control
in an experiment. One also has the partial molar volume
∂V
vi = . (3.38)
∂ni T,P,nj ,i6=j
It shall be soon seen that there are other natural ways to think of the volume per mole.
Now in general one would expect to find
ui = ui (T, P, n1 , n2 , . . . , nN ), (3.39)
vi = vi (T, P, n1, n2 , . . . , nN ). (3.40)
This is the case for what is known as a non-ideal mixture. An ideal mixture is defined as a
mixture for which the partial molar properties ui and vi are not functions of the composition,
that is
ui = ui (T, P ), if ideal mixture, (3.41)
vi = vi (T, P ), if ideal mixture. (3.42)
An ideal mixture also has the property that hi = hi (T, P ), while for a non-ideal mixture
hi = hi (T, P, n1 , . . . , nN ). Though not obvious, it will turn out that some properties of an
ideal mixture will depend on composition. For example, the entropy of a constituent of an
ideal mixture will be such that
si = si (T, P, n1, n2 , . . . , nN ). (3.43)
• Each constituent possesses a partial pressure which sums to form the total pressure of
the mixture.
The above characterize a Dalton model for any gas, ideal or non-ideal. One also takes for
convenience
It is more convenient to deal on a molar basis for such a theory. For the Dalton model,
additional useful quantities, the species mass concentration ρi , the mixture mass concentra-
tion ρ, the species molar concentration ρi , and the mixture molar concentration ρ, can be
defined. As will be seen, these definitions for concentrations are useful; however, they are not
in common usage. Following BS, the bar notation, ·, will be reserved for properties which
are mole-based rather than mass-based. As mentioned earlier, the notion of a partial molal
property is discussed extensively in the chemical engineering literature and has implications
beyond those considered here. For the Dalton model, in which each component occupies the
same volume, one has
Vi = V. (3.44)
The mixture mass concentration, also called the density is simply
m kg
ρ= , . (3.45)
V m3
One can find a convenient relation between species molar concentration and species mole
fraction by the following operations
ni n
ρi = , (3.49)
V n
ni n
= , (3.50)
nV
= yi ρ. (3.51)
A similar relation exists between species molar concentration and species mass fraction via
ni m Mi
ρi = , (3.52)
V m Mi
=m
z }|i{
m ni Mi
= , (3.53)
V mMi
=ci
z}|{
mi 1
= ρ , (3.54)
m Mi
ci
= ρ . (3.55)
Mi
The specific volumes, mass and molar, are similar. One takes
V V
v = , v= , (3.56)
m n
V V
vi = , vi = . (3.57)
mi ni
This definition of molar specific volume is not the partial molar volume defined in the chem-
ical engineering literature, which takes the form vi = ∂V /∂ni |T,P,nj ,i6=j .
For the partial pressure of species i, one can say for the Dalton model
N
X
P = Pi . (3.58)
i=1
Pi V = ni RT, (3.59)
ni RT
Pi = , (3.60)
V
N
X XN
ni RT
Pi = , (3.61)
i=1 i=1
V
| {z }
=P
N
RT X
P = ni . (3.62)
V i=1
| {z }
=n
Here n is the total number of moles in the system. Additionally R is the universal gas
constant with value
kJ J
R = 8.314472 = 8.314472 . (3.65)
kmole K mole K
Sometimes this is expressed in terms of kB , the Boltzmann constant, and N , Avogadro’s
number:
R = kB N , (3.66)
molecule
N = 6.02214179 × 1023 , (3.67)
mole
J
kB = 1.380650 × 10−23 . (3.68)
K molecule
Example 3.2
Compare the molar specific volume defined here with the partial molar volume from the chemical
engineering literature.
∂V
vi = . (3.69)
∂ni T,P,nj ,i6=j
δki = 0, k 6= i, (3.78)
δki = 1, k = i. (3.79)
Contrast this with the earlier adopted definition of molar specific volume
V
vi = . (3.80)
ni
So, why is there a difference? The molar specific volume is a simple definition. One takes the
instantaneous volume V , which is shared by all species in the Dalton model, and scales it by the
instantaneous number of moles of species i, and acquires a natural definition of molar specific volume
consistent with the notion of a mass specific volume. On the other hand, the partial molar volume
specifies how the volume changes if the number of moles of species i changes, while holding T and P
and all other species mole numbers constant. One can imagine adding a mole of species i, which would
necessitate a change in V in order to guarantee the P remain fixed.
n = nA + nB . (3.81)
PA VA = nA RTA , (3.82)
PB VB = nB RTB . (3.83)
PA V = nA RT, (3.84)
PB V = nB RT. (3.85)
PV
n = , (3.87)
RT
PA V
nA = , (3.88)
RT
PB V
nB = . (3.89)
RT
Now n = nA + nB , so one has
PV PA V PB V
= + . (3.90)
RT RT RT
P = PA + PB . (3.91)
That is the total pressure is the sum of the partial pressures. This is a mixture rule for
pressure
One can also scale each constituent ideal gas law by the mixture ideal gas law to get
PA V nA RT
= , (3.92)
PV nRT
PA nA
= , (3.93)
P n
= yA , (3.94)
PA = yA P. (3.95)
Likewise
PB = yB P. (3.96)
Now, one also desires rational mixture rules for energy, enthalpy, and entropy. Invoke Trues-
dell’s principles on a mass basis for internal energy. Then the total internal energy U (with
U = mu = mA uA + mB uB , (3.97)
m mB
A
= m uA + uB , (3.98)
m m
= m (cA uA + cB uB ) , (3.99)
u = cA u A + cB u B . (3.100)
H = mh = mA hA + mB hB , (3.101)
m mB
A
= m hA + hB , (3.102)
m m
= m (cA hA + cB hB ) , (3.103)
h = cA hA + cB hB . (3.104)
It is easy to extend this to a mole fraction basis rather than a mass fraction basis. One can
also obtain a gas constant for the mixture on a mass basis. For the mixture, one has
S = ms = mA sA + mB sB , (3.113)
m mB
A
= m sA + sB , (3.114)
m m
= m (cA sA + cB sB ) , (3.115)
s = cA sA + cB sB . (3.116)
Note that sA is evaluated at T and PA , while sB is evaluated at T and PB . For a CPIG, one
has
T PA
sA = so298,A + cP A ln −RA ln , (3.117)
To Po
| {z }
≡soT,A
T yA P
= so298,A + cP A ln − RA ln . (3.118)
To Po
Likewise
T yB P
sB = so298,B + cP B ln −RB ln . (3.119)
To Po
| {z }
≡soT,B
Here the “o” denotes some reference state. As a superscript, it typically means that the
property is evaluated at a reference pressure. For example, soT,A denotes the portion of the
entropy of component A that is evaluated at the reference pressure Po and is allowed to vary
with temperature T . Note also that sA = sA (T, P, yA) and sB = sB (T, P, yB ), so the entropy
of a single constituent depends on the composition of the mixture and not just on T and P .
This contrasts with energy and enthalpy for which uA = uA (T ), uB = uB (T ), hA = hA (T ),
hB = hB (T ) if the mixture is composed of ideal gases. Occasionally, one finds hoA and hoB
used as a notation. This denotes that the enthalpy is evaluated at the reference pressure.
However, if the gas is ideal, the enthalpy is not a function of pressure and hA = hoA , hB = hoB .
If one is employing a calorically imperfect ideal gas model, then one finds for species i
that
o yi P
si = sT,i − Ri ln , i = A, B. (3.120)
Po
Example 3.3
Initially calorically perfect ideal gases A and B are segregated within the same large volume by
a thin frictionless, thermally conducting diaphragm. Thus, both are at the same initial pressure and
temperature, P1 and T1 . The total volume is thermally insulated and fixed, so there are no global heat
or work exchanges with the environment. The diaphragm is removed, and A and B are allowed to mix.
Assume A has mass mA and B has mass mB . The gases are allowed to have distinct molecular masses,
MA and MB . Find the final temperature T2 , pressure P2 , and the change in entropy.
(mA RA + mB RB ) T1
P2 = , (3.137)
(mA RA + mB RB ) PT11
= P1 . (3.138)
Likewise
≥ 0. (3.150)
We can also scale Eq. (3.148) by Rn to get
1 ∆S nA nB
= −
n ln yA2 + n ln yB2 , (3.151)
n
R |{z} |{z} |{z}
=∆s =yA2 =yB2
∆s
= − (yA2 ln yA2 + yB2 ln yB2 ) , (3.152)
R
yA2 yB2
= − (ln yA2 + ln yB2 ), (3.153)
yA2 yB2
= − ln (yA2 yB2 ) . (3.154)
For an N-component mixture, mixed in the same fashion such that P and T are constant,
this extends to
N
X
∆S = −R nk ln yk , (3.155)
k=1
N
!
X nk
= −R nk ln PN ≥ 0, (3.156)
k=1 i=1 ni
| {z }
≤0
N
X nk
= −R n ln yk , (3.157)
k=1
n
XN
nk
= −Rn ln yk , (3.158)
k=1
n
XN
m
= −R yk ln yk , (3.159)
M k=1
N
X
= −Rm ln ykyk , (3.160)
k=1
= −Rm (ln y1y1 + ln y2y2 + · · · + ln yN
yN
), (3.161)
y1 y2 yN
= −Rm ln (y1 y2 . . . yN ) , (3.162)
N
!
Y
= −Rm ln ykyk , (3.163)
k=1
N
!
∆s ∆s Y
= = − ln ykyk . (3.164)
R R k=1
There is a fundamental dependency of the mixing entropy on the mole fractions. Because
0 ≤ yk ≤ 1, the product is guaranteed to be between 0 and 1. The natural logarithm of such
a number is negative, and thus the entropy change for the mixture is guaranteed positive
semi-definite. For the entropy of mixing, Truesdell’s third principle is not enforced.
Now if one mole of pure N2 is mixed with one mole of pure O2 , one certainly expects
the resulting homogeneous mixture to have a higher entropy than the two pure components.
But what if one mole of pure N2 is mixed with another mole of pure N2 . Then we would
expect no increase in entropy. However, if we had the unusual ability to distinguish N2
molecules whose origin was from each respective original chamber, then indeed there would
be an entropy of mixing. Increases in entropy thus do correspond to increases in disorder.
where
cv ≡ cA cvA + cB cvB . (3.167)
Similarly for enthalpy
where
cP ≡ cA cP A + cB cP B . (3.170)
For the entropy
The mixture behaves as a pure substance when the appropriate mixture properties are de-
fined. One can also take
cP
k= . (3.174)
cv
N
X
M = yi Mi , (3.175)
i=1
XN
ρ = ρi , (3.176)
i=1
1 1
v = PN 1
= , (3.177)
i=1 vi
ρ
N
X
u = ci u i , (3.178)
i=1
XN
h = ci hi , (3.179)
i=1
=R
N N
z }| { N
R X X yi Mi Ri R X
R = = ci Ri = = yi , (3.180)
M i=1 i=1
N
X M i=1
yj Mj | {z }
=1
j=1
| {z } =M
N
X
cv = ci cvi , (3.181)
i=1
cv = cP − R, if ideal gas, (3.182)
XN
cP = ci cP i , (3.183)
i=1
cP
k = , (3.184)
cv
XN
s = ci si , (3.185)
i=1
yi Mi
ci = , (3.186)
M
Pi = yi P, (3.187)
ρi = ci ρ, (3.188)
v 1
vi = = , (3.189)
ci ρi
V = Vi , (3.190)
T = Ti , (3.191)
hi = hoi , if ideal gas, (3.192)
Pi
hi = ui + = ui + Pi vi = ui + Ri T , (3.193)
ρi | {z }
if ideal gas
Z T
hi = ho298,i + cP i (T̂ ) dT̂ , (3.194)
298
| {z }
if ideal gas
Z T
cP i (T̂ ) Pi
si = so298,i + dT̂ −Ri ln , (3.195)
Po
| {z T̂
298
}
=soT,i
| {z }
if ideal gas
yi P Pi
si = soT,i − Ri ln = soT,i − Ri ln , (3.196)
Po Po
| {z } | {z }
if ideal gas if ideal gas
Ri T
Pi = ρi Ri T = ρRi T ci = , (3.197)
vi
| {z }
if ideal gas
X N
ci RT
P = ρRT = ρRT = , (3.198)
i=1
Mi v
| {z }
if ideal gas
N
X Z T
h = ci ho298,i + cP (T̂ ) dT̂ , (3.199)
298
|i=1 {z }
if ideal gas
P
h = u+ = u + P v = |u +{zRT} , (3.200)
ρ
if ideal gas
N Z N
!
X T
cP (T̂ ) P Y
s = ci so298,i + dT̂ − R ln − R ln yiyi .
298 T̂ Po
|i=1 {z i=1
}
if ideal gas
(3.201)
These relations are not obvious. A few are derived in examples here.
Example 3.4
Derive the expression h = u + P/ρ.
Start from the equation for the constituent hi , multiply by mass fractions, sum over all species,
and use properties of mixtures:
Pi
hi = ui + , (3.202)
ρi
Pi
ci h i = ci u i + ci , (3.203)
ρi
N
X XN XN
Pi
ci h i = ci u i + ci , (3.204)
i=1 i=1 i=1
ρi
N
X N
X N
X ρi Ri T
ci h i = ci u i + ci , (3.205)
i=1 i=1 i=1
ρi
| {z } | {z }
=h =u
N
X
h = u+T ci Ri , (3.206)
i=1
| {z }
=R
= u + RT, (3.207)
P
= u+ . (3.208)
ρ
Example 3.5
Find the expression for mixture entropy of the ideal gas.
Z T
cP i (T̂ ) Pi
si = so298,i + dT̂ − Ri ln , (3.209)
298 T̂ Po
Z T
cP i (T̂ ) Pi
ci si = ci so298,i + ci dT̂ − ci Ri ln , (3.210)
298 T̂ Po
N
X N
X XN Z T N
X
cP i (T̂ ) Pi
s= ci si = ci so298,i + ci dT̂ − ci Ri ln , (3.211)
i=1 i=1 i=1 298 T̂ i=1
Po
N
X Z T N
X N
X
ci cP i (T̂ ) Pi
= ci so298,i + dT̂ − ci Ri ln , (3.212)
i=1 298 i=1 T̂ i=1
Po
Z T N
X
cP (T̂ ) Pi
= so298 + dT̂ − ci Ri ln . (3.213)
298 T̂ i=1
Po
All except the last term are natural extensions of the property for a single material. Consider now the
last term involving pressure ratios.
N N !
X Pi X Pi P P
− ci Ri ln = − ci Ri ln + R ln − R ln , (3.214)
i=1
Po i=1
Po Po Po
N !
X Ri Pi P P
= −R ci ln + ln − ln , (3.215)
i=1
R Po Po Po
N !
X R/Mi Pi P P
= −R ci P N ln + ln − ln , (3.216)
i=1 j=1 cj R/Mj
Po Po Po
X !
N ci /Mi Pi P P
= −R PN ln + ln − ln , (3.217)
P P P
i=1 j=1 c j /M j o o o
| {z }
=yi
N !
X Pi P P
= −R yi ln + ln − ln , (3.218)
i=1
Po Po Po
N yi !
X Pi P P
= −R ln − ln + ln , (3.219)
i=1
Po Po Po
YN y ! !
Pi i Po P
= −R ln + ln + ln , (3.220)
i=1
Po P Po
N y ! !
Po Y Pi i P
= −R ln + ln , (3.221)
P i=1 Po Po
N
! !
Po 1 Y yi P
= −R ln PN PN (Pi ) + ln , (3.222)
P i=1 yi Po i=1 yi i=1 Po
N y! !
Y Pi i
P
= −R ln + ln , (3.223)
i=1
P Po
YN y ! !
yi P i P
= −R ln + ln , (3.224)
i=1
P Po
N
! !
Y yi P
= −R ln yi + ln . (3.225)
i=1
Po
The extra entropy is not found in the theory for a single material, and in fact is not in the form suggested
by Truesdell’s postulates. While it is in fact possible to redefine the constituent entropy definition in
such a fashion that the mixture entropy in fact takes on the classical form of a single material via
RT
the definition si = so298,i + 298 (cP i (T̂ )/T̂ ) dT̂ − Ri ln (Pi /Po ) + Ri ln yi , this has the disadvantage of
predicting no entropy change after mixing two pure substances. Such a theory would suggest that this
obviously irreversible process is in fact reversible.
N
X
cv = yi cvi = cv M, (3.232)
i=1
cv = cP − R, (3.233)
XN
cP = yi cP i = cP M, (3.234)
i=1
cP
k = , (3.235)
cv
XN
s = yi si = sM, (3.236)
i=1
ρi
ρ i = yi ρ = , (3.237)
Mi
V v 1
vi = = = = vi Mi , (3.238)
ni yi ρi
∂V V
vi = = = v = vM , (3.239)
∂ni P,T,nj |n {z }
if ideal gas
Pi = yi P, (3.240)
RT
P = ρRT = , (3.241)
| {z v }
if ideal gas
RT
Pi = ρi RT = , (3.242)
vi
| {z }
if ideal gas
P
h = u+ = u + Pv = u
| +{zRT} = hM, (3.243)
ρ
if ideal gas
o
hi = hi , if ideal gas, (3.244)
Pi
hi = ui + = ui + Pi v i = ui + P vi = ui + RT = hi Mi , (3.245)
ρi | {z }
if ideal gas
Z T
o
hi = h298,i + cP i (T̂ ) dT̂ = hi Mi , (3.246)
298
| {z }
if ideal gas
Z T
cP i (T̂ ) yi P
si = so298,i + dT̂ −R ln , (3.247)
Po
| 298
{z T̂ }
=soT,i
| {z }
if ideal gas
o yi P
si = sT,i − R ln = si Mi , (3.248)
Po
| {z }
if ideal gas
N Z N
!
X T
cP (T̂ ) P Y
s = yi so298,i + dT̂ − R ln − R ln yiyi = sM. (3.249)
i=1 298 T̂ Po i=1
| {z }
if ideal gas
Considering an air water vapor mixture, one models the water vapor as an ideal gas and
expresses the total pressure as
P = Pa + Pv . (3.257)
Here v denotes vapor and a denotes air. A good model for the enthalpy of the water vapor
is to take
hv (T, low P ) = hg (T ). (3.258)
Some definitions:
• absolute humidity: ω, the mass of water present in a unit mass of dry air, also called
humidity ratio,
mv
ω ≡ , (3.260)
ma
Mv nv
= , (3.261)
Ma na
Pv V
Mv RT
= , (3.262)
Ma PRT aV
Mv Pv
= , (3.263)
Ma Pa
kg
18.015 kmole Pv
= kg
, (3.264)
28.97 kmole Pa
Pv
= 0.622 , (3.265)
Pa
Pv
= 0.622 . (3.266)
P − Pv
• dew point: temperature at which the vapor condenses when it is cooled isobarically.
• saturated air: The vapor in the air-vapor mixture is at the saturation temperature and
pressure.
• relative humidity: The ratio of the mole fraction of the vapor in the mixture to the mole
fraction of vapor in a saturated mixture at the same temperature and total pressure:
nv
n
φ ≡ ng , (3.267)
n
nv
= , (3.268)
ng
Pv V
RT
= Pg V
, (3.269)
RT
Pv
= . (3.270)
Pg
Here the subscript g denotes saturated gas values. Combining, one can relate the
relative humidity to the absolute humidity:
ωPa
φ= . (3.271)
0.622Pg
Example 3.6
(from Çengal and Boles) A 5 m × 5 m × 3 m room contains air at 25 ◦ C and 100 kPa at a relative
humidity of 75%. Find the
• partial pressure of dry air,
• absolute humidity (i.e. humidity ratio),
• masses of dry air and water vapor in the room, and the
• dew point.
P = Pa + Pv . (3.272)
Pv = φPg . (3.273)
Here Pg is the saturation pressure at the same temperature, which is 25 ◦ C. At 25 ◦ C, the tables give
So
Pv = 0.75(3.169 kPa) = 2.3675 kPa. (3.275)
So from the definition of partial pressure
Pa = P − Pv , (3.276)
= 100 kPa − 2.3675 kPa, (3.277)
= 97.62 kPa. (3.278)
Now for the absolute humidity (or specific humidity), one has
Pv
ω = 0.622 , (3.279)
Pa
2.3675 kPa
= 0.622 , (3.280)
97.62 kPa
kg H2 O
= 0.0152 . (3.281)
kg dry air
Now for the masses of air and water, one can use the partial pressures:
Pa V
ma = Ma , (3.282)
RT
Pa V
= , (3.283)
Ra T
(97.62 kPa) 75 m3
= kJ
, (3.284)
8.314 kmole
28.97 kg
K
(298 K)
kmole
mv = ωma , (3.290)
= (0.0152)(85.61 kg), (3.291)
= 1.3 kg. (3.292)
Now the dew point is the saturation temperature at the partial pressure of the water vapor. With
Pv = 2.3675 kPa, the saturation tables give
Example 3.7
(adopted from BS). An air-water vapor mixture enters the cooling coils of an air conditioner unit.
The inlet is at P1 = 105 kPa, T1 = 30 ◦ C, φ1 = 0.80. The exit state is P2 = 100 kPa, T2 = 15 ◦ C,
φ2 = 0.95. Liquid water at 15 ◦ C also exits the system. Find the heat transfer per kilogram of dry air.
At state 2, the mass flow of water is in both liquid and vapor form.
The first law for the control volume give
dEcv
= Q̇cv − Ẇcv +ṁa ha1 + ṁv1 hv1 − ṁa ha2 − ṁv2 hv2 − ṁl2 hl2 , (3.296)
|{z}
| dt
{z } =0
=0
Q̇cv + ṁa ha1 + ṁv1 hv1 = ṁa ha2 + ṁv2 hv2 + (ṁv1 − ṁv2 )hl2 , (3.297)
Q̇cv ṁv1 ṁv2 ṁv1 − ṁv2
+ ha1 + hv1 = ha2 + hv2 + hl2 , (3.298)
ṁa ṁa ṁa ṁa
Q̇cv
+ ha1 + ω1 hv1 = ha2 + ω2 hv2 + (ω1 − ω2 )hl2 , (3.299)
ṁa
Q̇cv
= ha2 − ha1 − ω1 hv1 + ω2 hv2 + (ω1 − ω2 )hl2 , (3.300)
ṁa
= cpa (T2 − T1 ) − ω1 hv1 + ω2 hv2 + (ω1 − ω2 )hl2 . (3.301)
Now at the inlet, one has from the definition of relative humidity
Pv1
φ1 = . (3.302)
Pg1
Here Pg is the saturated vapor pressure at the inlet temperature, T1 = 30 ◦ C. This is Pg1 = 4.246 kPa.
So one gets
Pv2
ω2 = 0.622 , (3.312)
P2 − Pv2
1.620 kPa
= 0.622 , (3.313)
100 kPa − 1.620 kPa
= 0.0102. (3.314)
dmair
cv
= ṁa1 − ṁa2 , air (3.317)
dt
| {z }
=0
dmwater
cv
= ṁv1 + ṁl − ṁv2 , water (3.318)
dt
| {z }
=0
dEcv
= Q̇cv − Ẇcv +ṁa1 ha1 + ṁv1 hv1 + ṁl hl − ṁa2 ha2 − ṁv2 hv2 . (3.319)
dt }
| {z |{z} |{z}
=0 =0
=0
Example 3.8
(adopted from BS). The pressure of the mixture entering and leaving the adiabatic saturater is
14.7 psia, the entering temperature is 84 F, and the temperature leaving is 70 F, which is the adiabatic
saturation temperature. Calculate the humidity ratio and the relative humidity of the air-water vapor
mixture entering.
lbm H2 O
= 0.012895 . (3.339)
lbm dry air
Here hv1 was estimated as the saturated vapor value at T = 84 F by interpolating the tables. In the
absence of more information regarding the initial vapor state, this estimate is good as any. The value
of hl is estimated as the saturated liquid value at T = 70 F. Now
Pv1
ω1 = 0.622 , (3.340)
P1 − Pv1
P1 ω1
Pv1 = , (3.341)
0.622 + ω1
(14.7 psia)(0.0124895)
= , (3.342)
0.622 + 0.0124895
= 0.28936 psia. (3.343)
For the relative humidity
Pv1
φ1 = , (3.344)
Pg1
0.28936 psia
= , (3.345)
0.584 psia
= 0.495. (3.346)
0.0025
gmofDryAir)
Specifi
/gm
mWater/
c
Volum
0%
ity
10
(am
)
id
(°°
e0.990
HumidityRatio(a
(a
um
ture
eH
era
m° /K
20
%
ativ
mp
60
Re l
nTe
gDryfi
atio
50%
tur
rSa
r
ullo
tB
We
0.010
sob
98'°
98
200
10%
0% 0.000
15 20 25 30 35 40 45 50
Dry
yBu
Bulb
lbTTem
lb empe
perature(C)
Figure 3.1: Psychrometric chart for air-water mixtures,
https://commons.wikimedia.org/wiki/File:Psychrometric.SeaLevel.SI.svg.
CC BY-NC-ND. 12 March 2024, J. M. Powers.
Chapter 4
Mathematical foundations of
thermodynamics
δw = P dv, (4.2)
can be shown to be an inexact differential so that the work is indeed a function of the process
involved. Here we use the notation δ to emphasize that this is an inexact differential.
Example 4.1
Show the work is not a state function.
Let us employ the mode of argumentation commonly known as reductio ad absurdum in which we
show the denial of the contention leads to an absurdity. If work were a state function, one might expect
it to have the form
113
114 CHAPTER 4. MATHEMATICAL FOUNDATIONS OF THERMODYNAMICS
∂w ∂w
δw = dv + dP. (4.4)
∂v P ∂P v
∂w
= P, (4.5)
∂v P
∂w
= 0. (4.6)
∂P v
df (P )
0 = v+ , (4.9)
dP
df (P )
= −v. (4.10)
dP
Equation (4.10) cannot be: a function of P only cannot be a function of v. So, w cannot be a state
property:
w 6= w(P, v). (4.11)
Now, if the differential in Eq. (4.12), when set to a differential dy, can be integrated to form
the function
y = y(x1 , x2 , . . . , xN ), (4.13)
Now, if the algebraic definition of Eq. (4.13) holds, what amounts to the definition of the
partial derivative gives the parallel result that
∂y ∂y ∂y
dy = dx1 + dx2 + · · · + dxN . (4.15)
∂x1 xj ,j6=1 ∂x2 xj ,j6=2 ∂xN xj ,j6=N
Now, combining Eqs. (4.14) and (4.15) to eliminate dy, one gets
∂y ∂y ∂y
ψ1 dx1 + ψ2 dx2 + · · · + ψN dxN = dx1 + dx2 + · · · + dxN .
∂x1 xj ,j6=1 ∂x2 xj ,j6=2 ∂xN xj ,j6=N
(4.16)
∂y ∂y ∂y
ψ1 = , ψ2 = ,..., ψN = . (4.18)
∂x1 xj ,j6=1 ∂x2 xj ,j6=2 ∂xN xj ,j6=N
So when dy is exact, one says that each of the ψi and xi are conjugate to each other.
From here on out, for notational ease, the j 6= 1, j 6= 2, . . . , j 6= N will be ignored in
the notation for the partial derivatives. It becomes especially confusing for higher order
derivatives, and is fairly obvious for all derivatives.
If y and all its derivatives are continuous and differentiable, then one has for all i =
1, . . . , N, and k = 1, . . . , N that
∂2y ∂2y
= . (4.19)
∂xk ∂xi ∂xi ∂xk
Now from Eq. (4.18), one has
∂y ∂y
ψk = , ψl = . (4.20)
∂xk xj ∂xl xj
Taking the partial of the first of Eq. (4.20) with respect to xl and the second with respect
to xk , one gets
∂ψk ∂2y ∂ψl ∂2y
= , = . (4.21)
∂xl xj ∂xl ∂xk ∂xk xj ∂xk ∂xl
Because by Eq. (4.19) the order of the mixed second partials does not matter, one deduces
from Eq. (4.21) that
∂ψk ∂ψl
= . (4.22)
∂xl xj ∂xk xj
This is a necessary and sufficient condition for the exact-ness of Eq. (4.12). It is a gen-
eralization of what can be found in most introductory calculus texts for functions of two
variables.
For the Gibbs equation, (4.1), du = −P dv + T ds, one has
y → u, x1 → v, x2 → s, ψ1 → −P ψ2 → T. (4.23)
and one expects the natural, or canonical form of
u = u(v, s). (4.24)
Here, −P is conjugate to v, and T is conjugate to s. Application of the general form of
Eq. (4.22) to the Gibbs equation (4.1) gives then
∂T ∂P
=− . (4.25)
∂v s ∂s v
∂u ∂u
−P = , T = . (4.26)
∂v s ∂s v
PN
If the general differential dy =i=1 ψi dxi is exact, one also can show
R B PN
• The path integral yB − yA = A i=1 ψi dxi is independent of the path of the integral.
• The integral around a closed contour is zero:
I I X N
dy = ψi dxi = 0. (4.27)
i=1
• The function y can only be determined to within an additive constant. That is, there
is no absolute value of y; physical significance is only ascribed to differences in y. In
fact now, other means, extraneous to this analysis, can be used to provide absolute
values of key thermodynamic variables. This will be important especially for flows
with reaction.
Example 4.2
Show the heat transfer q is not a state function. Assume all processes are fully reversible.
So
∂u ∂u
δq = dv + dT + P dv, (4.32)
∂v T ∂T v
∂u ∂u
= + P dv + dT. (4.33)
∂v T ∂T v
| {z } | {z }
≡M ≡N
= M dv + N dT. (4.34)
Now by Eq. (4.22), for δq to be exact, one must have
∂M ∂N
= . (4.35)
∂T v ∂v T
This reduces to
∂2u ∂P ∂2u
+ = . (4.36)
∂T ∂v ∂T v ∂v∂T
This can only be true if ∂P/∂T |v = 0. But this is not the case; consider an ideal gas for which
∂P/∂T |v = R/v. So δq is not exact.
Example 4.3
Show conditions for ds to be exact in the Gibbs equation.
du = T ds − P dv, (4.37)
du P
ds = + dv, (4.38)
T T
1 ∂u ∂u P
= dv + dT + dv, (4.39)
T ∂v T ∂T v T
1 ∂u P 1 ∂u
= + dv + dT. (4.40)
T ∂v T T T ∂T v
| {z } | {z }
≡M ≡N
Because x and z are independent, so are dx and dz, and the coefficients on each in Eq. (4.52)
must be zero. Therefore from the coefficient on dx in Eq. (4.52)
∂x ∂y
− 1 = 0, (4.53)
∂y z ∂x z
∂x ∂y
= 1. (4.54)
∂y z ∂x z
So
∂x 1
= ∂y
, (4.55)
∂y z ∂x z
∂x ∂y ∂x
+ = 0, (4.56)
∂y z ∂z x ∂z y
∂x ∂x ∂y
= − ,. (4.57)
∂z y ∂y z ∂z x
So
∂x ∂y ∂z
= −1. (4.58)
∂z y ∂x z ∂y x
If one now divides Eq. (4.48) by a fourth differential, dw, one gets
dx ∂x dy ∂x dz
= + . (4.59)
dw ∂y z dw ∂z y dw
∂x ∂x ∂y
= , (4.60)
∂w z ∂y z ∂w z
∂x
∂w z ∂x
∂y
= , (4.61)
∂w z
∂y z
∂x ∂w ∂x
= . (4.62)
∂w z ∂y z ∂y z
∂x ∂x
dx = dy + dw. (4.63)
∂y w ∂w y
∂x ∂x ∂x ∂w
= + . (4.64)
∂y z ∂y w ∂w y ∂y z
Example 4.4
Apply Eq. (4.64) to a standard P − v − T system and let
∂x ∂T
= . (4.65)
∂y z ∂v s
Now by definition
∂u
cv = , (4.67)
∂T v
so
∂T 1
= . (4.68)
∂u v cv
Now by Eq. (4.26), one has ∂u/∂v|s = −P , so one gets
∂T ∂T P
= − . (4.69)
∂v s ∂v u cv
For an ideal gas, u = u(T ). Inverting, one gets T = T (u), and so ∂T /∂v|u = 0, thus
∂T P
=− . (4.70)
∂v s cv
For an isentropic process in an ideal gas, one gets
dT P RT
= − =− , (4.71)
dv cv cv v
dT R dv
= − , (4.72)
T cv v
dv
= −(k − 1) , (4.73)
v
T vo
ln = (k − 1) ln , (4.74)
To v
T v k−1
o
= . (4.75)
To v
However, v and s may not be convenient for a particular problem. There may be other
combinations of variables whose canonical form gives a more convenient set of independent
variables for a particular problem. An example is the enthalpy from Eq. (1.20):
h ≡ u + P v. (4.77)
dh = du + P dv + v dP. (4.78)
dh {z − v dP} = −P dv + T ds.
| − P dv (4.79)
=du
So
dh = T ds + v dP. (4.80)
h = h(s, P ). (4.81)
This exercise can be systematized with the Legendre transformation,1 which defines a set of
second order polynomial combinations of variables. Additional and more general develop-
ment of the Legendre transformation is given in the Appendix, p. 295.
Consider again the exact differential Eq. (4.14):
τ1 = τ1 (ψ1 , x2 , x3 , . . . , xN ) = y − ψ1 x1 , (4.83)
τ2 = τ2 (x1 , ψ2 , x3 , . . . , xN ) = y − ψ2 x2 , (4.84)
..
.
τN = τN (x1 , x2 , x3 , . . . , ψN ) = y − ψN xN , (4.85)
τ1,2 = τ1,2 (ψ1 , ψ2 , x3 , . . . , xN ) = y − ψ1 x1 − ψ2 x2 , (4.86)
τ1,3 = τ1,3 (ψ1 , x2 , ψ3 , . . . , xN ) = y − ψ1 x1 − ψ3 x3 , (4.87)
..
.
N
X
τ1,...,N = τ1,...,N (ψ1 , ψ2 , ψ3 , . . . , ψN ) = y − ψi xi . (4.88)
i=1
Each τ is a new dependent variable. Each τ has the property that when it is known as a
function of its N canonical variables, the remaining N variables from the original expression
(the xi and the conjugate ψi ) can be recovered by differentiation of τ . In general this is not
true for arbitrary transformations.
Example 4.5
Let y = y(x1 , x2 , x3 ). This has the associated differential form
z = y − ψ1 x1 . (4.90)
So the original expression had three independent variables x1 , x2 , x3 , and three conjugate variables
ψ1 , ψ2 , ψ3 . Definition of the Legendre function z with canonical variables ψ1 , x2 , and x3 allowed
determination of the remaining variables x1 , ψ2 , and ψ3 in terms of the canonical variables.
For the Gibbs equation, (4.1), du = −P dv+T ds, one has y = u, two canonical variables,
x1 = v and x2 = s, and two conjugates, ψ1 = −P and ψ2 = T . Thus N = 2, and one can
expect 22 − 1 = 3 Legendre transformations. They are
It has already been shown for the enthalpy that dh = T ds + v dP , so that the canonical
variables are s and P . One then also has
∂h ∂h
dh = ds + dP, (4.99)
∂s P ∂P s
∂h ∂h
T = , v= . (4.100)
∂s P ∂P s
From Eq. (4.100), a second Maxwell relation can be deduced by differentiation of the first
with respect to P and the second with respect to s:
∂T ∂v
= . (4.101)
∂P s ∂s P
The relations for Helmholtz and Gibbs free energies each supply additional useful relations
including two new Maxwell relations. First consider the Helmholtz free energy
a = u − T s, (4.102)
da = du − T ds − s dT, (4.103)
= (−P dv + T ds) − T ds − s dT, (4.104)
= −P dv − s dT. (4.105)
So the canonical variables for a are v and T . The conjugate variables are −P and −s. Thus
∂a ∂a
da = dv + dT. (4.106)
∂v T ∂T v
So one gets
∂a ∂a
−P = , −s = . (4.107)
∂v T ∂T v
∂P ∂s
= . (4.108)
∂T v ∂v T
g = |u +{zP v} −T s, (4.109)
=h
= h − T s, (4.110)
dg = dh − T ds − s dT, (4.111)
= (T ds + v dP ) −T ds − s dT, (4.112)
| {z }
=dh
= v dP − s dT. (4.113)
So for Gibbs free energy, the canonical variables are P and T while the conjugate variables
are v and −s. One then has g = g(P, T ), which gives
∂g ∂g
dg = dP + dT. (4.114)
∂P T ∂T P
So one finds
∂g ∂g
v= , −s = . (4.115)
∂P T ∂T P
∂v ∂s
=− . (4.116)
∂T P ∂P T
Example 4.6
If we have the canonical form for h(s, P ) of
R/cP
P s
h(s, P ) = cP To exp + (ho − cP To ) , (4.117)
Po cP
and cP , To , R, Po , and ho are all constants, derive both thermal and caloric state equations P (v, T )
and u(v, T ).
du = T ds − P dv, (4.136)
∂u ∂s
= T , (4.137)
∂T v ∂T v
∂s
cv = T . (4.138)
∂T v
Likewise,
dh = T ds + v dP, (4.139)
∂h ∂s
= T , (4.140)
∂T P ∂T P
∂s
cP = T . (4.141)
∂T P
One finds further useful relations by operating on the Gibbs equation:
du = T ds − P dv, (4.142)
∂u ∂s
= T − P, (4.143)
∂v T ∂v T
∂P
= T − P. (4.144)
∂T v
So one can then say
Consequently, u is not a function of v for an ideal gas, so u = u(T ) alone. Because Eq. (4.77),
h = u + P v, for an ideal gas reduces to h = u + RT
Now return to general equations of state. With s = s(T, v) or s = s(T, P ), one gets
∂s ∂s
ds = dT + dv, (4.151)
∂T v ∂v T
∂s ∂s
ds = dT + dP. (4.152)
∂T P ∂P T
cv ∂P
ds = dT + dv, (4.153)
T ∂T v
cP ∂v
ds = dT − dP. (4.154)
T ∂T P
cv − cP ∂P ∂v
0 = dT + dv + dP, (4.155)
T ∂T v ∂T P
∂P ∂v
(cP − cv ) dT = T dv + T dP. (4.156)
∂T v ∂T P
Now divide both sides by dT and hold either P or v constant. In either case, one gets
∂P ∂v
cP − cv = T . (4.157)
∂T v ∂T P
Now because T > 0, (∂v/∂T |P )2 > 0, and for all known materials ∂P/∂v|T < 0, we must
have
cP > cv . (4.159)
Example 4.7
For an ideal gas find cP − cv .
∂P R ∂v R
= , = . (4.160)
∂T v v ∂T P P
RR
cP − cv = T , (4.161)
vP
R2
= T , (4.162)
RT
= R. (4.163)
cP (T ) − cv (T ) = R. (4.164)
For the ratio of specific heats for a general material, one can use Eqs. (4.138) and (4.141)
to get
∂s
cP T ∂T P
k= = , then apply Eq. (4.55) to get
∂s
(4.165)
cv T ∂T v
∂s ∂T
= , then apply Eq. (4.57) to get (4.166)
∂T P ∂s v
∂s ∂P ∂T ∂v
= − − , (4.167)
∂P T ∂T s ∂v s ∂s T
∂v ∂s ∂P ∂T
= . (4.168)
∂s T ∂P T ∂T s ∂v s
∂v ∂P
k= . (4.169)
∂P T ∂v s
The first term can be obtained from P − v − T data. The second term is related to the
isentropic sound speed of the material, which is also a measurable quantity.
Example 4.8
For a calorically perfect ideal gas with gas constant R and specific heat at constant volume cv ,
find expressions for the thermodynamic variable s and thermodynamic potentials u, h, a, and g, as
functions of T and P .
du = T ds − P dv, (4.170)
T ds = du + P dv, (4.171)
= cv dT + P dv, (4.172)
dT P
ds = cv + dv, (4.173)
T T
dT dv
= cv +R , (4.174)
Z Z T vZ
dT dv
ds = cv + R , (4.175)
T v
T v
s − s0 = cv ln + R ln , (4.176)
T0 v0
s − s0 T R RT /P
= ln + ln , (4.177)
cv T0 cv RT0 /P0
R/cv
T T P0
= ln + ln , (4.178)
T0 T0 P
1+R/cv R/cv
T P0
= ln + ln , (4.179)
T0 P
1+(cP −cv )/cv (cP −cv )/cv
T P0
= ln + ln , (4.180)
T0 P
k k−1
T P0
= ln + ln . (4.181)
T0 P
So
k k−1
T P0
s = s0 + cv ln + cv ln . (4.182)
T0 P
u = uo + cv (T − To ). (4.183)
h = u + P v, (4.184)
= u + RT, (4.185)
= uo + cv (T − To ) + RT, (4.186)
= uo + cv (T − To ) + RT + RTo − RTo , (4.187)
= uo + RTo +cv (T − To ) + R(T − To ), (4.188)
| {z }
=ho
= ho + (cv + R)(T − To ). (4.189)
| {z }
=cP
So
h = ho + cP (T − To ). (4.190)
Example 4.9
Find a general expression for u(T, v) if
RT a
P (T, v) = − 2. (4.194)
v−b v
recalling that
∂u ∂u ∂P
= cv , =T − P. (4.196)
∂T v ∂v T ∂T v
So we have
∂u a
= , (4.200)
∂v T v2
a
u(T, v) = − + f (T ). (4.201)
v
Here f (T ) is some as-of-yet arbitrary function of T . To evaluate f (T ), take the derivative with respect
to T holding v constant:
∂u df
= = cv . (4.202)
∂T v dT
We also find
Z T
1 1
h = u + P v = uo + cv (T̂ ) dT̂ + a − + P v, (4.206)
To vo v
Z T
1 1 RT v a
h(T, v) = uo + cv (T̂ ) dT̂ + a − + − . (4.207)
To vo v v−b v
Example 4.10
A van der Waals gas with
J
R = 200 , (4.208)
kg K
Pa m6
a = 150 , (4.209)
kg2
m3
b = 0.001 , (4.210)
kg
J J
cv = 350 + 0.2 (T − (300 K)), (4.211)
kg K kg K2
begins at T1 = 300 K, P1 = 1 × 105 Pa. It is isothermally compressed to state 2 where P2 = 1 × 106 Pa.
It is then isochorically heated to state 3 where T3 = 1000 K. Find 1 w3 , 1 q3 , and s3 − s1 . Assume the
surroundings are at 1000 K.
Recall
RT a
P = − , (4.212)
v − b v2
so at state 1
200 kgJ K (300 K) 150 Pa m
6
5
kg2
10 Pa = − . (4.213)
v1 − 0.001 m
3 v12
kg
kg kg2 kg3
−0.15 + 150 3 v1 − 60100 6 v1 + 100000 9 v13 = 0.
2
(4.214)
m m m
m3
v1 = 0.598 , physical, (4.215)
kg
m3
v1 = 0.00125 − 0.0097i not physical, (4.216)
kg
m3
v1 = 0.00125 + 0.0097i not physical. (4.217)
kg
200 kgJ K (300 K) 150 Pa m
6
kg2
106 Pa = − . (4.218)
v2 − 0.001 m
3 v22
kg
The physical solution is v2 = 0.0585 m3 /kg. Now at state 3 it is known that v3 = v2 and T3 = 1000 K.
Determine P3 :
J 6
200
(1000 K)
kg K 150 Pakgm2
P3 = − 2 , (4.219)
m3 m3
0.0585 kg − 0.001 kg 0.0585 m
3
kg
R2 R3 R2
Now 1 w3 = 1 w2 + 2 w3 = 1
P dv + 2 P dv = 1 P dv because 2 → 3 is at constant volume. So
Z v2
RT a
1 w3 = − dv, (4.222)
v1 v − b v2
Z v2 Z v2
dv dv
= RT1 −a 2
, (4.223)
v1 v − b v1 v
v2 − b 1 1
= RT1 ln +a − , (4.224)
v1 − b v2 v1
0.0585 m
3
− 0.001 m
3
J kg kg
= 200 (300 K) ln
kg K m3
0.598 kg − 0.001 kg m3
!
Pa m6 1 1
+ 150 3 − 3 , (4.225)
kg2 0.0585 m kg 0.598 m kg
J J
= −140408 + 2313 , (4.226)
kg kg
J
= −138095 , (4.227)
kg
kJ
= −138.095 . (4.228)
kg
u3 − u1 = 1 q3 − 1 w3 , (4.238)
1 q3 = u3 − u1 + 1 w3 , (4.239)
kJ kJ
1 q3 = 292 − 138 , (4.240)
kg kg
kJ
1 q3 = 154 . (4.241)
kg
T ds = du + P dv, (4.242)
1 P
ds = du + dv, (4.243)
T T
1 a P
= cv (T ) dT + 2 dv + dv, (4.244)
T v T
1 a 1 RT a
= cv (T ) dT + 2 dv + − 2 dv, (4.245)
T v T v−b v
cv (T ) R
= dT + dv, (4.246)
T v−b
Z T3
cv (T ) v3 − b
s3 − s1 = dT + R ln , (4.247)
T1 T v1 − b
Z 1000 290 kgJ K
= + 0.2
J dT + R ln v3 − b , (4.248)
T kg K 2 v1 − b
300
J 1000 K J
= 290 ln + 0.2 ((1000 K) − (300 K))
kg K 300 K kg K2
0.0585 m3
− 0.001 m3
J kg kg
+ 200 ln , (4.249)
kg K m3
0.598 kg − 0.001 kg m3
J J J
= 349 + 140 − 468 , (4.250)
kg K kg K kg K
J
= 21 , (4.251)
kg K
kJ
= 0.021 . (4.252)
kg K
Is the second law satisfied for each portion of the process? First look at 1 → 2:
u2 − u1 = − 1 w2 ,
1 q2 (4.253)
1 q2 = u2 − u1 + 1 w2 , (4.254)
Z T2 !
1 1 v2 − b 1 1
= cv (T ) dT + a − + RT1 ln +a − . (4.255)
T1 v1 v2 v1 − b v2 v1
v2 − b
1 q2 = RT1 ln , (4.256)
v1 − b
0.0585 m3
− 0.001 m3
J kg kg
= 200 (300 K) ln , (4.257)
kg K 0.598 m3
− 0.001 m3
kg kg
J
= −140408 . (4.258)
kg
v2 − b
s2 − s1 = R ln , (4.259)
v1 − b
0.0585 m3
− 0.001 m3
J kg kg
= 200 ln , (4.260)
kg K 0.598 m 3
− 0.001 m 3
kg kg
J
= −468.0 . (4.261)
kg K
1 q2
s2 − s1 ≥ ? (4.262)
T
J
J −140408 kg
−468.0 ≥ , (4.263)
kg K 300 K
J J
−468.0 ≥ −468.0 ok. (4.264)
kg K kg K
Next look at 2 → 3:
2 q3 = u3 − u2 + 2 w3 , (4.265)
Z Z v3
T3 1 1
=
c v (T ) dT + a − + P dv , (4.266)
T2 v2 v3 v2
| {z } | {z }
=0 =0
Z T3
= cv (T ) dT, (4.267)
T2
Z 1000 K
J J
= 290 + 0.2 T dT, (4.268)
300 K kg K kg K2
J
= 294000 . (4.269)
kg
RT a
P = − . (4.277)
v − b v(v + b)T 1/2
It is modestly more accurate than the van der Waals equation in predicting material behavior.
Example 4.11
For the case in which b = 0, find an expression for u(T, v) consistent with the Redlich-Kwong state
equation.
RT a
P = − 2 1/2 . (4.278)
v v T
∂u ∂P
= T − P, (4.279)
∂v T ∂T v
R a RT a
= T + 2 3/2 − − 2 1/2 , (4.280)
v 2v T v v T
3a
= . (4.281)
2v 2 T 1/2
Integrating, we find
3a
u(T, v) = − + f (T ). (4.282)
2vT 1/2
Here f (T ) is a yet-to-be-specified function of temperature only. Now the specific heat is found by the
temperature derivative of u:
∂u 3a df
cv (T, v) = = 3/2
+ . (4.283)
∂T v 4vT dT
df
≡ cvo (T ). (4.284)
dT
Integrating, then one gets
Z T
f (T ) = C + cvo (T̂ ) dT̂ . (4.285)
To
1/2
Let us take C = uo +3a/2/vo/To . Thus we arrive at the following expressions for cv (T, v) and u(T, v):
3a
cv (T, v) = cvo (T ) + , (4.286)
4vT 3/2
Z !
T
3a 1 1
u(T, v) = uo + cvo (T̂ ) dT̂ + 1/2
− . (4.287)
To 2 vo To vT 1/2
For an ideal gas, Z = 1. For substances with a simple molecular structure, Z can be
tabulated as functions of the so-called reduced pressure Pr and reduced temperature Tr . Tr
and Pr are dimensionless variables found by scaling their dimensional counterparts by the
specific material’s temperature and pressure at the critical point, Tc and Pc :
T P
Tr ≡ , Pr ≡ . (4.289)
Tc Pc
Often charts are available which give predictions of all reduced thermodynamic properties.
These are most useful to capture the non-ideal gas behavior of materials for which tables are
not available.
U = U(V, S, ni ). (4.290)
The extensive version of the Gibbs law in which all of the ni are held constant is
dU = −P dV + T dS. (4.291)
Thus
∂U ∂U
= −P, = T. (4.292)
∂V S,ni ∂S V,ni
In general, because U = U(S, V, ni ), one should expect, for systems in which the ni are
allowed to change that
XN
∂U ∂U ∂U
dU = dV + dS + dni . (4.293)
∂V S,ni ∂S V,ni i=1
∂ni S,V,nj
∂U
µi ≡ , (4.294)
∂ni S,V,nj
N
X
dU = −P dV + T dS + µi dni . (4.295)
i=1
Obviously, by its definition, µi is on a per mole basis, so it is given the appropriate overline
notation. In Eq. (4.295), the independent variables and their conjugates are
x1 = V, ψ1 = −P, (4.296)
x2 = S, ψ2 = T, (4.297)
x3 = n1 , ψ3 = µ1 , (4.298)
x4 = n2 , ψ4 = µ2 , (4.299)
.. ..
. .
xN +2 = nN , ψN +2 = µN . (4.300)
Equation (4.295) has 2N +1 −1 Legendre functions. Three are in wide usage: the extensive
analog to those earlier found. They are
H = U + P V, (4.301)
A = U − T S, (4.302)
G = U + P V − T S. (4.303)
Differentiating each defined Legendre function, Eqs. (4.301-4.304), and combining with
Eq. (4.295), one finds
N
X
dH = T dS + V dP + µi dni , (4.305)
i=1
XN
dA = −S dT − P dV + µi dni , (4.306)
i=1
XN
dG = −S dT + V dP + µi dni , (4.307)
i=1
XN
dΩ = −P dV − S dT − ni dµi . (4.308)
i=1
Thus, canonical variables for H are H = H(S, P, ni ). One finds a similar set of relations as
∂U ∂H
T = = , (4.309)
∂S V,ni ∂S P,ni
∂U ∂A ∂Ω
P = − =− =− , (4.310)
∂V S,ni ∂V T,ni ∂V T,µi
∂H ∂G
V = = , (4.311)
∂P S,ni ∂P T,ni
∂A ∂G ∂Ω
S = − =− =− , (4.312)
∂T V,ni ∂T P,ni ∂T V,µi
∂Ω
ni = − , (4.313)
∂µi V,T,µj
∂U ∂H ∂A ∂G
µi = = = = . (4.314)
∂ni S,V,nj ∂ni S,P,nj ∂ni T,V,nj ∂ni T,P,nj
∂T ∂P
= − , (4.315)
∂V S,ni ∂S V,ni
∂T ∂V
= , (4.316)
∂P S,ni ∂S P,ni
∂P ∂S
= , (4.317)
∂T V,ni ∂V T,ni
∂V ∂S
= − , (4.318)
∂T P,ni ∂P T,ni
∂µi ∂S
= − , (4.319)
∂T P,nj ∂ni V,nj
∂µi ∂V
= , (4.320)
∂P T,nj ∂ni V,nj
∂µl ∂µk
= , (4.321)
∂nk T,P,nj ∂nl T,P,nj
∂S ∂P
= , (4.322)
∂V T,µi ∂T V,µi
∂ni ∂nk
= (4.323)
∂µk V,T,µj ,j6=k ∂µi V,T,µj ,j6=i
If m = 1, one has
f (λx1 , . . . , λxN ) = λf (x1 , . . . , xN ). (4.325)
Thermodynamic variables are examples of homogeneous functions.
Differentiate both sides of Eq. (4.327) with respect to λ, while holding P , T , and nj constant,
to get
G(T, P, n1 , n2 , . . . , nN ) =
∂G d(λn1 ) ∂G d(λn2 ) ∂G d(λnN )
+ +···+ , (4.328)
∂(λn1 ) nj ,P,T dλ ∂(λn2 ) nj ,P,T dλ ∂(λnN ) nj ,P,T dλ
∂G ∂G ∂G
= n1 + n2 + · · · + nN , (4.329)
∂(λn1 ) nj ,P,T ∂(λn2 ) nj ,P,T ∂(λnN ) nj ,P,T
∂G ∂G ∂G
G(T, P, n1, n2 , . . . , nN ) = n1 + n2 + · · · + nN ,
∂n1 nj ,P,T ∂n2 nj ,P,T ∂nN nj ,P,T
(4.330)
XN
∂G
= ni . (4.331)
i=1
∂ni nj ,P,T
Recall now the definition partial molar property, the derivative of an extensive variable with
respect to species ni holding nj , i 6= j, T , and P constant. Because the result has units per
mole, an overline superscript is utilized. The partial molar Gibbs free energy of species i, g i
is then
∂G
gi ≡ , (4.332)
∂ni nj ,P,T
so that
N
X
G= g i ni . (4.333)
i=1
Using the definition of chemical potential, Eq. (4.314), one also notes then that
N
X
G(T, P, n1 , n2 , . . . , nN ) = µi ni . (4.334)
i=1
The temperature and pressure dependence of G must lie entirely within µi (T, P, nj ), which
one notes is also allowed to be a function of nj as well. Consequently, one also sees that the
Gibbs free energy per unit mole of species i is the chemical potential of that species:
g i = µi . (4.335)
Using Eq. (4.333) to eliminate G in Eq. (4.303), one recovers an equation for the energy:
N
X
U = −P V + T S + µi ni . (4.336)
i=1
XN
∂U
U = ni (4.338)
i=1
∂ni nj ,V,S
XN
∂H
H = ni , (4.339)
i=1
∂ni nj ,P,S
XN
∂A
A = ni , (4.340)
i=1
∂ni nj ,T,V
XN
∂S
S = ni . (4.341)
i=1
∂ni nj ,U,T
These expressions do not formally involve partial molar properties because P and T are not
constant.
Take now the appropriate partial molar derivatives of G for an ideal mixture of ideal
gases to get some useful relations:
G = H − T S, (4.342)
∂G ∂H ∂S
= −T . (4.343)
∂ni T,P,nj ∂ni T,P,nj ∂ni T,P,nj
N
X
H = nk hk (T, P ), (4.344)
k=1
N
!
∂H ∂ X
= nk hk (T, P ) , (4.345)
∂ni T,P,nj ∂ni k=1
N
X ∂nk
= hk (T, P ), (4.346)
∂ni
k=1 |{z}
=δik
N
X
= δik hk (T, P ), (4.347)
k=1
= hi (T, P ). (4.348)
Here, the Kronecker delta function δki has been again used. Now for an ideal gas one further
has hi = hi (T ). The analysis is more complicated for the entropy, in which
N
X
o Pk
S = nk sT,k − R ln , (4.349)
Po
k=1
XN
o yk P
= nk sT,k − R ln , (4.350)
k=1
P o
N !!
X P nk
o
= nk sT,k − R ln − R ln PN , (4.351)
k=1
Po q=1 nq
N N
!
X P X nk
= nk soT,k − R ln −R nk ln PN , (4.352)
k=1
Po k=1 q=1 nq
N
∂S ∂ X o P
= nk sT,k − R ln
∂ni T,P,nj ∂ni k=1 Po
N
!!
∂ X nk
−R nk ln PN , (4.353)
∂ni T,P,nj q=1 nq
k=1
XN
∂nk o P
= sT,k − R ln
∂ni
k=1 |{z}
Po
=δik
N
!!
∂ X nk
−R nk ln PN , (4.354)
∂ni T,P,nj q=1 nq
k=1
N
!!
P ∂ X nk
= soT,i − R ln −R nk ln PN . (4.355)
Po ∂ni T,P,nj k=1 q=1 nq
Evaluation of the final term on the right side requires closer examination, and in fact, after
tedious but straightforward analysis, yields a simple result which can easily be verified by
direct calculation:
N
!! !
∂ X nk ni
nk ln PN = ln PN . (4.356)
∂ni T,P,nj k=1 q=1 nq q=1 nq
g i = hi − T si . (4.361)
∂g
gi = g + n . (4.369)
∂ni T,P,nj
Consider temporarily the assumption that the work and heat transfer are both reversible.
Thus, the irreversible entropy production must be associated with internal chemical reaction.
Now the first law for the entire system gives
dU = δQ − δW, (4.373)
= δQ − P dV, (4.374)
δQ = dU + P dV. (4.375)
Because the system is closed, there can be no species entering or exiting, and so there is
no change dU attributable to dni . While within the system the dni may not be 0, the net
contribution to the change in total internal energy is zero. A non-zero dni within the system
simply re-partitions a fixed amount of total energy from one species to another. Substituting
Eq. (4.375) into Eq. (4.372) to eliminate δQ, one gets
1
dS ≥ (dU + P dV ), (4.376)
T | {z }
=δQ
T dS − dU − P dV ≥ 0, (4.377)
dU − T dS + P dV ≤ 0. (4.378)
Eq. (4.378) involves properties only and need not require assumptions of reversibility for
processes in its derivation. In special cases, it reduces to simpler forms.
For processes which are isentropic and isochoric, the second law expression, Eq. (4.378),
reduces to
dU|S,V ≤ 0. (4.379)
For processes which are isoenergetic and isochoric, the second law expression, Eq. (4.378),
reduces to
dS|U,V ≥ 0. (4.380)
Now using Eq. (4.295) to eliminate dS in Eq. (4.380), one can express the second law as
N
!
1 P 1X
dU + dV − µ dni ≥ 0, (4.381)
T T T i=1 i
| {z }
=dS U,V
N
X
1
− µi dni ≥ 0. (4.382)
T i=1
| {z }
irreversible entropy production
The irreversible entropy production associated with the internal chemical reaction must be
the left side of Eq. (4.382). Often the irreversible entropy production is defined as σ, with
Now, while most standard texts focusing on equilibrium thermodynamics go to great lengths
to avoid the introduction of time, it really belongs in a discussion describing the approach
to equilibrium. One can divide Eq. (4.382) by a positive time increment dt to get
N
1 X dni
− µ ≥ 0. (4.384)
T i=1 i dt
This will hold if a model for dni /dt is employed which guarantees that the left side of
Eq. (4.385) is negative semi-definite. One will expect then for dni /dt to be related to the
chemical potentials µi .
Elimination of dU in Eq. (4.378) in favor of dH from dH = dU + P dV + V dP gives
| − P dV
dH {z − V dP} −T dS + P dV ≤ 0, (4.386)
=dU
dH − V dP − T dS ≤ 0. (4.387)
dH|P,S ≤ 0. (4.388)
For the Helmholtz and Gibbs free energies, one analogously finds
dA|T,V ≤ 0, (4.389)
dG|T,P ≤ 0. (4.390)
The expression of the second law in terms of dG is especially useful as it may be easy in an
experiment to control so that P and T are constant. This is especially true in an isobaric
phase change, in which the temperature is guaranteed to be constant as well.
Now one has
N
X
G = ni g i , (4.391)
i=1
XN
= ni µi . (4.392)
i=1
P
One also has from Eq. (4.307): dG = −S dT +V dP + N i=1 µi dni , holding T and P constant
that
N
X
dG|T,P = µi dni . (4.393)
i=1
Here the dni are associated entirely with internal chemical reactions. Substituting Eq. (4.393)
into Eq. (4.390), one gets the important version of the second law which holds that
N
X
dG|T,P = µi dni ≤ 0. (4.394)
i=1
In terms of time rates of change, one can divide Eq. (4.394) by a positive time increment
dt > 0 to get
N
X
∂G dni
= µi ≤ 0. (4.395)
∂t T,P i=1
dt
Example 4.12
Consider an equilibrium two-phase mixture of liquid and vapor H2 O at T = 100 ◦ C, x = 0.5. Use
the steam tables to check if equilibrium properties are satisfied.
In a two-phase gas liquid mixture one can expect the following reaction:
That is one mole of liquid, in the forward phase change, evaporates to form one mole of gas. In the
reverse phase change, one mole of gas condenses to form one mole of liquid.
Because T is fixed at 100 ◦ C and the material is a two phase mixture, the pressure is also fixed at a
constant. Here there are two phases at saturation; g for gas and l for liquid. Equation (4.394) reduces
to
µl dnl + µg dng ≤ 0. (4.397)
Now for the pure H2 O if a loss of moles from one phase must be compensated by the addition to
another. So one must have
dnl + dng = 0. (4.398)
Hence
dng = −dnl . (4.399)
So Eq. (4.397), using Eq. (4.399) becomes
At this stage of the analysis, most texts, grounded in equilibrium thermodynamics, assert that µl = µg ,
ignoring the fact that they could be different but dnl could be zero. That approach will not be taken
here. Instead divide Eq. (4.401) by a positive time increment, dt ≥ 0 to write the second law as
dnl
(µ − µg ) ≤ 0. (4.402)
dt l
One convenient, albeit naı̈ve, way to guarantee second law satisfaction is to let
dnl
= −κ(µl − µg ), κ ≥ 0, convenient, but naı̈ve model (4.403)
dt
Here κ is some positive semi-definite scalar rate constant which dictates the time scale of approach to
equilibrium. Note that Eq. (4.403) is just a hypothesized model. It has no experimental verification; in
fact, other more complex models exist which both agree with experiment and satisfy the second law.
For the purposes of the present argument, however, Eq. (4.403) will suffice. With this assumption, the
second law reduces to
−κ(µl − µg )2 ≤ 0, κ ≥ 0, (4.404)
which is always true.
Eq. (4.403) has three important consequences:
• Differences in chemical potential drive changes in the number of moles.
• The number of moles of liquid, nl , increases when the chemical potential of the liquid is less than that
of the gas, µl < µg . That is to say, when the liquid has a lower chemical potential than the gas, the
gas is driven towards the phase with the lower potential. Because such a phase change is isobaric and
isothermal, the Gibbs free energy is the appropriate variable to consider, and one takes µ = g. When
this is so, the Gibbs free energy of the mixture, G = nl µl + ng µg is being driven to a lower value. So
when dG = 0, the system has a minimum G.
• The system is in equilibrium when the chemical potentials of liquid and gas are equal: µl = µg .
The chemical potentials, and hence the molar specific Gibbs free energies must be the same for
each constituent of the binary mixture at the phase equilibrium. That is
gl = gg . (4.405)
Now because both the liquid and gas have the same molecular mass, one also has the mass specific
Gibbs free energies equal at phase equilibrium:
gl = gg . (4.406)
This can be verified from the steam tables, using the definition g = h − T s. From the tables
kJ kJ kJ
gl = hl − T sl = 419.02 − ((100 + 273.15) K) 1.3068 = −68.6 , (4.407)
kg kg K kg
kJ kJ kJ
gg = hg − T sg = 2676.05 − ((100 + 273.15) K) 7.3548 = −68.4 . (4.408)
kg kg K kg
The two values are essentially the same; the difference is likely due to table inaccuracies.
Let us look at this in another way that does not require a naı̈ve model for dnl /dt. Say at t = 0,
we have no moles of vapor and zero moles of liquid. Then, as a phase transition occurs, we might
instantaneously have nl moles of liquid and no − nl moles of vapor. Then the total G of the mixture is
dG
= gl − g g = 0, (4.410)
dnl
gl = gg , (4.411)
gl = gg . (4.412)
Example 4.13
This example is adopted from BS. A container has liquid water at 20 ◦ C, 100 kPa, in equilibrium
with a mixture of water vapor and dry air, also at 20 ◦ C, 100 kPa. Find the water vapor pressure and
the saturated water vapor pressure.
Now at this temperature, the tables easily show that the pressure of a saturated vapor is Psat =
2.339 kPa . From the previous example, it is known that for the water liquid and vapor in equilibrium,
one has
gliq = gvap . (4.413)
Now if both the liquid and the vapor were at the saturated state, they would be in phase equilibrium
and that would be the end of the problem. But they have slight deviations from the saturated state.
One can estimate these deviations with the standard formula
dg = −s dT + v dP. (4.414)
The tables will be used to get values at 20 ◦ C, for which one can take dT = 0. This allows the
approximation of dg ∼ v dP . So for the liquid,
Z
gliq − gf = v dP ∼ vf (P − Psat ), (4.415)
Pvap
gf + vf (P − Psat ) = gg + RT ln . (4.421)
| {z } Psat
=gliq | {z }
=gvap
The pressure is very near the saturation pressure. This justifies assumptions that for such mixtures, one
can take the pressure of the water vapor to be that at saturation if the mixture is in equilibrium. If the
pressure is higher, the pressure of the vapor becomes higher as well. Figure 4.1 shows how the pressure
of the equilibrium vapor pressure varies with total pressure. Clearly, a very high total pressure, on the
order of 1 GPa is needed to induce the vapor pressure to deviate significantly from the saturation value.
5
10
Pvap(Pa)
4
10
3
10 5 6 7 8 9
10 10 10 10 10
P (Pa)
Figure 4.1: Pressure of water vapor as a function of total pressure for example problem.
Example 4.14
At high temperatures, collisions between diatomic nitrogen molecules induce the production of
monatomic nitrogen molecules. The chemical reaction can be described by the model
N2 + N2 ⇌ 2N + N2 . (4.426)
Here one of the N2 molecules behaves as an inert third body. An N2 molecule has to collide with
something, to induce the reaction. Some authors leave out the third body and write instead N2 ⇌ 2N,
but this does not reflect the true physics as well. The inert third body is especially important when
the time scales of reaction are considered. It plays no role in equilibrium chemistry.
Consider 1 kmole of N2 and 0 kmole of N at a pressure of 100 kPa and a temperature of 6000 K.
Using the ideal gas tables, find the equilibrium concentrations of N and N2 if the equilibration process
is isothermal and isochoric.
nN2 RT
P1 = , (4.427)
V
nN2 RT
V = , (4.428)
P1
kJ
(1 kmole) 8.314 kmole K (6000 K)
= , (4.429)
100 kPa
= 498.84 m3 . (4.430)
Initially, the mixture is all N2 , so its partial pressure is the total pressure, and the initial partial pressure
of N is 0.
Now every time an N2 molecule reacts and thus undergo a negative change, 2 N molecules are
created and thus undergo a positive change, so
1
−dnN2 = dnN . (4.431)
2
This can be parameterized by a reaction progress variable ζ, also called the degree of reaction, defined
such that
dζ = −dnN2 , (4.432)
1
= dnN . (4.433)
2
As an aside, one can integrate this, taking ζ = 0 at the initial state to get
Now for the reaction, one must have, for second law satisfaction, that
which is always true. Obviously, the reaction ceases when dζ/dt = 0, which holds only when
Away from equilibrium, for the reaction to go forward, one must expect dζ/dt > 0, and then one
must have
The chemical potentials are the molar specific Gibbs free energies; thus, for the reaction to go forward,
one must have
2g N ≤ g N2 . (4.448)
Substituting using the definitions of Gibbs free energy, one gets
2 hN − T sN ≤ h N2 − T s N2 , (4.449)
y N P y N2 P
2 hN − T soT,N − R ln ≤ hN2 − T soT,N2 − R ln , (4.450)
Po Po
yN P y N2 P
2 hN − T soT,N − hN2 − T soT,N2 ≤ −2RT ln + RT ln , (4.451)
Po Po
yN P y N2 P
−2 hN − T soT,N + hN2 − T soT,N2 ≥ 2RT ln − RT ln , (4.452)
Po Po
2 2
yN P Po
≥ RT ln , (4.453)
Po2 P yN2
2
yN P
≥ RT ln . (4.454)
yN2 Po
At the initial state, one has yN = 0, so the right hand side approaches −∞, and the inequality holds.
At equilibrium, one has equality.
2
yN P
−2 hN − T soT,N + hN2 − T soT,N2 = RT ln . (4.455)
yN2 Po
n2N RT
= , (4.460)
nN2 Po V
2
(2 ( nN2 |t=0 − nN2 )) RT
= , (4.461)
n N2 Po V
(2 (1 kmole − nN2 ))2 (8.314)(6000)
= .(4.462)
n N2 (100)(498.84)
So
1Q2 = U2 − U1 , (4.478)
5 5
= 2.60966 × 10 kJ − 1.555964 × 10 kJ, (4.479)
5
= 1.05002 × 10 kJ. (4.480)
Heat needed to be added to keep the system at the constant temperature. This is because the nitrogen
dissociation process is endothermic.
One can check for second law satisfaction in two ways. Fundamentally, one can demand that
Eq. (4.372), dS ≥ δQ/T , be satisfied for the process, giving
Z 2
δQ
S2 − S1 ≥ . (4.481)
1 T
For this isothermal process, this reduces to
1Q2
S2 − S1 ≥ , (4.482)
T
(nN2 sN2 + nN sN )|2
1Q2
− (nN2 sN2 + nN sN )|1 ≥ , (4.483)
T
y N2 P yN P
nN2 soT,N2 − R ln + nN soT,N − R ln
Po Po
2
y N P y N P 1Q2
− nN2 soT,N2 − R ln 2
+ nN soT,N − R ln ≥ , (4.484)
Po Po T
1
P N P N
nN2 soT,N2 − R ln 2
+ nN soT,N − R ln
Po Po
2
P N P N 1Q2
− nN2 soT,N2 − R ln 2
+ nN soT,N − R ln ≥ , (4.485)
Po Po 1 T
n N2 RT n N RT
nN2 soT,N2 − R ln + nN soT,N − R ln
Po V Po V
2
nN2 RT nN RT 1Q2
− nN2 soT,N2 − R ln + nN soT,N − R ln ≥ . (4.486)
Po V |{z} Po V T
=0 1
Now at the initial state nN = 0 kmole, and RT /Po /V has a constant value of
kJ
RT 8.314 kmole K (6000 K) 1
= 3
=1 , (4.487)
Po V (100 kPa)(498.84 m ) kmole
so Eq. (4.486) reduces to
nN2 n
N
nN2 soT,N2 − R ln + nN soT,N − R ln
1 kmole 1 kmole 2
n
o N2 1Q2
− nN2 sT,N2 − R ln ≥ ,
1 kmole 1 T
(4.488)
((0.888143) (292.984 − 8.314 ln (0.88143)) + (0.223714) (216.926 − 8.314 ln (0.223714)))|2
105002
− ((1) (292.984 − 8.314 ln (1)))|1 ≥ ,
6000
(4.489)
kJ kJ
19.4181 ≥ 17.5004 . (4.490)
K K
Indeed, the second law is satisfied. Moreover the irreversible entropy production of the chemical reaction
is 19.4181 − 17.5004 = +1.91772 kJ/K.
For the isochoric, isothermal process, it is also appropriate to use Eq. (4.389), dA|T,V ≤ 0, to
check for second law satisfaction. This turns out to give an identical result. Because by Eq. (4.302),
A = U − T S, A2 − A1 = (U2 − T2 S2 ) − (U1 − T1 S1 ). Because the process is isothermal, A2 − A1 =
U2 −U1 −T (S2 −S1 ). For A2 −A1 ≤ 0, one must demand U2 −U1 −T (S2 −S1 ) ≤ 0, or U2 −U1 ≤ T (S2 −S1 ),
or S2 − S1 ≥ (U2 − U1 )/T . Because 1Q2 = U2 − U1 for this isochoric process, one then recovers
S2 − S1 ≥ 1Q2 /T.
Example 4.15
Consider the same reaction
N2 + N2 ⇌ 2N + N2 . (4.491)
for an isobaric and isothermal process. That is, consider 1 kmole of N2 and 0 kmole of N at a pressure
of 100 kPa and a temperature of 6000 K. Using the ideal gas tables, find the equilibrium concentrations
of N and N2 if the equilibration process is isothermal and isobaric.
V1 = 498.84 m3 . (4.492)
The volume will change in this isobaric process. Initially, the mixture is all N2 , so its partial pressure
is the total pressure, and the initial partial pressure of N is 0.
A few other key results are identical to the previous example:
and
2g N ≤ g N2 . (4.494)
Substituting using the definitions of Gibbs free energy, one gets
2 hN − T sN ≤ h N2 − T s N2 , (4.495)
yN P y N2 P
2 hN − T soT,N − R ln ≤ hN2 − T soT,N2 − R ln , (4.496)
Po Po
yN P y N2 P
2 hN − T soT,N − hN2 − T soT,N2 ≤ −2RT ln + RT ln , (4.497)
Po Po
yN P y N2 P
−2 hN − T soT,N + hN2 − T soT,N2 ≥ 2RT ln − RT ln , (4.498)
Po Po
2 2
yN P Po
≥ RT ln . (4.499)
Po2 P yN2
n2N
= , (4.506)
nN2 (nN + nN2 )
2
(2 ( nN2 |t=0 − nN2 ))
= , (4.507)
nN2 (2 ( nN2 |t=0 − nN2 ) + nN2 )
2
(2 (1 kmole − nN2 ))
= . (4.508)
nN2 (2 (1 kmole − nN2 ) + nN2 )
This is a quadratic equation for nN2 . It has two roots
nN2 = 0.882147 kmole physical; nN2 = 1.11785 kmole, non-physical. (4.509)
The second root generates more N2 than at the start, and also yields non-physically negative nN =
−0.235706 kmole. So at equilibrium, the physical root is
nN = 2(1 − nN2 ) = 2(1 − 0.882147) = 0.235706 kmole. (4.510)
Again, the diatomic species is preferred. As the temperature is raised, one could show that the
monatomic species would come to dominate.
The volume at equilibrium is
(nN2 + nN )RT
V2 = , (4.511)
P
kJ
(0.882147 kmole + 0.235706 kmole) 8.314 kmole K (6000 K)
= , (4.512)
100 kPa
= 557.630 m3 . (4.513)
The volume has increased because there are more molecules with the pressure and temperature being
equal.
The molar concentrations ρi at equilibrium, are
Here νi represents the net number of moles of species i generated in the forward reaction.
This negation of the term on the left side of Eq (4.521) is sometimes defined as the chemical
affinity, α:
N
X
α≡− µi νi . (4.522)
i=1
Figure 4.2: Equilibrium composition of air at low density and various temperatures. Figure
from W. E. Moeckel and K. C. Weston, 1958, Composition and thermodynamic properties
of air in chemical equilibrium, NACA Technical Note 4265.
dominant in air. Both of these major components begin to dissociate at higher temperatures.
For T > 6000 K, we no longer find diatomic N2 and O2 , but instead find their monatomic
components N and O. At higher temperatures still, the molecule loses electrons, and positive
ions remain.
This chapter will further develop notions associated with the thermodynamics of chemical
reactions. The focus will be on single chemical reactions.
MT = MT · φ, or M = φT · M. (5.2)
Here M is the vector of length N containing the molecular masses, M is the vector of length
L containing the elemental masses, and φ is the matrix of dimension L × N containing the
number of moles of each element in each species. We can call it the species-element matrix.
Generally, φ is full rank. If N > L, φ has rank L. If N < L, φ has rank N. In any problem
involving an actual chemical reaction, one will find N ≥ L, and in most cases N > L. In
163
164 CHAPTER 5. THERMOCHEMISTRY OF A SINGLE REACTION
isolated problems not involving a reaction, one may have N < L. In any case, M lies in the
column space of φT , which is the row space of φ.
Example 5.1
Find the molecular mass of H2 O.
Here, one has two elements H and O, so L = 2, and one species, so N = 1; thus, in this isolated
problem, N < L. Take i = 1 for species H2 O. Take l = 1 for element H. Take l = 2 for element O.
From the periodic table, one gets M1 = 1 kg/kmole for H, M2 = 16 kg/kmole for O. For element 1,
there are 2 atoms, so φ11 = 2. For element 2, there is 1 atom so φ21 = 1. So the molecular mass of
species 1, H2 O is
φ11
M1 = M1 M2 , (5.3)
φ21
= M1 φ11 + M2 φ21 , (5.4)
kg kg
= 1 (2) + 16 (1), (5.5)
kmole kmole
kg
= 18 . (5.6)
kmole
Example 5.2
Find the molecular masses of the two species C8 H18 and CO2 .
Here, for practice, the vector matrix notation is exercised. In a certain sense this is overkill, but
it is useful to be able to understand a general notation. One has N = 2 species, and takes i = 1 for
C8 H18 and i = 2 for CO2 . One also has L = 3 elements and takes l = 1 for C, l = 2 for H, and l = 3
for O. Now for each element, one has M1 = 12 kg/kmole, M2 = 1 kg/kmole, M3 = 16 kg/kmole. The
molecular masses are then given by
φ11 φ12
M1 M2 = M1 M2 M3 φ21 φ22 , (5.7)
φ31 φ32
8 1
= 12 1 16 18 0 , (5.8)
0 2
= 114 44 . (5.9)
That is for C8 H18 , one has molecular mass M1 = 114 kg/kmole. For CO2 , one has molecular mass
M2 = 44 kg/kmole.
5.2 Stoichiometry
5.2.1 General development
Stoichiometry represents a mass balance on each element in a chemical reaction. For example,
in the simple global reaction
2H2 + O2 ⇌ 2H2 O, (5.10)
one has 4 H atoms in both the reactant and product sides and 2 O atoms in both the reactant
and product sides. In this section stoichiometry will be systematized.
Consider now a general reaction with N species. This reaction can be represented by
N
X N
X
νi′ χi ⇌ νi′′ χi . (5.11)
i=1 i=1
Here χi is the ith chemical species, νi′ is the forward stoichiometric coefficient of the ith
reaction, and νi′′ is the reverse stoichiometric coefficient of the ith reaction. Both νi′ and νi′′
are to be interpreted as pure dimensionless numbers.
In Equation (5.10), one has N = 3 species. One might take χ1 = H2 , χ2 = O2 , and
χ3 = H2 O. The reaction is written in more general form as
(2)χ1 + (1)χ2 + (0)χ3 ⇌ (0)χ1 + (0)χ2 + (2)χ3 , (5.12)
(2)H2 + (1)O2 + (0)H2 O ⇌ (0)H2 + (0)O2 + (2)H2 O. (5.13)
Here, one has
ν1′ = 2, ν1′′ = 0, (5.14)
ν2′ = 1, ν2′′ = 0, (5.15)
ν3′ = 0, ν3′′ = 2. (5.16)
It is common and useful to define another pure dimensionless number, the net stoichio-
metric coefficients for species i, νi . Here νi represents the net production of number if the
reaction goes forward. It is given by
νi = νi′′ − νi′ . (5.17)
For the reaction 2H2 + O2 ⇌ 2H2 O, one has
ν1 = ν1′′ − ν1′ = 0 − 2 = −2, (5.18)
ν2 = ν2′′ − ν2′ = 0 − 1 = −1, (5.19)
ν3 = ν3′′ − ν3′ = 2 − 0 = 2. (5.20)
With these definitions, it is possible to summarize a chemical reaction as
N
X
νi χi = 0. (5.21)
i=1
It remains to enforce a stoichiometric balance. This is achieved if, for each element, l =
1, . . . , L, one has the following equality:
N
X
φli νi = 0, l = 1, . . . , L. (5.24)
i=1
That is to say, for each element, the sum of the product of the net species production and
the number of elements in the species must be zero. In vector notation, this becomes
φ · ν = 0. (5.25)
One may recall from linear algebra that this demands that ν lie in the right null space of φ.
Example 5.3
Show stoichiometric balance is achieved for −2H2 − O2 + 2H2 O = 0.
Here again, the number of elements L = 2, and one can take l = 1 for H and l = 2 for O. Also
the number of species N = 3, and one takes i = 1 for H2 , i = 2 for O2 , and i = 3 for H2 O. Then for
element 1, H, in species 1, H2 , one has
φ11 = 2, H in H2 . (5.26)
φ12 = 0, H in O2 , (5.27)
φ13 = 2, H in H2 O, (5.28)
φ21 = 0, O in H2 , (5.29)
φ22 = 2, O in O2 , (5.30)
φ23 = 1, O in H2 O. (5.31)
PN
In matrix form then, i=1 φli νi = 0 gives
ν
2 0 2 1 0
ν2 = . (5.32)
0 2 1 0
ν3
This is two equations in three unknowns. Thus, it is formally under-constrained. Certainly the trivial
solution ν1 = ν2 = ν3 = 0 will satisfy, but one seeks non-trivial solutions. Assume ν3 has a known value
ν3 = ξ. Then, the system reduces to
2 0 ν1 −2ξ
= . (5.33)
0 2 ν2 −ξ
The inversion here is easy, and one finds ν1 = −ξ, ν2 = −ξ/2. Or in vector form,
ν1 −ξ −1
ν2 = − 1 ξ = ξ − 1 , ξ ∈ R1 . (5.34)
2 2
ν3 ξ 1
Again, this amounts to saying the solution vector (ν1 , ν2 , ν3 )T lies in the right null space of the coefficient
matrix φli . Here ξ is any real scalar. If one takes ξ = 2, one gets
ν1 −2
ν2 = −1 , (5.35)
ν3 2
In summary, the stoichiometric coefficients are non-unique but partially constrained by mass conserva-
tion for each element. Which set is chosen is to some extent arbitrary, and often based on traditional
conventions from chemistry. But others are equally valid.
There is a small issue with units here, which will be seen to be difficult to reconcile.
In practice, it will have little to no importance. In the previous example, one might be
tempted to ascribe units of kmoles to νi . Later, it will be seen that in classical reaction
kinetics, νi is best interpreted as a pure dimensionless number, consistent with the definition
of this section. So in the context of the previous example, one would then take ξ to be
dimensionless as well, which is perfectly acceptable for the example. In later problems, it
will be more useful to give ξ the units of kmoles. Multiplication of ξ by any scalar, e.g.
kmole/(6.02 × 1026 ), still yields an acceptable result.
Example 5.4
Balance an equation for hypothesized ethane combustion
ν1 C2 H6 + ν2 O2 + ν3 CO2 + ν4 H2 O = 0. (5.40)
φ11 = 2, C in C2 H6 , (5.41)
φ12 = 0, C in O2 , (5.42)
φ13 = 1, C in CO2 , (5.43)
φ14 = 0, C in H2 O, (5.44)
φ21 = 6, H in C2 H6 , (5.45)
φ22 = 0, H in O2 , (5.46)
φ23 = 0, H in CO2 , (5.47)
φ24 = 2, H in H2 O, (5.48)
φ31 = 0, O in C2 H6 , (5.49)
φ32 = 2, O in O2 , (5.50)
φ33 = 2, O in CO2 , (5.51)
φ34 = 1, O in H2 O. (5.52)
PN
So the stoichiometry equation, i=1φli νi = 0, is given by
ν1
2 0 1 0 0
ν
6 0 0 2 2 = 0 .
ν3 (5.53)
0 2 2 1 0
ν4
Here, there are three equations in four unknowns, so the system is under-constrained. There are many
ways to address this problem. Here, choose the robust way of casting the system into row echelon form.
This is easily achieved by Gaussian elimination. Row echelon form seeks to have lots of zeros in the
lower left part of the matrix. The lower left corner has a zero already, so that is useful. Now, multiply
the top equation by 3 and subtract the result from the second to get
ν1
2 0 1 0 0
ν
0 0 −3 2 2 = 0 .
ν3 (5.54)
0 2 2 1 0
ν4
Now divide the first by 2, the second by 2 and the third by −3 to get unity in the diagonal:
ν
1 0 12 0 1 0
0 1 1 1 ν2 0 .
2 ν3 = (5.56)
0 0 1 − 32 0
ν4
So-called bound variables have non-zero coefficients on the diagonal, so one can take the bound variables
to be ν1 , ν2 , and ν3 . The remaining variables are free variables. Here one takes the free variable to be
Again, one finds a non-unique solution in the right null space of φ. If one chooses ξ = 6, then one gets
ν1 −2
ν2 −7
= , (5.59)
ν3 4
ν4 6
Example 5.5
Consider stoichiometric balance for a propane oxidation reaction which may produce carbon monox-
ide and hydroxyl in addition to carbon dioxide and water.
ν1 C3 H8 + ν2 O2 + ν3 CO2 + ν4 CO + ν5 H2 O + ν6 OH = 0. (5.62)
φ11 = 3, C in C3 H8 , (5.63)
φ12 = 0, C in O2 , (5.64)
φ13 = 1, C in CO2 , (5.65)
φ14 = 1, C in CO, (5.66)
φ15 = 0, C in H2 O, (5.67)
φ16 = 0, C in OH, (5.68)
φ21 = 8, H in C3 H8 , (5.69)
φ22 = 0, H in O2 , (5.70)
φ23 = 0, H in CO2 , (5.71)
φ24 = 0, H in CO, (5.72)
φ25 = 2, H in H2 O, (5.73)
φ26 = 1, H in OH, (5.74)
φ31 = 0, O in C3 H8 , (5.75)
φ32 = 2, O in O2 , (5.76)
φ33 = 2, O in CO2 , (5.77)
φ34 = 1, O in CO, (5.78)
φ35 = 1, O in H2 O, (5.79)
φ36 = 1, O in OH. (5.80)
Multiplying the first equation by −8/3 and adding it to the second gives
ν1
ν2
3 0 1 1 0 0
ν3 0
0 0 − 83 − 38 2
1 = 0 . (5.82)
ν4
0 2 2 1 1 1 ν5 0
ν6
Dividing the first row by 3, the second by 2 and the third by −8/3 gives
ν1
ν2
1 0 3 31 1
0 0
0
0 1 1 1 1 1 ν3 0 .
2 2 2 ν4 = (5.84)
0 0 1 1 − 43 − 38
ν5 0
ν6
Here, one finds three linearly independent vectors in the right null space. The dot product of any vector
in the right null space with any vector in the row space should be zero. For example, take the dot
product of the third row space vector with the first right null space vector and find
0
4
−8
0 2 2 1 1 1
8 = 0 + 8 − 16 + 8 + 0 + 0 = 0. (5.88)
0
0
To simplify the notation, take ξˆ1 = ξ1 /8, ξ̂2 = ξ2 /8, and ξ̂3 = ξ3 /8. Then,
ν1 0 −2 −1
ν2 4 −10 −7
ν3
= ξˆ1 −8 + ξ̂2 6 + ξˆ3 3 . (5.89)
ν4 8 0 0
ν5 0 8 0
ν6 0 0 8
The most general reaction that can achieve a stoichiometric balance is given by
(−2ξˆ2 − ξ̂3 )C3 H8 + (4ξˆ1 − 10ξˆ2 − 7ξˆ3 )O2 + (−8ξˆ1 + 6ξˆ2 + 3ξˆ3 )CO2 + 8ξˆ1 CO + 8ξˆ2 H2 O + 8ξˆ3 OH = 0. (5.90)
(2ξˆ2 + ξ̂3 )C3 H8 + (−4ξˆ1 + 10ξˆ2 + 7ξˆ3 )O2 ⇌ (−8ξˆ1 + 6ξˆ2 + 3ξˆ3 )CO2 + 8ξˆ1 CO + 8ξˆ2 H2 O + 8ξˆ3 OH. (5.91)
This will be balanced for all ξ̂1 , ξ̂2 , and ξˆ3 . The values that are actually achieved in practice depend
on the thermodynamics of the problem. Stoichiometry only provides some limitations.
A slightly more familiar form is found by taking ξˆ2 = 1/2 and rearranging, giving
(1 + ξ̂3 ) C3 H8 + (5 − 4ξˆ1 + 7ξˆ3 ) O2 ⇌ (3 − 8ξˆ1 + 3ξˆ3 ) CO2 + 4 H2 O + 8ξˆ1 CO + 8ξˆ3 OH. (5.92)
One notes that often the production of CO and OH will be small. If there is no production of CO or
OH, ξ̂1 = ξ̂3 = 0 and one recovers the familiar balance of
One also notes that stoichiometry alone admits unusual solutions. For instance, if ξ̂1 = 100, ξ̂2 = 1/2,
and ξˆ3 = 1, one has
This reaction is certainly admitted by stoichiometry but is not observed in nature. To determine
precisely which of the P infinitely many possible final states are realized requires a consideration of the
N
equilibrium condition i=1 νi µi = 0.
Looked at in another way, we can think of three independent classes of reactions admitted by the
stoichiometry, one for each of the linearly independent null space vectors. Taking first ξ̂1 = 1/4, ξˆ2 = 0,
ξˆ3 = 0, one gets, after rearrangement
as one class of reaction admitted by stoichiometry. Taking next ξˆ1 = 0, ξ̂2 = 1/2, ξˆ3 = 0, one gets
as the second class admitted by stoichiometry. The third class is given by taking ξˆ1 = 0, ξ̂2 = 0, ξˆ3 = 1,
and is
C3 H8 + 7O2 ⇌ 3CO2 + 8OH. (5.97)
In this example, both ξ and ξˆ are dimensionless.
In general, one can expect to find the stoichiometric coefficients for N species composed
of L elements to be of the following form:
N
X −L
νi = Dik ξk , i = 1, . . . , N. (5.98)
k=1
with vectors which lie in the right null space of φli . Multiplication of ξk by any constant
gives another set of νi , and mass conservation for each element is still satisfied.
For later use, we associate νi with the number of moles ni , and consider a differential
change in the number of moles dni from Eq. (5.98) and arrive at
N
X −L
dni = Dik dξk , i = 1, . . . , N. (5.99)
k=1
If there is not enough air to burn all the fuel, the mixture is said to be rich. If there is
excess air, the mixture is said to be lean. If there is just enough, the mixture is said to be
stoichiometric. The equivalence ratio, Φ, is defined as the actual fuel-air ratio scaled by the
stoichiometric fuel-air ratio:
Factual Astoichiometric
Φ≡ = . (5.103)
Fstoichiometric Aactual
The ratio Φ is the same whether F ’s are taken on a mass or mole basis, because the ratio of
molecular masses cancel.
Example 5.6
Calculate the stoichiometry of the combustion of methane with air with an equivalence ratio of
Φ = 0.5. If the pressure is 0.1 MPa, find the dew point of the products.
ν1′ CH4 + ν2′ (O2 + 3.76N2 ) ⇌ ν3′′ CO2 + ν4′′ H2 O + ν5′′ N2 , (5.104)
or
ν1 CH4 + ν2 O2 + ν3 CO2 + ν4 H2 O + (ν5 + 3.76ν2)N2 = 0. (5.105)
Here one has N = 5 species and L = 4 elements. Adopting a slightly more intuitive procedure for
variety, one writes a conservation equation for each element to get
ν1 + ν3 = 0, C, (5.106)
4ν1 + 2ν4 = 0, H, (5.107)
2ν2 + 2ν3 + ν4 = 0, O, (5.108)
3.76ν2 + ν5 = 0, N. (5.109)
Now, one might expect to have one free variable, because one has five unknowns in four equations.
While casting the equation in row echelon form is guaranteed to yield a proper solution, one can often
use intuition to get a solution more rapidly. One certainly expects that CH4 will need to be present
for the reaction to take place. One might also expect to find an answer if there is one mole of CH4 . So
take ν1 = −1. Realize that one could also get a physically valid answer by assuming ν1 to be equal to
any scalar. With ν1 = −1, one gets
0 1 0 0 ν2 1
0 0 2 0 ν3 4
= . (5.111)
2 2 1 0 ν4 0
3.76 0 0 1 ν5 0
One easily finds the unique inverse does exist, and that the solution is
ν2 −2
ν3 1
=
ν4 2 . (5.112)
ν5 7.52
If there had been more than one free variable, the four by four matrix would have been singular, and
no unique inverse would have existed.
In any case, the reaction under stoichiometric conditions is
Now Φ = 0.5, so
Check:
1
Factual = = 0.0525. (5.120)
4 + 4(3.76)
For the dew point of the products, one needs the partial pressure of H2 O. The mole fraction of
H2 O is
2
y H2 O = = 0.0499. (5.121)
1 + 2 + 2 + 15.04
So the partial pressure of H2 O is
From the steam tables, the saturation temperature at this pressure is Tsat = Tdew point = 32.88 ◦ C . If
the products cool to this temperature in an exhaust device, the water could condense in the apparatus.
U2 − U1 = − 1W2 ,
1Q2 (5.124)
Z 2
= 1Q2 − P dV, (5.125)
1
= 1Q2 − P (V2 − V1 ), (5.126)
1Q2 = U2 − U1 + P (V2 − V1 ), (5.127)
= H2 − H1 , (5.128)
= Hproducts − Hreactants . (5.129)
So
X X
1Q2 = ni hi − ni hi . (5.130)
products reactants
In this reaction, one measures that 1Q2 = −393522 kJ for the reaction of 1 kmole of C
and O2 . That is the reaction liberates such energy to the environment. So measuring the
heat transfer can give a measure of the enthalpy difference between reactants and products.
Assign a value of enthalpy zero to elements in their standard state at the reference state.
Thus, C and O2 both have enthalpies of 0 kJ/kmole at T = 298 K, P = 0.1 MPa. This
enthalpy is designated, for species i,
o o
hf,i = h298,i , (5.131)
and is called the enthalpy of formation. So the energy balance for the products and reactants,
here both at the standard state, becomes
o o o
= nCO2 hf,CO2 − nC hf,C − nO2 hf,O2 ,
1Q2 (5.132)
o kJ kJ
−393522 kJ = (1 kmole)hf,CO2 − (1 kmole) 0 − (1 kmole) 0 .
kmole kmole
(5.133)
o
Thus, the enthalpy of formation of CO2 is hf,CO2 = −393522 kJ/kmole, because the reaction
involved creation of 1 kmole of CO2 .
Often values of enthalpy are tabulated in the forms of enthalpy differences ∆hi . These
are defined such that
o o
hi = hf,i + (hi − hf,i ), (5.134)
| {z }
=∆hi
o
= hf,i + ∆hi . (5.135)
Lastly, one notes for an ideal gas that the enthalpy is a function of temperature only,
and so does not depend on the reference pressure; hence
o o
hi = hi , ∆hi = ∆hi , if ideal gas. (5.136)
Example 5.7
(adopted from BS). Determine the heat of reaction of the following irreversible reaction in a steady
state, steady flow process confined to the standard state of P = 0.1 MPa, T = 298 K:
All components are at their reference states. Table A.10 gives properties, and one finds
Qcv = nCO2 hCO2 + nH2 O hH2 O − nCH4 hCH4 − nO2 hO2 , (5.139)
kJ kJ
= (1 kmole) −393522 + (2 kmole) −285830
kmole kmole
kJ kJ
−(1 kmole) −74873 − (2 kmole) 0 , (5.140)
kmole kmole
= −890309 kJ. (5.141)
A more detailed analysis is required in the likely case in which the system is not at the
reference state.
Example 5.8
(adopted from Moran and Shapiro) A mixture of 1 kmole of gaseous methane and 2 kmole of oxygen
initially at 298 K and 101.325 kPa burns completely in a closed, rigid, container. Heat transfer occurs
until the final temperature is 900 K. Find the heat transfer and the final pressure.
The combustion is stoichiometric. Assume that no small concentration species are generated. The
global reaction is given by
CH4 + 2O2 → CO2 + 2H2 O. (5.142)
The first law analysis for the closed system is slightly different:
1Q2 = U2 − U1 , (5.144)
= nCO2 uCO2 + nH2 O uH2 O − nCH4 uCH4 − nO2 uO2 , (5.145)
= nCO2 (hCO2 − RT2 ) + nH2 O (hH2 O − RT2 ) − nCH4 (hCH4 − RT1 ) − nO2 (hO2 − RT1 ), (5.146)
= hCO2 + 2hH2 O − hCH4 − 2hO2 − 3R(T2 − T1 ), (5.147)
o o o o
= (hCO2 ,f + ∆hCO2 ) + 2(hH2 O,f + ∆hH2 O ) − (hCH4 ,f + ∆hCH4 ) − 2(hO2 ,f + ∆hO2 )
−3R(T2 − T1 ), (5.148)
= (−393522 + 28030) + 2(−241826 + 21937) − (−74873 + 0) − 2(0 + 0)
−3(8.314)(900 − 298), (5.149)
= −745412 kJ. (5.150)
Now V2 = V1 , so
The pressure increased in the reaction. This is entirely attributable to the temperature rise, as the
number of moles remained constant here.
• The adiabatic flame temperature can be well over 5000 K for seemingly ordinary mix-
tures.
• Dilution of the mixture with an inert diluent lowers the adiabatic flame temperature.
The same effect would happen in rich and lean mixtures.
• Preheating the mixture, such as one might find in the compression stroke of an engine,
increases the adiabatic flame temperature.
• Consideration of the presence of minor species lowers the adiabatic flame temperature.
Example 5.9
A closed, fixed, adiabatic volume contains a stoichiometric mixture of 2 kmole of H2 and 1 kmole
of O2 at 100 kPa and 298 K. Find the adiabatic flame temperature and final pressure assuming the
irreversible reaction
2H2 + O2 → 2H2 O. (5.158)
This is an extremely high temperature. At such temperatures, in fact, one can expect other species to
co-exist in the equilibrium state in large quantities. These will include H, H2 , O2 , OH, O, HO2 , and
H2 O2 . Detailed calculation reveals the energetics of the chemical reactions are such that the presence
of minor species induces the achieved temperature to be considerably lower than 5725 K, in fact near
3200 K.
The final pressure is given by
nH2 O RT2
P2 = , (5.173)
V
kJ
(2 kmole) 8.314 kmole K (5725 K)
= , (5.174)
74.33 m3
= 1280.71 kPa. (5.175)
2 kmole kmole
ρ H2 O = = 2.69 × 10−2 . (5.176)
74.33 m3 m3
Example 5.10
Consider a variant on the previous example in which the mixture is diluted with an inert, taken
here to be N2 . A closed, fixed, adiabatic volume contains a stoichiometric mixture of 2 kmole of H2 ,
1 kmole of O2 , and 8 kmole of N2 at 100 kPa and 298 K. Find the adiabatic flame temperature and
the final pressure, assuming the irreversible reaction
U2 − U1 = 1Q2 − 1W2 ,
= 0,
nH2 O uH2 O − nH2 uH2 − nO2 uO2 + nN2 (uN2 2 − uN2 1 ) = 0,
nH2 O (hH2 O − RT2 ) − nH2 (hH2 − RT1 ) − nO2 (hO2 − RT1 ) + nN2 ((hN2 2 − RT2 ) − (hN2 1 − RT1 )) = 0,
2hH2 O − 2 hH2 − hO2 +R(−10T2 + 11T1 ) + 8( hN2 2 − hN2 1 ) = 0,
|{z} |{z} | {z } | {z }
=0 =0 =∆hN2 =0
At this point, one begins an iteration process, guessing a value of T2 and an associated ∆hH2 O . When
T2 is guessed at 2000 K, the left side becomes −28006.7. When T2 is guessed at 2200 K, the left side
becomes 33895.3. Interpolate then to arrive at
T2 = 2090.5 K. (5.181)
The inert diluent significantly lowers the adiabatic flame temperature. This is because the N2 serves as
a heat sink for the energy of reaction. If the mixture were at non-stoichiometric conditions, the excess
species would also serve as a heat sink, and the adiabatic flame temperature would be lower than that
of the stoichiometric mixture.
The final pressure is given by
(nH2 O + nN2 )RT2
P2 = , (5.182)
V
kJ
(2 kmole + 8 kmole) 8.314 kmole K (2090.5 K)
= , (5.183)
272.533 m3
= 637.74 kPa. (5.184)
The final concentrations of H2 O and N2 are
2 kmole kmole
ρ H2 O = = 7.34 × 10−3 , (5.185)
272.533 m3 m3
8 kmole kmole
ρ N2 = = 2.94 × 10−2 . (5.186)
272.533 m3 m3
Example 5.11
Consider a variant on the previous example in which the diluted mixture is preheated to 1000 K.
One can achieve this via an isentropic compression of the cold mixture, such as might occur in an engine.
To simplify the analysis here, the temperature of the mixture will be increased, while the pressure will
be maintained. A closed, fixed, adiabatic volume contains a stoichiometric mixture of 2 kmole of H2 ,
1 kmole of O2 , and 8 kmole of N2 at 100 kPa and 1000 K. Find the adiabatic flame temperature and
the final pressure, assuming the irreversible reaction
2H2 + O2 + 8N2 → 2H2 O + 8N2 . (5.187)
T2 = 2635.4 K. (5.191)
The preheating raised the adiabatic flame temperature. The preheating was by (1000 K) − (298 K) =
702 K. The new adiabatic flame temperature is only (2635.4 K) − (2090.5 K) = 544.9 K greater.
The final pressure is given by
(nH2 O + nN2 )RT2
P2 = , (5.192)
V
kJ
(2 kmole + 8 kmole) 8.314 kmole K (2635.4 K)
= , (5.193)
914.54 m3
= 239.58 kPa. (5.194)
The final concentrations of H2 O and N2 are
2 kmole kmole
ρ H2 O = 3
= 2.19 × 10−3 , (5.195)
914.54 m m3
8 kmole kmole
ρ N2 = 3
= 8.75 × 10−3 . (5.196)
914.54 m m3
Example 5.12
Consider a variant on the previous example. Here allow for minor species to be present at equilib-
rium. A closed, fixed, adiabatic volume contains a stoichiometric mixture of 2 kmole of H2 , 1 kmole
of O2 , and 8 kmole of N2 at 100 kPa and 1000 K. Find the adiabatic flame temperature and the final
pressure, assuming reversible reactions.
Here, the details of the analysis are postponed, but the result is given which is the consequence of a
calculation involving detailed reactions rates. One can also solve an optimization problem to minimize
the Gibbs free energy of a wide variety of products to get the same answer. In this case, the equilibrium
temperature and pressure are found to be
kmole
minor product ρ H2 = 1.3 × 10−4 , (5.198)
m3
kmole
minor product ρH = 1.9 × 10−5 , (5.199)
m3
kmole
minor product ρO = 5.7 × 10−6 , (5.200)
m3
kmole
minor product ρ O2 = 3.6 × 10−5 , (5.201)
m3
kmole
minor product ρOH = 5.9 × 10−5 , (5.202)
m3
kmole
major product ρ H2 O = 2.0 × 10−3 , (5.203)
m3
kmole
trace product ρHO2 = 1.1 × 10−8 , (5.204)
m3
kmole
trace product ρ H2 O 2 = 1.2 × 10−9 , (5.205)
m3
kmole
trace product ρN = 1.7 × 10−9 , (5.206)
m3
kmole
trace product ρNH = 3.7 × 10−10 , (5.207)
m3
kmole
trace product ρNH2 = 1.5 × 10−10 , (5.208)
m3
kmole
trace product ρNH3 = 3.1 × 10−10 , (5.209)
m3
kmole
trace product ρNNH = 1.0 × 10−10 , (5.210)
m3
kmole
minor product ρNO = 3.1 × 10−6 , (5.211)
m3
kmole
trace product ρNO2 = 5.3 × 10−9 , (5.212)
m3
kmole
trace product ρ N2 O = 2.6 × 10−9 , (5.213)
m3
kmole
trace product ρHNO = 1.7 × 10−9 , (5.214)
m3
kmole
major product ρ N2 = 8.7 × 10−3 . (5.215)
m3
The concentrations of the major products went down when the minor species were considered. The
adiabatic flame temperature also went down by a significant amount: 2635 − 2484.8 = 150.2 K. Some
thermal energy was necessary to break the bonds which induce the presence of minor species.
n1 = ν1 ζ + n1o , (5.230)
n2 = ν2 ζ + n2o , (5.231)
n3 = ν3 ζ + n3o , (5.232)
n4 = ν4 ζ + n4o . (5.233)
One can also eliminate the parameter ζ in a variety of fashions and parameterize the
reaction one of the species mole numbers. Choosing, for example, n1 as a parameter, one
gets
n1 − n1o
ζ= . (5.234)
ν1
Eliminating ζ then one finds all other mole numbers in terms of n1 :
n1 − n1o
n2 = ν2 + n2o , (5.235)
ν1
n1 − n1o
n3 = ν3 + n3o , (5.236)
ν1
n1 − n1o
n4 = ν4 + n4o . (5.237)
ν1
Written another way, one has
n1 − n1o n2 − n2o n3 − n3o n4 − n4o
= = = = ζ. (5.238)
ν1 ν2 ν3 ν4
P
For an N-species reaction, N i=1 νi χi = 0, one can generalize to say
Now, from the previous chapter, one manifestation of the second law is Eq. (4.394):
N
X
dG|T,P = µi dni ≤ 0. (5.243)
i=1
Now, one can eliminate dni in Eq. (5.243) by use of Eq. (5.239) to get
N
X
dG|T,P = µi νi dζ ≤ 0, (5.244)
i=1
XN
∂G
= µi νi ≤ 0, (5.245)
∂ζ T,P i=1
= −α ≤ 0. (5.246)
Then for the reaction to go forward, one must require that the affinity be positive:
α ≥ 0. (5.247)
One also knows from the previous chapter that the irreversible entropy production takes the
form of Eq. (4.382):
N
1X
− µ dni ≥ 0, (5.248)
T i=1 i
N
1 X
− dζ µi νi ≥ 0, (5.249)
T i=1
N
1 dζ X
− µ νi ≥ 0. (5.250)
T dt i=1 i
PN
In terms of the chemical affinity, α = − i=1 µi νi , Eq. (5.250) can be written as
1 dζ
α ≥ 0. (5.251)
T dt
Now one straightforward, albeit naı̈ve, way to guarantee positive semi-definiteness of the
irreversible entropy production and thus satisfaction of the second law is to construct the
chemical kinetic rate equation so that
X N
dζ
= −k µi νi = kα, k ≥ 0, provisional, naı̈ve assumption (5.252)
dt i=1
This provisional assumption of convenience will be supplanted later by a form which agrees
well with experiment. Here k is a positive semi-definite scalar. In general, it is a function of
temperature, k = k(T ), so that reactions proceed rapidly at high temperature and slowly at
low temperature. Then certainly the reaction progress variable ζ will cease to change when
the equilibrium condition
N
X
µi νi = 0, (5.253)
i=1
α = 0. (5.254)
Now, while Eq. (5.253) is the most compact form of the equilibrium condition, it is not
the most commonly used form. One can perform the following analysis to obtain the form
in most common usage. Start by equating the chemical potential with the Gibbs free energy
per unit mole for each species i: µi = g i . Then employ the definition of Gibbs free energy
for an ideal gas, and carry out a set of operations:
N
X
g i νi = 0, at equilibrium, (5.255)
i=1
N
X
(hi − T si )νi = 0, at equilibrium. (5.256)
i=1
For the ideal gas, one can substitute for hi (T ) and si (T, P ) and write the equilibrium con-
dition as
Z Z
N
X o
T T
c ( T̂ ) y P
h298,i + o
cP i (T̂ ) dT̂ −T s298,i +
Pi
d T̂ −R ln
i
νi = 0.
| T̂ Po
i=1 | 298 {z }
298
{z }
o
=soT,i
=∆h T,i
| {z }
| {z }
o
=hT,i =hT,i =sT,i
(5.257)
Now writing the equilibrium condition in terms of the enthalpies and entropies referred to
XN
o o yi P
hT,i − T sT,i − R ln νi = 0, (5.258)
i=1
P o
XN XN
o o yi P
hT,i − T sT,i νi = − RT νi ln , (5.259)
i=1 | {z } i=1
P o
=goT,i =µoT,i
N
X N
X νi
yi P
− g oT,i νi = RT ln , (5.260)
i=1 i=1
Po
| {z }
≡∆Go
X N ν
∆Go yi P i
− = ln , (5.261)
RT i=1
Po
YN ν !
yi P i
= ln , (5.262)
i=1
Po
Y N ν
∆Go yi P i
exp − = , (5.263)
RT Po
| {z } i=1
≡KP
N
Y ν
yi P i
KP = , (5.264)
i=1
Po
PNi=1 νi Y n
P
= yiνi . (5.265)
Po i=1
So
N
Y ν
Pi i
KP = , at equilibrium. (5.266)
i=1
Po
Only at equilibrium does the property KP also equal the product of the partial pressures
as in Eq. (5.266). The subscript P for pressure comes about because it is also related to
the product of the ratio of the partial pressure to the reference pressure raised to the net
stoichiometric coefficients. Also, the net change in Gibbs free energy of the reaction at the
reference pressure, ∆Go , which is a function of T only, has been defined as
N
X
∆Go ≡ g oT,i νi . (5.268)
i=1
The term ∆Go has units of kJ/kmole; it traditionally does not get an overbar. If ∆Go > 0,
one has 0 < KP < 1, and reactants are favored over products. If ∆Go < 0, one gets KP > 1,
and products are favored over reactants. One can also deduce that higher pressures P push
the equilibrium in such a fashion that fewer moles are present, all else being equal. One can
also define ∆Go in terms of the chemical affinity, referred to the reference pressure, as
One can also define another convenient thermodynamic property, which for an ideal gas
is a function of T alone, the equilibrium constant Kc :
PNi=1 νi
Po ∆Go
Kc ≡ exp − , generally valid. (5.270)
RT RT
So
N
Y
Kc = ρi νi , at equilibrium. (5.274)
i=1
One must be careful to distinguish between the general definition of Kc as given in Eq. (5.270),
and the fact that at equilibrium it is driven to also have the value of product of molar species
concentrations, raised to the appropriate stoichiometric power, as given in Eq. (5.274).
d ρi ω̇i
= . (5.275)
dt ρ ρ
Multiplying both sides of Eq. (5.275) by molecular mass Mi and using the definition of mass
fraction ci then gives the alternate form
dci ω̇i Mi
= . (5.276)
dt ρ
Because the density is constant for the isochoric system, Eq. (5.275) reduces to
dρi
= ω̇i , isochoric. (5.278)
dt
Then, experiment, as well as a more fundamental molecular collision theory, shows that the
evolution of species concentration i is given by
≡ω̇i
z }| {
!
dρi −E
N
Y 1 Y νk
N
β νk′
= νi aT exp ρk 1 − ρk , isochoric system. (5.279)
dt RT Kc
| {z } | k=1{z } | k=1 {z }
≡k(T ) forward reaction reverse reaction
| {z }
≡r
This relation actually holds for isochoric, non-isothermal systems as well, which will not be
considered in any detail here. Here some new variables are defined as follows:
• a: a kinetic rate constant called the collision frequency factor. Its units will depend
on the actual reaction and could involve various combinations of length, time, and
temperature. It is constructed Pso that dρi /dt has units of kmole/m3 /s; this requires it
N
to have units of (kmole/m3 )(1− k=1 νk ) /s/Kβ .
′
• E: the activation energy. It has units of kJ/kmole, though others are often used, and is
fit by both experiment and fundamental theory to account for the strong temperature
dependency of reaction.
In Eq. (5.279) that molar concentrations are raised to the νk′ and νk powers. As it does not
make sense to raise a physical quantity to a power with units, one traditionally interprets the
values of νk , νk′ , as well as νk′′ to be dimensionless pure numbers. They are also interpreted
in a standard fashion: the smallest integer values that actually correspond to the underlying
molecular collision which has been modeled. While stoichiometric balance can be achieved
by a variety of νk values, the kinetic rates are linked to one particular set which is defined
by the community.
Equation (5.279) is written in such a way that the species concentration production rate
increases when
• The temperature increases; here, the sensitivity may be very high, as one observes in
nature,
• The species concentrations of species involved in the forward reaction increase; this
embodies the principle that the collision-based reaction rates are enhanced when there
are more molecules to collide,
or
1 Y νk′′
N
Y N
β −E ν′
r = aT exp ρk k − ρk . (5.281)
RT Kc
| {z } k=1
| {z } | k=1 {z }
≡k(T ), Arrhenius rate
| forward reaction{z reverse reaction }
law of mass action
This equation was advocated by van’t Hoff in 1884; in 1889 Arrhenius gave a physical justifi-
cation. The units of k(T ) actually depend on the reaction. This is a weakness of P
the theory,
3 (1− N ′
k=1 νk ) /s.
and precludes a clean non-dimensionalization. The units must be (kmole/m )
In terms of reaction progress, one can also take
1 dζ
r= . (5.283)
V dt
The factor of 1/V is necessary because r has units of molar concentration per time and ζ
has units of kmoles. The over-riding importance of the temperature sensitivity is illustrated
as part of the next example. The remainder of the expression involving the products of the
species concentrations is the defining characteristic of systems which obey the law of mass
action. Though the history is complex, most attribute the law of mass action to Waage and
Guldberg in 1864.1
Last, the overall molar production rate of species i, is often written as ω̇i , defined as
ω̇i ≡ νi r. (5.284)
1
P. Waage and C. M. Guldberg, 1864, “Studies Concerning Affinity, Forhandlinger: Videnskabs-Selskabet
i Christiania, 35. English translation: Journal of Chemical Education, 63(12): 1044-1047.
As νi is considered to be dimensionless, the units of ω̇i must be kmole/m3 /s. So we can say
for isochoric systems combining Eqs. (5.284) and (5.278) that
dρi
= νi r, isochoric. (5.285)
dt
Example 5.13
Study the nitrogen dissociation problem considered in an earlier example, see p. 152, in which at
t = 0 s, 1 kmole of N2 exists at P = 100 kPa and T = 6000 K. Take as before the reaction to be
isothermal and isochoric. Consider again the elementary nitrogen dissociation reaction
N2 + N2 ⇌ 2N + N2 , (5.286)
cm3 K1.6
a = 7.0 × 1021 , (5.287)
mole s
β = −1.6, (5.288)
cal
E = 224928.4 . (5.289)
mole
3 1.6
3
21 cm K 1m 1000 mole m3 K1.6
a = 7.0 × 10 = 7.0 × 1018 , (5.290)
mole s 100 cm kmole kmole s
cal J kJ 1000 mole kJ
E = 224928.4 4.186 = 941550 . (5.291)
mole cal 1000 J kmole kmole
At the initial state, the material is all N2 , so PN2 = P = 100 kPa. The ideal gas law then gives at t = 0
Thus, the volume, constant for all time in the isochoric process, is
1
−dnN2 = dnN , (5.297)
2
1
−(nN2 − nN2 |t=0 ) = (nN − nN |t=0 ), (5.298)
| {z } 2 | {z }
=1 kmole =0
nN = 2(1 kmole − nN2 ), (5.299)
nN 1 kmole nN2
= 2 − , (5.300)
V V V
1 kmole
ρN = 2 − ρ N2 , (5.301)
4.9884 × 102 m3
−3 kmole
= 2 2.00465 × 10 − ρ N2 . (5.302)
m3
Now the general equation for kinetics of a single reaction, Eq. (5.279), reduces for N2 molar con-
centration to
dρN2 −E ′
νN ′ 1
= νN2 aT β exp (ρN2 ) 2
νN
(ρN ) 1− νN2
(ρ ) (ρN )νN
. (5.303)
dt RT K c N2
′ ′
Realizing that νN 2
= 2, νN = 0, νN2 = −1, and νN = 2, one gets
dρN2 β −E 2 1 ρ2N
= − aT exp ρ N2 1 − . (5.304)
dt RT K c ρ N2
| {z }
=k(T )
Figure 5.1 gives a plot of k(T ) which shows its very strong dependency on temperature. For this
problem, T = 6000 K, so
7.0 × 1018 −1.1325 × 105
k(6000) = exp , (5.308)
60001.6 6000
m3
= 40071.6 . (5.309)
kmole s
10 4
-4
10
10 -12
10 -20
10 -28
10 -36 T (K)
1000 1500 2000 3000 5000 7000 10000
PN
For this system, because i=1 νi = 1, this reduces to
Po −(2goN − goN2 )
Kc = exp , (5.311)
RT RT
o o !
Po −(2(hN − T soT,N ) − (hN2 − T soT,N2 ))
= exp , (5.312)
RT RT
100 −(2(597270 − (6000)216.926) − (205848 − (6000)292.984))
= exp ,(5.313)
(8.314)(6000) (8.314)(6000)
kmole
= 0.000112112 . (5.314)
m3
(5.315)
= f (ρN2 ). (5.316)
The system is at equilibrium when f (ρN2 ) = 0. This is an algebraic function of ρN2 only, and can be
plotted. Figure 5.2 gives a plot of f (ρN2 ) and shows that it has three potential equilibrium points. It
is seen there are three roots. Solving for the equilibria requires solving
kmole
2 !
m3 1 2 2.00465 × 10−3 m3 − ρN2
0 = − 40071.6 ρ2N2 1− kmole
. (5.317)
kmole 0.000112112 m3
ρ N2
unstable
equilibrium
1
-1
unstable stable,
equilibrium physical
equilibrium
-2
Figure 5.2: Forcing function, f (ρN2 ), which drives changes of ρN2 as a function of ρN2 in
isothermal, isochoric problem.
By inspection of the topology of Fig. 5.2, the only stable root is 0.00178121 kmole/m3 . This root
agrees with the equilibrium value found in an earlier example for the same problem conditions. Small
perturbations from this equilibrium induce the forcing function to supply dynamics which restore the
system to its original equilibrium state. Small perturbations from the unstable equilibria induce non-
restoring dynamics. For this root, one can then determine that the stable equilibrium value of ρN =
0.000446882 kmole/m3 .
One can examine this stability more formally. Define an equilibrium concentration ρeqN2 such that
f (ρeq
N2 ) = 0. (5.319)
df 1 d2 f
f (ρN2 ) ∼ f (ρeq
N2 ) + (ρN2 − ρeq
N2 ) + (ρ − ρeq 2
N2 ) + . . . (5.320)
| {z } dρN2 ρN2 =ρeq
2 dρ2N2 N2
N 2
=0
Now the first term of the Taylor series is zero by construction. Next neglect all higher order terms as
small so that the approximation becomes
df
f (ρN2 ) ∼ (ρN2 − ρeq
N2 ). (5.321)
dρN2 ρN2 =ρeq
N2
concentration (kmole/m3)
0.002
0.001
t (s)
0.001 0.002 0.003 0.004 0.005
Figure 5.3: ρN2 (t) and ρN (t) in isothermal, isochoric nitrogen dissociation problem.
Because the derivative of a constant is zero, one can also write the equation as
d df
(ρN2 − ρeq
N2 ) ∼ (ρN2 − ρeq
N2 ). (5.323)
dt dρN2 ρN2 =ρeq
N2
Here C is some constant whose value is not important for this discussion. If the slope of f is positive,
that is,
df
> 0, unstable, (5.326)
dρN2 ρ =ρeq
N2 N2
the equilibrium will be unstable. That is a perturbation will grow without bound as t → ∞. If the
slope is zero,
df
= 0, neutrally stable, (5.327)
dρN2 ρ =ρeq
N2 N2
the solution is stable in that there is no unbounded growth, and moreover is known as neutrally stable.
If the slope is negative,
df
< 0, asymptotically stable, (5.328)
dρN2 ρN2 =ρeq
N 2
the solution is stable in that there is no unbounded growth, and moreover is known as asymptotically
stable.
A numerical solution via an explicit technique such as a Runge-Kutta integration is found for
Eq. (5.315). The solution for ρN2 , along with ρN is plotted in Fig. 5.3.
Linearization of Eq. (5.315) about the equilibrium state gives rise to the locally linearly valid
d
ρN2 − 0.00178121 = −1209.39(ρN2 − 0.00178121) + . . . (5.329)
dt
This has local asymptotically stable solution
ρN2 = 0.00178121 + C exp (−1209.39t) . (5.330)
Here C is some integration constant whose value is irrelevant for this analysis. The time scale of
relaxation τ is the time when the argument of the exponential is −1, which is
1
τ= = 8.27 × 10−4 s. (5.331)
1209.39 s−1
One usually finds this time scale to have high sensitivity to temperature, with high temperatures giving
fast time constants and thus fast reactions.
The equilibrium values agree exactly with those found in the earlier example; see Eq. (4.469). Here
the kinetics provide the details of how much time it takes to achieve equilibrium. This is one of the key
questions of non-equilibrium thermodynamics.
So a global density decrease of the inert material due to volume increase of a fixed mass
system induces a concentration decrease of each species. As an aside, we could also analyze
inert isobaric mixtures as follows, first dividing both sides of Eq. (5.338) by ρi :
1 dρi 1 dρ
= , (5.339)
ρi dt ρ dt
d d
(ln ρi ) = (ln ρ) , (5.340)
dt dt
ρ ρ
ln i = ln , (5.341)
ρio ρo
ρi ρ
= , inert, isobaric mixture. (5.342)
ρio ρo
The ratio of instantaneous species molar concentration to its initial value is the same as the
ratio of the instantaneous mass density to its initial value.
Extending to a material with a single reaction rate r, one can naı̈vely, but correctly, add
the just-described inert expansion effect, Eq. (5.338), to the isochoric Eq. (5.285) to get
dρi ρ dρ
= νi r + i . (5.343)
dt ρ dt
dρi ρi dρ
− = νi r, (5.344)
dt ρ dt
1 dρi ρ dρ νi r
− 2i = , (5.345)
ρ dt ρ dt ρ
d ρi νi r ω̇i
= = , generally valid. (5.346)
dt ρ ρ ρ
Equation (5.346) is identical to Eq. (5.275) which was postulated as the most generally valid
form for a spatially homogeneous chemical reaction. It is actually valid for general systems
with variable density, temperature, and pressure.
However, in this section, it is required that pressure and temperature be constant. Now
differentiate the isobaric, isothermal, ideal gas law to get the density derivative.
N
X
P = ρi RT. (5.347)
i=1
XN
dP dρ
= RT i , (5.348)
dt
|{z} i=1
dt
=0
N
X dρ i
0 = , (5.349)
i=1
dt
XN
ρi dρ
= νi r + , (5.350)
i=1
ρ dt
N
X N
1 dρ X
= r νi + ρ. (5.351)
i=1
ρ dt i=1 i
N
X
∆n ≡ νk . (5.357)
k=1
Now use Eq. (5.356) to eliminate the density derivative in Eq. (5.343) to get
PN !
dρi ρ ρRT r k=1 νk
= νi r + i − , (5.358)
dt ρ P
| {z }
dρ/dt
yi
z }| {
N
ρi RT X
= r νi − νk
, (5.359)
|{z}
reaction effect P
| {zk=1 }
expansion effect
or
dρi
= r νi − yi ∆n . (5.360)
dt |{z} | {z }
reaction effect expansion effect
There are two terms dictating the rate change of species molar concentration. The first, a
reaction effect, is precisely the same term that drove the isochoric reaction. The second is
due to the fact that the volume can change if the number of moles change, and this induces
an intrinsic change in concentration. The term ρi RT /P = yi , the mole fraction.
Example 5.14
Study a variant of the nitrogen dissociation problem considered in an earlier example, see p. 157, in
which at t = 0 s, 1 kmole of N2 exists at P = 100 kPa and T = 6000 K. In this case, take the reaction
to be isothermal and isobaric. Consider again the elementary nitrogen dissociation reaction
N2 + N2 ⇌ 2N + N2 , (5.361)
cm3 K1.6
a = 7.0 × 1021 , (5.362)
mole s
β = −1.6, (5.363)
cal
E = 224928.4 . (5.364)
mole
At the initial state, the material is all N2 , so PN2 = P = 100 kPa. The ideal gas law then gives at t = 0
In this isobaric process, one always has P = 100 kPa. Now, in general
P = RT (ρN2 + ρN ); (5.372)
P
ρN = − ρ N2 , (5.373)
RT
100 kPa
= kJ
− ρ N2 , (5.374)
8.314 kmole K (6000 K)
kmole
= 2.00465 × 10−3 − ρ N2 . (5.375)
m3
Then the equations for kinetics of a single isobaric isothermal reaction, Eq. (5.359) in conjunction
with Eq. (5.280), reduce for N2 molar concentration to
!
dρN2 −E ν ′ ′ 1 ρN RT
= aT β exp (ρN2 ) N2 (ρN )νN 1 − (ρ )νN2 (ρN )νN ν N2 − 2 (νN2 + νN ) .
dt RT K c N2 P
| {z }
=r
(5.376)
′ ′
Realizing that νN 2
= 2, νN = 0, νN2 = −1, and νN = 2, one gets
!
dρN2 2 ρ RT
−E 1 ρ
= aT β exp ρ2N2 1 − N
−1 − N2 . (5.377)
dt RT K c ρ N2 P
| {z }
=k(T )
The temperature dependency of the reaction is unchanged from the previous reaction:
β −E
k(T ) = aT exp , (5.378)
RT
!
3 1.6 kJ
18 m K −1.6 −941550 kmole
= 7.0 × 10 T exp kJ
, (5.379)
kmole s 8.314 kmole KT
7.0 × 1018 −1.1325 × 105
= exp . (5.380)
T 1.6 T
The system is at equilibrium when f (ρN2 ) = 0. This is an algebraic function of ρN2 only, and can be
plotted. Figure 5.4 gives a plot of f (ρN2 ) and shows that it has four potential equilibrium points. It is
seen there are four roots. Solving for the equilibria requires solving
2 !
m3 2 1 2.00465 × 10−3 kmole
m3 − ρ N2
0 = 40130.2 ρ N2 1 −
kmole 0.000112112 kmole
m3
ρ N2
!
kJ
ρ 8.314 kmole K (6000 K)
× −1 − N2 . (5.391)
100 kPa
By inspection of the topology of Fig. 5.2, the only stable, physical root is 0.001583 kmole/m3 . Small
perturbations from this equilibrium induce the forcing function to supply dynamics which restore the
unstable 3 unstable
equilibrium equilibrium
2
-1
stable,
stable,
physical
non-physical
equilibrium
equilibrium
Figure 5.4: Forcing function, f (ρN2 ), which drives changes of ρN2 as a function of ρN2 in
isothermal, isobaric problem.
system to its original equilibrium state. Small perturbations from the unstable equilibria induce non-
restoring dynamics. For this root, one can then determine that the stable equilibrium value of ρN =
0.000421 kmole/m3 . A numerical solution via an explicit technique such as a Runge-Kutta integration
is found for Eq. (5.391). The solution for ρN2 , along with ρN is plotted in Fig. 5.5.
Linearization of Eq. (5.391) about the equilibrium state gives rise to the locally linearly valid
d
ρN2 − 0.001583 = −967.073(ρN2 − 0.001583) + . . . (5.393)
dt
Again, C is an irrelevant integration constant. The time scale of relaxation τ is the time when the
argument of the exponential is −1, which is
1
τ= = 1.03 × 10−3 s. (5.395)
967.073 s−1
The time constant for the isobaric combustion is about a factor 1.25 greater than for isochoric combus-
tion under the otherwise identical conditions.
The equilibrium values agree exactly with those found in the earlier example; see Eq. (4.515).
Again, the kinetics provide the details of how much time it takes to achieve equilibrium.
0.002
concentration (kmole/m3)
0.0015
0.001
0.0005
t (s)
0.001 0.002 0.003 0.004 0.005
Figure 5.5: ρN2 (t) and ρN (t) in isobaric, isothermal nitrogen dissociation problem.
One can easily show that the isochoric reaction rate model, Eq. (5.279), satisfies the principle
of mixture mass conservation. Begin with Eq. (5.279) in a compact form, using the definition
of the reaction rate r, Eq. (5.281) and perform the following operations:
dρi
= νi r, (5.396)
dt
d ρci
= νi r, (5.397)
dt Mi
d
(ρci ) = νi Mi r, (5.398)
dt
X L
= νi Ml φli r, (5.399)
|l=1 {z }
=Mi
L
X
= Ml φli νi r, (5.400)
l=1
XN XN X L
d
(ρci ) = Ml φli νi r, (5.401)
i=1
dt i=1 l=1
d
N L X N
X X
ρ ci = Ml φli νi r, (5.402)
dt i=1
l=1 i=1
| {z }
=1
X L
X N
dρ
= r Ml φli νi . (5.403)
dt l=1
|i=1{z }
=0
Therefore, we get
dρ
= 0. (5.404)
dt
PN
The term i=1 φli νi = 0 because of stoichiometry, Eq. (5.24).
Here the elemental mole density, ρl e , for element l has been defined. So the element concen-
tration for each element remains constant in a constant volume reaction process:
dρl e
= 0, l = 1, . . . , L. (5.412)
dt
One can also multiply by the elemental mass, Ml to get the elemental mass density, ρel :
ρel ≡ Ml ρl e , l = 1, . . . , L. (5.413)
Because Ml is a constant, one can incorporate this definition into Eq. (5.412) to get
dρel
= 0, l = 1, . . . , L. (5.414)
dt
The element mass density remains constant in the constant volume reaction. One could also
simply say because the elements’ density is constant, and the mixture is simply a sum of the
elements, that the mixture density is conserved as well.
dm
= 0, (5.415)
dt
d
(ρV ) = 0, (5.416)
dt
dρ
V = 0, (5.417)
dt
dρ
= 0. (5.418)
dt
dU
= Q̇ − Ẇ . (5.419)
dt
But there is no heat transfer or work in the adiabatic isochoric process, so one gets
dU
= 0, (5.420)
dt
d
(mu) = 0, (5.421)
dt
du dm
m +u = 0, (5.422)
dt dt
|{z}
=0
du
= 0. (5.423)
dt
Thus for the mixture of ideal gases, u(T, ρ1 , . . . , ρN ) = uo. One can see how reaction rates
affect temperature changes by expanding the derivative in Eq. (5.423)
N
!
d X
ci u i = 0, (5.424)
dt i=1
X N
d
(ci ui ) = 0, (5.425)
i=1
dt
XN
dui dci
ci + ui = 0, (5.426)
i=1
dt dt
XN
dui dT dci
ci + ui = 0, (5.427)
i=1
dT dt dt
XN
dT dci
ci cvi + ui = 0, (5.428)
i=1
dt dt
N N
dT X X dci
ci cvi = − ui , (5.429)
dt i=1 i=1
dt
| {z }
=cv
N
X
dT d Mi ρi
cv = − ui , (5.430)
dt i=1
dt ρ
N
X
dT dρi
ρcv = − ui Mi , (5.431)
dt i=1
dt
XN
= − ui Mi νi r, (5.432)
i=1
PN PN
dT r i=1 νi ui i=1 ui ω̇i
= − =− . (5.433)
dt ρcv ρcv
N
X
∆U ≡ νi ui , (5.434)
i=1
dT r∆U
=− . (5.435)
dt ρcv
The rate of temperature change is dependent on the absolute energies, not the energy dif-
ferences. If the reaction is going forward, so r > 0, and that is a direction in which the net
molar energy change is negative, then the temperature will rise.
Solving for the reaction dynamics in an adiabatic isobaric system requires some non-obvious
manipulations. First, the first law of thermodynamics says dU = dQ − dW . Because the
process is adiabatic, one has dQ = 0, so dU + P dV = 0. Because it is isobaric, one gets
d(U + P V ) = 0, or dH = 0. So the total enthalpy is constant. Then
d
H = 0, (5.436)
dt
d
(mh) = 0, (5.437)
dt
dh
= 0, (5.438)
dt!
N
X
d
ci hi = 0, (5.439)
dt i=1
N
X d
(ci hi ) = 0, (5.440)
i=1
dt
XN
dhi dci
ci + hi = 0, (5.441)
i=1
dt dt
XN
dhi dT dci
ci + hi = 0, (5.442)
i=1
dT dt dt
N
X N
dT X dci
ci cP i + hi = 0, (5.443)
i=1
dt i=1
dt
N N
dT X X dci
ci cP i + hi = 0, (5.444)
dt i=1 i=1
dt
| {z }
=cP
N
dT X d ρi Mi
cP + hi = 0, (5.445)
dt i=1
dt ρ
N
dT X d ρi
cP + hi Mi = 0. (5.446)
dt i=1
dt ρ
Now use Eq (5.346) to eliminate the term in Eq. (5.446) involving molar concentration
derivatives to get
N
dT X νi r
cP + hi = 0, (5.447)
dt i=1
ρ
PN PN
dT r i=1 hi νi i=1 hi ω̇i
= − =− . (5.448)
dt ρcP ρcP
So the temperature derivative is known as an algebraic function. If one defines the net
enthalpy change as
XN
∆H ≡ hi νi , (5.449)
i=1
one gets
dT r∆H
=− . (5.450)
dt ρcP
Now differentiate the isobaric ideal gas law to get the density derivative.
N
X
P = ρi RT, (5.451)
i=1
XN N
dP dT X dρ
= ρi R + RT i , (5.452)
dt
|{z} i=1
dt i=1
dt
=0
N N
dT X X ρi dρ
0 = ρ + T νi r + , (5.453)
dt i=1 i i=1 ρ dt
N N N
1 dT X X 1 dρ X
= ρ +r νi + ρ. (5.454)
T dt i=1 i i=1
ρ dt i=1 i
Solving, we get
PN PN
dρ − T1 dT
dt i=1 ρi − r i=1 νi
= PN ρi . (5.455)
dt i=1 ρ
1 r
PN
hi νi PN PN
i=1 ρi − r i=1 νi
i=1
dρ T ρcP
= PN ρi . (5.456)
dt i=1 ρ
Now recall that ρ = ρ/M and cP = cP M, so ρ cP = ρcP . Then Eq. (5.456) can be reduced
slightly:
=1
z }| {
XN
PN
hi νi ρ i PN
i=1
cP T
− i=1 νi
dρ i=1
ρ
= rρ PN , (5.457)
dt i=1 ρi
PN P
hi νi
i=1 cP T − N i=1 νi
= rρ PN , (5.458)
ρ
PN i=1 i
hi
i=1 νi cP T − 1
= rρ P
, (5.459)
RT
N N
ρRT X hi X hi
= r νi − 1 = rM νi −1 , (5.460)
P i=1 cP T i=1
cP T
P
where M is the mean molecular mass. For exothermic reaction N i=1 νi hi < 0, so exothermic
reaction induces a density decrease as the increased temperature at constant pressure causes
the volume to increase.
Then using Eq. (5.460) to eliminate the density derivative in Eq. (5.343), and changing
the dummy index from i to k, one gets an explicit expression for concentration evolution:
X N
dρi ρ hk
= νi r + i rM νk −1 , (5.461)
dt ρ k=1
cP T
N
ρi X hk
= r νi + M νk −1
, (5.462)
ρ cP T
| {z } k=1
=yi
N
X !
hk
= r νi + yi νk −1 . (5.463)
k=1
cP T
PN
Defining the change of enthalpy of the reaction as ∆H ≡ k=1 νk hk , and the change of
P
number of the reaction as ∆n ≡ N k=1 νk , one can also say
dρi ∆H
= r νi + yi − ∆n . (5.464)
dt cP T
Exothermic reaction, ∆H < 0, and net number increases, ∆n > 0, both tend to decrease the
molar concentrations of the species in the isobaric reaction.
Lastly, the evolution of the adiabatic, isobaric system, can be described by the simul-
taneous, coupled ordinary differential equations: Eqs. (5.448, 5.456, 5.463). These require
numerical solution in general. One could also employ a moreP fundamental treatment as a
differential algebraic system involving H = H1 , P = P1 = RT N i=1 ρi and Eq. (5.343).
N
! N
!
dS V Y ν′ 1 Y νi
= k(T ) ρi i 1− ρ α ≥ 0, (5.471)
dt U,V T i=1
Kc i=1 i
V
= rα, (5.472)
T
α dζ
= . (5.473)
T dt
Consider now the affinity α term in Eq. (5.470) and expand it so that it has a more useful
form:
N
X N
X
α=− µi νi = − g i νi , (5.474)
i=1 i=1
XN
Pi
= − g oT,i + RT ln νi , (5.475)
i=1
Po
XN X N νi
Pi
= − g oT,i νi −RT ln , (5.476)
P o
|i=1 {z } i=1
=∆Go
N νi
−∆Go X Pi
= RT − ln , (5.477)
Po
| RT
{z } i=1
=ln KP
N ν !
Y Pi i
= RT ln KP − ln , (5.478)
i=1
Po
N
Y ν !
1 Pi i
= −RT ln + ln , (5.479)
KP i=1
P o
N
Y ν !
1 Pi i
= −RT ln , (5.480)
KP i=1 Po
PNi=1 νi N
Po Y ρ RT νi
= −RT ln RT i , (5.481)
Kc i=1
P o
N
!
1 Y νi
= −RT ln ρ . (5.482)
Kc i=1 i
Equation (5.482) is the common definition of affinity. Another form can be found by em-
To see clearly that the entropy production rate is positive semi-definite, substitute
Eq. (5.482) into Eq. (5.470) to get
N
! N
! N
!!
dS V Y νi′ 1 Y νi 1 Y νi
= k(T ) ρi 1− ρ −RT ln ρ ≥ 0,
dt U,V T i=1
Kc i=1 i Kc i=1 i
(5.486)
N
! N
! N
!
Y νi′ 1 Y 1 Y
= −RV k(T ) ρi 1− ρi νi ln ρ i νi ≥ 0. (5.487)
i=1
Kc i=1
Kc i=1
Obviously, if the forward rate is greater than the reverse rate R′ − R′′ > 0, ln(R′ /R′′ ) > 0,
and the entropy production is positive. If the forward rate is less than the reverse rate,
R′ − R′′ < 0, ln(R′ /R′′ ) < 0, and the entropy production is still positive. The production
rate is zero when R′ = R′′ .
The affinity α can be written as
′
R
α = RT ln . (5.494)
R′′
And so when the forward reaction rate exceeds the reverse, the affinity is positive. It is zero
at equilibrium, when the forward reaction rate equals the reverse.
nA = νA ζ + nAo , (5.496)
|{z} |{z}
=−1 =no
nB = νB ζ + nBo . (5.497)
|{z} |{z}
=1 =0
Thus
nA = −ζ + no , (5.498)
nB = ζ. (5.499)
Now no is constant throughout the reaction. Scale by this and define the dimensionless
reaction progress as ζ̂ ≡ ζ/no to get
nA
= −ζ̂ + 1, (5.500)
no
|{z}
=yA
nB
= ζ̂. (5.501)
no
|{z}
=yB
yA = 1 − ζ̂, (5.502)
yB = ζ̂. (5.503)
yA + yB = 1. (5.509)
Now r = (1/V )dζ/dt = (1/V )d(no ζ̂)/dt = (no /V )d(ζ̂)/dt = ρo dζ̂/dt. So the reaction
dynamics can be described by a single ordinary differential equation in a single unknown:
!
dζ̂ 1 ζ̂
ρo = kρo (1 − ζ̂) 1 − , (5.514)
dt Kc 1 − ζ̂
!
dζ̂ 1 ζ̂
= k(1 − ζ̂) 1 − . (5.515)
dt Kc 1 − ζ̂
(5.522)
N
! N
! N
!
dA Y νi′ 1 Y νi 1 Y νi
= RV T k(T ) ρi 1− ρ ln ρ ≤ 0.
dt T,V i=1
Kc i=1 i Kc i=1 i
(5.523)
For the assumptions of this section, Eq. (5.523) reduces to
! !
dA 1 ζ̂ 1 ζ̂
= RT kρo V (1 − ζ̂) 1 − ln ≤ 0, (5.524)
dt T,V Kc 1 − ζ̂ Kc 1 − ζ̂
! !
1 ζ̂ 1 ζ̂
= kno RT (1 − ζ̂) 1 − ln ≤ 0. (5.525)
Kc 1 − ζ̂ Kc 1 − ζ̂
Because the present analysis is nothing more than a special case of the previous section,
Eq. (5.525) certainly holds. One questions however the behavior in the irreversible limit,
1/Kc → 0. Evaluating this limit, one finds
dA 1
lim = kno RT
(1 − ζ̂) ln +(1 − ζ̂) ln ζ̂ − (1 − ζ̂) ln(1 − ζ̂) + . . .
≤ 0.
1/Kc →0 dt T,V | {z } Kc
>0 | {z }
→−∞
(5.526)
Now, performing the distinguished limit as ζ̂ → 1; that is the reaction goes to completion,
one notes that all terms are driven to zero for small 1/Kc . Recall that 1− ζ̂ goes to zero faster
than ln(1 − ζ̂) goes to −∞. The entropy inequality is ill-defined for a formally irreversible
reaction with 1/Kc = 0.
Thermochemistry of multiple
reactions
This chapter will extend notions associated with the thermodynamics of a single chemical
reactions to systems in which many reactions occur simultaneously.
However, each reaction has a stoichiometric coefficient. The j th reaction can be summarized
in the following ways:
N
X N
X
χi νij′ ⇌ χi νij′′ , j = 1, . . . , J, (6.2)
i=1 i=1
XN
χi νij = 0, j = 1, . . . , J. (6.3)
i=1
219
220 CHAPTER 6. THERMOCHEMISTRY OF MULTIPLE REACTIONS
Stoichiometry for the j th reaction and lth element is given by the extension of Eq. (5.24):
N
X
φli νij = 0, l = 1, . . . , L, j = 1, . . . , J. (6.4)
i=1
The net change in Gibbs free energy and equilibrium constants of the j th reaction are defined
by the extensions of Eqs. (5.268, 5.267, 5.270):
N
X
∆Goj ≡ g oT,i νij , j = 1, . . . , J, (6.5)
i=1
−∆Goj
KP,j ≡ exp , j = 1, . . . , J, (6.6)
RT
PNi=1 νij
Po −∆Goj
Kc,j ≡ exp , j = 1, . . . , J. (6.7)
RT RT
N
X
µi νij = 0, j = 1, . . . , J, (6.8)
i=1
In terms of the chemical affinity of each reaction, the equilibrium condition is simply the
extension of Eq. (5.254):
αj = 0, j = 1, . . . , J. (6.11)
At equilibrium, then the equilibrium constraints can be shown to reduce to the extension
of Eq. (5.266):
N
Y ν
Pi ij
KP,j = , j = 1, . . . , J, (6.12)
i=1
Po
For isochoric reaction, the evolution of species concentration i due to the combined effect
of J reactions is given by the extension of Eq. (5.279):
≡ω̇i
z }| {
!
dρi
J
X −E j
N
Y 1 Y νkj
N
βj ′
νkj
= νij aj T exp ρk 1 − ρk , i = 1, . . . , N.
dt RT Kc,j k=1
j=1 | {z } | k=1{z } | {z }
≡kj (T ) forward reaction reverse reaction
| {z }
≡rj =(1/V )dζj /dt
(6.14)
The extension to isobaric reactions is straightforward, and follows the same analysis as for
a single reaction. Again, three intermediate variables which are in common usage have been
defined. First one takes the reaction rate of the j th reaction to be the extension of Eq. (5.280)
Y !
−E j
N 1 Y νkj
N
βj
′
νkj
rj ≡ aj T exp ρk 1 − ρk , j = 1, . . . , J, (6.15)
RT K c,j
| {z } | k=1{z } | k=1
{z }
≡kj (T ) forward reaction reverse reaction
1 dζj
rj = . (6.17)
V dt
Here ζj is the reaction progress variable for the j th reaction.
Each reaction has a temperature-dependent rate function kj (T ), which is an extension of
Eq. (5.282):
βj −E j
kj (T ) ≡ aj T exp , j = 1, . . . , J. (6.18)
RT
The evolution rate of each species is given by ω̇i , defined now as an extension of Eq. (5.284):
J
X
ω̇i ≡ νij rj , i = 1, . . . , N, ω̇ = ν · r. (6.19)
j=1
For a set of adiabatic, isochoric reactions, one can show the extension of Eq. (5.435) is
PJ PN
dT j=1 rj ∆Uj i=1 ui ω̇i rT · ∆U uT · ω̇
=− =− =− =− , (6.24)
dt ρcv ρcv ρcv ρcv
where the energy change for a reaction ∆Uj is defined as the extension of Eq. (5.434):
N
X
∆Uj = ui νij , j = 1, . . . , J, ∆U = (uT · ν)T = ν T · u. (6.25)
i=1
Similarly for a set of adiabatic, isobaric reactions, one can show the extension of Eq. (5.450):
PJ PN
dT j=1 rj ∆Hj i=1 hi ω̇i
=− =− , (6.27)
dt ρcP ρcP
where the enthalpy change for a reaction ∆Hj is defined as the extension of Eq. (5.449):
N
X T
∆Hj = hi νij , j = 1, . . . , J, ∆H = (h · ν)T = ν T · h. (6.28)
i=1
T
rT · ∆H = rT · ν T · h = (ν · r)T · h = ω̇ T · h = h · ω̇. (6.29)
Moreover, the density and species concentration derivatives for an adiabatic, isobaric set can
be shown to be extensions of Eqs. (5.460, 5.464):
XJ XN
dρ hi
= M rj νij −1 , (6.30)
dt j=1 i=1
cP T
XJ
dρi ∆Hj
= rj νij + yi − ∆nj , (6.31)
dt j=1
c P T
where
N
X
∆nj = νkj . (6.32)
k=1
In a similar fashion to that shown for a single reaction, one can further sum over all
reactions and prove that mixture mass is conserved, element mass and number are conserved.
Example 6.1
Show that element mass and number are conserved for the multi-reaction formulation.
XJ
d
(φli ρi ) = φli νij rj , (6.35)
dt j=1
X N XN X J
d
(φli ρi ) = φli νij rj , (6.36)
i=1
dt i=1 j=1
N
! J X N
d X X
φli ρi = φli νij rj , (6.37)
dt i=1 j=1 i=1
| {z }
=ρl e
J
X N
X
dρl e
= rj φli νij , (6.38)
dt j=1 i=1
| {z }
=0
dρl e
= 0, l = 1, . . . , L, (6.39)
dt
d
(Ml ρl e ) = 0, l = 1, . . . , L, (6.40)
dt
dρl e
= 0, l = 1, . . . , L. (6.41)
dt
It is also straightforward to show that the mixture density is conserved for the multi-reaction, multi-
component mixture:
dρ
= 0. (6.42)
dt
The proof of the Clausius-Duhem relationship for the second law is an extension of the
single reaction result. Start with Eq. (5.465) and operate much as for a single reaction model.
N
1X
dS|U,V = − µ dni ≥ 0, (6.43)
T i=1 i
| {z }
irreversible entropy production
N
dS V X dni 1
= − µ ≥ 0, (6.44)
dt U,V T i=1 i dt V
N
V X dρi
= − µ ≥ 0, (6.45)
T i=1 i dt
N J
V X X
= − µ νij rj ≥ 0, (6.46)
T i=1 i j=1
N J
V XX
= − µ νij rj ≥ 0, (6.47)
T i=1 j=1 i
J N
V X X
= − rj µ νij ≥ 0, (6.48)
T j=1 i=1 i
J N N
! N
V X Y νij′ 1 Y νij X
= − kj ρ 1− ρ µi νij ≥ 0, (6.49)
T j=1 i=1 i Kc,j i=1 i i=1
J N N
! N
!!
V X Y ν′ 1 Y ν 1 Y νij
= − kj ρi ij 1− ρi ij RT ln ρ ≥ 0,
T j=1 i=1
Kc,j i=1
Kc,j i=1 i
(6.50)
J N N
! N
!
X Y ν′ 1 Y νij 1 Y νij
= −RV kj ρi ij 1− ρ ln ρ ≥ 0. (6.51)
j=1 i=1
Kc,j i=1 i Kc,j i=1 i
Note that Eq. (6.48) can also be written in terms of the affinities (see Eq. (6.10)) and reaction
progress variables (see Eq. (6.17) as
J
dS 1 X dζj
= αj ≥ 0. (6.52)
dt U,V T j=1 dt
Similar to the argument for a single reaction, if one defines extensions of Eqs. (5.488,
5.489) as
N
Y ′
νij
R′j = kj ρi , (6.53)
i=1
N
kj Y νij′′
R′′j = ρ , (6.54)
Kc,j i=1 i
then it is easy to show that
rj = R′j − R′′j , (6.55)
and we get the equivalent of Eq. (5.493):
J
X
dS R′j
= RV R′j − R′′j ln ≥ 0. (6.56)
dt U,V j=1
R′′j
Because kj (T ) > 0, R > 0, and V ≥ 0, and each term in the summation combines to
be positive semi-definite, one sees that the Clausius-Duhem inequality is guaranteed to be
satisfied for multi-component reactions.
N
X
φli (nio − ni ) = 0, l = 1, . . . , L. (6.57)
i=1
Here recall nio is the initial number of moles of species i in the mixture, and φli is the number
of moles of element l in species i. If one interprets nio − ni as −νij , the negative of the net
P
mole change, Eq. (6.57) becomes − N i=1 φli νij = 0, equivalent to Eq. (6.4).
One can now use the method of constrained optimization given by the method of Lagrange
multipliers to extremize G subject to the constraints of element conservation. The extremum
will be a minimum; this will not be proved, but it will be demonstrated. Define a set of L
Lagrange multipliers λl . Next define an augmented Gibbs free energy function G∗ , which is
simply G plus the product of the Lagrange multipliers and the constraints:
L
X N
X
∗
G =G+ λl φlk (nko − nk ). (6.58)
l=1 k=1
Now when the constraints are satisfied, one has G∗ = G, so assuming the constraints can be
satisfied, extremizing G is equivalent to extremizing G∗ . To extremize G∗ , take its differential
with respect to ni , with P , T and nj constant and set it to zero for each species:
L N
!
∂G∗ ∂G ∂ X X
= + λl φlk (nko − nk ) = 0, i = 1, . . . , N,
∂ni T,P,nj ∂ni T,P,nj ∂ni T,P,nj l=1 k=1
| {z }
=µi
(6.59)
L
X N
X ∂nk
= µi − λl φlk = 0, i = 1, . . . , N, (6.60)
∂ni T,P,nj
l=1 k=1
| {z }
δki
L
X N
X
= µi − λl φlk δki = 0, i = 1, . . . , N, (6.61)
l=1 k=1
XL
= µi − λl φli = 0, i = 1, . . . , N. (6.62)
l=1
Next, for an ideal gas, one can expand the chemical potential so as to get
L
X
Pi
µoT,i + RT ln − λl φli = 0, i = 1, . . . , N, (6.63)
Po
| {z } l=1
=µi
! L
ni P 1 X
o
µT,i + RT ln PN − λl φli = 0, i = 1, . . . , N. (6.64)
k=1 nk
Po
l=1
| {z }
=Pi
PN
Recalling that k=1 nk = n, in summary then, one has N + L equations
L
X
ni P
µoT,i + RT ln − λl φli = 0, i = 1, . . . , N, (6.65)
n Po l=1
N
X
φli (nio − ni ) = 0, l = 1, . . . , L. (6.66)
i=1
in N + L unknowns: ni , i = 1, . . . , N, λl , l = 1, . . . , L.
Example 6.2
Consider a previous example problem, see p. 157, in which
N2 + N2 ⇌ 2N + N2 . (6.67)
Take the reaction to be isothermal and isobaric with T = 6000 K and P = 100 kPa. Initially one has
1 kmole of N2 and 0 kmole of N. Use the extremization of Gibbs free energy to find the equilibrium
composition.
First find the chemical potentials at the reference pressure of each of the possible constituents.
o o o
µoT,i = goi = hi − T soi = h298,i + ∆hi − T soi . (6.68)
kJ
µoN2 = 0 + 205848 − (6000)(292.984) = −1552056 , (6.69)
kmole
kJ
µoN = 472680 + 124590 − (6000)(216.926) = −704286 . (6.70)
kmole
To each of these one must add
ni P
RT ln ,
nPo
to get the full chemical potential. Now P = Po = 100 kPa for this problem, so one only must consider
RT = 8.314(6000) = 49884 kJ/kmole. So, the chemical potentials are
n N2
µN2 = −1552056 + 49884 ln , (6.71)
n N + n N2
nN
µN = −704286 + 49884 ln . (6.72)
n N + n N2
Then one adds on the Lagrange multiplier and then considers element conservation to get the
following coupled set of nonlinear algebraic equations:
n N2
−1552056 + 49884 ln − 2λN = 0, (6.73)
n N + n N2
nN
−704286 + 49884 ln − λN = 0, (6.74)
n N + n N2
nN + 2nN2 = 2. (6.75)
Example 6.3
Consider a mixture of 2 kmole of H2 and 1 kmole of O2 at T = 3000 K and P = 100 kPa. Assuming
an isobaric and isothermal equilibration process with the products consisting of H2 , O2 , H2 O, OH, H,
and O, find the equilibrium concentrations. Consider the same mixture at T = 298 K and T = 1000 K.
The first task is to find the chemical potentials of each species at the reference pressure and
T = 3000 K. Here one can use the standard tables along with the general equation
o o o
µoT,i = goi = hi − T soi = h298,i + ∆hi − T soi . (6.79)
kJ
µoH2 = 0 + 88724 − 3000(202.989) = −520242 , (6.80)
kmole
kJ
µoO2 = 0 + 98013 − 3000(284.466) = −755385 , (6.81)
kmole
kJ
µoH2 O = −241826 + 126548 − 3000(286.504) = −974790 , (6.82)
kmole
kJ
µoOH = 38987 + 89585 − 3000(256.825) = −641903 , (6.83)
kmole
kJ
µoH = 217999 + 56161 − 3000(162.707) = −213961 , (6.84)
kmole
kJ
µoO = 249170 + 56574 − 3000(209.705) = −323371 . (6.85)
kmole
to get the full chemical potential. Now P = Po = 100 kPa for this problem, so one must only consider
RT = 8.314(3000) = 24942 kJ/kmole. So, the chemical potentials are
n H2
µ H2 = −520243 + 24942 ln , (6.86)
n H2 + n O 2 + nH2 O + nOH + nH + nO
n O2
µO 2 = −755385 + 24942 ln , (6.87)
n H2 + n O 2 + nH2 O + nOH + nH + nO
n H2 O
µ H2 O = −974790 + 24942 ln , (6.88)
n H2 + n O 2 + nH2 O + nOH + nH + nO
nOH
µOH = −641903 + 24942 ln , (6.89)
n H2 + n O 2 + nH2 O + nOH + nH + nO
nH
µH = −213961 + 24942 ln , (6.90)
n H2 + n O 2 + nH2 O + nOH + nH + nO
nO
µO = −323371 + 24942 ln . (6.91)
n H2 + n O 2 + nH2 O + nOH + nH + nO
Then one adds on the Lagrange multipliers and then considers element conservation to get the following
n H2
−520243 + 24942 ln − 2λH = 0, (6.92)
nH2 + nO2 + nH2 O + nOH + nH + nO
n O2
−755385 + 24942 ln − 2λO = 0, (6.93)
nH2 + nO2 + nH2 O + nOH + nH + nO
n H2 O
−974790 + 24942 ln − 2λH − λO = 0, (6.94)
nH2 + nO2 + nH2 O + nOH + nH + nO
nOH
−641903 + 24942 ln − λH − λO = 0, (6.95)
nH2 + nO2 + nH2 O + nOH + nH + nO
nH
−213961 + 24942 ln − λH = 0, (6.96)
nH2 + nO2 + nH2 O + nOH + nH + nO
nO
−323371 + 24942 ln − λO = 0, (6.97)
nH2 + nO2 + nH2 O + nOH + nH + nO
2nH2 + 2nH2 O + nOH + nH = 4, (6.98)
2nO2 + nH2 O + nOH + nO = 2. (6.99)
These non-linear algebraic equations can be solved numerically via a Newton-Raphson technique.
The equations are sensitive to the initial guess, and one can use ones intuition to help guide the
selection. For example, one might expect to have nH2 O somewhere near 2 kmole. Application of the
Newton-Raphson iteration yields
At this relatively high value of temperature, all species considered have a relatively major presence.
That is, there are no truly minor species.
Unless a very good guess is provided, it may be difficult to find a solution for this set of non-
linear equations. Straightforward algebra allows the equations to be recast in a form which sometimes
Then solve these considering ni , exp (λO /24942), and exp (λH /24942) as unknowns. The same result is
recovered, but a broader range of initial guesses converge to the correct solution.
One can verify that this choice extremizes G by direct computation; moreover, this will show
that the extremum is actually a minimum. In so doing, one must exercise care to see that element
conservation is retained. As an example, perturb the equilibrium solution above for nH2 and nH such
that
Leave all other species mole numbers the same. In this way, when ξ = 0, one has the original equilibrium
solution. For ξ 6= 0, the solution
PN moves off the equilibrium value in such a way that elements are
conserved. Then one has G = i=1 µi ni = G(ξ).
The difference G(ξ) − G(0) is plotted in Fig. 6.1. When ξ = 0, there is no deviation from the value
predicted by the Newton-Raphson iteration. Clearly when ξ = 0, G(ξ) − G(0), takes on a minimum
value, and so then does G(ξ). So the procedure works.
At the lower temperature, T = 298 K, application of the same procedure yields very different
results:
40
30
20
10
Figure 6.1: Gibbs free energy variation as mixture composition is varied maintaining element
conservation for mixture of H2 , O2 , H2 O, OH, H, and O at T = 3000 K, P = 100 kPa.
PNi=1 νij N
Y
Po −∆Goj νkj
Kc,j = exp = ρk , j = 1, . . . , J. (6.134)
RT RT k=1
With some effort it can be shown that not all of the J equations are linearly independent.
Moreover, they do not possess a unique solution. However, for closed systems, only one of
the solutions is physical, as will be shown in the following section. The others typically
involve non-physical, negative concentrations.
Nevertheless, Eqs. (6.134) are entirely consistent with the predictions of the N + L equa-
tions which arise from extremization of Gibbs free energy while enforcing element number
constraints. This can be shown by beginning with Eq. (6.64), rewritten in terms of molar
concentrations, and performing the following sequence of operations:
! L
ni /V P X
µoT,i + RT ln PN − λl φli = 0, i = 1, . . . , N, (6.135)
k=1 nk /V
Po
l=1
! L
ρiP X
µoT,i + RT ln PN − λl φli = 0, i = 1, . . . , N, (6.136)
k=1 ρ k
P o
l=1
X L
ρi P
µoT,i + RT ln − λl φli = 0, i = 1, . . . , N, (6.137)
ρ Po l=1
X L
o RT
µT,i + RT ln ρi − λl φli = 0, i = 1, . . . , N, (6.138)
Po l=1
XL
o RT
νij µT,i + νij RT ln ρi − νij λl φli = 0, i = 1, . . . , N,
Po l=1
j = 1, . . . , J, (6.139)
N
X N
X X N XL
RT
νij µoT,i + νij RT ln ρi − νij λl φli = 0, j = 1, . . . , J, (6.140)
Po
|i=1 {z } i=1 i=1 l=1
=∆Goj
N
X X L X N
RT
∆Goj + RT νij ln ρi − λl φli νij = 0, j = 1, . . . , J, (6.141)
i=1
Po l=1
|i=1 {z }
=0
N
X
RT
∆Goj + RT νij ln ρi = 0, j = 1, . . . , J. (6.142)
i=1
Po
Here, the stoichiometry for each reaction has been employed to remove the Lagrange multi-
In summary,
N
Y
ρi νij = Kc,j , j = 1, . . . , J, (6.148)
i=1
which is identical to Eq. (6.134), obtained by equilibrating each of the J reactions. Thus,
extremization of Gibbs free energy is consistent with equilibrating each of the J reactions.
Here, the N × N projection matrix P is symmetric, has norm of unity, has rank of N − L,
has N − L eigenvalues of value unity, and L eigenvalues of value zero. And, while application
of a general projection matrix to ν · r filters some of the information in ν · r, because the
N × (N − L) matrix D spans the same column space as the N × J matrix ν, no information
is lost in Eq. (6.157) relative to the original Eq. (6.154). Mathematically, one can say
P · ν = ν. (6.158)
1 dζ
= r. (6.159)
V dt
dρ 1 dζ
= ν· . (6.160)
dt V dt
Take as an initial condition that the reaction progress is zero at t = 0 and that there are an
appropriate set of initial conditions on the species concentrations ρ:
ζ = 0, t = 0, (6.161)
ρ = ρo , t = 0. (6.162)
Then, because ν is a constant, Eq. (6.160) is easily integrated. After applying the initial
conditions, Eq. (6.162), one gets
1
ρ = ρo + ν · ζ. (6.163)
V
Last, if J = (N − L), either approach yields the same number of equations, and is equally
concise.
Example 6.4
Consider the full time-dependency of a problem considered in a previous example in which the
equilibrium state was found; see Sec. 5.3.3.4. A closed, fixed, adiabatic volume contains at t = 0 s a
stoichiometric mixture of 2 kmole of H2 , 1 kmole of O2 , and 8 kmole of N2 at 100 kPa and 1000 K.
Find the reaction dynamics as the system proceeds from its initial state to its final state.
This problem requires a detailed numerical solution. Such a solution was performed by solving
Eq. (6.14) along with the associated calorically imperfect species state equations for a mixture of
eighteen interacting species: H2 , H, O, O2 , OH, H2 O, HO2 , H2 O2 , N, NH2 , NH3 , N2 H, NO, NO2 , N2 O,
HNO, and N2 . The equilibrium values were reported in a previous example.
0
H2
HO
10
O2 2 N2
H2
−5
10 hydrogen- O2 NO
−10 NO2
oxygen-related OH
10 N N2O
species H
O
−10
10 evolution −20
HO2 10 nitrogen
H2O2
HO
2
NO and nitrogen
−15
10 10
−30 oxides species
evolution
H2O2 a) b)
−20 −40
10 10
−6 −4 −2 −6 −4 −2
10 10 10 10 10 10
t (s) t (s)
3000 250
2500
200
P (kPa)
T (K)
2000
150
1500 induction
time
c) d)
1000 100
−6 −4 −2 −6 −4 −2
10 10 10 10 10 10
t (s) t (s)
Figure 6.2: Plot of a) ρH2 (t), ρH (t), ρO (t), ρO2 (t), ρOH (t), ρH2 O (t), ρHO2 (t), ρH2 O2 (t), b) ρN (t),
ρNO (t), ρNO2 (t), ρN2 O (t), ρN2 (t), c) T (t), and d) P (t) for adiabatic, isochoric combustion of
a mixture of 2H2 + O2 + 8N2 initially at T = 1000 K, P = 100 kPa.
The dynamics of the reaction process are reflected in Fig. 6.2. At early time, t < 4 × 10−4 s, the
pressure, temperature, and major reactant species concentrations (H2 , O2 , N2 ) are nearly constant.
However, the minor species, e.g. OH, NO, HO2 and the major product, H2 O, are undergoing very
rapid growth, albeit concentrations whose value remains small. In this period, the material is in what
is known as the induction period.
After a certain critical mass of minor species has accumulated, exothermic recombination of these
minor species to form the major product H2 O induces the temperature to rise, which accelerates further
the reaction rates. This is manifested in a thermal explosion. A common definition of the end of the
induction period is the induction time, t = ti , the time when dT /dt goes through a maximum. Here
one finds
ti = 4.53 × 10−4 s. (6.164)
At the end of the induction zone, there is a final relaxation to equilibrium. Often the induction time is
called the ignition delay time.
Here we give further details of kinetics. These notes are also used to introduce a separate
combustion course and have some overlap with previous chapters.
Let us consider the reaction of N molecular chemical species composed of L elements via
J chemical reactions. Let us assume the gas is an ideal mixture of ideal gases that satisfies
Dalton’s law of partial pressures. The reaction will be considered to be driven by molecular
collisions. We will not model individual collisions, but instead attempt to capture their
collective effect.
An example of a model of such a reaction is listed in Table 7.1. There we find a N = 9
species, J = 37 step irreversible reaction mechanism for an L = 3 hydrogen-oxygen-argon
mixture from Maas and Warnatz,1 with corrected fH2 from Maas and Pope.2 The model has
also been utilized by Fedkiw, et al.3 We need not worry yet about fH2 , which is known as
a collision efficiency factor. The one-sided arrows indicate that each individual reaction is
considered to be irreversible. For nearly each reaction, a separate reverse reaction is listed;
thus, pairs of irreversible reactions can be considered to model reversible reactions.
In this model a set of elementary reactions are hypothesized. For the j th reaction we have
the collision frequency factor aj , the temperature-dependency exponent βj and the activation
energy E j . These will be explained in short order. Other common forms exist. Often
reactions systems are described as being composed of reversible reactions. Such reactions are
usually notated by two sided arrows. One such system is reported by Powers and Paolucci4
reported here in Table 7.2. Both overall models are complicated.
1
Maas, U., and Warnatz, J., 1988, “Ignition Processes in Hydrogen-Oxygen Mixtures,” Combustion and
Flame, 74(1): 53-69.
2
Maas, U., and Pope, S. B., 1992, “Simplifying Chemical Kinetics: Intrinsic Low-Dimensional Manifolds
in Composition Space,” Combustion and Flame, 88(3-4): 239-264.
3
Fedkiw, R. P., Merriman, B., and Osher, S., 1997, “High Accuracy Numerical Methods for Thermally
Perfect Gas Flows with Chemistry,” Journal of Computational Physics, 132(2): 175-190.
4
Powers, J. M., and Paolucci, S., 2005, “Accurate Spatial Resolution Estimates for Reactive Supersonic
Flow with Detailed Chemistry,” AIAA Journal, 43(5): 1088-1099.
239
240 CHAPTER 7. KINETICS IN SOME MORE DETAIL
PN
(mol/cm3 )(
′
1−νM,j − ′
i=1 νij ) kJ
j Reaction aj s Kβj
βj Ej mole
Table 7.1: Third body collision efficiencies with M are fH2 = 1.00, fO2 = 0.35, and
fH2 O = 6.5.
CC BY-NC-ND. 12 March 2024, J. M. Powers.
241
PN
(mol/cm3 )(
′
1−νM,j − ′
i=1 νij ) cal
j Reaction aj s K βj βj Ej mole
13
1 H2 + O2 ⇋ OH + OH 1.70 × 10 0.00 47780
2 OH + H2 ⇋ H2 O + H 1.17 × 109 1.30 3626
3 H + O2 ⇋ OH + O 5.13 × 1016 −0.82 16507
4 O + H2 ⇋ OH + H 1.80 × 1010 1.00 8826
5 H + O2 + M ⇋ HO2 + M 2.10 × 1018 −1.00 0
6 H + O2 + O2 ⇋ HO2 + O2 6.70 × 1019 −1.42 0
7 H + O2 + N2 ⇋ HO2 + N2 6.70 × 1019 −1.42 0
8 OH + HO2 ⇋ H2 O + O2 5.00 × 1013 0.00 1000
9 H + HO2 ⇋ OH + OH 2.50 × 1014 0.00 1900
10 O + HO2 ⇋ O2 + OH 4.80 × 1013 0.00 1000
11 OH + OH ⇋ O + H2 O 6.00 × 108 1.30 0
12 H2 + M ⇋ H + H + M 2.23 × 1012 0.50 92600
13 O2 + M ⇋ O + O + M 1.85 × 1011 0.50 95560
14 H + OH + M ⇋ H2 O + M 7.50 × 1023 −2.60 0
15 H + HO2 ⇋ H2 + O2 2.50 × 1013 0.00 700
16 HO2 + HO2 ⇋ H2 O2 + O2 2.00 × 1012 0.00 0
17 H2 O2 + M ⇋ OH + OH + M 1.30 × 1017 0.00 45500
18 H2 O2 + H ⇋ HO2 + H2 1.60 × 1012 0.00 3800
19 H2 O2 + OH ⇋ H2 O + HO2 1.00 × 1013 0.00 1800
Table 7.2: Nine species, nineteen step reversible reaction mechanism for an H2 /O2 /N2 mix-
ture. Third body collision efficiencies with M are f5 (H2 O) = 21, f5 (H2 ) = 3.3, f12 (H2 O) = 6,
f12 (H) = 2, f12 (H2 ) = 3, f14 (H2 O) = 20.
7.1.1 O − O2 dissociation
One of the simplest physical examples is provided by the dissociation of O2 into its atomic
component O.
7.1.1.1.1 Mathematical model The reactions describe oxygen dissociation and recom-
bination in a pair of irreversible reactions:
13 : O + O + M → O2 + M, (7.1)
14 : O2 + M → O + O + M, (7.2)
with
−2
17 mole K kJ
a13 = 2.90 × 10 3
, β13 = −1.00, E 13 = 0 (7.3)
cm s mole
−1
mole K kJ
a14 = 6.81 × 1018 3
, β14 = −1.00, E 14 = 496.41 . (7.4)
cm s mole
The irreversibility is indicated by the one-sided arrow. Though they participate in the overall
hydrogen oxidation problem, these two reactions are in fact self-contained as well. So let us
just consider that we have only oxygen in our box with N = 2 species, O2 and O, J = 2
reactions (those being 13 and 14), and L = 1 element, that being O.
Recall that in the cgs system, common in thermochemistry, that 1 erg = 1 dyne cm =
10−7 J = 10−10 kJ. Recall also that the cgs unit of force is the dyne and that 1 dyne =
1 g cm/s2 = 10−5 N. So for cgs we have
10
erg kJ 10 erg erg
E 13 = 0 , E 14 = 496.41 = 4.96 × 1012 . (7.5)
mole mole kJ mole
5
W. G. Vincenti and C. H. Kruger, 1965, Introduction to Physical Gas Dynamics, Wiley, New York.
The standard model for chemical reaction induces the following two ordinary differential
equations for the evolution of O and O2 molar concentrations:
dρO β13 −E 13 β14 −E 14
= −2 a13 T exp ρO ρO ρM +2 a14 T exp ρ O2 ρ M , (7.6)
dt RT RT
| {z } | {z }
=k13 (T ) =k14 (T )
| {z } | {z }
=r13 =r14
dρO2 β13 −E 13 β14 −E 14
= a13 T exp ρO ρO ρM − a14 T exp ρ O2 ρ M . (7.7)
dt RT RT
| {z } | {z }
=k13 (T ) =k14 (T )
| {z } | {z }
=r13 =r14
Here we use the notation ρi as the molar concentration of species i. Also a common usage for
molar concentration is given by square brackets, e.g. ρO2 = [O2 ]. The symbol M represents
an arbitrary third body and is an inert participant in the reaction. We also use the common
notation of a temperature-dependent portion of the reaction rate for reaction j, kj (T ), where
βj Ej
kj (T ) = aj T exp . (7.8)
RT
The reaction rates for reactions 13 and 14 are defined as
r13 = k13 ρO ρO ρM , (7.9)
r14 = k14 ρO2 ρM . (7.10)
We will give details of how to generalize this form later. The system Eq. (7.6-7.7) can be
written simply as
dρO
= −2r13 + 2r14 , (7.11)
dt
dρO2
= r13 − r14 . (7.12)
dt
Even more simply, in vector form, Eqs. (7.11-7.12) can be written as
dρ
= ν · r. (7.13)
dt
Here we have taken
ρO
ρ = , (7.14)
ρ O2
−2 2
ν = , (7.15)
1 −1
r13
r = . (7.16)
r14
Equations (7.24-7.25) with Eqs. (7.22) represent two non-linear ordinary differential equa-
tions with initial conditions in two unknowns ρO and ρO2 . We seek the behavior of these two
species concentrations as a function of time.
Systems of non-linear equations are generally difficult to integrate analytically and gen-
erally require numerical solution. Before embarking on a numerical solution, we simplify as
much as we can. Note that
dρO dρ
+ 2 O2 = 0, (7.26)
dt dt
d
ρO + 2ρO2 = 0. (7.27)
dt
We can integrate and apply the initial conditions (7.22) to get
ρO + 2ρO2 = b
ρO + 2b
ρO2 = constant. (7.28)
The fact that this algebraic constraint exists for all time is a consequence of the conservation
of mass of each O element. It can also be thought of as the conservation of number of
O atoms. Such notions always hold for chemical reactions. They do not hold for nuclear
reactions.
Standard linear algebra provides a robust way to find the constraint of Eq. (7.28). We
can use elementary row operations to cast Eq. (7.16) into a row-echelon form. Here our goal
is to get a linear combination which on the right side has an upper triangular form. To
achieve this add twice the second equation with the first to form a new equation to replace
the second equation. This gives
d ρO −2 2 r13
= . (7.29)
dt ρO + 2ρO2 0 0 r14
Obviously the second equation is one we obtained earlier, d/dt(ρO + 2ρO2 ) = 0, and this
induces our algebraic constraint. We also note the system can be recast as
1 0 d ρO −2 2 r13
= . (7.30)
1 2 dt ρO2 0 0 r14
This is of the matrix form
dρ
L−1 · P · = U · r. (7.31)
dt
Here L and L−1 are N × N lower triangular matrices of full rank N, and thus invertible.
The matrix U is upper triangular of dimension N × J and with the same rank as ν, R ≥ L.
The matrix P is a permutation matrix of dimension N × N. It is never singular and thus
always invertable. It is used to effect possible row exchanges to achieve the desired form;
often row exchanges are not necessary, in which case P = I, the N × N identity matrix.
Equation (7.31) can be manipulated to form the original equation via
dρ −1
=P
| · L · U} ·r.
{z (7.32)
dt
=ν
This gives the dependent variables in terms of a smaller number of transformed dependent
variables in a way which satisfies the linear constraints. In vector form, the equation becomes
b + D · ξ.
ρ=ρ (7.35)
Here D is a full rank matrix which spans the same column space as does ν. Note that ν
may or may not be full rank. Because D spans the same column space as does ν, we must
also have in general
φ · D = 0. (7.36)
We also note that the term exp(−E j /RT ) is a modulating factor to the dynamics. Let
us see how this behaves for high and low temperatures. First for low temperature, we have
−E j
lim exp = 0. (7.38)
T →0 RT
At high temperature, we have
−E j
lim exp = 1. (7.39)
T →∞ RT
And lastly, at intermediate temperature, we have
−E j Ej
exp ∼ O(1) when T =O . (7.40)
RT R
A sketch of this modulating factor is given in Figure 7.1. Note
• for small T , the modulation is extreme, and the reaction rate is very small,
• for T ∼ E j /R, the reaction rate is extremely sensitive to temperature, and
• for T → ∞, the modulation is unity, and the reaction rate is limited only by molecular
collision frequency.
Now ρO and ρO2 represent molar concentrations which have standard units of mole/cm3 .
So the reaction rates
dρO dρO2
and
dt dt
3
have units of mole/cm /s.
After conversion of E j from kJ/mole to erg/mole we find the units of the argument of
the exponential to be dimensionless. That is
Ej erg mole K 1
= ⇒ dimensionless. (7.41)
RT mole erg K
Here the brackets denote the units of a quantity, and not molar concentration. Let us get
units for the collision frequency factor of reaction 13, a13 . We know the units of the rate
(mole/cm3 /s). Reaction 13 involves three molar species. Because β13 = −1, it also has an
extra temperature dependency. The exponential of a unitless number is unitless, so we need
not worry about that. For units to match, we must have
mole mole mole mole
s = [a13 ] K−1 . (7.42)
cm3 cm3 cm3 cm3
0.00150
O2
0.00100
0.00070
0.00050
Figure 7.2: Molar concentrations versus time for oxygen dissociation problem.
dρO
= −(1.16 × 1014 )ρ2O (ρO + ρO2 ) + (1.77548 × 1010 )ρO2 (ρO + ρO2 ), (7.50)
dt
dρO2
= (5.80 × 1013 )ρ2O (ρO + ρO2 ) − (8.8774 × 109 )ρO2 (ρO + ρO2 ), (7.51)
dt
mole
ρO (0) = 0.001 , (7.52)
cm3
mole
ρO2 (0) = 0.001 . (7.53)
cm3
These non-linear ordinary differential equations are in a standard form for a wide variety
of numerical software tools. Solution of such equations are not the topic of these notes.
r
1.5 x 10-16
1.0 x 10-16
5.0 x 10-17
t (s)
10-11 10-10 10-9 10-8 10-7 10-6
Figure 7.3: Dimensionless residual numerical error r in satisfying the element conservation
constraint in the oxygen dissociation example.
We can check how well the numerical solution satisfied the algebraic constraint of element
conservation by plotting the dimensionless residual error r
ρO + 2ρO2 − b
ρO − 2b
ρO2
r= , (7.56)
b
ρO + 2b
ρO 2
as a function of time. If the constraint is exactly satisfied, we will have r = 0. Any non-zero
r will be related to the numerical method we have chosen. It may contain roundoff error
and have a sporadic nature. A plot of r(t) is given in Figure 7.3. Clearly the error is small,
and has the character of a roundoff error. In fact it is possible to drive r to be smaller by
controlling the error tolerance in the numerical method.
7.1.1.1.2.2 Pressure versus time We can use the ideal gas law to calculate the
pressure. Recall that the ideal gas law for molecular species i is
Pi V = ni RT. (7.57)
Here Pi is the partial pressure of molecular species i, and ni is the number of moles of
molecular species i. We also have
ni
Pi = RT. (7.58)
V
By our definition of molecular species concentration that
ni
ρi = . (7.59)
V
So we also have the ideal gas law as
Pi = ρi RT. (7.60)
Now in the Dalton mixture model, all species share the same T and V . So the mixture
temperature and volume are the same for each species Vi = V , Ti = T . But the mixture
pressure is taken to be the sum of the partial pressures:
N
X
P = Pi . (7.61)
i=1
P (t = 0) = RT (b
ρO + b
ρO2 ), (7.64)
erg mole mole
= 8.31441 × 107 (5000 K) 0.001 3
+ 0.001 , (7.65)
mole K cm cm3
dyne
= 8.31441 × 108 , (7.66)
cm2
= 8.31441 × 102 bar. (7.67)
This pressure is over 800 atmospheres. It is actually a little too high for good experimental
correlation with the underlying data, but we will neglect that for this exercise.
At the equilibrium state we have more O2 and less O. And we have a different number
of molecules, so we expect the pressure to be different. At equilibrium, the pressure is
The pressure has dropped because much of the O has recombined to form O2 . Thus there
are fewer molecules at equilibrium. The temperature and volume have remained the same.
A plot of P (t) is given in Figure 7.4.
8 x 108
7 x 108
10-11 10-10 10-9 10-8 10-7 10-6 t (s)
7.1.1.1.2.3 Dynamical system form Now Eqs. (7.50-7.51) are of the standard form
for an autonomous dynamical system:
dy
= f(y). (7.72)
dt
Here y is the vector of state variables (ρO , ρO2 )T . And f is an algebraic function of the state
variables. For the isothermal system, the algebraic function is in fact a polynomial.
Equilibrium
The dynamical system is in equilibrium when
f(y) = 0. (7.73)
This non-linear set of algebraic equations can be difficult to solve for large systems. For
common chemical kinetics systems, such as the one we are dealing with, there is a guarantee
of a unique equilibrium for which all state variables are physical. There are certainly other
equilibria for which at least one of the state variables is non-physical. Such equilibria can
be quite mathematically complicated.
Solving Eq. (7.73) for our oxygen dissociation problem gives us symbolically from Eq. (7.6-
7.7)
−E 13 e e e β13 −E 14
−2a13 exp ρO ρO ρM T + 2a14 exp ρeO2 ρeM T β14 = 0, (7.74)
RT RT
β13 −E 13 e e e β14 −E 14
a13 T exp ρO ρO ρM − a14 T exp ρeO2 ρeM = 0. (7.75)
RT RT
We notice that ρeM cancels. This so-called third body will in fact never affect the equilib-
rium state. It will however influence the dynamics. Removing ρeM and slightly rearranging
Eqs. (7.74-7.75) gives
β13 −E 13 e e β14 −E 14
a13 T exp ρO ρO = a14 T exp ρeO2 , (7.76)
RT RT
β13 −E 13 e e β14 −E 14
a13 T exp ρO ρO = a14 T exp ρeO2 . (7.77)
RT RT
These are the same equations! So we really have two unknowns for the equilibrium state ρeO
and ρeO2 but seemingly only one equation. Rearranging either Eq. (7.76) or (7.77) gives the
result
β14 −E 14
e e
ρO ρO a14 T exp RT
e = = K(T ). (7.78)
ρO2 a T β13 exp −E 13
13 RT
That is, for the net reaction (excluding the inert third body), O2 → O+O, at equilibrium the
product of the concentrations of the products divided by the product of the concentrations
of the reactants is a function of temperature T . And for constant T , this is the so-called
equilibrium constant. This is a famous result from basic chemistry. It is actually not complete
yet, as we have not taken advantage of a connection with thermodynamics. But for now, it
will suffice.
We still have a problem: Eq. (7.78) is still one equation for two unknowns. We solve
this be recalling we have not yet taken advantage of our algebraic constraint of element
conservation, Eq. (7.28). Let us use the equation to eliminate ρeO2 in favor of ρeO :
1 b
ρeO2 = ρO − ρeO + b
ρO2 . (7.79)
2
So Eq. (7.76) reduces to
β13 −E 13 e e β14 −E 14 1 b e b
a13 T exp ρO ρO = a14 T exp ρ − ρO + ρO2 . (7.80)
RT RT 2 O
| {z }
=ρeO
2
Equation (7.80) is one algebraic equation in one unknown. Its solution gives the equilibrium
value ρeO . It is a quadratic equation for ρeO . Of its two roots, one will be physical. We note
that the equilibrium state will be a function of the initial conditions. Mathematically this
is because our system is really best posed as a system of differential-algebraic equations.
Systems which are purely differential equations will have equilibria which are independent
of their initial conditions. Most of the literature of mathematical physics focuses on such
systems of those. One of the foundational complications of chemical dynamics is that the
equilibria is a function of the initial conditions, and this renders many common mathematical
300 000
200 000
100 000
-200000
notions from traditional dynamic system theory to be invalid Fortunately, after one accounts
for the linear constraints of element conservation, one can return to classical notions from
traditional dynamic systems theory.
Consider the dynamics of Eq. (7.24) for the evolution of ρO . Equilibrating the right
hand side of this equation, gives Eq. (7.74). Eliminating ρM and then ρO2 in Eq. (7.74) then
substituting in numerical parameters gives the cubic algebraic equation
This equation is cubic because we did not remove the effect of ρM . This will not affect
the equilibrium, but will affect the dynamics. We can get an idea of where the roots are by
plotting f (ρO ) as seen in Figure 7.5. Zero crossings of f (ρO ) in Figure 7.5 represent equilibria
of the system, ρeO , f (ρeO ) = 0. The cubic equation has three roots
mole
ρeO = −0.003 , non-physical, (7.82)
cm3
mole
ρeO = −0.000518944 , non-physical, (7.83)
cm3
mole
ρeO = 0.000442414 , physical. (7.84)
cm3
The physical root found by our algebraic analysis is identical to that which was identified
by our numerical integration of the ordinary differential equations of reaction kinetics.
Stability of equilibria
We can get a simple estimate of the stability of the equilibria by considering the slope of
f near f = 0. Our dynamic system is of the form
dρO
= f (ρO ). (7.85)
dt
• Near the first non-physical root at ρeO = −0.003, a positive perturbation from equi-
librium induces f < 0, which induces dρO /dt < 0, so ρO returns to its equilibrium.
Similarly, a negative perturbation from equilibrium induces dρO /dt > 0, so the system
returns to equilibrium. This non-physical equilibrium point is stable. Stability does
not imply physicality!
• Perform the same exercise for the non-physical root at ρeO = −0.000518944. We find
this root is unstable.
• Perform the same exercise for the physical root at ρeO = 0.000442414. We find this
root is stable.
In general if f crosses zero with a positive slope, the equilibrium is unstable. Otherwise, it
is stable.
Consider a formal Taylor series expansion of Eq. (7.85) in the neighborhood of an equi-
librium point ρ3O :
d df
(ρO − ρeO ) = f (ρeO ) + (ρO − ρeO ) + . . . (7.86)
dt | {z } dρO ρO =ρeO
=0
7.1.1.1.3 Effect of temperature Let us perform four case studies to see the effect of
T on the system’s equilibria and it dynamics near equilibrium.
• T = 3000 K. Here we have significantly reduced the temperature, but it is still higher
than typically found in ordinary combustion engineering environments. Here we find
mole
ρeO = 8.9371 × 10−6 , (7.92)
cm3
τ = 1.92059 × 10−7 s. (7.93)
The O concentration is effectively zero at room temperature, and the relaxation time
is effectively infinite. As the oldest star in our galaxy has an age of 4.4 × 1017 s, we see
that at this temperature, our mathematical model cannot be experimentally validated,
so it loses its meaning. At such a low temperature, the theory becomes qualitatively
correct, but not quantitatively predictive.
At this high temperature, O become preferred over O2 , and the time scales of reaction
become extremely small, under a nanosecond.
7.1.1.2.1.1 Kinetics For the reversible O −O2 reaction, let us only consider reaction
13 from Table 7.2 for which
13 : O2 + M ⇌ O + O + M. (7.100)
Units of cal are common in chemistry, but we need to convert to erg, which is achieved via
7
cal J 10 erg erg
E 13 = 95560 4.186 = 4.00014 × 1012 . (7.102)
mole cal J mole
Here we have used equivalent definitions for k13 (T ) and r13 , so that Eqs. (7.103-7.104) can
be written compactly as
dρO
= 2r13 , (7.105)
dt
dρO2
= −r13 . (7.106)
dt
ρO + 2ρO2 = b
ρO + 2b
ρO2 = constant. (7.117)
This gives the dependent variables in terms of a smaller number of transformed dependent
variables in a way which satisfies the linear constraints. In vector form, the equation becomes
b + D · ξ.
ρ=ρ (7.119)
Once again φ · D = 0.
Here Po = 1.01326 × 106 dyne/cm2 = 1 atm is the reference pressure. The net change of
Gibbs free energy at the reference pressure for reaction 13, ∆Go13 is defined as
We further recall that the Gibbs free energy for species i at the reference pressure is defined
in terms of the enthalpy and entropy as
o
g oi = hi − T soi . (7.122)
o
It is common to find hi and soi in thermodynamic tables tabulated as functions of T .
We further note that both Eqs. (7.103) and (7.104) are in equilibrium when
1 e e e
ρeO2 ρeM = ρ ρ ρ . (7.123)
Kc,13 O O M
If Kc,13 > 1, the products are preferred. If Kc,13 < 1, the reactants are preferred.
Now, Kc,13 is a function of T only, so it is known. But Eq. (7.124) once again is one
equation in two unknowns. We can use the element conservation constraint, Eq. (7.117) to
reduce to one equation and one unknown, valid at equilibrium:
ρeO ρeO
Kc,13 = (7.125)
b
ρO2 + 21 (b
ρO − ρeO )
Using the element constraint, Eq. (7.117), we can recast the dynamics of our system by
modifying Eq. (7.103) into one equation in one unknown:
dρO −E 13
= 2a13 T β13 exp
dt RT
1 b 1 b 1 1 b
×(bρ
2 2
O + ( ρO − ρO )) (bρO + ( ρ O − ρO ) + ρO ) − ρO ρO (bρ O + ( ρO − ρ O ) + ρO ) ,
| {z }|
2
2 {z } K c,13 |
2
2 {z }
=ρO2 =ρM =ρM
(7.126)
7.1.1.2.2 Example calculation Let us consider the same example as the previous sec-
tion with T = 5000 K. We need numbers for all of the parameters of Eq. (7.126). For O, we
find at T = 5000 K that
o erg
hO = 3.48382 × 1012 , (7.127)
mole
erg
soO = 2.20458 × 109 . (7.128)
mole K
So
erg erg
g oO = 3.48382 × 10 12
− (5000 K) 2.20458 × 10 9
,
mole mole K
erg
= −7.53908 × 1012 . (7.129)
mole
For O2 , we find at T = 5000 K that
o erg
hO2 = 1.80749 × 1012 , (7.130)
mole
erg
soO2 = 3.05406 × 109 . (7.131)
mole K
So
erg erg
g oO2 = 1.80749 × 1012 − (5000 K) 3.05406 × 109
mole mole K
13 erg
= −1.34628 × 10 . (7.132)
mole
0.00150
O2
0.00100
0.00070
0.00050
O
0.00030
t (s)
10-11 10-9 10-7 10-5 10-3
Figure 7.6: Plot of ρO (t) and ρO2 (t) for oxygen dissociation with reversible reaction.
40 000
30 000
20 000
10 000
-20000
Figure 7.7: Plot of f (ρO ) versus ρO for oxygen dissociation with reversible reaction.
Again, A is an arbitrary constant. Obviously the equilibrium is stable. Moreover the time
constant of relaxation to equilibrium is
1
τ= = 4.77156 × 10−8 s. (7.146)
2.09575 × 107
This is consistent with the time scale to equilibrium which comes from integrating the full
equation.
1: N + NO ⇌ N2 + O, (7.147)
2: N + O2 ⇌ NO + O. (7.148)
similar to our results for O2 dissociation, N2 and O2 are preferred at low temperature. As
the temperature rises N and O begin to appear, and it is possible when they are mixed for
NO to appear as a product.
7.1.2.1.1 Standard model form Here we have the reaction of N = 5 molecular species
with
ρNO
ρN
ρ=
ρ N2 . (7.149)
ρO
ρ O2
The first row of φ is for the N atom; the second row is for the O atom.
6
Yakov Borisovich Zel’dovich, 1915-1987, prolific Soviet physicist and father of thermonuclear weapons.
Here, we have
β1 Ta,1
k1 = a1 T exp − , (7.153)
T
β2 Ta,2
k2 = a2 T exp − . (7.154)
T
Here the matrix ν has dimension N × J which is 5 × 2. The model is of our general form
dρ
= ν · r. (7.156)
dt
Our stoichiometric constraint on element conservation for each reaction φ · ν = 0 holds
here:
−1 1
−1 −1
1 1 2 0 0
0 0
φ·ν = 1 0 = . (7.157)
1 0 0 1 2 1 0 0
1
0 −1
We get 4 zeros because there are 2 reactions each with 2 element constraints.
7.1.2.1.2 Reduced form Here we describe non-traditional, but useful reductions, using
standard techniques from linear algebra to bring the model equations into a reduced form in
which all of the linear constraints have been explicitly removed.
Let us perform a series of row operations to find all of the linear dependencies. Our aim
is to convert the ν matrix into an upper triangular form. The lower left corner of ν already
has a zero, so there is no need to worry about it. Let us add the first and fourth equations
to eliminate the 1 in the 4, 1 slot. This gives
ρNO −1 1
ρN −1 −1
d
ρN
= 1
r1
0 r2 . (7.158)
dt 2
ρNO + ρO 0 2
ρ O2 0 −1
Next, add the first and third equations to get
ρNO −1 1
ρN −1 −1
d
ρNO + ρN = 0
r1
2 1 r2 . (7.159)
dt
ρNO + ρO 0 2
ρ O2 0 −1
Now multiply the first equation by −1 and add it to the second to get
ρNO −1 1
−ρNO + ρN 0 −2
d
ρNO + ρN = 0
r1
2 1 r2 . (7.160)
dt
ρNO + ρO 0 2
ρ O2 0 −1
Next multiply the fifth equation by −2 and add it to the second to get
ρNO −1 1
−ρNO + ρN 0 −2
d
= 0
r1
ρNO + ρN2 1 r2 . (7.161)
dt
ρNO + ρO 0 2
−ρNO + ρN − 2ρO2 0 0
Next add the second and fourth equations to get
ρNO −1 1
−ρNO + ρN 0 −2
d
= 0
r1
ρNO + ρN2 1 r2 . (7.162)
dt
ρN + ρO 0 0
−ρNO + ρN − 2ρO2 0 0
Next multiply the third equation by 2 and add it to the second to get
ρNO −1 1
−ρNO + ρN 0 −2
d
ρNO + ρN + 2ρN = 0
r1
2 0 r2 . (7.163)
dt
ρN + ρO 0 0
−ρNO + ρN − 2ρO2 0 0
A way to think of this type of row echelon form is that it defines two free variables, those
associated with the non-zero pivots of U: ρNO and ρN . The remain three variables ρN2 , ρO
and ρO2 are bound variables which can be expressed in terms of the free variables.
The last three of the ordinary differential equations are homogeneous and can be inte-
grated to form
The constants C1 , C2 and C3 are determined from the initial conditions on all five state
variables. In matrix form, we can say
ρNO
1 1 2 0 0 ρN
C1
0 1 0 1 0 ρN = C2 . (7.168)
2
−1 1 0 0 −2 ρO C3
ρO2
Considering the free variables, ρNO and ρN , to be known, we move them to the right side
to get
2 0 0 ρN2 C1 − ρNO − ρN
0 1 0 ρO = C2 − ρN . (7.169)
0 0 −2 ρO2 C3 + ρNO − ρN
Solving, for the bound variables, we find
1 1 1
ρ N2 2
C 1 − ρ
2 NO
− ρ
2 N
ρO = C2 − ρN . (7.170)
1 1 1
ρ O2 − 2 C3 − 2 ρNO + 2 ρN
We can rewrite this as
1 1
ρ N2 C
2 1
− 2 − 12
ρO = C2 + 0 −1 ρNO . (7.171)
ρN
ρ O2 − 12 C3 − 21 12
We can get a more elegant form by defining ξNO = ρNO and ξN = ρN . Thus we can say our
state variables have the form
ρNO 0 1 0
ρN 0 0 1
1 ξNO
ρN = 1 C1 + − 1 −2 . (7.172)
2 2 2 ξN
ρ O C2 0 −1
ρO2 − 21 C3 − 21 1
2
ρO2 b
ρO2 − 21 12 =ξ
| {z } | {z } | {z }
=ρ b =D
=ρ
Here the matrix D is of dimension N × R, which here is 5 × 2. It spans the same column
space as does the N × J matrix ν which is of rank R. Here in fact R = J = 2, so D has the
same dimension as ν. In general it will not. If c1 and c2 are the column vectors of D, we
see that −c1 − c2 forms the first column vector of ν and c1 − c2 forms the second column
vector of ν. Note that φ · D = 0:
1 0
0 1
1 1 2 0 0 − − =
1 1
0 0
φ·D= . (7.175)
1 0 0 1 2 2
0 −1
2 0 0
− 21 12
Equations (7.165-7.167) can also be linearly combined in a way which has strong phys-
ical relevance. We rewrite the system as three equations in which the first is identical to
Eq. (7.165); the second is the difference of Eqs. (7.166) and (7.167); and the third is half of
Eq. (7.165) minus half of Eq. (7.167) plus Eq. (7.166):
Equation (7.176) insists that the number of nitrogen elements be constant; Eq. (7.177)
demands the number of oxygen elements be constant; and Eq. (7.178) requires the number
of moles of molecular species be constant. For general reactions, including the earlier studied
oxygen dissociation problem, the number of moles of molecular species will not be constant.
Here because each reaction considered has two molecules reacting to form two molecules, we
are guaranteed the number of moles will be constant. Hence, we get an additional linear
constraint beyond the two for element conservation. Because our reaction is isothermal,
isochoric and mole-preserving, it will also be isobaric.
We will also need thermodynamic data. The data here will be taken from the Chemkin
database.8 Thermodynamic data for common materials is also found in most thermodynamic
texts. For our system at 6000 K, we find
erg
g oNO = −1.58757 × 1013 , (7.186)
mole
erg
g oN = −7.04286 × 1012 , (7.187)
mole
erg
g oN2 = −1.55206 × 1013 , (7.188)
mole
erg
g oO = −9.77148 × 1012 , (7.189)
mole
erg
g oO2 = −1.65653 × 1013 . (7.190)
mole
Again, omitting details, we find the two differential equations governing the evolution of
8
R. J. Kee, et al., 2000, “The Chemkin Thermodynamic Data Base,” part of the Chemkin Collection
Release 3.6, Reaction Design, San Diego, CA.
10 -5
10 -6
NO
10 -7
t (s)
10 -10 10 -9 10 -8 10 -7 10 -6 10 -5
Evaluation of Eqs. (7.214-7.215) near the physical root, root 3, yields the system
d ρNO − 7.336 × 10−7 −2.129 × 106 −4.155 × 105 ρNO − 7.336 × 10−7
= .
dt ρN − 3.708 × 10−8 2.111 × 105 −3.144 × 107 ρN − 3.708 × 10−8
| {z }
=J= ∂ρ |
∂f
e
(7.216)
It is the eigenvalues of the Jacobian9 matrix J that give the time scales of evolution of the
concentrations as well as determine the stability of the local equilibrium point. Recall that
we can usually decompose square matrices via the diagonalization
J = S · Λ · S−1 . (7.218)
Here, S is the matrix whose columns are composed of the right eigenvectors of J, and Λ is
the diagonal matrix whose diagonal is populated by the eigenvalues of J. For some matrices
9
after Carl Gustav Jacob Jacobi, 1804-1851, German mathematician.
(typically not those encountered after our removal of linear dependencies), diagonalization
is not possible, and one must resort to the so-called near-diagonal Jordan form. This will
not be relevant to our discussion, but could be easily handled if necessary. We also recall the
eigenvector matrix and eigenvalue matrix are defined by the standard eigenvalue problem
J · S = S · Λ. (7.219)
We also recall that the components λ of Λ are found by solving the characteristic polynomial
which arises from the equation
S · z ≡ ρ − ρe , (7.221)
and using the decomposition Eq. (7.218), Eq. (7.217) can be rewritten to form
d
(S · z) = |S · Λ{z· S−1} · (S · z), (7.222)
dt | {z }
J
ρ−ρe
dz
S· = S · Λ · z, (7.223)
dt
dz
S−1 · S · = S−1 · S · Λ · z, (7.224)
dt
dz
= Λ · z. (7.225)
dt
Eq. (7.225) reduces to the diagonal form
dz
= Λ · z. (7.226)
dt
This has solution for each component of z of
Here, our matrix J, see Eq. (7.216), has two real, negative eigenvalues in the neighborhood
of the physical root 3:
1
λ1 = −3.143 × 107 , (7.230)
s
1
λ2 = −2.132 × 106 . (7.231)
s
Thus we can conclude that the physical equilibrium is linearly stable. The local time con-
stants near equilibrium are given by the reciprocal of the magnitude of the eigenvalues.
These are
Evolution on these two time scales is predicted in Fig. 7.8. This in fact a multiscale problem.
One of the major difficulties in the numerical simulation of combustion problems comes in
the effort to capture the effects at all relevant scales. The problem is made more difficult as
the breadth of the scales expands. In this problem, the breadth of scales is not particularly
challenging. Near equilibrium the ratio of the slowest to the fastest time scale, the stiffness
ratio κ, is
τ2 4.691 × 10−7 s
κ= = = 14.75. (7.234)
τ1 3.181 × 10−8 s
Many combustion problems can have stiffness ratios over 106 . This is more prevalent at
lower temperatures.
We can do a similar linearization near the initial state, find the local eigenvalues, and
the local time scales. At the initial state here, we find those local time scales are
So initially the stiffness, κ = (2.403 × 10−8 s)/(2.123 × 10−8 s) = 1.13 is much less, but the
time scale itself is small. It is seen from Fig. 7.8 that this initial time scale of 10−8 s well
predicts where significant evolution of species concentrations commences. For t < 10−8 s, the
model predicts essentially no activity. This can be correlated with the mean time between
molecular collisions–the theory on which estimates of the collision frequency factors aj are
obtained.
We briefly consider the non-physical roots, 1 and 2. A similar eigenvalue analysis of root
1 reveals that the eigenvalues of its local Jacobian matrix are
1
λ1 = −1.193 × 107 , (7.237)
s
1
λ2 = 5.434 × 106 . (7.238)
s
Thus root 1 is a saddle and is unstable.
For root 2, we find
1
λ1 = 4.397 × 107 + i7.997 × 106 , (7.239)
s
1
λ2 = 4.397 × 107 − i7.997 × 106 . (7.240)
s
-5
-10
-15
spiral
-20 source
2
-4 -3 -2 -1 0 1 2
Figure 7.9: NO and N phase portraits for T = 6000 K, P = 2.4942 × 106 dyne/cm2
Zel’dovich mechanism.
The eigenvalues are complex with a positive real part. This indicates the root is an unstable
spiral source.
A detailed phase portrait is shown in Fig. 7.9. Here we see all three roots. Their local
character of sink, saddle, or spiral source is clearly displayed. We see that trajectories are
attracted to a curve labeled SIM for “Slow Invariant Manifold.” A part of the SIM is
constructed by the trajectory which originates at root 1 and travels to root 3. The other
part is constructed by connecting an equilibrium point at infinity into root 3. Details are
omitted here.
10 -6
NO
10 -8
10 -10
10 -12
N
t (s)
10 -11 10 -9 10 -7 10 -5 0.001 0.1 10
Figure 7.10: ρNO and ρN versus time for Zel’dovich mechanism at T = 1500 K, P =
6.23550 × 105 dyne/cm2 .
mass-spring-damper systems with stiff springs. Here we will examine the effect of tempera-
ture and pressure on time scales and stiffness. We shall also look simplistically how different
numerical approximation methods respond to stiffness.
7.1.2.2.1 Effect of temperature Let us see how the same Zel’dovich mechanism be-
haves at lower temperature, T = 1500 K; all other parameters, including the initial species
concentrations are the same as the previous high temperature example. The pressure how-
ever, lowers, and here is P = 6.23550×105 dyne/cm2 , which is close to atmospheric pressure.
For this case, a plot of species concentrations versus time is given in Figure 7.10.
At T = 1500 K, we notice some dramatic differences relative to the earlier studied
T = 6000 K. First, we see the reaction commences in around the same time, t ∼ 10−8 s. For
t ∼ 10−6 s, there is a temporary cessation of significant reaction. We notice a long plateau
in which species concentrations do not change over several decades of time. This is actually
a pseudo-equilibrium. Significant reaction recommences for t ∼ 0.1 s. Only around t ∼ 1 s
does the system approach final equilibrium. We can perform an eigenvalue analysis both
at the initial state and at the equilibrium state to estimate the time scales of reaction. For
this dynamical system which is two ordinary differential equations in two unknowns, we will
always find two eigenvalues, and thus two time scales. Let us call them τ1 and τ2 . Both
these scales will evolve with t.
At the initial state, we find
τ1 = 2.37 × 10−8 s, (7.241)
τ2 = 4.25 × 10−7 s. (7.242)
The onset of significant reaction is consistent with the prediction given by τ1 at the initial
state. Moreover, initially, the reaction is not very stiff; the stiffness ratio is κ = 17.9.
At equilibrium, we find
mole
lim ρNO = 4.6 × 10−9 , (7.243)
t→∞ cm3
mole
lim ρN = 4.2 × 10−14 , (7.244)
t→∞ cm3
and
The slowest time scale near equilibrium is an excellent indicator of how long the system takes
to relax to its final state. Near equilibrium, the stiffness ratio is large, κ = τ2 /τ1 ∼ 3.8 × 105 .
This is known as the stiffness ratio. When it is large, the scales in the problem are widely
disparate and accurate numerical solution becomes challenging.
In summary, we find the effect of lowering temperature while leaving initial concentrations
constant
• lowers the pressure somewhat, slightly slowing down the collision time, and slightly
slowing the fastest time scales, and
• slows the slowest time scales many orders of magnitude, stiffening the system signifi-
cantly, because collisions may not induce reaction with their lower collision speed.
b mole
ρNO = b
ρN = b
ρN2 = b
ρO2 = b
ρO = 10−8 . (7.247)
cm3
With this decrease in number of moles, the pressure now is
dyne
P = 6.23550 × 103 . (7.248)
cm2
This pressure is two orders of magnitude lower than atmospheric. We solve for the species
concentration profiles and show the results of numerical prediction in Figure 7.11 Relative to
the high pressure P = 6.2355 × 105 dyne/cm2 , T = 1500 K case, we notice some similarities
and dramatic differences. The overall shape of the time-profiles of concentration variation
is similar. But, we see the reaction commences at a much later time, t ∼ 10−6 s. For
t ∼ 10−4 s, there is a temporary cessation of significant reaction. We notice a long plateau
in which species concentrations do not change over several decades of time. This is again
actually a pseudo-equilibrium. Significant reaction recommences for t ∼ 10 s. Only around
10- 8
NO
10 -10
10 -12
10 -14
N
10 -16
t (s)
10- 8 10 - 6 10- 4 0.01 1 100
Figure 7.11: ρNO and ρN versus time for Zel’dovich mechanism at T = 1500 K, P =
6.2355 × 103 dyne/cm2 .
t ∼ 100 s does the system approach final equilibrium. We can perform an eigenvalue analysis
both at the initial state and at the equilibrium state to estimate the time scales of reaction.
At the initial state, we find
The onset of significant reaction is consistent with the prediction given by τ1 at the initial
state. Moreover, initially, the reaction is not very stiff; the stiffness ratio is κ = 17.9.
Interestingly, by decreasing the initial pressure by a factor of 102 , we increased the initial
time scales by a complementary factor of 102 ; moreover, we did not alter the stiffness.
At equilibrium, we find
mole
lim ρNO = 4.6 × 10−11 , (7.251)
t→∞ cm3
mole
lim ρN = 4.2 × 10−16 , (7.252)
t→∞ cm3
(7.253)
and
By decreasing the initial pressure by a factor of 102 , we decreased the equilibrium concentra-
tions by a factor of 102 and increased the time scales by a factor of 102 , leaving the stiffness
ratio unchanged.
In summary, we find the effect of lowering the initial concentrations significantly while
leaving temperature constant
• lowers the pressure significantly, proportionally slowing down the collision time, as well
as the fastest and slowest time scales, and
7.1.2.2.3 Stiffness and numerics The issue of how to simulate stiff systems of ordinary
differential equations, such as presented by our Zel’dovich mechanism, is challenging. Here a
brief summary of some of the issues will be presented. The interested reader should consult
the numerical literature for a full discussion. See for example the excellent text of Iserles.10
We have seen throughout this section that there are two time scales at work, and they are
often disparate. The species evolution is generally characterized by an initial fast transient,
followed by a long plateau, then a final relaxation to equilibrium. We noted from the phase
plane of Fig. 7.9 that the final relaxation to equilibrium (shown along the green line labeled
“SIM”) is an attracting manifold for a wide variety of initial conditions. The relaxation onto
the SIM is fast, and the motion on the SIM to equilibrium is relatively slow.
Use of common numerical techniques can often mask or obscure the actual dynamics.
Numerical methods to solve systems of ordinary differential equations can be broadly cat-
egorized as explicit or implicit. We give a brief synopsis of each class of method. We cast
each as a method to solve a system of the form
dρ
= f(ρ). (7.256)
dt
• Explicit: The simplest of these, the forward Euler method, discretizes Eq. (7.256) as
follows:
ρn+1 − ρn
= f(ρn ), (7.257)
∆t
so that
– easy to program, because Eq. (7.258) can be solved explicitly to predict the new
value ρn+1 in terms of the old values at step n.
10
A. Iserles, 2008, A First Course in the Numerical Analysis of Differential Equations, Cambridge Uni-
versity Press, Cambridge, UK.
• Implicit: The simplest of these methods, the backward Euler method, discretizes
Eq. (7.256) as follows:
ρn+1 − ρn
= f(ρn+1 ), (7.259)
∆t
so that
– more difficult to program because a non-linear set of algebraic equations, Eq. (7.260),
must be solved at every time step with no guarantee of solution,
– requiring potentially significant computational time to advance each time step,
– capable of using very large time steps and remaining numerically stable,
– suspect to missing physics that occur on small time scales τ < ∆t,
– in general better performers than explicit methods.
A wide variety of software tools exist to solve systems of ordinary differential equations.
Most of them use more sophisticated techniques than simple forward and backward Euler
methods. One of the most powerful techniques is the use of error control. Here the user
specifies how far in time to advance and the error that is able to be tolerated. The algorithm,
which is complicated, selects then internal time steps, for either explicit or implicit methods,
to achieve a solution within the error tolerance at the specified output time. A well known
public domain algorithm with error control is provided by lsode.f, which can be found in
the netlib repository.11
Let us exercise the Zel’dovich mechanism under the conditions simulated in Fig. 7.11,
T = 1500 K, P = 6.2355 × 103 dyne/cm2 . Recall in this case the fastest time scale near
equilibrium is τ1 = 7.86 × 10−5 s ∼ 10−4 s at the initial state, and the slowest time scale is
τ = 3.02 × 101 s at the final state. Let us solve for these conditions using dlsode.f, which
uses internal time stepping for error control, in both an explicit and implicit mode. We
specify a variety of values of ∆t and report typical values of number of internal time steps
Table 7.3: Results from computing Zel’dovich NO production using implicit and explicit
methods with error control in dlsode.f.
selected by dlsode.f, and the corresponding effective time step ∆tef f used for the problem,
for both explicit and implicit methods, as reported in Table 7.3.
Obviously if output is requested using ∆t > 10−4 s, the early time dynamics near t ∼
−4
10 s will be missed. For physically stable systems, codes such as dlsode.f will still provide
a correct solution at the later times. For physically unstable systems, such as might occur in
turbulent flames, it is not clear that one can use large time steps and expect to have fidelity
to the underlying equations. The reason is the physical instabilities may evolve on the same
time scale as the fine scales which are overlooked by large ∆t.
A ⇌ B. (7.261)
11
Hindmarsh, A. C., 1983,“ODEPACK, a Systematized Collection of ODE Solvers,” Scientific Com-
puting, edited by R. S. Stepleman, et al., North-Holland, Amsterdam, pp. 55-64. Source code at
https://www.netlib.org/alliant/ode/prog/lsode.f
Let us take A and B to both be calorically perfect ideal gases with identical molecular masses
MA = MB = M and identical specific heats, cvA = cvB = cv ; cP A = cP B = cP . We can
consider A and B to be isomers of an identical molecular species. So we have N = 2 species
reacting in J = 1 reactions. The number of elements L here is irrelevant.
ρA (0) = b
ρA , (7.264)
ρB (0) = 0. (7.265)
For our alternate compact linear algebra based form, we note that
−E 1
r = a exp ρA − ρ , (7.266)
RT Kc B
and that
d ρA −1
= (r). (7.267)
dt ρB 1
Performing the decomposition yields
d ρA −1
= (r). (7.268)
dt ρA + ρ B 0
Expanded, this is
1 0 d ρA −1
= (r). (7.269)
1 1 dt ρB 0
Combining Eqs. (7.262-7.263) and integrating yields
d
(ρ + ρB ) = 0, (7.270)
dt A
ρA + ρB = b
ρA , (7.271)
ρB = b
ρA − ρA . (7.272)
U = nA uA + nB uB , (7.278)
n n
A B
= V uA + uB , (7.279)
V V
= V (ρA uA + ρB uB ) , (7.280)
PA PB
= V ρA h A − + ρB hB − , (7.281)
ρA ρB
= V ρA hA − RT + ρB hB − RT , (7.282)
o
o
= V ρA cP (T − To ) + hTo ,A − RT + ρB cP (T − To ) + hTo ,B − RT ,(7.283)
o o
= V (ρA + ρB )(cP (T − To ) − RT ) + ρA hTo ,A + ρB hTo ,B , (7.284)
o o
= V (ρA + ρB )((cP − R)T − cP To ) + ρA hTo ,A + ρB hTo ,B , (7.285)
o o
= V (ρA + ρB )((cP − R)T − (cP − R + R)To ) + ρA hTo ,A + ρB hTo ,B , (7.286)
o o
= V (ρA + ρB )(cv T − (cv + R)To ) + ρA hTo ,A + ρB hTo ,B , (7.287)
o o
= V (ρA + ρB )cv (T − To ) + ρA (hTo ,A − RTo ) + ρB (hTo ,B − RTo ) , (7.288)
= V (ρA + ρB )cv (T − To ) + ρA uoTo ,A + ρB uoTo ,B . (7.289)
And because we have assumed the molecular masses are the same, MA = MB , the mole
fractions are the mass fractions, and we can write on a mass basis
Returning to Eq. (7.292), our energy conservation relation, Eq. (7.277), becomes
b b b o b o
0 = V (ρA + ρB )cv (T − To ) + (ρA − ρA )uTo ,A + (ρB − ρB )uTo ,B . (7.295)
0 = (b
ρA + bρB )cv (T − To ) + (ρA − b ρA )uoTo ,A + (ρB − b
ρB )uoTo ,B , (7.296)
ρ −b ρA o ρ −b ρB o
0 = cv (T − To ) + A uTo ,A + B uT ,B , (7.297)
b
ρA + b
ρB b
ρA + b ρB o
ρ − ρA uoTo ,A b
b ρ − ρB uoTo ,B
T = To + A + B . (7.298)
b
ρA + bρ B cv b
ρA + b ρB cv
(b
ρA − ρA ) o o
T = To + (hTo ,A − hTo ,B ). (7.301)
b
ρ A cv
o o
We see then that if hTo ,A > hTo ,B , that as ρA decreases from its initial value of b
ρA that T
will increase. We can scale Eq. (7.301) to form
o o !
T ρA hTo ,A − hTo ,B
=1+ 1− . (7.302)
To b
ρA cv To
We also note that our caloric state equation, Eq. (7.293) can, for yAo = 1, yBo = 0 as
with
∆Go = g oB − g oA , (7.308)
o o
= hB − T soB − (hA − T soA ), (7.309)
o o
= (hB − hA ) − T (soB − soA ), (7.310)
o o
= (hTo ,B − hTo ,A ) − T (soTo ,B − soTo ,A ). (7.311)
So
o o !
hTo ,A − hTo ,B − T (soTo ,A − soTo ,B )
Kc = exp , (7.312)
RT
o o !!
cv To hTo ,A − hTo ,B − T (soTo ,A − soTo ,B )
= exp , (7.313)
RT cv To
o o !!
1 1 hTo ,A − hTo ,B T (soTo ,A − soTo ,B )
= exp − . (7.314)
k − 1 TTo cv To To cv
Here we have used the definition of the ratio of specific heats, k = cP /cv along with R =
cP − cv . So we can solve Eq. (7.273) by first using Eq. (7.314) to eliminate Kc and then
Eq. (7.301) to eliminate T .
τ = at. (7.315)
ρB ρc − ρA
λ= = A . (7.325)
b
ρA b
ρA
Thus,
λ = 1 − z. (7.326)
The right side of Eq. (7.331) is at equilibrium for values of λ which drive it to zero.
Numerical root finding methods show this to occur at λ ∼ 0.920539. Near this root, Taylor
series expansion shows the dynamics are approximated by
d
(λ − 0.920539) = −0.17993(λ − 0.920539) + . . . (7.332)
dτ
Thus the local behavior near equilibrium is given by
Here C is some arbitrary constant. Clearly the equilibrium is stable, with a time constant
of 1/0.17993 = 5.55773.
Numerical solution shows the full behavior of the dimensionless species concentration
λ(τ ); see Figure 7.12. Clearly the product concentration λ is small for some long period of
time. At a critical time near τ = 2.7×106 , there is a so-called thermal explosion with a rapid
increase in λ. The estimate of the time constant near equilibrium is orders of magnitude less
than the explosion time, 5.55773 ≪ 2.7 × 106 . Thus, linear analysis here is a poor tool to
estimate an important physical quantity, the ignition delay time. Once the ignition period
is over, there is a rapid equilibration to the final state. The dimensionless temperature
plot is shown in Figure 7.13. The temperature plot is similar in behavior to the species
1.0
0.8
0.6
0.4
0.2
0 6 6 6 6 6
1 x 10 2 x10 3 x 10 4 x10 5 x 10
Figure 7.12: Dimensionless plot of reaction product concentration λ versus time τ for adia-
batic isochoric combustion with simple reversible kinetics.
10
6 6 6 6 6
1 x 10 2 x10 3 x 10 4 x10 5 x 10
Figure 7.13: Dimensionless plot of temperature θ versus time τ for adiabatic, isochoric
combustion with simple reversible kinetics.
concentration plot. At early time, the temperature is cool. At a critical time, the thermal
explosion or ignition delay time, the temperature rapidly rises. This rapid rise, coupled
with the exponential sensitivity of reaction rate to temperature, accelerates the formation
of product. This process continues until the reverse reaction is activated to the extent it
prevents further creation of product.
λ = ǫλ1 + ǫ2 λ2 + ǫ3 λ3 + . . . (7.335)
Here we will assume 0 < ǫ ≪ 1 and that λ1 (τ ) ∼ O(1), λ2 (τ ) ∼ O(1), . . . , and will define ǫ
in terms of physical parameters shortly. Now with this assumption, we have
1 1
= . (7.336)
1 + qλ 1 + ǫqλ1 + ǫ qλ2 + ǫ3 qλ3 + . . .
2
Long division of the term on the right side yields the approximation
1
= 1 − ǫqλ1 + ǫ2 (q 2 λ21 − qλ2 ) + . . . , (7.337)
1 + qλ
= 1 − ǫqλ1 + O(ǫ2 ). (7.338)
So
Θ
exp − ∼ exp −Θ(1 − ǫqλ1 + O(ǫ2 )) , (7.339)
1 + qλ
∼ e−Θ exp ǫqΘλ1 + O(ǫ2 ) . (7.340)
We have moved λ from the denominator to the numerator of the most important exponential
term.
12
J. D. Buckmaster and G. S. S. Ludford, 1983, Lectures on Mathematical Combustion, SIAM, Philadelphia.
This is supplemented by the initial condition λ1 (0) = 0. Separating variables and solving,
we get
exp(−qλ1 ) dλ1 = dτ∗ , (7.349)
1
− exp(−qλ1 ) = τ∗ + C. (7.350)
q
Applying the initial condition gives
1
− exp(−q(0)) = C, (7.351)
q
1
− = C. (7.352)
q
So
1 1
− exp(−qλ1 ) = τ∗ − , (7.353)
q q
exp(−qλ1 ) = −qτ∗ + 1, (7.354)
1
exp(−qλ1 ) = −q τ∗ − , (7.355)
q
1
−qλ1 = ln −q τ∗ − , (7.356)
q
1 1
λ1 = − ln −q τ∗ − . (7.357)
q q
For q = 10, a plot of λ1 (τ∗ ) is shown in Fig. 7.14. We note at a finite τ∗ that λ1 begins to
exhibit unbounded growth. In fact, it is obvious from Eq. (7.348) that as
1
τ∗ → ,
q
that
λ1 → ∞.
That is there exists a finite time for which λ1 violates the assumptions of our asymptotic
theory which assumes λ1 = O(1). We associate this time with the ignition delay time, τ∗i :
1
τ∗i = . (7.358)
q
Let us return this to more primitive variables:
1 −1 1
exp τi = , (7.359)
ǫ ǫ q
ǫ exp 1ǫ
τi = , (7.360)
q
exp Θ
= . (7.361)
Θq
1.2
0.8
0.4
For our system with Θ = 20 and q = 10, we estimate the dimensionless ignition delay time
as
exp 20
τi = = 2.42583 × 106 . (7.362)
(20)(10)
This is a surprisingly good estimate, given the complexity of the problem. Recall the nu-
merical solution showed ignition for τ ∼ 2.7 × 106 .
In terms of dimensional time, the ignition delay time prediction becomes
exp Θ
ti = , (7.363)
aΘq
!
1 RTo cv To E
= o o exp . (7.364)
a E hTo ,A − hTo ,B RTo
Ignition is suppressed if the ignition delay time is lengthened, which happens when
• the activation energy E is increased, because the exponential sensitivity is stronger
than the algebraic sensitivity,
o o
• the energy of combustion (hTo ,A − hTo ,B ) is decreased because it takes longer to react
to drive the temperature to a critical value to induce ignition, and
• the collision frequency factor a is decreased, which suppresses reaction.
The initial mass fractions are calculated via ci = Mi ρi /ρ. They are
To avoid issues associated with numerical roundoff errors at very early time for species
with very small compositions, the minor species were initialized at a small non-zero value
near machine precision; each was assigned a value of 10−15 mole. The minor species all have
ρi = (10−15 mole)/(0.3061251 cm3 ) = 3.26664 × 10−15 mole/cm3 . They have correspondingly
small initial mass fractions.
We seek the reaction dynamics as the system proceeds from its initial state to its final
state. We use the reversible detailed kinetics mechanism of Table 7.2. This problem requires
a detailed numerical solution. Such a solution was performed by solving the appropriate
equations for a mixture of nine interacting species: H2 , H, O, O2 , OH, H2 O, HO2 , H2 O2 ,
and N2 . The dynamics of the reaction process are reflected in Figs. 7.15-7.16.
At early time, t < 10−7 s, the pressure, temperature, and major reactant species con-
centrations (H2 , O2 , N2 ) are nearly constant. However, the minor species, e.g. OH, HO2 ,
and the major product, H2 O, are undergoing very rapid growth, albeit with math fractions
whose value remains small. In this period, the material is in what is known as the induction
period.
After a certain critical mass of minor species has accumulated, exothermic recombination
of these minor species to form the major product H2 O induces the temperature to rise, which
accelerates further the reaction rates. This is manifested in a thermal explosion. A common
definition of the end of the induction period is the ignition delay, t = ti , the time when dT /dt
goes through a maximum. Here one finds
0 H
10 O
H2
O2
OH
H2O
−5 H 2 O2
10 HO 2
N2
−10
10
−15
10
−10 −5 0
10 10 10
t (s)
Figure 7.15: Plot of mass fractions cH2 (t), cH (t), cO (t), cO2 (t), cOH (t), cH2 O (t), cHO2 (t),
cH2 O2 (t), cN2 (t), for adiabatic, isochoric combustion of a mixture of 2H2 + O2 + 3.76N2
initially at To = 1542.7 K, Po = 2.8323 × 106 Pa.
7
4000 10
3500
3000
2500
P (Pa)
T (K)
2000
1500
1000
500
6
0 −10 −5
10 −10 −5
10 10 10 10
t (s) t (s)
Figure 7.16: Plot of T (t) and P (t), for adiabatic, isochoric combustion of a mixture of
2H2 + O2 + 3.76N2 initially at To = 1542.7 K, Po = 2.8323 × 106 Pa.
0
10
H
O
−5 H2
10
O2
OH
H2O
H 2 O2
10
−10 HO 2
N2
2 4 6 8 10
t (s) −7
x 10
Figure 7.17: Plot near thermal explosion time of mass fractions cH2 (t), cH (t), cO (t), cO2 (t),
cOH (t), cH2 O (t), cHO2 (t), cH2 O2 (t), cN2 (t), for adiabatic, isochoric combustion of a mixture of
2H2 + O2 + 3.76N2 initially at at To = 1542.7 K, Po = 2.8323 × 106 Pa.
At the end of the induction zone, there is a final relaxation to equilibrium. The equilib-
rium mass fractions of each species are
We note that because our model takes N2 to be inert that its value remains unchanged.
Other than N2 , the final products are dominated by H2 O. The equilibrium temperature is
3382.3 K and 5.53 × 106 Pa.
This appendix will consider two peripheral topics: Legendre transformations, first introduced
in Sec. 4.3, and the method of least squares.
Legendre transformations
Here we will draw upon the work of Zia, et al.14 to better understand Legendre transforma-
tions in general and how they relate to thermodynamics. The method is general and is often
applied to other problems in physics, especially in dynamics.
Consider a function F (x). Let us insist that F be such that
d2 F
≥ 0.
dx2
That is to say, its slope increases or does not change as x increases. Such a function may
possess a minimum value, and can be considered to be convex. Formally, one can say that
F is convex when the region above it, known as the epigraph, forms a so-called convex set.
A set is convex if special linear combinations of any of its elements also reside within the
set. Examples of convex and non-convex functions are shown in Fig. 7.18. For the convex
function F (x) = x2 , shown in Fig. 7.18a, we have d2 F/dx2 = 2 > 0. We also see that
special linear combinations of any points within the epigraph will lie within the epigraph.
Mathematically, this is expressed as follows: for s ∈ [0, 1] we must have F (sx1 + (1 − s)x2 ) ≤
sF (x1 ) + (1 − s)F (x2 ). This is illustrated by a sample line whose interior points all lie within
the epigraph. For the non-convex function F (x) = x3 − x2 − x, shown in Fig. 7.18b, we have
d2 F/dx2 = 6x − 2. Obviously, this is not positive for all x, and so the function is non-convex.
And we see that lines exist connecting points within the epigraph that contain points outside
of the epigraph. Hence it is non-convex.
Let us define the slope of F as w. Mathematically, we can say
dF
= w(x).
dx
Because F is a function of x, its derivative also is a function of x. Now because we have
insisted that w is increasing as x increases, we can always find a unique inverse such that
14
R. K. P. Zia, E. F. Redish, and S. R. McKay, 2009, “Making sense of the Legendre transform,” Ameri-
can Journal of Physics, 77(7): 614-622.
295
F(x) F(x)
2.0 2
0.990m
e
%
ativ
°
60
//KKgDr
Rel
n
atio
50%
epigraph epigraph
tur
yfir
1.5
rSa
1
llo
tBu
We
sob
1.0
0
0.5
-1
0.0
convex function non-convex
30 35 function 40
- 0.5 -2
- 1.0 - 0.5 0.0 0.5 1.0 - 1.0 - 0.5 0.0 0.5 1.0 1.5 2.0
x x
a) b)
Figure 7.18: Plots illustrating a) the convex function F (x) = x2 and b) the non-convex
function F (x) = x3 − x2 − x.
x(w) exists. That is to say, just as the slope is a function of x, x can be identified as a
function of slope. Now let us define the Legendre transformation G as
Given F (x), and convexity, it is always possible to compute G. We can write this in a form
that will be more useful for thermodynamics:
Here we see a symmetry in the relationship. Now differentiate G with respect to w. We get
dG dx dF dx
=w + x(w) − .
dw dw dx dw
Rearrange to get
dG dx dF
= w− +x(w)
dw dw dx
| {z }
0
Now because dF/dx = w, we get dG/dw = x. Thus we get a set of symmetric relations
dF dG
= w, = x.
dx dw
We could also say then that
dG dF
G(w) + F (x) =
.
dw dx
It can remarkably be shown that the Legendre transformation of G returns us to the original
function F .
296
We also have
d2 F dw d2 G dx
2
= , 2
= .
dx dx dx dw
So
d2 F d2 G dw dx
= = 1.
dx2 dw 2 dx dw
Example 7.1
Find the Legendre transformation of F (x) = ex .
w(x) = ex , x(w) = ln w.
For the inverse to be real valued, we must require that w > 0, which is the case for w = ex . So
Example 7.2
Find the Legendre transformation of F (x) = x(ln x − 1). The function F has the same form as the
Legendre transformation of ex studied in the previous example.
x(w) = ew .
297
The Legendre transformation is
The Legendre transformation of the Legendre transformation returns us to the original function. Thus,
the transform is its own inverse.
Example 7.3
Find the Legendre transformation of F (x) = αx2 /2, where α is a positive scalar constant α > 0.
Example 7.4
Find the Legendre transformation of F (x) = x2 /(2α), with α > 0. The function F has the same
form as the Legendre transformation of αx2 /2 studied in the previous example.
298
We also see that
d2 F 1
= ,
dx2 α
so F is convex. We then get that
x(w) = αw.
The Legendre transformation is
Example 7.5
Apply the formalism of Legendre transformations to a linear mass spring system.
dU
f= = kx.
dx
Thus, we associate f with the traditional spring force kx. We see
d2 U
= k,
dx2
and because k > 0, U (x) is convex.
Now we also see
f
x(f ) = .
k
Let us let V stand in for G and define the Legendre transformation
V (f ) = xf − U (x).
So 2
f2 1 f
V (f ) = − k .
k 2 k
299
Thus,
1 f2
V (f ) = .
2 k
Note that
dV f
= = x(f ).
df k
Our Legendre transformation then gives
U (x) + V (f ) = f x,
1 2 1 f2
kx + = f x,
2 2 k
(kx)2 + f 2 − 2kf x = 0,
(f − kx)2 = 0.
This yields
f = kx,
as it must. We can think of U (x) as a potential associated with the position. We can think of V (f ) as
a potential associated with the control parameter f , the spring force.
Example 7.6
Consider the Legendre transformation in terms of traditional Lagrangian and Hamiltonian dynamics
for a particle of mass m with position q, velocity v = q̇, momentum p = mv = mq̇, Lagrangian L(q̇, q, t),
and Hamiltonian H(p, q, t). The particle moves in a potential field with the potential a known function
of position, V (q). We consider the so-called “coordinates” to be p and q̇
We typically think of Hamiltonian systems as non-dissipative; for particle dynamics, this implies
that the sum of the kinetic energy T = mv 2 /2 and potential energy V = mgz is a constant independent
of time:
H = T + V = constant.
The kinetic and potential energy can change with time, but their sum does not for the non-dissipative
system. Here z is a traditional position coordinate that we take to be z = q. And we simply require the
potential V to be a general function of position only: V (q). The kinetic energy in terms of momentum
p is
1 m2 v 2 p2
T = mv 2 = = .
2 2m 2m
So
p2
H(p, q, t) = + V (q).
2m
For the Legendre transformation formalism, we wish to consider the variables p and q̇. Let us then
define the “slope”
∂H p
= = q̇.
∂p q,t m
300
With the definition of momentum p = mv, we see that
q̇ = v.
∂2H 1
= > 0,
∂p2 q,t m
because m > 0.
The Legendre transformation allows us to define the Lagrangian L in terms of the Hamiltonian H:
So
(mq̇)2
L(q̇, q, t) = pq̇ − + V (q) .
2m
Simplifying, we get
2 mq̇ 2
L(q̇, q, t) = mq̇ − + V (q) ,
2
or
q̇ 2 mv 2
L(q̇, q, t) = m − V (q) = − V (q) = T − V.
2 2
So we have
H = T + V, L = T − V.
We could then say
L(q̇, q, t) + H(p, q, t) = pq̇.
Note that
∂L ∂H p
= mq̇ = mv = p, = = v = q̇.
∂ q̇ q,t ∂p q,t m
Now with L = L(q, q̇, t), we have the total derivative
∂L ∂L ∂L
dL = dq + dq̇ + dt.
∂q ∂ q̇ ∂t
Now p = ∂L/∂ q̇, so
∂L ∂L
dL = dq + p dq̇ + dt.
∂q ∂t
With the product rule, we say
∂L ∂L
dL = dq + d(pq̇) − q̇ dp + dt.
∂q ∂t
Rearranging gives
∂L ∂L
d (L − pq̇) = dq − q̇ dp + dt.
∂q ∂t
∂L ∂L
d (pq̇ − L) = − dq + q̇ dp − dt.
∂q ∂t
Recalling pq̇ − L = H, we get
∂L ∂L
dH = − dq + q̇ dp − dt.
∂q ∂t
301
Scale by dt to get
dH ∂L dp ∂L
=− q̇ + q̇ − .
dt ∂q dt ∂t
dH ∂L d ∂L ∂L
=− q̇ + q̇ − .
dt ∂q dt ∂ q̇ ∂t
Now because H = pq̇ − L, we have
∂H ∂L
=− .
∂t p,q̇ ∂t p,q̇
And because for non-dissipative systems H is constant, it does not vary with time, so we also have
∂L/∂t = 0, along with dH/dt = 0. Thus, after scaling by q̇, we get the well-known Euler-Lagrange
equation:
∂L d ∂L
− = 0.
∂q dt ∂ q̇
We can get the so-called Hamilton’s equations for a Hamiltonian system by the following. For a
constant Hamiltonian, we have H = H(p, q). So
∂H ∂H
dH = dp + dq = 0.
∂p q ∂q p
Scale by dt to get
dH ∂H ∂H
= ṗ + q̇ = 0.
dt ∂p q ∂q p
For this to hold, we must require ṗ = −∂H/∂q. Thus, we have the well-known Hamilton’s equations:
∂H ∂H
= q̇, = −ṗ.
∂p q ∂q p
Let us see how this fits with thermodynamics. We will need some unusual notation to
identify the similarities. Let us take
P̃ ≡ −P, h̃ ≡ −h, ã ≡ −a, g̃ ≡ −g.
Our definition of enthalpy, h = u + P v, can be rewritten as
u − h = −P v.
Now recall from Eq. (4.76) that the canonical form for u is u(v, s), and that from Eq. (4.81),
the canonical form for h is h(P, s). Also invoking Eq. (4.26) to eliminate P and Eq. (4.100)
to eliminate v, we get
∂u ∂h
u(v, s) − h(P, s) = .
∂v s ∂P s
302
In terms of P̃ and h̃, we could say
∂u ∂ h̃
u(v, s) + h̃(P̃ , s) = .
∂v s ∂ P̃ s
Recognizing that s is a frozen variable here, we see that this is precisely the form of the
Legendre transformation G(w) + F (x) = wx, with dF/dx = w, dG/dw = x. We could also
say
u(v, s) = v P̃ (v, s) − h̃(P̃ (v, s), s),
to be consistent with the original form G(w) = wx(w) − F (x(w)). So u is playing the role
of G, v is playing the role of w, P̃ is playing the role of x, and h̃ is playing the role of F .
Rearranging gives
(−h̃(P̃ (v, s), s)) = u(v, s) + (−P̃ (v, s))v,
h(P (v, s), s)) = u(v, s) + P (v, s)v.
For convexity, we need d2 F/dx2 ≥ 0. This extends to
∂ 2 h̃
≥ 0.
∂ P̃ 2 s
303
non-convexity can be introduced by non-ideal effects such as reflected in a van der Waals
equation of state.
Let us repeat this for a and g. First for Helmholtz free energy from Eq. (4.102) we get
u − a = T s.
Recognizing that v is a frozen variable here, we once again see that this is precisely the form
of the Legendre transformation G(w) + F (x) = wx, with dF/dx = w, dG/dw = x.
For Gibbs free energy, from Eq. (4.110), we can say
h − g = T s.
Recognizing that P is a frozen variable here, we once again see that this is precisely the form
of the Legendre transformation G(w) + F (x) = wx, with dF/dx = w, dG/dw = x.
One important application of data analysis is the method of least squares. This method is
often used to fit data to a given functional form. The form is most often in terms of polyno-
mials, but there is absolutely no restriction; trigonometric functions, logarithmic functions,
Bessel functions can all serve as well. Here, we will restrict ourselves to strictly scalar
functions of the form
x = f (t; aj ), j = 1, . . . , M,
where x is a dependent variable, t is an independent variable, f is an assumed functional
form, and aj is a set of M constant parameters in the functional form. The analysis can
304
easily be extended for functions of many variables. General mathematical background is
given by Strang.18
Mathematically, the fundamental problem is, given
• an assumed functional form for the curve fit f (t; aj ) that has M parameters aj , j =
1, . . . , M,
find the best set of parameter values aj so as to minimize the least squares error between the
curve fit and the actual data points. That is, the problem is to find aj , j = 1, . . . , M, such
that v
u N
uX
ℓ2 = ||xi − f (ti ; aj )||2 ≡ t (xi − f (ti ; aj ))2 ,
i=1
• substitutes each data point into the assumed form so as to form an over-constrained
system of equations,
• uses straightforward techniques from linear algebra to solve for the coefficients that
best represent the given data if the problem is linear in the coefficients aj ,
• uses techniques from optimization theory to solve for the coefficients that best represent
the given data if the problem is non-linear in aj .
The most general problem, in which the dependency aj is non-linear, is difficult, and
sometimes impossible. For cases in which the functional form is linear in the coefficients aj or
can be rendered linear via simple transformation, it is possible to get a unique representation
of the best set of parameters aj . This is often the case for common curve fits such as straight
line, polynomial, or logarithmic fits.
Let us first consider polynomial curve fits. Now, if one has say, ten data points, one can
in principle, find a ninth order polynomial that will pass through all the data points. Often
times, especially when there is much experimental error in the data, such a function may
be subject to wild oscillations, that are unwarranted by the underlying physics, and thus is
18
G. Strang, 1988, Linear Algebra and its Application, Harcourt Brace Jovanovich, Orlando, Florida.
19
Pythagoras of Samos, c. 570 B.C.-495 B.C., Ionian Greek philosopher and mathematician.
305
not useful as a predictive tool. In such cases, it may be more useful to choose a lower order
curve that does not exactly pass through all experimental points, but that does minimize
the error.
This is the most common method used when one has equal confidence in all the data.
Example 7.7
Find the best straight line to approximate the measured data relating x to t.
t x
0 5
1 7
2 10
3 12
6 15
x = a1 + a2 t,
where a1 and a2 are the terms to be determined. Substituting each data point to the assumed form,
we get five equations in two unknowns:
5 = a1 + 0a2 ,
7 = a1 + 1a2 ,
10 = a1 + 2a2 ,
12 = a1 + 3a2 ,
15 = a1 + 6a2 .
This is an over-constrained problem, and there is no unique solution that satisfies all of the equations!
If a unique solution existed, then the curve fit would be perfect. However, there does exist a solution
that minimizes the error, as is often proved in linear algebra textbooks (and will not be proved here).
The procedure is straightforward. Rearranging, we get
1 0 5
1 1 7
a1
1 2
a2 = 10 .
1 3 12
1 6 15
AT · A · a = AT · b,
−1
a = AT · A · AT · b.
306
Substituting, we find that
−1
1 0 5
1 1 7
a1 1 1 1 1 1 1 1 1 1 1
= 1 2
10 = 5.7925 .
a2 0 1 2 3 6 0 1 2 3 6 1.6698
1 3 12
1 6 15
20
16
Data Points
12
x
x = 5.7925 + 1.6698 t
4
2 4 6
t
Figure 7.19: Plot of x − t data and best least squares straight line fit.
If one has more confidence in some data points than others, one can define a weighting
function to give more priority to those particular data points.
307
Example 7.8
Find the best straight line fit for the data in the previous example. Now however, assume that we
have five times the confidence in the accuracy of the final two data points, relative to the other points.
Define a square weighting matrix W:
1 0 0 0 0
0 1 0 0 0
W= 0 0 1 0 0 .
0 0 0 5 0
0 0 0 0 5
A·a = b,
W·A·a = W · b,
T
(W · A) · W · A · a = (W · A)T · W · b,
−1
T T
a = (W · A) · W · A (W · A) · W · b.
x = 8.0008 + 1.1972 t.
A plot of the raw data and the best fit straight line is shown in Fig. 7.20.
When the measurements are independent and equally reliable, W is the identity matrix.
If the measurements are independent but not equally reliable, W is at most diagonal. If the
measurements are not independent, then non-zero terms can appear off the diagonal in W.
It is often advantageous, for instance in problems in which one wants to control a process in
real time, to give priority to recent data estimates over old data estimates and to continually
employ a least squares technique to estimate future system behavior. The previous example
does just that. A famous fast algorithm for such problems is known as a Kalman20 Filter.
It is common and useful at times to fit data to a power law form, especially when the
data range over wide orders of magnitude. For clean units, it is advisable to scale both x
20
Rudolf Emil Kálmán, 1930-2016, Hungarian-American electrical engineer.
308
20
12
x
8
x = 8.0008 + 1.1972 t
2 4 6
t
Figure 7.20: Plot of x − t data and best weighted least squares straight line fit.
and t by characteristic values. Sometimes this is obvious, and sometimes it is not. Whatever
the case, the following form can usually be found
a2
x(t) t
= a1 .
xc tc
Here, x is a dependent variable, t is an independent variable, xc is a characteristic value of x
(perhaps its maximum), tc is a characteristic value of t (perhaps its maximum), and a1 and
a2 are curve fit parameters. This fit is not linear in the coefficients, but can be rendered so
by taking the logarithm of both sides to get
a2
x(t) t t
ln = ln a1 = ln(a1 ) + a2 ln .
xc tc tc
Often times one must not include values at t = 0 because of the logarithmic singularity
there.
Example 7.9
An experiment yields some data, shown next.
t(s) x(nm)
0.0 0.0
1 × 10−3 1 × 100
1 × 10−2 5 × 101
1 × 100 3 × 105
1 × 101 7 × 109
1 × 102 8 × 1010
309
Analyze.
A plot of the raw data is shown in Fig. 7.21. Notice that the linear plot obscures the data at small
x 1010
8 1012
a) b)
6
108
x (nm)
x (nm)
4
104
2
100
20 40 60 80 100 10-4 10-2 100 102
t (s) t (s)
time, while the log-log plot makes the trends more clear. Now, to get a curve fit for the log-log plot, we
assume a power law form. We first eliminate the point at the origin, then scale the data, in this case
by the maximum values of t and x, and take appropriate logarithms to get to following values.
t x
t (s) x (nm) t/tmax x/xmax ln tmax ln xmax
1 × 10−3 1 × 100 1 × 10−5 1.25 × 10−11 −11.5129 −25.1053
1 × 10−2 5 × 101 1 × 10−4 6.25 × 10−10 −9.2013 −21.1933
1 × 100 3 × 105 1 × 10−2 3.75 × 10−6 −4.6052 −12.4938
1 × 101 7 × 109 1 × 10−1 8.75 × 10−2 −2.3026 −2.4361
1 × 102 8 × 1010 1 × 100 1 × 100 0.0000 0.0000
310
In matrix form, this becomes
1 −11.5129 −25.1053
1 −9.2013 −21.1933
ln a1
1
−4.6052 = −12.4938
a 2
.
1 −2.3026 −2.4361
1 0.0000 0.0000
This is of the form
A · a = b.
T
As before, we multiply both sides by A and then solve for a, we get
−1
a = AT · A · AT · b.
Solving, we find
0.4206
a= .
2.2920
So that
ln a1 = 0.4206, a2 = 2.2920,
or
a1 = 1.5228.
So the power law curve fit is
2.2920
x(t) t
= 1.5228 ,
8.000 × 1010 nm 100 s
or 2.2920
11
t
x(t) = 1.2183 × 10 nm .
100 s
A plot of the raw data and curve fit is shown in Fig. 7.22.
As long as the assumed form for the curve fit is linear in the coefficients, it is straight-
forward to extend to high order curve fits as demonstrated in the following example.
Example 7.10
An experiment yields the data that follows.
t x
0.0 1.0
0.7 1.6
0.9 1.8
1.5 2.0
2.6 1.5
3.0 1.1
Find the least squares best fit coefficients a1 , a2 , and a3 if the assumed functional form is
311
1012
x (nm) 108
104
100
10-4
10-4 10-2 100 102
t (s)
Figure 7.22: Plot of raw x − t data and power law curve fit to the data: x(t) =
2.2920
(1.2183 × 1011 nm) 100t s .
1. x = a1 + a2 t + a3 t2 ,
2. x = a1 + a2 sin 6t + a3 sin t
3 .
Plot on a single graph the data points and the two best fit estimates. Which best fit estimate has the
smallest least squares error?
• x = a 1 + a 2 t + a 3 t2 :
We substitute each data point into the assumed form and get the following set of linear equations
312
This is of the form
A · a = b.
As before, we multiply both sides by AT and then solve for a to get
−1
a = AT · A · AT · b.
Solving, we find
0.9778
a = 1.2679 .
−0.4090
So the best quadratic curve fit to the data is
This form has applied a bit of intuition. The curve looks like a sine wave of wavelength 6 that has
been transposed. So we suppose it is of such a form. The term a1 is the transposition; the term on
a2 is the fundamental frequency, also known as the first harmonic, that fits in the domain; the term
on a3 is the second harmonic; we have thrown that in for good measure.
We substitute each data point into the assumed form and get the following set of linear equations
0.0 0.0
1.0 = a1 + a2 sin + a3 sin ,
6 3
0.7 0.7
1.6 = a1 + a2 sin + a3 sin ,
6 3
0.9 0.9
1.8 = a1 + a2 sin + a3 sin ,
6 3
1.5 1.5
2.0 = a1 + a2 sin + a3 sin ,
6 3
2.6 2.6
1.5 = a1 + a2 sin + a3 sin ,
6 3
3.0 3.0
1.1 = a1 + a2 sin + a3 sin .
6 3
This can be rewritten as
1 0.0 0.0 1.0
1 0.1164 0.2312 1.6
a1
1 0.1494 0.2955
a2 = 1.8 .
1 0.2474 0.4794 2.0
a3
1 0.4199 0.7622 1.5
1 0.4794 0.8415 1.1
This is of the form
A · a = b.
As before, we multiply both sides by AT and then solve for a, we get
−1
a = AT · A · AT · b.
313
Solving, we find
1.0296
a = −37.1423 .
21.1848
So the best curve fit for this form is
t t
x(t) ∼ 1.0296 − 37.1423 sin + 21.1848 sin .
6 3
two-term sinusoidal
2 curve fit
quadratic
x
polynomial
curve fit
1 2 3
t
Figure 7.23: Plot of x − t data and two least squares curve fits x(t) ∼ 0.9778 + 1.2679t −
0.4090t2 , and x(t) ∼ 1.0296 − 37.1423 sin (t/6) + 21.1848 sin (t/3).
314
Bibliography
M. M. Abbott and H. C. van Ness, 1989, Thermodynamics with Chemical Applications, Second Edition,
Schaum’s Outline Series in Engineering, McGraw-Hill, New York.
First published in 1972, this is written in the style of all the Schaum’s series. It has extensive solved
problems and a crisp rigorous style that is readable by undergraduate engineers. It has a chemical
engineering emphasis, but is also useful for all engineers.
J. D. Anderson, 2019, Hypersonic and High-Temperature Gas Dynamics, Third Edition, AIAA, Reston,
Virginia.
First published in 1989, this readable text includes details of how equilibrium and non-equilibrium
thermodynamics influences hypersonic fluid mechanics.
J. D. Anderson, 2020, Modern Compressible Flow with Historical Perspective, Fourth Edition, McGraw-Hill,
New York.
First published in 1982, this readable text includes essential thermodynamics for high Mach number
fluid mechanics.
P. Atkins, 2010, The Laws of Thermodynamics: a Very Short Introduction, Oxford University Press, Oxford.
This is a short book by an eminent chemist summarizing the foundations of thermodynamics for an
interested general reader.
A. Bejan, 2016, Advanced Engineering Thermodynamics, Fourth Edition, John Wiley, Hoboken, New Jersey.
This is an advanced undergraduate text first published in 1988. It gives a modern treatment of
the science of classical thermodynamics. It does not confine its attention to traditional engineering
problems, and considers applications across biology and earth sciences as well; some readers will
find parts of the discussion to be provocative, as received wisdom is occasionally challenged. The
thermodynamics of irreversible processes are discussed in detail.
R. S. Berry, S. A. Rice, and J. Ross, 2000, Physical Chemistry, Second Edition, Oxford University Press,
Oxford.
This is a rigorous general text in physical chemistry at a senior or first year graduate level. First
appearing in 1980, it has a full treatment of classical and statistical thermodynamics as well as
quantum mechanics.
C. Bohren and B. Albrecht, 2023, Atmospheric Thermodynamics, Second Edition, Oxford University Press,
Oxford.
This is readable advanced undergraduate text on thermodynamics as it applies to the atmosphere,
especially weather. First appearing in 1998, it adopts an irreverent style to treat some very serious
and important subjects with rigor.
L. Boltzmann, 1995, Lectures on Gas Theory, Dover, New York.
This is a detailed monograph by the founding father of statistical thermodynamics. This edition is a
translation of the original German version, Gastheorie, from 1896-1898.
315
C. Borgnakke and R. E. Sonntag, 2017, Fundamentals of Thermodynamics, Ninth Edition, John Wiley,
New York.
This classic and popular undergraduate mechanical engineering text, which in earlier editions was
authored by J. G. van Wylen and Sonntag, has stood the test of time and has a full treatment of
most classical problems. Its first edition appeared in 1965. The tenth edition unfortunately relegates
mixture theory to an on-line supplement.
M. Born, 1949, Natural Philosophy of Cause and Chance, Clarendon Press, Oxford.
This monograph is a summary of the Waynflete Lectures delivered at Oxford University in 1948 by
the author, the winner of the 1954 Nobel Prize in physics. The lectures consider various topics, and
include an important chapter on thermodynamics with the author’s earlier defense of the approach of
Carathèodory playing a prominent role.
R. Boyle, 2003, The Sceptical Chymist, Dover, New York.
This is a reprint of the original 1661 work of the famous early figure of the scientific revolution.
P. W. Bridgman, 1943, The Nature of Thermodynamics, Harvard, Cambridge.
This short monograph by the winner of the 1946 Nobel Prize in physics gives a prosaic introduction
to thermodynamics which is directed at a scientifically literate audience who are not interested in
detailed mathematical exposition.
H. B. Callen, 1985, Thermodynamics and an Introduction to Thermostatistics, Second Edition, John Wiley,
New York.
This advanced undergraduate text, an update of the 1960 original, has an emphasis on classical physics
applied to thermodynamics, with a few chapters devoted to quantum and statistical foundations.
S. Carnot, 2005, Reflections on the Motive Power of Fire, Dover, New York.
This is a translation of the author’s foundational 1824 work on the heat engines, Réflexions sur la
Puissance Motrice du Feu et sur les Machines propres à développer cette Puissance (“Reflections on
the Motive Power of Fire and on Machines Fitted to Develop that Power”). Also included is a paper
of Clausius.
Y. A. Çengel and M. A. Boles, 2023, Thermodynamics: An Engineering Approach, Tenth Edition, McGraw-
Hill, Boston.
This popular undergraduate mechanical engineering text which first appeared in 1989 has most of the
features expected in a modern book intended for a large and varied audience.
S. Chandrasekhar, 2010, An Introduction to the Study of Stellar Structure, Dover, New York.
This monograph, first published in 1939, on astrophysics is by the winner of the 1983 Nobel Prize
in physics. It has large sections devoted to rigorous axiomatic classical thermodynamics in the style
of Carathèodory, highly accessible to engineering students, which show how thermodynamics plays a
critical role in understanding the physics of the heavens.
R. J. E. Clausius, 2008, The Mechanical Theory of Heat, Kessinger, Whitefish, Montana.
This is a reprint of the 1879 translation of the 1850 German publication of the great German scientist
who in many ways founded classical thermodynamics.
S. R. de Groot and P. Mazur, 1984, Non-Equilibrium Thermodynamics, Dover, New York.
This is an influential monograph, first published in 1962, which summarizes much of the work of the
famous Belgian school of thermodynamics. It is written at a graduate level and has a strong link to
fluid mechanics and chemical reactions.
316
E. Fermi, 1936, Thermodynamics, Dover, New York.
This short 160 page classic clearly and efficiently summarizes the fundamentals of thermodynamics.
It is based on a series of lectures given by this winner of the 1938 Nobel Prize in physics. The book
is highly recommended, and the reader can benefit from multiple readings.
R. P. Feynman, R. B. Leighton, and M. Sands, 1963, The Feynman Lectures on Physics, Volume 1, Addison-
Wesley, Reading, Massachusetts.
This famous series documents the introductory undergraduate physics lectures given at the California
Institute of Technology by the lead author, the 1965 Nobel laureate in physics. Famous for their clarity,
depth, and notable forays into challenging material, they treat a wide range of topics from classical
to modern physics. Volume 1 contains chapters relevant to classical and modern thermodynamics.
J. B. J. Fourier, 2009, The Analytical Theory of Heat, Cambridge, Cambridge.
This reprint of the 1878 English translation of the 1822 French original, Théorie Analytique de la
Chaleur is a tour de force of science, engineering, and mathematics. It predates Carnot and the
development of the first and second laws of thermodynamics, but nevertheless successfully develops
a theory of non-equilibrium thermodynamics fully consistent with the first and second laws. In so
doing, the author makes key advances in the formulation of partial differential equations and their
solution by what are now known as Fourier series.
J. W. Gibbs, 1957, The Collected Works of J. Willard Gibbs, Yale U. Press, New Haven.
This compendium gives a complete reproduction of the published work of the monumental American
engineer and scientist of the nineteenth century, including his seminal work on classical and statistical
thermodynamics.
W. T. Grandy, 2008, Entropy and the Time Evolution of Macroscopic Systems, Oxford, Oxford.
This monograph gives an enlightening description, fully informed by both historical and modern
interpretations, of entropy and its evolution in the context of both continuum and statistical theories.
E. A. Guggenheim, 1933, Modern Thermodynamics by the Methods of Willard Gibbs, Methuen, London.
This graduate-level monograph was important in bringing the work of Gibbs to a wider audience.
E. P. Gyftopoulos and G. P. Beretta, 2005, Thermodynamics: Foundations and Applications, Dover, Mine-
ola, New York.
This beginning graduate text has a rigorous development of classical thermodynamics.
C. S. Helrich, 2009, Modern Thermodynamics with Statistical Mechanics, Springer, Berlin.
This is an advanced undergraduate text aimed mainly at the physics community. The author includes a
full treatment of classical thermodynamics and moves easily into statistical mechanics. The author’s
undergraduate training in engineering is evident in some of the style of the text which should be
readable by most undergraduate engineers after a first class in thermodynamics.
J. O. Hirschfelder, C. F. Curtis, and R. B. Bird, 1954, Molecular Theory of Gases and Liquids, Wiley, New
York.
This comprehensive tome is a valuable addition to any library of thermal science. Its wide ranging
text covers equations of state, molecular collision theory, reactive hydrodynamics, reaction kinetics,
and many other topics all from the point of view of careful physical chemistry. Much of the work
remains original.
J. R. Howell and R. O. Buckius, 1992, Fundamentals of Engineering Thermodynamics, Second Edition,
McGraw-Hill, New York.
This is a good undergraduate text for mechanical engineers; it first appeared in 1987.
317
W. F. Hughes and J. A. Brighton, 1999, Fluid Dynamics, Third Edition, Schaum’s Outline Series, McGraw-
Hill, New York.
Written in the standard student-friendly style of the Schaum’s series, this discussion of fluid mechanics
includes two chapters on compressible flow which bring together fluid mechanics and thermodynamics.
It first appeared in 1967.
F. P. Incropera and D. P. DeWitt, 1981, Fundamentals of Heat and Mass Transfer, First Edition, John
Wiley, New York.
This first edition of 1981 has evolved over several years and is the standard undergraduate text
in heat transfer; the first edition however properly recognizes its original authors. The interested
thermodynamics student will find the subject of heat transfer builds in many ways on a classical
thermodynamics foundation. Especially relevant to thermodynamics are chapters on boiling and
condensation, heat exchangers, as well as extensive tables of thermal properties of real materials.
J. H. Keenan, F. G. Keyes, P. G. Hill, and J. G. Moore, 1985, Steam Tables: Thermodynamic Properties
of Water Including Vapor, Liquid, and Solid Phases, John Wiley, New York.
This is a version of the original 1936 data set which is the standard for water’s thermodynamic
properties.
J. Kestin, 1966, A Course in Thermodynamics, Blaisdell, Waltham, Massachusetts.
This is a foundational textbook that can be read on many levels. All first principles are reported in a
readable fashion. In addition the author makes a great effort to expose the underlying mathematical
foundations of thermodynamics.
R. J. Kee, M. E. Coltrin, P. Glarborg, and H. Zhu, 2018, Chemically Reacting Flow: Theory, Modeling,
and Simulation, Second Edition, John Wiley, Hoboken, New Jersey.
This comprehensive text gives an introductory graduate level discussion of fluid mechanics, thermo-
chemistry, and finite rate chemical kinetics. The focus is on low Mach number reacting flows, and
there is significant discussion of how to achieve computational solutions.
C. Kittel and H. Kroemer, 1980, Thermal Physics, Second Edition, Freeman, San Francisco.
This is a classic undergraduate text, first introduced in 1970; it is mainly aimed at physics students.
It has a good introduction to statistical thermodynamics and a short effective chapter on classical
thermodynamics. Engineers seeking to broaden their skill set for new technologies relying on micro-
scale thermal phenomena can use this text as a starting point.
D. Kondepudi and I. Prigogine, 2015, Modern Thermodynamics: From Heat Engines to Dissipative Struc-
tures, Second Edition John Wiley, New York.
This is a detailed modern exposition which exploits the authors’ unique vision of thermodynamics with
both a science and engineering flavor. The authors, the second of whom is one of the few engineers
who was awarded the Nobel Prize (chemistry 1977, for the work summarized in this text), often
challenge the standard approach to teaching thermodynamics, and make the case that the approach
they advocate, with an emphasis on non-equilibrium thermodynamics, is better suited to describe
natural phenomena and practical devices than the present approach, which is generally restricted to
equilibrium states.
K. K. Kuo, 2005, Principles of Combustion, Second Edition, John Wiley, New York.
This is a readable graduate level engineering text for combustion fundamentals. First published in
1986, it includes a full treatment of reacting thermodynamics as well as discussion of links to fluid
mechanics.
K. J. Laidler, 1987, Chemical Kinetics, Third Edition, Prentice-Hall, Upper Saddle River, NJ.
This is a standard advanced undergraduate chemistry text on the dynamics of chemical reactions. It
first appeared in 1965.
318
L. D. Landau and E. M. Lifshitz, 2000, Statistical Physics, Part 1, Volume 5 of the Course of Theoretical
Physics, Third Edition, Butterworth-Heinemann, Oxford.
This book, part of the monumental series of graduate level Russian physics texts, first published in
English in 1951 from Statisticheskaya fizika, gives a fine introduction to classical thermodynamics as
a prelude to its main topic, statistical thermodynamics in the spirit of Gibbs.
B. H. Lavenda, 1978, Thermodynamics of Irreversible Processes, John Wiley, New York.
This is a lively and opinionated monograph describing and commenting on irreversible thermodynam-
ics. The author is especially critical of the Prigogine school of thought on entropy production rate
minimization.
A. Lavoisier, 1984, Elements of Chemistry, Dover, New York.
This is the classic treatise by the man known as the father of modern chemistry, translated from the
1789 Traité Élementaire de Chimie, which gives the first explicit statement of mass conservation in
chemical reactions.
G. N. Lewis and M. Randall, 1961, Thermodynamics and the Free Energy of Chemical Substances, Second
Edition, McGraw-Hill, New York.
This book, first published in 1923, was for many years a standard reference text of physical chemistry.
H. W. Liepmann and A. Roshko, 2002, Elements of Gasdynamics, Dover, New York.
This is an influential text in compressible aerodynamics that is appropriate for seniors or beginning
graduate students. First published in 1957, it has a strong treatment of the physics and thermody-
namics of compressible flow along with elegant and efficient text. Its treatment of both experiment
and the underlying theory is outstanding, and in many ways is representative of the approach to
engineering sciences fostered at the California Institute of Technology, the authors’ home institution.
J. C. Maxwell, 2001, Theory of Heat, Dover, New York.
This is a short readable book by the nineteenth century master, first published in 1871. Here the
mathematics is minimized in favor of more words of explanation.
M. J. Moran, H. N. Shapiro, D. D. Boettner, and M. B. Bailey, 2018, Fundamentals of Engineering Ther-
modynamics, Ninth Edition, John Wiley, New York.
This is a standard undergraduate engineering thermodynamics text, and one of the more popular.
First published in 1988, it has much to recommend it including good example problems, attention to
detail, good graphics, and a level of rigor appropriate for good undergraduate students.
P. M. Morse, 1969, Thermal Physics, Second Edition, Benjamin, New York.
This is a good undergraduate book on thermodynamics from a physics perspective. It covers classical
theory well in its first sections, then goes on to treat kinetic theory and statistical mechanics. It first
appeared in 1964.
I. Müller and T. Ruggeri, 1998, Rational Extended Thermodynamics, Springer-Verlag, New York.
This modern, erudite monograph gives a rigorous treatment of some of the key issues at the frontier
of modern continuum thermodynamics.
I. Müller and W. Weiss, 2005, Entropy and Energy, Springer-Verlag, New York.
This is a unique treatise on fundamental concepts in thermodynamics. The author provide mathe-
matical rigor, historical perspective, and examples from a diverse set of scientific fields.
I. Müller, 2007, A History of Thermodynamics: the Doctrine of Energy and Entropy, Springer-Verlag,
Berlin.
The author gives a readable text at an advanced undergraduate level which highlights some of the
many controversies of thermodynamics, both ancient and modern.
319
I. Müller and W. H. Müller, 2009, Fundamentals of Thermodynamics and Applications: with Historical
Annotations and Many Citations from Avogadro to Zermelo, Springer, Berlin.
The author presents an eclectic view of classical thermodynamics with much discussion of its history.
The text is aimed at a curious undergraduate who is unsatisfied with industrial-strength yet narrow
and intellectually vapid textbooks.
W. Nernst, 1969, The New Heat Theorem: Its Foundation in Theory and Experiment, Dover, New York.
This monograph gives the author’s exposition of the development of the third law of thermodynamics.
It first appeared in English translation in 1917 and was originally published in German.
S. Paolucci, 2019, Undergraduate Lectures on Thermodynamics, BreviLiber.
This short book, the first of two on this subject, gives small individual chapters covering material
similar to what one might find in a traditional lecture.
S. Paolucci, 2019, Undergraduate Lectures on Intermediate Thermodynamics, BreviLiber.
This short book, the second of two on this subject, gives small individual chapters covering material
similar to what one might find in a traditional lecture.
W. Pauli, 2000, Thermodynamics and the Kinetic Theory of Gases, Dover, New York.
This is a monograph on statistical thermodynamics by the winner of the 1945 Nobel Prize in physics.
It is actually derived from his lecture course notes given at ETH Zurich, as compiled by a student in
his class, E. Jucker, published in 1952.
M. Planck, 1990, Treatise on Thermodynamics, Dover, New York.
This brief book, which originally appeared in German in 1897, gives many unique insights from the
great scientist who was the winner of the 1918 Nobel Prize in physics. It is rigorous, but readable by
an interested undergraduate student.
H. Poincaré, 1892, Thermodynamique: Leçons Professèes Pendant le Premier Semestre 1888-89, Georges
Carré, Paris.
This text of classical undergraduate thermodynamics has been prepared by one of the premier math-
ematicians of the nineteenth century.
J. M. Powers, 2016, Combustion Thermodynamics and Dynamics, Cambridge, New York.
This is a graduate level text on combustion and its interplay with both thermodynamics and dynamics.
It uses a large portion of the present course notes as a foundation for its discussion of thermodynamics
of reacting mixtures.
J. M. Powers, 2024, Mechanics of Fluids, Cambridge, Cambridge, UK.
This is a graduate level text on fluid mechanics. It has significant treatment of the thermodynamics
of single phase fluids including treatments of the first and second laws of thermodynamics, non-ideal
gases, the Clausius-Duhem relation, and Onsager reciprocity.
I. Prigogine, 1967, Introduction to Thermodynamics of Irreversible Processes, Third Edition, Interscience,
New York.
This is a famous book that summarizes the essence of the work of the Belgian school for which the
author was awarded the 1977 Nobel Prize in chemistry. This book first appeared in 1955.
W. J. M. Rankine, 1908, A Manual of the Steam Engine and Other Prime Movers, Seventeenth Edition,
Griffin, London.
This in an accessible undergraduate text for mechanical engineers of the nineteenth century. It first
appeared in 1859. It contains much practical information on a variety of devices, including steam
engines.
320
L. E. Reichl, 2016, A Modern Course in Statistical Physics, Fourth Edition, John Wiley, New York.
This full service graduate text has a good summary of key concepts of classical thermodynamics and
a strong development of modern statistical thermodynamics.
O. Reynolds, 1903, Papers on Mechanical and Physical Subjects, Volume III, The Sub-Mechanics of the
Universe, Cambridge, Cambridge.
This volume compiles various otherwise unpublished notes of Reynolds and includes his detailed deriva-
tions of general equations of conservation of mass, momentum, and energy employing his transport
theorem.
W. C. Reynolds, 1968, Thermodynamics, Second Edition, McGraw-Hill, New York.
This is an unusually good undergraduate text written for mechanical engineers. The author has
wonderful qualitative problems in addition to the usual topics in such texts. A good introduction
to statistical mechanics is included as well. This particular edition is highly recommended; the first
edition appeared in 1965.
W. C. Reynolds and P. Colonna, 2018, Thermodynamics: Fundamentals and Engineering Applications,
Cambridge, New York.
This is a modern and thorough update of the 1965 edition of the first author. It is an outstanding
treatment of modern thermodynamics.
S. I. Sandler, 2017, Chemical, Biochemical, and Engineering Thermodynamics, Fifth Edition, John Wiley,
Hoboken, New Jersey.
This is an advanced undergraduate text in thermodynamics from a chemical engineering perspective
with a good mathematical treatment. It first appeared in 1977.
E. Schrödinger, 1989, Statistical Thermodynamics, Dover, New York.
This is a short monograph written by the one of the pioneers of quantum physics, the co-winner of
the 1933 Nobel Prize in Physics. It is based on a set of lectures delivered to the Dublin Institute for
Advanced Studies in 1944, and was first published in 1946.
A. H. Shapiro, 1953, The Dynamics and Thermodynamics of Compressible Fluid Flow, Vols. I and II, John
Wiley, New York.
This classic two volume set has a comprehensive treatment of the subject of its title. It has numerous
worked example problems, and is written from a careful engineer’s perspective.
J. M. Smith, H. C. van Ness, and M. Abbott, 2017, Introduction to Chemical Engineering Thermodynamics,
Eighth Edition, McGraw-Hill, New York.
This is probably the most common undergraduate text in thermodynamics in chemical engineering.
It first appeared in 1959. It is rigorous and has went through many revisions.
A. Sommerfeld, 1956, Thermodynamics and Statistical Mechanics, Lectures on Theoretical Physics, Vol. V,
Academic Press, New York.
This is a compilation of the author’s lecture notes on this subject. The book reflects the author’s
stature of a leader of theoretical physics of the twentieth century who trained a generation of students
(e.g. Nobel laureates Heisenberg, Pauli, Debye and Bethe). The book gives a fine description of
classical thermodynamics with a seamless transition into quantum and statistical mechanics.
J. W. Tester and M. Modell, 1997, Thermodynamics and Its Applications, Third Edition, Prentice Hall,
Upper Saddle River, New Jersey.
First appearing in 1974, this entry level graduate text in thermodynamics is written from a chemical
engineer’s perspective. It has a strong mathematical development for both classical and statistical
thermodynamics.
321
K. S. Thorne and R. D. Blandford, 2017, Modern Classical Physics: Optics, Fluids, Plasmas, Elasticity,
and Statistical Physics, Princeton University Press, Princeton, New Jersey.
This comprehensive text is accessible to advanced undergraduates who appreciate a physics perspec-
tive. It treats a wide range of topics that include and transcend thermodynamics.
C. A. Truesdell, 1980, The Tragicomic History of Thermodynamics, 1822-1854, Springer Verlag, New York.
This idiosyncratic monograph has a lucid description of the history of nineteenth century thermal
science. It is written in an erudite fashion, and the reader who is willing to dive into a difficult subject
will be rewarded for diligence by gain of many new insights.
C. A. Truesdell, 1984, Rational Thermodynamics, Second Edition, Springer-Verlag, New York.
This is an update on the evolution of classical thermodynamics in the twentieth century. The book
itself first appeared in 1969. The second edition includes additional contributions by some contempo-
raneous leaders of the field.
S. R. Turns and D. C. Haworth, 2020, An Introduction to Combustion: Concepts and Applications, Fourth
Edition, McGraw-Hill, Boston.
This is a senior-level undergraduate text on combustion which uses many notions from thermodynam-
ics of mixtures. It first appeared in 1996.
S. R. Turns and L. L. Pauley, 2020, Thermodynamics: Concepts and Applications, Second Edition, Cam-
bridge University Press, New York.
This is a well prepared undergraduate text on thermodynamics. It first appeared in 2006.
H. C. van Ness, 1983, Understanding Thermodynamics, Dover, New York.
This is a short readable monograph from a chemical engineering perspective. It first appeared in 1969.
W. G. Vincenti and C. H. Kruger, 1976, Introduction to Physical Gas Dynamics, Krieger, Malabar, Florida.
This graduate text on high speed non-equilibrium flows contains a good description of the interplay
of classical and statistical mechanics. There is an emphasis on aerospace science and fundamental
engineering applications. It first appeared in 1965.
F. M. White, 2020, Fluid Mechanics, Ninth Edition, McGraw-Hill, Boston.
This standard undergraduate fluid text draws on thermodynamics in its presentation of the first law
and in its treatment of compressible flows. It first appeared in 1979.
B. Woodcraft, 1851, The Pneumatics of Hero of Alexandria, London, Taylor Walton and Maberly.
This is a translation and compilation from the ancient Greek of the work of Hero (10 A.D.-70 A.D.).
The discussion contains descriptions of the engineering of a variety of technological devices including
a primitive steam engine. Other devices which convert heat into work are described as well.
L. C. Woods, 1975, The Thermodynamics of Fluid Systems, Clarendon, Oxford.
This graduate text gives a good, detailed survey of the thermodynamics of irreversible processes,
especially related to fluid systems in which convection and diffusion play important roles.
322