You are on page 1of 43

Chemical Product and Process

Modeling
Volume 3, Issue 1 2008 Article 24

Thermodynamics and the Simulation Engineer


Marco A. Satyro∗


Virtual Materials Group, Inc., msatyro@ucalgary.ca

Copyright 2008
c The Berkeley Electronic Press. All rights reserved.
Thermodynamics and the Simulation
Engineer∗
Marco A. Satyro

Abstract

In this paper a brief, commented introduction to thermodynamics highlighting its connections


with the simulation of chemical processes is presented. With these connections established I pro-
ceed to comment on current challenges faced by process and product modelers and conclude with a
peek into the future of process and product design and the role of thermodynamics as the unifying
discipline for engineers.

KEYWORDS: thermocynamics, process simulation, simulation engineering, product and process


design


The permission of Virtual Materials Group, Inc. to publish this work is gratefully acknowledged.
The author is indebted to the kindness and patience of the editors due to delays in delivering the
manuscript caused by work constraints.
Satyro: Thermodynamics and the Simulation Engineer

Introduction
It is a well-known fact that process simulators are heavily dependent on
thermodynamics and physical property models. It is also well known that process
simulators usually have large tomes of written documentation talking about
physical properties and thermodynamics. Experience has shown me that at least
these sections of manuals go mostly unread although the simulation results are
completely dependent on the underlining thermodynamic models for phase
equilibrium and physical property calculations as defined by the chosen property
package.
Property packages are the workhorses of process and reservoir simulators
and in a nutshell they are a collection of methods bounded by the first and second
laws of thermodynamics and designed to solve a specific class of problems.
Going deeper, the essence of a property package is its ability to determine the
state of minimum free energy of a system, and when that is determined it must
have the ability to compute physical and transport properties and provide an
accurate accounting of mass as it distributes across different phases. For example,
a set of components and chemical reactions and a couple of independent variables
such as temperature and pressure define a chemical system. A properly
constructed property package will then determine the number of phases at
thermodynamic equilibrium, the amounts and compositions1 of each phase at
equilibrium and physical properties of each phase. In actuality, thermodynamic
related calculations consume the lion’s share of clock cycles during any non-
trivial simulation and they govern the numbers and the quality of the numbers one
obtains.
As process simulators have evolved, more and more sophistication has
been built into unit operation models, graphical user interfaces and also
thermodynamic models. One interesting shift that has happened over the years is a
change in the knowledge background of engineers using simulators. When I
started in this business close to 30 years ago, engineers using simulators would be
quite proficient at programming, process engineering and thermodynamics (Chien
and Null, 1972; Null, 1980). As engineering practice evolves, engineer’s interests
also change and nowadays we have just as bright engineers, but with a rather
different background. Most young engineers these days do little programming and
they do not seem to see much chemistry or physical chemistry during their
formative years. Thus good part of the engineering they learn comes from the
ubiquitous use of simulation tools (Svrcek and Satyro, 2006)

1
Thermodynamic equilibrium here is applied on its broadest sense and the state of minimum free
energy is sought not only by varying the amounts of moles chemical components in the phases due
to diffusion but also compositional variations due to chemical reactions.

Published by The Berkeley Electronic Press, 2008 1


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

Extensive use of simulators is not necessarily a bad thing; simulators are


wonderful tools to teach natural gas and refinery technologies, heat integration,
distillation and absorption and to boot simulators are great virtual laboratories.
That said, simulation is not the real thing, and good users of process simulation
are the ones who combine expert use of the computer tools with deep knowledge
of the rudiments of chemical engineering – thermodynamics, physical chemistry,
stoichiometry and reaction engineering. These rudiments were very well
established more than half a century ago and they are the most valuable
commodities today. Computers are notorious for generating a large amount of
information, but sometimes that information does not make sense2.
Since most engineers being trained today are proficient users of simulators
from the onset of their professional careers and will use them routinely throughout
their professional lives, I think it is appropriate to use the term simulation
engineer for the majority of process engineers today. Our objective is to take a
high level tour of the field and we will stop to examine an interesting issue here
and there. During the process we will stress fundamental points I believe are
important for the intelligent use of process simulators.
It is important to stress that by necessity this is a biased paper since it is
based on my experience helping engineers model gas plants, refineries,
petrochemical and chemical complexes. Nevertheless, the same principles apply
to many other interesting areas that use simulation like pharmaceutical synthesis,
corrosion engineering and reservoir engineering.

Why Thermodynamics is Relevant for Simulation?


Although we all state the fact that thermodynamics is important for process
simulation, the fundamental reason has to do with a special feature of classical
thermodynamics. Classical thermodynamics is a model free science and as such, it
is not bound by the accuracy of specific models and theories. This way of looking
at Nature provides us with a powerful tool because even though we may not be
able to understand the details related to a specific fluid or process, classical
thermodynamics allow us to write exact relationships between quantities of
interest, and with these relationships (and experiments or models or as it is usually
the case a combination of models and data) we can perform material and energy
balances.
This power comes with strings attached. Perhaps the biggest issue with the
framework provided by classical thermodynamics is the fact that implicitly we
always assume processes to be reversible. We know that real processes are
irreversible. These irreversibilities may come from different sources such as
poorly designed equipment (such as in distillation trays that show large

2
John Prausnitz put it best: “The quality of a computer is speed, not intelligence.”

http://www.bepress.com/cppm/vol3/iss1/24 2
Satyro: Thermodynamics and the Simulation Engineer

entrainment or weeping) to energy dissipation coming from internal friction due


to viscosity. The snag about trying to model thermodynamic processes taking into
account irreversibilities is a question of ignorance. Bluntly stated, we usually do
not know the minute details that are required to model a process taking into
account dissipative forces, at least in a readily accessible and cheap way.
When performing simulations we are usually interested in making broad
(but correct from the first and second laws point of view) statements related to the
material and energy balances a real operating unit would exhibit, but we usually
have rudimentary knowledge about the actual equipment being used (we may
know the number of trays and reflux ratio of a certain tower, but may have only
vague information about the tray geometry and performance curve of the reflux
pump). Even if we had this information easily available, the appropriate
computation of equipment performance invariably involves detailed fluid flow
analysis, and although computational fluid dynamics is an important tool, one
usually can not justify using such a sophisticated tool to complete the material
balance for a simple separation system3.
Thus enters the reversibility assumption. If we assume that the processes
can be represented as a reversible process, we then can use classical
thermodynamics to provide us with the relationships we need to calculate material
and energy balances. For example, when computing the duty for a condenser we
can simply calculate the difference between the enthalpy at the outlet and inlet of
the heat exchanger. The enthalpy is a thermodynamic property that can be defined
by the specification of two state variables (for example temperature and pressure)
and the composition of the fluid. With this we can calculate useful information
that allows us to proceed with design work.
Naturally this came with a price. We do not know exactly how the
equipment is going to perform in the field because we are not accounting for non-
idealities such as pressure drops, heat losses, etc. In exchange we can do useful
engineering work by making simple but judicious guesses about pressure drops,
mass transfer efficiencies, mechanical efficiencies and the like.
This is, in essence, the reason why thermodynamics is such a basic
ingredient of process simulation. Its formal structure allows is to draw exact
relationships between variables and correlate and extend data in a precise and
consistent manner. The assumption of reversibility is the key feature used to
create these precise relationships, and the relationships of classical

3
One could be malicious here and mention that even if we knew the construction details of a
certain piece of equipment and we could calculate the fluid flow phenomena accurately, this does
not mean we actually are modeling the real plant. Stories abound where distillation trays collapse
and are collected at the bottom of the tower during maintenance. I fondly recall a case where we
found a fossilized owl inside an ethanol concentration tower and no computational sophistication
could have taken that into account.

Published by The Berkeley Electronic Press, 2008 3


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

thermodynamics allow us to write mathematical expressions for the phase


equilibrium, chemical equilibrium and energy balance problems we encounter
when we try to express the behavior of a chemical plant using mathematical
models.
The use of this idealized, but precise framework allows us to move on
with the analysis of complex chemical engineering problems in a formal way
while adding an immense amount of order in the process. The removal of rate
issues from the problem makes it solvable using a very small amount of
information. For example computerized analysis of complex distillation problems
are now routine (Rong and Turunen 2006 a,b), and an example is shown in figure
1.
Incidentally, a thermodynamic property that is important for the
conceptual design of separation systems was discovered by the careful analysis of
the differential equations that describe the structure of simple distillation systems
(Ostwald, 1900; Schreinemakers, 1901; Doherty and Malone, 2001). This
property is the residue curve map and it describes the structure of the separation
space of mixtures and it has an especially strong significance for the design of
extractive and azeotropic distillation. A closely related thermodynamic property is
the distillation boundary present in a multi-component system. These boundaries
divide the separation space into regions and tell us what types of separations are
possible and impossible based on elementary information such as feed
composition and desired operating pressure ranges. It is quite remarkable that by
using the condition of thermodynamic equilibrium and simple models for the
liquid phase so much is already known and can be efficiently used for screening
of processes (Foucher et. al, 1991). Figure 2 shows a typical residue curve,
distillation boundary and distillation regions for azeotropic distillation.

Property Package Fundamentals


Now that we have a good idea of why classical thermodynamics is important for
process simulation we proceed to examine how thermodynamics is used for the
solution of material and energy balance problems. This is usually done using a
computer construct called a property package. Property package is the generic
term we use when we refer to a set of computer routines used to model the
thermo-physical behavior of fluids. Property packages are designed to solve the
following problems:

1. Determine the number of phases a fluid has when its global composition
and two extra intensive variables (such as temperature, pressure, molar
enthalpy, molar entropy or molar volume) are specified.
2. Determine the composition of each phase at equilibrium

http://www.bepress.com/cppm/vol3/iss1/24 4
Satyro: Thermodynamics and the Simulation Engineer

3. Determine the pressure, temperature and composition of each phase at


equilibrium
4. Determine volume of each phase
5. Determine the molar enthalpy, entropy and heat capacity of each phase
6. Determine transport properties of each phase
7. Determine interfacial properties

You will notice that transport properties are really not thermodynamic
properties since they depend on rate processes (e.g. rate of momentum
distribution as related to viscosity) but it is common to have these properties
estimated by property packages due to their importance for equipment design
(Ely, J.F. and Hanley, H.J.M, 1981).

The equilibrium problem


Classical thermodynamics provides us with the tools needed for the precise
mathematical formulation of thermodynamic equilibrium, and when we create a
property package we usually start by writing the models that, when applied to this
mathematical formulation, will give us in return useful numbers that describe
some important characteristic of the system. We start with the combined form of
the first and second laws of thermodynamics as shown by equation 1 (Prigogine
and Defay, 1954).

nr ⎛ ⎞
nc
⎛ ∂U ⎞ ∂U
dU = TdS − pdV + ∑ ⎜⎜ ⎟⎟ dni + ∑ ⎜ ⎟ dε j + ... 1
i ⎝ ∂ni ⎠ S ,V ,n

j =1 ⎝ ∂ε j

j ⎠ S ,V ,n ,ε k

Equation 1 is the starting point for pretty much all arguments related to
phase equilibrium and it captures the heat transfer, mechanical work and diffusion
phenomena as applied to a given system. It can be augmented to include
gravitational, electrical, magnetic, surface effects due to external force fields as
desired, but the idea is always the same – we want to express the energy content
and mass content of a body based on fluxes that enter or leave it carrying mass
and energy. At thermodynamic equilibrium the fluxes are zero and equation 1
give us, in principle, all the necessary tools we need to determine the equilibrium
condition based on the value of a certain set of variables, in this case the entropy
S, the volume V, the composition and the chemical reactions happening in the
body.
The snag with equation 1 is related to its natural variables. It is an
unfortunate fact that we do not have an “entropy meter” that could be used like a
thermometer, and an alternative variable has to be sought for the solution of
practical engineering problems. From a process engineering perspective, the best

Published by The Berkeley Electronic Press, 2008 5


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

variables are the ones easily measured, and those are usually the temperature and
pressure of a system. Fortunately we have exact mathematical techniques that
allow us to change the variables present in equation 1 by more convenient
variables. This technique is called Legendre transform (Modell and Reid, 1974)
and it is a procedure designed to replace the independent variable by its derivative
without loss of information. When we apply this technique to equation 1 we get
all the common expressions we are used to such as enthalpy, Helmholtz energy
and Gibbs free energy expressed in different independent variable sets and
completely equivalent from a mathematical perspective.
Although each of these expressions for the energy of a system is
equivalent, the independent variable set is not. It just happens that the Gibbs free
energy is the equivalent to equation 1 when the independent variables are
temperature and pressure, and it is written as equation 2. The only difference
between equations 1 and 2 is a reflection of our limitations as human beings.
Nature “knows” what the state of equilibrium should be and always works
towards it. We don’t and in order to try to figure it out we are bound by things we
can measure.

nr ⎛
nc
⎛ ∂G ⎞ ∂G ⎞⎟
dG = − SdT − Vdp + ∑ ⎜⎜ ⎟⎟ dni + ∑ ⎜ dε j + ... 2
i ⎝ ∂ni ⎠T , p ,n
⎜ ⎟
j =1 ⎝ ∂ε j ⎠
j T , p ,n ,ε k

Equation 2 is the starting point for the construction of virtually all process
engineering calculations and from a purely formal point of view we are done. We
end up spending most of our time creating models for the free energy as functions
of temperature, pressure and composition and solving the resulting equations for
some equilibrium situation we are interested in.
Note that equation 2 provides us with a rather complete description of
Nature. When a process has given up everything it can and it has exhausted its
sources of temperature gradients we will have no heat fluxes across it (dT = 0),
we will not be able to extract any mechanical work out of it due to movement of a
phase (dp = 0), all mass transfer that could happen will have happened and no
⎛ ∂G ⎞
compositional gradients will exist ( ⎜⎜ ⎟⎟ = 0, all i ) and all chemical reactions
⎝ ∂ni ⎠
that could happen will have happened to maximum possible extent
⎛ ∂G ⎞
(⎜ ⎟ = 0, all j ). This static picture of Nature painted by classical
⎜ ∂ε ⎟
⎝ j⎠
thermodynamics is its great achievement due to its precise mathematical
formulation and generality. It is also its greatest handicap since rates are

http://www.bepress.com/cppm/vol3/iss1/24 6
Satyro: Thermodynamics and the Simulation Engineer

convenient ignored, and introduces at a fundamental level the need for


engineering knowledge by users of process simulation software.
You see, according to equation 2, when we mix hydrogen and oxygen at
ambient conditions we will form water and release energy. This is an accurate
statement and there’s no escaping to its conclusion as long as time is not
important. Oxygen and hydrogen react at extremely slow rates and a bottle filled
with these two gases will need eons in order for any water to form. A small spark
or the addition of a catalyst will change the situation and allow the reaction to
proceed at an appreciable rate and allow the mixture to fulfill its fate as predicted.

Thermodynamics, Process and Product Modeling


The true power of thermodynamics is expressed when equation 2 is combined
with a good empirical model for phase equilibrium and physical property
estimation. There are many successful examples of thermodynamic modeling
applied to all types of chemical processes and the thermodynamics section of your
favorite process simulator will provide you with an inventory of useful
thermodynamic methods that have found their way into mainstream process
modeling. Instead of looking at an encyclopedic listing of successful
thermodynamic models, we will look at the main features of a successful class of
model that provides the thermodynamic background for the modeling of processes
of an important industrial segment – natural gas and oil processing.

Thermodynamics of Hydrocarbon Systems


To a great extent, our economy is a hydrocarbon based economy and in a quiet
and unassuming manner, this industry is shaped by equation 2 and simple but very
successful thermodynamic models that allow engineers to play the whole gamut –
conceptual design, design, detailed design, optimization, control design and
operator training – with confidence. This engineering success depends on two
simple observations:

1 Real molecules, in average, attract each other


2 Real molecules can not be compressed forever

The actual explanation for observation 1 is quite involved (Parsegian, 2006)


and can be achieved in a satisfactory manner only if we resort to a quantum
mechanical interpretation of molecular electrical field averages. For our purposes
it suffices to acknowledge that this is an experimental feature of Nature.
Observation 2 is more easily understood and is related to the fact that, to a first
approximation, molecules have a definitive size and therefore infinite
compression is not possible.

Published by The Berkeley Electronic Press, 2008 7


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

Using these two observations, scientists have been working on the creation of
models for fluids for hundreds of years, and these models can be expressed
conceptually by equation 3.
r
p = f (T , V , x ) 3

Equation 3 is what is called an equation of state and it is a model designed


to be plugged into equation 2 and in turn provide us with a compact description of
Nature. Before we proceed, did you ever ask yourself why equation 3 is written
that way? Mathematically we could just as easily write equation 4.
r
V = g (T , P, x ) 4

Equation 4 is in a form ideal for chemical engineers since usually pressure


and temperature are their bread and butter coordinates, it is kind of surprising to
see that equations of state are usually not cast in this form. A moment of
reflection shows that pressure and temperature are indeed convenient coordinates,
but they do not uniquely define the state of a saturated system. If you imagine a
pot of water boiling at one atmosphere pressure, the temperature will be 100
degrees centigrade, but we have two physically possible densities for the system,
one corresponding to a saturated vapor and another corresponding to a saturated
liquid. Equation 4 can provide one of the volumes, but not both. Equation 3 on the
other hand can model this phenomenon since two different values of volume can
correspond to a single pressure, in this case the vapor pressure of water at the
normal boiling temperature. This is illustrated in figure 3 and is a basic structural
characteristic of industrially useful equations of state.
The simplest mathematical structure that accommodates observations 1
and 2 and allows the prediction of liquid-vapor phase equilibrium was initially
proposed by van der Waals in 1873 (Rowlinson, 1988) and is expressed by
equation 5.

RT a
p= − 2 5
V −b V

Equation 5 is the essence of gas processing modeling thermodynamics and it


captures all the essential phenomena that are important for process modeling.
When combined with equation 2 it can be used for the computation of phase
equilibrium, it provides values for phase volumes and it also allows the prediction
of critical points, cricondentherms and cricondenbars thus providing engineers
with an useful map of the thermodynamic space for the design of gas processing
facilities. Naturally, the structure provided by equation 5 captures the basic

http://www.bepress.com/cppm/vol3/iss1/24 8
Satyro: Thermodynamics and the Simulation Engineer

physics of simple fluids, but it is highly simplified, and as written it is not


accurate enough for technical work. By judicious use of empiricism, industrial
strength equations of state can be constructed by noting that three boundary
conditions should be met:

1 Critical temperatures and pressures for individual components must be


modeled accurately
2 Individual component vapor pressures must be modeled accurately
3 Individual component liquid densities must be modeled accurately

An example of simple equation of state with these characteristics is the


Advanced Peng-Robinson model (Virtual Materials Group, 2004) where equation
5 is modified to equation 6.

RT a cα (T )
p= − 6
V − b V (V + b) + b(V − b)

When the equation of state is cast as described by equation 6 the term ac is


determined using the individual component critical pressure and temperature thus
ensuring exact representation of the critical temperature and pressure (the critical
volume is estimated and usually not very accurate and therefore estimated by the
model). The function α is an empirical function of temperature and its parameters
are determined to provide optimal representation of individual component vapor
pressures. In some instances these parameters can be generalized as a function of
molecular parameters (most commonly the acentric factor for hydrocarbons and
non-polar substances) and a simple prescription for handling undefined
components such as oil fractions can be quickly constructed.
Usually simple cubic equations of state can not represent liquid volumes
accurately because they overestimate the critical compressibility of fluids and
consequently liquid densities are underestimated. There’s a voluminous amount of
literature about this (Patel and Teja, 1982; Trebble and Bishnoi, 1986) but a
simple way to skirt the issue is to define an empirical parameter called the volume
translation (Peneloux et. al, 1982) where the volume calculated by equation 6 is
shifted by a constant computed to match the liquid density exactly at some
convenient condition (in gas processing and refining this is usually at 1 atm and
60 F) and the prescription for the equation of state is completed by equation 7.

Vt = V + δ 7

Published by The Berkeley Electronic Press, 2008 9


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

The next step is to extend the model from pure components to mixtures.
This is another subject that is extensively covered in the technical literature and
many good ideas as well as not so good ones have been published. For our
purposes it suffices to say that many hydrocarbon systems can be effectively
modeled for engineering design using remarkably simple mixing rules. Mixing
rules for cubic equations of state is another subject that has an immense volume of
literature attached to it. Here we simply state that effective modeling for
hydrocarbon systems can be achieved using the mixing rules derived by van der
Waals using simple physical arguments and summarized in modern form by
equations 8 and 9.

a = ∑∑ ai a j xi x j (1 − kij )
nc nc
8
i =1 j =1
nc
b = ∑ xi bi 9
i =1

The interaction parameter term kij in equation 8 accounts for the imperfect
adherence to the energy field of a van der Waals fluid when compared to the
energy field of a real fluid and they can be generalized for hydrocarbons and also
hydrocarbons and important inorganic gases such as carbon dioxide nitrogen and
hydrogen. Judicious usage of interaction parameters also allows the modeling of
water and hydrocarbon mixtures to a reasonable degree of accuracy4. Figure 4
illustrates how well these simple models can represent complex, polar molecules
(Virtual Materials Group, 2004). Figure 5 illustrates the effectiveness of this
simple, semi-empirical solution to the natural gas modeling problem (Cota et al,
2007).
Before we proceed, it is instructive to examine equation 6 in conjunction
to equation 2. By applying Legendre transformations on equation 2 and assuming
that no chemical reactions happen in the system we can express equation 2 in a
form where temperature and volume are the independent variables. When we do
that we have a different5 expression for the system energy using volume and

4
It is important to qualify this statement. Equation 8 implies a single interaction parameter per
binary pair, and this means that only one side of the mutual solubility curve can be modeled using
this type of simplistic model. In practice this corresponds to us having to choose on modeling
accurate solubilities of water in hydrocarbons or vice versa. The common practice is to model the
solubility of water in hydrocarbon and predict the solubility of hydrocarbon in water. This usually
means that the solubility of hydrocarbon in the water phase is grossly underestimated and has
significant impact on the simulation of water purification and disposal systems.
5
We stress here. Different but completely equivalent to internal energy or Gibbs free energy. We
just have a different set of independent variables.

http://www.bepress.com/cppm/vol3/iss1/24 10
Satyro: Thermodynamics and the Simulation Engineer

temperature as independent variables. This form is called the Helmholtz energy


and it is shown in equation 10 (Shaw and Satyro, 2006).

⎛ ∂U ⎞
dA = − SdT − pdV + ∑ ⎜⎜ ⎟⎟ dni 10
i ⎝ ∂ni ⎠T ,V ,n
j

Assuming a system of constant composition and taking the volume partial


derivative with respect to the volume at constant temperature we have equation
11.

⎛ ∂A ⎞
p = −⎜ ⎟ 11
⎝ ∂V ⎠T

Thus we can plug into equation 10 the equation of state and we can calculate
differences between Helmholtz energies at different volumes via integration of
equation 10 as suggested by equation 12.

V
ΔA = − ∫ pdV 12
V0

Using standard thermodynamic relationships you can show that equation 13


holds.

⎛ ∂ΔS ⎞
ΔS = −⎜ ⎟ 13
⎝ ∂T ⎠V

Therefore, as long as we can establish a reference point for the integration of


equations 12 and 13 and our model for p is reliable, we can calculate
thermodynamic properties such as enthalpies and free energies and these in turn
are the basis for energy balances and equilibrium reactor calculations performed
by simulators. For example, we can write the enthalpy as equation 14.

H = H 0 + ΔA + TΔS + RT (Z − 1) 14

And the calculation is complete when H0 is defined. It is common to use


the ideal gas state as the reference state for fluid phase calculations. At this
hypothetical state molecules behave like non-interacting hard points. They have
no volume, they generate no potential fields and their collisions are perfectly

Published by The Berkeley Electronic Press, 2008 11


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

elastic. Thus the energy content depends only on the average kinetic, rotational
and vibrational energies of individual molecules and it is a function of
temperature only. This is conveniently expressed using equation 15.

T
H = ∫ C p0 dT
0
15
T0

In equation 15 we have a reference temperature for integration and this is


an arbitrary choice, although careful choice slightly simplifies chemical reactor
modeling6. In a nutshell, by combining the fundamental equation 2, a model for
the Helmholtz energy, equation 11 (or if you feel more comfortable a model for
fluid pressure, equations such as 6) and the ability do calculate a reference state
for the integration of the energy equation 12 we are in business and a self-
contained description of reality is available for efficient simulations of fluid
systems. This is the essence of a property package and now it should be clear that
they come from a well defined procedure bounded by thermodynamic principles.
With it we can compute the number of phases at equilibrium, their compositions,
densities, enthalpies and so on7.
The salient feature with the approach outlined here is really equation 11. It
tells us that we can create Helmholtz energy models that are suited to a specific
problem and then we can use the same framework to compute physical properties
perform phase equilibrium calculations and compute material and energy balances
of interest. For example, this infrastructure supports complex equations of state
developed for accurate representation of important substances like water and
industrial gases (Lemmon et. al, 2002). It also supports equations of state
designed to better capture the molecular behavior of fluids such as SAFT
(Chapman, 1990).
More recently, equations of state more complex than cubics, but simple
enough for automated parameter determination have been proposed (Span and
Wagner 2003 a,b,c). These equations provide an accurate representation of the
Helmholtz energy surface and better qualitative description of volumetric

6
It is common to use the ideal gas enthalpy of formation at 25 C and 1 atm as the reference point
for the evaluation of ideal has enthalpies. This simplifies the equations for modeling reactors but
adds the complication of having to have (or have estimation methods) for ideal gas enthalpies of
formation of all components in a simulation. This can be particularly troublesome when dealing
with oil pseudo-components.
7
This is a convenient point to remind the reader that solving in principle and solving efficiently
and reliably are two different things. The efficient computation of phase equilibrium is usually
done using flash algorithms and this is a skillful combination of thermodynamics, mathematics,
numerical methods and individual talent (Michelsen, 1982ab; Nghiem, Li and Heidemann, 1985).

http://www.bepress.com/cppm/vol3/iss1/24 12
Satyro: Thermodynamics and the Simulation Engineer

properties than achievable with cubics and it is expected that they will play a
significant role in the natural gas processing industry in the next decade.
It is evident that the construction of complex models will always partly
rely on quality physical property data. The last five years has seen a significant
development in this area, and notable advances are the SOURCE database
(Frenkel et. al, 2001) maintained by the Thermodynamics Research Center at the
National Institute of Standards and Technology and the Thermo Data Engine
(Frenkel et. al, 2005), also maintained by TRC’s group. TDE allows on evaluation
of physical property on demand, creation of complex equations of state and last
but not least immediate validation of the results as shown in figure 6.
The quality of the predictions from simple equations of state for natural
gas processing is so good, and the model reliability is so high that this class of
industry can be modeled accurately from the well bore down to the tail gas
cleaning stage, and derivative works such as on-line monitoring are now common
place (Sitter et al, 2006).

Thermodynamics and Statistical Mechanics


The previous section briefly examined the structure of property packages from a
thermodynamic perspective and how a simple model can be used for the modeling
of an entire industry. Thus the requirements for a good model with industrial
significance boil down to two observations:

1 The model needs to provide a good description of the Helmholtz


energy of a fluid or fluid mixture
2 The model must include a way of calculating a reference state

This is usually how we leave the problem at undergraduate level and this is
also usually the level users interact with process simulation tools. There’s nothing
wrong with this and lots of solid and creative engineering work is done using
models like cubics. That said, I always have an intellectually unsatisfying
experience with reference states and the empirical (albeit brilliant) shape of van
der Waals type equations of state.
The remedy for the lack of satisfaction8 comes from statistical mechanics. In
essence, statistical mechanics concern itself with the interactions of a very large
number of particles and the properties of the system emerge as average properties
of these very many interacting particles. Statistical mechanics started with
Maxwell and Boltzmann and it is now a well developed tool used in classical and
quantum physics. The neat thing about statistical mechanics is that it provides a

8
As Mick Jagger said “you can get no satisfaction”. But statistical mechanics provides a pleasant
stop gap.

Published by The Berkeley Electronic Press, 2008 13


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

natural connection between microscopic and macroscopic properties, and


equations of state can be created in a methodical way. This ability of creating
models taking into account molecular interactions is important and allows
systematic modeling of systems currently hard or impossible to model using
simpler tools like surface phenomena and folding phenomena in complex
molecules. This has obvious implications in protein molecule modeling and no so
obvious in the processing of bitumen as we will see shortly. Statistical mechanics
pinnacle is equation 16, and as Richard Feynman noted many years ago
(Feynman, 1972) it is all “downhill from here” by creating applications or uphill
to derive the fundamental expression.

A = − RT ln Q 16

Q is the partition function of the system and it encodes all the different ways a
system can dispose of its energy, and if you combine equation 16 with equation
11 you can get an immediate connection between microscopic phenomena as
described by the partition function and classical thermodynamics. It is common to
factor the partition function into one part that deals with energy associated with
movement and another part that deals with internal molecular interactions that
include vibration, rotation and electronic transitions9. When you factor these
contributions to the energy of the system the partition function is expressed as
equation 17.

ZQ int
Q = 3N 17
Λ N!

The Qint term describes the internal movements inside a molecule with Z is
usually called the configurational partition function and describes the translational
movement of molecules. In general the movement is a function of the position
since potential energies are in action and Z is expressed, in general, by equation
18.

9
This is a useful, but not quite correct assumption. Imagine long, polyatomic molecules colliding.
It is clear that collisions will induce internal movements in the molecules after a collision and the
factoring of translation and intermolecular phenomena is not entirely rigorous. That said, this is
usually a good starting point. A remarkable thing with statistical mechanics is that by using very
simple models we can get reasonable answers. Sometime ago someone told me that statistical
mechanics is the science of getting reasonable answers from unreasonable assumptions. Be that is
as it may, it is neat to see how much insight on complex problems we can get by imagining
interacting balls, springs and potential energies.

http://www.bepress.com/cppm/vol3/iss1/24 14
Satyro: Thermodynamics and the Simulation Engineer

U
1 −
3N ∫ ∫
Z= ... e kT dq 3 N 18
N!h

What is important from a modeling perspective is that by separating the


partition function into one section that depends on internal motion and another
part that depends in external interactions we can devote attention in describing the
structure of a molecule separately from the task of modeling the interactions
between molecules. For example, if we concentrate on systems where the
potential energy in the fluid is zero we can model ideal gas properties of complex
molecules and accurately determine important properties such as ideal has heat
capacities that in turn can be used for the modeling of real systems by providing
the reference state for the computation of enthalpies as we saw before. On the
other hand, we can dedicate our attention to the modeling of potential energies
between molecules and derive new equations of state. It is significant that
statistical mechanics can be used to create equations of state for any aggregation
state of matter such as gases, liquids, solids, and complex interactions between
not only molecules but also the substract in the background. This has important
repercussions on the modeling of surface phenomena and semiconductor
materials properties.
For example, by careful choice of intermolecular potential and internal
movement we can reconstruct the van der Waals equation. This technique is also
useful for the construction of new equations of state or activity coefficient models
as shown elegantly by Sandler and co-workers (Sandler et al, 1986).
A last thought related to this bridge between macroscopic and microscopic
phenomena. The partition function formulation is completely general and can be
extended to quantum phenomena. Therefore, a consistent formal structure
connecting the dimensions usually explored by chemical engineers and
dimensions explored by physicists exists. As we move towards the processing of
more exotic materials or processing at more extreme conditions of temperature we
will see a significant interplay between physics and engineering. Albeit at its
infancy, ultra cold chemistry related to Bose-Einstein condensates is being
developed. At these extreme conditions, atoms can be manipulated, moved and
separated using light – optical tweezers (Boyer et. al, 2006). It is quite possible
that this will find its way into mainstream life through quantum computers. If this
is to come into being engineers will have to be involved for the creation of large
scale processing facilities and readily available predictions for the physical
properties of such materials will be necessary.
This type of molecular manipulation also opens some interesting
opportunities for the development of new catalysts, or catalyst based processes.
Recently Horn and coworkers (Horn et. at, 2006) show that molecules that are
illuminated by laser pulses with duration considerably shorter than their rotational

Published by The Berkeley Electronic Press, 2008 15


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

period undergo alignment immediately after the pulse quits. This in turn can be
used for the creation of aligned systems of molecules in gas phase, and even the
remote possibility of aligning gas molecules and collimate them into catalyst sites
presents incredible economic potential related to increased yields for kinetically
controlled processes.

Process Economics
Thermodynamics play the key role in the economic evaluation of processes that
involve usage of energy. Simply put, thermodynamics establish the base line for
reversible processes. Since we know that no real process is reversible, the study of
reversible processes tells us the very best that can be obtained and therefore
establishes an absolute and impartial basis for the creation, evaluation and
ranking of processes. It is true that much can be done in optimizing kinetic
parameters in processes and that many processes operate optimally from an
economical perspective when driven into conditions away from what would be
optimal from thermodynamic equilibrium due to kinetic constraints10. That said, if
we did not have an absolute scale to measure the efficiency of processes we
would be at the mercy of belief and not science. Thermodynamics is the provider
of this absolute scale and the guardian against ill founded belief. In a nutshell,
thermodynamics forces us to be honest. Without its guiding principles applicable
to any natural process we would be forced to deal with a virtually endless need for
experimental evidence when studying new processes and proposing new
solutions.
It is amusing to see how this fundamental characteristic of
thermodynamics has made its way into mainstream thought and is used in
economics. For example, important issues nowadays are related to creating
processes that are optimal from an economic point of view taking into account
ecological effects. For example, Barbiroli (Barbiroli, 2006) wrote an interesting
article on eco-efficiency where a key point is the definition of a “state of bliss”
which is the “highest ideal condition in any choice”. The only way of objective
studying these conditions is through the usage of limits imposed by
thermodynamics.
Talking about environmentally important issues, thermodynamics plays a
key role in the most important current issue related to the environment – carbon
dioxide. From an orthodox point of view thermodynamics provides assistance in
the optimization of combustion processes and in turn in the optimization of the

10
The production of ammonia and synthetic fuels are interesting case studies related to this.
Perhaps more interesting is to study the influence of these processes in the economic history of the
world. It is sobering for technically minded people to see the type of immense impact chemical
technologies have on the world. Two highly recommended reads are Tooze’s (Tooze, 2006) and
Levi’s (Levi and Debenedeti, 2006) books.

http://www.bepress.com/cppm/vol3/iss1/24 16
Satyro: Thermodynamics and the Simulation Engineer

useable energy per pound of fuel consumed. Thermodynamics also provides the
basic framework for the analysis of non-conventional processes such as fuel cells
and hydrogen based solutions. Steeneveldt and coworkers (Steeneveld et. al,
2006) provide an interesting overview on processes for the capture and storage of
carbon dioxide which are convenient summarized in figure 7.
It is interesting to note that one of the most important processes for carbon
dioxide removal is alkanolamine based processes. These processes are interesting
for several reasons. From a thermodynamic point of view they involve a series of
modeling techniques. It starts with the usage of equation 2 as the unifying
principle, which is then fleshed out by experimentally determined equilibrium
constants for the ionic acid-base reactions in aqueous phase. This is
complemented by rigorous physics used to describe the limiting activity
coefficients for ions in the aqueous phase and topped off by statistical mechanical
based equations for activity coefficients, necessary to bring the model predictions
in line with experimental measurements within error ranges acceptable for process
design.
The interesting feature of this type of process is that it is not
thermodynamically controlled, but rather, the kinetics of absorption of carbon
dioxide and hydrogen sulfide are of fundamental importance for the designer, and
these features are not described by the thermodynamic model. Therefore we
supplement the model with the necessary kinetic information related to the carbon
dioxide and hydrogen sulfide and in turn this information makes its way into the
equations that describe the material and energy balances of distillation towers in
the form of efficiencies11. A sample illustration in how different the carbon
dioxide and hydrogen sulfide efficiencies are in a typical amine sweetening unit is
shown in figure 8. Alternatively, models using first principle mass transfer models
are available (Krishnamurthy and Taylor, 1985; Taylor and Krishna, 1993) where
the material and energy balances are solved simultaneously with mass transfer
rate expressions. Although this class of model does not require efficiencies or
equivalent HETP values, it does require mass and heat transfer coefficients for the
particular type of tray or packing being used. Its main advantages over simpler
efficiency based models lie of its ability of predicting more accurate temperature
profiles where vapor and liquid phases are not at equilibrium and rigorous
integration of the material and energy balance for the simulation of packed
towers. Rate based models have evolved considerably over the years and a state of

11
The kinetics of carbon dioxide absorption is very different from the kinetics of hydrogen sulfide
absorption. Carbon dioxide reacts much more slowly than hydrogen sulfide with alkanolamines
and therefore the amount of carbon dioxide absorbed in the amine solution per unit of time is
much less than what could be absorbed from a purely thermodynamic point of view. The
efficiency of carbon dioxide absorption is low and this can be used for the creation of all types of
processes that remove hydrogen sulfide preferentially from a gas stream.

Published by The Berkeley Electronic Press, 2008 17


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

the art implementation is provided by Taylor and co-workers (Taylor et al, 1994)
and is also available in several commercial products such as Aspen’s
RATEFRAC (Aspen Technology, 1994) and PROTREAT (Sulfur Experts, 2008).
A series of fascinating problems related to the modeling and optimization
of this class of processes is related to thermodynamic irreversibilities. The basic
idea of this type of process is to absorb the acid gases using a solution of
alkanolamine in an absorber at high pressure and then drive the acid-base
equilibrium back by heating up the solution rich in acid gases at low pressures.
This works in general quite well, but there are other chemicals in the gas other
than carbon dioxide and hydrogen sulfide such as nitrogen oxides, sulfur oxides
and hydrogen chloride. These chemicals react with the alkanolamines forming
stable salts that can not be decomposed via heating in the amine regeneration
tower and the gas absorption capability of the amine solution decreases with time.
Another interesting issue to be explored has its roots on purely
thermodynamic grounds. Since it takes lots of energy to bring aqueous solutions
to boil in the amine regeneration step, why not replace the solvent by a solvent
that can be more easily boiled off such as methanol? This is an idea being
explored by Park and co-workers (Park et. al, 2006), and it is amenable to the
same principles used for the thermodynamic modeling of aqueous solutions of
alkanolamines and consequent design, optimization and comparison of alternative
processes and finally informed decision related to the economic deployment of
such alternative carbon dioxide and gas sweetening processed.

Process Optimization and Process Integration


Process optimization and process integration methodologies go hand in hand with
the thermodynamic model of the chemicals of interest as well as empirical
knowledge from the engineer performing the optimization studies. Failure to
accept the roles of empiricism and fundamental thermodynamic relationships is
doomed to produce large case studies with little substance combined with
significant doses of embarrassment. Over the years I have had the chance to
collect my fair share of horror stories and I will not bother you with an
encyclopedic listing. But I will tell you one of my favorites – distillation design
by 10. This is a true story – and I saw it repeated many times with slight
variations. Someone was designing a distillation tower to separate a certain
mixture. The distillation algorithm would not converge and I was called to help
diagnose what was going on. The mixture being separated had an azeotrope – it
was not an exotic mixture or anything and the data was readily available in a
standard handbook – and the tower layout looked strange to me so I asked the
designer how he was going about designing the tower. To my absolute dismay the
tower was being designed like this:

http://www.bepress.com/cppm/vol3/iss1/24 18
Satyro: Thermodynamics and the Simulation Engineer

1 Set the number of trays to 10


2 Put the feed in the middle
3 Set the reflux to 3 and the distillate flow corresponding to the desired
recovery
4 Run the simulation
5 If the simulation converged great
6 If tower did not converge then increase the number of trays by 10,
place the feed in the middle and run the simulation again until it
converged.

I know it sounds like I am making this up but it is true – I saw it12. The
designer had several papers published and is an expert in the process. The
problem is, there was no way that the approach would work because there was an
azeotrope in the tower and the desired specification was impossible. No matter
how many trays were added to the tower, the thermodynamic model built into the
simulator would steadfastly refuse to allow the second law of thermodynamics to
be broken. I solved the problem using pencil and paper in an hour or so and
graphical techniques chemical engineers learn in undergraduate distillation
courses and then used the simulator to polish the results. Similar situations
happened several times, including one where a scientist kept asking me if the
azeotrope problem would not go away if he kept increasing the size of the
equipment he intended to use to separate the mixture in question.
It sounds amusing, but it is in reality quite tragic. In the heat of the battle folks
rely on computerized tools as if they were some kind of substitute for thinking.
Simulators are certainly great thinking aides, but they are not able to think for
anyone, and this is perhaps the most important point I would like to get across in
the paper. Deep understanding of the fundamentals of chemical engineering,
chemistry, physics and physical-chemistry are your best guarantee for a
successful career in chemical engineering. As tools grow in complexity,
understanding of the basics grows in importance because you must be able to spot
non-sense, either coming from the simulator or as a consequence of poor choices
or poor thermodynamic modeling. I am a true believer that simulation engineers
should also be conceptual engineers where they express their talent by exercising
creative thinking buttressed by knowledge of basic science.
In my experience, the most fruitful advice for process optimizers and
integrators is that before they embark on significant work involving computers
that they spend the time understanding the basic chemistry of the process and ask

12
Mike Malone from the University of Massachusetts put it best. When having lunch a while ago
we were talking about conceptual design and some of the new techniques he and Mike Doherty
were developing. I will never forget Mike looking at me and saying something like “You know
Marco, if the only tool you have is a hammer soon enough all problems look like nails”.

Published by The Berkeley Electronic Press, 2008 19


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

themselves how good the thermodynamic model is and have a clear understanding
of the limitations of such model. You see, this is the ideal moment for seeding the
field with useful documentation about the model, clearly state assumptions, list
experimental data used to define the model and report areas where the data is
faulty, incomplete or of questionable quality. Moving forward, this opportunity
offered by the time used to create the thermodynamic model allows us to ask good
questions about the process such as what could be kinetically limited constraints,
what are potential side products, what are potential regulatory issues we may
have, what are potential alternative solvents, etc. It is impossible to place a truly
comprehensive list of questions since problems will always be different and you,
the designer, will always know better. What is important is to build this
questioning behavior in your thinking process and make the most of the time you
have to think about the fundamentals of your process. After you are satisfied with
your basic understanding of the physics and chemistry of the process then
simulators are wonderful tools that allow you to piece together your ideas and get
your design done reliably, effectively and conscientiously.

Current and Future Research


Important problems involving thermodynamics abound and opportunities for
fruitful research appear virtually everywhere. I will not even try to provide an
encyclopedic13 listing of opportunities since you can get these by simply cracking
up any chemical engineering, chemistry, and physics or economics magazine.
Instead I would like to mention two opportunities separated by a vast chasm of
application. They illustrate extreme aspects of the art but intersect at a
fundamental thermodynamic modeling interface.

Volumetric behavior of bitumen and solvent mixtures


We can think of bitumen as a widely known thing since references to it are
present in the Bible and by rights we should know quite a bit about it. This is not
quite the case. Bitumen is a complex class of hydrocarbon mixtures that involve
light and heavy organic molecules. The most important operation related to
bitumen processing is how to get it out of the rock and get the hydrocarbon
materials to processing plants. A couple of basic questions we need to be
answered when designing facilities, are what is the density of the fluids being
processed and when I mix fluids does the temperature go up or down? When we
look at facilities that are processing thousands of tons of material per hour these
questions translate into large capital investment and energy costs.
One would suppose that we can answer these questions quickly and
reliably, especially if we could have a few more points on the distillation curve

13
It may sound as if I do not like encyclopedias. Nothing could be farther from the truth.

http://www.bepress.com/cppm/vol3/iss1/24 20
Satyro: Thermodynamics and the Simulation Engineer

and a better characterization of the heavy end. Mahan and co-workers (2006)
measured the excess volumes of mixtures of heavy residues and paraffins used to
simulate industrial plants using naphtha as a solvent. Since bitumen is processed
in different ways depending on how it is extracted, but blended when fed to the
upgrading facility, measurements were conducted in different ways. Some
mixtures were produced at a certain temperature and then heated to a desired
temperature. Some mixtures were first heated and then mixed and then moved to
the desired temperature.
From a conventional thermodynamic perspective, the measured excess
volumes should be identical within some measurement error. After all we all
know that if we are careful and patient enough we should get pretty close to
thermodynamic equilibrium and therefore the measurement of physical properties
should be independent of the path of measurement. Not only the results are
different depending on how the mixtures are prepared, they are qualitative
opposite to one another as shown in figure 9. A definitive interpretation for these
results is not yet available, but a couple of ideas come to mind. The first is that
some kind of polymorphism may happen depending on the thermal history of the
material, akin to the protein folding of complex molecules that result in a
molecular shape change and therefore the volumetric behavior is different. A
second thought is that the approach to equilibrium is so slow that even after
several days no significant changes can be detected in the lab and much longer
equilibration times are necessary.
If this is the case the residence times of these materials in the plant would
suggest that we are not free to choose the way we perform experiments, but
rather, we have to tailor experiments to mimic the way the plant is run because the
mixing process is kinetically controlled. This fact is important for simulations,
i.e., the way the process is conducted. This is not an alien concept when we are
dealing with slow chemical reactions, but in this case we are perhaps considering
an entry in a physical property handbook where, in order to make the measured
properties relevant we have to specify that this is the excess volume for this
bitumen and this solvent mimicking the thermal history of a certain processing
technology.
This may be a necessary piece of information for the design of processing
plants, but it does underline some significant lack of basic knowledge about what
we are processing. The definitive answer to this conundrum is connected to
finding ways of gathering a better understanding of the materials being processed,
better understanding of the time constants involved in the process when compared
to thermodynamically meaningful equilibration times and on top of this, how to
perform these basic measurements in an efficient and cost effective way. Given
the importance of energy producing technologies and its vast impact on the
environment techniques for efficient hydrocarbon characterization and modeling
for process design and optimization will only grow in importance.
Although not connected to bitumen processing, it is interesting to note that
recycling provides a significant amount of hydrocarbon materials available for
processing, and that there are significant opportunities related to process design
and product design related to recycling. In order to explore these opportunities,
thermodynamic modeling of polymeric materials after they are produced and
subject to normal usage and finally discared will be important (Bai et. al, 2007).

Published by The Berkeley Electronic Press, 2008 21


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

Statistical Mechanics and Process Engineering


An important area of process engineering deals with interactions between
molecules and surfaces such as adsorption and catalysis. For example, zeolite
based surfaces can be used for separation of alkane isomers and recovery of
valuable high octane chemical species, separation of aromatic isomers like
xylenes, separation of sulfur compounds such as mercaptans and gas dehydration.
In a related field, membranes can be used for separation of carbon dioxide from
natural gas mixtures and creation of efficient separation processes for ethanol and
water separations.
Contrary to conventional separation methods such as distillation and
absorption where reliable thermodynamic models such as equations of state or
activity coefficient based methods exist, surface based processes depend on
interactions between substract and molecules. This is an area where classical
thermodynamics has little to offer and statistical mechanical methods shine. If you
recall equation 17, it encodes all the necessary structure to model the movement of
molecules as well as interactions between molecules and any other source of
potential energy such as the adsorbing substract. Since classical models are of
little use, good part of the art related to surface chemistry deals with experiments
trying to determine the molecule-substract interactions and summarize this
experimental information in models that can be used for process design. An
alternative way to go about solving this problem is based on the creation of
models that take into account molecular and substract conformations and
determination of molecule-substract interactions. This interesting approach is
discussed in detail by Fuchs and coworkers (Fuchs et. al, 2006) and illustrated in
figure 10.
The statistical mechanical tools can be used for virtually any molecular
process and as we move towards more exotic materials such as organic metals,
conductive liquids, ionic liquids, superconductors and conductive films the proper
usage of modeling tools tempered with good understanding of the fundamentals
of chemistry, physics and physical chemistry will be the differentiating factor
between good and bad designs. This is poignantly demonstrated by Saito and
Yoshida (Saito and Yoshida, 2007). The scientist who learns how to navigate the
phase envelope of superconductive materials will have a good opportunity of
designing the hot new commodity products of the next decade. Hand in hand with
this, the engineer who can design the process that will bring these products to the
mass market will change the world.

Integrating Thermodynamics into Processes, Products and


Appliances
A long time ago Arthur Clarke wrote that if technology is sufficiently advanced it
is indistinguishable from magic. I think these days if technology is sufficiently

http://www.bepress.com/cppm/vol3/iss1/24 22
Satyro: Thermodynamics and the Simulation Engineer

advanced it is indistinguishable from an appliance. You see, my first computer


had a 10 MB hard drive and I had lots of respect, care and tenderness towards it.
If you are old enough you will recall a program called park that you had to run
before moving your desktop. Today I have a 160 GB iPod that I simply toss
around. Electronic storage technology is sufficiently advanced and therefore we
treat it as just another appliance – we simply don’t care about it. We use it, very
much like we use a pencil.
I do not think we have reached the point where fluid phase
thermodynamics and flash calculation technology is advanced enough that we can
consider it an appliance. But there are a few specific aspects of it where we are
very close, and we at Virtual Materials Group are working on making the
embedding of this technology in other applications a reality. For example, we now
routinely use flash calculations and phase diagram tracing on-line to warn plant
operators of potentially dangerous operating conditions when compressing sour
gases or avoiding liquid dropping in retrograde conditions or creating ersatz
sensors to provide operators with physical properties that can not be measured or
telling refinery operators what’s the best way to operate units for changing feed
stocks. This allow us to continuously learn new interesting twists and turns related
to how thermodynamic models are used and how they must behave in real time
applications (Sitter et. al, 2006). A sample of this type of thermodynamically
based appliance is shown in figure 11.

Thermodynamics and Education


By now it should be self-evident that thermodynamics is a rather pervasive
discipline and it underlines the calculations performed by simulation programs
and acts as the unifying force bring order to a significant amount of data collected
in chemical and petroleum engineering. It should also be clear that mathematical
prowess is not enough for efficient use of thermodynamics. Mathematical skill
has to be tempered with empirical knowledge, and it is the combination of theory
and data that makes thermodynamics a formidable tool.
The interplay between theory and empirical observation defines, in my
opinion, how thermodynamics should be taught from a process engineering
perspective. Unfortunately, in my experience the amount of theory and
empiricism currently taught are not enough to allow recent graduates to work
efficiently using modern process simulation tools. When I took chemical
engineering in the early 1980’s we learned quite a bit of chemistry and more
importantly physical-chemistry. I do not recall very complex labs, but we did
perform experiments and somehow we did learn some rudiments about the
behavior of fluids. My limited experience lecturing shows that students nowadays
are just as bright and hardworking as I recall from my student years, but many
students lack empirical knowledge. Perhaps this happens because chemical
engineers are not taught as much physical-chemistry as previous generations were
taught. A simple example should suffice. Recently in one graduate level
examination I asked for students to comment on the accuracy of a simple activity

Published by The Berkeley Electronic Press, 2008 23


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

coefficient model when applied to ethanol and water mixtures. The model
incorrectly predicted that ethanol and water would phase split at 25 C and 1
atmosphere pressure. My students could perform the necessary mathematics to
find out that the model would predict phase splitting and they had no problems
computing the phase fractions and phase compositions.
But they did have problems commenting on the accuracy of the model. In
this particular occasion I even had a student mentioning that he did not have time
to go to the library to look for liquid-liquid equilibrium data for ethanol and water
solutions. This lack of (basic) empirical knowledge seems to be making an
insidious penetration in the educational process. Since in thermodynamics we still
need a degree of empirical knowledge for effective usage of tools and models, it
seems to me that students would benefit quite a bit by having a more extensive
training in physical-chemistry.
Naturally, we do not want to create a situation where we are only reliant
on empirical information. On the contrary, models are getting better all the time
thanks for advancements in theory and in the available experimental databases
available for determining model parameters. That said, it is always useful to have
a certain amount of empirical information in your mind to help you judge the
quality of computer models before embarking on a large simulation project.
Sometimes this will avoid embarrassment. Sometimes it will motivate you to look
for more data, more applicable to the conditions of the problem at hand or perhaps
discard the thermodynamic model you originally intended to use.
I think that we should train the best conceptual engineers we can, and part
of the art related to this task is learning when to be critical of results coming from
computerized tools, how to refine models and how to do this in a time and
resource sensible manner.

In conclusion
Thermodynamics provides the underlining logical principle that allow us to
collect, interpret and organize experimental data. This same logical principle
allows us to draw exact relationships between quantities without the need for a
molecular model. These relationships are derived from the first and second laws
of thermodynamics and although axiomatic, these two laws seem to apply to
every natural process at any scale. Combined with this logical structure, educated
use of empiricism allows us to create useful models that can be used to calculate
material and energy balances and design equipment, processes and new products.
Some of the most successful models currently used owe their origin to theories
created in the 19th century and are a testament to how far a simple theory can be
taken as long as its development is done sensibly using basic engineering
principles. A formal structure in the form of statistical mechanics can be used for
the creation of new equations of state and models for the interactions of molecules

http://www.bepress.com/cppm/vol3/iss1/24 24
Satyro: Thermodynamics and the Simulation Engineer

and substract and combined with the logical structure provided by classical
thermodynamics and software it will play a key role in the development of new
processes and products of relevance for the 21st century.

Nomenclature
a attractive coefficient for cubic equations of state
A Helmholtz energy
b covolume for cubic equations of state
Cp isobaric heat capacity
f some function
g some function
G Gibbs free energy
h Planck’s constant
H enthalpy
k interaction parameter, Boltzman constant
N number of particles
n number of moles
P Pressure
q generalized position
Q Partition function
R gas constant
S Entropy
T Absolute temperature
U Internal energy, potential energy
x mole fraction vector
V Volume
Z compressibility factor, configurational partition function

Subscripts
i component index
j reaction index or component index
t volume translated variable
0 reference state

Greek Letters
Δ difference
δ volume translation factor
ε reaction extent
Λ de Broglie wave number

Published by The Berkeley Electronic Press, 2008 25


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

Figure 1 Petlyuk towers provide a highly integrated environment for separation


using distillation with a single reboiler and condenser with savings between 20
and 50% when compared to a standard arrangement using multiple towers. Rong
and Turunen (Rong and Turunen, 2006) show that this type of tower provides a
unique thermodynamic equivalent side column structure and provide a rational
procedure for tackling a combinatorial problem.

http://www.bepress.com/cppm/vol3/iss1/24 26
Satyro: Thermodynamics and the Simulation Engineer

Figure 2 Residue curves and distillation boundaries are thermodynamic


properties defined by vapor-liquid equilibrium, simple models for the liquid phase
and differential equations describing simple batch distillation. Note that the
distillation boundary separates two different regimes for the distillation depending
on the overall feed. One where the bottoms is rich in benzene and another where
the bottoms is rich in n-propanol. In both cases the distillate is rich in i-propanol
and benzene azeotrope.

Published by The Berkeley Electronic Press, 2008 27


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

Figure 3 The structure of Nature is such that equations of state explicit in


temperature and volume are more useful than equations of state explicit on
pressure and temperature due to phase equilibrium phenomena. Note that the
saturation pressures for the vapor or liquid steam are the same at the same
temperature but the phase volumes are quite different. Curves calculated using
VMGSim (Virtual Materials Group, Inc., 2007)

http://www.bepress.com/cppm/vol3/iss1/24 28
Satyro: Thermodynamics and the Simulation Engineer

Glycols Vapor Pressure Prediction

1.E+04

1.E+03

1.E+02
Vapor Pressure

1.E+01

1.E+00

1.E-01

1.E-02

1.E-03
0 100 200 300 400
Temperature (°C)

EG DEG TEG

Figure 4 Note that polar substances can also be modeled and important
aspects of processing such as dehydration and hydrate depression can also be
simulated and optimized (Virtual Materials Group, Inc., 2004)

Published by The Berkeley Electronic Press, 2008 29


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

TEG Dehydration NLG Recovery


APR for Natural Gas APR for Natural Gas

Gas Gathering MDEA Plant Claus Plant


APR for Natural Gas Amine Package Claus Package

Figure 5 Simple but sensible flow sheet modeling in action. Properly tuned cubic equations of state together with
reaction models for sweetening and sulfur processing can be used to model entire processes.

http://www.bepress.com/cppm/vol3/iss1/24 30
Satyro: Thermodynamics and the Simulation Engineer

Figure 6 Typical results calculated using NIST’s Thermo Data Engine. Note the detailed quality information
provided by uncertainty values for each experimental data point. Experimental data is for water thermal conductivity as
a function of temperature and pressure (Frenkel et al, 2004)

Published by The Berkeley Electronic Press, 2008 31


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

Figure 7 Fuel / carbon dioxide sequestration (adapted from Steeneveld et. al, 2006)

http://www.bepress.com/cppm/vol3/iss1/24 32
Satyro: Thermodynamics and the Simulation Engineer

Figure 8 DEA sweetening plant absorption tower simulation showing the tray efficiencies in absorber for H2S and
CO2. Note that the efficiencies of carbon dioxide and hydrogen sulfide are very different from 1 and reflect kinetic
effects that can not be predicted by classical thermodynamics. Also observe the maximum in temperature caused by the
acid-base chemical reactions on the trays. Results from VMGSim (Virtual Materials Group, Inc., 2007)

Published by The Berkeley Electronic Press, 2008 33


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

Excess Volumes for solvent / Athabasca Bitumen


Vaccum Bottoms at 25 C

0.035

0.030

0.025
Excess Volume, cm3/g

0.020

0.015
prepared
0.010
annealed
0.005

0.000
0 0.2 0.4 0.6 0.8 1
-0.005

-0.010

-0.015
Mass Fraction ABVB

Figure 9 Different excess volumes are measured depending on how the heavy hydrocarbon and solvent sample was
prepared, based on data from Mahan et al, 2006.

http://www.bepress.com/cppm/vol3/iss1/24 34
Satyro: Thermodynamics and the Simulation Engineer

Water adsorption on Na52 fausjite

300
Number of water molecules adsorbed per

250

200
micro cm3

Experiment
150
Simulation

100

50

0
0 500 1000 1500 2000 2500 3000
Pressure, Pa

Figure 10 Adsorption of water in Na52Y fausjite at 300 K, based on data from Fuchs et. al, 2006

Published by The Berkeley Electronic Press, 2008 35


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

Figure 11 On-line monitoring of sour gas compressor trains. Note that for optimal operation the temperature after
the compressor intercooler has to be always above the gas dew point and therefore the intercooler temperature after
stage 4 is not feasible. Based on data from Sitter et al, 2006.

http://www.bepress.com/cppm/vol3/iss1/24 36
Satyro: Thermodynamics and the Simulation Engineer

Bibliography

ASPEN PLUS Reference Manual – Volume I, Aspen Technology, Cambridge,


MA (1994)

Barbiroli, G.; “Eco-efficiency or/and eco-effectiveness? Shifting to innovative


paradigms for resource productivity”; International Journal of Sustainable
Development & World Ecology; 15 (2006), 391-395

Boyer, V.; Godun, M.; Smirne, G.; Cassettari, D.; Chandrashekar, C.M.; Deb,
A.B.; Laczik, Z.J. and Foot, C.J.; “Dynamic manipulation of Bose-Einstein
condensate with spatial light modulator”; Physical Review A 73, 031402(R), 2006

Bai, X.; Isaac, D.H. and Smith, K.; “Reprocessing Acrylonitrile-Butadiene-


Styrene Plastics: Structure-Property Relationships”; Polymer Engineering and
Science; 2007

Chapman, W.G.; “Prediction of the thermodynamic properties of associating


Lennard-Jones fluids: Theory and simulation”, J. Phys. Chem., 93, 4299-4304
(1990)

Chien, H.H.Y. and Null, H.R.; “Generalized Multicomponent Equation for


Activity Coefficient Calculation”; AIChE J.; Vol. 18, No.6, November 1972

Cota, R.; Hay, G.; Jacobs, G.; Krenz, R.; van der Lee, J.; Li, Y.-K.; Svrcek, W.Y
and Satyro, M.A.; “Addressing Challenges in Accurate Simulation of Natural Gas
Production”; Hydrocarbon Engineering, February, 2007

Doherty, M.F. and Malone, M.F.; “Conceptual Design of Distillation Systems”;


McGraw-Hill, 2001

Ely, J.F. and Hanley, H.J.M.; “A Computer Program for the Prediction of
Viscosity and Thermal Conductivity in Hydrocarbon Mixtures”; Thermophysical
Properties Division, National Engineering Laboratory, National Bureau of
Standards, Boulder, Colorado, 1981

Feynman, R.P.; “Statistical Mechanics – A Set of Lectures”; Adison-Wesley


Publishing Company, 1972

Published by The Berkeley Electronic Press, 2008 37


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

Foucher, E.R.; Doherty, M.F. and Malone, M.F.; “Automatic Screening of


Entrainers for Homogeneous Azeotropic Distillation”; Ind. Eng. Chem. Research,
30, 760-772 (1991)

Frenkel, M.; Dong, Q.; Wilhoit, R. C.; Hall, K. R. TRC SOURCE Database: A
Unique Tool for Automatic Production of Data Compilations. Int. J. Thermophys.
2001, 22, 215-226.

Frenkel, M.; Chirico, R. D.; Diky, V.; Yan, X.; Dong, Q.; Muzny, C. ThermoData
Engine (TDE): Software Implementation of the Dynamic Data Evaluation
Concept. J. Chem. Inf. Model. 2005, 45, 816-838.

Fuchs, A.H.; Boutin, A.; Teuler, J.M.; Di Lella, A.; Wender, A.; Tavitian, B. and
Ungerer, P.; “Development and Application of Molecular Simulation Methods for
the Screening of Industrial Zeolite Adsorbents”; Oil and Gas Science and
Technology – Rev. IFP. Vol. 61 (2006), No. 4, pp 571-578

Horn, C.; Wollenhaupt, K.; Krug, M.; Baumert, T.; de Nalda, R. and Banares, L.;
“Adaptive control of molecular alignment”; Physical Review A 73, 031401(R)
(2006)

Krishnamurthy, R. and Taylor, R.; “A Nonequilibrium Stage Model of


Multicomponent Separation Processes Part I: Model Description and Method of
Solution”; AIChE J., 31, 449-456 (1985)

Lemmon, E.W.; McLinden, M.O. and Huber, M.L.; REFPROP 7.0; NIST
Standard Reference Database 23 Version 7; 2002

Levi, P and Debenedeti, L.; “Auschwitz Report”, Verso Books, 2006

Maham, Y.; Zhang, X.; Zabeti, P.; Goodkey, J.; Allain, M. and J. M. Shaw;
Presented at the 7th international conf. on petroleum phase behavior and fouling,
Asheville, USA, June 25-29, 2006.
Michelsen, M.L.; “The isothermal flash problem. Part I. Stability”; Fluid Phase
Equilibria, Volume 9, Issue 1, December 1982, Pages 1-19
Michelsen, M.L.; “The isothermal flash problem. Part II. Phase-split calculation”;
Fluid Phase Equilibria, Volume 9, Issue 1, December 1982, Pages 21-40

Modell, M. and Reid, R.C.; “Thermodynamics and Its Applications”; Prentice


Hall, 1974

http://www.bepress.com/cppm/vol3/iss1/24 38
Satyro: Thermodynamics and the Simulation Engineer

Nghiem, L.X.; Li, Y.-K. and Heidemann, R.A.; “Application of the tangent plane
criterion to saturation pressure and temperature computations”; Fluid Phase
Equilibria, Volume 21, Issues 1-2, 1985, Pages 39-60

Null, H.R.; “Phase Equilibrium in Process Design”; Robert E. Kriger Publishing


Company, Huntington, New York, 1980

Ostwald, W. Dampfdrucke Ternarer Gemische, Abahndlungen der Mathematisch-


Physischen der Konig Sachsischen. Ges. Wiss. 1900, 25, 413.

Park, S.W.; Choi, B.S. and Lee, J.W.; “Chemical absorption of carbon dioxide
with triethanolamine in non-aqueous solutions”; Korean J. Chem. Eng.; 23(1),
138-143 (2006)

Patel, N.C. and Teja, A.S.; “A new cubic equation of state for fluids and fluid
mixtures”; Chem. Eng. Sci., 37, 463-473 (1982)

Peneloux, A.; Rauzy, E. and Freze, R.; “A consistent correction for Redlich-
Kwong-Soave volumes”; Fluid Phase Eq. 8, 7-23 (1982)

Persegian, V.A.; “Van der Waals Forces – A Handbook for Biologists, Chemists,
Engineers and Physicists”; Cambridge University Press, 2006

Prigogine, I. and Defay, R.; “Chemical Thermodynamics”; Longman, Green and


Co Ltd, 1954

Rong, B.G. and Turunen, I.; “A New Method for Synthesis of


Thermodynamically Equivalent Structures for Petlyuk Arrangements”; Chemical
Engineering Research and Design, 84(A12): 1095-1116, 2006

Rowlinson, J.S.; “J.D. van der Waals: On the Continuity of the Gaseous and
Liquid States”; Elsevier Science Publishers, 1988

Saito, G. and Yoshida, Y.; “Development of Conductive Organic Molecular


Assemblies: Organic Metals, Superconductors and Exotic Functional Materials”;
Bull. Chem. Soc. Jpn. Vol. 80, No. 1, 1-137 (2007)

Sandler, S.I.; Kee, K.H. and Kim, H.; “The generalized van der Waals partition
function as a basis for equations of state, their mixing rules and activity
coefficient models”; ACS Symp. Ser.; 300, 180-200, 1986

Published by The Berkeley Electronic Press, 2008 39


Chemical Product and Process Modeling, Vol. 3 [2008], Iss. 1, Art. 24

Schreinemakers, F. A. H. Dampfdrucke Ternarer Gemische. Z. Phys. Chem. 1901,


36, 257, 413, 711.

Shaw, J.M. and Satyro, M.A.; Lecture notes for Advanced Chemical Engineering
Thermodynamics CHE 624 and ENCH 633, The University of Alberta and the
University of Calgary, 2006

Sitter, J.R.; Hay, G. and Neumeister, L.; Hydrocarbon Engineering, February


2006

Span, R.; Wagner, W. Equations of State for Technical Applications. I.


Simultaneously Optimized Functional Forms for Nonpolar and Polar Fluids. Int.
J. Thermophys. 2003, 24, 1-39. (a)

Span, R.; Wagner, W. Equations of State for Technical Applications. II. Results
for Nonpolar Fuids. Int. J. Thermophys. 2003, 24, 41-109. (b)

Span, R.; Wagner, W. Equations of State for Technical Applications. III. Results
for Polar Fuids. Int. J. Thermophys. 2003, 24, 111-161. (c)

Steeneveldt, R.; Berger, B. and Torf, T.A.; “CO2 Capture and Storage Closing the
Knowing- Doing Gap”; Chemical Engineering Research and Design, 84(A9),
739-763 (2006)

Sulfur Experts;
http://www.sulphurexperts.com/AmineExperts/SoftwareAx/SoftwareAxBlank.ht
m (last accessed on March 30th 2008)

Svrcek, W.Y. and Satyro, M.A.; "Process Simulation – From Large Computers
and Small Solutions to Small Computers and Large Solutions," Chemical Product
and Process Modeling: Vol. 1, (2006)

Taylor, R. and Krishna, R.; “Multicomponent Mass Transfer”; John Wiley and
Sons, New Youk (1993)

Taylor, R.; Koojiman, H.A. amd Hung, J.-S.; “A Second Generation Non
Equilibrium Model for Computer Simulation of Multicomponent Separation
Processes”; Computers and Chemical Engineering, 18, 205-217 (1994)

http://www.bepress.com/cppm/vol3/iss1/24 40
Satyro: Thermodynamics and the Simulation Engineer

Tooze, A.; “The Wages of Destruction - The Making and Breaking of the Nazi
Economy”; Penguin, 2006

Trebble, M.A. and Bishnoi, P.R.; “Accuracy and consistency comparisons of ten
cubic equations of state for polar and non-polar compounds”; Fluid Phase Eq., 29,
465-474 (1986)

Virtual Materials Group, Inc. “APR for Natural Gas Validation Documentation”;
Calgary, Alberta, Canada, 2004

Virtual Materials Group, Inc.; VMGSim version 2.8.7; Calgary, Alberta, Canada,
2007

Published by The Berkeley Electronic Press, 2008 41

You might also like