You are on page 1of 59


discussions, stats, and author profiles for this publication at:

Second-Law Analysis in Heat

Transfer and Thermal Design
DOI: 10.1016/S0065-2717(08)70172-2





Adrian Bejan
Duke University

Available from: Adrian Bejan

Retrieved on: 17 February 2016


Second-Law Analysis in Heat Transfer

and Thermal Design
Department of Mechunicul Engineering. University of Colorudo. Boulder. Colorado

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A. Second-Law Analysis in Engineering . . . . . . . . . . . . . . . . . . . .
B. Second-Law Analysis in Heat Transfer . . . . . . . . . . . . . . . . . . .
11. Irreversibility, Entropy Generation, and Lost Available Work
(Lost Exergy). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A. Opensystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
B. The Absolute Temperature Factor To . . . . . . . . . . . . . . . . . . . .
111. Local Entropy Generation in Convective Heat Transfer . . . . . . . . . . . .
A. Conductive versus Viscous Effects . . . . . . . . . . . . . . . . . . . . .
B. Entropy Generation Profiles and Maps . . . . . . . . . . . . . . . . . . .
C. The Entropy Generation Number Ns . . . . . . . . . . . . . . . . . . . .
D. The Impact of Heat Transfer Augmentation on Entropy Generation . . .
IV. Entropy Generation Minimization in Heat Exchanger Design . . . . . . . . .
A. Heat Exchangers with Zero Pressure Drop . . . . . . . . . . . . . . . . .
B. Heat Exchanger Geometry for Minimum Irreversibility . . . . . . . . . .
C. Sensible Heat Units for Energy Storage . . . . . . . . . . . . . . . . . .
V. Thermal Insulation Systems . . . . . . . . . . . . . . . . . . . . . . . . . . .
A. Minimization of Entropy Generation in an Insulation System
of Fixed Identity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
B. Engineering Applications . . . . . . . . . . . . . . . . . . . . . . . . . .
VI. Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Note Added in Proof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .





I. Introduction
The objective of this contribution is to summarize an important contemporary trend in the field of heat transfer and thermal design. This trend is
represented by the infusion of the second law of thermodynamics and its

Copyright @ 1982 by Academic &ss, Inc.

AU tifits of reproduction in any form reserved.
ISBN-0- 12-020015-5


design-related concept of entropy generation minimization. This new

trend is important and, at the same time, necessary, if the heat transfer
community is to contribute to a viable engineering solution to the energy



At the root of the growing interest in the thermodynamic irreversibility

of heat transfer lies the emphasis placed today on energy conservation
and the efficient use of energy. In any power plant, for example, the thermodynamic nonideality (irreversibility) of any of its engineering components causes a decrease in the net power output of the cycle. Likewise, in
a refrigeration plant the irreversibility accumulated over various components leads to an increase in the mechanical power input required by the
refrigeration cycle. Either way, the thermodynamic irreversibility of components such as heat exchangers, mixers, turbines, and compressors
amounts to a penalty in otherwise available work or, on a unit time basis,
available mechanical power. From an engineering standpoint, it makes
good sense to first identify the irreversibility associated with various components and, second, to design for less irreversibility in order to avoid the
imminent loss of available mechanical power.
The above conclusion follows directly from the sirrzultaneous consideration of the first and second laws of thermodynamics,* as we demonstriate
in greater detail in the next section. This is certainly not a new conclusion,
its first statement and engineering use dating back to the work of Gouy [I]
and Stodola [2]. Since then, the same principle was restated by others,
who in the process clarified its implications and made it more accessible
to engineering practice [3-61.
Today, there is a growing consensus that irreversibility analysis is a
powerful approach, in fact, the only one, to deciding which installation or
process is efficient or inefficient [7, 81. In addition, through irreversibility
minimization the engineer can make specific design decisions aimed at
conserving available work. Citing only a few examples, this approach and
its relation to cost minimization was exploited by Tribus and Evans [9] in
a cornerstone study of the thermoeconomics of seawater desalination.
More recently Kestin ct ( I / . [ 10, 113 established the thermodynamic faundations for evaluating the available work potential of geothermal installations. Gaggioli, Wepfer, and Elkouh [ 121 relied on second-law concepts to
show the great margin for improvement present in the contemporary de* In the thermal engineering literature, this procedure is recognized simply as secondlaw analysis.

St-cO N I ) - L A W

sign of heating, ventilating. and air-conditioning (HVAC) systems. Leidenfrost [ 13. 141 used the second law to analyze the potential for conserving available energy in a variety of power consuming processes.
The growing interest in irreversibility analysis in engineering is paralleled by a new emphasis placed on t h e teaching of the second law and its
applications in engineering thermodynamics. A significant effort in this direction was made by Kestin [IS. 161. Lu [17], and CravalhoandSmith[18].

The place occupied by heat transfer and thermal design in the greater
picture described so far is central. Engineering components and devices
for heat transfer are inherently irreversible. For example. a two-fluid
counterflow heat exchanger draws its irreversibility from two distinct
mechanisms, namely. heat transfer across the stream-to-stream temperature difference and fluid friction (pressure drop) in the two flow passages.
We shall consider this example in detail in Section I l l . For now, it is necessary to keep in mind how common and indispensable heat exchangers
are in power systems and in many other applications. This underlines the
important connection which muht be made between heat transfer and fluid
mechanics, on the one hand, and thermodynamic irreversibility. on the
The ultimate motive behind the infusion of entropy generation analysis
in heat transfer and thermal design is economic. Consider for a moment
the many factors which affect the decision of whether one thermal system
design is better than another [ 191. A number of these factors are political
in nature, but, to a large extent, the decision is the result of a cost-benefit
analysis, which takes into account the expense associated with manufacturing the device (capital cost) and the expense associated with running
the device (operating cost). The combination of design parameters which
yields the minimum cost subject to various constraints represents the economic optimum design. An important component in the cost analysis is
the degree of thermodynamic ineffectiveness exhibited by the device.
This component is the cost of electrical power required by the device
when it functions in t h e sense o f a wor-k-absorbing thermal system. or the
revenue derived from the sale of electrical power when the device functions in the work-producing mode. I t is clear that minimizing irreversibility in the thermal system yields a decrease in the operating cost. This
effect is usually accompanied by a parallel increase in the capital cost,
which demonstrates that the least irreversible design is not necessarily the
economic optimum. However. in order to make a sound economic deci-


sion, the thermal designer must understand the true thermodynamic performance of the heat transfer device; in other words, the designer must
know the irre versibiliry picture .
The work which lies ahead of heat transfer engineers, researchers, and
educators is to finally make the connection, to fill the gap, between the
traditional practice of heat transfer and the contemporary implications of
the second law. This is an activity which must be pursued in engineering
schools as well as in industrial circles, at the fundamental and applied
level as well. In order to be able to minimize the thermodynamic irreversibility of heat transfer equipment, engineers must first understand the fundamentals of the entropy generation mechanism. Also, they must understand precisely "where" in the heat transfer device irreversibility is being
produced, and "how much" is being produced locally. This requirement
is very similar to the relevance of "local" Nusselt number and skin friction information to the conceptual design of efficient and compact heat exchangers.
In this contribution the reader is exposed to a summary of first steps in
the direction of second-law analysis in basic heat transfer and thermal tiesign. The presentation is intended to introduce the inexperienced reader
to the new methodology of irreversibility analysis and irreversibility minimization in heat transfer. At the same time, the article is intended to show
the experienced heat transfer researcher the wealth of research opportunity which exists in this growing sector of heat transfer. The monograph
stops short of investigating the relationship between irreversibility minimization in thermal design and economic optimization of heat transfer
equipment. However, a number of thoughts in this direction are offered in
Section VI.
In writing the article. this author made a special effort to bring together
as much of the existing heat transfer irreversibility work as possible.
Despite the relative absence of publications on this subject, it is possible
that many workers in heat transfer have entertained second-law ideas
over the recent past. If so, it is hoped that through this article a more effective dialogue is initiated.

11. Irreversibility, Entropy Generation, and Lost Available Work

(Lost Exergy)

It has been pointed out already that the concept of thermodynamic irreversibility and its relation to the one-way destruction of available work
are not new. Brief expositions of this subject are found in some of the


most popular engineering thermodynamics textbooks [20-221. However,

its use in engineering is still sporadic, particularly in heat transfer and in
thermal design. For this reason this author finds it necessary to review the
irreversibility concept and its implications in the area of energy conservation.

Consider the open thermodynamic system shown schematically in Fig.

1, which is the most general model of a heat transfer device such as a heat

exchanger. The system is said to operate in steady state and steady flow.
This means that the thermodynamic state of matter surrounded by the
control surface does not vary with time, although it can vary from one
point t o another inside the control surface. In addition, the mass streams
piercing the control surface. mi and m k , are constant in time. The thermodynamic state at each of the inlet or outlet ports is represented by properties averaged over the port cross section; in other words, the bulk flow
model applies. The energy transfer interactions exhibited by the system
are shaft (shear) work transfer at a rate W , and heat transfer at a rate Qo.
The positive sense of these interactions is assumed as shown in Fig. I .
The system is in thermal communication with the environment, which is
modeled as a heat reservoir o f temperature T o . Most importantly, however, the control surface is drawn to include the actual heat transfer device p/rr.s those immediately adjacent parts of the environment affected by
the functioning of the device. This choice makes the heat transfer interaction Qo reversible. as it takes place across an infinitely small temperature
difference. As will soon become apparent, this choice is motivated by the





-- '1












I . Open thermodynamic system f o r the second-law analysis of a heal transfer


need to identify all the thermodynamic irreversibility associated with the

device. In the words of various writers, the irreversibility internul and exterm1 to the actual heat transfer device is incorporated in the aggregate
system delineated by the control surface. (For an instructive discussion of
this procedure, see Van Wylen and Sonntag [21].)
With all these assumptions made, we are still not in a position to
address the question of what impact the system irreversibility has on the
work transfer experienced by the system. We must make additional assumptions, for example, that the inlet and outlet thermodynamic states do
not change with the varying degree of irreversibility of the system. In
other words, the property pairs ( h , s ) ~and ( h , s ) k are considered fixed.
The freedom exercised in making this selection is what makes the notion
of lost available work a rrlative concept. We return to this comment later
in this section.
The statements of the first and second laws of thermodyamics for the
system of Fig. 1 are, respectively,

In writing the first law we assumed that the changes in kinetic energy and
gravitational potential energy experienced by streams mi and mk are negligible compared with the enthalpy changes retained in Eq. (2.1). The negative of the right-hand 4ide of Eq. (2.2) is the net rate of entropy generation in the system, S,,,, , a quantity which is always positive and in fhe
reversible limit equal to zero:


C (ms)i - Qo/To


Eliminating Qo between statements (2.1) and (2.2) one finds



describing an upper bound for the work transfer* of which the system of
Fig. 1 is capable. As indicated in Eq. (2.4), the algebraically maximum
work W,,, is achieved through reversible operation, i.e.,


9 mi(h



- ToS)k



* In what follows we use the terms"work transfer" and "heat transfer" for W and Q o ,
respectively, instead of the unit time terminology of "mechanical power transfer" and "heat
transfer rate."


This brings us back to the original question addressed by the analysis,

namely. the magnitude of the available work lost as a result of system irreversibility. Writing for lost ciwiltihlc~work
WLa,= w,,, - w
and combining Eqs. (2. I ) and (2.2) with notation (2.3) yields. finally.



Expression (2.7) is a well-known result discussed in the Introduction, result referred to by some authors as the Goiry-Stodolci r h e o r m [23]. We
use the same name in this monograph. There exists a general lack of
agreement concerning the terminology associated with result (2.7). For
example, for the difference (W,,,,,, - W ) Keenan [24] introduced the term
irrtwrsihility. This term will not be used here in order to avoid confusing
the irreversible features of a thermodynamic system with the quantity
irreversibility whose units are [ W ] .
Some authors prefer to discuss the above concepts by making extensive
use of the property "availability." defined as b = h - T,s. From Eq.
(2.51, the maximum available work becomes equal to the drop in availability across the system. Once more, the lost available work is that share of
the maximum available work no longer present due to system irreversibility. The situation is shown schematically in Fig. 2. Such diagrams are
usually found in availability analyses of complex engineering systems,
where the lost available work can be broken down according to the irreversibility of individual components [23].
The property "availability." made popular in the United States through
the work of Keenan. has a number of equivalents proposed overseas. In
Europe. the term "exergy" introduced by Rant [25] finds wider accepAVAI LAB1 LIT Y





FIG.2. Schematic representation of availability analysis of a thermodynamic system.


tance. Earlier, the thermodynamics literature was acquainted with such

terms as die griisste Nuf,-.arbeit (maximum useful work, Clausius [26]),
dip technische Arheitsf2ihigkeit (capability of performing work, BOSnjacovit [27]), and c>tiorgir utilisahle (useful energy, Danieus [28]).
The last term is perhap5 responsible in part for todays fashionable reference to an energy crisis or to energy conservation when, in effect,
we all mean available work or exergy instead of energy [9].
In this review article we zero in on result (2.7) and conclude that the
available work lost (destroyed) through the irreversible operation of the
system of Fig. 1 is proportional to the rate of entropy generation in the
system. The proportionality factor in this case is the absolute temperature
To of the heat reservoir (environment) with which the system exchanges
From a thermal design optimization point of view, the direct route
toward minimizing the loss of available work is through the systematic
elimination of irreversibility sources in the system, i.e., through minimizing S,,, . This is the important conclusion which motivates the work
summarized in this article. However, a thermal designer is usually not satisfied with simply minimizing the rate of entropy generation. Equally important is the price paid as lost available work, at the end of the entropy
generation minimization effort. In the next paragraph we demonstrate
that the simple statement (2.7), used for calculating W,ost, obscures a
number of subtleties relative to the meaning and the value of the temperature factor T o . These observations are usually not present in published
discussions of the Gouy -Stodola theorem.

Instead of Fig. 1, let u s focus on a system in communication with more
than one heat reservoir. The new system is shown in Fig. 3, where it is assumed that the system executes an integral number of cycles experiencing
the net energy interactions Qi (i = 0, 1, ..., n ) and W . For simplicity, we
consider a closed system. Again, the system boundary is chosen such that
all irreversibilities, if present, are located inside the system.
The analysis presented below is based on a communication by Jackson
[29]. In a manner identical to the preceding analysis, we apply the first and
second laws of thermodynamics to obtain

W = i Q i



S t c or.

Fit,. 3 . Closed system operating




c y c l e 5 while in thermal communication wilh n heat


At this point we arbitrarily eliminate thejth heat transfer interaction, Q,,

between Eqs. (2.8) and (2.9).This operation yields





- Tj/Zi)

(2.1I )


The lost available work. W , , ,


W , follows directly from Eq5. ( 2 . 8 )and


WI,,s,j = 7iLYgt.,,

( . j = 0, 1,

..., n )


We conclude that lost available work is a quantity which dr,pends on

our choice of reference heat reservoir, hence, subscriptj in Eq. (2.13). In
fact, Eq. (2.13) allows us to calculate not one but ( n + I ) magnitudes, all
representing available work lost with respect to a succession of ( n + I )
absolute temperature levels. Since the rate of entropy generation is the
same in all cases. result (2.13) implies

= (7J/To)w,0Sf,o




Equation (2.14) constitutes a device for converting one lost available

work value into another, depending on the choice of reference temperature. Furthermore, since one or more of the heat transfer interactions Ql
of Fig. 3 may be zero, according to Eq. (2.14) it is possible to define lost
available work relative to a heat reservoir with which the system does not
interact at all. Therefore, based on convention alone, it is possible to refer
lost useful work to a universal absolute temperature for all systems which
may enter our consideration. This is actually what is done, a popular
choice of universal temperature being the absolute temperature of the
standard atmosphere (298.16 K , i.e., 25C). In cryogenic heat transfer the
reference temperature i s usually taken as To = 300 K.
In conclusion, the important thought in maximizing available work is to
reduce the entropy gciiirrution in the> thermodynamic system. The notion
of lost available work carries with it the specification of an absolute temperature level relative to which the lost work is defined.
Since the lost availahle work is proportional to entropy generation, this
article focuses on entropy generation and its minimization through
thermal design. As summarized by London [30], the list of components
and phenomena responsible for entropy generation in engineering
systems is practically open-ended. A partial listing of entropy generation
sources includes [30] the following:
flow impact, shock, and fluid friction;
solid friction;
free expansion of a gas (blowdown, explosion);
flow throttling;
5 . mixing of dissimilar fluids, where the fluids can be dissimilar with
respect to temperature, pressure, or composition;
6. heat transfer across a finite temperature difference;
7. phase change where the initial conditions are not conditions of
equilibrium, e.g.. supercooled liquid, superheated liquid, supercooled vapor;
8. solution of a solid in a liquid;
9. plastic (inelastic) deformation of a solid;
10. electromagnetic histeresis;
11. joule ( P R ) heating in an electrical conductor;
12. virtually all chemical reactions that occur with any rapidity.

In this article we focus on the mechanism commonly responsible for entropy generation in heat exchange processes, namely, heat transfer across
a finite temperature difference and fluid friction. We also discuss specific
analytical methods aimed at minimizing entropy generation in engineering
components for heat exchange processes.




111. Local Entropy Generation in Convective Heat Transfer

It is instructive to examine first the entropy production mechanism at

the continuum level. before tackling the more complicated design task of
entropy production minimization at the heat transfer component level. To
do this we focus on convective heat transfer, the heat transfer mode most
commonly encountered in heat transfer equipment such as heat exchangers. The thermodynamics of irreversible processes in continuous
systems is well established and, tor a consistent exposition of the subject,
the reader is urged to get acquainted with any of the contemporary books
on irreversible thermodynamics [31, 321. A very good introduction can be
found in Chapter 24, "Irreversible Processes in Continuous Systems." in
Kestin's thermodynamics treatise (331.
Consider the local rate of entropy production inside a fluid engaged in
convective heat transfer without internal heat generation. If the fluid is
Newtonian and incompressible. and if it obeys the Fourier law of heat
conduction, the volumetric rate of entropy generation in Cartesian coordinates is [34]
S"' = ( k / 7 ' 9 [(dT/d.r)' t (i,T/il?.Y

+ (dT/dz)2] + ( p / T ) @


Function @ appearing in Eq. (3.1) is the viscous dissipation function,

better known from the energy equation for convective heat transfer [35]:

In Eq. ( 3 . 2 ) .us, r u , and I > , are the local velocity components in the Cartesian system.
From examining Eqs. ( 3 . I ) and (3.2) it is evident that high temperature
and velocity gradients are the features responsible for entropy generation
in the convective heat transfer situation considered here. Another important effect likely to influence S"' .. is the variation in absolute temperature
through the continuum. For the same temperature and velocity gradients,
S"' increases as the temperature of the medium decreases. This effect is
exploited in Section V . in the minimization of entropy generation in
systems exposed to large variations in absolute temperature. In most heat
transfer applications. however. it is safe to replace T appearing as denom* Note that for the remainder of the al-trcle we are dropping the subscrrpt "gen" in the
symbol for entropy generation rate. The prrnles indicate the local character of S. 1.e.. per
unit volume S " ' (W/m.'). per unit area .Y'" (Wlm'). or per unit length S ' t W / m ) .



inator in Eq. (3.1) by an average, characteristic, absolute temperature of

the medium. This approximation is consistent with the constant property
model relied upon in writing Eq. (3.1). It should also be pointed out that
the temperature gradients appearing in Eq. (3.1) may be written in terms
of a relative temperature 0 = T - T*, where T* is the reference absolute
temperature characteristic to the system.
An important property made visible by the volumetric entropy generation formula (3. l) is the separation of the two entropy generation mechanisms present, heat transfer in the direction of a nonzero temperature gradient and fluid friction. The relative importance of the two contributions
to S"' requires special attention. Interpreting Eq. (3.1) as

= SEnductive

+ s:l;,,,,,


one can show that the relative order of magnitude of the two terms is [34]




with the following notation:



Pr = v / a ,
T =


the Eckert number


the Prandtl number


the temperature difference number (3.7)

Here, u* and 0 * are the characteristic velocity and temperature difference

for the convective heat transfer system at hand. The "temperature difference number," Eq. (3.7), is a dimensionless parameter always present
in the second-law treatment of heat transfer processes. In most applications of engineering interest we find T + 1.
In heat transfer analyses it is often convenient and possible to neglect
the viscous dissipation term in favor of the conduction and convection
terms in the energy equation [35]:

This is particularly true in the case of subsonic gas flow. It is easy to show
that the order of magnitude of p@ relative to the conduction part in Eq.
(3.8) is dictated by the dimensionless group EcPr [36]. In other words, in
convective heat transfer problems in which the energy conservation statement may be considered without the viscous dissipation term, the group
EcPr is much smaller than unity.

St-cON [>-LAWA Y A L Y S I S


Comparing this conclusion with Eq. (3.4) we see that the energy argument by which Eq. ( 3 . 8 ) is regularly simplified has absolutely no bearing
on the question on whether or not the viscous dissipation effect is negligible in the makeup of entropy generation. It is quite possible to have a
heat transfer problem where, although the energy equation can be simplified according to EcPr e I , the entropy generation figure is in fact dominated by viscous effects. This is the case in which T is so small that the aggregate group EcPr/T is actually of order one. This observation is motivated by the fact that sometimes, perhaps suggested by first-law rules of
thumb and the desire to simplify the analysis, the viscous effect is
neglected ( I priori in the second-law treatment of convective heat transfer


h o t I I t.s A N D


The local entropy generation formula. Eq. (3. I ) , or equivalent forms for
other systems of coordinates [ 371. can be applied to known solutions for
the velocity and temperature fields in convective heat transfer. Thus. entropy generation profiles or maps may be constructed, maps illustrating
the areas and features of the flow which act as sources of thermodynamic
irreversibility. Below, we reproduce two examples which are based on
known textbook solutions in laniinar heat transfer.
As a first example, consider the Poiseuille flow through a round tube
with uniform heat flux 4" around its circumference (see insert of Fig. 4).
The velocity and temperature profiles for this flow are particularly
simple [37]:
u, =




(4"r,/X) [ -4 (.u/xo)


+ +(r/ro)4]


L!,,,,,= (ro2/4p)( - tlP/d.r)


x,,/ro =

r ~ , i ~ L ~ , , ,=J ( Pe,

the Pkclet number


Combining these expressions with the equivalent of Eq. ( 3 . I ) in cylindrical coordinates, one obtains the entropy generation profile in the tube

= (2R

R")' + 16/Pe2 + ( 4 E c P r / ~R'



where R = r / r o . In writing Eq. (3.13) we made the assumption that the

temperature variation across the pipe cross section is small compared
with the local absolute temperature 7" . The local entropy generation rate




FIG.4. Entropy generation profiles for heat transfer to laminar pipe flow [34].

depends on the radial position R, on the Peclet number Pe, and on the
group EcPr/7, which determines the relative importance of viscous effects. The Peclet number governs the relative importance of irreversibility
due to conduction in the longitudinal direction. We see that when Pe < 4,
the axial conduction contribution dominates the radial conduction effect.
Figure 4 displays a family of entropy generation profiles in the cross
section, for cases in which the axial conduction effect is negligible, Pe 4
4.The value of ECPr/T increases gradually to the point where viscous effects take over. In all cases, the pipe wall region acts as a strong source of
irreversibility. When E3cPr/T = 0, the maximum S occurs inside the
the same place where due to the wall curvature effect
fluid, at R =
the maximum temperature gradient is located. As viscous effects take
over, the point of highest irreversibility migrates toward the wall and, for
EcPr/.r > f , it coincides with the wall.
As a second example, consider the development of laminar momentiim
and thermal boundary layers along a flat plate. The situation is shown
schematically in the horizontal plane of the isometric drawing of Fig. 5 . At
a sufficiently large distance away from the solid wall the velocity and temperature are uniform, vX,= and T , . The wall temperature is constant, I*.
The entropy generation surface presented in Fig. 5 is based on the c1,assical Blasius-Pohlhausen solution [38] which, for the special case Pr = 1,



FIG. 5 . Entropy generation \urface foi- laminar boundary flow and heat transfer over a
flat plate [34].

reduces to




V J 1 J . X =



The similarity variable r) is equal to v ~ c ~ , ~ / ( u x while

) ] ~ ~ *f (, q ) I \ the function tabulated by Howarth [39] Combining solutions (3.14) and (3.15)



with the S"' formula (3.11, and neglecting the terms associated with gradients in the longitudinal direction x, yields

( S " ' / k )(vT*/f3muz,,)2


+ Ec/7)fl12/Rez


Here, Re, is the Reynolds number u,,,x/v. The three-dimensional display

of Fig. 5 was constructed using (xz),,,/v) and ( y u , , , / v ) in the horizontal
plane andfl12/Re, in the vertical direction. It is evident that the entropy
generation is limited to the boundary layer. In the y direction, S""is
highest, reaching a maximum at the solid wall. The longitudinal variation
of S"' is as l / x , indicating that like all gradients in the boundary layer solution S"' blows up at the origin.
Entropy generation profiles of the type presented in Figs. 4 and 5 may
be constructed for other basic laminar flow solutions. More examples of
such plots may be found in Bejan [34]. However instructive, this visualization technique is limited to laminar solutions for which analytical expressions (exact or approximate) for velocity and temperature may be
available. In turbulent flow one has to rely on average (integral) heat
transfer and friction results, condensed in experimental correlations. 'The
method by which the experimental heat transfer and fluid friction information is combined to reveal the irreversibility picture is presented in the
next paragraph.



Consider a duct of arbitrary cross section, shown schematically as a

round tube in Fig. 6. The sketch represents an infinitesimally short passage in a heat exchanger, for example. The heat flux per unit length q ' , the
mass flow rate m ,the wetted perimeter p , and the cross-sectional area A
are specified. In steady state a finite temperature difference AT exists
between the wall and the bulk temperature of the fluid. The schematic
Fig. 6 also applies to ducts with the cross section varying periodically in


- 1






j j


FIG. 6. Infiiiitesimally short duct in a heat transfer device.

St C O N 0-1 A W A N A L Y S I S


the flow direction. I n such ciiseh, q ' represents the heat flux averaged
over a length L containing an integral number of periods, whereas A becomes the minimum flow area.
It can be shown that for ( i t / \ ' prrrc' . s r r h s r m n c . c ~ flowing through the
system, the rate of entropy generation per unit length is [40]

We recognize here the contribution due to fluid friction in the duct. the
first term in Eq. (3.17), and the irreversibility due to heat transfer across a
nonzero temperature difference. I n many cases the temperature difference number 7 = 1777' is much smaller than unity; hence. one can
write ap prox imatel y

Although the A P and A 7 contributions appear separate in Eqs. (3.17)

and (3.18). they are in fact tightly interrelated through the geometry and
flow parameters of the system. The relations establishing this interdependence are the standard definitions [ 4 1 ]

q ' / ( pA r )


dP/il.\- = 2fG2//pD






/ I : , ,/c,G





where h,,, , I ) , G . j'. and St are. respectively, the average heat transfer
coefficient. hydraulic diameter ( 4 A I p ) . mass velocity, friction factor. and
Stanton number. Therefore, if the friction and heat transfer information is
available (f'and St). i t is possible based on Eq. (3.17) to evaluate the rate
of entropy generation integrated over the duct cross section. This statement holds for both laminar ( i t i d turbulent flow.
Since the momentum transfer information is condensed in dimensionless groups such as friction factor. skin friction coefficient. or drag
coefficient, and since heat transfer results are expressed similarly in the
form of Nusselt or Stanton numbers, it is appropriate to define a dimensionless group for second-law analysis in heat transfer, the ctirrop\' gc'tic'rn t i o t i nirttihrr N , . This new group is defined in a manner similar to friction



factor and Nusselt number:

N\. =

actual entropy generation rate

characteristic entropy transfer rate


Thus, if in the system of Fig. 6 the heat transfer rate is fixed, the characteristic entropy transfer interaction across the system boundary is q ' / T .
Consequently, the entropy generation number derived from Eq. (3.1'7) is
N.5, ==

+ 7/( 1 + T )

(rn/pq') (- d P / d ~ )


Similarly, local entropy generation numbers Ns- can be introduced based

on Eqs. (3.13) and (3.16).
There are systems in which it is possible to identify a characteristic entropy generation rate, for example, the minimum S in a thermal insulation
system of fixed geometry (Section V). In such cases, a more appropriate
way to define the entropy generation number is

actual entropy generation rate

characteristic entropy generation rate


An important characteristic of the local entropy generation rate for a

general duct is that its dependence on various design parameters is nonmonolonic'. This feature is exemplified by the fact that a parameter variation which yields a decrease in one of the two contributions to S ' is lik.ely
to induce an increase in the remaining contribution. For this reason, it is
difficult to predict in advance the changes induced in S ' or Nst through the
implementation of a certain design modification. Care must be exercised
before deciding whether a parameter change is indeed beneficial with
regard to minimizing entropy generation.
As a first example. consider the net effect of reducing the wall-fluid
temperature difference. For this, it is convenient to combine Eqs. (3..25)
and (3.19)-(3.23) into an expression showing N s , as an explicit function of
7 = A T / T . The end result is
N,s. = (5'/32) (fRez/St3)/.r3

+ ~ / ( 1 + T)


where parameter J plays the role of "duty" parameter for the duct:

= q'p/mp(c,n7':!


It is apparent that the AP and AT terms in Eq. (3.27) are coupled via the
temperature difference number T. Consequently, a minimization of the entropy generation number requires the optimum selection of 7 with respect
to the sum of friction and heat transfer contributions. In the limit A 3 0,

SI-C O N I>- I , 4 W A N A L Y S1S

The entropy generation nunihcr f o r :I general duct. a s a function ot the will-fluid

temperature difference and the cornhind parameter A . Eq. (3.30, [40].
Fie;. 7 .

the optimum is described by

Top1 ==

A I12




where the dimensionless parameter A . not to be confused with the flow

area of Fig. 6, is

[ ( 3 / 3 7 )(.f/St)]l'2(Re/St)J


For a given duty parameter .I. parameter A varies approximately as

Re/%. This is due to the fact that for common duct geometries the
Reynolds analogy between momentum and heat transfer holds to t h e extent that the group (JYSt)'" may be regarded as constant over the range
10' < Re < lo5 [42]. Figure 7 is a three-dimensional logarithmic plot of
Eq. (3.27). When T < T , , ~ , the
~ . heat transfer A T losses are small compared
with the fluid friction losses which account for most of N s . . In this region,
if the combined parameter A remains constant, the entropy generation
number N,s, increases sharply as T decreases: Nhv, T ". Conversely,
when T > T,,,,~, , N, dominated by losses due to inadequate thermal con-



tact. In this region, as A is kept constant, Nst varies roughly as Nst -- T .

The minimum is thus shallower (less critical) on the T > T~~~ side of the
Nst surface.
Too often, designers aim at maximizing the ratio h,,/(pumping power)
in order to improve the performance of a heat transfer passage. As a second example, we critically examine the meaning of this procedure from
the point of view of minimizing entropy generation. We show that, by itself, this procedure is irrelevant (or, at best, incomplete) with respect to
minimizing N s r.
A dimensionless group proportional to the ratio of heat transfer coefficient to pumping power is

(h,,pTp/m) (- dP/dx)-


Expressing the entropy generation number in terms of R and A as independent parameters, we find

N s r = 32/R23
+ A (R/3) [ l

+ A (R/3)]-

( 3 ..33)

This result is shown plotted as a three-dimensional surface in Fig. 8. The

features of this surface are very similar to the features presented in Fig. 7.
Thus, for constant A , there is an optimum ratio R for which Nst is a min-



FIG. 8. The entropy generation number as a function of A and the ratio of heat transfer
coefficient to fluid pumping power R , Eq. (3.31) [40].

imum: for sufficiently small value\ of A . the optimum ratio i \




Based o n Fig. 8 and Eq. (3.33). we conclude that increasing the ratio of
heat transfer coefficient to pumping power ( R )is not ,sii,jfic.ic,iir to ensure
improved thermodynamic performance, Since N s , depends o n more than
just R . the true effect of a proposed design change can only be evaluated
by estimating the changes induced in R and A , and eventually in N S , .



H ~ . A IH&\scr

GFN t R 4 I I O N

The entropy generation number i s an important parameter in deciding

the true merit of a proposed design change aimed at enhancing heat
transfer in a heat exchanger apparatus. The research on heat transfer augmentation techniques i s advancing rapidly, the more recent advances
having been reviewed by Bergles [43-451. According to the systematic ordering proposed by Bergles [44. 451, these techniques belong to two large
classes: ( I ) passive techniques requiring no external mechanical power,
and ( 2 ) active techniques which do require the use of external mechanical
power. The most popular passive techniques are those employing treated
surfaces, roughened surfaces. extended surfaces, swirl flo* devices. displaced promoters of heat transfer. and. finally, additives for liquids and
gases. The active techniques rely o n mechanical aids such as rotating heat
exchangers, surface and fluid vibration. electrostatic fields. and injection
or suction.
The task of evaluating the worth of a proposed augmentation technique
may be as important as conceptualizing and developing the technique. Of
the many evaluation criteria proposed (see, e.g., Webb and Eckert [46]
and Bergles r f a / . [47] ), the ratio of heat transfer coefficients (augmented
surface/reference surface) evaluated at constant fluid pumping power is
commonly preferred. However. a s demonstrated in the preceding paragraph, this procedure is only partially relevant in that it focuses only on
the possible payoff derived from an increased heat transfer coefficient,
saying nothing about the importance of a possible penalty associated with
the fluid pumping power loss.
We can assess the merit of a given augmentation technique by analyzing its effect o n the degree of thermodynamic irreversibility characterizing the heat exchanger in which the technique will be incorporated. In
this sense, if a technique leads to reduced entropy generation. the technique is effective. The evaluation procedure consists of comparing the




i-X , i


FIG. 9. Heat transfer augmentation by finning a bank of tubes in cross flow: (a) smooth
tubes; (b) finned tubes.

rate of entropy generation present in the augmented duct, Sa,with the

entropy generation in the reference, untouched, duct, So. In the example of Fig. 9, the finned bank of tubes represents the augmented design,
vis B vis the bank of smooth tubes used as reference.
The parameter describing the effect of augmentation on irreversibility is
the entropy generation number Ns,a, which follows from definition (3.26),


Ns,a = Sk/SA

Subscript a is used here to denote the fact that Ns is associated sole/.y

with the effect of augmentation. The reference and the augmented duct
are identical except for those parameters affected by the addition of the
heat transfer augmentation feature. For example, in Fig. 9 the spacing and
tube diameters are the same in both cases, and so is the working fluid, the
flow rate, m , and the heat transfer rate per unit length in the direction of
flow, q . Affected by the addition of circular fins is the hydraulic diameter, the wetted perimeter, the Reynolds number, the friction factor, and
the Stanton number.
A more useful version of Eq. (3.35) is
Ns,a =


+ [+o/(l + &)I

(Ns,iip - N S A T )


Sh.A*/SA,aT =

(Sb/sta) DalDo






2 fStRe(p2cpT/p2D4) ( w z / ~ ) ~ (3.39)

Ns,AT =


(Do/Da> (&/A,)*

Forms (3.37)-(3.39) are all based on the standard definitions for a duct,
Eqs. (3.19)-(3.23), plus the simplified expression (3.18). A special position is occupied by parameter +o, which describes the role played by fluid
friction in the irreversibility of the reference design. As shown below,
is a crucial parameter, which, in addition to the specific augmentation

technique, determines whether augmentation will indeed decrease the

rate of entropy generation in the duct (i.e., N.y,a< I ) .
The entropy generation number. E q . (3.36), has been arranged to show
the special forms assumed by N,,,;,in the two extremes, 4,
0 and do -+
=. As expected, in a situation dominated by heat transfer irreversibility
(& -+ 0) the entropy generation number will be proportional among other
things to the ratio of heat transfer coefficients, Eq. ( 3 . 3 7 ) . Conversely,
when the irreversibility is dominated by AP effects (4, + x ) , Ns., will
vary as the friction factor ratio. I t is necessary to point out that both
ratios, St,(Re,)/St,( Re,) and .til(Re21)Lfi,(
Re,), are function4 of the reference Reynolds number Re, since. for constant m ,

Relations (3.36)-(3.40)are general and can be used to evaluate the impact of augmentation on the irreversibility of ducts of diverse geometries.
As an example, we illustrate this procedure by considering the augmentation technique presented in Fig. Y . There are three finned-tube designs
( a l . a2. a3) being compared with the reference bank of smooth tubes. The
pertinent geometric parameters have been summarized in Table I [48-50].
Despite a conscious effort to compare tube bank geometries which differ
only with regard to tin geometry. it was impossible to find friction and
heat transfer data for smooth and finned tubes having exactly the same
array geometry ( X , , A',).
The cases considered in Table I have approxi-

Refe ire nce

Augmented \urface


(\moot h )


X , , transverse-tube pitch ratio

X , , longitudinal-tube pitch
Fin density number of fins per
one-diameter unit length
X , . fin height. relative to rube
diamete IFin thickne\s. relative to tube
Source of friction and heat
transfer data


2 00

I .91



1 (HI


2 00



5 hl


I .74

I 74


Sui-lxe n o .
10- I I I>
148. 491


Surface no.
(48. 501

0 0 155
Surface no
10-80 ,A
[4x. 501


0.01 17
Surface no.
10-92 A
[48, 501




l L l L


- I1



Re, x I 0-3
F I G . 10. Entropy generation number due to heat transfer augmentation, N,,,, versus reference surface parameters Re, and 4, .

mately the same array geometry, an inconsistency which does not affect
the main conclusions of this example.
The results of evaluating the available work conservation potential of
finning are summarized in Figs. 10 and 11. The entropy generation
number, Eq. (3.36), was plotted in Fig. 10 based on friction factor and
Stanton number data compiled by Kays and London [48]. It is evident
that, depending on the value of & , the augmentation techniques under
study can either decrease or increase the rate of available work destruction in the heat transfer device. There exists a special class of reference





Re, x I O - ~

FIG. 11. The function &,,(Ke,) for Ns.a= I . corresponding to three different finning




designs (Re,, +o) in which the implementation of a given augmentation

technique has absolutely no effect on irreversibility. This class is defined
by setting Ar,s,a= I in Eq. ( 3 . 3 6 )and Fig. 10. The result is represented in
Fig. 1 1 as a curve &(Re,) for each of the finned designs under consideration. The curve &(Re,,) divides the field +o - Re, into two distinct domains: only below this curve heat transfer augmentation is thermodynamically advisable.
In conclusion, we find that the evaluation of an augmentation technique
without specifying the design which is to benefit from the technique, is
inc*ornp/rti,from the second-law viewpoint of conserving available work.
Whether an augmentation technique is valuable depends on the design in
which i t is to be incorporated. This means that Re, and 6,or 4 ' / m , L) [Eq.
(3.39)] must be known before deciding the merit of augmenting heat
transfer in a duct.
The second conclusion is that graphs such as Figs. 10 and 1 I can be
used t o assess the relative effectiveness of different augmentation techniques with respect to conserving available work. In Fig. I I. for example,
three fin designs were considered for the job; in addition, we could have
considered the impact of other techniques, such as roughening the tube
surface or using pin fins. for the same design.

IV. Entropy Generation Minimization in Heat Exchanger Design

In this section we increase the degree of complexity of the heat transfer
apparatus, by addressing the question of irreversibility minimization in
heat exchangers. The topic of heat exchanger design, however, suffers
from a traditional bias toward first-law anlaysis and away from secondlaw considerations. The very name "heat exchanger" is suggestive of the
fact that the function of the apparatus might be to transfer a certain
amount of heat between two bodies (fluid streams, most often) at different
temperatures. This is not generally true. For example, in power and
refrigeration cycles the true function of heat exchange equipment is to
allow various components of the cycle to communicate with one another
in the least irreversible way possible (see Section V.B.3). This observation is supported directly by the Gouy-Stodola theorem ( 2 . 7 ) .In connection with this observation, i t i4 instructive to first consider the following
Figure 12 shows schematically a Brayton cycle heat engine with regenerative heat exchanger. The high-temperature end of the cycle (heater.
expander) must communicate with the low-temperature end (cooler. compressor) in order to exchange low-pressure fluid for high-pressure fluid.
The most efficient communication is established when: ( 1 ) there is no











FIG.12. Countertlow heat exchanger for a Brayton cycle heat engine [42].

pressure drop in the regenerative heat exchanger; (2) the inlet to the
heater is already as hot as possible, i.e., as hot as the expander outlet; and
(3) the inlet to the cooler is as cold as the compressor outlet. This limiting
case corresponds to a completely reversible regenerator (AT = 0, A P =
0). From this example it is apparent that the effective stream-to-stream
heat exchange in the regenerator is only one of the by-products of its true
function, that of allowing the hot and cold ends of the power cycle to cornmunicate (trade fluid) in the least irreversible manner. As discussed in
Section V,B,3, the function of the regenerative heat exchanger is to
practically insulate the hot end of the cycle from the cold end.
McClintock [51] appears to have been the first to recognize that the
concept of irreversibility minimization has a definite place in heat exchanger design. In a generally unnoticed conference article, McClintock
establishes the connection between Keenans irreversibility quantity
as a means of measuring thermodynamic nonideality [24] and the engineering task of designing efficient heat exchangers. He discusses the irre* This remark is amplified by the often-quoted example of a naturally occurring counterflow heat exchanger, the blood counterflow in the long legs of wading birds [52]. More visibly than anywhere else, the job of the counterflow is not to exchange heat but to permit the
birds body and its foot to exchange oxygenated blood for oxygen-depleted blood, as required by metabolism. There are reasons to believe that the birds metabolism is ultimately
geared onto the conservation of available work in its organs, particularly in the legs.



versibility minimization procedure in terms of a heat exchanger passage

of infinitessimally short flow length. Assuming that t h e wall-fluid temperature difference A 7 is fixed, McClintock demonstrates analytically the
existence of clear trade-offs regarding the selection of duct geometric
parameters such as hydraulic diameter. His analytical conclusions are
qualitatively similar t o those presented graphically in Figs. 7 and 8. From
a practical engineering point of view, his design conclusions are not
immediately applicable due to the artificial nature of the constant A 7 constraint placed at the basis of his study.
Le Foll [S3] used an irreversibility analysis similar to McClintock's to
evaluate the thermodynamic effectiveness of convective heat transfer
from gas-cooled nuclear reactors. Le Foll proposes the use of St"/J'as a
performance evaluation criterion (figure of merit).

A. Ht-.41 E X C H A N G E RWS I I H


The simplest irreversibility analysis of the process in a heat exchanger

is based on the assumption of Lero pressure drop in the flow passages.
Accordingly, the only source o f entropy generation appearing in the analysis is the transfer of heat across the nonzero temperature difference
which exists between the two fluid streams when the heat exchanger area
is finite. Although the practice of neglecting the fluid friction irreversibility is not generally supported by order of magnitude arguments such as
the discussion centered around Eq. (3.41, it is instructive to first consider
this highly idealized limit for the insight it provides into the relationship
between the thermodynamic irreversibility and the choice of heat exchanger design parameters.
Irreversibility analyses of two-stream (parallel and counterflow) heat
exchangers with zero pressure drop have been reported by a number of
authors [23, 30, 54. 551. Here we summarize an example illustrated by
Tribus [54], who considered the irreversibility in a heat exchanger with
equal and constant capacity rates, C',
= Cmin= mc, (Fig. 13). The
overall heat transfer coefficient I/ is also a constant in the analysis. Tribus
shows that in this limit the rate of entropy generation takes a particularly
simple form:

where T I , Tz are the inlet absolute temperatures of the two streams and
N,, is the number of heat transfer units UAlmc,. This result is shown
plotted in Fig. 13 as Sarn/tnc.,,versus the heat exchanger effectiveness E,
which in the example shown here is given by

Nt,/( 1 + N,,)




0.2 0.5






10 a


FIG. 13. Entropy generation rate in a balanced counterflow heat exchanger with zero
pressure drop. (After Tribus [54].)

The symmetry of Eq. (4.1) is reflected graphically in Fig. 13. The entropy generation rate is invariant to the transformation T 1 / T z+ T z / T , ,
meaning that in Fig. 14 the absolute temperature may be measured in
either direction on the abscissa. In other words, Tl may be assigned to
either the warm inlet or the cold inlet without changing the appearance of
result (4.1).
The entropy generation rate reaches a clear maximum when N,, = 1,
maximum given by
Sec"/ln('P =

In[* + i(Tl/TZ + Tz/T,)I


This feature is not intuitively obvious since we would expect the heat






FIG. 14. Schematic of counterflow heat exchanger with imbalanced capacity rates [42].

Sc( Y

N I)- 1,A w

A NA L Y s I s


transfer irreversibility to decrease monotonically with the increasing heat

exchanger area ( N , , ) . I n the range N , , > 1, the behavior is as expected,
the entropy generation rate decreasing in the direction of smaller
stream-to-stream temperature differences (higher N , , ' s ) . Below N , , = 1 ,
however, the stream-to-stream temperature difference is of the order of
IT1 - T,I and is relatively insensitive t o changes in N , , . Consequently, as
Nl, decreases the net heat exchange between the two streams decreases,
and since the temperature difference is roughly constant, so does the entropy generation rate.
Golem and Brzustowski [5S] examined the irreversibility of heat exchangers using the Reistad effectiveness eK defined as [%]
EK =

availability (exergy) gained by the cold stream

availability (exergy) lost by the warm stream


In the limiting case of reversible heat exchanger operation. eHis equal to

unity. Neglecting the irrever5ibility associated with frictional pressure
drops, Golem and Brzustowski showed that the Reistad effectiveness reduces to

where subscripts C. H. out, and in refer to the cold stream, hot stream.
outlet, and inlet, respectively. The "+" sign applies to counteiflow and
sign to parallel flow. l h i s expression holds for ideal gases and
incompressible liquids. The same authors extended the eRconcept to the
local level, showing that when the longitudinal temperature variations
Tc(.r), T&) are known. one can evaluate locally the relative loss of availability (exergy) due to heat transfer across the stream-to-stream temperature difference.
L b _ ' '


H t . 4 1 E X C H A N ( ; t R GF.OMI I K Y F-OK M I N I M U M I R R E V t R S I R I I . I I Y

A more realistic approach to the second-law analysis and synthesis of

heat exchangers must take into account the irreversibility due to fluid friction, focusing on the strong ( , o / ~ p / i ~which
exists between heat transfer
and fluid friction irreversibilities. This was done in a recent article by this
author 1421. This i s the first instance in which the procedure of entropy
generation minimization is presented in the context of a complete heat exchanger system, namely, the class of countedow heat exchangers for
gas-to-gas applications. A new design method, thr, tirrnihrr (?I' c'ritropy
gericrrrtiou rrriits N.$.is proposed in lieu of the traditional number of heat
transfer units N , , [48].
Consider the schematic representation of Fig. 14, in which indices 1 and



2 have been associated with the minimum and maximum capacity rates
(rnc,) in counterflow. Defining the entropy generation number N, in the
manner indicated in Eq. (3.24),

the rate of destruction of available work in the heat exchanger can be

written as

= (Cmin/Crnax)


MTz/T,) +


[I -

(1 -



- (Cmin/Cmax) (1


(1 - T , / T J 2
x ex~[-Ntu(l - C m i n / C m a x ) l
( 1 - T1/TdI (1 - (Cmin/Crnax)
x exP[-Ntu(l - Crnin/Cmax)l}

(Af/P)I + (R/cp)Z (AP/P)z


Here, R is the ideal gas constant, the rest of the symbols having been defined in Fig. 14 and in the Nomenclature.
Equation (4.7) is based on the assumption that the heat exchanger is
nearly ideal, in other words, (AP/P)l,ze 1 and 1 - E 6 1, where E is
the effectiveness (Tl,out- T1)/(Tz - T I ) .This assumption makes visible in
Eq. (4.7) the two mechanisms responsible for entropy generation. The
first two terms account for irreversibility due to heat transfer betwee:n
streams, across a nonzero temperature difference. The last two terms,
individually, represent the fluid friction effect in each of the two ducts of
the heat exchanger. It is easy to see that in the Ntu + limit the second
term vanishes, and the AT irreversibility is due solely to flow imbalance
(Cmax> Cmin).Figure 15 shows how the imbalance contribution to N F ,
the first term in Eq. (4.71, varies with the capacity ratio and the absolute
temperature ratio. From a practical fiesign viewpoint, it is important 1.0
know the magnitude of the imbalance contribution when one seeks to
minimize N, by increasing the N,, and by decreasing the friction APs: the
imbalance component tells the designer when he or she has reached the
point of diminishing returns in the minimization of overall Ns, since in the
limit N,, -+ 00, APl,z-+ 0. the imbalance term is the sole survivor in
Eq. (4.7).
A further simplification of the Ns Eq. (4.7) is achieved when one con. Applying
siders the case of nearly balanced capacity rates, Cmin= CmaX
the calculus limits as C,,in C,,,, the number of entropy generation
units becomes




1, 1,

F I G . IS. The entropy generation due l o capacity rate imbalance in a counterflow heat exchanger [4?].


= [(~nla,/(,nln)

I I [(];/TI) - I - In



[ ( T 2 / T 1 ) 1 2- (7,/7;)1]2 N,,

+ (cmin/crnax)
(Klc,,,),( A P / P ) ,
N,?, = (C,,,,,,/C,,,,,,)


(T1/T2)*12N& + ( R / ( . , ) , ( A P / P ) ,

In this limit Nshas separate terms describing the rate of entropy generation o n each of the two sides of the heat exchanger. Moreover, the irreversibility of each side N s , , , appears a s the sum of one contribution due to
heat exchanger A T and one contribution due to frictional AP losses:





This form is similar to the f o r m of N,, for an elementary heat exchanger

passage. Eq. (4. I). Equation ( 4 . 8 ) is pivotal in the design process, as it
permits the minimization of N , , for each side of the heat exchanger. separately. Furthermore. the analytical form of N,, and N,, is similar (identical in the balanced flow case). s o that the design procedure for each side
is the same.


Nirttrtwr c ! f

E I I / ~ oG ~
< , t\ l ~
( , / i l / i o t l 1 J t l i t . y .fbr OtIc S i t l ~ K,,,

I t is instructive to examine the manner in which various heat exchanger

parameters affect the irreversibility of one side. Recalling that from the



definitions of number of heat transfer units and friction factor we can

write [48]
Ntm,, = (4L/D) St

= f(4L/D)




the irreversibility per side assumes the general form





61.2 (;)l,,




Equation (4.15) contains the usual heat exchanger notation, where, in addition, we defined



= Cmin/Crnax


Crnin/ Cm ax



[ ( T , / T , Y - (Tl/T2)]2,

g = G/(2pP)



Thus, g plays the role of dimensionless mass velocity, while T~ has the
same significance as the temperature difference number ( 7 ) relative to the
difference between inlet temperatures, TI and T,.
The dependence of N s , , , on design parameters such as Re, 4L/D, andl g
is shown qualitatively on the three-dimensional logarithmic plot of Fig.
16. The graphic construction of Fig. 16 is actually based on empirical data
on turbulent flow inside smooth tubes, where both St and f a r e proportional to Rep0.,; one can construct qualitatively similar three-dimensional
plots for other heat exchanger surfaces.
2. Optimum Duct Geonrrtry, 4LID

From Fig. 16 and Eq. (4.15) it is evident that number of entropy generation units always increases when g increases, with 4L/D and Re remaining fixed. Unlike g, the duct slenderness ratio 4L/D plays a definite
trade-off role: for constant g and Re, there exists a clearly defined optimum 4L/D for which the resulting N8,,*is a minimum. The optimum condition for each side is described by


2g[abr0 (R/c,) (jC/St)]1/2


For common heat exchanger surfaces the group (f/St)1/2is only a weak
function of the Reynolds number [42]. Therefore, Eq. (4.20) establishes a



FIG. Ih. Number of entropy generation units for one side of the heat exchanger, as a
function of L / r , , . g. and hRt.
(hence r , , = 11/41 [4!].

one-to-one correspondence between the mass velocity g and the lowest

rate of entropy generation achievable in the heat exchanger duct.
There are various ways in which the N s optimization of a heat exchanger may be conducted. The three dimensionless design parameters
for one side of the heat exchanger, 4 L / D , g , and Re (Fig. 16). define a
three-dimensional space of possible design conditions. If the degree of
thermodynamic irreversibility of each side N,,,, is specified in advance,
then, via Eq. (4.15). the number of unknowns for each side i \ reduced to
two. If, as in Eqs. (4.19)and (4.20).N S , , *is not specified but. for a given g
and Re, the ratio 4L/D is chosen such that N,,,, is minimized. the number
of design unknowns per side is again reduced to two. In practice, the
number of independent design parameters may be less than two per side
due to additional design constraints. The design procedure subject to two
constraints. constant heat transfer area and constant heat exchanger volume, is presented in Bejan [42]. I n addition, Bejan [42] develops the
complete design procedure for minimum heat transfer area subject to fixed
N,,,, . This procedure is applied eventually to design of a specific regenerator for a Brayton cycle.



As a second example consider the minimization of irreversibility i.n a
sensible heat unit for energy storage. Traditionally, the thermal design
and optimization of a sensible heat storage unit relies on the view that the
system thermal Performance can be assessed based on how much thermal
energy the unit can store. In short, a unit is considered more efficient than
another if-for the same heat input and the same amount of storage
material-it is capable of storing more thermal energy. This point of view
is generally accepted and serves as basis for testing and evaluating the
thermal performance of sensible heat (fluid and solid) storage units [57].
Bejan [S8] analyzed t h e performance of such units by treating them as
systems intended to store available work, the function they perform in
most power applications.
Consider, for example, the system shown schematically in Fig. 17. It
consists of a large liquid bath of mass M and specific heat C placed in an
insulated vessel. Hot gas enters the system through one port, is cooled by
flowing through a gas-liquid heat exchanger immersed in the bath, and is
eventually discharged into the atmosphere. Gradually, the bath temperature T as well as the gas outlet temperature To,, rise, approaching the hot
gas inlet temperature 7', . It is assumed initially that the bath temperature
equals the environment temperature T o . The bath is filled with an incompressible liquid such as water or oil. The stream m carries an ideal gas, for
example, high-temperat ure steam or air. The stream of hot gas is supplied
continuously at T , and P o ; before entering the unit, the stream is compressed to Po + AP in order to overcome the pressure drop caused by
friction in the heat exchanger.
The time dependence of the bath temperature and the gas outlet temperature can be derived analytically and the result is available in the engineering literature [S9]. Of interest here is the total amount of entropy generated from the beginning of the charging process until an arbitrary time r
(S/mc,t) = (R/c,,) In(l

+ 0-'{ln[l

+ AP/Po)+ T
+ ~ ( -1 c ~ " ) ]

+ T)


~ ( 1 e-"")}


In Eq. (4.21),T is the dimensionless temperature difference ( T , - T o ) / T o ,

while 8 is a number proportional to the elapsed time, i.e., the total quantity of hot gas used over a given time period t :

= mc,t/MC


Parameter y is shorthand notation for

y =



.1 1 q u ~ dh s a l

Fic, 17 Schematic 01 wnwhle hedt

ature g ~ [ \i x ]

(trill l o i


energy storage from d 5tre.inl ot hlgh temper-

where A is the total heat transfer area on the gas side, U is the overall heat
transfer coefficient based on A . and UA/(rwc,) is the heat exchanger
number of transfer u n i t \ N,, .
According to the Gouy-Stodola theorem (2.7), T,S given by Eq. (4.71)
equals the available work destroyed due to process irreversibility. We can
compare the lost portion of the ;ivailable work with the maximum available work associated with the same amount of hot gas, m(',,/.

ct',,,,, =



[T -


+ TI]


Equation (4.24)is based o n calculating the drop if gas availability from the
inlet condition T , P,, , before compression to Po + A P . to the "dead"
state T o , P o . Dividing Eqs. (4.7I ) and (4.74)side by side, one arrives at the
number of entropy generation units
for the 0 -+ (3 heating (charging)
From Eqs. (4.21)and (4.24). i t is easy to see that .'V, will again
have two parts, one describing irreversibility caused by friction. the other
irreversibility due to heat transfer across the nonideal heat exchanger

The dimensionless time c) plays iin important role in determining the

size of the AT portion o f N s . Figure I X shows the dependence of /V,s.ATon
8, N,, and T for a sequence of discrete cases. It is clear that for a given
N , , and T there exists an optimum time 8 when the A 7 irreversibility
reaches a minimum. In the c) + 0 limit the entire available work content
of the hot stream is dissipated by heat transfer to the liquid bath. which is




0 5

8FIG. 18. Entropy generation caused by heat transfer across a finite temperature difference, as a function of heat exchanger N , , , temperature difference T, and charging time 8:
(a) T = 0 ; ( b ) T = I : (c) T = 2 [5X].

S t u )hi I )- LAw A NA L Y S Is



T, - T<,


Optimum \tornge ternperdure l o r niinirnurn destruction of avdilnble uork during

the chdrging process 1581
Fit, I9

initially at 7 0 . In the c) + x limit, the gas stream exits the heat exchanger as hot as it enters: its available work content is again dissipated
entirely by direct heat transfer to atmospheric temperature.
The optimum time t),,,,
and the corresponding Ny,yT can be determined
as functions of N,,, and T. An interesting result associated with this optimum situation is the optimum temperature of the sensible heat u n i t , reproduced here in Fig. 19. We see that when the storage unit operates at
the optimum, the total amount of thermal energy stored as sensible heat at
the end of the charging proce\s. hlC(
- To),is far from the maximum
storage capability of the liquid bath. M C ( T , - To). In accordance with
Fig. 18, if one seeks t o heat the liquid bath until the stored thermal energy
reaches its maximum ( 7 ' - 7 . x ) ,one runs the risk of storing none or a very
small fraction of the total available work drawn from the high-temperature
Other design trade-offs are brought to light by the entropy generation
analysis sketched above. For example, in Eq. (4.25), increasing the heat
exchanger N,, has the effect of decreasing NsSaTand that of increasing
N,.,, . Therefore. i t is possible to determine an optimum number of heat
transfer units N,, , which leads l o the mimimum overall N,, . The optimum



N,, is derived purely from thermodynamic considerations before proceeding with the actual design of the heat exchanger. The analytical details of
finding the optimum N , , are left to the reader who is encouraged to consult Bejan [ 5 8 ] .
In conclusion, the irreversibility analysis summarized above outlines
the inherent inefficiency of storing available work as sensible heat. Even
under the most efficient of operating conditions, 8 =
and using a
pressure drop-free heat exchanger, the unit stores less than 50% of the
available work delivered by the high-temperature source. Again, the irreversibility analysis reveals the heat exchanger trade-off between losses
due to fluid friction AP and heat transfer across a finite AT, trade-off cliscussed in the preceding section on counterflow heat exchangers. Despite
the simplicity of the sensible heat unit analyzed (Fig. 17), the conclusions
reached here apply qualitatively to other, more complex, units such as
solid units for sensible heat storage [60].
V. Thermal Insulation Systems

It is appropriate to focus on thermal insulation systems in the closing

stages of this review since, as illustrated below, this class of thermal
systems is unexpectedly wide and complex. For a start, we should try to
answer the question of what practical purpose might a thermal insulation
serve. One generally accepted answer is that the thermal insulation a:<a
system prevents two bodies (surfaces) of different temperatures from
coming into direct thermal contact. Figure 20 schematically illustrates
three different types of thermal systems which qualify as thermal insulations according to this generally accepted view.
The classical case of a low thermal conductivity material sandwiched between two bodies TH (high temperature) and TL (low temperature) is presented in Fig. 20a. Although the two bodies may communicate mechanically, the warm body is in direct contact with a body of equal temperature, TH. The same can be said about the cold body T L .A power plant and
refrigeration plant operating in cycles, while in thermal communication
with two heat reservoirs, are shown in Fig. 20b and c , respectively. In
these examples, again, the warm reservoir sees the warm end of the
cycle, while the cold reservoir is in contact with the cold end of the same
cycle. So, although the two bodies TH, TLcommunicate to exchange heat,
they do not make direct contact, thanks to the presence of cyclic devices.
Therefore, heat engines and refrigerators function as thermal insulation
systems in the sense of the definition stated in the preceeding paragraph.
Two noteworthy conclusions result from the discussion based on Fig.






Q, I



Flc,. 2 0 Three examples of thern1;il in\ulation \ystems preventing two surfaces of different temperatures from corning into diicci thermal contact: (a) conducting layer: ( h ) heat
engine: (CI refrigerator.

20. The first conclusion is that thermal insulation systems. as a class. are
varied, numerous. and complex. There is considerably more to this class
than the textbook-style example shown in Fig. 20a. The second conclusion which must be stressed in this article is that the heat transfer from 7-k,
to 7; is ~ r o lticc~es.wr-ilyc . o r i . s f t i r r / across the insulation. In a heat engine.
for example, the heat transfer decreases in magnitude, a decrease which
accounts for the net work output of the engine. Therefore. applying again
the Gouy-Stodola theorem ( 2 . 7 ) . i t makes perfect sense to inquire into
the possibility of reducing the irreversibility of an insulation system. particularly in large-scale power- refrigeration applications. The objective of
this design procedure is to squecre the most available work out of a given
insulation system, in the manner illustrated in Fig. 20b.
A. M I N I M I L A T I O N


1 N S l I L A l I O N S l S I t - M Of-

G t N F R A T l O N IN A N

FIX1 1)


Consider the one-dimensional insulation system in Fig. 7 1 . The heat

flow path length (i.e., the insulation thickness) t and the crosh-sectional
area A are given. In any plane .t-. the heat transfer q is proportional to the
local temperature gradient:
y = Ah



where the proportionality factor h( 73 is the "effective" thermal conductivity o f the insulation material. Although Eq. (5.1) is identical to the
Fourier law of heat conduction, in this treatment its meaning is considerably more general. as it describes systems in which the heat transfer
mechanism is not necessarily conduction. In the analysis, the local temperature T and heat transfer y are controlled externally by placing the insulation in local thermal equilibrium with the T end of a reversible



FIG. 21. Analysis of entropy generation in a continuous, one-dimensional, thermal insulation system: (a) the "brake" effect buried in an insulation; (b) local extraction of available
work [62].

(Carnot) cyclic device operating between T and T , (this device is shown in

Fig. 21b). Allowing for ii variable heat transfer function q(Z') is the necessary degree of freedom for being able to extract the available work internally dissipated in a "constant q" insulation (this dissipative process is illustrated in Fig. 21a).
The quantity of interest is the total rate of entropy generation in the insulation:

where the integral obtained was based on an entropy generation analysis

around the cross-hatched element of Fig. 21b. There are two ways in
which S can be minimized, the immediately obvious being suggested by
integral (5.2). One effective method consists of decreasing the effective
thermal conductance ( k A / t ) ;this approach motivates the continuing quest
for inexpensive low thermal conductivity materials. A more subtle way of
minimizing S applies to insulations of fixed geometry and constitution [61,
621, i.e., systems in which the thermal conductance has already been decreased to the limit imposed by economic considerations. For systerns

where t . A , and k ( 7 )are known, the con4titutive relation ( 5 . I i becomes an

integral constraint to be met by the heat transfer function y ( 7 i

In mathematical terms, the ta\k of minimizing the irreversibility iissociated with a one-dimensional continuous insulation of fixed identity boils
down to minimizing integral ( 5 . 2 ) subject to constraint (5.3) by properly
choosing the unknown function q( 7'). The result to this variational calculus problem is [63]
y,,,,,(T)= 7 A 1 ' ( A / / )

which. combined with Eq. ( 5 . 2 ) . yields






'I I




In conclusion, the optimum inurlation design is characterized by ti heat

transfer function which i , ( / r i ~ swith the absolute temperature, The function 4( 7') is controlled externally by removing a known fraction ~ l at
y each
intermediate temperature interval 117 across the insulation sy4teni.
The practical implications o t the above optimum are best illustrated by
considering a system in which X is constant, k = k , . and comparing the
traditional design in which 4 is constant with the optimum dekign in which
4 varies according to Eq. (5.4). This comparison is presented in Table 11.
We see that the optimum design is characterized by a heat transfer rate
which is proportional t o the local absolute temperature. The optimum
heat transfer function is achieved by removing a constant share dy over
each temperature interval (17'. 'lhe measure of how inferior the constant 4



FIG. 22. Entropy generation number of constant q insulation system, as a function of

absolute temperature ratio [ J ~[ h 2 ] .

design is relative to the optimum design is the entropy generation number


s* /Smin


As shown in Table 11, N , depends on the absolute temperature ratio

p T = T 2 / T 1 .This relationship is shown plotted on Fig. 22 as curve n = 0.
Figure 22 was constructed for the more general case, where the effective
thermal conductivity k ( T ) obeys a power law relationship over the te:mperature range covered by the insulation, k ( T ) = kip,". Thus, in the more
general case, N , depends on both p T and n [62].
In conclusion, decreasing the thermal conductance A k , / t without providing intermediate heat transfer dq/dT is the only approach to a better
thermal performance only in the limit p T + 1. As soon as the temperature
difference ( T , - T , ) is of the same order of magnitude o r greater than T , ,
the thermodynamic performance of a given insulation (fixed A k , / t ) can be
improved drarnaricril/y by appropriately controlling the local heat flow
dependence on absolute temperature, q ( T ) . Figure 22 is a comparison of
only one practical design ( q = constant) with the theoretical optimum. As
illustrated by the variational approach to determining the optimum de* This ratio is directly relaid to the temperature difference number T encountered earlier,
+ T. where T = ( T 2 - 7 , ) / T , .

pT = 1

S L c o N I )- LAw A N A L Y sI s


sign. there exists an infinity ofde\igns in which q ( T ) may or may not come
close to q,,,,,(T).The ability of achieving a heat transfer distribution y ( T )
which closely resembles 4(,,,41)depends on the amount and distribution of
lateral heat transfer effect.
We learn by examining Fig. 2 2 that insulation systems exposed to large
absolute temperature ratios p,.\tand to benefit most from a de\ign where
the heat leak function is controlled externally via principle ( 5 . 4 ) . Cryogenic insulations are examples of such systems. Below. we review a
number of applications in which low-temperature systems are thermodynamically optimized from the viewpoint of irreversibility minimiz a t ion

discussed in this monograph.

B. E N G I N E [ - R I APPI
N G I(


Considerable research is prrseiitly focused on the large scale applications of superconductivity and their possible role in the energy projects of
the future [64]. For example. giant superconducting magnets are being
developed for the purpose of creating the magnetic confinement vessel for
plasma in fusion reactors [ 6 5 ] . hven larger superconducting systems are
contemplated a s energy storage systems for peak power-shaving applications [66]. In the electric power indu\try, significant resource\ are being
invested in the development of rotating electric machinery w i t h superconducting windings [67].
Since superconducting magnet\ must operate at liquid helium temperatures. applications of the type listcd above require thermodynamically efficient insulation \ystems to provide and maintain a working lowtemperature environment. We show below that the conceptual design of
such systems follows closely the principle of entropy generation minimization presented in the preceeciing section.

M l J c . / l l r l l i c ~ t rSlr/pot

Large superconducting magnet4 must be supported from room temperature. The supports must be thick and strong to withstand the static
loading experienced by the cold space. At the same time, these supports
must be short enough to provide adequate stiffness to the assembly.
Overall. the task of thermodynamically optimizing this class of supports
consists of minimizing the irreversibility of an insulation system of fixed
geometry and identity. with t h e final objective of minimizing the refrigerator power input required to keep the cold space cold.
In practice, the optimum design of cryogenic mechanical supports may
involve the use of a stream of cold gas (helium) flowing in parallel with the
conduction path, again\t the conduction heat flow (see Fig. 23). If the








- 4


A , k(T)


FIG. 23. Intermediate cooling effect provided by a stream of cold helium gas in counterflow with the conduction heat flow through a structural support [69].

capacity rate of this stream is mc,, then the support heat leak varies according to dq/dT = rnc,, . If the support conductivity is temperature independent and if the stream flow rate is selected such that m c , =
( A k l / r )In(T,/T,), then, from Table 11, the lateral cooling provided by ithe
cold stream is the optimum required for the job. In reality, the conductivity of most structural materials varies strongly with the temperature,
particularly at low temperatures. Consequently, the one-stream intermediate cooling technique, which has been popular for some time starting
with the work of Scott [68], can only approximately approach the irreversibility minimum ( 5 . 5 ) . In such cases, one still faces the design question of using the best capacity rate mc,, which will differ from
( A K l / t ) In(T,/T,) listed in Table I1 for k = constant materials. The optimum capacity rate mc,, for actual structural supports is reported in Bejan
and Smith [69].
Another important aspect of the optimum design of cryogenic supports
is the design and fabrication of the stream-to-conducting solid heat exchanger which houses the intermediate cooling effect dictated by 1he
variational principle. This is a difficult task when we think that bringing a
gaseous stream in good thermal contact with a conduction heat current in
a solid piece implies machining holes and channels in the solid, thereby
weakening the structural member. The entropy generation minimization
in this class of heat exchangers is discussed in Bejan and Smith [69]. The
heat exchanger fabrication is less difficult and more economic if, instead
of building a continuous heat exchanger from T1 to T, between gas stream
and mechanical support. one builds a succession of discrete heat exchangers (cooling stations). The optimum heat leak function required by
Eq. (5.4) is therefore approximated by a stepwise-varying function.
Hilal and Boom [70] addressed the question of optimum cooling and
spatial positioning of di.vcrete heat exchangers on conducting mechanical

supports. Their work was motivated by the need to design leastirreversible support structure\ for ;I stadium-size superconducting energy
storage system [MI.
A s shown i n Fig. 24. Hilal and Boom considered a
support stretching from 7 , = 1.8 K (magnet temperature) to I , , , =
300 K (room temperature), the support being cooled at ( N t I ) discrete
points (nodes. cooling stations) hy contact with a refrigeration plant. Consequently. the conduction heat flow along the support has a stepwise variation in temperature. The refrigerator power input associated with cooling
each point along the support i4
Wi = A ~ i [ ( l l / A . r l ) ll-,/A.rl-l]

for i

2 . .... !Y + I


W , = Aylll/A.r

for i




and Axi is the support length between the ith and (i + I ) t h cooling stations
(Fig. 24). A s in Fig. 23. A is the support cross-sectional area. actor Ci appearing in Eq. (5.9) is the "Carnot efficiency" of the refrigerator cooling
node i; note that C', = 1 if the refrigerator is reversible. and I ' , '
, 1 if the
refrigerator is irreversible [7 I ] .
The total refrigerator power expen\e to be minimized is

.v+ 1

1 W l / A = C DilAxi



\ i I

W/A =








Equation (5.11) muht be minimized subject to the constraint of fixed support length,





which is a constraint equivalent to integral (5.3) in the continuous cooling

problem considered in the previous section. The minimization procedure
is based on t h e Lagrange multiplier method and is described in detail i n
* The "Carnot efliclency . of il refrigerator i \ defined a s the ratio of the actual refrigerator
power requirement divided by the Carnot (minimum) power requirement





TN+, T H



Ti = T c

\ \ \

FIG.24. Schematic of striictural support cooled at a number of discrete points. (After

Hilal and Boom [70].)

Hilal and Boom [70]. The optimum node temperatures T 2 , ..., TNfl are
given by the solution to the following set of equations:
i = 2 ,..., N + 1
( 1 / D f F l ) a 7 ; + ( l / D : )-jf = 0 ,
The optimum node-to-node spacing is then calculated from
. v i1

AxJL = Dii2,/E




The minimum refrigerator power requirement corresponding to design

Eqs. (5.14) and (5.15) is






Hilal and Boom [70] applied this optimization procedure to various designs of superconducting systems. For example, for a 10,000 MW hr energy storage solenoid supported by a Narmco epoxy -fiberglass structure
with two intermediate heat exchangers (nodes) the optimum node temperatures are T2 = 1I . I K , T3 = 70.3 K, while the optimum spacing between
cooling stations is AxJL = 0.305, Ax2/L = 0.325, Ax.JL = 0.370.
For the minimum refrigerator power requirement the authors found
WinLIA = 94.5 Wlcm, a value used later in the cost analysis of the proposed system.
It is worth pointing out that the choice of calculating and minimizing the

total refrigerator power. as d o n e in Eq. (5.1 I ) , is analogous t o minimizing

the total entropy generated i n t he space residing below room temperature.
As in the derivation which led to the statement of the ~ ; ~ i / ~ \ ~ - . S ~ ~ i ( l o l ( i
t i i ~ ~ o r o t t(i3 . 7 ) , it is n o t difficult to prove that thr r q f k i , y c t ~ i / ( i ipoit'"r
tic~rtlrclt o tiiuintciiii ( i c ~ t i l t l, \ p i ( (' i ~ o l d(heloir* e n \ ~ i r t i t i t ? i v t ~t ct~ t i i p o r t r t i r r )r
ryirci1.v t l i c (.old . s p t i c - c cirtrop\' g(,tii,rotioii rate times ~tiot i t ~ . c o l r i t ct c~r i i p r r otidrc' 1 ) t t / i c t ~ i i \ - i r o t i t ? i o iI iO~ i ~ . l i r cI1 l h t , rc;fiig[>rotorriyjects i i v t i t . T h e cold
space entropy generation rate must include the entropy generation associated with the leakage of heat from environment into the cold space.
Before leaving the subject ot' mechanical supports for low -temperature
systems. it should be recognized that due to manufacturing constraints
the single-stream continuous cooling method of Fig. 23 can be implemented less expensively in a rn;inner similar t o the method described by
Hilal and Boom (Fig. 24). Specitically. the cold stream t i i ( ' , , i \ forced to
make thermal contact with the conducting support discretely. :it a number
of points (cooling stations). T h e optimum spacing of cooling stations is
found based o n an optimization procedure involving the use of 1,agrange
multipliers (721. At present. the design and construction of mechanical
support\ for rotating superconducting windings are based o n the discontinuous version of the method of b$g. 23. T h e most recent support system
of this hind was designed, constructed. and tested by Tepper t r l . 1731.

2 . Mirltrplc. ,

Optiiili I





K t 1 rlitiI/( l t I


It is known from elementary heat transfer that the effective conductivity of a stack of parallel radiating shields is proportional to the local
shield absolute temperature c u b e d . I n Fig. 22 this class of insulations is
approximately accounted for b y the 1 1
3 curve. Therefore. in cryogenic
radiation shielding more than anywhere else, t h e minimization of irreversibility according t o rule ( 5 . 4 ) i s likely to yield important savings in available work.
Imagine an insulation consisting of (,V - 1) radiation shields suspended
in the evacuated space separating 'I.,and T 2 . If the number of shields is
very large. the insulation may be regarded a s continuous and the variational result (5.4) applies unchanged. I f the number of shields i s small so
that the temperature varies :rppreciably between two adjacent shields.
the variational result does n o t apply and must be replaced with a similar
principle valid f o r discontinuous insulations. This derivation i s presented
in detail in Bejan [67]. T h e insul:ition irreversibility reaches i t s minimum
when the ( N
1 ) shield temper-attires are externally controlled such that

d J / ( t 7 ' ' ' = 0.

I =

I. 2.

.... N





2 vAF



[(l/T(i-l)) - ~ / p ]


i= 1

In these expressions, v is Boltzmanns constant, A is the shield area, and

F is an effective view factor accounting for the emissivities of the two surfaces. In combination with Eq. (5.18), system (5.17) is sufficient for determining the optimum set of shield temperatures T i ) which
yields the minimum rate of entropy generation S (note that N and A remain fixed). This
statement replaces the variational result (5.4) presented earlier in this section for t and A fixed. As one might have suspected, the number of shields
plays the role of insulation thickness (or heat flow path length). The subject of optimum radiation shield temperatures is covered extensively in
the cryogenic heat transfer and thermal design literature (see, e.g.,
It is necessary to note at this point that the second-law aspects of
thermal radiation constitute an important topic in fundamental heat
transfer research today. Petela [78] showed for the first time that the exergy (available work) carried by thermal radiation is


+ i(T0/T)4 - 4To/3T]e


where e , T, and To are the radiation intensity, the emitter temperature,

and the reference (environment) temperature, respectively. The same result was reported later by Press [79]. In connection with the maximum
available work which could, theoretically, be harvested from solar racliation, Parrott [80] showed that the solar exergy is


S(T0/T5)4- (4T0/3T5)( 1

cos 8)14]e


In this expression, Ts is the equivalent sun temperature (5800 K) and IS is

the half-angle of the cone subtended by the suns disk (0.0047 rad). Result
(5.20) is of pivotal importance in the field of solar power engineering,
where it should serve as a basis for determining the efficiency of solar
power installations (see, e.g., Parrott [80] and Kreider [81]).
3. Counterflow H E U Exchangers
Fucing a Lurgr End-to-End
Temprrat ure RNt io
We return now to the subject of regenerative counterflow heat exchangers considered in detail in Section IV. This time we base the discussion on the balanced-flow heat exchanger schematic of Fig. 25. A straightforward entropy generation calculation in the dashed element may be





FIG 2 5 Counterflou heat exchangci

end direction [h2]


thcrmtl insulation system in the urmiend-cold

used to \how that the quantity


= !)I(,,



plays the role of convective heat transport in the T2+ 7 ,direction, in

place of q in t h e general insulation system of Fig. 21b. In addition. one can
prove that 4 is proportional to the local temperature gradient [Q]. as in
Eq. (5.1).
(1 = [ ( r r r c ,,Y/Up] dT/dx


which demonstrates that the group ( r ~ c , , ) ~ / ( U

is pthe
) equivalent of ( L A )
and that the length of the heat exchanger, L , plays t h e role of insulation
thickness. t . The longitudinal heat transport is inversely proportional to
the heat transfer area Pr [i.e..the N,,, = U p t / ( m c p ) ]and
, so is the heat exchanger irreversibility. Counterflow heat exchangers are good insulations
in the hot end-cold end direction by promoting effective heat transfer in
the stream-to-stream direction.



The general conclusions reached in the preceding section apply also to

counterflow heat exchangers. This type of insulation is improved first by
increasing the heat transfer area Pt. When this approach is no longer feasible economically, intermediate cooling can further improve the thermodynamic performance. In many applications, the effective conductance
(mc,)2/Uptis relatively insensitive to temperature changes; hence, the effective conductivity li is constant. Consequently, the entropy generation
minimization principle (5.4)recommends a uniform intermediate cooling
effect so that q becomes proportional to the absolute temperature or, according to Eq. (5.21).




This conclusion is in very good agreement with the engineering practice

of cooling at intermediate temperatures the main counterflow heat exchanger of a helium liquefaction plant or refrigerator [82, 831. In Fig. 26,

-300 K


I +



FIG. 26. Intermediate cooling of the main counterflow heat exchanger of a helium refrigerator.



on the left side, we show schematically the layout of a helium refrigerator

with three expansion engines distributed along the main heat exchanger.
On the right side of Fig. 96 we decomposed the 5 K-300 K section of the
refrigerator: it is evident that the "minirefrigerator" associated with each
expansion engine serves the function of cooling the main counterflow heat
exchanger. which carries heat by convection from room temperature to
the cold space.
In conclusion, we reviewed a number of engineering applications in
which the concept of irreversibility minimization is a central part of the
heat transfer design philosophv. The thermodynamic optimization of a
thermal insulation system does not end with minimizing the heat transfer
rate by decreasing the effective thermal conductance. At least as important to an efficient operation is the external control of the heat flow distribution 4 ( 7 ' ) . In the applications assembled in this section, the external
control of 4 ( T )was effected via "intermediate cooling," as in Figs. 23. 24,
and 26. Intermediate cooling. a s an energy-saving technique, is standard
in the emerging technology of large-scale superconducting systems. The
instances in which it is applied are considerably more numerous than the
examples presented here: for example, Agsten [84] applied this technique
to the least irreversible design of cryogenic current cables for superconducting magnets.
V1. Concluding Remarks

This review article was devoted to the introduction of second-law analysis in heat transfer, and entropy generation minimization in thermal design. The presentation proceeded from the derivation of the G o u y Stotiolli rhrorrm ( 2 . 7 ) ,the basi\ for entropy generation minimization in
the conceptual design of heat transfer equipment. Appropriate analytical
tools, such as the entropy generation number N , in Section 111, were devised for the task of estimating the destruction of available work in the
processes involving heat transfer. However, the entropy generation
number concept is considerably more general since it can be used to quantitatively describe the degree o f irreversibility of engineering components
and processes which do not draw their irreversibility solely from heat
The examples considered in this article ranged from the irreversibility
associated with some of the most fundamental convective heat transfer
processes, to the minimum irreversibility design of one-dimensional insulations such as the main counterflow heat exchanger of a helium liquefaction plant. This article summarized only the beginning of what can be



done, case by case, to revise today's caloric-biased heat transfer and

thermal design methodology, from the new exergy conservation viewpoint offered by the second law of thermodynamics. To extend this revision process to other pivotal examples from heat transfer is a timely contribution which should be camed out, not as part of this review article,
but as solutions to individual problems in the day-to-day thermal engineering practice.
An important area, intentionally left out of the defined objective of this
article, is the union of irreversibility analysis with cost analysis, with the
ultimate goal of producing economic optimum designs of required thermlal
systems [85-891. This area should be investigated in future research. As
pointed out in the Introduction, thermodynamic and economic optima do
not generally coincide, although, as available work (exergy) becomes an
increasingly expensive commodity, the two optima more closely parallel
one another.
In heat transfer, the earliest attempt to cast the irreversibility production figure of heat exchangers in economic terms is due to McClintock
[ 5 13. He recognized that irreversibility minimization alone is not a sufficient basis for preferring one design over another. To account for this difficulty, McClintock introduced the notions of cost per unit of available
work lost due to heat transfer across a finite temperature difference, C h T ,
and cost per unit of available work lost due to frictional pressure drop,
. terms of our entropy generation expressions for a general heat exchanger duct, Eq. (3.17). one can calculate in dollars the total cost associated with process irreversibility:
C,y* = C,S;p



McClintock argued that in the heat exchanger for a well-defined application, the unit costs CAPand CAT are known (in general, CM # CAT). Consequently, one can base the heat exchanger design optimization procedure on minimizing Eq. (6.l), the cost attributed directly to thermodynamic inefficiency. With CAPand CAT fixed, this optimization procedure
becomes analytically identical to the entropy generation minimization
procedure illustrated at various points in this monograph.
More recently, Fadden [90] proposed a similar procedure for taking into
account the cost of available work destroyed by fluid friction in piping and
valves. He addressed the simpler class of ducts in which the irreversibility
is due entirely to the AP term appearing in Eq. (3.17). In lieu of cost formula (6. l ) , Fadden proposes the use of an entropy increase cost (EIC) per
unit length of piping,





where C,,,is the current cost of electric power and the product T,S Ah represents the available power lost due to fluid friction. Conceptually,
Fadden's proposal is an extreme case of the more general cost formula visualized by McClintock. the extreme case where the cost of heat transfer
irreversibility is negligible compared with the cost of fluid friction irreversibility.
The above schemes for irreversibility cost calculation and minimization
stress one major difference between entropy generation analysis and cost
analysis. As pointed out by Urdaneta and Schmidt [86], the entropy generation analysis is irz\wriuJrt with time and place for a given process. The
invariant, fundamental character of second-law analysis stresses t h e importance of the material assembled in the present monograph. This material serves as starting point for thermal design engineers engaged not only
in irreversibility minimization. but also in cost minimization. The real
need for a reference work on the emerging subject of second-law analysis
in heat transfer was the motivation behind the writing of thi\ review.

NOhll N( I A I URE








coefficients. Eq (4.16)
parameter. Eq. 13.31). also
heat exchanger area. Eq.
availability. energy. h - /(,s
coefficients. Eq. (4.17)
specific heat at constant pre\sure
capacity rate. m ,
C'arnot efficiency of refrigerator
unit cost of irreversibility
due t o fnction
unit cost of irrevercibility due
to heat transfer
unit CQSt of lost available work
hydraulic diameter
function, E q . ( 6 . 1 2 )
Eckert number, E q . (3.5)
friction factor
dimensionlehs mass velocity.
Eq. (4.18)
mass velocity
specific enthalpy
average heat transfer coefficient

conductivity integral. Eq.

duty parameter. Eq. 13.28)
thermal conductivity
length of heat exchanger Row
mass flow rate
entropy generation number
entropy generation number
attributed to heat transfer
number of heat transfer units
wetted perimeter
pressure drop
Pkclet number. Eq. ( 3 . 1 2 )
Prandtl number. I ' / U
heat Row function. Eq 15.1)
heat transfer interaction
local heat transfer rate.
(W/m). ( W i m ' )
radial position
tube radius
dimensionles\ radial position
r / r o , also ideal gas constant
Reynolds number, Eq. 13.21)

specific entropy

similarity variable, Eqs. (3.14)

and (3.15)
temperature difference, Eq.
characteristic temperature
temperature difference, Eq.
dimensionless time, Eq. (4.22)
kinematic viscosity
absolute temperature ratio,

rate of entropy generation, S,,,

entropy generation due to fluid
entropy generation due to heat
Stanton number. t q . (3.22)
time, Eq. (4.22);alco insulation
thickness, Eq. (5.3)
absolute temperature
characteristic temperature
reference temperature
temperature difference
high temperature
low temperature
overall heat transfer coefficient, Eq. ( 4 . 2 3 )
velocity components
characteristic velocity
lost available work
maximum available work
axial position
transversal position. also heat
exchanger parameter, Eq.
thermal diffusivity
factor, Eq. (5.9)


temperature difference number, Eq. (3.7)

irreversibility ratio S,/SaT,
Eq. (3.39)
viscous dissipation function,
Eq. (3.2)


heat transfer. augmented


This article is partially based on research supported by Contract No. N00014-79-C-0006
from the Office of Naval Rewarch, the Power Program.
I . M. Gouy, Sur I'energie utilisable. J . Phys. (Paris) 8, 501 (1889).
2. A. Stodola, "Steam and Gas Turbines." McGraw-Hill, New York, 1910.
3. J . H . Keenan, "Thermodynamics," Chapter XVII. MIT Press, Cambridge, Massachusetts. 1941.
4. G. Darrieus, The rational definition of steam turbine efficiencies. E~iginrc,ring(London).
p. 283 (1930).
5. R. W . Haywood, A critical review of the theorems of thermodynamic availability, with
concise formulationb. Part 2 . Irreversibility. J . M w h . E n g . Sci. 16, 258 (1974).
6. R. A . Gaggioli, The concept of available energy. Chrni. G i g . Sci. 16, 87 (1961).
7. C. A. Berg, A technical basis for energy conservation. Mcwh. E u g . 96, 30 (1974).
8. R. H . Socolow et ( I / . , Etficient use of energy. Phys. Today 28, 23 (1975).



9. M. I ribus and R . Evans, The thermo-economics of sea-water conver\ion ( ' o t i f r i h . U i i i i . C ' ( d / ; f ' , , W " i f c r / ? ~ \ o i t r .('cur 61, N o . 62-53 (1963).
10. J . Kestin. "Available Work i n Geothermal Energy,'' Rep. No. C A I MEC/!O. Div.
Eng.. Brown Univer\ity. Providence, Rhode Island, 1978.
I I . J . Kestin, R. DiPippo. and H . 1- Khalifa. Hybrid geothermal-fossil power plant\. .I.Ic(.h.
t'/rif. 100, 28 (1978).
12. R. .4. Gaggioli, W J . Wepfer. and A F. Plkouh. Availableenergy analy4isfor HVAC. I .
Inefficiencies in ii dual-duct \v\teni Winter Ann. Meet. ASMF.. ASME Publ.. San
Francisco. California ( I97H).
13. W. Leidenfrost. The use of heat pumps in reducing fuel consumption for nonsular climale control of building\. E/ic,r,kr\ 3, 83 (1978).
14. W . Leidenfrost. K . H . Lee. and H. Korenic. Conservation of energ) by \econd law
anal) sis of a powei- conwrving p r t ~ e s s/..ur,ryy
5 , 47 (1980).
IS. J . Kestin. "A Course in ~fherniodynanlics." revised printing. Vol 1. C'hapter 9.
McGraw-Hill. New Yorh. 1979.
16. J . K e h n . A simple unitied approiich to the first and second law5 of thermodynamics.
Pure A p p I . C'herii. 22, S l I (1970).
17. P. C . IN. A belated revolution in teaching engineering thermodynamics !-Second
according to Kestin. N6rli A/i,iic
O I I ) A I P I . So(.. E H X .Edrrc . . /Y7X
18. E . G Cravalho and J . 1.. Smith. Jr.. "'l'hermodynamics (An Introduction)." issued
since I969 as Clas\ Note\ in the Department of Mechanical Engineering. Maszachusetts
Institute of Technology. Cambridge. MA
19. W . P Stoecker. "l>e\ign of Thermal Sy\tenis." McGraw-Hill. New York, 1971.
20. W . C . Reynolds and H . C . Perkin\. "hngineering Thermodynamic\." Chapter 7.
McGraw-Hill. New York. 1977
21. G . J . Van Wylen and R F Sonntag. "Fundamentals of Classical 'Therniodynami
Chapter 8. Wiley. New York, 1973
22. J . Kestin. "A Cour\e in 'Thrrmodyiianiics." revised printing, Vol. I . Chapter 13.
McGraw-Hill. New York. 1979.
23. V. Radcenco. "Criteria for Optimiting Irreversible Thermal Proce\se\. I-Aitura 'lehnica. Bucharest, 1977 (in Romanian).
24. J . H . Keenan, Availability and irrever\ihility in thermodynamics. H r ./ 4 1 4 P / ~ \ , Y2,.

183 IIYSI).
25. 2 . Rant. Exergie em n e w \ Wort fur "technische Arbeitsfahigkeit." / . , l r \ c h I r r ~ t ~ t i ic,itru c \ . 22. 36 ( 1 9 % ) .
26. R. Clausius, "Die mechanische Warnietheorie." Braunschweig. 1887.
27. F. Bo\njacovif. ' T e c h n i \ c h e I'hermody namik," Vol. I . Steinkopf. Dre\den and
Leipzig, 1935.
28. G . Lkirrieub. L't5volution de\ ceiitr;ile\ thermiques et la notion d'energie utili\able. .S(.i.
1/10.(Ptrri,) 15, 2 M I 193 I ).
29. J . D Jackson, Fint year therniodk n;imic\ notes. University of Manche\ter. Simon
Engineering Laboratone\. 1978 (pi-ivatecommunication).
30. A. I.. London, Clas\ note\ for c'our\e M E 270A. "Engineering 'I'herniodynamics."
Stanford University. Department 0 1 Mechanical Engineering. I977 (private communication).
31. S . K. DeGroot and P. Marur. '.Noii-tquilibrium Thermodynamic\.' Wile) (Interscience). New Yorh. 1967.
32. I . Prigogine. "Introduction to the 'I hei-modynamics of Irreversible Proces\e\." Wiley
(Inter\cience), Nem k'orh, 1967
33. J . K e d i n . "A Courw in ~ I ' h e r i i i ~ i d ) i i ~ i n i i crevised
printing. Vol. 11. McGraw-Hill,
New York. 1979.



34. A . Bejan, A study of entropy generation in fundamental convective heat transfer. J .

Hear Transfer 101, 718 (1979).
35. H. Schlichting, Boundary Layer Theory, 6th ed. McGraw-Hill, New York, 1968.
36. J . Kestin and P. D. Richardson. heat transfer across turbulent, incompressible boundary
layers. l i i r . J. Heut Mu.\.\ litrrisfer 6 , 147 (1963).
37. R. B. Bird, W. E. Stewart. and E . N. Lightfoot, Transport Phenomena. Wiley, New
York. 1960.
38. E. Z. Pohlhausen, Der Warmeaustausche zwischen festen Kopern and Fliissigkeiten
mit kleiner Reibung und kleiner Warmeleitung. Z. Angew. M a t h . M r c h . 1, 115 (1921).
39. L. Howarth, On the solution of the laminar boundary layer equations. Proc. R. Sot,.
London. Ser. A 164, 547 ( 1938).
40. A. Bejan, General criterion for rating heat exchanger performance. l n r . J. Heor M a s s
Trunsfer 21, 655 (1978).
41. W. M. Rohsenow and H. Y . Choi, Heat, Mass and Momentum Transfer. PrenticeHall, Englewood Cliffs. New Jersey, 1961.
42. A Bejan, The concept of irreversibility in heat exchanger design: Counterfiow heat exchangers for gas-to-gas applications. J . Hear Transfer 99, 374 (1977).
43. A. E. Bergles and R. L. Webb, Bibliography on augmentation of convective heat and
mass transfer. Prc,i.icw..c ffeuf Mass Transfer 4, 61 (1978).
44. A. E. Bergles, Survey and evaluation of techniques to augment convective heat and
mass transfer. Prog. Haur Muss Transfer 1, 331 (i969).
45. A. E. Bergles, Recent developments in convective heat transfer augmentation. Appl.
MtJch. Rev. 26, 675 (1973).
46. R. L. Webb and E. R. G. Eckert, Application of rough surfaces to heat exchanger design. f n r . J . Hear M a s s 7rurrsfer 15, 1647 (1972).
47. A. E. Bergles, A. R. Blumenkrantz, and J. Taborek, Performance evaluation criteria for
enhanced heat transfer surfaces. Heut Transfer, Proc. lnr. Heat Transfer Conf., 5 t h ,
1974 Vol. 2 , p. 1 I 239 (1974).
48. W. M. Kays and A . L. London, Compact Heat Exchangers. McGraw-Hill, New
York, 1964.
49. E. D. Grimison, Correlation and utilization of new data on flow resistance and heat
transfer for cross-flow of gases over tube banks. Trans. ASME 59, 583 (1937).
50. S. L. Jameson, Tube spacing in finned-tube banks. Trans. ASME 67, 633 (1945).
51. F. A . McClintock. The design of heat exchangers for minimum irreversibility. ASME
Annu. Meer. Pap. No. SI-A-108(1951).
52. A . P. Fraas and M. N . Ozisik, Heat Exchanger Design. Wiley, New York, 1963.
53. J. Le Foll, Etudes experimentales sur le transport de chaleur par les gaz en convection
forcte. Houille Blaizclir 1, 34 (1957).
54. M. Tribus, Class notes for Thermoeconomics. Massachusetts Institute of Technology, 1978 (private communication).
55. P. J. Golem and T . A . Brzustowski, Second-Law analysis of energy processes. Pad 11.
The performance of simple heat exchangers. Trans. Can. SOC.Much. Eng. 4, 219
( 1976- 1 977).
56. G. M. Reistad. Availability: Concepts and applications. Ph.D. Thesis, University of
Wisconsin, Madison (1970).
57. G. E. Kelly and 3. E. Hill. Method of Testing for Rating Thermal Storage Devices
Based on Thermal Performance, Rep. NBS IR-74-634. U.S. Natl. Bur. Stand., Washington, D.C., 1975.
58. A. Bejan. Two thermodynamic optima in the design of sensible heat units for energy
storage. J. Heur Trtrit.s$,r 100, 708 (1978).



59. A. c'. Mueller, Heal exchanger\. Irr "Handbook of Heat Transfer." ( W . M . Rohsenow
and 1. P. Hartnett. ed\.). pp. 18-31. McCiraw-Hill, New York. 1973.
60. F. W . Schmidt. R . R. Somers. II.J . Srego. and D. H. Laananen. Design optimization of
a single fluid. solid sensihle heat \lorage unit. J. Heut 7rtinsfbr 99, 174 11977).
61. A . Bqjan and J. L. Smith, J r . . Thermodynamic optimization of mechanical supports for
cryogenic apparatus. ( ri,o,cc>rii(.\ 14. 158 (1974).
62. A. Be,ian. A general variational principle for thermal insulation 5yctem design. I n / . J.
Hrcrr Mci\v 7rari,\,tc,r 22, 219 (1979)
63. K . Rektorys. "Survey o f .Applicable Mathematics." MIT Press, Cambridge, Mascachusetts. 1969.
64. S. Foner and B. B. Schwartz. e l l \ . .'Superconducting Machines and Devices, Large
Systems Applications." Plenum, New York. 1974.
65. C . D. Henning, Superconductivit) tor mirror fusion. / E E E 7>trr7s. \!(ry!i MAC-15, 525
66. "Wisconsin Superconductive C'nergv Storage Report," Vol. I. Univer\ity o f Wisconsin. Madison. 1977.
67. S. H . Minnich o / t i / . . Design studie\ of aupei-conductinggenerators. / E . E t . I r i r r r c . .l.ltr,yri.
MAG-IS, 703 (1979).
68. R. B . Scott. "Cryogenic Engineering." Van Nostrand-Reinhold. Princeton. New
Jersek, 1959.
69. A . Bejan and J . L. Smith. J r . , Heat exchanger\ for vapor-cooled conducting supports of
cryostats. .Ad\.. C'r\.(rg.Eire. 21, 247 1 1975).
70. M. A. Hilal and R . W Boom. Optiiniration ot mechanical supports for large buperconductive magnets. A d i . ('rvo,y. Erip 22, 224 11977).
71. T. R . Strobridge and D.B. Chelton. Size and power requirements of4.2K refrigerators
Adi.. ( ' r \ , o g . E n g . 12, 576 (19671.
72. A. BeJan. Discrete cooling of low heat leak \upports t o 4.2 K. ('rJo,qctiit .\ IS, 2 9 0 (1975).
73. K . A . Tepper, J. L. Smith. Jr.. and F C ' . Rumore, Experimental simulation ofcryogenic
system for a large superconducting rolor. . 4 d i . C r y o g . E t i g . 23, I 1 8 (19771.
74. J . A. Piavanas. 0. P Roberts. and I). I . Wang. Multishielding-an advanced wperinsulation technique. Adi' C'rni,g.
10. 197 (1965).
75. A. Cavallini, E. Fornasieri. and R. Zecchin. 'Thermodynamic optimization o f thermal insulation in cryogenich. Pre\ented at the XlVth Congr. Int. Inst. Refrig.. Moscow (1976).
76. M. A . Hilal and G . E . Mclntosh. ('ryogenic design for large superconductive energy
storage magnets. A d . . (.rio,c. Eri,<,. 21, 69 ( 1975).
77. Y. M . Eyssa and 0. Okasha. 7'herniod)naniic optimization of thermal radiation shields
for a cryogenic apparatus. ('rvtjycriil,\ IS, 305 (1978).
78. R. Petela. Exergy of heat radiation. . I . Iletri 7rtrnsfi.r 86, 187 ( 1964).
79. W. Press. 'Theoretical maximum tor energy from direct and diffuse radiation. R'crritre
( L , O t i d < J ! I ) 264, 734 (IY76)
80. J . E . Parrot. Theoretical upper limit to the conversion efficiency of w l a r energq. So/.
Erierpv 21, 227 (197X).
81. J. F . Kreider. Second-law analysis of solar-thermal processes. Erit.r,&rvH c ~ .3. 325
82. J . E . Ahern. "The Exergy Method (11 t:nec-gy Systems Analysis.'' Wiley. N e w York.
83. R. H. Kropschot, B . W. Birmingham. and I). B . Mann, eds.. "Technology of Liquid
Helium,'' N.B.S. Monogr. No. I I I . 0 . S . Nail. Bur. Stand.. Washington. D.C.. 1968.
84. R. Agsten. Thermodynamic optimization of current leads into low temperature regions.
Crvogcriii,.\ 13, 141 (19731.



85. R. Gaggioli, L. Rodriguez. and W. Wepfer, "A Thermodynamic-economic Analysis of

the Synthane Process." Final rep. Marquette University (1978).

86. A. Urdaneta and P. S . Schmidt, Evaluation of energy utilization analysis methods for
industrial processes. / t i "Energy Use Management" (R. Fazzolare and C. Smith, eds.),
Presented at the XlVth Congr. Int. Inst. Refrig., Moscow (1976). Pergamon, Oxford,
87. Y. M. El-Sayed and R . B. Evans, Thermo-economics and the design of heat systems. J .
E n g . Po\ 92, 27 (1970).
88. Y. M .El-Sayed and A . J . Aplenc, Application of the thermo-economic approach to the
analysis and optimization of a vapor compression desalting system../. Eng. P m t w 92, 17
89. H. Fruth, Cost determination and comparison of nuclear and fossil fueled dual purpose
power and desalination plants. Proc. f n t . S y m p . Wutrr Uosolin.. Ist, IY65. Presented at
the XIVth Congr. I n t . Inst. Refrig., Moscow (1976).
90. W. Fadden, A way to pinpoint energy loss. Consulr. Eng. 18, 106 (1962).
91. A. Bejan, D. W. Kearney . and F. Kreith, Second law analysis and synthesis of solar collector system. A S M E J . .So/wEnergv Eng. 103, 23 (1981).
92. A. Bejan, On the buckling property of inviscid jets and the origin of turbulence, Lett.
Hemt Mrrs.s 7runsji.r 8, I87 (1981).
93. A . Bejan, Comments on "Viscous buckling of thin fluid layers," P h y s . N u i d s 24(9),
1764 (1981).

Note Added in Proof

The author would like to use this opportunity to bring to the reader's attention a number of
developments which have taken place since the completion of this article.
In the area ofsolar r n c ~ rciiKineering.
attention is drawn to a study of the second-law aspects of solar collector operation [91]. In this study, solar collectors are for the first time analyzed as installations for the efficient harvesting and delivery of exergy (not energy). It is
shown that the fraction of exrrgy delivered by solar collector systems is adversely affected
by heat transfer irreversibilities occurring between the sun and the collector, between the
collector and the ambient air. and inside the collector. The study uses as working examples
an isothermal collector. a stream-cooled (nonisothermal) collector, and the design of the
collector-user heat exchanger. to show how solar collector systems can be designed for
maximum exergy delivery.
In the area of,crrch, attention is drawn to the discovery of the buckling
property of inviscid fluid fibers of finite thickness (jets, wakes, shear layers) which serves
as origin for the turbulent motion of fluids [92]. This property follows from the fact that
inviscid fluid layers cannot generate entropy [Eq. (3. I)], hence, they possess electric properties analogous to those of slender beams in longitudinal compression [93]. It has been shown
that the natural behavior of any inviscid fiber is to buckle and to execute a "meandering"
motion in which the wavelength is a precise multiple of the fiber thickness. The buckling
property accounts for observed features of turbulent flow and, most important, it proviides
for the first time a theoretical explanation for the transition from laminar to turbulent flow.