Proc. R. Soc. A (2011) 467, 10121028 doi:10.1098/rspa.2010.

0293 Published online 13 October 2010

An alternative criterion in heat transfer optimization

BY QUN CHEN1,3, *, HONGYE ZHU2 , NING PAN3
1 Key

AND

ZENG-YUAN GUO1

Laboratory for Thermal Science and Power Engineering of Ministry of Education, Department of Engineering Mechanics, and 2 Institute of Nuclear and New Energy Technology, Tsinghua University, Beijing 100084, Peoples Republic of China 3 Biological and Agricultural Engineering Department, University of California, Davis, CA 95616, USA
Entropy generation is recognized as a common measurement of the irreversibility in diverse processes, and entropy generation minimization has thus been used as the criterion for optimizing various heat transfer cases. To examine the validity of such entropy-based irreversibility measurement and its use as the optimization criterion in heat transfer, both the conserved and non-conservative quantities during a heat transfer process are analysed. A couple of irreversibility measurements, including the newly dened concept entransy, in heat transfer process are discussed according to different objectives. It is demonstrated that although thermal energy is conserved, the accompanied system entransy and entropy in heat transfer process are non-conserved quantities. When the objective of a heat transfer is for heating or cooling, the irreversibility should be measured by the entransy dissipation, whereas for heat-work conversion, the irreversibility should be described by the entropy generation. Next, in Fouriers Law derivation using the principle of minimum entropy production, the thermal conductivity turns out to be inversely proportional to the square of temperature. Whereas, by using the minimum entransy dissipation principle, Fouriers Law with a constant thermal conductivity as expected is derived, suggesting that the entransy dissipation is a preferable irreversibility measurement for heat transfer.
Keywords: heat transfer; irreversibility; optimization; entransy dissipation; entropy generation

1. Introduction
Heat transfer occurs in about 80 per cent of all the energy utilization systems, so improving the heat transfer performance signicantly promotes the energy conservation in most thermal systems, through either increasing the heat ow rate for a given volume of facility, or reducing the cost of equipment with given heat load (Webb & Bergies 1983; Bergles 1997, 1988). Besides, enhancing the
*Author for correspondence (chenqun@tsinghua.edu.cn).
Received 8 June 2010 Accepted 14 September 2010

1012

This journal is 2011 The Royal Society

Criterion in heat transfer optimization

1013

heat transfer effectively raises the operation reliability of electronic devices where electricity-generated heat has frequently posed a serious problem (Chen 1996). Thus, heat transfer improvement (or optimization) has become one of the critical issues for the efciency of energy utilization, and heat transfer research has propelled rapid advancement in various heat transfer enhancement technologies, including using extended surfaces, stirrers and external electric or magnetic elds (Bergles 1988, 1997; Webb 1994; Karcz et al. 2005; Schfer et al. 2005). In the theoretical study of heat transfer, based on the concepts of entropy and entropy generation, Gyarmati (1970) derived Fouriers Law with the criterion of minimum entropy generation and showed that entropy generation is the irreversibility measurement for any heat transfer process. Bejan (1979, 1996) and Charach & Rubinstein (1989) used this criterion for the optimization of convective heat transfer process and heat exchangers for different applications. On the other hand, there are some scholars who questioned if the entropy generation is the universal irreversibility measurement for heat transfer or if the minimum entropy generation is the general optimization criterion for all heat transfer processes, regardless of the nature of the applications. For instance, Bertola & Cafaro (2008) found that when satisfying the Onsager reciprocal relation, the principle of minimum entropy production (Prigogine 1967) could be tenable only if there is zero generalized ow under a non-zero generalized force, or the thermal conductivity should be inversely proportional to the square of the absolute temperature during steady-state heat conduction. By analysing the relationship between the efciency and the entropy generation in 18 heat exchangers with different structures, Shah & Skiepko (2004) demonstrated that even when the system entropy generation reaches the extremum, the efciency of the heat exchangers can be at either the maximum or the minimum, or anything in between. In addition, the so-called entropy generation paradox (Bejan 1996; Hesselgreaves 2000) exists when the entropy generation minimization is used as the optimization criterion for counter-ow heat exchanger. That is, enlarging the heat exchange area from zero simultaneously increases the heat transfer rate and improves the heat exchanger efciency, but does not reduce the entropy generation rate monotonouslythe entropy generation rate increases at rst and then decreases. Therefore, it was speculated that the optimization criterion of minimum entropy generation is not always consistent with the heat transfer improvement. Recently, Guo et al. (2007) introduced the concepts of entransy and entransy dissipation to measure, respectively, the heat transfer capacity of a system and the loss of such capacity during the process. Moreover, Guo et al. (2007) proposed the entransy dissipation extremum as an alternative optimization criterion for a heat transfer process not involved in a thermodynamic cycle, and, consequently, developed the extremum principle of entransy dissipation to optimize the processes in heat conduction (Guo et al. 2007; Chen et al. 2009a ), heat convection (Meng et al. 2005; Chen et al. 2007) and thermal radiation (Wu & Liang 2008). The contribution of this present paper is to further examine the physical implications of both entropy generation and entransy dissipation, compare their differences and, more importantly, examine their applicability to heat transfer optimization in applications of different nature.
Proc. R. Soc. A (2011)

1014

Q. Chen et al.

2. Irreversibility of heat transfer (a ) Conserved and non-conserved quantities in a transport process

After intensive study, we found that all transport processes contain two different types of physical quantities owing to the existing irreversibility, i.e. the conserved ones and the non-conserved ones, and the loss or dissipation in the nonconserved quantities can then be used as the measurements of the irreversibility in the transport process. Taking an electric system as an example, although both the electric charge and the total energy are conserved during electric conduction, the electric energy, however, is not conserved and it is partly dissipated into the thermal energy owing to the existence of the electrical resistance. Consequently, the electrical energy dissipation rate is often regarded as the irreversibility measurement in the electric conduction process. Similarly, for a viscous uid ow, both the mass and the momentum of the uid, transported during the uid ow, are conserved, whereas the mechanical energy, including both the potential and kinetic energies, of the uid is turned into the thermal energy owing to the viscous dissipation. As a result, the mechanical energy dissipation is a common measure of irreversibility in a uid ow process. The above two examples show that the mass, or the electric quantity, is conserved during the transport processes, while some form of the energy associated with them is not. This loss or dissipation of the energy can be used as the measurement of irreversibility in these transport processes. However, an irreversible heat transfer process seems to have its own particularity, for the thermo-energy always remains constant during transfer and it does not appear to be readily clear what the non-conserved quantity is in a heat transfer process. Non-equilibrium thermodynamics (Gyarmati 1970; Kreuzer 1981) seems to offer an answer by suggesting that the entropy or available energy (exergy) is the non-conserved quantity in a heat transfer process; that is, entropy can be generated or exergy can be dissipated during the process. Also it is interesting that, in general, all other physical energies (mechanical, electric, acoustic, etc.) can turn into thermal energy as the nal form, except we rarely, if ever, use entropy or exergy to measure the irreversibility in other physical processes besides heat.

(b ) Irreversibility measurements in heat transfer

Before discussing the irreversibility in a heat transfer process, it is necessary to distinguish the objectives of heat transfer, for, as shown below, the irreversibility may have different implications for different objectives. The objectives in heat transfer can be classied into two categories: one is to use thermal energy as a form of energy to perform work, and the other is directly using thermal energy for warming up or cooling down the temperature. When performing work, i.e. in heatwork conversion, the heat transfer process is a link in a thermodynamic cycle. Whereas for heating or cooling, the heat transfer process is apparently much simpler. We will demonstrate that it is necessary to use different concepts and quantities to describe the irreversibilities in two such distinctive processes. Based on the analogy between heat conduction and electric conduction, Guo et al. (2007) introduced a physical quantity, termed entransy, to study a heat
Proc. R. Soc. A (2011)

Criterion in heat transfer optimization

1015

transfer process not involved in heatwork conversion. The denition of entransy G is 1 G = UT , (2.1) 2 where T is the temperature and U the internal energy of the system. The expression of entransy is analogous to that of electric energy Ee in a capacitor, 1 Ee = Qe V , (2.2) 2 where Qe and V are the electric quantity and the electric potential, respectively. As shown in equations (2.1) and (2.2), the entransy G in the heat transfer should be viewed as a special form of energy with the dimension (JK). Further, accompanying the electric charge, the electric energy is transported during electric conduction. Similarly, along with the heat, the entransy is transported during heat transfer too. For example, in a heat conduction process, the thermal energy conservation equation can be expressed as vT , = V q +Q (2.3) vt where r is the density, cv the constant-volume specic heat, t the time, q the heat the internal heat source. Multiplying both sides of equation ow density and Q (2.3) by temperature T gives an equation that can be viewed as the conservation equation of the entransy in the heat conduction: rcv rcv T that is vT , = V (qT )+q VT + QT vt (2.4)

vg , (2.5) = V (qT ) fh + g vt where g = G /V = (1/2) uT is the specic entransy, V the volume, u the specic internal energy, qT the entransy ow density, g the entransy change owing to heat source and fh can be taken as the entransy dissipation rate per unit volume, expressed as VT . (2.6) fh = q

The left term in either equation (2.4) or (2.5) is the time variation of the entransy stored per unit volume, consisting of three items shown on the right: the rst represents the entransy transferred from one (or part of the) system to another (part), the second term can be considered as the local entransy dissipation during the heat conduction and the third is the entransy input from the internal heat source. It is clear from equation (2.5) that the entransy is dissipated when heat is transferred from high temperature to low temperature. Thus, heat transfer is irreversible from the viewpoint of entransy, and the dissipation of entransy can hence be used as a measurement of the irreversibility in heat transfer. Figure 1 is a schematic diagram of a case of one-dimension steady-state heat conduction. During the heat transfer, the entropy (or exergy) and entransy are transported with the heat. Among the system parameters, the thermal 2 , but neither the energy is conserved during the entire process, i.e. q 1 q entropy nor the entransy is conservedthe entropy is generated and the entransy
Proc. R. Soc. A (2011)

1016
q1

Q. Chen et al.
g1 s1 T1 T2 q2 T1 > T2, q1 = q2, g2 g1 > g2, s2 s1 < s2,

Figure 1. The schematic of one-dimension steady-state heat conduction.

is dissipated, i.e. s 1 < s 2 and g 1 > g 2 . Now we have two physical quantities for measuring the irreversibility in heat transfer, and we will demonstrate that, because of the intrinsic complex nature of heat transfer, we need both in dealing with the aforementioned different objectives in heat transfer. For heatwork conversion, the entropy generation or the exergy dissipation is a better irreversibility measurement, whereas for heating or cooling, the entransy dissipation is preferable. The physical distinctions between them will be further elucidated later in this article.

3. Irreversibility of heat transfer and Fouriers Law

In non-equilibrium thermodynamics, the thermodynamic force and thermodynamic ow for heat transfer are chosen to ensure that the scalar product of them equals the entropy generation rate Sgen = k |VT |2 /T 2 . Thus, based on the Onsager theory, the linear phenomenological heat transfer law can be generally expressed as 1 , (3.1) q = LV T where L is the phenomenological coefcient for heat transfer process and a constant unrelated to temperature. The phenomenological law in equation (3.1) is also termed the entropy picture of heat transfer by Gyarmati (1970), which is different from the Fourier picture of heat transfer, i.e. Fouriers Law: q = k VT (3.2) where k is the thermal conductivity. Comparison between equations (3.1) and (3.2) gives the relationship between the different phenomenological coefcients in the two representations L constant = (3.3) 2 T T2 That is, if the phenomenological law of heat transfer in non-equilibrium thermodynamics is to be consistent with Fouriers Law, the thermal conductivity k has to be proportional to 1/T 2 . According to the entropy picture of heat transfer, Prigogine (1967) proposed in 1947, the least energy dissipation principle, i.e. the minimum entropy generation principle, which is expressed for systems that satisfy the linear phenomenological L = kT 2 or k =
Proc. R. Soc. A (2011)

Criterion in heat transfer optimization

1017

law and the Onsager reciprocal relations as: during evolution towards a stationary state the entropy production decreases and takes its lowest value compatible with external constraints when this stationary state is reached (p. 85). From this minimum entropy generation principle, Fouriers Law was then derived. However, in the derivation, the phenomenological coefcient L in the entropy picture was treated as a constant, requiring the thermal conductivity be inversely proportional to the square of the absolute temperature during steady-state heat conduction. But in actuality, the thermal conductivity of most materials used in engineering is largely a constant, independent of temperature under normal conditions. This led to the conclusion by Prigogine (1967) himself that Fouriers Law derived using the minimum entropy production principle is not the most desirable. In contrast, according to the expression of entransy dissipation in equation (2.6), the temperature gradient is thought as the generalized force and the heat ux as the generalized ow for heat conduction, i.e. X = VT and J = q (3.4) (3.5)

Here, the scalar product of the generalized force and generalized ow equals to the entransy dissipation rate. Then based on the Onsager theory, the generalized force representation of the dissipation function j can be expressed as L (VT )2 , j= 1 2 (3.6) where the constant, L is the heat transfer phenomenological coefcient. Moreover, the linear phenomenological law of the generalized force and generalized ow is q = L VT . (3.7) Substituting equation (3.7) into equation (3.6) yields q VT . j = 1 2 Comparing equation (2.6) with equation (3.8) gives 2j fh = 0. (3.9) In order to search for the extremum of the dissipation function in equation (3.6) with the constraint from equation (3.9), a Lagrange function, F , is constructed as F = j + A(2j fh ). (3.10) (3.8)

where A is the Lagrange multiplier. The variation of F with respect to the general force, as shown in equation (3.4), is dF = (2A 1) vfh vj A dVT . vVT vVT (3.11)

For F to be stationary, the term within the square bracket must vanish, that is, (2A 1)
Proc. R. Soc. A (2011)

vj vfh A = 0. vVT vVT

(3.12)

1018

Q. Chen et al.

Meanwhile, the partial differential of the entransy dissipation and the dissipation function, as shown in equations (2.6) and (3.6), with respect to the temperature gradient VT are, respectively fh vfh = q = vVT VT and vj 2j = L VT = . vVT VT Substituting equations (3.13) and (3.14) into equation (3.12) yields (2A 1)2j Afh = 0. (3.15) (3.14) (3.13)

Equating equation (3.9) to equation (3.15), we obtain the Lagrange multiplier A = 1. Then equation (3.11) can be rewritten as dF = vfh vj vVT vVT dVT . (3.16)

This is the requirement for the extreme value of the force representation in the dissipation function and satises both the linear phenomenological law and the Onsager reciprocal relation. Because j is a positive-denite function and d2 F = d2 (j fh )J = L > 0, the extreme value determined by equation (3.16) is the minimum, i.e. (j fh )q = min, dq = 0, dVT = 0 (3.17)

Since the temperature gradient VT is arbitrary, the term in the parentheses should be equal to 0. = 0. (3.18) L VT + q Now when the linear phenomenological coefcient L* is chosen to equal the thermal conductivity k , equation (3.18) is exactly Fouriers Law for heat transfer. In other words, for a given heat ux distribution, of all the possible temperature distributions, the actual solution is the one that satises the minimum entransy dissipation. Similarly, from the general ux representation of the dissipation function, we can also prove that for a given temperature distribution, compared with all the other possible heat ux distributions, the actual solution satises the least action principle based on the entransy dissipation. Therefore, for an arbitrary boundary condition, heat will always ow along the path with the least action of the minimum entransy dissipationa truly natural law as expected. The aforementioned deduction process is similar to that based on nonequilibrium thermodynamics, and the differences between them lie in that the least action here refers to the entransy dissipation rate, while the least action used in non-equilibrium thermodynamics is in terms of the entropy generation rate. Furthermore, Fouriers Law derived here requires a constant thermal conductivity, whereas in non-equilibrium thermodynamics, the thermal conductivity has to be inversely proportional to the squared temperaturesomething deviating grossly from existing facts.
Proc. R. Soc. A (2011)

Criterion in heat transfer optimization

1019

4. Optimization of heat transfer with different objectives (a ) Entransy dissipation extremum principle
Heat transfer optimization aims for minimizing the temperature difference at a given heat transfer rate, d(DT ) = df (x , y , z , t, T , k , q , r, cp , ) = 0 or maximizing the heat transfer rate at a given temperature difference = dg (x , y , z , t, T , k , q , r, cp , ) = 0. dQ (4.2) (4.1)

In conventional heat transfer analysis, it is difcult to establish the relationship between the local temperature difference (or local heat transfer rate) and the other related physical variables over the entire heat transfer area, so the variational methods in equations (4.1) and (4.2) are not practically usable. However, in addition, to be the expression of least action in heat transfer, the entransy dissipation in equation (2.6) is also a function of the local heat ux and local temperature gradient in the heat transfer area, and thus the variational method will become usable if written in terms of the entransy dissipation (Cheng 2004). Integrating the conservation equation of the entransy equation (2.4) over the entire heat transfer area gives rcT
U

vT dV = vt

V (qT )dV +
U U

q VT dV +
U

dV . QT

(4.3)

For a steady-state heat conduction problem, the left term in equation (4.3) vanishes, i.e. 0=
U

V (qT )dV +
U

q VT dV +
U

dV . QT

(4.4)

If there is no internal heat source in the heat conduction domain, equation (4.4) is further reduced into q VT dV =
U U

V (qT )dV .

(4.5)

By transforming the volume integral to the surface integral on the domain boundary according to Gausss Law, the total entransy dissipation rate in the entire heat conduction domain is deduced as Fh =
U

q VT dV =
G

q G T dS =

G+

q in Tin dS

q out Tout dS ,

(4.6)

where G+ and G represent the boundaries of the heat ow input and output, respectively. t, The continuity of the total heat owing requires a constant total heat ow Q t = Q
Proc. R. Soc. A (2011)

G+

q in dS =

q out dS .

(4.7)

1020

Q. Chen et al.

We further dene the ratio of the total entransy dissipation and the total heat ow as the heat ux-weighted average temperature difference DT DT = Fh = t Q
G+

q in T dS t in Q

q out T dS . t out Q

(4.8)

For one-dimensional heat conduction, equation (4.8) is reduced into DT = (Tin Tout ), exactly the conventional temperature difference between the hot and cold ends. Using the heat ux-weighted average temperature difference dened in equation (4.8) and applying the divergence theorem, a new expression for optimization of a steady-state heat conduction at a given heat ow rate can be constructed as t d(DT ) = d k |VT |2 dV = 0. (4.9) Q
U

It shows that when the boundary heat ow rate is given, minimizing the entransy dissipation leads to the minimum in temperature difference, that is, the optimized heat transfer. Conversely, to maximize the heat ow at a given temperature difference, equation (4.9) can be rewritten as t = d DT dQ
U

1 2 |q | dV = 0, k

(4.10)

showing that maximizing the entransy dissipation leads to the maximum in boundary heat ow rate. Similarly, for a steady-state heat dissipating process with internal heat source in equation (4.4), the total entransy dissipation rate in the entire heat conduction domain is derived as Fh =
U

q VT dV =
U

dV QT

q out Tout dS .

(4.11)

The heat generated in the entire domain will be dissipated through the boundaries, i.e. t = Q
U

dV = Q

q out dS .

(4.12)

Again the heat ux-weighted average temperature is dened as the entransy dissipation over the heat ow rate DT = Fh = t Q Q T dV t Q q out T dS . t out Q (4.13)

And thus the optimization of the process is achieved when t d(DT ) = d Q

U

k |VT |2 dV = 0,

(4.14)

which means that in a heat dissipating process, minimizing the entransy dissipation leads to the minimum-averaged temperature over the entire domain. Based on the results from equations (4.9), (4.10) and (4.14), it can be concluded that the extremes in entransy dissipation lead to the optimized heat transfer performance at different boundary conditions.
Proc. R. Soc. A (2011)

Criterion in heat transfer optimization

L highconductivity material adiabatic H Q uniform heat source T0 W

1021

Figure 2. Two-dimensional heat conduction with a uniformly distributed internal heat source.

(b ) Application to a two-dimensional volume-point heat conduction

We will apply our proposed approach to practical cases where heat transfer is used for heating or cooling such as in the so-called volume-point problems (Bejan 1997) of heat dissipating for electronic devices as shown in gure 2. A uniform distributes in a two-dimensional device with length and internal heat source Q width of L and H , respectively. Owing to the tiny scale of the electronic device, the joule heat can only be dissipated through the surroundings from the point boundary area such as the cooling surface in gure 2, with the opening W and the temperature T0 on one boundary. In order to lower the unit temperature, a certain amount of new material with high thermal conductivity is introduced inside the device. As the amount of the high thermal conductivity material (HTCM) is given, we need to nd an optimal arrangement so as to minimize the average temperature in the device. According to the new extremum principle based on entransy dissipation, for this volume-point heat conduction problem, the optimization objective is to minimize the volume-average temperature, the optimization criterion is the minimum entransy dissipation, the optimization variable is the distribution of the HTCM and the constraint is the xed amount of the HTCM, i.e. k (x , y )dV = constant.
U

(4.15)

By the variational method, a Lagrange function, P, is constructed =

U

[k |VT |2 + Bk ]dV ,

(4.16)

where the Lagrange multiplier B remains constant because of a constant thermal conductivity. The variation of P with respect to temperature T gives k VT dT n dS
G U

V (k VT )dT dV = 0.

(4.17)

Proc. R. Soc. A (2011)

1022

Q. Chen et al.

Because the boundaries are either adiabatic or isothermal, the surface integral on the left side of equation (4.17) vanishes, that is, k VT dT n dS = 0.
G

(4.18)

Moreover, owing to a constant entransy output and a minimum entransy dissipation rate, the entransy input reaches the minimum when d
U

dV = QT
U

dT dV = 0. Q

(4.19)

Substituting equations (4.18) and (4.19) into equation (4.17) in fact gives the thermal energy conservation equation based on Fouriers Law = 0. V (k VT ) + Q (4.20)

This result shows again that the irreversibility of heat transfer can be measured by the entransy dissipation rate Fh . The variation of P with respect to thermal conductivity k gives (4.21) |VT |2 = B = constant This means that in order to optimize the heat dissipating process, i.e. to minimize the volume-average temperature, the temperature gradient should be uniform. This in turn requires that the thermal conductivity be proportional to the heat ow in the entire heat conduction domain, i.e. the HTCM be placed at the area with the largest heat ux. As an example, the cooling process in a low-temperature environment is analysed here. For the unit shown in gure 2, L = H = 5 cm, Q = 100 W cm2 , W = 0.5 cm and T0 = 10 K. The thermal conductivity of the unit is 3 W (mK)1 , and that for the HTCM is 300 W (mK)1 occupying 10 per cent of the whole heat transfer area. The implementary steps are as follows: (A) Local optimization (1) Divide the entire heat transfer region into several parts; (2) set all the parts composed by the original substrate material with low thermal conductivity; (3) numerically simulate the temperature eld and the entransy dissipation rate in the entire heat transfer region; (4) nd the part with the highest temperature gradient, and ll it with the HTCM; (5) numerically simulate again the temperature eld and the total entransy dissipation rate in the heat transfer region; (6) compare the total entransy dissipation rates with and without the HTCM element just lled in. If the total entransy dissipation rate is reduced, go to step 8. Otherwise go to step 7; (7) remove the HTCM element from the part that is just lled in and ll it in the part where the temperature gradient is the next largest. Then go to step 5; (8) judge whether all the HTCM has been used up. If so, go to (B) global optimization. Otherwise, go to step 4. (B) Global optimization In the above steps, the HTCM is distributed by iteration. Since local optimum may not assure a global optimum, the results obtained above need further adjustment. The detailed steps are:
Proc. R. Soc. A (2011)

Criterion in heat transfer optimization

(a) (b)

1023

(c)

Figure 3. Different arrangements of HTCM. (a ) Simple uniform HTCM arrangement, (b ) HTCM arrangement using the extremum principle of entransy dissipation and (c ) HTCM arrangement using the principle of minimum entropy generation.

(9) Choose one of the parts lled with the HTCM one after another; (10) remove the HTCM element from the chosen part and ll it in with the original substrate material; (11) numerically simulate the temperature eld and the entransy dissipation rate in the entire heat transfer region; (12) nd the part with the highest temperature gradient, and ll it with the HTCM; (13) numerically simulate again the temperature eld and the total entransy dissipation rate in the heat transfer region; (14) compare the total entransy dissipation rates with and without using the HTCM. If the total entransy dissipation rate is reduced, go to step 16. Otherwise go to step 15; (15) remove the HTCM element from the part that is just lled in and ll it in the part where the temperature gradient is the next largest. Then go to step 13; (16) judge if all the HTCM lled parts in series have been checked and the arrangement of them has not been changed: if not, go to step 9 and continue the optimization steps, or else end the optimization. For example, after dividing the whole heat transfer area into 40 40 parts, gure 3b shows the distribution of the HTCM according to the extremum principle of entransy dissipation, where the black area represents the HTCMthe same hereinafter. The HTCM with a tree structure absorbs the heat generated by the internal source and transports it to the isothermal outlet boundarysimilar in both the shape and function of actual tree roots. For a xed amount of HTCM, gure 4a ,b compares the temperature distributions between a uniform distribution of HTCM shown in gure 3a , and the optimized distribution in gure 3b based on the extremum principle of entransy dissipation. The average temperature in the rst case is 544.7 K
Proc. R. Soc. A (2011)

1024
(a) 650

Q. Chen et al.
(b) 75 65 600 550 500 450 350 250 150 (c) 180 170 160 150 140 130 150 120 110 180 170 160 150 140 130 120 25 150 55 55 45 35 25 25 55 35 55 45 65 75

Figure 4. The temperature elds obtained from different arrangements of HTCM. (a ) From the uniform HTCM arrangement in gure 3a , (b ) from the optimized arrangement of HTCM in gure 3b and (c ) from the HTCM arrangement in gure 3c .

while the temperature in the second-optimized case is 51.6 K, a 90.5 per cent reduction! It clearly demonstrates that the optimization criterion of entransy dissipation extremum is highly effective for such applications. Furthermore, as shown in gure 4b , the temperature gradient eld is also less uctuating in the optimized case.

(c ) The same case optimized using the minimum entropy approach

For comparison, we also treated the same problem using the minimum entropy V = constant, and what generation principle. Again, the constraint is U k (x , y )d we seek is also the optimal HTCM arrangement, except in this case that: (i) the optimization criterion is the minimum entropy generation and (ii) the corresponding energy conservation equation should be added as a constraint, because it is not implied in the principle of minimum entropy generation when the thermal conductivity is constant.
Proc. R. Soc. A (2011)

Criterion in heat transfer optimization

1025

Introducing the corresponding Lagrange function =

U

|VT |2 ) dV , + B k + C (V k VT + Q T2

(4.22)

where B and C are also the Lagrange multipliers. The constraint of thermal conductivity is the isoperimetric condition, and consequently B is a constant. C is a variable related to space coordinates. The variation of P with respect to temperature T gives V (k VC ) = 2k |VT |2 2Q + 2. 3 T T (4.23)

While the variation of P with respect to thermal conductivity k yields VC VT VT 2 = B = constant T2 (4.24)

Likewise, equation (4.24) gives the guideline for optimization based on the criterion of minimum entropy generation. Meanwhile, the most optimization steps used are the same as in the rst case, except that: (i) change nding the part with extreme temperature gradient to that with the extreme in absolute value of VA3 VT (VT 2 /T 2 ); (ii) replace the criterion of the total entransy dissipation rate by that of the total entropy generation rate. Figure 3c shows the distribution of HTCM based on the principle of minimum entropy generation. Comparison of gure 3b ,c shows that although the distributions of HTCM are similar between the two results in most areas, the root-shape structure from the extremum principle of entransy dissipation is not directly connected to the heat ow outlet, leaving some parts with the original material in between them so that the heat cannot be transported smoothly to the isothermal outlet boundary. Figure 4c gives the optimized temperature distribution obtained by the minimum entropy generation. Because the lowthermal conductivity material is adjacent to the heat outlet, the temperature gradient grows larger and thus lowers the entire heat transfer performance. The averaged temperature of the entire area is 150.899.2 K higher than that obtained by the extremum principle of entransy dissipation. From the denition, it is easy to nd that in order to decrease the entropy generation, we have to both reduce the temperature gradient and raise the temperature, thus leading to the arrangement of HTCM shown in gure 3c . In addition, according to the principle of minimum entropy generation, the optimization objective of a steady-state heat dissipating process can be expressed as: d D 1 Q T =d
m U

k |VT |2 dV = 0, T2

(4.25)

where (D(1/T ))m = ((1/T ) (1/T0 ))m is the equivalent thermodynamics potential difference that represents the generalized force in the entropy picture for heat transfer. Thus, minimizing the entropy generation equals to minimizing
Proc. R. Soc. A (2011)

1026

Q. Chen et al.

Table 1. Optimized results obtained by the optimization criteria of minimum entropy generation and entransy dissipation extremum. optimization results optimization criterion extremum entransy dissipation minimum entropy generation Fh (WK)1 5.5 104 1.58 105 Sgen (W K1 ) 100.7 81.7 Tm (K) 51.6 150.8 Tmax (K) 83.0 194.9 D 1 T /(1/K )
m

2.2 102 7.1 103

the equivalent thermodynamics potential difference (D(1/T ))m , leading to the highest exergy transfer efciency. That is, the minimum entropy generation principle is equivalent to the minimum exergy dissipation during a heat transfer process. To facilitate the comparison between the two results from gure 4b ,c , table 1 lists the key ndings side by side, obtained, respectively, by the optimization criteria of the minimum entropy generation and the entransy dissipation extremum. It indubitably shows in the table that the proposed entransy-based approach is more effective than the entropy-based one in heat transfer optimization, for the former leads to a result with signicantly reduced mean temperature than that by the latter (51.6 versus 150.8 K), and much lower maximum temperature (83.0 versus 194.9 K). Whereas the entropybased approach is preferred in exergy transfer optimization, as it results in a signicantly lower equivalent thermodynamic potential (7.1 103 K1 versus 2.2 102 K1 ). Besides heat conduction, we (Chen et al. 2009b ) also compared the two criteria in heat convection optimization. Our results indicate that both principles are applicable to convective heat transfer optimization, subject, however, to different objectives. The minimum entropy generation principle works better in searching for the minimum exergy dissipation during a heatwork conversion, whereas the entransy dissipation extremum principle is more effective for processes not involving heatwork conversion, in minimizing the heat transfer ability dissipation. In addition, based on the concept of the entransy dissipation rate, we (Chen et al. 2009a ) introduced the non-dimensional entransy dissipation rate and employed it as an objective function to analyse the thermal transfer process in a porous material. Moreover, some of us (Guo et al. 2010) dened the equivalent thermal resistance of a heat exchanger to measure the irreversibility of heat transfer in the processes of heating or cooling. After establishing the relationship between the heat exchanger effectiveness and the thermal resistance, Guo et al. found that reducing the thermal resistance leads to a monotonic increase in the heat exchange effectiveness. Guo et al. also demonstrated that the irreversibility in a heat exchanger is more effectively represented by its thermal resistance, while the so-called entropy generation paradox occurs if using the entropy generation criterion.
Proc. R. Soc. A (2011)

Criterion in heat transfer optimization

1027

5. Conclusions
In an electric conduction process or in a uid ow process, although the electron or the uid mass is conserved, the electric energy or the mechanical energy is dissipated. These two non-conserved quantities are used as the irreversibility measurements for their respective processes. Similarly, in heat transfer processes, the thermal energy is conserved. When the heat transfer involves heatwork conversion, exergy is the dissipated quantity, whereas in the heating or cooling process, we have demonstrated that the newly dened entransy is the dissipated quantity. Correspondingly, the objectives of heat transfer are classied into two different categories: one is direct heating or cooling and the other is for heatwork conversion. The irreversibility of the former case should be measured by the entransy dissipation rate while the latter should be measured by the rate of entropy generation or exergy dissipation. More generally, Fouriers Law derived by the minimum entransy dissipation principle satises a constant thermal conductivity, whereas by the principle of minimum entropy production, the thermal conductivity has to be inversely proportional to the squared temperaturesomething that contradicts the existing facts. Thus, the entransy dissipation is a better alternative for the measurement of irreversibility in heat transfer. For the testing case of a volume-point heat conduction problem with uniform internal heat source, if a xed amount of HTCM is used to minimize the volume-average temperature, the entransy dissipation extremum has shown to be a more effective optimization criterion than the minimum entropy generation.
The present work is supported by the National Natural Science Foundation of China (grant no. 51006060) and the Postdoctoral Scientic Fund of China (grant no. 200902080).

References
Bejan, A. 1979 Study of entropy generation in fundamental convective heat transfer. J. Heat Transf. Trans. ASME 101, 718725. Bejan, A. 1996 Entropy generation minimization: the new thermodynamics of nite-size devices and nite-time processes. J. Appl. Phys. 79, 11911218. (doi:10.1063/1.362674) Bejan, A. 1997 Constructal-theory network of conducting paths for cooling a heat generating volume. Int. J. Heat Mass Transf. 40, 799816. (doi:10.1016/0017-9310(96)00175-5) Bergles, A. E. 1988 Some perspectives on enhanced heat transfersecond-generation heat transfer technology. J. Heat Transf. 110, 10821096. (doi:10.1115/1.3250612) Bergles, A. E. 1997 Heat transfer enhancementthe encouragement and accommodation of high heat uxes. J. Heat Transf. Trans. ASME 119, 819. (doi:10.1115/1.2824105) Bertola, V. & Cafaro, E. 2008 A critical analysis of the minimum entropy production theorem and its application to heat and uid ow. Int. J. Heat Mass Transf. 51, 19071912. (doi:10.1016/j.ijheatmasstransfer.2007.06.041) Charach, C. & Rubinstein, I. L. 1989 On entropy generation in phase-change heat conduction. J. Appl. Phys. 66, 40534061. (doi:10.1063/1.343989) Chen, G. 1996 Heat transfer in micro- and nanoscale photonic devices. Annu. Rev. Heat Transf. 7, 157.
Proc. R. Soc. A (2011)

1028

Q. Chen et al.

Chen, Q., Ren, J. & Meng, J. 2007 Field synergy equation for turbulent heat transfer and its application. Int. J. Heat Mass Transf. 50, 53345339. (doi:10.1016/j.ijheatmasstransfer. 2007.10.001) Chen, Q., Wang, M., Pan, N. & Guo, Z. Y. 2009a Irreversibility of heat conduction in complex multiphase systems and its application to the effective thermal conductivity of porous media. Int. J. Nonlinear Sci. Numer. Simul. 10, 5766. Chen, Q., Wang, M., Pan, N. & Guo, Z. Y. 2009b Optimization principles for convective heat transfer. Energy 34, 11991206. (doi:10.1016/j.energy.2009.04.034) Cheng, X. 2004 Entransy and its applications in heat transfer optimization. PhD thesis, Tsinghua University, Beijing. Guo, Z., Zhu, H. & Liang, X. 2007 Entransya physical quantity describing heat transfer ability. Int. J. Heat Mass Transf. 50, 25452556. (doi:10.1016/j.ijheatmasstransfer.2006.11.034) Guo, Z., Liu, X., Tao, W. & Shah, R. 2010 Effectiveness-thermal resistance method for heat exchanger design and analysis. Int. J. Heat Mass Transf. 53, 28772884. (doi:10.1016/ j.ijheatmasstransfer.2010.02.008) Gyarmati, I. 1970 Non-equilibrium thermodynamics: eld theory and variational principles. New York, NY: Springer. Hesselgreaves, J. E. 2000 Rationalisation of second law analysis of heat exchangers. Int. J. Heat Mass Transf. 43, 41894204. (doi:10.1016/S0017-9310(99)00364-6) Karcz, J., Cudak, M. & Szoplik, J. 2005 Stirring of a liquid in a stirred tank with an eccentrically located impeller. Chem. Eng. Sci. 60, 23692380. (doi:10.1016/j.ces.2004.11.018) Kreuzer, H. 1981 Nonequilibrium thermodynamics and its statistical foundations. Oxford, UK: Clarendon Press. Meng, J., Liang, X. & Li, Z. 2005 Field synergy optimization and enhanced heat transfer by multi-longitudinal vortexes ow in tube. Int. J. Heat Mass Transf. 48, 33313337. (doi:10.1016/ j.ijheatmasstransfer.2005.02.035) Prigogine, I. 1967 Introduction to thermodynamics of irreversible processes, 3rd edn. New York, NY: Interscience. Schfer, M., Karaszen, B., Uludag, Y., Yapici, K. & Ugur, . 2005 Numerical method for optimizing stirrer congurations. Comput. Chem. Eng. 30, 183190. (doi:10.1016/ j.compchemeng.2005.08.016) Shah, R. K. & Skiepko, T. 2004 Entropy generation extrema and their relationship with heat exchanger effectivenessnumber of transfer unit behavior for complex ow arrangements. J. Heat Transf. Trans. ASME 126, 9941002. (doi:10.1115/1.1846694) Webb, R. 1994 Principles of enhanced heat transfer. New York, NY: John Wiley & Sons, Inc. Webb, R. & Bergies, A. 1983 Heat transfer enhancement: second generation technology. Mech. Eng. 115, 6067. Wu, J. & Liang, X. 2008 Application of entransy dissipation extremum principle in radiative heat transfer optimization. Sci. China Ser. E Technol. Sci. 51, 13061314. (doi:10.1007/ s11431-008-0141-6)

Proc. R. Soc. A (2011)