International Journal of Heat and Mass Transfer 53 (2010) 4112–4122

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

Natural convective flow and heat transfer of supercritical CO2 in a rectangular

circulation loop
Xin-Rong Zhang a,b,*, Lin Chen a, Hiroshi Yamaguchi b
a
Department of Energy and Resources Engineering, College of Engineering, Peking University, Beijing 100871, China
b
Energy Conversion Research Center, Department of Mechanical Engineering, Doshisha University, Kyo-Tanabeshi, Kyoto 610-0321, Japan

a r t i c l e i n f o a b s t r a c t

Article history: Fluid dynamics and heat transfer of supercritical CO2 natural convection are important for nuclear engi-
Received 13 April 2009 neering and new energy system design etc. In this paper, in order to study the flow and heat transfer
Received in revised form 29 April 2010 behavior of supercritical CO2 natural circulation system, a computational simulation on a closed natural
Accepted 7 May 2010
circulation loop (NCL) model has been carried out. The fluid temperature in the loop varies between
Available online 3 June 2010
298.15 K and 323.15 K, which is across the CO2 critical temperature, and the density is found to be in
the range of 250–800 kg/m3. The results show a small temperature difference of 25 °C between heating
Keywords:
and cooling sources can induce a mass flow with the Reynolds number up to 6  104 using supercritical
Carbon dioxide
Supercritical fluid
CO2 fluid. A periodic reversal flow pattern is found and presented in this paper. Enhanced heat transfer
Natural convection phenomenon is also found for the supercritical CO2 natural convective flow. The mechanisms to this
Heat transfer enhancement and the heating effect on the flow are also discussed in detail in the present study.
Periodic flow Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction operation pressure in some energy systems, such as heat pump

and solar thermal collector. It is seen that the density, thermal con-
Application system utilizing supercritical CO2 as working fluid ductivity and viscosity drop dramatically when the temperature
is one of the most popular research issues in recent years. The increases across the critical temperature. The specific heat curve
key problem is how to get higher efficiency based on better under- forms a high peak near the critical temperature.
standing about the mechanisms and the design of new systems These thermal physical properties help to explain the flow and
and/or new processes. Now both in the laboratory and industrial heat transfer behaviors exhibited in relevant studies using super-
stage, supercritical CO2 offers a promising choice, it can be em- critical CO2 as working fluids. The basic performance and flow
ployed in the process of deposition and preparation of materials property of supercritical CO2 have been discussed originally by Hall
[1,2], nuclear reactor applications [3,4], chemical extraction [5,6], [11], Protopopov [12], and later by Jiang et al. [13], Jackson [14], re-
cryogenic refrigeration [7], and for the heat pump systems [8– cently by He et al. [15,16]. Besides, a lot of studies on the thermal
10]. The main reason that supercritical CO2 is chosen in these and hydraulic performances of supercritical fluids have also been
applications is that CO2 is non-flammable, environmental benign, carried out by other researchers. Bernard Zappoli has made a com-
and generally displays high efficiency. However, there still exist prehensive review on these methods and results [17].
difficulties in understanding the behavior of supercritical CO2 sys- Utilizing CO2 as working fluid in closed systems can also be seen
tem, due to the temperature-sensitive thermal properties and the as a way of CO2 storage which is also very helpful to solve the envi-
geometric complexities of devices, which have been discussed in ronmental problems such as global warming and ozone layer
various previous studies [4,5,8]. depletion. Previous researches focus on basic flow properties and
The thermal physical properties of CO2 under supercritical con- heat transfer behaviors of supercritical CO2 [18–20]. Now the
ditions vary greatly even when there is only a quite small change in emphasis has partly shifted toward its application in engineering
temperature. Its critical temperature is 304.13 K, with the critical systems where the more complicated factor of geometric design
pressure is 7.38 MPa. Fig. 1 shows the thermal physical properties is involved. For example, Zhang et al. [21,22] systematically stud-
of CO2 at 9.0 MPa. This pressure of 9.0 MPa represents a typical ied the performance of a solar energy powered supercritical CO2
cycle through both experiments and numerical simulations. Nikitin
et al. [23] experimentally studied the thermal hydraulic perfor-
* Corresponding author at: Department of Energy and Resources Engineering, mance of a CO2 circulation loop.
College of Engineering, Peking University, Beijing 100871, China. Tel.: +86 10
Most of the systems using supercritical CO2 as working fluid
82529066; fax: +86 10 82529010.
E-mail address: zhxrduph@yahoo.com (X.-R. Zhang). have shown advantages in both heat transfer behaviors and the

0017-9310/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2010.05.031
 X.-R. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 4112–4122 4113

Nomenclature

A area V dimensional velocity

cp specific heat capacity x x coordinate location
D diameter of pipe X dimensionless axial coordinate (X = x/L0)
E energy y y coordinate location
g gravitational acceleration
Gr Grasholf number Greek letters
h heat transfer coefficient a thermal diffusivity
H length of vertical pipes b volumetric expansion coefficient
L heating (cooling) length of a pipe k thermal conductivity
L0 total length of a horizontal pipe l dynamic viscosity
L1 adiabatic pipe length on horizontal pipe U   ð ðs  
dissipation function, rÞ  VÞ
m_ mass flow rate
s shear tensor, 
sxx sxy
Nu Nusselt number syx syy
P external surface force m kinetic viscosity
p pressure of the fluid q density of fluid
Pr Prandlt number
q heat flux Subscripts
QW boundary heat input b bulk
Ra Rayleigh number r radial direction
Re Reynolds number ref reference value, bulk value
t time wall wall value
T temperature of fluid x local, value of specific axial location
u velocity

compactness of system design. But the mechanisms of the behav- for particular system do exist. However, Kumar and Gopal’s study
iors, especially their instabilities are still not clearly known be- used a one-dimensional steady state model and many strict
cause of the complexities in both geometry and boundary assumptions were made, thus the detailed behavior and inferences
conditions. For CO2 systems, Yoshikawa et al. [24] numerically are to some extent limited.
and experimentally studied a CO2 circulation system and found In the present system, a closed rectangular circulation loop uti-
unstable behavior of velocity field, and recommended two or three lizing supercritical CO2 is set. The system works under simple heat-
dimensional model for more detailed knowledge. Chatoorgoon et ing–cooling condition and no pumping devices are used. Such a
al. [25] conducted numerical experiments and developed a non- simplified model can be considered as a fundamental study and
dimension parameter to describe the flow stabilities of supercriti- the purpose is threefold: (1) to investigate the flow behavior and
cal CO2 in a natural convection loop. Jain and Rizwan-uddin [26] stability of such a ‘‘without-valve” natural convection system; (2)
numerically studied the stability threshold in power-curve of to study the heat transfer performance of such a supercritical
two-phase supercritical CO2 flow in a natural circulation loop. CO2 based system; and (3) to discuss the mechanisms and infer-
Those works got some useful results, which fit quite well with ences of the system. With these aims set, this study can improve
respective experiments. However, their systems are mostly rectan- our understanding about natural convection energy conversion
gular loops where an inlet and outlet boundary condition is as- and other supercritical CO2 based implicational systems.
sumed and the results can be used only through case-by-case
analysis, and the general flow behaviors and stabilities for super- 2. Physical model and approaches
critical CO2 flow are yet to be learned in further studies. Most re-
cently, Kumar and Gopal [27] simulated a closed sub-critical CO2 2.1. Physical model
based natural circulation loop, and reported that optimal designs
In the present study, a rectangular circulation loop is set up and
shown in Fig. 2 to investigate the flow behaviors and stabilities of

L1 Constant temperature wall; TL L1

L
CO2
adiabatic

D
H
g

y Plane 1
x L
0 L1 Constant temperature wall; TH L1
L0

Fig. 1. Thermal physical properties of CO2 around critical temperature (9.0 MPa). Fig. 2. Schematic of the problem (NCL) studied.
4114 X.-R. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 4112–4122

natural convection supercritical CO2 system. As shown in Fig. 2, the the average kinetic viscosity along a pipe:
circulation loop is two-dimensional with constant diameter of D. RL
The geometry for the loop is H  L0 for vertical pipe and horizontal 0
mx dx
m¼ ð7Þ
pipe. The middle parts of the upper and lower pipes (with length of L
L) are heating and cooling wall sections, respectively. Other pipe where L is the length of heating or cooling wall.
walls are assumed to be adiabatic areas. The present model estab- When referring to the velocity field, the local main flow velocity
lished here is also based on a solar collecting system using super- and the reference velocity is:
critical CO2 as working fluid [28]. In the present physical model, RA *
the temperatures of heating and cooling walls are set constant, ur jqr V dA
ux ¼ R0 * ð8Þ
TH and TL, respectively. The thickness of the tube wall is assumed A
0
jqr V dA
to be zero. The gravity acceleration is fixed at g as shown in Fig. 2.
In order to investigate the behaviors and flow structures in the and
simplified two-dimensional system, some walls and planes are set RL
and named: the upper cooling wall of upper pipe is named as wall 0
ux dx
u1 ¼ ð9Þ
1, the lower cooling wall of upper pipe is named as wall 2, the L
upper heating wall of lower pipe named as wall 3 and the lower the local reference thermal conductivity is defined as:
heating wall of lower pipe named as wall 4. Wall 1 and wall 2 RA
are the cooling walls with a constant temperature of TL; wall 3 kr jqr V  dAj
kx ¼ R0 A ð10Þ
and wall 4 are heating walls with a constant temperature of TH. jqr V  dAj
0
It is also shown in Fig. 2 that one plane is set near the right end
of the lower pipe, which is named as plane 1, to monitor the vari- And local bulk temperature and the reference temperature of a
ation of some important variables before/after the main heat trans- heating pipe:
fer wall sections. And to minimize the effect of 90° flow direction RA *

change at corners, semicircle elbows are set to connect the four 0

T r jqr V dAj
T b;x ¼ R * ð11Þ
A
pipes together. jqr V dAj
0

2.2. Governing equations and defined parameters and

RL
0
T b;x dx
In this numerical analysis the continuity equation and incom- T ref ¼ ð12Þ
L
pressible Navier–Stokes equations are solved coupled with energy
equation on controlled volume. This assumption of incompressible and the Reynolds number defined:
flow is reasonable considering the small velocity (0–1 m/s) in the u1 D
loop and the large acoustic speed (102 m/s) of CO2 fluid under cur- Re ¼ ð13Þ
m
rent model conditions. Then the governing equations are written as
following: the local Nusselt number along the heating or cooling wall is de-
Continuity equation: fined as:

@q * hx D
þ r  ðq V Þ ¼ 0 ð1Þ Nux ¼ ð14Þ
@t kx
Momentum equation for incompressible flow The heat flux on wall is defined:
 
*   @T
@V * * * qx ¼ kx ð15Þ
þ ðq V V Þ ¼ rP þ r lðr  V Þ þ q!
g ð2Þ @y W
@t
and the local heat transfer coefficient:
Energy equation
qx
@ðqEÞ * dP hx ¼ ð16Þ
þ r  ðq V EÞ ¼ rðkrTÞ þ bT þU ð3Þ T wall  T b;x
@t dt
then the Nusselt number of a heat transfer wall is:
where U represents the viscous dissipation function; and E is de-
RL
fined as Nux dx
0
Z Nu ¼ ð17Þ
T
V 2 L
E¼ C p dT þ ð4Þ
To 2 the Grasholf number and Rayleigh number:
3
T0 is the reference value used in Eq. (4); gd ðT wall  T ref Þb
Gr ¼ ð18Þ
In this paper, some physical parameters are defined to describe m2
the flow and heat transfer. The mass flow rate monitored of a plane 3
gd ðT wall  T ref Þb
is: Ra ¼ ð19Þ
ma
Z A *
where Tref is as defined in Eq. (12); b is the volumetric expansion
_ ¼
m qr V dA ð5Þ
0 coefficient of the supercritical CO2 flow; a is local thermal diffusiv-
ity of the supercritical CO2 fluid.
where qr is local density along the radial direction.
In addition, a dimensionless parameter X is defined, which rep-
The local kinetic viscosity is defined as:
resents the length measured from the left corner of a horizontal
RA * pipe (total length = L0):
0
mr jqr V dA
mx ¼ R * ð6Þ x
A X¼ ð20Þ
0
jqr V dA L0
 X.-R. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 4112–4122 4115

2.3. Thermal–physical properties of supercritical CO2 fluid 2.4. Simulation details

In the present simulation an operation pressure of 9.0 MPa is as- An unsteady flow condition is applied in this study as the pre-
sumed and the thermal–physical properties of CO2 are defined as a vious studied supercritical CO2 systems [24–27] generally showed
function of both temperature and pressure. Several pre-calcula- unsteady behavior. And after several testing simulations this un-
tions have shown the maximum pressure difference in the loop steady model was finally decided. In this simulation, finite-volume
is smaller than 2  103 Pa, which is only 0.022% compared with method which is based on integration over the controlled volume
the operation pressure of 9.0 MPa. In addition, it is also reported is used to solve the model equations subject to the initial and
in a study of Kumar and Gopal [27] with a similar set closed CO2 boundary conditions in this numerical solution. The governing
circulation loop the relative pressure difference is smaller than equations are solved concurrently using the coupled method. The
0.3% (with the operation pressure 5 MPa). So the calculation of pressure-implicit with splitting of operators (PISO) algorithm is
thermal physical properties with fixed pressure of 9.0 MPa is suffi- used to solve the coupling model between velocity and pressure.
ciently precise in this study. Hence the thermal–physical proper- That is also used in previous relevant studies [31–33]. The momen-
ties of CO2 are simplified to be only a function of temperature. In tum terms and energy terms in the governing equations are iter-
this study these properties are obtained from the NIST Standard ated with a secondary-order upwind scheme which uses the
Reference Database (REFPROP) Version 8.0 [29] and are arranged upstream values and gradients to compute the control volume face
in piecewise-polynomial (quartic polynomial) functions of temper- values. On the pipe walls a non-slip boundary condition is used and
ature at 9.0 MPa. This is because the number of options for the in- a laminar flow model is taken. The numerical model has been pre-
let/outlet boundary conditions is severely limited under the NIST viously used and validated in a number of studies, especially in
database. And for the balance of computation resources cost and natural heat transfer cases [34,35].
in order to avoid these disadvantages, the properties of the super- A structured non-uniform grid system is used to solve the dis-
critical CO2 were written as piecewise polynomial functions of cretized equations. The grid space is set smaller and finer near
temperature at specific pressures. The same method has been used the pipe wall and at the elbows, in order to more accurately simu-
in other studies [20,30]. The NIST database and the interpolation late the situation near the wall areas where large parameter gradi-
function are compared in Fig. 3. After examining the values from ents exist. In this numerical study, a typical mesh of 16 points for
NIST and the interpolation functions it is found that the standard the radial direction is used and the pipe length direction about
residual of each property is less than 1% and the R-square of the every 2 mm a mesh point is set. The total number of mesh points
fit for each physical property exceeds 0.995. used here is about 1.0  105. In this study, an adaptive time meth-
od is used and the time step varies within the range of 103–0.1 s
according to the converge conditions. For validation, we used the
present simulation method to solve the flow of the famous general
natural convection loop model put forward by Vijayan [36], where
water convection is assumed. Results showed good agreement
with the Re–Gr analytical correlation proposed by Vijayan (see Ta-
ble 1), where only small derivations were found because of three
dimensional effects.

3. Simulation results and discussion

This model is the first step investigating into supercritical CO2

flow in natural circulation loops. The present parameters are cho-
sen based on the application of our previous solar collect experi-
mental model [28], the parameters for the geometry are set as:
L0 = 5.0 m, L = 3.0 m, L1 = 1.0 m, H = 2.0 m, D = 0.006 m; for con-
stant temperature walls, TH = 323.15 K and TL = 298.15 K, which
temperature is across the critical temperature of CO2 fluid. And
the initial temperature and pressure for CO2 fluid are 298.15 K
and 9.0 MPa. The time used in this study is measured from the
steady beginning with zero velocity.

3.1. Overall flow behaviors

For the supercritical CO2 heat transfer model in the present pa-
per, calculation work with the above mentioned initial and bound-
ary conditions was conducted. The flow time is promoted to about
1500 s and a periodic reversal flow pattern is found.

Table 1
Validation comparison between present model and general Vijayan [36] solution.

Heat input Re (present Gr (present Re (Vijayan

(W) model) model) solution)
1000 15.1 654,000 13.5
6000 270 2,140,000 283
Fig. 3. Comparison of CO2 properties from the NIST database and the piecewise 9000 364 3,210,000 343
polynomial functions at 9.0 MPa.
4116 X.-R. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 4112–4122

Fig. 4. Variations of average temperature and average velocity of plane 1 with flow Fig. 5. Variations of average temperature and average velocity of plane 1 in one
time. cycle.

The recorded average velocity and temperature as defined in lyze the dominant factor, the average pressures of several cross
Eqs. (8) and (11) are shown in Fig. 4 against the flow time. It can sections are respectively recorded and their differences are calcu-
be seen from Fig. 4 that at the beginning 40 s the recorded velocity lated and presented in Fig. 6. In Fig. 6, ‘Up’ or ‘Low’ means the
on plane 1 stays at nearly zero value and the temperature main- upper or lower horizontal pipe, and the label ‘0.1’ or ‘0.9’ repre-
tains at about 298 K. It can be considered as the heat cumulating sents the location X of cross sections along a pipe as defined in
time and then after the quiescent state temperature and velocity Eq. (20). The cross sections at X = 0.1 and X = 0.9 represent the dif-
overshoots are seen. This kind of overshoot is typical and validated ferences before and after heating or cooling areas.
in the beginning step of natural convection studies [24,37]. Then A detailed comparison between Figs. 5 and 6 shows that the
the flow begins to fluctuate forth and back and at about 500 s it velocity change is strictly dependent on the pressure change. It is
reaches a periodic change mode. It is interesting that according seen that during the time of 1191–1200 s the pressure difference
to Fig. 4 flow reversal phenomenon is seen in the loop. increases with flow time and reaches its peak and the velocity on
The recorded mass flow rate varies between about 0.018 kg/s the plane 1 increases. Then the pressure differences begin to drop
and 0.018 kg/s in the loop. And the maximum Reynolds number of and the velocity continues increasing but the acceleration de-
6.0  104 is obtained and the number ranges between 4  104 creases as a consequence of pressure drop. Till about 1204 s the
and 6  104 during the steady repetitive flow time after the over- velocity reaches its peak value about 0.9 m/s and the pressure dif-
shoots. The velocity can reach as high as 0.9 m/s, which is much ference becomes zero. After that the same principle applies: the
higher than natural pipe flow using water flow [37] and CO2 exper- acceleration of flow is directly determined by the pressure field
imental natural convection studies [24]. This can be ascribed to the in the loop. Further calculation shows that the acceleration of the
large density and viscous difference of water and CO2 flow, as well flow velocity can be written as linear function of the pressure dif-
as the damping forces in the experimental devices like much more ferences, the correlation coefficient between the four pressure dif-
elbow loss and the viscous loss. ferences and the acceleration of flow are 0.9620, 0.9534, 0.9426
In the present model strong natural convection flow is induced and 0.9776. This high correlation for the present closed loop is
by using supercritical CO2 fluid, only by a temperature difference of rarely seen in water cases. With low viscosity and high density
25 °C. The values of the Reynolds number are rarely seen in natural flow, the wall friction factor is not as important a determinant fac-
convection systems using normal working fluids, such as water and tor as the change of pressure field.
air, by such a small temperature difference. So in the present study, In order to analyze the periodical changes of pressure distribu-
the high Reynolds number and flow rates are of particular future tion in the circulation loop, the variation of average temperature
interest in the study of supercritical systems, and of potential and static pressure are recorded at the position of [Up(0.1)],
importance for low-grade heat utilization and high efficient system
design.

3.2. Discussions on the periodic reversal flow

The calculated average time periods for repetitive temperature

and velocity are of the same value of 52.8 s at the interval of 1000–
1400 s, where the flow can be considered as steadily periodic. In
this paper, we take out one typical period in the time 1180–
1240 s for flow reversal analysis and other flow periods all show
the same mode. Fig. 5 shows the evolution of average temperature
and velocity on plane 1 in that period. Plane 1 as shown in Fig. 2 is
located near the outlet of heating area on lower pipe. This kind of
repetitive flow reversal flow and its temperature variations are
very much similar with that of a previous natural convection study
using water fluid [38].
From Fig. 5, it is seen that flow reversal happens at the mid of a Fig. 6. Evolution of pressure differences of some cross sections located in upper and
cycle. Some critical time points are also noted in the figure. To ana- lower pipes.
 X.-R. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 4112–4122 4117

[Low(0.1)], [Up(0.9)] and [Low(0.9)]. Calculations show that for the For previous study the enhanced heat transfer for supercritical
positions of [Up(0.1)] and [Low(0.1)], the temperature and pres- CO2 flow is also found and generally ascribed to the thermal–phys-
sure are positively correlated with correlation coefficients 0.6742 ical properties of supercritical CO2 fluid [30,34]: the thinner ther-
and 0.8874; the positions of [Up(0.9)] and [Low(0.9)] negatively mal boundary layer, the decrease in viscosity, the increase in
correlated with correlation coefficients 0.7530 and 0.6910, heat capacity, and also the not fully-developed flow field. Those
respectively. This shows a moderate correlation between tempera- factors are also found in this simulation. Nevertheless, the present
ture and pressure in the loop and it can be concluded that the dis- model temperature range (298.15–323.15 K) falls around the sen-
tribution of temperature and density field partly determines the sitive critical temperature region of supercritical CO2 fluid. Thus
pressure field, and then the pressure field in the loop induces the non-uniform distribution of buoyancy forces in the radial pipe
mass flow. Also the pressure field can be affected by the flow rever- direction can be generated. Here the time point of 1196 s is taken
sal process. Thus a periodic flow pattern is found. out as an example. The density field in the heat transfer areas is
The overall system stability is of crucial importance to the de- shown in Fig. 8. In this figure, density difference between the
sign and operation of application systems. This is the first repeti- boundary layer and the main flow is greater than 150 kg/m3, which
tive flow reversal flow found for supercritical CO2 based natural is rarely seen in normal fluid flow. That density difference induces
circulation loop. The knowledge on this point is quite limited. For- great buoyancy forces in the loop.
mer studies concerning stabilities of rectangular natural circula- For the heating pipe (see Fig. 8a), the fluid near the walls is
tion loops generally used water as working fluid and some heated and its density decreases. The density difference between
dominant factors are clarified: Vijayan et al. [38] reported in their the main flow and the near-wall boundary layer is about 180 kg/
experiment repetitive reversal flow in rectangular natural circula- m3, which is more than 25% greater than the main flow. Therefore,
tion loops (NCLs), and analyzed the effect of heater power. Cha- a strong buoyancy force is generated along the radial direction of
toorgoon [39] also reported the high flow fluctuations of a the pipe, and the velocity vectors in the boundary layer always
supercritical water case. Then Cammarata et al. [37,40] analyzed show an upward direction as shown in Fig. 9a. For the cooling pipe
the effects of loop inner diameter, heat input, aspect ratio and grav- (see Fig. 8b), the fluid near the walls is cooled down and its density
ity intensity on stability. Misale et al. [41] examined loop incline increases. The density difference between the near-wall boundary
angle. Recently, Pilkhwal [42] used a simplified one dimensional layer and the main flow is about 150 kg/m3, which is about 25% of
model and found a repetitive flow figure very much similar with the average density in main flow. And the fluid is dragged down
the results in the present study. and the velocity vectors in the boundary layer always show down-
In the above studies, the stability of rectangular NCLs is gener- ward direction as shown in Fig. 9a.
ally attributed to five factors: heat input, loop diameter, gravity Therefore it is easy to understanding the velocity profiles of
intensity, loop inclination angle as well as aspect ratio. However, respective positions on heating and cooling pipes in Fig. 9a. It is
this study is only the first step toward the analysis of supercritical presented in Fig. 9a that the velocity vectors near walls on cooling
CO2 rectangular NCLs. Future point should be laid on the effect of pipe (wall 1 and 2) are always downward direction, while near
various relevant factors in this kind of loop and research into the walls on heating pipe (wall 3 and 4) are always upward direction.
typical laws controlling this phenomenon. However, as shown in Fig. 9b, no such phenomenon is found in the
non-heated region of the pipes. So we can decide that this observed
phenomenon is a result of the heating or cooling of wall.
3.3. Heat transfer behavior As a result of the above discussion, the downward flow near
wall 1 and the upward flow near wall 4 directly drive the heat
Calculated Nusselt numbers for heat transfer walls in the cho- and mass flow into the main flow field. In this case the heat and
sen flow period are shown in Fig. 7. It can be seen from Fig. 7 that: mass are directly transported to the main flow and are carried
(1) the Nusselt numbers also change periodically and generally the away quickly. However, for wall 2 and wall 3 their boundary flow
values on the four walls is above 10 and even greater than 30. The direction is toward the wall so the heat and mass transport is
time averaged Nusselt number is about 14, which is a relative high partly restricted. So the overall heat transfer of wall 1 and wall 4
Nusselt number for natural convective flow system with similar is enhanced and wall 2, wall 3 are restricted (see Nusselt numbers
pipe diameter; (2) an interesting phenomenon is the coherence in Fig. 7).
behavior between wall 1 and wall 4, and that between wall 2 In addition, during the time interval of the flow reversal, the
and wall 3 for calculated Nusselt number. flow reversal occurs for the boundary layer and the main flow as
shown in Fig. 10. The counter-flow time period is also shown in
this figure. In this period the counter-flow may be more turbulent
near the boundary layer, which can also contribute to the heat
transfer enhancement. It can be seen from Fig. 10 that for all the
time in the period the high density differences are maintained
and the streamlines’ conditions are the same for the near wall
areas. It should be noted that in the time of 1212.5 s (see
Fig. 10b), which is in the time of flow reversal, the bi-direction flow
is found. This can greatly contribute to the mix of the boundary and
the main flow, therefore the high peak Nusselt numbers are seen
(see Fig. 7). This kind of chaotic flow is also of importance in chem-
ical extraction and material preparation process [24].
Fig. 11 shows the Grasholf number and Rayleigh number calcu-
lated for the circulation loop. It is seen from Fig. 11 that as the flow
is unsteady both numbers change with time. The Grasholf number
is in the range of 2.25  108–4.5  108 and the Rayleigh number
varies between 1  109 and 2.75  109. As the Grasholf number
represents the strength of buoyancy force against the viscous force,
Fig. 7. Evolution of average Nusselt number of walls (1, 2, 3, 4) in one typical cycle. it can be inferred that the dominant factor is buoyancy forces. Ray-
4118 X.-R. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 4112–4122

Fig. 8. Density and flow streamlines at mid-lower (cooling) and mid-upper (heating) pipes (t = 1196 s).

Fig. 9. Velocity field at mid-lower and mid-upper pipes (t = 1196s).

 X.-R. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 4112–4122 4119

Fig. 10. Variation of density field and flow streamlines with time.

leigh number can be used to classify the laminar and turbulent region. From Fig. 11 it can be generally decided that the present flow
flows: when the number is smaller than 108 the flow is laminar is in a transition flow phase. This is very interesting for a self driven
flow, if the number is greater than 1010 the flow has become turbu- supercritical flow based natural convection system induced by only
lent, otherwise if it lies between 108 and 1010 it is transition flow a temperature difference of 25 °C. In one previous research of Shan
4120 X.-R. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 4112–4122

Fig. 10 (continued)

Fig. 11. Grasholf number and Rayleigh number of the loop in one typical cycle.

et al. [30] with supercritical CO2 closed chamber natural convection of temperature sensitive thermal–physical properties of supercrit-
very high Grasholf number was found and it is also concluded that ical CO2 fluid.
this kind of strong self induced convection of supercritical CO2 A small temperature difference of 25 °C for supercritical CO2
should have great importance in future energy systems design and fluid in this natural convective system without any pumping de-
applications. Further research works including experiments should vices can induce a mass flow with a Reynolds number up to
be done to get more knowledge of the behaviors of such systems, 6  104. This indicates the potential for supercritical CO2 to be used
including the heat transfer behaviors and the stabilities. as a high efficient natural convection working fluid.
Enhanced heat transfer efficiency is maintained in this super-
4. Concluding remarks critical model. The heating and cooling effects on the supercritical
fluid are evident and this contributes to the potential of high effi-
Heat transfer behaviors and flow pattern of a supercritical CO2 ciency. Besides traditional factors considered being the reason in
based rectangular circulation loop have been numerically studied. supercritical heat transfer process, the distribution of density field,
Based on the obtained results and discussions, the following re- especially the great density difference found in radial direction can
marks are made: be of great importance. The buoyancy forces induced boundary
For the present natural convection model, a repetitive reversal flow direction can greatly influence the heat transfer process.
flow is found, which is directly driven by the periodic change of All the results above strengthen the belief that utilizing CO2 as
pressure field. This pressure field is due to the typical distributions working fluid can help to promote a more efficiently compact
 X.-R. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 4112–4122 4121

mechanical design of heat transfer process and will also find use in [8] Petter Neksfit, Hfivard Rekstad, G. Reza Zakeri, Per Arne Schiefloe, CO2-heat
pump water heater: characteristics, system design and experimental results,
other systems or engineering applications. More studies should be
Int. J. Refrig. 21 (1998) 172–179.
conducted to get further knowledge of such systems. [9] Petter Neksa, CO2 heat pump systems, Int J. Refrig. 25 (2002) 421–427.
[10] Karl Ochsner, Carbon dioxide heat pipe in conjunction with a ground source
heat pump (GSHP), Appl. Therm. Eng. 28 (2008) 2077–2082.
Acknowledgements [11] W.B. Hall, Heat transfer near the critical point, in: Advances in Heat Transfer,
vol. 7, Academic Press, 1971.
The support of the National Science Foundation of China [12] V.S. Protopopov, Generalized correlations for local heat transfer coefficient for
turbulent flow of water and carbon dioxide at super-critical pressure in
(50706002) is gratefully acknowledged. uniformed heated tubes, Teplofizika Vysokikh Temperatur 15 (1977) 815–821
(in Russian).
[13] P.X. Jiang, Z.P. Ren, B.X. Wang, Convective heat and mass transfer of water at
Appendix A supercritical pressures under heating or cooling conditions in vertical tubes, J.
Therm. Sci. 4 (1995) 15–25.
[14] J.D. Jackson, Some striking features of heat transfer with fluids at pressures and
For 280 K < T < 330 K
temperatures near the critical point, in: Keynote Paper for International
X Conference on Energy Conversion and Application (ICECA), Wuhan, China,
qðTÞ ¼ aij T j ðfor 280 K < T < 313:5 K : i ¼ 1; 2001.
[15] S. He, Pei-Xue Jiang, Yi-Jun Xu, Run-Fu Shi, W.S. Kim, J.D. Jackson, A
for 313:5 K < T < 330 K : i ¼ 2; j ¼ 0; 1; 2; 3; 4Þ computational study of convection heat transfer to CO2 at supercritical
X
kðTÞ ¼ aij T j ðfor 280 K < T < 310:25 K : i ¼ 3; pressures in a vertical mini tube, Int. J. Therm. Sci. 44 (2005) 521–530.
[16] S. He, W.S. Kim, J.D. Jackson, A computational study of convective heat transfer
for 310:25 K < T < 316:5 K : i ¼ 4; to carbon dioxide at a pressure just above the critical value, Appl. Therm. Eng.
28 (2008) 1662–1675.
for 316:5 K < T < 330 K : i ¼ 5; j ¼ 0; 1; 2; 3; 4Þ [17] Bernard Zappoli, Near-critical fluid hydrodynamics, C.R. Mecanique 331 (2003)
X 713–726.
C p ðTÞ ¼ aij T j ðfor 280 K < T < 306:5 K : i ¼ 6; [18] G. Scalabrin, L. Piazza, Analysis of forced convection heat transfer to
for 306:5 K < T < 314 K : i ¼ 7; supercritical carbon dioxide inside tubes using neural networks, Int. J. Heat
Mass Transfer 46 (2003) 1139–1154.
for 314 K < T < 330 K : i ¼ 8; j ¼ 0; 1; 2; 3; 4Þ [19] P.X. Jiang, Y.J. Xu, J. Lv, R.F. Shi, S. He, J.D. Jackson, Experimental investigation of
X convection heat transfer of CO2 at super-critical pressures in vertical mini-
lðTÞ ¼ aij T j ðfor 280 K < T < 306:5 K : i ¼ 9; tubes and in porous media, Appl. Therm. Eng. 24 (2004) 1255–1270.
[20] M. Sharabi, W. Ambrosini, S. He, J.D. Jackson, Prediction of turbulent
for 306:5 K < T < 314 K : i ¼ 10; j ¼ 0; 1; 2; 3; 4Þ convective heat transfer to a fluid at supercritical pressure in square and
triangular channels, Ann. Nucl. Energy 35 (2008) 993–1005.
where aij(i = 1, . . . , 10; j = 0, . . . , 4) are the fitting parameters and [21] X.R. Zhang, H. Yamaguchi, K. Fujima, M. Enomoto, N. Sawada, Experimental
are shown in the following table. performance analysis of supercritical CO2 thermodynamic cycle powered by
solar energy, in: American Institute of Physics Conference Proceedings, vol.
Piecewise-polynomial fitting of properties for simulation of
832, 2006, pp. 419–424.
CO2:

j=0 j=1 j=2 j=3 j=4

i=1 7.27E+06 99645.7081 511.6921 1.1677 9.99E04
i=2 6.60E+07 812904.8 3753.5337 7.70336 0.00593
i=3 2.60E+08 3.57E+06 18381.067 42.05866 0.03608
i=4 4.05E+11 5.17E+09 2.48E+07 52716.85 42.07046
i=5 2.95E+09 3.62E+07 1.66E+05 340.3015 0.26091
i=6 0.05212 0.00232 6.18E06 0 0
i=7 167528.955 2166.73553 10.50851 0.02265 1.83E05
i=8 365.6572 3.35668 0.01027 1.05E05 0
i=9 3.25E04 4.43E06 1.04E08 0 0
i = 10 0.26065 0.00241 7.43E06 7.63E-09 0

References [22] X.R. Zhang, H. Yamaguchi, K. Fujima, M. Enomoto, N. Sawada, Theoretical

analysis of a thermodynamic cycle for power production using supercritical
carbon dioxide, Energy 32 (2007) 591–599.
[1] J.M. Blackburn, D.P. Long, A. Cabanas, J.J. Watkins, Deposition of conformal
[23] K. Nikitin, Y. Kato, L. Ngo, Printed circuit heat exchanger thermal–hydraulic
copper and nickel films from supercritical carbon dioxide, Science 294 (2001)
performance in supercritical CO2 experimental loop, Int. J. Refrig. 29 (2006)
141–145.
807–814.
[2] E. Kondoh, Deposition of Cu and Ru thin films in deep nanotrenches/holes
[24] Syouichirou Yoshikawa, Richard L. Smith Jr., Hiroshi Inomata, Yukihiko
using supercritical carbon dioxide, Jpn. J. Appl. Phys. 43 (Part 1) (2004) 3928–
Matsumura, Kunio Arai, Performance of a natural convection circulation
3933.
system for supercritical fluids, J. Supercrit. Fluids 36 (2005) 70–80.
[3] V. Dostal, P. Hejzlar, M.J. Driscoll, The supercritical carbon dioxide power cycle:
[25] V. Chatoorgoon, A. Voodi, P. Upadhye, The stability boundary for supercritical
comparison to other advanced power cycles, Nucl. Technol. 154 (2006) 283–
flow in natural-convection loops Part II: CO2 and H2, Nucl. Eng. Des. 235 (2005)
301.
2581–2593.
[4] Dong Eok Kim, Moo Hwan Kim, Jae Eun Cha, Seong O. Kim, Numerical
[26] Prashant K. Jain, Rizwan-uddin, Numerical analysis of supercritical flow
investigation on thermal–hydraulic performance of new printed circuit heat
instabilities in a natural circulation loop, Nucl. Eng. Des. 238 (2008) 1947–
exchanger model, Nucl. Eng. Des. 238 (2008) 3269–3276.
1957.
[5] P. Bondioli, C. Mariani, E. Mossa, A. Fedelli, A. Muller, Lampante olive oil
[27] K. Kiran Kumar, M. Ram Gopal, Steady-state analysis of CO2 based natural
refining with supercritical carbon dioxide, J. Am. Oil Chem. Soc. 69 (1992) 477–
circulation loops with end heat exchangers, Appl. Therm. Eng. (2008),
480.
doi:10.1016/j.applthermaleng.2008.08.002.
[6] F.C.V.N. Fourie, C.E. Schwarz, J.H. Knoetze, Phase equilibria of alcohols in
[28] X.R. Zhang, H. Yamaguchi, An experimental study on evacuated tube solar
supercritical fluids Part I. The effect of the position of the hydroxyl group for
collector using supercritical CO2, Appl. Therm. Eng. 28 (2008) 1225–1233.
linear C8 alcohols in supercritical carbon dioxide, J. Supercrit. Fluids 47 (2008)
[29] NIST Standard Reference Database-REFPROP, Version 8.0, 2006.
161–167.
[30] Xiaoying Shan, David P. Schmidt, James J. Watkins, Study of natural convection
[7] H. Yamaguchi, X.R. Zhang, K. Fujima, Basic study on new cryogenic
in supercritical CO2 cold wall reactors: simulations and experiments, J.
refrigeration using CO2 solid–gas two phase flow, Int. J. Refrig. 31 (2008)
Supercrit. Fluids 40 (2007) 84–92.
404–410.
4122 X.-R. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 4112–4122

[31] R.I. Issa, Solution of the implicit discretized fluid flow equations by operator [37] L. Cammarata, A. Fichera, I.D. Guglielmino, A. Pagano, On the effect of gravity
splitting, J. Comput. Phys. 62 (1986) 40–65. on the bifurcation of rectangular closed-loop thermosyphon, Heat Mass
[32] R.I. Issa, A.D. Gosmann, A.D. Watkind, The computation of compressible and Transfer 40 (2004) 801–808.
incompressible recirculating flows by a non-iterative scheme, J. Comput. Phys. [38] P.K. Vijayan, H. Austregesilo, V. Teschendorff, Simulation of the unstable
62 (1986) 66–82. oscillatory behavior of single-phase natural circulation with repetitive flow
[33] X.R. Zhang, S. Maruyama, H. Yamaguchi, Laminar natural convection hear reversals in a rectangular loop using the computer code ATHLET, Nucl. Eng.
transfer from a vertical baffled plate subjected to a periodic oscillation, ASME J. Des. 155 (1995) 623–641.
Heat Transfer 127 (2005) 733–739. [39] Vijay Chatoorgoon, Stability of supercritical fluid flow in a single-channel
[34] X.R. Zhang, S. Maruyama, S. Sakai, Numerical investigation of laminar natural natural-convection loop, Int. J. Heat Mass Transfer 44 (2001) 1963–1972.
convection on a heated vertical plate subjected to a periodic oscillation, Int. J. [40] L. Cammarata, A. Fichera, A. Pagano, Stability maps for rectangular circulation
Heat Mass Transfer 47 (2004) 4439–4448. loops, Appl. Therm. Eng. 23 (2003) 965–977.
[35] X.R. Zhang, H. Yamaguchi, Forced convection heat transfer of [41] M. Misale, P. Garibaldi, J.C. Passos, G. Ghisi, Bitencourt, Experiments in a single-
supercritical CO2 in a horizontal circular tube, J. Supercrit. Fluids. 41 phase natural circulation mini-loop, Exp. Therm. Fluid Sci. 31 (2007) 1111–1120.
(2007) 412–420. [42] D.S. Pilkhwal, W. Ambrosini, N. Forgione, P.K. Vijayan, D. Saha, J.C. Ferreri,
[36] P.K. Vijayan, Experimental observations on the general trends of the steady Analysis of the unstable behaviour of a single-phase natural circulation loop
state and stability behaviour of single-phase natural circulation loops, Nucl. with one-dimensional and computational fluid-dynamic models, Annal. Nucl.
Eng. Des. 215 (2002) 139–152. Energy 34 (2007) 339–355.