You are on page 1of 43

Review

A critical review of models of coupled heat and mass transfer


in falling-lm absorption
Jesse D. Killion, Srinivas Garimella *
Department of Mechanical Engineering, 2030 H.M. Black Engineering Building, Iowa State University, Ames, IA 50011, USA
Received 9 May 2000; accepted 28 August 2000
Abstract
In absorption space-conditioning systems, the performance of the absorber is critical to the overall system perfor-
mance, size, and rst-cost. The objective of this paper is to provide a comprehensive review of the signicant eorts
that researchers have made to mathematically model the coupled heat and mass transfer phenomena that occur during
falling-lm absorption. A detailed review of the governing equations, boundary conditions, assumptions, solution
methods, results, and validation of these investigations is presented. This review excludes experimental work in this
area, the eect of additives, and the eect of non-absorbable gases. It is shown that most work found in the literature
has focused on the particularly simplied case of absorption in laminar vertical lms of water-lithium bromide. Fewer
researchers have considered the important situations of wavy lms, turbulent lms, and lms on horizontal tubes.
Investigations of the ammonia-water uid pair have been generally more empirical in nature and/or restricted to ver-
tical laminar lms. This review is used to highlight key areas which need attention such as lm and vapor hydro-
dynamics, especially the non-periodicity, instability, and recirculatory motion of waves in the vertical wall case and
droplets and waves in the horizontal tube case. Also the potential interaction of the heat and mass transfer process on
the lm hydrodynamics, surface wetting, heat transfer in the vapor phase, and common simplications to the govern-
ing equations should all be considered carefully. Finally, emphasis must be placed on experimental validation of the
local conditions and transfer processes within the absorber, not just overall transport values. # 2001 Elsevier Science
Ltd and IIR. All rights reserved.
Keywords: Refrigerating system; Absorption system; Absorber; Falling lm; Heat transfer; Mass transfer; Modelling survey
Absorption a lm tombant : examen critique des mode les de
transfert de chaleur et de transfert de masse
Re sume
Dans les syste mes de conditionnement d'air a absorption, la performance de l'absorbeur est de terminante pour la per-
formance globale et la taille du syste me ainsi que l'investissement initial. L'objectif de cet article est de donner une vue
d'ensemble des travaux des chercheurs sur la mode lisation mathe matique des phe nome nes de transfert de chaleur et de
masse lors de l'absorption a lm tombant. On pre sente une analyse des e quations qui s'appliquent, les conditions limites, les
hypothe ses utilise es, les solutions trouve es, les re sultats et la validation de ces e tudes. Cette analyse exclut les travaux
0140-7007/01/$20.00 # 2001 Elsevier Science Ltd and IIR. All rights reserved.
PI I : S0140- 7007( 00) 00086- 4
International Journal of Refrigeration 24 (2001) 755797
www.elsevier.com/locate/ijrefrig
* Corresponding author. Tel.: +1-515-294-8616; fax: +1-515-294-3261.
E-mail address: garimell@iastate.edu (S. Garimella).
expe rimentaux dans ce domaine, l'eet des additifs, et les eets des gaz non absorbables. On montre que la plupart des
travaux publie s est axe e sur le cas simplie de l'absorption dans les lms laminaires verticaux a eau/bromure de lithium.
Peu de chercheurs ont examine les lms ondule s, les lms turbulents, et les lms en e coulement sur les tubes horizontaux.
Des e tudes sur le couple actif ammoniac-eau sont en ge ne ral de nature empirique et/ou limite es aux lms laminaires ver-
ticaux. Cet examen est destine a souligner les domaines cle s sur lesquels il faudrait se concentrer tels que les dynamiques
des lms et de la vapeur, surtout en termes de non pe riodicite , de l'instabilite et l'eet de recirculation des ondes dans un
lm vertical, et des gouttelettes et des ondes dans le cas du tube horizontal. On doit e galement conside rer l'interaction
potentielle entre le transfert de chaleur et de masse sur les dynamiques des lms, la mouillabilite supercielle, le transfert
de chaleur dans la phase vapeur, et les e quations simplie es couramment utilise es. Enn, on doit mettre l'accent sur la
validation expe rimentale des conditions locales et les processus de transfert a l'inte rieur de l'absorbeur et non simplement
des valeurs globales de transmission. #2001 Elsevier Science Ltd and IIR. All rights reserved.
Mots cles : syste me frigorique ; syste me a absorption ; absorbeur ; lm tombant ; transfert de chaleur ; transfert de masse ;
mode lisation - enque te
Nomenclature
a,b constants in linear equilibrium model
C
A
molar concentration of species A
C
P
mass specic heat
D
AB
binary uid mass diusivity
h
a
heat of absorption
i mass specic enthalpy
k thermal conductivity
Le Lewis number=D
AB
/o
m
.
A
. m
.
B
mass ux of species A, B
Pe Peclet number=RePr/4=uo/o
Pr Prandtl number=v/o
Re Reynolds number=4uo/v, (varies by author)
Sc Schmidt number=v/D
AB
t time
T temperature
u stream-wise velocity
v transverse velocity
x stream-wise coordinate
X
A
mole fraction of species A
y transverse coordinate
Greek symbols
o thermal diusivity=k/,C
P
o lm thickness
c turbulent eddy diusivity
v kinematic viscosity
, mass density
Subscripts/superscripts
A, B species A, or B
e equilibrium
i interface
y transverse direction
0 inlet
indicates mean value
Contents
1. Introduction
2. Problem description
3. Organization
4. Nonvolatile absorbent
4.1. Laminar lm ow assumptions for vertical or horizontal tubes
4.1.1. Nakoryakov and Grigor'eva 19771997 [17,1926]
4.1.2. Kholpanov, Malyusov, and Zhavoronkov 1982 [33]
4.1.3. Grossman 1983, 1987 [14,29]
4.1.4. Andberg and Vliet 19831987 [3741]
4.1.5. Le Go, Ramadane, Barkaoui, Chen, and Le Go 1986 [48]
4.1.6. Kawae, Shigechi, Kanemaru, and Yamada 1989 [51]
4.1.7. Brauner, Moalem Maron, and Meyerson 1989, 1991 [5254]
756 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
4.1.8. van der Wekken and Wassenaar 1988 [55]
4.1.9. Habib and Wood 1990 [32]
4.1.10. Yang and Wood 1992 [57]
4.1.11. Hajji and Worek 1992 [58]
4.1.12. Ramadane, Aoufoussi, Le Go [60]
4.1.13. Conlisk 19921995 [6166]
4.1.14. Ibrahim and Vinnicombe 1993 [31]
4.1.15. Choudhury, Hisajima, Ohuchi, Nishiguchi, Fukushima, and Sakaguchi 1993 [69]
4.1.16. Lu, Li, Li, and Yu-Chi 1996 [70]
4.1.17. Jernqvist and Kockum 1996 [71]
4.1.18. Conlisk and Mao 1996 [72]
4.2. Laminar lm ow for other absorber geometries
4.2.1. Conlisk 1994b [64]
4.2.2. Yang and Jou 1995 [75]
4.2.3. Conlisk 1996a, 1996b [76,77]
4.2.4. Summary of laminar water-lithium bromide lms
4.3. Turbulent lm ow
4.3.1. Grossman and Heath 1984, 1986 [79,80]
4.3.2. Faghri and Seban 1988, 1989 [84,85]
4.3.3. Yu ksel and Schlu nder 1987 [86]
4.3.4. Kholpanov and Kenig 1993 [91]
4.3.5. Summary of turbulent water-lithium bromide lms
4.4. Wavy lm ow
4.4.1. Burdukov et al. 1980, Nakoryakov et al. 1982a, 1982b [21,22,99]
4.4.2. Kholpanov 1987, 1990 and Tsvelodub 1989 [103105]
4.4.3. Uddholm and Setterwall 1988 [106]
4.4.4. Morioka and Kiyota 1991 [109]
4.4.5. Yang, Jou, and Chen 1991, Yang and Jou [111,112]
4.4.6. Jernqvist and Kockum 1996 [71]
4.4.7. Sabir, Suen, and Vinnecombe 1996 [114]
4.4.8. Patnaik and Perez-Blanco 1996b [28]
4.4.9. Other wavy lm models for uncoupled transport
4.4.10. Summary of wavy water-lithium bromide lms
4.5. Assumed transport coecients
4.5.1. Wassenaar and Westra 1992 [30]
4.5.2. Patnaik, Perez-Blanco, and Ryan 1993 [68]
4.5.3. Patnaik and Perez-Blanco 1993 [118]
4.5.4. Kirby and Perez-Blanco 1994 [119]
4.5.5. Tsai and Perez-Blanco 1998 [123]
5. Volatile absorbent
5.1. Neglected vapor-phase mass transfer resistance
5.1.1. Ruhemann 1947 [16]
5.1.2. Briggs 1971 [129]
5.1.3. Stagnant lm models 19531975
5.2. Neglected liquid-phase mass transfer resistance
5.2.1. Colburn and Drew 1937 [134]
5.2.2. Price and Bell 1973
5.2.3. Kang and Christensen et al. 19931997 [125,136,141143]
5.2.4. Garrabrant and Christensen 1997 [147]
5.2.5. Garimella 1999 [127]
5.3. Mass transfer in both phases
5.3.1. Kholpanov, Kenig, and Malyusov 1985, [128,152]
5.3.2. Perez-Blanco 1988 [124]
5.3.3. Conlisk 1994b and Conlisk and Mao 1996 [64,72]
5.3.4. Potnis, Gomezplata, Papar, Anand, and Erickson 1997 [153]
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 757
1. Introduction
The absorber found in absorption heat pumps is
widely acknowledged as the most critical part of a heat
pump system, both in terms of cycle performance and
system cost. Consequently, the complex heat and mass
transfer phenomena occurring in the absorber have been
the subject of a great amount of research, especially
within the last 25 years. Creating useful mathematical
models of the absorber has proven to be a dicult
challenge due to the complex and coupled nature of the
transport phenomena. As absorption occurs, heat and
mass are transferred through and between the liquid and
vapor phases. The driving forces for these transport
phenomena change as the process progresses both due
to changes in the local temperature and concentration
gradients and due to changes in the liquid-vapor inter-
face equilibrium condition. The governing equations are
thus mathematically coupled. In addition, the hydro-
dynamics within the absorber have a profound eect on
its performance; for example, turbulence, droplet phe-
nomena, and wave formation all have the eect of
redistributing the species and energy within the absor-
ber. In order to solve the transport equations, a detailed
understanding of two-phase, multi-component ow in
complex geometries is also required.
The objective of this paper is to review the signicant
eorts that have been made to develop mathematical
models of the coupled heat and mass transfer accom-
panying falling-lm absorption in heat-driven heat
pumps. The focus of this paper is primarily on analy-
tical and numerical models and excludes experimental
work in this area. The most common absorber geome-
tries involve a falling lm because of its potential for
high heat and mass transfer rates with minimal pressure
drops; bubble, spray and many other less widely used
geometries are not included in this review. The two most
common working uid pairs, water-lithium bromide
and ammonia-water, are considered; many other work-
ing uid pairs can be modeled using the solution meth-
ods established for these two pairs. The enhancing
eects of chemical reactions sometimes encountered
with less common working uids is not included in this
review. Additionally, models which do not simulta-
neously consider the heat and mass transfer processes,
such as the case of isothermal absorption where the heat
eect of absorption is negligible, are not applicable to
this review. Finally, additives (surfactants) used to
enhance the transfer rates in absorbers, surface tension-
driven Marangoni instabilities, Rayleigh instabilities
arising from buoyancy forces, and the eect of system
contaminants such as non-absorbable gases are also not
included in this review. The eect of additives used to
increase mass transfer coecients and alter the lm
hydrodynamics is studied by Michel and Perez-Blanco,
Jung et al., Kim et al., Ziegler and Grossman and
Kulankara et al. [15]. Non-absorbable contaminants
are studied using mathematical models by Kenig et al.,
Yang and Chen, Chen and Vliet, Ameel et al., Gross-
man and Gommed and Yang and Jou [611].
2. Problem description
Figs. 14 show typical schematics of the solution
domains for absorption on falling lms for the large
range of lm Reynolds numbers that may be used in
absorbers. The lm ow direction is generally down-
ward; however the vapor ow and coolant ow direc-
tions depend on the design of the absorber. Typically,
the coolant is in counter- or cross-ow with the solution
for vertical and horizontal tubes, respectively. The
vapor migrates toward the liquid, across the interface
and into the lm; in addition it may be in general co- or
counter-ow with the lm (in Figs. 14, coolant and
vapor are shown in counter-ow with the lm).
Heat and mass transfer for a dierential control
volume can be described by the following conservative
equations (in two spatial dimensions and one temporal
dimension):
,
oi
ot
u
oi
ox
v
oi
oy

=
o
ox
k
oT
ox

o
oy
k
oT
oy

(1)
5.3.5. Kim 1998 [126]
5.3.6. Gommed, Grossman, and Koenig 1999, 2000 [15,154]
5.3.7. Summary of ammonia-water models
6. Conclusions
7. Uncited references
Acknowledgements
References
758 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
oC
A
ot
u
oC
A
ox
v
oC
A
oy

=
o
ox
D
AB
oC
A
ox

o
oy
D
AB
oC
A
oy

(2)
assuming only two species, A and B, exist within the
system and the mass densities of the two phases are
constant [12]. The above equations also assume negli-
gible dissipation eects, pressure gradients, transport of
mass due to energy uxes (thermal diusion or Soret
eect, see [12]), transport of energy due to mass uxes
(Dufour eect, see [12,13]), and transport of energy due
to interdiusion [14,15].
The coupling between these equations is realized
through two boundary conditions at the interface
between the liquid lm and the vapor. First, in almost
all cases, investigators assume that vapor pressure equi-
librium prevails at the interface; that is, whenever con-
tact is established between the liquid and the vapor, the
concentrations at the interface adjust such that the
vapor pressure of the each species in the lm at the
interface is equal to the partial pressure of each species
in the vapor at the interface. The dierences between the
equilibrium concentrations at the interface and the bulk
concentrations in the two phases induce the concentra-
tion gradients that drive the mass transfer processes.
These concentration dierences are noted in Fig. 2.
Fig. 1. Typical schematic for smooth laminar lm (Re<20
approximate Re values from Patnaik and Perez-Blanco, [116]).
Fig. 1. Sche ma typique d'un lm laminaire lisse (Re < 20
valeurs Re approximatives d'apre s Patnaik et Perez-Blanco,
[116]).
Fig. 2. Typical schematic for smooth turbulent lm (4000<
Re approximate Re values from Patnaik and Perez-Blanco,
[116]).
Fig. 2. Sche ma typique d'un lm turbulent lisse (4000 < Re
valeurs Re approximatives d'apre s Patnaik et Perez-Blanco,
[116]).
Fig. 3. Typical schematic for lm with capillary waves (20<
Re<200 approximate Re values from Patnaik and Perez-
Blanco, [116]).
Fig. 3. Sche ma typique d'un lm avec ondes capillaire (20 < Re
< 200 valeurs Re approximatives d'apre s Patnaik et Perez-
Blanco, [116]).
Fig. 4. Typical schematic for lm with roll waves (200<
Re<4000 approximate Re values from Patnaik and Perez-
Blanco, [116]).
Fig. 4. Sche ma typique a ondes en rouleaux (200 < Re < 4000
valeurs Re approximatives d'apre s Patnaik et Perez-Blanco,
[116]).
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 759
The second coupling boundary condition is an energy
balance at the interface. As mass is transported across
the interface, it must change phase and join a mixture
with dierent relative concentrations. The change in
enthalpy from the vapor state to the solution state
(generally called the heat of absorption) is released at
the interface causing the interface temperature to rise.
This generates temperature gradients in both the solu-
tion and vapor which drive the heat transfer processes;
this can be seen in Fig. 1. In addition, the temperature
of the interface aects the interface concentration
through the chemical equilibrium condition; as the tem-
perature increases, the gradients driving the mass trans-
fer processes decrease. Thus, the heat and mass transfer
processes are inextricably linked. Summarizing this in
1947, Ruheman wrote ``it is not possible to calculate an
absorber completely without taking both transfer coef-
cients into account, however dierent their initial roles
. . . may appear to be at the outset'' [16].
The earliest eorts to model absorption in falling
lms that accounted for the coupling of the heat and
mass transfer processes, essential where the heat of
absorption is signicant as in absorption heat pumps,
concentrated on the highly idealized situation of water
vapor in contact with an aqueous lithium bromide lm
of uniform thickness with uniform stream-wise velocity
on an isothermal vertical plate [17]. Subsequently,
investigators have attempted to include more realistic
ow characteristics in their models such as parabolic
velocity proles for laminar lms, turbulent eddy pro-
les for turbulent lms, and transient lm thickness and
velocity proles for wavy-laminar lms; however, even
the most sophisticated of these models still includes
many simplifying assumptions as will be discussed in
detail below. Nevertheless, signicant work has been
done in this area which provides a strong basis for con-
tinued development.
3. Organization
The following sections are categorized rst according
to working uid pair, then by assumption of the ow
regime. This particular assumption has a preeminent
eect on any modeling results because the temperature
and concentration gradients that drive the heat and
mass transfer are so strongly aected by the uid
motion within the lm and vapor. Within each section,
papers are ordered chronologically by date of rst pub-
lication, although, for clarity, multiple papers by the
same author(s) may be grouped together. For each
model reviewed, an attempt has been made to summar-
ize the governing equations, boundary conditions and
other key assumptions utilized. The solution techniques
and the results of the models are compared whenever
possible with emphasis on model validation. Although it
is the intention of the present study to highlight as many
of the relevant and signicant contributions in this area
as possible, it is likely that the listing here is not com-
prehensive. For instance, publications from journals
that are not translated into English are not presented
here.
4. Nonvolatile absorbent
Many absorption heat pumps utilize water as the
refrigerant and a salt such as lithium bromide as the
absorbent. Water-lithium bromide has been used in
absorption heat pumps since the 1950s [18]. Because
lithium bromide is essentially nonvolatile, the vapor is
pure in the absence of non-absorbable gases and so
there is no resistance to mass transfer in the vapor
phase. In contrast, due to the volatility of both compo-
nents in an ammonia-water system, both ammonia and
water may be present in the vapor phase and thus the
potential for mass transfer resistance in the vapor phase
exists.
4.1. Laminar lm ow assumptions for vertical or
horizontal tubes
The earliest modeling attempts assumed laminar lms
and, of the investigations covered by this paper, the
majority fall under this category. This is due to the fact
that this assumption allows the most simplied models
to be developed without assuming transport coecients
a priori, a common technique that is reviewed in sub-
sequent sections of this paper. However, as will be dis-
cussed in detail later, the assumption of laminar ow is
dicult to justify because of the inherent instabilities of
a falling lm even at very low Reynolds numbers.
Nevertheless, it is important to understand how the
laminar lm problem has been solved before examining
the attempts at incorporating the more complex hydro-
dynamics.
4.1.1. Nakoryakov and Grigor'eva 19771997 [17,1926]
Nakoryakov and Grigor'eva [17,26] take a signicant
step forward in modeling the absorption process. They
consider the case of steady absorption in a smooth
laminar lm falling down an isothermal, impermeable
vertical plate. The list of assumptions they employ is
quite long, but many are still current practice and will
be enumerated here. In order to simplify the problem,
the following assumptions are made:
. Vapor pressure equilibrium is instantaneously
established at the liquid-vapor interface and pre-
vails along the entire interface. This chemical
equilibrium assumption establishes the interface
concentrations via continuity of the vapor pressure
760 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
of each species at the interface temperature, which
is analogous to the enforcement of continuous
temperature throughout the two contacting phases.
. All thermophysical properties of the liquid and
vapor can be considered constant. That is, the
eect of temperature and concentration on prop-
erties such as thermal conductivity, specic heat,
density, viscosity, diusivity, and the heat of
absorption is negligible. This assumption allows,
among other things, the terms D
AB
and k in the
governing partial dierential equations to be fac-
tored out of the dierential terms. Eqs. (1) and (2)
can, therefore, be simplied to the following form:
u
oT
ox
v
oT
oy
= o
o
2
T
ox
2
o
o
2
T
oy
2
(3)
u
oC
A
ox
v
oC
A
oy
= D
AB
o
2
C
A
ox
2
D
AB
o
2
C
A
oy
2
(4)
. Diusion and thermal conduction in the stream-
wise direction are negligible. This common
assumption is well justied by comparing the rate
of diusion/conduction to the velocity in the
stream-wise direction. This yields the following
simplications:
u
oT
ox
v
oT
oy
= o
o
2
T
oy
2
(5)
u
oC
A
ox
v
oC
A
oy
= D
AB
o
2
C
A
oy
2
(6)
. Due to the low velocity associated with absorp-
tion, the transverse velocity, v, is negligible,
therefore:
u
oT
ox
= o
o
2
T
oy
2
(7)
u
oC
A
ox
= D
AB
o
2
C
A
oy
2
(8)
. Also, because the anticipated absorption rates are
low compared to the solution ow rate, changes in
the lm thickness are negligible.
. Either the net mass ux in the transverse direction
is zero or the concentration of water in the solu-
tion is very low. Fick's rst law of binary diusion
can be written:
m
.
A
C
A
m
.
A
m
.
B
( ) = ,D
AB
oC
A
oy
(9)
where component A corresponds to lithium bro-
mide and B to water [12]. However, close exam-
ination of the equations reveals that Nakoryakov
and Grigor'eva apparently assume:
m
.
B
= ,D
AB
oC
A
oy
(10)
which is only true if m
.
A
= m
.
B
or if C
A
=1 and
C
B
=0. If m
.
A
= m
.
B
, there is equal and opposite
mass ux of lithium bromide and water across the
interface, often described as counter- or equimolar-
diusion. If C
A
=1 and C
B
=0, the lm is com-
posed of pure lithium bromide and innitely dilute
in water. Neither of these assumptions is well jus-
tied given the solubility limits and low volatility
of LiBr. Because of their equivalence, the terms
innite water dilution and equimolar diusion will be
used interchangeably in this review also see sections
4.1.7 [5254], and 4.4.8 [28] for further discussion.
. The velocity in the laminar lm is uniform. Also
see section 4.1.3 [29] for further discussion.
. The pressure is constant at all points within the
absorber; consequently the interface equilibrium
concentration in the lm is purely a function of
interface temperature. For the range of tempera-
tures encountered, this relationship is approxi-
mately linear: C
i
=aT
i
+b; where a and b are
constants also see sections 4.1.3 [29] and section
4.1.4 [31] for further discussion.
. Heat transfer to the vapor is negligible; all the
energy released during absorption goes into heat-
ing the liquid lm also see section 4.1.9 [32] for
further discussion. This assumption is question-
able since the vapor enters the absorber at a much
lower temperature than the solution; however,
given the low heat capacity and typically low heat
transfer coecients of the vapor phase, the eect
of this assumption on the overall results is prob-
ably small.
. The inlet solution concentration and temperature
proles are uniform.
Other ancillary assumptions include complete wetting
of the isothermal wall, no shear or surface tension
eects at the solution-vapor interface, no buoyancy,
natural convection or inertia forces, and no slip or con-
centration gradients at the wall.
Using these simplifying assumptions, Grigor'eva and
Nakoryakov [26] provide a solution using Fourier
separation of variables techniques. Evaluation of the
solutions requires the determination of a large number
(4050) of eigenvalues for the series solution to con-
verge. Nevertheless, expressions for temperature and
concentration throughout the vertical lm, heat transfer
at the wall and solution-vapor interface, and average
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 761
lm temperature and concentration are presented. An
interesting conclusion of their work [26] is that besides
the boundary conditions (solution inlet concentration
and temperature, wall temperature, and the equilibrium
constants), the solution for this particular case depends
only on four non-dimensional parameters:
. the Lewis number, Le=D
AB
/o, which is the ratio
of mass diusivity to thermal diusivity,
. the Prandtl number, Pr=v/o, which is the ratio of
kinematic viscosity to thermal diusivity,
. the Reynolds number, Re=4uo/v, where o is the
lm thickness, or Peclet number, Pe=RePr/4, and
. the dimensionless group h
a
a/C
P
, where h
a
is the
heat of absorption, a is a constant from the equi-
librium relation (dimension T
1
), and C
P
is the
solution specic heat.
Noting the complexity of using the resulting analy-
tical solution, Nakoryakov and Grigor'eva [17] also
develop approximate solutions. The basis for this
method is the use of two more simplifying assumptions.
First, it is assumed that the temperature prole in the
lm is linear, i.e. conduction dominates convection or
o
2
T,oy
2
= 0. Second, the concentration prole is
assumed based on an approximate boundary layer
solution. As a result, much simpler expressions for the
interface and average temperatures and concentrations,
and heat uxes are developed. These solutions are only
valid where the preceding two assumptions hold, which
is generally far downstream from the inlet, but not so
far that the concentration proles deviate from the
boundary layer solutions. A comparison of these solu-
tions with the exact analytical solutions is presented as a
function of non-dimensional downstream position in
Grigor'eva and Nakoryakov [26]. It can be seen that
local absorption rates predicted using the approximate
method are generally lower than those from the exact
solution, and the predicted interface conditions become
increasingly divergent closer to the lm inlet.
Nakoryakov and Grigor'eva [19,20] also develop a
simplied solution that is appropriate for the region
near the inlets. The assumptions required are estimates
of the thicknesses of the developing thermal and con-
centration boundary layers. A slight modication to the
assumed velocity prole is also introduced. Within the
thermal boundary layer developing from the isothermal
wall, it is assumed that the velocity prole varies line-
arly. Within the thermal and concentration boundary
layers developing from the solution-vapor interface, the
velocity prole is considered uniform. Within these
developing boundary layers, similarity solutions for the
temperature and concentration proles are given. With
these assumptions Nakaryakov and Grigor'eva develop
expressions for temperatures, concentrations, and heat
and mass uxes throughout the lm from the inlet to the
point where the growing boundary layers meet. An
expression for this length is also given, which, according
to Nakoryakov and Grigor'eva, is about 75 times the
lm thickness for typical operating conditions inside a
water-lithium bromide absorber. Beyond this length, the
thermal boundary layers from the interface and wall
meet and the temperature prole becomes pre-
dominantly linear allowing the use of the approximate
solutions presented in Nakoryakov and Grigor'eva [17].
It is interesting to note that, within the developing
boundary layer region, the interface temperature and
concentration remain constant at the equilibrium values
derived from the inlet conditions.
Nakoryakov and Grigor'eva [23] compare the inter-
face temperatures and concentrations predicted by the
previously developed analytical model of Grigor'eva
and Nakoryakov [26] and the approximate solutions of
Nakoryakov and Grigor'eva [17,19,20]. Additionally,
they compare their results to those of Grossman [29]
which will be discussed in section 4.1.3 but are essen-
tially an extension of the problem to include a fully
developed parabolic velocity prole in the lm. In the
initial region, all three solutions have the same interface
temperature and concentration because the cooling
eect of the wall does not reach the interface; the inter-
face conditions are determined by the inlet conditions.
Just after the thermal boundary layers meet, the diver-
gence between the solutions becomes the greatest.
Finally, all three solutions converge to the same asymp-
totic value downstream. Although the dierences are
not quantied, Nakoryakov and Grigor'eva [23] claim
``the comparison of these results with the exact solutions
make it clear that for engineering calculations these
simple formulae are quite acceptable.''
Nakoryakov and Grigor'eva [24] relax the assump-
tion that lm thickness is constant and allow it to vary
with absorbed mass. The problem is again solved in two
parts: the developing boundary layer region and the
region with a fully developed temperature prole. In the
initial region, the solution remains almost unchanged;
the interface temperature and consequently concentra-
tion remain constant while the thermal boundary layers
are developing. However, the absorption rate is highest
in this region which leads to the highest rate of lm
thickness growth. Even so, according to the authors, the
eect of the lm growth on the transport processes is
small since the lm thickness only changes by a few
percent. An expression for the lm thickness in this
region is presented based on the similarity solutions
developed. For the positions outside this initial region,
it is assumed that the temperature prole is approxi-
mately linear and an expression for interface concentra-
tion as a function of downstream position is developed
along with an iterative method for calculating the lm
thickness. It is shown that typical lm thickness growth
is small (<10%) even for relatively large downstream
distances. The dierences in interface concentration
762 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
predicted using the approximate solutions with and
without variable lm thickness are shown to be quite
small.
Nakoryakov et al. [25] develop an expansion series
solution for the problem considered in [17] but for both
an isothermal and adiabatic wall. In addition, more
detailed expressions for nding eigenvalues are pre-
sented for each case. The authors present comparisons
of the interface and wall temperatures and concentra-
tions predicted by these solutions and those predicted by
Grossman [29] for a parabolic velocity prole (see Fig. 5
for non-dimensional interface conditions versus non-
dimensional downstream position for both the adiabatic
and isothermal wall case). It is again noted that the
initial and asymptotic values of all solutions are the
same; for the adiabatic wall, the asymptotic values of
concentration depend on the dimensionless group h
a
a/
C
P
.
4.1.2. Kholpanov, Malyusov and Zhavoronkov 1982
[33]
Kholpanov et al. [33] present an analysis similar to
that developed by Nakoryakov and Grigor'eva [19,20]
but allow for the eect of tangential shear stress or a
surface tension gradient at the solution-vapor interface.
The solution is only valid for regions near the beginning
of the absorption process where the thermal and con-
centration boundary layers developing from the inter-
face do not reach the wall since the transfer of heat and
mass from the interface into the lm is assumed to be
conned to a thin boundary layer. They also assume
that the velocity prole within this thin boundary is
uniform, even in the presence of shear and surface ten-
sion gradients. The authors develop expressions for
concentration and temperature within the lm that they
state reduce to the inlet region approximations developed
by Nakoryakov and Grigor'eva [19,20] for the case of
no interface shear or surface tension gradients. Addi-
tionally, local and mean heat and mass uxes are also
developed from these expressions.
4.1.3. Grossman 1983, 1987 [14,29]
Grossman [29] uses essentially the same simplifying
assumptions found in Grigor'eva and Nakoryakov [26].
He solves the equations:
u
oT
ox
= o
o
2
T
oy
2
(11)
u
oC
A
ox
= D
AB
o
2
C
A
oy
2
(12)
for a constant thickness lm falling down a vertical
adiabatic or isothermal wall. The main dierence is that
Grossman assumes the fully developed, laminar, Nusselt
solution for the velocity prole:
u(y) =
3
2
u 2
y
o

y
o

2

(13)
where u is the average stream-wise velocity. Another
condition Grossman uses is that, for the isothermal
case, the temperature of the lm at the inlet is equal to
the wall temperature. Therefore no thermal boundary
layer develops from the wall, and the only boundary
layer development is inward from the liquid-vapor
interface.
Grossman gives some additional justication for two
of the assumptions he shares with Grigor'eva and
Nakoryakov. He cites justication for using a linearized
model for the equilibrium relationship between interface
concentration and temperature in a prior work [34] and
states ``the validity of the linear relation was checked for
two common absorbents, LiBrH
2
O and LiClH
2
O,
and found to be very good under the above limitations
for a wide range of temperatures and concentrations.''
He also points out that the assumption of constant heat
Fig. 5. Dimensionless interface temperature,
i
=(T
i
T
0
)/
(T
e
T
0
), and concentration, ,
i
=(C
i
C
0
)/(C
e
C
0
), versus
normalized length, =x/(Peo) for Le=0.0001 (1),=0.001
(2),=0.01 (3) and (Leh
+
a
(C
e
C
0
))/(C
p
*(T
e
T
0
))=(Leh
a
a)/
C
p
=0.01. T
e
corresponds to equilibrium interface temperature
at inlet concentration, C
0
; C
e
corresponds to equilibrium inter-
face concentration at inlet temperature, T
0
. Dash-double-dot
lines correspond to an adiabatic wall, dash-dot lines to iso-
thermal wall. Dashed line is adiabatic wall prediction from
Grossman [29], solid line is isothermal wall prediction from
Grossman [29] (adapted from Nakoryakov and Grigor'eva [25]).
Fig. 5. Tempe rature a l'interface adimensionnelle
i
= (T
i
T
0
)/
(T
e
T
0
), et la concentration, g
i
= (C
i
C
0
)/(C
e
C
0
) en
fonction de la longueur normalise e, ..... = x/Ped) pour Le =
0,0001 (1), = 0,001 (2), = 0,01 (3) et (Leh*
a
(C
e
C
0
)/
(C
p
*(T
e
T
0
))= (Leh
a
a)/Cp = 0,01. T
e
correspond a la tem-
pe rature a l'e quilibre a l'interface a la concentration d'entre e, C
0
;
C
e
correspond a la tempe rature a l'e quilibre a la concentration
d'entre e, T
0
. Les courbes en hachure /deux points repre sentent un
mur adiabatique, et la courbe en hachure /un point repre sente un
mur isothermique. La courbe hachure e repre sente une pre vision
d'un mur adiabatique selon Grossman [29] et la courbe continue
repre sente une pre vision d'un mur isothermique selon Grossman
[29] (adapte a partir des travaux de Nakoryakov et Grigor'eva
[25]).
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 763
of absorption is quite reasonable because its dependence
on concentration and temperature is ``very weak.'' This
is presumably because the phase-change enthalpy is the
dominant term in the heat of absorption.
Grossman uses two methods to solve the problem.
First, using the Fourier method as in Grigor'eva and
Nakoryakov [26], he seeks a series expansion solution.
He discusses extensively the steps required for deter-
mining the eigenvalues for both the adiabatic and iso-
thermal wall case and gives an example of the
eigenvalues computed for the adiabatic and isothermal
wall cases with typical values for the relevant para-
meters. He too points out that in some instances, parti-
cularly near the solution inlet, a very large number of
eigenvalues is required for series convergence. He notes
that the assumptions of uniform inlet concentration
prole and interface equilibrium together are contra-
dictory and therefore form a discontinuity at the lm
interface at the inlet.
Second, Grossman uses a numerical technique based
on nite dierence methods. Again, the discontinuity at
the inlet causes diculty in obtaining a solution. Con-
sequently, he develops a similarity solution for the con-
centration and temperature within the lm for the range
where the thermal eect of the wall does not reach the
interface (constant interface temperature and con-
centration) similar to Nakoryakov and Grigor'eva
[19,20]. He proposes that this solution can be used near
the inlet where both the numerical and analytical solu-
tions are dicult to obtain, and either the analytical or
numerical solution can be used in the downstream
region. In his review of heat and mass transfer in falling
lms, Grossman [35] also points out that in this region,
the situation ``is adequately described by Higbie's pene-
tration theory, as conrmed experimentally by Emmert
and Pigford [36].''
The results of the numerical and analytical solutions
are reported to be ``in excellent agreement.'' Grossman
presents the results of his model in several ways. Inter-
face, bulk and wall temperatures and concentrations are
plotted against downstream position for the adiabatic
and isothermal cases. Proles of concentration and
temperature through the lm are presented at a number
of positions. A comparison between the non-dimen-
sional interface temperature and concentration pre-
dicted by his solution and that of Nakoryakov and
Grigor'eva can be seen in Fig. 6; again the inlet region
and asymptotic values agree well, but the values in the
intermediate range show discrepancies. According to
Grossman the assumption of uniform velocity prole
leads to a 40% underprediction of the length required to
achieve a certain temperature or concentration level (i.e.
degree of concentration boundary-layer development) and
a 20% deviation in predicted heat and mass transfer coe-
cients. Grossman also presents results with several dierent
values for the heat of absorption and Lewis number.
Grossman [14] uses essentially the same assumptions
as Grossman [29] except that the term for change in
energy due to interdiusion is retained in the governing
energy equation. Solutions are developed for a region
near the inlet using the similarity technique as in
Nakoryakov and Grigor'eva [19,20]; the stream-wise
velocity is assumed to be uniform since the heat and mass
transfer processes in this region are restricted to a thin
layer near the interface. Several assumptions are made
about the sensitivity of the liquid enthalpy, density, and
specic heat to temperature and concentration. The
similarity solutions developed depend on a constant that is
determined by numerically evaluating an integral expres-
sion based on the boundary conditions. Expressions for
non-dimensional heat and mass transfer coecients are
developed along with a deviation factor for quantifying
the impact of interdiusion on heat transfer. A non-
dimensional group involving the product of the mass
transfer driving force in the liquid and the sensitivity of
volumetric specic heat to concentration is introduced.
It is shown that the deviation factor is nearly propor-
tional to this non-dimensional group. The sensitivity of
the deviation factor to this non-dimensional group
depends on the heat of absorption and the Lewis num-
ber. Grossman notes that as the non-dimensional heat
of absorption is increased, or the Lewis number
reduced, the magnitude of the deviation decreases. For
low Lewis numbers, typical of LiBr systems, the devia-
tion factor is only a few percent.
4.1.4. Andberg and Vliet 19831987 [3741]
Andberg and Vliet [37] utilize a nite dierence for-
mulation to solve the absorption problem for a laminar
lm falling down a vertical, isothermal wall. Although
most of the assumptions are the same as those used by
the investigators mentioned above, there are several
noteworthy dierences. First, the change in energy due
to interdiusion is not neglected. They also assume a
fully developed, parabolic stream-wise velocity prole
and allow the lm thickness, ow-rate and velocity to
increase as vapor is absorbed. A coordinate transfor-
mation is utilized to t a ``rectangular'' solution domain
to the thickening lm. Interface equilibrium is assumed,
but the formulation for this is not linearized as was the
case in Grigor'eva and Nakoryakov [26] and Grossman
[29]. Instead, the relationship given by McNeely [42] is
used. Andberg and Vliet also give further justication
for the assumption that there is no heat transferred from
the liquid to the vapor by noting the large dierence in
the thermal conductivity of the two phases and the fact
that the vapor phase generally moves toward the inter-
face as absorption occurs.
Heat and mass uxes, lm mass ow rate, interface
and bulk concentrations and temperatures are plotted
against downstream position. Temperature and con-
centration proles are presented at a variety of locations.
764 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
Additionally, comparisons are made with the experi-
mental data of Burdukov et al. [43]. Generally con-
centrations and temperatures agree within experimental
error except close to the solution inlet where the pre-
dicted heat and mass transfer rates are the highest. They
suggest that the isothermal wall assumption may be
inappropriate in this region: ``in reality, the heat transfer
resistance to the cooling water prevents this high heat
ux, and thus the absorption and the change in con-
centration near the inlet are much more gradual'' than
predicted.
Andberg and Vliet [38] utilize their earlier model [37]
to develop a relationship for absorber length based on
desired concentration change and operating conditions
such as solution mass ow rate, inlet concentration,
inlet temperature and wall temperature. By dening
absorption percentage as the predicted concentration
change compared to the concentration change that
could be achieved given an innitely long isothermal
wall, they nd that the only variable which has a sig-
nicant eect on the required length for a desired
absorption percentage is solution mass ow rate.
Andberg and Vliet [37] and Andberg [39] are perhaps
the rst to develop a specialized solution for the hor-
izontal tube case. They extend the numerical techniques
utilized in the above work to the case of a bank of tubes
using a coordinate system t to the shape of the lm
around the tube. The hydrodynamics of the solution
owing over the tubes is treated in great detail but they
still make some key assumptions. First, the ow
between successive tubes is assumed to be a planar jet
(sheet) that falls at the speed resulting from free-fall
between the tubes which does not contribute to the
absorption process. Although the authors admit that
drop-wise ow or isolated columnar jets between the
tubes would be more likely at the solution ow rate
considered, they suggest that the planar jet is the best
approximation that allows a 2-D solution to the pro-
blem. Second, it is assumed that within the planar jet,
the solution is well mixed, resulting in uniform tem-
perature and concentration within the lm at the top of
each tube. Third, to solve the hydrodynamics of the
region where the planar jet transitions to the horizontal
tube surface, boundary layer approximations of the
Fig. 6. Dimensionless interface temperature,
i
=(T
i
T
0
)/(T
e
T
0
), and concentration, ,
i
=(C
i
C
0
)/(C
e
C
0
), versus normalized
length, =x/(Peo) for Le = 0.001 and normalized heat of absorption, l = (Leh
+
a
(C
e
C
0
)),(,C
+
p
(T
e
T
0
)), as indicated. T
e
corresponds
to equilibrium interface temperature at inlet concentration, C
0
; C
e
corresponds to equilibrium interface concentration at inlet tem-
perature, T
0
. Dashed lines correspond to an adiabatic wall, solid lines to isothermal wall. Lines labeled ``slug ow'' solution are from
Grigor'eva and Nakoryakov [26] (adapted from Grossman [29]).
Fig. 6. Tempe rature a l'interface adimensionnelle y
i
= (T
i
T
0
) et concentration g
i
= (C
i
C
0
)/(C
e
C
0
) en fonction de la longueur .....
z= x/(Ped) pour Le = 0,001 et la chaleur d'absorption normalise e l = (Leh*
a
(C
e
C
0
))/(C*p(T
e
T
0
), comme indique . T
e
correspond a
la tempe rature a l'e quilibre a l'interface a la concentration d'entre e, C
0
; C
e
correspond a la concentration a l'e quilibre a l'interface a la
tempe rature a l'entre e, T
0
; les courbes hachure es repre sentent un mur adiabatique et les courbes continues repre sentent un mur iso-
thermique. Les courbes de signe es slug ow solution sont fonde es sur celles de Grigor'eva et Nakoryakov [26] (adapte es des travaux de
Grossman [29]).
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 765
NavierStokes equations are applied assuming constant
properties; it is noted that this turning-jet region
accounts for less that 0.1% of the entire solution
domain. The thermophysical properties in all other
regions depend on the local composition and tempera-
ture within the lm, although boundary layer forms of
the NavierStokes equations are still used. Another
change justied by the results of their vertical wall
model is that the authors neglect the transport of energy
due to interdiusion. Also, the assumption of innite
dilution of water in the lm is eliminated by using the
expression:
m
.
B
=
,D
AB
C
A
oC
A
oy
(14)
Andberg [39] describes the nite dierence formulations
in great detail in his dissertation. Also, a model for the
top tube is presented in which the liquid addition is by a
falling mist landing on the top half of the tube, as
opposed to a localized columnar jet. Results from the
models are given for one operating condition with three
dierent solution inlet temperatures (46, 39, and 32

C).
The tubes have a 19-mm (0.75-inch) OD. The heat
transfer model extends through the wall and to a cool-
ing medium (water) with a constant temperature (30

C)
and prescribed heat transfer coecient. The develop-
ment of velocity, temperature, and concentration pro-
les are presented along with interface, bulk and wall
temperatures and concentrations, and heat and mass
uxes around the tubes. It is noted that the hydro-
dynamics are quite similar to the classical Nusselt solu-
tion and the concentration proles all generally exhibit
the same characteristic shape. The eect of inlet tem-
perature on absorption rate is shown to be signicant.
At the lowest inlet temperature, the absorption rate near
the inlet is approximately an order of magnitude greater
than at the highest temperature. The total absorption
rate over the whole tube is approximately 80% greater.
It is shown that at the highest inlet temperature, sig-
nicant absorption does not occur until approximately
one third of the way around the tube.
Andberg and Vliet [41] and Andberg [39] suggest a
simplied model that retains the key characteristics of
the more complex model for horizontal tubes. The
Nusselt solution for hydrodynamics, constant thermo-
physical properties, and a similarity solution for the
concentration prole developed from the detailed model
are all used to simplify the analysis. The concentration
proles are modeled with a power-law approximation
for the error function near the liquid-vapor interface.
The authors note, as in the models above that use a
similarity solution, that there is ``good theoretical basis
for using an error function prole near the surface
where u constant.'' They state that, in terms of overall
concentration change within the lm, the simplied
model results are on average within 3% of the detailed
model with a worst case error of 10.5%. Finally, the
authors compare their results to data from commercially
available absorbers that utilize a geometry similar to
that used in the model. An average error of 6.6% is
reported based on overall concentration change. In a
subsequent publication by Cosenza and Vliet [44], in
which a large amount of experimental data was
obtained to validate the above models, fair agreement is
reported between the analytical model and experimental
data although the analytical model consistently under-
predicts the measured data. It should be noted also that
in a recent experimental study, Kim et al. [45] noted that
the values of mass diusivity used by Andberg and Vliet
and presumably in much of the literature are sig-
nicantly higher (up to a factor of 3 depending on con-
centration) than the values determined by two other
experimental studies, Kashiwagi [46] and Enderby [47].
4.1.5. Le Go, Ramadane, Barkaoui, Chen and Le Go
1986 [48]
Le Go et al. [48] give a brief summary of several
methods for modeling absorption heat and mass trans-
fer. In other works not translated into English [49,50]
they developed analytical, numerical, and approximate
models of laminar falling lm absorption over an adia-
batic wall, however these are not reviewed here. Le Go
et al. [48] discuss some of the features, limitations, and
usefulness of the dierent solution methods including
analytical and approximate methods, numerical meth-
ods with Laplace transforms and nite dierence
approximations. They also briey discuss methods for
wavy and turbulent lms. The authors also refer to
other publications (in French) in which they have uti-
lized each of the solution methods, compared them with
each other and with experimental data. One interesting
concept presented by the authors is the idea of ``tie
lines'' and ``trajectories''. A ``trajectory'' is a line on a
temperature-concentration plot representing the pro-
gression of temperature and concentration of a parti-
cular location within the lm, such as the interface, bulk
or wall, as the lm proceeds along the wall. ``Tie lines''
connect the trajectories at any given cross-section of the
lm. This is a unique plotting method not found else-
where in the literature.
4.1.6. Kawae, Shigechi, Kanemaru and Yamada 1989 [51]
Kawae et al. [51] develop a nite dierence model of
laminar, vertical falling-lm absorption which is very
similar to that of Andberg and Vliet [37]. Some of the
key features and assumptions of this model are: fully
developed parabolic stream-wise velocity prole, trans-
verse velocity equal to zero, energy equation allowing
for interdiusion, equimolar diusion in Fick's law, lm
thickness varying with absorbed mass, no vapor shear at
the interface, no heat transfer to the vapor, uniform
766 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
inlet concentration and temperature proles, and an
isothermal wall. A dierence between Kawae et al. [51]
and the formulation in Andberg and Vliet [37] is that
Kawae et al. do not assume constant thermophysical
properties (except implicitly in the assumption of lm
hydrodynamics). To speed up the calculation by avoid-
ing an iterative procedure they use an approximation
where, at each solution point, they use the temperature
and concentration calculated at an adjacent point to
estimate the thermophysical properties at the current
location. They model a baseline case where the inlet lm
is in equilibrium with the vapor and present plots of
temperatures, concentrations, uxes and lm thickness.
They show that allowing for variable properties has a
small eect on the results, especially near the inlet where
absorption is slow in this case. They nd, for instance,
that the prediction of total mass absorbed for a 10-m
tall plate is about 5% higher when assuming constant
properties.
Kawae et al. also perform several parametric studies
using the model with constant properties. They docu-
ment the eect of each parameter on total mass absor-
bed by plotting it against downstream position. The
eects of absorber pressure, wall temperature, inlet
concentration and Lewis number are shown to be
straightforward. The inlet lm thickness (or ow rate)
and temperature have more complicated eects that
change as ow length is increased. For instance,
according to their model, if the inlet temperature is
reduced, initial absorption rates are higher, but this
trend actually reverses after about 0.2 m downstream.
The complex relationship between inlet lm thickness
and total absorption rate is discussed (also considering
the wall temperature) and it is shown that for any given
plate length and wall temperature, there may be a lm
thickness that provides optimum total absorption rate.
4.1.7. Brauner, Moalem Maron and Meyerson 1989,
1991 [5254]
Brauner et al. [52] present solutions valid near the
inlet region using the similarity technique as in Nakor-
yakov and Grigor'eva [19,20]. The key dierence is that
the application of Fick's law of diusion at the interface
is formulated without assuming innite dilution in the
liquid lm. Regarding the term which was neglected in
Fick's law, Brauner et al. [53] state, ``the convective term
. . . can be omitted for either X
A
0 or N
Ay
=N
By
,
neither of which holds for the case of hygroscopic con-
densation. For instance, the minimum molar fraction of
water which corresponds to a saturated salt solution of
MgCl
2
, CaCl
2
, LiBr, NaOH, is about X
A
0.8.'' All
other assumptions are identical to Nakoryakov and
Grigor'eva [19,20].
The solutions presented are valid near the inlet of the
lm; the eect of heat transfer to the wall is not included
and the velocity in the boundary layers developing from
the liquid-vapor interface is assumed to be uniform. For
determining the range of validity of the solution, Brau-
ner et al. suggest that the uniform velocity assumption
deviates only 10% from a parabolic prole for the third
of the lm closest to the interface. Thus, to quantify the
length over which their assumptions hold, they present a
relationship for calculating the distance along the lm at
which the developing concentration boundary layer reaches
any arbitrary penetration depth. However the development
of the thermal boundary layer, which is much faster in
water-lithium bromide lms, is not discussed.
The equations presented for the temperature and
concentrations within the lm require an iterative solu-
tion. Once the solutions are obtained, relationships are
presented for calculating heat and mass transfer coe-
cients. A comparison with the expressions developed by
Nakoryakov and Grigor'eva [19,20] shows signicant
deviations in temperature and concentration when the
nite concentration of the lm is considered. Brauner et
al. show that including this eect leads to predictions of
enhanced transfer rates and penetration depth of the
concentration boundary layer. An expression for the
enhancement factor is also presented.
Brauner [54] develops new solutions to the vertical
laminar lm absorption problem for any downstream
location for both adiabatic and isothermal wall cases.
Her assumptions are the same as those of Grossman [29]
with two signicant extensions. First, as in her previous
work, the application of Fick's law of diusion at the
interface is formulated without assuming innite dilu-
tion of the water in the lm. Second, the lm thickness is
allowed to vary with absorbed mass; the resulting
transverse velocity component, v, is not neglected. This
leads to the following forms of the governing equations:
u
oT
ox
v
oT
oy
= o
o
2
T
oy
2
(15)
u
oC
A
ox
v
oC
A
oy
= D
AB
o
2
C
A
oy
2
(16)
She divides the solution domain into three regions: (1)
the region where both thermal and concentration
boundary layers are developing from the liquid-vapor
interface, (2) the region where the thermal boundary
layer has reached the wall but the concentration prole
is still developing, and (3) the region where both
boundary layers are fully developed. Within each
region, she assumes parabolic proles of temperature
and concentration which satisfy the required boundary
conditions. The temperature prole may reduce to a
linear prole in the isothermal wall case. These assumed
prole shapes are then substituted into the continuity,
energy, and species conservation equations. The equations
are translated into integral form and require numerical
integration. By varying the parameter corresponding to
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 767
solution concentration, she demonstrates that the solu-
tions for the lm temperature, concentration, heat and
mass uxes all change signicantly when the concentra-
tion of the lm is taken into account. Additionally by
examining the non-dimensional coecients of heat and
mass transfer, she shows that the heat transfer coecient
is unaected by the solution concentration and that the
modication to the predicted heat transfer is due to the
coupling of the heat and mass transfer problems.
4.1.8. van der Wekken and Wassenaar 1988 [55]
van der Wekken and Wassenaar [55] present an
extension to the problem solved by Grossman [29].
Instead of assuming either an adiabatic or isothermal
wall, they extend the heat transfer model to a constant
temperature cooling medium (representing the case of
cross-ow coolant) and use the coolant heat transfer
coecient as a variable for study. They do not assume
innite dilution of the water in the lm in their for-
mulation of Fick's law. The rest of their assumptions
are identical to Grossman [29] including parabolic, fully
developed velocity prole, constant thermophysical
properties, constant lm thickness (transverse velocity is
zero), no heat transfer to the vapor, linearized equili-
brium model, etc. The problem is solved using nite
dierence methods.
For values of the model parameters corresponding to
water-lithium bromide absorption, van der Wekken and
Wassenaar show the variation of temperatures and
concentrations against downstream position for four
values of non-dimensional coolant heat transfer coe-
cient: 0, 0.1, 10, and o. Plots of non-dimensional lm
heat and mass transfer coecients are also presented.
The authors note that the coecients may vary by up to
50% when a parameter such as the non-dimensionalized
heat of absorption, Lewis number, coolant temperature
or coolant heat transfer coecient is doubled or
increased by an order of magnitude, but that for any
given problem, the parameters would only vary by ``tens
of percent'' along the lm. This leads them to conclude
that curves of the transfer coecients predicted using
average values of the problem parameters would be
sucient for a design model.
4.1.9. Habib and Wood 1990 [32]
Habib and Wood [32] create a numerical model for
co-current absorption on a laminar, vertical lm that
includes heat and momentum transfer in the vapor
phase. Many typical assumptions are applied including
constant lm thickness, isothermal wall, and innite
dilution in Fick's Law. The authors include many of the
terms typically neglected by other authors in the gov-
erning equations [for example Eqs. (1) and (2) without
the transient terms]. They allow for pressure gradients
along the absorber and interfacial shear between the two
phases, although presumably these terms are very small
since the Reynolds numbers for the lm and vapor are
both 30, and the Nusselt solution for lm thickness is
used. A unique boundary condition applied by the
authors is that the stream-wise gradients in velocity,
temperature, and concentration at the exit of the absor-
ber must be zero; no justication is given for this
assumption. The governing equations are discretized
using the approach of Patankar [56] with grid density
increasing near the interface. For the baseline operating
condition, temperature and concentration proles
across the lm and temperature proles in the vapor are
plotted at a variety of downstream conditions. It is
shown that both the temperature in the lm and the
vapor approach the wall temperature at the absorber
exit. The absorption rate along the length of the absor-
ber shows a maximum just after the absorber inlet (pre-
sumably where the eect of the wall temperature is rst
transmitted to the interface) and an exponential decline
thereafter. It is shown that as the inlet lm temperature
is reduced to the wall temperature, the maximum
absorption rate increases and its position approaches
the inlet. Other variations in inlet lm concentration,
absorber pressure and wall temperature are shown to
result in straightforward changes in absorption rate.
The authors conclude that ``the temperature in the gas
phase can not be neglected as was suggested by previous
investigators'' though no quantication of the attendant
errors is presented. Furthermore, the authors show very
close agreement between their predictions of interface
and bulk lm concentration along the absorber and
those of Andberg and Vliet [37].
4.1.10. Yang and Wood 1992 [57]
Yang and Wood [57] also present an extension to the
problem solved by Grossman [29], although they only
consider the isothermal wall case, by allowing the inlet
solution temperature to deviate from the temperature of
the wall. This problem is very similar to the one solved
by Andberg and Vliet [37] and the authors note that
they obtain excellent agreement with the results of
Andberg and Vliet. However they point out that their
formulation of the problem is simplied from the one
considered by Andberg and Vliet since the assumptions
are essentially the same as Grossman [29].
The authors present plots of the interface and bulk
temperatures and concentrations against downstream
position for a baseline case and for cases with diering
lm Reynolds numbers. They also present a comparison
of total absorption rate against downstream position
from the model and from the authors' previous experi-
mental work. In the experimental data the lm is
reported to have waves and the vapor to contain 5% air,
two counteracting inuences. Perhaps fortuitously, the
comparisons reveal that the model predicts the experi-
mental results ``reasonably well'' except at the lowest
lm Reynolds numbers.
768 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
4.1.11. Hajji and Worek 1992 [58]
Hajji and Worek [58] consider the transient absorp-
tion of water vapor into a stagnant lm of aqueous
lithium bromide. An abridged version of this analysis
can also be found in a previous work by Hajji and
Lavan [59] in which the authors also considered surface
tension-driven instabilities. They suggest that the results
can be adapted to steady absorption on a falling lm
over a at plate by substituting downstream position/
stream-wise velocity (x/u) for time (t) in the resulting
expressions; presumably this would be analogous to
assuming uniform velocity within the lm. Some
assumptions they employ include no heat transfer to the
vapor phase, constant thermophysical properties, and
innite dilution in Fick's law. Since the lm is stagnant,
the governing equations are a function of depth and
time only. At the bottom of the lm, two boundary
conditions are considered: either an isothermal wall, or
a constant heat and mass ux wall (which includes
adiabatic and impermeable). The governing equations
are non-dimensionalized and analytical solutions are
developed using Fourier expansion series methods.
Exact analytical expressions for calculating the eigen-
values are presented for several ranges of system
parameters, and two sets of eigenvalues are calculated
for the case representing a waterlithium bromide
system. Results using several values of Le are plotted
along with experimental data obtained from the litera-
ture. The authors show that Le=0.015 provides the
best t to the experimental data for a water-lithium
bromide system. Additional results are presented as
curves of temperature, concentration and uxes against
time. The eects of Le, the heat of absorption and the
boundary conditions at the bottom of the lm are
investigated.
It should be noted that Grossman [35] in his review
on heat and mass transfer in lm absorption discusses
other solutions to absorption on stagnant lms. He
points out that the solution of Nakoryakov and Gri-
gor'eva [19,20] for coupled heat and mass transfer can
be modied for the case of a stagnant lm by replacing
terms of the form x/u with t. He then points out that the
resulting expression for mass transfer coecient is the
same as that obtained from Higbie's penetration theory
for isothermal absorption except that the equilibrium
concentration at the interface, which drives the mass
transfer, is lowered as a function of the non-dimensional
heat of absorption and the Lewis number.
4.1.12. Ramadane, Aoufoussi and Le Go 1992 [60]
Ramadane et al. [60] briey describe an implicit nite
dierence model developed for laminar falling lm
absorption. Their assumptions are very similar to Gross-
man [29] including a parabolic velocity prole, linearized
interface equilibrium model, and constant thermo-
physical properties. One key dierence is that coolant side
heat transfer is also modeled, either using an appropriate
heat transfer correlation or by including a partial dier-
ential equation for this transfer mechanism. Typical pro-
les of temperature and concentration through the lm
are plotted for a variety of downstream positions. Addi-
tionally, the authors present experimentally determined
values for the heat transfer coecient between the lm
interface and bulk for lms with Reynolds numbers ran-
ging from about 15 to 70. The predictions of the model are
compared with the experimental data albeit with a slightly
dierent Prandtl number for each. The authors note that
the experimentally derived average lm heat transfer coef-
cient increases with Reynolds number because of the
development and increasing importance of waves. The
authors conclude that the model shows satisfactory agree-
ment with the experimental data for Re<100 although
no quantication of the error is presented.
4.1.13. Conlisk 19921995 [6166]
Conlisk [6163] presents the development of a solu-
tion to the vertical, laminar lm absorption problem
employing the Laplace transform technique. His
assumptions are essentially the same as Grossman [29]
except that the lm thickness is not required to be constant,
the transverse velocity component, v, is not neglected
in the species conservation equation although it is
neglected in the energy equation, the wall temperature
can either be constant or an empirically derived function
of downstream position, and innite dilution of water in
the lm is not assumed. It is assumed that mass transfer
is conned to a boundary layer (i.e. doesn't reach the
wall) but this assumption is justied for moderate
lengths; for a typical operating condition, Conlisk sug-
gests that this length is about 3-m, increasing or
decreasing with Reynolds number. Conlisk notes that
the use of the Laplace transform eliminates the diculty
often encountered in nding solutions at the inlet dis-
continuity. Considering an expansion series formulation
for the lm thickness along the absorber, he mathema-
tically demonstrates that, ``to a leading order'', the lm
thickness is constant because the mass transfer rate is
limited by the physical parameters of the problem.
However, his solution for the concentration prole
depends on the leading order term for variation in lm
thickness (absorption rate) which is determined
approximately from the inlet and equilibrium conditions
for the isothermal case. For the non-isothermal case, a
more complex relationship is needed for determining
this leading order term of the lm thickness variation.
Determining the temperature prole requires numeri-
cally evaluating a Laplace inversion integral. Conlisk
shows that in the region near the inlet where heat
transfer is conned to a boundary layer near the inter-
face, a solution similar to the solution for the con-
centration distribution can be determined which does
not require numerical evaluation.
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 769
Conlisk [61] presents the results of the model in sev-
eral ways. For both an isothermal and non-isothermal
wall, plots of temperature proles are given at various
downstream locations for two solution ow rates. Also
the leading order term for absorption rate (lm thick-
ness variation) is plotted for the case of constant and
variable wall temperature and curve ts are given which
can be used to simplify subsequent analyses. Compar-
isons with experimental data are also presented for two
operating conditions. The total absorption rate pre-
dicted is within 20% of the experimental values, over-
predicting the experimental data in both cases. The
author concludes that the potential of waves to aect
the results is small.
Conlisk [62] presents the results of varying other
parameters of the problem such as wall temperature,
solution ow rate, and tube length. An interesting con-
clusion of this work is that ``for tube lengths less than
about one meter, limited increases in the amount of
water absorbed may be expected beyond a solution ow
rate of about 0.6 kg/min,'' which tacitly assumes com-
plete wetting. Conlisk [63] also concludes that the results
suggest ``that the use of long tubes to increase heat and
mass transfer area is not an ecient means to increase
the absorption rate to the tube surface'' primarily since
the highest absorption rates occur near the inlet. ``The
implication is that a bank of short tubes operating at
lower individual Reynolds numbers will be most e-
cient for the absorption process.''
Noting that the temperature proles predicted by the
above model become substantially conduction dominated
at a small downstream length, Conlisk [65] presents fully
analytical solutions for the problem by assuming that the
temperature prole within the lm must be linear. The
total absorption rate at the two operating conditions used
for comparison with experimental data in Conlisk [61] is
within 1% of the original model. It is shown that at posi-
tions very near the lm inlet, the prediction of local
absorption rate diers greatly between the two models.
But because this region is small compared to the length of
the tube (1.5 m in this case) the eect on overall model
performance is small. Based on the analytical expressions
developed, Conlisk [65] shows that as the initial thickness
of the lmis reduced, the heat transfer coecient increases
but the mass transfer coecient decreases, although no
further physical interpretation of why this happens is
given. Also, Conlisk notes that the total absorbed mass
ux is almost completely independent of the dierence
between the lm inlet and wall temperatures. Presumably
this is because the sensible heat load associated with inlet
temperature dierences is small compared to the heat load
of absorption for long absorbers. This conclusion is in
sharp contrast to the ndings of Andberg and Vliet [40] for
the horizontal tube case where the lm ow length is small.
Conlisk [66] extends his earlier model [65] to include
heat transfer to the coolant by assuming specic heat,
mass ow rate and average heat transfer coecient
(author suggests Gnielinski [67]) of the coolant, as well
as wall thickness and conductivity. Obtaining the solu-
tion requires numerical evaluation of two integrals, and
two methods for predicting absorbed mass are given. It
is demonstrated that the model results are not overly
sensitive to the choice of correlation for heat transfer
coecient. Additionally, comparison with four sets of
experimental data reveals that total mass absorption
rate is generally predicted within 10%. Conlisk also
notes that only a 10% increase in absorbed mass is
achieved from a four-fold increase in coolant ow rate,
and suggests that ``attempting to increase the mass
absorbed by modifying the coolant-side ow has limited
return.'' This is presumably due to the relatively sig-
nicant thermal resistance of the falling lm, compared
to the coolant-side resistance.
4.1.14. Ibrahim and Vinnicombe 1993 [31]
Ibrahim and Vinnicombe [31] develop what they call a
hybrid method for solving the problem of vertical,
laminar lm absorption. By combining the analytical
solutions proposed by Nakoryakov and Grigor'eva
[19,20] near the inlet and along the lm interface with a
nite dierence scheme in other regions of the lm, they
develop a method that accurately predicts the results of
more complex solutions such as Grossman [29] and van
der Wekken and Wassenaar [55] but requires less com-
putational eort. The assumptions they employ are
essentially the same as van der Wekken and Wassenaar
[55] except that the coolant is assumed to be in either co-
or counter-ow with the lm, and the heat transfer
coecient is derived from the Dittus-Boelter correlation
[156]. Also, a test case is solved to determine the eects
of linearizing the interface equilibrium condition. They
show that there are essentially no dierences for a
practical range of interest. For the test case, only very
far downstream of the solution inlet can the eect of
linearization be discerned.
In order to compare their results with those of
Grossman [29] and van der Wekken and Wassenaar
[55], the authors model a case where the coolant tem-
perature is xed. They show that there are slight dis-
crepancies between all three models but that in general,
the authors' model agrees very well with one or the
other previous models for predicted mass transfer coef-
cient versus downstream position. Ibrahim and Vinni-
combe demonstrate the computational savings of their
hybrid scheme by showing that a typical nite dierence
scheme requires much smaller discretization lengths
than the hybrid model before the solution demonstrates
grid-size independence.
Ibrahim and Vinnicombe also apply the eectiveness/
number-of-transfer-units method of quantifying heat
exchanger performance to the absorber model. Since other
authors (see section 4.5.3 [118] may dene the parameters
770 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
of this analysis dierently for analyzing an absorber, the
denitions are reviewed here. Ibrahim and Vinnecombe
dene the eectiveness as the absorbed mass compared
to the mass that could be absorbed given an innitely
large absorber; capacity ratio as the product of specic
heat and mass ow rate of the coolant divided by that of
the solution lm; and number of transfer units (NTU)
as the product of an overall heat transfer coecient and
absorber area divided by the capacity of the coolant.
They compare the eectiveness of co- and counter-ow
coolant arrangements versus NTU and nd that the
counter-ow arrangement is generally more eective,
although the dierence is not as pronounced as in con-
ventional heat exchangers. They also demonstrate that
for any given absorber size, a capacity ratio between 20
and 40 will provide nearly the overall eectiveness of an
isothermal cooling surface. Capacity ratios above this
(greater coolant ow rates) will not appreciably improve
absorber eectiveness; this agrees with the ndings of
Conlisk [66]. Another interesting conclusion the authors
make is that although absorption rates decrease as the
heat of absorption increases, the total refrigerant load
that could be supplied by a heat pump increases.
4.1.15. Choudhury, Hisajima, Ohuchi, Nishiguchi,
Fukushima and Sakaguchi 1993 [69]
Choudhury et al. [69] examine the problem of
absorption in laminar lms falling over a horizontal
tube in a manner similar to Andberg and Vliet [40].
Dierences between their model and the model of And-
berg and Vliet [40] include the assumption of constant
thermophysical properties, a hydrodynamic formulation
not based on boundary layer formulations, and an iso-
thermal tube wall. Some assumptions in common with
Andberg and Vliet are the planar jet inlet at the top of
the tube, smooth laminar lm, and no heat transfer to
the vapor. They also solve the problem using a nite
dierence formulation on a grid t to the lm shape,
and neglect absorption in the planar jet. However, the
lm inlet conditions are assumed to be in equilibrium
with the vapor. Consequently absorption does not begin
until a certain distance around the tube where the heat
transfer to the wall is able to reduce the interface tem-
perature. Plots of iso-concentration lines and isothermal
lines on the solution domain are shown for two lm ow
rates. Additionally, uxes, temperatures and concentra-
tions around the tube are shown. The eect of ow rate
and tube diameter on heat transfer coecient and total
absorption rate are examined. It is shown that an opti-
mum absorption rate (assuming complete wetting) is
predicted to occur at some low solution ow rate
depending on tube diameter.
4.1.16. Lu, Li, Li and Yu-Chi 1996 [70]
Lu et al. [70] also develop a model for laminar lm
absorption over horizontal tubes. The assumptions are
similar to Choudhury et al. [69] and Andberg and Vliet
[40]. They assume constant physical properties, and
constant coolant temperature with xed heat transfer
coecient based on the Dittus-Boelter correlation.
However, at the top of the tube, the lm is assumed to
have a velocity equal to the velocity of a body freely
falling between the tubes. This assumption neglects
momentum losses that would occur upon impact. Also,
the inlet is assumed to have uniform temperature and
concentration proles but is not necessarily in equili-
brium with the vapor as was assumed by Choudhury et
al. [69]. Results from the model are not presented.
Instead, experimental results with one smooth and two
dierent spirally grooved tubes are used to correlate two
model coecients: an eective wetting area coecient,
and a mass transfer enhancement coecient. The
authors found that only the spirally grooved tubes at the
highest ow rates (Re=36) have complete wetting. The
smooth tube wetting coecient may be as low as 40%.
The mass transfer enhancement for the smooth tube is
identically equal to 1 for all cases. The enhancement fac-
tor for the spirally-grooved tubes may be as high as 8.
4.1.17. Jernqvist and Kockum 1996 [71]
Jernqvist and Kockum [71] solve the problem of ver-
tical laminar lm absorption without several of the
assumptions commonly found in the literature. For
instance, they allow for variable thermophysical prop-
erties, developing laminar velocity prole, variable lm
thickness (increasing with absorption), non-linear equi-
librium model, energy ux due to interdiusion, and
nite dilution of water in the solution. Although many
of these assumptions are relaxed individually in previous
models, this is the rst attempt to eliminate so many
simultaneously. Nonetheless, several assumptions are
still employed including no shear at the interface, no
heat transfer to the vapor, and an isothermal wall. Also
the transverse velocity is neglected in the governing
equations. Solutions are obtained using a nite dier-
ence formulation (note that the hydrodynamics in the
lm are simultaneously calculated to account for the
variable property eect). One interesting eect of this is
that the change in interface velocity with downstream
position is a compound eect due to the developing
hydrodynamics, the increasing mass ow rate due to
absorption, and the changing temperature and con-
centration in the lm aecting the uid viscosity and
density. Concentration proles at several downstream
positions and lm thickness versus downstream position
are presented. Comparison with experimental data
reveals that the laminar theory does not accurately pre-
dict the trends in the measured results in all cases. For
instance, as lm ow rate is increased, the model pre-
dicts reduced absorption rates, but the data exhibit an
increase in absorption rate. The authors propose a x
based on mixing within the lm due to waves. Their
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 771
discussion of wave eects will be presented in the wavy
lm section of this paper.
4.1.18. Conlisk and Mao 1996 [72]
Conlisk and Mao [72] consider the problem of tran-
sient absorption into an initially constant thickness lm
surrounding a horizontal tube. In this case there is no
lm inlet to the system, but solutions are a function of
time. As time proceeds, the lm begins to sag around
the tube due to gravity, while simultaneously absorbing
vapor. The authors actually consider the case of binary
vapor but that feature will be discussed in the section of
this paper concerning ammonia-water. Conlisk and
Mao treat the water-lithium bromide uid pair as a
special case of the solution. The solution to the lm
hydrodynamics is the classical Nusselt solution, how-
ever the lm thickness varies at each point with time. At
the bottom of the tube, surface tension eects are not
considered; where a droplet would form, intsead a dis-
continuity occurs limiting the length of time for which
solutions are valid as noted by the authors. The authors
employ many typical assumptions including constant
thermophysical properties, no heat transfer to the
vapor, and a linearized interface equilibrium model. The
solution method is similar to Conlisk [61,65]. Three
regions for the solution are considered: (1) where the
thermal boundary layer is just developing, (2) where the
thermal boundary layer has reached the wall but is not
fully developed, and (3) where the temperature is fully
conduction dominated. Solutions of case (2) require
numerical analysis and are not presented by Conlisk and
Mao [72]. In the other two regions, analytical solutions
are obtained but still require some numerical evaluation.
Several plots of concentration, temperature, heat and
mass uxes are presented for typical operating condi-
tions with water-lithium bromide lms. Conlisk and
Mao note that ``absorption is more rapid on the top of
the tube where the lm is thinning,'' although it is
unclear how this conclusion would be aected by dro-
plet or jet impingement which would normally occur in
this region.
4.2. Laminar lm ow for other absorber geometries
A few attempts to solve the coupled equations of heat
and mass transfer governing absorption have been made
considering specialized absorber geometries. Some of
these are reviewed in this section. The following models
do not consider the eects of waves or turbulence and
are therefore classied as laminar ow in spite of the
fact that the lm ow pattern is not classically laminar.
4.2.1. Conlisk 1994b [64]
A vertical tube with lengthwise utes is considered by
Conlisk [64] due to its potential to increase transfer rates
by altering the laminar ow pattern from that seen in
the ow over a smooth vertical tube. ``The basic physi-
cal eect of the utes is to induce the ow to vary on a
length scale much shorter than the length of the tube.''
The premise of the enhancement to heat and mass
transfer is that the surface tension of the lm interacting
with the periodic curvature of the utes tends to draw
the lm from the crests of the utes into the troughs
thus inducing a sideways velocity component and
maintaining a thin-lm situation on the ute crests (the
Gregorig [73] eect). Solutions for the hydrodynamics in
this situation are given in Johnson and Conlisk [74]. For
typical applications, these solutions are stated to be
valid up to some large distance down the lm where the
lm in the troughs becomes too thick. Except for the
dierent hydrodynamics, Conlisk [64] uses the same
assumptions as Conlisk [61]. He considers the problem
in three regions of boundary layer development as in
Conlisk and Mao [72] and develops solutions using the
Laplace transform for the region near the inlet and far
downstream for both the mass transfer and heat transfer
problems; the intermediate region requires a fully
numerical solution.
4.2.2. Yang and Jou 1995 [75]
Utilizing essentially the same assumptions as Gross-
man [29], Yang and Jou [75] model the absorption pro-
cess for the case when the lm ow is contained within a
porous medium on the surface of an isothermal wall.
The hydrodynamics and eective thermal conductivity
of the uid-saturated porous material are based on
models developed in previous work. The system of
equations is solved numerically. Temperatures, con-
centrations and absorption rates are presented as a
function of downstream position for a baseline operat-
ing condition. By comparing the predictions to the
smooth lm model of Yang and Wood [57], the authors
show that absorption rates are enhanced especially for
lm Reynolds numbers greater than 60; this is without
considering the enhanced wetting characteristics of a
porous medium. Additionally, parameters describing
the porous medium including conductivity and porosity
are varied and the eect on absorption rate is shown.
4.2.3. Conlisk 1996a, 1996b [76,77]
Conlisk [76,77] analyzes the hydrodynamics and
absorption for a lm owing down a vertical tube with
spine ns. To simplify the solution for the hydro-
dynamics, surface tension eects are neglected. The
premise of spine-n tubes is similar to the uted tubes in
that the laminar ow is altered on a short length scale.
Eectively, the smooth tube model of Conlisk [66],
which assumes a linear temperature prole and mass
transfer conned to a thin boundary layer, is used in the
region between the spines. The model also includes heat
transfer to a cooling medium with constant heat transfer
coecient. Conlisk reports that some enhancement in
772 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
the absorption rate is gained from the spine-n tubes
compared to the smooth tube; however ``the enhance-
ment is not large and is signicantly decreased with
increasing inuence of surface tension through capillar-
ity'', the last conclusion being derived from experi-
mental data.
4.2.4. Summary of laminar water-lithium bromide lms
Table 1 contains a summary of the vertical wall
models described above. Acronyms are used for the
common assumptions to show the progression of
attempts to relax assumptions. It should be noted that
the tables presented here are not meant to contain a
comprehensive list of all assumptions used; rather they
primarily display the key assumptions which have been
modied and removed by authors over the years. It is
clear, even with this partial listing, that no models have
been developed without many simplifying assumptions.
For instance, many recent models still use the assump-
tion of an isothermal or adiabatic wall. It may be possi-
ble, although probably cumbersome, to extrapolate to
more realistic wall conditions from these solutions. More
useful models will incorporate realistic coolant models
since this is a key parameter over which a designer has
control. Further discussion of the assumptions can be
found in the conclusions section. Table 2 summarizes the
horizontal tube models. The horizontal-tube case has
received less attention in spite of its wide use in absor-
bers. Currently, the assumptions made to model ow
both between and around the tubes are often signicant
over-simplications. Thus, this is an area that warrants
further development. In general, although they provide
valuable insight into the problem, the practical usefulness
of laminar-lm models is not well established since lami-
nar lms are rarely encountered in actual absorbers.
4.3. Turbulent lm ow
Much work has been done to investigate pure heat
transfer in turbulent lms and absorption in turbulent
lms when the heat of absorption is negligible. How-
ever, only a few investigators have attempted to solve
the coupled heat and mass transfer problem of absorp-
tion for the case of turbulent falling lms. As is typically
done for the case of either heat or mass transfer alone,
the turbulent motion of the lm is accounted for by
modifying the physical constants a and D
AB
in the gov-
erning equations using empirical correlations, not by
modeling the highly complex and random motion within
the lm. This approach is physically justiable since the
a and D
AB
basically account for the random motion
within the lm on the molecular scale; the modied
values simply account for random motion on the larger
turbulence scale. Thus extending laminar lm models to
turbulent lm models is reasonably straightforward
since the assumption of a smooth interface is used for
both (turbulent lms may actually have a rough surface
with chaotic, high frequency, low amplitude waves as
seen in Miller [78]). A typical schematic of the solution
domain for a turbulent falling lm, including a repre-
sentative prole of turbulent eddy diusivity (discussed
below), can be seen in Fig. 2. Grossman [35] points out
that the transition to turbulence in falling lms is a more
gradual process than in pipe or channel ow. He sug-
gests that the transition to turbulent ow may begin at
Reynolds numbers as low as 200, and that when
Re>1600, the ow is completely turbulent. Other
investigators have suggested higher transition Reynolds
numbers. Most probably, this is due to diering desig-
nations of chaotic-wavy ow and fully turbulent ow.
4.3.1. Grossman and Heath 1984, 1986 [79,80]
Many of the assumptions used in Grossman and
Heath [79] mirror those found in Grossman [29] with a
signicant exception for the treatment of turbulence.
The mass and thermal diusivities in the governing
equations are modied to include the eect of turbu-
lence:
D
AB.turbulent
= D
AB
c (17)
o
turbulent
= o c (18)
where c, the turbulent eddy diusivity, is a function of
transverse position within the lm. From the above
denitions, it appears that the eddy diusivity is
assumed to aect each transport process identically. The
eddy diusivity prole the authors use to characterize
the transport enhancement is a combination of three
proles found in the literature. Near the wall, the Van
Driest [81] prole is used up to its recommended region
of applicability. From there, a prole suggested by Rie-
chardt [82] is scaled to intersect the Van Driest prole
and is used through the core of the lm. Near the inter-
face a prole suggested by Lamourelle and Sandall [83]
which accounts for damping near the interface is used.
The point at which the near-interface and core proles
intersect is used to dene their respective regions of
applicability. The eddy diusivity as dened above is
zero at both the liquid-vapor interface and the wall due
to the damping of the turbulent motion that occurs near
these boundaries (this is the prole illustrated in Fig. 2).
Grossman [35] points out that the prole near the
interface is both the most important to absorption heat
and mass transfer, and the least well understood. He
also reviews several recent approaches investigators
have proposed for modeling the eddy diusivity prole
throughout the lm. He notes that some investigators
suggest that the damping near the interface is primarily
due to surface tension forces while others suggest that
the eddies are damped by viscosity. Based on his review,
Grossman concludes that surface tension may play a
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 773
Table 1
Vertical wall, laminar water-lithium bromide lms
a
Tableau 1
Mur vertical, lms laminaires eau/bromure de lithium
a
Author Date Solution type Hydrodynamics Domain Wall Other
Nakoryakov
and Grigor'eva
(1977) [17] Analytical (SS) SFI, USVP, NTV Unrestricted ISO CFT, LIEM, EMD, NVHT, CTP,
NIE, LFTP
(1977) [26] Analytical (ES) SFI, USVP, NTV Unrestricted ISO CFT, LIEM, EMD, NVHT, CTP, NIE
(1980) [19,20] Analytical (SS) SFI, USVP, NTV Film inlet ISO CFT, LIEM, EMD, NVHT, CTP, NIE
(1995) [24] Analytical (SS) SFI, USVP, NTV Film inlet and
downstream
ISO LIEM, EMD, NVHT, CTP, NIE,
LFTP downstream
(1997) [25] Analytical (ES) SFI, USVP, NTV Unrestricted ISO or adiabatic CFT, LIEM, EMD, NVHT, CTP, NIE
Kholpanov et al. (1982) [33] Analytical (SS) SFI, USVP, NTV,
vapor shear
Film inlet NA CFT, LIEM, EMD, NVHT, CTP, NIE
Grossman (1983) [29] Numerical and
analytical (ES,
SS near lm inlet)
SFI, PSVP, NTV Unrestricted ISO or adiabatic CFT, LIEM, EMD, NVHT, CTP, NIE
(1987) [14] Analytical (SS) SFI, USVP, NTV Film inlet NA CFT, LIEM, EMD, NVHT, CTP
Andberg and Vliet (1983) [37] Numerical SFI, PSVP, NTV Unrestricted ISO EMD, NVHT, CTP
(1983) [38] Approximate
correlation
SFI, PSVP, NTV Unrestricted ISO EMD, NVHT, CTP
van der Wekken
and Wassenaar
(1988) [55] Numerical SFI, PSVP, NTV Unrestricted Isothermal coolant CFT, LIEM, NVHT, CTP, NIE
Kawae et al. (1989) [51] Numerical SFI, PSVP, NTV Unrestricted ISO EMD, NVHT
Brauner et al. (1989) [52] Analytical (SS) SFI, USVP, NTV Film inlet NA CFT, LIEM, NVHT, CTP, NIE
Brauner (1991) [54] Numerical (SS
near inlet)
SFI, PSVP Unrestricted ISO or adiabatic CFT, LIEM, NVHT, CTP, NIE
Habib and Wood (1990) [32] Numerical SFI, PSVP Unrestricted ISO CFT, EMD, CTP, NIE
Yang and Wood (1992) [57] Numerical SFI, PSVP, NTV Unrestricted ISO or adiabatic CFT, LIEM, EMD, NVHT, CTP, NIE
Hajji and Worek (1992) [58] Analytical (ES) Stagnant lm NA ISO or ux CFT, LIEM, EMD, NVHT, CTP, NIE
Ramadane et al. (1992) [62] Numerical SFI, PSVP, NTV Unrestricted Coolant CFT, LIEM, EMD, NVHT, CTP, NIE
Conlisk (1992,3,
1994) [6163]
Analytical
(LaPlace integrals)
and numerical
SFI, PSVP Moderate lengths Fixed temp. prole LIEM, NVHT, CTP, NIE
(1995) [65] Analytical SFI, PSVP Moderate lengths Fixed temp. prole LIEM, NVHT, CTP, NIE, LFTP
(1995) [66] Analytical SFI, PSVP Moderate lengths Fixed coolant htc LIEM, NVHT, CTP, NIE, LFTP
Ibrahim and
Vinnecombe
(1993) [31] Hybrid (analytical
and numerical)
SFI, PSVP, NTV Unrestricted Fixed coolant htc CFT, LIEM (tests this assumption),
EMD, NVHT, CTP, NIE
Jernqvist and
Kockum
(1996) [71] Numerical SFI, NTV, developing Unrestricted ISO NVHT
a
SS=similarity solution; ES=expansion series; SFI=smooth lm interface; USVP=uniform stream-wise velocity prole; PSVP=parabolic stream-wise velocity prole;
NTV=negligible transverse velocity; ISO=Isothermal wall; CFT=constant lm thickness; LIEM=linear interface equilibrium model; EMD=equimolar diusion at interface;
NVHT=negligible vapor-phase heat transfer; CTP=constant thermophysical properties; NIE=negligible interdiusion eects; LFTP=linear lm temperature prole.
7
7
4
J
.
D
.
K
i
l
l
i
o
n
,
S
.
G
a
r
i
m
e
l
l
a
/
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
R
e
f
r
i
g
e
r
a
t
i
o
n
2
4
(
2
0
0
1
)
7
5
5

7
9
7
role under certain conditions, but that it is probably of
secondary importance compared to the eect of viscos-
ity.
Grossman and Heath [79] solve the problem numeri-
cally, but as in Grossman [29], develop a similarity
solution for the region near the inlet to overcome the
discontinuity at this location. The similarity solution is
identical except that it accounts for the dierent velocity
in the lm. Numerical results are presented for both the
isothermal and adiabatic wall cases. The concentration
and temperature at the interface, wall and within the
bulk are plotted against normalized downstream posi-
tion. They show that the transport process within the
lm is much faster than in the laminar case (see Fig. 7).
The authors show that the higher the lm Reynolds
number (for Re>10
4
), the greater the enhancement of
transport within the lm. The eects of heat of absorp-
tion, Prandtl number and Schmidt number on interface
conditions and transfer coecients are also presented
and corrections for the transfer coecients are found in
an errata [80]. One interesting conclusion of this work is
that the heat and mass transfer coecients are nearly
constant over almost the entire range of downstream
positions. Only near the inlet do the coecients vary
from their asymptotic values. As expected, the mass
transfer coecient is primarily determined by Reynolds
number and Schmidt number whereas the heat transfer
coecient depends on Prandtl number.
4.3.2. Faghri and Seban 1988, 1989 [84,85]
Noting the independence of transfer coecients to
position in the results of Grossman and Heath [79],
Faghri and Seban [84] seek a purely analytical solution
to the coupled heat and mass transfer problem. The
basic assumptions are identical to Grossman and Heath
[79]. It is shown that by additionally assuming that the
concentration distribution near the interface is indepen-
dent of downstream position, the asymptotic values of
interface heat and mass transfer coecients given by
Grossman and Heath [80] can be derived from the
assumed eddy diusivity prole near the interface; in
other words, assuming an eddy diusivity prole is
nearly equivalent to assuming the asymptotic values of
the heat and mass transfer coecients. The authors
assume that the Dittus-Boelter correlation holds [156] for
the wall-side heat transfer coecient and a closed-form
solution is fully dened for both the isothermal and
constant heat ux (including adiabatic) wall cases. It is
shown that the solution predicts the numerical results of
Grossman and Heath [79] for a Reynolds number of
10
4
; at a higher Reynolds number, where the assump-
tion of constant transfer coecients is even more justi-
able, the agreement is not as good. The authors note
``nothing decisive can be said about these uncertainties.''
In a follow-up work, Faghri and Seban [85] construct
a numerical model similar to Grossman and Heath [79] T
a
b
l
e
2
H
o
r
i
z
o
n
t
a
l
t
u
b
e
s
,
w
a
t
e
r
-
l
i
t
h
i
u
m
b
r
o
m
i
d
e

l
m
s
a
T
a
b
l
e
a
u
2
T
u
b
e
s
h
o
r
i
z
o
n
t
a
u
x
,

l
m
s
e
a
u
/
b
r
o
m
u
r
e
d
e
l
i
t
h
i
u
m
a
A
u
t
h
o
r
D
a
t
e
S
o
l
u
t
i
o
n
t
y
p
e
H
y
d
r
o
d
y
n
a
m
i
c
s
D
o
m
a
i
n
W
a
l
l
O
t
h
e
r
A
n
d
b
e
r
g
a
n
d
V
l
i
e
t
(
1
9
8
6
,
1
9
8
7
)
[
3
9

4
1
]
N
u
m
e
r
i
c
a
l
P
l
a
n
a
r
j
e
t
,
S
F
I
,
d
e
v
e
l
o
p
i
n
g
P
S
V
P
F
i
l
m
o
n
t
u
b
e
I
s
o
t
h
e
r
m
a
l
c
o
o
l
a
n
t
N
V
H
T
,
N
I
E
C
h
o
u
d
h
u
r
y
e
t
a
l
.
(
1
9
9
3
)
[
6
9
]
N
u
m
e
r
i
c
a
l
P
l
a
n
a
r
j
e
t
,
S
F
I
,
d
e
v
e
l
o
p
i
n
g
P
S
V
P
F
i
l
m
o
n
t
u
b
e
I
s
o
t
h
e
r
m
a
l
N
V
H
T
,
N
I
E
,
C
T
P
,
i
n
l
e
t

l
m
/
v
a
p
o
r
e
q
u
i
l
i
b
r
i
u
m
L
u
e
t
a
l
.
(
1
9
9
6
)
[
7
0
]
N
u
m
e
r
i
c
a
l
P
l
a
n
a
r
j
e
t
,
S
F
I
,
n
o
i
m
p
a
c
t
l
o
s
s
F
i
l
m
o
n
t
u
b
e
I
s
o
t
h
e
r
m
a
l
c
o
o
l
a
n
t
N
V
H
T
,
N
I
E
,
C
T
P
,
e
m
p
i
r
i
c
a
l
w
e
t
t
i
n
g
a
n
d
m
a
s
s
t
r
a
n
s
f
e
r
e
n
h
a
n
c
e
m
e
n
t
C
o
n
l
i
s
k
a
n
d
M
a
o
(
1
9
9
6
)
[
7
2
]
A
n
a
l
y
t
i
c
a
l
(
L
a
P
l
a
c
e
i
n
t
e
g
r
a
l
s
)
a
n
d
n
u
m
e
r
i
c
a
l
S
a
g
g
i
n
g

l
m
,
S
F
I
,
P
S
V
P
F
i
l
m
o
n
t
u
b
e
I
s
o
t
h
e
r
m
a
l
L
I
E
M
,
N
V
H
T
,
C
T
P
,
N
I
E
a
S
S
=
s
i
m
i
l
a
r
i
t
y
s
o
l
u
t
i
o
n
;
E
S
=
e
x
p
a
n
s
i
o
n
s
e
r
i
e
s
;
S
F
I
=
S
m
o
o
t
h

l
m
i
n
t
e
r
f
a
c
e
;
U
S
V
P
=
u
n
i
f
o
r
m
s
t
r
e
a
m
-
w
i
s
e
v
e
l
o
c
i
t
y
p
r
o

l
e
;
P
S
V
P
=
p
a
r
a
b
o
l
i
c
s
t
r
e
a
m
-
w
i
s
e
v
e
l
o
c
i
t
y
p
r
o

l
e
;
N
T
V
=
n
e
g
l
i
g
i
b
l
e
t
r
a
n
s
v
e
r
s
e
v
e
l
o
c
i
t
y
;
C
F
T
=
c
o
n
s
t
a
n
t

l
m
t
h
i
c
k
n
e
s
s
;
L
I
E
M
=
l
i
n
e
a
r
i
n
t
e
r
f
a
c
e
e
q
u
i
l
i
b
r
i
u
m
m
o
d
e
l
;
E
M
D
=
e
q
u
i
m
o
l
a
r
d
i

u
s
i
o
n
a
t
i
n
t
e
r
f
a
c
e
;
N
V
H
T
=
n
e
g
l
i
g
i
b
l
e
v
a
p
o
r
-
p
h
a
s
e
h
e
a
t
t
r
a
n
s
f
e
r
;
C
T
P
=
c
o
n
s
t
a
n
t
t
h
e
r
m
o
p
h
y
s
i
c
a
l
p
r
o
p
e
r
t
i
e
s
;
N
I
E
=
n
e
g
l
i
g
i
b
l
e
i
n
t
e
r
d
i

u
s
i
o
n
e

e
c
t
s
;
L
F
T
P
=
l
i
n
e
a
r

l
m
t
e
m
p
e
r
a
t
u
r
e
p
r
o

l
e
.
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 775
to further check the validity of the analytical solution.
They nd satisfactory agreement with their previous
results and note that the dierences are ``consistent quali-
tatively with the fact that the average interface transfer
numbers exceed the asymptotic values'' due to the small
enhancement during the development length near the inlet.
4.3.3. Yuksel and Schlunder 1987 [86]
Yu ksel and Schlu nder [86] consider much the same
problem as Grossman and Heath [79] with the following
exceptions: Fick's law is formulated allowing for nite
dilution of water in the lm, interfacial shear stress due
to laminar counter-ow of vapor to the liquid lm is
included in the hydrodynamics, the transverse velocity
term (v) is included in the governing equations, and the
lm thickness varies with downstream position based on
absorbed mass. Shared assumptions include constant
thermophysical properties and no heat transfer to the
vapor. They also consider three wall boundary conditions:
isothermal, constant heat ux (includes adiabatic), and
constant coolant heat transfer coecient. Additionally,
a slightly modied eddy diusivity prole is used; and
by choosing a turbulent Prandtl number and Schmidt
number not equal to 1, the eect of the eddy diusivity
is not identical in each transport process. In the near-
wall region, the Van Driest equation is modied
according to Seban and Faghri [87] but with one con-
stant changed. And near the interface, either a prole
proposed by Hubbard et al. [88] or by Carrubba [89] is
used depending on Reynolds number (the correlation by
Carruba is actually for lower Reynolds numbers corre-
sponding to a transient wavy ow regime). A discussion
of the merits of several dierent diusivity models is
also presented. Another unique aspect of this model is
that the inlet concentration prole is chosen to be an
eigth-order polynomial function based on the boundary
layer solution of absorption into a semi-innite medium
instead of a uniform prole.
The authors compare their results to two sets of
experimentally derived transfer coecients. The rst,
Blangetti [90], does not account for the sensible heating
load due to the change in bulk lm temperature, leading
to articially high heat transfer coecients. The second,
by the authors, accounts for the change in bulk lm
temperature in the calculation of heat transfer coe-
cients. The second set also includes a full set of mea-
sured mass transfer coecients. The authors show that
models using the eddy proles proposed by Grossman
and Heath [79], Blangetti [90], and Seban and Faghri
[87] lead to overpredictions of the heat transfer coe-
cients. The model by the authors, however, accurately
predicts both sets of overall heat transfer coecients
and the measured mass transfer coecients over a range
of lm Reynolds numbers from about 100 to 1200.
Fig. 7. Dimensionless bulk concentration, ,, versus normalized length, =(xPr)/(Peo), for Pr=10, Sc=2000, and Re as indicated.
Solid lines correspond to an isothermal wall prediction, dashed lines to an adiabatic wall (adapted from Grossman and Heath [79]).
Fig. 7. Concentration adimensionnelle, ... en fonction de la longueur normalise e, ..... (xPr)/(Ped) ou Pr = 10, Sc = 2000, et Re comme
indique . Les courbes continues correspondent a une pre vision de mur isothermique, ; les courbes hachure es correspondent a un mur adia-
batique (adapte s des travaux de Grossman et Heath [79]).
776 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
4.3.4. Kholpanov and Kenig 1993 [91]
Kholpanov and Kenig [91] extend the model pro-
posed by Grossman and Heath [79] to the case of mul-
ticomponent absorption where mass and heat transfer
resistance in the vapor phase are still neglected. The
governing equations are arranged in matrix form and
solved using nite dierence approximations. A discus-
sion of analytical solutions in the entrance region and
far downstream is also presented. Results for a two
component vapor absorbing into a three component
lm are given for both the isothermal and adiabatic wall
case. One interesting conclusion proposed by the
authors is that a uniform stream-wise velocity can be
used with minimal change to the solution, presumably
because the velocity prole for turbulent ow is rela-
tively constant for the bulk of the lm.
4.3.5. Summary of turbulent water-lithium bromide lms
Table 3 contains a summary of the turbulent lm
models that are reviewed here. The turbulent lm case
inherently requires assumptions about the turbulent
motion within the lms based on experimental work; the
models will only be as good as the eddy diusivity cor-
relations. However general agreement on the best eddy
diusivity prole to use is lacking. It would be benecial
if continued experimental research in the area of turbu-
lent lms took into account the combined heat and mass
transfer process, and attempts were made to corrobo-
rate predictions made using these eddy diusivity pro-
les with experimental data.
4.4. Wavy lm ow
As was stated earlier, it is well known that falling
lms are inherently unstable, even at low Reynolds
numbers [9294]. This instability leads to the growth of
waves visible on the lm surface even in the absence of
any disturbances due to the motion of the vapor. A
recent review of the literature on the hydrodynamics of
wavy falling lms is presented by Miller [78]. Com-
menting on several key investigations, Miller states
``their results imply that all vertical falling lms are
naturally wavy''.
The waves that may develop on a falling lm are not
all alike. Many investigators have suggested categoriza-
tion methods based on the non-dimensional Reynolds
and Kapitza numbers. A simple classication is pro-
posed by Brauner [27] who suggests the existence of two
primary types of waves: (1) capillary waves which are
characterized by their low amplitude, sinusoidal shape,
regular frequency, and tendency of the wave-front to be
aligned perpendicular to the ow direction (see Fig. 3),
and (2) inertial waves, often called roll waves, which can
obtain high amplitudes (several times the underlying
lm thickness), have a steep wave-front and long wave-
back, may travel as solitary waves with long spans of T
a
b
l
e
3
V
e
r
t
i
c
a
l
w
a
l
l
,
t
u
r
b
u
l
e
n
t
w
a
t
e
r
-
l
i
t
h
i
u
m
b
r
o
m
i
d
e

l
m
s
a
T
a
b
l
e
a
u
3
M
u
r
v
e
r
t
i
c
a
l
,

l
m
s
t
u
r
b
u
l
e
n
t
s
e
a
u
/
b
r
o
m
u
r
e
d
e
l
i
t
h
i
u
m
a
A
u
t
h
o
r
D
a
t
e
S
o
l
u
t
i
o
n
t
y
p
e
H
y
d
r
o
d
y
n
a
m
i
c
s
D
o
m
a
i
n
W
a
l
l
O
t
h
e
r
G
r
o
s
s
m
a
n
a
n
d
H
e
a
t
h
(
1
9
8
4
,
6
)
[
7
9
,
8
0
]
N
u
m
e
r
i
c
a
l
(
S
S
n
e
a
r

l
m
i
n
l
e
t
)
S
F
I
,
N
T
V
U
n
r
e
s
t
r
i
c
t
e
d
I
s
o
t
h
e
r
m
a
l
o
r
a
d
i
a
b
a
t
i
c
C
F
T
,
L
I
E
M
,
E
M
D
,
N
V
H
T
,
C
T
P
,
N
I
E
,
c
u
s
t
o
m
3
-
p
a
r
t
e
d
d
y
d
i

u
s
i
v
i
t
y
p
r
o

l
e
F
a
g
h
r
i
a
n
d
S
e
b
a
n
(
1
9
8
8
)
[
8
4
]
A
n
a
l
y
t
i
c
a
l
S
F
I
,
N
T
V
A
w
a
y
f
r
o
m
i
n
l
e
t
I
s
o
t
h
e
r
m
a
l
o
r
q
=
c
o
n
s
t
.
C
F
T
,
L
I
E
M
,
E
M
D
,
N
V
H
T
,
C
T
P
,
N
I
E
,
a
s
s
u
m
e
d
e
d
d
y
d
i

u
s
i
v
i
t
y
p
r
o

l
e
(
1
9
8
9
)
[
8
5
]
N
u
m
e
r
i
c
a
l
S
F
I
,
N
T
V
U
n
r
e
s
t
r
i
c
t
e
d
I
s
o
t
h
e
r
m
a
l
o
r
q
=
c
o
n
s
t
.
C
F
T
,
L
I
E
M
,
E
M
D
,
N
V
H
T
,
C
T
P
,
N
I
E
,
a
s
s
u
m
e
d
e
d
d
y
d
i

u
s
i
v
i
t
y
p
r
o

l
e
Y
u

k
s
e
l
a
n
d
S
c
h
l
u

n
d
e
r
(
1
9
8
7
)
[
8
6
]
N
u
m
e
r
i
c
a
l
S
F
I
,
v
a
p
o
r
s
h
e
a
r
U
n
r
e
s
t
r
i
c
t
e
d
A
d
i
a
b
a
t
i
c
i
s
o
t
h
e
r
m
a
l
w
a
l
l
/
c
o
o
l
a
n
t
N
V
H
T
,
C
T
P
i
n
t
r
a
n
s
v
e
r
s
e
d
i
r
e
c
t
i
o
n
,
N
I
E
,
c
u
s
t
o
m
3
-
p
a
r
t
e
d
d
y
d
i

u
s
i
v
i
t
y
p
r
o

l
e
K
h
o
l
p
a
n
o
v
a
n
d
K
e
n
i
g
(
1
9
9
3
)
[
9
1
]
N
u
m
e
r
i
c
a
l
,
m
u
l
t
i
-
c
o
m
p
o
n
e
n
t

l
m
/
v
a
p
o
r
S
F
I
,
N
T
V
U
n
r
e
s
t
r
i
c
t
e
d
I
s
o
t
h
e
r
m
a
l
o
r
a
d
i
a
b
a
t
i
c
C
F
T
,
L
I
E
M
,
E
M
D
,
N
V
H
T
,
C
T
P
,
N
I
E
,
e
d
d
y
d
i

u
s
i
v
i
t
y
p
r
o

l
e
f
r
o
m
G
r
o
s
s
m
a
n
a
n
d
H
e
a
t
h
a
S
S
=
s
i
m
i
l
a
r
i
t
y
s
o
l
u
t
i
o
n
;
E
S
=
e
x
p
a
n
s
i
o
n
s
e
r
i
e
s
;
S
F
I
=
s
m
o
o
t
h

l
m
i
n
t
e
r
f
a
c
e
;
U
S
V
P
=
u
n
i
f
o
r
m
s
t
r
e
a
m
-
w
i
s
e
v
e
l
o
c
i
t
y
p
r
o

l
e
;
P
S
V
P
=
p
a
r
a
b
o
l
i
c
s
t
r
e
a
m
-
w
i
s
e
v
e
l
o
c
i
t
y
p
r
o

l
e
;
N
T
V
=
n
e
-
g
l
i
g
i
b
l
e
t
r
a
n
s
v
e
r
s
e
v
e
l
o
c
i
t
y
;
C
F
T
=
c
o
n
s
t
a
n
t

l
m
t
h
i
c
k
n
e
s
s
;
L
I
E
M
=
l
i
n
e
a
r
i
n
t
e
r
f
a
c
e
e
q
u
i
l
i
b
r
i
u
m
m
o
d
e
l
;
E
M
D
=
e
q
u
i
m
o
l
a
r
d
i

u
s
i
o
n
a
t
i
n
t
e
r
f
a
c
e
;
N
V
H
T
=
n
e
g
l
i
g
i
b
l
e
v
a
p
o
r
-
p
h
a
s
e
h
e
a
t
t
r
a
n
s
f
e
r
;
C
T
P
=
c
o
n
s
t
a
n
t
t
h
e
r
m
o
p
h
y
s
i
c
a
l
p
r
o
p
e
r
t
i
e
s
;
N
I
E
=
n
e
g
l
i
g
i
b
l
e
i
n
t
e
r
d
i

u
s
i
o
n
e

e
c
t
s
;
L
F
T
P
=
l
i
n
e
a
r

l
m
t
e
m
p
e
r
a
t
u
r
e
p
r
o

l
e
.
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 777
smooth lm in between successive waves (see Fig. 4), or
may interact with each other to form complex three-
dimensional wave patterns with wave-fronts that are not
necessarily aligned across the ow direction. Inertial
waves may also contain regions of recirculation (see
Fig. 8, [27,95,96]) that play a role in the transport pro-
cesses between the lm interface and bulk. It is also
important to note that wave shape generally travels
several times faster than the uid in the lm (this speed
ratio is called wave celerity) and consequently wavy
lms have a higher mass ow rate than a smooth lm
with an equivalent mean lm thickness. The majority of
the mass ow may be carried within the wave shape
itself leaving only a fraction of the ow in the thin lm
substrate between and underneath the waves [27].
To model the absorption process in the presence of
wavy falling lms, investigators must make assumptions
regarding the lm hydrodynamics which involve a
number of simplications. In each case presented in this
section, the hydrodynamics are assumed a priori,
neglecting the potential inuences of the heat and mass
transfer processes on the hydrodynamics. However,
several investigators, for example Cosenza and Vliet [44]
and Miller and Keyhani [97], have noted signicant dif-
ferences in the observed hydrodynamics and surface
wetting when absorption is occurring; this is generally
attributed to surface tension gradients resulting from
uneven absorption over the lm surface. It is also typi-
cal to neglect the actual development of the waves from
an initially smooth lm in favor of using an asymptotic
steady-state solution. Also, often only a small number
of wavelengths are modeled representing a short length
of the absorber. Models have been developed for
absorption in the presence of both capillary and inertial
wavy-laminar lms, but in either case most investigators
assume that the waves are periodic. Experimental
investigations of wavy lms reveal that the waves them-
selves, especially at Reynolds numbers greater than 100,
are unstable and often develop into seemingly chaotic
three-dimensional patterns [97] perhaps requiring a sta-
tistical description. The general conclusion of all of
these investigations is that the presence of waves sig-
nicantly enhances the transport processes compared
with a laminar lm. The exact mechanism attributed to
the enhancement and its quantitative value depend on
the investigator.
4.4.1. Burdukov et al. 1980, Nakoryakov et al. 1982a,
1982b [21,22,99]
One of the simplest methods for modeling the eect of
waves is to assume that a passing wave completely mixes
the lm, resulting in uniform concentration and tem-
perature proles at regular intervals corresponding to
the characteristic wavelength. This assumption allows
results such as those of Nakoryakov and Grigor'eva
[19,20] for the absorption near the inlet region of a fall-
ing lm absorber to be applied to the regions in between
each wave. This technique is used in Burdukov et al. [99]
and Nakoryakov et al. [21,22].
Burdukov et al. [99] and Nakoryakov et al. [21] pre-
sent the data from experimental measurements of
absorption rate on a vertical tube absorber. It is noted
that beyond a position of 350 millimeters down the
tube, the absorption rate and lm temperature become
nearly constant. This does not agree with the theoretical
solutions of Nakoryakov and Grigor'eva [17] for lami-
nar lms which predict that the transfer rates should
vary as the inverse of the square root of the downstream
position. The authors also note the presence of ``irre-
gular three-dimensional waves'' in this region beyond
Fig. 8. Contours of the streamfunction for a solitary roll wave
with wave-peak to substrate-lm thickness ratio of 4. Note
wave scale is reduced in the vertical direction, actual wave
slopes are not as steep [adapted from Jayanti and Hewitt [96]).
Fig. 8. Contours de l'e coulement d'un rouleau solitaire d'une onde
avec un maximum de l'onde sur e paisseur du lm substrat, de 4.
Notez que l'e chelle de l'onde est re duite dans le sens vertical et
que les pentes re elles des ondes sont moins fortes (travaux de
Jayanti et Hewitt [96] adapte s).
778 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
350 millimeters. To account for the waves, the authors
adopt the postulate of Kholpanov et al. [100] that the
eect of the waves is to completely mix the lm. They
use photography to determine the characteristic wave-
length and then assume that the laminar theory for the
inlet region holds within each wavelength. Comparison
with experimental data shows that this wavy model at
least qualitatively predicts the trends in the data.
Nakoryakov et al. [22] present a new set of experi-
mental results for vertical falling lm absorption and a
slightly modied model to account for lm waviness.
The dierence from the previous model is that instead of
using photography to characterize the average wave-
length, the authors use existing experimental correla-
tions for wave velocity [101] and for wave frequency
[102] to calculate the wavelength. This wavy lm model
predicts the observed trends in the experimental data
better than a laminar lm model, but as in the previous
models, quantitative agreement is only fair.
4.4.2. Kholpanov 1987, 1990 and Tsvelodub 1989 [103
105]
Kholpanov [103] rst presents the results of solving a
simplied version of the Navier-Stokes equations for the
wave motion. The predicted wave amplitude is shown to
agree reasonably well with some experimental data from
the literature. A plot of non-dimensional ow rate
against wave number shows local minima and maxima,
the maxima corresponding to the ``optimal'' regime
which will be established. By assuming a parabolic
stream-wise velocity prole throughout the lm, the
author develops a boundary layer solution for mass
transfer only, and another solution for combined heat
and mass transfer. Neither the temperature boundary
layer nor the concentration boundary layer developing
from the interface is allowed to reach the wall; thus the
solution is insensitive to wall conditions. For the pure
mass transfer problem, the model results are compared
with data from ``long pipes'' and show good agreement.
No validation or discussion of limitations is given for
the combined heat and mass transfer problem. How-
ever, since the transport processes are even more rapid
due to presence of waves, the entry region where
absorption is insensitive to wall conditions will pre-
sumably be shorter than for the smooth lm case.
Kholpanov also considers wavy ow over a rough sur-
face with gas shear using a linear stream-wise velocity
prole and ow along the inside of a rotating cylinder.
A very similar problem is also considered by Tsvelodub
[105]. He assumes a constant stream-wise velocity within
the boundary layers and develops analytical relation-
ships for heat and mass transfer coecients in terms of
non-dimensional parameters.
Kholpanov [104] analytically develops expressions for
coupled heat and mass transfer coecients for absorp-
tion on a wavy lm with a purely sinusoidal surface. A
constant stream-wise velocity is assumed within the lm,
and wall conditions are not specied. Wavy lm transfer
coecients are expressed as a correction to the laminar
lm transfer coecients. The expressions for these cor-
rections are functions of wave amplitude and frequency
(or phase velocity), which the author suggests may be
determined from experimental data or from approx-
imate expressions presented.
4.4.3. Uddholm and Setterwall 1988 [106]
Uddholm and Setterwall [106] model the absorption
process on a vertical isothermal wall where the lm
exhibits inertial wavy-laminar ow (roll waves). Their
assumption for the hydrodynamics within the wave is
based on early work by Brauner et al. [107,108] who
divide a typical roll wave into four sections: wave front,
wave back, wave trail, and substrate (see Fig. 4). They
use a nite dierence scheme to solve the problem.
Other assumptions include constant thermophysical
properties, innite dilution of water within the lm, and
no heat transfer to the vapor. The front of the wave is
apparently modeled assuming a parabolic velocity pro-
le although within the wave back, the laminar sub-
strate is assumed to be developing; a solution for
boundary layer thickness and velocity prole for a wall
suddenly set in motion is used for the wave back. The
wave trail is assumed to have a velocity prole derived
from a fth-order polynomial for the stream function
proposed by Brauner et al. [108]. Finally, in the sub-
strate region between waves, a parabolic velocity prole
is assumed.
Results of photography of falling lms allowed
Uddholm and Setterwall to develop a correlation for
wave frequency based on downstream position. Their
experimental results show that near the inlet, the wave
frequency may be 20 Hz, but it reduces to about 10 Hz
further along the tube as the roll waves develop. The
authors consider four dierent models for the wave fre-
quency: constant 20 Hz waves, constant 9 Hz waves,
linearly varying 209 Hz waves, and no waves. The
predictions of average overall heat transfer coecients
are compared with experimental data from a water-
lithium bromide heat transformer. When assuming only
low frequency (9 Hz) roll waves, the model over-predicts
the experimental data. The other three assumptions
under-predict the data by varying degrees. The pre-
dicted heat transfer coecient is between 35 and 120%
greater with waves than with a laminar lm model.
4.4.4. Morioka and Kiyota 1991 [109]
Morioka and Kiyota [109] model absorption in the
presence of capillary waves. The key assumptions of
Morioka and Kiyota are that the lm surface is sinu-
soidal, the stream-wise velocity prole is parabolic [93],
the wave speed and length are determined from a linear
stability analysis by Pierson and Whitaker [110], but the
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 779
wave amplitude is arbitrarily assumed to be 30% of the
average lm thickness to ensure what the authors felt
are reasonable velocity proles within the lm. Other
assumptions include constant thermophysical properties,
isothermal wall, linear interface equilibrium model, and
no heat transfer to the vapor phase. The authors note
that for a concentration range of 5460%, the error from
the linear equilibrium assumption is 0.12

C.
A numerical solution is generated using a grid struc-
ture t to the shape of the wavy lm over three wave
lengths and governing equations transformed in such a
way as to eliminate cross-derivatives. Local absorption
rates for lm Reynolds numbers of 20, 50 and 100 are
presented along with the predictions from a smooth lm
model. Temperature and concentration proles shown
at the wave crests and troughs for a Reynolds number
of 20 and 50 demonstrate the development of a
``depression'' in the concentration prole due to the
motion of the waves and the enhanced absorption rates.
The authors point out ``the absorption mass ux shows
a maximum at the troughs of the wave where the lm
thickness is minimum. At these points, the mass ux is
approximately 3 times larger than the smooth ow
case.'' The authors propose that the enhanced absorp-
tion rates are due to both the increase in the transverse
velocity and the action of the waves eectively exposing
and covering sections of an underlying lm. They note
that their predicted absorption rate is nearly indepen-
dent of Reynolds number for the range considered. On
average, they report that the absorption rate in wavy
lms is 70 to 140% higher than for the smooth laminar
lm case. In a later experimental study which includes
photographs of wavy falling lms on vertical tubes,
Morioka et al. [98] measure absorption rates that are
generally higher than their predicted values and sensitive
to Reynolds number.
4.4.5. Yang, Jou and Chen 1991, Yang and Jou 1993
[111,112]
Yang et al. [111] and Yang and Jou [112] also con-
sider the problem of absorption on a lm with capillary
waves. The hydrodynamic solution is presented in an
earlier work [113] but is based on the following
assumptions: the stream-wise velocity prole is para-
bolic, the wave velocity and wave number correspond to
the ``most unstable wave'' determined by the linear sta-
bility analysis of Pierson and Whitaker [110], and the
solution is the asymptotic, periodic, steady-state wave
pattern. The prole of the lm interface is described by
a Fourier expansion series, the coecients of which are
determined numerically. The resulting proles vary with
Reynolds number (up to 500), and it is noted that the
solution for the wave prole always corresponds to the
maximum wave amplitude and minimum average lm
thickness at any given ow rate. The resulting proles
are not purely sinusoidal as in Morioka and Kiyota
[109], but exhibit shortened wave fronts and elongated
wave backs, suggesting characteristics of roll waves;
however, the amplitude is much lower than what is
normally labeled a roll wave, perhaps due to the linear
model used in the stability analysis. A comparison of
predicted wave lengths and amplitudes with several sets
of experimental measurements from the literature shows
fair agreement. The authors develop a numerical solu-
tion for absorption using several other assumptions
including: constant thermophysical properties, no heat
transfer to the vapor, innite dilution in Fick's law, and
an isothermal wall. To speed up the calculation proce-
dure, the results of a smooth lm model are used as
initial conditions for the wavy lm model. For a base-
line operating condition, bulk concentration plotted
against time at three downstream positions shows that
the solution becomes periodic after about the second
wave period. Total absorption rate plotted against
downstream position reveals that the absorption rates
predicted with the wavy model are about double the results
with a smooth lm. It is also shown that absorption rate
increases with Reynolds number, the eect diminishing
somewhat for higher Reynolds numbers. This trend does
not agree with the conclusions of Morioka and Kiyota
[109]; the discrepancy may be due to the fact that the
hydrodynamic solution of Yang and Wood varies with
Reynolds number and is not purely sinusoidal.
4.4.6. Jernqvist and Kockum 1996 [71]
Jernqvist and Kockum [71] suggest that their model
for smooth laminar lm absorption may be extended to
wavy (or turbulent) ow regimes by assuming complete
mixing within the lm at regular intervals (similar to
Burdukov et al. [99] and Nakoryakov et al. [21,22]).
They introduce a mixing density parameter and suggest
that it could be related to Reynolds number to account
for changing lm hydrodynamics, although no quanti-
tative relationships are presented. As was mentioned
before, the laminar lm predictions of absorption rate
decrease with increasing lm ow rate, under-predicting
the experimental data which show the opposite trend.
The authors show that by increasing the mixing density
parameter at higher Reynolds numbers, the model can
be made to predict absorption rates closer to the
experimental results. Interestingly though, the error in
heat transfer rate at the wall, which is over-predicted for
even the purely laminar case, is increased when the
mixing density is increased. The fact that wall heat
transfer rates are over-predicted even when absorption
rates are under-predicted may suggest that the assump-
tion of no heat transfer to the vapor is leading to
appreciable errors.
4.4.7. Sabir, Suen and Vinnecombe 1996 [114]
Sabir et al. [114] generate numerical solutions for the
problem of absorption on lms with capillary waves.
780 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
They assume the hydrodynamics proposed by Penev et
al. [115] including a stream-wise velocity prole which
the authors point out does not allow for any circulation
of liquid within a wave (described as surface renewal).
This prole is based on a parabolic function with coef-
cients that are calculated from a truncated expansion
series of sinusoidal functions. The lm thickness prole
is a similar truncated expansion series of sinusoidal
functions. Other assumptions used include: constant
thermophysical properties, isothermal wall at solution
inlet temperature, linearized interface equilibrium
model, and no heat transfer to the vapor. Plots of bulk
concentration and temperature against downstream
position reveal that initially, the wavy lm shows great
enhancement over the smooth lm assumption. How-
ever, further downstream, the results of the two models
converge as the limit of absorption is approached. The
authors also note that the enhancement of mass transfer
appears to be greater than that of heat transfer. They
attribute this to the fact that the mass transfer is gov-
erned by conditions at the interface where wave eects
are strongest, while heat transfer occurs across the entire
lm and the enhancement near the wall due to the waves
is less. The authors conclude that the enhancement
eect is due to the additional transverse velocity com-
ponents occurring in the wavy ow, not mixing within
the waves, although they eliminate the latter possibility
purely by assumption.
4.4.8. Patnaik and Perez-Blanco 1996b [28]
Patnaik and Perez-Blanco [28] model the absorption
process on a lm with roll waves (Reynolds numbers
between 200 and 1000) numerically. The hydro-
dynamics, specied in Patnaik and Perez-Blanco [116],
are adapted from the work of Brauner [27] who provides
an expression for the length and lm thickness of four
parts of a roll wave: front, back, trail and substrate. A
three-term Fourier series is used to provide a smooth,
periodic function that approximates the lm thickness
prole of Brauner. In order to determine wave fre-
quency and celerity, Patnaik and Perez-Blanco apply a
Fourier transform to video images of falling lms of
aqueous lithium bromide with Reynolds numbers
between 200 and 300. Although the images contain
many frequency components, a consistently identiable
13-Hz wave is chosen to be representative for modeling
purposes. It is shown that this frequency agrees with
experimental and theoretical work presented by Brauner
[27]. Patnaik and Perez-Blanco also assume that the
stream-wise velocity prole is parabolic at all points
within the lm. To model the absorption process, Pat-
naik and Perez-Blanco make several additional
assumptions including constant thermophysical proper-
ties, non-linear interface equilibrium, isothermal wall,
and no heat transfer to the vapor. The interface energy
balance is formulated assuming equimolar diusion, but
the authors suggest that compared to the transverse
velocity of the lm due to the waves, the convective
eect of non-equimolar diusion is negligible.
Predictions of bulk concentration and temperature
are plotted against downstream position at a xed time
and against time at a xed downstream position for one
operating condition. It is shown that periodic steady
state is achieved at about 0.47 s. After this time the
length of the concentration boundary layer development
region is shown to uctuate around 0.12 m. Beyond this
position, the concentration even at the wall is less than
the inlet value. The authors attribute the enhanced
transport rates to the transverse components of the lm
velocity, and their relative phasing with peaks in the
stream-wise velocity. A comparison of overall non-
dimensional heat and mass transfer coecients is made
with several other models at Reynolds numbers from 20
to 600. It is shown that at low Reynolds number there is
good agreement with Higbie's penetration theory.
Agreement with other experimental data and theoretical
models is also discussed. A large degree of scatter can be
observed among this collection of data. The non-
dimensional mass transfer coecients are shown to be
up to almost four times those predicted with a smooth
lm model. A local maximum in the predicted mass
transfer coecient is attributed to the changing wave
frequency and the limited length of the model (0.35 m).
4.4.9. Other wavy lm models for uncoupled transport
Although not expressly within the scope of this
review, it is worth noting some other signicant mathe-
matical investigations into the eect of wavy lms on
uncoupled heat and mass transfer processes due to their
advanced methods of modeling the lm hydrodynamics.
For instance, Wasden and Dukler [95,117] apply a nite
dierence grid to measured wave interface proles and
modeling the isothermal absorption process in the pre-
sence of these waves. The numerically calculated wall
shear stress and wave velocity are shown to agree with
measured results. These same calculations show that
within the waves recirculation regions may develop
depending on the wave amplitude (which is generally 25
times the underlying lm thickness, see Fig. 8). Absorp-
tion across the interface is modeled using an averaged
wave prole which includes the ripples that often develop
just ahead roll waves (sometimes called push waves). Pre-
dicted mass transfer rates are 1.52.5 times the smooth
lm predictions. The authors note that the signicant
transverse velocities caused by the waves ``in conjunction
with recirculation within the large waves . . . produce
transfer rates for large waves that are several times larger
than predicted for quasiparallel velocity elds.'' This
suggests that the assumption of a parabolic stream-wise
velocity prole may lead to erroneous results.
Jayanti and Hewitt [96] assume several shapes for
waves on a thin lm and numerically study the eect of
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 781
amplitude and distortion from the sinusoidal shape on
the hydrodynamics and heat transfer. They too nd that
recirculation regions develop within narrow, high
amplitude waves. They show that the enhancement of
the heat transfer process is almost entirely due to the
eective thinning of the lm (a wavy lm transports the
same amount of liquid with a thinner average lm
thickness). It is shown that the heat transfer rate across
the lm can be closely approximated by assuming that
conduction is the only heat transfer mechanism and
using the lm thickness prole of the wavy lm, even in
the presence of waves with a large recirculation region.
Overall heat transfer coecients may be increased by up
to 35% over a smooth lm depending on wave shape.
The results of Wasden and Dukler and Jayanti and
Hewitt support the conclusion of Sabir et al. [114] that
mass transfer may be enhanced more than heat transfer;
however, unlike Sabir et al., they point to the importance
of considering the true recirculating nature of roll waves in
the absorption process, especially in light of the coupling
of the heat and mass transfer processes in absorption.
4.4.10. Summary of wavy water-lithium bromide lms
Table 4 contains a summary of the absorption models
reviewed here that include a model for lm motion in
the presence of waves. Each of these models predicts
signicantly increased absorption rates due to the
enhanced mixing eect of waves, underscoring the
importance of assumptions about lm hydrodynamics.
There is currently no model that can account for the
complete eect of waves in a comprehensive manner.
Understanding the details of the behavior of wavy lms
is still an area that needs signicant additional research;
as newer models for the hydrodynamics emerge, these
should be incorporated into absorption modeling eorts.
Key areas which need attention are summarized in the
conclusions section but must include the non-periodicity
and instability/development of waves, recirculation within
large waves, and the potential interaction of the heat and
mass transfer process with the lm hydrodynamics.
4.5. Assumed transport coecients
A distinction can be made between the work reviewed
above and the papers in this section. Many investigators
have taken correlations for heat and mass transfer
coecients derived from experimental work and devel-
oped absorption models from them. These models
include all types of assumed hydrodynamics and are
primarily used as design tools that are computationally
simple but provide reasonable results due to an ade-
quate representation of the physical situation.
4.5.1. Wassenaar and Westra 1992 [30]
Wassenaar and Westra [30] consider the transient
response of a vertical falling-lm absorber to step T
a
b
l
e
4
V
e
r
t
i
c
a
l
w
a
l
l
,
w
a
v
y
w
a
t
e
r
-
l
i
t
h
i
u
m
b
r
o
m
i
d
e

l
m
s
a
T
a
b
l
e
a
u
4
M
u
r
v
e
r
t
i
c
a
l
,

l
m
s
o
n
d
u
l
e

s
e
a
u
/
b
r
o
m
u
r
e
d
e
l
i
t
h
i
u
m
a
A
u
t
h
o
r
D
a
t
e
S
o
l
u
t
i
o
n
t
y
p
e
H
y
d
r
o
d
y
n
a
m
i
c
s
D
o
m
a
i
n
W
a
l
l
O
t
h
e
r
K
h
o
l
p
a
n
o
v
(
1
9
8
7
)
[
1
0
3
]
A
n
a
l
y
t
i
c
a
l
(
b
o
u
n
d
a
r
y
l
a
y
e
r
S
S
)
F
r
o
m
s
t
a
b
i
l
i
t
y
a
n
a
l
y
s
i
s
N
e
a
r
i
n
l
e
t
N
A
L
I
E
M
,
E
M
D
,
N
V
H
T
,
C
T
P
,
N
I
E
T
s
v
e
l
o
d
u
b
(
1
9
8
9
)
[
1
0
5
]
A
n
a
l
y
t
i
c
a
l
U
S
V
P
N
e
a
r
i
n
l
e
t
N
A
L
I
E
M
,
E
M
D
,
N
V
H
T
,
C
T
P
,
N
I
E
K
h
o
l
p
a
n
o
v
(
1
9
9
0
)
[
1
0
4
]
A
n
a
l
y
t
i
c
a
l
S
i
n
u
s
o
i
d
a
l
,
U
S
V
P
N
e
a
r
i
n
l
e
t
N
A
L
I
E
M
,
E
M
D
,
N
V
H
T
,
C
T
P
,
N
I
E
U
d
d
h
o
l
m
a
n
d
S
e
t
t
e
r
w
a
l
l
(
1
9
8
8
)
[
1
0
6
]
N
u
m
e
r
i
c
a
l
B
r
a
u
n
e
r
r
o
l
l
w
a
v
e
s
,
P
S
V
P
,
N
T
V
U
n
r
e
s
t
r
i
c
t
e
d
I
s
o
t
h
e
r
m
a
l
E
M
D
,
N
V
H
T
,
C
T
P
,
N
I
E
M
o
r
i
o
k
a
a
n
d
K
i
y
o
t
a
(
1
9
9
1
)
[
1
0
9
]
N
u
m
e
r
i
c
a
l
S
i
n
u
s
o
i
d
a
l
c
a
p
i
l
l
a
r
y
w
a
v
e
s
,
P
S
V
P
U
n
r
e
s
t
r
i
c
t
e
d
I
s
o
t
h
e
r
m
a
l
L
I
E
M
,
N
V
H
T
,
C
T
P
,
N
I
E
Y
a
n
g
a
n
d
J
o
u
(
1
9
9
1
,
3
)
[
1
1
1
,
1
1
2
]
N
u
m
e
r
i
c
a
l
C
a
p
i
l
l
a
r
y
w
a
v
e
s
,
P
S
V
P
U
n
r
e
s
t
r
i
c
t
e
d
I
s
o
t
h
e
r
m
a
l
E
M
D
,
N
V
H
T
,
C
T
P
,
N
I
E
S
a
b
i
r
e
t
a
l
.
(
1
9
9
6
)
[
1
1
4
]
N
u
m
e
r
i
c
a
l
C
a
p
i
l
l
a
r
y
w
a
v
e
s
,
P
e
n
e
v
e
t
a
l
.
p
s
u
e
d
o
P
S
V
P
U
n
r
e
s
t
r
i
c
t
e
d
I
s
o
t
h
e
r
m
a
l
L
I
E
M
,
N
V
H
T
,
C
T
P
,
N
I
E
P
a
t
n
a
i
k
a
n
d
P
e
r
e
z
-
B
l
a
n
c
o
(
1
9
9
6
)
[
2
8
]
N
u
m
e
r
i
c
a
l
B
r
a
u
n
e
r
r
o
l
l
w
a
v
e
s
,
P
S
V
P
,
U
n
r
e
s
t
r
i
c
t
e
d
I
s
o
t
h
e
r
m
a
l
E
M
D
,
N
V
H
T
,
C
T
P
,
N
I
E
a
S
S
=
s
i
m
i
l
a
r
i
t
y
s
o
l
u
t
i
o
n
;
E
S
=
e
x
p
a
n
s
i
o
n
s
e
r
i
e
s
;
S
F
I
=
s
m
o
o
t
h

l
m
i
n
t
e
r
f
a
c
e
;
U
S
V
P
=
u
n
i
f
o
r
m
s
t
r
e
a
m
-
w
i
s
e
v
e
l
o
c
i
t
y
p
r
o

l
e
;
P
S
V
P
=
p
a
r
a
b
o
l
i
c
s
t
r
e
a
m
-
w
i
s
e
v
e
l
o
c
i
t
y
p
r
o

l
e
;
N
T
V
=
n
e
g
l
i
g
i
b
l
e
t
r
a
n
s
v
e
r
s
e
v
e
l
o
c
i
t
y
;
C
F
T
=
c
o
n
s
t
a
n
t

l
m
t
h
i
c
k
n
e
s
s
;
L
I
E
M
=
l
i
n
e
a
r
i
n
t
e
r
f
a
c
e
e
q
u
i
l
i
b
r
i
u
m
m
o
d
e
l
;
E
M
D
=
e
q
u
i
m
o
l
a
r
d
i

u
s
i
o
n
a
t
i
n
t
e
r
f
a
c
e
;
N
V
H
T
=
n
e
g
l
i
g
i
b
l
e
v
a
p
o
r
-
p
h
a
s
e
h
e
a
t
t
r
a
n
s
f
e
r
;
C
T
P
=
c
o
n
s
t
a
n
t
t
h
e
r
m
o
p
h
y
s
i
c
a
l
p
r
o
p
e
r
t
i
e
s
;
N
I
E
=
n
e
g
l
i
g
i
b
l
e
i
n
t
e
r
d
i

u
s
i
o
n
e

e
c
t
s
;
L
F
T
P
=
l
i
n
e
a
r

l
m
t
e
m
p
e
r
a
t
u
r
e
p
r
o

l
e
.
782 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
changes in operating parameters such as cooling water
temperature, vapor ow rate, and inlet lm concentra-
tion. A common feature of models using assumed
transport coecients is that the governing equations are
written in terms of temperatures, concentrations, and
velocities that are averaged in the transverse direction
for each phase of the system. For instance, the lm
temperature is taken to be the average temperature of
the entire lm including the interface, bulk and wall
regions. This formulation requires the use of transport
coecients which ``implicitly assume proles u(y), T(y)
and w(y) [v(y) in the present nomenclature]''. Further-
more, Wassenaar and Westra assume that the transient
changes do not aect the assumed transport coecients.
The authors point out at least one limitation of this
approach is that, in the model, perturbations to the lm
inlet conditions travel along the absorber at the mean
lm velocity, whereas the liquid at the interface of even
a laminar lm travels 50% faster than the mean lm
velocity. Thus inlet perturbations reach the absorber
lm outlet later in the model because of the 1-D
assumption. But, the authors suggest that the choice of
numerical formulation can help overcome this dis-
crepancy by introducing numerical diusion. Other
assumptions employed by the authors are very typical:
no heat transfer to the vapor, thermodynamic equili-
brium at the interface, nite dilution of the water in the
lm, linearized interface equilibrium model. It is worth
mentioning that the equilibrium model, although line-
arized, is dierent from most other models because it
accounts for the variation of pressure in the absorber, a
key feature required for accurately predicting the tran-
sient behavior of the system according to the authors.
The equation for average lm thickness allows for the
possibility of laminar, wavy and turbulent ow regimes
through the use of constants that must be assumed
based on ow rate. Similarly, the choice of appropriate
transport coecients must also be made a priori. Was-
senaar and Westra develop several non-dimensional
parameters that can be used to characterize the transient
response of the system. Parametric studies show that,
due to the coupled nature of the problem, the change in
system response is not directly proportional to the
change in any given parameter. Three perturbations to
the system are studied for a baseline operating condi-
tion. It is shown that the system parameters including
absorber pressure, coolant and lm outlet temperature,
and lm outlet concentration respond in dierent
sequences depending on the actual perturbation, and
that they may exhibit overshoot before settling to a new
steady condition.
4.5.2. Patnaik, Perez-Blanco and Ryan 1993 [68]
Patnaik et al. [68] utilize a number of dierent corre-
lations for heat and mass transfer coecients to develop
a simplied model for steady vertical-lm absorption
that applies to both laminar and wavy-laminar lm ow
regimes. Their basic assumptions include constant ther-
mophysical properties evaluated at the bulk-averaged
temperature and concentration, linear temperature pro-
le within the lm, no heat transfer to the vapor phase
and no vapor shear. Dierent correlations for the falling
lm heat and mass transfer coecients are used in the
entrance regions than in the fully developed ow
regions. The average lm thickness is assumed from
Nusselt's solution even in the wavy-laminar condition
and the liquid lm initially enters in equilibrium with
the vapor. Governing equations are again derived using
transversely-averaged values for velocity, temperature
and concentration. The solution is computed numeri-
cally along the length of the absorber. The model is used
to develop performance charts which present total
absorption rate, heat transfer rate, concentration
change, lm outlet temperature and outlet subcooling as
a function of downstream position for a variety of
coolant and solution mass ow rates. These perfor-
mance charts provide a quick absorber design tool and
method for predicting absorber performance, given
basic operational conditions. Additionally, it is pointed
out that the modeling technique used by Patnaik et al.
could be extended as transport coecients for other
absorber geometries become available.
4.5.3. Patnaik and Perez-Blanco 1993 [118]
Patnaik and Perez-Blanco [118] examine the use of
eectiveness/number-of-transfer-units analysis techni-
ques for simplifying the solutions to absorption pro-
blems for vertical falling-lm absorbers. By dening
several new eectiveness ratios (described below), the
authors are able to develop a straightforward relation-
ship between number of absorber transfer units and the
dened eectiveness ratios. Several assumptions are
required to develop this relationship, however. First, it
is assumed that the overall heat and mass transfer coef-
cients are known and constant. Second, the tempera-
ture prole within the lm is assumed to be linear,
although it is shown that it drops out during the solu-
tion development. Another key assumption is that the
relationship between bulk temperature and downstream
position is known. Using both experimental data and
the results of other vertical plate models, an exponential
relationship incorporating the driving force for mass
transfer at the inlet conditions is proposed. The authors
note that this assumption ``needs to be investigated fur-
ther, and the overall sensitivity of the results to changes
in this assumption established.'' Other assumptions the
authors use are fairly common including a linear equili-
brium model, no heat transfer to the vapor, and con-
stant thermophysical properties. The eectiveness ratios
dened include: a mass transfer eectiveness, the total
mass absorbed divided by the maximum that could be
absorbed given innite absorber size; a sensible heat
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 783
eectiveness, the ratio of the change in solution tem-
perature to the temperature dierence between the
solution and coolant at the inlets; and a maximum sen-
sible-to-latent heat load eectiveness, the maximum sen-
sible heat load if the solution were to reach the coolant
temperature divided by the heat of absorption that
would be released given an innite absorber size. When
designing a new absorber, design targets may be used to
help determine these eectiveness ratios; however, they
are interrelated by the heat and mass transfer coe-
cients, the interface equilibrium model, and the degree
of subcooling of the lm. Using parameters covering a
range of reasonable operating conditions for a water-
lithium bromide absorber, the authors demonstrate the
potential usefulness and simplicity of the method in
determining absorber performance.
4.5.4. Kirby and Perez-Blanco 1994 [119]
Kirby and Perez-Blanco [119] consider the case of
absorption around horizontal tubes. Although their
model uses assumed heat and mass transfer coecients
to simplify the analysis, they are the rst to incorporate
the combined eects of lm ow, droplet formation and
droplet fall which are common features of absorbers
using horizontal tubes. The absorber is divided into
three regions: the lm ow region, the droplet formation
region, and the droplet fall region. In the droplet for-
mation regime on the underside of the tubes, the aver-
age distance between droplets and the average volume
of the droplets when they fall are determined from
Taylor instability analysis of Taghavi-Tafhreshi and
Dhir [120]. The residence time of the liquid solution in
each regime can thus be calculated given an overall mass
ow rate. For the droplet fall regime, simple free-fall is
assumed, initially from rest. For the lm, the Nusselt
equations for lm thickness and velocity are used. One
key assumption about the droplet formation is that, as
liquid solution reaches the droplet, it is deposited as a
thin spherical shell on the surface of the old droplet
core. Thus the temperature of the liquid solution enter-
ing the forming droplet is used to set the driving force
for heat transfer and mass transfer (through the equili-
brium condition) into the droplet. The authors suggest
that this is the ``fresh surface'' assumption described by
Clift et al. [121]. However, Clift et al. state that, in the
``fresh surface'' model, ``fresh uid elements are
assumed to arrive at the interface to provide the increase
in area.'' Therefore, the interpretation by Kirby and
Perez-Blanco seems somewhat questionable. Correla-
tions for the mass transfer coecients in each of the
three regimes are taken from the literature. In the dro-
plet formation region, the best correlation available is a
time-averaged empirical expression based upon work
with liquid-liquid extraction where droplet formation
occurs at the end of nozzles. One of the required para-
meters for this correlation is the velocity of the liquid in
the nozzle, for which the authors apparently substituted
some characteristic stream-wise velocity from the lm
ow regime. Two correlations for heat transfer are used,
one for the coolant inside the tubes and one for the
falling lm. Presumably heat transfer during the droplet
formation region is neglected implying that adiabatic
absorption takes place in this region. For typical oper-
ating conditions of a water-lithium bromide absorber
using a six-pass serpentine horizontal-tube arrangement,
plots of average solution temperature and concentration
are given throughout the absorber. It is shown that a
major portion of the mass transfer takes place in the
droplet formation region, while the heat transfer occurs in
lm ow around the tubes. Additionally, the absorption
that takes place between the tubes signicantly increases
the solution temperature from the bottom of one tube to
the top of the next. Results are presented for various
coolant and solution ow rates, coolant inlet temperature,
solution inlet concentration, and absorber pressure.
In a later work, Atchley et al. [122] present a more
detailed description of the model and compare the
results to experimental data with three dierent tube
spacing arrangements. The agreement of overall heat
duty, absorbed mass, and heat transfer coecient is
shown to be much better when the space between the
tubes is large than when it is small. The authors attri-
bute the dierences to the wetting characteristics due to
the eect of droplet impact and splashing. The authors
also extend the model to tubes with enhanced surfaces
by selecting correction factors for the assumed heat and
mass transfer coecients.
4.5.5. Tsai and Perez-Blanco 1998 [123]
Tsai and Perez-Blanco [123] extend the model of Pat-
naik et al. [68] for vertical falling-lm absorption to
determine the asymptotic limit to absorption perfor-
mance that could be achieved by active mixing techni-
ques. The authors note that most enhancement
techniques, whether active or passive, can be considered
as mixing processes. The authors attribute this asymp-
totic behavior to the fact that ``the nite heat transfer
coecients of the absorber wall and the coolant do not
allow innite heat of absorption released to the lm''
even if resistance to mass transfer is eliminated due to
mixing. To establish transfer coecients applicable to a
fully mixed lm, the transient problem of transfer into a
semi-innite medium is used. The resulting expressions
for transfer coecients contain the square root of
exposure time in the denominator. Thus as mixing fre-
quency is increased, the exposure time is reduced, and
the transfer coecients increase. For a typical absorber
(19-m, 0.75-inch diameter vertical tube) at typical oper-
ating conditions, the absorption rate is plotted against
mixing frequency for a variety of lm Reynolds num-
bers. It is shown that as Reynolds number is increased,
the asymptotic value of absorption rate increases (with
784 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
diminishing results) but the concentration dierence
resulting from this maximum absorption rate decreases.
Based on this trend, the authors suggest that the opti-
mum Reynolds number would be a compromise
between these two eects. It is shown that, in theory, an
order of magnitude increase in absorption rate is possi-
ble over current best practice.
5. Volatile absorbent
The ammonia-water uid pair is also used in heat-
driven heat pumps, with ammonia as the refrigerant and
water as the absorbent. Although ammonia is more
volatile than water, the ammonia entering an absorber
in a heat pump will rarely be pure. Furthermore, within
the absorber itself, it is not uncommon for water to
evaporate even while ammonia is absorbed [124127].
The fact that two species may be present in both the
vapor and the liquid means that concentration gradients
and resistance to mass transfer may be present in both
phases making the problem more complex than that of
water-lithium bromide systems. As will be seen, to sim-
plify their models, many authors neglect the resistance
to mass transfer in the liquid phase; however, some
neglect the resistance in the vapor phase. Authors who
do not neglect the transport in either phase often dis-
agree on which mass transfer resistance dominates.
Conlisk [64] and Kholpanov et al. [128] both discuss the
conditions required for the problem to be simplied to a
single-phase mass transfer problem as in water-lithium
bromide systems.
Although ammonia-water systems have been studied
longer than water-lithium bromide systems, the state-of-
the-art modeling techniques are not as well developed as
those for water-lithium bromide systems. Historically,
models of ammonia-water absorption have relied heav-
ily on correlations for heat and mass transfer coe-
cients. Only in recent years have models been developed
for laminar smooth interface lms based on the gov-
erning equations without using such correlations [15,
64,128]. Models that use empirical coecients usually
incorporate one-dimensional lm models and account
for waviness, turbulence, and other lm-ow phenom-
ena only through the choice of specic correlation.
5.1. Neglected vapor-phase mass transfer resistance
Authors who neglect the resistance to mass transfer in
the vapor phase when modeling ammonia-water
absorption essentially solve the same problem as for a
water-lithium bromide system.
5.1.1. Ruhemann 1947 [16]
Ruhemann [16] presents a method for analyzing the
coupled heat and mass transfer inside a coiled-tube
absorber. The model assumes pure ammonia in the
vapor phase and constant overall heat and mass transfer
coecients throughout the absorber. The driving
potential for heat transfer is taken as the dierence
between the average lm temperature and the cooling
water temperature for a given vertical position within
the absorber (the 1.4-m tall absorber is divided into
0.01-m segments for the analysis). Similarly, the driving
potential for mass transfer is the dierence between the
pressure of the ammonia vapor at the interface, which,
based on the assumptions, is the same as the bulk pres-
sure, and the average partial pressure of ammonia in the
lm. Although this model is greatly simplied, Ruhe-
mann uses it to point out the salient features of the
coupling between the heat and mass transfer in the
absorber. He demonstrates how using either transfer
coecient alone would lead to large errors, and how
raising one transfer coecient and lowering the other
may net essentially the same overall absorber perfor-
mance despite radical changes to the local conditions
and processes within the absorber.
5.1.2. Briggs 1971 [129]
Briggs [129] models the absorption process consider-
ing only the heat transfer process by assuming a ``cti-
tious absorbing liquid temperature'' based on the
assumption of complete equilibrium between the vapor
and liquid. Although the obvious limitations of the
model are duly noted, Briggs suggests that it is still use-
ful as a quantitative design tool. He also claims that
``the main resistance to mass transfer is met as the
molecules of ammonia migrate in the liquid phase''. His
main objective is to evaluate the benets and drawbacks
of the direction of vapor ow, whether co-, counter- or
cross-current with the lm. The eects on heat transfer
and mass transfer coecients of each conguration are
discussed along with other important practical con-
siderations such as design feasibility and pressure
drop.
5.1.3. Stagnant lm models 19531975
Several investigators study the transient absorption of
ammonia into a still, semi-innite ammonia-water lm.
Although clearly these models are of limited use in fall-
ing-lm absorption, they are included here because they
may be useful as a basis for understanding how the
assumption of the lm hydrodynamics augments the
predicted results, particularly near the inlet where the
thermal and concentration boundary layers do not
reach the wall. Typical assumptions made by this group
of investigators are: pure ammonia in the vapor phase,
no heat transfer to the vapor phase, and constant ther-
mophysical properties. An analytical solution for the
heat and mass uxes and interface temperature is
developed by Danckwerts [130] by additionally assum-
ing no volume change in the liquid and constant inter-
face concentration (based on interface equilibrium at the
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 785
initial lm temperature). Carslaw and Jaeger [131]
extend the solution by allowing the interface concentra-
tion to maintain equilibrium during the absorption
process. Chiang and Toor [132] extend the analysis fur-
ther to the case where the interface moves due to volume
change of the liquid, which results in increased predic-
tions of absorption rate. They present a comparison
between their solution and experimental data obtained
by absorbing a saturated mixture of water and ammonia
into a pure water jet. Exposure times range from 0.0009
to 0.02 seconds, and the results show good agreement.
Verma and Delancey [133] point out that the coordinate
transform used in the case of a moving interface creates
a ``pseudo Dufour eect'' that may be signicant, and
they extend the solution of Chiang and Toor to account
for this. They demonstrate reasonable agreement
between their solution and experimentally measured
absorption rates and interface temperatures (exposure
times up to 400 s). For one operating condition, it is
reported that by neglecting the ``pseudo Dufour eect''
the predicted absorption rate would change by 23% and
the interfacial temperature rise would be 30

C higher.
5.2. Neglected liquid-phase mass transfer resistance
All the authors in this section utilize the method of
Colburn and Drew [134] for analyzing the combined
heat and mass transfer during binary-vapor condensa-
tion. The key assumption here is that the liquid phase is
well mixed, that is, there is no resistance to mass trans-
fer in the liquid lm. The ColburnDrew method pre-
supposes that heat and mass transfer coecients are
available for the given conditions. This method is
implemented in a computer program by Price and Bell
[135] for the case of a co-current binary-vapor con-
denser. Kang and Christensen [136] adapt this method
to model the performance of an ammonia-water absor-
ber, and extend it to a variety of absorber geometries in
subsequent work.
5.2.1. Colburn and Drew 1937 [134]
The ColburnDrew method requires iteration and is
generally valid over a small area; thus analysis of an
entire absorber or condenser requires application of the
method at many incremental areas. The basic steps for
the Colburn-Drew method are as follows. Initially, esti-
mates must be made for the composition of the
condensing vapor being absorbed, the interfacial tem-
perature and the interfacial concentrations (liquid and
vapor side). Equilibrium at the interface is usually used
to relate the interface temperature and concentrations;
choosing either the interface temperature or one of the
concentrations then xes the remaining interface condi-
tions. Once these estimates are made, the absorption
mass and heat uxes can be calculated. The mass ux is
a function of the mass transfer coecient, the con-
centration of the vapor in the bulk and at the interface,
and the assumed composition of the condensing vapor.
The heat ux at the interface is determined from the
assumed heat transfer coecient in the vapor, the vapor
temperature at the interface and in the bulk, and the
condensation rate and condensate composition. The
initial guesses of the composition of the condensing vapor
and the interface conditions are rened by balancing the
heat ux at the interface with the heat ux across the
lm for each increment while enforcing conservation of
species. The heat ux across the lm is determined from
the interface temperature, heat transfer coecient, and
either a wall or coolant temperature depending on the
denition of the heat transfer coecient. Sherwood and
Pigford [137] also present a similar method.
5.2.2. Price and Bell 1974 [135]
Price and Bell [135] incorporate the ColburnDrew
method for analyzing binary-vapor condensation into a
computer program for a co-current condenser. The heat
transfer coecient in the vapor phase is determined
from the Chilton and Colburn [138] correlation for
vapor owing over a smooth plate. A heat transfer
coecient accounting for the enhancement due to ow
over a liquid interface is also considered. This is based
on the Martinelli-Nelson correlation for the pressure
gradient for two-phase ow in spite of the fact that
uniform pressure is assumed within the condenser. The
vapor-phase mass transfer coecient is evaluated in a
similar manner. The liquid-lm heat transfer coecient
is taken as the greater the Nusselt [139] correlation for
laminar ow down a vertical surface or a curve-tted
Colburn correlation for turbulent lm ow. Initial
guesses for the interface temperature and the composi-
tion of the condensing vapor are rened through ``a
double trial and error calculation'' where the iterative
loop to calculate the condensing vapor composition is
embedded within an iterative loop to calculate the
interface temperature. Results are presented for both a
methanol-water system and an n-butane n-octane system.
Interface and bulk vapor temperatures, bulk vapor com-
position, and heat ux are plotted against cumulative
condenser area. Comparisons are made between the results
of single and two-phase transfer coecients and an
approximate analysis method proposed by Bell and Ghaly
[140]. Although these results are not for an ammonia-water
system, the algorithm developed by Price and Bell is
used in most subsequent applications of the Colburn
Drew method to ammonia-water absorber modeling.
5.2.3. Kang and Christensen et al. 19931997
[125,136,141143]
Kang and Christensen [136] adapt the algorithm
developed by Price and Bell [135] to modeling the
ammonia-water absorption process in a vertical uted-
tube absorber. The absorber features a falling lm inside
786 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
spirally uted vertical tubes with upward owing vapor.
The coolant ows upward in an annulus formed by a
larger smooth tube surrounding the uted tube. The
heat and mass transfer coecients that are used in the
vapor phase are the same as in Price and Bell [135] as
suggested by Chilton and Colburn [138]. To calculate
the overall heat transfer coecient from the lm inter-
face to the coolant, Kang and Christensen use the cor-
relation of Chun and Seban [144] originally developed
for thin falling-lm evaporation for the lm heat trans-
fer coecient, and the correlation of Garimella and
Christensen [145] for the coolant heat transfer coe-
cient in the annulus formed by a spirally uted tube
surrounded by a smooth tube. They use the model to
design a hydronically cooled absorber, and one portion
of a typical absorber found in a Generator Absorber
Heat Exchange (GAX) cycle heat pump. The authors
conclude that the heat transfer rate in the vapor and in
the coolant play an important role in the design of the
absorber. The model predicts that in some sections, a
small amount of water is desorbed while ammonia is
absorbed. They also conclude that to increase absorp-
tion rate, the vapor velocity should be maximized while
avoiding ooding or entrainment. Kang et al. [141] and
Kang and Christensen [142] extend the previous model
to the case where the lm ows within a porous med-
ium inside a smooth tube and spirally uted tube,
respectively. This is accomplished by substituting cor-
relations for the heat transfer coecient within a por-
ous medium into the model. The eect of porosity and
thermal conductivity of the porous material are inves-
tigated and optimum values determined. Kang et al.
[143] extend the model of Kang and Christensen [136] to
a combined generator-absorber found in GAX systems.
They show that in some portions of the absorber, water
vapor desorbs while ammonia is absorbed; during the
desorption process, both water and ammonia are des-
orbed simultaneously. Additionally they conclude that
``heat transfer resistance in the liquid side is a dominant
factor during the desorption process while both heat
transfer in the liquid lm and mass transfer in the vapor
ow are considerable during the absorption process.''
Palmer and Christensen [146] extend the model of Kang
and Christensen [136] to the case where a proprietary
insert is placed in the spirally uted tube to form an
annulus for the falling lm absorption. Kang et al. [125]
use the modeling technique of Kang and Christensen
[136] to analyze each component in an absorption heat
pump. They illustrate why the ux of ammonia and
water may be in opposite directions at positions within
an absorber or a desorber but not within the other
components such as the condenser or dephlegmator.
5.2.4. Garrabrant and Christensen 1997 [147]
Ammonia-water absorption is modeled within a novel
absorber geometry by Garrabrant and Christensen [147]
using the analysis techniques of Kang and Christensen
[138] as a basis. The absorber consists of a perforated
plate bent accordion-style and sandwiched between two
vertical plates [148]. The angle of the surfaces of the
perforated plate promotes lm ow; puddling occurs
where the perforated plate meets the vertical plates.
When the liquid pool reaches a perforation, liquid falls
as droplets to the surface below and lm ow occurs
again, thus regulating the solution ow. Vapor generally
ows upward, although the arrangement of the per-
forations promotes cross-current ow with the falling
lm. Heat is removed through the vertical plates by
coolant owing upward through oset-strip n cavities.
Correlations were used for the transfer of heat between
the vapor and liquid lm, the vapor and vertical plates,
the liquid lm to the wall (using an eective n e-
ciency), and the wall to coolant. A correlation by Sher-
wood and Pigford [149] is used as a baseline for the
mass transfer coecient in the vapor phase; the lm is
assumed to be well mixed. The assumed wetted area, n
eciency and mass transfer coecient are adjusted
based on experimental testing. Expressions for the for-
mer two corrections are developed based on liquid
solution inlet concentration, which aects surface ten-
sion, and non-dimensional ow rate. The correction to
the mass transfer coecient is formulated as a function
of non-dimensional vapor ow rate. The local con-
centration within the absorber, predicted by the model
with the above-mentioned correction factors, is pre-
sented for one operating condition.
5.2.5. Garimella 1999 [127]
A novel, miniaturized absorber design that results in
substantial size reduction over conventional geometries
is proposed by Garimella [127]. The absorber consists of
a stacked lattice-style arrangement of very small dia-
meter tubes through which coolant ows. Several par-
allel tubes are arranged in a horizontal plane to form a
square array. Several of these arrays are positioned one
above the other, oriented such that the tubes in each
array are perpendicular to those in the adjacent arrays.
A structure of distribution chambers holds the tubes in
the desired arrangement and distributes the coolant; the
exact pass arrangement of the coolant is determined by
pressure drop considerations. The liquid solution is dis-
tributed as droplets to the top array of horizontal tubes.
By alternating lm and droplet ow the liquid proceeds
to the bottom of the absorber. Vapor is introduced at
the bottom of the absorber and therefore ows counter-
current with the liquid solution. Additionally the dis-
tribution of the coolant starts with the bottom rows and
therefore ows generally counter-current to the falling
liquid solution. Garimella develops a model of this
absorber using the ColburnDrew method. The heat
and mass transfer coecients in the vapor are derived
from a correlation by Zhukauskas [150] for cross-ow
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 787
over staggered tube banks. Absorption is assumed to
only occur during the lm ow regime; the adiabatic
absorption of the falling droplets is neglected. The cor-
relation for falling-lm heat transfer over horizontal
tubes by Dorokhov and Bochagov [151] and a constant
coolant side heat transfer coecient are used. Garimella
states that the model is limited by the assumption of
transfer coecients and that a more accurate model
would require further development of these correlations
or of ``a comprehensive ow-mechanism based theory
for absorption heat and mass transfer''. The model is
used to predict overall performance of the absorber and
local conditions and uxes within the absorber for one
design layout at one typical operating condition. The
heat transfer resistance in the coolant is uniquely low
due to very small tube-side hydraulic diameters; the
coolant heat transfer coecient was about 57% of that
of the falling lm. With plots of ow rates, tempera-
tures, concentrations, and absorption rates throughout
the absorber, he shows that through a major portion of
the absorber, a small amount of water is desorbed as
ammonia is absorbed. However, near the top of the
absorber (the lm inlet) a larger rate of water absorp-
tion occurs, due in part to the reduced ammonia con-
centration in the vapor and the large driving
temperature dierence.
5.3. Mass transfer in both phases
5.3.1. Kholpanov, Kenig and Malyusov 1985, 1986
[128,152]
Kholpanov et al. [128] numerically solve the coupled
equations governing heat and mass transfer with a bin-
ary liquid and vapor in counter-ow without assuming
any transport coecients. The simplied form of the
governing transport equations used can be found in
Eqs. (7) and (8); however, unlike the water-lithium bro-
mide case, these equations are applied in both phases.
Equilibrium is assumed at the interface, and the condi-
tions are determined by iteratively balancing the heat
and mass transfer at the interface while enforcing
equilibrium. The equilibrium condition is linearized,
although it contains an extra term to account for the
concentration in the vapor phase. Other assumptions
include either an isothermal or adiabatic wall, constant
thermophysical properties, and laminar ow in both
phases. It is not clear exactly what velocity proles are
considered and if they allow for interfacial shear; they
simply state that ``the problem is investigated for a ow
of moderate velocity, although the method of solution
is applicable for any specied velocity prole''. The
assumed hydrodynamics are independent of the heat
and mass transfer processes. The governing equations
are transformed using non-dimensional parameters. The
eect of these non-dimensional parameters on the
absorption process is shown by plotting bulk
concentration in the liquid phase against downstream
position for the isothermal-wall case where it is noted
that the average lm temperature ``remains practically
constant and close to T
w
[the temperature of the wall]''.
For the adiabatic-wall, both the average liquid tem-
perature and concentration are plotted. Kholpanov et
al. discuss the instances where the problem degenerates
to a single-phase problem, that is, when the transport
resistance in one of the phases can be neglected. It is
shown that this is essentially based on two non-dimen-
sional parameters. One is the ratio of mass diusivities
in the two phases and the other depends on the relative
thicknesses and velocities of the two phases. One inter-
esting conclusion of the parametric studies is that ``with
a decrease in the quantity characterizing the volume of
the liquid lm the resistance to mass transfer gradually
shifts from the gas phase to the liquid phase.'' This cor-
roborates a similar conclusion by Conlisk [65] that the
mass transfer coecient of a falling lm decreases with
lm thickness. Some dimensional results are given for
the absorption of water from air into aqueous diethy-
lene glycol.
Kholpanov et al. [152] solve a similar problem for the
case of multi-component liquid and vapor. The govern-
ing equations are formulated as matrices and discussion
of the characteristics of the matrices of coecients is
provided. The solution procedure is signicantly dier-
ent than in the above work; it is assumed that heat and
mass transfer in both phases occur within a boundary
layer and an approximate analytical solution is devel-
oped. Clearly due to the conclusions in Kholpanov et al.
[128] about the temperature in the liquid phase, the
results presented here are only valid in a very small
region near the lm inlet.
5.3.2. Perez-Blanco 1988 [124]
Perez-Blanco [124] develops a model for the absorp-
tion of ammonia into a lm falling around coiled tubes.
The model uses assumed transfer coecients for each
transport process in the absorber. Instead of using a
typical equilibrium relationship for the interface condi-
tions, Perez-Blanco de-couples the concentrations at the
interface ``to avoid extremely high calculated mass
transfer rates''. It is not immediately evident what fea-
tures of this model lead to this unique requirement. The
equation used to relate the de-coupled interface con-
centrations is a simple function of the mass transfer
coecients and bulk concentrations in each phase.
The interface temperature is determined by a similar
relationship between the heat transfer coecients and bulk
temperatures in each phase. Once the interface tem-
perature is determined, the liquid-side interface con-
centration is determined by a property relation using the
overall absorber pressure. This allows the vapor inter-
face concentration to be determined from the afore-
mentioned de-coupling relationship. Although details
788 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
are not given about the correlations used for each
transfer coecient, the mass transfer coecient in the
falling lm is adapted from the penetration theory. A
correction is applied for small Reynolds numbers based
on surface tension; the result more than doubles the
original coecient. Two operating conditions are mod-
eled representing typical ranges. By varying each of the
transfer coecients and the solution ow rate, Perez-
Blanco determines that, within the range of operating
conditions considered, the mass transfer resistance in
the liquid lm is the most important in controlling the
entire absorption process, and that the coolant heat
transfer coecient can inuence the results if it is below
a certain threshold. The model is shown to predict outlet
concentration, overall heat duty, and absorbed mass
within about 10% for one operating condition measured
in an actual absorber. Perez-Blanco also discusses the
potential for water desorption in the presence of
ammonia absorption and how it aects absorber per-
formance by changing the heat load.
5.3.3. Conlisk 1994b and Conlisk and Mao 1996 [64,72]
In seeking an analytical solution to the ammonia-
water absorption process based on the governing trans-
port equations, Conlisk [64] and Conlisk and Mao [72]
extend their solutions for water-lithium bromide
absorption and show that water-lithium bromide may
be considered a special case of the two-phase binary
heat and mass transfer problem. Conlisk [64] considers
the case of a laminar lm on a vertical uted tube and
Conlisk and Mao [72] consider an initially stagnant,
uniform lm on a smooth horizontal tube; the solution
method is the same in both cases. The basic assumptions
used can be found in the discussion of these papers in
the water-lithium bromide section of this paper, but
they include no heat transfer to the vapor, a linear
model for the interface equilibrium, and connement of
mass transfer to a boundary layer near the interface (a
questionable assumption in the ammonia-water case).
Conlisk discusses when the mass transfer in both phases
must be analyzed and when one phase would dominate
the absorption process. He shows that this can be
determined by examining two sets of non-dimensional
groupings, one that pertains to the size of the mass
transfer boundary layers in the vapor and liquid and the
other that pertains to the mass transfer driving poten-
tials in each phase. For the dierent cases, Conlisk
develops expressions for the solutions which he admits
are ``not easy to solve'' in some cases. This is due in part
to the fact that, for ammoniawater, the interface equi-
librium condition includes one more unknown. Conlisk
and Mao [72] develop the solutions further and present
plots of absorption rate, lm concentration and tem-
perature, and heat and mass transfer coecients against
non-dimensional time at various positions around a
horizontal tube.
5.3.4. Potnis, Gomezplata, Papar, Anand and Erickson
1997 [153]
Potnis et al. [153] use a modied version of the Col-
burnDrew method for analyzing a coiled, uted-tube
ammonia-water GAX absorber. Absorption occurs on
the outside of the coiled tubes with falling lm ow
around the tubes. Unlike the previous applications of
the ColburnDrew method to ammonia-water absor-
bers, Potnis et al. do not assume that the liquid lm is
fully mixed. Instead they estimate the mass transfer
coecient in the falling lm from heat and mass transfer
analogies. Plots of temperatures, concentrations, heat
and mass uxes and transfer coecients against down-
stream position illustrate the results of the model for
one operating condition. Potnis et al. note that ``the
vapor-phase mass transfer coecients are signicantly
lower than those of the liquid phase, indicating that the
mass transfer resistance resides primarily in the vapor
phase. However, the estimates based on the bulk and
diusional transport show that the liquid-phase mass
transfer resistance can be as high as 25% of the total
resistance and, therefore, should not be considered neg-
ligible compared to the vapor-phase resistance for an
ammonia-water system.''
5.3.5. Kim 1998 [126]
Kim [126] numerically solves the problem of ammo-
nia-water absorption on a smooth vertical wall by
assuming transfer coecients for the vapor phase and
proles of velocity, temperature, and concentration in
the liquid phase. The velocity prole within the lm
accounts for interfacial shear between the vapor and the
lm. The proles for lm velocity, temperature, and
concentration are all assumed to be parabolic, and the
mass transfer is assumed to be conned to a boundary
layer near the lm surface (Kim states that the con-
centration boundary layer may grow to 50% of the
liquid lm thickness in the situation considered). The
velocity and temperature proles are assumed to be fully
developed at the inlet to the absorber, but the boundary
layer thickness for the mass transfer is a calculated
quantity, as is the lm thickness. Furthermore, equili-
brium is assumed to prevail at the interface, the wall is
assumed to be either isothermal or adiabatic, and the
thermophysical properties are evaluated at the local
bulk temperatures and concentrations. A baseline oper-
ating condition is modeled to investigate the eect of co-
and counter-current vapor ow. The eect of inlet lm
and vapor concentration, wall conditions, and vapor
ow rate on absorption is also presented. Kim shows
that in many cases, the model predicts that the mass of
water evaporating from the lm is greater than the mass
of ammonia being absorbed. Additionally in some cir-
cumstances, both ammonia and water may be evapor-
ating which Kim suggests can be avoided by proper
control of operating conditions such as vapor velocity.
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 789
He also concludes that ``the bulk motion of vapor con-
trols the ammonia mass ux in the vapor mixture'' but
goes on to say ``the main mass transfer resistance lies in
the solution lm''. These results appear to be somewhat
divergent from similar types of models.
5.3.6. Gommed, Grossman and Koenig 1999, 2000
[15,154]
Gommed et al. [15] develop a model for ammonia-
water absorption on a falling lm inside a vertical tube
with co-current vapor ow. The dierential equations of
momentum, heat, and mass transfer are solved simulta-
neously using an implicit nite-volume method. A
coordinate transformation is applied to the governing
equations which eliminates discontinuities and ensures a
rectangular solution domain in spite of the changing lm
thickness. The model allows for interfacial shear between
the two phases, and downstream pressure changes are
calculated; presumably the interface is smooth and ow
is laminar in both phases. Other assumptions the authors
use are that the pressure and thermophysical properties
are constant in the radial direction, interface equilibrium
holds, and the wall temperature is known (isothermal
wall is used for the results presented). The energy equa-
tion is formulated to allow for the ux of energy due to
interdiusion, although the authors conclude that this
eect is negligible except for a small region near the lm
inlet. After applying the coordinate transformation
which is based on the axisymmetric geometry and the
stream function, the three equations governing heat,
mass, and momentum transfer are presented in a single
general form; the generic variables in this expression are
then dened for each of the three transfer processes. The
authors outline the computational steps for the iterative
solution of the problem. For each iteration, either initial
guesses or results of a previous iteration are used to rst
iteratively solve the momentum equations in the two
phases, then the energy equations (note heat transfer in
the vapor phase is not neglected), and nally the mass
transfer equations. Results are presented for an operat-
ing condition dened by Kang and Christensen [136] for
the hydronically coupled absorber in a GAX system.
Gommed et al. illustrate that the interface temperature
remains virtually xed along the absorber as does the
temperature prole within the lm which is essentially
linear especially at greater downstream positions. Not
only does the bulk ammonia concentration in the liquid
lm increase with downstream position, but the ammo-
nia concentration at the interface does as well pre-
sumably due to changing pressure. Both the absorption
rate and heat transfer coecient in the liquid lm are
relatively high near the inlet but reduce to relatively
constant values at greater downstream positions. It is
shown that decreasing the wall temperature, entrance
liquid bulk temperature, or entrance liquid bulk ammo-
nia content increases the absorption rate. It is also
shown that the absorption rate increases slightly as mass
diusivity in the liquid phase is increased, a quantity the
authors suggest is not well dened. The authors also
state that absorption rate increases with vapor ow rate
which is consistent with the conclusion of Potnis et al.
[153] that the dominant resistance to mass transfer
occurs in the vapor phase.
Gommed et al. [154] again model co-current absorp-
tion on a laminar falling-lm with smooth interface
inside a vertical tube. They describe their model as ``an
improvement on and extension of'' the model of Gom-
med et al. [15]. The fundamental assumptions are iden-
tical and the methodology is similar. The authors use
the implicit nite volume method and coordinate trans-
formation techniques of Patankar and Spalding [155].
They also include a more accurate numerical scheme for
the interdiusion terms where Gommed et al. [15] used
approximations. Using the same baseline condition as
before, the authors show that the inlet conditions of the
liquid and vapor aect the absorption process sig-
nicantly. The authors note that, near the inlet, the
subcooling of the lm is more important than the actual
wall temperature and suggest it is worth considering
short, adiabatic absorbers with subcooled liquid. The
results also show that the temperature prole in the
liquid lm becomes linear quickly and remains nearly
constant along the length of the lm (isothermal wall
assumed). The temperature proles in the vapor phase
reveal that convection is dominant. Again, the authors
show that, although the interface temperature remains
nearly constant, the interface (and bulk) concentration
of ammonia increases with downstream position. The
authors also demonstrate the eect of solution ow rate.
For a 50% reduction in lm ow rate, the absorption
rate is initially lower, but by the end of the absorber (1-
m downstream), the absorption rate has recovered and
surpassed the baseline case. For a doubled lm ow
rate, the overall absorption rate is higher than the base-
line case for any downstream position considered.
Another interesting conclusion the authors make is that
changing the solution thermal conductivity has a much
greater eect than changing its diusion coecient.
Furthermore, the authors show that, although increas-
ing the vapor ow rate increases the absorption rate, the
coecients of diusion and heat transfer through the
lm have a much stronger eect. They state that ``the
absorption process is controlled by the liquid solution
side,'' and ``the dominant conduction mechanism on the
liquid side causes the heat transfer coecient to
approach a constant value in the fully developed
region''. Finally, the authors again suggest that the
eect of interdiusion is negligible.
5.3.7. Summary of ammonia-water models
Table 5 contains a summary of ammonia-water
absorption models that do not rely on assumed transport
790 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
coecients. It can be seen that recent models are elim-
inating many of the standard simplications; unfortu-
nately the list of works in this area is quite short. And
conspicuously missing are models including wavy lms
and lms over horizontal tubes. As ammonia-water is a
viable and widely used working uid pair recognized as
a promising candidate for residential and light com-
mercial applications where both heating and cooling are
required [15], the explanation for this void may be the
added complexity of the problem. Certainly, this is an
area that needs further investigation.
6. Conclusions
A large body of work exists in the eld of modeling
falling lm absorption; however, there appears to be a
lack of agreement among authors in several areas and,
furthermore, there are still some key deciencies that
can only be addressed through continued enhancement
of the existing modeling techniques. The key areas that
need further investigation are itemized below:
. Film hydrodynamics is an area that clearly needs
improved understanding. The often conicting
conclusions of many of the authors clearly show
that this key assumption has an extremely large
impact on the predicted results. The reasons for
adopting a laminar lm assumption for the initial
work in this area were clear. But given the inher-
ent instability of falling lms and the fact that
absorbers are often designed specically to avoid
smooth laminar lm ow, continuing to use these
types of simplifying assumptions is not particu-
larly useful.
. In most of the existing work, the hydrodynamics
are assumed a priori, but as was mentioned
before, experimental evidence suggests the heat
and mass transfer processes may signicantly
aect the hydrodynamics by altering the local
thermophysical properties and gradients, espe-
cially near the interface. This suggests that the
hydrodynamics should be considered simultaneously
in a coupled manner with the heat and mass transfer
processes, although this certainly adds signicant
complexity to the problem. The studies that have
attempted to do this have been for smooth-inter-
face, laminar lms and may or may not allow for
property variation across the lm. None have
considered the case of a rippled or wavy lm
where the eects might be even more pronounced.
. Modeling absorption in the presence of wavy lm
ow is still an area where more work needs to be
done. The existing models make many simplifying
assumptions, have had little validation, and gen-
erally do not account for the development of the T
a
b
l
e
5
V
e
r
t
i
c
a
l
w
a
l
l
,
a
m
m
o
n
i
a
-
w
a
t
e
r
a
T
a
b
l
e
a
u
5
M
u
r
v
e
r
t
i
c
a
l
,
a
m
m
o
n
i
a
c
-
e
a
u
A
u
t
h
o
r
D
a
t
e
S
o
l
u
t
i
o
n
t
y
p
e
H
y
d
r
o
d
y
n
a
m
i
c
s
D
o
m
a
i
n
W
a
l
l
O
t
h
e
r
K
h
o
p
a
n
o
v
e
t
a
l
.
(
1
9
8
5
)
[
1
5
2
]
A
n
a
l
y
t
i
c
a
l
(
S
S
)
S
F
I
,
U
S
V
P
,
N
T
V
N
e
a
r
i
n
l
e
t
N
A
C
F
T
,
L
I
E
M
,
E
M
D
,
C
T
P
,
N
I
E
(
1
9
8
6
)
[
1
2
8
]
N
u
m
e
r
i
c
a
l
S
F
I
,
U
S
V
P
/
P
S
V
P
,
N
T
V
U
n
r
e
s
t
r
i
c
t
e
d
I
s
o
t
h
e
r
m
a
l
o
r
a
d
i
a
b
a
t
i
c
C
F
T
,
L
I
E
M
,
E
M
D
,
C
T
P
,
N
I
E
C
o
n
l
i
s
k
(
1
9
9
4
)
[
6
4
]
A
n
a
l
y
t
i
c
a
l
(
L
a
P
l
a
c
e
i
n
t
e
g
r
a
l
s
)
a
n
d
n
u
m
e
r
i
c
a
l
F
l
u
t
e
d
t
u
b
e
,
S
F
I
,
P
S
V
P
M
o
d
e
r
a
t
e
l
e
n
g
t
h
s
F
i
x
e
d
t
e
m
p
.
p
r
o

l
e
L
I
E
M
,
N
V
H
T
,
C
T
P
,
N
I
E
C
o
n
l
i
s
k
a
n
d
M
a
o
(
1
9
9
6
)
[
7
2
]
A
n
a
l
y
t
i
c
a
l
(
L
a
P
l
a
c
e
i
n
t
e
g
r
a
l
s
)
a
n
d
n
u
m
e
r
i
c
a
l
S
a
g
g
i
n
g

l
m
,
S
F
I
,
P
S
V
P
H
o
r
i
z
o
n
t
a
l
t
u
b
e
I
s
o
t
h
e
r
m
a
l
L
I
E
M
,
N
V
H
T
,
C
T
P
,
N
I
E
G
o
m
m
e
d
e
t
a
l
.
(
1
9
9
9
)
[
1
5
,
1
5
4
]
N
u
m
e
r
i
c
a
l
S
F
I
,
c
o
u
p
l
e
d
,
c
o
-
c
u
r
r
e
n
t
s
h
e
a
r
U
n
r
e
s
t
r
i
c
t
e
d
I
s
o
t
h
e
r
m
a
l
C
T
P
i
n
t
r
a
n
s
v
e
r
s
e
d
i
r
e
c
t
i
o
n
a
S
S
=
s
i
m
i
l
a
r
i
t
y
s
o
l
u
t
i
o
n
;
E
S
=
e
x
p
a
n
s
i
o
n
s
e
r
i
e
s
;
S
F
I
=
s
m
o
o
t
h

l
m
i
n
t
e
r
f
a
c
e
;
U
S
V
P
=
u
n
i
f
o
r
m
s
t
r
e
a
m
-
w
i
s
e
v
e
l
o
c
i
t
y
p
r
o

l
e
;
P
S
V
P
=
p
a
r
a
b
o
l
i
c
s
t
r
e
a
m
-
w
i
s
e
v
e
l
o
c
i
t
y
p
r
o

l
e
;
N
T
V
=
n
e
g
l
i
g
i
b
l
e
t
r
a
n
s
v
e
r
s
e
v
e
l
o
c
i
t
y
;
C
F
T
=
c
o
n
s
t
a
n
t

l
m
t
h
i
c
k
n
e
s
s
;
L
I
E
M
=
l
i
n
e
a
r
i
n
t
e
r
f
a
c
e
e
q
u
i
l
i
b
r
i
u
m
m
o
d
e
l
;
E
M
D
=
e
q
u
i
m
o
l
a
r
d
i

u
s
i
o
n
a
t
i
n
t
e
r
f
a
c
e
;
N
V
H
T
=
n
e
g
l
i
g
i
b
l
e
v
a
p
o
r
-
p
h
a
s
e
h
e
a
t
t
r
a
n
s
f
e
r
;
C
T
P
=
c
o
n
s
t
a
n
t
t
h
e
r
m
o
p
h
y
s
i
c
a
l
p
r
o
p
e
r
t
i
e
s
;
N
I
E
=
n
e
g
l
i
g
i
b
l
e
i
n
t
e
r
d
i

u
s
i
o
n
e

e
c
t
s
;
L
F
T
P
=
l
i
n
e
a
r

l
m
t
e
m
p
e
r
a
t
u
r
e
p
r
o

l
e
.
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 791
waves along the absorber, for the possibility of
uid recirulation within the wave, or for the non-
periodicity of actual waves at higher Reynolds
numbers. Most models also simply assume and
impose velocity and thickness proles within the
lm rather than deriving them.
. Models for absorption on horizontal tubes have
not included waves, there are few studies on the
inception and development of waves on horizontal
tubes, and the few studies on droplets use empirical
transport correlations. The literature lacks much
useful information on absorption phenomena in
pendant forming droplets, but this region may in
fact account for a large portion of the total mass
transfer in horizontal tube systems. Additionally,
models of horizontal-tube absorbers should
include the eect of droplet impact on the top of
tubes and the resulting concentration redistribu-
tion and heat transfer enhancement. In general,
much more work has been done on vertical lm
absorption, although horizontal tubes are more
widely used in absorber designs.
. Consideration of surface wetting limits are rarely
incorporated into absorption models and some-
times lead to suggestions for impracticably low
lm ow rates. When the eect of surface wetting
is included, it is usually by empirical correlation
with experimental results, not by predictive meth-
ods. Physically representative, predictive methods
for surface wetting may also require the inclusion
of the local thermophysical property variation in
the presence of absorption, which requires simul-
taneous solution with the heat and mass transfer
problem, to account for eects such as surface
tension gradients formed by preferential absorp-
tion. Additionally, fully three-dimensional models
may be required to account for the spatially and
temporally varying wetting phenomena.
. The interaction of vapor ow and lm ow has only
been included as simple shear forces on a smooth
interface, although some of the transport coe-
cients assumed may, in part, account for this eect.
In some instances, it may be warranted to include
the vaporliquid hydrodynamic interaction in more
detail, especially in the presence of surface waves.
. There appears to be a general agreement that
vapor pressure equilibrium prevails at the interface,
although one model was developed without this
assumption. A linear model of interface equili-
brium appears to be reasonable, though not ideal,
for most operating conditions in a waterlithium
bromide absorber. Less evidence is available on
the eect of this assumption in ammonia-water
systems. It may be possible to use more realistic
equilibrium models in numerical analyses with the
advanced capabilities of today's computers.
. In general, ammonia-water absorption models are
not as well developed as those for waterlithium
bromide systems, especially for wavy lms, tur-
bulent lms, and lms on horizontal tubes.
. Neglecting heat transfer to the vapor phase is a
very common assumption. However, some inves-
tigators have concluded that heat transfer to the
vapor phase is quite important, particularly in
situations with a large temperature dierence
between the liquid lm and vapor and/or high
vapor-phase heat transfer coecients.
. Simplications to the governing equations must be
considered carefully. As models begin to include
greater detail of spatial variation of parameters, it
may be necessary to reevaluate some of the math-
ematical simplications that are often made by
assuming constant thermophysical properties
(including constant total mass density of each
phase) and negligible transverse velocities. Addi-
tionally, care should be taken, when applying
the laws of diusion, to account for the frame
of reference which is used at the interface. This
was a point of inconsistency among several
authors.
. It is often dicult to compare the predictions of
dierent models to each other because of dier-
ences in the assumed geometries and operating
conditions and the limited amount of predicted
data that is typically presented. It would be bene-
cial if authors presented the results in a con-
sistent format at a variety of operating conditions,
including those used in previous work, and with
enough detail that meaningful comparisons could
be made. Also, parametric studies are typically
done in one-at-a-time fashion from a chosen
baseline condition. Investigation of the potential
interaction between model parameters has only
been done on a limited basis but may shed light
on why investigators sometimes come to conict-
ing conclusions.
. Experimental validation is very limited and usually
based on overall performance predictions. As was
illustrated as early as 1947 by Ruhemann, sig-
nicant variations in transfer coecients can
combine to net only slight changes in outlet con-
ditions despite large changes within the absorber.
Using overall performance as a basis for evaluating
model accuracy masks the real limitations of a
model, particularly when the absorption resistance
is not the dominant resistance. This implies that
future experimental studies must be conducted
with a particular emphasis on obtaining data with
enough detail to validate the local transfer pro-
cesses at many locations within the absorber, not
just overall transport values. This trend is now
starting to appear in the literature.
792 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
. It should be noted that this review does not
include the eects of heat and mass transfer
enhancing additives that are often an important
and required aspect of achieving the performance
found in modern absorbers. Incorporating the
eect of these additives, which often work by
altering the surface tension and promoting lm
waviness, will be important to the applicability of
these models.
Acknowledgements
The authors gratefully acknowledge support from the
National Science Foundation through Grant Number
9875010 for this research.
References
[1] Michel JW, Perez-Blanco H. Inuence of surfactant
additive on absorption heat pump performance. ASH-
RAE Transactions 1985;91(Part 2B):1847.
[2] Jung SH, Sgamboti C, Perez-Blanco H. An experimental
study of the eect of some additives on falling lm
absorption. International Absorption Heat Pump Con-
ference 1993;31:4955.
[3] Kim KJ, Berman NS, Wood BD. Experimental investi-
gation of enhanced heat and mass transfer mechanisms
using additives for vertical falling lm absorber. AES
International Absorption Heat Pump Conference
1993;31:417.
[4] Ziegler F, Grossman G. Heat-transfer enhancement by
additives. International Journal of Refrigeration
1996;19(5):301309.
[5] Kulankara S, Verma S, Herold KE. Theory of heat/mass
transfer additives in absorption chillers. Proceedings of
the ASME Advanced Energy Systems Division, AES
1999;39:199206.
[6] Kenig YA, Kholpanov LP, Malyusov VA, Zhavoronkov
NM. Calculation of multicomponent mass exchange in
the presence of inert components. Institute of New Che-
mical Problems (English translation of Teoreticheskie
Osnovy Khimicheskoi Tekhnologii) 1983;16(6):47784.
[7] Yang R, Chen JH. A numerical study of the non-
absorbable eects on the falling liquid lm absorption.
Warme-und Stoubertragung 1991;26(4):21923.
[8] Chen W, Vliet GC. Eect of an inert gas on the heat and
mass transfer in a vertical channel with falling lms.
Solar Engineering 1995;2:124957.
[9] Ameel TA, Habib HM, Wood BD. Eects of a non-
absorbable gas on interfacial heat and mass transfer for
the entrance region of a falling lm absorber. Journal of
Solar Energy Engineering 1996;118:459.
[10] Grossman G, Gommed K. Heat and mass transfer in lm
absorption in the presence of non-absorbable gases.
International Journal of Heat and Mass Transfer
1997;40(15):3595606.
[11] Yang R, Jou TM. Non-absorbable gas eect on the wavy
lm absorption process. International Journal of Heat
and Mass Transfer 1998;41:365768.
[12] Bird RB, Stewart WE, Lightfoot EN. Transport Phe-
nomena. New York: Wiley, 1960.
[13] Delancey GB, Chiang SH. Dufour eect in liquid sys-
tems. AIChE Journal 1968;14(4):6645.
[14] Grossman G. Analysis of interdiusion in lm absorp-
tion. International Journal of Heat and Mass Transfer
1987;30(1):2058.
[15] Gommed K, Grossman G, Koenig M. Numerical model
of ammonia-water absorption inside a vertical tube. Pro-
ceedings of the International Sorption Heat Pump Con-
ference, Munich, Germany 2426 March 1999: 27581.
[16] Ruheman, M. (1947). The transfer of heat and matter in
an ammonia absorber. Transactions of the Institute of
Chemical Engineers, 1947: 1720.
[17] Nakoryakov VE, Grigor'eva NI. Combined heat and
mass transfer during absorption in drops and lms.
Journal of Engineering Physics 1977;32(3):2437.
[18] Herold KE, Radermacher R, Klein SA. Absorption
Chillers and Heat Pumps. Boca Raton: CRC Press LLC,
1996.
[19] Nakoryakov VE, Grigor'eva NI. Calculation of heat and
mass transfer in nonisothermal absorption of the initial
portion of a downowing lm. Theoretical Foundations
of Chemical Engineering (English translation of Teor-
eticheskie Osnovy Khimicheskoi Tekhnologii) 1980;14
(4):4838.
[20] Nakoryakov VY, Grigor'yeva NI. Combined heat and
mass transfer in lm absorption. Heat Transfer-Soviet
Research 1980;12(3):1117.
[21] Nakoryakov V, Ye, Burdukov AP, Bufetov NS, Gri-
gor'eva NI, Dorokhov AR. Coecient of heat and mass
transfer in falling wavy liquid lms.. Heat Transfer-
Soviet Research 1982;14(3):611.
[22] Nakoryakov V, Ye, Bufetov NS, Grigor'yeva NI. Heat
and mass transfer in lm absorption.. Fluid Mechanics-
Soviet Research 1982;11(3):97115.
[23] Nakoryakov VE, Grigoryeva NI. Heat and mass transfer
in lm absorption. Russian Journal of Engineering
Thermophysics 1992;2:116.
[24] Nakoryakov VE, Grigor'eva NI. Heat and mass transfer
in lm absorption with varying liquid-phase volume.
Theoretical Foundations of Chemical Engineering
1995;29(3):2238.
[25] Nakoryakov VE, Grigor'eva NI. Analysis of exact solu-
tions to heat- and mass-transfer problems for absorption
with lms or streams. Theoretical Foundations of Che-
mical Engineering 1997;31(2):11926.
[26] Grigoryeva NI, Nakoryakov VE. Exact solution of com-
bined heat and mass transfer problems during lm absorp-
tion. Journal of Engineering Physics 1977;33(5):134953.
[27] Brauner N. Modelling of wavy ow in turbulent free
falling lms. Int J Multiphase Flow 1989;15(4):50520.
[28] Patnaik V, Perez-Blanco H. A study of absorption
enhancement by wavy lm ows. International Journal
of Heat and Fluid Flow 1996;17(1):717.
[29] Grossman G. Simultaneous heat and mass transfer in
lm absorption under laminar ow. International Jour-
nal of Heat and Mass Transfer 1983;26(3):35771.
[30] Wassenaar RH, Westra JJW. Dynamic model of a lm
absorber with coupled heat and mass transfer. International
Journal of Heat and Mass Transfer 1992;35(1):8799.
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 793
[31] Ibrahim GA, Vinnicombe GA. A hybrid method to ana-
lyse the performance of falling lm absorbers. Interna-
tional Journal of Heat and Mass Transfer
1993;36(5):138390.
[32] Habib HM, Wood BD. Simultaneous heat and mass
transfer for a falling lm absorber the two phase ow
problem. Solar Energy Engineering, Proceedings of the
12th Annual International Solar Energy Conference,
ASME, p. 6167, 1990.
[33] Kholpanov LP, Malyusov VA, Zhavoronkov NM.
Hydrodynamics and heat- and mass-transfer in a liquid
lm in the presence of a gas stream or of surface tension.
Theoretical Foundations of Chemical Engineering (Eng-
lish translation of Teoreticheskie Osnovy Khimicheskoi
Tekhnologii) 1982;16(3):2017.
[34] Grossman G. Simultaneous heat and mass transfer in
absorption/desorption of gases in laminar liquid lms.
Proceedings of the AIChE Winter Annual Meeting,
Orlando, FL, 1982.
[35] Grossman G. Heat and mass transfer in lm absorption.
Handbook of Heat and Mass Transfer, Vol. 2, Chapter
6, pp. 21157, Houston: Gulf Publishing, 1986.
[36] Emmert RE, Pigford RL. Chemical Engineering Progress
1954;50:87.
[37] Andberg JW, Vliet GC. Nonisothermal absorption of
gases into falling liquid lms. ASME-JSME Thermal
Engineering Joint Conference Proceedings 1983;2:42331.
[38] Andberg JW, Vliet GC. Design guidelines for water-
lithium bromide absorbers. ASHRAE Transactions
1983;89:2.
[39] Andberg JW. Absorption of vapors into liquid lms fall-
ing over cooled horizontal tubes. PhD Dissertation,
University of Texas, Austin, 1986.
[40] Andberg JW, Vliet GC. Absorption of vapors into liquid
lms owing over cooled horizontal tubes. Second
ASME-JSME Thermal Engineering Joint Conference,
Honolulu, Hawaii, 2227 March, Vol. 2, No. 0, pp. 533
41, 1987a.
[41] Andberg JW, Vliet GC. A simplied model for absorption
of vapors into liquid lms owing over cooled horizontal
tubes. ASHRAE Transactions, 1987;93(2):245463.
[42] McNeely LA. Thermophysical properties of aqueous
solutions of lithium bromide. ASHRAE Transactions
1979;85:41334.
[43] Burdukov AP, Bufetov NS, Dorokov AR. Absorption
on a falling liquid lm. Izv Sib Otd Akad Nuak SSSR,
seriya tekhn Nauk 1979;13(3):4852.
[44] Cosenza F, Vliet GC. Absorption in falling water/LiBr
lms on horizontal tubes. ASHRAE Transactions
1990;96:1.
[45] Kim KJ, Berman NS, Chau DSC, Wood BD. Absorp-
tion of water vapour into falling lms of aqueous lithium
bromide. International Journal of Refrigeration 1995;
18(7):48694.
[46] Kashiwagi T. The activity of surfactant in high-perfor-
mance absorber and absorption enhancement. Reito,
1985; 60(687): 729 (in Japanese, referenced by Kim et
al., 1995).
[47] Enderby JE. Neutron scattering from ionic solutions.
Annual Reviews in Physics and Chemistry 1983; 34: 155
85 (referenced by Kim et al., 1995).
[48] Le Go H, Ramadane A, Barkaoui M, Chen Y, Le Go
P. Modeling the coupled heat and mass transfer in a
falling lm: application to absorbers and desorbers in
absorption heat pumps. Heat Transfer, Proceedings of the
International Heat Transfer Conference, Vol. 4, pp 1971
1976, New York: Hemisphere Publishing Corp. 1986a.
[49] Le Go H, Ramadane A, Le Go P. Modelisations des
transferts couples de matiere et de chaleur dans
l'absorption gas-liquide en lm ruisselant laminaire.
International Journal of Heat and Mass Transfer
1985;28(11):200517.
[50] Le Go H, Ramadane A, Le Go P. Un modele simple
de la penetration couplee de chaleur et de matiere dans
l'absorption gasliquide en lm ruisselant laminaire.
International Journal of Heat and Mass Transfer
1986;29(4):62534.
[51] Kawae N, Shigechi T, Kanemaru K, Yamada T. Water
vapor evaporation into laminar lm ow of a lithium
bromide-water solution (inuence of variable properties
and inlet lm thickness on absorption mass transfer rate).
Heat Transfer Japanese Research, 1989; 18(3): 5870.
[52] Brauner N, Moalem Maron D, Meyerson H. Coupled
heat condensation and mass absorption with comparable
concentrations of absorbate and absorbent. International
Journal of Heat and Mass Transfer 1989;32(10):1897906.
[53] Brauner N, Moalem Maron D, Meyerson H. The eect
of absorbate concentration level in hygroscopic con-
densation. International Communications Heat and
Mass Transfer 1988;15(3):26979.
[54] Brauner N. Non-isothermal vapour absorption into fall-
ing lm. Int J Heat and Mass Transfer 1991;34(3):767
84.
[55] van der Wekken BJC, Wassenaar RH. Simultaneous heat
and mass transfer accompanying absorption in laminar
ow over a cooled wall. Int J Refrig 1988;11:707.
[56] Patankar SV. Numerical heat transfer and uid ow.
New York: Hemisphere Publishing Corp, 1980.
[57] Yang R, Wood BD. A numerical modeling of an
absorption process on a liquid falling lm. Solar Energy
1992;48(3):1958.
[58] Hajji A, Worek WM. Transient heat and mass transfer in
lm absorption of nite depth with nonhomogeneous
boundary conditions. International Journal of Heat and
Mass Transfer 1992;35(9):21018.
[59] Hajji A, Lavan Z. Eect of lm depth and interfacial
convection on transient heat and mass transfer in lm
absorption. Heat Transfer 1990, Proceedings of the
International Heat Transfer Conference, p. 347352,
Hemisphere Publishing Corp., 1990.
[60] Ramadane A, Aoufoussi Z, Le Go H. Experimental
investigation and modeling of gas-liquid absorption with
a high thermal eect. Distillation and Absorption, Insti-
tution of Chemical Engineers Symposium Series
1992;1:A4510A459.
[61] Conlisk AT. Falling lm absorption on a cylindrical
tube. AIChE Journal 1992;38(11):171628.
[62] Conlisk AT. The use of boundary layer techniques in
absorber design. ASME Proceedings of the International
Absorption Heat Pump Conference 1993;31:16370.
[63] Conlisk AT. Semi-analytical design of a falling lm
absorber. Journal of Heat Transfer 1994;116:10558.
794 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
[64] Conlisk AT. Structure of falling lm heat and mass
transfer on a uted tube. AIChE Journal 1994;40(5):756
66.
[65] Conlisk AT. Analytical solutions for the heat and mass
transfer in a falling lm absorber. Chemical Engineering
Science 1995;50(4):65160.
[66] Conlisk AT. The eect of coolant ow parameters on the
performance of an absorber. ASHRAE Transactions
1995;101(2):7380.
[67] Gnielinski V. New equations for heat and mass transfer
in turbulent pipe and channel ow. International Che-
mical Engineering 1976;16:359.
[68] Patnaik V, Perez-Blanco H, Ryan WA. A simple analy-
tical model for the design of vertical tube absorbers.
ASHRAE Transactions 1993;99(2):6980.
[69] Choudhury SK, Hisajima D, Ohuchi T, Nishiguchi A,
Fukushima T, Sakaguchi S. Absorption of vapors into
liquid lms dlowing over cooled horizontal tubes. ASH-
RAE Transactions 1993;99(2):819.
[70] Lu Z, Li D, Li S, Yu-Chi B. A semi-empirical model of
the falling lm absorption outside horizontal tubes. Pro-
ceedings of the International Ab-Sorption Heat Pump
Conference 1996;2:47380.
[71] Jernqvist A, Kockum H. Simulation of falling lm
absorbers and generators. Proceedings of the Interna-
tional Ab-Sorption Heat Pump Conference Montreal,
Canada, 1720 September, Vol. 1, p. 3118, 1996.
[72] Conlisk AT, Mao J. Nonisothermal absorption on a
horizontal cylindrical tube 1. The Film Flow. Chemi-
cal Engineering Science 1996;51(8):127585.
[73] Gregorig R. Hauktkondensation an Funegewellton
Oberachen bei Berucksichtigung der Ober-
achenspannungen. Z Angew Math Phys 1954;5:36.
[74] Johnson RE, Conlisk AT. Laminar lm condensation/
evaporation on a vertically uted tube. Journal of Fluid
Mechanics 1987;184:254.
[75] Yang R, Jou D. Heat and mass transfer of absorption
process for the falling lm inside a porous medium.
International Journal of Heat and Mass Transfer
1995;38(6):11216.
[76] Conlisk AT. Prediction of the performance of a splined-
tube absorber Part 1: governing equations and
dimensional analysis. ASHRAE Transactions
1996;102(1):11627.
[77] Conlisk AT. A model for the prediction of the perfor-
mance of a splined-tube absorber Part 2: model and
results. ASHRAE Transactions 1996;102(1):12837.
[78] Miller WA. The experimental analysis of aqueous lithium
bromide vertical falling lm absorption. Dissertation,
University of Tennessee, Knoxville, 1998.
[79] Grossman G, Heath MT. Simultaneous heat and mass
transfer in absorption of gases in turbulent liquid lms.
International Journal of Heat and Mass Transfer
1984;27(12):236576.
[80] Grossman G, Heath MT. Errata. International Journal
of Heat and Mass Transfer 1986;29(3):505.
[81] van Driest ER. On turbulent ow near a wall. Jounal of
Aeronautical Science 1956;23:100711.
[82] Reichardt Z. Vollstandige Darstellung der Turbulenten
Geschwindigkeitsverteilung in Glatten Leitungen. Z
Angew Math Mech 1951;31:208.
[83] Lamourelle AP, Sandall OC. Gas absorption into a tur-
bulent liquid. Chemical Engineering Science
1972;27:103543.
[84] Faghri A, Seban R. Heat and mass transfer to a turbu-
lent liquid lm. International Journal of Heat and Mass
Transfer 1988;31(4):8914.
[85] Faghri A, Seban R. Heat and mass transfer to a turbu-
lent falling lm II. International Journal of Heat and
Mass Transfer 1989;32(9):17968.
[86] Yu ksel ML, Schlu nder EU. Heat and mass transfer in
non-isothermal absorption of gases in falling lms, Part
II: theoretical description and numerical calculation of
turbulent falling lm heat and mass transfer. Chemical
Engineering and Processing 1987;22(4):20313.
[87] Seban RA, Faghri A. Evaporation and heating with tur-
bulent falling liquid lms. Journal of Heat Transfer,
Transactions ASME 1976;98:315.
[88] Hubbard GL, Mills AF, Chung DK. Heat transfer across
a turbulent falling lm with cocurrent vapor ow. Journal
of Heat Transfer, ASME Transactions 1976;98:31920.
[89] Carrubba G. Dissertation, TU Berlin, 1976.
[90] Blangetti F. Dissertation, Universitat Karlsruhe (TH),
1979.
[91] Kholpanov LP, Kenig EY. Coupled mass and heat
transfer in a multicomponent turbulent falling liquid
lm. International Journal of Heat Mass Transfer
1993;36(14):364757.
[92] Fulford GD. The ow of liquids in thin lms. Adv Che-
mical Engineering 1964;5:151.
[93] Kapitza PL. Wave ow of thin layers of a viscous uid.
Collected Papers of P.L. Kapitza, Oxford: Pergamon
Press, 1965.
[94] Benjamin TB. Wave formation in laminar ow down an
inclined plane. Journal of Fluid Mechanics 1957;2:55474.
[95] Wasden FK, Dukler AE. Insights into the hydro-
dynamics of free falling wavy lms. AIChE Journal
1989;35(9):137990.
[96] Jayanti S, Hewitt GF. Hydrodynamics and heat transfer
of wavy thin lm ow. International Journal of Heat and
Mass Transfer 1997;40(1):17990.
[97] Miller WA, Keyhani M. The correlation of coupled heat
and mass transfer experimental data for vertical falling
lm absorption. Proceedings of the ASME Advanced
Energy Systems Division, Vol. 39, Session 7B, 1999.
[98] Morioka I, Kiyota M, Nakao R. Absorption of water
vapor into a lm of aqueous solution of LiBr falling
along a vertical pipe. JSME International Journal Series
B 1993;36(2):3516.
[99] Burdukov AP, Bufetov NS, Deriy NP, Dorokhov AR,
Kazakov VI. Experimental study of the absorption of
water vapor by thin lms of aqueous lithium bromide.
Heat Transfer Soviet Research 1980;12(3):11823.
[100] Kholpanov AP, Shkadov VYa, Zhavoronkov NM. Mass
transfer in a wavy liquid lm. Teor Osnovy khim Tekhn
1967;1(1):735.
[101] Alekseenko SV, Nakoryakov VYe, Pokusaev BG. Waves
on the surface of a vertically falling liquid lm. Preprint of
the Thermophysics Institute, Siberian Division of the USSR
Academy of Sciences, Nos. 3679, Novosibirsk, 1979.
[102] Brauer H. Stro mung und Wa rmeu bergang bei Riesell-
men. VDI Forschungsheft, Dusseldorf, 40 S, 1956.
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 795
[103] Kholpanov LP. Heat and mass transfer and uid mechan-
ics in thin liquid lms. Theoretical Foundations of Chemi-
cal Engineering (English translation of Teoreticheskie
Osnovy Khimicheskoi Tekhnologii) 1987;21(1): 5562.
[104] Kholpanov LP. Heat and mass transfer in systems with a
phase interface. Fluid Mechanics-Soviet Research 1990;
19(1):10918.
[105] Tsvelodub OYu. Nonisothermal absorption of a gaseous
impurity in a wavy liquid lm. Soviet Journal of Applied
Physics 1989;3(4):549.
[106] Uddholm H, Setterwall F. Model for dimensioning a
falling lm absorber in an absorption heat pump. Inter-
national Journal of Refrigeration 1988;11:415.
[107] Brauner N, Moalem Maron D. Modeling wavy ow in
inclined thin lms. Chemical Engineering Science
1983;38(5):77588.
[108] Brauner N, Dukler A, Moalem Maron D. Interfacial
structure of thin falling lms: piecewise modelling of the
waves. Physico-Chemical Hydrodynamics 1985;6:87113.
[109] Morioka I, Kiyota M. Absorption of water vapor into a
wavy lm of an aqueous solution of LiBr. JSME Inter-
national Journal Series II 1991;34(2):1838.
[110] Pierson FW, Whitaker S. Some theoretical and experi-
mental observations of the wave structure of falling
liquid lms. Industrial and Engineering Chemistry Fun-
damentals 1977;14(4):401.
[111] Yang R, Jou D, Chen J. Numerical study of heat and
mass transfer on the wavy lm absorption process. Pro-
ceedings of the Intersociety Energy Conversion Engi-
neering Conference 1991;2:55965.
[112] Yang R, Jou D. Heat and mass transfer on the wavy lm
absorption process. Canadian Journal of Chemical
Engineering 1993;71:5338.
[113] Yang R, Wood BD. A numerical solution of the wavy
motion on a falling liquid lm. Canadian Journal of
Chemical Engineering 1991;69(3):7238.
[114] Sabir H, Suen KO, Vinnicombe GA. Investigation of
eects of wave motion on the performance of a falling
lm absorber. International Journal of Heat and Mass
Transfer 1996;39(12):246372.
[115] Penev V, Krylov SV, Boyadjiev CH, Vorotilin VP. Wavy
ow of thin liquid lm. International Journal of Heat
and Mass Transfer 1968;15:1389406.
[116] Patnaik V, Perez-Blanco H. Roll waves in falling dilms:
an approximate treatment of the velocity eld. Interna-
tional Journal of Heat and Fluid Flow 1996;17(1):6370.
[117] Wasden FK, Dukler AE. A numerical study of mass
transfer in free falling waves. AIChE Journal 1990;
36(9):137990.
[118] Patnaik V, Perez-Blanco H. A counterow heat-exchan-
ger analysis for the design of falling-lm absorbers.
ASME International Absorption Heat Pump Conference
1993;31:20916.
[119] Kirby MJ, Perez-Blanco H. A design model for hor-
izontal tube water/lithium bromide absorbers. ASME
Heat Pump and Refrigeration Systems Design, Analysis
and Applications 1994;32:110.
[120] Taghavi-Tafreshi K, Dhir VK. Taylor instability in boil-
ing, melting and condensation or evaporation. Interna-
tional Journal of Heat and Mass Transfer 1980;23:1433
45.
[121] Clift R, Grace JR, Weber ME. Bubbles, Drops and Par-
ticles. New York: Academic Press, 1978.
[122] Atchley JA, Perez-Blanco H, Kirby MJ, Miller WA. An
experimental and analytical study of advanced surfaces
for absorption chiller absorbers. Gas Research Institute
Report GRI-95/0498, 1998.
[123] Tsai B, Perez-Blanco H. Limits of mass transfer
enhancement in lithium bromide-water absorbers by
active techniques International Journal of Heat and
Mass Transfer 1998;41(15):240916.
[124] Perez-Blanco H. A model of an ammonia-water falling
lm absorber. ASHRAE Transactions 1988;94(1):46783.
[125] Kang YT, Chen W, Christensen RN. A generalized com-
ponent design model by combined heat and mass transfer
analysis in NH3-H20 absorption heat pump systems.
ASHRAE Transactions: Symposia, 1997: 44453.
[126] Kim B. Heat and mass transfer in a falling lm absorber
of ammonia-water absorption systems. Heat Transfer
Engineering 1998;19(3):5363.
[127] Garimella, S. Miniaturized heat and mass transfer tech-
nology for absorption heat pumps. Proceedings of the
International Sorption Heat Pump Conference, Munich,
Germany, 2224 March 1999: 66170.
[128] Kholpanov LP, Kenig EYa, Malyusov VA. Combined
heat and mass transfer for opposing lm ows of liquid
and gas.. Journal of Engineering Physics (English trans-
lation of Inzhenerno-Fizicheskii Zhurnal) 1986;51(1):
768773.
[129] Briggs SW. Concurrent, crosscurrent, and countercurrent
absorption in ammonia-water absorption refrigeration.
ASHRAE Transactions 1971;77(1):1715.
[130] Danckwerts PV. Applied Scientic Research, 3-A:385
[131] Carslaw HS, Jaeger JC. Conduction of heat in solids. 2nd
ed. England: Clarendon Press, Oxford University Press,
1959 (pp. 76).
[132] Chiang SH, Toor HL. Gas absorption accompanied by a
large heat eecct and volume change of the liquid phase.
AIChE Journal 1964;10(3):398402.
[133] Verma SL, Delancey GB. Thermal eects in gas absorp-
tion. AIChE Journal 1975;21(1):96102.
[134] Colburn AP, Drew TB. The condensation of mixed
vapours. AIChE Transactions 1937;33:197212.
[135] Price BC, Bell KJ. Design of binary vapor condensers
using the colburn-drew equations. AIChE Symposium
Series-Heat Transfer-Research and Design 1974;70(138):
163171.
[136] Kang YT, Christensen RN. Development of a counter-
current model for a vertical uted tube GAX absorber.
ASME International Absorption Heat Pump Conference
Proceedings, p. 716, 1993.
[137] Sherwood TK, Pigford RL. Absorption and extraction.
New York: McGraw-Hill, 1952 (pp. 94).
[138] Chilton TH, Colburn AP. Mass transfer (absorption)
coecients: prediction from data on heat transfer and
uid friction. Industrial Engineering Chemistry 1934;
26(11):11837.
[139] Nusselt W. Die oberaechenkondensateion des Wasser-
dampfes. VDI Zeitschrift 1916; 60: 5416, 56975.
[140] Bell KJ, Ghaly MA. An approximate generalized design
method for multicomponent/partial condensers. AIChE
Symposium Series 1973;69(131):729.
796 J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797
[141] Kang YT, Christensen RN, Vafai K. Analysis of absorp-
tion process in a smooth-tube heat exchanger with a porous
medium. Heat Transfer Engineering 1994;15(4): 4255.
[142] Kang YT, Christensen RN. Combined heat and mass
transfer analysis for absorption in a uted tube with a
porous medium in conned cross ow. ASME Thermal
Engineering Conference 1995;1:25160.
[143] Kang YT, Chen W, Christensen RN. Design of a com-
bined component in GAX heat pump systems: absorber/
desorber. Proceedings of the International Heat Pump
Conference, Montreal, Quebec, Canada, 1720 Septem-
ber, Vol. 2, p. 45361, 1996.
[144] Chun KR, Seban RA. Heat transfer to evaporating
liquid lms. Journal of Heat Transfer 1971;91:3916.
[145] Garimella S, Christensen RN. Experimental investigation
of heat transfer characteristics of annuli with spirally
uted inner tubes. ASME Winter Meeting, HT-11D, 28
November New Orleans, LA, 1993.
[146] Palmer SC, Christensen RN. Experimental investigation and
model verication for a GAX absorber. Proceedings of the
International Heat Pump Conference, Montreal,Quebec,
Canada, 1720 September, Vol. 1, pp. 367374, 1985.
[147] Garrabrant MA, Christensen RN. Modeling and experi-
mental verication of a perforated plate-n absorber for
aqua-ammonia absorption systems. Proceedings of the
ASME Advanced Energy Systems Division, AES
1997;37:33747.
[148] Christensen RN, Garimella S, Kang YT, Garrabrant
MA. Perforated-n heat and mass transfer device. Patent
Number 5,704,417, 1998.
[149] Sherwood TK, Pigford RL, Wilke CR. Mass transfer.
New York: McGraw-Hill, 1975.
[150] Zhukauskas A. Heat transfer from tubes in cross ow. In
Hartnett JP, Irvine Jr., TF, editors. Advances in heat
transfer, Vol. 8. New York: Academic Press, 1972.
[151] Dorokhov AR, Bochagov VN. Heat transfer to a lm
falling over horizontal cylinders. Heat Transfer Soviet
Research 1983;15(2):96101.
[152] Kholpanov LP, Kenig E, Ya, Malyusov VA. Coupled
simultaneous heat and mass transfer in multicomponent
two-phase mixtures.. Journal of Engineering Physics
(English translation of Inzhenerno-Fizicheskii Zhurnal)
1985;49(3):105763.
[153] Potnis SV, Gomezplata A, Papar RA, Anand G, Erick-
son DC. GAX component simulation and validation.
ASHRAE Transactions 1997;103(1):4549.
[154] Gommed K, Grossman G, Koenig MS. Numerical study
of absorption in a laminar falling lm of ammonia-water.
Forthcoming in ASHRAE Transactions 2001.
[155] Patankar SV, Spalding DB. Heat and mass transfer in
boundary layers. London: Morgan-Grampian, 1967.
[156] Dittus FW, Boelter IMK. Publications on Engineer-
ing. Berkley: University of California, 1930 (Vol. 2, pp.
443).
J.D. Killion, S. Garimella / International Journal of Refrigeration 24 (2001) 755797 797

You might also like