0263–8762/02/$10.00+0.

00
# Institution of Chemical Engineers
Trans IChemE, Vol 80, Part A, April 2002

HEAT TRANSFER CHARACTERISTICS OF

ICE MELTING IN WATER AND SALT SOLUTIONS
G. L. QUARINI 1 and Y. C. CHANG 2
1
Mechanical Engineering Department, University of Bristol, UK.
2
School of Mechatronics, Korea University of Technology and Education, Korea.

I
ce cubes were placed in identical beakers, one containing tap water and the other con-
taining a sodium chloride salt solution. Both beakers and their original contents were at
20± C. As expected, the ice cubes  oated and slowly melted, however, it was found that the
ice cube in the water melted 25% faster than in the brine, contrary to expectation. A second
series of tests were undertaken in which the ice cubes were made to sink to the bottom by
placing them in a coarse string bag with weights attached. It was found that the submerged ice
cube in the brine solution melted four times faster than that submerged in tap water.
Computational Fluid Dynamics, has shown that the paradoxical experimental Ž ndings can
all be explained by considering the strength and direction of buoyancy driven currents.

Keywords: natural convection; buoyancy; phase change; melting; CFD.

INTRODUCTION tational Fluid Dynamic (CFD) simulation of the processes

involved which go some way to providing a sound explana-
Interest in crushed ice as a means of controlling and main-
tion for the counter intuitive observations.
taining environmental conditions around food materials has
existed for some time; for example, the Ž shing industry has
stored and transported some of its most valuable catch in it.
EXPERIMENTS WITH FLOATING AND
This interest is expanding as the market for chilled ready
SUBMERGED ICE CUBES
meals increases and the use of in-situ rapid cooling is
beginning to be used as part of the manufacturing process An experimental programme was devised to monitor the
of a number of industries1. The growing interest in ice as a time required for ice cubes to melt when immersed in a Ž xed
cooling medium has lead to a number of research papers volume of nominally stagnant water, initially at ambient
being published studying different aspects of the problem; temperature. The parameters investigated were the position
the freezing=melting process of water=ice in simple topolo- of the ice cubes in the water and the quantity of freezing
gies has been investigated by Hill and Kucera2 and Roy and point depressant in the water. Different concentrations of salt
Sengupta3 among others, while the heat transfer character- solutions (brine) were used, all at the same temperature.
istics of pumped ice slurries have been investigated by Snoek A data logger with 8 channels was used to record the
and Bellamy4 for a simple tube  ow and more recently for temperature at 10-second increments. The K-type thermo-
topologies as complex as plate heat exchangers by Bellas couples were used and placed at depths of 2, 17, 32, 47, 62,
et al.5. Ice slurries are ‘easier’ to deal with if the slurry has 77, 92 and 107 mm from the bottom of a glass beaker. The
a temperature range rather than a unique temperature at beaker was placed on an insulating plastic slab. The thermo-
which it suddenly changes from a liquid to a solid. The couple at 107 mm was 5 mm above the level of the water
easiest way to modify the freezing=melting temperature of an (550 ml). 18 ml of tap water were frozen to form a ‘standard’
ice=water mixture is to add a freezing point depressant such ice cube and, for the  oating experiment, placed in the
as salt. The addition of salt lowers the freezing point below beaker and left to melt. The data logger was stopped after all
0± C. Experience suggests that the addition of salt to ice will ice was melted. For the submerged measurements, the
speed up the melting process, hence, frosty roads are gritted ice was put into a string bag (with weights attached) and
with sand=salt mixtures to melt the ice and increase road grip. allowed to sink to the bottom of the beaker. The start
A series of experiments reported below observed that temperature was approximately 18 to 19± C for all experi-
melting could be delayed if salt is added to an ice=water ments. Minor variations in room temperature and conduc-
mixture, contrary to expectations. This Ž nding has an impact tion of heat down the thermocouple wires were possible, but
on the design of crushed ice storage vessels. However, believed to be of secondary importance in determining the
before the information can be used there is a need to rate of melting of the ice cubes.
understand the physical phenomena responsible for the Six experiments were undertaken; in the Ž rst two experi-
paradoxical Ž ndings. This paper presents a possible ments, tap water was placed in the beaker, in the next two,
reason, based on simple physical arguments, and a Compu- the beaker was Ž lled with a salt-water solution containing

320
 HEAT TRANSFER CHARACTERISTICS OF ICE 321

3% by weight of salt, while the last two used a 10% brine out the ice melting process. This is not the case for the brine
solution. In each experiment an ice cube (made from tap solution, where the  uid temperature at the top of the beaker
water, and hence containing virtually no salt) was either decreases monotonically as the ice melts. This is believed to
allowed to  oat or submerged in the  uid in the beaker. arise from buoyancy forces which tend to keep the most
Temperature measurements were taken from each of the dense  uids at the bottom and the least dense at the top. For
thermocouples from the time the ice cube was introduction the tap water experiment, any melt from the cube will be
to the time all the ice had melted. colder, and therefore, denser than the surrounding water,
The thermocouples provided a temperature-position distri- hence, this cold water will tend to quickly sink to the
bution evolution in the  uid with time after the introduction bottom, leaving the cube surrounded by relatively warm
of the ice cube. For the case of the  oating ice cube, it was water. This is not the case for the brine experiment where
found that the temperature of the water adjacent to the cube the density of warm brine is much higher than that of the
did not drop as fast as that of the brine. However, the coldest water. Thus, any cold water resulting from melting
temperature of the water lower down the beaker dropped ice is ‘buoyant’ in a brine environment, and cannot sink. The
faster than that of the brine. This effect is illustrated in ice cube, therefore, generates its own cold water environ-
Figure 1. Almost the reverse was found for the case of the ment which tends to stay with it.
submerged cube; the temperature of the water adjacent to the This ‘cold layer’ at the top of the beaker slows down the
submerged cube dropped faster than that of the brine. melting process, for the brine and  oating cube experiment.
The time taken to completely melt the ice cube for the Two effects slow the rate of melting, the Ž rst is due to a
different boundary conditions are illustrated in Figure 2. reduced driving temperature difference between the ice and
This Ž gure clearly shows the dramatic differences in the the surrounding (now very) cold water. The second results
melt times (almost an order of magnitude difference from the stagnation effects of the low buoyancy  ow
between the  oating and submerged ice cube in brine). velocities and accompanying low heat transfer coefŽ cients.
The rate of melting will be faster for the tap water case
where the ice is surrounded by warm water (hence a large
INTERPRETATION OF EXPERIMENTS
driving temperature difference) and a higher level of buoy-
It has been observed that, for the  oating cube in tap ancy driven mixing.
water experiments, the temperature of the water near the top The same arguments apply to the submerged ice cube
of the beaker remains at an almost constant value through- case. However, in this case the cold water generated by the

Figure 1. Sketches of the temperature-time evolution for the ice cube  oating in water and brine respectively (numeric key refers to distance in mm from
bottom).

Figure 2. Time taken to completely melt the ice cube for the different boundary conditions.

Trans IChemE, Vol 80, Part A, April 2002

322 QUARINI and CHANG

melting ice tends to stay at the bottom and so provide a cold

blanket for the ice cube in tap water. While for the sub-
merged cube in brine, the cold but buoyant water generated
by the melting cube  oats to the top, and thus, a warm
environment is maintained around the cube. The results in-
dicate that for a 10% salt solution, a  oating ice cube
survives nine times longer than for a submerged cube in
the same solution. (This is almost an order of magnitude
difference). Further, comparing the ice cube life times for
the cubes  oating in tap water and 10% salt solution
indicates that the cube in the brine lasts approximately
50% longer. On the other hand, if the cubes are made to
occupy the bottom of the beaker, then, the life time
comparison reverse, and the cube at the bottom of the tap
water Ž lled beaker lasts typically four times longer than that
in the brine Ž lled one.
These differences in melting rates are considerable, it is
therefore, important that the mechanisms controlling heat
Figure 3. Showing the iso-surface at 16± C; the cube at the top is at 0± C,
transfer rates be understood if we are to develop optimized while all the surfaces of the beaker are at 20± C.
engineering processes and equipment in which phase
changes and buoyancy are present. To do this, a three-
dimensional Computational Fluid Dynamics (CFD) study
has been undertaken and is reported below.

COMPUTATIONAL FLUID DYNAMICS

The commercial CFD code CFX4.36 was used as the
vehicle to examine the buoyancy driven  ows within the
beaker. Although the melting process itself was not
modelled, the cold sink resulting from the ice cube was
represented by a solid held at 0± C of dimensions equal to the
original ice cube. The beaker walls were maintained at
20± C. The air-water interface at the top was modelled as
a 20± C zero shear surface. The liquid in the beaker was
assumed to have constant thermodynamic and transport
properties equal to those of water at 10± C, except for the
density. In order to represent the buoyancy forces resulting
from density differences, the Bousiesqu approximation is
adopted, in which the density is linearly related to tempera-
ture by:
Figure 4. Velocity  ow vectors.
r = r0 (1 ¡ b(T ¡ T0 )) (1)
where b is the volumetric expansion coefŽ cient.
This equation describes a situation in which the density
increases with reduction in temperature, and gives a reason-
able representation of the situation in which ice  oats on
warm water. The ice is effectively a heat sink, which gives
rise to a cold negatively buoyant plume, as shown in Figure 3.
The approximations used and their validity are discussed
below.
The associated  ow proŽ les (see Figure 4) show strong
bulk mixing throughout the beaker. The resultant tempera-
ture distribution is such that the ice cube sits in relatively
warm water; buoyancy forces ensures that warm water rises
towards the cube, while cold water drops away from it. This
is shown in Figure 5.
The situation for ice  oating on brine is quite different.
Cold melting water from the ice is less dense than the warm
brine, and hence tends to  oat rather than fall. A simple
representation of this is to reverse the density-temperature
dependence. This is not a true representation of the phenom-
ena, however it does reproduce the correct direction of the
buoyancy forces. With this approximation, the predicted Figure 5. Isothermal contour plot for the ice in water situation.

Trans IChemE, Vol 80, Part A, April 2002

 HEAT TRANSFER CHARACTERISTICS OF ICE 323

 ow and temperature distributions are very different from

those of ice and water.
Figures 6 and 7 show iso temperature surfaces of 10± and
16± C.
From these Ž gures it is clear that the cold  uid stays in the
neighborhood of the ice cube. Further evidence of this is
seen in the next two Ž gures (Figures 8 and 9), which present
the temperature and  ow distribution in the brine Ž lled
beaker.
As the cold water stays in the vicinity of the ice cube, the
temperature around the cube falls, and the available driving
temperature difference between the ice and its immediate
surrounding decreases.
Even though the approximations used do not represent the
mass source effect of the melting ice, and at best just capture
the  avour of the heat sink, the resultant CFD predictions
suggest that the ice will melt at a faster rate when  oating in
water than brine. With the approximations described above
and the implicit assumption that the buoyancy driven  ows
remain laminar, the CFD calculations have been used to
predict the  owing heat transfer rates between the ice cube Figure 8. Flow distribution in brine Ž lled beaker.

Figure 6. Iso-surface of T = 10± C. Figure 9. Temperature distribution in brine Ž lled beaker.

and the beaker walls. The calculations predict that the heat
transfer rate is 12.24 watts in the case of the ice cube
 oating in water, and 2.47 watts for the ice cube in brine.
The CFD computations appear to give the correct quali-
tative trends (a  oating ice cube will melt faster in water
than in brine), and hence, gives credence to the buoyancy
dominated heat transfer explanation that has been suggested.
However, it is worthwhile examining the various approx-
imations that have been made in undertaking the CFD
computations. Firstly, in order to save on computing
resources, the CFD calculations have been undertaken as
though the process took place under steady state conditions.
This is obviously incorrect; for the authors system the
temperature of the  uid in the beaker decreases as the ice
cube melts. This has the effect of reducing the driving
temperature difference as the melting process proceeds.
Indeed, the calculations are likely to be ‘more’ correct for
Figure 7. Iso-surface of T = 16± C. the situation near the start of the experiment, when the

Trans IChemE, Vol 80, Part A, April 2002

324 QUARINI and CHANG

buoyancy driven  ows have just been established, but before CFD computations have been undertake to cast light on
the  uid in the beaker has signiŽ cantly cooled. Secondly, it this paradoxical behaviour. These indicate that buoyancy
has been assumed that the ice cube provides just a heat sink forces determine whether natural convection currents are
at 0± C, without any mass transfer resulting from phase enhanced or inhibited and whether the cold water is
change. This is not as bad an approximation as may Ž rst convected away from the ice cube or forms a cold ‘blanket’
appear. The volume change in transforming from ice to around it. These convection currents together with different
water is quite small (approximately a 10% reduction). The degrees of thermal stratiŽ cation and are responsible for
accompanying heat sink resulting from a phase change is vastly differing melting rates.
large (latent heat of fusion of 333 kJ kg¡1 as compared to It is important that designers and operators of ice making
the relatively modest speciŽ c heats of 2.1 kJ kg¡1 K¡1 for and storing equipment be aware of these buoyancy effects,
ice and 4.2 kJ kg¡1 K¡1 for water). Thus, there is a rela- especially in an era of energy efŽ ciency and climate change
tively modest momentum  ux resulting from phase change, levy.
whilst there is a considerable buoyancy driving force that
results from the ice being able to maintain its temperature at
0± C in a warm sea of  uid.
REFERENCES
As a result of these approximations, it is difŽ cult to
attempt to obtain an exact comparison with experimental 1. Metz, P. and Margen, P., 1987, The feasibility and economics of slush ice
measurement. Nonetheless, the computed heat transfer rates cooling systems, ASHRAE Trans, 932(2): 1672–1686.
2. Hill, J. M. and Kucera, A., 1983, Freezing a saturated liquid inside
show qualitative agreement with the observed experimental a sphere, Int J Heat Mass Trans, 26: 1631–1636.
trends. 3. Roy, S. K. and Sengupta, S., 1987, The melting process within spherical
Salinity and its effect on buoyancy forces is important in enclosures, Trans ASME, 109: 460–462.
determining the melt rates of ice in oceans. Holland and 4. Snoek, C. W. and Bellamy, J., 1997, Heat transfer measurements of ice
slurry in tube  ow, Experimental Heat Transfer, Fluid Mechanics and
Jenkins7 report that both the temperature and the salt Thermodynamics, pp 1993–1997.
concentration in the regions around ice shelf have to be 5. Bellas, J., Chaer, I. and Tassou, S. A., 2001, Heat transfer and pressure
accounted for in order to obtain a representative buoyancy drop of ice slurries in plate heat exchangers, UK Heat Transfer Confer-
 ux. The rather simple experiments reported here emphasis ence (Nottingham, UK), paper C2.
how important it is to correctly represent both temperature 6. AEA Technology, 1995, CFX 4 Flow solver user guide.
7. Holland, D. M. and Jenkins, A., 1999, Modelling thermodynamic ice-
and salt concentration effects on buoyancy driven  ows. ocean interactions at the base of an ice shelf, J Phys Oceano, 29: 1787–
1800.
CONCLUSIONS
Experiments have been reported which show that there
ADDRESS
can be large differences in the time required to melt ice in
what nominally appears to be similar thermal boundary Correspondence concerning this paper should be addressed to Professor
conditions. Heat transfer rates in a melting system are G. L. Quarini, Mechanical Engineering Department, University of Bristol,
Queen’s Building, University Walk, Bristol, BS8 1TR, UK.
affected by the presence of freezing point depressants in E-mail: [email protected]
the liquid and the location of the ice. Further, contrary to
intuition, it has been found that  oating ice lasts longer in The manuscript was received 19 November 2001 and accepted for
‘warm’ water than in brine at the same temperature. publication after revision 15 December 2001.

Trans IChemE, Vol 80, Part A, April 2002