Documente Academic
Documente Profesional
Documente Cultură
d
400076, India
ite
Virkeshwar12@gmail.com
ed
Atul Srivastava
Indian Institute of Technology Bombay
Department of Mechanical Engineering, Indian Institute of Technology Bombay, Mumbai,
py
400076, India
atulsr@iitb.ac.in
Co
Shyamprasad Karagadde
Indian Institute of Technology Bombay
ot
Department of Mechanical Engineering, Indian Institute of Technology Bombay, Mumbai,
400076, India
tN
s.karagadde@iitb.ac.in
ASME Membership: 000100787849
rip
Abstract:
sc
nu
Natural convection during solidification of liquids is known to impact the freezing characteristics
and also lead to defect formation. In this study, we report the findings of real-time interferometric
Ma
observation of bottom-cooled solidification of pure water in a cubical cavity. The results show first
ed
quantitative evidence of full-field thermal history during solidification, clearly depicting the
pt
anomalous expansion of water below 4°C. Furthermore, based on the strength of natural
ce
convection, characterized by the Rayleigh number, we identify and report four distinct regimes of
Ac
solidification, namely – conduction dominated, early convection, front instability and sustained
convection. A critical Rayleigh number that initiates an instability in the solidifying front has been
relating to the initiation of flow. The study shows full-field quantitative evidence of a well-known
solidifying interfaces.
d
Interferometry
ite
ed
1. Introduction
py
Solidification is an exothermic phase change phenomenon, which has numerous implications in
Co
natural and industrial processes, e.g., the formation of rocks from magma, freezing in oceans and
moist soil, crystal growth, polymer production, freeze-drying of food, and more importantly,
ot
manufacturing of alloy components [1–5]. The process of solidification of pure materials, which
tN
is primarily influenced by heat removal conditions, has been extensively studied since the early
rip
works of Stefan [6]. Furthermore, natural convection during the phase change process has shown
sc
transparent, allows real-time observations on the solidification processes using optical imaging
ed
principles. Unlike most other materials, natural convection in water, under thermally stable
pt
conditions, is strongly influenced by the anomalous expansion of water (maximum density at 4°C)
ce
between 4°C and 0°C. The hydrogen bonding is known to play a key role in the anomalous
Ac
behavior of water, changing several of its thermo-physical and optical properties [7,8].
The effect of natural convection during solidification in transparent liquids with experiment and
analytical studies have been reported by numerous researchers using the Rayleigh number [9–12].
Several researchers have reported the influence of natural convection on liquid-solid interface
shape during solidification and melting in transparent liquids (Weaver and Viskanta [13], Epstein
and Cheung [14], Cheng et al. [15], Bathelt and Viskanta [16], Ezan et al. [17]). During
solidification, the occurrence of defects such as freckles has been related to Rayleigh number [18]
in multi-component alloys. Boger and Westwater [19] investigated experimental and analytical
d
analyses of freezing and melting of water by controlling the heat transfer at the bottom and top
ite
surfaces, to understand the influence of buoyancy forces in the liquid. Using the observations of
ed
interfacial growth rates and temperatures (measured at a few predetermined locations in liquid),
py
Rayleigh number estimates corresponding to the density anomaly were provided. Gilpin [20]
Co
performed solidification of water with different supercooled surfaces to observe the influence on
dendritic ice formation and reported that high supercooling condition leads to lower convective
ot
velocities in liquid. Merker et al. [21] provided analytical investigation by Galerkin’s method to
tN
calculate the Rayleigh number, estimated with the total height of liquid rather than the unstable
rip
region. Tanaka and Miyata [22] have reported turbulent natural convection field in water by top
sc
cooled and bottom heated conditions, but without phase change, whereas Osorio et al. [23] have
nu
reported the effect of inclination to study the influence of gravity on solidification and Rayleigh
Ma
number between hot and cold vertical walls. Bgara and Viskanta [24] reported natural convection
between two thermally controlled plates and suggested that Boussinesq approximation is
ed
insufficient to describe the phenomenon and the polynomial density variation with temperature be
pt
used instead. Gau and Viskanta [25] performed top surface freezing of N-octadecane and
ce
visualized flow, temperature fluctuations in the field by thermocouple measurements and interface
Ac
evolution. The flow in the melt was measured using particle tracking with the help of aluminum
particles, which resulted in a reduction of the growth rate and curved interface due to the formation
of convection rolls. Ezan and Kalfa [26] have reported the influence of the cooling boundaries
(side, bottom, top) on the interface shape, and described the conduction and convection dominated
regimes. Their numerical simulations show the changes in positions of circulation zones in
In the aforementioned literature, it is evident that there is a paucity of data on full-field density and
d
non-intrusive temperature measurements, and furthermore, on the onset of natural convection and
ite
its subsequent stabilization. Thus, real-time observations of full field data are required for
ed
developing a detailed understanding of the process. Recently, the study of thermal and
py
compositional distribution in liquids using non-intrusive optical imaging techniques has become
Co
vital in obtaining insights to real-time dynamics of several thermo-solutal convection and phase
change processes [27–36]. Although the optical imaging of microstructure during solidification
ot
tN
has been very well established, the transient full-field information of temperature and composition
in liquid has not been fully explored. In this context, optical interferometry has shown its potential
rip
and is still being developed for a variety of applications. In one of the very early studies, Tankin
sc
from the bottom. By observing isotherms obtained from interferogram, the authors reported a
Ma
critical value of Rayleigh number that describes the initiation of flow on the basis of the first
appearance of the instability of liquid zone. A critical Rayleigh number of 480 was reported by the
ed
authors. Farhadieh and Tankin [38] studied freezing of sea water from the top in the rectangular
pt
cavity using interferometry and calculated the Rayleigh number based on temperature difference
ce
between top and bottom plate without taking into consideration the length-scale corresponding to
Ac
solidification in ammonium-chloride and water solution and depicted the compositional mapping
in the liquid phase. The above-presented literature assumes the existence of a critical Rayleigh
number associated with the initiation of flow, and there is little information on the transient
(dynamic) variation of Rayleigh number and its influence on the stability of the interface during
With this background, this work presents full-field, time-dependent temperature and temperature
d
gradient information during solidification of pure water in a bottom cooled cavity, and quantifies
ite
the time history of thermal Rayleigh number during the solidification process. Information on the
ed
transient variation of Rayleigh numbers available from interferometric observation has been
py
utilized to ascertain when the density-anomaly driven natural convection affects the ice growth.
Co
The following section describes the experimental set-up and optical configuration of the Mach-
ot
tN
density behavior and the resulting natural convection. Four distinct regimes of heat transfer are
identified, which provide a new insight into the process. The influence of natural convection on
rip
the evolution of the solid fraction is explained with the help of analytical and numerical solutions.
sc
Finally, the correlation of Rayleigh number with the initiation of flow and the interface
nu
perturbation is discussed.
Ma
The experiments have been performed with a well insulated 8 mm thick perspex glass rectangular
ed
cavity. The internal solidification cell has inner dimensions of 110 mm in height, 47 mm in width,
pt
and 40 mm in thickness. It is housed in a low conductivity perspex glass enclosure (cover cell) to
ce
minimize the condensation effects during the experiments. Under reference conditions, the optical
Ac
path lengths of the two arms of the interferometer have been matched by placing a compensation
chamber (of dimensions identical to that of the solidification cell) in the reference arm of the
interferometer. The isometric, front and side views of the solidification cell are shown in Fig. 1.
At the top of the solidification cell, provision has been made to compensate for volumetric
expansion of ice. A 1.2 mm thick copper plate is attached at the bottom for maintaining the
(Model-no 490-1402, RS Components). For a given DC input voltage, the peltier element
d
maintains a constant temperature difference. The ‘hot-side’ of the peltier plate is mounted on a
ite
copper heat exchanger. Each surface of the peltier is glued by thermal grease to maintain good
ed
thermal contact. A heating-cooling circulating bath with ethylene glycol and water mixture is used
py
to maintain the temperature of the hot-side. The peltier modules are controlled by a TEC
Co
(Thermoelectric Cooler) controller (TEC-1123-HV, Meerstetter Engineering GmbH) with the
ot
tN
Optical windows (50 mm square cross-section and 6 mm thick) are fitted at 32.5 mm above
the bottom copper plate, along two parallel faces of the internal solidification cell and cover cell,
rip
which are aligned with the direction of the laser source. Optical windows are made of high-quality
sc
transparent quartz material (flatness: /6). The other two sides and top of the cell are covered by
nu
insulating polypropylene foam material. The bottom part of the setup with the peltier and copper
Ma
heat exchanger is covered by polypropylene foam to reduce heat interaction from the surroundings.
K-type thermocouples are inserted at the top) boundary of the cell visual zones, the top surface of
ed
the internal cell, and at the surface of the bottom copper plate, with an accuracy of ±0.2°C. K-type
pt
thermocouples are connected to the desktop computer via a data-logger (TC-08, Pico Technology)
ce
and all temperatures recorded from the data-logger are stored in the desktop computer.
Ac
Solidification
Cell
Expansion Space
d
Optical Internal
ite
Window Cell
ed
Cover Cell
py
Copper plate
Peltier
Outlet
Co
y Heat Exchanger 10mm
Inlet
z
ot
(a) (b) (c)
tN
Fig. 1. Isometric (a), front (b) and side (c) views of the solidification cell showing the key
components.
rip
For in situ investigations of the solidification process one of the refractive index-based
sc
imaging techniques, namely the Mach-Zehnder interferometry, has been employed in the present
nu
work. An isometric view of the experimental setup which consists of the Mach-Zehnder
Ma
interferometer is an optical device which works by the division of amplitude, and provides
ed
variations in the solidification cell due to the changes in the optical path lengths. As shown in the
ce
figure, the interferometer consists of a laser source (He-Ne laser), a spatial filter, two beam splitters
Ac
(BS1 and BS2), two mirrors (M1 and M2), CCD camera and desktop computer. All these
To measure the temperature field in the liquid zone during solidification, a 13 mW average power
output He-Ne laser as a coherent light source (wavelength λ = 632.8 nm, spot size: 16.67 µm) is
employed. The monochrome light beam is filtered to minimize aberrations in the beam by use of
a spatial filter. A collimating lens (60 mm diameter) is employed on the filtered and diverging laser
d
beam to collimate it into a beam of constant diameter. The beam splitters reflect 50% of the laser
ite
light beam and transmit the remaining 50% light. To prevent oxidation, a silicon dioxide layer is
ed
employed on the mirrors which are coated with 99.9% pure silver. Beam splitters and mirrors are
py
oriented at 45° with respect to parallel laser light and are placed to form a rectangle.
Co
Laser Spatial
ot
Collimating
Filter Lens
tN
BS1
rip
Beam 1
M1
Beam 2
sc
M2
nu
Ma
BS2
CCD
Camera
ed
Reference
Solidification Cell Cell
pt
ce
Ac
Desktop computer
Fig. 2. Isometric view of Mach-Zehnder interferometer.
The parallel light from collimating lens goes to the first beam splitter BS1 and it splits this coherent
light into two parts- transmitted (Beam 1) and reflected (Beam 2) beams, which pass through the
HT-17-1193 Karagadde Page 8 of 34
reference and the test cells respectively. Beam 1 and Beam 2 go to mirror M1 and M2 respectively
and mirrors turn these beams at a right angle. Both beams from mirrors are directed through the
corresponding solidification and reference cells and they meet at the second beam splitter (BS2).
There is no change in the fluid condition in the reference cell. But due to a change in temperature
d
in the solidification cell compared to the reference cell, there is a change of path length (due to the
ite
relative changes in the refractive index values) between the beams passing through the test and
ed
reference cells of the interferometer.
py
Beams (Beam 1 and Beam 2) superpose with each other at the beam splitter to produce the
Co
interference pattern. The resultant interference patterns are captured by a monochrome CCD
camera (Thor labs) with a pixel resolution of 1024 × 768. The interferometer is operated in infinite
ot
tN
fringe setting mode. Due to the presence of thermal gradients in the solidification cell, interference
patterns consisting of black and white alternative layers are formed which are called fringes. In the
rip
present setting, the fringe patterns recorded in the form of interferograms represent lines of
sc
constant temperatures, i.e. isotherms. The vibration isolation optical table is pressurized by the
nu
external air compressor to avoid ground vibrations. The recorded monochrome images are stored
Ma
in the form of gray-scale images. The data is acquired at a speed of 15 fps, which is considerably
De-ionized water is carefully siphoned into the solidification and reference cells to ensure
pt
no air is entrapped in the cell at an initial temperature of 25.5°C. The interferometer is aligned for
ce
infinite fringe setting after flow and temperature within the cell are stabilized. The frames are
Ac
captured by the CCD camera and stored in the desktop computer. The bottom of the solidification
cell is maintained at -21°C and images are captured at various time instants during the course of
the experiments to record the transient evolution of the temperature field and solidification front
position.
d
The fringes obtained from the interferometric experiment contain information of the refractive
ite
index field. The refractive index of liquids is a function of wavelength, temperature, and
ed
concentration [40]. In the present set of experiments, a monochromatic laser (λ=632.8 nm) and
py
pure water are used therefore, the refractive index is predominantly a function of temperature only.
Co
To minimize the effects of superimposed noise arising from slight misalignment and optical
aberrations, image processing techniques are employed. In the acquired images, fringes contain
ot
information of the temperature field and this information is extracted by a fringe thinning
tN
technique [32]. As the temperature difference between two consecutive fringes in the image is the
rip
same, the minimum or maximum intensity points in the image correspond to the dark or bright
sc
bands respectively. Along any dark band, the minimum intensity points are manually joined by a
nu
fringe thinning algorithm, and this gives the thinned fringe image. This choice of the algorithm
Ma
was based on the studies reported by Mishra et al. [41], where it was shown that the free hand
drawing technique gives the least error when the fringes are considerably curved.
ed
The temperature difference between two consecutive fringes (ΔT) can be obtained as Eq. (1):
pt
T
ce
dn
L (1)
dT
Ac
where λ is the wavelength of the laser light source, L is the dimension of the cell along the direction
dn
of the propagation of laser beam, is the variation of refractive index of water with temperature
dT
dn
[42]. Since the fluid temperature decreases during the course of the experiment, values are
dT
evaluated using the average of temperatures obtained from the top boundary of the optical window
(visual region) and either the bottom boundary (visual region) or the ice interface. In this study,
d
the method of obtaining two-dimensional temperature distribution in the liquid field is followed
ite
from References [33,41,43,44].
ed
py
4. Results and discussion
Co
The experimental results have been documented in the form of two-dimensional interferometric
images of solidification of pure water during bottom cooling. This section presents images of the
ot
temperature field and temperature gradient as retrieved through the recorded interferograms,
tN
Rayleigh number in the convection region, and critical Rayleigh number on basis of perturbation
rip
Temperature distribution
nu
Early solidification
Ma
The evolution of solidification and natural convection in the liquid is shown via the transient
evolution of the interferometric images recorded at different time instants of the experimental run
ed
time in Fig. 3. Fig. 3(a) shows the interferometric image in the liquid zone after 1.5 hours from
pt
initialization of the boundary conditions. It is observed that ice does not appear in the observation
ce
area of the cell at 1.5 hours. All isotherms are almost straight, which indicates that the heat transfer
Ac
is predominantly through conduction in the beginning. It is to be mentioned that the visual region
is 32.5 mm above the bottom, this location of the window is chosen to capture the maximum effects
of convection, which initiates after certain initial solidification. During the experiment,
temperature is continuously decreasing at the top of the visual boundary. The presence of closely-
spaced fringes in the bottom region of the solidification cell indicates the existence of relatively
strong temperature gradients compared to the top section. As one moves upward in the y-direction,
the fringe density decreases, which in turn indicates gradual decrease in the strength of the thermal
d
gradients. Since isotherms are completely horizontal in the field of view (x=0 and x=40 mm), heat
ite
loss from the side boundaries can reasonably be expected to be negligible. Fig. 3(d) shows
ed
temperature contours for the interferometric image shown in Fig. 3(a) obtained through the
py
quantitative analysis, with a maximum of 22.83°C and a minimum of 7.68 °C.
Co
After 2.15 hours, the process of ice formation is clearly seen in the field of view (marked
using a yellow line) and is shown in Fig. 3(b). Curved isotherms are seen near the interface, which
ot
tN
indicate that there is a local variation in the density due to the anomalous expansion of water below
4°C. This density inversion initiates local convection in this region. Beyond three to four curved
rip
fringes near the solidification interface, the rest of the fringes are almost horizontal, indicating
sc
diffusion dominated behavior elsewhere. The temperature contours for this phenomenon are
nu
presented in Fig. 3(e). The image captured after 2.5 hours is shown in Fig. 3(c). The behavior is
Ma
similar to that observed in Fig. 3(b). In both the figures, Fig. 3(b, c), it is clear that the interface of
ice is almost flat, but during initial stages of solidification, negligible wall effects are to be
ed
observed. Fig. 3(f) is the two-dimensional temperature field corresponding to Fig. 3(c). The
pt
maximum temperatures in the field of view of Fig. 3(d, e, f) are respectively 22.83°C, 20.65°C,
ce
18.71°C (at the top region), while the minimum temperatures are 7.68°C, 0.0023°C, 0.0015°C.
Ac
This analysis clearly outlines how the beginning of convective flow takes place in the conduction
dominated regime. The positions of the freezing interface (0 °C isotherm) from the acquired raw
images and the temperature values evaluated through the quantitative analysis are in complete
d
ite
ed
py
(a) t = 1.5 hrs. (d)
Co
ot
tN
rip
sc
T°C
ed
pt
ce
Ac
agreement with each other. This, in turn, also provides a verification of the interferometric data
reduction scheme for the present study. Although analytical solution of the Stefan’s problem is a
possible means of further verification, the solution is not applicable to the present system due to,
(i) unit order of the height to width ratio (2.75), (ii) natural convection due to anomalous density
d
behaviour.
ite
Density anomaly driven natural convection:
ed
Fig. 4 shows the formation of convection-rolls and the corresponding two-dimensional
py
temperature contours in the liquid field after the initial solidification. The presence of the bulk
Co
convective movement of the fluid is evident in the form of circular fringe patterns in the images
shown. This observation is particularly highlighted through the interferogram shown in Fig. 4(a).
ot
At 2.75 hours, the density difference between isotherms at 0 and 4°C is increased to such a level
tN
that it leads to the formation of convective rolls and localized mixing (circular fringes shown in
rip
Fig. 4(a)) of relatively warmer water (4°C) which moves down to mix with colder water at (0°C).
sc
The temperature inside the inner most circular fringe is almost constant as no fringe is visible
nu
within. Fig. 4(d) represents temperature contours for interferometric image Fig. 4(a) obtained
Ma
through quantitative analysis. At 3.4 hours, these convective rolls increase in their strength thereby
forming larger circular fringes in the visual region as shown in Fig. 4(b). This mixing is found to
ed
extend till 4.3 hours and leads to larger convective length scale as shown in Fig. 4(c). Fig. 4(e, f)
pt
show the corresponding temperature contours for interferograms shown in Fig. 4(b, c). Beyond the
ce
convective regime (more than 4°C) straight fringes indicate conduction heat transfer in all the
Ac
figures. It can be observed from Fig. 4 that the solid-liquid interface is still flat. The maximum
temperatures at the top region of Fig. 4 (d, e, f) are respectively 17.71°C, 14.35°C, 9.8°C while the
d
ite
ed
interface at 0°C
py
(a) t = 2.75 hrs. (d)
Co
ot
tN
rip
sc
T°C
Ma
ed
pt
ce
Ac
interface is maintained at 0°C. Because of the higher density of water at 4 °C, a downward motion
of water initiates gets initiated and thus isothermal fringes show a dip at the center of Fig. 4c.
Interfacial instability
After 4.6 hours (Fig. 5(a, d)), the temperature of liquid of the top most layer reaches 6.98°C
d
ite
(measured by thermocouple) and there are no straight fringes in the field of view of the
interferogram, indicating complete convection in the visual region. At 4.95 hours (Fig. 5(b, e)),
ed
the interface of ice begins to curve. It is clearly seen that the initiation of curved interface is
py
accompanied by the upward deformation of fringes assisted by flow in the upward direction at the
Co
interface. This leads to a corresponding increase in local interface growth rate and thus a
ot
perturbation develops. This is sustained till the end of the experiment (5.8 hours, Fig. 5(c, f)). The
tN
maximum temperatures at the top region of Fig. 5 (d, e, f) are respectively 6.98°C, 5.23°C, 4.13°C
while the lowest temperature in the liquid is 0°C. These images clearly indicate the formation of
rip
dominating convective mixing flow and the resulting onset of the non-uniform solidifying
sc
interface.
nu
In Fig. 6, a plot of non-dimensional temperature (θ) of the liquid zone is shown with respect to
Ma
non-dimensional height (η). The non-dimensional height is defined as the ratio of the height of the
ed
liquid zone at a given instant of time to the maximum height of the liquid zone. The non-
pt
dimensional temperature (θ) is the ratio of the difference between the temperature at any given
ce
vertical height and the freezing point of the liquid (0°C), to the difference between initial liquid
Ac
temperature (25.5°C) and freezing point of the liquid (0°C). To obtain a temperature along the
vertical direction, the temperature values are measured at fourteen equidistant vertical sections
along the width of the visual region have been horizontally averaged. The resultant width-averaged
temperature profiles have been shown in Figure 6 at different time instants of the experiment.
d
ite
interface at 0°C
ed
(d
py
(a) t = 4.6 hrs.
)
Co
ot
tN
rip
sc
T°
Ma
C
ed
pt
ce
Ac
d
ite
ed
py
Co
ot
tN
rip
Fig. 6. Non-dimensional temperature (θ) with respect to non-dimensional height (η) of liquid
sc
It is seen from the Figure 6 that the temperature profile for t =1.5 hours has a non-uniform curve
Ma
in the visual region. At the top of the visual region temperature profile indicates that the effects of
the cooling have not fully reached these heights yet. The observed change in the slope of the profile
ed
may be explained with reference to the corresponding interferometric image (shown earlier in Fig.
pt
3a) wherein the fringe spacing relatively wider near the top of liquid zone. At 2.5 and 2.75 hours,
ce
the temperature profile at the interface begins to alter due to the initiation of convective motion
Ac
and convective rolls respectively. At 2.5 and 2.75 hours, the temperature profile beyond the
convective regime shows a curvy profile because the cooling effect has not uniformly propagated
throughout the visual region (Fig. 3(c) and 4(a)), where the top of the visual region has a higher
fringe spacing compared to the middle section. At 3.4 hours, temperature propagation is uniform
in the visual region because the fringes are uniformly spaced beyond the convection region. A
typical ‘S’ shaped temperature curve is clearly observed at 4.6 hours due to mixing in the
convective regime. The plot at 4.95 hours is almost linear as convective mixing is established (Fig.
5 (b)) and it leads to the homogenization of the temperature gradient. During the formation of the
d
perturbed interface, there is stabilized convection in the liquid.
ite
Temperature Gradient
ed
The temperature gradient has been calculated to develop further insights into the solidification
py
process and interfacial heat balance. As the effective temperature gradient in the x direction is
Co
negligible, the gradient along the height (y direction) has been evaluated using a first order estimate
ot
at every location using the distance between the isotherms in the interferogram. Fig. 7(a) shows
tN
the temperature gradient contours at 1.5 hours, in which there is a lower value of temperature
gradient near the top liquid layer. From Fig. 3(a), it is seen that the isotherms near the bottom are
rip
closely spaced compared to those at the top, an observation which is consistent with the relative
sc
spacing of the fringes in the interferogram recorded during this period of the experimental run time
nu
(shown earlier in Figure 3(a)). Initially, the liquid zone can be split up into three regions in Fig.
Ma
7(b); a) upper b) middle and c) lower regions. The relatively smaller values of gradients are seen
in the upper (conduction dominated but the effects of cooling have not fully reached these heights
ed
yet) and lower (due to convection initiation above the interface, high inter-fringe spacing in Fig.
pt
3b) regions. However, the middle region shows significantly stronger gradients.
ce
In Fig. 7(c), three similar regions of low-high-low temperature gradients are formed as in Fig. 7(b).
Ac
But at the lower region of Fig. 7(c), the values of gradients are relatively lesser compared to that
in Fig. 7(b) due to the formation of convection rolls, as has earlier been shown in interferogram of
d
ite
ed
Moving solidification front
py
(a) t = 1.5 hrs. (b) t = 2.5 hrs.
Co
ot
tN
rip
sc
°C/m
ed
pt
ce
Ac
Fig. 4(a). Formation of convection rolls leads to better fluid mixing which in turn homogenizes
Fig. 7(d) shows upper and lower regions on either side of the mid-height, with nearly uniform
gradients in the upper region (conduction), and convective region that is characterized by non-
d
uniform temperature distribution in the lower half. At 4.95 hours, when the perturbation in the
ite
interface shape is observed, the gradient ahead of the interface in the liquid region is considerably
ed
lower than those on either side (with respect to a vertical line passing through the center of the
py
interface) of the curved interface (Fig. 7f). This can be also verified with the help of the energy
Co
balance at the interface that is given by the Stefan’s condition.
Ts T (2)
ot
ks kl l s HV
y y
tN
In Eq. (2), ks, kl are the thermal conductivities of solid and liquid respectively, Ts and Tl are the
rip
temperatures in solid and liquid phases at the interface respectively, ρs is the density of the solid,
sc
H is the latent heat of fusion, and V is the velocity of the solidifying interface. At any instant, the
nu
first term of Eq. (2) (heat flux in the solid) is constant and kl, ρs, H are taken as constant. Thus,
with an increase in the liquid gradient value, the velocity (V) will decrease and vice versa.
Ma
Fig. 8 shows the variation of the width-averaged temperature gradients (°C/m) as a function of
ed
non-dimensional height (η) at different time instants of the experimental run time. During the
pt
initial stages (t = 2.5 and 2.75 hours), the heat transfer phenomenon in most of the cell (beyond
ce
the convective region) is dominated by conduction and the maximum gradients are observed in the
Ac
lower regions of the cell (shown inside the dashed circle in Fig. 8). During the final stages of the
experiment, the bulk movement of the fluid (convective mixing) leads to the homogenization of
the thermal gradients, thereby reducing the strength of the temperature gradients throughout the
bulk fluid region (shown inside the dashed box in Fig. 8). On the other hand, the gradients increase
near the ice interface (bulk behaviour excluding the location of the curved portion) leading to a
d
ite
ed
py
Co
ot
tN
rip
sc
nu
Fig. 8. Temperature gradient (°C/m) in liquid with respect to non-dimensional height (η) of liquid
Ma
Since the heat loss at the side boundaries is negligible, further insights can be obtained by
comparing the experimental observations with a one-dimensional heat balance formulation at the
pt
solidifying interface. The instantaneous calculation of the energy balance (Eq.(2)) at the interface
ce
has been performed under the assumption that the profile of the temperature gradient in the solid
Ac
is linear and thermal conductivity of ice is 2.27 W/m-K at a mean temperature (-10.5°C) of bottom
(-21°C) and interface of ice (0°C) [45]. Thermal gradient of liquid is calculated using an average
of data of the five pixels near the interface. The left (net conductive heat flux) and right (latent
heat release) hand side terms of interfacial energy balance (Eq. (2)) are found to be within 15% of
each other. This analysis, performed on the basis of energy balance, further establishes the validity
and correctness of the temperature and gradient field distribution as retrieved through the
d
Evolution of the solid fraction
ite
From the analytical solution (in the absence of gravity) of one-dimensional heat conduction
ed
equation (Stefan’s problem), it is well-known that the fraction of solid varies with the square root
py
of time [6,46]. In Fig. 9, the transient evolution of the solid fraction, obtained from the analytical
Co
solution is in close agreement with experimental results until 2.6 hours. Note that the figure shows
ot
data points from 2.15 hrs, as the field of view is positioned at a distance of 32.5 mm from the base
tN
of the solidification cell. We call this period of conduction dominated heat transfer as Regime I.
rip
sc
nu
Ma
ed
pt
ce
Ac
I Regime II III IV
Subsequently, the experimental trend of solid fraction deviates from the analytical solution and
continues to remain so till the end of solidification. Thus, it can be concluded that convection
phenomenon arising from the density anomaly plays a vital role in reducing the solidification rate.
d
Critical Rayleigh number for front instability
ite
The Rayleigh number (Ra) is a dimensionless parameter used to quantify the strength of natural
ed
convection in a fluid. In this experiment, Ra is formulated with the change in density that causes
py
natural convection and is given by the following relation.
Co
gh3 (3)
Ra
ot
In Eq. (3), Δρ is the density difference of the unstable region in the liquid, g is gravitational
tN
acceleration, h is the characteristic length scale of the unstable region in the liquid (where the
rip
characteristic length scale is defined as the distance between the layers of maximum density (4°C)
to the ice interface density (0°C)), α is thermal diffusivity and µ is dynamic viscosity. Following
sc
the Boussinesq approximation, the equation for the density of water as a function of temperature
nu
Fig. 10 shows a transient development of the Rayleigh number with the elapsed time of the
pt
experiment. The experiment has been repeated five times (interferograms and corresponding
ce
Rayleigh number plots for five different experiment have been presented in the Appendix) and the
Ac
error bars in Fig. 10 are obtained with 95% confidence level. We identify four different regimes
of heat transfer during the solidification phenomena. Regime I, the previously mentioned
conduction dominated reime, where convection is very weak (due to a low h) and exists only near
the interface (Fig. 3(b, c)). The Ra in this regime is below 8000 and does not undergo any
significant change till 2.6 hours. Subsequently in Regime II, when the convection initiates due to
the anomalous density behavior, Ra gradually increases and at 4.7 hours (i.e. the end of Regime
II) reaches a value of close to 110000±11600. This marks the beginning of Regime III, till about
d
5.15 hours and Ra shows a little increase (from 110000±11600 to 118000±9000), and the
ite
solidifying front becomes curved (Fig. 5b).
ed
py
I Regime II III IV
Co
ot
tN
rip
sc
nu
Ma
ed
pt
Fig. 10. Time evolution of Rayleigh number during solidification of water: The four regimes of
ce
solidification are marked and a representative interferometric image for each of the four regimes
is shown in the inset.
Ac
Until the beginning of Regime III, the solidifying interface is flat and the direction of the flow at
the central region is downward (Figs. 4c and 5a). However, as the length scale h (of Eq. 3) becomes
large enough, buoyant convection is initiated ahead of the interface at the central location. During
this process, the height h does not change momentarily, but the change in energy (due to cooling)
is invested in forming a curved interface and the direction of flow flips. The localized thermal
gradient distribution (Fig. 7f), resulting from convective mixing, leads to the perturbation at the
ice interface which is presented in Fig. 5(b, c). Thus, we observe a negligible increase in Ra during
d
this regime, and this critical value leads to initiation of the perturbation. This is termed as the
ite
critical Rayleigh number above which the solidifying interface becomes curved. From this study,
ed
the critical Ra during solidification of pure water is thus approximately equal to 110000±11600 (
py
O 105 ). Finally, sustained convection occurs in Regime IV which continues until the end of the
Co
experiment and the perturbation is unaltered as the upward flow ahead of the curved interface
ot
drags the colder fluid from the surroundings. As the experiment progresses, the temperature of the
tN
top fluid layer reaches 4°C and thus Ra continues to increase. The supplementary video shows the
rip
The primary purpose of this study was two-fold: (i) to perform and establish non-intrusive, full-
sc
field temperature measurement of transparent liquids during solidification, and (ii) to verify the
nu
technique on a well-known pure system such as water. Although water was used as an initial test
Ma
system, we observed that the results of the full-field temperature data and the associated natural
convection within the temperature range of 0-4 °C (solely due to the anomalous expansion of
ed
water) has not been reported in the literature. The non-uniform interface formation can be
pt
attributed to the buoyant convection, and interestingly leads to an almost invariant Rayleigh
ce
number for a certain duration. The present study has led to the development of a critical Rayleigh
Ac
number that changes the morphology of the solidifying interface. In addition, the present study has
also provided an experimental evidence of the fact that the convective phenomena arising from the
density anomaly plays a significant role in reducing the rate of solidification. This optical
5. Concluding remarks:
d
In this study, an in situ optical imaging is used to investigate the transient development of the
ite
solidifying ice-water interface in a bottom cooled cavity. Mach-Zehnder interferometry and fringe-
ed
thinning methods are used to extract the full-field temperature data. The following conclusions
py
may be drawn on the basis of the present experimental study:
Co
Higher temperature gradients are observed in the middle of the solidification cell, which
ot
tN
The Rayleigh number is observed to gradually increase throughout the experiment, based
on which the solidification process can be categorized into four distinct regimes, viz., (i)
rip
conduction dominated, (ii) early convection (iii) front instability, and (iv) sustained
sc
convection.
nu
In the conduction dominated regime, the growth of solid obtained through experimental
Ma
initialization) marks the beginning of reduced growth rate and deviation from analytical
ed
solution.
pt
After initial convection, a sudden increase in Rayleigh number is observed, which leads to
ce
a perturbation in the interface shape. This is a critical Rayleigh number beyond which
Ac
curved interfaces are formed due to the anomalous behaviour of water. In our experiments,
The data and methodology can be further used to validate numerical simulations which
Acknowledgments
d
Authors gratefully acknowledge the financial support from the Department of Science and
ite
Technology (DST), India (Grant: EMR/2015/001140). Partial support from another DST project
ed
(Grant: SR/S3/MERC/0030/2012) for building the Mach-Zehnder interferometer and development
py
of data reduction schemes is also thankfully acknowledged. V. Kumar thanks his colleagues S. S.
Co
Rao and N. Dama for their assistance with the experimental set-up and writing.
ot
Nomenclature tN Greek symbols
T Temperature (oC)
x, y, z Coordinate axes
Fig. 1 Isometric (a), front (b) and side (c) views of the solidification cell showing the key
components.
Fig. 2 Isometric view of Mach-Zehnder interferometer.
Fig. 3 Conduction dominated interferometric images during solidification of water from
d
bottom cooling at (a) 1.5 hrs. (b) 2.15 hrs. and (c) 2.5 hrs.; (d), (e) and (f) are the
ite
corresponding temperature fields (°C).
ed
Fig. 4 Convection rolls formed in interferometric images during solidification of water
from bottom cooling at (a) 2.75 hrs. (b) 3.4 hrs. and (c) 4.3 hrs.; (d), (e), and (f) are
py
the corresponding temperature fields (°C).
Co
Fig. 5 Non-uniform interface forming interferometric images during solidification of water
from bottom cooling at (a) 4.6 hrs. (b) 4.95 hrs. and (c) 5.8 hrs.; (d), (e), and (f) are
ot
the corresponding temperature fields (°C). tN
Fig. 6 Non-dimensional temperature (θ) with respect to non-dimensional height (η) of
liquid during solidification of water.
rip
Fig. 7 Temperature gradients (°C/m) contour in liquid during solidification of water from
bottom at different time instants; (a) 1.5 hrs., (b) 2.5 hrs., (c) 2.75 hrs., (d) 3.4 hrs.,
sc
Fig. 8 Temperature gradient (°C/m) in liquid with respect to non-dimensional height (η) of
liquid during solidification of water.
Ma
References
[1] Ehrhard, P., Riley, D. S., and Steen, P. H., 2001, Interactive Dynamics of Convection and
Solidification.
[2] Ghenai, C., Mudunuri, A., Lin, C. X., and Ebadian, M. A., 2003, “Double-Diffusive
Convection during Solidification of a Metal Analog System (NH4Cl–H2O) in a
Differentially Heated Cavity,” Exp. Therm. Fluid Sci., 28(1), pp. 23–35.
[3] Wang, S. Y., Lin, C. X., and Ebadian, M. A., 1999, “Vortex Flow of Low Concentration
d
NH4Cl–H2O Solution during the Solidification Process,” Int. J. Heat Mass Transf., 42, pp.
ite
4153–4163.
[4] Srivastava, A., Muralidhar, K., and Panigrahi, P. K., 2004, “Comparison of Interferometry,
ed
Schlieren and Shadowgraph for Visualizing Convection around a KDP Crystal,” J. Cryst.
Growth, 267(1–2), pp. 348–361.
py
[5] Kehtarnavaz, H., and Bayazitoglu, Y., 2017, “Solidification of Binary Mixture in a Finite
Planar Medium: Saline Water,” J. Heat Transfer, 107(4), pp. 964–966.
Co
[6] Hahn, D. W., and Ozisik, M. N., 2012, “Phase-Change Problems,” Heat Conduction, John
Wiley & Sons, pp. 452–495.
[7] Tanaka, H., 1998, “Simple Physical Explanation of the Unusual Thermodynamic Behavior
of Liquid Water,” Phys. Rev. Lett., 80, pp. 5750–5753.
ot
tN
[8] Lin, S., Gao, D. Y., and Yu, X. C., 1990, “Thermal Stresses Induced by Water Solidification
in a Cylindrical Tube,” J. Heat Transfer, 112(4), pp. 1079–1082.
rip
[9] Wettlaufer, J. S., Worster, M. G., and Huppert, H. E., 1997, “Natural Convection during
Solidification of an Alloy from above with Application to the Evolution of Sea Ice,” J. Fluid
sc
via Rayleigh Number Method for Single-Crystal Nickel-Base Superalloy Castings,” Metall.
Mater. Trans. A, 31(10), pp. 2545–2557.
Ma
[12] Chakraborty, P. R., and Dutta, P., 2013, “Study of Freckles Formation During Directional
Solidification Under the Influence of Single-Phase and Multiphase Convection,” J. Therm.
pt
Rectangular Cavity,” Int. Commun. Heat Mass Transf., 13, pp. 245–252.
[14] Epstein, M., and Cheung, F. B., 1983, “Complex Freezing-Melting Interfaces in Fluid
Ac
726.
[17] Ezan, M. A., Erek, A., and Dincer, I., 2011, “A Study on the Importance of Natural
Convection During Solidification in Rectangular Geometry,” J. Heat Transfer, 133(10), p.
102301(1-9).
[18] Karagadde, S., Yuan, L., Shevchenko, N., Eckert, S., and Lee, P. D., 2014, “3-D
Microstructural Model of Freckle Formation Validated Using in Situ Experiments,” Acta
Mater., 79, pp. 168–180.
d
[19] D.V.Boger, and J.W. Westwater, 1967, “Effect of Buoyncy on the Melting and Freezing
ite
Proccess,” J. Heat Transfer, 89(1), pp. 81–89.
[20] Gilpin, R. R., 1976, “The Influence of Natural Convection on Dendritic Ice Growth,” J.
ed
Cryst. Growth, 36(1), pp. 101–108.
[21] Merker, G. P., Waas, P., and Grigull, U., 1979, “Onset of Convection in a Horizontal Water
py
Layer with Maximum Density Effects,” Int. J. Heat Mass Transf., 22(4), pp. 505–515.
[22] Tanaka, H., and Miyata, H., 1980, “Turbulent Natural Convection in a Horizontal Water
Co
Layer Heated from below,” Int. J. Heat Mass Transf., 23(9), pp. 1273–1281.
[23] Osorio, A., Avila, R., and Cervantes, J., 2004, “On the Natural Convection of Water near
ot
Its Density Inversion in an Inclined Square Cavity,” Int. J. Heat Mass Transf., 47(19–20),
pp. 4491–4495.
tN
[24] Braga, S. L., and Viskanta, R., 1992, “Transient Natural Convection of Water near Its
Density Extremum in a Rectangular Cavity,” Int. J. Heat Mass Transf., 35(4), pp. 861–875.
rip
[25] Gau, C., and Viskanta, R., 1983, “Flow Visualization during Solid- Liquid Phase Change
Heat Transfer I . Freezing in a Rectangular Cavity,” Int. Commun. Heat Mass Transf., 10(3),
sc
pp. 173–181.
[26] Ezan, M. A., and Kalfa, M., 2016, “Numerical Investigation of Transient Natural
nu
Convection Heat Transfer of Freezing Water in a Square Cavity,” Int. J. Heat Fluid Flow,
61(Part B), pp. 438–448.
Ma
[30] Yabuki, T., Hamaguchi, T., and Nakabeppu, O., 2012, “Interferometric Measurement of the
Liquid-Phase Temperature Field around an Isolated Boiling Bubble,” J. Therm. Sci.
Ac
for Simultaneous Measurements of Temperature and Solutal Fields,” Int. J. Heat Mass
Transf., 98, pp. 662–674.
[33] Varma, S. S., Rao, S. S., and Srivastava, A., 2017, “Simultaneous Measurement of Thermal
and Solutal Diffusivities of Salt-Water Solutions from a Single-Shot Dual Wavelength
Interferometric Image,” Exp. Therm. Fluid Sci., 81, pp. 123–135.
[34] Tsushima, N., Narumi, A., Nakane, I., Kashiwagi, T., and Akisawa, A., 2005,
“Visualization of Transient Solidification Process of Aqueous Solution by Dual Wavelength
d
Holographic Interferometry,” J. Heat Transfer, 127(8), p. 801.
ite
[35] Spatz, T. L., and Poulikakos, D., 1992, “A Two-Wavelength Holographic Interferometry
Study on the Solidification of a Binary Alloy Around a Horizontal Pipe,” J. Heat Transfer,
ed
114(4), p. 998.
[36] El-Wakil, M. M., Myers, G. E., and Schilling, R. J., 1966, “An Interferometric Study of
py
Mass Transfer From a Vertical Plate at Low Reynolds Numbers,” J. Heat Transfer, 88(4),
pp. 399–406.
Co
[37] Tankin, R. S., and Farhadieh, R., 1971, “Effects of Thermal Convection Currents on
Formation of Ice,” Int. J. Heat Mass Transf., 14(7), pp. 953–961.
ot
[38] Farhadieh, R., and Tankin, R. S., 1972, “Interferometric Study of Freezing of Sea Water,”
J. Geophys. Res., 77(9), pp. 1647–1656.
tN
[39] Vikas, D., Basu, S., and Dutta, P., 2012, “In-Situ Measurements of Concentration and
Temperature during Transient Solidification of Aqueous Solution of Ammonium Chloride
rip
Using Laser Interferometry,” Int. J. Heat Mass Transf., 55(7–8), pp. 2022–2034.
[40] Goldstein, R. J., 1996, Fluid Mechanics Measurements, Taylor and Francis.
sc
[41] Mishra, D., Muralidhar, K., and Munshi, P., 1998, “Performance Evaluation of Fringe
Thinning Algorithms for Interferometric Tomography,” Opt. Lasers Eng., 30(3–4), pp. 229–
nu
249.
[42] Abbate, G., Bernini, U., Ragozzino, E., and Somma, F., 1978, “The Temperature
Ma
Dependence of the Refractive Index of Water,” J. Phys. D. Appl. Phys., 11, pp. 1167–1172.
[43] Mishra, D., Muralidhar, K., and Munshi, P., 2013, “Measurements of Three Dimensional
Temperature Field in Fluids Using Laser Interferometry,” Def. Sci. J., 49(3), pp. 243–255.
ed
[44] Rao, S. S., and Srivastava, A., 2016, “Interferometric Study of Natural Convection in a
Differentially-Heated Cavity with Al2O3-Water Based Dilute Nanofluids,” Int. J. Heat
pt
[46] Dantzig, J. A., and Rappaz, M., 1994, “Analytical Solutions for Solidification,”
Ac
Appendix:
Five experiments were performed to check the reproducibility of the convective flow pattern and
non-uniform interface. In all of these experiments, the expeimental setup and the optical
d
intrumentation setup (interferometer) remained undisturbed and experiments were performed with
ite
identical conditions as described in the manuscript. The temperature of the fluid and Rayleigh
ed
numbers were calculated as expalined in sections 3 and Eq. (3) respectively. In the first trial the
py
non-uniform interface is developed between 4.6 hours and 4.85 hours. Similarly in the subsequen
Co
trials, the non-uniform interface was found to develop between 4.61 hours and 5.00 hours, 4.40
hours and 4.83 hours, 4.36 hours and 4.90 hours, 4.68 hours and 5.2 hours respectively. The
ot
corresponding critical Rayleigh number for each experimental trails is 95900, 124000, 117140,
tN
102200 and 111900. The non-uniformity in the interface shape was found to sustain till the end,
rip
d
ite
ed
8mm 8mm
Trial 1 Ra critical =95900 Trial 2 Ra critical =124000
py
Co
ot
tN
rip
8mm 8mm
sc
8mm
Ac
Fig. A1. Repeated experiment of non-uniform interface (marked in yellow circles) and corresponding
Rayleigh Number with each experiment.