RESEARCH ARTICLE

In situ LIF temperature measurements in aqueous ammoniumchloride solution during uni-directional solidification

M. Behshad Shafii • Chee L. Lum •

Manoochehr M. Koochesfahani

Received: 12 June 2009 / Revised: 16 September 2009 / Accepted: 24 September 2009 / Published online: 14 October 2009

� Springer-Verlag 2009

Abstract We present in situ whole-field measurements of

the temperature field using laser-induced fluorescence in a

study of bottom-chilled uni-directional solidification of

aqueous ammonium chloride. We utilize a two-color, two-

dye, ratiometric approach to address the significant spatial

and temporal variations of laser sheet intensity field due to

refractive index variations caused by the evolving con-

centration and temperature fields. In our work we take

advantage of two temperature sensitive fluorescent dyes

with opposite temperature sensitivities in order to increase

the overall sensitivity and temperature resolution of the

measurements. The resulting temperature sensitivity (about

4% K-1) is more than a factor of two higher than the

original work of Sakakibara and Adrian (Exp Fluids 26:7–

15, 1999) with a sensitivity 1.7% K-1. In situ measure-

ments of the temperature field during solidification are

presented, along with temperature characteristics of some

of the complex flow features, such as plumes and fingers.

1 Introduction

A wide variety of products are now made of metal alloys

such as automobile engine blocks and the compressor and

turbine section of an aircraft engine. The major issues for

these alloys are melt-related defects and inhomogeneities

such as segregation defects and freckles formed during the

solidification process. The solidification process of a binary

melt is influenced by fluid motion in the liquid, whose

flow pattern strongly depends on both thermal and solutal

buoyancy forces. Studies of these multiphase transport

phenomena include stability and numerical analyses of the

coupled nonlinear governing equations based on simplifying

assumptions, exemplified by works such as Heinrich et al.

(1989), Worster (1991), Neilson and Incropera (1991),

Worster (1992), Schneider and Beckermann (1995), Schulze

and Worster (1998) and Feltham and Worster (1999).

An experimental approach to studying the solidification

of metallic alloys has been the use of transparent analogs of

metallic alloys. A popular analog is the binary aqueous

ammonium chloride (NH4Cl-H2O) system which has pro-

ven to be useful for flow visualization and morphological

studies and has been utilized by many researchers because

of its convenient solidification temperature and similar

behavior to metallic alloys (e.g., Chen and Chen 1991;

Magirl and Incropera 1993; Beckerman and Wang 1996;

Liu and Hellawell 1999). Previous studies indicate that

convective phenomena associated with solidification of

aqueous ammonium chloride solution cover a large range

of length scales that evolve dynamically. These features

include the vertical growth of a solid/liquid interface along

with the development of numerous fine structures of salt

fingers, followed by the appearance of a number of plumes

ejecting water-rich fluid jets from channels in the mushy

zone of the growing dendritic crystal mass.

M. B. Shafii � C. L. Lum � M. M. Koochesfahani (&)

Department of Mechanical Engineering,

Michigan State University, East Lansing,

MI 48824, USA

e-mail: koochesf@egr.msu.edu

Present Address:M. B. Shafii

Mechanical Engineering Department,

Sharif University of Technology,

Azadi St., 14588-89694 Tehran, Iran

Present Address:C. L. Lum

BD Medical, Franklin Lakes,

NJ 07417, USA

123

Exp Fluids (2010) 48:651–662

DOI 10.1007/s00348-009-0758-7

The general features described above are illustrated in

Fig. 1 for the specific geometry that is the subject of the

current study. This figure depicts a shadowgraph visuali-

zation of the solidification process in a hypereutectic (26%

by weight; 26% w) ammonium chloride (NH4Cl) solution

within a rectangular container at room temperature whose

base plate is cooled to -14�C to impose a vertical tem-

perature gradient and initiate uni-directional solidification

(Lum et al. 2001; Lum 2001). The details of the experi-

mental setup will be provided later in this paper. The

cooling of the base plate starts at time zero and small-scale

features, termed ‘‘fingers’’, begin to form at about 4 min.

As the cooling of fluid continues, more ascending water-

rich fingers develop due to a need to reject more solute

during the crystallization process. This is accompanied by a

stronger convection and the fingers are discernible all the

way to the top of the test section (see image at 10 min).

The large Lewis number of aqueous ammonium chloride

helps in maintaining the identity of the water-rich fingers

(Lewis number is defined by Le = a/D, where a and D are

thermal and mass diffusivity, respectively). During the

initial stages of cooling, rapid formation of ammonium

chloride crystals occurs which settle to form the mushy

layer. As time progresses, well-defined chimney-like

structures begin to form which provide outlets from the

mush for rejecting lighter fluid into the surroundings

through ascending water-rich plumes. The small fingers in

the earlier stages die down as more plumes develop.

Since the thermal buoyancy force is one of the main

forces that plays an essential role in uni-directional solid-

ification, it is important to characterize the in situ tem-

perature distribution during the solidification process.

Unfortunately, there exist very little non-intrusive quanti-

tative experimental data on the in situ temperature field

during the solidification process of ammonium chloride.

Previous measurements include thermocouple measure-

ments of temperature histories in the test container (Magirl

and Incropera 1993) and at the container side wall (Wirtz

et al. 1998). Non-intrusive measurements with thermo-

chromic liquid crystal paints (TLC) have also been repor-

ted on the surface of a Hele-Shaw test cell (Solomon and

Hartley 1998). Any inference on the temperature in the

interior was indirect and based on the assumption of rapid

diffusion of heat across the narrow gap of the test cell. The

use of encapsulated TLC particles (10–15 lm in diameter)

has been reported by Nishimura et al. (1992) and Ni-

shimura and Wakamatsu (2000) for generating qualitative

temperature maps based on the color change of TLC; the

quantitative temperature distributions were measured by

thermocouples, however. We also note that TLC tends to

give a high temperature resolution, but over a rather narrow

temperature range, e.g., 2.5�C range in Nishimura et al.

(1992).

Motivated by the discussion above, we present here in

situ whole-field measurements of the temperature field

using laser-induced fluorescence (LIF) in a study of bot-

tom-chilled uni-directional solidification process of aque-

ous ammonium chloride. This work was first reported

briefly by Shafii and Koochesfahani (2003) and extensive

details can be found in Shafii (2005). In our work we have

employed a two-color, two-dye, ratiometric approach to

address the significant spatial and temporal variations of

laser sheet intensity field due to refractive index variations

10 minutes

50 minutes

90 minutes 70 minutes

30 minutes

4 minutes

10 mm

nucleator base plate

Fig. 1 Shadowgraph visualization of the unidirectional cooling of

ammonium chloride at different times during the solidification

process (Lum 2001). The entire test section is shown. Cooling is

from the bottom and the growth of the solidification front is

highlighted by the increasing thickness of the (black) mushy zone

652 Exp Fluids (2010) 48:651–662

123

caused by the evolving concentration and temperature

fields. The traditional approach uses a temperature sensi-

tive dye along with another with little temperature sensi-

tivity (e.g., see Sakakibara and Adrian 1999). In our work,

however, we take advantage of two temperature sensitive

fluorescent dyes, with opposite temperature sensitivities

(Fluorescein and Kiton Red, at 514.5 nm excitation), in

order to increase the overall sensitivity and temperature

resolution of our measurements. An approach of this type

has also been recently reported by Sutton et al. (2008)

using other dye pairs for the 532-nm excitation of Nd-YAG

laser. Another distinction in our study is that we focus on a

‘‘cold’’ application of LIF with temperatures below ambi-

ent, whereas most previous studies have addressed heated

fluid measurements with temperatures above ambient.

In the following sections the LIF technique for mea-

suring temperature is briefly described, along with a dis-

cussion of our dye selection, calibration, and further details

of implementation. The experimental test facility and the

procedure for creating the uni-directional solidification of

ammonium chloride are given next. Finally, sample mea-

surement results are provided in terms of in situ quantita-

tive temperature maps at different times during the

solidification process, along with characterization of tem-

perature profiles at selected locations in the test section.

2 Temperature measurement using laser induced

fluorescence (LIF)

The use of LIF in fluid flows takes advantage of the fact that,

for dilute solutions and unsaturated laser excitation energies,

the fluorescence emission intensity I at a given point is given

by I ¼ AIiCeU; where A is a parameter representing the

detection collection efficiency, Ii the local incident laser

intensity, C the concentration of the dye, e the absorption

coefficient, and U the photoluminescence quantum effi-

ciency. For some molecules, the absorption coefficient e

and/or the quantum efficiency U are temperature dependant,

allowing the measurement of the emission intensity of tracer

molecules to be used to quantify the temperature field in

a fluid flow. The temperature sensitive fluorescent dye,

Rhodamine B, has been one of the most common tracers

used in various applications (see e.g., Sakakibara et al. 1993;

Sato et al. 1997; Coppeta and Rogers 1998; Sakakibara and

Adrian 1999, 2004; Lemoine et al. 1999; Coolen et al. 1999;

Kim and Kihm 2001; Lavieille et al. 2001; Saeki and Hart

2001; Kim et al. 2003). Comparative studies of commonly

used fluorescent dyes in liquids can be found in Coppeta and

Rogers (1998) and Saeki and Hart (2001).

In its simplest form of implementation, the emission

from a single fluorescent dye, over a single spectral band,

can be used to measure the temperature field (e.g., Lemoine

et al. 1999). The difficulty in this case is that any spatial

and temporal variations in the local incident laser intensity

need to be separately accounted for, something that is

impossible to do if the variations are caused by the flow

itself. A good example is the solidification process which is

the focus of our investigation. This flow is characterized by

variable temperature and concentration fields that cause a

variable index of refraction field within the flow, causing

the incident laser intensity field within the flow to vary in

time and space. Figure 2 illustrates the recorded fluores-

cence intensity, before and during solidification, from a

region in the test section that is illuminated by a laser sheet.

The smoothly varying intensity field before the start of

solidification, which is characteristic of the spatial intensity

distribution of the laser sheet, develops significant

‘‘streaks’’ after the start of the solidification process. The

strengths and positions of these streaks change with time.

A common method to eliminate the effect of in situ

variations of incident laser intensity is to use a ratiometric

approach. In the two-color, single-dye, approach, emission

from two different spectral bands with different temperature

sensitivities are utilized (see Lavieille et al. 2001). A three-

color approach, using emission from three different spectral

Fig. 2 Fluorescence intensity over an imaged area, a before solidification and b during solidification

Exp Fluids (2010) 48:651–662 653

123

bands, has also been reported for measurements with large

optical paths (Lavieille et al. 2004). The most popular

method, however, has been the two-color, two-dye, approach

where two dyes with different temperature sensitivities are

employed (e.g., Coppeta and Rogers 1998; Sakakibara and

Adrian 1999, 2004; Kim and Kihm 2001; Kim et al. 2003).

The details of this approach are given in the cited references

and will not be repeated here. Briefly, however, for fixed

concentrations of a mixture of two fluorescent dyes the

fluorescence intensity ratio R of the two dyes is cast in the

form R � I1=I2 ¼ K � FðTÞ: In this expression, I1 and I2 are

the fluorescence intensities emitted by the two dyes, K is a

constant that contains information about the specific exper-

imental setup (i.e., dye concentrations and optical detection

parameters), and F(T) represents the overall temperature

response function that is connected to the temperature

dependence of the absorption coefficients and quantum

efficiencies of the two dyes. An experimental calibration is

used to establish the fluorescence intensity ratio referenced

to its value at a reference temperature To, i.e., R(T)/R(To), as

the primary calibration curve that is used to measure the fluid

temperature. The following section provides details on the

selection of the fluorescent dyes used in this study and the

calibration of the temperature response curve.

2.1 Dye selection and calibration

The most common fluorescent dyes used in the two-color,

two-dye, ratiometric method are the temperature sensitive

Rhodamine B and the nearly temperature insensitive

Rhodamine 110. The overall temperature sensitivity of this

dye pair is 1.6–1.7% K-1 (Sakakibara and Adrian 1999). In

our investigation, the vertical temperature variation in the

liquid melt is about 20�C, but the transverse temperature

variation across fingers and plumes is expected to be much

smaller and on the order 2�C. It is, therefore, desirable to

improve the temperature sensitivity of the fluorescent dye

pair. To accomplish this, we take advantage of two tem-

perature sensitive dyes, one with positive and the other

with negative temperature sensitivities. Most fluorescent

molecules, however, have a negative temperature sensi-

tivity, i.e., fluorescence intensity decreases with increasing

temperature. Fluorescein (disodium fluorescein), on the

other hand, has a positive temperature sensitivity when

excited at 514.5 nm (green line of argon-ion laser) due to

the increase with temperature of its absorption at this

wavelength (Coppeta and Rogers 1998). The second dye

was selected to be Kiton Red because its absorption and

emission spectra are red-shifted in comparison with Rho-

damine B and this reduces the spectral conflicts in the two-

dye approach.

The relative absorption and emission spectra of Fluo-

rescein and Kiton Red are depicted in Fig. 3. To minimize

spectral conflicts during detection, appropriate spectral

bands were selected using two optical filters (Edmond

Optics). Fluorescein’s fluorescence was captured over

the wavelength range 535–545 nm (optical filter with full-

width half-maximum FWHM of 10, 540 nm center wave-

length), with minimal influence from Kiton Red’s emission.

The fluorescence emission of Kiton Red was captured by a

long pass filter over wavelengths greater than 600 nm, little

affected by fluorescein’s emission. The negligible leakage

through the filters was confirmed by preparing the Fluo-

rescein solution and measuring its emission by the fluo-

rescein camera, followed by introducing the Kiton Red dye

into the solution and measuring the emission intensity

again. The results showed the intensity change was about

0.1%, indicating there was negligible leakage through the

fluorescein filter. The same experiment was carried out by

preparing the solution with Kiton Red and then adding

Fluorescein into the solution, resulting in the same con-

clusion as before, i.e., leakage through the kiton red filter

was negligible. Note, in Fig. 3, that the Fluorescein emis-

sion over the selected spectral band is partially absorbed by

Kiton red, a spectral conflict that is difficult to avoid in the

two-color, two-dye approach. To reduce this conflict, the

molar concentrations of Fluorescein and Kiton red were

minimized at 5 9 10-7 M. The practical implication of this

spectral conflict, and possible additional self-absorption, is

that the detected emission intensity can become dependent

on the optical path length L within the liquid (i.e., the dis-

tance between laser sheet illumination and the viewed

test section wall). This influence is negligible in our study,

as will be quantified later through calibration experiments.

An important issue that needs consideration is the pH of

the solution. The working fluid in our study, Aqueous

Filter 2

Filter 1

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

400 450 500 550 600 650 700

Kiton Red, EmissionKiton Red, AbsorptionFluorescein, EmissionFluorescein, Absorption

Wavelength (nm)

Rel

ativ

e In

ten

sity

, I

Fig. 3 Absorption and emission spectra of Fluorescein and Kiton

Red. Green arrow designates argon-ion laser excitation at 514.5 nm.

The spectral bands of the optical filters used for detection are noted by

areas in shaded gray

654 Exp Fluids (2010) 48:651–662

123

ammonium chloride solution, is acidic and its pH is mea-

sured to be as low as 4.3 for a 26% w solution (Shafii

2005). There is also a slight variation of solution pH with

temperature (about 6% change over about 10�C; see Shafii

2005). It is known that fluorescence of Fluorescein is pH

dependent, increases significantly over the pH range 3–7

and becomes insensitive to it above pH of about 7. On the

other hand, Kiton Red does not exhibit a pH dependent

fluorescence over the pH range 3–10 (Coppeta and Rogers

1998). In order to operate in the pH-insensitive range of

Fluorescein, the pH of ammonium chloride solution was

always raised to about 7.8 by adding Sodium Hydroxide

(NaOH).

The calibration of the overall temperature response of

the two-color, Fluorescein-Kiton Red, system was carried

out in the same test section, and with the same optical and

detection configuration as the actual solidification experi-

ments; see Fig. 4. The test section was filled with the

aqueous solution (Ammonium Chloride 19% wt., 5 9 10-7

M Fluorescein, and 5 9 10-7 M Kiton Red) at room

temperature. This particular concentration of NH4Cl was

selected based on the phase diagram of ammonium chlo-

ride (Shafii 2005) to avoid solidification of the liquid

solution down to -15�C. The calibration was carried

out over the (4–20�C) range, however, because previous

experimental results of Wirtz et al. (1998) suggested that

the temperature of mush/melt interface was always above

4�C. The temperature of the fluid was lowered by lowering

the temperature of the base plate, whose temperature was

controlled by a recirculating constant-temperature bath (see

Section 3 for further details). The solution was stirred

continuously to ensure a uniform temperature over the

imaged area in the test section. The temperature of the

solution was independently monitored with a thermocou-

ple. The fluid was illuminated with a laser sheet (argon-ion

laser at 514.5 nm) and fluorescence images of Fluorescein

and Kiton Red were recorded by two aligned 8-bit,

640 9 480 pixel Pulnix cameras (TM-9701) at 30 fra-

mes s-1. Each camera was fitted with appropriate optical

filter (see earlier description) to capture the emission from

one dye. Calibration intensity data were averaged over 30

images and over a 32 9 32 pixel region of the image.

Figure 5 shows the measured intensity variation of both

dyes with temperature for two different optical path lengths

L. Measurements are also included for water solution (i.e.,

without NH4Cl) to assess the influence ammonium chloride

on the temperature response of these dyes. Intensity data of

each dye are normalized by their respective intensity at

reference temperature of 20�C. As expected for 514.5 nm

excitation, the fluorescence intensity of Fluorescein

decreases with decreasing temperature, whereas the emis-

sion of Kiton Red increases, instead. The normalized

intensity ratio, R(T)/R(20�C), was calculated from the data

in Fig. 5 and the results are illustrated in Fig. 6. Several

observations are noteworthy in this figure. First, the pres-

ence of ammonium chloride has a very minor affect on the

temperature response (i.e., calibration curve) compared to

that for the native aqueous solution. Second, the effect of

spectral conflict is small, but observable when we compare

the calibration curve for the two optical path lengths

L = 1 cm and 7 cm. Our studies of solidification typically

used optical path lengths in the range 1–3 cm (Shafii 2005)

and the results reported here are for L & 1 cm. Generally

speaking, before each experiment the appropriate calibra-

tion curve corresponding to its optical path length was first

determined. The final observation from Fig. 6 is that the

temperature sensitivity of the two-color, Fluorescein-Kiton

Red, system is about 4% per �C, significantly higher than

the 1.7% for the original Rhodamine B-Rhodamine 110

pair (Sakakibara and Adrian 1999). As mentioned earlier,

the approach of using two dyes with positive and negative

Argon-Ion Laser Mirror

Mirror

Spherical lens

Cylindrical lens

CCD Camera

CCD Camera

Green Filter for Fluorescein

Red Filter for Kiton Red

Beam Splitter

Test Section

Fig. 4 Schematic of optical setup for calibration and actual

experiments

0.7

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

5.0 7.5 10.0 12.5 15.0 17.5 20.0

Fluorescein (without NH4Cl; L = 1 cm)

Fluorescein (with NH4Cl; L = 1 cm)

Fluorescein (with NH4Cl; L = 7 cm)

Kiton Red (without NH4Cl; L = 1 cm)

Kiton Red (with NH4Cl; L = 1 cm)

Kiton Red (with NH4Cl; L = 7 cm)

Temperature, T (oC)

No

rmal

ized

Inte

nsi

ty,

I / I 20

OC

Fig. 5 Fluorescence intensity variation with temperature for Kiton

Red and Fluorescein (514.5 nm excitation; pH & 7.8; both dye

concentrations are 5 9 10-7 M; L represents the optical path length)

Exp Fluids (2010) 48:651–662 655

123

temperature sensitivities has also been recently reported by

Sutton et al. (2008) using other dye pairs with the 532-nm

excitation of Nd-YAG laser.

Figure 7 depicts the results from a preliminary test to

assess the effectiveness of the radiometric approach,

described above, for in situ mapping of the temperature

field in the uni-directional solidification problem under

investigation. Shown are the inferred temperature fields if

each dye (Fluorescein or Kiton Red) is used by itself using

its own single-dye calibration curve. The non-uniform

illumination field and ‘‘streaking’’ due to the index of

refraction variations render the results unreliable. The

temperature map using the ratiometric method does not

suffer from the artifacts caused by the streaks and the

nonuniform laser intensity field. The remaining oblique

streak on the upper portion of the image was a consequence

of the very low intensities of less than 20 counts in that

region (see the laser sheet intensity distribution in Fig. 2),

which make it difficult for the ratiometric approach to

remove the high intensity gradients in the streak. The

temperature field illustrated in Fig. 7 shows evidence of a

spanwise set of cold fingers protruding into the ambient

ammonium chloride solution. Section 4 provides more

specific examples of the evolution of the temperature field.

3 Experimental setup

The solidification system utilized is similar to the one

described in Magirl and Incropera (1993). The schematic of

the test section is shown in Fig. 8 and is identical to the one

described in Wirtz et al. (1998), Lum (2001, and Lum et al.

1.0

1.2

1.4

1.6

1.8

5 10 15 20

Exponential FitWith NH

4Cl; L = 1 cm

Without NH4Cl; L = 1 cm

With NH4Cl; L = 7 cm

Temperature, T (oC)

R(T

) / R

(20o

C)

Fig. 6 Variation with temperature of fluorescence intensity ratio, R,

of Kiton Red and Fluorescein (514.5 nm excitation; pH & 7.8; both

dye concentrations are 5 9 10-7 M; L represents the optical path

length)

Fluorescein Kiton Red

T (oC)

Ratiometric

Fig. 7 Sample temperature

map during solidification

experiment, arrived at using

only Fluorescein, only Kiton

Red, and ratiometric approach

Coolant inlet

Coolant outlet

Quartz container

Nucleators

Shrinkage compensator

11.17 mm 11.11 mm

89 mm

Row 1

Row 2

Row 3

11.71m

m14.29

mm

52mm

Fig. 8 Schematic of test section, stainless steel base plate, and

location of nucleators

656 Exp Fluids (2010) 48:651–662

123

(2001). An enclosure made of 1/400 quartz glass measuring

52 mm 9 89 mm 9 100 mm was placed over a stainless

steel base plate and sealed with high vacuum grease and

gaskets to form the test section. The base plate served as a

heat exchanger with a cooling chamber underneath whose

temperature was maintained by a constant-temperature bath

(Neslab RTE-140) that recirculated through the cooling

chamber a coolant mixture of 50:50 water and automotive

anti-freeze solution. A Plexiglas lid was placed above the

quartz enclosure and was fitted with shrinkage compensa-

tors to allow additional fluid to be supplied during the

course of the experiment to compensate for solidification

shrinkage. This ensured that the amount of fluid within the

test section at any time was kept constant. Eleven nucleators

of size 1.59 mm 9 1.59 mm 9 3.18 mm were embedded

in the base plate on a regular pattern to induce finite dis-

turbances and create convective plumes in geometrically

regular locations. The fluid used in these studies was a 26%

wt aqueous solution of ammonium chloride (i.e., a hyper-

eutectic ammonium chloride solution according to its phase

diagram) mixed with 5 9 10-7 M Fluorescein, and

5 9 10-7 M Kiton Red. The fluid in the test was initially at

room temperature (about 23�C) and the solidification pro-

cess was initiated by lowering the temperature of the base

plate to -14�C. The two-color, two-dye, approach requires

the dye concentrations in the solution to remain fixed.

According to the phase diagram of the hypereutectic

ammonium chloride solution used here, cooling of the fluid

results in the phase change (crystallization) of ammonium

chloride; the water itself does not freeze. As a result, we do

not expect variations of dye concentration to develop. This

was confirmed by comparing the temperature measured by

LIF against that from a thermocouple at specific locations in

the solution. The results agreed to better than 0.2�C.

The optical arrangement was discussed previously; see

Fig. 4. The beam of an argon-ion laser, operating at single

line (514.5 nm) and 1.5 W, was manipulated into a laser

sheet (thickness about 0.3 mm) using appropriate optics.

Images were obtained with two CCD cameras (Pulnix

TM9701, 8-bit, 640 9 480 pixel, 30 frames s-1), each fit-

ted with the appropriate optical filter (see Sect. 2.1). The

cameras viewed the same field of view through a cube

beamsplitter and were aligned to one pixel accuracy. The

framing rate of the camera used here is much faster than

needed. The convective flow generated in this uni-direc-

tional solidification problem is very slow (Lum 2001; Lum

et al. 2001) and it can be considered ‘‘frozen’’ over several

consecutive images. For example, the plumes shown in

Fig. 1 ascend at about 2 mm s-1. Therefore, temporal

averaging of data over consecutive frames could be used for

reducing the random error in the instantaneous tempera-

ture measurements. These issues will be discussed more

fully when measurements of temperature distributions are

presented in Sect. 4. In the early stages of the solidification

process (approximately the first 8 min) the flow convection

is exceedingly small in the liquid melt and possible

photobleaching of the fluorescent dyes needs to be consid-

ered. It was observed that the intensity of fluorescein would

drop about 1% upon two-seconds of continuous exposure to

laser. Therefore, a shutter was used to limit the exposure of

the solution to the laser to 0.5 s at a time. The bleaching rate

of Kiton Red was measured to be about one tenth of Fluo-

rescein’s and was not a factor. The effect of dye bleaching

in the worst case (i.e., roughly the initial 8 min of the

solidification process) would correspond to a temperature

error of about 0.06�C, based on the temperature sensitivity

of current approach.

The estimation of temperature requires calculation of

the intensity ratio at the same spatial location. To accom-

plish this, a reference target (1 mm grid pattern) placed in

the image plane was used to establish the relative dis-

placement between the pixels of one camera relative to

those of the other. This was done to sub-pixel accuracy

(better than 0.1 pixel) using the spatial correlation proce-

dure described by Gendrich and Koochesfahani (1996).

The resulting displacement field was used to define a third

order two-dimensional polynomial mapping function by a

least-squares procedure. A bi-linear interpolation scheme

was used to estimate intensities at sub-pixel locations.

4 Results and discussion

Before carrying out quantitative temperature measure-

ments, it would be prudent to establish that the presence of

fluorescent dyes in the liquid melt did not influence the

solidification process. Experiments were carried out with

and without the addition of fluorescent dyes at four times

the nominal concentration and shadowgraph image time

series, similar to Fig. 1, were analyzed. It was noted that

the timing of various events (e.g., fingers, plumes), the

wavelength of plumes and the rise speed of plumes all

remained unchanged. The growth of the mushy zone, cal-

culated from the shadowgraph images and shown in Fig. 9,

also confirmed that the low concentration of the dyes used

here did not influence the solidification process under

investigation.

We now discuss representative temperature measure-

ments. The imaged area selected was 25 9 19 mm2 in size

and was chosen to be located on the first row nucleators

approximately 2.25 cm above the cooling base plate. See

Fig. 10 for a schematic of the imaged region. The spatial

resolution of the measurement in the imaged plane was

about 40 lm per pixel. The measured temperature maps at

different times during the solidification process are shown

Fig. 11. The approximate location of the imaged area

Exp Fluids (2010) 48:651–662 657

123

relative to the mushy zone is indicated by the red box on

the shadowgraph images that are included in the figure.

Temperature maps in Fig. 11 are shown in three dif-

ferent ways corresponding to different temperature accu-

racy levels. The data in the left column represent the

instantaneous map, calculated from a single image pair of

the two CCD detectors corresponding to a temporal aver-

aging time of 30 ms. The noise characteristics of the CCD

cameras, measured using a highly stable light source,

corresponded to a rms intensity fluctuation of about 2.5%

of the mean intensity for the conditions of these experi-

ments (Shafii 2005). Assuming the noise of the two

detectors are uncorrelated, the temperature sensitivity of

current two-dye approach would suggest a temperature

measurement rms uncertainty of 0.8�C based on a single

image pair. This accuracy is comparable to the 1.4 K 95%

confidence level reported in Sakakibara and Adrian (1999).

As mentioned earlier, the convective flow generated in our

uni-directional solidification problem is very slow (Lum

2001; Lum et al. 2001) and it can be considered ‘‘frozen’’

over several consecutive images. Therefore, we take

advantage of temporal averaging over 10 consecutive

frames (i.e., 300 ms temporal averaging) in order to reduce

the random error in the temperature measurement to about

0.25�C. These results are shown in the temperature maps in

the middle column of Fig. 11. Comparison of temperature

maps in the left and middle columns clearly show the

justification of ‘‘frozen’’ flow state over the 300 ms inte-

gration period and the resulting reduction of the tempera-

ture measurement noise. The temperature maps in the right

column represent a further spatial filtering (8 9 8 pixel

area), using the Wiener method in Matlab, in order to

reduce the remaining striations in the spatial maps and a

further improvement in temperature accuracy by a factor of

eight, estimated at 0.03�C. As can be seen in Fig. 11,

because of the high resolution of the original images the

corresponding degradation of spatial resolution by spatial

filtering does not impact the ability to clearly resolve the

essential thermal features in this problem.

The temperature maps of Fig. 11 reveal the presence of

four fingers in the imaged area at time t = 17 min and one

of them has a lower temperature than the rest (i.e., the third

finger from the left located at X = 1.5 cm). This finger is

right above a nucleator and later becomes a plume, while

the adjacent fingers weaken and decline in number. We

note that the number of fingers visible in the temperature

map is significantly less that those in the shadowgraph

images. This is because the shadowgraph technique shows

a field of view that is integrated across the depth of the test

section, whereas the LIF temperature maps are obtained

over a single plane illuminated by a thin laser sheet.

Generally speaking, at each time the temperatures are

cooler at the bottom of the image than the top, as expected

from the mean vertical temperature gradient that is

imposed. Also, as time progresses the temperature field

over the imaged plane becomes cooler over the entire plane

as the mush zone grows and gets close to the bottom of the

images area. We note that at t = 40 min three of the four

fingers subside and only one finger survives as a plume. At

this time, the mushy zone height is about 1.8 cm, 0.45 cm

below the imaged area. The missing portion of the plume

0

5

10

15

20

25

30

0 20 40 60 80 100 120

With Fluorescein and Kiton RedTrial #2 (without fluorescent dyes)Trial #1 (without fluorescent dyes)

Time (min)

Hei

gh

t o

f M

ush

y R

egio

n (

mm

)

Fig. 9 Growth of the mushy region with time. Kiton Red and

Fluorescein concentrations are 2 9 10-6 M; pH & 7.8

Imaged Area 25 x 19 mm

2

2.25 cm

10 cm

8.9 cm

Imaged area

Fig. 10 Location and size of the imaged area for the temperature

maps in Fig. 11

658 Exp Fluids (2010) 48:651–662

123

on the top of the temperature map is likely due to the

helical motion of the plume, causing the center of plume to

move out of the illuminated laser sheet. At time

t = 60 min, the mushy zone has reached the bottom of the

imaged area and one distinct plume is clearly observed.

The feature on the right side of the plume is speculated to

be a finger that is about to form. At this time the mush

interface has a temperature of 8.4�C, whereas the temper-

ature of the plume at the exit of the mush is cooler and is

about 7.8�C.

Temperature profiles along a horizontal line at selected

times are extracted from the temperature maps of Fig. 11 in

T (oC)

X (cm)X (cm)X (cm)

Y (c

m)

t = 17min

T (oC)

X (cm)X (cm)X (cm)

Y (c

m)

t = 30min

T (oC)

Y (c

m)

X (cm)X (cm)X (cm)

t = 40min

T (oC)

Y (c

m)

X (cm)X (cm)X (cm)

t = 60min

Fig. 11 Temperature maps at different times during the solidification

process over the region indicated in Fig. 10. Left column represents

the instantaneous map, middle column is average of ten consecutive

images (i.e., 300 ms temporal average), and right column represents

further spatial filtering over 8 9 8 pixel area. The approximate

location of the imaged area relative to the mushy zone is indicated by

the red box on the shadowgraph images

Exp Fluids (2010) 48:651–662 659

123

order to quantify the temperature profiles across the finger

and plumes. Results are shown in Fig. 12 across a line at

Y & 0.3 cm in the temperature maps (about 2.6 cm above

the cooling plate). Data are shown for both 300 ms tem-

poral averaging and those after 8 9 8 pixel spatial filtering.

When compared with the spatially smoothed profiles, the

un-smoothed data indicate about 0.3�C rms fluctuation,

consistent with the previous estimate of temperature

accuracy for data with 300 ms temporal averaging. Tem-

perature profiles in Fig. 12 clearly show that the location of

the finger that later becomes the plume change very little

during the solidification process (minimum peak always

remains at X & 1.5 cm). We also note that the minimum

temperature in the center of the finger and the plume

illustrated in Fig. 12 is typically two degrees cooler than

the surroundings.

It is also instructive to examine the vertical variation of

temperature differential DT between the ambient and the

finger and plume. Figure 13 shows the measured difference

between the two indicated vertical lines at t = 17 min and

t = 60 min. The temperature difference right at the exit of

the mushy zone (Y = 0 cm and t = 60 min) is about

0.8�C. The maximum temperature difference reaches about

2.7�C and occurs when the plume is formed. 5 Conclusions

A two-color, two-dye, ratiometric laser induced fluores-

cence method was implemented for carrying out in situ

whole-field measurements of the temperature field in a

study of bottom-chilled uni-directional solidification of

aqueous ammonium chloride. The traditional approach

uses a temperature sensitive dye along with another with

little temperature sensitivity. In our work, however, we

took advantage of two temperature sensitive fluorescent

dyes, with opposite temperature sensitivities, in order to

increase the overall sensitivity and temperature resolution

of our measurements. The resulting temperature sensitivity

was more than a factor of two higher than the original work

of Sakakibara and Adrian (1999).

Initial measurements of the in situ temperature field

during solidification of ammonium chloride solution were

presented for the first time. Many interesting characteristics

of the process were quantified, including formation and

evolution of fingers and plumes and the temperature dis-

tributions within them. It was shown that as the plume

moves upward the maximum temperature difference

between the center of the plume and the surroundings

reaches about 2.7�C.

We note that even though the locations of the pixels of

one CCD camera relative to the pixels of the second

camera were determined with sub-pixel accuracy (better

than 0.1 pixel), some streaks and striations still remained

in the final temperature maps. The reasons for the

remaining streaks have been discussed by Sakakibara and

4

6

8

10

12

14

16

18

20

22

24

0 0.5 1.0 1.5 2.0 2.5

t = 60mint = 40mint = 30mint = 17min

X (cm)

Tem

per

atu

re, T

(oC

)

Y (

cm)

X (cm)

Fig. 12 Time variation of the horizontal temperature profile along

the indicated dashed line (Y & 0.3 cm). Symbols are data points

(every other 3 points shown) from the middle column of temperature

maps in Fig. 11 (i.e., 300 ms temporal average). Solid lines represent

the 8 9 8 pixel spatial average data (right column of temperature

maps in Fig. 11)

Y (

cm)

X (cm)

0

1

2

3

4

5

0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

t = 60mint = 17min

Y (cm)

Tem

per

atu

re D

iffe

ren

ce,

∆T

(oC

)

Fig. 13 The vertical profile of temperature difference DT across the

indicated dashed lines at two times during the solidification process.

One of every five data points is shown

660 Exp Fluids (2010) 48:651–662

123

Adrian (2004) where they also suggest a blurring method

for improvement. Subsequent to the results we presented

here, improvements were achieved in terms of both

removing the streaks and also temperature accuracy by

using higher resolution CCD cameras with better dynamic

range (Cooke Corp. Pixelfly, 12-bit, 1,024 9 1,392 pixel).

Combination of better detection signal-to-noise ratio and

higher pixel density led to an increase in temperature

accuracy by a factor of four compared to those presented

here (see Shafii 2005). Sample of a temperature map

acquired with improved detection is shown in Fig. 14.

This figure illustrates the temperature map of ammonium

chloride solution at t = 9 min over an imaged area

30 9 22 mm2 located above the second nucleator on the

second row. The spatial resolution of the measurement in

the plane of illumination is about 21 lm per pixel, a

factor of two better than before. Temperature variations

within fingers are shown in great detail and clarity, with

little evidence of streaks.

Acknowledgments This work was supported by NASA, Grant No.

NAG8-1690. The assistance of Dr. Shahram Pouya in the preparation

of the manuscript is gratefully acknowledged.

References

Beckerman C, Wang CY (1996) Equiaxed dendritic solidification

with convection: Part III Comparisons with NH4Cl-H2O exper-

iments. Metall Mater Trans A 27A:2784–2795

Chen CF, Chen F (1991) Experimental study of directional solidifi-

cation of aqueous ammonium chloride solution. J Fluid Mech

227:567–586

Coolen MCJ, Kieft RN, Rindt CCM, Van Steenhoven AA (1999)

Application of 2D LIF temperature measurements in water using

a Nd:YAG laser. Exp Fluids 27:420–426

Coppeta J, Rogers C (1998) Dual emission laser induced fluorescence

for direct planar scalar behavior measurements. Exp Fluids

25:1–15

Feltham DL, Worster MG (1999) Flow-induced morphological

instability of a mushy layer. J Fluid Mech 391:337–357

Gendrich CP, Koochesfahani MM (1996) A spatial correlation

technique for estimating velocity fields using molecular tagging

velocimetry (MTV). Exp Fluids 22(1):67–77

Heinrich JC, Felicelli S, Nandapurkar P, Poirier DR (1989) Thermo-

solutal convection during solidification of alloys: part II.

Nonlinear convection. Metall Trans B 20B:883–891

Kim HJ, Kihm KD (2001) Application of a two-color laser induced

fluorescence (LIF) technique for temperature mapping. In:

Proceedings of ASME IMECE, 11–16 November, New York,

USA

Kim HJ, Kihm KD, Allen JS (2003) Examination of ratiometric laser

induced fluorescence thermometry for microscale spatial mea-

surement resolution. Int J Heat Mass Transf 46:3967–3974

Lavieille P, Lemoine F, Lavergne G, Lebouche M (2001) Evaporation

and combusting droplet temperature measurements using two-

color laser-induced fluorescence. Exp Fluids 31:45–55

Lavieille P, Delconte A, Blondel D, Lebouche M, Lemoine F (2004)

Non-intrusive temperature measurements using three-color laser-

induced fluorescence. Exp Fluids 36:706–716

Lemoine F, Antoine Y, Wolff M, Lebouche M (1999) Simultaneous

temperature and 2D velocity measurements in a turbulent heated

jet using combined laser-induced fluorescence and LDA. Exp

Fluids 26:315–323

Liu S, Hellawell A (1999) Experiments with constrained chimney-

plume flows in the system ammonium chloride-water: compar-

ison with unconstrained case. J Fluid Mech 388:21–48

Lum CL (2001) Velocity field measurement for a uni-directional

solidification of an ammonium chloride (NH4Cl) solution using

molecular tagging velocimetry (MTV). Dissertation, Michigan

State University

Lum CL, Koochesfahani MM, McGrath JJ (2001) Measurements of

the velocity field with MTV during the solidification of an alloy

analog with mushy region. Proc ASME/IMECE 31:1–7

Magirl CS, Incropera FP (1993) Flow and morphological conditions

associated with unidirectional solidification of aqueous ammo-

nium chloride. Trans ASME 115:1036–1043

Neilson DG, Incropera FP (1991) Unidirectional solidification of a

binary alloy and the effects of induced fluid motion. J Heat

Transf 34:1717–1732

Nishimura T, Wakamatsu M (2000) Natural convection suppression

and crystal growth during unidirectional solidification of a

binary system. Heat Transf Asian Res 29:120–131

Nishimura T, Fujiwara M, Miyashita H (1992) Visualization of

temperature fields and double-diffusive convection using liquid

crystals in an aqueous solution crystallizing along a vertical wall.

Exp Fluids 12:245–250

Saeki S and Hart DP (2001) Investigation on YAG (532) laser dyes

for oil film thickness and temperature measurement. In:

Proceedings of the third pacific symposium of flow visualization

and image processing (Paper index number F3096), 18–21

March 2001, Maui, Hawaii, USA

Sakakibara J, Adrian RJ (1999) Whole field measurement of

temperature in water using two-color laser induced fluorescence.

Exp Fluids 26:7–15

Sakakibara J, Adrian RJ (2004) Measurement of temperature field of a

Rayleigh–Benard convection using two-color laser-induced

fluorescence. Exp Fluids 37:331–340

Sakakibara J, Hishida K, Maeda M (1993) Measurements of thermally

stratified pipe flow using image-processing techniques. Exp

Fluids 16(2):82–96

T (oC)t = 9 min

X (cm)

Y (c

m)

Fig. 14 Temperature map of imaged area (30 9 22 mm2) using

1,024 9 1,392 pixel, 12-bit CCD cameras (300 ms temporal average,

8 9 8 pixel spatial average). Gray region indicates the area below the

laser sheet where no data is available

Exp Fluids (2010) 48:651–662 661

123

Schneider MC, Beckermann C (1995) A numerical study of the

combined effects of microsegregation, mushy zone permeability

and flow, caused by volume concentration and thermosolutal

convection, on macrosegregation and eutectic formation in

binary alloy solidification. Int J Heat Mass Transf 38:3455–3473

Schulze TP, Worster MG (1998) A numerical investigation of steady

convection in mushy layers during the directional solidification

of binary alloys. J Fluid Mech 356:199–220

Shafii MB (2005) Temperature measurement of aqueous ammonium

chloride solution during solidification process using laser-

induced fluorescence. Dissertation, Michigan State University

Shafii M, Koochesfahani MM (2003) Temperature measurement of

aqueous ammonium chloride solution during solidification

process using laser induced fluorescence (LIF). Bull Am Phys

Soc 48(10):180

Solomon TH, Hartley RR (1998) Measurements of the temperature

field of mushy and liquid regions during solidification of aqueous

ammonium chloride. J Fluid Mech 358:87–106

Sutton JA, Fisher BT, Fleming JW (2008) A laser-induced fluores-

cence measurement for aqueous fluid flows with improved

temperature sensitivity. Exp Fluids 45:869–881

Sato K, Kasagi N, Suzuki, Y (1997) Combined velocity and scalar

field measurement with the simultaneous use of PIV and

scanning LIF. In: Proceedings of 10th International symposium

on transport phenomena in thermal science and process

engineering, Kyoto 1997, vol 2, pp 541–546

Wirtz K, Koochesfahani MM, McGrath JJ, Benard A (1998)

Molecular tagging velocimetry applied to buoyancy-driven

convective phenomena during solidification. Proc ASME Heat

Transf Division ASME Pap HTD 361–4:103–110

Worster MG (1991) Natural convection in a mushy layer. J Fluid

Mech 224:335–359

Worster MG (1992) Instabilities of the liquid and mushy regions

during directional solidification of alloys. J Fluid Mech 237:649–

669

662 Exp Fluids (2010) 48:651–662

123