Experiments in Fluids 37 (2004) 341–349
DOI 10.1007/s00348-004-0822-2
Application of LIF to investigate gas transfer near the air-water interface
in a grid-stirred tank
Herlina, G. H. Jirka
Abstract The interaction between oxygen absorption into
liquids and bottom shear-induced turbulence was inves-
tigated in a grid-stirred tank using a laser-induced fluo-
rescence (LIF) technique. The LIF technique enabled
visualization as well as quantification of planar concen-
tration fields of the dissolved oxygen (DO) near the air-
water interface. Qualitative observation of the images
provided more insight into the physical mechanism con-
trolling the gas transfer process. The high data resolution
is an advantage for revealing the concentration distribu-
tion within the boundary layer, which is a few hundreds of
a micrometer thick. Mean and turbulent fluctuation
characteristics were obtained and compared with previous
results.
1 For practical use, the thickness d in Eq. (1) gives an
Introduction
Gas transfer from air into liquids is an important process
occurring in nature as well as in industrial facilities. One
case is the oxygen absorption into polluted rivers. This
reaeration process is very critical to the aquatic habitat
because it recovers the deficit of dissolved oxygen in
polluted rivers. This study is motivated by this problem
and has the objective of gaining more understanding of the
physical mechanisms that control the oxygen absorption
process.
The flow conditions in nature are typically turbulent,
and it is well known that turbulence enhances the gas
transfer process. Many researchers tried to explain the
physical mechanism of the process using conceptual
models. Some of the models frequently mentioned in the
literature are briefly described here. The ‘‘film model’’ by
Lewis and Whitman (1924) assumed a stagnant film on
each side of the interface where only molecular diffusion
Received: 25 July 2003 / Accepted: 6 April 2004
Published online: 27 May 2004

Springer-Verlag 2004
Herlina (&), G. H. Jirka (&)
Institute for Hydromechanics, University of Karlsruhe,
Kaiserstr.12, 76128 Karlsruhe, Germany
E-mail: herlina@ifh.uka.de
E-mail: jirka@uka.de
This project was funded by the ‘‘German Science Foundation’’
(DFG Grant No. Ji18/7–1). The authors would like to express their
appreciation for the constructive suggestions from anonymous
reviewers.
341
takes place. This leads to a linear concentration profile
within the stagnant films and the following relation could
be deduced
D
KL ¼ ð1Þ
d
with KL the transfer velocity, D the molecular diffusivity
and d the thickness of the stagnant film. It is clear that this
model oversimplifies the actual mechanism. However, it
should be noted that Lewis and Whitman’s (1924) actual
purpose of proposing the ‘‘film model’’ was to estimate the
relative importance of gas-phase and liquid-phase resis-
tances between different gases with varying solubility. The
linear assumption was sufficient for that purpose. In the
literature, however, the ‘‘film model’’ has often been used
to explain the actual mechanism of gas transfer.
estimate on the lower bound of the boundary layer
thickness, although it may underestimate the actual value
by a factor of ten (Lee 1977). Higbie (1935) developed a
‘‘penetration model’’ which is based on a surface renewal
effect. His idea was that the turbulence in the bulk would
bring up fresh packages of liquid to the surface, where gas
transfer takes place for a certain renewal time T.
Danckwerts (1951) modified Higbie’s model by allowing T
to follow a random distribution. The random function s(T)
should be determined experimentally for individual tur-
bulence conditions. Fortescue and Pearson (1967) elabo-
rated the surface renewal model by introducing the ‘‘large
eddy model’’. Their results showed that the largest tur-
bulent eddies dominate the gas transfer process. Lamont
and Scott (1970) on the other hand suggested that small
eddies are the dominant mechanism controlling the
transfer process.
Despite intensive research efforts, the physical mecha-
nism of the interaction between the turbulence and the gas
transfer process still remains unclear. For example, the
question of which eddy size contributes more to the gas
transfer process is still open. It is, therefore, the objective
of this study to gain more understanding of the actual
mechanism controlling the gas transfer process in a tur-
bulent environment. In general, there are three major
types of turbulence generation in the environment, namely
wind-induced turbulence, bottom shear-induced turbu-
lence and buoyant convection turbulence (Brutsaert and
Jirka 1984). Our study focuses on the second type of tur-
bulence generation.
For slightly soluble gases, including oxygen, a boundary
layer of ten to hundreds lm thick on the liquid side
 controls the gas transfer process. This makes measure- This set-up allows larger scales of turbulent eddies to grow
ments near the interface very difficult. Employing intrusive and thus a wider spectrum of turbulent eddies can be
measurement sensors would disturb the actual process. observed.
Here, a laser induced fluorescence (LIF) technique based
on the oxygen quenching phenomenon was used to mea- 2
sure the concentration fields near the interface. Besides Experimental set-up and technique
being non-intrusive, the high data density obtained with
this technique is also an advantage for elucidating the 2.1
profile near the boundary with its limited thickness. Turbulence generation
Moreover, 2D visualization with good temporal resolution A grid-stirred tank system was employed to produce tur-
(see e.g. Münsterer and Jähne 1998) of the DO distribution bulence similar to that generated by bottom shear. This
342
would be of great interest to give more insight into the type of turbulence generation was chosen because it is
actual process. characterized by zero-mean velocity and by turbulence
The oxygen quenching method employed here has been homogeneity. In comparison with open channel flows, the
developed by Vaughan and Weber (1970) to measure zero-mean velocity in the tank reduces the intricacy in
oxygen concentration in living tissues. Wolff et al. (1990) performing measurements since no mean flow is present.
were the first who adopted this method to study gas In addition, the results of Brumley and Jirka (1987)
transfer processes. They measured 1D oxygen concentra- showed that the grid-stirred tank is suitable for investi-
tion profile near the air-water interface in a small chamber. gation in the area of gas transfer and direct data com-
Since then, LIF based on the oxygen quenching technique parison is allowed.
has been applied by other investigators in the field of gas A schematic illustration of the experimental set-up is
transfer. Some of these works, which are relevant to our shown in Fig. 1. The tank was made of Perspex, with a
study, are shortly mentioned here. Wolff and Hanratty 50·50 cm2 bottom and a depth of 65 cm. Glass windows
(1994) employed this method to study oxygen absorption were installed at the side walls to enhance optical access. A
in a stratified air-water flow with smooth and wavy vertically oscillating grid installed near the bottom of the
interfaces. Münsterer et al. (1995) performed measure- tank generated the turbulence, and consisted of an alu-
ments in a circular wind/wave flume using two types of LIF minium plate perforated to form a 7 bar·7 bar grid. The
technique simultaneously. The first technique was LIF centre-to-centre mesh size, M, was 63.5 mm, and the hole
based on the oxygen quenching method and the second size was 50·50 mm2, resulting in a mesh solidity of 36%.
was a Fluorescein technique employing HCl as tracer. They
could obtain 2D concentration information with very high
time resolution up to 200 frames/s using the Fluorescein-
HCl method. On the other hand, only 1D concentration
profiles could be measured with the oxygen quenching
method. The oxygen quenching method was then
improved by Duke and Hanratty (1995) to measure not
only 1D profiles but also 2D concentration fields in an air-
water flow channel with wavy and flat interfaces. Other 2D
measurements have been performed recently by Woodrow
and Duke (2001) in a wave tank.
All applications of the 2D LIF-oxygen quenching tech-
nique mentioned above had been performed in channels
with very shallow water depth (less than 30 mm) and the
interrogation area were relatively near the side walls (less
than 40 mm). Recently, Lee (2002) and Schladow et al.
(2002) applied the 2D LIF technique in a deep water tank
(0.6 m) to visualize the oxygen transfer across the air-
water interface driven by natural convection. However, it
should be noted that in their case the thermal boundary
layer was between 2 and 4 mm so they had not looked into
the small scale of ten to hundreds of lm thick, which is the
scale that must be elucidated in the other previous LIF
works as well as in our present case.
The application of the LIF technique in this paper dif-
fers from previous works in that it was applied for the first
time in a grid-stirred tank that enables the simulation of
gas transfer process at the interface induced by bottom
shear turbulence. The water depth of 473 mm, which was
considerably deeper than previous set-ups, and the inter-
rogation window which was about 250 mm away from the
sidewalls, made the application of LIF more complicated. Fig. 1. Experimental set-up and LIF configuration
This solidity was considered optimal to avoid secondary
motions and inhomogeneity (Hopfinger and Toly 1976).
The vertical distribution of the turbulence in such a
grid-stirred tank can be related to the grid parameters as
has been shown by Hopfinger and Toly (1976):
UHT ¼ 0:25 f S1:5 M 0:5 z
1
s ð2Þ
where f being the frequency of the oscillation, S is the
stroke, and z s is the distance from the centre of the stroke
towards the water surface.
Thompson and Turner (1975) found that the longitu-
dinal integral length scale L¥ proportional to zs is:
L1 ¼ 0:1zs ð3Þ
Brumley and Jirka (1987) used 2 L¥ and UHT evaluated
at the water surface as the length and velocity scales,
respectively, to characterize different turbulence levels for
the transfer process. The Reynolds number ‘‘Re’’ can thus
be defined by
UHT 2L1
Re ¼ ð4Þ
m
where m is the kinematic viscosity.
In this study, the grid was operated with a 50 mm
stroke at 3 Hz. The water depth was approximately
473 mm with a temperature of 10 C. The position of the
centre of the grid stroke was 200 mm from the bottom or
273 mm from the water surface. The Reynolds number for
the present experimental condition was therefore 325
based on Eq. (4).
Preliminary experiments showed that the undulations
at the water surface were typically smaller than 30 lm, so
that the water surface could be considered as flat.
2.2
LIF application
The present LIF technique takes advantage of the oxygen
quenching phenomenon. Vaughan and Weber (1970)
showed that the fluorescence lifetime of pyrene butyric
acid (PBA) is quenched when oxygen is present in the
solution. The fluorescence lifetime reduces from 160 ns in
the absence of oxygen to 65 ns in air-saturated solution
(Wolff and Hanratty 1994). The fluorescence lifetime s is
proportional to the fluorescence intensity F and depends
on the quencher concentration Q as described by the
Stern-Vollmer equation:
Fo s o
¼ ¼ 1 þ Ksv Q ð5Þ
F s
where so and Fo are the fluorescence lifetime and inten-
sity, respectively, in the absence of the quencher, Ksv the
Stern-Vollmer quenching constant. Hence, the quencher
(in this case dissolved oxygen) concentration level can be
determined by measuring the intensity of the emitted
fluorescence.
Figure 1 shows the LIF system configuration which
consists of tracer, illumination system, image recorder and
synchronization unit. The optical set-up was in general
similar to the one developed by Duke and Hanratty and
Woodrow and Duke (2001), except that here the light sheet
was aligned from above the water surface instead from the
bottom of the tank. The illumination from above was
similar to Münsterer et al. (1995), only here the laser beam
was expanded into a sheet. The individual components of
the optical set-up used in this study are described as fol-
lows.
The test tank was filled with 2·10)5 Molar PBA, which
served as DO tracer. A pulsed nitrogen laser (MNL 801)
with a mean energy power of 0.4 mJ was used to excite the
PBA solution. The laser emission was at 337.1 nm, which
was ideal since the absorption peak of PBA is at 340 nm. A
343
UV-mirror was placed in front of the laser to guide the
laser beam into the centre of the test tank. The laser beam
was then expanded to a sheet by a combination of two
spherical and one cylindrical (f=)25.4 mm) lenses. The
configuration of the illumination system was so that the
sheet excited a 2D flow field at the centre of the tank. A
disadvantage of directing the laser sheet from above the
water surface is that dust particles floating on the water
surface could partially block the laser light as it travels into
the test tank. This means that additional noise is added to
the recorded intensity images when particles happened to
float at the water surface within the measurements win-
dow. An ideal configuration to avoid this type of noise is if
the laser light enters the test tank from the side walls.
However, a laser with a very high power would be there-
fore required so that sufficient intensity could still reach
the measurement area. For the laser power available, the
present illumination configuration was an optimal solu-
tion. A requirement was of course, to keep the surface as
clean as possible. Therefore, the surface was cleaned with a
suction device prior to the experiment.
The intensity of the PBA fluorescence was captured
with a FlowMaster CCD camera. It has a resolution of
1280·1024 pixels and 12 bit. A macro objective (Compo-
non-S 5.6/135 from Schneider Optics) was used to zoom
into an investigation area of approximately 8·7 mm2 from
a distance of about 30 cm. Hence, the measurement res-
olution was as small as 6 lm. The PBA fluorescence lies
between 370–410 nm. An optical bandpass filter was
mounted in front of the camera in order to block out the
scattered laser light and all ambient light. Hence, it was
ensured that only the fluorescent light could pass through.
The camera was tilted about 10 so that the viewing point
was slightly from below the water surface. This adjustment
was necessary to eliminate interference from the water
surface meniscus at the side wall facing the camera. The
recording interval of the camera was 4 Hz. Restricted by
the RAM capacity of the computer, the maximum number
of successive images in one experiment is limited to 300.
A PTU board synchronized all LIF activities including
the synchronization between the laser pulse and camera.
The commercial software DaVis provided the interface for
communications between the user and the LIF system.
DO concentration in the bulk were measured using the
InoLab Level 3 Oximeter. It was installed approximately
10 cm below the centre of the grid stroke. Nitrogen bub-
bles were introduced into the test tank to purge the oxygen
from the water. An initially low DO concentration of about
0.6 mg/l was achieved after 20 min of bubbling. The
measurements started shortly after that.
 3
Image processing
Several processes must be performed on the raw images
before the concentration information can be extracted.
These processes include noise removal, water surface
detection, correction of laser attenuation, and correction
of optical blurring near the interface. The image process-
ing procedure was, in general, similar to that described in
more detail by Woodrow and Duke (2001). The specific
procedure and optimal parameters for our case are dis-
344 cussed as follows.
First, 10 iterations of filtering in the horizontal x and
vertical z directions were carried out to remove noise. An
adaptive filter with filter size of 9 pixels was applied in the
z direction. In the x direction, the filter size was 19 pixels.
The different filtering size was chosen since the changes
along the horizontal direction are not as abrupt in the
vertical direction. The advantage of the adaptive filter is
that it preserves strong gradients but it requires longer
processing time.
The next processing step was detecting the water sur-
face (z = 0). As mentioned previously, the camera was
slightly tilted so that the reflection on the water surface
was also recorded on the image. The location of the water
surface could thus be detected by searching for the
reflective symmetry line. Evaluating a single vertical col-
umn, the symmetry point would be at the zero-crossing of
the first derivative of the profile. The Canny method was
used to detect the zero-crossing. Provided that the surface
could be considered flat, the surface detection algorithm Fig. 2. Vertical intensity profile from raw image (a) and processed
was then performed on a restricted area where it was image (b)
expected to occur in order to reduce the running time of
the program.
The intensity of the laser light decays exponentially where Fi and Fb are the fluorescence intensity at the
when it travels through the water as described by the interface and the bulk, respectively.
Lambert-Beer theory. Consequently, the fluorescence re- Performing the conversion from fluorescence values to
sponse must be corrected to take account of this intensity DO levels on all vertical profiles, a whole planar contour
attenuation. The Lambert-Beer exponent was determined map of the instantaneous DO concentration was obtained.
by fitting an exponential curve on each vertical intensity A series of such contour maps is shown in the supplied
profile in the bulk region. The fitted exponents from figures.
twenty randomly chosen images were then averaged. The
Lambert-Beer effect on each instantaneous image was then 4
corrected by dividing the recorded intensity with the Verification of the LIF set-up
averaged exponent. Preliminary experiments with no grid movements were
The filtering and correction of laser attenuation reduce performed using the same arrangements as the actual
the sharp gradient at the interface. Assuming that directly experiments in order to verify the current LIF set-up.
below the water surface only molecular diffusivity occurs Five sets of images were recorded and the oxygen
(i.e. a linear concentration distribution), the blurring effect concentration was varied between 0.9 to 7.3 mg/l. The
is corrected by extrapolating the sharpest gradient up to images were processed with the same image-processing
the interface (see Fig. 2b). Figure 2 shows an example of a algorithm described in the previous section, except that no
vertical intensity profile before and after the image pro- transformation from fluorescence intensity to concentra-
cessing. Finally, the concentration level of the dissolved tion was done. The normalized fluorescence intensities
oxygen can be extracted. Assuming no gas phase resis- obtained for each level of oxygen concentration is plotted
tance, the dissolved oxygen concentration at the interface in Fig. 3. The data points are the mean value from 100
is equal to the saturation concentration Cs . The conver- instantaneous images recorded in each set. The mean
sion from each vertical profile of the measured fluores- value was taken in order to rule out the variation of the
cence intensity F values to DO concentration values C can incident laser light intensity that was about ±4%. Mün-
be described by the following relation: sterer et al. (1995) found that the best fit of the Stern-
Vollmer equation (Eq. 5) yielded a Ksv value of 683 ± 70 l/
Cs
Cb mol. The curve is included in the graph and it agrees well
C ðz Þ ¼ ½Fb
F ðzÞ þ Cb ð6Þ
Fb
Fi with the present data points.
 5
Discussion
The fluorescence images captured enabled qualitative
visualization of the flow field and also quantitative analysis
of the dissolved oxygen concentration distribution
including the statistical mean and turbulence properties.

5.1
Qualitative observations
Figures 4 and 5 show a sequence of 12 instantaneous DO
fields. The images were recorded at 4 Hz. In Fig. 5, how- 345
ever, only every third recorded image is shown, thus the
time interval between the shown images is 0.75 s. The dark
and light colour scaling in the images correspond to the
regions with high and low oxygen concentration, respec-
Fig. 3. Relation of fluorescence intensity to oxygen concentration tively. Therefore, the following general description for all
images is expected. The darkest layer occurs at the water
It should be noted that the effect of the inconsistency in surface where the oxygen concentration reaches 100%
the incident laser light would be eliminated in the final saturation. In the vicinity of the interface, a very thin dark
step where fluorescence intensity is transformed into DO layer indicating the boundary layer in which the oxygen
concentration since for each vertical profile the intensities concentration decreases rapidly from saturated to the
are normalized when applying Eq. 6. It is difficult to concentration in the bulk region can be observed. Below
determine the individual contribution to the total error this boundary layer, the images are dominated with light
from each component of the whole LIF set-up. However, colour corresponding to the low oxygen concentration in
the standard deviation from DO images with the same the bulk region that is constantly being mixed by the
concentration suggests that the resolution of the current turbulence generated by the oscillating grid.
set-up is about 5%. It is interesting to observe the actions occurring in the
instantaneous images. The turbulence obviously plays a

Fig. 4. A sequence of DO field with peeling process

346
Fig. 5. A sequence of DO field with a small eddy structure
role in the gas transfer process as can be compared to
Fig. 6, which is one typical example of images taken in a
quiescent environment (i.e. no grid movements). There
were no dark strikes found penetrating the bulk region in
all images taken in quiescent environment. This fact
ensures that the phenomena visible in Figs. 4 and 5 are
indeed the effect of the turbulence and not optical, pro-
cessing, or flow artifacts.
Figure 4 shows a transport process in which portions of
the surface layer with higher oxygen concentration was
peeled off by a turbulent structure and then transported
Fig. 6. Typical instantaneous image with no grid movements
into the bulk region. This observation agrees with the idea
of the surface renewal model as has also been observed by
other investigators (e.g. Woodrow and Duke 2001). The
presence of turbulence obviously aids the oxygen
absorption into the water body.
In the sequence presented in Fig. 5, a very small tur-
bulent structure is visible. It can be observed that as soon
as the structure impinges on the water surface (about the
third frame), the vertical growth of the structure is
inhibited and consequently it undergoes horizontal
stretching and finally diffuses.
For comparison purposes, the characteristic length
scale of the smallest turbulent eddies is the Kolmogorov
scale g. Following Brumley and Jirka (1987), g is defined as
g ¼ 2L1 Re
3=4 ð7Þ
Based on this equation, the Kolmogorov scale for the
present turbulent condition is about 0.7 mm. The large
structures on the other hand would be theoretically of
order integral scale L¥=27 mm. The eddy captured in
Fig. 5 is therefore of intermediate scale since its size is
about 4 mm. The size of the eddy initiating the peeling
process shown in Fig. 4 is obviously larger than the
observation window.
Comparing the observations in Figs. 4 and 5, the
present results suggest that both large and small structures
contribute to the transfer process.
5.2
Mean and turbulence characteristics
Two consecutive series of measurements with 300 instan-
taneous images each were performed. As can be inter-
preted from the DO fields shown in Figs. 4 and 5, the
vertical distribution of the DO concentration changes over
time and space including also in the bulk region outside
the boundary layer. The time mean averaged and standard
deviation profiles from the 600 instantaneous images are
shown in Fig. 7.
Assuming an exponential distribution as proposed by 347
Chu and Jirka (1992), the boundary layer thickness ze may
be defined as the depth where the value of the normalized
concentration ( C ) Cb) = ( Cs ) Cb) is equal to 1/e = 37%.
The instantaneous thickness changes with time and space
due to the turbulence structure impinging on the bound-
ary as can be observed in Figs. 4 and 5. The time mean
averaged boundary layer thickness for the present exper-
imental condition was approximately 420 lm.
It is convenient to present the vertical distribution in a
non-dimensional form to allow comparison with other
experiments despite the different turbulence characteris-
tics. Here the boundary layer thickness z e is used to scale
the depth and the normalized concentration is presented
in the form of ( C ) Cb) = ( Cs) Cb ). The profiles are
shown in Fig. 8. The data points from Chu and Jirka
(1992) with Re =360 and Atmane and George (2002) are
incorporated in the figure for comparison. They both
performed gas transfer measurements in experimental

Fig. 8. Normalized concentration profile mean (a) and rms (b)

facilities similar to the present study. The mean distribu-

tion is exponential and shows excellent agreement with the
previous investigations. The exponential distribution also
shows good agreements with the solution of the mass flux
equation when incorporating the surface renewal model
assumption (see e.g. Jähne and Haussecker (1998)).
The peak of the fluctuations are found to occur around
the boundary layer thickness. The fluctuation values of the
present data show some discrepancy with the previous
data. The present values are higher compared to that
measured by Chu and Jirka (1992). It should be noted that
both Atmane and George (2002) and Chu and Jirka (1992)

Fig. 7. Time averaged concentration profile mean (a) and rms (b) Fig. 9. Bulk concentration measurement
 Table 1. Typical KL values in
various facilities and Investigators Turbulence Generation ze K L, t K L, ze
comparison of turbulence (lm) (cm/s) (cm/s)
generation
Present Stirred grid 420 6.7Æ10)4 4.45Æ10)4
Chu and Jirka (1992) Stirred grid 260 7.9Æ10)4 7.3Æ10)4
Atmane and Georage (2002) Microjets
Test 4 366 3.7Æ10)4
Test 5 300 5.8Æ10)4
Woodrow and Duke (2001) Wave tank
Flat surface 581 1.5Æ10)4 3.7Æ10)4
Wavy surface 336 8.8Æ10)4 5.8Æ10)4

348
used an invasive polarographic oxygen microprobe to actual mechanism. The presence of turbulence causes
measure the DO level. The probe was rotated on a hori- constant changes of the boundary layer thickness and also
zontal level to obtain the statistics of the vertical concen- may transport portions of rich oxygen layers into the bulk.
tration distributions. This means that the data points This of course would affect the gas transfer process.
obtained by them were discrete, and not simultaneous.
The advantages of the present measurements over Chu and 6
Jirka (1992) and Atmane and George (2002) are the non- Conclusions
invasive technique and the capability of providing The interaction between gas transfer process with the
instantaneous planar concentration information with a bottom shear induced turbulence was studied experi-
good temporal resolution. mentally in a grid-stirred tank using the non-intrusive LIF
technique. Besides providing high data density of the
5.3 concentration fields near the interface, a qualitative insight
Gas transfer velocity (KL) of the physical mechanism controlling the gas transfer
The gas transfer rate is commonly parameterized by the process is made possible. In particular, a peeling process
transfer velocity KL. of high DO layers near the interface initiated by large scale
turbulent structures could be observed. In some
J ¼ K L ð Cs
Cb Þ ð8Þ sequences, small turbulent structures approaching the
Kolmogorov scale were visible. The mean concentration
where J is the gas flux per unit area; K L is related to the
profile showed an exponential distribution and the fluc-
reaeration coefficient K2 (KL= K2Æ H, H =water depth ) that
tuation profile reached a maximum around the boundary
can be determined using the conservation of mass equa-
layer thickness. The present results show that the inter-
tion
action of large and small structures is present, and it may
dCb play a crucial role in the transfer process. Further work in
¼ K2 ðCs
Cb Þ ð9Þ progress aims at linking this technique with simultaneous
dt
near-surface PIV measurements which allows direct mea-
Integrating Eq. (9) with an initial condition (Cs ) Cb)o surements of the turbulent mass flux.
at time t = to yields
ln ðCs
Cb Þo
ln ðCs
Cb Þ ¼ K2 ðt
to Þ ð10Þ
References
In the literature, K L values are normally reported at a Atmane MA, George J (2002) Gas transfer across a zero-shear surface:
reference temperature of 20 C. A correction factor of a local approach. Geophys Monogr 127, Gas Transfer at Water
1.0241 proposed by Churchill (1961) can be used for the Surfaces, pp 255–259
Brumley BH, Jirka GH (1987) Near-surface turbulence in a grid-stir-
temperature T differences red tank. J Fluid Mech 183:236–263
Brutsaert WH, Jirka GH (1984) Gas transfer at water surfaces.
K2 ðT Þ ¼ K2;20  1:0241ðT
20Þ ð11Þ D. Reidel Publishing Co
Chu CR, Jirka GH (1992) Turbulent gas flux measurements below the
In this study, the increase of oxygen concentration in air-water interface of a grid-stirred tank. Int J Heat Mass Transfer
the bulk region was measured for approximately 1.5 h. 35(8):1957–1968
Based on the above equations, the K L,t value is determined Churchill M (1961) Effect of water temperature on stream reaeration. J
Sanitary Eng Div 5(87):59–71
as the gradient obtained from the graph in Fig. 9 times the Danckwerts PV (1951) Significance of liquid-film coefficients in gas
water depth in the tank which yields a value of absorption. Indust Eng Chem 43:1460–1467
6.7·10)4 cm/s at 20 C. Duke SR, Hanratty TJ (1995) Measurement of the concentration field
The transfer velocity KL,ze estimated based on Eq. (1) resulting from oxygen absorption at a wavy air-water interface.
Air-Water Gas Transfer: 3rd Int Symp on Air-Water Gas Transfer,
using the measured ze is about 4.45·10)4 cm/s. pp 627–635
For comparison, KL values obtained from the present Fortescue GE, Pearson JR (1967) On gas absorption into a turbulent
and other researchers are shown in Table 1. It can be seen liquid. Chem Eng Sci 22:187–216
that when turbulence is present, the transfer velocity cal- Higbie R (1935) The rate of absorption of a pure gas into a still liquid
during short periods of exposure. AIChE Trans 31:365–390
culated using the ‘‘film model’’ (KL,ze ) is usually under- Hopfinger EJ, Toly JA (1976) Spatially decaying turbulence and its
estimated compared to the actually measured (KL,t ) one. relation to mixing across density interfaces. J Fluid Mech
This is expected as the ‘‘film model’’ oversimplifies the 78:155–175
Jähne B, Haussecker H (1998) Air-water gas exchange. Annu Rev
Fluid Mech 30:443–468
Lamont JC, Scott DS (1970) An eddy cell model of mass transfer into
the surface of a turbulent liquid. AIChE J 16:513–519
Lee M (2002) Visualization of oxygen transfer across the air-water
interface using a fluorescence oxygen visualization method. Water
Res 36:2140–2146
Lee Y (1977) Microprobe method for studying gas-liquid oxygen
transfer in various surfactant systems. PhD Thesis, Purdue Uni-
versity
Lewis WK, Whitman W (1924) Principles of gas absorption. Ind Eng
Chem 16
Münsterer T, Jähne B (1998) LIF measurement of concentration
profiles in the aqueous mass boundary layer. Exp Fluids 25:190–
196
Münsterer T, Mayer H, Jähne B (1995) Dual-tracer measurements of
concentration profiles in the aqueous mass boundary layer. Air-
Water Gas Transfer: 3rd Int Symp on Air-Water Gas Transfer, pp
637–648. Application of LIF to investigate gas transfer 21
Schladow S, Lee M, Hrzeler B, Kelly P (2002) Oxygen transfer across
the airwater interface by natural convection in lakes. Limnol
Oceanogr 47(5):1394–1404
Thompson SM, Turner JS (1975) Mixing across an interface due to
turbulence generated by an oscillating grid. J Fluid Mech 67:349–
368
Vaughan WM, Weber G (1970) Oxygen quenching of pyrenebutyric
acid fluorescence in water, a dynamic probe of the microenvi-
ronment. Biochemistry 9:464–473
Wolff LM, Hanratty TJ (1994) Instantaneous concentration profiles of
oxygen accompanying absorption in a stratified flow. Exp Fluids
16:385–392
Wolff LM, Liu ZC, Hanratty TJ (1990) A fluorescence technique to
measure concentration gradients near an interface. Proc 2nd Int 349
Symp on Gas Transfer at Water Surfaces, pp 210–218
Woodrow PT, Duke SR (2001) Laser-induced fluorescence studies of
oxygen transfer across unsheared flat and wavy air-water inter-
faces. Ind Eng Chem Res 40:1985–1995