0

Mass Transfer between Solid

0 and Fluid Streams
0 Mechanism and Eddy Distribution
0 Relationships in Turbulent Flow

C. S. LIN’, R. W. MOULTON, AND G. L. PUTNAM

University of Washington, Seattle, Wash.

A LTHOUGH numerous empirical methods have been developed

for predicting rates of mass and heat transfer between
phases in flowing fluids, a thorough understanding of the
tude of 88,/($)1’3
. \.
; 6, is the thickness of the sublaminar layer
I

The transfer of matter at S8’<ySSd is by turbulence, although

fundamental transfer mechanism is very necessary for the design the momentum is transferred by a viscous mechanism. The
of various kinds of industrial processes. This paper presents a analysis is intriguing, but i t is crude and lacks supporting experi-
theoretical analysis of mass transfer between a turbulent fluid mental evidence. T h e analysis presented in this paper involves
stream and the wall. It gives a method for determining the ratio the concept of introduction into the sublaminar layer of a small
of the turbulent t o molecular mass transfer a t the vicinity of the appropriate amount of eddy which decreases very rapidly to
wall and the concentration distribution from the wall to the main zero a t the wall.
body of the fluid.
-4s shown by Colburn ( 4 ) and Sherwood ( d 4 ) , the processes of Velocity and Eddy Distribution in Straight Tubes
mass transfer and heat transfer are similar, and hence all the-
retical principles treated previously for heat transfer may be re- I n a turbulent fluid, momentum is transferred by both viscosity
garded as applicable t o mass transfer. The analogy between and velocity fluctuation ( 9 ) . T h e shearing stress for two di-
momentum and mass transfer was first deduced b y Reynolds (U), mensional flow may be expressed as
postulating the exact similarity between momentum exchange
and material exchange. Unfortunately, i t is only correct when
the Schmidt group is unity. Prandtl (19) and Taylor ( 2 5 ) later
extended the analogy to include a laminar layer near the wall where u’and v’ are fluctuation velocities in the direction of and
for mass transfer with all values for Schmidt group. More re- perpendicular t o the main flow. A bar over the product denotes
cently von KBrm&n (IZ), Boelter, Martinelli, and Jonassen ( d ) , the time average of the product. The fluctuation velocities can
and Reichardt ( $ 0 ) analyzed the problem more extensively on the be correlated in terms of mixing length and eddy viscosity as
basis of the velocity distribution measurements in straight tubes.
The fluid is divided into three layers-namely, the very thin
sublaminar layer adjacent t o the wall, the turbulent core, and the
buffer region between the core and laminar layer. I n the sub- where 7is the mean square root of v’. The mixing length repre-
laminar layer, in which there is no turbulence or eddy diffusion, all sents the mean free path of the turbulent pulsation (9) and was
material is transferred by molecular motion. The velocity distri- also explained by Sherwood (3.3). The experimentally measured
bution and mass transfer in the turbulent core are controlled by velocity distribution determines the value of eddy viscosity, e,
eddy diffusion of momentum and eddy diffusion of mass. I n a t any point in a tube. The velocity distribution was Correlated
buffer region, the combined action of molecular and eddy dif- U YU* P
by Nikuradse (18) with the parameter u+ = - against y+ = ~

fusion determines the velocity distribution and mass transfer. U* c1

Recent theories were found satisfactory for predicting the mass
transfer rate in turbulent gas streams. For turbulent liquid
streams (14, 16), where the molecular diffusion is very slow in
where u* = 4;
comparison with t h a t in gases, the theories cannot be adequately
applied. and 70 =fi pu2 (4)
The concept of the existence of a sublaminar layer for molecu-
The follswing equations express the velocity distribution in
lar diffusion may not be true. Ultramicroscopic observations
the three fluid layers in von K&rm&n’sanalysis ( I $ ) :
of fluid particles adjacent t o the wall by Fage and Townend ( 8 )
revealed that the velocity component fluctuations vertical to the Sublaminar layer (0 5 y+ 5 5) u+ = y+
wall do not cease t o exist until they reach the .cvall. Unfor- Buffer region (5 5 y+ 5 30) u+ = -3.05 + 5.00 In y+
tunately, the slight amount of eddies close to the wall can be Turbulent core (y+ 2 30) +
u+ = 5.5 2.5 In y+
neither correlated quantitatively from these observations nor The amount of 2ddy viscosity introduced a t O<y+ 55 in the
detected appreciably by the velocity distribution measurements. present analysis is assumed as
However, when the diffusivities of transferring materials are low,
the material transferred by small eddies becomes appreciable.
By analogy with the laminar diffusion boundary layer, Levich
t = (E5Y (5)

( I S ) proposed t h a t in a turbulent boundary layer the laminar It will be shown later t h a t the assumed relation is appropriate.
diffusion layer aa‘ (called diffusion sublayer) would be the magni- Combining Equations 1and 2 gives

1 Present address, Department of Chemical Engineering, Massachusetts

tnstitute of Technology, Cambridge, Mass.
636
 March 1953 INDUSTRIAL AND ENGINEERING CHEMISTRY 637

Equations 5 and 11 represent the eddy

distribution from the boundary of the
turbulent core t o the wall. It might be
eo
possible t o find a single, function t o de-
scribe the eddy and velocity distribu-
tions from the boundary of the turbu-
lent core t o the wall. However, in
I5 order t o avoid the integration and dif-
ferentiation of a complicated single func-
U* tion, two separate, simple’functions will
satisfy the conditions as well.
10

Mass Transfer and Concentration

x
Profiles
5 I n a turbulent fluid, material is trans-
ferred by both molecular diffusion and
concentration fluctuation (9). T h e rate
of transfer for a two-dimensional flow in
0
I 5 20 100 1000
a steady state can be expressed as
Y+
Figure 1. Generalized Velocity Distribution for Turbulent Flow of Fluid in Pipes

Near the wall, the shearing stress, T, may be assumed t o be con- where C’ is the concentration fluctuation. A bar over the prod-
stant and equal t o t h a t a t the wall, TO. Setting r = TO, and in- uct denotes the time average of the product. T h e concentration
tegrating Equation 6 with the substitution of Equations 3 and 5 : fluctuation may be correlated in terms of mixing length and eddy
diff usivity as
(7)

On carrying out the integration by replacing dy with dy, and Combining Equations 12 and 13, there is obtained
evaluating C so that u+ = 0 when y+ = 0, which is the boundary
condition, from Equation 7 : dC
N = (D + E ) -d y

Comparison of Equations 2 and 13 gives

E=€
since the mixing lengths for eddy diffusion of momentum and for
eddy diffusion of mass are equal at identical flow conditions (9).

Equation 8 is plotted in Figure 1, and the region Oy<+<5 is

designated as the wall layer. There is only a slight change of
the curve from the curve u+ = y+ . The concept of the existence
of a pure laminar layer, however, is completely eliminated. In
the turbulent core, von K&rmBn’s relationship is retained. I n
order to satisfy the velocity and eddy conditions a t y+ = 5 ,
the velocity distribution at the buffer layer is expressed as

or
u+ = -3.27 + 5 In (y+ + 0.205) (9)

J.
u+ = 4.77 + 5 In (9 + 0.041) (sa)

The velocity distribution data of Nikuradse (18), Reichardt (%I),

and Rothfus, Monrad, and Seneca1 (23)are shown in Figure 1.
Differentiation of Equation 9a gives 4Y
Figure 2. Comparison of M a s s Transfer Data in Turbulent
5, PU$ Liquid Stream with Equation 17
(10)

The mass transfer equation can be deduced from Equations 6

The shearing stress in the buffer region is also assumed t o be con-
and 14. I n the wall layer and buffer region, Equation 14 is
stant and equal t o r0. Combining Equations 3, 6, and 10, and
integrated with the use of eddy distribution expressions as indi-
eliminating d u l d y :
cated by Equations 5 and 11; the shearing stress and mass flux
are assumed t o be constant. I n the turbulent region, the varia-
tion of shearing stress and mass flux are assumed similar t o the
638 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 45, No. 3

extent that 7 / N is constant, and p and D negligible in comparison particular annular space as a circular tube. The mass transfer
with pc and E. When the mass transfer coefficient is defined as coefficient data in wetted wall columns by Chilton and Colburn
(31, Barnet and Kobe ( I ) , Jackson and Ceaglske (IO), and
N
k, = -
c. -cw. (16) Johnstone and Pigford ( 1 1 ) are shown in Figure 3. The
Schmidt group varies from 0.54 t o 1.86.
the following result is obtained with the assumption that E = a: If the equality of thermal eddy diffusion, EIl, and eddy vis-
cosity, E, is assumed, the heat transfer equation obtained with the
‘“-“+D = -
U
f2 (17) procedure described is
h
where C,CT+H = 2
f (18)

+D = 1 + 4;
f [ *
1 4j 5 (pD)s/aF ($1 + 5 In
1 + 5.64 E-
pD
6.64 1 + 0.041 E
( PD)
- 4.77

1
where +H is the same function as +D in Equation
17 with E replaced by %*!
PD k
The heat transfer coefficients of Eagle and
and Ferguson (7) in water streams, corrected for isothermal condi-
tions. are shown in Figure 4. The heat transfer data of Colburn

F .E = -1l n
(PD) 2 1
[1 + &
- -E(’)”’
5
+
(3
(&)2($)
PjS +
and Coghlan for nitrogen and hydrogen gas mixtures (6) are alao
given in Figure 4. Xumerous data for. the heat transfer of oil
CPP
14.5 pD with high 7 in literature were not used, since the viscosity of
lo( E )‘/a -1 the liquid changes very appreciably with the temperature. In
.\/$ tar,-‘ 14.5 PD T& general, the theory is in good agreement with experimental re-
43 +6 sults both for mass and heat transfer, over wide ranges of the
Schmidt and Prandtl groups. The analysis approaches von
A graphic representation of F K&rm&n’sanalysis for the turbulent mass transfer in gas streams
-i.e., Schmidt group in the vicinity of unity. I n other words.
the small amount of the eddy diffusion by turbulence in this case
becomes negligible in comparison with the molecular diffusion in
the wall layer.
The eddies close to the wall (O<y+ 5 5 ) are so small that it is
impossible t o obtain any information from the velocity profile
measurements as in the region 5y<y+ 5 3 3 . The assumed eddy
distribution relationship in Equation 5 is arbitrary and was
found on the basis of simplicity and the best agreement of the
over-all mass transfer coefficient equation with expwimental data

-
Figure 3. Comparison of Mass Transfer Data in
Turbulent Gas Stream with Equation 17

Equation 17 is plotted in Figures 2 and 3. Experimental mass

transfer coefficient data for the dissolution rates of cast solid tubes
in the water stream by Linton and Shermood (16), and for the IO
limiting current densities of diffusion-controlled electrode re- 3x10’ IO‘ IO‘ IO

actions by Lin, Denton, Gaskill, and Putnam ( I C ) , in a n annular P

&
*L

cell are shown in Figure 2. The Schmidt group varies from 320 Figure 4. Comparison of Heat Transfer
to 3200. The agreement of experimental values with the theo- Data in Turbulent Stream with Equation 18

t o that in a circular pipe. It seems

 March 1953 INDUSTRIAL AND ENGINEERING CHEMISTRY 639
where the turbulent core is included In order to avoid the complica-
tion of the over-all mass transfer equation, the authors have ex-
cluded the detailed analysis in the turbulent core without losing
the benefit of simple assumptions. If the investigator is in-

- (%)’’’(&) (s) (m)

Ifa Y+ terested in the concentration profiles for lower Schmidt groups
than those indicated in Figures 5 and 6, they can be expressed
without difficulty by adding Boelter, Martinelli, and Jonasson’s
analysis ( d ) for the turbulent core to the present analysis.
rpj
4 2 tan-’ <3 +6 Conclusions
I n the buffer region A small amount of eddy in the sublaminar layer has a marked
influence on the mass transfer rate for

(-$)*’’F ($) +51n

1f -$ (v - 0.959) any matserial, the molecular diffusion of
which is very low in comparison with

c - c,
-=
1 + 0.041 !!-
PD (20)
the kinematic viscosity of the fluid. T h e
general mass transfer equation (Equa-
c*v. - c. + 17), derived on the basis of the new

t
1 5.64 1- tion
4+{y($)2‘3t7 pD - 4.77 eddy distribution relationship intro-
duced at the vicinity of the wall, agrees
well with Dublished exnerimental data
available in- literature with the Schmidt
Derivations of groups varying from 0.54 t o 1.86 (in the gas streams) t o
Equations 17, 320 t o 3200 (in the liquid streams). It also agrees with heat
18, 19, and 20 transfer data corrected for isothermal conditions. The validity
are available of this eddy relation, however, is inseparably connected with the
(16).At moder- assumption that the eddy mass diffusion is equal to the eddy
ate flow rates viscosity. ABindicated by concentration profiles in Figures 5 and
and high 6, there is no definite laminar film for pure molecular diffusion
Schmidt num- near the wall. The laminar type of diffdsion film becomes thin-
bers, the value ner when the molecular diffusivity decreases.
d;may benep I n this analysis, the shearing stress and the mass flux are as-
sumed to be constant near the wall. However, marked deviation
lecied in com- could be expected from the actual behavior at low flow rates where
parison with the the buffer region is relatively thick, although the error will be
valuein thelarge greatly diminished as the flow rate increases. It might be pos-
bracket in the sible to apply the Reichardt’s method (10)for correction of the
denominator of shearing stress variation near the wall, but the final mass transfer
E q u a t i o n s 19 equation will be much more complex.
and 20. Then
-
CW. - c.
“ canbe
e x p r e s s e d di-
rectly as a func-
tion of y+. A
plot of
c - c,
c
LIV.- c a Fi ure 5. Generalized Concentration Dis- 06
versus y+ is tritution for Mars Transfer in Turbulent
given in Figure Stream at Moderate Flow Rate
5 with the use !j*OISTANCE FROM WALL
of t h e E q u a - 02
dmDIAMETER OF PIPE
tions 19 and 20. I
A definite laminar diffusion film does not exist near the wall. o 0o v 5 10 IS EO 25 30 35 40 45 E
The zone of linear concentration gradient becomes narrower and 3x103

closer to the wall at higher Schmidt numbers. I n other words,

. the pure molecular diffusion region starts a t a position where the
magnitude of eddy diffusion is considerably less than the magni-
Figure 6. Concentration Distribution for M a r s Transfer
in Turbulenf Stream

tude of molecular diffusion. The laminar diffusion layer, there- Reynolds number = 10,000
fore, becomes narrower as the molecular diffusion value decreases.
A plot of concentration gradients for various Schmidt groups
a t Reynolds number = 10,000 is shown in Figure 6. Cmax. The behavior of mass transfer at the distance far away from
here replaces C,,., since a t high Schmidt groups CaV.approaches the wall, usually less important, has only been briefly considered.
Cmax.. The concentration gradient for the low Schmidt group Evidently, the concept of different fluid layers is not correct
was talcen from the analysis of Boelter et al. ( 2 ) who include the and can be eliminated. The use of the fluid layer concept, how-
ratio of eddy t o molecular diffusivity in the turbulent core. As ever, has been helpful in simplifying the analysis. Recently,
mentioned above, the analysis presented in this paper approaches Rannie i n “Heat Transfer Studies Relating t o Rocket Power-
the von KArmBn’s analysis (1%) a t low Schmidt group. I n Plant Development’’ (6) has developed a velocity distribution re-
other words, with high diffusivity value, the wall layer becomes a lation which applies both t o the region previously called the sub-
laminar diffusion layer. The concentration gradient in the inter- laminar layer and t o the buffer region, and furthermore joins the
mediate Schmidt groups can also be expressed if the analysis of logarithmic distribution in the turbulent core. The ratio of
640 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 45, No. 3
eddy viscosity to kinematic viscosity close t o the wall derived Fage, A , , and Townend, H. C. H., Ibid., A135, 656 (1932).
from Rannie’s expression varies with the second power of the dis- Goldstein, S.,“Modern Development in Fluid Mechanics,”
London, Oxford University Press, 1938.
tance parameter, y+. Rannie’s expression has undoubtedly the Jackson, M. L., and Ceaglske, N. H., IND. ENG.CHEM., 42, 1188
advantage, but the mass transfer equation derived with the use (1950).
of Rannie’s eddy distribution relationship deviates markedly Johnstone, H. F., and Pigford, R. L., Trans. A m . Inst. Chenz.
from the experimental data in liquid streams where the Schmidt Engrs., 38, 25 (1942).
KLrmLn, T . yon, Trans. A m . SOC.Mech. Engrs., 61, 705 (1935).
group is extremely high. Levich, B., Acta Physicochim. U.R.S.S., 17, 256 (1942).
Lin, C. S.,Denton, E. B., Gaskill, H. L., and Putnam, G. L.,
Acknowledgment IND. ENG.CHEM., 43, 2136 (1951).
Lin, C. S., Moulton, R. W., and Putnam, G. L., deposited with
The authors are indebted to B. H. Sage and W. G. Schlinger for the American Documentation Institute, Washington 25,
numerous suggestions and criticisms of the manuscript. D. C., Doc. 32845 (1952).
Linton, W. H., Jr., and Sherwood, T. K., Chem. Eng. Progr., 46,
258 (1950).
References Martinelli, R. C., Trans. Am. SOC.Mech. Engrs., 69, 547 (1947).
Xikuradse, J., VDI-Forschungsheft. H356 (1932).
(1) Barnet, W. I., and Kobe, K. A., IND.ENG.CHEM., 33,436 (1941).
( 2 ) Boelter, L. M. K., Wfartinelli, R. C., and Jonassen, F., Trans.
Prandtl, L., P h y s i k . Z . , 29, 487 (1928).
Am. Soc. Mech. E ? L Q ~63,
s . , 447 (1941). Reichardt, R., Natl. Advisory Comm. Aeronaut., Tech. Mem.
(3) Chilton, T . H., and Colburn, A. P., ISD.ENG.CHEM.,26, 1183 1047 (1943).
(1934). Reynolds, O., Proc. Manchester Literary and Phzlosophical Soc ,
(4) Colburn, A. P., Ibid., 22, 967 (1930). 8 (1874).
(5) Colburn, A. P., and Coghian, C. A., Trans. Am. Soc. X e c h . Rothfus, R. R., Moniad, C. C., and $enecal, V. E.. ISD. ETG.
Engrs., 63, 561 (1941). CHEX, 42, 2511 (1950).
( 6 ) Dunn, L. G., Powell, TT. B., an’d Seifert, H. S., “Heat Transfer SherTTood, T. K.. IND.ENC.CHEM.,42, 2077 (1950).
Studies Relating t o Rocket Po%-er-Plant Development,” Sherwood, T. K., Trans. Am. Inst. Chem. Engrs., 36, 817 (1940).
Third Angl0-~4mericanAeronautical Conference, published Taylor, G. I., Great Britzin Advisory Comm. Aeronaut., R e p t .
by RoyaI Aeronautical Society, England, 1951. Mem. 2272 (1916-17).
(7) Eagle, A., and Ferguson, R. M., Pioc. Roy. Soc. London, A127,
540 (1930). RECEIVED
for review July 7 , 1051. ACCEPTEDOctober 23, 1992.

(Mass Transfer between Solid Wall and Fluid Streams)

Interferometric Measurements of Concentration Profiles in

Turbulent and Streamline HOW

A LTHOUGH the rate of mass or heat transfer between solid

walls and turbulent fluid streams has been the subject of
considerable study, the fundamental mechanism of the trans-
will reveal the nature of these small eddies. Since in the liquid
stream the molecular diffusion of material is very low in compari-
son with the molecular diffusion of momentum, the quantitative
ferring process in the vicinity of the wall is still inconclusive. effect of these small eddies becomes appreciable and can be de-
Progress made in the development of the analogies between mass tected in the concentration distribution measurements.
or heat and momentum transfer by employing concepts of laminar For streamline flow in a tube, the material is transferred by
films existing adjacent t o the wall has not been proved because of molecular diffusion with parabolic velocity distribution (10, 15)
lack of experimental evidence ( 7 ) ,and because of failure to explain and hence the theoretical calculation of the concentration dis-
the experimental results of mass transfer rates in turbulent liquid tribution is possible (5). For a flat plate, mass transfer coef-
streams (10, 13) where the diffusivities of materials are very low ficients and concentration profiles can be predicted with the use
in comparison with those in gaseous streams. Instead of using of boundary layer concepts (1).
the sublaminar layer for pure molecular diffusion, the authors, The purposes of this work were to measure quantitatively the
in the preceding paper ( 1 1 ) , have developed a new eddy distri- concentration distributions near the surfaces of concentration-
bution relationship in the vicinity of the wall. The equation polarized electrodes in turbulent fluid streams and t o verify the
derived for calculating the mass transfer coefficients in turbulent eddy distribution relationship introduced by the authors for the
fluid streams agrees well with the experimental results of mass analogy between mass transfer and momentum transfer ( 1 1 )
transfer rates covering the whole range of Schmidt groups avail- Since the electrochemical reaction on the so-called concentration-
able in the literature. polarized electrode is diffusion-controlled, the rate of deposition
hIeasurements of mass and thermal eddy diffusion in turbulent or reactions of materials or ions depends on the rate of supply of
fluid streams have been contributed by several investigators materials or ions from the main body of the fluid to the surface
(14, 16), but in order to reveal the existence of the small eddies by diffusion and convection (10). The concentration distribu-
adjacent to the wall, accurate experimental concentration or tions in streamline flow were also measured, in order to demon-
temperature distribution data are very desirable. Moreover, strate the laminar diffusion boundary layer on a flat surface.
the eddy diffusion of momentum near the wall is extremely low in
comparison with the molecular diffusion of momentum, and i t Experimental
seems impossible to obtain any quantitative information about
these small eddies from the velocity profile measurements. Tem- The concentration distribution near the surface of the electrode
perature or concentration distribution measurements encounter was measured by means of light interference. Local changes of
similar difficulties when the molecular diffusion of heat or material concentration in the fluid, with resultant changes in the refractive
has the same order of magnitude as the molecular diffusion of index, cause the displacement of interference fringes. The
momentum-i.e., the Prandtl or Schmidt group is near 1 or less measurement of the displacement of the local fringes gives an
than 1. It is believed, however, that measurements of con- accurate quantitative value of local concentration a t any point in
centration profiles in a liquid stream very close to the a a l l the fluid. The optical method is superior t o the ordinary eani-