thousand gal of acid. Passing the raffinate through a 0.

5-hr
holdup tank reduced this loss to less than 0.10 gal/thousand
gal of acid.
Chemi cal Reagent Costs. Total reagent costs for the
process, based on the continuous bench-scale tests, are con-
servatively estimated to be less than $1.00 per lb of Us08
(Table 111). Since the major cost is that of the relatively ex-
pensive solvent, the economic success of the process depends
on keeping solvent losses at a tolerable level. Although solvent
loss by entrainment in the aqueous raffinate in tests i n the
laboratory coiitinuous unit was low (less than 0.01% of the
wet-process acid volume after the deentrainment tank) , it
will be necessary to demonstrate that a significant loss of
solvent does not occur during cleanout of the solid material
that accumulates in the settlers at the aqueouq organic inter-
face. These losses can be more accurately determined in a
small pilot plant located at a plant site where the process can
be ruii continuously with fresh acid feed.
l i terature Cited
Baes, C. F., J r., J . Phys. Chem., 60, 805 (1936).
Blake, C. 9., J r., Bam, C. F., J r., Brown, K. B., I nd. Eng. Chem.,
50, 1763-7 (1958).
Blake, C. A., J r., Homer, D. E., Schmitt, J . M., U.S. At. Energy
Cronan, C. S., Chenz. Eng. , 66 (91, 108-11 (1959).
Ferguson, D. E., US. At. Energy Comm., Rept. ORNL-4272,
Comm., Rept. ORNL-2259, 1959.
1968.
Ferguson, D. E., ibid., ORNL-4422, 1969.
Ferguson, D. E., ibid., ORNL-4572, 1970.
Greek, B. F., Allen, 0. W., Tynan, D. E., I nd. Eng. Chem., 49,
Hurst, F. J ., Crouse, D. J ., U. S. At. Energy Comm., Rept.
Hurst, F. J ., Crouse, D. J ., Brown, K. B., ibid. , ORNL-TM-
628-38 (1957).
ORNL-2952, 1960.
2.522. 1969.
I - - -
Kennedy, R. H., in "Processing of Low-Grade Uranium Ores,"
pp 220-6, International Atomic Energy Agency, Vienna,
Austria, 1967.
Latimer, W. M., "Oxidation Potentials," Prentice-Hall, New
York, N.Y., p 44.
Long, R. S., Ellis, D. A., Bailes, R. H., Proc. Znt. Conf . Peaceful
Uses At. Energy, Geneva, Switzerland, 8 , 77-80 (1955).
RECEI VED for review March 29, 1971
ACCEPTED J uly 19, 1971
Presented at the Division of Nuclear Chemistry and Technology,
161st Meeting, ACS, Los Angeles, Calif., April 1971. Research
sponsored by the U. S. Atomic Energy Commission under con-
tract with the Union Carbide Corp.
Activity Coefficients of Strong Electrolytes in
Aqueous Solutions
Herman P. Mei ssnerl and Jefferson W. Tester
Deparfment of Chemical Engineering, Siassachusetts Institute of Technology, Cambri dge, Mass. 02139
The val ue of yi , the mean activity coefficient of an el ectrol yte in aqueous solution, vari es with the ionic
strength p. If the charges on the anions and cations present are, respectively, L- and L+, then the term
yi i s r eadi l y cal cul ated. When this term yTtl/'+'- is pl ot t ed agai nst p f or various strong electrolytes
at 25"C, a curve fami l y forms such that, gi ven a single experi mental val ue of y* at a known ionic strength
for an el ectrol yte, its val ue of y+at any other ionic strength can b e approxi mat ed. For use, when no such
experi mental point is avai l abl e, a method appl i cabl e to a number of electrolytes i s described f or pr e-
dicting T~at an ionic strength of 2. By this empirical procedure, approxi mat e values of y+ can b e predi ct ed
at any concentration f or many strong electrolytes at 25' C.
l / Z , Z -
M u c h of t'lie experiment'al data available for Ti , the
activity coefficients of strong electrolytes at various concen-
trations in aqueous solutions, is well summarized in the stan-
dard references (Harned and Owen, 1958; Robinson and
Stokes, 1965). The Debye-Hilckel equat'ion applies to t,hese
data below about 0.01m, while the modified equations of
Scatchard and others, involving empirical constants to be de-
termined for each electrolyte, apply at higher concentrations.
Pitzer and Brewer (1961) have presented a procedure for in-
terpolating yL values for a number of electrolytes.
These activity coefficient data in general are frequently pre-
sented as in Figure 1, showing curves of y* at 25OC for several
1 To whom correspondence should be addressed
arbitrarily selected electrolytes plotted against concentrations
from 0. to 2.0m. Inspect,ion indicates that the change of yh
with molality differs widely from one electrolyte to the next.
The object here is to present a simple graphical method for
representing the act,ivity coefficients of strong electrolytes in
aqueous solution, applicable from low concentrations to
saturated solutions. To use this method, a value of y i for the
electrolyte in question must be known at a concentration well
above those to which t,he Debye-Hiickel equation applies.
When such an experimental value is not available, a procedure
for predicting yi at a known concentration is presented. This
procedure appears applicable to chlorides. bromides. iodides,
acetates, nitrates, chlorates, and hydroxides. In combina-
tion, these relat,ionshjps make it possible to predict yh values
128 Ind. Eng. Chem. Process Des. Develop., Vol. 1 1, No. 1 , 1972
4 KOH 1!5
I .o 2 .o
m
Figure 1. The effect of concentration on the mean activity
coefficient yi for aqueous solutions ut 25C
0 IO. 20.
)I
Figure 2. The effect of ionic strength p on r at 25C (line
identification numbers correspond to those of Figure 1)
at any concentration for a number of pure electrolytes in
aqueous solution at 25C.
Acti vi ty Coefficients vs. Ionic Strength
I n the correlations pr'oposed, extensive use is made of the
term I?, defined as follows:
Here z+ and z- are the charges on the cations and anions of
the electrolyte i n question. Concentrations are expressed as p ,
the ionic strength, defined as 0.5m ( Y + Z + ~ +v - z- ' ) where Y+
and v- are the moles of cations and anions, respectively,
formed upon complete clissociation of 1 mole of electrolyte,
while m is the electrolyte molality i n g-mol/1OOO g HzO.
Thus, in an m molal solution of AlC13, z+ is 3, z- is 1, Y+ is 1,
Y- is 3, and p is 0.5m (1 ><3* +3 X 12) or 6m.
c O ' I 0.7
c
XHCl
OKCI
OCaC12
0 LizS04
Figure 3. The effect of ionic strength p on r at 25C in the
low concentration region with the Debye-Huckel behavior
shown
The data of Figure 1 are replotted on Figure 2, usiag the
coordinates log r vs. p , A wider range of p values is covered in
Figure 2 t,han in Figure 1, while additional data (Harned and
Owen, 1958; Robinson and St'okes, 1965; Rloore, 1962;
Pitzer and Brewer, 1961) at' relatively low concentrations are
presented in Figure 3. I n contrast to the divergence and curve
crossing of Figure 1, the curves on Figures 2 and 3 fall into a
single family having tmhese charact'eristics:
All curves converge at low ionic strengths and terminate at
a value for r of unity when p is zero. Also plotted on Figure 3
is the dotted liiie representing the Deybe-Huckel limiting law
(Debye and Huckel, 1923), written in terms of r as follows:
log r = -0.510 p l / z (2)
As expected, all curves converge on the deb ye-Kickel line as
p approaches zero.
With few exceptions, the curves of Figure 2 usually do not
C~ORS, each remaining bet'ween its neighboring curves regard-
less of the electrolyte involved. I ncluded here are the 1:l
electrolytes like NaC1, 1 : 2 electrolytes like Na2S04, 2 : 1
electrolytes like CaC12, 2 : 2 types like CuSO4, 3: 1 types like
AICI,, and 3:2 types like Al,(SOI)3. Locating a single point on
Figure 2, therefore, serves to locate the entire curve. Earlier
plots using the coordinates of Figure 2, such as the one of
Guggenheim (1933, are restricted to ionic strengths below
unitmy, making it less evident that a curve family exists.
Inspection of Figure 3 shows t.hat at any p value below
about 0.2, the term F is similar enough for all electrolytes so
that it cannot with corifideiice be assigned to a single curve.
On the other hand, these curves have diverged sufficiently at
a p value of unity or greater, so that the parent curve is easily
selected. Knowing yrt at' a p value of unity or greater, there-
fore, makes i t possible to 1ocat)e the entire curve within the
family of Figure 2 or 3. For example, assume that y i is to
be predicted for RlgClz iii aqueous solution at a p of 9, know-
ing that when p is 2, yh is 0.5, corresponding to a r of 0.707
by Equat'ioii 1. When we locate this known point on Figure
2 and extrapolate to a p of 9, r is 1.45, corresponding to a yrt
value of 2.09 which is wit'hiii 10% of the experimental value
for yrt of 2.32.
I n view of the consistency of the curves, it also seems reason-
able to assume that Figure 2 applies to supersaturated solu-
tions. Thus, knori ng yk to be 0.269 for KKOs at a molality of
3, then by extrapolation on Figure 2, yk is 0.126 at a p of 8,
Ind. Eng. Chem. Process Des. Develop., Vol. 11, No. 1, 1972 129
+2.0
+1.6
+1.2
+0.8
2 +0.4
0
-
0
- 0.4
- 0.8
-1.2
0 4.0 80 120 160 20.0
P
Figure 4. Generalized plot of r as a function of ionic
strength p for use at 25C and over an ionic strength
range of 2.0 to 20.0
a concentration well beyond the 3.7 molality of a saturated
K x03 solution.
For simplicity, curves for only 21 electrolytes are shown in
Figures 1 and 2. However, the generalizations just presented
apply to most of the 120 electrolytes for which experimental
data of yk vs. p are summarized in the standard references
(Harned aiid Omen, 1958; Robinson and Stokes, 1965; Wu
and Hamer, 1969). Not surprisingly, the notable exceptions
are sulfuric acid, sodium and potassium hydroxide, thorium
nitrate, and most of the cadmium and zinc salts, whose lines
would intersect neighboring lines on Figure 2. I n addition,
lithium chloride, magnesium iodide, uranyl fluoride, lithium
nitrate, and berylliuni sulfate show some curve crossing.
Worki ng Charts of l o g r vs. p
While y+ values can be determined by extrapolation on
Figure 2, as illustrated above, it is more convenient to use
Figures 4 and 5 as working plots for this purpose. The lines
on these working plots were located by interpolation of the
data for the over 110 selected electrolytes. Figure 4 is best
utilized for ionic strengths above 2, and Figure 5 for ionic
strengths ranging from 0.1 to 2. At p values below 0.1, the
curves have converged so that the experimental data on
Figure 3 apply with sufficient accuracy. The curves on Figures
4 aiid 5 have been chosen at intervals which appear convenient
for interpolation, and have been smoothed to conform to what
appears to be the general family configuration. Such smooth-
ing recognizes that for any two neighboring curves I and I1
drawn through experimental data on Figure 2, the term log
rI /log rII is reasonably linear in p, at least ovei shorter
ranges of concentration. This behavior is consistent with
Scatchards observations (Scatchard, 1936; Pitzer and Brewer,
1961) that the term log yr/log Y I I is linear in m for two 1: 1
electrolytes, for two 2: 1 electrolytes, and so forth.
Closer scrutiny of Figure 1 shows that it contains several
curve families, one for the 1 : 1 salts, another for the 1 : 2 curves,
another for the 2: 1 curves, another for the 1 : 3 salts, and so
forth. The curve crossover which occurs on Figure 1 because of
the dissimilar pattern of the various curve families is largely
+O.IO
0
-0.10
c
,g -0.20
0
-
-0.30
-0.40
-0.50
0 0.50 I .oo 1.50 2.00
F
Figure 5. Generalized plot of r as a function of ionic
strength p for use at 25C and over an ionic strength range
of 0.1 to 2.0
eliminated by replotting as in Figures 2 to 5. Curve extrapola-
tion within each family on Figure 1 could of course be carried
out, and would have the advantage of yielding yh directly,
rather than r, but would lose the guidance afforded on Figures
2 to 5 by the neighboring curves for the other electrolyte
types.
Preci si on
Since the curves of Figures 4 and 5 diverge as p increases,
extrapolation from a higher to a lower p value is always more
successful than the reverse. The success attained by extrap-
olation also depends on the range of p involved, and on the
electrolyte selected for study. Thus, extrapolation should not
be undertaken for those electrolytes showing excessive devia-
tions from the curves of Figures 4 and 5, namely the chlorides,
bromides, and iodides of zinc and cadmium as well as sulfuric
acid and thorium nitrate. Excluding these materials, 59 elec-
trolytes were found for which experimental yf values were
reported (Harned and Owen, 1958; Robinson and Stokes, 1965;
Wu and Hamer, 1969) up to a p of at least 9. Figure 4 was used
to extrapolate the experimental y+ value at a p of 2 to a p of 9.
The average deviation of the predicted from the experimental
y* value at a p of 9 was 18%. Several of the 59 electrolytes
included in this calculation showed considerable curve cross-
ing, with differences between the experimental and predjcJ ed
values for yi in excess of 25%, as listed: LiN03 (74%), LiCl
(3i%), CaCl? (%%I, Mg(C104)~(54%), MgIz (46%), BeS04
(41%), NaOH (42%), CuCh (30%), Mg(Ac)? (32761, Ba(Ach
(39%)) and UOZ(XO& (25.5%).
Extrapolation over a range of p values smaller than from 2
to 9 would, of course, have resulted in smaller errors. Further
consideration of these data shows greater success in prediction
with the electrolytes of lower ionic charge. Thus, for the 59
electrolytes considered in the above test, the 43 1: 1, 1 : 2, and
2: 1 types showed an average deviation of 16.5% while the 15
3: 1, 2: 1, and 3:2 types showed a deviation of 22%.
Deviations of the experimental r curves of electrolytes hav-
ing a (z+z-) product of 3 or greater from the generalized
curves limit the success attained in extrapolating yh values.
These deviations are not large, as is evident from inspection of
130 Ind. Eng. Chem. Process Des. Develop., Vol. 1 1, No. 1, 1972
\ I -*- - I I I
lo 20
t
Figure 6. The effect of ionic strength p on I at 25C f or
19 2 : 2, 3 : 1, and 3 : i! electrolytes. Lines from Figure 4 ar e
repl ot t ed as dot t ed lines
, 0.8-
>
+
>
X
0.6 -
L=
0.4 -
0.2
-
c
u-
OO 0.2 0.4 0.6 0.8 I .o
r
wv.cAV-
Figure 7. Cross plots of IMv+Xv- vs. r Mu+c l u- [ X - = I-
and NO3-(ll)] at 25 C and an ionic strength of 2.0
Figure 6, where experimental values of r for 2 : 2, 3 : 2, and 3 : 1
electrolytes are plotted on an expanded ordinate of r, along
with dotted lines taken directly from Figure 4. Errors in pre-
dicting y& caused by these deviations are magnified by the
( z+z- ) product in the exponent of Equation 5. Thus, when
extrapolating for a 2: 2 electrolyte, the term ( z+z- ) equals 4,
and consequently, an error of 1Oyo in r from Figures 4 and 5
results in a 40% error in y5. This error, which obviously be-
conies larger as the ( z+z- ) product increases, is reduced by
UK of the expanded ordiiiate scale of Figure 6, and so im-
proved success can be attained by replacing Figure 4 by Figure
MutClu-
AgCl
BaCle
BeCl,
CaClz
CdClz
CeC13
CoClz
CrC13
CsCl
CuClz
EuC13
FeC12
HC1
KC1
LiCl
AlC13
Table 1. Valuer of IM,+cl,- at p
Equals 2
0. 550 MgC12
0. 673 MnClz
0. 630 NaC1
rMyLCly- Mvt Cl v-
0. 710 NdC13
0. 676 NHiCl
0. 639 PbClz
0. 292 NiCl2
0. 690 PrC13
0. 666 RbCl
0. 496 scc13
0. 641 SrClz
0. 646 ThC14
0. 679 U02Clz
1. 009 Yc13
0. 573 ZnCls
0. 921
rM,+Cl,-
0. 707
0. 670
0. 668
0. 639
0. 570
0. 691
0. 599
0. 638
0. 546
0. 657
0. 660
0. 717
0. 733
0. 646
0. 612
6. For example, starting with the experimental value of r at a
p of 2 for MgS04, r can be predicted at a H of 9, using either
Figure 4 or Figure 6. The difference between this predicted r
and the experimental value is 29% with Figure 4 but only 2%
with Figure 6, which is approximately the accuracy of reading
the chart itself.
Estimating Y& Values
As just pointed out, Figures 4 and 5 can be used to predics
at various ionic strengths only when a known value of T~ it
available for the electrolyte in question at some given concen-
tration well outside the Debye-Huckel region. When such an
experimental reference value is lacking, the term log r, and
consequently yi itself, can often be predicted by the procedure
illustrated in Figure 7. This figure applies always at 25OC at a
value for of 2, with the ordinate referring to the iodide or the
nitrate of a particular cation, and the abscissa to the chloride
of the same cation. Each of the 13 iodide points plotted, there-
fore, involves a different single cation, with the lowest point
representing cadmium, the next lowest point cesium, the
third rubidium, and so forth. The points on Figure 7 can be
identified by their abscissa values-i.e., r for Mv+C1,- from
Table I. Inspection shows that the iodide points on Figure 7
not only fall on a single line, but that this line is straight.
Consequently, from Figure 7, when we know r for the chloride
of a given cation at an ionic strength of 2, the value of r for
the iodide of this cation (at an ionic strength of 2) can pre-
sumably be obtained. Similarly, when we know r for the
iodide, the r value for the chloride of the same cation can be
obtained when p is 2.
The same type of correlation just described, relating r
values for the iodides and chlorides, exists between other
anions and the chlorides. Thus when I values for bromides are
plotted against F values of the chlorides of these same ele-
ments, always at 25OC and an ionic strength of 2, a new
straight line is obtained. Similarly, r values of the sulfates,
fluorides, and nitrates (excepting nitric acid, zinc, and
cadmium) are plotted against the I- values of the correspond-
ing chlorides in Figures 7 and 8. I n each case, a straight line
forms, regardless of the z+ and z- values of the electrolyte
involved.
The lines on Figures 7 and 8 are all of the form
rMv+Xv- =Ar MV+ Cl v - +B (3)
Ind. Eng. Chem. Process Des. Develop., Vol. 1 1, No. 1, 1972 131
I I 1 I I finds that the contributions of charge-charge and charge-
molecule effects to log yrt all vary as the product (z+z),
while the molecule-molecule effects are relatively inde-
pendent of ion charge. I n other words, for a given solvent, T~
is a function of p (z+z), and noncharge factors unique to each
solvent-electrolyte system. A similar dependence between
variables is involved in the equation of Davies (1938) which
followed the work of Guggenheim and others (Guggenheim,
1935; Guggenheim and Turgeon, 1955; Guggenheim and
Stokes, 1969; Wu and Hamer, 1969) :
-
-
-
a
0.4
0.4 0.6 0.0 I .o
rM,,+CIv-
Figure 8. Cross plots of r Mv + Xv - vs. r Mv.Cl v- ( X- =F- and
S042-) at 25C and an ionic strength of 2.0
where r Mu+Cl v- refers to the chloride of a given cation M while
I Mv+Xv- represents the combination of M with the anion x,
always at the ionic strength of 2. Table I presents values of
r Mv+Cl u- while Table 11presents values of A and B for various
anions to be used in Equation 3. These tabulated A and B
values were obtained by least square calculations, while the
standard deviations a* of the best fit are listed in Table 11.
From Figures 7 and 8, or Equation 3, values of I? at an ionic
strength of 2 can be predicted for many substances. Thus
knowing r for Pb(NO& to be 0.426 at a p of 2, the value of r
for PbClz at this concentration (which would represent a
highly supersaturated solution) can be predicted from Figure
7, and is equal to 0.599.
Di scussi on
Deviations from ideal solution behavior for electrolytes,
leading to T* values lower or higher than unity, are the result
of interactions of ions with other ions and with the solvent.
Scatchard (1959) discusses these under three headings, namely
the charge-charge (electrostatic) interactions of the Debye-
Hiickel theory, Lcharge-molecule effects such as are en-
countered in salting out behavior, and the molecule-mole-
cule effects involved in solvation-association reactions. I n
developing his final equation 2 (Scatchard, 1959) , Scatchard
I n this equation, log y* is again indicated to be a function of
p( z +z - ) , and the empirical constant, B, which changes from
one solvent- electrolyte system to the next.
There appears to be no theoretical conflict between all this
earlier work and the correlations of Figures 2-5, wherein I
(which includes the effect of ion charges as is evident from
Equation 5) is shown to be a function only of p and of the
electrolyte itself in aqueous solutions. The curves of Figures
2-5 fit the experimental data more successfully than the
analytical equations mentioned, presumably because the latter
are derived on assumptions regarding the force-distance rela-
tionships between ions, while no such assumptions need be
made in the graphical approach.
The relationships shown in Figures 7 and 8 and by Equation
4 are again empirical. The selection of chloride as the reference
ion was arbitrary, but was made because of the relative abun-
dance of data for chloride systems. The success encountered
in predicting r values at an ionic strength of 2 by Equation 4
has already been discussed. Presumably equal success will be
encountered with other anions such as the cyanides, but this
remains to be established. To use Equation 4 for electrolytes
involving other untested anions, values for A and B in Equa-
tion 4 must be available, requiring experimental data on at
least two compounds. I n the cyanide system, these might be
KCN and NaCN.
The selection of an ionic strength of 2 as a basis for Equa-
tion 4 and Figures 7 and 8 was also arbitrary, since equivalent
relations exist at other ionic strengths. Moreover, the iodide-
chloride data at ionic strengths both at unity and 4 yield
points which fall approximately onto the line of Figure 7 and
so have the A and B values of Table I. Other systems, however,
such as the sulfate-chloride electrolytes, form different lines
from those of Figure 7 for ionic strengths of unity and 4. The
Table II. Values of A and B for Equation 3; p =2.0
Anion Cation
F- Na, K, Rb, Cs
C1-
Br-
I-
NO*-
504-
OAc-
C104-
OH- K, Na, Li
H, Na, Li, K, Cs, Rb, Mg
Ca, Sr, Ba, Co, Cd, Zn
Na, K, Rb, NHI , Mg, Ca,
Sr, Co, UOZ
Li, 4, Rb, Cs, Na, Al,
Mg, Cu, Mn, Ni, U0zb
Na, Rb, K, Cs, Mg, Ba
H, Li, Na, Mg, Ba, Ca
(acetate)
~UO2F2 omitted. * H2SO4 omitted.
A
-1. 482
1
1. 259
1. 648
2. 248
0. 100
-1. 707
1. 223
- 1. 043
B
1. 521
0
-0. 121
- 0. 313
- 0. 918
0. 443
1. 830
-0.085
1.437
d
0.00018
0
0.00078
0.00196
0.00272
0.00056
0.01565
0.01065
0.00174
No. of
points
4
13
13
9
. . .
11
6
6
3
132 Ind. Eng. Chem. Process Des. Develop., Vol. 1 1, No. 1, 1972
choice of an ionic strength of 2 for Equation 4 and for these
cross plots is therefore a compromise, since at smaller p values
the differences between r for various electrolytes diminish,
thereby increasing uncertainty in locating the parent curve,
while at larger p values fewer data on y* become available.
Finally, values of yh for electrolytes can be measured very
accurately by methods such as freezing point depression and
half cell potentials. In consequence, experimental values of yf
are often reported to three significant figures. An equal suc-
cess cannot, of course, be obtained by the empirical method
described here, and so this procedure should be used only when
direct measurements are lacking.
Nomenclature
a =electrolyte activity, mo1,/1000 g solvent
uk =ykmk =mean electrolyte activity, mo1/ 1000 g solvent
yk =mean activity coefficient at 25C
r =(y+)b*+z-
m =molality, mo1/1000 g solvent
mi: =rn(v+y+v-y-)l/v
V + =moles of completely dissociated catioiis
V- =moles of completely dissociated anions
Y =Y+ +Y -
p =ionic strength = / z x(Vini)zi2
a
zt =charge on the cation
z- =charge on the anion
Literature Cited
lhvies, C. W., J . Chem. Soc., 1938, p 2093.
Debye, P., Huckel, E., Physik. Z., 24, 185, 305 (1923).
Guggenheim, E. h., Phil. Mag., 19, 588 (1935).
Guggenheim, E. A, , Stokes, R. H., Equilibrium Pro erties of
ilqueous Solutions of Single Strong Electrolytes, $ergamon
Press, New York, K.Y., 1969.
Guggenheim, E. A. , Turgeon, J . C., Trans. Faraday Soc., 51, 747
(1955).
Harned, H. S., Owen, R. B., Physical Chemistry of Electrolyte
Solutions. No. 95. 3rd ed.. Reinhold. New York. N.Y.. 1958.
hloore, W. J ., P h& d Chemistry, 3rd ed., Prentide-Hall,
Pitzer, K. S., Brewer, L., Thermodynamics, McGraw-Hill,
Robinson, R. A. , Stokes, It. H., Electrolyte Solutions, 2nd ed.
Scatchard, G., Chem. Rev., 19, 309 (1936).
Scatchard, G., in The Structure of Electrolyte Solutions, p 9,
Wu, Y-C., Hamer, W. J ., Electrochemical Data, Part XIII,
Englea-ood Cliffs, N.J ., 1962.
Kew York, N.Y., 1961.
rev., Academic Press, New York, N.Y., 1965.
W. J . Ilamer, Ed., Wiley, New York, N.Y., 1959.
Vatl. Bur. Std., 1Lept. 10002 (1969).
RECEI VED for review March 31, 1971
ACCEPTED J uly 22, 1971
Presented in part at the Meeting of the Electrochemical Society,
Washington, D.C., May 11, 1971.
Heat Transfer to Air-Solids Suspensions
in Turbulent Flow
Shafik E. Sadek
Dynatech RI D Co., 99 Erie St., Cambridge, Mass. 02189
A correlation of heat transfer coefficients to air-solids suspensions based on the principles of similarity i s pre-
sented. Published data indicate that a single parameter, particle area per unit gas volume, i s sufficient to
define reasonably well the relative increase in heat transfer coefficients. The scatter in the available data
prevents the weak dependency on another parameter, the pipe Reynolds number, from being accurately
defined. The correlation represents data in air covering pipe Reynolds numbers from 4000-80,000, particle
diameters from 20-600 1.1, pipe diameters from 17-1 02 mrn, and solids loading ratios up to 300 in vertical
transport.
T he effect of particulate matter on gas-phase heat transfer
has recently received widespread attention. Particles sus-
pended in a gas were discovered to increase heat transfer rates,
often significantly above those expected from the higher heat
capacities of the mixtui~es and the resulting greater driving
forces (Koble et al., 1951; Farbar and Xorley, 1957).
The promise of high heat transfer rates to gas-solids suspen-
sions during transport led to a number of studies to gain an
understanding of the mechanisms involved. At first, the aug-
mentation of heat transfer mas explained on the basis of the
mixtures behaving as a homogeneous mixture (Depew and
Farbar, 1963; Tien, 1961). When experimental data indicated
that this theory was inadequate, i t was concluded that the
particles disturbed the velocity profile and consequently the
temperature profile, thus reducing the effective thickness of
the boundary layer (Depew and Farbar, 1963; J epson et al.,
1963; Leva, 1959; Peskin and Briller, 1970). Even though this
is the most widely accepted theory for explaining the aug-
mentation effect, no satisfactory correlation was developed on
this basis. Theoretical developments did not correlate the data
properly olving to the complexity of the details of the aug-
mentation mechanism (Peskin and Briller, 1970), while
empirical correlations fell short from being general (Danaiger,
1963; Koble et al., 1951; hlickley and Trilling, 1949). It was
pointed out that a compromise between these two approaches
was the most promising approach toward generating a useful
correlation (Peskin and Briller, 1970). Such a compromise is
offered by the principles of similarity: Dimensionless param-
eters governing the process are first defined based on a sim-
plified understanding of the augmentation mechanisms fol-
Ind. Eng. Chem. Process Des. Develop., Vol. 1 1 , No. 1, 1972 133