United States
Environmental Protection
Agency
Water Engineering
Research Laboratory
Cincinnati OH 45268
Research and Development
EPA/600/S2-85/093 Sept. 1985
v>ERA Project Summary
Prediction of Selectivity for
Activated Carbon
Adsorption of Trace Organic
(Homologue) Contaminants
Georges Belfort, Cheng-Sheng Lee, Roger J. Weigand, and
Diane M. Neuhaus
3**
Preferential adsorption of organic
compounds onto activated carbon from
dilute aqueous solutions was studied
to develop a comprehensive theoretical
basis for predicting adsorption of mufti-
component solutes. The study com-
pared different carbons and investi-
gated their adsorption behavior with
variable aqueous solution properties,
and the differences between gas and
liquid phase adsorption. The overall ob-
jective was to develop and test the
comprehensive solvophobic theory.
Experimental adsorption isotherms
of a series of alcohols on five different
carbons were measured and compared.
Isotherms were run for 4-n-propyl phe-
nol at various concentrations of ammo-
nium suffate salt and methanol as the
additives to increase or decrease sur-
face tension of solution, respectively.
Also, isotherms for three linear car-
boxylic acids were measured at differ-
ent pH's to determine the effect on ad-
sorption.
An experimental glass vacuum sys-
tem has been designed and built to
measure adsorption isotherms in the
gas phase. Measurements of adsorp-
tion isotherms for four alcohols were
conducted in the aqueous and gas
phases and confirmed the theoretical
model.
Predictive techniques for multi-
component adsorption were developed
based on the use of an equation of
state. They were compared with the
Ideal Adsorbed Solution (IAS) theory
using competitive phenol adsorption
data.
This Project Summary was devel-
oped by EPA's Water Engineering Re-
search Laboratory, Cincinnati, OH, to
announce key findings of the research
project that is fully documented in a
separate report of the same title (see
Project Report ordering information at
back!.
Background
Although activated carbon (AC) ad-
sorption has been widely used for odor
and color removal in the water industry,
only recently has it been seriously con-
sidered for removal of dissolved organ-
ics from water supplies. This recent in-
terest is caused primarily by the
growing concern for potential carcino-
genic, mutagenic, and teratogenic com-
pounds found in drinking waters.
With respect to solid-liquid adsorp-
tion, a major limitation of the various
equilibrium theories of adsorption (be-
sides the adaptation of the partial solu-
bility parameter theory) is that they
were orginally derived from gas and
vapor phase adsorption and thereby a
priori ignored the presence of the sol-
vent during solute adsorption.
Recent attempts to include the sol-
vent effect in aqueous phase adsorption
include a semi-empirical approach
based on partial solubility parameters
called the net adsorption energy ap-
proach. Unfortunately, the arbitrary
choice of the relative values for the acid
and base hydrogen bonding solubility
parameters is without theoretical justifi-
cation and renders this approach some-
what capricious.
-------
During the past several years, a con-
certed effort has been made in our
Rensselaer Polytechnic Institute labora-
tory to understand the role of the sol-
vent during liquid-solid approach. A
theoretical basis has been derived for
predicting a priori the preferential ad-
sorption of organic compounds in the
same homologous series on activated
carbon from dilute aqueous solution.
This basis has been derived and used
with confidence to rank-order single so-
lute isomers (alcohols, ketones, and
phenols) with respect to their adsorp-
tion potential. Clearly, the ability to pre-
dict the effects of structure on the ad-
sorption of organic molecules between
different homologous series from dilute
aqueous solutions could be of great
value in the design and operation of
large-scale commercial plants.
Objectives
The objectives of this study were as
follows:
1. To compare different activated car-
bons using a well characterized,
model homologous series (linear al-
cohols),
2. To evaluate the effects of surface
tension on adsorption by changing
the concentration of ammonium sul-
fate salt and methanol,
3. To evaluate the effects of pH on car-
boxy lie acids adsorption,
4. To develop further the comprehen-
sive solvophobic theory by estimat-
ing the value of the gas phase term.
This involved measuring the gas
phase adsorption of the same ho-
mologous series as that measured
from the liquid phase, and
5. To develop predictive techniques
based on the use of an equation of
state for multicomponent adsorption
from mixtures.
Adsorption capacity for specific, sin-
gle, organic solutes of a homologous
series is thought to be a direct function
of: (1) the adsorbate properties, (2) the
solution conditions, and (3) the nature
of the adsorbent. Little quantitative in-
formation is known about the surface
characteristics of carbon, the solution
properties, and their influence on or-
ganic adsorptive selectivity. The gen-
eral thrust of this study is to extend both
the theoretical formulations and the ex-
perimental data base to develop a com-
prehensive theoretical basis including
the important factors on adsorption for
predicting a priori the preferential ad-
sorption of organic compounds on acti-
vated carbon from aqueous solution.
Materials and Methods
Adsorbents and Adsorbates
Crushed granular activated carbons
(PAC) were the major adsorbents used
in this study. Comparisons were made
of adsorption isotherms of three
aliphatic alcohols on five different gran-
ular activated carbons. Cleaning proce-
dures of the PAC were an important as-
pect of obtaining reproducible
isotherms.
The adsorbates used included three
homologous series of alkyl phenols,
aliphatic alcohols, and linear carboxylic
acids—all of the highest grade available
(>99% purity). Stock solutions of the in-
dividual alcohols or phenols were made
up with standard 0.01 m phosphate
buffer (pH 7.00) and stored in amber
glass or aluminum-covered bottles. Car-
boxylic acids were prepared in phos-
phate buffer solutions of pH 4.75 and
7.00.
For the liquid phase, several innova-
tions have been introduced into the ad-
sorption isotherm procedures in an at-
tempt to improve reproducibility and
accuracy and to reduce solute losses.
Thus completely filled and capped
stainless steel tubes containing the ad-
sorbate/adsorbent mixture were rotated
360° end-over-end at 2 rpm for 24 hr at
20 ± 0.5°C. The tubes were then ultra-
centrifuged at 20,000 rpm for 20 to
30 min at 20°C, thereby spinning down
the carbon. The tubes were then
opened, and the supernatant was ana-
lyzed directly.
Vacuum System
The vacuum system (made of glass)
was used in the gas phase adsorption/
desorption experiments. The amount of
alcohol adsorbed onto the activated car-
bon was measured directly using an
electrobalance. The final mass of the
sample and the final pressure in the sys-
tem were used to represent one point
on the isotherm. Successive points
were determined by admitting small
doses of alcohol and recording the equi-
librium mass and pressure readings
until the final system pressure was
nearly equal to the vapor pressure of
the pure alcohol. The desorption
isotherm was then determined in a sim-
ilar, stepwise fashion by heating and
evacuating the carbon samples. Mass
and pressure readings were recorded at
equilibrium.
General Solvophobic Approach
The solvophobic (c<}>) theory de-
scribes the tendency of a surrounding
solvent medium to influence aggrega-
tion or dissociation of those molecules
with considerable microsurface areas
exposed to the solvent medium.
In the solvophobic treatment, adsorp-
tion is considered a reversible reaction
between the adsorbate molecules, S,,
and the activated carbon, C, to form the
adsorbed complex, S\C, at the surface of
the carbon, S, + C ±? SjC. The effect of
the solvent on this reaction is obtained
by subtracting the standard free energy
change for the reaction in the gas phase
from that in the presence of the solvent
(taking as standard states X£ = 1, p£ = 1
atmosphere ideal gas). This process re-
sults in a net free energy change, with
AGFso'ivent effect) expressing the effect of
the solvent on the association adsorp-
tion reaction.
Conceptually, Sinanoglu posed a two-
step dissolution process. First, a hole or
cavity needs to be prepared in the sol-
vent to accommodate the solute, car-
bon, or adsorbed complex molecule.
Second, after the molecule is placed
into the cavity, it interacts with the sol-
vent. Quantitatively, this process is ex-
pressed as follows:
A — net .f. assoc . — assoc
AG(so|vent effect) = AG(so|vent) ~
= RT
(1)
or
A f* ^o* A f* net « r^ not
Ao(solvent effect) = Akj.SiC ~ AGj/Si
-AG'
net
'J.C
(2)
where Kf = Pk/Xk is the Henry's con-
stant for the kth species, and j represents
each type of interaction. After specify-
ing each interaction such as the cavity,
van der Waals', and electrostatic terms
plus two correction terms for polymer
mixing and reduced electrostatic effects
because of the presence of the solvent,
the following expression is obtained
from Equations 1 and 2 for the overall
standard free energy change, viz.
.—assoc _
AG(Solvent) -
(3)
AGvdw + AGes
AGred]s?c-8i-c
RT ln(RT/P0V)
-------
where the last term is called the cratic
term and results from an entropy or free
volume reduction. AG^^nt) is related
to the experimental equilibrium con-
stant, Ksdvenu = XSiC/XSiXc, which itself
will be related tojhe experimental ad-
sorption capacity pifor solute Sj later in
this analysis. AGfgSf can be calculated
by multiplying the spreading pressure
by the molar area occupied by the ad-
sorbate in a surface layer, a\
= -ar>r
(4)
The spreading pressure can be obtained
from gas phase isotherms. Each term in
the square bracket in Equation 3 can be
calculated explicitly from known phys-
iochemical parameters obtained from
the literature.
For the comprehensive c-model,
each of the terms in Equation 3 is calcu-
lated explicitly; for the simplified
model, the thermodynamic microsur-
face area change of the reaction, AA, in
the cavity term is assumed to be propor-
tional to the cavity surface area, ISA, of
the specific sorbate and homologous
series, AA = g TSA. Thus in this study.
In Q°b is correlated with TSA, where Q°b
is the initial slope at low solute concen-
tration for the Langmuirian adsorption
isotherm. Also, included in the compre-
hensive report are (1) the adsorption be-
havior as related to surface tension at
various concentrations of ammonium
sulfate salt and methanol, (2) pH effect,
and (3) a comparison of a nonideal mul-
ticomponent adsorption model based
on the use of an equation of state with
the ideal adsorbed solution theory.
Experimental Results
Comparison Between Different
Carbons
In Table 1, the Langmuir adsorption
parameters Q° and lnQ°b are correlated
for the linear alcohols with molecular
weight. Very good correlations are ob-
tained. To test the simplified version of
the solvophobic theory for different car-
bons, the molecular cavity surface area
of each solute (i.e., the total surface
area, or TSA) was calculated using the
MDS program. The TSA for each alkyl
alcohol is then correlated with the Lang-
muir adsorption parameters in an at-
tempt to verify the theory as a predictive
measure of solute adsorption. In Table
1, we observe that the adsorptivities for
the linear alcohols also correlate well
with TSA. However, TSA does not sig-
nificantly correlate with adsorption
parameters any better than molecular
weight for linear alcohols, as expected.
In Table 2, the Langmuir adsorption
parameters Q° and lnQ°b are listed with
characteristics of adsorbents for five dif-
ferent carbons. Neither the saturation
adsorption nor the adsorption capacity
has any apparent correlation with sur-
face area of adsorbents. The surface en-
ergy of adsorbent should have an im-
portant effect on the adsorption
process.
The Differences Between Gas
and Liquid Phase Adsorption
Adsorption isotherms were measured
in the gas phase and in the liquid phase
for four linear alcohols (butanol-
heptanol) and three branched alcohols
(2-methyl 1-butanol, 2-methyl 3-
pentanol, 3-3-dimethyl 1-butanol) on
the same activated carbon. All of these
isotherms fit the Langmuir form very
well, and the parameters of this model
versus molecular weight of the linear
alcohol appear in Figure 1. The satu-
rated adsorption capacity, Q°, increases
linearly with molecular weight for ad-
sorption from solution, whereas the op-
posite is seen for gas adsorption. The
amount of water adsorbed from the
aqueous phase can also be estimated
by assuming complete wetting of the
entire carbon surface (1014 m2/g). By
assuming spherical molecules in hexag-
onal packing on the surface, we can cal-
culate the amount of water necessary to
cover the remaining surface not cov-
ered by the alcohol. The adsorption ca-
pacity for water also appears on Fig-
ure 1 and is seen to decrease linearly as
the alcohol adsorption increases. Since
the water curve is much higher than the
solute curve, we know that the solute
never forms a complete monolayer on
the surface in liquid phase adsorption.
The energy term, b, increases with
molecular weight in all cases except
one. The anomalous result for the ad-
sorption of heptanol from the gas phase
is probably caused by steric exclusion
effects or incomplete adsorption, since
extreme hysteresis was observed for its
isotherm during desorption.
General Solvophobic Approach
The solvent-effect free energies,
AGsofvent effect'for 12 linear and branched
alcohols have been calculated previ-
ously. The gas phase free energies are
Table 1.
Correlation Coefficient r for Langmuir Adsorption Parameters versus Molecular Weight and Total Surface Area for the Alkyl Alchohols
on Five Different Carbons'
Filtrasorb 400
W20
D10
SA4
SA Plus
Adsorption
Parameter
Q°
lnQ°b
Molecular
Weight
0.99
0.97
TSA
A2
0.99
0.97
Molecular
Weight
0.85
0.96
TSA
A2
0.85
0.94
Molecular
Weight
0.87
0.97
TSA
A2
0.87
0.97
Molecular
Weight
0.79
0.93
TSA
A2
0.79
0.93
Molecular
Weight
0.99
0.98
TSA
A2
0.99
0.98
'The alcohols studied were butanol, pentanol, and hexanol.
Table 2. Langmuir Adsorption Parameters for the Alkyl Alcohols on Five Different Carbons
Adsorbents
Filtrasorb
W20
D10
SA4
SA Plus
Surface
Area
-------
; 2
i
o
14
12
10
120
100 fr
50 r*
60 3
40 I
a
20
• Vapor
• Solute in Solution
* Water
0.9976
0.9964
0.9934
70 80 90 100 110
'Molecular Weight
120
Figure 1.
Langmuir model parameters
versus molecular weight for the
adsorption of linear alcohols
. from gas and liquid phases.
calculated using Equation 4 and added
to the solvent-effect free energies to
yield the free-energy change in solution
(see Equation 3).
Equation 4 is shown as the theoretical
line on Figure 2, and the experimental
points for the four linear alcohols fall
very close to this line. Figure 2 also
shows the experimental points for these
compounds when adsorbed from the
liquid phase. The c theory predicts a
linear relationship between InCTb and
SOLVENT. This is shown by the best fit
line, which correlates the data quite well
(r = 0.979). Thus the reversal in the
order of preferential adsorption result-
ing from the presence of water is ob-
served experimentally and substanti-
ated theoretically.
For this reversal to occur, the domi-
nant mechanism for equilibrium ad-
sorption onto activated carbon must be
different in the vapor and aqueous
phases. Although little quantitative in-
formation is known about the surface
characteristics of activated carbon and
its influence on organic adsorptive se-
lectivity, essentially two types of inter-
actions are thought to dominate. The
first can be characterized by van der
Waals physical interactions, and it
occurs on a majority of the surface (on
basal planes). The second type of sur-
face interaction, which occurs at the
more reactive edges of the microcrys-
I
I
a
Solvent
r = 0.979
best fit
theory
-25
-20
-15
-10
-5
AG//?r
Figure 2. lnQ°b and Inbp, as a function of AG"f£,cm and AGSSJ00, respectively.
tallites, can be characterized by attrac-
tive polar interactions resulting from,
for example, hydrogen bonding and
electrostatic forces. These specific inter-
actions result from the surface hetero-
geneity and the presence of oxides, hy-
droxyls, and other groups on the
surface. The adsorption capacities for
alcohols are much higher for gas ad-
sorption than for liquid phase adsorp-
tion (Figure 1). Thus gas adsorption is
most likely dominated by van der Waals
dispersion interactions.
Surface Tension Effect on
Adsorption
Adsorption isotherms were measured
in the liquid phase for 4-n-propyl-
phenol in solvents of varying surface
tension. The surface tension was
changed by adding varying concentra-
tions of (1) ammonium sulfate salt to
increase the solvent surface tension, or
(2) methanol to decrease the solvent
surface tension. Figure 3 shows the ad-
sorption capacity, Q°, and the ad-
sorbability, lnQ°b, as functions of salt
concentration and surface tension. In
both cases the first four points follow a
linear increasing trend as predicted by
the c theory- The theory therefore fits
the data well up to a certain salt concen-
tration (around 1.143 M). Above this
concentration, the data show a de-
crease in adsorption, deviating from the
theory. This tendency may be explained
by entropic effects. At high salt concen-
trations, the salt molecules may hinder
the movement of the organic toward the
carbon surface, or they may induce a
precipitation. The discussion of adsorp-
tion behavior at various concentrations
of methanol is also included in the com-
prehensive report.
pH Effect on Adsorption
Adsorption isotherms were measured
at different pH values (7.00 and 4.75) in
the aqueous phase for three linear car-
boxylic acids on the same activated car-
bon. All of these isotherms fit the Lang-
muir form very well. A large increase in
absorbability occurred as a result of the
lower pH for each acid studied. This can
be explained by noting that a low pH
increases the relative amount of a-
polar, undissociated carboxylic acid,
and this relatively non-polar species is
more easily adsorbed than the dissoci-
ated form.
Multi-Component Adsorption
Figure 4 contains the experimental bi-
nary adsorption data and the predicted
4
-------
results for an initial concentration of the
alkyl phenols in the mixtures. Solid lines
and dotted lines are calculated using the
IAS theory and the theory based on the
equation of state, respectively. Agree-
ment between predicted and experi-
mental results is good for the latter the-
ory using van der Waals Equation,
especially in the higher surface loading
region compared with that predicted by
the IAS theory. The IAS theory appears
to be most suitable for quantitative pre-
diction of binary adsorption of small or-
ganic molecules with similar adsorption
characteristics.
Conclusions
The woHcreported here involves the
development of a comprehensive theo-
retical basis for predicting the preferen-
tial adsorption of organic compounds
onto activated carbon from dilute
aqueous solution. A detailed analysis
including the characteristics of adsorb-
ents, the effect of solution properties,
and the differences between gas aad
liquid phase adsorption yields an in-
sight into the adsorption process. This,
together with the solvophobic theory,
provides a quantitative description of all
the important factors on adsorption.
The solvophobic theory is tested
using single-solute aqueous-phase ad-
sorption isotherms of a series of alco-
hols on different carbons. Correlations
with adsorption capacity for the simpli-
fied solvophobic theory parameter, ab-
sorbate molecular area (ISA), are all in
good agreement with theory for all five
different carbons. Despite the steric hin-
drance (pore size and pore size distribu-
tion), both the adsorption capacity and
adsorbability show no apparent correla-
tion with surface area of adsorbents. It
seems that the surface energy of the ad-
sorbent has an important effect on the
adsorption process.
Adsorbability of 4-n-propylphenol in-
creases linearly with increasing ammo-
nium sulfate concentration (or surface
tension) up to a critical concentration of
about 1.143 M. At higher concentrations
the adsorbability decreases, deviating
from the theory. This result is explained
by incomplete adsorption and precipi-
tation arguments. Adsorbability de-
creases linearly with the volume per-
cent of methanol present in the solvent.
Solvent surface tension is not a linear
function of methanol concentration, so
adsorbability does not correlate linearly
with surface tension. The variability of
the parameter K" in the solvophobic the-
ory is invoked to explain this result;
Surface Tension (erg/cm2
72 73 74 75 76 77
0.5 1.0 1.5 2.0
SO* Concentration (mole/L)
Figure 3. Adsorption capacity and adsorb-
ability versus salt concentration
and solvent surface tension for
4-n-propylphenol.
however, competitive effects resulting
from the methanol may also be present.
The adsorption capacity of carboxylic
acids exhibit a definite increase as the
pH is decreased from 7.00 to 4.75. This
can be explained by noting that a low
pH increases the relative amount of a-
polar undissociated carboxylic form.
The solvophobic theory is also tested
using carboxylic acids at different pH.
Correlation with adsorption capacity for
the simplified solvophobic theory
parameter, TSA, is only in good agree-
ment with theory at pH 7.00. Surface in-
teractions at low pH probably compli-
cate matters.
The solvophobic theory successfully
correlates both gas and liquid phase ad-
sorbabilities (Inbp, and lnQ°b respec-
tively) for the alcohols. It also predicts
the reversal in order of preferential ad-
sorption with molecular weight because
of the presence of liquid water. The
complete and simplified c theory is
also used to correlate adsorbability
from solution for homologous series of
alcohols, ketones and phenols. In all
three cases, the simplified theory is
found to be sufficient, and molecular
weight can be ruled out as a correlating
parameter for adsorbability.
Experiment
IAS Theory
van der Waals Equation
fa) 4-n-methylphenol
1-2\- (initial cone.: 0.0760 mmole/L)
(b) 4-n-propylphenol
(initial cone. .-0.0652 mmole/L)
1.0
0.8
£
0.4 •
0.2
1.0 2 3 4 56
Figure 4. Adsorption from aqueous multi-solute system (4-n-methylphenol and 4-n-propyl-
phenol) at low initial concentrations.
-------
The adsorption isotherms of dilute
aqueous solutions containing 4-n-
methylphenol and 4-n-propylphenol are
measured over a concentration range of
0.1 - 0.001 mole/L on activated carbon
at 20°C. The experimental results agree
better with equilibrium predictions
based on a two-dimensional adsorbed
phase represented by a van der Waals
Equation of state than by an ideal ad-
sorbed solution theory, especially for
the weak adsorbate. It appears that the
ideal adsorbed solution theory is reli-
able for multicomponent systems
where solute adsorption loading is low.
When solute adsorption loading is high,
the non-ideality of mixing in the ad-
sorbed phase is related to constants in
the two-dimensional equation of state
to allow for solute-solute interactions
on the surface.
Recommendations
Myers and Sircar have recently re-
ported an asymptotic principle of corre-
spondence for physical adsorption
equilibria on heterogeneous and mi-
croporous surfaces. They propose a di-
mensionless group [mRT In (p/p0)/AG]
as a universal function of 6 (fractional
filling of micropores); where m is the
saturation capacity and AG is the free
energy of immersion of the adsorbent in
the pure liquid sorbate. It may be possi-
ble to predict AG for homologous series
of adsorbate, using the solvophobic
theory. In addition to the free energy of
immersion approach, gas phase ad-
sorption data have been used in our lab-
oratory to estimate the dispersion con-
tribution of the surface free energy of
the adsorbent. Combining the statistical
mechanical theories of adsorption from
solution with the relationship between
interfacial tension and tensions, it can
be shown that both the total surface
area of the solute and the surface free
energy of the adsorbent are important
factors for rank-ordering adsorption po-
tential. A linear relationship between
total surface area of solute and Gibbs
free energy change of adsorption from
solution has been demonstrated by the
solvophobic theory. Thus a comprehen-
sive experimental data-base of surface
free energy for adsorption from the gas
phase and its influence on organic ad-
sorption selectivity will provide insight
into the adsorption process. This, to-
gether with the solvophobic theory, will
provide a quantitative description of the
role of the solvent and the effects of
structural modifications of organic
molecules on adsorption.
The full report was submitted in fulfill-
ment of Cooperative Agreement No.
CR-809686-01-0 by Rensselaer Polytech-
nic Institute under the sponsorship of
the U.S. Environmental Protection
Agency.
Georges Be/fort, Cheng-ShengLee, Roger J. Weigand, andDianeM. Neuhausare
with Rensselaer Polytechnic Institute. Troy, NY 12180-3590.
Richard A. Dobbs is the EPA Project Officer (see below).
The complete report, entitled "Prediction of Selectivity for Activated Carbon
Adsorption of Trace Organic (Homologue) Contaminants," (Order No. PB 85-
243 160/AS; Cost: $11.95, subject to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Water Engineering Research Laboratory
U.S. Environmental Protection Agency
Cincinnati, OH 45268
U. S. GOVERNMENT PRINTING OfflCE:1985/559-l 11/20692
-------
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
EPA/600/S2-85/093
0000329 PS
------- |