GENERAL0ELECTRIC
RESEARCH
AND
DEVELOPMENT
CENTER
SCHENECTADY, NEW YORK
IMMOBILIZED LIQUID MEMBRANES
FOR SULFUR DIOXIDE SEPARATION
FINAL REPORT
Contract No. PH-86-68-76
Prepared for
Department of Health, Education, and Welfare
U.S. Public Health Service
National Air Pollution Control Administration
5710 Wooster Pike
Cincinnati, Ohio 45227
W. J. Ward III and C. K. Neulander
General Electric Research and Development Center
Schenectady, New York
March 1970
S-70-1053
-------
IMMOBILIZED LIQUID MEMBRANES
FOR SULFUR DIOXIDE SEPARATION
FINAL REPORT
Contract No. PH-86-68-76
Prepared for
Department of Health, Education, and Welfare
U.S. Public Health Service
National Air Pollution Control Administration
5710 Wooster Pike
Cincinnati, Ohio 45227
W. J. Ward III and C. K. Neulander
General Electric Research and Development Center
Schenectady, New York
March 1970
S-70-1053
-------
ABSTRACT
An immobilized liquid membrane has been developed which at 100°C has an
S02 flux of
3.2 X 10-3 cc (STP)
sec, cm2, cm Hg 6. P
and an S02/ CO2 separation factor of 14. The membrane operated for one month
at 100°C with no change in permeation properties. The membrane could not
withstand a pressure difference of 1 atm across it. However, a technique was
recently developed which may make it possible to overcome this limitation.
The low S02/ CO2 separation factor makes the use of this membrane for
removing S02 from power plant stack gases economically unattractive. An
economically attractive system has been conceived for treatment of higher S02
concentration, and lower total flow stack gases, such as those emitted from
many ore- smelting processes.
-------
IMMOBILIZED LIQUID MEMBRANES FOR SULFUR DIOXIDE SEPARATION';'
W. J. Ward III and C. K. Neulander
INTRODUCTION
Today, sulfur oxide pollution is a substantial
threat to people and agriculture in many areas of the
world. With present control practices the rate of
S02 discharge in the United States will double by 1980.
Past research efforts on means of controlling S02
pollution have concentrated on methods for removing
S02 from combustion gases and extracting sulfur from
coal and residual oil. Sulfur dioxide elimination from
combustion gases has been the subject of more re-
search than any other gas purification operation, but
the results have not been satisfactory. Much of the
past and present research has centered around the
identification of the nature and magnitude of the prob-
lem. While this has been useful, what are clearly
needed are methods of control. Several first gener-
ation processes currently under development may be
economically tolerable in certain critical areas.
However, entirely new approaches to the problem
must be developed if problems beyond the 1975 to
1980 period are to be met. The following is a report
on the initial developmental work on one such approach.
Over the years there have been many attempts to
develop semipermeable membranes for separation
processes, and patents on potential processes date
back more than a century. However, it is only in the
past 10 years that membranes have been developed
which make certain membrane separation processes
within range of being economically and technically
feasible. Although several polymers such as silicone
rubber and cellulose acetate are useful membrane
materials, in general, polymeric materials are not
desirable as semipermeable membranes since most
are relatively impermeable to all gases and liquids.
Several years ago, W. L. RobbO) considered liquids
as potential membrane materials, and based on this
concept a highly effective immobilized liquid mem-
brane for CO2 removal from a life-supporting envi-
ronment was developed by Ward and Robb. (2)
The transfer of a solute through polymeric and
liquid membranes occurs by a solution! diffusion
mechanism. That is, the solute dissolves in the
membrane, diffuses through it, and desorbs on the
other side. The permeability of a nonreacting solute
in a membrane is thus the product of the diffusion
coefficien t and the solubility of the solute in the mem-
brane. A standard set of units for permeability,
which is consistent with this definition, is
cc(STP), cm thickness
sec, cm2, cm Hg LIP
In the case of gas permeation, the driving force
for the transfer of a gas through a membrane is a
partial pressure difference in that gas across the
membrane. In the case of S02 removal from stack
gases, there are only two reasonable ways of main-
taining a partial pressure difference in S02' One is
to evacuate one side of the membrane, and the other
is to sweep away with a condensable gas the S02
permeating through the membrane. A combination
of these two methods is also possible.
Liquids are under consideration here as mem-
brane materials since they have large diffusion co-
efficients compared to polymers; and also, for cer-
tain gases, they can have enormous solubilities. Thus
liquids can be highly permeable to, and highly selec-
tive for, certain gases, particularly polar ones such
as S02' Liquid membranes can be made in a variety
of ways. Very stiff gels can be used as membranes.
In some cases a polymer membrane can be swelled
with a liquid to the extent that the liquid controls the
rate of permeation. In some cases a high-grade
microporous filter can be impregnated with the liquid,
and this will serve as the liquid membrane. In the
present work, fabricating the liquid membrane was
particularly difficult because it had to be very thin.
In general, packaging membranes in the form of
usable hardware is a formidable problem which only
recently is receiving intensive consideration. The
design and fabrication of membrane packages was not
within the scope of this work. However, immobilized
liquid membrane packages are being made in this
Laboratory, and from this work it can be assumed
that packaging densities of the order of 100 to 500
square feet of active membrane area per cubic foot
of volume can be achieved.
The present investigation was concerned with
developing an immobilized liquid membrane which
was highly permeable to and selective for S02 at
elevated temperatures, and devising an appropriate
membrane system for removal of S02 from stack gases.
MEMBRANE DEVELOPMENT
Membrane development included selection of the
optimum membrane liquid and fab rication of an im-
mobilized film of this liquid. The primary require-
ments of the membrane were:
1.
The S02 flux must be of the order of
4 X 10-3
cc(STP)
sec, cm2, cm Hg LIP
Thus for a liquid film 1 mil thick, the S02 perme-
ability must be
-------
','
cc(STP), cm
cm2, cm Hg liP'
10.000 X 10-9
sec,
As a point of comparison. silicone rubber, one of
the most permeable membrane materials, has an
S02 permeability at 25° and 100°C of
1500 X 10-9 and 750 X10-9
cc(STP), cm
cm2, cm IIg liP
sec.
respectively. The membrane must also be highly
selective for S02' In order that it have application
for S02 removal from combustion gases, which
typically contain of the order of O. 2% S02' the S02/C02
separation factor (ratio of S02-to-C02 permeability)
of the membrane would have to be of the order of
several hundred. Oxygen 3l1d nitrogen permeabilities
are not a problem since they would be at least 10
times less than CO2, For the membrane to have ap-
plication for more concentrated stack gases such as
2% to 10% S02 which might be produced in a smelter,
the S02/C02 separation factor would have to be at
least 10. Silicone rubber has S02/ CO2 separation
factors at 25° and 100°C of 5 and 2.5, respectively.
2. The membrane must be able to operate at
elevated temperature. In this work an operating
temperature of at least 100°C has been the goal.
3. The membrane liquid must be chemically
inert and nonvolatile at the operating temperature.
4. The membrane must have an effective life
of at least one year and perhaps more.
When the diffusion coefficient (D) of a permeant
is independent of concentration and time, and when
the solubility (S) follows Henry's law, the following
relation obtains
PI'
(D)(S) .
(1)
This equation holds well enough to serve as a guide
in selecting suitable liquids for the gas separation
desired here. Since the diffusion coefficient is of
the order of 10-5 cm2/sec for most solutes in most
liquids, one is primarily concerned with finding
liquids witr high S02 solubilities and low CO2 solu-
bilities. Table I is a summary of the solubility data
obtained on what were considered to be the most
promising membrane liquids. Besides being polar,
all of the liquids were chemically inert and nonvolatile
at 100°C.
From Table I it is seen that the polyethylene
glycols (PEG) have the highest S02 solubility of all
the liquids tested. Thus this material was more
thoroughly characterized. The solubility of CO2 and
S02 in Union Carbide Carbowax 600 polyethylene
glycol was measured from 75° to 150°C, 3l1d the
Yfhese units are used throughout.
results are shown in Fig. 1. The solubility of N2 and
NO at 100°C were found to be 3.2 x 10-3 and 4.8 X 10-3
moles/ t, respectively, at 1 atm of gas over the solu-
tion. It is clear from these data that at 100°C a film
of PEG is sufficiently selective for S02 to be useful
in treating stack gases having of the order of 5% S02'
but is not suitable for very dilute stack gases.
It was found that at 100°C oxygen reacted with
the PEG although the rate of attack was substantially
reduced by excluding light. If it is assumed that for
each mole of oxygen consumed a mole of PEG is
destroyed as a useful membrane material, then the
rate of degradation was 0.23% per day. This rate was
insensitive to the oxygen pressure over the solution
in the range of 10 to 40 mm Eg. Various antioxidants
were added to the solution. The most effective was
Ionox-330 (Shell Chemical Co.) at 0.2%, which de-
creased the oxidation rate to 0.16% per day. Thus it
appears that oxygen degradation is not a serious
problem.
Assuming a diffusivity for S02 of 1 x 10-5, the
permeability of S02 in PEG is 2.95 x 10-6 at 100°C.
Based 011 the flux requirements for S02 stated above,
in order to have an acceptable throughput for S02 a
film of pure PEG would have to be approximately 8iJ.
thick. Immobilizing the PEG will lower the S02 dif-
fusivity by a factor of 2 to 3, and it can be reasonably
assumed that an immobilized film must be less than
3iJ. thick for it to be of practical value.
The first task in fabricating a membrane of PEG
was to find a material that would thicken or gel it.
Numerous water- soluble polymers were tested, but
none was suitable. It was found that a 5% solution of
hydroxymethyl cellulose (HMC), marketed as Cello-
size by Union Carbide, would not dissolve in PEG,
but that a film cast from an aqueous solution of Cello-
size and PEG was clear and reasonably strong. The
HMC film was swollen with PEG, and the PEG not in
the polymer matrix simply adhered to the swollen
polymer film. The permeability of films containing
90%. PEG and 10% HMC was determined. The films
were backed by a I-mil silicone rubber membrane to
avoid forcing the liquid out of the polymer matrix
when a pressure difference was applied. The resis-
tance of the silicone rubber membrane to nitrogen
3l1d carbon dioxide permeation was known, and it was
small compared to the l'esist311Ce of the PEG film.
The composite film was supported on a porous stain-
less steel disk and the permeation measurements
were made in a vacuum system with a pressure dif-
ference of 1 atm across the film. The permeability
was determined by allowing the gas passing through
the film to build up a prcssure in the vacuum system
of approximately 500iJ. Ifg over a measured time
interval. i\t room temperature and 100°C the follow-
ing data we re obtained:
Permcability of CO2
CO2/N2 separation factor
Calculated cliffusivity of CO2
25°C
17x10 9
48
7 x 10-7
100°C
56 X 10-9
"" 15
5 x 10-6
2
-------
TABLE I
Survey of Potential Membrane Materials
Temp. S02 Solubility
Liquid or Solution Tested ~ mol/2., at 1 atm SO? Comments
1. Carbowax 600 100 0.766
(Polyethylene glycol) 80 1. 26
2. Carbowax 4000 (PEG) 100 O. 697
3. Diphenyl sulfoxide 100 .656
4. Two aromatic hydrocarbons made
by GE on laboratory scale
#1 10D .425
#2 100 . 195
5. Versamide #115, polyamide Solution decomposed when
made by General Mills contacted with 802, 1000
6. 5'70 solution of silicone stopcock 100 . 158
grease in 6-ring polyphenyl ether
7. Dioctyl phthalate 100 .246
8. 20 gms phenyl sulfone in 24 cc 100 .328
of tricresol phosphate
9. Phenyl sulfone 135 Very low 802 solubility
10. 2 gms benzene sulfonamide in 100 .298
10 cc tricresol phosphate
11. 2 gms sulfanilamide in 100 .595
10 cc Carbowax 400 (PEG)
12. 10 gms sulfanilamide in 100 0.356
10 cc Carbowax 400 (PEG)
13. I gm saccharin in 10 cc 100 Solution unstable
Carbowax 400 (PEG)
14. 3 gms nicotinamide in 10 cc 80 1. 26
Carbowax 600 (PEG)
15. 20 gms polyethyleneimine 100 802 reacted with solu-
in 10 cc Carbowax 600 (PEG) tion giving H2S
16. Polyphenyl ether 100 0.18
(Monsanto OS 138) 75 .29
17. Dow Corning 704 100 . 17
Diffusion Pump Fluid
18. GE Aromatic hydrocarbon 100 . 15
19. Tricresol phosphate 100 .37
20. Uvinul N-539 100 .25
(GAF)
Thus at room temperature the PEG/HMC film was
quite impermeable, but at 100°C it approached the
permeability of pure PEG. It was concluded that a
suitable film could be made from the PEG-HMC
combination.
a microporous layer of Teflon on a porous support.
A colloidal dispersion of 0.25fl Teflon particles was
Sprayed onto a 75fl thick Solvinert(3) membrane
having 0.25fl pores. The Solvinert film was attached
to a hot plate set at 150°C. A Teflon layer approxi-
mately 25fl thick was deposited. After spraying, the
composite Teflon-Solvinert film was heated in air at
200°C for approximately 1/2 hour. This temperature
was sufficient to drive the wetting agent (Triton X-100)
Obviously it was necessary to have a mechanical
support for the liquid film. It was found that a non-
wetting porous backing could be made by depositing
3
-------
o CONC. 802 vs liT AT I ATM 802
oCONC. C02 vs 1 T AT I ATM C02
1.0
-..
-
W
....J
0
::::!:
C\J
0
U
C\J
0
en
u 0.1
:z
0
u
~
0.01
0.002
0.0025
liT
0.003
Fig. 1 S02' CO2 solubility in polyethylene glycol.
out of the Teflon layer, making it nonwetting. This
microporous Teflon film was completely nonwetting
to PEG at 100°C for an indefinite time if there was no
pressure difference across the PEG.
Clearly, if an immobilized PEG film 3~ thick was
to be made, it would have to be formed directly on the
Teflon support. This required that the surface of the
Teflon be made wetting. It was found that this could
be accomplished by spraying a very dilute PEG-HMC
solution on the Teflon film. A suitable spraying solu-
tion consisted of
0.02% Cellosize QP-4400 (HMC)
0.6% Carbowax 600 (PEG)
99.4% Water
A loading of approximately 50 mg/cm2 of spraying
solution was sufficient to render the Teflon surface
wetting. At this point there was not a continuous
liquid film on the Teflon support. To form such a
film, the wetting Teflon film was dipped into and with-
drawn from a dilute aqueous solution of PEG and HMC.
When the water evaporated, a continuous immobilized
TABLE II
DippIng Solutions Used to Form
Immobilized PEG Films
Solution
cc of 1% Cellosize
(QP-IOOM) Solution
H,O(cm')
Carbowax (grams)
1
2
3
4
5
6
100
100
100
100
100
100
20g Carbowax 600
20g C arbowax 4000
20g Carbowax 4000
20g Carbowax 4000
20g Carbowax 600
20g Carbowax 4000
150
100
125
150
200
250
FEED GAS IN FEED GAS OUT
~ t
CHANNELl!l- - -- -------- ----iSIMMOBILIZED
FOR FLOW~ M LIQUID
OF GASES- S02 PERMEATION MEMBRANE
I~T----------- ILJ
. +
SWEEP GAS OUT SWEEP GAS IN
Fig. 2 Permeation cell used for testing immobilized
liquid membranes.
film of PEG remained on the porous Teflon support.
Six different aqueous PEG/HMC dipping solutions were
used. These are listed in Table II.
Permeation measurements were made on the
films produced from the solutions listed in Table n.
The system shown in Fig. 2 was used. The S02 or
CO2 concentrations in the exiting feed and sweep gases
were measured, and knowing the flow rates of each
stream, the S02 or CO2 flux was calculated. Results
of these runs are presented in Table III. Solution if 3
was found to be optimum. This gave a 3~ film which
had constant permeation properties during a one-
month life test (run if 8, Table III). This is a remark-
able performance for a liquid film 3~ thick.
The major limitation of this membrane is that
when a pressure difference was applied across it, the
backing was wetted 3l1d the film was destroyed. As
discussed below, substantial adv311tages in the system
could be gained if the membrane could be operated
with a pressure difference. At the end of this program
a method of accomplishing this was developed. ':' It was
found that through the use of Cabosil, an extremely
fine grade of silica marketed by Cabot Corporation,
a film of polyethylene glycol could be sufficiently
gelled so as to prevent wetting of two porous backing
materials when pressure differences as high as 25 psi
were applied at 100°C. Suitable backing materials
were Millipore Solvinert and a porous polypropylene
film sold by the Celanese Corporation.
':'The work described below was done by Jerry Meldon,
a summer employee of General Electric Research and
Development Center, who is currently a graduate stu-
dent at the Massachusetts Intitute of Technology.
4
-------
T AI3LE III
Permeation Runs Made on PEG/HMC Films.
Unless Otherwise Noted.
All Runs at 100°C
Dipping Soln.
(Table II )
[sec,
S02 Flux
cc (STP) 1
cm2, cm Bg 6.P.
r
i.sec,
CO2 Flux
cc (STP) I
cm2, em Hg 6.P J
Comment
Run
1
1
1. 5 X 10 - 3 to
3.0X10-3
Variable S02 flux due to variable
mixing of :?O-mil gas films ad-
jacent to membrane. Filled gas
flow channels with 20-mil screen
for subsequent runs
2 5 4.3x10-3
3 6 3. 7 x 10 -3
4 6 3.9 X 10-3 1.5 X 10-3
Proof that film could be made
with Carbowax 4000
S02 fluxes at 70° and 100°C were
equal. Low SOZ/ CO2 separation
factor indicates film was not
continuous
5
2
1. 97 X 10-3 (lOO°C)
2.0 X 10-3 (70°C)
O. 15S X 10-3 (100°C)
0.131 X 10-3 (70°C)
SOz/C02 separation indicates
continuous film was achieved.
Films were probably not con-
tinuous in runs 1-3
6
4
No s02/c02 separation. There-
fore solution #4 does not give
continuous film
7
2
1. 61 X 10-3
After 11 days
2.9 X10-3
0.123 X 10-3
After 11 days
0.51 X 10-3
Flux changes with time. This
was found to be due to seepage
of oil from the constant temper-
ature bath into the membrane
backing. Cell was run i.'1 an oven
in subsequent run
S
3
3.2x10-3
0.22 X10-3
Film was run for 1 month with
no change in fluxes
The procedure finally arrived at by Meldon for
preparing the membranes is as follows:
A film of Cabosil, PEG, and BEC could be cast
onto porous polypropylene or Solvinert by the follow-
ing procedure:
1. 2.4 grams of Cabosil BS- 5 were dispersed
in 50 ml water in a blender at low speed for 2 min-
utes;
A sheet of backing membrane was taped
onto a liS-inch copper plate on a hot
plate at SO°C. A mixture containing half
the above concentrations of Cabosil, BEC,
and PEG was sprayed on, using an air-
brush. When !he water evaporated a
gelled Carbow8...x film was left which was
not continuous. A continuous film was
then cast by dipping the sprayed mem-
brane into the above concentrated mix-
ture. The membrane was then allowed
to dry in air.
2. 12 gi'ams of Carbowax (PEG) 4000 were dis-
solved in SO ml water and blended into (1) at low
speed for 1 minute; and
3. 2 grams of Cello size (BEC) QP-100M were
dissolved in 200 ml water and blended into (2) at low
speed for 1 minute, followed by high speed for several
seconds.
The resulting mixture was 0.7% Cabosil, 3.6%
PEG, and 0.6% BEC, and the rest water.
The resulting film was approximately 1 mil thick.
It displayed the same selectivity and permeability as
5
-------
the films previously prepared with a TeflonjSolvinert
substrate. Furthermore, whereas the films pre-
pared previously without Cabosil wet the backing under
a pressure difference of 1 atm, no wetting with the
Cabosil-gelled films was found with pressure differ-
ences of 25 psi. Assuming the feasibility of casting
thinner films which is reasonable but remains to be
demonstrated, the required S02 flux could be achieved
with this film.
SYSTEMS DEVELOPMENT
Concurrently with the development of a suitable
membrane, the technical and economic feasibility of
several S02-removal systems were studied. Various
methods of providing the necessary driving force for
efficient S02 permeation were investigated. These
included removal of the S02 from the permeate side
of the membrane by both physical and chemical means.
The systems work also included a study of the pos-
sible by-products and processes which could be in-
corporated into the systems planning. Economic
analyses were prepared on those concepts which
appeared to be technically feasible, and rough attempts
at systems optimizations were made. The calculations
were often upgraded as new information was made
available from the membrane research and develop-
ment.
Economic analyses were based on equipment
cost-estimating nomographs and data from several
sources, (4-7) and were updated to 1967 utilizing
Chemical Engineering Magazine's annual plant cost
and equipment cost indices. Total plant cost figures
were estimated by using an over-all multiplicative
factor (of 4.0) on the total purchase price of the major
pieces of equipment used in each process. Operating
cost data included depreciation at the rate of 14'% of
fixed capital charges per year.
Several schemes can be conceived to maintain a
low partial pressure of S02 on the back, permeate
side of the membrane. Chemical means of removing
the S02 would require a reactive sweep vapor to be
employed. The reaction between the S02 and the re-
active medium must be rapid, and, if the reaction
product itself is not commercially attractive, it must
be easily rever,sible. Many cases have been reported
in the literature (see Appendix A) detailing the
thermally reversible reactions between S02 and
various unsaturated olefinic-type compounds. Most
of the cases so described involved a liquid phase,
catalyzed reaction within a tightly sealed pressure
vessel, and resulted in a series of rather complex
polymerization products (polysulfones). Reaction
times were long (several hours) and yields were gen-
erally low. Two papers (Refs. A3 and A4 of Appendix
A) describe the formation of sulfinic acids from sulfur
dioxide and hydrocarbons (both olefins and paraffins)
in the gas phase. However, the reaction takes place
only in the presence of ultraviolet light, and again it
was relatively slow. No information on the ease of
thermal reversibility of these reaction products was
given. Thus, because of the complicated chemistry,
inapplicable system conditions, and relatively slow
reaction rates required by the above ".'eactions, it
was decided to rule out this scheme for removing S02
from the back side of the membrane.
Physical methods of doing this include diluting
and sweeping the permeating gases from the back
surface of the membrane with another gas or vapor,
maintaining the back side of the membrane under
vacuum, or a combination of the above two, i. e.,
using a sweeping medium at subatmospheric pressures.
Major equipment requirements for the above methods
include boilers and condensers if a condensable vapor
is used as the sweep medium, and gas compressors
and associated vacuum equipment if subatmospheric
pressures are involved. Additional equipment would
be required depending on the recovered final product.
Since the driving force for permeation for a given
component is a difference in partial pressure across
the membrane, operating at reduced pressure on the
sweep side would enable a more concentrated S02 gas
to be collected (for a constant total volumetric flow).
Also, the mass of sweep gas required for a given
separation would be lower, which becomes an im-
portant consideration if a condensable vapor is used
for the sweeping stream. In this case, the 502-
containing sweep stream may be concentrated still
further by condensing the vapor. Obviously, additional
effort must be expended to compress and recover these
off-gases.
System analyses soon indicated that a swept sys-
tem would be most feasible, both technically and
economically (see below). Thus, a computer program
was assembled to enable rapid calculation of mem-
brane area and feed and sweep gas compositions and
flow rates. The program used as a model a counter-
current, plug flow operating system (see Fig. 3). In-
put data included initial feed and sweep flow rates
and compositions, an estimated feed exit flow rate
and composition, operating pressures, and the desired
reduction of 502 concentration in the feed gas stream.
A five-component gas stream was considered (502,
CO2, H20, O2, N2). Based on the experimentally
determined (or estimated) permeabilities of these
compounds in the membrane, the fllL"X of each compo-
nent was calculated over an area increment small
relative to the total area required for the given sep-
aration. The mathematical equations required for the
calculations were based on two relations: one de-
scribing the fllL"X of any component across a differential
area element by the definition of permeability [Eq. (2)],
the second being a mass balance about the incremental
area element [Eq. (3)].
N dx ~ (pf pSy) (2)
= 6 x
dA
N (x xo) 5 (y Yi) (3)
An assumption was made [included in Eqs. (2) and (3)]
that the total gas flow rate (both feed and sweep) was
constant over a single area increment. This was
justifiable by the use of a small area increment and
6
-------
Ni I
xi ~
pf
t N'~
--1 Xii
o
N, x
Membrane
'-_.-
:: r
,
'" s.
Iy:
S, Y
ps
Nomenclature
N
S
Feed flow rate (at standard conditions)
Sweep flow rate (at standard conditions)
Feed component concentration
Sweep component concentration
x
y
p
Operating pressure
SubscriPts:
i
initial (inlet) value
final (outlet)value
o
Superscript:
f
feed side condition
sweep side condition
s
"
known value
assumed value
Fig. 3 Model for membrane area calculation.
by the fact that the major portion of the feed gas
streams under consideration was composed of es-
sentially impermeable components.
Equation (3) was used to describe the sweep con-
centration (y) as a function of feed concentration (x)
and known constants (xo, Yi' N, 8). This relation was
incorporated into Eq. (2), which could then be inte-
grated in closed form to enable the calculation of a
pseudo-feed concentration for each component over
the area increment. The flux of each component was
calculated, from which the total flow rate and gas
composition were recalculated at the interior point.
The calculation was continued automatically until the
feed 802 concentration equaled the preset input value,
at which point the other component concentrations
were compared with the described input values. If
they were not in agreement, the program was rerun
with a corrected set of estimated feed outlet conditions.
The area elements were added after each step to ob-
tain the total membrane area requirements.
Different processes were considered for utilizing
the recovered 802, These included various methods
of preparing sulfur, sulfuric acid, and liquid 802,
Only those processes which seemed to have the inherent
advantage of being easily integrated with the rest of
the membrane system were studied in detail.
Because of the major contributions to 802 emis-
sions from the combustion of fossil fuels for power
generation (46% of the total 802 emissions in the U. 8.,
according to a 1966 U. 8. D. H. E. W. estimate), first
consideration was given to the applicability of a mem-
brane system to recover the 802 from power plant
stack gases. The systems detailed generally above
were evaluated, and several variations of each pro-
cess scheme were considered. Large- and medium-
size power plants were investigated. The initial 802
content of the stack gases was assumed to be 0.2 v / 0,
and a design basis of 90% removal of the pollutant was
used. The water vapor and CO2 contents of the stack
gas were assumed to be 6.0'10 and 14.5%, respectively.
Whereas the membrane developed has a high perme-
ability for 802, the membrane separation factors for
802/H20 and 802/C02 are approximately 2.0 and 14.0,
respectively. Thus, because of the unfavorable con-
centration ratios in the stack gas, coupled with the
above separation factors, a high degree of concen-
tration of the 802 in the permeate stream is not pos-
sible. Because of the low initial concentration of S02
in the stack gas (0.2%), the desired rate of removal
(90,%> and the high stack gas flow rate membrane area
requirements were of the order of 7.5 to 8 million
square feet for a 1000 MW power plant, for reasonable
values of the total sweep flow rate and operating pres-
sure conditions.
System analyses indicated that the use of turbine
exhaust steam (at 1.5 inches Hg) as a sweeping medium
for removing the permeating gases from the membrane
interface would be the most economical system. Steam,
at this pressure, would not permeate through the mem-
brane in the reverse direction. Since only a small
fraction of the plant steam consumption would be re-
quired for the membrane system [~2.4% of the total
turbine requirements( 8)], the capital and operating
cost for producing the membrane package steam might
be charged to the power plant itself rather than to the
recovery operation.
The effort which must be expended for merely
separating and concentrating the S02 (to a maximum
of about 10%) was great. Table IV summarizes the
cost data for the two plants considered. A membrane
cost of $1/ft2 was used, which amounted to about 25%
of the total plant investment. A subsequent reaction
TABLE IV
Power Plant Operation
Plant Size Plant Cost Operating Cost
440 MW $2L 66 x 106 0.662 mil/b.."W-hr
($50. 4/kW)
10 0 0 MW $33.42 x 106 0.447 mil/b.."W-hr
(33.4/kW)
7
-------
of the concentrated S02 with a hydrocarbon additive
(based on a potential but commercially untried pro-
cess developed by the now defunct Thiogen Co. ) was
assumed with the final product being sulfur. Credit
is taken for the sulfur produced (at $40 per long ton).
No other schemes were known which might be able to
handle the low pressure, moist, and still relatively
dilute gas stream. Labor costs were not included in
the operating data. Because of the costs involved,
and the uncertainty involved in the above reaction
process, the power plant applications study was ter-
minated, with the conclusion that the present mem-
brane cannot be applied efficiently and effectively to
the separation of S02 from the combustion gases.
Consideration was next given to gas effluents of
higher S02 content and lower flows, such as are
found in various ore-smelting operations (which
account for about 12% of the S02 emissions in the U. S.).
A range of S02 concentrations was considered (4.0%
to 10.0%) and a total stack gas flow of about 20, 000
SCFM was used as a basis. In this case, because of
the low CO2 and H20 vapor contents of the stack gas
(about 6.5% and 3.5%, respectively) a relatively ef-
ficient membrane separation could be performed.
The most economical system was one in which a con-
densable' impermeable sweep vapor carried the
permeating S02 from the membrane, with sub-
sequent condensation of the entire sweep stream to
produce liquid S02 as a salable product. Economic
evaluations were prepared for both atmospheric
and vacuum sweeping of the membrane system.
For comparison purposes, a hydrocarbon (e. g.,
heptane) was considered as the sweeping medium,
BASIS'STACK GAS FLOW = 2.0 X 104 SCFM SC2 CONC. AT STACK INLET = 10%
4.0 S02 CONC. AT STACK OUTLET = 1.0 %
on
b
3.5
""
,....:
LL
< 3.5
""
,....:
LL
~
a::
-------
TABLE V
Smelter Operation to Produce Liquid 802
Stac k S02
Cone entration
(%)
S\veep
Pressure
~
Plant Cost
~-
Operating CO::Jl
- ($)
10
1. 458 X 10'
O..tD3 " 10"
(0.:30 ('lIb S02)
1. .3~1c!: xl 06
O. 4.59 ( 10.
(0. 469/lb SO,)
1.:383 xl0'
0.443 xlOG
(0. 6S,,/lb S02)
10
1.152xl06
0.2
0.318 xl06
(0.22 9/lb SO,)
Basis: 90% recovery of 302 from stack gas.
20. 000 SCFM stack gas flow.
TABLE VI
Hydrocarbon Solubility in PolyetJ1ylene Glycol
(Umon Carbldels Carbowax -tOOO)
Hydrocarbon
(Moles )
~ @ 1 aim HC Pressure
T 0 lOaDe
Solubility
T '" 7SoC
Pentane
0.11
Hexane
0.28
.36
.2.3
Heptane
Cyclomethylhexane
.67
.42
2 - H eptene
Freon 113
. 16
FC-75 (3M fluorocarbon)
2 -Chloroethanol
0.0:3
1. 63 (may be in
error)
Tetrahydropyran
0.61
co,
SO,
.027
1. 34
.017
.55
A search for a sweep vapor to be used in treating
gases from ore-smelting operations was made. Since
at the time of this study the membrane backing became
wet when a pressure difference was applied, atmo-
spheric pressure operation was assumed. This put
very stringent requirements on the physical properties
of the vapor, and a suitable one has not yet been found.
The vapors considered were listed in Table VI together
with their solubilities. With the renewed possibility
of vacuum operation, many new vapors may be tried,
and it may yet be possible to find a sufficiently im-
permeable hydrocarbon vapor for use in the membrane
system.
Alternatively, one can use water vapor as the
sweep vapor at a pressure approximately equal to the
partial pressure of water in the gas to be cleaned.
This pressure can be adjusted to a convenient range
by humidifying the stack gas prior to S02 removal.
This may be done as a by-product of scrubbing the gas
to remove dust and S03'
Due to the high permeability of the membrane to
water vapor, the vapor may pass through the membrane
in either direction depending on the local gradient in
vapor pressure. However, water has the great ad-
vantage of being low in cost and nonpolluting. There-
fore, the low-pressure steam sweeping approach
appears to be the most promising one available at
present. The cost of producing by-product S02 in this
manner should not be very different from the costs as
shown in Table V. A schematic diagram of such a
system is shown in Fig. 6. Experiments on low-pres-
sure steam sweeping are in progress in connection
with another application of membranes and will yield
operational experience with this approach.
0.4 -1.0 % S02
100'C
12,000 Ib/hr. STEAM
100'C
P = 0.2 atm.
DUST
S03 20'C
100'C
~LlNG
~TER
9GOO Ib/hr. H20
LOW PRESSURE BOILER
12 x lOG Btu/hr.
S02
Fig.6 Smelter system schematic.
The fact that hydrocarbons have the high solu-
bilities in polyethylene glycol as shown in Table VI
suggests that they should have permeabilities about
equal to that of S02' This offers the interesting pos-
sibility of using this membrane to remove hydrocarbon
vapors from air or other impermeable gases. A
number of pollution control applications of this pos-
sibility are obvious, such as cleaning air laden with
solvent vapors.
CONCLUSIONS
A number of conclusions from this work can now
be made. The immobilized liquid membrane which has
been developed is substantially more permeable to and
selective for 802 than any other membrane. It is
usable for extended periods at temperatures up to 100°C.
Because the S02/ CO2 separation factor is not sub stan -
tially higher than 10 to 15, this membrane is not suit-
able for removing S02 from combustion gases. How-
ever, with the new advance making possible a mem-
brane which will not break through under a pressure
difference, it appears that a commercially attractive
system might be developed which would be applicable
for treating gases typical of those produced in ore-
smelting operations.
9
-------
RE COMMENDA TIONS
The preliminary system's analyses show that an
economically attractive system for removing S02
from smelter effluents may be available if the liquid
membrane can withstand a pressure difference. Thus,
the work of Meldon should be continued with the goal
of producing a thin, immobilized liquid film with the
same flux as the present 3fl film. A multilayer mem-
brane package should then be attempted. Current
costs of the Solvinert backing are extremely high;
thus a cheaper method of supporting the membrane
must be found. The Celanese Celgard porous poly-
propylene may be a possible backing material.
ACKNOWLEDGEMENT
The authors wish to acknowledge many helpful
discussions and key suggestions of P. H. Kydd and
E. E. Bostick.
F.EFERENCES
1.
W. Robb et al., U. S. Patent 3,335,545.
2.
W. Ward and W. Robb, Science, 156, 1481 (1967).
3.
Solvinert filters are a class of solvent- resistant
microporous membranes manufactured by
Millipore Corp., Bedford, Mass.
4.
Chemical Engineering Cost Files, 1958 to present.
5.
M. Peters, Plant Design and Economics for
Chemical Engineers, McGraw-Hill Book Co.,
Inc., New York (1958).
6.
C. H Chilton, ed., Cost Engineering in the Pro-
cess Industries, McGraw-Hill Book Co., Inc.,
New York (1960).
7.
H. C. Bauman, Fundamentals of Cost Engineering
in the Chemical Industry, Reinhold Publishing
Corp., New York (1964).
8.
Mechanical Engineer's Handbook, 6th ed.
9.
Oil, Paint, and Drug Reporter (April 1968).
10
-------
A10.
All.
A12.
APPENDD< A
BIBLIOGRAPHY OF S02-I-lYDROCARBON HEACTION SYSTEM
AI.
H. J. Backer and J. A. Bottema, "Addition de
Sulfite au Dimethylbutadiene, " Rec. Trav.
Chim., ~, 294 (1932).
A2.
H. J. Backer, J. Strating, and C. M. H. Kool,
"Quelques Sulfones Cycliques, Formees par
Addition D'Anhydride Sulfureux a Des
Butadienes, " Rev. Trav. Chim., 58, 77 8
(1939). -
A3.
F. S. Dainton and K. J. Ivin, "The Photochem-
ical Formation of Sulphinic Acids from Sulphur
Dioxide and Hydrocarbons, " Trans. Faraday
Soc, 46, 374 (1950).
A4.
F. S. Dainton and K. J. Ivin, "The Kinetics of
the Photochemical Gas-Phase Reactions Be-
tween Sulphur Dioxide and n - Butane and 1-
Butene Respectively, " Trans. Faraday Soc.,
46, 382 (1950).
A5.
F. S. Dainton and K. J. Ivin, "The Kinetics of
Polysulfone Formation. I. The Formation
of 1-Butene Polysulfone at 25°C," Proc. Roy.
So c., A 212, 96 (1952).
A6.
F. S. Dainton and K. J. Ivin, "The Kinetics of
Polysulfone Formation. II. The Formation
of 1- Butene Polysulfone in the Region of the
Ceiling Temperature, " ibid., p.207.
A7.
A. H. Frazer, "Thermal Stability of the Co-
polymer of S02 and cis, cis-1, 5-cyclooctadiene, "
J. Polymer Sci., A2, 4031 (1964).
A8.
D. S. Frederick, H. D. Cogan, and C. S.
Marvel, "The Reaction Between S02 and
Olefins. 1. Cyclohexene," J. Am. Chem.
~, 1815 (1934).
Soc. ,
A9.
M. Hunt and C. S. Marvel, "The Reaction Be-
tween S02 and Olefins. II. Propylene," J. Am.
Chern. Soc., ~, 1691 (1935).
C. S Marvel and L. H. Dunlap, "Vinyl Chloride
Polysulfone, " J. Am. Chern. Soc., B, 2709
(1939).
C. S. Marvel and W. W. Williams, "Acetylene
Polysulfones. XI. The Compound ClOH16S02
from 1- Pentynepolysulfone and Some Experi-
ments on Other Acetylene Polysulfones. "
ibid., p.2710.
C. S. Marvel and W. W. Williams, "Polysul-
fones. XII. The Synthesis of 3, 4- and 2,5-
Di-n-propyltetrahydrothiophene-1, 1-dioxides, "
ibid., p.2714.
A13.
A14.
A15.
A16.
A17.
A18.
11
L. L. Hyden, F. J. Glavis, and C. S. Marvel,
"Reaction Between S02 and Olefins and
Acetylenes. VI. Ascaridole as a Catalyst
for the Reaction, " J. Am. Chern. Soc.. ~,
1014 (1937).
L. L. Ryden and C. S. Marvel, "Polysulfones
from Acetylene and S02' " J. Am. Chern. Soc.
~, 2047 (1936).
W. F. Seyer and E. G. King, "Systems of
Sulfur Dioxide and Hydrogen Derivatives of
Benzene, " J. Am. Chern. Soc., ~, 3140
(1933).
H. Staudinger, French Patent 698, 857,
July 11, 1930.
H. Staudinger and B. Ritzenthaler, "Dber
hochpolymere Verbundungen, 104. Mitteil.:
Uber die Anlagerung von Schwefeldioxyd an
Athylen-Derivate, " Ber., 458 (1935).
C. Walling, Free Radicals in Solution, John
Wiley and Sons, Inc., New York (1957).
-------
APPENDIX B
Breakdown of Smelter Operation Cost':'
A. Capital Cost Summary A B C D
Major Equipment
Heat Exchan;:>:ers $163,000 $157,600 $152,600 $134,900
Compressor, drive 21,000 15,000 11 , 000 10,000
Boiler 77,000 77,000 77,000 77,000
Separator 9,000 9,000 9,000 9,000
Drying Tower 10,000 10,000 10,000 10,000
280,000 268,600 259,600 240,900
Total Investment (co. 4.0 x 'Squipment Cost)
$1,120,000 $1,074,400 $1,038,400 $ 963,600
Membrane Cost (?, $1/ft2) 338,000 320,500 345,000 189,000
Total Plant Investment $1,458,000 $1,394,900 $1,383,400 $1,152,600
B. Operating Cost Summary
Raw Materials
Fuel 92,400 92,400 92,400 13,400
Vlater 96,400 92,700 90,000 27,500
Power 70,300 51,100 39,500 94,000
Depreciation (:::: lL10' of TPI) 204,000 195,100 193,600 161,000
...,'0
Taxes or Insurance (.\] nOt of TPI) 29,200 27,900 27,600 23,000
4'.)/0
Total Operating (not including labor) 492,300 459,200 443,100 318,900
Cost (i!lb 302) 0.30 0.46 0.63 0.22
Case 1'.:
Case B:
20,000 SCFM in stack gas ~ 27,000
90% removal of initial stack 302
10% initial 302 concentration
7% initial 302 concentration
SCFM in sweep gas for cases A, B, C
* Basis:
Case C:
Case D:
4% initial S02 concentration
10% initial 302 concentration;
Sweep pressure = 0.2 atm
1~
------- |