X-/EPA
United States
Environmental Protection
Agency
Industrial Environmental Research
Laboratory
Research Triangle Park NC 27711
Research and Development
EPA-600/S2-82-067 August 1982
Project Summary
Supercritical Fluid
Regeneration of Activated
Carbon Used for Volatile-
Organic-Compound Vapor
Adsorption
Christopher P. Eppig, Richard P. de Filippi, and Rosemary A. Murphy
The report gives results of a program
to develop a sound fundamental
technical base for supercritical-fluid
regeneration of activated carbon
applied to adsorption of volatile
organic compounds. The process is
based on using a supercritical fluid to
desorb granular activated carbon
containing adsorbed vapor contami-
nants. The desorption is at high
pressure, and is followed by distillation
of the fluid/adsorbate mixture at
subcritical conditions for removing
contaminants and recovering fluid
solvent for recycle. Key unit operations
characterized were (1) desorption
from the carbon bed. and (2) distilla-
tion of the solvent/adsorbate solution.
The process was characterized for
both leaded and unleaded gasoline
vapors, representing contaminants
from gasoline storage and distribution
facilities; and for ethanol and methyl
ethyl ketone (MEK) vapors, repre-
senting many solvent finishing oper-
ations. Commercial vapor-phase acti-
vated carbon loaded with gasoline
vapor components, ethanol, or MEK
vapor could be completely regenerated
with supercritical CO2 at 1500 psia
and 50°C. A carbon adsorption/
supercritical COa regeneration system
to purify 10,000 scfm of air containing
a quarter of the lower explosive limit
of MEK was estimated to have a price
of $530,000 (December 1980), a
yearly operating cost of $285,000,
and a MEK recovery value (pure MEK
is recovered) of $1,386,000.
This Project Summary was devel-
oped by EPA's Industrial Environmen-
tal Research Laboratory, Research
Triangle Park, NC, 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).
Introduction
Activated carbon has been shown to
be effective for atmosphere contamina-
tion control because of its ability to
adsorb a wide range of organic com-
pounds. The principal drawback of
existing processes using carbon for the
control of organic vapor emissions has
been the requirement for a large carbon
i nventory to compensate for the inabil ity
of the regeneration technique to keep
the carbon working capacity from
declining over the course of multiple
adsorption/regeneration cycles. How-
ever, process studies have shown that
supercritical fluids (fluids in the region
above their critical temperatures and
pressures) can rapidly and effectively
-------
regenerate activated carbon loaded
with a broad range of organic com-
pounds. The use of this regeneration
method could favorably influence the
economics of carbon used in emission
control.
The essential features of the super-
critical fluid process are set forth in a
simplified process flowsheet, Figure 1.
In the adsorption cycle, aircontaminated
with organic vapor flows through the
granular activated carbon bed until
breakthrough. Loaded carbon is then
transferred to the pressure desorption
vessel, and the desorption cycle begins.
Supercritical fluid at conditions which
favor the desorption of the component
adsorbed on the carbon flows through
the carbon bed. The effluent regenerant
stream is rendered subcritical by
alterations of its pressure and/or
temperature, and a vapor/liquid sepa-
ration (distillation) of this stream is
effected to recover the solvent fluid for
recycle. The regeneration system is
efficient because the supercritical fluid
has high solubility for the carbon
adsorbates, favorable mass transfer
properties for rapid desorption, and high
volatility for subsequent separation of
solutes. CO2 is particularly suitable as
the solvent; its critical temperature and
pressure (31.0°C and 72.8 atm) are
economically attainable, it has high
solubilities for organic compounds, it
is fairly dense at process conditions
(therefore power requirements for
compression are reasonable), and it is
non-flammable and non-toxic. Where
desirable, adsorbate can be recovered in
reasonably pure form without a water/
adsorbate separation process.
The objective of this program was to
develop process information and eco-
nomic estimates for this emission
control system. The two key unit
operations characterized were:
(1) The desorption of the carbon bed
with the supercritical fluid.
(2) Distillation of the solvent/adsor-
bate mixture.
The program involved combination of an
analytical and experimental approach to
obtain proper characterization of pa-
rameters for the design of each unit
operation. Specifically, the desorption
operation was characterized for four
test adsorbates. Two of these were
volatile organic compounds (VOCs)
representing important industrial sol-
vents: ethanol and methyl ethyl ketone.
The two other adsorbates were vapors
from leaded and unleaded gasoline. For
each, a range of regeneration conditions
were evaluated using CO2 as the
regenerating solvent.
The process analysis and design
phase included evaluation and selection
of COa/adsorbate separation processes.
Efficiency, low-cost, and component
purities were criteria used in the
evaluation.
Calgon type BPL granular activated
carbon (GAC) was chosen as the carbon
for the experimental work of this
investigation. This GAC was chosen
because it is in widespread use in
commercial vapor-phase applications.
Gasoline and Industrial
Volatile Organic Compound
(IVOC) Adsorption/
Desorption Studies
The experimental adsorption/desorp-
tion studies included a series of
screening runs, followed by evaluation
Carbon Adsorption System
VOC
Air i
Adsi
(Batch Continuous)
''1
Y
1
w y
T f
K
i
/G
iacte/7
rrto
vAe/-
X^\
f
4 >
Purified Air
Out of Adsorber
/
7
~ 1
AC Adsorber (.
IA
\
y
v
X
y
1A(
-/
is
? Adsorber
A
^V ^^\ '^s
i
/)
Carbon
Conveyor
System
> >
v
I
Carbon Regeneration System
(Batch) |
^
j
|l
I
^
\\
1
1
7 \
1
*T] to VOC/SCl
pq 1 > ^onaratiftn
System
Pressure Vessel .
— ,
Regenerant \
1
VJ^KS ' v^\\\ I
Figure 1. Simplified process flow sheet for supercritical-fluid regeneration of GA C used for VOC adsorption.
2
-------
of the Affects of regeneration process
parameters. This was followed by a
series of multi-cycle runs to establish
regenerability in a bench-scale simula-
tion of realistic operation of theadsorp-
tion/desorption process.
Screening Studies
A six-cycle adsorption/regeneration
test on Calgon BPL 12x30 mesh GAC
was carried out. A British Petroleum
Company 89-octane no-lead gasoline
(purchased at a local service station)
was used as the vapor source. Air
saturated with the gasoline at tempera-
tures of 22-24°C flowed at approximately
500 cc/min through a 0.95-cm-diameter
23-cm-long high-pressure pipe packed
with 7.5 g carbon. The adsorptions were
not terminated at a particular break-
through point, but rather after 30
minutes of adsorption. Regenerations
were run immediately after adsorption.
The tests occurred over a 6-day period,
with the column open to the atmosphere
when it was not adsorbing or being
regenerated; the humidity of the air was
not controlled. The carbon loadings,
regeneration conditions, and residual
weights for each cycle are presented in
Table 1.
The first three regenerations were
run with liquid CO2, but above the
critical pressure. The 53 standard liters
(SL) of C02 used in the third run left
more residual on the carbon than did the
96 and 85 SL of C02 used in the first and
second regenerations, respectively. The
fourth regeneration was run with
supercritical CO2 and had the lowest
residual, 0.013 g/g carbon. The working
capacity for this cycle was 26.1 g/100 g
carbon.
The fifth and sixth regenerations
were run withFreon-13(chlorotrifluoro-
methane; Tc = 28.9°C, Pc - 38.2 atm).
The fifth regeneration with 92 SL of
liquid Freon-13 at 24°C had a greater
residual, 0.036 g/g carbon, than did the
first cycle (96 SL of liquid CO2 at 24°C,
residual of 0.018 g/g carbon). The
residual left by 108 SL of supercritical
Freon-13, 0.015 g/g carbon, was very
close to the 0.013 g/g carbon residual
left by the supercritical C02.
These preliminary data showed that
both liquid CO2 and Freon-13 can
effectively regenerate GAC which had
adsorbed gasoline vapors, with C02
leaving a lower residual than Freon
when approximately 100 SL is used.
Supercritical C02 and Freon left very
low residuals when about 100 SL was
used; supercritical CC>2 reduced the
residual of 0.043 g/g, left after three
Table 1. Six-Cycle Adsorption/Regeneration Carbon Calgon BPL 12 x 30 Mesh
Gasoline = B.P. No-Lead, 89 Octane (7.15 g GAC in Column)
Cycle Loading & Residual
Regeneration Conditions
1 Loading 1.98 g 0.277 g/g carbon
Residua/ 0.13 g 0.018 g/g
2 Loading 1.99 g 0.278 g/g
Residual 0.2O g 0.028 g/g
3 Loading 1.96 g 0.274 g/g
Residual 0.3J g 0.043 g/g
4 Loading 1.96 g 0.274 g/g
Residual 0.09 g 0.013 g/g
5 Loadingl.84 g 0.257 g/g carbon
Residual 0.26 g 0.036 g/g
6 Loading 1.93 g 0.270 g/g
Residual 0.11 g 0.015 g/g
T = 24° C Liquid
P= JSOOpsi COz
96 SL @ 10 SL/min
T = 26°C Liquid
P= 1600 psi COz
85 SL @ JO SL/min
T= 24° C Liquid
P= 1600 for 20 SL CO2
P~WOOfor33SL
53 SL@ 10 SL/min
T ~ 50°C Supercritical
P= 1600 psi CO 2
102 SL@ 10 SL/min
T = 24°C
P = 1200 psi
92 SL @ 1O SL/min
T ~ 50°C Supercritical
P = 1200 psi CF3CI
108 SL @ 10 SL/min
cycles with liquid CO2, to 0.013 g/g.
Preliminary testing was also done
with an experimental GAC from Amoco,
GX-31. Air saturated at 23°C with the
BP no-lead gasoline flowed for 30
minutes through a 1.7-cm diameter 26-
cm-long high-pressure pipe packed
with 7 72 g of Amoco GX-31 pelletized
16-32 mesh carbon. The regeneration
was run with 110 SL of supercritical
CO2 at 50°C and 1400 psi.
The working capacity was 0.60 g/g
carbon. On the basis of g gasoline/cc
carbon, the capacity of the Amoco
carbon was 0.6 g/g carbon x 0.248 g
carbon/cc = 0.15 g/cc, whereas the
Calgon carbon was 0.25 g/g carbon x
0.455 g carbon/cc = 0.11 g/cc, a factor
of 1.4 difference. Results for the
adsorption/SCF regeneration on GX-
31 were:
Loading and
Residual
Loading: 502 g;
0.650 g/g carbon
Residual. 0.39 g;
0.051 g/g
Regeneration
Conditions
T ~ 50°C
P = 1400 psi Super-
critical CO
110SL@ 10 SL/min
When the activated carbon column
was removed from the high-pressure
desorption apparatus after regenera-
tion, the glass fiber plug at the desorption-
effluent end of the column was black-
ened with fines from the carbon. Work
was terminated with the GX-31 carbon
at that time.
Toluene comprised up to 19 wt
percent of the heel formed during a
1000-cycle test reported in the literature
(Manos and Kelly (1977)). Thus, it was
felt that the ability to desorb it from GAC
would be a good indication of the ability
of the near-critical fluid regeneration
technique to maintain high carbon
working capacity. Additionally, toluene
is widely used as an industrial solvent,
in paper-coating operations, for instance.
Thus, toluene was selected for a
preliminary pure component VOC study.
Air saturated at 23°C with toluene
flowed through the 0.9-cm-diameter
column packed with 7.47 g GAC. Three
regenerations were run at the conditions
given below:
-------
Cycle
Loading & Residual
Regeneration Conditions
1 Loading: 1.78 g; 0.238 g/g carbon
Residual: 0.07 g; Og/g
2 Loading: 1.64 g; 0.220 g/g
Residual: 1.66 g; 0.222 g/g
3 Loading: 1.66 g; 0.222 g/g
Residual: 0.22 g; 0.029 g/g
T = 24°C Liquid
P = 1300 psi CO2
150SL@10SL/min
T = 24°C Gaseous
P = 500 psi CO 2
101 SL@10SL/min
T = 24°C Liquid
P= 1200 psi C02
138SL@10SL/min
The first and third regenerations with
liquid C02 were effective in removing
toluene from the GAC. The gaseous
COa was totally ineffective for desorbing
toluene.
The following grid generalizes these
results with those for the gasoline vapor
desorptions:
Hydrocarbon
Desorption
from
Activated
Fluid State Carbon
C02 gaseous totally
T = 24°C ineffective
P = 500 psi
CO2, near critical very good
Freon-13 T = 24° C
C02 supercritical excellent
Freon-13 T - 50°C
Thus, preliminary results indicated that
COa and Freon in the near-critical or
supercritical states were effective in
regenerating activated carbon loaded
with hydrocarbons, while gaseous CO2
was totally ineffective.
Process Studies with Gasoline
Additional experiments were done
to determine what conditions of temper-
ature and pressure were suitable to use
as regeneration conditions for the
multicycle adsorption/regeneration
cycles. Four of the 7-g capacity columns,
packed with the BPL 12x30 mesh GAC,
were loaded with BP 89-octane no-lead
gasoline vapors. Loadingwasallowedto
continue until column effluent hydro-
carbon concentration was 80 percent of
that in the column influent. The columns
were then regenerated with super-
critical COa at different conditions.
Regeneration Conditions
Pressure, Temperature,
Column atm °C
1
2
3
4
103
103
207
207
25
50
50
25
Approximately 180 SL of C,Q2 was used
for each regeneration; the flow rate was
15 SL/min. Three adsorption/regener-
ation cycles were run; results are
shown in Table 2. After cycle 1, columns
1 and 4 had residuals of 0.05 g; whereas
columns 2 and 3 had residuals of -0.01
and -0.04 g, respectively. The negative
residuals suggest that the virgin carbon
was initially slightly contaminated with
volatile species, or that the water
loading baseline varied slightly. After
cycle 2, columns 1 and 4 had residuals
of 0.07 g, and columns 2 and 3 had
residuals of -0.01 and -0.03 g, respec-
tively. After cycle 3, columns 1 and 4
had residuals of 0.11 and 0.06 g,
respectively, and columns 2 and 3 had
residuals of 0.03 and 0.02 g, respectively.
These data indicate that over three
cycles, regenerations run at 207 atm
are not significantly different than those
run at 103 atm, and that regenerations
run at 50°C are more complete than
those run at 25°C.
Based on these results, regeneration
conditions of P = 103 atm and T - 50°C
were selected for the multicycle adsorp-
tion/regeneration experiments.
Process Studies with
Industrial VOCs
After the selection of ethanol (EtOH)
and methyl ethyl ketone (MEK) as the
two industrial volatile organic com-
pounds (IVOCs) for study, preliminary
experiments were done to show that
GAC loaded with these compounds
could be regenerated by liquid C02. The
procedure employed in the gasoline
adsorption/regeneration cycles was
used.
Air saturated with the VOC at 24°C
flowed at approximately 800 cc/min
through a 1.7-cm diameter, 16-cm long
high-pressure pipe packed with 15.46 g
of Calgon 12x30 mesh BPL carbon. The
adsorptions were not terminated at a
particular breakthrough point, but
rather after SOminutesof adsorptionfor
the MEK, or after 60 minutes for the
EtOH. Regenerations were run immedi-
ately after adsorption, with the column
open to the atmosphere when it was
being transferred between the adsorp-
tion and regeneration equipment; the
humidity of the VOC carrier air was not
controlled.
The MEK was Union Carbide com-
mercial grade (99+ percent) from a 55
gal. drum; the EtOH was Publicker
Industries Company 200 proof "Pharm-
co" brand. The carbon loadings, regene-
ration conditions, and residual weights
for each cycle are given in Table 3.
All four regenerations were run with
liquid COa at 109 atm. In each cycle, the
VOC was completely removed from the
column; over four cycles there was no
indication of residual buildup. The
working capacity at 24°C of 0.21 g
MEK/g carbon was approximately 80
percent of the capacity for no-lead
gasoline vapors (0.26 g BP 89-octane
no-lead/g carbon); the 0.16 g EtOH/g
carbon working capacity was about 60
percent of the capacity for the no-lead
gasoline vapors. Thus, preliminary
studies indicated that liquid C02 was
v.ery effective as a regenerant for GAC
loaded with MEK or EtOH.
Process studies and multicycle re-
generations were run using at least 100
SL CO2 for the 8-g carbon bed (about
22.5 IbCO2/lb GAC). Because it isdesir-
able to operate with minimum solvent
usage, two tests were run with MEK at
lower COa usages to determine if lesser
quantities would give adequate regene-
ration. The results are given in Table 4:
14 SL (3.4 Ib C02/lb GAC) gave essen
tially the same capacity recovery as the
larger CO2 volumes.
Multicycle Tests
Multiple-cycle adsorption/regene-
rations were run on a 0.9-cm-diameter
by 23.5-cm-long column packed with
Calgon BPL 12x30 mesh GAC. Equili-
brium vapor of a BP 89-octane unleaded
gasoline in air was loaded at 200
cc/min onto the GAC column; loading
was terminated when the hydrocarbon
content of the column effluent reached
-------
Table 2. Three-Cycle Adsorption/Regeneration
Column
Weight
Carbon
Gasoline Loading
(9/9!
Regeneration
Pressure/ Temperature
Residual
After Regeneration
1
1
2
3
4
Column
Column
7.41 g
7.66
7.39
7.20
1.80g 0.243
1.93 0.252
1.86 0.252
1.76 0.244
Gasoline Loading
2 (g/g)
1500psi/25°C
1500/50
3000/50
3000/25
Regeneration
Pressure/ Temperature
0.05 g
-0.01
-0.04
0.05
Residual
After Regeneration
2
1
2
3
4
1.86g
1.97
1.86
1.81
0.251
0.257
0.252
0.251
1500psi/25°C
1500/50
3000/50
3000/25
0.07g
-0.01
-0.03
0.07
Gasoline Loading
3 (g/g)
Regeneration
Pressure/ Temperature
Residual
After Regeneration
3
1
2
3
4
1.88g
1.91
1.82
1.80
0.254
0.249
0.246
0.250
1500psi/25°C
1500/50
3000/50
3000/25
0.1 1g
0.03
0.02
0.06
Carbon - Calgon BPL 12 x 30 mesh
Gasoline = B.P. no-lead 87-octane
Adsorption column influent flow rate =180 ml/min
Adsorption termination when effluent hydrocarbon concentration = 80% of influent concentration
Desorption CO2 flow rate =15 SL/min
Desorption CO2 volume = 180 SL
Table 3. Four-Cycle Absorption/Regeneration
(carbon = 15.46 g Calgon BPL 12 x 30 Mesh)
Cycle VOC Loading and Residual
Regeneration Conditions1'
1 MEK Loading 3.28 g 0.212 g/g carbon T = 23°C
Residua 1-0.05 g 0 g/g
2 MEK Loading 3.33 g 0.215 g/g
Residual-O.OSg 0 g/g
3 EtOH Loading 2.43g 0.157 g/g
Residual -0.09 g 0 g/g
4 EtOH Loading 2.44 g 0.158 g/g
Residua/ -006 g 0
9/9
P = 1600 psi
127 SL@ 10 SL/min
T = 25°C
P = 1600 psi
137 SL @ 10 SL/min
T = 24°C
P = 1600 psi
124 SL @ 20 SL/min
T = 24°C
P = 1600 psi
122 SL @ 20 SL/min
Liquid COZ
Liquid COS
Liquid C02
Liquid CO2
"For all regenerations, the last approximately 30 SL are Psat <** 900 psi.
Table 4. Reduced COz Usage Regeneration Tests
After MEK Loading
After Regeneration
Cycle
1
2
Column
Mass
438.41 g
438.35
MEK
mass
2.84g
2.78
g/g Carbon
0.376
0.368
SL COz
20
14
Column
Mass
435. 83 g
436.06
MEK
Residual
0.26g
0.49
Residual
g/g Carbon
0.034
0.065
Mass Carbon fCalgon BPL 12x30 Mesh}: 7.56 g
Mass Carbon and Column : 435.57g
Column Loaded with Equilibrium MEK- vapor in Air at 23°C (~ 1 x 105 ppmv)
Regeneration Conditions: T = 50°C
P = 100 atm
30 percent of the Column influent, as
determined with the FID. After loading,
the column was regenerated with CC>2
at 103 atm and about 50°C; approxi-
mately 100 SL of COa was used at a flow
rate of 15-20 SL/min. After regenera-
tion, constant humidity air flowed
through the column at 2.5 L/min for
about 1 min to desorb residual COz.
In the 25-adsorption/regeneration-
cycle test for unleaded gasoline, the
initial mass of the column and the virgin
GAC (equilibrated with constant humid-
ity air) was 435.42 g After 25 cycles,
the mass was 435.42 g, i e., the carbon
was 100 percent regenerated. The
working capacity of the carbon was 0.24
g/g carbon; using as little as 89 SL C02
appeared to completely regenerate the
carbon.
Another series of multiple-cycle
adsorption/regenerations were run
using BP 89-octane leaded gasoline.
The same adsorption procedures as
with the unleaded gasoline were
employed. The air/leaded gasoline
stream had a smaller flow rate: about
150 vs 200 cc/min for unleaded.
Similar regeneration conditions were
used to desorb the leaded gasoline as
for unleaded.
-------
In the 25-adsorption/regeneration-
cycle test for leaded gasoline, the initial
mass of the column and the virgin GAC
was 436.30 g. After 25 cycles, the mass
was 436.28 g; i.e., the carbon was 100
percent regenerated (the experimental
precision is ±0.02 g). There were
gasoline residuals on the carbon after
regenerations 6 and 11; cycles 6 and 11
had relatively high initial gasoline
loadings, 0.249 and 0.254 g/g, respec-
tively. The working capacity of the
carbon was 0.25 g/g carbon.
After the 25th cycle of the leaded
gasoline vapor testing, additional
adsorption/regeneration cycles were
carried out with ethanol vapor. Equili-
brium vapor of 200 proof ethanol in air,
about 6x104 vol. ppmat 23°C, flowed at
1.5 SL/min through the GAC column.
The loading was terminated when the
ethanol content of the column effluent
reached 80 percent of the column
influent, as determined with the FID.
After loading, the column was regen-
erated with COa at 103 atm, and
temperatures slightly lower than used
for the gasoline vapor regeneration, 41
to 51°C. The carbon was 100 percent
regenerated after each cycle; the
working capacity was about 0.32 g/g.
Thus, after 50 adsorption/regenera-
tion cycles, 25 cycles with leaded
gasoline vapor adsorbate and 25 cycles
with ethanol vapor adsorbate, the
carbon retained 100 percent of its virgin
working capacity.
Further testing on this carbon bed
included multicycle tests with MEK.
Equilibrium vapor of Doe and Ingalls
Industries Co. industrial grade MEK in
air, about 1.13 x 10s vol. ppm at 22°C,
flowed at 0.8 SL/min through the GAC
column. The loading was terminated
when the MEK content of the column
effluent reached about 80 percent of the
column influent. After loading, the
column was regenerated with COa at
103 atm, and temperatures of 40-64°C.
A residual of 0.02 g, which could not
be removed with up to 235 SL C02 (at
50°C and 103 atm), was left on the
carbon. After the first four cycles, it was
discovered that the MEK had attacked
the rubber gaskets in the flowmeter
upstream of the carbon bed; some of the
rubber was probably adsorbed on the
carbon and caused its 0.02 g increase in
mass.
While 100-110 SL C02 at 50°C and
103 atm could completely regenerate
carbon loaded with gasoline vapors or
EtOH, approximately 150 SL COa was
necessary to regenerate the MEK-
loaded carbon. With 120 SL CO2
regeneration, or about a 0.07 g residual,
the carbon working capacity was about
0.33 g MEK/g carbon; with 150 SL of
regeneration, or a 0.02 g residual, the
carbon working capacity was approxi-
mately 0.34g MEK/g carbon.
Figure 2 shows the working'capacity
as a function of cycle number over the
entire 75-cycle test.
Adsorbate/Regenerant
Separation
The effluent stream from a carbon
column undergoing regeneration with a
supercritical fluid must have nearly
complete separation of the adsorbate
from the regenerant so that the regen-
erant may be recycled. These paragraphs
discuss separation of COa regenerant by
distillation from the different test
adsorbates: gasoline vapor components
(GVCs), EtOH, and MEK.
COz/GVC Separation
The feed stream to a COa/GVC
separation system would be at typical
carbon regeneration temperature and
pressure: 100 atm and 50°C. The
stream would be between 1 and 10
mole percent hydrocarbon in COa (most
likely at the low end of that range). The
GVCs would be approximately 50 mole
percent n-butane, 30 mole percent
isopentane, 2 mole percent n-hexane,
with the remainder being light paraffins,
olefins, and aromatics (Cs-Ca). The
heavy key component is n-butane, and
design calculations were done on this
basis.
Vapor/liquid equilibrium data for the
COa-n-butane system at 34 atm* are
available (Olds et al., 1949; Poettman
and Katz, 1945).
To achieve a separation with
XCO2 bottoms < 0.01
XCO2 distillate > 0.99,
a graphical calculation using the
McCabe-Thiele method indicated that
the minimum number of theoretical
stages is seven.
A column design with a low reflux
ratio is feasible because of the high COz
"Slightly below the maximum pressure at which
COz-n-butane separation can be effected -- 37 atm,
the critical pressure of n-butane
50 i
40-
c
o
I 30 i
8
4*.
\
|
£
O)
20-
10-
O Leaded Gasoline Vapor Components
O Ethanol
A Methyl Ethyl Ketone
10
20
30
40
50
60
T
70
8C
Cycle Number
Figure 2.
Carbon working capacity vs. cycle number for 75-cycle test with supei
critical COz regeneration.
-------
feed concentration (XCO2> 0.9)and high
relative COz-n-butane volatility. The
feed stream at 100 atm and 50°C is
flashed to 34 atm, and in so doing is
about 75 percent vaporized, if the
mixture's thermodynamic properties
are assumed to be equalto those of pure
CO2. A D/V (= total overhead product
rate/CO2 phase rate) = 0.6, or L/V (=
slope of upper operating line) = 0.4
when XCC>2 distillate =1.0 gives nine
theoretical stages. The external reflux
ratio at the top of the column is 0.67.
The column has the feed introduced
between stages 7 and 8 and has a par-
tial condenser. The distillate (vapor at
0°C and 34 atm) is compressed to 100
atm and 50°C in a two-stage compres-
sor. FigureS isaflowsheetforthisdistil-
lation system.
The column design cannot be opti-
mized with respect to reflux ratio versus
total number of stages (column height)
until the height equivalent to a theo-
retical plate (HETP) is known. The HETP
is preferably obtained by experiment;
although there are correlations in the
literature (e.g., Murch, 1953), that can
be used for preliminary design esti-
mates. One would expect low mass
transfer resistances between the liquid
and vapor phases at the temperatures
and pressures under consideration, and
hence relatively small HETPs.
The energy requirement of the sample
nine-stage system shown in Figure 3
was estimated to be 4000 Btu/lb-mol
feed including the recycle compressor.
This assumes that the compressor and
cooler are run electrically and that the
reboiler is steam heated; the conversion
factor between electrical and steam
energy is 0.8/0.34.
Alternatively, a vapor-recompression
distillation system can be used for a
more energy-efficient CO2/butane
separation. Figure 4 is a flowsheet for
this system. In this scheme, the entire
overhead vapor stream V is fed to
compressor Cl where it is compressed to
the recycle pressure (100 atm). A
portion of the heat of compression of
compressor discharge stream 1 is used
to heat stream 1 in partial reboiler PRBI.
The compression ratio, compressor
efficiency, and heat exchanger flow
rates and efficiency determine the
maximum temperature which can be
obtained in the reboiler. If this reboiler
temperature is insufficient to achieve
the desired bottoms purity, stream B
would have to be flashed to a lower
pressure and fed to another separation
system.
Reflux at the top of the column is
provided by splitting stream 2, cooling
split-stream 2a in reflux exchanger RE
and flashing back to column pressure by
means of valve X. For instance, if stream
2 were at 100 atm and 50°C,2a could be
cooled to about 30°C with cooling water
in RE, and upon flashing to 34 atm,
would be about 70 percent liquefied.
The recycle compressor cost would
likely be a sizable fraction of the fixed
capital investment for this vapor-
recompression system, and this an
optimal design would tend toward a
relatively small reflux ratio and more
theoretical stages.
COz/IVOC Separation
To carry out the design calculations
for a system for removal and recovery of
MEK or EtOH, performance of vapor/
liquid separations for the binaries
CO2/MEK and C02/EtOH are required.
Detailed design calculations were done
for an MEK adsorption/regeneration
process. In these paragraphs, the bases
for the CO2/MEK separation design are
given, plus the fundamental information
needed for a COa/EtOH vapor/liquid
separation design.
No experimental data for the C02/
MEK system are available, and only
fragmentary information has been
published for CO2/EtOH. Thus, two-
component phase-equilibrium relation-
ships had to be used to predict the
required data from thermodynamic
properties of the pure components. A
common method, the Lewis and Randall
rule, failed to give reasonable agreement
with the published C02/EtOH data.
The Peng-Robinson modification of
the Redlich-Kwong equation of state
was shown recently to predict vapor-
liquid equilibria successfully near the
critical point. This approach was tried
for both C02/EtOH and C02/MEK, and
C02 Recycle: XCo, > 0.99
COa Recycle
Compressor
100 Atm
50 C
34 Atm
(T=0°C)
V
150°C
34 Atm
D/V = 0.6
L/V = 0.4
L'/V'=1.15
Figure 3.
D-0.89F mols
B*0.11 F mols
V = 1.48 F mols
L =0.59 F mols
V' = 0.73F
L' = 0.84 F
COi-butane separation flow sheet.
7
-------
Figure 4. Vapor-recompression dis-
tillation for COz-butane
separation.
it was found to adequately model the
known C02/EtOH data. Thus, it was
used for the C02/MEK predictions.
Calculations for COa/MEK were done at
both 800 and 300 psia, in preparation
for the design case.
The COa/MEK vapor/liquid equili-
brium data were then used to fix the
separation system design. The results
show:
(1)feed composition (desorber
effluent) of 0.95 C02 mol fraction
(0.05 MEK mol fraction), at a
pressure of 102 atm (1500 psia)
and a temperature of 50°C (122°F),
a C02 purity of 0.999995 can be
achieved at a distillation pressure
of 54.4 atm (800 psia), repre-
senting 90.6 mol percent of the
feed. The bottoms composition is
0.49 CO2 mol fraction. Five recti-
fying stages and one stripping
stage are needed, and the reflux
ratio is 0.06.
( 2) feed composition of 0.49
CO2 mol fraction, at 54.4 atm (800
psia) and 50°C (122°F), an over-
head product of 0.991 CC>2 mol
fraction is obtained at 20.4 (300
psia), representing 38 mol percent
of the feed. The bottoms composi-
tion is 0.18 CO2 mol fraction. Two
stripping stages are needed, with
no reflux requirement.
Process Design
These paragraphs describe the process
design of a full-scale system to treat
10,000 scf m of air with 4500 ppmv MEK
vapors. System operation is described
first; the design of important components,
such as adsorption and desorption
vessels, is then presented, followed by
the capital and operating costs for the
system.
System Operation
The adsorption/desorption system
described here is based on the use of
supercritical CO2 to remove adsorbed
MEK vapors from GAC. The adsorption
and desorption vessels are separate
because of significant differences in
design criteria; thus, the carbon is
transported from one to the other. The
desorption cycle includes separation of
MEK dissolved in the CO2 effluent by
distillation and flash units at several
pressure levels, to minimize the loss of
CO2, to recoverthe maximum amount of
MEK adsorbate, and to maximize the
purity of CO2 return to the desorption
vessel.
The adsorption cycle operation is
shown schematically in Figures 5A-D.
Carbon adsorption vessels CA1 and
CA2 and ancillary carbon hoppers H1-
H4 are connected to carbon regeneration
vessel CRV and hoppers H5 and H6 by
tubular conveyor TC. The MEK-contain-
ing air stream, flowing at 10,000 scfm
with a head of 8 in. water, is alternately
cycled over a 15-minute adsorption
period between CA1 and CA2.
Figure 5A shows a 10-min phase in
the adsorption cycle. Air flows through
carbon bed CA1, regenerated carbon
from H6 is transferred by TC to H1, and
MEK-loaded carbon in H4 (previously in
CA2) is moved through TC to H5. At the
end of 10-mmutes, the system is as
shown in Figure 5B: H1, H3, and CRV
contain regenerated carbon. In the next
21/2 minutes, CRV is depressurized, the
carbon is transferred to H6 and the
carbon from H3 is moved into CA2. In
the following 21/2 minutes, CRV is filled
with the carbon in H5 and repressurized.
The system is then as shown in Figure
5C. The air flow is then switched to CA2,
and the cycle begins again, as shown in
Figure 5D.
Figure 6 is a flow schematic of the
desorption and solvent recovery portion
of the system. The CRV shown in Figure
6 is the same as that shown in Figure 5.
The desorption cycle operates as
follows. Desorption at 100 atm and
50°C is carried out with continuous C02
flow for 10 minutes. At the completion
of regeneration, CRV is first depressur-
ized into FV-1, the primary distillation
unit, until CRV pressure falls to 54 atm.
At that point, depressurization continues
with CO2 flow into FV-2, the secondary
flash vessel, until CRV pressure falls to
20 atm. Finally, CRV is depressurized to
2 atm with CC»2 flow directed to low-
pressure accumulator LPA. Following
this, the remaining C02 is vented, and
the carbon is discharged to hopper
H-6.
CRV is then charged with carbon from
hopper H-5. During the CRW down-
time, high-pressure accumulator HPA is
recharged from about 50 atm to 100 atm
by continuing operation of compressor
C-1. After the carbon charge is received
by CRV, the valve between HPA and the
CRV is opened, the pressure equalizes
at about 50 atm, and the valve is closed.
At that point, the CO2 flow from
compressor C-1 is redirected to pressur-
ize the CRV to 100 atm as the desorption
cycle is initiated.
During the desorption cycle, the
desorber effluent, stream No. 1, leaves
CRV at 100 atm and 59°C. It is flashed
through a valve to 54 atm and 20°C
(stream No. 2) to still FV-1, a small
packed column with the equivalent of
five theoretical plates in the rectifying
section, and one theoretical plate in the
stripping section. The overhead from
FV-1 is fed to the suction side of
compressor C-1 to increase the pressure
to 100 atm, with a concomitant tem-
perature rise to 65°C.TheC-1 discharge
is then fedtoreboilerFV-1 to provide the
heat of vaporization, in a standard vapor
recompression cycle. Approximately 6
mole percent of this stream is fed back
to the still as reflux by being cooled to
27°C, and being flashed to 54 atm
where its quality becomes about 95
percent liquid, 5 percent vapor. At this
point, a bleed of noncondensable gases
may be taken, if necessary (FV-4), and
stream No. 15, the column reflux, is
reintroduced into FV-1. The remainder
of the recompressed C02 (stream No. 5)
returns to the column, undergoing
slight cooling from 57°C to 50°C, where
it is reintroduced to CRV as clean
makeup solvent.
The still bottoms from FV-1 leaves at
54 atm and 50°C, containing about 49
mole percent C02. It is directed to flash
vessel FV-2, maintained at 20 atm and
21 °C. The overhead from FV-2 is fed to
compressor C-2, along with the re-
quired C02 makeup (availableat 20atm)
and the discharge from compressor C-3,
which empties low-pressure accumulator
LPA from a pressure of 2 atm to atmos-
pheric pressure. The C-2 discharge
(stream No. 8), at 54 atm and 82°C, pro-
vides the heat of vaporization for the
flash in FV-2. The cooled 54-atm ef-
fluent (stream No. 9)isthenaddedtothe
8
-------
CRV
H6
HI
CA1
CA2
CRV
Figure 5A
Figure 5C
HI
CA1
H3
CA2
H5
CRV
\H6
CA1
Figure SB
Figure 5D
Y///A Denoted MEK-loaded Carbon l&Sxa Denotes Regenerated Carbon
Figure 5. Adsorption cycle operation schematic.
FV-1 feed to ensure maximum removal
of MEK from the CO2.
The bottoms liquid from flash vessel
FV-2 is discharged into small flash
vessel FV-3, which operates at the
pressure of LPA, cycling between 2 atm
and atmospheric pressure. Heat is
added to provide essentially complete
stripping of COa from the final MEK
discharge, using an electric heater.
Thus, the MEK discharge is essentially
free of C02. The FV-3 overhead is
directed to LPA.
System Design
The system is designed to handle
10,000 cfm containing 25 percent of the
lower explosive limit (LEL) of MEK,
giving a feed concentration of 4,500
ppm. The capacity of GAC for MEK at 1.1
x 105 ppm feed was measured experi-
mentally as 0.34 Ib/lb.
For design purposes, it was assumed
that the carbon would adsorb 75
percent of this level (0.26 Ib/lb carbon),
since the MEK adsorption isotherm is
relatively concentration independent at
these concentration levels.
A 15-minute adsorption/desorption
cycle was chosen by running optimiza-
tion calculations with regard to vessel
costs.
The above parameters fix the carbon
bed size at 480 Ib GAC, or about 18 ft3 A
figure of 20 ft3 was used to fix the bed
dimensions of both the adsorption and
desorption vessel. In the adsorber, it
was assumed that the bed depth would
be about 4 in., utilizing a square vessel
with a dimension of about 7 ft on the
side. The desorber was dimensioned to
give reasonable pressure-vessel pa-
rameters, and was taken as a 2 ft
diameter, 7 ft long vessel.
Sizing of the flash vessels was based
on established capacity correlations for
packed or open vessels. Sizing of the
high-and low-pressure accumulators
was based on the gas volumes involved.
Figure 7 shows the material balance
on COa flow around the CC>2 recovery
system.
Process Economics
Vendor's estimates or estimates
based on standard correlations were
used to obtain costs of the individual
equipment items. Following the cost-
estimating practice described by Guthrie
(1969), installation factors were used to
9
-------
Tubular Conveyor
From Adsorption
System
CRV
Tubular Conveyor
To Adsorption
System
MEK Discharge
Process Conditions
Pressure PSIA
Temperature °C
Temperature °F
Stream No.
123456789 10 11 12 13 14 15
1500 900 800 1500 1500 300 300 800 800 14.7 to 30 14.7 to 3O 300 300 .1500 800
50 20 50 66 57 21 47 82 60 16 16 82 -18 50 18
122 68 122 150 135 70 117 180 140
60
60 180 0 122
65
Legend
C-1 - Compressor
C-2 - Compressor
C-3 • Compressor
CA V - Carbon Regeneration Vessel
FV-1 • Flash Vessel
FV-2 - Flash Vessel
FV-3 - Flash Vessel
H-5 - Carbon Feed Hopper
H-6 - Carbon Discharge Hopper
HP A - High Pressure Accumulator
LPA - Low Pressure Accumulator
FV-4 - Flash Vessel
RE - Reflux Exchanger
Figure 6. Supercritical COz VOC carbon regeneration system.
give so-called module costs for each
item; these were summed to give the
installed costs for the system. As
recommended by Guthrie, a 10 percent
contingency and a 3 percent con-
tractor's fee were added, for a system
cost of $530,000. These costs were for
December 1980.
System operating costs are shown in
Table 5. These total about 2C/lb of GAC
regenerated.
Capital costs may be compared with
those published for other carbon regen-
eration systems. The highest-cost
system among competitive processes is
about $450,000. That system, which
involves steam regeneration, produces
a mixture of MEK and water, which has
a low value as solvent to be recycled to
the industrial operation. The MEK
discharge from the COa system is
essentially free of water, and thus may
be valued at close to makeup MEK
prices.
On this basis, it is possible to
calculate an approximate payout for the
system, assuming full value for recovery
of MEK. The calculation shows that the
annual MEK value, reduced by the
annual operating costs of the entire
carbon regeneration system, comes to
$1,100,900/year. Dividing this into the
capital cost of $530,000 gives a rough
payout time of approximately 0.48
years. Thus, if good-purity MEK can be
recovered from the system, the carbon
regeneration process provides an
opportunity for a rapid payout source of
makeup industrial solvent.
References
Guthrie, K.M., "Capital Cost Esti-
mating," Chem. Eng., 76 (6), 114
(1969).
Manos, M.J., and W.C. Kelly, "Control
Characteristics of Carbon Beds for
Gasoline Vapor Emissions", EPA-600/2-
77-057 (NTIS PB 268650), February
1977.
Murch, D., "Height of Equivalent
Theoretical Plate in Packed Fractiona-
tion Columns — An Empirical Correla-
tion," Ind. and Eng. Chem., 45, 2616
(1953).
10
-------
FV1
145 Ib/min
154 Ib/min
9 Ib/min
12.2 Ib/min
FV2
1.6 Ib COz makeup/min
5.7 Ib/min
6.5 Ib/min
1.7 Ib/min
FV3
MEK
Discharge
Figure 7.
3.2 Ib/min HP A
C02 flow rates for MEK/COz separation system.
4.9 Ib/min
Table 5. Process Cost Estimate
Case: Supercritical COz Regeneration of Activated Carbon Used for MEK Vapor
Adsorption
Capacity: 10,000 scfm air laden with 0.25 LEL MEK = 72,000 Ib MEK/day;
46.100 Ib GAC regenerated/day
Capital Investment: $532,000 (Basis: December 1980}
Operation: 33O days/yr
Variable Costs
Unit/Day
$/Unit S/day C/lb MEK Recovered
Olds, R.H., H.H. Reamer, B.H. Sage,
and W.N. Lacy, "Phase Equilibria in
Hydrocarbon Systems — The n-Butane-
Carbon Dioxide System," Ind. andEng.
Chem.,41 (3), 475 (1949).
Poettman, F.H., andD.L Katz, "Phase
Behavior of Binary Carbon Dioxide-
Paraffin Systems," Ind. and Eng. Chem.,
37(9), 847(1945).
Electricity
Make-up CO? :
Cooling H2O
Make-up Carbon:
Bern/variable Costs
Operating Labor :
Supervision
Labor Overhead:
Maintenance
Fixed Costs
Plant Overhead :
Depreciation
Taxes &
Insurance
2208 kWh 0.05
2304 Ib 0.08
2.448 Mgal. 0 15
10% Carbon Inventory a/yr
Total Variable Costs:
'/2 man/shift, 3 shift/ day @
$11.00/hr
V* man/yr @ S30,000/yr
60% of Labor & Supervision
4% of Capital Investment/yr
Total Semivariables:
40% of Labor & Supervision
10% of Capital Investment/yr
2% of Capital Investment/yr
Total Fixed Costs:
Total Operating Cost:
110.40
184.32
0.37
1.45
296.54
132.00
22.73
92.84
64.48
312.05
61.89
161.21
32.24
255.34
0.920
1.536
0.003
0.012
2471
1.100
0.189
0.774
0.537
2.600
0.516
1.343
0.269
2.128
$863.93/ day
7.199 C/lb MEK recovered
1.87 C/lb GAC regenerated
For fixed-bed adsorber system.
11
U. S. GOVERNMENT PRINTING OFFICE: I982/559-09V0482
-------
Christopher P'. Eppig, Richard P. de Filippi. and Rosemary A. Murphy are with
Arthur D. Little, Inc.. Cambridge. MA 02140.
Bruce A. Tichenor is the EPA Project Officer (see below).
The complete report, entitled "Supercritical Fluid Regeneration of Activated
Carbon Used for Volatile-Organic-Compound Vapor Adsorption," (Order No.
PB 82-228 974; Cost: $12.00. 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:
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Postage and
Fees Paid
Environmental
Protection
Agency
EPA 335
Official Business
Penalty for Private Use $300
------- |