United States
Environmental Protection
Agency
Industrial Environmental Research
Laboratory
Research Triangle Park NC 27711
Research and Development
EPA-600/S7-83-053 Feb. 1984
Project Summary
Coal Gasification/Gas Cleanup
Test Facility: Volume IV. A
Mathematical Model of the
Packed Column Acid Gas
Absorber
R. M. Kelly, R. W. Rousseau, and J. K. Ferrell
The report describes a mathematical
model for adiabatic operation of a
packed-column absorber designed to
remove acid gases from coal gasification
crude product gas. It also gives results
of experiments with a small pilot-scale
coal gasification/gas cleaning facility
designed to test the model. The model
predictions compared well with the
actual absorber liquid temperature pro-
file and outlet gas composition. The
model is useful for the evaluation of the
effect of changes in process variables
on absorber column performance and
(hence) for column design.
This Project Summary was developed
by EPA's Industrial Environmental
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
As a part of a continuing research
program on the environmental aspects of
fuel conversion, the U.S. Environmental
Protection Agency (EPA) has sponsored a
research project on coal gasification at
North Carolina State University. The
facility used for this research is a small
coal gasification/gas cleaning pilot plant.
The overall objective of the project is to
characterize the gaseous and condensed
phase emissions from the gasification/
gas cleaning process, and to determine
how emission rates of various pollutants
depend on adjustable process parameters.
The plant, described in detail in Volume
I of this report series, consists of a
fluidized bed reactor; a cyclone and
venturi scrubber for particulates, con-
densables, and solubles removal; and
absorption and stripping columns for acid
gas removal and solvent regeneration.
The plant has a nominal capacity of 50
IbVhr of coal feed for steady state
operation. A schematic diagram of the
gasifier, the acid gas removal system
(AGRS), and other major components is
shown in Figure 1.
In an initial series of runs on the
gasifier, a pretreated Western Kentucky
No. 11 coal was gasified with steam and
oxygen. The results of this work and a
detailed listing of the project objectives
are given in Volume II and were presented
at the EPA Symposium on Environmental
Aspects of Fuel Conversion Technology
V, in St. Louis MO, September 1980.
Volume III gave a detailed discussion of
a series of runs made with a New Mexico
subbituminous coal. Part of Volume III
summarized a mathematical model of the
operation of the packed tower absorber of
the acid gas removal system. This model
may be used to estimate the height of
packing required to remove a specified
component from a gas stream or recover
a specified component for a column with
) Readers more familiar with metric units may use
the conversion factors at the end of this Summary.
-------
/Vz Purge
A
Coal
Feed
Hopper
Filter
F Venturi
4 (Scrubber
Nz Purge^ \
Gasifier
2 Purge
'Char
Receiver
Nz Purge
Steam
Plant Water •
Sour-Gas
Compressor
Mist
Eliminator
Heat
Exchanger
PCS Tank
^®
f>
Acid'Gas
Flash Tank
•r i
; r — -t
T \
Heater
i i — .
i
/V2 H
X
X
X
Stripper
Drain
Circulation
Pump
^
Solvent Pump
S = Sample Port
Figure 1. Pilot plant facility.
a fixed packing height. Furthermore,
parametric studies involving a variety of
process variables can be performed
which may lead to an improved design
and/or operation.
This report gives a detailed description
of the absorber model and its develop-
ment, and compares model-based predic-
tions with experimental data from the
pilot plant.
Packed Absorption Column
Model
Several assumptions typically have
been made in developing a physical
model for non-isothermal absorption in a
packed tower. Since axial dispersion is
usually neglected, fluids are assumed to
move through the column in plug flow.
This is probably valid for small diameter
columns with low to moderate liquid flow
rates. Another assumption often made is
that the column approaches adiabatic
operation; most industrial columns have
been shown to operate nearly adiabati-
cally. Departures from adiabatic operation,
due to either heat losses or planned heat
removal, can be included in the model on
a case-by-case basis.
Figure 2 is a schematic of a packed
tower for which the following differential
material and energy balances may be
written:
d L/3z - d G/3z = d b/3 t (1 )*
3 (LHL)/9z - 3 (GHe)/3z
- He, = (bHb)/3t (2)
The resulting system of coupled, partial
differential equations is very difficult to
solve, and several approaches are
possible. At steady state. Equations 1 and
2 become ordinary differential equations
and, if adiabatic operation is assumed,
the following equations may be written:
dL,dz - dG/dz = 0 (3)
d(LHL)/dz - d(GHG)dz = 0 (4)
The steady state assumption leads to a
two point boundary value problem with
neither boundary condition completely
specified. One solution involves an
iterative calculation in which the condi-
tions at one end of the column are
assumed and an incremental calculation
is initiated at the other end. When the
calculation reaches the end of the column
where the initial assumptions were
"(Nomenclature of equation elements is defined at
the end of this Summary.
made, calculated and assumed values are
compared. If agreement between these
values is unsatisfactory, an iterative
procedure must be used to adjust
assumed values. Algorithms have been
suggested for both single-solute and
multicomponent systems, although ex-
perimental verification of only the former
has been reported.
Aside from the assumptions that the
column is operating adiabatically and the
axial dispersion is negligible, the single-
solute calculational procedure is essen-
tially rigorous. While more simplifying
assumptions could be made (e.g., ignoring
the resistance of one phase to mass and
heat transfer), they could lead to serious
error in the calculation. Certain situations
may allow a less rigorous approach, but it
is often difficult to determine this without
first performing a more detailed calcula-
tion.
Description of Computer
Programs
Two separate but similar computer
programs were used in this study.
SIMPAK was limited to the simple case of
a single transferring solute in an inert,
insoluble carrier gas and a volatile
-------
/.2= I /.JZ
J=1
To,
G7= I Gj,
/J2
7-Q2
G2=
* = O for Adiabatic
Case
n
L1= I /.j,
J=1
Figure 2. Mutt/component gas absorption/stripping in an adiabatic packed column.
solvent, while MCOMP relaxed these
restrictions.
SIMPAK receives as input the condition
and composition of the entering gas and
liquid streams and the specified removal
efficiency of the key component. The key
component can be chosen from Table 1, a
list of 13 compounds. Methanol, the
liquid used in this study, must be
designated as the solvent. The inert
carrier gas can also be any compound in
Table 1, except methanol. In general, the
key component should be appreciably
soluble in methanol, while the inert
component should be sparingly soluble in
methanol.
For multicomponent absorption, the
computer program MCOMP has been
developed. The program is a modification
of the computer code used by H. M.
Feintuch. Although any number of
components can be handled, the 13 Table
1 components, designated as the most
Table 1. Components Included in Model
important in coal gasification applications,
have been made a part of the model.
Because any of the 13 components can
transfer in either direction, the problem
becomes much more complicated than
for single-solute absorption. Although
the algorithm used in this case is similar
to SIMPAK, including multicomponent
mass transfer requires certain modifica-
tions.
After input of all necessary information
(see Table 2), the calculation is initiated
by assuming the outlet gas temperature
and outlet gas flow rates for all compo-
nents, except the key component, which
has been specified. The outlet gas
temperature is assumed initially to be
equal to the inlet liquid temperature.
After the flow rates of all exiting gas
stream components are assumed, mater-
ial and -energy balances are solved to
determine the conditions of the exiting
liquid stream.
The calculation then begins at the
bottom of the column. The total amount of
key component to be removed in the
column is divided by the number of
packing segments to be used. The height
required to achieve the removal of the key
component in each segment is determined
by calculating the temperature and
concentration gradients.
Physical, Transport, and
Thermodynamic Properties
The success of any modeling hinges on
either the availability of accurate physical,
transport, and thermodynamic property
data or the ability to predict this informa-
tion. Surprisingly, relatively little effort
has been invested in expanding existing
physical property data bases and improving
predictive methods. This is especially true
for mixture properties and at other than
ambient conditions. Both SIMPAK and
MCOMP have subroutines that compute
mixture physical property information as
needed in the model calculation.
Description of the performance of a
packed column requires both transfer
coefficients and interfacial areas for mass
and heat transfer. These quantities, along
with column hydraulics, will determine
packing effectiveness.
Butane
Carbon Dioxide
Carbon Monoxide
Carbonyl Sulfide
Ethane
Ethylene
Hydrogen
Hydrogen Sulfide
Methane
Methanol
Nitrogen
Propane
Propylene
-------
Table 2. Information Input To MCOMP (Corresponding to Conditions of Integrated Run
AMI-53/GO-73J
Gas Mass Vel.
mol/hr/ft2
17.88
Pressure
Atmos.
17.890
Comp. No.
1
2
3
4
5
6
7
8
9
10
11
12
13
Gas Temp. In
°F
42.53
The Key
Comp.
1
Comp.
C02
HiS
COS
MEOH
H*
CO
/V2
CW4
C*H<
Cz//6
Catfe
CaHa
CtHio
Liq. Mass Vel.
mol/hr/ft2
127.46
Inlet Gas
Mol. Fract
0.2213
0.0017
0.0001
0.0000
0.3292
0.1724
0.2112
0.0529
0.0045
0.0065
0.0000
0.0000
0.0000
Liq. Temp. In
°F
-4.28
Fract Key
Comp. Absorbed
0.99900
Inlet Liquid
Mol. Fract
0.0000
0.0000
O.OOOO
1.0OOO
0.0000
0.0000
O.OOOO
0.0000
O.OOOO
O.OOOO
O.OOOO
O.OOOO
O.OOOO
Binary Gas Diffusivity Option:
Binary Liquid Diffusivity Option:
Mixture Liquid Diffusivity Option:
Transfer Coefficient Option:
Packing Height
Packing Type
Packing Increments
Interface Area Coor.
Maximum Iterations
Convergence Tolerance
Convergence Accel.
Wilke-Lee Modification to Chapman-Enskog
Wilke-Chang Modification to Stokes-Einstein
Equation
Modification of Holmes et al. to Wilke-
Chang Equation
Onda et al.
7.10ft
1/4-in. ceramic Intalox saddles
10.0
1.00
0
0.0000/00 0.0001000 0.0150000 0.0050OOO
0.50000
Although the literature contains a great
deal of information about the performance
of certain packings with particular
systems, prediction of the performance of
packings for which data are not available
is difficult at best.
Mass transfer data, obtained from a
particular column and system, are often
reported as the product of the transfer
coefficient and the interfacial area
because it is often difficult, if not
impossible, to separate the two quantities.
However, using data obtained on one
system to predict mass transfer rates in
another often requires estimation of both
quantities.
While many investigators have studied
the effectiveness of various packings, few
have tried to generalize their results to
other operating conditions and systems.
For this reason, design methods for
packed columns still rest to a large extent
on experience and vendor advice. Large
safety factors have to be used to ensure
that the column meets design expecta-
tions.
To be able to predict column perfor-
mance accurately, mass transfer coeffi-
cients in the liquid and gas phases must
be known. If this information is not
available for the system being used,
correlations based on experimental data
can be used to predict them. The accuracy
of the correlations depends on the
accuracy of the data used in their
development. This is especially important
because experimental difficulties make
the measurement of mass transfer
coefficients susceptible to significant
error.
While it is relatively easy to determine
the dry surface area associated with a
particular type and size of packing, it is
difficult to determine the amount of the
packing surface that is effectively used in
an irrigated section of packing. Several
factors are important: the type of process
for which the column is used influences
the amount of surface area; the surface
tension of the liquid passing through the
packing determir.es the wetting char-
acteristics; and the shape, size, and
material used for packing.
Because there is no widely accepted
correlation for either mass transfer
coefficients or interfacial area in packed
columns, four options have been included
in SIMPAK and MCOMP. The option to be
used in a particular computation is
chosen in the program input.
Experimental Method
Details of the gasifier and the entire
pilot plant are provided in Volume I. Data
required to compare experimental results
with model predictions were generated in
two ways: (1) syngas runs involved the
use of a gas-mixing manifold where gas
was fed from bottles and metered to the
absorber through a flow controller; and
(2) integrated runs involved the use of the
fluidized bed gasifier to manufacture gas
to be fed to the absorber. The procedures
used to operate the absorber and the rest
of the acid gas removal system were
similar in both types of runs.
After steady state was achieved, gas
was sampled at the inlet to the absorber
(sour gas) and at the outlets from the
absorber (sweet gas), the flash tank (flash
gas), and the stripper (acid gas). After this
first sampling period, the system was
operated for at least an hour before a
second set of samples was taken. After
the second set of samples, the system
was usually shut down and the run
concluded.
After conclusion of the run, collected
gas samples were analyzed and used to
check mass balances around the entire
AGR system, comprised of the absorber,
flash tank, and stripper. Results from
these mass balances were then used to
assess the quality of the run. Deviations
of more than 10% for either the overall
balance or the individual component
balances usually resulted in disallowing
the results of the run. Some judgment
was used, however, in analyzing the
results, especially if the reason for the
discrepancy in the material balance was
known. Also, for several components fed
to the system at concentrations less than
1 mole %, mass balance deviations of
greater than 10% may have resulted in
disallowing the run results. As a rule,
however, the mass balance results from a
particular run met the given criteria. The
mass balance results for both the syngas
and integrated runs are included in
Volume II.
Discussion of Results
Comparison of Experimental
Data to SIMPAK
In general, model predictions compared
well with experimental data. The experi-
mental results from syngas Run AM-50,
typical of the better runs in this series, are
-------
illustrated in Figure 3. It shows excellent
agreement between the computed liquid
temperature profile and the experimental
data.
Agreement between measured syngas
temperature profiles and calculated
values showed that the model satisfac-
torily predicted column performance. The
plots of the height of packing versus the
liquid temperature show that theshape of
the predicted curve closely follows the
experimental data. Uncertainties in
temperature measurement probably
account for the infrequent differences
between predicted and experimental
results.
Table 3 compares predicted and meas-
ured temperatures of liquid leaving the
absorber for completed runs satisfying
the closure requirements on mass
balances. That these compare closely
confirms the accuracy of the flow meter
calibrations, the sampling and analytical
techniques, and the validity of the
assumptions made in estimating the heat
of solution of COz in methanol from
Henry's Law constants.
Comparison of Experimental
Data to MCOMP
Using the method established for
comparing results from SYNGAS runs to
the prediction from SIMPAK, results from
integrated runs were compared to
predictions from MCOMP. For each
integrated run, the computed liquid
temperature profile was compared to the
experimental data. The measured sweet
gas compositions were also compared to
predicted results. Results are shown in
Table 4, which compares predicted and
measured outlet liquid temperatures.
Agreement between these two sets of
values is an indication of the accuracy of
the adiabatic assumption as well as a
check on sampling and analytical tech-
niques. In all cases, the correlation of K.
Onda et at. was used to estimate the mass
transfer coefficient and interfacial area.
Because MCOMP handles up to 13
components, considerably more compu-
tation time was required than was the
case with SIMPAK. For this reason, 10
packing increments were chosen to
economize on computer time.
In all the comparisons of model
predictions to experimental results, CQz
was the key component. While any of the
components in the model could be used
as the key component, the absorption of
CO2 could be followed most easily
through the column liquid temperature
profile. The other components were
either sparingly soluble in methanol or
-30
JO
15
Figure 3.
Comparison of data from Syngas run AM-50 with model calculations using the
Onda correlation and 50 packing increments.
Table 3. Comparison of Predicted and Experimental Outlet Liquid
Run TL In. °F TL Out, °F
24
25
26
27
32
33
34
38
50
55
-33.69
-34.02
-33.30
-28.55
-31.79
-28.91
-29.33
-18.62
- 6.33
- 4.17
6.36
3.73
9.68
14.41
11.34
5.78
15.75
10.70
11.17
29.01
Temperatures (T\.)
Computed TL Out, °F
6.91
1.53
9.33
9.34
6.78
17.11
7.69
12.03
32.18
Table 4. Comparison of Predicted and Experimental Outlet Liquid Temperatures (TL) For
Integrated Runs
AGRS
Run No.
30
35
36
37
43
44
45
47
52
53
57
59
60
Gasifier
Run No.
56
59
60
61
68B
69
70
71B
72
73
76
78
79
Mass
Balance
100.0
102.4
104.0
103.3
103.8
102.3
103.0
104.8
101.4
103.1
99.2
101.8
95.5
TLln
°F
-28.27
-34.65
-27.75
-22.20
-34.07
0.84
-34.07
-33.07
-30.70
- 4.28
-32.70
- 2.92
-16.00
TLOut
°F
4.83
0.72
-1.72
7.19
-4.53
24.16
-19.23
-17.13
-15.11
8.72
-2.45
9.50
5.98
Computed
TL Out, °F
5.47
3.11
-0.84
8.22
-2.82
a
-19.74
-17.35
-14.81
10.13
a
9.29
2.44
'Runs failed to converge.
-------
•c
.§!
1.0
present in such small quantities that their
absorption did not affect the measured
temperature profile. 3 0
A series of integrated runs were made
using gas produced from a devolatilized
,char to feed the absorber. The model
predictions of MCOMP compared well
with the actual liquid temperature profile
and outlet gas composition for all runs. In ^
these runs, the transfer of eight com- £
pounds (including the solvent) was .c1 2'°
considered. The difference between the "S
actual and predicted H2S and COS outlet
gas concentrations was attributed to the
presence of both compounds in the
recirculated methanol fed to the absorber.
This resulted from the inefficient stripping
of these compounds from the rich
methanol and led to a reduced driving
force at the top of the absorber. Figure 4
uses data from run AMI-30/GO-56 to
illustrate the comparison of model predic-
tions with data from these runs.
Another series of runs were made in
the early 1980s, using a feed gas
generated by the gasification of a New
Mexico subbituminous coal. Gases
produced from gasification of this coal
were characterized by fairly low amounts
of H2S and COS, and relatively high levels
of both aromatic and aliphatic hydrocar-
bons. Only compounds that could be
analyzed with some accuracy and that
were important in mass balance calcula-
tions were included in the computer 7.5
model runs. An example of the comparison
of model predictions to data obtained in
these runs is given in Figure 5 for Run
AMI-60/GO-79.
In general, the computer prediction
compared well with the experimental
data. Comparisons of predicted and
experimental outlet liquid temperatures,
and indications of the closure of mass 5
balances for all integrated runs are given £
in Table 4. -5
Figures 6 and 7 show the variation in £
composition of the gases as they move S?
through the absorber for AMI-53. Two a:
graphs are presented because of the *>o.5
range of compositions encountered; note
that the acid gases H;>S, COS, and C02 are
readily absorbed, but that the product
gases CO, H2, and CH4 increase in
concentration through the column,
mainly due to the concentrating effect of
removing large amounts of the acid
gases.
Minimum Liquid/Gas Ratio
Specification of the liquid flow rate is
required for absorber design. The mini- Figure 5.
mum liquid/gas ratio, evaluated from
CO2
H2S
COS
MEOH
H2
CO
Nz
CH^
C2H4
C2He
Inlet
24.600
0.910
0.042
— -
33.170
21.060
18.500
1.640
—
Outlet
1.460
0.048
0.003
trace
43.190
28.480
24.890
1.950
^~—
— —
Predicted
Outlet
1.378
0.002
trace
0.013
44.459
27.698
24.345
2.085
—
Concentrations in Mole Percent
O
-35
-30 -25
-20
-15
-10
-5
w
Figure 4. Comparison of data from integrated run AM-30/GO-56 with model calculations
using the Onda correlation and 10 packing increments.
:i.o
Inlet
CO2
HtS
COS
MEOH
H2
CO
A/2
CH,
C2/Y4
CtHi
21.740
0.214
0.008
32.690
17.300
20.770
6.630
0.417
0.465
Outlet
trace
0.026
0.002
trace
43.010
21.900
27.620
7.440
0.015
0.023
Predicted
Outlet
0.029
trace
trace
O.042
42.264
22.133
26.840
8.252
0.205
0.235
Concentrations in Mole Percent
Comparison of data from integrated run AMI-60/GO-79 with model calculations
using the Onda correlation and 10 packing increments.
-------
2.0
.1
7.0
10 20
Mole Percent in Gas Phase
30
40
Figure 6.
Calculated gas composition profile for conditions corresponding to integrated run
AMI-53/GO-73.
physical property data, provides a useful
limit in setting design flow rates.
SIMPAK was modified to estimate the
minimum liquid rate for all syngas and
integrated runs made for this study. The
subroutine MINLG, included in SIMPAK,
systematically brackets the minimum
liquid rate until it can be determined to
within 0.5 Ib moles/hr/ft2.
Comparison of Transfer
Coefficient Correlations
Correlations of K. Onda et al. were used
to estimate mass transfer coefficients
and interfacial area in the computer runs
comparing model predictions to experi-
mental results. The correlations were
chosen because they were developed
with data obtained for small packing sizes
and for cases where organic liquids,
including methanol, were used. Earlier
results showed that these correlations
yielded excellent agreement between
predicted and experimental data.
The following points can be made
concerning other correlations of transfer
coefficients that were tested in this study.
The approach taken by W. R. Bollesand J.
R. Fair seems to be the most sound: it
correlates the product of the transfer
coefficient and the interfacial area.
However, Bolles and Fair suggested that
the scatter in the data used to develop the
correlation requires a safety factor of
1.70. They also suggested using a safety
'actor of 2.23 with the Onda correlation.
Results of this study indicate significant
differences between the results of
various researchers in developing gen-
eralized correlations for mass transfer in
packed columns. This is partly the result
of the scatter in the data used to develop
these correlations. While Fair's approach
I
.2.0-
I
N'
7.0-
COSxIO
avoids the problem of correlating the
interfacial area and transfer coefficient
separately, an effective generalized correl-
ation for each quantity would be more
desirable so that these results can be
extended to a wider variety of conditions.
At this point, it appears that additional
experimental data and better methods of
correlation would improve design proced-
ures for packed columns.
Recommentations for
Application of Model
The mathematical model of the packed
absorber described here has been used to
analyze the results of the experimental
runs made for this study. Other objectives
in developing this model included its use
to extend the experimental results to
situations not attainable in pilot plant
operation. Future work will extend the
packed absorber model to stripper
operation and culminate in the develop-
ment of a simulation package to describe
the performance of the entire system.
A valuable use of a computer model is
to evaluate the effect of changes in
process variables on the final column
design. This may require a complex
optimization procedure or may be fairly
simple if only one variable is involved. In
either case, the computer model becomes
the tool for the evaluation.
C2//6
7000 2000 3000 4000 5000 6000 7000
Gas Composition (PPM-MolesJ
8000 9000 10000
Figure 7.
Calculated gas composition profile for conditions corresponding to integrated run
AMI-53/GO-73.
-------
s
5 3
O Used in Experimental Run
AM-50
AM-32
40 SO 60 70 80 90 100 no 120 130 140
L. Ib moles/hr ft*
Figure 8. Effect of Solvent Flow Rate on Packing Requirements for conditions corresponding
to Syngass runs AM-32 and AM-50.
R. Kelly, R. Rousseau, and J. Ferrell are with North Carolina State University,
Raleigh, NC 27650.
N. Dean Smith is the EPA Project Officer (see below}.
The complete report, entitled "Coal Gasification/Gas Cleanup Test Facility:
Volume IV. A Mathematical Model of the Packed Column Acid Gas Absorber,"
(Order No. PB 84-113 083; Cost: $ 17.50, 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
Syngas Runs AM-32 and AM-50 were
used as base cases for SIMPAK to
illustrate the use of the models developed
here in a parametric study. Figure 8
shows the application of these models to
examine the effect of liquid flow rate on
required packing height. This kind of
analysis can be performed with any of the
specified operating variables in the
process. That AM-50 was made at a
higher inlet liquid temperature is partly
responsible for the greater sensitivity of
the packing height to the liquid flow rate.
The solubility of CO2 in methanol at the
inlet liquid temperature of AM-50 is
much less than for AM-32.
Nomenclature
b total liquid holdup in column
b, holdup of j in liquid
G total gas molar flow rate
G, gas flow rate of j
Hb specific enthalpy of liquid holdup
Hex enthalpy loss to surroundings
HG specific enthalpy of gas
HL specific enthalpy of liquid
j chemical species j
L total liquid molar flow rate
L| liquid flow rate of j
t time
TL liquid temperature
x, mole fraction of j in liquid
y, mole fraction of j in gas
z height of packing
Conversion Factors
Readers more familiar with metric
units may use the following factors to
convert to metric equivalents:
Non-metric Times Yields metric
°F
ft
ft2
in.
Ib.
5/9(°F-32)
30.48
0.09
2.54
0.45
°C
cm
m2
cm
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
CHICAGO IL 60604
U.S. GOVERNMENT PRINTING OFFICE: 1984.759-102/84
------- |