PB82-141797
    Octanol/Water Partition Coefficients  and  Aqueous
    Solubilities of Organic Compounds
    (U.S.)  National Bureau of Standards
    Washington,  DC
   Prepared for

   Environmental Protection Agency
   Washington,  DC
    Dec  81
U.S. tepert&sst eff Ces*5»rcs
nations) Tectaica) Information Semct

-------
NBSiR 81-2406
U.S. DEPARTMENT OF COMMERCE
National Bureau of Standards
Center for Chemical Physics
Chatnical Thermodynamics Division
WaahSneton. OC 20234
Dscembar 1981

-------
NBS-M4A IREV. 2->CI
U.S. DEPT. Of COUU.
BIBLIOGRAPHIC DATA
SHtETlSe* instructions)
1. PUBLICATION OR
REPORT NO.
NBSIR-81-2406

2. Performini Otfan. Report No
* u .1 T 21 1 7 ^ f
December 1981
4. TITLE AND SUBTITLE ..•,. . c n
"Octanol/Water Partition Coefficients and Aqueous Solubilities of Organic
Compounds"
s. AUTHORISE Stanley P. Wasik, Yadu B. Tewari, Hichele M. Miller,
and Daniel E. Marti re 356
t. PERFORMING ORGANiZATION (if joint or otter than NBS.
NATIONAL BUREAU OF STANDARDS
DEPARTMENT OF COMMERCE
WASHINGTON, D.C. 20234
see instructions)

095
7. Contract'Cranl No.
1. Type or Report &
Period Covered
  Office of Toxic Substances
  Environmental  Protection  Agency
  Washington,  DC 20460
10. SUPPLEMENTARY NOTES
   [~"| Document describe) a computer protram: SF-IBS, FlPS Software Summary, Is attached.
 1. ABSTRACT (A 200->vord or fesj factual summary o/ most significant information. II document includes a significant
   bibliography of literature survey, mention it here)

            A generator column method  for measuring  the octanol/water partition

      coefficient, K  , ,  and the  aqueous solubility,  C*, is described.  When water

      is  pumped through a generator column packed with solid support coated with
      an  organic stationary phase, an aqueous  solution is generated that 1s in
      equilibrium with the stationary phase.   The solute concentration in the
      elutea aqueous  phase was measured either by high pressure  liquid
      chromatography  or by solvent extraction  followed by gas  chromatographic
      analysis.
            Aqueous solubilities  and octanol/water partition coefficients of organic
      solutes, falling Into 7 general chemical  classes, have been systematically
      determined using the modified generator  column method.   From thermodynamics
      an  equation is  derived relating KO/W to  the volume-fraction-based solute

      activity coefficient 1n water (YJ). the  latter being determinate from Cs-
      For each class  of compounds, excellent linear correlations are found between
      log K / and  log Y", w1th  sl°Pes close to the theoretical  value of unity.
            o/w           s
12. KEY WORDS [Six to twelve entries; alphabetical Oi'der; capitalize only proper names; anil separate key words by s«—ucoions)
       Activity  coefficients;  aqueous solubility; gas chromatography; generator
       column; high pressure liquid chromatography; and octanol/water partition
       coefficients
U. AVAILABILITY

   JS_j Unlimited
   (~1 For Official Distribution. Do Not Release to N.TIS

   r~~) Order From Superintendent of Documents, U.S. Government Princint Office. Washlnfton, D.C.
      20402.

   ft" Order From National Technical Info-nation Service (NTIS). Springfield, VA. 22161
14, NO. OF
   PRINTED PAGES
     66
15. Price

-------
NBSIR81-2406

OCTANOL/WATER PARTITION
COEFFICIENTS AND AQUEOUS
SOLUBILITIES  OF ORGANIC COMPOUNDS
Stanley P. Wasik, Yadu B. Tewari. Michele M. Miller.
and Daniel E. Martire
U.S. DEPARTMENT OF COMMERCE
National Bureau of Standards
Center for Chemical Physics
Chemical Thermodynamics Division
Washington, DC 20234
December 1981
 Prepared for
 Office of Toxic Substances
 Environmental Protection Agency
 Washington, DC 20460
 U.S. DEPARTMENT OF COMMERCE. Malcolm SaJdrisa, Secretory
 NATIONAL BUREAU OF STANDARDS, Errant Ambter.

-------
       OCTANOL/WATER PARTITION

      COEFFICIENTS AND AQUEOUS

       SOLUBILITIES OF ORGANIC

              COMPOUNDS
Stanley P. Wasik, Yadu B.  Tewarl, and
          Michele M.  Miller
 Chemical Thermodynamics Division
     Center for Chemical Physics
    National Bureau of Standards
       Washington, D.C.  20234
                 and
          Daniel E. Marti re
        Chemistry Department
        Georgetown University
       Washington, D.C.  200b7
                Under
 Interagency Agreement EPA-80—X0985
                   ,-t

-------
                               ABSTRACT
     A generator column method for measuring  the octanol/water  partition
coefficient, K  , ,  and the aqueous solubility,  c",  is  described.   When water
is pumped through a generator column packed with solid  support  coated  with
an organic stationary phase, an aqueous solution is  generated  that  is  in
equilibrium with the stationary phase.  The solute concentration  in the
eluted aqueous phase was measured either by high pressure liquid
chromatography or by solvent extraction followed by  gas chromatographic
analysis.
     Aqueous solubilities and octanol/water partition coefficients  of  organic
solutes, falling into 7 general chemical classes, have been systematically
determined using the modified generator column method.   From thermodynamics
an equation is derived relating KQ.  to the volume-fraction-based solute
activity coefficient in water (Y*), the latter being determinate from c"
                                5                                       5
For each class of compounds, excellent linear correlations are found between
log K    and log YI with slopes close to the theoretical value of unity.

-------
                                 ACKNOWLEDGMENT

     The authors gratefully acknowledge the financial  support of their
work by the Environmental Protection Agency.

-------
                                TABLE OF CONTENTS
                                                                 Page
ABSTRACT 	     1
ACKNOWELDGMENT 	     ii
TABLE OF CONTENTS	     1ii
LIST OF TABLES   	     v
LIST OF FIGURES	     vi
DISCLAIMER	     V11
I.  .SUMMARY	     1
II. INTRODUCTION 	    2
III. COMPARISON OF EXPERIMENTAL METHODS  	     6
IV.  EXPERIMENTAL
     A.   Aqueous Solubility and Partition Coefficients 	  8
     B.   Solute Activity Coefficient in Octanol 	    15
V.  VALIDATION
     A.   Stir-Flask  Equilibrations  	     17
     8.   Solubility	    18
     C.   Partition Coefficient	     18
     D.   Activity Coefficients	     21
VI.  RESULTS AND DISCUSSION	     24
VII.   RECOMMENDATIONS FOR FUTURE RESEARCH
     A.   Salinity Dependence  	      40
     B.   Temperature Dependence  	      40
     C.   Head-Space  Measurements  	     41
     0.   Tests for Other Classes  	     41
     E.    Predictive  Schemes  	      42
 VIII.   REFERENCES	     43
                                     ill

-------
APPENDIX 1	45



APPENDIX II	47
                                      1v

-------
                                 LIST OF TABLES
Table                                                                 Page
1    Aqueous Solubility of n-Propylbenzene	 .  .  .  .   19
2    Octanol/Water Partition Coefficient, K .  . of n-Propylbenzene  .   2P
3    Activity Coefficients and Octanol/Water Partition Coefficients
     of Organic Compounds at 25.0°C 	   23
4    Aqueous Solubility and Octanol/Uater Partition Coefficients
     of Organic Compounds:  Aromatic Hydrocarbons 	  25
4-A  Aqueous Solubilities and Octanol/Water Partition Coefficients
     of Organic Compounds:  Miscellaneous Aromatic Compounds ....   26
5    Aqueous Solubilities and Octanol/Water Partition Coefficients
     of Organic Compounds:  Unsaturated Hydrocarbons  	   27
6    Aqueous Solubilities and Octanol/Water Partition Coefficients
     of Organic Compounds:  Halogenated Hydrocarbons  	   28
 7    Aqueous Solubilities and Octanol/Water Partition Coefficients
     of Organic Compounds:  Normal Hydrocarbons  	  29
 8    Aqueous Solubilities and Octanol/Water Partition Coefficients
     of Organic Compounds:  Aldehydes  and  Ketones  	  30
 9    Aqueous Solubilities and Octanol/Water Partition Coefficients
     of Organic Compounds:  Esters  ,  .	31
 10  Aqueous Solubilities and Octanol/VJater Partition Coefficients
     of Organic Compounds:  Alcohols  .  .	   32
 11  Coefficients of the Regression  Equation and the  Correlation
     Coefficient  of the Solutes  	   33

-------
                            LIST OF FIGURES
Figure
  1       Generator Column 	       9
  2       Schematic of HPLC System	«	       11
  3       Generator Column and Collecting Vessel 	     14
  4       Experimental Log K  .  Versus Literature Log KQ/M	     35
  5       Experimental Log C* Versus Literature Log C* 	     36
  6       Log KQyw Versus Log £w for the Literature Values	     38
  7       Log K  .  Versus Log Iw for the Experimental Values ....     39
                                    vl

-------
                              DISCLAIMER

     Certain cwrenercial  equipment,  instruments,  and materials are identified
in this paper in order to specify adequately  the experimental procedure.
In no case does such identification imply recommendation  or endorsement by
the National Bureau of Standards, nor does it imply that  the material,
instruments, or equipment identified are necessarily  the  best available
for the purpose.
                                           vii

-------
                                   I.   SUMMARY
     A chromatogrrtphic method for measuring the octanol/water partition
coefficient, K  . , and the aqueous solubility, c", of hydrophobic substances
              O/ W                               5
is described.  When water is pumped through generator columns containing
a solid support coated with e.n organic stationary phase, an aqueous solution
is generated which is in equilibrium with the stationary phase.  When the
organic phase is pure solute, the concentration of the solute in the eluted
aqueous phase is the solute solubility. When the organic phase is octanol
containing a solute of concentration C°  and the concentration of the solute
in the aqueous  phase is determined  to be c", then the octanol/water
partiton coefficient is given as K.  = C°/C*.  Validation of the generator
column method was made by corr.paring K  >  and C*  values of n-propylbenzene
carefully measured by the conventional shake-flask method with those
measured by  the generator column method.  Additional validation of the
method was made by analyzing K,  and C^ data measured by the generator
column method in terms of the thermodynamically derived equation.
          log KQ/w = -log $" -  log  Y°
where Y°  is the solute activity coefficient, on a volume fraction basis, in
octanol as measured by the GC met'iod and $" is the solute volume fraction in
water which  is  determined from  C* data.
     Thermodynamic theory predicts  that plots of log K,,  versus log \\n
                                                      I.I/W            y_
should Jiave  slopes of unity if  the  data are accurate while the coefficient
of determination gives precision of the measurements.  Analysis of the data
for  62 compounds representing seven chemical classes gave slopes in the
range 0.93 to 1.08 with an average  coefficient of determination 0.994.
     Our studies indicate that  the  C* ar.d K  ,  data obtained by using the
generator column method agree very  well with the shake-flask literature  data
obtained experimentally.  Thus  these results establish that the data are
accurate and the generator column method is a valid method for measuring
Cs and Ko/W

-------
                                  II.   INTRODUCTION
     Measurement of the aqueous solubility, C*. and the octanol/water
partition coefficient, K .  , are important for  determining the fate
and transport of organic chemicals in the environment.  K ,  may be
considered as a quantitative measure of the hydrophobicity of the compound.
In this respect, it has been used to predict bicconcentration of
organic pollutants in trout muscle [Neely et al_., 1974] and for making
other chemical and biological correlations [Leo e_t al_., 1971].  The aqueous
solubility gives an insight into environmental  movement and distribution and
the potenti''! for biodegradation by microorganisms in soil surfaces, water,
arid sewage treatment plants as well as other degradation pathways such as
photolysis, hydrolysis, and oxidation.
     Octanol/water partition coefficients and aqueous solubilities are
often measured by a sfake-flask method.  Solubilities are determined
by shaking the solute with water, allowing the two phases to separate,
and then determining the solute concentration in the aqueous phase by
an appropriate analytical technique.  Octanol/water partition coefficients
are determined by shaking a known volume of octanol containing a small
amount of solute with water,  allowing the two phases to separate,
and measuring the solute concentration in  the aqueous phase, Cw, and
in the octanol phase, C° . K.  is determined as the ratio C°/c"
                       S    O/ W                             S  S •
      In any shake-flask experiment with a  hydrophobic substance,
the solute concentration in the aqueous phase is low.  Consequently
the measurement of this concentration can  be seriously affected by
the presence of colloidal dispersions (emulsions), adsorption of
the solute onto surfaces of transfer vessels, and loss of a volatile
solute into the atmosphere.  These errors  can be reduced, but only
at a  cost of inconvenience and increased time, by centrifuging or

-------
replacing the shaking by gentle stirring, by prerinsing the glassware
with the aqueous phase, and by avoiding exposure of the aqueous phase to
the atflRrephere.  A rapid, convenient method for preparing equilibrated
aqueous phases without significant errors from such effects would be
desirable.
     For this purpose we have adapted the generator column method
described by Hay e£ a_L [1978].  A generator column has a solid support
coated with a solid or liquid stationary phase.  When water is pumped
through the column, an aqueous solution  is generated which is equilibrated
with the stationary phase.  May e£ al_. [1973} used their generator
columns, packed with glass beads, mainly to measure solubilities
of  solid polycyclic aromatic hydrocarbons; the only liquid stationary
phase used by the.., was benzene [Hay et^ <*!•, 1978].  We have redesigned
the ge.'.erator column, changed the  support  to silanized silica in order
to  increase the coverage, and extended the stationary phase to hydrocarbon-octane1
mixtures.
      In  this report we discuss the work  done under Interagency Agreement
EPA-80-2095  in the period October  1,  1980  to September 30, 1981.  The objective
of  this work was  to develop and validate a chromatographic method to measure
C*  and K .  and to make measurements  on  selected classes of organic compounds
in  order to obtain accurate and precise  K  .  and C* values.
      In  the  third section of this  report,  "Comparison o* the Experimental
Methods,"  the shake-ftask method  is compared with the generator column
method.   The advantages and  limitations  of both methods are discussed.
      The fourth section  includes experimental details of the generator
column method and the gas-liquid chromatographic method for determination of
activity coefficients and cctanol/water  parti ton coefficients of organic
compounds.

-------
     In the fifth section, "Direct Experimental Validation of the Generator
Column Method," the work done to establish a bench-mark C* and KQ/w measurement
for n-propyl benzene is discussed.  A detailed error analysis investigating
every experimental parameter was made in order to arrive at the best
and Cw values.  These values were then compared with the values measured
by the snake-flask method.
      In the sixth section, "Results and Discussion," K    and C" values are
given  for over 90 compounds representing seven types of organic
compounds.   Further validation of the yenerator column method is made by
analyzing the data using  linear regression analysis to the thermodynamically
derived  equation log  K  ,  = c log YJ * d» where c and d are series constants
and i* is the activity  coefficient of solutes in water saturated with octanol.
The accuracy and precision of the data is discussed in terms of the value of
the slope,  c, and  the coefficient of detenvi nation, r  ,  for the regression
analysis of the data  to the different chemical classes.
      In  the seventh  section, "Recommendations for Future Work," we
recommend  that  additional K  .  and C* measurements  be made in aqueous NaCl
solutions  and simulated sea water  in order  to establish the salinity dependence
of these properties.  We also  recommend  that C* measurements be made at different
 temperatures in order to establish the temperature  dependence of solubility
and that solute activity coefficients be rr-easured by d head-space method
 to further validate the generator  column method.    There  is also a need
 for KQ.  and C* measurement on classes of compounds not studied in this report.
      In Appendix I,  the thermodynamic derivation  is given  for the correlating
 equation
          log KQ/W = -  log **    -  log Y°
 where $*  is the volume fraction of  solute  in water saturated with octanol
 and Y?  Is the  activity coefficient, on  a volume  fraction  basis, of  the

-------
solute in octanol saturated with water.
     Appendix II includes the proposed test method for determination
of the aqueous solubility of organic compounds.

-------
                    III.   COMPARISON OF EXPERIMENTAL METHODS
     Generator columns have several advantages over shake-flask methods
for equilibrating water with a liquid organic phase.  The flow rate  of water
 through the column can be made slow enough to avoid colloidal dispersions,
while the large interfacial area between the organic and aqueous phases  allows
rapid equilibration.  When the column is part of a continuous and closed flow
system, the system walls become equilibrated with the aqueous solution  and
errors from adsorption are avoided; also there is no exposure of the solution
to the atmosphere, thus loss of volatile solutes is avoided.  Only about 1 ml
of the organic phase is sufficient to load the column, and the column will
generate samples of the aqueous solution of whatever volume is needed for
analysis until the organic phase has been stripped.  Once a column has  been
loaded with a given organic phase, it is a simple matter to vary other
parameters such as pH or salinity of the aqueous phase to determine these
effects on the aqueous solubilities of organic compounds.
     For compounds having  relatively large aqueous  solubilities  (>1.0 H)
and low KQ^w values (<100), there  is no particular  advantage  1n  the
generator column method over  the shake-flask method, since colloidal
dispersion and surface adsorption  are no longer an  experimental  problem.
However, the generator column method is still preferred because  it is more
convenient to thermostat a generator colics, than •>  fUsk that must be shaken.
The generator column method requires no special skill of the operator.
The generator column  is easy  and Inexpensive  to construct and generally
yields reproducible and accurate data at desired temperatures; whereas,
one will require a  thermostated centrifuge to obtain accurate results
using  the shake-flask method.

-------
     In summary, the generator column method is much better than the  shake-flask
       in that the measurement time is much less than the shake-flask method.
Although equipment is required to measure the solute concentration in the
two phases (HPLC or GLC) the generator column method with its
closed flow system makes the measurements more convenient and avoids  the
intrinsic errors associated with the shake-flask method.

-------
                                 IV.  EXPERIMENTAL
 A.  Aqueous  solubility  and  partition coefficients
 Keagents
      The  compounds  (solutes)  for which aqueous solubilities and
 octane 1/water partition coefficients (K  ,  's) were measured were
 obtained  from various commercial sources with purities >98S.  The octanol
 was purified by washing successively with  0.1 K H^SO^, 0.1 H NaOHt and
 distilled water. It was then dried with CaClg, filtered and distilled
 at atmospheric pressure. HPLC grade methanol and water were used to
 prepare the  mobile  phases for liquid chromatography.  Ordinary distilled
 water was used for  the  aqueous phase solutions.  Octanol saturated with
 water (w-octanol) was used  for making up the solute in octanol solution.
 This  solution was made  up by  stirring 10 mL of water with 250 ml of octanol for
 several hours.   The lower excess water layer was removed by a siphon.  For the
 *o/w  measuremsnts,  a \% (by weight) solute in w-octanol was used for coating
 the solid support in the generator colurm.
      The  design of  the  generator column and the method used to determine the
 solute concentration, c", in  the aqueous phase depended on whether the
                        5
 solute absorbed the U.V.  For compounds which absorbed in the U.V., high
 pressure  liquid chromatography,  HPLC, was  used to determine C*; for compounds
 which did not absorb in the U.V., solvent  extraction followed by gas
 chromatographic analysis was  used to determine C*.
 Generator Colurn
      The  design of  tne  generator column is shown in Figure 1.   The column
 consists  of  a 6 mm  (outside diameter) Pyrex tube joined to a short enlarged
 section of 9 mm Pyrex tubing.  To pack the column a plug of silanized glass
 wool was  inserted at one end of the 6 mm tube.   Silanized diatomaceous silica
 support (about  O.Sg 100-120 mesh Chromosorb W cleaned  by Soxhlet extraction
 with ethanol and dried) was poured into the tube with  tapping  and retained
with a second plug of silanized  glass wool.
                                     8

-------
                      Inlet
            20 cm
                                 Glass  Wool
         Support (100-120 mesh

                      Chromsorb W)
X>r*—
                              6 mm
                                 Glass Wool
                            =-9 mm
                     Outlet
Figure 1. Generator column used to generate the aqueous phase.  The column is 20
        cm. long and packed with 100-120 mesh ehromosorb W for the solid support.

-------
     The column was coated with a liquid solute by pulling approximately
2 ml of the liquid through the clean dry support with gentle suction.   The
column was coated with a solid solute by dissolving the solute in a volatile
solvent (102 v/v) and then pouring the mixture into the column until all
the solid support was saturated as evident by the appearance of the volatile
solution at the base of the column.  The solvent was then distilled off
under reduced pressure.
HPLC Method
     The HPLC analytical system  is shown schematically in Figure 2.  Two
reciprocating piston pumps delivered the mobile phase (water or a methanol-
water mixture)  through two 6-port high  pressure rotary valves and a
30  X O.b cm C-18 analytical column to an ultraviolet absorption detector
operating at 254 nm.  Chromatogram peaks were  recorded and  integrated with
a recording integrator.
     One of the 6-port valves  was  sample injection valve  used for  injecting
samples of either  the octanol  phase  for analysis, or standard solutions
of  the  solute  in methanol  for  determining  response factors.  The sample loop
volume  was determined by a spectrophotometric  method using  basic chromate
 solution which are known tc be suitable as absorption  standards  [Edlsbury,
 1967J.  The method consisted of measuring  the  absorbance  at 373  rtm of  three
 loopfuls  of ar» aqueous  stock solution of K^CrO^ (1.3%  by  weight) diluted  to 50 ml
with  0.2  percent KOH,  and measuring  the absorbance of  the same stock
 solution  -\fter diluting 1:500 with 0.2 percent KOH.  From these  absorbances
 the loop  volume was calculated to be (35.7 +, 0.1) mL.
      The  other 6-port valve 1n the system  served as  a  switching  valve  for
 the extractor column which was used to remove solute from the aqueous
 solutions.   The extractor column was a 6.6 X 0.6 cm  stainless steel tube
                                      10

-------
  WeOH
                       Sample Injection Valvo
Switching Valve
                                                                                Switching PiitJern:
               MPLC
               Pumps
                                Sample Loop
                                                                            T
                                                                       Weighing Bottle
                                                                          or Waste
                                                     Extractor Column
                                                                            Analytical
                                                                            Column
                                                                                                  • •  .!• L.
                                     Plollcr
                                     Integrator
                                       UV
                                     Detector
                                                                                                    i
                                                ?
                                              Waste
Figure 2.   Schematic of the HPLC analysis system.   The  generator column outlet 1s connected at A and
           when  a known amount of aqueous phase 1s  collected  in the weighing bottle, the switching
           valve 1s switched and the HPLC pumps transfer  the  mobile phase through the extractor column
           and Into the analytical column.  The syringe injection Is used for calibration of the UV
           detector.

-------
with stainless steel end fittings containing 5-i.rn frits, filled with
a superficially porous reverse phase packing (Bondapak C,n/Corasil,
Waters Associates).  Aqueous solution from a generator column entered
the switching valve through the 1/16 in (1.6 mm) stainless steel tubing
labelled "A" in Figure 2.  A water bath was used to thermostat
to +_ 0.05°C the generator column, the switching valve, the extractor
column, and the interconnecting stainless steel tubing.
     The procedure for analyzing the aqueous phase was first to flow
the aqueous solution to waste with the switching valve in tne inject
position In order to equilibrate internal surfaces with the solution, thus
insuring that the analyzed sample would not be depleted by solute
adsorption on surfaces upstream from the valve.  At the same time, water
was pumped from the HPLC pumps in order to displace methanol from the
extractor column.  The switching valve was next cnangei! to the load position
to divert a sample of the solution through the extractor column, and the
liquid  leaving this column was collected in a weighing bottle.  During
this extraction step, the mobile phase was changed to a methanol-water
mixture to condition the analytical column.  After the desired volume
of sample had been extracted, the switching valve was returned to the
Inject  position for elution and analysis.  Provided that there was no
breakthrough of solute from the extractor column during the extraction
step, the chromatographic peak represented all of the solute in the
sample.  The solute concentration in the aqueous phase was calculated
from the peak area and the weight of the extracted liquid collected in
the weighing bottle.
                                    12

-------
     The octanol/water partition coefficient was calculated from the solute
concentration in the octa"ol and aqueous phases by dividing the concentration
in the octanol phase by the concentration in the aqueous phase.  The water
flow through the generator column was 2 mL/min.  Further experiments
with a generator column at 25°C show that within experimental error
(standard deviation 3» or less) the measured aqueous concentration was
independent of flow rate (in the range 0.5-2 mL/min).
GC Method
     In the GC method aqueous solutions from the generator column entered
a collecting vessel (Figure 3) containing a known weight of extracting solvent
which was immiscible in water.  The outlet of  the generator column was positioned
such that the aqueous phase always entered below the extracting solvent.
After the aqueous phase was collected,  the collecting vessel was stoppered
with a 3/8" Teflon Swagelok cap, and the quantity of aqueous phase was determined
by weighing.  The solvent and the aqueous phase were equilibrated by either
stirring with a  glass enclosed magnet or by rotating the collecting vessel
at the rate of two revolutions per minute for  five minutes.  It was found
that rotating the collecting vessel  for longer than five minutes did not
Increase the amount of solute extracted into the extracting  solvent,  A small
amount of the solvent was then Injected into a gas chromatograph equipped
with flame  ionization detector, and  the solvent concentration  In aqueous
phase was determined from a calibration '•urve  constructed  using known
concentrations of the solute.   In order to  determine the partition  coefficient.
Ko/w* samPles of tlie 1*  (by weight)  solute  in  n-octanol solution were injected
Into the GC,  and concentration of the solute in the octanol  phase was
similarly calculated from a calibration curve. As for  the HPLC method, our
experimental  error  1n the GC method  was less than 3%.
                                     13

-------
             Bath
             Return
                                 Generator
                                 Column
                                 Inlet
                                    Generator
                                    Column
                                       To Constant
                                       Temperature
                                       Bath
                                    Extracting
                                    Solvent
Figure 3.  Generator' column and collecting vessel used for analysis of the
         aqueous phase by the GC-solvent extraction method.
                            14

-------
B.   Solute activity coefficient__1n_ octanpl
     The stationary phase in octanol, was coated onto the support material,
Chromosorb W-HP, 100/120 mesh, and the weight percent of coating was
determined by an ashing method.  A stainless steel column (1/8" OD) was then
packed with a known amount of the coated support and was connected to a
Hewlett-Packard 5830A gas chromatograph equipped with a flame ionization
detector (FID) for the determination of the solute retention time.  The
temperature of the column was controlled by circulating water through a
copper tubing jacket around the column and by a Haake Model FK temperature
regulator which regulated the water  temperature to 25.0 + 0.05 °C.
     The column inlet pressure, measured with a precalibrated pressure
gauge  (range 0-15 PSI), was kept constant  during a run by regulating with
a precision valve.   In order to measure the carrier gas flow rate  the column
was  disconnected from the FID, just  before and after the experiment, and
a soap bubble  flowmeter was connected  to the column outlet.  The carrier
gas  was presaturated with n-octanol  in order to reduce bleeding of the
stationary phase.
      Since the retention  times  decreased as the experiment  progressed due
to column  loss of  the n-octanol,  toluene was used  as a reference retention
time standard  and  Injected with each solute.  The  measured  retention ti;r.es
were then  corrected for  bleeding  using the reference solute retention time.

-------
     The solute specific retention volumes  (V  }  were  calculated  by  using
the following equation                           _
          „    V   V!a
             "      '        '
                                a         
-------
                                 V.  VALIDATION
     Because of the general lack of agreement in reported C^ and K .
                                                           s      o/w
values, the first steps toward validation of the generator column method
was to compare generator column data with data measured in this laboratory
by the shake-flask method.  Propylbenzene was chosen as the compound
to establish a bench mark measurement because it is very stable, absorbs
1n the U.V., has a typical log K  .  and log C* value, and can be
obtained pure (>99.9 %) from commercial sources.
A.  Shake-Flask Equilibrations
     For each equilibration about  10 ml of water and about 4 ml of the
organic phase were placed in a closed Pyrex vessel and stirred with a glass
enclosed stirring bar.  For the solubility measurements the organic phase
was the pure solute.  For the K  .  measurements the organic phase was
a }% (by weight) octanol solution.  All precautions were taken to eliminate
systematic error.  All transfer vessels and tubing were prerinsed with
the aqueous phase to avoid solute  adsorption on the walls.  The equilibration
flask was stoppered at all times to avoid exposing the aqueous phase to the
atmosphere.  To prevent emulsion formation in the aqueous phase vigorous
stirring was avoided.  Most of the organic layer was confined to a narrow
neck at the top of the vessel.  Above the neck was a 3/8-inch tube septum
for sampling the organic layer.  The equilibrated aqueous phase was
transferred to the HPLC system for analysis with the aid of nitrogen  pressure.
                                      17

-------
B.  Solubility
     Table 1 lists our solubility results for propylbenzene.   The aqueous
samples were 2-6 nt in volume.  The shake-flask and generator column results
for comparable temperatures agree within experimental uncertainty.  Our
value at 25°C> 4.3 X 10~4 M, is similar to the values for this temperature
reported by Andrews and Keefer [1950] (4.6 X 10"4 M)  and by Hermann [1972]
(5.0 X 10"4 M).  The temperature dependence of the solubility suggests
that there is a minimum in the neighborhood of 20°C.   Solubility minima
near 18°C have been observed for benzene and its methyl, dimethyl, and
ethyl derivatives [Bohon and Claussen, 1951].  In the case of propylbenzene,
G^l Sill-  [1976] predict from the temperature dependence of the heat
of solution  that there is a solubility minimum at (18.6 +.0.2) °C which is
consistent with our results.
C.   Partition Coefficient
     The octanol-water partition coefficient, KQ.,  was calculated from the
measured propylbenzene concentrations in the equilibrated octanol and
aqueous  phases by dividing  the concentration in the  octanol phase by the
concentration  In the  aqueous  phase.
     The volumes of  the  aqueous  phase samples used to obtain  the  partition
coefficient values  listed  in  Table Z were  in the  range 1-4 ml for the
shake-flask results  and  10-24 ml for the generator column results.  The
water  flow rate  through  the generator columns was 2  mL/min.   Further
experiments with a  generator  column at  25°C  showed that within experimental
uncertainty (standard deviation  3% or less)  the measured aqueous
concentration  was  Independent of the water flow rate (in the  range 0.5-2
rnL/min)  and was  independent of the volume  of sample  passed through the
extractor column  (in  the range 2-24 ml).   These experiments show  that
under  our experiment.!  conditions  the equilibration  was complete  in the

-------
                                    Table 1
                     Aqueous Solubility of n-Propylbenzene

                   t/°Ca          nb        C", 10'4 Mc
                 stir-flask, stirred 16 h
                   ambient      4       4.27 + 0.06
                 Generator  column
                   15.0     3       4.26 + 0.05
                   20.0     2       4.25*0,12
                   25.0     3       4.32 1 0.02
                   30.0     3       4.45 + 0.05

Equilibration temperature.  Ambient is  (23 + 1)°C.
 Number of aqueous solution samples  analyzed.
cAqueous concentration; the mean value and the  confidence  limits
 at a 95% confidence level  are listed.
                                      19

-------
                           Table 2
Octanol/Uater Partition Coefficient,  K.  , of n-Propylbenzene
Method Used Cone, of propyl-
benzene 1n octanol
(Moles/ liter)
Stir-flask 0.222
0.223
0.220
Generator column 0.038
0.086
0.223
0.223
0.223
0.223
0.223
0.223
Equilibrium
Temperature
(°C)
23 ±1
23+1
23 + 1
25.0
25.0
25.0
10.0
15.0
20.0
30.0
35.0
Number of Number of
Hours Stirred Octanol
Phase samples
Analyzed
24 4
41 4
44 3
4
4
4
4
4
4
4
4
Number of
Aqueous
Phase Samples
Analyzed
4
3
4
9
4
6
2
2
2
4
3
L°9 *o/w

3,73 £0.02
3.72 +0.01
3.71 +0.01
3.69 +0.01
3.70 +0.01
3.72 +0.01
3.71 +0.01
3.72 +0.01
3.74 + 0.01
3.72 +0.01
3.68 + 0.01

-------
generator column, and that there was no significant breakthrough of
propylbenzene from the extractor column.
     Three shake-flask equilibrations with different stirring times were
carried out (Table 2).  With increased stirring time, there was a decrease
in the partition coefficient values of about the same size as the experimental
uncertainty.  There is good agreement between these values and the generator
column results.
     The partition coefficient 1s seen to vary with temperature (Table
2) having a maximum In the vicinity of 20°C.  However, further studies are
required to generalize this phenomenon.  It Is Interesting to note that while
the solubility and the partition coefficient of propylbenzene both vary
with temperature, their product over the temperature range in which both
quantities were measured  (15-30°C) 1s essentially constant (2.26 + 0.05 M).
This suggests that the heat of solution of propylbenzene in octanol Is small.
     The partition coefficient measurements at 25°C (Table 2) show a small
but systematic decrease of log KQ/W with decreasing concentration in the
octanol phase.  Extrapolation to Infinite dilution gives log K  .  = 3.69 + 0.01
                                                              o/w        —
at 25°C.  Iwasa ejt al_. [1965] report log KQ/W = 3.68 + 0.01 from shake-flask
measurements at ambient temperature, 1n excellent agreement with our
results.
Activity Coefficients
     In Appendix I an expression Is derived showing the relationship of
109 Ko/w with the 8<3UCOUS solubility in term of solute volume fraction
           *°9 Ko/w " " '°9 *s   " ^°9 Ys
where **   Is the solute volume fraction 1n water saturated with octanol

-------
and Y?   is the solute activity coefficient,  on a  volume  fraction basis,
in octanol saturated with water.  This equation provides  an  indirect method
to determine K.  from the solute activity coefficient in octanol and  its
aqueous solubility.
     The solute activity coefficient in octanol, Y?  » can be determined from  the
retention volume [Conder and Young, 1979] of the solute eluting from a G.C.
column containing octanol saturated with water as  the stationary phase.
This method for determination of Kr»w has been documented in our recent
publication [Tewari et al_., in press].  Since Y?   "Is measured at infinite
                    "                           5
dilution and $*   is also at Infinite dilution, the K ,  determined using
the above equation is also at infinite dilution.
     The value for log Y?   determined by the G.C. method for a number of
solutes are listed in Table 3.  Using Y°   and Y"    determined from
the aqueous solubility and solute molar volume, octanol/water partition
coefficients for the solutes were calculated and compared with the
experimental KQ^W.  These two quantities are listed in the last two
columns of table 3.  An examination of these two columns  shows an excellent
agreement between the experimental KQ/W and the calculated K..  Based
on our measurements, there is a complete consistency In the above
equation.  Thus these results suggest that our data are precise and
accurate and that the generator column is a valid method  for determination
of the aqueous solubility and octanol/water partition coefficient of  organic
compounds.
      In conclusion, we have demonstrated In this section  that the
generator column method for determining C* and K  ,  is a  valid and an
accurate method for measuring these quantities.
                                   22

-------
                Table 3
Activity Coefficients and Octanol/Water Partition
   Coefficients of Organic Compounds  at 25.0°C
Solutes
Benzene
Toluene
Ethyl benzene
o-Xylene
m-Xylene
Chlorobenzert
-------
                           VI.  RESULTS AND DISCUSSION
     The experimentally determined values of K .   and C* and the  calculated
           ^J
values of YS are listed in Tables 4 through 10 for several  classes  of  compounds.
The notations a and b refer, respectively, to the HPLC and  GC methods  of  analysis.
The average experimental uncertainties in KQ,W and C* are around  l.OX.  An
asterisk is used to indicate that the literature  data are calculated values
according to Hansch et al_. (1968).  Our results are generally in  good  agreement
with available experimental literature data obtained by the conventional
shake-flask method.
     The present data were subjected to a linear regression of log  K .  •
c log YJ + d for each class of compound and for all compounds taken together.
The results of these analyses are shown in Table 11.
     The present equations are Inadequate to explain K  . and C*  data  of
aromatic solids listed in Table 4A.  Analysis of these data 1s, therefore,
not Included in Table 11.
     According to equation 9 of Appendix I the slope for each class should be
unity and the negative of the intercept Is log Y° (assumed  to be  relatively
constant).  Examination of Table 11 shows that the slopes are indeed close to
unity, ranging between 0.91 and 1.08.  These small deviations are most likely
caused by slight variations 1n log >J within each class (Tewari et  aJL, in
press).  For example, the average of the difference between log TW  and log K .
for the aromatic compounds is 0.55 for those with aliphatic substltuents  and 0.68
for those with halogen substltuents, whereas the least-squares value of log
Ys Is 0.77 when all the aromatic compounds are Included. Therefore, the  Intercepts
represent only approximate (averaged) values of log Y? for  each class.
     These findings show ttoit the octanol/water partition coefficient  of  a
solute may be estimated with reasonable accuracy  from knowledge of  Its  aqueous
                                   24

-------
                                                       Table 4

                                   AQUEOUS SOLUBILITIES AND OCTANOL/KATER PARTITION

                                           COEFFICIENTS OF ORGANIC COMWXAOS
                                                                     rrr
                                                                                  Aromatic Hydrocarbons
                                Concentration
                                                          1091;
     Solute
                   This Study     literature Value
                       IH)              IM)
CMorooenzefte    2.62 X 10
                             2.84 X  10
3.56
                   This Study-J
2.98*
                                                                                                      0/,
                                                                                                 Literature
Toluene
Etnylbenzene
o-xylene
•-Kylefvc
p-Xylene
n-Propyl benzene
n-Butylbenzene
n-Pentyl benzene
n-Heiy I benzene
6.28
1.76
2.08
1.51
2.02
4.34
1.03
2.S9
6.27
X
X
X
X
X
X
X
X
X
lO"3
10°
10-3
10-3U)
,0-3(0)
)a-4CO)
,o-«f.b>
10-5U)
,«•»(•)
5.80 X
1.52 X
1.6) X
1.38 X
1.47 X
4.99 X
8.79 X
«
••>
10
10
10
10
10
10
10
-
*
.3
-3(c)
-3(c)
-3{c,
-4(f)
•5(c)


3
3
3
3
3
4
4
S
5
.17
.66
.60
.73
.60
.22
.79
.35
.92
4 Ccl&S
«.* VJ
3.13<*>
J.I|W.
3.20**'
3io(«l*b)
• lo
3.69f'-b>
4.2B(0)
4CA»*»
• W
5«?Z
2.63
3.15
3.12
3.20
3.15
3.63
4.26
—
—
(f)

to-3'''
lo-3'*)
io-2l«)
io-^-»
3.56 2.98(tJ 2.99tf>
3.96 3.28<§) 3.25(f)
3.47 2.67*'* —
3.46 2.7/1*1
3.52 1.961** —
2.50 1.B5'*1
         Indicates that the literature daU are calculated values  according to Hansch 5J. aJL, 1968.
      '*' W>LC eethod of jnalysis.
     (b)
     (c)
    GC neUxxl of analysis.
    Sutton and Calder. 197i.
td) NcAulUfe. 1966.
    Hansch et.aK,  1968.
    Hansch and Lto. 1979.
           et »i...  1980.
     	... and  Burger. 1955

(J)
      (f)
      (9)
      (h)
         Average standard deviation of log i* M«urenents Is 0.04.
         Average standard deviation of log KO/W Masurenents Is 0.04.
                                                          25

-------
         wuenn JOIMIIITUS. AND OCTA*WWTM MKTIIIOM comictwTS
                               Of OftSMIC aM>OUIC>S
                                             MlictlUntouf AroMtfc
          txnttnt
SOW..
n-Hcplylbwuenc
n-OctyU«um
Hc«Me Otf 1 beftieti*
Aqueous CcACtntritlMi
Thtt study L1t*r«urt
(H) («)
Z.S* X IIT'14'
|.4« X 10'7'*'
I.4S I 10-6(*»
.« ,;»»
.
t.ze
;.u
—
109 "o/*
T«1s Stud/'^' L1ttr«tur«
—
—
4.«1">
(UpMluleive
                      2.J9 I 10'
                               *U)
I XT*'*' ~.
                                    Ml  I  IO*4'*'
                      J.S1
                            W
                              **'*'
p-Fluorut«fiiirl
     chlorld*
 •-Dlcnlorgocnicnc
 •-DtcMarofettucfW
 p'DIChlorottflMM
 1,Z,)-tricii1«raMnttiit
 1 .2 ,4-THmthy IbcfUtnt
                        i.Cl
                                                                *.»
                                                                    <*>        -.
                        6.40 X
                        1.90 I 10*Zt*>
 2.J.*.Trl««tliyl
                                                                *.«
                                                                    <•>
         pJwoo!
                                                                ,.„
                                                                    (»)
      |*| Indlcim tfwt tlw lUcratw* tftU 
-------
                                          Table 5
                     AQJEOUS SOLUBILITIES AND OCTANOL/WATER PARTITION
                        COEFFICIENTS OF ORGANIC COMPOUNDS
                                             Unsaturated Hydrocarbons
Solute
1-Hexene
1-Heptene
VOctene
l*Nonene
1-Pentyne
1-Hexyne
Aqueous Concentrations
This Stwdy
(H)
8.28 X 10~^b) 5
1.85 X 10~Atb*
3.65 X 10"5(b) 2
8.85 X 10"6fbJ
1.54 X 10"2tb)
8,37 X 10'3tb)
Literature
value
(H)
on w l/vH^i
• Q7 A III
—
.40 X 10"5

3.98
4.58
5.24
5.81
2.81
3.01
109 Ko/»
Th1s^^ Literature
Study Value
3.47- )J 2.70*e)*
3.99*b)
4.88tb^ 3.70^e^*
5.35^
2.12
-------
                                               Table 6
                          AQUEOUS SOLUBILITIES AND OCTANOL/WATER PARTITION
                                  COEFFICUNTS Of ORGANIC COMPOUNDS
                                                       Halogenated  Hydrocarbons
Solute
1-Chlorobutane
1-Cfilcroheptane
I-Bromobutane
l-Brofliopentane
l-Brwnohexane
1 -Brow/heptane
1-Bromooctane
Bremochloromethane
l-8romo-2-chloro-
propane
1-Iodofoeptane
Trichloroethylene
4-Brou»- 1 -butene
Ally! bromide
Aqueous Concentration
This Study Literature
Value
(H) (M)
9.43 X }Q~3M 7.19 X 10*3{e)
1.01 X 10'4(b)
6.34 X 10'3(b) 4.13 X 10'3{e)
8.38 X lfl-4{b)
1.56 X 10'4(b)
3.71 X 10-5(b)
8.65 X 10'6(b)
0.129(b)
1.42 X 10-2(b)
1.55 X 10'2 2.60^*
3.49^>
3.8Qfb>
4.36
2.18
2.53^)
2.53<'>
1.79(4>
(a)
(b)
(e)
    Indicates that the literature data are calculated values according to Hansch «t aj...  196«.
    HPLC method of analyls.
    GC method of analysis.
    SutLon and Calder, 1975.
^d) McAullffe. 1966.
te) Hansch et aj... 1963.
f Hansch and Leo, 1979.
    Mac Kay et. a\_., 1980.
    Reddick and Burner. 1955.
(9)
(h)
(J)
    Average standard deviation of log YJ "easurenents  Is 0.03.
    Average standard deviation of 109 KO/W measurements  is 0.03.
                                              28

-------
                                     Table 7
                AQUEOUS SOLUBILITIES AND OCTANOL/HATER PARTITION
                        COEFFICILNTS OF OR3ANIC COMPOUNDS
                                             Normal Hydrocarbons
Aqueous Concentration log y* log K.
Solute
n-Pentane
n-Hexane
n-Heptana
n-Octane
This Study
(M)
5.65 X 10"4(b)
1.43X10"4(b)
3.57 X 10"5(b)
6(b)
9,66 X 10 b
Literature
Value
(M)
5.39 X 10'4(d) ' 4.19
1.13 X 10~4^ 4.73
3.05 X 10~(dJ 5.28
5.97 X 10'6(d) 5.80
This"*
Study
3.62"1'
4.11'"'
4.66'
3.50*
4.00(6'*
(a)
(b)
(c)
(e)
(f)
(9)
(h)
Indicates that the literature data are calculated values according to Hansch et. al., 1968.
HPLC method of analysis.
GC method of analysis.
Sutton and Calder, 1975.
McAullffe, 1966.
Hansch et aj_., 1968.
Hansch and Leo, 1979.
MacKay et al., 1980.
    Reddlck and Burger,  1955.
*  ' Average standard deviation  of log y* measurement?  is 0.06.
                                       s
    Average standard deviation  of log K ,  measurements Is 0.02.

-------
                                         Table 8
                    AQUEOUS SOLUBILITIES AND OCTANOL/WATER PARTITION
                           COEFFICIENTS OF ORGANIC COMPOUNDS
                                  Aldehydes and Ketones
Solute
2-Butanone
3- Pen ta none
2-Heptanone
2-Octanone
2-Nonanone
2-Decanone
Acetal
2-Furaldehyde
Aqueous Concentrations
This Study
(H)
1.89
0.53
3.57 X 10'2
8.85 X 10~3
3(a)
2.61 X 10 J
5.03 X 10"4
0.75
0.81
Literature
Value
(M)
—
0.43
3.79 X 10'2*e)
...
—
—
0.42(h)
0.85(e)
log Y*

0.77
1.25
2.30
2.86
3.35
4.02
0.97
1.17

This")
Study
0.69
0.99{b)
1.98
2.76
3.81
0.52
.52<"
109 Ko/w
Literature
Value
0.29(e)
0.79^*
1.73^>*
...
—
...
— .
—
(b)
(c)
(d)
(e)
(O
(9)
00
(D
Indicates that the literature data are calculated  values according to Kansch et al_., 1968.
HPLC method of analysis.
GC method of analysis.
Sutton and Calder, 1975.
HcAullffe. 1966.
Hansch ej^ajk, 1968.
Hansch and Leo, 1979.
HacKay et al... 1980.
Reddick and Burger. 1955.
Average standard deviation of log Y" measurements  Is O.OS.
Average standard deviation of log KQ,W measurements is 0.03.
                                             30

-------
                                        Table 9
                   AQUEOUS SOLUBILITIES AND OCTANOL/WATER PARTITION
                           COEFFICIENTS OF ORGANIC COMPOUND.-.
                                                     Esters
Aqueous Concentrations
Solute
Methyl nonanoate
Methyl decanoate
Ethyl acetate
n-Propyl acetate
n-Butyl acetate
Ethyl prcpionate
2-Bromoethyl
acetate
This Study
(H)
1.33 X 10~*
5(b)
2.05 X 10 3
0.726{b)
0.200tb)
5.77 X 10'*
0.148
fl.2«W
Literature
Value
(H)
—

0.912^
0.185te)
0.203te)
0.187(h)
—
w(1)
S • O/ "
Thls*^ Literature
Study Value
4.58 4.32{b)
5.36 4.41tb)
1.15 0.68{b) 0.73le)*
1.64 1.24^ 1.23<«>*
2.12 1 .82 1 .73
1.77 1.43(b)
11.1 i 11 '^ ' _--.
.DJ 1 . 1 1 -»»
(a)
(b)
* '
Indicates that the literature data are calculated values according to Hansch et^al_., 1968.
HPLC method of analysis.
GC method of analysis.
Sutton and Calder, 1975.
McAuUffe, 1966.
Hansch et aj.., 1968.
Kaii&Ch uiiU Leo, 1979.
MacKay et aj.., 1980.
Reddlck and Burger, 1955.
Average standard deviation of log y" measurements Is 0.04.
Average standard deviation of log K.  measurements Is 0.02.
                                              31

-------
                                                  Table 10
                              AQUEOUS SOLUBILITIES AND OCTANOL/WATER PARTITION
                                 COEFFICIENTS OF ORGANIC COMPOUNDS
                                                           Alcohols
Solute

Tnls Study
(H)
wH)
Aqueous Concentration Log YS v '
Literature
Value
L°9 K0/w
This")
Study

Literature
Value
1-Butanol
0.854
                              (b)
                                                                     1.0
                                                       0.785
                                                             (b)
            0.84v
1-Pentanol
O.U3
                              (b)
                                            0.249
                         (h)
                                             1.84
1.53
                                                                                    (b)
1.34
                                                                        (O*
1-Hexanol
4.14 X 10"
                                             2.28
                                                                                2.03
                                                                                     (b)
            1.84*
1-Heptanol
1.13 X 10
                                  .-21
                                                                     2.80
                                                        2.57
                                                                                     (b)
            2.34
                                                                        (e)*
1-Nonanrl
7.35 X 10'
                                             3.89
                                                                                3.77
                                                                                     (b)
2 Ethyl-1,3-hexanedlol   1.56 X 10"
                                                                      3.61
                                                        3.22
                                                                                     (b)
     (a)
     (b)
     (c)
     (d>
     (e)
     (f)
     (9)
     (h)
     (1)
     (j)
         Indicates that the literature data are calculated values  according  to  Hansch ejt al_. , 1968.
         HPLC method of analysis.
         GC method of analysis.
         Sutton and Calder, 1975.
         McAultffe. 1966.
         Hansch et aj_. , 1968.
         Hansch and Leo, 1979.
         MacKay et aj[. , 1980.
         Reddlck and Burger, 1955.
         Average standard deviation of log Y"
         Average standard deviation of log
                     measurements 1s 0.03.
                       measurements Is 0.03.
                                               32

-------
to
CO
                                              Table 11
                            Coefficients of the Regression Equation and
                            the Coefficient of Determination for the Solutes
Type of
Compounds
Aromatic
Hydrocarbons
Unsaturated
Hydrocarbons
Halogenated
Hydrocarbons
Normal
Hydrocarbons
Aldehydes and
Ketones
Esters
Alcohols
All compounds
Number of
Compounds
18
6
13
4
8
7
6
62
Slope {c}
1
1
0
0
1
0
1
0
.056
.024
.907
.972
.079
.932
.030
.944
+ 0.026
+0.061
+ 0.033
+_ 0.016
+ 0.065
+ 0.055
+ 0.011
+0.018
-0.
-0.
-0.
-0.
-0.
-0.
-0.
-0.
Coefficient of 2
Intercept(d) Determination {r )
768
585
323
468
465
285
348
311
i o.ioo
+ 0.267
+ 0.133
+ 0.081
+ 0.155
+ 0.167
+ 0.112
+ 0.066
0.
0.
0.
0.
0.
0.
0.
0.
990
986
986
999
978
982
994
980

-------
solubility, its molar volume and the regression equation for compounds
of its class.  Also, this study establishes that the generator column
method coupled with either HPLC or GC modes of analysis provides an accurate
and rapid method for systematic determination of K. and C* for organic
compounds.
     Equation 9 of Appendix I was rigorously derived from thermodynamic
principles.  Plots of log K,  versus log Y^ having slopes of unity are a
strong indication of the accuracy of the data.  The scatter or precision of
the data should be reflected in the coefficient of determination obtained from
a linear regression analysis of the data.  A coefficient near unity indicates
good precision in the measurements.
     The fact that the  slopes of the log KQ.  versus log y* plots for the
seven different classes of compounds are near unity (Table 11) is a strong
argument in  favor of tne generator column method giving accurate Cw and K  .
values and 1s a further validation of the method.
      It 1s Interesting  to analyze literature values of  K  .  and C* 1n the saute
manner and compare  these results with those obtained by the generator column
method.   In  Figure 4 are plotted lo<, KQ,W values measured by the generator
column against  literature log KQ.W values measured  by  the shake-flask method.
The  points that are obviously off 1n the literature log KQ/w range of 2 to 4
are  alkanes  that  have  been  calculated by the group  addttivUy method and are
iiOt  experimental  values.
      In  Figure  5  are plotted log C* values measured by  the generator column
method against  literature  log C* values measured by the shake-flask method.
Th*1  agreement is  much  better than that  1n Figure 5  because all the literature
values are experimental values.
                                    34

-------
                           6.
u»
U1
t—»
ce.
UJ
o.
X
                     C9
                     O
                           5.
                     ~    4.
      3.
                           2.
                           1.
                                                                                r~ » 0,913
                                                                     Log K.  (Exp) -.1.148 Log K,

                                                                          0/w  (Lit) -0.135      0/w
                                         1.
                                 2.          3.          4.

                                   LOG KQ/W  (LITERATURE)
5.
6.
                            Figure 4.   Experimental  Log KQyw Versus Literature Log

-------
CO
                   Ul
                   Cu
                   X
                   LJ
                  C3
                  O
                        5.
                   «!    4.
                                                                    I           rc = 0.993      u
                                                                    •Log C" (Exp) = -0.970 Log C* (Lit)
                                                                         5     +0.011
                            L/  I
                                       1.
2.          3.           4.

   - LOG C* (LITERATURE)
                          Figure 5.  Experimental Log C^ Versus Literature   Log C
                                                                                  ,w

-------
     In Figure 6 are plotted literature log K .   values versus  literature
log 1/C* values.  The large scatter in the points indicates  the accuracy
and/or the precision of the measurements is not  very good.   In  Figure  7 are
plotted log K  ,  versus log 1/C* where both K .   and C* were measured  by  the
generator column method.  The error bars indicate the precision of the
measurements.  A linear regression analysis of the data gives a slope  of
0.994 with a coefficient of determination of 0.998 indicating that the
measurements are accurate and precise.
     In conclusion we feel that the Figures 4 through 7 and the regression
analysis of the data on the seven different groups of compounds show that the
slopes of the  log KQ,W versus log Y* Plots and the coefficients of
determination  of these plots are all approximately equal to unity.  This  is
further validation of the generator column method.
                                   37

-------
     5.
     4.
     3.
ts
o
                                      •            r2 « 0.903

                                       Log KQ/W • -0.657 Log q +0.993
    2.
    1.
                  1.
      3.


- Log C*
                                               4.
Figures
                  Log KQ/W Vers0s -Log C^ for the Literature  Values.

-------
CJ
                          5.
                          4.
                     I   3.
                    g
                          1.
                       r  » 0.994
         Cog KQ/w = -0.891 Log cj  + 0.730
                                                                I
                         I
I
                                        1.
2.
                           Figure 7.  Log KQ/w Versus -Log
    3.          4.           5.

 -  Log C*

C^  for cfTe experimental Values,
            6.

-------
                     VII.  RECOMMENDATIONS FOR FUTURE RESEARCH
     In any research endeavor it is frequently the case that more new
questions are asked than old ones answered.  This is the case for this
work.  Two studies were initiated that are incomplete:  (1) salinity
dependence of K  .  and Cw and (2) the temperature dependence of K .   and
C .  Information obtainted from these studies is important for determining
 Vr
the rate and transport of organic substances in the environment.
Sal i ni ty Dependence
     It can be shown from thermodynamics that the octanol/salt water
partition coefficient K-sw is related to the concentration of the aqueous
salt solution, C$, by equation 1:
          log KQ/SW - log K0/w * ks Cs                                (1)
where k$ 1s the  salting out coefficient which is characteristic of the  solute.
     Thermodynamics also shows that for compounds of similar chemical
structure k$ is  related to the solute molar volume 9 by equation 2:
          ks » k, + k2 7                                              (2)
where kj and k~  are series constants*  Combining equations 1 and 2
          log KQ/SW - log K0/w Mk, + k2 7) Cs                       (3)
     To date we  have data on two different classes of compounds, the
n-alcohols and the alkylbenzenes.  These data were taken in order to
test equation 3.  It would be desirable to have data on other types of
compounds in order to generalize equation 3.
Temperature Dependence
     The temperature dependence of aqueous solubility may be expressed  as
                              AH
          d log  Cj/dO/T) * - -^
AH   is the heat  of solution, and R Is gas constant.  Very little is  known
                                    40

-------
about the heats of solution of hydrophobic substances. Since water temperature
varies from -400C (in the artic) to 40°C (near the equator), it is desirable
to study temperature dependence in order to make accurate predictions of the
aqueous solubility of organic compounds.
Head-Space Measurements
     During the course of this work it wc-.s realized that it would be
desirable to have another method (other than shake-flask) for measuring
*o/w for volatile P°lap compounds.  The shake-flask method gives unreliable
results because of the relative large solubility of water in these compounds.
The head-space method for determining K.  [Brinckman and Bellama, 1978]
avoids these experimental errors because there Is no solute-water interface
in these measurements.  These results should be very helpful In the development
of a reliable predictive scheme.
Tests for Other Classes
     Additional reliable data on monofunctional groups should be generated to
develop a sound prediction scheme.  Compounds containing multifunctional
groups should be studied and the results analyzed using additivity rules for
functional groups to assess the effects of steric hindrance.  Also, the aqueous
solubility of organic mixtures should be studied in order to develop a sound
predictive scheme for mixtures.
     HacKay ejtaU [19&0] have developed an equation which may be used
to predict the aqueous solubility and octanol/water partition coefficient
of rigid  (solid) molecules.  However, predictions are generally poor.
Here our  study will generate data for rigid and flexible molecules in the solid
state and will develop suitable expressions to explain their aqueous solubilities.
     In the study Just completed* there was an attempt to choose
compounds to represent a wide selection of chemical types which would be
of environmental interest.  Unfortunately one important group of chemicals
                                      41

-------
was left out, the organometals [Brinc^n and Bel lama,  1978].   The occurrence
of these compounds in water Is known to cause sickness  and deaths.  It is,  therefore,
desirable to determine the ^ueous solubility and octanol/water partition
coefficient for this class of compounds.
     This study will, of course, include other classes  of compounds
of immediate interest to EPA.
Predictive Schemes
     Success of a predictive scheme depends crucially on the accuracy
of the limited experimental data on which the calculations must rest.
Hansch et aV [1968] have developed a predictive scheme based on group
contributions.  However, there has been some concern on their reported *
values.  Here our goal will be to develop a more comprehensive predictive
scheme.  This Is based on the experimental results obtained in this
laboratory and should also include data on compounds containing multifunctional
groups that have not been measured to date.  K  .  and C* measurements should
be made on compounds containing multifunctional groups that have obvious
steric hinderence.  KQ.  measurements should be taken on compounds of
increasing complexity in order to determine where the group additivity breaks
down.
     Using thermodynamics and a simple model, log K  .  and log cj will be
related to some simple physical properties (such as, solute molar volumes,
boiling points, and carbon numbers).  Our preliminary study suggests
that log KQ/w and  log cj may be correlated to the solute molar volumes.
     Organic pollutants are often multlcomponent mixtures.  It is,
therefore, desirable to develop thermodynamic relations which can predict
the aqueous  solubility of each component in the mixture.  These will be
modeled on hydrocarbon-hydrocarbon systems but may be used to predict
the solubility of  complex mixtures such as DOT-kerosene systems.
                                      42

-------
                           VIII.   REFERENCES
Andrews, L. and R. Keefer.  1950.  Cation Complexes  of  Compounds Containing
     Carbon-Carbon Bonds.   VIII.  Further Studies on  the Argentation of
     Substituted Benzenes.   J.  Am. Chem. Soc.  72:  5C34-5037.
Bohon, R. and W. CUussen.  1951.  The Solubility of Aromatic  Hydrocarbons
     in Water. J. Am. Chem. Soc.  73: 1571-1578.
Brinckman, F. and J. Sena^a, eds. 1978. Organoroetals and Organometalloids.
     ACS Symposium Series, Washington, D.C. 314-326.
Conder, J. and C. Young. 1979.  Solution Thermodynamics,  Chapter  5  in  Physico-
     chemical Measurements by Gas Chromatography.  John Wiley and Sons,
     New York. 154-221.
DeVoe, H., H. Miller, and S. Waslk. 1981. Generator Columns  and  High  Pressure
     Liquid Chromatography for Determining Aqueous Solubilities  and
     Octanol-Water Partition Coefficients of Hydroohic Substances. N.B.S.
     J. Res.  86:  361-366.
Edisbury, 0.  1967. Practical Hints of Absorption Spectrometrv (Ultraviolet
     and Visible). Plenum Press,  New York.
Gill, S.,  Nichols, and  I. Wadso.  1976.  Calorlmetrlc Determination  of
     Enthalpies  of Solution of Slightly Soluble Liquids. J.  Chem.  Thermo.
     8: 445-452.
Hansch, C. and A. Leo.  1979. Substituent Constants for Correlation
     Analysis in Chemistry and Bloloqy.  John Wiley and Sons, New York.
Hansch, C. and W. Dunn, III. 1972.  Linear  Relationships Between
     Llphophilic Character and Biological Activity of Drugs. J.  Pharm.
     Sci.  6JL: 1-19.
Hansch, C..  J. Quinlan, and G. Lawrence. 1968. The Linear Free-Enemy
     Relationship Between Partition Coefficients and the Aqueous Solubility
     of Organic Liquids. J. Org.  Chem.  33:  347-350.
                                    43

-------
Hermann, R.  1972. Theory of Hydrophobic Bonding.  II.  The Correlation
     of Hydrocarbon Solubility in Water with Solvent  Cavity Surface
     Area. J. Fhys. Chem. 76; 2754-2759.
Iwasa, J., T. Fujita, and C. Hansch. 1965. Substituent Constants for
     Aliphatic Functions Obtained from Partion Coefficients. J.  Med.
     Chem. 8.: 150-153.
Leo, A., C. Hansch, and D. Elkins. 1971. Partition Coefficients  and Their
     Uses.  Chem. Rev. 71_: 525-616.
MacKay, 0., A. Bobra, W. Shiu, S. Yalkowsky. 1980. Relationships Between
     Aqueous Solubility and Octanol-Water Partition Coefficients. Chemosphere.
     2: 701-711.
May, W., S. Wasik, and D. Freeman. 1978. Determination of the Aqueous
     Solubility of Polynuclear Aromatic Hydrocarbons  by a Coupled Column
     Liquid Chromatographic Technique.  Anal. Chem. 50: 175-179.
May, W., S. Wasik, and D. Freeman. 1978. Determination of the Solubility
     Behavior of Some Polycycllc Aromatic Hydrocarbons In Water. Anal.
     Chem. 50: 997-1000.
McAuliffe, C. 1966. Solubility 1n Water of Paraffin,  Cycloparaffln, plefin.
     Acetylene, Cycloolefin, and Aromatic Hydrocarbons. 0. Phys. Chem.
     70: 1267-1275.
Neely,  W., D. Branson, and G. Blau. 1974. Partition Coefficient to Measure
     Bioconcentration Potential of Organic Chemicals in Fish. Environ.
     Sci. Technol. 8: 1113-1115.
Reddick, J. and W. Burger. 1955. Techniques of Organic Chemistry vol. 2.
     Wiley Interscience  Inc., New York.
Sutton, C., and J.A. Calder. 1975. Solubility of Alkylcenzenes in Distilled
     and Seawater at 25.0°C. J. Chem. Eng. Data. 20;  320-322.
Tewari, Y. M. Miller, and S. Hasik. Calculation of Aqueous Solubility of
     Organic Compounds.  In press.
                                  44

-------
                              APPENDIX I

     The octanol/water partion coefficient,  K.  , is defined as  the
ratio of molar concentration of solute s  in  octanol  saturated with water,
C°, to its concentration in water saturated  with octanol, C*, under
 S                                                         5
equilibrium condition in an octanol/water system:
               Ko/wsCs'Cs                                     <')
Furthermore, the chemical potential of the solute in water, u",  and  in octanol,
ti°, may be expressed by the following equations:
               p* - M* + RT In Y" • *J                           (2)

               pj • u* + RT In Y° . *°                           (3)
where  j.  is the chemical potential of pure solute, R is the gas constant,
T  is the  temperature of the system in Kelvin, $w and $° are, respectively,
                                               5      S
the volume fraction of solute in water and in octanol, and yw and Y° are the
solute activity coefficients on a volume fraction basis, where Y  approaches
unity  as  t  approaches unity, in water and in octanol, respectively.
     Under equilibrium condition, u* = u°, hence
                                                                 (4)
 Similarly  for a  solute  in equilibrium with its aqueous solution (jiw * ys)
 the  activity  coefficient may also be expressed as
               *s  *s  '  t *J
 where Y| and  ^  are,  respectively, the solute activity coefficient and
 volume fraction  In solute.  For a solute of low aqueous solubility YS
                                      45

-------
approaches unity as $* approaches unity,  hence
                    rs

                w ,w
or             Y*    * ($*)                                      (&)


At infinite dilution solute concentration in aqueous phase


(C*) and in octanol phase (C°) may be expressed by the following
  S                         *


equations
          and  C. - */?


where V^ and 7° are, respectively, the solute partial molar


volume  In water and in octanol.


     Combining equations 1, 4, and 7 we get
           K    » —  —
           *o/w  vo  vo

                   5                 w


 or log K0/w -  log  Y" -  log Y° * 1°9 r|-                         (8)

                                     s



 But for all practical purposes 9* = y°  > v$, the molar volume of


 pure solute.  Hence equation  (8) reduces to


           log  KQ/W » log YJ -  log Y®                             (9)
                                     46

-------
                                   APPENDIX  II
                         Proposed Method for Determining
                     Aqueous Solubility of Organic Compounds

     Scope - This method covers the determination of aqueous  solubility of
both liquid and solid compounds.  There are no restrictions on the type of
compounds.
     Summary of Method  - Generator columns packed with a solid support and
loaded with an organic substance make it possible to rapidly  and conveniently
equilibrate water with the organic phase by coupling the generator column
to an extractor column for high pressure  liquid chromatographic analysis
of the aqueous solution.   Errors from surface adsorption and loss to the
atmosphere are avoided.  Another method for determining the solute
concentration in the aqueous phase is by  solvent extraction of the aqueous
pnase followed by gas chromatographic analysis of the solvent extract.
     Apparatus
     1.   Generator column - Two different designs were used depending on whether
the  eluted  aqueous phase was analyzed  by  HPLC  (Procedure A) or by solvent
extraction  followed  by GC analysis of  solvent  extract  (Procedure B).
     The design  of the generator column for procedure A 1s shown  in Figure
 II-l.  The  column consists  of  a 1/4-Inch O.D. Pyrex tube joined  to a short
enlarged section of  9mm  Pyrex  tubing which  In  turn  is  connected to another section
of 1/4-Inch O.D,  Pyrex tublnq.  Connections to inlet Teflon tubing
 (1/8-Inch O.D.)  and  to outlet  stainless steel  tubing (1/16-inch O.D.) are by
means  of stainless steel fittings with Teflon  ferrules.
      2.   Constant temperature  bath with circulation pump-bath should be
 capable of  controlling  temperature to  * 0.05°C.
                                       47

-------
                 Inlet
        20 cm
                          Glass Wool
                          Support (100-120 mesh

                                     Chromsorb W)
                        6 mm


                       — Glass WOOL
                          mm
                 ~T  i
                 rru
                    I
                 Outlet

Figure II-l  Generator column.
                         48

-------
     3.  High pressure liquid chromatograph equipped with  a  UV  detector.
     4.  Extractor column - 6.6 X 0.6 cm stainless steel  tube with  end  fittings
containing 5 urn frits filled with a superficially porous  phase  packing
(Bondapack C,a/corasil Waters Associates).
     5.  Two 6-port high pressure rotary valves.
     6.  The design of the generator column for procedure 8 is  shown in
Figure H-2.  The column consists of a 1/4-inch Pyrex tube jointed  to a short enlarged
enlarged section of  9mm O.D. Pyrex tubing.   The outlet end of the 9mm 0.0.
tubing is connected to a long section of 3mm O.D. stairless steel tubing.
'i>e column is enclosed in a water jacket for temperature control.
     7.  Collecting tube for procedure B (Figure II-2) -  8X3/4 inch
section of Pyrex tubing with a flat bottom 1s connected to a short  section of
3/8-inch O.D. Pyrex tubing.  The collecting tube Is sealed with a 3/8-inch
Teflon cap  fitting.
     8.  Gas chromatograph with hydrogen flame detector.
     Procedure A - HPLC Method
     Procedure A covers the determination of the aqueous solubility of
compounds which absorb In the U.V.  The HPLC analytical system is shown schematically
In Figure II-3.  Two  reciprocating piston pumps deliver the mobile phase (water
or methanol-water mixture) through two 6-port high pressure rotary valves  and a
30 X 0.6 cm C-18 analytical column to an ultraviolet adsorption detector operating
at 254 nm.  Chromatogram peaks are recorded and Integrated with a recording
Integrator.
1.   Determination of response factor
     a.   Prepare known concentrations (mole/L (M)) solute in methanol
          (standard solution).
     b.   Inject samples of  standard solution into HPLC system using a
          calibrated  sample  loop.
                                     49

-------
       Bath •*
       Return
       Collecting
       Vessel
                          Cc Jurnn
                          Inlet
X
                     n
                             Generator
                             Column
p=
                                To Constant
                                Temperature
                                Bath
                              Extracting
                              Solvent
                             -V/ater Containing Solute
Figure II-2  Design of generator column for GC method.
                         50

-------
 MeOH
                      Sample Injection Valve
Switching Valve
                                                                              Switching Pattern:
                               Sample Loop
                                                                           T
                                                                     Weighing Bottle
                                                                        or Waste
                                                    Extractor Column
                     Analytical
                     Column
Load
Inject
Figure I1-3   Schematic of HPLC - generator column flow system.

-------
     c.    Adjust organic solvent (nethanol)/water  ratio  to obtain a reasonable
          retention time for solute peak.
     d.    Obtain an average peak area for  several  injections  of  standard
          sample at a set absorbance unit  full  scale (AUFS).
     e.    Calculate response factor from following equation:
               RF « (conc.(H)) (average area)(AUFS)
2.   Loading of Generator Column
     a.    The design of the generator column is shown in Figure  INI.   The
          column consists of a 6mm of Pyrex tube jointed to a short enlarged
          section of 9mm Pyrex tubing.
     b.    To pack the column a plug of silanized glass wool  Is inserted into
          one end of the 6mm Pyrex tubing.  Silanized diatomaceous silica
          support (about 0.5g 100-120 mesh chromosorb W cleaned  by Soxhlet
          extraction with ethanol and dried) is poured into  the  tube with  tapping
          and retained with a second plug of silanized glass  wool.
     c.   If the solute is a liquid, the column is loaded by  pulling  the
          liquid solute through the dry support with gentle  section.   If the
          solute Is a solid, a IX solution of the  solid in a  volatile  solvent
          is added to the dry packing.  The solvent is then  distilled  off  the
          column under reduced pressure.
3.   Analysis of Solute
     a.   Pump water to the generator column by means of a minioump.
     b.   With the switching valve (Figure II-3) In the inject position, pump
          water through the generator column at a  flow rate of 1 ml/m1n for
          approximately 5 minutes.
     C.   Switch HPLC pump to 100* water.
     i.   Weigh a 25 mL weighing bottle.
                                     52

-------
e.   Once the water reaches the HPLC detector which is indicated  by a  negative
     reading on the detector, simultaneously place the weighing bottle at
     the waste position and turn the switching valve to the load  position.
f.   Switch HPLC pump back to organic (methanol)/water mixture.
g.   After collecting approximately 10 ml of water in weighing bottle
     turn the switching valve back to inject position.
h.   Replace the weighing bottle with the waste container.
1.   Turn on the integrator.
j.   Weigh the weighing bottle containing water sample.
k.   Determine average area/g. of water collected at a selected AUFS
     setting from several collections.
1.   Calculate the solute concentration in water by the following
     equation:
          solute concentration (M) = (RF) (average area/g){AUFS)(997g/L)
          (volume of sample loop in L)
Procedure B - GC Method
1.   Determination of Calibration Curve
     a.   Prepare solute standard solutions of varying concentrations
     b.   Select a column and optimum GC operating conditions for suitable
          resolution between the solute and solvent and the solute and
          extracting solvent.
     c.   Inject a known volume of each standard solution into iaiecton
          part of GC.
     d.   Determine average area/ul for each standard solution.
     e.   Determine linear regression equation of concentration vs area/pi
          for 1n the following form
          conc(M) • a ££2>+ b
                                53

-------
2.   Loading of Generator Column
     a.   Generator column is packed and loaded with solute in  the  same manner
          as for the HPLC method.
     b.   Approximately 20 cm of straight stainless steel  tubing is connected
          to the bottom of the generator column and a beaker is placed beneath
          the tubing to collect the water.
     c.   The top of the generator column is connected to a water reservoir
          (Figure I1-4) using Teflon tubing.
     d.   Air or nitrogen pressure (5 P.S.I.) from an air or nitrogen cylinder
          is applied to the water reservoir thus forcing water from the  reservoir
          through the column to the beaker.
     e.   Water is collected in the oeaker for approximately 10 min. while
          the solute concentration In water equilibrates.
3.   Collection and Extraction of the Sotute
     a.   During the equilibration t1met a known weight of extracting solvent
          1s added to a collection vessel which contains a glass-enclosed
          stirring rod and can be capped.  The extracting solvent should completely
          cover the bottom of the collection vessel.
     b.   Record the weight of collection vessel with cap and extracting
          solvent.
     c.   Remove the cap  from the collection vessel and place  it under the
          generator column so that water  from the generator column enters below
          the  level of the extracting solvent (Figure II-2).
     d.   When the collection vessel Is filled, remove it from under tne
          generator column, replace the cap* and weigh the filled vessel.
     e.   Determine the weight of water collected.
                                      54

-------
rr "E^r->
( To
Compressed
Gas Cylinder

i i
                             Water
                   (ED
                        To Generator
                        Column Inlet
Figure II-4  Water reservoir for GC method.
                   55

-------
     f.   Place the collection vessel  on a  magnetic  stirrer and  stir  the
          contents for approximately 30 min.  controlling  the rate of  stirring
          so as not to break the meniscus between  the extracting solvent
          and water layers.
4.   Analysis of the solute
     a.   After stirring remove a known volume of  extracting solvent  from
          the vessel using a microliter syringe and  inject into  the G.C.
     b.   Record the peak area/pL injected  and from  the regression equation
          of the calibration line, determine  the concentration of solute
          In extracting solvent.
     c.   The concentration of solute in water (H),  c" ,  is determined  from
          the following equation
          where c" is the concentration of solute in extracting solvent
          (M), dH 0 and d*s are the densities of water and extracting  solvent,
          respectively,  and ggs and gH Q are the grams of extracting  solvent
          and water, respectively, contained in the collection vessel.
     d.   Replicate injections are made from each collection vessel  to determine
          an average concentration in water for the vessel.
     e.   At least one additional collection vessel Is prepared 1n a similar
          manner for each generator column.
                                   56

-------