EPA-600/2-76-220
December 1976
Environmental Protection Technology Series
EXTRACTION OF CHEMICAL POLLUTANTS FROM
INDUSTRIAL WASTEWATERS WITH
VOLATILE SOLVENTS
Robert S. Kerr Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Ada, Oklahoma 74820
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into five series. These five broad
categories were established to facilitate further development and application of
environmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
This report has been assigned to the ENVIRONMENTAL PROTECTION
TECHNOLOGY series. This series describes research performed to develop and
demonstrate instrumentation, equipment, and methodology to repair or prevent
environmental degradation from point and non-point sources of pollution. This
work provides the new or improved technology required for the control and
treatment of pollution sources to meet environmental quality standards.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-76-220
December 1976
EXTRACTION OF CHEMICAL POLLUTANTS FROM
INDUSTRIAL WASTEWATERS WITH VOLATILE SOLVENTS
by
Jonathan P. Earhart, Kwang W. Won,
C. Judson King, and John M. Prausnitz
Department of Chemical Engineering
University of California
Berkeley, California 94720
Grant No. R801030
Project Officer
Jack H. Hale
Industrial Section
Robert S. Kerr Environmental Research Laboratory
Ada, Oklahoma 74820
ROBERT S. KERR ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ADA, OKLAHOMA 74820
-------
DISCLAIMER
This report has been reviewed by the Robert S. Kerr
Environmental Research Laboratory, U.S. Environmental
Protection Agency, and approved for publication. Approval
does not signify that the contents necessarily reflect the
views and policies of the U.S. Environmental Protection
Agency, nor does mention of trade names or commercial pro-
ducts constitute endorsement or recommendation for use.
n
-------
FOREWORD
Man and his environment must be protected from the
adverse effects of toxic and hazardous chemicals, metals,
and other forms of pollution; including the unwise manage-
ment of residuals. Efforts to protect the environment
require a focus that recognizes the interplay between the
components of our physical environment--air, water, and land.
The Robert S. Kerr Environmental Research Laboratory con-
tributes to this focus through multidiscipiinary programs
engaged in
- studies on the removal of environmental contaminants
from waste discharges to the environment, and
- the search for ways to prevent contamination and to
recycle valuable resources.
The information presented in this grant report is the
culmination of research on a method designed to reduce the
contaminent load of certain industrial wastewaters and recover
valuable chemicals.
m
-------
ABSTRACT
This report presents the results of an experimental
study and general process evaluation of the extraction
of organic pollutants from wastewaters from petroleum
refineries and petrochemical plants, using volatile
solvents. Three basic approaches are considered; the
first uses a single extraction step with a volatile
solvent; the second uses successive extractions with
a less volatile, polar solvent followed by extraction
with a volatile solvent; and the third uses extraction
with a mixture of polar solvent and volatile solvent,
followed by extraction with the volatile solvent alone.
The principal volatile solvents considered are iso-
butylene and isobutane.
Equilibrium distribution coefficients were measured
for numerous solutes distributing between water and
various solvents. These reflect solvent capacities
and sel activities.
A miniplant extraction system was constructed,
using both RDC and spray-column extractors, coupled
with solvent circulation and regeneration systems. The
miniplant was used to test and demonstrate extraction
processes for seven different real industrial waste-
waters, and for various synthetic waters.
Volatile-solvent extraction is evaluated and compared
with other competitive means of handling high-oxygen-demand
wastewaters -- e.g., steam stripping -- and is found to
be economically promising for cases where the organics
IV
-------
load is high, where recovery of organic solutes is
desireable, or where the water presents special problems
for biological treatment systems.
This report was submitted in fulfillment of Grant
No. R801030 by the University of California, Department
of Chemical Engineering, under the sponsorship of the
Environmental Protection Agency. Work was completed as
of September 1975.
-------
TABLE OF CONTENTS
Page
LIST OF FIGURES xii
LIST OF TABLES xv
ACKNOWLEDGEMENTS xix
SECTIONS
I. CONCLUSIONS 1
II. RECOMMENDATIONS 4
III. INTRODUCTION AND BACKGROUND 6
Categories of Waste Water
Treatment Processes 6
General Considerations in Solvent
Extraction for Waste Water Treatment 8
Previous Applications of Solvent
Extraction for the Recovery of
Phenolic Compounds 10
Recovery of Non-Phenolic Pollutants by
Solvent Extraction 16
Types of Extraction Devices 18
Objectives of This Research 19
IV. GENERAL PROCESS CONSIDERATIONS 21
Considerations Pertaining to Steam
Stripping 22
Typical Cost of Steam Stripping 24
Considerations Pertaining to Solvent
Extraction 29
Choice of Volatile Solvent 33
Considerations in Volatile Solvent
Extraction 38
Typical Cost of Volatile Solvent
Extraction 43
Comparison of Volatile Solvent Extraction
with Steam Stripping 50
VII
-------
Dual Solvent Extraction Processes 56
V. BASES FOR EXPERIMENTS AND INDUSTRIAL
WASTEWATERS EMPLOYED' 61
Industrial Waste Water Samples 63
Lube oil refining waste water 63
Cresylic acid recovery waste water 66
Ethylene quench waste water 67
Oxychlorination waste water 69
Phenol-Formaldehyde Resin Manufacture
Wastewater 73
Hydrofiner Condensate Wastewater 74
Styrene Manufacture Wastewater 75
Choice of Type of Mini plant Extractor 76
Axial Mixing and Mass Transfer in
Continuous Extractors 78
VI. EXPERIMENTAL APPARATUS AND PROCEDURES 91
Analytical Methods SI
Equilibrium Determinations 98
Spray Column Extractor 98
Rotating Disc Contactor 102
Solvent Regeneration Apparatus 109
Chemicals Used 116
Experimental Procedure 118
VII. SPRAY COLUMN - EXPERIMENTAL RESULTS 127
Method of Data Reduction 129
Estimation of physical properties 130
Estimation of hold-up and Peclet
numbers 132
Circulating-drop estimate of mass
transfer 133
Osci11ating-drop estimate of mass
transfer 1 33
Experimental mass transfer results 133
Choice of Dispersed Phase ^39
vn i
-------
Choice of Type of Volatile Solvent 147
Interaction Among Solutes 150
Regeneration of Loaded Solvent 156
Industrial Waste Waters 159
Overall Process Feasibility 165
VIII. RDC EXTRACTOR - EXPERIMENTAL RESULTS 167
Test Run to Check RDC Correlations 171
Experiments on Lube Oil Refining Waste
Water 182
Summary of Experiments on Lube Oil Refining
Waste Water 219
Experiments on Ethylene Quench Waste Water 223
Experiments on Oxychlorination Waste Water 227
Experiments on Phenol-Formaldehyde Resin
Manufacture Waste Water 233
Experiments on Hydrofiner Condensate
Waste Water 238
Experiments on Waste Water from Styrene
Manufacture 241
IX. EFFECTS OF SCALE ON RDC DESIGN 244
Steps in Developing the Design of an
ROC Extractor 244
Considerations in Changing Scale of
an RDC 246
Sizing a Pilot Plant RDC 258
X. SUMMARY AND DISCUSSION OF APPLICATION 261
Strategy of Process Selection 262
Examples Illustrating Stretegy of
Process Selection 270
XI. NOMENCLATURE 273
XII. REFERENCES 277
XIII. PUBLICATIONS 285
-------
APPENDICES
A. BASIS FOR COST ESTIMATES 236
Estimated Total Plant Investment 286
Annual Operating Costs 288
B. HYDRODYNAMICS, AXIAL MIXING AND MASS
TRANSFER IN CONTINUOUS EXTRACTORS 291
Spray Column Extractors 291
Rotating disc contactors 302
C. DEVELOPMENT OF THE DISPERSION MODEL 316
D. ALTERNATIVE PROCESSES FOR VOLATILE
SOLVENT DISTILLATION 340
E. DISTRIBUTION COEFFICIENTS FOR ORGANIC
SOLUTES BETWEEN ISOBUTYLENE OR
ISOBUTANE AND WATER 354
Introduction 354
Apparatus 354
Procedure 356
Sampling 357
Chemical Analysis 359
Results 362
Correlation of Results: Theory of Dilute
Solutions 362
Evaluation of Size Parameter V* 379
Correlation of Characteristic Volume V* ...381
F. DISTRIBUTION OF PHENOLIC SOLUTES BETWEEN
POLAR ORGANIC SOLVENTS AND WATER 390
Introduction 390
Experimental 391
Effect of Solute Concentration 392
Thermodynamic Analysis 404
Distribution Coefficient at High
Dilution 415
-------
Additional Solvents 418
Thermodynamic Relations 418
Data Correlation 423
Distribution Coefficients for Phenol
Derivatives 429
Nomenclature 432
References 433
G. EXPERIMENTAL DATA FROM SPRAY COLUMN
EXTRACTOR 434
Estimates of Physical Properties 434
Computer Programs 435
List of Experimental Data 450
H. EXPERIMENTAL DATA FROM RDC EXTRACTOR 473
Estimates of Physical Properties 474
Computer Programs 474
List of Experimental Data 486
I. LABORATORY EXTRACTIONS OF
OXYCHLORINATION WASTE WATER 502
Initial Experiment 503
Results and Discussion on Initial
Experiment 503
Additional Experiments 506
Additional Results and Discussion 506
J. METRIC CONVERSION TABLE 509
-------
LIST OF FIGURES
Figure Pagt
1. The Phenolsolvan Process 14
2. Steam Stripping Process 25
3. Typical Volatile Solvent Process 34
4. Detailed Volatile Solvent Process 42
5. Total Cost of EDC Recovery 47
6. Breakdown of Costs for EDC Recovery
(Excluding Extractor) 48
7. Details of Volatile Solvent Extraction
Process 52
8. Dual Solvent Process (Separate Solvent Cycles)57
9. Dual Solvent Process (Linked Solvent Cycles) 58
10. Rotating Disc Contactor 77
11. Comparison of Plug Flow and Dispersion Models 84
12. Plot of Equation 3 to Illustrate Additivity of
Resistances 87
13. Volatile Solvent Sampler 95
14. Spray Column Extractor 100
15. Distributor Plates 101
16. Rotating Disc Contactor 103
17. Mid-Column Bearing 105
18. Lower Part of PDC 107
19. Upper Part of RDC 108
20. Volatile Solvent Regenerator 110
21. Miniplant Evaporator ill
22. Reflux Accumulator 114
23. Photograph of the Entire Miniplant 115
24. Pilot Plant Flow Diagram 123
25. Mass Transfer Rates 135
26. Effect of Physical Properties 140
-------
27. Experimental Data for Run RS13 18°
28. Effect of Rotational Speed on Phenol Removal
at 10 GPM Waste Water Flow 250
29. Extractor Size for 10 GPM Waste Water Flow .. 252
30. Extractor Cost for 10 GPM Waste Water Flow .. 253
31. Effect of Rotational Speed on Phenol Removal
at 100 GPM Waste Water Flow 255
32. Extractor Size for 100 GPM Waste Water Flow . 256
33. Extractor Cost for 100 GPM Waste Water Flow . 257
Cl . Basis for Dispersion Model 318
C2. Distribution of Phenol Between Water and
n-Butyl Acetate 333
Dl. Simplest Alternative 344
D2. Details of Simplest Alternative 345
D3. Alternative with Feed Vaporizer 347
D4. Details of Alternative with Feed Vaporizer .. 348
D5. Alternative with Side Stream Boiler 350
D6. Details of Alternative with Side Stream
Boiler 351
El. Schematic of Equilibrium Cell Assembly 355
E2. Indalloy Encapsulation Sampling Device 358
E3. Chemical Analysis Instrumentation. Gas
Chromatograph and Induction Oven 360
E4. Induction Oven for Melting of Indalloy
Sample Capsule 361
E5. Distribution Coefficients vs. Characteristic
Volumes for n-Alkanes Between Water and
n-Butane (and n-Heptane) at 25°C 371
E6. Distribution Coefficients for Acetates
Between Water and C, Hydrocarbons at 25°C ...373
E7. Distribution Coefficients for Ketones Between
Water and C^ Hydrocarbons at 25°C 374
E8. Distribution Coefficients for Aldehydes
Between Water and C4 Hydrocarbons at 25°C ...375
E9. Distribution Coefficients for Phenolics
Between Water and C. Hydrocarbons at 25°C ...376
xiii
-------
E10. Effect of Carbon Number on the Characteristic
Volumes for n-Alkanes and n-Primary Alcohols
FT. Distribution Coefficients for Phenol and
Butyl Acetate Between Water and Butyl
Acetate at 25°C 393
F2. Distribution Coefficients for Phenol and
Methyl Isobutyl Ketone Between Water and
Methyl Isobutyl Ketone at 25°C 394
F3. Distribution Coefficients for Phenol and
Isopropyl Ether Between Water and Isopropyl
Ether at 25°C 395
F4. Distribution Coefficients for Phenol and
Butyl Acetate Between Water and Butyl
Acetate at 45°C 396
F5. Distribution Coefficients for Resorcinol
Between Water and Butyl Acetate at 25°C 397
F6. Effect of Solvation Equilibrium Constant, K ,
and that of Aqueous Concentration on
Distribution Coefficients for Phenol at 25°C 411
F7. Effect of Aqueous Phenol Concentration and
that of Volume Ratio (VR/V.) on Distribution
Coefficients for Phenol13 M at 25°C 412
F8. Concentration Dependence of Distribution
Coefficients for Phenol at 25° C. Predic-
tions by Chemical Theory and by Physical
Theory 414
F9. Experimental and Predicted Distribution
Coefficient for Phenol Between Water and
Diethyl Ketone at 25°C 416
F10. Effect of Solvation Equilibrium Constant K
and of Molar Volume Ratio VR/V. on s
Distribution Coefficients for Phenol at
High Dilution 422
Fll. Effect of Molar Volume of Polar Solvent on
the Solvation Equilibrium Constant K for
Phenol ! 427
xiv
-------
LIST OF TABLES
Table Page
1. Material Balance for n-Butyl Acetate
Recovery by Steam Stripping ................. 26
2. Cost Estimates for n-Butyl Acetate Recovery
by Steam Stripping ........... ............... 27
3. Solvent Solubility Losses ................... 32
4. Factors Affecting Solvent Choice ............ 36
5. Capital Costs for EDC Recovery .............. 45
6. Material Balance for n-Butyl Acetate Recovery
by Solvent Extraction ....................... 53
7. Cost Estimates for n-Butyl Acetate Recovery
by Solvent Extraction ....................... 54
8. Physical Properties for Run SS12A ........... 131
9. Predicted Mass Transfer Rates ............... 134
10. Experimental Mass Transfer Rates ............ 137
11. Comparison of Runs with Different Phases
Dispersed ................................... 1 42
12. Comparison of Runs with Different Phases
Dispersed ................................... 144
13. Additional Runs with Water as the Dispersed
Phase ....................................... 145
14. Comparison of Results When Using Different
Volatile Solvents ........................... 148
15. Interaction Among Solutes ................... 152
16. Interaction Among Solutes ................... 154
17. Extraction of n-Butanol and Propioni tril e ...158
18. Extraction of Phenol and o-Cresol ........... 160
19. Extraction of Lube Oil Refining Waste ....... 162
20. Extraction of Cresylic Acid Recovery Waste ..164
21. Results from RDC Test Run RS13 .............. 173
22. Physical Properties for Run RS13 ............ 174
23. Effect of Glfi on Prediction of Hydrodynamic
Characteristics of RDC ...................... 178
xv
-------
24. Effect of G-,0 on the Experimental Estimates
°f No« •••••• '"
25. Comparison of Predicted and Experimental
Solute Concentrations for Z = 0.034 and
fi = 0 2 181
b!8 u'^
26. Effect of G1Q on the Prediction of N,, for
the RDC ...1? ? ""S3
27. Effect of G,Q on the Prediction of Nc for
the RDC ...!? ? 134
28. Results from Run RS2 136
29. Results from Runs RS1A and RS1B 188
30. Results from Run RS3 191
31. Experimental Estimates of Nnu( for Run RS3 ...192
u w
32. Results from Run RS4 195
33. Calculated Effectiveness of the Dual Solvent
Process 199
34. Results from Run RS6A 201
35. Results from Run RS6B 202
36. Experimental Estimates of NQW for Run RS6A ..205
37. Experimental Estimates of NQW for Run RS6B ..205
38. Results from Run RS7A 207
39. Results from Run RS7B 208
40. COD and TOD for Runs RS6 and RS7 210
41. Overall Removals for the Dual Solvent
Process 211
42. Results from Run RS8 214
43. Experimental Estimates of NQW for Run RS8 ...216
44. Results from Run RS9 218
45. Theoretical and Experimental Mass Transfer
Estimates for Extraction with n-Butyl
Acetate 220
46. Results from Run RS10 225
47. Results from Run RS11 226
48. Experimental Estimates of N for Each Solute
in Runs RS10 and RS11 228
xvi
-------
49. Results from Run RS12
50. Results from Run RS15 234
51. Results from Run RS16 235
52. Results from Run RS17 239
53. Results from Run RS18 240
54. Results from Run RS19 243
55. Physical Properties for RDC Design Sample
Calculation 248
56. Summary of RDC Designs 260
Cl. Results from Sample Numerical Calculation ... 334
C2. Linear Approximation for K. Varying with
Solute Concentration 339
Dl. Costs for Simplest Alternative 345
D2. Costs for Alternative with Feed Vaporizer ... 348
D3. Costs for Alternative with Side Stream
Boi ler 351
El. Distribution Coefficients and Characteristic
Volumes for Acetates. Distribution Between
Water and Isobutylene (and Isobutane) at 25°C .. 363
E2. Distribution Coefficients and Characteristic
Volumes for Ketones. Distribution Between
Water and Isobutylene (and Isobutane)at 25°C.. 364
E3. Distribution Coefficients and Characteristic
Volumes for Aldehydes. Distribution Between
Water and Isobutylene (and Isobutane) at 25°C.. 365
E4. Distribution Coefficients and Characteristic
Volumes for Phenolics. Distribution Between
Water and Isobutylene (and Isobutane)at 25°C.. 366
E5. Distribution Coefficients and Characteristic
Volumes for n-Alkanes Between Water and n-
Butane (and n-Heptane) at 25°C 37C
E6. Constants in Equation (E8) with q = V*0'7 ...377
E7. Characteristic Volumes V* for Some Other
Organic Solutes (cm3/mol) 378
E8. Relative Characteristic Group Volume AV* for
Equation (E13) 384
E9. Experimental and Calculated V* for Some
Polar Fluids 335
xvi i
-------
Fl. Distribution Coefficients for Phenolics
Between Water and Two Organic Solvents
at High Dilution at 25°C 398
F2. Distribution Coefficients for Phenol and Butyl
Acetate Between Water and Butyl Acetate
F3. Distribution Coefficient for Phenol and Iso-
propyl Ether at 25°C 400
F4. Distribution Coefficients for Phenol and
Methyl Isobutyl Ketone Between Water and
Methyl Isobutyl Ketone at 25°C 401
F5. Distribution Coefficients for Phenol Between
Water and 1 ,2-Dichloroethane at 25°C 402
F6. Distribution Coefficients for Resorcinol
Acetate Between Water and Butyl Acetate at 25° 403
F7. Margules Constants for the Aqueous Phase
[Equations (F6) and (F7)] and Solubilities in
Solute-Free Water of Polar Solvents at 25°C..407
F8. Distribution Coefficients for Phenol at High
Dilution Between Water and Polar Solvent at
25°C 419
F9. Properties of Polar Solvents: Density, Solu-
bility in Water, Distribution Coefficient
for Phenol, K and Solvation Equilibrium
Constant, KS 424
F10. Reduced Solvation Equilibrium Constant £
for Nine Types of Polar Solvents :....428
Fll . Calculation of Distribution Coefficients for
Phenol Derivatives: Group Contribution,
JlnK0? for Methyl, Chlorine and Hydroxyl
GroHps 430
G. Spray Column Data 451
H. RDC Extractor Data 487
II. Results from Extraction of a Prepared Water
Solution using 2-Ethyl Hexanol 504
12. Results from Extractions of Prepared Water
Solutions which Contained Chloral Hydrate ...508
xvm
-------
ACKNOWLEDGEMENTS
This work was carried out in the Department of
Chemical Engineering of the University of California,
Berkeley, California. In addition to the principal
authors, important contributions were made by Richard
K. Herz, Ho-Yan Wong and James M. Wardell, all acting
as Research Assistants, and by James N. Michaels and
Stephen Loftus, acting as Laboratory Helpers. Thanks
are also due to the support staff of the College
of Chemistry, notibly the shops, for their contributions
of services and advice, and to Ms. Bonni Maunder for
her conscientious typing of this report. We are also
grateful for the frequent advice and assistance of
Messrs. L. Frank Mayhue and Jack H. Hale of the EPA staff
xix
-------
SECTION I
CONCLUSIONS
Solvent extraction is attractive for treatment
of wastewaters from petroleum refining and petrochemi-
cal plants in cases where there is a high value
of recovered chemicals, where constituents of the
water pose special problems for biological treatment,
and/or when a particular water stream has a much
higher organics content than other wastewaters
from the plant. A principal advantage of solvent
extraction is the opportunity for recovery, rather
than degradation, of dissolved and suspended organics.
Volatile solvents are attractive for extraction
processes for wastewater treatment because of their
low solubility in water at atmospheric pressure
and because of their ease of regeneration. An
analysis of candidate volatile solvents points
toward the general attractiveness of C. hydrocarbons
-- e.g., isobutylene and isobutane.
Measurements of equilibrium distribution coeffi-
cients and miniplant extraction runs show that
direct extraction with volatile solvents is effective
in many cases. There is a strong economic incentive
for the use of relatively low ratios of solvent
flow to water flow, e.g., 0.1. Because of this,
there is often an incentive for dual solvent processes,
-------
wherein a less volatile, polar solvent is used
to remove the primary pollutants from water and
then a volatile solvent is used to remove residual
amounts of the polar solvent. This may be accomplished
by contacting the water with the two solvents sequen-
tially, or by contacting the water first with a
mixture of the solvents, followed by extraction
with the volatile solvent alone.
The miniplant was used to carry out solvent-
extraction treatment of seven real, industrial
wastewaters. (1) High removals of phenol, cresols
and COD were obtained from a lube-oil refining
wastewater by dual-solvent extraction using n-butyl
acetate and isobutylene. (2) A substantial reduction
in COD was obtained for a condenstate from a process
manufacturing cresylic-acids from spent caustic,
using direct extraction with isobutylene. (3)
For quench water from an ethylene plant, a consider-
able reduction in COD and effective removal of
suspended solids were obtained using direct extraction
with isobutane. (4) For wastewater from an oxychlori-
nation ethylene-dichloride plant, chloral was effectively
removed using octanol as a solvent; the effluent
is then subjected to isobutane extraction for octanol
removal. It was found that extraction with octanol
is rate-limited by a chemical reaction of chloral
with the solvent. (5) For wastewater from the
manufacture of phenol-formaldehyde resins, effective
removal of phenol was achieved by dual-solvent
extraction using n-butyl acetate and isobutylene.
(6) Dissolved organics were effectively removed
from a hydrofiner condensate water by extraction
with methyl isobutyl ketone, followed by volatile-solvent
-------
extraction for reclamation of the ketone solvent.
(7) For wastewater from styrene manufacture, very
complete removal of dissolved aromatics and COD
was achieved by extraction with isobutylene. Scale-
up procedures for industrial units were considered.
These reflect key variables for further study and
show the desirability of on-stream pilot testing
with much higher water throughputs than could be
used in the present work.
In some cases, suspended solid material was
effectively removed during extraction with isobutylene
or isobutane; in other cases it was not. Factors
leading to effective removal of suspended solids
in wastewater extraction processes deserve further
study.
-------
SECTION II
RECOMMENDATIONS
1. The results obtained in this work are highly
encouraging. It is now evident that double-solvent
extraction (one polar solvent, one volatile nonpolar
solvent) and, in some cases, direct volatile-solvent
extraction provide a promising and economic technique
for removal of organic pollutants from industrial
wastewaters. Therefore, this technique merits further
study toward development of a large-scale process
suitable for industrial application.
2. One area in which we do not have a good under-
standing of the fundamentals involved is in the re-
moval of dispersed liquid pollutants. With the
ethylene quench waste water, the dispersed organics
were not removed by isobutylene extraction but were
well removed by isobutane extraction. A study of
the factors which make for good coalescence between
solvent and pollutant droplets would help determine
when dispersed pollutants could be recovered by
solvent extraction.
3. Several families of organic solutes are similar
to phenol and acetic acid with respect to the difficulties
in recovery by steam stripping. Further work is needed to
-------
determine which polar solvents would be good for
these solutes in a dual solvent process. The organic
acids from propionic to pentanoic would be solutes
which are particularly promising candidates for
recovery by dual solvent extraction.
4. Finally, the need for an intermediate-scale
pilot plant study is discussed in Section IX.
In addition to providing data for scale-up to a
commercial RDC extractor, such a study would be
useful in determining the validity of the general
correlations describing the operation of an RDC. In
particular the unexpectedly low removal found in this
work for solutes having an extraction factor less than
1 should be investigated further. By using an RDC
extractor with sample points all along the column,
the validity of the axial dispersion model for cases
where E is less than 1 could be checked. Answers to
these uncertainties would improve our ability to
evaluate the cost of an RDC extractor in the treat-
ment of many additional waste water streams.
-------
SECTION III
INTRODUCTION AND BACKGROUND
This report deals with solvent extraction as a
method of removal of pollutants from waste waters
from petroleum refining and petrochemical manufacture
processes. Attention is confined, for the most part,
to removal of organic solutes. The use of volatile
solvents is emphasized, either alone or in combination
with other solvents.
Categories of. Waste Water Treatment Processes.
Even when we limit consideration to organic
pollutants in industrial waste waters, there are a
variety of possible processing techniques. These
processes can be categorized into two broad groups.
First, there are the non-recovery techniques. In
these processes no attempt is made to recover the
pollutants in a concentrated form to help offset the
processing costs associated with waste treatment.
Some examples of non-recovery processes are biological
oxidation, carbon adsorption with thermal-oxidative
regeneration, direct incineration, deep well
injection, and solvent extraction where the loaded
solvent is either destroyed (e.g., by burning) or
-------
recycled to some part of the plant where the
pollutants are destroyed by existing process steps.
The second category of waste water treatment
processes is the recovery techniques. In these
processes the pollutants are recovered in a concen-
trated form so that they may be sold or used to
replace the purchase of make-up chemicals. Some
examples of recovery processes are solvent extrac-
tion with a secondary separation process to regenerate
the solvent, steam stripping, and carbon adsorption
with caustic washing for regeneration. The processing
costs are generally higher than for the non-recovery
processes because of such factors as the cost of
distilling a loaded solvent stream, the cost of
heat to raise the entire water stream to its boiling
point, or the cost of required chemicals. Thus the
recovered pollutant must be of sufficient value and
quantity for a recovery technique to be economically
competitive with a non-recovery process.
Although there are many possible waste water
processing techniques, the petroleum, petrochemical
and organic chemical industries presently rely
extensively on biological oxidation processes
(Beychok, 1967). Those aqueous process effluents
which have an appreciable biological oxygen demand
are usually combined and treated in a central
facility. Only the most highly contaminated,
toxic, or non-biodegradable streams are treated
on an individual basis. The potential advantage
of pollutant recovery from many of these individual
streams has recently been described (Fox, 1973).
-------
General Considerations in Solvent Extraction for
Waste Water Treatment.
The application of solvent extraction to selected
water effluents from individual processes can produce
valuable recovered pollutants, can solve problems
with non-biodegradable or toxic pollutants, and will
reduce the load on the central biological oxidation
plant. However, the cost of waste treatment will
make solvent extraction feasible only when applied
to certain water effluents.
The cost to process a given flow of polluted
water by biological oxidation or carbon adsorption
tends to increase as the pollutant concentration
increases. This increase in cost results from the
need to supply more oxygen and to provide a greater
residence time for biological oxidation, or from
the need to regenerate the carbon more frequently.
However, the cost to remove a fixed fraction of the
pollutants from a given flow of polluted water by
solvent extraction tends to remain approximately
constant as the pollutant concentration increases
since the necessary solvent flow does not change
greatly. Also, the profit from recovered pollutant
increases proportionally to the concentration of
pollutant. These factors tend to make solvent
extraction most economical when applied to the most
highly contaminated streams and biological oxi-
dation or carbon adsorption best for dilute streams.
One of the goals of this work is to identify
particular pollutants and particular waste water
problems which will be most economically controlled
by solvent extraction.
8
-------
Solvent extraction has been infrequently
applied to waste water treatment. The principal
application has been for the recovery of phenol
and higher molecular weight phenolic compounds.
Other applications include acetic acid recovery
and oily water treatment. The reasons for this
limited application include: (1) the losses of
solvent due to its solubility in the purified water
can result in unacceptable costs for solvent make-up
or solvent recovery, (2) the equipment and energy
for solvent regeneration can be very expensive, and
(3) synthesizing a good process requires the
designer to choose from a large number of alter-
natives such as type of solvent, type of extractor,
method of solvent regeneration, and method of
solvent removal from the purified water. One
approach to minimizing solubility and regeneration
costs is to use a very volatile, slightly soluble
solvent like isobutane. Development of this
approach has been a major part of this research.
Another factor which may have constrained the
development of solvent extraction processes is
the need to identify the chemical nature of the
major pollutants present in the waste water to
be able to evaluate and understand this advanced
processing technique. There is a considerable
literature describing waste streams from petroleum
(Beychok, 1967) and chemical (Jones, 1971) industries,
but the waste waters are usually characterized
in terms of overall pollution impact parameters
such as biological oxygen demand (BOD), chemical
oxygen demand (COD), turbidity, dissolved solids,
and oil concentration. The phenol concentration
is the only individual pollutant often identified,
-------
and this usually includes all phenolic compounds
instead of only phenol. In the combined streams
feeding the central biological oxidation process,
the identification of individual pollutants is
difficult because of the large number of pollutants
present. However, for an individual stream
where the number of major pollutants is much
smaller, techniques such as those based on gas
chromatography either with flame ionization detec-
tion (Herz, 1972) or in conjunction with mass
spectroscopy or infrared analysis (Sugar and
Conway, 1968) have been successfully used. Also,
the company that developed the offending process
should be able to provide a list of the pollutants
most likely to be present. With these approaches
now available, pollutant identification should
not be a limiting constraint to process development.
Previous Applications of Solvent Extraction for
the Recovery of Phenolic Compounds.
The extensive literature on the removal and
recovery of phenolic compounds by solvent extraction
has recently been reviewed (Kiezyk and MacKay, 1971)
By considering the details of the various processes
which have been developed for recovery of phenol and
other pollutants, we can begin to develop generaliza-
tions about the application of solvent extraction.
Phenol is a particularly severe organic pollutant
not only because of its contribution to the BOD
of the waste water, but also because it imparts
a medicinal taste and odor to drinking water which
can be detected in the concentration range 0.05
10
-------
to 0.10 ppm and, when the water is chlorinated,
the resulting chlorinated phenols can be detected
at about 0.005 ppm (Beychok, 1967).
First consider the physical characteristics
of phenol which have led to its preferred treatment
by solvent extraction. In biological oxidation
processes, soecial precautions must be taken since
the oxidation bacteria may be killed by phenolic
concentrations over 50 ppm (McKinney, 1967).
However, in the waste waters from the by-product
coke industry, from phenolic resin manufacture,
from phenol production plants, and from certain
petroleum refining operations, the concentration
of phenolic compounds can be several thousand
ppm or higher. At these concentrations a recovery
process should be favored. Phenol forms a minimum
boiling azeotrope with water which contains only
9.2 weight % phenol (Weast, 1970), so steam strip-
ping will not be as successful as it is in other
applications. Carbon adsorption with chemical
regeneration has been applied (Fox, 1973), but
the processing costs become large at such high
concentrations. Solvent extraction of phenol
may be the best process primarily because all other
alternatives are more expensive for phenol removal
than with other, more common pollution problems.
The earliest large-scale use of solvent
extraction for phenol recovery was associated with
the by-product coke industry in Germany. Rhodes
(1949) gives a detailed description of the three
types of solvent extraction processes which were
developed. The earliest process used benzene to
extract phenolics from the waste water from coke
11
-------
ovens. Initially the phenol was recovered by
distillation, but this proved too costly and later
regeneration of the benzene utilized a back-
extraction into a sodium hydroxide solution. The
process thus recovered the phenolics as a concen-
trated aqueous solution of sodium phenolate.
Eighteen of these plants were still in use in
1968 (Wurm, 1968). The second process used tricresyl
phosphate as a solvent for phenol extraction. This
solvent has a distribution coefficient for ohenol
about 8 times larger than that for benzene (Tupholme,
1933); therefore, a smaller volume of solvent per
unit of water treated was possible. Distillation
was used to separate the phenolics from the solvent.
Since tricresyl phosphate has a very high boiling
point, vacuum distillation was used, and the phenol
was recovered as the distillate. This resulted
in a very pure phenol product, but less volatile
phenolics and other organics tended to build up
in the recirculated solvent causing problems of
increased viscosity and decreased phenol capacity.
These problems led to the failure of this process
after it was applied for a short time in several
small plants in Germany.
The third process described by Rhodes is
known as the Phenolsolvan process. This process
is still being offered commercially, and it will
likely be used in several of the presently
planned coal gasification projects which utilize
Lurgi gasifiers (Beychok, 1974). The solvent
initially used was isobutyl acetate which has
a capacity for phenol extraction about 3 times
better than that of tricresyl phosphate. In the
12
-------
late 1950's a number of process changes were
incorporated including a change in solvent to
isopropyl ether. Although this change resulted
in a 60% reduction in distribution coefficient,
the advantages of lower cost and of higher volatility
both in solvent regeneration and in solvent recovery
from the purified water were claimed to justify
the change (Wurm, 1968).
The solubility of isopropyl ether is about
8000 ppm; therefore, it must be recovered from the
purified water. As shown in Figure 1 (Wurm, 1968),
this need for solvent recovery complicates the
process. The primary step is the countercurrent
contacting of feed water with regenerated
isopropyl ether in a multiple-stage mixer-settler.
The loaded solvent is then regenerated in a two-
stage distillation column, D, where the lower
portion operates as a steam stripper to remove
traces of solvent from the product phenolic mixture.
The three absorbers, A, B, and C, are used for
recovering dissolved solvent. In C the isopropyl
ether is stripped from the purified water using
an inert gas. The isopropyl ether is next absorbed
from the inert gas by contacting with phenol in
B. The small amount of phenol picked up by the
inert gas is then absorbed into the feed water in
A. Several other methods of dissolved solvent
recovery including steam stripping and extraction
with xylene have also been developed, but the
illustrated process is the preferred arrangement
for treating coke oven condensate (Wurm, 1968).
During petroleum refining, phenolic waste
waters are generated from catalytic and thermal
cracking operations. This water, which typically
13
-------
Feed ^
Water
\
J
A
^
i
m
™
i
— ^
I
1 TI K 1 B |
• i } |
— >
i LT-H ;
L_J 1
r i
* — r-p^
*3k 1
1 — n lu *"s
s i
+ i
~rn 1
T2 JL.L
f
1 Mixer- ^
Settler
^^
/
C
^L«M
1 Purified
^ Water
Equipment Legend
A,B,C, & D - Columns
Tl & T2 - Tanks
S - Steam Supply
Flow Legend
TT X.
Recovered Water
pnenoi Isopropyl
17 x. V, ^^-K-
jj uner
Phenol
Stripping
Gas
Figure 1. The Phenosolvan Process
-------
contains about 300 ppm of phenolic compounds, has
been extracted either with crude oil or with light
catalytic cycle oil to remove most of the pheno-
lics. The operation using crude oil is actually
a modification of the normal crude oil desalting
operation where a waste water rather than fresh
water is used to remove salt from the incoming
crude oil (Beychok, 1967). The extracted phenols
are subsequently destroyed in downstream refinery
operations. The process using light catalytic
cycle oil is offered to the petroleum industry
under the Phenex trademark (Lewis, 1968). The
cycle oil after treatment is blended with distillate
fuels where the phenols are beneficial in inhibiting
oxidation and improving color stability. In
both cases, the extraction is conducted in a
mixing valve followed by an electrostatic coalescer.
Since the phenols are not recovered in either
process, these techniques are examples of solvent
extraction in a non-recovery process.
About half the U. S. production of phenol is
by the cumene process in which cumene is reacted
with air to produce phenol and by-product acetone
(Stobaugh, 1966). Several side-products including
mesityl oxide are also produced, and a waste water
stream containing phenol and acetone is generated.
A logical route for reducing the pollutant content
of this waste water is extraction using the reac-
tant cumene stream as the solvent (Witt and Forbes,
1971). Bewley (1969) has determined that a mixture
of 20% mesityl oxide in cumene will increase the
phenol distribution coefficient about 10 times
over that for pure cumene. The loaded cumene
15
-------
will contain phenol, mesityl oxide, and acetone,
all of which are normal products of the cumene-
to-phenol reaction. Therefore, the existing separa-
tion sequence used in the main process can equally
well serve to regenerate the phenol extraction sol-
vent at a minimal increase in cost. Since the
recovered phenol ends up in the primary product, this
is a recovery technique with a unique opportunity for
solvent regeneration in the main processing equipment
Recovery of Non-Phenolic Pollutants by Solvent
Extraction.
The recovery of acetic acid from the effluent
water from cellulose acetate manufacture and from
semi-chemical pulping plants is of significant
economic importance (Brown, 1963). These waste
waters contain sodium acetate at 2-20%. After
acidification, the resulting acetic acid can
be economically recovered by solvent extraction.
The black liquor from semi-chemical pulping plants
is extracted with methyl ethyl ketone for the
recovery of both acetic and formic acids (Weaver
and Biggs, 1961). The cellulose acetate waste
water is usually extracted with ethyl acetate,
although ethyl and isopropyl ether have also been
used. In this application, the simultaneous
extraction of water is a very important considera-
tion, and the steps in solvent regeneration and
acetic acid concentration often involve azeotropic
distillations. Acetic acid extraction is more
difficult than phenol extraction in that the
distribution coefficients are much lower.
16
-------
Two processes have been developed which use
solvent extraction for the treatment of oily
waste waters. Strausser and Kurland (1970)
have patented a process for the purification of
ethylene plant process quench water. This quench
water contains a very stable emulsion of aromatic
hydrocarbons, light olefinic polymers, suspended
heavy tar polymers, and coke particles. The
water stream is contacted in a mixer-settler with
an aromatic distillate which contains 40 to 70%
benzene and which is a natural by-product of the
ethylene production process. The extraction after
complete settling and filtration produces a water
suitable for recycle to the quench tower. The
loaded aromatic solvent is separated from the
extracted tars in the existing distillation train.
King (1970) has patented a similar process for
treating the aqueous condensate from styrene manu-
facture. The condensate from the dehydrogenation
of ethylbenzene to produce styrene contains a
small quantity of dissolved styrene which tends
to polymerize and plug the process equipment.
In this extraction process, the water is treated
with fresh ethylbenzene which extracts the styrene
into the organic phase where it is less likely
to polymerize. The loaded solvent is recycled
to the dehydrogenation reactor, and the water is
steam stripped to recover dissolved ethylbenzene.
Both processes recover the pollutants by using
an existing process stream as solvent and existing
distillation equipment for regeneration.
Two other applications of solvent extraction
for waste water treatment are mentioned by Jones
17
-------
(1971) without details. The aqueous effluent
from a rubber processing plant was treated by
extraction with benzene for the removal of thiazole.
Salicyclic acid and other hydroxy aromatic acids
were extracted from a waste water using methyl
isobutyl ketone.
The preceding discussion comprises the known
applications of solvent extraction for industrial
waste water treatment. Clearly there must be
many other organic chemicals that cause problems
by their presence in an aqueous effluent but
that are not now treated by solvent extraction.
One of the objectives of this research is to
determine if there are fundamental barriers
which have limited wider application of solvent
extraction.
Types of Extraction Devices.
The number and variety of liquid-liquid
contactors that have been proposed is considerable,
varying from the simplest spray towers to high-
performance centrifugal types. The well-established
extractors are described in the standard works
(Treybal, 1963) and in several review articles
(Hanson, 1968; Akell , 1966; Reman, 1966). The
special requirements for use in treating waste
water do not eliminate many types of extractors from
consideration, although the presence of solid parti-
cles in a waste water stream might cause plugging
problems in a packed tower.
For the recovery of phenol, commercial instal-
lations using a packed column extractor (Edmonds
18
-------
and Jenkins, 1954), a pump-mix mixer-settler (Wurm,
1968), and a rotating disc contactor (Misek and
Rozkos, 1966) have been described. The use of a
centrifugal extractor for phenol recovery has
also been recommended (Kaiser, 1955). In the
Phenex process (Lewis, 1968) a mixing valve followed
by an electrostatic coalescer is used. A packed
column of special design to overcome fouling problems
has been used in the recovery of acetic acid
from pulping waste water (Weaver and Biggs,
1961). Mixer-settlers are the suggested type of
extractor for oily water treatment (Strausser
and Kurland, 1970).
Objectives of This Research.
The goals of this project were (1) to
generate sufficient data so that a realistic eco-
nomic evaluation can be made for the use of solvent
extraction on selected actual waste waters, (2) to
determine in what cases the use of a volatile sol-
vent like isobutane can lead to better economics,
(3) to identify types of extractors that are likely
to be most useful, and (4) to try to generalize the
procedures for choosing a "good" solvent extraction
process. We expect this study to help identify cases
where solvent extraction should be a promising
alternative to the present treatment method. The
approach has been a combination of conceptual
process designs, experimental extractions of actual
and synthetic waste waters, and measurement and
correlation of equilibrium distribution data.
Since the present state-of-the-art of solvent
extraction almost always requires pilot-scale
19
-------
study when developing a new process (Treybal ,
1966), a small pilot plant was built and used to
study many different organic pollutants and several
different solvents. Realizing that the use of
actual waste water may uncover problems in process
development which were not apparent when treating
synthetic mixtures of solutes in water, we
contacted several companies and asked for informa-
tion on their particular waste water problems.
Generally these contacts were productive and in
some cases resulted in our being provided with
large samples of actual waste water which could
be treated in the pilot plant. The results of
these and other experiments are described in
succeeding sections.
20
-------
SECTION IV
GENERAL PROCESS CONSIDERATIONS
The net cost of waste treatment is the primary
basis for choosing between feasible alternative
processes. If any recovery process is to be
preferred over biological oxidation for the elimi-
nation of a biodegradable pollutant, then the
value and quantity of the recovered pollutant
must be substantial. Because of the large volume
of water to be treated, a relatively low concentra-
tion of only 1% pollutant in water can lead to
a significant value. This can be simply illustrated
for the case of phenol when it is assumed that
95% of the feed phenol is recovered and sold at
one-half the current market price of 9-1/2 cents/lb.
The value of the recovered pollutant is $3.75 per
thousand gallons of water, quite a substantial value,
The above estimate is conservative as compared
to the general case because many pollutants are
more valuable than phenol. Sale at one-half market
price, which was assumed since the pollutant will
likely be impure, is also conservative for cases
where the recovered pollutant is simply added back
to the main process. For example, this direct
recycle would be feasible if the recovered pollutant
is a product of the main process or is a solvent
in the main process. From these approximate
21
-------
considerations we would expect that a general
purpose recovery process which could be operated
for several dollars per thousand gallons of water
treated would be a useful alternative to biological
oxidation. Steam stripping and solvent extraction
are two such processes.
Considerations Pertaining to Steam Stripping
Steam stripping is a general purpose waste
treatment process because at low concentrations
most dissolved organic chemicals are more volatile
than water. However, to avoid a vacuum column,
the entire waste water stream must be heated to
the atmospheric boiling point. This requires
a substantial amount of energy (e.g., as steam),
although some of this heat can be recovered in a
feed-bottoms heat exchanger. If the waste water
stream is corrosive, increasing the temperature
as required in steam stripping will aggravate the
problem.
Generally with all but the lowest molecular
weight organic compounds, an azeotrope will occur
at some concentration in the water-pollutant
binary system. Although the process is referred
to as steam stripping, reflux will be required to
produce a recovered organic stream with a concentra-
tion approaching the azeotropic composition. The
process is actually a distillation, and the azeo-
trope limits the purity of the concentrated organic
product. With the higher molecular weight compounds,
the critical solution temperature will be above
the binary atmospheric azeotropic temperature,
22
-------
and a heterogeneous azeotrope will be formed.
This generally makes the separation easier because,
when the distillate condenses, two phases are
formed. The organic phase can be taken as the
recovered pollutant product, and the water phase
can be either mixed with the feed or used as reflux
The above description refers to dissolved
organic pollutants in water. When the organic
pollutant is present as a liquid dispersion, then
its removal by steam stripping may be more dif-
ficult. Unless the organic phase separates from
the water in the distillate, the greatest purity
of recovered pollutant is determined by the rela-
tive vapor pressures of water and pollutant. When
a vapor mixture of a highly insoluble organic
and water is condensed, the resulting liquid may
be an emulsion of very fine organic droplets
suspended in the water. When this is the case,
the waste water problem is not yet solved.
A process which recovers dissolved n-butyl
acetate from a large flow of water is shown in
Figure 2. This binary system forms a hetero-
geneous azeotrope, and advantage is taken of this
fact in the process arrangement. One source of
such a waste water is a process which uses n-butyl
acetate as solvent for the extraction of phenol.
Another source of this type of waste stream but
at a much smaller flow rate is an n-butyl acetate
production plant which is based on the following
reaction:
n-C4HgOH + CH3COOH
23
-------
The waste stream contains water present in the
reactant acetic acid as well as that produced in
the reaction (Faith, et al., 1965).
Typical Cost of Steam Stripping
A process design and cost estimate have been
completed for the process shown in Figure 2. The
vapor-liquid equilibrium data of Weller, et al. (1963)
were used to estimate the relative volatilities
in the stripping column. The capital and operating
costs were determined as described in Appendix A.
The following assumptions were made to define
the process:
1. Feed water contains 0.6% n-butyl acetate.
2. Product water contains 50 ppm n-butyl
acetate.
3. Feed water flow rate is 100 GPM.
4. Feed water temperature is 80°F.
5. The temperature at the bottom of the
column is 230°F.
6. The amount of stripping steam injected
directly is that needed to saturate the
subcooled feed plus 1.4 times the
additional amount needed at minimum
flow. The flow of vapor decreases
substantially at the feed stage.
7. The distillate is condensed and cooled
to 180°F before phase separation occurs.
8. Overall stripping stage efficiency
is 75%.
The results of the material balance and the opera-
ting conditions are given in Table 1; in Table 2
24
-------
ro
en
Condenser
H
Steam
Feed Water
1
_« Recovered
Butyl Acetate
Decanter
Stripping
Column
D
Feed-Bottoms
Heat
Exchangers
B
Product
Water
Figure 2. Steam Stripping Process
-------
Table 1. Material Balance for n-Butyl Acetate Recovery by Steam Stripping
Stream (Figure 2) A BC D E FG H
Temperature (°F) 80 88 180 230 143 218 180 230
Condition Liq. Liq. Liq. Liq. Liq. Vap. Liq. Vap.
n-Butyl Acetate 4.96 4.96 5.04 0.04 0.04 5.00 4.92 0.00
(Ib/min)
Water (Ib/min) 826.6 826.6 832.4 874.8 874.8 6.1 0.3 48.5
-------
Table 2. Cost Estimates for n-Butyl Acetate Recovery
by Steam Stripping
Capital Costs:
Total Plant Investment = $41,700
Equipment Item Percent of Plant Investment
Stripping Column 23
Condenser 4
Decanter 4
Feed-Bottoms Exchangers 69
Operating Costs;
Treatment Cost = $0.61/1000 gal.
Cost Item $/year
Stripping Steam 18,600
Capital Equipment 10,800
27
-------
the estimated capital and operating costs
are presented. As described in detail in Appendix
A, the "Total Plant Investment" includes the
cost of all major equipment (stripping column,
feed-bottoms exchangers, condenser, and decanter),
all auxiliary equipment (piping, concrete,
instruments, etc.), labor for material erection
and equipment installation, indirect items (freight,
construction overhead, engineering, etc.), and an
18% factor for contingency plus contractor's
fee. Costs for site development, off-site invest-
ment, and working capital are not included. The
"Treatment Cost" includes utilities, maintenance,
supplies, depreciation, insurance, taxes, and
an 8% return on investment. As discussed in
Appendix A, the cost of operating labor and
laboratory charges have not been included.
For this steam stripping process, the principal
cost is associated with increasing the temperature
of the entire stream to 230°F. About 69% of
the investment is for the feed-bottoms exchangers,
and 87% of the stripping steam is that required
to increase the temperature of the feed water
stream from the feed-bottom exchanger outlet
temperature up to its bubble point. The overall
operating cost, $0. 61/thousand gallons, compares
favorably with the value of the recovered n-butyl
acetate, $7.40/thousand gallons.
It is well known that low stage efficiencies
can occur for systems which have a large relative
volatility (Fair, 1973), and in this case the relative
volatility ranges from 170 to 200. If the overall
stripping stage efficiency were 20% instead of
28
-------
75% as assumed above, the overall operating cost
would increase to $0.69/thousand gallons, and
the total plant investment would increase to
$55,200 with the stripping column contributing
42% of this total .
Considerations Pertaining to Solvent Extraction.
In contrast to steam stripping, solvent
extraction is not considered a general purpose
waste treatment process. The usual approach is
to choose a specific solvent for use in treating
each individual waste water. For example, Kiezyk
and MacKay (1973) screened 26 solvents for the single
application of treating phenol-containing
waste waters. What is needed is a more universal
solvent. If this solvent is selected to eliminate
the need for recovery of the solvent which is
dissolved in the purified water and to make the
method of solvent regeneration a clear choice
(e.g., distillation), then the difficulty in
synthesizing a new process is greatly reduced.
This approach was taken in this research to try
to develop solvent extraction into a general
purpose waste water treatment process.
The bases for solvent selection in the more
common solvent extraction separations (Treybal ,
1963) do not always apply in waste water extrac-
tion. The distribution coefficient, K^ (= ppm
pollutant in solvent/ppm pollutant in water), is
the most important consideration in choosing a
solvent. Since solute concentrations in the
waste water are usually in the range from 2%
in the feed water down to 10 ppm in the purified
29
-------
waste water, using the value of K, at infinite
dilution, K°?, for each pollutant independently
is usually satisfactory. Even in cases where K.
is significantly different from its value at
infinite dilution, it may still be valid to analyze
a process using only K^. «d directly affects the
required ratio of solvent mass flow rate to water
mass flow rate, F/F. through the extraction factor,
s w
E (= KjF /F ). Sl'nce E should equal 2 or larger
for a practical process giving a high extraction
efficiency, a large value of K. allows efficient
pollutant removal at a low value of FS/FW- A large
value of K. also allows efficient extraction with a
less thoroughly regenerated solvent. The driving
force for mass transfer of pollutant from the
water into the solvent (which would be used with
an overall water-phase mass transfer coefficient)
is given by the following relationship between
K. and the bulk concentrations of pollutant:
driving force = ppm in water - PP" In solvent
Kd
At the solvent inlet end of a counter-current
extractor, the concentration of pollutant in the
purified water will likely be 10 - 100 ppm. If
Kd = 1000, then the regenerated solvent could
contain up to 1% pollutant. However, if K. = 1,
then the regenerated solvent must contain less
than 10 ppm pollutant. If the solvent is more
volatile than the pollutant and if solvent
regeneration is carried out by distillation, then
rectification will be required to produce this
pure a solvent.
30
-------
Although K. is very important in solvent
selection, it is also a function of the particular
pollutant under consideration. Other factors
that are not so dependent on the pollutant will
have more influence on the choice of a universal
solvent. The solvent should be highly insoluble
in water to eliminate the need for recovery of
the solvent dissolved in the purified water. The
information in Table 3 illustrates the importance
of solvent solubility for a number of potential
solvents. These data indicate that the paraffinic,
olefinic, and higher aromatic hydrocarbons may
be suitable for a general purpose process.
The solvent should be chosen to have a
density much different from that of water so that
countercurrent flow or settling in a mixer-settler
will proceed easily. All the solvents in Table
3 have a density less than 0.9 gm/cc , except
ethylene dichloride which has a density of 1.26
gm/cc. For some pollutants the density difference
between solvent and water may decrease signifi-
cantly as the concentration of pollutant in the
solvent increases.
Distillation is the most common method for
separating the pollutant from the loaded solvent.
If we limit consideration to distillation as a
method for solvent regeneration, then the
normal boiling point of the solvent is another
important property. Vie could consider a solvent,
such as a higher aromatic, with a boiling point
above that of the pollutant so that the pollutant
would be recovered as a distillate. The major
advantage of this mode of operation is the
31
-------
Table 3. Solvent Solubility Losses
Solvent Water Solubility Market Price Loss
(% at 20°C) (cents/lb) ($/1000 gal)
Methyl Isobutyl 1.9
Ketone
Isopropyl Ether 1.2°
n-Butyl Acetate 0.68C
Q
n-Hexanol 0.58
Ethylene 0.81C
Dichloride
Benzene 0.17
Toluene 0.06
Isobutylene 0.053*
Xylenes 0.02d
Isobutane 0.015C
n-Hexane 0.001(
15*5
24.50
m
15
12%
9
9
8
6
7
4
8
11.50
8.50
6.00
6.00
1.25
0.40
0.26
0.12
0.05
0.0?
a Data from Chemical Marketing Reporter
(March 25, 1974).
b Assuming no solvent entrainment.
c Data from Union Carbide (1973).
d Data from API (1963).
e Estimated from Matsuura and Sourirajan (19'
32
-------
elimination of the need to boil the entire solvent
recycle stream. However, a major disadvantage
is the accumulation of high boiling impurities
from the waste stream. These impurities would be
expected to change Kd and the physical properties
of the solvent. The other alternative is to choose
a solvent with a much lower boiling point than
the pollutant. We will refer to this option as
volatile solvent extraction.
Choice of Volatile Solvent.
Consider the C~ to Cg paraffinic and olefinic
hydrocarbons as possible universal solvents. Such
a volatile solvent process would probably be
arranged as in Figure 3 (shown for isobutylene
as the volatile solvent). The pressure in the
extractor will be slightly above the vapor pressure
of pure solvent (to avoid vaporization due to
temperature fluctuation) at the extractor tempera-
ture. The pressure in the distillation column
will be determined by the temperature in the con-
denser. To avoid refrigeration and to use cooling
water, the condensation temperature should be
about 110°F or higher. These factors make the
Cg hydrocarbons the favored choice. However,
the volatility of the solvent relative to the pollu-
tant should be large to make the distillation easy.
This factor makes the C3 hydrocarbons appear
better. A compromise is necessary in the choice
of best solvent, as is illustrated by the following
approximate calculations.
33
-------
Waste
Water
Loaded
Isobutylene
Distillation
Column
Extractor
T
Pollutants
Recycle Isobutylene
Isobutylene
Vapor
Holding
Tank
Purified
Water
Figure 3. Typical Volatile Solvent Process
34
-------
For the case where a hydrocarbon solvent
extraction is used to recover dissolved ethylene
dichloride from a waste water, we will assume
that binary liquid mixtures of ethylene dichloride
and each hydrocarbon behave as regular solutions
(Hildebrand, et al . , 1970). Won (1973) has
determined the distribution coefficient for ethylene
dichloride between water and isobutylene to be
about 70. In dilute solutions the distribution
coefficient, K,, is related to the activity
coefficients of ethylene dichloride in the water,
(H20) (HC)
Yrnr , and in the hydrocarbon, Ynr > anc' to
molecular weights of water and hydrocarbon, MWH
and MM,,,,, as follows:
MWH20
MWRC
(H20)
YEDC
Y(HC)
YEDC
(HC)
For isobutylene, we calculate YEDC from the
regular solution theory by neglecting the effect
of water in the hydrocarbon phase. This allows
(H20)
Ycn;: to be calculated using the experimental
EUL (H?0)
value for Krf. We assume that YED5 does not change
from hydrocarbon solvent to hydrocarbon solvent.
Therefore, K. for each solvent may be determined
by again applying regular solution theory. The
results of this estimation are shown in the first
column of Table 4 for a series of hydrocarbons
from C3 to Cg. Solubility parameters and molar
volumes were taken from Chao and Seader (1961).
Also for the purpose of illustrating trends,
the Chao-Seader (1961) estimation method may
35
-------
Table 4. Factors Affecting Solvent Choice
uo
CT)
Hydrocarbon
Solvent
Propane
Propylene
n-Butane
i-Butane
1-Butene
i-Butylene
n-Pentane
1-Pentene
n-Hexane
1-Hexene
K
—
65
70
66
66
70
70
67
70
67
68
Column
Pressure
(psia)
212
253
60
83
72
74
19
23
17
17
Distillate
Temperature
<'F>
110
110
110
110
110
110
110
110
164
155
Relative
Volatility
(when XEDC=0)
8.3
9.2
4.4
5.8
5.1
5.4
2.0
2.9
0.78
0.99
Bottoms
Tempera1
<°F>
347
367
247
269
262
263
183
193
188
187
Relative
(when XE[)C=0.97)
16.8
14.8
18.8
19.6
17.5
17.6
11.9
13.7
4.5
5.3
-------
be used to calculate the relative volatilities
of mixtures of hydrocarbons and other non-polar
compounds (assumed to include ethylene dichloride).
It is assumed that the solvent regeneration column
will operate at slightly above atmospheric
pressure or at a higher pressure if that is
necessary to condense the distillate at 110°F
using cooling water. Thus fixing the column
pressure, the volatility of solvent relative to
ethylene dichloride is calculated at a point
approaching pure solvent (i.e., like the distillate
composition) and at a point where the mole fraction
of ethylene dichloride is 0.97 (i.e., like the
recovered pollutant product composition). These
values are also included in Table 4 along with
the distillate and bottom product bubble point
temperatures.
If we compare the estimated distribution
coefficients for the C3 to Cg hydrocarbons, we
see that there are no significant differences
except that the olefins are slightly better
solvents than the paraffins. Experimental
differences between olefins and paraffins are
somewhat larger (Won and Prausnitz, 1974), but the
olefins, being more reactive, can not always be
considered (e.g., for treating acidic waste waters).
Comparing the C3 and C. hydrocarbons shows that the
relative volatility favors the C-,'s and the operating
pressure favors the C.'s. However, the relative vola-
tility of 4 to 6 for the C.'s should be large enough to
make the separation easily, so the C.'s are the
better choice. The Cg hydrocarbons are good in
37
-------
terms of low pressure and high distillate temperature
(requiring a smaller condenser), but the relative
volatility is too small, even forming azeotropes.
The choice between the C. and C,- hydrocarbons
is not as clearly indicated, but the slight increase
in cost to operate the column at 60 to 83 psia
should be smaller than the cost associated with
the more difficult separation in the upper portion
of the column in the case of the C5's. Also,
the C. hydrocarbons are generally more readily
available. On the basis of this simple comparison,
we chose to study the C. hydrocarbons, isobutylene
and isobutane. The iso- rather than the normal-
isomers were chosen because they are less soluble
in water. An even more readily available solvent
would be a mixture of these C^ paraffins and
olefins.
Considerations in Volatile Solvent Extraction.
Many of the factors discussed previously
for a general solvent extraction process also
apply to volatile solvent extraction. These
factors can be illustrated by estimating the
costs of treating a typical waste water. For
this purpose, we will consider the recovery of
ethylene dichloride from a water solution using
isobutylene as the volatile solvent. This pol-
lutant occurs in the waste waters from the produc-
tion of vinyl chloride and is non-biodegradable.
The loaded solvent from the extractor is a
dilute, subcooled liquid consisting of a wide-
boiling mixture. During the distillation, all
38
-------
the volatile solvent must be boiled and then
condensed making heat economy very important.
If the loaded solvent is added directly to a simple
distillation column having a single reboiler,
then all the heat required to vaporize the solvent
must be supplied at the temperature of the reboiler
From Table 4, if the distillation column were
operated at 74 psia using cooling water for
condensing, then the reboiler will operate at
263°F. As is discussed in detail in Appendix D,
other alternatives result in a much lower cost
for solvent regeneration.
One alternative which substantially reduces
the regeneration cost utilizes a total feed vapor-
ization. In this alternative most of the heat
to boil the solvent can be supplied at a lower
temperature using exhaust steam. However, more
reflux is required than for a subcooled liquid
feed resulting in larger amounts of heat supplied
and removed.
A second alternative, which is better than
either of the above two, utilizes a subcooled
liquid feed; however, several stages below the
feed point the downflowing liquid is drawn off
and partially vaporized. The exact location of
this side stream boiler is subject to optimization,
but it is chosen so that the temperature of the
boiler is low enough to be driven with exhaust
steam. With ethylene dichloride the optimum
location results in a side stream boiler tempera-
ture of about 128°F. This last alternative allows
supply of most of the heat using exhaust steam
39
-------
but does not increase the amount of reflux over
that required for a subcooled liquid feed. Other
savings such as a lower cost for the distillation
column are described in Appendix D.
A practical mode of operation of the distil-
lation column condenses the distillate at 1 2 0 ° F ,
uses a vapor flow in the portion of the column below
the side boiler 20% above minimum vapor flow,
and locates the side stream boiler several stages
below the feed point. This alternative is partic-
ularly attractive when the relative volatility
is low at a high concentration of pollutant, as
is the case with phenol in n-butyl acetate.
The distribution coefficient for the extraction
of ethylene dichloride with isobutylene is about
70. Therefore, the volatile solvent extraction
process can be operated with a solvent-to-water flow
ratio'of 1:30 and still have the extraction factor
large enough for efficient pollutant removal.
At these low solvent flow rates, the heat released
in the condenser is small enough that the purified
waste water can be used as coolant rather than
using cooling water. This reduces the operating
cost for cooling water, but in all the alternatives
the condenser represents a substantial capital
cost. This cost can also be reduced through the
use of a direct contact condenser.
Direct contact condensation has seen infre-
quent use in industry. It was studied by Harriott
and Wiegandt (1964) in conjunction with a desal-
ination process. These authors measured volumetric
heat transfer coefficients for the condensation
of methylene chloride with water in a packed
40
-------
bed and on a single sieve plate. Their data can
readily be used to estimate the design of a direct
contact condenser using the cool purified waste
water to condense the overhead vapors (isobutylene)
from the distillation column.
For cost estimation purposes, we consider
a short, packed bed with the purified waste water
being distributed on top of the packing and the
vapors entering from below. Both the condensed
organic and the water flow out the bottom of this
device as a two-phase mixture and are settled
in a separate tank. Other types of contacting
devices could also be used and are expected to
result in comparable costs; they may be preferred
if suspended solids are present. Devices which
allow vapors to condense on a film of water should
eliminate a potential problem in which the small
condensed organic droplets could form a stable
emulsion in the purified water.
Since the purified waste water leaving the
extractor is already saturated with volatile
solvent, its direct contact with the distillate
will not result in additional pollutants being
added to the purified water. In fact, the direct
contact condenser will act as an additional
extraction stage (at a higher temperature where
K. is usually lower). This arrangement of the
volatile solvent process is shown in Figure 4.
The quantitative effects of replacing the normal
condenser by a direct contact condenser on the
operation and cost of the process are discussed
below.
-------
X)
o
en
-------
The substantial cost of regenerating the
solvent and the opportunity to use a direct
contact condenser clearly favor operation with
as low a solvent-to-water flow ratio, F /F , as
possible. The lower limit is determined at the
point where the extraction factor, E (= K.F /F ),
Q S W
becomes so low that the cost of the extractor off-
sets the savings in volatile solvent regeneration.
For pollutants having a large K, (i.e., greater
than 30), the optimum FS/FW will likely be 0.1
or less. However, not all types of extraction
equipment will operate efficiently at such low
values of F /F . Mixer-settlers can operate well
O f>
since settled solvent can be recycled to the previous
mixer without destroying the counter-current
operation, but these devices are very expensive
if many extraction stages are required. The ro-
tating disc contactor (RDC) has been used success-
fully with a low flow ratio (Misek and Rozkos, 1966).
The design of an RDC without detailed mass transfer
data is only approximate, and therefore for pre-
liminary estimates of cost a mixer-settler will be
used.
Typical Cost of Volatile Solvent Extraction.
A preliminary cost evaluation has been
completed for the recovery of ethylene dichloride
(EDC) from a waste water by extraction with
isobutylene. The following assumptions were made:
1. Feed water contains 0.8% EDC.
2. Extraction removal efficiency is 95%.
3. Feed water flow rate is 100 GPM.
43
-------
4. Feed water temperature is 80° F.
5. Recovered EDC residue contains 1.5 wt.
% i sobu ty1ene.
6. Regenerated solvent contains 0.56 wt. %
EDC. This will reduce the driving force
at the water outlet of the extractor
by 20% from that if a pure solvent had
been used .
7. Distillate is condensed at 120°F
using cooling water which is heated from
80°F to 100°F, or using direct contact
condenser when possible to do so without
heating the purified water above 100°F.
8. Distillation column operates as described
in Appendix D with a 75^ Murphree vapor
stage efficiency.
9. in-minute liquid hold-up is used in
both water and reflux tanks.
10. 3-ninute liquid hold-up is used in the
bottom of the distillation column.
11. A multiple-stage mixer-settler is used for
extraction with a solvent recycle, if nec-
3
essary to give a volumetric ratio of 0.6 ft
solvent per ft' water, and a stage
efficiency of 85%. Mixer-settler costs
are estimated from the data of Treybal
(1963) .
Based on these assumptions using K^ = 70.0
and estimating vapor liquid equilibria from the
Chao-Seader correlation, costs were estimated for
F /F from 0.08 to 1.5 using the methods of
Appendix A. The distribution of capital costs
is shown in Table 5 for the two extremes of F^/Fw-
44
-------
Table 5. Capital Costs for EDC Recovery
Capital Cost
Total Plant Investment
Equipment Item
Distillation Column
Condenser
Side Stream Boiler
Reboiler
Tanks
Pumps
Extractor
W0-08
$108,700
Fs/Fw=1.50
$156,100
Percent of Total
12
3*
2
1
5
9
68
12
29
20
1
8
7
23
Direct contact condenser + settling tank
45
-------
For the low flow ratio, the major capital cost
item is the extractor. For the high flow ratio,
the major capital cost is associated with heat
exchangers to vaporize and condense the volatile
sol vent.
The operating costs for this volatile solvent
extraction process are divided into an amortized
capital cost for the extractor (including a
prorated portion of maintenance, repairs, supplies,
depreciation, insurance, taxes, and return on
investment) and a cost associated with all other
items (including the remaining portion of the
above capital related costs plus chemicals and
utilities, but not including operating labor
and laboratory costs). Figure 5 shows such a
division of costs for ethylene dichloride
extraction and illustrates the advantages of
operating at a low flow ratio. The discontinuities
in the curve for the mixer-settler result from
a requirement for abrupt increases in the number
of stages to achieve at least 95% recovery as
the flow ratio decreases. The discontinuity
in the curve for "other" costs results from the
use of purified waste water for cooling instead
of cooling water and the use of a direct contact
condenser. Figure 6 shows a breakdown of the
regeneration costs. For comparison, the recovered
EDC is estimated to be worth $5.70/thousand gallons
at its full market value.
At values of F /F = 0.135 and less, the
purified waste water can be used to condense all
the distillate vapors, making the direct contact
condenser practical. The reduction in cost when
46
-------
Operating
Cost
(TM gai.)
I I I I
I I
Amortized Cost of
Mixer-Settler
™««i««^"^™^™^"^^^^^"^^™^^^
I i I i r t t
0.5 c / 1.0
VFW
Figure 5. Total Cost of EDC Recovery
1.5
-------
00
Operating
Cost
( /M gal.)
I I
j—i
Isobutylene Losses
± « » i—_i—i—i—L
0.5
1.0
i—L
1.5
Figure 6. Breakdown of Costs for EDC Recovery
(Excluding Extractor)
-------
using the direct contact condenser is greater
than just the reduction in cooling water and con-
denser cost. The volumetric heat transfer
coefficient is so large (i.e., about 50,000 BTU/hr
3
ft °F) that the optimum distillate temperature
is reduced to about 1° above the final temperature
of the purified waste water. This results in
a reduction in pressure within the distillation
column and a reduction in cost for the solvent
recycle pump, for the distillation column (through
both a higher relative volatility and a smaller
shell thickness), and for both boilers (through
an increase in temperature driving force).
There is a slight increase in cost due to a larger
feed water pump and operation of the extractor
at higher pressure.
Figure 5 can be used to determine a first-
order approximation to the costs of extracting
other pollutants. The cost of solvent regeneration
depends primarily on FS/FW for any pollutant with
a volatility similar to that of ethylene dichloride.
The cost of extraction depends primarily on K.
and on the required removal efficiency. Therefore,
for 95% removal efficiency, the curve for extractor
costs can be moved horizontally to a point giving
the same cost vs. E as in Figure 5 (where K. = 70.0)
If K, is less than 70.0, the curve would be moved
to the right to provide an estimate for extractor
costs. This argument leads to the conclusion that
the optimum F_/Flf will increase as K. decreases.
S W Q
49
-------
Comparison of Volatile Solvent Extraction with
Steam Stripping.
For very volatile pollutants having a low
Kd with the volatile solvent, steam stripping
is expected to be less expensive than volatile
solvent extraction. As higher molecular weight
pollutants are considered, extraction is more
competitive. This is especially true if the
pollutant is present as an emulsion, if the pol-
lutant forms an azeotrope with a high fraction of
water, or if the waste water would become highly
corrosive at increased temperatures. However,
the cost of volatile solvent extraction can be
comparable to or less than steam stripping even
when the latter is known to be a practical
process.
A process design and cost estimate have been
completed for a process in which isobutylene is
used to extract n-butyl acetate (Figure 7). The
distribution coefficient for n-butyl acetate into
isobutylene is 168 (Appendix E). The assumptions
about water flow rate, feed and product concentra-
tions, and feed water temperature are the same
as those for the previously described cost esti-
mate for the recovery of n-butyl acetate by steam strip-
ping. The recovered n-butyl acetate was assumed
to contain 2.0 wt. % isobutylene, and the regen-
erated solvent was assumed to contain 0.168 wt. %
n-butyl acetate (based on a driving force at the
water outlet of the extractor 20% less than the
driving force if pure isobutylene had been used).
50
-------
A direct contact condenser is used since the optimum
F /F is 0.11 (requiring 2 mixer-settler stages).
Table 6 gives the material balance and operating
conditions for the conceptual process, and Table
7 presents the cost estimates. The cost for
volatile solvent extraction is $0.71 per thousand
gallons as compared to $0.61 per thousand gallons
for steam stripping.
Realizing the approximate nature of the
estimates, these costs are essentially equal.
For the extraction process a majority of the cost
is associated with the extractor, the cost of
which might be reduced by using a different type
of contactor (e.g., an RDC). Also, the assumption
of a 75% stage efficiency for both steam stripping
(where the relative volatility ranges from 170 to
200) and for n-butyl acetate-isobutylene distillation
(where the relative volatility ranges from 15 to
54) is probably optimistic in both cases, but
certainly is more likely to be correct when the
relative volatility is lower. Considering these
factors, volatile solvent extraction does appear
to be an economically attractive alternative to
steam stripping.
To estimate the cost of extracting ethylene
dichloride, n-butyl acetate, or other potential
pollutants, it is necessary to have available
estimates for K. and for the binary vapor-liquid
equilibria between pollutant and volatile
solvent. Appendix E reports values for K^ at high
dilution between water and C^ hydrocarbons for a
variety of esters, aldehydes, ketones, and phenols.
For relatively non-polar polutants (e.g., ethylene
51
-------
Feed ^_
Water /\
Direct
Contact
Condenser
and Settler
t H
I"
^XC
^%^
]
Two
Loaded
[sobutylene
Stage
Extractor
F
•«— '
Isobut.
Make-up
J
Reflux
«.
c
^
J
~*" Distillation
4—— ^ f* A 1 1 1 m rt
' ^ uoiumn
,^ ^ Side
^ — Rnilpr "*" Exhaust Steam
T
' ^ Reboiler ^ 100 psig Steam
l_
' fc< Recovered Butyl Acetate
Holding
•" ** - ianK
Purified Water
Figure 7. Details of Volatile Solvent Extraction Process
-------
Ui
CO
Table 6. Material Balance for n-Butyl Acetate Recovery by Solvent Extraction
Stream A BCDEFGHJKL
Temperature (°F) 80 80 96 96 96 97 80 80 97 97 330
Pressure (psia) 58 57 57 15 15 57 57 58 58 58 58
Condition liq. liq. liq. liq. vap. liq. liq. liq. liq. vap. liq.
Water (Ib/min) 826.6 826.6 826.6 826.6 =0 =0 =0 =0 -0 =0 =0
Isobutylene 0 0.5 0.5 0.2 0.3 91.3 0.6 90.8 4.5 95.2 0.1
(Ib/min)
Butyl Acetate 4:96 0.04 0.04 0.04 0 0.15 0 5.07 0.01 0.16 4.92
(Ib/min)
-------
Table 7. Cost Estimates for n-Butyl Acetate Recovery
by Solvent Extraction
Capital Costs;
Total Plant Investment = $78,600
Cost Item Percent of Total
Extractor 63
Distillation Column 11
Direct Contact Condenser 5
and Settler
Side Stream Boiler 1
Reboiler 1
Tanks 8
Pumps 11
Operating Costs;
Total Treatment Cost = $0.71/1000 gal.
Cost Item Annual Cost
Chemicals $10/100/yr
Utilities 3,500/yr
Capital Equipment 20,300/yr
54
-------
dichloride), vapor-liquid equilibria can be estimated
using the Chao-Seader correlation (1961). This
approach is based on regular solution theory for
prediction of liquid activity coefficients and is
not reliable for polar components. However, an
alternative approach based on the Wilson equation
(1964) for liquid activity coefficients should give
reliable estimates.
For a binary solution, the Wilson equation
has two adjustable parameters and is of a form
which allows the effect of temperature on the
activity coefficients to be estimated. This is
an important feature since at constant pressure
the temperature varies over a wide range with
composition for a mixture of a volatile solvent
and a much-less-volatile pollutant. The two Wilson
parameters may be evaluated by a simple iterative
calculation using the two infinite dilution activ-
ity coefficients. The infinite dilution activity
coefficient for the pollutant in the volatile
solvent may be estimated for many pollutants from
Kd at 25°C and the activity coefficient at infinite
dilution for the pollutant in water (Pierotti, et al.,
1959). The infinite dilution activity coefficient
for the volatile solvent in the pollutant at 25°C
may be estimated by the method of Helpinstill and
Van Winkle (1968) or the method of Weimer and Prausnitz
(1965). Vapor phase nom'deal i ties can adequately be
approximated using the Redlich-Kwong equation of state,
as is done in the Chao-Seader method. This procedure
was used in the estimation of vapor-liquid equilib-
ria for the separation of n-butyl acetate from
isobutylene. The procedure may also be extended
55
-------
to multicomponent mixtures when more than one
pollutant is extracted, provided some estimate
can be made for the pollutant-pollutant binary
mi xtures.
Dual Solvent Extraction Processes.
The attractiveness of volatile solvent extrac-
tion for the recovery of n-butyl acetate, a known
good solvent for phenol extraction (Kiezyk and
MacKay, 1973) leads to the consideration of a
dual solvent extraction process for the recovery
of phenol. Such a process is illustrated in Figure
8. The phenol is removed from the waste water by
extraction with n-butyl acetate (Kd = 57; Appendix F)
at a low flow ratio. The phenol is separated from
n-butyl acetate by distillation as was done in the
original Phenolsolvan process (Wurm, 1968). The
dephenolized water contains about 6800 ppm dissolved
n-butyl acetate, which is recovered by
extraction with isobutylene. The recovered n-butyl
acetate is separated from the isobutylene by
distillation and is recycled to the n-butyl acetate
solvent cycle. Other polar solvents could also
be considered for such a dual solvent process,
with the best choice being a solvent which is easily
recovered by volatile solvent extraction as well
as being a good solvent for phenol.
An alternative dual solvent process is shown
in Figure 9, where the phenolic waste water is
first extracted with a mixture of volatile solvent
and n-butyl acetate, and then the volatile solvent
is used to recover the dissolved n-butyl acetate.
56
-------
Waste
Water
Loaded P . S
Polar
Solvent
Extractor
T
P.S.- Pollutants
Splitter
Pollutants
Recycle P
I *• 1
_ — —j— — -—
Volatile
Solvent
Extractor
I
V.S. - P.S.
Splitter
Recycle _£. S_. j
I ^- V.S. Vapor
Holding
Tank
Purified
Water
Figure 8. Dual Solvent Process (Separate Solvent Cycles)
57
-------
Waste
Water
Loaded Mixed
Solvent
Mixed
Solvent
Extractor
Volatile
Solvent
Extractor
r
V.S. - P.S.
Splitter
Recycle P.SS,
P. S.- Pollutants
Splitter
T
Pollutants
t
i
• Recycle V.S. I
^ V.S. Vapor
Holding
Tank
Purified
Water
Figure 9. Dual Solvent Process (Linked Solvent Cycles)
58
-------
This process with linked solvent cycles has several
potential advantages over the dual solvent pro-
cess with separate solvent cycles while still
requiring only two distillation columns. For
extraction with pure n-butyl acetate, the density
of phenol-containing solvent becomes close to
that of water. This small density difference leads
to large extraction equipment because of slow
settling in mixer-settlers or low countercurrent
flows to avoid flooding in continuous extractors.
However, dilution of the n-butyl acetate with
isobutylene greatly increases the density difference
and allows operation in smaller, less expensive
extractors. The extraction with a solvent mixture
also results in a lower solubility of n-butyl
acetate in the dephenolized waste water, thus
requiring less thorough volatile solvent extrac-
tion in the second step. However, the dual solvent
process with separate solvent cycles, Figure 8,
has the potential advantage that the hot n-butyl
acetate vapors from the phenol-polar solvent splitter
may be more easily used to provide heat for the
volatile solvent-polar solvent splitter.
Although illustrated for the recovery of
phenol, the dual solvent processes may also be
attractive for other pollutants. The conditions
when a second polar solvent should be considered
are a low Kd (e.g., less than 5) for the pollutant
in the volatile solvent and a high K^ (e.g.,
greater than 20) for the pollutant in a polar
solvent. Almost all slightly soluble polar solvents
will show a high Kd in the volatile solvent because
the same factor leading to low water solubility
59
-------
(i.e., a high activity coefficient in water) also
leads to a high K ..
The examination of both single volatile
solvent extraction and alternative dual solvent
extraction processes constitutes the major oortion
of the experimental work to be described in this
report.
60
-------
SECTION V
BASES FOR EXPERIMENTS AND
INDUSTRIAL WASTEWATERS EMPLOYED
Experiments were carried out to determine equi-
librium distribution coefficients for various important
solutes between various solvents and water. Appendix
E reports methods and results for various families of
compounds using isobutylene, isobutane and other sub-
stances as solvents. Appendix F reports distribution
coefficients for phenolic solutes with various solvents.
The demonstration program using the miniplant
extraction facility consisted of two portions. The
first series of experiments was conducted in a
spray tower extractor for the purpose of studying
the overall process feasibility of volatile sol-
vent extraction. During these experiments volatile
solvents were used to treat a variety of actual
and synthetic waste waters. The objectives of
these initial experiments included (1) the quanti-
tative measurement of pollutant removal efficiencies
for both solvent-phase dispersed and aqueous-phase
dispersed, (2) the identification of possible
solvent degradation reactions when isobutylene
was used as solvent, (3) the determination of
whether emulsified liquids and particulate solids
in actual waste water could be removed through
61
-------
flotation by solvent droplets, and (4) the veri-
fication that the loaded solvent could be suffi-
ciently well separated from the pollutants in a
short, packed distillation column so that the
solvent could be recycled.
During the second portion of the experimental
program, a rotating disc contactor (RDC) was
used in place of the spray column since the results
of earlier experiments and conceptual process
calculations had shown that operation with a low
solvent-to-water flow ratio (Fe/F) would lead
5 W
to a more economical process. The system of using
a C, hydrocarbon as solvent is characterized by a
low solvent viscosity and a large difference in
density between solvent and water. These proper-
ties along with a lower limit on the practical
size of holes in the distributor plate of a spray
column lead to a low dispersed phase hold-up
at all values of Fe/Fl( when water is dispersed
5 w
and at low values of FS/FW when the solvent is
dispersed. A low hold-up and thus a low inter-
facial area leads to poor extraction efficiency.
In an RDC, the dispersed phase hold-up can be
increased by increasing the disc rotational speed
which results in smaller drops and improved
extraction efficiency. Although this is an overly
simplified comparison, the result is that the
RDC is expected to work satisfactorily even when
FC/F|; is low (e.g., F/F^ = 0,1) (Misek and
S W 5 W
Rozkos, 1966). Also the RDC is more characteristic
of modern, high-performance types of extractors,
so the collection of mass transfer data will be
more useful for scale-up and estimation of commercial
62
-------
feasibility. During the second series of experiments,
effort was mainly directed toward treating several
actual waste waters that we believed to be promising
candidates for economical treatment by solvent
extraction at low flow ratios.
An important aspect of the experimental
approach for actual waste waters involved the
identification of the individual pollutants
present. To understand, evaluate and improve the
solvent extraction process, it was desirable to
determine the removal efficiency for each of the
identified major pollutants. Seven separate types
of waste water samples were obtained from petroleum
and chemical companies and were treated in the mini-
plant.
In this Section the industrial waste waters
which we studied are first described with respect
to the identification of pollutants and to the
logical approach for treatment by solvent ex-
traction. The importance of axial dispersion
(backmixing) on the performance of solvent extraction
equipment is discussed, and a method of estimating
the individual-phase resistances to mass transfer
from miniplant data is developed.
Industrial Waste Water Samples
Lube oil refining waste water. One
waste water sample came from a lube oil refining
plant. This waste water is produced during solvent
refining and deasphalting (by the Duo-Sol process)
and dewaxing (by MEK-benzene extraction) of lubri-
cating oils in a typical petroleum refinery. The
63
-------
waste waters produced by these operations would
be expected to contain phenol and cresols, which
are used as a mixed solvent in the Duo-Sol process,
and methyl ethyl ketone (MEK) and benzene, which
are used as solvents for dewaxing. In addition
to the dissolved organic pollutants, the waste
water contained a small amount of floating oil
and fine, black, solid particles (possibly asphal-
tenes) which did not settle on standing and which
were not removed by filtering with qualitative
grade fi1ter paper.
The procedure used to identify the typical
pollutants will be described here as it was used
with slight modification for all the actual waste
waters. A GC analysis (Porapak Q at 200°C) showed
that the waste water contained five major
dissolved components. By comparing retention
times in the GC, four of the five pollutants were
tentatively identified as MEK, benzene, phenol,
and o-cresol. The fifth component had a retention
time similar to that of acetone (which could enter
the process as an impurity in MEK). A standard
solution was prepared which contained these five
components in about the same concentration as the
waste water; the pH of this synthetic sample was
the same as the actual waste. Equal volumes of
the actual and synthetic samples were contacted
in a separatory funnel with equal volumes of
highly purified n-butyl acetate, and the aqueous
raffinates from both extractions were analyzed
using the GC. Since the percentage extracted
for each of the five peaks was identical for
both samples, this was taken as confirmation of
64
-------
the pollutant identification. Had some uncer-
tainty existed at this point, a second comparative
extraction using a different solvent could have
been used. This simple procedure was useful
because the number of expected pollutants was
not too large and because their expected identity
was known. Had more major pollutants been present
or had no estimate been available of what pollutants
might be present, then a more sophisticated analysis
such as that of Herz (1972) or Sugar and Conway
(1968) could have been required.
In addition to the chromatographic analysis,
the chemical oxygen demand (COD) was measured
by a standard method (described in Section VI).
A theoretical COD was determined from the GC
analysis by assuming that all identified components
were completely oxidized to C02 and H20. The
ratio of theoretical to measured COD averaged
1.10 for the lube oil refining waste water. This
indicates that there may have been some loss of
volatile components or have been an incomplete
oxidation of organics in the measured COD and that
there were no components present in large concen-
tration other than those identified by the GC
analysis.
Samples of this lube oil refining waste
water were obtained on four occasions. The concen-
tration of components in three of these samples
fell into the following ranges:
Acetone: 25 to 40 ppm
MEK: 108 to 232 ppm
Benzene: 37 to 170 ppm
65
-------
Phenol: 17,000 to 23,200 ppm
o-Cresol: 1,200 to 2,660 ppm
This is probably the composition during normal
operation. The fourth sample was found to contain
a much different composition. Acetone and benzene
were absent, MEK was present at about 100 times
higher concentration than normal, and phenol and
o-cresol were present at about one-half their normal
concentrations. In addition, the appearance was
milky white rather than black as in the normal
samples. This variation in composition was
probably the result of an upset in the dewaxing
plant. It illustrates another problem not en-
countered in tests with synthetic waste water
samples, that of variation in feed water composi-
tion.
Since the major pollutants in this waste
water were polar organic compounds which are
difficult to extract with hydrocarbon solvents,
treatment had to use either volatile solvent
extraction at a high value of F /F or polar sol-
3 W
vent extraction as part of a dual solvent process.
Both approaches were studied using isobutylene
as volatile solvent and n-butyl acetate as polar
sol vent.
Cresylic acid recovery waste water. The
second waste water sample came from a cresylic
acid recovery plant which reprocesses the caustic
treating effluent produced during petroleum
refining. Since caustic treating is used to extract
phenolic compounds from petroleum, the waste
water was analyzed by GC (Porapak Q at 220°C) for
66
-------
phenolic compounds. It contained phenol, o-cresol ,
m, p-cresol (isomer separation was not possible
by GC), and at least two isomers of xylenol. It
also contained several low molecular weight compounds
at low concentrations. In addition, this waste
water contained a small amount of suspended solids
(about 50 ppm) and about 700 ppm of dissolved
inorganic substances (based on a commercial analysis
supplied by the company). The concentrations of
the phenolic compounds were as follows:
Phenol: 579 ppm
o-Cresol: 307 ppm
m, p-Cresol: 291 ppm
'. . i _ ._ _ ^ _ _ « « •»
Xylenols: 227 ppm
The ratio of theoretical to measured COD averaged
0.86 which indicates the presence of additional
oxidizable pollutants.
We were able to obtain this waste water in
limited quantities during the period when the
spray column was being studied. Since the cresols
and xylenols have distribution coefficients
between water and isobutylene in the range from 3 to
7, they are easier to remove than phenol, which
shows a distribution coefficient of 0.7 (Appendix E).
Volatile solvent extraction using isobutylene was
thus studied in the miniplant.
Ethylene quench waste water. The third
polluted water sample was ethylene quench water
produced in a naphtha-pyrolysis, olefin plant.
This cracking reaction uses a complex hydrocarbo
67
-------
mixture as feed and is expected to produce a variety
of organic side products, most of which are only
slightly soluble in water (e.g., aromatic hydro-
carbons, heavy tars, etc.) or are produced in small
quantities (e.g., phenolics). Analysis using the
GC (Porapak Q at 220°C) showed eleven major
components. Five components were identified as
benzene; toluene; m, p-xylene; o-xylene or ethyl-
benzene; and phenol. Three others correspond to
light oxygenated compounds having retention times
in the GC corresponding to methanol or formalde-
hyde, acetone, and n-propanol. Since removal of
these pollutants is not practical by volatile
solvent extraction in a single-solvent process,
and since they are present at less than 100 ppm,
a positive identification was not attempted.
The remaining three components appeared to be
high-molecular-weight phenolics and hydrocarbons.
When the samples were received, they contained a
substantial amount of floating and settled or-
ganic phases, as well as additional suspended
solids which resulted in a high turbidity. The
compositions of the aromatic pollutants and phenol
were as follows:
Benzene: 71 to 81 ppm
Toluene: 40 to 44 ppm
Xylenes: 34 to 40 ppm
Phenol: 67 to 68 ppm
.The ratio of theoretical to measured COD averaged
0.33 which indicates the importance of the unidenti-
fied pollutants.
68
-------
Since phenol and the light oxygenated compounds
could be recovered only by dual-solvent extraction,
and since these components can be readily removed at
these low concentrations by biological oxidation,
this waste water was treated only by single-solvent,
volatile solvent extraction using two separate
solvents in different runs, isobutane and isobutylene
This treatment would be expected to remove the
non-biodegradable, aromatic hydrocarbons even when
operating at low values of F /F.
s w
Oxychlorination waste water. The fourth
industrial waste water is from an oxychlorination
plant which produces ethylene dichloride by the
following gas-phase, chemical reaction:
CH2=CH2 + 2HC1 + 1/2 02—»-CH2Cl CHgCl + H20
The products of the reaction along with some un-
reacted HC1 are condensed, and the aqueous phase
is decanted and sent to waste. Therefore, the
waste water was expected to be acidic, to be
nearly saturated with ethylene dichloride, and
to contain any polar side products of the reaction
which would likely distribute significantly into
the aqueous phase. Because these side products
could include a very large number of possible
chlorinated and/or oxygenated hydrocarbons con-
taining 1 or 2 carbon atoms, the qualitative
identification of pollutants was complicated. The
company which supplied the waste water suggested
that ethylene dichloride, chlorform, and chloral
{CC13 CHO) could be present.
69
-------
Samples of oxychlon'nation waste water were
obtained on three occasions. They showed a
pronounced variation in the composition of minor
pollutants; this may be the result of sample
collection and storage procedures as much as
actual variations due to changing conditions in
the plant operation. The ranges of concentrations
of HC1 and organic pollutants were found to be
as follows:
HC1: 1.49 to 5.78 wt. %
(determined by titration)
chloral: 14,100 to 16,900 ppm
ethylene dichloride: 1500 to 3360 ppm
ethanol: 290 to 520 ppm
acetaldehyde: 0 to 100 ppm
monochloroacetaldehyde : 0 to 300 ppm
The COD was not measured due to the high chloride
concentration which interferes.
The identification of the organic compounds
was based on correspondence of GC residence times
(Porapak Q at 210°C) and equivalence of fraction
extracted when standard and actual solutions were
contacted in consecutive experiments with highly
purified 2-ethyl hexanol and 1 ,1 ,2-trichloroethane.
These solvents were utilized because of differences
in polarity and because they boil at much higher
temperatures than any of the organic pollutants
and thus do not interfere with the GC analysis.
Solvents used for this purpose had to be purified
by repeated water washing followed by distillation
so that impurities in the solvents would not
interfere with pollutant analysis.
70
-------
Chloral was clearly the most serious pol-
lutant in this waste water and also the component
most likely to provide a significant value from
its recovery. In aqueous solution, chloral is
present as chloral hydrate, CC13—CH(OH)2>
which could not be economically recovered by
volatile solvent extraction. In equilibrium experi-
emnts it was determined that n-octanol gave a distri-
bution coefficient for chloral in the range 15 to 50
depending on the concentration (K. increased as chloral
concentration decreased). Thus, a dual solvent process
using 2-ethyl hexanol (the least expensive form
of Cg alcohol) as polar solvent and isobutane
as volatile solvent was chosen for study. This
waste water is presently treated in commercial
operations by neutralization using any inexpensive
base (e.g., limestone), followed by steam stripping.
The neutralization may cause most of the chloral
to react with hydroxyl ion and to form chloro-
form and formate ion (Pfeil, et al., 1959),
in which case the chloroform but not the formate
would be recovered by steam stripping.
The presence of HC1 made this a very difficult
waste water to handle in the miniplant because
of severe corrosion. An initial attempt was made
to neutralize the HC1 by addition of solid NaOH
so that it could be treated, but the high pH
near the surface of solid NaOH led to decomposition
of the chloral even though the bulk of the solution
was still acidic. After trying several alternative
neutralization methods, we found that by holding
the waste water at 5°C while adding solid NaHCOg
until the pH increased to about 5.2 (the pH of
71
-------
a solution of 15,000 ppm chloral in pure water),
we got no appreciable chloral decomposition.
However, the COp which was evolved tended to sweep
away most of the volatile compounds (90% of the
acetaldehyde and 50% of the ethylene dichloride).
This neutralization using NaHCO~ was required only
to protect the miniplant, and in a commercial
plant the organics could be extracted from the
unneutralized waste water. However, since HC1
recovery is not economical in this commercial situa
tion, the HC1 would have to be neutralized after
chloral extraction.
72
-------
Phenol-Formaldehyde Resin Manufacture Wastewater
This wastewater was obtained form an inland plant
location, where the water effluent is currently in-
cinerated because of the absence of a plant biological
oxidation system. Analysis of the water using flame-
ionization chromatography with a Porapak-Q column
revealed three major peaks, corresponding to methanol ,
formaldehyde and phenol. These three components are
to be expected at high concentrations in the wastewater
on the basis of the known chemistry of this process.
Quantitative analysis of the wastewater by compari-
son with peaks from known, standard solutions gave
the following concentrations:
Component Concentration
Methanol 12,000 ppm
Formaldehyde 17,370 ppm
Phenol 48,270 ppm
The phenol concentration in this water (4.8 wt%) was
therefore the highest of any of the wastewaters
studied. Clearly, simple extraction with isobutylene
or isobutane would not be successful without a very
high solvent/water ratio. Consequently this water was
treated experimentally by the two dual-solvent processes.
n-Butyl acetate was used as solvent for a test of Process
II, and a mixture of 48.2% n-butyl acetate in isobutylene
was used as the solvent for a test of Process III. Since
numerous earlier experimental tests had shown that
n-butyl acetate is readily removed from the effluent
water by extraction with isobutylene or isobutane, the
second extraction step with the volatile hydrocarbon
solvent was not included in the testing program for
this water.
73
-------
Hydrofiner Condensate Wastewater
Wastewater samples were obtained from a nearby
petroleum refinery, and were identified as water
condensate from a hydrofining operation. This water
is known to contain large quantities of ammonia and
hydrogen sulfide, in addition to appreciable quantities
of dissolved organics. The phenol concentration in
the sample used for extraction runs was determined
through peak coincidence and comparison with standard
samples in flame-ionization chromatography with
Porapak-Q to be 400 ppm. Quite a bit of effort was
invested into searching for identification of several
other smaller peaks (Herz, 1973), without significant
success.
The measurement of the COD of this water was compli-
cated by the presence of hydrogen sulfide and other
sulfur-bearing compounds. A procedure was developed
wherein the steps of the COD analysis were carried out
as rapidly as possible, and the COD of the water was
thereby found to be 17,530 ppm. It was confirmed that
most of this was accounted for by hydrogen sulfide,
which was not expected to be removed readily by any of
the solvents under consideration. Methyl isobutyl
ketone was chosen as the solvent for the extraction
experiments carried out, and attention was focussed
upon the removal of the 400 ppm of phenol. In industrial
handling of this water, the hydrogen sulfide and
ammonia would be removed by a stripping process.
74
-------
Styrene Manufacture Wastewater
A sample of wastewater from a plant manufacturing
styrene by the dehydrogenatlon of ethylbenzene was
obtained through the services of the U. S. Environmental
Protection Agency and Dr. John H. Coco of Gulf South
Research Institute, New Orleans, La. Analysis of this
water by f1ame-ionization gas chromatography with
the Porapak Q column revealed only three significant
peaks. By comparison with standard solutions of known
concentration these were qualitatively and quantitatively
identified as
Benzene 345 ppm
Ethylbenzene 170 ppm
Styrene 10 to 20 ppm
This sample may have come from a plant employing
as a first step the process described by King (1970)
and discussed in Section III wherein wastewater con-
taining styrene is contacted with ethylbenzene
(and/or benzene) so as to remove most of the styrene
by extraction into the ethylbenzene (and/or benzene).
Since these compounds and any others expected to
be in this wastewater are refractory to chemical
oxidation as well as biological oxidation, COD
measurements for this water are potentially misleading.
Because these three compounds all have high
distribtuion coefficients into C4 hydrocarbons, this
water was treated by Process I, using isobutylene
as the sole solvent.
75
-------
Choice of Type of Mini pi ant Extractor
With the present state-of-the-art of solvent
extraction, the development of a new process
will always benefit from and will usually require
testing on a pilot plant scale (Treybal, 1966). A
spray column extractor was chosen for our initial
experimental studies because of its inherent simplic-
ity, its ability to handle waste waters containing
suspended solids without plugging, and its
flexibility of operation which makes it possible
to operate at a wide range of flow rates and
with either phase dispersed. In later studies,
when the desirability of operation at a low solvent-
to-water flow ratio had become apparent, we chose
an RDC (rotating-disc contactor -- see Figure 10) to
carry out experiments aimed at providing more quanti-
tative data on mass transfer rates and backmixing.
These data would be useful for scale-up and estimation
of commercial feasibility. A multiple-stage mixer-
settler with the means for recycling settled solvent
from the settler back to the previous mixer can also
be used to handle low flow ratios, but on a miniplant
scale such a device would be very complex because of
the need for multiple pumps and multiple agitator
drives. Also, in a large-scale installation the
settlers which would be required when treating
waste water streams of several hundred gallons
per minute become very large in horizontal cross-
section (but can be short in vertical height).
When many stages are required, such devices would
require a large floor area, which may not be
available for installing a waste water treatment
76
-------
Variable
Speed
Motor
Polluted
Water
Fresh
Solvent
\
Loaded
Solvent
H
Interface
Purified
Water
Figure 10. Rotating Disc Contactor
77
-------
plant, and would require a large capital investment.
One of the main advantages of mixer-settlers is
their reliability of scale-up because they normally
have a high stage efficiency on both scales. For
this reason we felt that by gathering data only on
an RDC we would be better able to develop a rec-
ommendation for choosing between these two contact-
ors than we would had we gathered data only on
a mixer-settler which would leave our understanding
of the RDC short of being able to make such a
recommendation.
To design any type of continuous, countercurrent
extractor, a design procedure is needed which
includes three main aspects (1) hydrodynamics, includ
ing droplet size estimation, flooding, and dis-
persed-phase hold-up, (2) axial mixing, and (3)
mass transfer. Such an understanding is also
needed to make the best interpretation of experi-
mental extraction data. A general discussion which
applies to the interpretation of data from both the
spray column and the RDC follows.
Axial Mixing and Mass Transfer in Continuous
Extractors
Solvent extraction devices can either be
staged, as with mixer-settlers where the two phases
are mixed and then settled within each stage, or
be continuous, as with a spray column or an RDC
where no separation of phases takes place except
at the ends of the extractor. Axial dispersion
or backmixing of one or both phases is an important
design consideration in continuous extractors.
78
-------
Axial mixing always tends to decrease the per-
formance of an extractor by decreasing the mass
transfer driving force below what would be predicted
if the fluids passed through the device in plug
flow. Mass transfer coefficients determined in
small-scale equipment without consideration to
axial mixing do not correspond to the coefficients
similarly determined in production-scale equipment.
The effects of axial mixing become more severe as
the ratio of diameter to height increases, and
usually become more severe with increasing diameter
when this ratio remains constant. Fortunately,
over the last decade, considerable progress has
been made in understanding and predicting the
effect of axial mixing.
Axial mixing in the continuous phase is the
sum of the effects of turbulent or eddy diffusion
along the axis of the extractor and of channeling
or radial diffusion resulting from a non-uniform
velocity field (Sleicher, 1959). One experimental
method for measuring axial mixing in the continuous
phase involves making a steady injection of a
tracer which is soluble only in the continuous
phase and measuring its concentration upstream
(with respect to the continuous phase) from the
point of injection. This method measures only
the eddy diffusion contribution to axial mixing
since channeling or radial diffusion does not
cause solute to move upstream. For the remainder
of this report this portion of the effect of axial
mixing will be referred to as "turbulent mixing."
Another method for measuring continuous-phase axial
mixing involves making a sudden injection of tracer
79
-------
and measuring the spreading of the peak by measuring
the tracer concentration as a function of time at two
downstream points. This method measures the contribution
from a non-uniform velocity field in addition to the tur-
bulent mixing contribution. Subsequently, the contribu-
tion to axial mixing due to a non-uniform velocity field
will be referred to as "channeling." In using
this term we do not mean to imply the type of
channeling that can occur in a small diameter
packed bed where most of the liquid flows close
to the wall, but rather we refer to the type
of non-uniform flow like that which occurs in an
open pipe in the laminar flow regime. The sum
of the effects of turbulent mixing and channeling
is the significant measure of axial mixing when
solute is being transferred from one phase to
the other in a solvent extraction column. Other
experimental methods for measuring axial mixing
are discussed in a review by Ingham (1971).
Axial mixing in the dispersed phase is
somewhat more complex than for the continuous
phase, and it includes the effects of turbulent
mixing, of mixing caused by the distribution of
droplet sizes and thus a distribution of residence
times (often called forward mixing), and of
coalescence and redispersion of the droplets
(Ingham, 1971). The various models which have
been proposed to account for axial mixing in
extraction columns have recently been reviewed
(Misek and Rod, 1971) and include the stage model
(Young, 1957), the backflow model (Sleicher, 1960;
Miyauchi and Vermeulen, 1963), the dispersion
80
-------
model (Sleicher, 1960; Miyauchi and Vermeulen, 1963),
and a combined model which includes forward mixing
effects (Olney, 1964). A random walk model has
been used in single-phase flow in packed-beds
(Jacques and Vermeulen, 1957; Cairns and Prausnitz,
1960).
The stage model uses only one parameter to
represent the mixing in both phases and is usually
inadequate (Misek and Rod, 1971). The backflow
and dispersion models have been shown to be equiv-
alent in the limit as the effective number of
stages for backflow becomes large (Hartland and
Mecklenburgh, 1966); both models require two
parameters to represent the backmixing, but the
stage model also defines a number of stages. The
combined model is much more complex as it requires
a knowledge of the dispersed phase droplet size
distribution. For the purpose of evaluating the
data taken in this report, we have chosen to use the
dispersion model. The details of the dispersion
model are developed in Appendix C.
If the counterflowing fluid phases are assumed
to move in plug flow with no axial mixing, then
the relationship between the terminal concentrations
can be expressed as follows (Colburn, 1939):
n = "° " (1)
Xwi - Xsi/Kd E ' exp C(1 • 1/E) Now] -1
where
n = Dimensionless concentration change defined
by the first equality,
81
-------
X = Pollutant concentration in the outlet
water phase,
Xwi = Po^utant concentration in the inlet
water phase,
X . = Pollutant concentration in the inlet
sol vent phase,
E = Extraction Factor = K^ • F /F ,
N = Number of overall water-phase transfer
units.
It is assumed in the derivation of this equation
that K., F , F and the overall water-phase mass
transfer coefficient per unit volume of extractor,
Kowa* are constant throughout the extractor. If
the solutions are dilute, these assumptions will
hold when the concentrations are measured in weight
fractions and the flow rates are measured in
weight per unit time. However, when F /F is small
s w
and extraction is efficiently carried out, then
the concentrations in the solvent phase can be
large even though X . is small. In this case
it is usually better to measure concentrations in
weight ratios (Ib. pollutant/lb. pollutant-free
fluid) and to measure flows as weight pollutant-
free fluid per unit time. As long as the two
fluids are mutually immiscible, flows measured in
this manner will be constant.
Equation (1) can be expressed simply as
follows:
n • f(E,Mow) (2)
Similarly, for the case where axial mixing is
82
-------
predicted by the dispersion model, an expression
is developed in Appendix C which is of the following
form:
" = f Pes> <3>
where
Pe = Water-phase Peclet number,
Pe = Solvent-phase Peclet number.
These two dimensionless parameters characterize
the extent of axial mixing in each phase. When
Pew = Pes = °°' ecluatl'on (3) 1>s tne same as tnat
for plug flow. When Pe « 0 for one or both
phases, equation (3) expresses the overall
concentration change when that chase is completely
mixed (i.e., has a concentration equal to its
outlet concentration at all points in the
extractor). The effect of axial mixing is
illustrated in Figure 11 where the case for Pe., =
W
10 and Pe = 20 is compared to that for plug
flow. For a given number of transfer units, the
decrease in removal efficiency due to axial mixing
is most significant when E is greater than 1.
Equation (3) may be used in the evaluation
of experimental data as well as in the design of
large-scale extractors. For many types of extrac-
tors, correlations are available in the literature
which relate Pe and Pe to the flow velocities,
W 5
droplet sizes, extractor dimensions, and physical
properties of the two phases. If the overall
extraction efficiency is measured experimentally
and used to calculate n. » then with Pe and Pe
w s
83
-------
— Plug Flow
0.07-
0.04-
ow
Figure 11. Comparison of Plug Flow and Dispersion Models
84
-------
determined from correlations and E determined from
the measured flow rates and the separately
measured distribution coefficient, equation (3)
may be solved implicitly to give a value of NQW
for each solute in the feed water.
The usual method of correlating mass transfer
data obtained from an experiment is to correlate
the individual mass transfer coefficients in the
separate phases. Thus, if a correlation is de-
veloped for the mass transfer coefficient in the
water phase and another for that in the solvent
phase, then by assuming that the resistances are
additive, the overall mass transfer coefficient
may be calculated at conditions different from
the actual experiment. The validity and limita-
tions in the additivity assumption have been
described by King (1964). Since the number of
transfer units is simply related to the mass
transfer coefficient per unit volume (k«a
where a is the interfacial area per unit volume),
the additivity of resistances can be expressed
as follows:
Now Nw Ns
where
= Number of water-phase transfer units,
's
w
= Number of solvent-phase transfer units.
There are a number of expressions available
which relate N and N to physical size and flow
W S
parameters and to solute diffusivities in the two
(4)
85
-------
phases; several such expressions are discussed
in Appendix B and are used for purposes of comparison
with data taken in this project. In a single experi-
ment where several solutes are simultaneously extracted,
the physical size and flow parameters are the same for
each solute. In many cases the solute-to-solute varia-
tion in diffusivity in the water phase (and in
the solvent phase) is small, so the solute-to-
solute variation in NW (and N ) is also small. The
influence of the resistance to mass transfer in one
phase on the resistance in the other phase can also
lead to minor variations in NW and NS from solute
to solute. However, according to equation (4) a
plot of l/nft vs 1/E should result approximately
o w
in a straight line since solute-to-solute variations
in N and N are much less than solute-to-solute
w s
variations in E.
As illustrated by equation (3) and in Figure 11,
once Pe and Pe are specified the dispersion model
W S
allows n to be calculated for each pair of values of
E and N . From the set of values of E and NQW which
give a specified value of n» a plot of 1/N vs 1/E may
be prepared. Such a plot is illustrated in Figure 12
for Pe = 10.0 and Pe = 20.0 where n, defined in equa-
W ^
tlon (1), is the parameter (actually Figure 12 is a
cross plot from Figure 11). Superimposed on Figure 12
is a straight line, as suggested by equation (4), for
a case where N., = 5.0 and N = 8.0 (typical of
W 5
operation in the miniplant RDC at a low flow
ratio). Thus for any value of E, the corresponding
value of N may be determined from the straight
ow
line using the ordinate, and the corresponding value
of n may be determined by interpolation between
86
-------
0
Figure 12. Plot of Equation 3 to Illustrate Additivity
of Resistances. Parameter is n.
87
-------
1i nes of constant n•
The values of N and N which result from
w s
the above type of analysis of experimental data
may be used to calculate the individual-phase mass
transfer coefficients from the following equations
once a is determined as described below:
a H
N. (F /A)
k, = -^ § (6)
5 a H
where A = Cross-sectional area of column,
H = Total height of column.
By comparing the values of the individual-phase
mass transfer coefficients with those predicted
from various models in the literature, the exper-
imental data can be used to infer the answers to
such questions as, "Do the dispersed-phase droplets
act as rigid spheres, or do they show internal
circulation?" and "Is interfacial turbulence
important, or do surface-active impurities inhibit
mass transfer?" The lack of a_ priori answers to
these types of questions is what makes pilot-scale
experimentation a necessary part of developing
a new solvent extraction process.
In this report the dispersion model has
been used to solve two distinct types of problems
which involve the analysis of continuous extraction
devices. In the Type 1 problem, the overall column
-------
height and the two phase velocities are specified.
If the removal efficiency is to be estimated, then
the physical properties for each phase are used with
three correlations for predicting (1) the hydro-
dynamics of the phases (from which the hold-up
and interfacial area per unit volume are calculated),
(2) the axial dispersivities , e and e (from which
Pe and Pe_ are calculated - see Appendix C),
w s
and (3) the individual-phase mass transfer coef-
ficients, k and k (from which N is calculated
Wo O W
using equations 4, 5, and 6). Equation (3) can
then be used directly to calculate n and the removal
efficiency. If the removal efficiency has been
measured and N is to be calculated, then the
physical properties and the same correlations are
used to estimate the hydrodynamics and Pew and Pes<
In this case equation (3) must be solved implicitly
for *„«•
In the Type 2 problem, the overall column
height is not specified, but rather it must be
calculated to provide a specified removal effi-
ciency. Since the column height enters into N ,
Pe , and Pe., equation (3) is more difficult to
w s
use. In this project a method has been used in
which H is assumed and the removal efficiency is
calculated as for a Type 1 problem. The correct
value for H is then determined by trial-and-error
calculation. Pratt (1975) has recently proposed
a more direct calculation which avoids this trial-
and-error procedure. The correlations which were
used for hydrodynamics, axial mixing, and mass
transfer for both the spray column and the rotating
-------
disk contactor for both types of problems are discussed
in detail in Appendix B, to which the reader is
referred at this point.
90
-------
SECTION VI
EXPERIMENTAL APPARATUS AND PROCEDURES
To carry out the experimental program de-
scribed in Section V, apparatus and procedures
had to be developed for (1) the analysis of aqueous-
and organic-phase samples taken during the treat-
ment of both actual and simulated waste waters,
(2) the measurement of equilibrium distribution co-
efficients for various solute-solvent combinations,
(3) the continuous, countercurrent contacting of
actual and simulated waste waters with both volatile
solvents and less-volatile, polar solvents, (4) the
continuous regeneration of volatile solvents by
distillation in equipment designed for operation
at moderate pressures (up to 50 psig), and (5) the
batch-wise regereration by atmospheric distillation
of less-volatile, polar solvents. During the
project the equipment and techniques were continually
modified and improved as problems were encountered
and solved. In the discussion to follow, the final
arrangement will be described along with the most
important modifications made during the project.
Analytical methods. The quantitative analyses
of aqueous-phase samples and samples of polar solvent
solutions during extraction mini-plant runs were made
by injection of 0.3 to 6yl samples into a Varian (Model
600-D) gas chromatograph equipped with a Varian (Model 328)
91
-------
temperature controller and a flame ionization
detector. For aqueous-phase samples, O.Syl injections
made using a Hamilton (Model 7001) microliter syringe
were found to be reproducible giving a precision
of ± 2 - 4% depending on the concentration and type of
organic solute present. Larger volume injections
using a Hamilton (Model 701) ten microliter syringe were
occasionally used for very dilute samples, but reproduce
bility fell to ±5 - 8% when using this larger syringe.
Injections greater than 8yl usually extinguished
the flame. For polar solvent solution samples,
0.5pl injections were always used as larger volumes
tended to saturate either the column or the detector
gtving an erratic peak size for the polar solvent.
Separate syringes for aqueous- and organic-phase
injections were used to minimize the difficulty
in flushing residue from the syringes.
The flow of helium carrier gas was maintained
at 20 - 30 ml/min through a 3 foot length of 1/8-inch,
stainless steel, packed chromatographic column. The
packing material found most universally useful
was Porapak Q (100 - 120 mesh) operated at tempera-
tures between 170 and 249°C. The temperature
was chosen to allow the separation of peaks while
giving a minimum time for the peak of longest
retention time to be completely emitted. In one
case when a water sample containing benzene and
n-butanol was analyzed, a 3 foot length of column
containing Porapak T (80 - 100 mesh) at 180°C was
used to provide an adequate separation of peaks
when it was found that they could not be separated
using Porapak Q.
92
-------
One problem often encountered after a column
had been used for several weeks was the appearance
of unexplained peaks. Such peaks, which would
appear even when pure distilled water was injected,
may result from an accumulation of organic material
on the first few inches of column that either reacts
with or is stripped by the water sample. The
retention time was reproducible, but the peak
size for these anomolous peaks usually varied by
about ±20%. As long as the unexplained peak did
not overlap with a peak from an important consti-
tuent in the sample this problem could be tolerated.
If necessary, the unexplained peaks could be elim-
inated by preparing, installing, and conditioning
a new column. During all analyses, occasional
pure water injections were made to check for
such anomolous peaks.
The hydrogen flow to the flame ionization
detector was maintained at 20 - 30 ml/min; the
air flow at about 300 rnl/min. A small flow
controller and a cartridge filter were installed
on both the helium and hydrogen flow lines. The
combination of a Porapak column and a flame ioni-
zation detector worked particularly well for
aqueous samples since the water retention time
was short (10 to 15 sec.) and the water response
was very small. Therefore, the water peak passed
quickly through the column and did not interfere
appreciably with the other peaks. In addition to
the analyses using gas chromatography, the chemical
oxygen demands (COD) of the feed waste water and of
the purified product water were determined for
the experiments in which industrial waste water
93
-------
was treated. The ASTM standard procedure (ASTM,
1970) was employed which uses potassium dichromate
in 50% sulfuric acid with silver sulfate as catalyst
to oxidize all organic substances in the sample.
To reduce the loss of volatile organics, the sample
was cooled in an ice water bath while the sulfuric
acid was being added to the sample (EPA, 1971).
The determination of COD was useful when the waste
water contained substances of very low volatility
which were not detected by gas chromatograohy.
The quantitative analyses of samples of
volatile solvent were made using a technique
developed by Fleck and Prausnitz (1968) and improved
in this work. During an extraction experiment
each sample was collected in a 60 ml sample holder,
as described below under "Experimental Procedure".
Following the experiment, the sample holder was
pressurized to a total pressure of about 50 psig
with dry nitrogen, was connected to another 60 ml
container by an 8 inch length of soft, Indalloy
(Indalloy #5 with a melting point of 134°C, Indium
Corporation of America) tubing (1/16" o.d. with
a 0.010" hole) as shown in Figure 13, and the empty
container and tubing were evacuated. Gaseous
volatile solvent was then added to the empty
container and tubing until the pressure was slightly
below the room temperature vapor pressure of the
volatile solvent. The valves on both ends of the
Indalloy tubing were opened and about 20 ml of
volatile solvent solution was allowed to flow
through the tubing. With the liquid still flowing,
the tubing was pinched off and cut at the center
94
-------
Vacuum
Vacuum
t
Volatile
Solvent -
Vapor
Valves
(
• - ^+~-
^ Indalloy Tubing J
•M
•^
>0
/
A - 60-ml Sample Holder with Glass Front
B - 60-ml Brass Container
Figure 13. Volatile Solvent Sampler
95
-------
with a Van'an swaging tool, and then the Indalloy
tubing was swaged into about 10 pieces each 3/4
to 1 inch in length. These encapsulated samples
of volatile solvent solution were dropped into
a beaker of water to check for leaking volatile
solvent and then were taken immediately for analysis.
A Perkin-Elmer (Model 990) gas chromatograph
equipped with dual columns and a temperature
programmer was used for quantitative analysis of
volatile solvent solutions and for all analyses
associated with the determination of equilibrium
distribution coefficients. An induction heater
was attached to the injection port of the chromato-
graph allowing the Indalloy-encapsulated samples
to be melted in a stream of carrier gas flowing
directly into one of the columns. As the low-melting,
Indalloy melted, the entire sample of volatile
solvent solution was vaporized into the chromato-
graph. The temperature of the chromatograph was
initially held at 150°C until the volatile
solvent peak was detected, then the temperature
was increased to a level allowing the solute
peaks to be quantified. The same type of packing
and the same gas flow rates were used with both
chromatographs.
The response factors of both chromatographs
were found to vary significantly with such variables
as gas flow rates and column temperature. Such
day-to-day variation made frequent calibrations
an essential part of the analytical procedure.
For the analysis of aqueous and polar solvent
solutions, a standard solution of approximately
the same composition as the unknown was prepared
96
-------
by weight, using an analytical balance which
reads to ± 0.1 mg. Three repetitive injections
were made with both the standard and the unknown
solutions, and the composition of the unknown was
determined from the average resoonses assuming
that peak area was proportional to concentration
over the slight concentration differences between
unknown and standard. Peak areas were determined
by a disc integrator except when peak overlap
occurred, in which case a procedure of cutting
out and weighing peaks was found to be considerably
more accurate. Typically the concentrations
determined in this manner are accurate to ± 2 - 4%.
The calibration procedure for volatile solvent
solutions was more difficult since the total
quantity of sample was not precisely controlled
with the encapsulated samples. A gas-tight, 0.5 cc
(Precision Sampling) syringe was used to inject
very accurate and precise volumes of pure volatile
solvent vapor into the gas chromatograph. By
measuring the pressure and temperature of the gas
in the syringe, the mass injected was determined
and a response factor curve was prepared. The
response factors for the solutes were determined
separately by injection of standard solutions of
these solutes in methanol using the microliter
syri nge.
With the response factors thus determined for
each component in the volatile solvent sample,
its composition could be calculated. Comparison
with prepared standard solutions of solutes in
volatile solvent indicated that the accuracy of
volatile solvent analyses was ±5%.
97
-------
The samples of both aqueous- and organic-
phase solutions were usually analyzed within two
days of when they were collected. During this
period the samples were stored at 4°C to minimize
biological oxidation. During the period in which
experiments with the spray column were conducted,
the non-volatile samples were stored in 5 cc vials
with a plastic cap and a paper seal. Later it
was determined that some organic solutes in aqueous
samples were absorbed into the oaper seals, causing
a decrease in concentration of as much as 1.5% per
day for a 300 ppm solution of n-butyl acetate.
For all subsequent experiments the paper seals
were reolaced by a Teflon disc which completely
eliminated this problem.
Equilibrium Determinations. Controlled measurements
of equilibrium distribution coefficients were made
for a large number of solute-solvent systems pertinent
to this work. The apparatus and procedure used and
the results obtained for these systems are reported
in Appendices E and F.
Spray column extractor. For the initial
experimental studies, a small spray column extractor
was constructed following the design suggestions
of Blanding and Elgin (1942). The distance between
the distributor plate and the main interface was
42 inches, and the column diameter was 1 inch.
The design capacity at 25% of flooding was for
a water flow rate of 1.0 gal/hr and for a C.
hydrocarbon flow rate of 3.4 gal/hr.
98
-------
As shown in Figure 14, the column was
built using 5 pieces of industrial glass pipe with
appropriate flanges and gaskets. The use of the
glass contactor made it possible to observe the
motion of the fluid phases, and thereby to judge
the accuracy of design drop size calculations,
the degree of turbulence, the possible flotation
of solid particles, and other performance charac-
teristics. All flow line and temperature well
connections were through two brass end plates
to eliminate the need for glass blowing with its
expected reduction in the strength of the tempered
glass pipe. As purchased, the maximum safe operating
pressure for the glass pipe was 50 psig, which
is only about 8 psi above the expected operating
pressure when using isobutane as solvent.
The design of the distributor plate has a
very important effect on the performance of
a spray column. For operation with the hydrocarbon
phase dispersed, the drops were formed by forcing
the solvent to flow up through 30 holes of 1/16
inch diameter (Figure 15A). The holes were formed
by drilling through a 1 inch diameter circular
disc of stainless steel without removing the burrs
formed by the exit of the drill. The flow line
connections were designed so that the column could
be inverted to study operation with the aqueous
phase dispersed. To obtain uniform drops of water,
'a different distributor plate was required.
Satisfactory operation was achieved when water
was forced to flow down through 7 holes of 1/16
inch diameter drilled through a Teflon disc of
1 inch diameter (Figure 15B). Holes drilled
99
-------
H
•E
•K
E -
F -
G -
tl —
I -
J -
Continuous Phase Inlet
Dispersed Phase Outlet
Continuous Phase Outlet
Dispersed Phase Inlet
Distributor Plate
Main Interface Position
2"-diameterX6" Glass Pipe
2" to 1" Glass Pipe
Reducer
l"-diameterX36" Glass
Pipe
1"-diameter Glass Tubing
(attached by epoxy resin)
K - Brass End Plate
All glass-to-glass and
glass-to-brass joints
separated by Teflon gaskets
and attached by standard
cast iron flanges.
D
Figure 14. Spray Column Extractor
TOO
-------
Stainless Steel
Disc with 30 Holes
of 1/16" Diameter
\
JUUUUK
y/
vXXXXXXX
*
I
'///<
Copper Shell
t
Solvent Inlet
A. Solvent Distributor Plate
Water Inlet
Copper Shell
Teflon Plug with
7 Holes of 1/16"
Diameter
B. Water Distributor Plate
Figure 15. Distributor Plates
101
-------
through copper did not result in aqueous drops of
a uniform size.
Rotating disc contactor. After initial
experiments, the spray column extractor was replaced
with an RDC to obtain effective operation and
representative mass transfer data while operating
with a low flow ratio. As shown in Figure 16,
the outer shell of the RDC was a 4 foot length of
3 inch diameter industrial glass pipe, and the
internal shaft, discs, stators, and support rods
as well as the external end plates were built
from type 316 stainless steel. The total 48 inch
length was divided into a 29 inch length of
active mass transfer region and two quiescent
zones of about 10 inches at either end for dis-
engagement of the phases. A stationary tubular
shield surrounding the rotating shaft extended 9
inches from each end plate to reduce turbulence
in these quiescent zones. The design capacity
was for a water flow rate of 5.0 gal/hr when the
mass ratio of solvent-to-water flow was 0.3 or less.
Not shown on Figure 16 but present on the pilot
plant RDC, a 4 foot length of 6 mm glass tubing
was connected by stainless steel tubing through
the end plates to serve as a sight glass. The
vertical distance between the level of the interface
within the sight glass and that in the column
divided by the vertical distance between the inter-
face in the column and the bottom rotating disc
provided an estimate of the solvent hold-up in
the mass transfer zone of the column.
102
-------
WI
In.
SO
USZ
MTZ
LSZ
WO
V^si
A - Upper End Plate
B - Mid-column Bearing
C - Lower End Plate
D - Dispersing Disc
E - Ball Bearing
I - Main Interface
M - Variable Speed
Electric Motor
WI - Polluted Water
Inlet
WO - Purified Water
Outlet
SI - Fresh Solvent
Inlet
SO - Loaded Solvent
Outlet
USZ - Upper Settling
Zone
MTZ - Mass Transfer
Zone
LSZ - Lower Settling
Zone
Figure 16. Rotating Disc.Contactor
103
-------
The rotating discs were 0.035 inch thick
with a 1/2 inch hole so that they would just slide
along the 1/2 inch rotating shaft. Discs of 1.50
and 1.75 inch diameter could be interchangeably
attached to the shaft. The discs were held firmly
in position by a threaded tightening nut at one
end, a fixed sleeve at the other end, and 1
inch lengths of tubing (0.506 inch i.d.; 0.563 inch
o.d.) which acted as spacers to separate each pair
of discs. The stators were 0.035 inch thick with
an outside diameter slightly less than the 3 inch
inside diameter of the glass pipe. Interchangeable
stators with holes of 2.00 and 2.25 inch diameter
were used. The stators were held in position by
3 rods of 3/16 diameter running the entire length
of the column. Spacers of 1 inch length were used
to separate the stators, and threaded tightening
nuts were used at each end of the rods.
To minimize vibration of the rotating shaft,
it was necessary to install a bearing midway
through the column. As with the stators, this
bearing was held in position by the three 3/16
inch rods plus two rubber o-rings which pressed
firmly against the glass pipe when the tightening
nuts on the rods were tightened. The details of
this bearing, built of Teflon with a 35% graphite -
65% Teflon solid inner bearing surface, are shown
in Figure 17. The cut-out area, which allowed the
fluids to flow past the bearing, provided an open
area about 24% of the total column cross-sectional
area. There were 12 discs above the bearing and 14
discs below it.
104
-------
Vertical Section A—A
Q,
B - Rotating Shaft
C - Rotating Disc
D - Stator Ring
E - Stator Support
Rod
F - Bolt to Hold
Bearing Parts
Together
G - Cutout Area
for Fluid Flow
H - Rubber 0-Rings
Top View
Materials Legend
Teflon
Teflon-
Graphite
Figure 17. Mid-Column Bearing
105
-------
The rotating shaft of the RDC was driven
by a 1/4 hp electric motor with a continuously
variable speed control (Minarik Model SL61 ) capable
of providing rotational speeds between 400 and
1600 RPM. The rotational speed was calibrated
with a stroboscope and was reproducible to ±5 RPM.
The motor was located about 12 inches below the
bottom end plate, and a ball bearing supported
the shaft just above the flexible coupling which
connected the motor shaft to the shaft of the RDC.
As shown in Figure 18, a rotating seal (U.S.
Seal Model PS-903) provided a positive seal where
the shaft passed through the bottom end plate.
Just above the seal, a Teflon-graphite insert
provided a bearing surface for the rotating shaft;
a similar bearing was installed in the upper end
plate, but no seal was required on the upper end.
The RDC was always operated with the solvent
phase dispersed. The solvent entered through a
length of 1/4 inch tubing which carried it into
the area between the rotating shaft and the stationary
tubing which surrounded the shaft. A short length
of Teflon tube was attached to the top of the
stationary stainless steel tubing; this Teflon
tube extended upward to within 1/16 inch of the
bottom rotating disc. Since the solvent preferen-
tially wetted the Teflon, a thin film of solvent
was emitted so that the shearing action of the
bottom disc broke it into small droplets.
When the column was in operation, the main
solvent-water interface was positioned about 2 inches
above the top disc (Figure 19). The inlet water
106
-------
t
Outlet
i or
Motor
Figure 18. Lower Part of RDC
Rotating Shaft
Stator Ring
Rotating Disc
Glass Pipe Wall
Slipon Teflon Cap
Stationary Tube
Teflon Gasket
Stainless Steel
End Plate
Teflon-Graphite
Bearing
Solvent Inlet
Stainless Steel
Cap
Fixed Ball Bearing
Flexible Coupling
107
-------
UT
1/16" Sample
Line
Stainless Steel
Cap
Solvent Outlet
Teflon-graphite
Bearing
Stainless Steel
End Plate
Stationary Tube
Glass Pipe Wall
Water Inlet Line
Main Interface
Sampling Point
Tightening Nut
Top Stator Ring
Figure 19. Upper Part of RDC
108
-------
flowed through a 1/4 inch stainless steel tubing
which extended downward to a position about 1 inch
above the top disc. A 1/16 inch diameter sample
line also extended downward from the top end
plate, allowing the aqueous phase to be sampled
just below the interface.
Solvent Regeneration Apparatus.
The apparatus used for the regeneration of
pollutant-loaded, volatile solvent included an
evaporator, a packed distillation column, a condenser,
a small reflux accumulator, and a reflux pump
arranged as shown in Figure 20. The equipment was
designed for a solvent flow rate of 3.4 gal/hr
and was thus oversized for the experiments using
a low solvent-to-water flow ratio. The regeneration
of less-volatile, polar solvent loaded with extracted
pollutants was carried out separately in an atmos-
pheric batch distillation apparatus.
The evaporator design took into account the
need for keeping unevaporated pollutants in the
liquid state before they are withdrawn, as well
as the need for variable and controlled temperature
differences between the heating medium and the
evaporating solvent. The loaded solvent was evap-
orated in a cylindrical, vertical copper tube
about 3 inches in diameter and 12 inches in length
as shown in Figure 21. The heat of vaporization was
supplied by a double-boiler arrangement, wherein
hot Freon-114 vapors were condensed onto the outside
of the tube wall. The Freon-114 was in turn
evaporated by a 1000 watt electrical heating
element (Chromalox Model Rl-1100) located inside
109
-------
Distillation
Column
Condenser
Reflux
Accumulator
Reflux
Pump
Regenerated
Solvent
Tank
A - Loaded Solvent Inlet
B - Pollutant Outlet
C - Cooling Water
Figure 20. Volatile Solvent Regenerator
110
-------
A - Pollutant-loaded
Solvent
B - Purified Solvent
Vapor
C - Pollutant Residue
Drain
D - Boiling Solvent
E - Concentrated
Pollutant
F - Boiling Freon 114
G - Condensing Freon
114
H - Electrical Heater
I - Heater Power
Control
J - Solvent Level
Gauge
K - Freon 114 Level
Gauge
L - Temperature Well
M - Freon Pressure
Gauge and Relief
Valve
Figure 21. Mini plant Evaporator
111
-------
a second, outer container, which surrounded the
lower 9 inches of the inner tube. The pressure,
and hence the condensation temperature, of the
Freon-114 could be varied by adjusting the power
input to the heater. The Freon-114 was charged
after thoroughly evacuating and flushing the outer
container, and since the vapor pressure at room
temperature is about 20 psig, no inert gas could
enter the evaporator. The concentrated pollutants
were intermittently withdrawn through a line at
the bottom of the inner evaporator tube. The
temperature of the solvent vapors leaving the
evaporator was measured by a thermistor in a
copper temperature well.
A 12 inch long by 2 inch diameter distillation
column packed with 1/4 inch ceramic Raschig rings
was mounted on top of the evaporator. The packing
was supported by a screen at the bottom, and the
liquid reflux was distributed over the top of
packing through 5 holes of 0.014 inch diameter.
The distillation column was covered on the outside
by foam rubber insulation.
The purified volatile solvent vapor leaving
the distillation column was condensed in a
double-pipe heat exchanger, using chilled water as
coolant. The heat transfer surface was a 30 inch
length of 1 inch diameter copper water pipe mounted
at a 45 degree angle, giving an inside surface
area of 0.67 sq. ft. Because of space limitation,
chilled water (about 2°C), rather than normal
cooling water, was circulated through the jacket
of this small condenser to provide the required
rate of heat transfer while maintaining the
112
-------
pressure of the condensing solvent below the safe
operating pressure of the glass extraction column.
The condensed solvent flowed by gravity into
the reflux accumulator shown in Figure 22. The
accumulator consisted of a 3/4 inch diameter copper
tube inside a 2 inch diameter copper pipe connected
so that the annular area between tubes acted to
store about 0.12 gal of liquid and so that the
inner tube acted as a weir allowing the overflow
to pass downward to the regenerated solvent tank.
A three-way ball valve arrangement allowed the
accumulated liquid to be emptied into the receiving
tank if it became contaminated with pollutant
(e.g., during startup).
A diaphragm metering pump (Lapp Pulsafeeder,
Model LS-10) with a very precise flow rate adjust-
ment was used to pump reflux liquid up through a
rotameter, through a flow restriction, and into
the distillation column at flows up to 0.61 gal/hr.
Because of the pulsating flow, the rotameter was
used only to check for proper operation; the flow
rate was determined from a separate calibration
of the pump control. A flow restriction con-
sisting of a 3/4 inch length of 0.012 inch dia-
meter tubing had to be installed between the
reflux pump and the distillation column to eliminate
vaporization in the pump. The pump was located
about 2 feet below the reflux accumulator to
provide the necessary suction head.
Figure 23 is a photograph of the entire
mini-plant. The RDC is located at the left, and
the other equipment is arranged as in Figure 20.
The dark colored tank below and just to the right
113
-------
B
A = From Condenser
B = Weir
C = 3-Way Ball Valve
D = To Reflux Pump
E = To Solvent Tank
F = Level Gauge
Figure 22. Reflux Accumulator
114
-------
Figure 23. Photograph of Entire Miniplant
115
-------
of the RDC is a 9-gallon water feed tank; the
light colored tank beside it is a 9-gallon volatile
solvent feed tank. Helium pressure was applied
above the liquids in the two feed tanks to force
the fluids to flow up through the equipment thus
avoiding the need for feed water and feed solvent
pumps. The other 9-gallon, light colored tank
is the regenerated solvent tank. The vertical
pipe just below the condenser was used as a 4-gallon
feed tank for the polar solvent experiments. The
instrument just to the right of this pipe tank is
an Atkins (Model 3L01J) thermistor reader which
was used to display the temperature measurements at
the top and the bottom of the extractor and at the
vapor outlet of the evaporator.
The pollutant-loaded polar solvent solutions
were not regenerated in the mini-plant because the
Freon-114 in the evaporator did not provide a high
enough temperature. This separation was made in
an atmospheric batch distillation apparatus
consisting of a 3 liter spherical glass boiling
flask driven by an electrical heating mantle,
a vertically mounted Allihn condenser which provided
reflux by circulation of air through the jacket,
and a water cooled Claisen head condenser.
Chemicals Used.
Volatile solvents were purchased from Matheson
Gas Products in pressure cylinders. C.P. grade
Isobutylene of 99.0% purity was found by gas
chromatographic analysis to contain about 0.04% of
a heavy oil (probably diisobutylene and higher
116
-------
polymerization products) and about 0.01% of tertiary-
butanol (the product of reacting water with iso-
butylene). Instrument grade isobutane of 99.5%
purity did not contain any detectable impurity
which could be extracted by water (such as t-butanol)
or separated by the miniplant distillation (such as
heavy oil). The isobutylene was purified before
use by running it through the miniplant using pure
water to wash out water soluble impurities and
the distillation to separate the heavy components.
This purification was necessary only because the
impurities could interfere with the analysis of
some pollutants and because we wanted to test for
solvent degradation which might occur during an
extraction experiment. Such impurities in the
make-up to a commercial installation would cause
no problem.
The n-butyl acetate used as a polar solvent
was purchased in two qualities: (1) purified
grade (Baker Chemical Co.), which contained 0.76%
n-butanol as the only appreciable impurity, and
(2) technical grade (Bryant Lab, Inc.), which
contained 8.26% n-butanol plus at much lower concen-
trations several other impurities which appeared
to be a higher molecular weight ester and several
low molecular weight alcohols. The purified grade
n-butyl acetate was used without purification
because the n-butanol did not interfere with pollu-
tant analyses in those experiments. The technical
grade solvent was purified by washing with water
in the RDC at a high water-to-sol vent flow ratio,
followed by a batch distillation. The early
fraction from the batch distillation was rewashed
117
-------
in a separatory funnel with water, then added
back to the distillation apparatus. The final
product from this treatment contained about 0.2%
n-butanol and almost no low-boiling impurities,
but still contained about 1% of a higher molecular
weight ester. The presence of 3% n-butanol in
the make-up solvent used in a commercial installation
could cause a problem if the rate of solvent loss
were high, but the steady state concentration of
n-butanol in the solvent would be low because it
would be washed out in the purified water.
The 2-ethyl hexanol used as a polar solvent
was practical grade (Matheson, Coleman, and Bell),
which contained less than 0.2% total impurities;
it was used without purification. The n-pentane
used in batch extractions was spectroquality
{Matheson, Coleman and Bell) of greater than 99%
purity and was used as purchased.
The various chemicals used to prepare synthetic
waste water solutions were reagent grade or purer
quality and were used without purification.
Experimental Procedure.
In a typical extraction experiment, whether
in the spray column or the RDC and whether using
volatile solvent or less-volatile polar solvent,
the procedure involved (1) preparing and charging
the water and solvent to the feed tanks,
(2) initializing flows in a specific manner that
would assure smooth operation and rapid approach
to steady state, (3) allowing sufficient time to
reach steady operation, (4) collecting final
118
-------
samples and recording measured flow rates, tempera-
tures, and pressures, and (5) terminating flows
in a manner that would leave all fluids in one
or another of the tanks with all other equipment
empty. Between experiments all pieces of equipment
which were in contact with pollutant-containing
fluids were disassembled and cleaned by water or
acetone followed by water washing.
The fluid samples were analyzed by gas
chromatography, resulting in data on the steady
state concentration of each pollutant in the feed
and product water, in the feed and product solvent,
and (with the RDC) in the water within the column
just below the main interface. From the determined
flows and concentrations a steady state material
balance could be calculated around the extractor.
After an experiment, especially if the steady
state balance did not check, the volumes and concen-
trations of the fluids remaining in the tanks would
be determined so that a "whole run" material
balance could be calculated. This would usually
show the location of any error in the "steady state"
material balance. Because of the low flow rate
of extracted pollutants leaving the bottom of the
evaporator, a steady state material balance around
the evaporator and distillation column was not
possible.
In preparation for an experiment in which
a simulated waste water was to be treated, up to
eleven 1-gallon bottles were filled with identical,
carefully measured amounts of each solute and
water. Since the water feed tank did not contain
an agitator, these bottles were shaken intermittently
119
-------
for several hours to assure complete dissolution,
and then about nine gallons were added to the
water feed tank. When industrial waste water was
to be treated, the preparation depended on the type
of water. Waste water from lube oil refining
was filtered through a coarse filter paper to
remove large solid particles which could plug
valves and rotameters, its pH was adjusted to
about 5, and it was added to the feed water tank.
Waste water from cresylic acid recovery was also
prepared in this manner. With the ethylene quench
water, the aqueous phase was separated from floating
and settled organic phase by syphoning through
a tube submerged in the sample bottle. Waste
water from the oxychlorination plant was neutralized
with NaHCO^. as described in Section V.
Before filling the feed water tank with the
waste water, it was washed with and then filled
with pure water. After pressurizing the feed
water tank with helium, the pure water was used
to fill the extraction column, thus purging the
air from it. When a volatile solvent was to be
used, about 500 cc of volatile solvent liquid
was then added to the extractor while allowing
some water to be displaced out the water-phase
outlet line. When a polar solvent was to be used,
a low-pressure dry nitrogen line was connected
through the solvent outlet line to pressurize the
column. Then, under the pressure of volatile
solvent vapor or dry nitrogen, about three-
quarters of the water in the column was removed
and discarded. The volume of the remaining water
120
-------
was recorded for use in calculating the whole
run material balance.
The volatile solvent was charged to the solvent
feed tank as a liquid by inverting the high pressure
cylinder so that its valve was below the level of
the liquid in the cylinder. Prior to transferring
the solvent, the solvent feed tank was evacuated
to about 1 mm Hg absolute pressure, and dry nitro-
gen was added to the high pressure cylinder to
provide a pressure driving force. When a mixture
of volatile solvent and less-volatile, polar
solvent was to be used, the solvent feed tank
was first evacuated, polar solvent was sucked
in, and then volatile solvent was added. After
both types of solvent had been added, the valves
on the tank were closed, and it was shaken by hand
to assure thorough mixing.
After the feed tanks were filled with waste
water and solvent and were pressurized with helium,
the extraction column was filled with waste water
to within about 2 inches of the top. Solvent
was then added to fill the column completely with
liquid. This method of filling the extractor
resulted in a concentration gradient at the begin-
ning of an experiment and thus decreased the amount
of time required to reach steady state.
The flow rates of both water and solvent
were next determined. With the solvent inlet and
outlet valves closed, the water flow was started
and controlled to the desired setting on the water
rotameter by adjusting the water outlet valve.
A volume of water was then collected during
a timed interval to establish the polluted water
121
-------
volumetric flow rate which corresponded to the
chosen rotameter setting. The solvent inlet
valve was opened, and the water inlet valve was
closed. The solvent flow was controlled to the
desired setting on the solvent rotameter by again
adjusting the water outlet valve. As the solvent
slowly displaced the water from the column, a
volume of the water was collected during a timed
interval to establish the solvent volumetric flow
rate which corresponded to the chosen rotameter
setting.
With both the inlet and outlet solvent valves
on the extractor closed, the waste water flow
rate was set at the desired value, with the inlet
valve completely open and the outlet valve used
for control with the flow being measured by the
rotameter between the feed tanks and the extractor
(Figure 24). The inlet solvent valve was then
opened completely, and the outlet solvent valve
was opened until the water flow into the column
was reestablished. Since the flow of solvent had
little effect on the column pressure, this procedure
does not change the rate at which water flows out
of the column; therefore, the interface was station-
ary. After about 10 minutes to allow the solvent
phase above the interface to come to approximately
the steady state concentration, the solvent outlet
valve was closed until the interface descended to
its desired location. If more than the original
9 gallons of waste water was required to complete
the run, all flows were stopped at this point,
122
-------
V
s
= Vent
= Sample Point
= Valve
Extractor
Product
Water
Tank
Condenser
He
Evaporator
Reflux Pump
8t Accumulator
Pollutants
Feed
Solvent
Tank
Flowmeters
Feed
Water
Tank
He
Figure 24. Pilot Plant 'Flow Diagram
Regen.
Solvent
Tank
123
-------
and the water feed tank was refilled. The above
precedure was then used to restart the flows.
By adjusting the water inlet and solvent inlet
valves and then reestablishing the water inlet
flow to the desired value by adjusting the solvent
outlet valve, any combination of solvent and water
flows could be set and still give a stationary
interface.
Before starting an experiment in which
volatile solvent was used, some pure volatile
solvent was passed through the column while it
was filled with pure water, and the reflux
accumulator was filled with pure liquid volatile
solvent by boiling it in the evaporator and con-
densing it in the condenser. Then, just before
the pollutant-containing volatile solvent was
first admitted to the evaporator, the reflux
pump was started and the evaporator heater acti-
vated. When this method was used, the only adjust-
ment that had to be made when the outlet solvent
valve on the extractor was opened was to increase
both the heater power and the cooling water flow
rate.
Once the apparatus was running, the only
adjustments necessary were to keep the solvent
and water flow rates constant by controlling the
solvent outlet valve and to keep the pressure in
the condenser constant by adjusting the heater
power. The approach to steady state was followed
by sampling the water outlet and analyzing for
the pollutant concentrations with the gas chroma-
tograph. Steady state operation usually required
124
-------
about 45 minutes, and the equipment could be oper-
ated for at least 30 additional minutes before the
feed water tank was emptied. During this
period the pollutants which accumulated in the
evaporator were intermittently withdrawn out the
bottom and were collected in a previously weighed
pressure vessel of about 500 ml volume.
After steady state was attained, final sam-
ples were taken of the outlet water, the outlet
solvent, and (with the RDC) the water within the
column. The solvent inlet and water outlet
valves were then closed until the interface ascended
to just below the top of the column, at which
time the solvent outlet valve was closed. The
column was then drained into the water receiving
tank, a glass carboy located within a fume hood
to eliminate the solvent vapors that were released
when the pressure was decreased across the water
outlet valve. The evaporator was allowed to
run until no more liquid was present in the evapor-
ation tube at which time the reflux pump was
stopped, the heater turned off, the condenser
water flow stopped, and the reflux accumulator
drained into the regenerated solvent tank.
After the run, the total quantity of water
treated was determined by measuring the volume of
the accumulated purified water and subtracting
the volume of the pure water initially present
in the column. The total quantity of solvent
used was determined from the volume in the re-
generated solvent tank. The total quantity of
concentrated pollutant residue was determined
125
-------
by weighing the collecting vessel. After
the concentrations of pollutants were determined
in these liquids, the whole run material balance
was readily calculated.
126
-------
SECTION VII
SPRAY COLUMN - EXPERIMENTAL RESULTS
The primary objective of experiments conducted
in the spray column was to prove the overall process
feasibility of volatile solvent extraction. The
majority of runs were conducted with isobutylene
as the dispersed phase; however, for comparison
water was dispersed into isobutylene in several
runs, and n-butane was dispersed in one experiment.
A variety of types of organic chemicals was extracted,
including one aromatic hydrocarbon (benzene), two
alcohols (n-butanol and isoamyl alcohol), two
esters (vinyl acetate and n-butyl acetate), two
ketones (acetone and methyl ethyl ketone), one
aldehyde (crotonaldehyde) , one nitrile (propionitrile) ,
one chlorinated hydrocarbon (ethylene dichloride),
and several phenolics (phenol; o-cresol ; m, p-cresol;
and xylenols). Most of these solutes are known to
be present in the waste water from chemical processing
plants, but they were also chosen to determine if
any class of organic chemicals might cause unexpected
problems such as an irreversible reaction with iso-
butylene. Most of the solutes studied were present
in synthetically prepared water solutions, but samples
of lube oil refining waste water and cresylic acid
recovery waste water were also tested. Many more indus
trial waste waters were tested with the RDC extractor
127
-------
(see Section VIII) .
Each experiment was conducted in the miniplant
spray column extractor as described in Section VI.
In addition to the types of solutes and the choice of
volatile solvent, the principal independent variables
were the water and solvent flow rates. The tempera-
ture was measured but not controlled. The principal
measured response was the concentration of each
solute in the product water. Approximate measurements
were made of the droplet size, by holding a scale
against the glass column, and of the time for a droplet
to travel from the distributor to the main interface,
by using a stopwatch. From the "whole run" material
balance (see Section VI), it was possible to estimate
the effectiveness of solvent regeneration. However,
the procedure for sampling and analyzing the pressur-
ized solvent was not perfected until after the
experiments in the spray column had been completed.
In this chapter we first describe the method
of reducing the data on solute removal efficiencies
to mass transfer coefficients. A sample calculation
is described for Run SS12A (SS = Spray column with
Solvent dispersed; 12 = sequence number; and A =
first setting of solvent and water flow rates).
The correlations which describe the hydrodynamics
and axial mixing in the spray column are shown to
provide a reasonable description of its operation.
The mass transfer models for circulating and
oscillating drops are shown to bracket the experi-
mental results for Run SS12A.
The experimental results are then described
and discussed with respect to answering such
questions as which phase should be dispersed and
128
-------
when are solute interactions important. The
results from several experiments indicate that the
solvent phase rather than the aqueous phase should
be dispersed. Experiments in which the performance
using n-butane as solvent was compared to the
performance using isobutylene show that isobutylene
is usually preferred. Isobutylene tends to give
both higher distribution coefficients and higher
mass transfer coefficients. While experiments did
not lead to a clear understanding of interactions
between solutes, the results did indicate that
such interactions can be important. Mass transfer
rates were often found to be higher than theoretically
predicted for oscillating drops as well as for
circulating drops. Runs are described in which
lube oil refining waste water and cresylic acid
recovery waste water were treated by volatile solvent
extraction. Further details on the method of
data reduction along with a listing of the computer
program used to make the calculations and the experi-
mental results for all 37 runs conducted in the
spray column are listed in Appendix G.
Method of Data Reduction.
Although the comparison of solute removal
efficiencies is a useful means of comparing various
modes of operation, a better comparison is possible
in the spray column experiments by reducing the data
to mass transfer coefficients (or equivalently,
numbers of transfer units) and then by comparing these
to theoretical and empirical predictions. This
procedure requires the use of the design methods
described in Section V and Appendix B to predict the
129
-------
effects of hydrodynamics, axial mixing, and mass
transfer rates. The procedure of data reduction is
described for Run SS12A, an experiment in which a
prepared waste water containing about 2000 ppm each of
acetone and n-butyl acetate and about 4000 ppm each
of methyl ethyl ketone (MEK) and crotonal dehyde was
treated with isobutylene as the dispersed phase.
Estimation of physical properties. The estima-
tion of solvent droplet size and solvent holdup
requires data on the densities of both phases, on
the interfacial tension, and on the aqueous-phase
viscosity. In addition, for estimation of mass
transfer rates, data are required for the solvent
viscosity and the diffusivities of each solute
through water and through isobutylene. The values
used at the measured temperature of 21.6°C are listed
in Table 8. The aqueous-phase density and viscosity
were taken as those listed by Weast (1970) for pure
water. The density of the solvent phase was taken
to be that of pure isobutylene (API, 1963), and the
solvent-phase viscosity was estimated by extrapolating
low-temperature (-110°F to +10°F) viscosity data for
n-butane, iso-butane, and 1-butene (API, 1963) and
assuming that the viscosity of isobutylene is the
product of the viscosity of 1-butene and the ratio
of viscosities of iso-butane to n-butane. The inter-
facial tension was estimated by plotting data for
several hydrocarbons for which data is available
(n-pentane, n-hexane, n-heptane, and n-octane) in
the manner suggested by Donahue and Bartell (1952).
These authors showed that the interfacial tension of
many systems fell on a single line when plotted as
130
-------
Table 8. Physical Properties for Run SS12A
Continuous-phase density = 0.9979 gm/cc
Continuous-phase viscosity = 0.9704 cp
Dispersed-phase density = 0.5914 gm/cc
Dispersed-phase viscosity = 0.182 cp
Interfacial tension = 41.5 dyne/cm
Solute diffusivities (105 x ft2/hr) and distribution
coefficients:
Solute DC Dd Kd
Acetone 4.22 26.2 0.63
MEK 3.87 24.0 2.49
Crotonaldehyde 3.88 24.0 2.48
n-Butyl Acetate 2.71 20.4 168.
131
-------
interfacial tension vs. the logarithm of the sum of
mole fraction solubility of water in the organic-
phase and mole fraction solubility of the organic in
water. The mutual solubilities for the hydrocarbon-
water binary systems were taken from API (1963).
The diffusivities were estimated as recommended by
Reid and Sherwood (1966) using the method of Scheibel
(1954) and assuming that the values at infinite
dilution would apply at the low concentrations in the
experiment. The distribution coefficients were taken
from Appendix E assuming the values at 25°C and at
infinite dilution would apply.
Estimation of hold-up and Peclet number. In
Run SS12A the isobutylene flow rate was 21.3 gal/hr
which when passed through the 30 holes of 1/16-inch
diameter in the distributor gave a discharge velocity
of 0.1236 ft/sec. Since this is considerably less
than the jetting velocity of 1.12 ft/sec predicted
by the method of Scheele and Meister (1968), their
correlation could reliably be used to estimate the
average drop size of 0.1819 inch. This agreed very
well with estimates made by visual observation. By
using the correlation of Minard and Johnson (1952),
operation during this run was estimated to be 11.9%
of flooding (i.e., 11.9% of the fluid velocities at
flooding with the same ratio of phase flows).
The method of Hughmark (1967) was used to estimate
the hold-up as 0 = 0.0272, and the correlation of
Henton, et al . (1973) predicted Pe = 2.017 as an
estimate of axial mixing. Assuming the velocity of
the rising droplets to be V./0 leads to a rise time
132
-------
of 6.6 sec which agrees with an average experimental
estimate of 7 sec.
Circulating-drop estimate of mass transfer.
Using the estimates of droplet size and hold-up, the
correlation of Ruby and Elgin (1955) was used to
estimate k for each solute, and equations (5) and
C
(B5) were used to calculate N (=N..). The dispersed-
\* w
phase mass transfer coefficient for circulating drops
was estimated by the equation of Kronig and Brink
(1950) which assumes no resistance to mass transfer
in the continuous phase. In Run SS12A the water
flow rate was 0.948 gal/hr, which when combined with
the solvent flow rate and the two fluid densities
led to FC/F1( = 1.331, from which the value of E for
o W
each solute was determined. The predicted values of
NW and NS were then combined according to equation
(4) to give the predicted value of N . The results
ow
of these estimates are listed in Table 9 and plotted
as the upper curve on Figure 25. R (= k /k^K. =
N /N E) is also included in Table 9.
Oscillating-drop estimate of mass transfer. A
similar procedure was followed using the equations of
Angelo} et al. (1966; 1968) for oscillating drops;
these results are also listed in Table 9 and plotted
as the lowest curve on Figure 25. The major factor
increasing the rate of mass transfer for oscillating
drops relative to circulating drops is a 3- to 5-fold
increase in N- .
Experimental mass transfer results. To compare
the experimental results with the predictions for
circulating drops and for oscillating drops, the
removal efficiency data were corrected for end
133
-------
Table 9. Predicted Mass Transfer Rates
Acetone MEK Crotonaldehyde
E 0.84
Circulating
N 2.46
w
N 0.61
s
N 0.42
ow
R 4.8
Oscillating
N 2.84
w
N 3.15
s
M_ 1.37
ow
R 1.1
3.31
Drops :
2.34
0.58
1.06
1.2
Drops :
2.72
3.02
2.14
0.27
3.30
2.35
0.58
1.06
1.2
2.73
3.02
2.14
0.27
224.
1.90
0.54
1.87
0.016
2.28
2.78
2.27
0.004
134
-------
3.0
2.5
2.0
o
1.5
1.0
0.5
I I I I I T
Run SSI2A
Oscillating Drops
I I i i I I
0 0.2 0.4 0.6 0.8 1.0 1.2
I/E
Figure 25. Mass Transfer Rates
135
-------
effects and converted to values of N . In using
ow
equation (14) to correct for mass transfer during
solvent droplet formation, the continuous-phase
concentration was assumed to be the measured outlet
aqueous-phase concentration, and the feed solvent
was assumed to be free of solutes. The correction
for mass transfer during water droplet formation was
made by assuming the continuous solvent-phase concen-
tration was that which would satisfy a material
balance around the spray column. The corrections
for mass transfer during droplet coalescence were
made assuming the continuous-phase concentration was
the feed concentration (i.e., neglecting the concentra-
tion jump) and assuming the dispersed-phase concentra-
tion was the concentration of the product. The
measured (before drop formation and after drop
coalescence) and corrected (for end effects)
aqueous-phase concentrations which are listed in
Table 10 show that these corrections are relatively
minor, which justifies the approximate nature of the
corrections. The corrected concentrations were used
to calculate n according to equation (1). Using
the previously estimated Pe = 2.017 and assuming that
Pes is infinite, equation (3) was solved implicitly
to give the values of N l( for each solute as listed
U W
in Table 10 and plotted on Figure 25. The experi-
mental results fall between the curves for oscillating
drops and for circulating drops for all solutes except
n-butyl acetate.
All the curves in Figure 25 deviate from straight
lines which would occur if the values of N (and
W
NS) were the same for each solute. The prediction
136
-------
Table 10. Experimental Mass Transfer Rates
Acetone MEK CrotonaIdehyde n-Butyl Acetate
CO
Ppm in Feed Water
Ppm Corrected* Feed
Ppm in Product Water
Ppm Corrected* Product
Ppm in Feed Solvent
Ppm Corrected* Solvent
Experimental N
ow
2058
2038
1307
1329
0.0
16.6
0.648
0.69
4167
4078
1579
1621
0.0
31.3
0.396
1.40
4422
4328
1620
1663
0.0
32.1
0.382
1.48
2212
2150
620
637
0.0
13.0
0.296
1.68
* for end effects
-------
for circulating drops shows curvature because of
differences in solute diffusivities . Additional
curvature would have been introduced had the correc-
tion to additivity of the individual-phase resistances
been included in the predicted curves. The correc-
tion due to the interaction of resistances is
largest at R=l and falls to zero at large and at
small values of R. The point R-l occurs near 1/E=0.25
for the circulating drop model and near 1/E=1.1 for
the oscillating drop model. The corrected curves
would lie slightly below those plotted, following
results presented by King (1965) for various simpli-
fied models. The curvature in the prediction for
oscillating drops is due to variations in solute
diffusivities. The fact that the experimental line
shows curvature like that for the oscillating drops
does not necessarily mean this model is most repre-
sentative.
Because of the complex nature of these calcu-
lations, it is not readily apparent how errors in
the assumed values of physical properties will
influence the comparison between experimental and
predicted values of N . It is unlikely that the
estimates for densities or viscosities are much in
error, but the interfacial tension could be sub-
stantially reduced by the presence of the solutes
and other impurities. If the interfacial tension
were 20.0 dynes/cm rather than 41.5 dynes/cm, the
solvent hold-up would be increased from 0.0272 to
0.0314, and the average droplet diameter decreased
from 0.1819 inch to 0.1494 inch. These changes would
result in the increased rate of mass transfer for
both circulating and oscillating drops shown by
138
-------
the dashed lines in Figure 26. The increase in
estimated hold-up would result in an increase in Pe
G
which causes a slight change in the experimental
curve. The estimated values of the solute diffusi-
vities could also be in error. Reid and Sherwood
(1966) found when using the Scheibel method that the
average error for the diffusivity of organics in water
was 11% and for the diffusivity of organics in other
organic solvents was 25%. The vertical bars on the
predicted curves in Figure 26 show the range of
changes in N caused by an 11% increase in all values
of D and a 25% increase in all values of D^
or by an 11% decrease in D 's and a 25% decrease in
D.'s. Changing diffusivities has almost no effect
on the experimental curve since they only enter into
the minor correction for end effects. Figure 26
shows that these errors in physical properties would
not change the location of the experimental curve
from that of lying between the curves for circulating
drops and for oscillating drops.
Choice of Dispersed Phase.
The results of Runs SS11 and SW3 give some
insight into how operation with isobutylene dispersed
compares with operation with water dispersed. In
both these experiments a single solute, ethylene
dichloride (measured K. = 70.0, by the method reported
in Appendix E), was present in the feed water at
about 3000 ppm. When isobutylene was dispersed in
two experiments in which the water flow rate was
held constant at 1.50 gal/hr, the removal efficiency
decreased from 87.2% with an isobutylene flow rate
of 1.76 gal/hr to 62.0% with an isobutylene flow
139
-------
3.0
2.5
2.0
1.5
0
I I
Run SSI2A
I I
cr = 41.5 dynes/cm
-------
rate of 0.700 gal/hr. This decrease must reflect
a decrease in k a since the major resistance to
C
mass transfer is in the water phase. The main
reason for this decrease is the decrease in solvent
hold-up and therefore in a. The results of comparing
the experimental with the predicted values for N
0 W
are shown in Table 11. For Runs SS11A and SS11B the
experimental N was larger than the prediction for
oscillating drops as well as that for circulating
drops, but the ratio of experimental N to predicted
o w
N , for oscillating drops was constant at about 2.2
O W
for these two runs.
When water was dispersed into isobutylene at
five different flow settings in Run SW3, the ethylene
dichloride removal efficiency ranged from 51 to 55%
and was nearly independent of solvent flow rate; it
showed only a slight variation with water flow rate.
Since the resistance to mass transfer was almost
entirely in the water phase, the solute removal
efficiency from a droplet of water falling through
the solvent is expected to be only slightly affected
by the solvent flow rate. When the water flow rate
was decreased, the droplet size decreased slightly
which accounts for a slightly improved removal
efficiency. The results of comparing the experimental
with the predicted values of N... are shown in
0 W
Table 11; they indicate that the experimental N
O W
fell between the values predicted for circulating
and for oscillating drops. The ratio of experimental
Nrtll to N«,, predicted for oscillating drops ranged
o w ow
from 0.82 to 0.92 for these five runs with the water
phase dispersed.
141
-------
Table 11. Comparison of Runs with Different Phases Dispersed
ro
(Solute = Ethylene Dichloride)
Run #
SS11A
SS11B
SW3A
SW3B
SW3C
SW3E
SW3D
Vw vs E Pec
(ft/hr) (ft/hr)
36.8 43.
36.8 17.
36.8 43.
36.8 30.
36.8 17.
27.2 43.
27.2 30.
2
2
2
2
2
2
2
48.8 2.93
19.4 2.19
49.1 2.57
34.3 1.79
19.5 1.02
66.3 2.35
46.3 1.64
% Removal
87.2
62.0
51.1
52.0
50.9
55.4
55.1
Now
(experimental)
3
1
0
0
0
0
0
.09
.24
.71
.74
.72
.80
.80
Now Now
(circulating) (oscillating)
1.15 1.
0.49 0.
0.26 0.
0.26 0.
0.26 0.
0.26 0.
0.26 0.
37
57
86
85
85
88
87
-------
Further insight into the question of which
phase to disperse was gained from Runs SS9 and SW1.
In these runs, isoamyl alcohol (measured K^ = 3.53)
was present at about 2000 ppm in the feed water in
addition to about 3000 ppm of ethylene dichloride.
The changes in removal efficiencies with changes in
isobutylene flow rate were similar to those in the
experiments where only ethylene dichloride was
present. However, a comparison of experimental values
of N with values of N oredicted for oscillating
ow ow •
drops as listed in Table 12 shows that the rate of
mass transfer was greater than that predicted for
oscillating drops for both water dispersed and solvent
dispersed. Comparing the values of N for ethylene
dichloride with and without isoamyl alcohol being
present shows that when the alcohol was present,
the rate of mass transfer was increased. This is
probably the result of interfacial turbulence being
promoted by the presence of the alcohol. The incon-
sistently high experimental values of N in Run
SS9C were traced to a sample bottle which did not
seal. The product water sample showed a concentration
of ethylene dichloride which was too low due to
volatilization losses; the loss of isoamyl alcohol
was also appreciable but was not as significant as
the loss of the more volatile ethylene dichloride.
One additional set of runs was completed with
water as the dispersed phase. In Run SW2 about 3000
ppm of ethylene dichloride and about 2000 ppm each
of n-butanol (measured K. = 0.76) and n-butyl
acetate (measured Kd = 168.0) were present in the
feed water. The results of a comparison of
experimental and predicted NQW are shown in Table 13.
143
-------
Table 12. Comparison of Runs with Different Phases Dispersed
Run #
SS9A
SS9B
SS9C
SW1B
SW1A
(Solutes
VM Ve Solute
W 5
(ft/hr) (ft/hr)
51.3 43.2 IAA
EDC
51.3 30.2 IAA
EDC
51.3 17.2 IAA
EDC
51.3 43.2 IAA
EDC
51.3 17.2 IAA
EDC
= iso-Amyl Alcohol and Ethylene Dichloride)
E
1.76
34.9
1.23
24.4
0.70
13.9
1.77
35.1
0.70
14.0
Pe % Removal
G
4.12 71.0
92.5
3.66 52.7
78.1
3.07 47.0
(98.4)
2.91 44.0
68.9
1.15 30.9
59.1
Now
(experimental)
2.28
3.96
1.24
2.02
1.53
(9,53)
0.74
1.19
0.65
0.94
Now
(oscillating)
0.79
1.00
0.56
0.72
0.33
0.42
0.69
0.88
0.68
0.87
-------
Table 13. Additional Runs with Water as the Dispersed Phase
-P.
en
(Solutes - n-Butyl Alcohol/ Ethylene Dichloride
Run * Vw
(ft/hr)
SW2A 36.8
SW2B 36.8
SW2C 27.2
V Solute
s
(ft/hr)
43.2 n-BuOH
EDC
n-BuAc
30.2 n-BuOH
EDC
n-BuAc
30.2 n-BuOH
EDC
n-BuAc
E Peg
0.53 2.57
49.0
118.
0.37 1.79
34.3
82.2
0.50 1.64
46.3
111.
% Removal
25.1
82.5
79.1
23.4
80.2
75.0
27.0
83.0
78.3
, and n-Butyl Acetate)
N
ow
(experimental)
0.47
1.80
1.59
0.61
1.69
1.41
0.62
1.85
1.55
N
OW
(oscillating)
0.46
0.86
0.75
0.46
0.86
0.75
0.47
0.88
0.76
-------
Although the theory for oscillating drops (Angelo,
et al., 1966 & 1968) has not been well tested,
the fact that the experimental values of NQW are larger
than predicted for oscillating (as well as for cir-
culating) drops may be another indication of inter-
facial turbulence.
It is interesting to compare the removal
efficiencies of ethylene dichloride and n-butyl
acetate in this experiment. Both predicted and
experimental values of N were larger for ethylene
dichloride than for n-butyl acetate, even though
the latter had a larger distribution coefficient.
The reason for this difference was the larger diffu-
sivity for ethylene dichloride. The extraction
factor for both these solutes was so large that
almost all the resistance to mass transfer was in
the aqueous phase where the diffusivity of ethylene
dichloride was predicted to be 3.37 x 10"5 ft2/hr
while that for n-butyl acetate was predicted to
be 2.51 x 10"5 ft2/hr.
These experiments in which the choice of which
phase to disperse was considered can now be summarized.
For cases where the flow rates of the two phases were
about equal, better performance was obtained by dis-
persing the volatile solvent. The mass transfer
rates when the solvent was dispersed were greater
relative to predictions for oscillating drops than
when the water phase was dispersed. Also, the
predictions for oscillating drops were larger with
the solvent dispersed, mainly because this mode of
operation gave a continuous phase with a much larger
viscosity leading to a longer drop residence time.
146
-------
Since the performance fell less sharply with decreas-
ing solvent flow when the water was dispersed than
when the solvent was dispersed, there is likely
a lower limiting ratio of solvent to water flow
below which dispersing the water will result in
a greater solute removal efficiency. However,
operation with a very low solvent-to-water flow
ratio in the spray column gives poor solute removal
efficiencies no matter which phase is dispersed,
so other types of extractors would likely be pre-
ferred. In general, it appears preferable to dis-
perse the solvent for all modes of operation giving
a high removal efficiency.
Choice of Type of Volatile Solvent.
The results from Run SS10 where n-butane was
dispersed, when compared to Run SS9 where isobutylene
was dispersed, give some insight into the relative
merits of these two volatile solvents. In Run
SS10 n-butane was used to treat a prepared waste
water containing about 3000 ppm of ethylene dichloride
(measured K. = 44.0) and about 2000 ppm isoamyl
alcohol (measured K^ = 1.41); this feed water had
nearly the same composition as that used for Run
SS9. The predicted values of N , for oscillating
o w
drops are compared to experimental values of N for
Runs SS10 and SS9 in Table 14. A more detailed study
showed that predicted values of N were essentially
W
the same for the two volatile solvents, while N was
about 7% larger for isobutylene than for n-butane.
For ethylene dichloride, where the resistance to mass
transfer was almost entirely in the aqueous phase,
the predicted values of N were essentially equal.
147
-------
Table 14. Comparison of Results When Using Different Volatile Solvents
(n-Butane used
Run * V V Solute
Mr 5
(ft/hr) (ft/hr)
SS10A 51.3 43.8 IAA
EDC
_ SS10B 51.3 31.4 IAA
-F*
CO
EDC
SS10C 51.3 17.4 IAA
EDC
SS9A 51.3 43.2 IAA
EDC
SS9B 51.3 30.2 IAA
EDC
in
Run
SS10; i-Butylene used in
Kd
1
44
1
44
1
44
3
70
3
70
.44
.0
.44
.0
.44
.0
.53
.0
.53
.0
0
20
0
15
0
8
1
34
1
24
E Pe % Removal
w
.68 3.85
.9
.49 3.47
.0
.27 2.88
.32
.76 4.12
.9
.23 3.66
.4
42.
74.
29.
59.
19.
42.
71.
92.
52.
78.
6
6
4
1
1
7
0
5
7
1
Run SS9)
N
ow
(experimental )
1.10
1
0
1
0
0
2
3
1
2
.75
.61
.07
.41
.63
.28
.96
.24
.02
N
ow
(oscillating)
0.63
1
0
0
0
0
0
1
0
0
.00
.46
.74
.26
.42
.79
.00
.56
.72
-------
However, for isoamyl alcohol, which had lower values
of K, in both solvents and therefore had a larger
resistance to mass transfer in the solvent phase,
both the lower Kd and the lower N for n-butane as
solvent resulted in substantially lower values being
predicted for N .
The experimental values of N in Table 14 show
ow
that the factor which was even more significant than
a larger value of K^ in isobutylene was the fact that
the ratio of experimental NQW to predicted NQ
was much larger for isobutylene than for n-butane.
Therefore, isobutylene would in general be the pre-
ferred choice as volatile solvent. Even for solutes
which show a very large value of Kd, the better
ratio of experimental to predicted N (if this factor
occurs for other waste waters) would result in
higher removal efficiencies with isobutylene. One
factor which does weight in favor of butane in terms
of process costs is the lower solubility and lower
solvent make-up.
Solvent reactivity could also affect the
choice of volatile solvent. In one early experiment
(to be described below as Run SS6) in which iso-
butylene was used to extract phenol from a waste
water, a solute was detected in the product water
which was not present in the feed water. This
solute was identified by chromatographic analysis
to be tert-butanol. Since a 2% phenol solution is
slightly acidic (pH=4) and since isobutylene is
known to react with water in acidic media to form
tert-butanol, an experimental program was carried
out to determine the importance of this reaction.
149
-------
Isobutylene does not undergo appreciable hydration
in 2% phenol solution, and the source of the tert-
butanol was found to be the feed isobutylene which
contained about 0.01% tert-butanol impurity. However,
during the program isobutylene was found to undergo
both hydration and dimerization in more acidic
solutions. Therefore, butane would be the preferred
volatile solvent for the treatment of a waste
water which is acidic and which can be reused without
neutralization if the organic solutes are first
removed. The highly acidic waste water from certain
steps in the production of vinyl chloride would
be an example of such a waste water.
Interactions Among Solutes.
The previous results showed that the presence
of isoamyl alcohol could improve the removal of
ethylene dichloride, even though both solutes were
present at such high dilution that their values of
Kd should equal Kd at infinite dilution. This is
an example of solute interactions. A series of
experiments was conducted in the spray column ex-
tractor to try to improve our understanding of these
interactions and also to try to determine under what
conditions interactions tend to increase values
°f NoW
In this series of experiments two solutes with
widely differing values of K ., acetone with K. = 0.63
and n-butyl acetate with K, = 168.0, (Appendix E),
were included so that the relative importance
of mass transfer in each phase could be estimated.
In Run SS15 about 2000 ppm each of only acetone and
n-butyl acetate were present. The results of an
150
-------
analysis of the mass transfer are shown in Table 15.
In Run SS15A the experimental values of NQW were much
larger than those predicted by the oscillating drop
model; the measured removal efficiency for acetone
was so large that according to the dispersion model
it should not have been possible even with an infi-
nite number of transfer units.
There are several possible explanations for
experimentally determined values of N being infinite.
Had the actual removal efficiency for acetone been
59.4% rather than the measured value of 71.2%,
then the experimental value of N would have been
0 W
4. Although the difference in concentration of
acetone in the product water (831 ppm for 59.4%
removal, rather than the 590 ppm measured) is much
larger than the expected analytical error, acetone
could have been lost through vaporization during the
short period between when the sample was collected
and when it was analyzed by gas chromatography.
Another possibility is that the extent of backmixing
was overestimated. If Pe., = » (plug flow in both
W
phases), the measured removal would have been
possible with N = 3.31 for acetone and N ,, = 5.15
0 W 0 W
for n-butyl acetate. However, even these values
are larger than what was estimated for oscillating
drops. Whatever the true explanation is, this run
and Run SS15B indicate that the rate of mass transfer
for the simultaneous extraction of acetone and n-butyl
acetate was far more rapid than predicted for oscil-
lating drops.
In Run SS7 isobutylene was used as the dis-
persed phase to extract a mixture of about 2000 ppm
151
-------
en
ro
Run ft Vw
(ft/hr)
SS15A 23.3
SS15B 23.3
SS7 21.6
SS12A 23.3
SS12B 23.3
SS16 23.3
Table 15.
Vg Solute
(ft/hr)
52.3 Acetone
n-BuAc
26.9 Acetone
n-BuAc
68.0 MEK
Crotonal .
52.3 Acetone
MEK
Crotonal.
n-BuAc
83.2 Acetone
MEK
Crotonal.
n-BuAc
52.3 Acetone
MEK
Crotonal.
n-BuAc
Interaction Aroong Solutes
E
0.84
224.
0.43
115.
4.65
4.63
0.84
3.31
3.30
224.
1.33
5.28
5.25
356.
0.84
3.31
3.30
224.
Pe % Removal
w
2.01 71.2
99.4
1.59 34.6
93.2
2.09 77.7
77.0
2.02 36.5
62.1
63.4
72.0
2.44 47.0
74.0
74.0
80.5
2.03 45.0
68.2
67.7
76.2
Now
(expt'l)
00
14.7
1.60
5.44
2.51
2.43
0.69
1.40
1.48
1.68
0.94
2.00
2.00
2.25
1.16
1.81
1.77
1.98
N
ow
(circul)
0.42
1.86
0.22
1.00
1.46
1.46
0.42
1.06
1.06
1.87
0.66
1.65
1.64
2.86
0.43
1.06
1.06
1.93
Now
(oscill)
1.36
2.26
0.72
1.19
2.95
2.95
1.37
2.14
2.14
2.27
2.15
3.36
3.36
3.57
1.39
2.12
2.12
2.32
-------
each of MEK (Krf = 2.49, Appendix E) and croton-
aldehyde (Krf = 2.48, Appendix E). As shown in
Table 15, the experimental values of N for each
ow
solute were about 84% of what was predicted for
oscillating drops. In Runs SS12 and SS16 all four
of these solutes were present. The results show that
all solutes gave experimental values of N less than
ow
those predicted for oscillating drops. For the case
of n-butyl acetate in Runs SS12A and SS12B, the
experimental values of NQW were even less than those
predicted for circulating drops. Thus, when MEK and
crotonaldehyde were added to the water stream containing
acetone and n-butyl acetate, the factors which caused
the rapid mass transfer during the extraction of acetone
and n-butyl acetate alone seem to have been damped out.
In Run SS13 a prepared feed water containing
about 2000 ppm each of acetone and n-butyl acetate
plus about 300 ppm of benzene (measured K. = 407.0)
and about 4000 ppm of n-butanol (measured K, = 0.76)
was treated by isobutylene extraction. The results
are compared with those for extraction of acetone
and n-butyl acetate alone in Table 16. In Run SS13A
as in Run SS15A the experimental removal efficiency
for acetone was larger than what was predicted for
perfect mass transfer, but when the solvent flow
was reduced the experimental values of N in both
0 W
Runs SS13B and SS15B were more nearly in agreement
with the prediction for oscillating drops. One
possible explanation is that the value of Pe increased
more rapidly with Vd than was predicted by equation (B6)
For the purpose of comparison, the last column in
Table 16 lists the experimental values of N
ow
153
-------
Table 16. Interaction Among Solutes
en
Run f Vw
(ft/hr)
SS13A 23.3
SS13B 23.3
SS15A 23.3
SS15B 23.3
V Solute
(ft/hr)
52.3 Acetone
n-BuOH
n-BuAc
Benzene
26.9 Acetone
n-BuOH
n-BuAc
Benzene
52.3 Acetone
n-BuAc
26.9 Acetone
n-BuAc
E Pe % Removal
w
0.84 2.01
1.01
224.
543.
0.43 1.58
0.52
115.
279.
0.84 2.01
224.
0.43 1.59
115.
74.3
77.9
99.7
98.1
28.0
35.3
90.7
93.1
71.2
99.4
34.6
93.2
N N
ow ow
(expt with (expt with
Pe finite)
CO
03
17.5
9.17
0.66
1.05
4.47
5.32
oo
14.7
1.60
5.44
Pew = °°)
4.17
3.42
5.70
3.91
0.52
0.74
2.35
2.63
3.31
5.15
0.88
2.66
N
ow
(predict
oscill)
1.37
1.38
2.27
2.62
0.71
0.72
1.18
1.36
1.36
2.26
0.72
1.19
-------
calculated by assuming plug flow for both phases.
Actual removal efficiencies of 61.0% and 66.7% for
acetone and n-butanol , respectively, would have been
necessary to give N = 5 for these solutes. Con-
sidering all this information leads one to conclude
that mass transfer was more rapid than predicted
for oscillating drops in both Runs SS13 and SS15.
Therefore, the addition of n-butanol and benzene
did not damp out the high rate of mass transfer found
for the extraction of acetone and n-butyl acetate
without other components.
The results of these experiments do not lead
to a clear understanding of this type of interaction
between solutes, but they show that it can be impor-
tant. The fact that this complicated interaction
took place even for feed water prepared from pure
chemicals indicates the need for mass transfer data
using actual samples of industrial waste water.
In Run SS5 isobutylene was used to treat a
waste water containing about 12,000 ppm each of
phenol (K. = 0.70, Appendix E) and vinyl acetate
(measured K. = 52.0). The measured removal
efficiency of vinyl acetate (at E = 69.2) was 99.2%
which corresponds to 13.2 for the experimental value
of N as compared to 2.53 predicted for oscillating
drops. The measured removal efficiency for phenol
(at E = 0.93) was 80.0%, which is not theoretically
possible even for infinite N and plug flow, if the
value of K. at infinite dilution applies. Using the
estimated value of Pe.. = 2.01, the dispersion model
w
predicts 64.2% removal for phenol at N . = 5. By
U W
a material balance the product isobutylene contained
8941 ppm vinyl acetate.
155
-------
The results shown in Appendix F and by Won (1974)
lead to the conclusion that the value of K™lx for
phenol distributing between water and a mixture of
isobutylene and n-butyl acetate is approximately
given by the following for low concentrations of
phenol :
i/mix RA /, i .IB m
Kd " XBA Kd + (1 ' XBA} Kd (7;
where XD. = wt. n-butyl acetate/(wt. n-butyl acetate
'BA
+ wt. isobutylene),
RA
Kj = distribution coefficient into pure
n-butyl acetate,
T R
Ki = distribution coefficient into pure
isobutylene.
The value of K")IX for phenol distributing between
water and a mixture of isobutylene and vinyl acetate
is about equal to that for a mixture of the same
weight fraction of isobutylene and n-butyl acetate.
Assuming KJ]A= K^A = 57.0, the value of K™1x for
phenol distributing between water and an organic
phase of the same composition as the product solvent
is 1.20, which is high enough to explain the 80.0%
removal of phenol. A calculation procedure that
qualitatively explains this result is described in
Appendix C.
Regeneration of Loaded Solvent.
At the completion of each of the experiments
described so far, a sample of the regenerated solvent
was analyzed using the Indalloy sampling method
described in Chapter IV. Unfortunately the method
of calibrating the gas chormatograph response was
156
-------
later found to be inaccurate. However, by assuming
that the response factors for each component did not
change with the quantity injected, a comparison
of the chromatogram for the regenerated solvent with
that for the loaded solvent provided an accurate
estimate of the efficiency of solvent regeneration.
The ratio of solute in the loaded solvent to that
in the regenerated solvent ranged from 22 for ethylene
dichloride to more than 200 for isoamyl alcohol.
This result is a good indication of the relative ease
of separating the solute from the volatile solvent,
and it also serves to demonstrate the feasibility
of solvent regeneration.
In Run SS8 a prepared feed water containing
about 4000 ppm propionitri1e (measured K. = 1.80)
and about 6000 ppm n-butanol (measured K. = 0.76)
was treated by isobutylene extraction. The removal
efficiency for propionitri1e varied from 91 to 96%
and that for n-butanol varied from 79 to 88%; the
results of an analysis of mass transfer are shown in
Table 17. Once again the experimental value of N
ow
was much larger than that predicted for oscillating
drops. However, when the concentrations of solutes in
the loaded solvent were compared to those in the regen-
erated solvent, we found that the concentration of propi
onitrile had only been reduced by 47% during solvent
regeneration. A likely explanation is that propi-
onitrile forms an azeotrope with isobutylene (esti-
mated to contain 0.1% propionitrile from the
estimate of the regenerated solvent composition)
which prohibits more complete regeneration of the
solvent. The concentration of n-butanol in the
157
-------
Table 17. Extraction of n-Butanol and Propionitrile
Run # V
SS8A
15.4
tn
00
SS8B
15.4
V
w s
(ft/hr) (ft/hr)
36.1
68.0
Solute E Pe
w
n-BuOH 1.06 1.15
Pr-CN 2.51
n-BuOH 1.99 1.48
Pr-CN 4.72
% Removal
N
N
ow - ow
(experimental) -(.oscillating)
78.5
90.7
88.1
96.1
00*
16.8
12.8
14.1
1.44
2.10
2.69
3.92
* Dispersion model predicts for E = 1.06, Pe =1.15, and N = 12, % Removal = 67.8
-------
loaded solvent was reduced by 97% in the same regener-
ation.
Of all the solutes studied, only propionitrile
showed a tendency to form an azeotrope. Since the
most important other nitrile pollutants (acetonitri1e
and acrylonitri1e) are more volatile than propioni-
trile, it is likely that all these nitriles will
form azeotropes with isobutylene. Thus, volatile
solvent extraction of the lower-molecu!ar-weight
nitriles is not feasible in a process using iso-
butylene as solvent and distillation for regeneration,
Industrial Waste Haters.
The initial runs made to establish and improve
the experimental apparatus and techniques were
conducted in preparation for treating samples of
lube oil refining waste water. For this reason
in Runs SSI, SS2 and SS4 a prepared feed water
containing about 2% phenol and 0.1 to 1.0% o-cresol
was treated by isobutylene extraction. In these
runs the phenol removal efficiency varied from 76
to 88%, and the o-cresol removal efficiency varied
from 95 to 97%. The results of an analysis of mass
transfer rates are shown in Table 18. The experimen-
tal values of N were much larger than N predicted
ow ow r
for oscillating drops in all cases; two results
showed that the measured phenol removal efficiency
should be impossible even with infinite N . The
results indicate that interfacial turbulence was
probably occurring during the extraction of these
relatively concentrated water streams.
In Runs SS3 and SS6 samples of lube oil
refining waste water were treated by isobutylene
159
-------
Table 18. Extraction of Phenol and o-Cresol
en
o
Run # Vw Vs
(ft/hr) (ft/hr)
SSI 19.4 82.4
SS2 22.2 71.6
SS4A 22.5 71.6
SS4B 22.5 52.5
Solute
Phenol
o-Cresol
Phenol
o-Cresol
Phenol
o-Cresol
Phenol
o-Cresol
E
1.76
12.1
1.34
9.21
1.32
9.05
0.97
6.63
Pe % Removal
w
2.03 79.1
95.1
2.18 88.2
96.6
2.23 85.6
96.1
1.96 76.0
95.3
Now
(expt'l)
4.76
6.66
00*
8.42
(37.5)*
7.56
00*
8.11
N
ow
(oscill)
2.52
4.02
1.89
3.00
1.90
3.03
1.40
2.23
* Dispersion model predicts for the listed values of E and Pe and N =5 for Phenol
w ow
Run SS2, % Removal = 74.8; Run SS4A, % Removal = 74.7; and Run SS4B, % Removal = 65.3
-------
extraction. The phenol and o-cresol removal effi-
ciencies in Run SS3 were 40% and 67%, respectively.
The substantial reduction in removal efficiency as
compared to the prepared waste waters was determined
to be due to two factors -- (1) the feed isobutylene
had not been properly regenerated and contained an
unknown amount of phenol and o-cresol from a previous
run, and (2) the pH of the waste water was about 9,
which has been shown by Beychok (1967) to lead to
a reduction in K. for phenolic compounds. These
results led to the installation of the refluxed
distillation column on the miniplant evaporator
and to the practice of acidifying this waste water
to a pH of about 4 prior to phenol extraction. The
isobutylene used in the extraction of prepared waste
waters had been either used fresh or regenerated by
back extracting the phenol and o-cresol into a dilute
caustic solution. Run SS3 also led to the realiza-
tion that solutes other than phenol and o-cresol
were present at lower concentrations; these were
later identified as acetone, MEK, and benzene.
In Run SS6 another sample of lube oil refining
waste water was treated by isobutylene extraction.
The results of an analysis of mass transfer rates
are shown in Table 19; the last column headed "Q"
is the ratio of experimental N to N predicted
ow ow
for oscillating drops. This is the first case where
Q was consistently less than 1 for some solutes
(acetone and benzene in Run SS6B) and greater than
1 for the other solutes. The values of Q for phenol
and o-cresol, the major pollutants, were fortunately
the largest. It is interesting to note that Q
was smallest for the solute with the lowest and
161
-------
Table 19. Extraction of Lube Oil Refining Waste
ro
Run f V
V
w s
(ft/hr) (ft/hr)
Solute
Pe
'w
% Removal
N
N
ow ow
(expt'l) (oscill)
SS6A 21.6 92.2 Acetone
Phenol
MEK
o-Cresol
Benzene
SS6B 21.6 52.0 Acetone
Phenol
MEK
o-Cresol
Benzene
1.60 2.35
1.77
6.31
12.2
1032.
0.90 1.87
1.00
3.56
6.86
582.
56.8
80.2
94.8
97.5
95.9
40.5
58.9
76.3
83.9
79.4
1.37
4.74
7.03
8.98
5.99
0.85
2.51
2.62
3.31
2.27
2.56
2.48
3.88
3.96
4.90
1.47
1.43
2.24
2.28
2.82
0.53
1.91
1.81
2.27
1.22
0.58
1.76
1.17
1.45
0.81
-------
the solute with the highest value of K.; this tends
to rule out a correlation between Q and E. Q
generally decreased as V. decreased at a constant
value of V .
L>
In Run SS14 cresylic acid recovery waste water
was treated by isobutylene extraction. The removal
efficiencies varied from 72% for phenol to almost
96% for xylenols. The results of analyzing the mass
transfer aspects of this run are shown in Table 20.
The values of Q were again greater than 1 for all
solutes.
In all three runs where isobutylene was used
to treat industrial waste water, the feed water
contained suspended, fine particles, but in no case
was there a substantial flotation effect. During
the treatment of lube oil refining waste water there
was a slight accumulation of suspended material at
the main interface. This accumulation did not
appear to be the fine particles, but rather a
sticky material that caused some problems of delayed
droplet coalescence.
The measurement of Chemical Oxygen Demand (COD)
for the feed and treated industrial waste waters can
yield useful data about the fate of dissolved and
suspended organic pollutants which are not observed
with the gas chrotnatograph. In Run SS6 the COD's
of the feed and product lube oil refining waste
water streams were not measured. However, subsequent
experiments in which this waste water was treated
(see Section VIII) show that COD calculated from
the amounts of identified pollutants agrees
well with the measured COD. In Run SS14 the measured
COD's of the feed and treated cresylic acid recovery
163
-------
Table 20. Extraction of Cresylic Acid Recovery Waste
V
V
Solute
Pe
% Removal
N
N
Q
(ft/hr) (ft/hr)
SS14 20.1 60.6 Phenol
m, p-Cresol
o-Cresol
Xylenols
w
ow
(expt'l)
1
4
8
12
.25 1.85
.82*
.57
.5*
71
91
89
95
.8
.3
.8
.6
5
6
4
7
.01
.12
.60
.40
ow
(oscill)
1
2
2
2
.78
.60
.83
.79
2.
2.
1.
2.
82
35
62
65
* E for m, p-Cresol calculated assuming Kd = Kd (for m-Cresol) =2.7, and E for Xylenols
calculated assuming Kfl = Kd (for 3, 5-Xylenol) = 7.0 (Appendix E).
-------
waste water were 4050 ppm and 1070 ppm (74% reduction)
These values can be compared to the COD's of the
feed and treated streams as calculated from the iden-
tified phenolic compounds which were 3840 ppm and
560 ppm (84% reduction). The nearly constant
difference between measured and calculated values
indicates that the components which account for this
difference were not well extracted. The fine solids
which were obviously not well removed probably contrib-
uted to the difference between measured and cal-
culated COD's.
Overall Process Feasibility.
In the experiments conducted in the spray
column extractor, a wide variety of organic solutes
was extracted by isobutylene, and in no case was
there an irreversible reaction between solvent and
solute. The only case of any solvent degradation
occurred in an auxilliary experiment where isobutylene
hydration was noted in a strongly acidic aqueous
soluti on.
The removal efficiency for most cases indicated
a mass transfer rate substantially larger than that
predicted for oscillating drops. However, an un-
explained damping out of the factors which caused this
high rate was noted for all combinations of solutes
containing MEK and crotonaldehyde. Even with
the two samples of industrial waste water, there
was no indication of an interfacial effect which
would inhibit mass transfer. The use of solvent
extraction did not lead to an appreciable flotation
of suspended solids.
165
-------
The short, refluxed distillation column
provided a very effective means for solvent regener^
ation. Only in the case of propionitri1e was
solvent regeneration not feasible, probably
because of an azeotrope.
166
-------
SECTION VIII
RDC EXTRACTOR - EXPERIMENTAL RESULTS
During the second portion of the experimental
program, several prepared aqueous solutions and
industrial waste waters were treated by solvent
extraction in the RDC as described in Section VI.
Experiments included runs using volatile solvents,
less-volatile polar solvents, and mixtures of
volatile and polar solvents. In these experiments
the solvent was always the dispersed phase, and
the solvent-to-water flow ratio (F /F ) was set
s w
at much lower values than with the spray column.
F /F was varied over the range from 0.1 to 0.3 in
S W
order to demonstrate solvent extraction under condi-
tions which would be most likely to lead to favorable
process economics.
In addition to the choices of solvent and the
value of F /F , the independent variables which could
be varied on the RDC included the water flow rate,
the disc rotational speed, the disc diameter, the
stator hole diameter, and the compartment height.
The principal measured responses were the solvent
hold-up, the concentrations of each solute in the
product water and in tjie water within the RDC just
below the main interface, and in most runs the
concentration of each solute in the loaded solvent.
167
-------
The temperature was measured but not controlled. An
approximate measurement was made of the maximum
stable solvent droplet diameter by holding a scale
against the glass column. From the "whole run"
material balance (see Section VI), it was possible
to estimate the effectiveness of solvent regeneration
in runs using a volatile solvent. The batch distil-
lation of polar solvent also gave a qualitative
measure of the ease of solvent regeneration.
The method by which the independent variables
were chosen in each experiment involved some trial
and error. Treybal (1963) cites the following ranges
for the ratio of column diameter to rotating disc
diameter, D/d., and the ratio of column diameter
to compartment height, D/H , as preferred proportions
c
for commercial RDC extractors:
D/d. = 1.5 to 3
D/HC = 2 to 8
The dimensions of RDC extractors in the studies
reviewed by Ingham (1971) fell into these ranges, and
the ratio of column diameter to stator hole diameter,
D/ds, in these studies fell into the following range:
D/ds = 1.2 to 1.6
The value of D was chosen to be as large as possible
and still be small enough so that the pressure within
the column could be contained by industrial glass pipe
For operation with isobutane as solvent this resulted
in D = 3 inches. For the initial tests based on
the above ranges of ratios, the other dimensions
were chosen as follows:
d. = 1 .50 inch
168
-------
ds = 2.25 inch
HC = 1.00 inch
In the studies reviewed by Ingham (1971) where
water was the continuous phase, the highest value of
V in each study ranged from 15 to 50 ft/hr. Since
for most industrial waste water samples we could
obtain only 5- to 10-gallon quantities we chose to
operate at the low end of this range with a maximum
V of 15 ft/hr and a minimum Vr of 6 ft/hr.
c c
Once the column dimensions, V , and F /F
\+ o W
had been chosen, the remaining independent variable
was the disc rotational speed, N. Correlations such
as that of Logsdail, et al. (1957) can be used to
estimate the value of N at the flooding point, but
as discussed in Section V flooding in an RDC does
not suddenly occur as N is increased. Also, the
value of the interfacial tension is required in the
correlation, and the interfacial tension for a waste
water can be substantially below the known value for
the solvent-water binary mixture.
For these reasons the operating value of N was
chosen by experimental observation as follows. With
the waste water and solvent flows established, N was
slowly increased. At low rotational speeds, the
solvent droplets were large and good contact between
water and solvent was not obtained. As N increased,
contact between phases improved until small solvent
droplets could be seen in the quiescent zone below
the bottom disc. When a value of N was determined
which resulted in the entrainment of what appeared
to be about 1% solvent in the water phase, N was
decreased by about 10% to establish its operating
169
-------
value. This procedure resulted in operation at a
practical set of conditions which should be reasonably
close to what would be used in a commercial extractor.
In this chapter we first describe an experiment
which was designed to determine how useful existing
correlations for RDC extractors would be in predicting
the observed performance with respect to hydrodynamics,
axial mixing, and mass transfer rates. The correlations
are shown to depend strongly on the choice of the
constant G,g (Eq. B12). The remaining experiments were
grouped according to the type of industrial waste water
being studied. The lube oil refining waste water was
treated using dual solvent extraction with n-butyl acetate
and isobutylene. Both process arrangements of dual
solvent extraction are demonstrated to result in
effective removal of phenol and o-cresol and moderate
removal of the other solutes. Experiments in which
the ethylene quench water was treated by volatile
solvent extraction show effective removal of aromatic
hydrocarbons; isobutane was a better solvent than
isobutylene for removal of dispersed organics.
Treatment of the oxychlorination waste water using
2-ethyl hexanol is shown to be complicated by a
simultaneous chemical reaction between the solvent
and the principal solute, chloral hydrate. The
experimental results for all runs conducted
in the RDC are listed in Appendix H; run numbers
start with the prefix RS to signify Rotating
disc extractor with Solvent dispersed.
170
-------
Test Run To Check RDC Correlations.
One experiment was conducted in which n-butyl
acetate was dispersed with F_/F = 0.1 into a water
s w
stream which contained methyl acetate (measured
K. = 3.64), ethyl acetate (measured Kd = 11.2) iso-
propyl acetate (measured K^ = 34.1), and o-cresol
(K. = 206, Appendix F). This prepared water was
not made up to simulate any particluar waste stream,
but rather the solutes and their concentrations
in the feed water were chosen to satisfy the
following criteria .which were established to
obtain the maximum amount of information about the
rates of mass transfer:
1. Solutes should cover a range of E from
about 0.3 to greater than 10,
2. Solutes and the solvent must be separable
using a gas chromatograph so that their concentrations
can be simultaneously determined,
3. The concentrations of all solutes must
be high enough in the product water so that their
concentrations can be accurately determined,
4. The concentrations of all solutes must
be low enough in the feed water to be completely
m i s c i b 1 e,
5. The concentrations of all solutes should
be as low as possible so that each value of K, can be
estimated at its infinite dilution value, and
6. Solutes should ideally be separable from
the solvent by distillation so that the solvent can
be regenerated.
Since n-butyl acetate was a solvent of particular
interest in this study, it was chosen for this test
171
-------
run. The choice of the lower molecular weight ace-
tates provided solutes with Kd from 3.6 to 34 while
satisfying requirements 2 through 6. Krf was
found to vary only a little with solute concentration
for these systems. Amy! acetate (measured K^ = 700)
could not be used as a solute showing a very high value
of K. because its solubility was too low to satisfy
both requirements 3 and 4. Cresol was chosen since
it had K, = 206 and still had a water solubility of
greater than 2%. However, the presence of about
5% o-cresol in the loaded solvent may have affected
K. at the upper end of the column, thus violating
requirement 5.
The results of this test run (Run RS13) are
shown in Table 21. The percentage of each solute
removed varied from 20.3% for methyl acetate (E=0.41)
to 95.5% for o-cresol (E=23.0). The n-butyl acetate
in the product water includes 4 ppm (of the 6600 ppm)
due to entrained n-butyl acetate phase. This estimate
was determined by collecting the n-butyl acetate
phase which appeared on the surface of the treated
water after it had separated in the receiving tank
on sitting for about five days.
At the column temperature of 21.6°C, the
physical properties of the pure solvent and of pure
water along with the solute diffusivities as estimated
by the method of Scheibel (1954) are listed in Table
22. The solvent density and viscosity are from
Toropov (1956). The interfacial tension at 20°C
was taken from Logsdail, et al. (1957), and the
correlation of Donahue and Bartell (1952) was used
to correct for the slight difference in temperature.
172
-------
Table 21. Results from RDC Test Run RS13
Water flow rate
Solvent flow rate
Rotating disc diameter
Stator hole diameter
Rotational speed
4.79 gal/hr
0.606 gal/hr
1.50 inch
2.25 inch
800 RPM
Analytical results (concentrations in ppm):
Methyl Ethyl i-Propyl o-Cresol
Acetate Acetate Acetate
Feed
Water
Product
Water
Percent
Removal
Water in
Column
Loaded
Solvent
227.
181.
20.3
235.
497.
270.
81.6
69.8
248.
1882.
676.
85.3
87.4
509.
5944.
5622.
255.
95.5
3349.
51120.
n-Butyl acetate in product water = 6600 ppm.
Measured solvent hold-up = 0.0428
Measured column temperature = 21.6°C
Estimated maximum drop diameter =0.05 inch
173
-------
Table 22. Physical Properties for Run RS13
Water-phase density = 0.9979 gm/cc
Water-phase viscosity = 0.9642 cp
n-Butyl Acetate density = 0.8792 gm/cc
n-Butyl Acetate viscosity = 0.708 cp
Interfacial tension = 13.9 dyne/cm
Solute diffusivities (10 x ft /hr) in water (D )
\^r
and in n-butyl acetate (D^) :
Solute D
C
Methyl, Acetate 3.93 6.45
Ethyl Acetate 3.40 5.97
i-Propyl Acetate 3.04 5.60
o-Cresol 3.02 5.57
174
-------
From the information in Tables 21 and 22 the
steady state material balance around the extraction
column was calculated. It showed that the total
quantity of each solute present in the product
water and in the loaded solvent exceeded the measured
quantity of solute in the feed water by 2.7% for
methyl acetate, by 4.1% for ethyl acetate, by 6.0%
for isopropyl acetate, and by 1.0% for o-cresol .
The closures for these solutes in this run were
slightly worse than for a typical run with a polar
solvent and were somewhat better than for a typical
run using a volatile solvent.
The inlet flows to the column were FS = 4.45
1b/hr and FW = 39.9 Ib/hr (FS/FW = 0.111). Based on
the concentration of n-butyl acetate in the purified
water, the quantity of solvent decreased by about
6% as it passed through the column. However, the
total solute extracted into the solvent almost
exactly balanced the dissolved solvent so that the
total solvent-phase mass flow rate was nearly constant
For this reason the analysis of this run using
the dispersion model was made in terms of total flows
and weight fractions rather than solvent flows and
weight ratios.
The experimentally measured fractional solvent
hold-up is an average value for the portion of the
column between the bottom disc and the main inter-
face. Visual inspection showed that there was
a variation in hold-up along the column. The drops
were not broken up to an equilibrium drop size
distribution at the bottom disc, but rather there
was an obvious increase in hold-up as the drops
175
-------
were broken up between the bottom disc and about the
fifth disc from the bottom. Between the fifth disc
and the twelfth disc the hold-up appeared to be
approximately constant. The hold-up increased
substantially between the twelfth disc and the mid-
column bearing because of the decrease in free flow
area offered by the bearing (about 24% free area
compared to CR = 0.52). For several discs above the
bearing the hold-up increased until it reached a
value about like that between the fifth and twelfth
discs from the bottom; then it remained approximately
constant up to the main interface. Therefore,
the measured hold-up is probably not equal to that
of the equilibrium drop size distribution.
This variation in hold-up along the column
made the quantitative analysis of the hydrodynamics
very difficult. In the procedure of Strand, et al .
(1962) as described in Appendix B, the one adjustable
parameter, G,g, is usually estimated from the
hold-up. For the n-butyl acetate-water system
Strand, et al . (1962) estimated Glg = 0.4 from hold-
up measurements. For several systems with water as
the continuous phase, but not including n-butyl
acetate-water, Olney (1964) found the ratio of
Sauter mean drop diameter to maximum stable drop
diameter was about 0.37, which when combined with
an estimated maximum stable drop diameter of 0.05
inch gives d = 0.02 inch. In Table 23 values of
hold-up, 0, Sauter mean drop diameter, d , inter-
facial area per unit volume, a, characteristic
velocity according to the method of Strand, et al .
(1962), V., and the two Peclet numbers, Pe and
176
-------
Pe., are listed for three values of G^g- G-,R =
0.0477 results from the experimentally determined
hold-up. G-jg = 0.2 gives a Sauter mean drop
diameter of about that estimated from the measured
maximum stable drop diameter. G,g = 0.4 corresponds
to the estimate of Strand, et al . (1962).
The data in Table 23 serve to show how sensi-
tive the predictions of the hydrodynamic character-
istics of an RDC are to the choice of G,g, and
how difficult it is to choose a value of G,g from
experimental data on a single experiment. The
correlation of Logsdail, et al. (1957) predicts
Vk = 128 ft/hr for this system; this fact along
with the measured drop size makes the value of
G,g = 0.2 appear reasonable. The importance of
the choice of G1Q on the prediction of N ,, is
i o ow
illustrated by the variation of a with G,g, if
one realizes that NQW is proportional to a.
Table 24 shows that the variation in G18 has
little effect on the evaluation of the experimental
NQW for each component. The experimental values
of N were calculated from the measured removal
0 W
efficiency, the values of E, and the two Peclet
numbers. Of the variables affecting NQW> only
Pec and Pe^ vary with G-jg, and the variation is not
large enough to affect the experimental N substan-
tially. The values of E and % removal based on the
analyses of feed and product water are included in
Table 24.
The experimental data for N calculated
O W
assuming GIS = 0.2 are plotted as 1/NQW vs. 1/E
177
-------
Table 23. Effect of G,g on Prediction of Hydrodynamic
Characteristics of RDC
G18
0.0477
0.20
0.40
Table
a
0.0428
0.0124
0.0073
24. Effect
dp VK a Pec
(inch) (ft/hr) (ft2/ft3)
0.005 55 606 8.41
0.021 149 42 8.19
0.043 242 12 8.15
of G,g on the Experimental
Ped
13.0
29.0
36.6
Estimates of N
ow
G18
0.0477
0.20
0.40
E
%
Methyl
Acetate
Now
0.35
0.34
0.34
0.41
20.3
Ethyl i-Propyl o-Cresol
Acetate Acetate
ow ow ow
2.7 3.4 4.4
2.5 3.2 4.3
2.5 3.2 4.3
1.25 3.80 23.0
69.8 87.4 95.5
Removal
178
-------
in Figure 27. The points for ethyl acetate, iso-
propyl acetate, and o-cresol fall within experi-
mental variation on a straight line, but the point
for methyl acetate is much above the line. The
removal efficiency for methyl acetate would have
to be 36% rather than 20% for that point to have
fallen on the line; this large a variation is much
more than what can be explained by experimental
variation. This observed low experimental value
of N for a component with E much less than 1
ow r
has been observed in several other experiments in
the RDC and may be the result of the failure of
the dispersion model to predict the concentration
changes for such solutes. Concentration profiles,
which were not measured in the present experiments,
would be required to verify this possibility.
Although concentration profiles over the entire
length of the column were not measured, after reaching
steady state one sample was taken of the continuous
aqueous phase from a point just below the main
interface. In terms of the dimensionless position
variable, Z, defined as the distance below the main
interface divided by the total length from the
main interface to the bottom disc, Z = 0.034 specifies
the position from which the sample was taken. The
measured solute concentrations were included in
Table 21; the solute concentrations predicted by
the dispersion model with G18 = 0.2 at the point
Z = 0.034 are compared to the measured concentrations
in Table 25.
The straight line through the data points for
the three solutes having E > 1 results in NW = 4.2
179
-------
3.0
2.8
2.6
2.4
2,2
2.0
1.8
1.6
o 1.4
§ 1.2
1.0
0.8
0.6
0.4
0.2
'I
0
I I
1
I I I I I I
0
0.4 0.8 1.2 1.6 2.0 2.4
I/E
Figure 27. Experimental Data for Run RS13
180
-------
Table 25. Comparison of Predicted and
Experimental Solute Concentrations
for Z = 0.034 and G,0 = 0.2
xcs
Solute Experimental Predicted % Difference
Concentration, Concentration,
ppm ppm
Methyl
Acetate 235 221 -6.0
Ethyl
Acetate 248 237 -4.4
i-Propyl
Acetate 509 515 1.2
o-Cresol 3349 3682 9.9
181
-------
and N = 5.0 for the 1 individual -phase numbers of
transfer units. In Tables 26 and 27 the values of
the N and N predicted by the various models for
mass transfer discussed in Chapter III are shown
as a function of Glg. If G]8 is assumed to be
approximately 0.2, then the model for turbulent
drops gives the best estimate for NW, and both the
models for circulating drops and for stagnant drops
give a value of N that is much too large. The
uncertainty in G,n prohibits one from making a clear
choice between models from these limited data.
The results of this single experiment have been
presented in detail to illustrate the successes
and failures of the presently available methods
for modeling an RDC. The ability to estimate the
individual-phase resistances to mass transfer
experimentally, which is possible by the collection
of mass transfer data on the simultaneous extraction
of several solutes, provides a severe test to the
present model. Clearly more experimental work
is needed to improve our ability to correlate
data from an RDC.
Experiments on Lube Oil Refining Waste Water
This waste water was the most thoroughly studied
sample of all the industrial waste water samples
tested in the RDC. The two types of dual solvent
processes (Section IV) were simulated by separate
extractions, first with pure n-butyl acetate or
with a mixture of n-butyl acetate and isobutylene.
Then the resulting purified water from the first
extraction was treated with pure isobutylene. The
182
-------
Table 26. Effect of G,g on the Prediction
of N for the RDC
w
Methyl Ethyl i-Propyl
Acetate Acetate Acetate
G18 Nw Nw Nw
Stagnant Drops - Equation (B22)
0.0477 5.9 5.9 5.9 5.9
0.20 1.1 1.1 1-1 LI
0.40 0.55 0.55 0.55 0.55
Circulating Drops - Equation (B24)
0.0477 280. 260. 250. 250.
0.20 16. 15. 14. 14.
0.40 4.2 3.9 3.7 3.7
Turbulent Drops - Equation (B26)
0.0477 34. 31. 29. 29.
0.20 2.4 2.1 2.0 2.0
0.40 0.69 0.63 0.58 0.58
183
-------
Table 27. Effect of G18 on the Prediction
of N for the RDC
s
Methyl Ethyl i-Propyl n
Acetate Acetate Acetate
G18 Ns Ns Ns Ns
Stagnent Drops - Equation (B23)
0.0477 890 830 780 770
0.2 15 14 13 13
0.4 2.2 2.0 1.9 1.9
Circulating Drops - Equation (B25)
0.0477
0.2
0.4
990.
34.
12.
930.
33.
11.
880.
32.
11.
870.
32.
11.
184
-------
initial tests of the RDC (Runs RS1A, RS1B, and RS2)
are also discussed in this section since the prepared
feed waters resemble the lube oil refining waste
water. The analysis of mass transfer characteristics
for all runs dealing with this waste water are
discussed together at the end of this section.
In Run RS2 n-butyl acetate was used to treat
a feed water prepared to contain methyl ethyl ketone
(MEK, measured Kd = 4.56), diethyl ketone (DEK, measured
Kd = 16.2) and phenol (Kd = 57.0, Appendix F) while
operating at a low flow ratio (Fg/Fw = 0.120). The
settings of independent variables and the results
of the extraction are shown in Table 28. This
experiment was conducted prior to the installation
of the level gauge used to measure solvent hold-up
and prior to the installation of the sampling port
for removing samples from within the column. Also,
the solvent feed was not directed through the annular
space between the rotating shaft and the bottom
stationary tube, but it was added through a tube which
extended to about 2 inches below the bottom disc.
The material balance for this run showed a
loss of 12.5% of the inlet MEK, a loss of 5.7%
of the inlet DEK, and a gain of 2.5% of the inlet
phenol. The concentration of n-butyl acetate in
the product water was also unexpectedly low compared
to its solubility. These results led to several
tests which showed that the paper seals in
the sample vials could absorb a considerable amount
of the volatile organic solutes from the aqueous
samples (see discussion in Section VI). Since
the sample of feed water was taken into a glass
stoppered bottle but the product water samples
185
-------
Table 28. Results from Run RS2
Water flow rate = 5.16 gal/hr
Solvent flow rate = 0.701 gal/hr
Rotating disc diameter = 1.50 inch
Stator hole diameter = 2.25 inch
Rotational speed = 617 RPM
Analytical results (concentrations in ppm):
MEK DEK Phenol
Feed water 2213. 4314. 6143.
Product wate^r 1135. 1238. 1228.
Loaded solvent 6508. 22860. 40910.
Percent removal 36.2 65.6 82.5
n-Butyl acetate in product water = 4986 ppm.
E 0.55 1.94 6.82
Now 0.88 1.6 2.3
Measured column temperature = 22.0 °C
G, „ assumed to be 0.2 giving Pe =8.53
IB W
and Peo =28.1
s
186
-------
were taken into the 5 cc sample vials, absorption
by the seal would account for the MEK and DEK losses
and for the low concentration of n-butyl acetate.
Due to the errors in the product water analyses,
the removal efficiencies listed in Table 28 were
calculated from the feed water analysis and the
loaded solvent analysis by assuming that the material
balance was exact. Using the inlet solvent and
water flows, the values of E were determined for
each solute. Assuming GIQ = 0.2 as found to be
reasonable for Run RS13 when n-butyl acetate was
used as solvent, the values of Pew and Pe were
W b
estimated. This allowed the values of NQW to be
calculated for each solute and allowed a plot of
1/N vs. 1/E to be made. The three points fell
almost exactly on a straight line resulting in NW =
2.5 and N = 2.4 as experimental estimates. The
values of NW and NS were about half as large as
found in Run RS13; this is probably due to using
fewer discs, having poorer solvent dispersion,
and using a lower rotational speed in Run RS2 than
in RSI 3.
In Runs RS1A and RS1B isobutylene was used
to treat a prepared feed water which contained
MEK (Kd = 2.49), DEK (Kd = 13.4), and n-butyl acetate
(Kd = 168) (see Appendix E) while operating at
two settings of the solvent flow rate (F/F.. = 0.354
5 W
and 0.117). The setting of the independent variables
and the results of the extraction are shown in Table 29.
The RDC was set up 'exactly as in Run RS2 just previously
described. As with the spray column experiments, the
material balance could not be checked because the
187
-------
Table 29. Results from Runs RS1A and RSIB
Water flow rate =2.28 gal/hr
Rotating disc diameter = 1.50 inch
Stator hole diameter = 2.25 inch
Rotational speed = 1430 RPM
Measured column temperature = 22.0 °C
Run RSIA Solvent flow rate = 1.36 gal/hr
Analytical results (concentrations in ppm):
Feed water
Product water
Percent removal
E
Run RSIB: Solvent
2109.
639.
69.7
0.88
flow rate
Analytical results (in
Feed water
Product water
Percent removal
E
2109.
1340.
36.5
0.29
4297.
183.
95.7
4.74
= 0.450
ppm) :
4297.
888.
79.3
1.57
4393.
77.3
98.2
59.4
gal/hr
4393.
288.
93.4
19.7
188
-------
volatile solvent sampling procedure had not been
perfected.
It is difficult to interpret this experiment
using the dispersion model. Subsequent experiments
described below which use isobutylene as a solvent
indicate that a reasonable range of values for G^g
is from 0.15 to 0.30. Values of Glg in this range
lead to Pe =1.9 which is characteristic of ex-
w
tensive axial mixing in the continuous phase.
Using this value for Pe,, and the corresponding
W
prediction for Pe between 24 and 29, the dispersion
model predicts that the observed removal efficiencies
for MEK in Runs RS1A and RS1B are not possible
even for infinite NQW. The model also predicts
large values of N01I for DEK (11 for RS1A and 7
ow
for RS2A) and for n-butyl acetate (10 for RS1A and
6 for RS2A). These values are not only larger
than expected, but they also decrease with increasing
E which violates the additivity of resistances concept
No explanation for these observations is available.
Although the quantitative explanation for
these results is unknown, visual observation agreed
with predicted values of Pew as considerable mixing
of the continuous phase was apparent. This obser-
vation led to the fabrication of a second set of
discs and stator plates having a disc diameter of
1.75 inch (1.50 inch for initial discs) and a
stator hole diameter of 2.00 inch (2.25 inch for
initial stator holes). For all subsequent tests
using volatile solvents, this second set of discs
and stator plates was used; this practice led to
-------
much larger values of Pe and better removal effi-
ciencies. These new dimensions are also within
typical ranges of commercial RDC extractors.
In Runs RS3 and RS4 a waste water which was
prepared to simulate the lube oil waste was extracted
first with n-butyl acetate (RS3) and then with
isobutylene (RS4). The feed water to Run RS3 which
contained acetone (measured Kd = 1.05), MEK (measured K^ :
4.56), phenol (Kd = 57.0, Appendix F), benzene
(measured Kd = 61.5), and o-cresol (Krf = 206,
Appendix F) was treated at a low flow ratio of n-butyl
acetate to water (FS/FW = 0.097). The settings of
independent variables and the results of the extraction
are shown in Table 30. The steady state material balance
closed within less than 6% for each of the five
components. The total solvent-phase flow rate
increased by almost 9% because of the addition of
solutes, while the flow of n-butyl acetate in
the solvent decreased by about 6% due to solubility
losses in the water. Therefore, the analysis was
calculated on a solute-free flow basis using the
inlet solvent flow rate.
The analysis of this run showed the same
conflict in choosing the best value of GIQ as was
encountered in Test Run RS13. Assuming as before
that G18 = 0.2, the method of Strand, et al . (1962)
was used to estimate Pe and Pe from which the
W 5
values of N listed in Table 31 were calculated
0 W
using the experimental removal efficiencies. The
ratio of solute-free flows was 0.0992 Ib. solvent/lb.
water from which the values of E were determined.
The percent removal values listed in Table 31 were
190
-------
Table 30. Results from Run RS3
Water flow rate
Solvent flow rate
Rotating disc diameter
Stator hole diameter
Rotational Speed
5.65 gal/hr
0.626 gal/hr.
1.50 inch
2.25 inch
805 RPM
Analytical results (concentrations in ppm)
Acetone MEK Phenol Benzene
13300. 169.
308. 30.5
97.7 82.0
6820. 19.0
Feed
water 38.0 217.
Product
water 34.3 126.
Percent
removal 9.7 41.9
Water in
column 32.5 181.
Loaded
solvent 44.9 974. 115000.
1292.
o-Cresol
2107.
25.4
98.8
649.
20300.
n-Butyl acetate in product water = 6110 ppm.
Measured solvent hold-up = 0.0630
Measured column temperature = 22.0°C
Estimated maximum drop diameter = 0.05 inch
191
-------
Table 31. Experimental Estimates of
NQw for Run RS3
Acetone MEK Phenol Benzene o-Cresol
E 0.10 0.45 5.65 6.10 20.4
% Removal 9.7 41.9 97.7 82.0 98.8
N 0.4 2.5 6.6 2.2 6.7
ow
G,g assumed to be 0.2; Pe = 9.55
and Pee =29.2
s
192
-------
calculated from the solute-to-water weight ratios
in the aqueous phase. Using Glg = 0.2 led to a
calculated solvent hold-up of 0.0126 which can be
compared to 0 = 0.0630, the measured value.
When the data from RS3 were plotted as 1/N
O W
vs 1/E, the points for phenol and o-cresol essentially
coincided. If a straight line were drawn through
this point and through the point for MEK, N = 7.3
w
and N = 8.4 were determined as the individual-
phase numbers of transfer units. The point for
acetone fell above the line through the points for
phenol, o-cresol, and MEK much as the point for
methyl acetate did in Run RS13 (Figure 27), but
in Run RS3 the removal efficiency for acetone would
only need to be changed from 9.7% to 10.3% to make
the point for acetone fall on the line. This
difference is within experimental uncertainty. The
point for benzene which also fell above the straight
line cannot be explained as experimental uncertainty.
A low removal efficiency for benzene when benzene
was present with this combination of solutes has
been observed in several experimental runs including
SS6 in the spray column. Chromatographic analysis
of the feed solvent showed that it was free of benzene.
In Run RS4 the collected water which had been
treated by n-butyl acetate extraction in RS3 was
treated with isobutylene at a low solvent-to-water
flow ratio (F /F = 0.100). The major solute was
s w
n-butyl acetate (Kd = 168, Appendix E) with smaller
amounts of acetone (Kd = 0.63, Appendix E), phenol
(Kd = 0.7, Appendix E), MEK (Kd = 2.49, Appendix E),
o-cresol (Kd = 4.8, Appendix E), and benzene
193
-------
(measured Krf = 407). The settings of independent
variables and the results of the treatment are shown
in Table 32. The dramatic improvement in column
performance due to changing to larger discs and to
stators with smaller holes is evident by comparing
n-butyl acetate removal efficiency in RS1B (93.4%)
with n-butyl acetate removal efficiency in RS4 (99.
The material balance closures for each solute were
within 15%; the main error was still believed to be
in the analyses of organic-phase samples containing
the volatile solvent.
Using the method of Strand, et al. (1962),
G,g = 0.337 was estimated from the measured solvent
hold-up. The variation of solvent hold-up along the
length of the RDC appeared to be much less pronounced
in this run than in runs using n-butyl acetate as
solvent. With the larger disc and smaller stator
holes the restriction of free flow area due to the
mid-column bearing was less severe (about 24% free
area compared to C^ = 0.25); this was probably the
major reason for a more uniform hold-up. The value
of G,o calculated from the average hold-up resulted
in a calculated average droplet size which agreed
with visual estimation.
Values of Nftll for each solute were estimated
o w
using Glg = 0.337 to estimate Pew = 4.92 and Pe = 32.0,
using Kd for pure isobutylene as solvent, and using
experimental removal efficiencies. Benzene removal
was again much lower than expected based on K..
d
N = 14.9 was estimated for n-butyl acetate (E = 16.8).
0 W
The dispersion model predicted that the observed
removal efficiencies for acetone, MEK, phenol and
194
-------
Table 32. Results from Run RS4
Water flow rate = 4.66 gal/hr
Solvent flow rate = 0.790 gal/hr
Rotating disc diameter =1.75 inch
Stator hole diameter = 2.00 inch
Rotational speed = 1430 RPM
Analytical results (concentrations in ppm):
Acetone Phenol MEK o-Cres. n-BuAc Benzene
Feed
water 29.9 605. 124. 72.8 5457. 68.5
Product
water 28.2 522. 83.3 17.0 10.8 60.4
Percent
removal 5.7 13.7 22.8 76.6 99.8 11.8
Water in
column 29.4 655. 114. 51.0 1324. 63.0
Measured solvent hold-up = 0.0066
Measured column temperature = 26.8°C
195
-------
o-cresol should not have been possible even with
an infinite value of NQW. The explanation for these
observations requires the consideration of interactions
between solutes.
Measurements and correlations were made of the distri
bution coefficient for phenol distributing between
water and mixtures of isobutylene and n-butyl acetate.
Equation (7) provides an approximation to this more
accurate but more complex correlation. In Run RS4
the feed water contained 5457 ppm of n-butyl acetate;
a material balance showed that the loaded solvent
contained 5.44% n-butyl acetate. The distribution
coefficient for phenol thus increased from about 0.7
at the bottom of the RDC to about 3.8 at the top of
the RDC. The numerical method described in Appendix
C was used to quantitatively estimate the effect
of n-butyl acetate on the expected removal of phenol
and o-cresol.
To make this calculation we first assumed that
the dispersion model with constant K^ gave a good
representation of the concentration profiles for
n-butyl acetate in each bulk phase. We then assumed
that N = N for n-butyl acetate (as has been found
W 5
approximately true in several other experiments in
the RDC) which, when combined with the experimental
value of N , and the additivity of resistances
o w
relationship (equation 4), allowed N (and N ) to
w s
be evaluated. These assumptions allowed the inter-
facial concentration of n-butyl acetate in the
isobutylene phase to be calculated throughout the
column, and thus the value of Kd for phenol was cal-
culated. Kd was estimated to vary from 0.71 at the
196
-------
solvent inlet to 3.6 at the solvent outlet. We
finally assume that NW and Ng for phenol are equal
to N and N determined for n-butyl acetate. Then
w s
using K- for phenol as a function of position in the
column, the numerical method discussed in Appendix
C was used to calculate the concentration profile
for phenol. This method led to the estimation that
23.6% of the phenol would be removed, which can
be compared to the experimentally determined value
of 13.7% removal. The dispersion model using Kd
for phenol distributing between water and pure iso-
butylene predicted only 7.0% removal. Even for a
case where values of K^ are thought to be nearly
constant, it can be noted by observing Figure 27
that the dispersion model tends to overestimate the
removal efficiency for solutes having a value of E
below 1 when the assumption of additivity of resis-
tances is used with N and N determined from solutes
having values of E greater than 1.
A similar calculation was made for o-cresol
by assuming equation (7) would provide a reasonable
estimate of the effect of solvent composition on K^
for o-cresol. In this case there is no experimental
evidence that equation (7) will be accurate. The
numerical method of calculation led to the estimation
that 69.1% of the o-cresol would be removed, which
can be compared to the experimentally determined
value of 76.6% removal and to the value using Kd = 4.8
of 45.3% removal. Similar calculations could
have been made for acetone and MEK, but the variation
of Kd with mixed solvent composition for these
solutes was unknown.
197
-------
For the purpose of demonstrating the effective-
ness of the dual-solvent process, the results from
Runs RS3 and RS4 can be combined to calculate an
overall removal efficiency. By assuming that the
measured removal efficiencies would be duplicated
in a two-step process, the concentrations of solutes
in the water phase after the n-butyl acetate extraction
and after both extractions were calculated as shown
in Table 33. The calculated concentrations after
both extractions differ from the concentrations
measured in Run RS4 since it was assumed that the
water feeding the second step contained the steady
state concentrations from the first step. Clearly
a substantial improvement in water quality is
possible by dual-solvent extraction.
In Runs RS6 and RS7 a sample of the lube oil
refining waste water supplied from an industrial
source was treated first by extraction using n-butyl
acetate (RS6), and then by extraction using isobutylene
(RS7) to recover the dissolved n-butyl acetate. As
discussed in Section V, this sample differed
appreciably from the typical composition of this
source of waste water and from the simulated lube
oil refining waste water treated in Runs RS3 and RS4.
In this sample MEK was present at about 100 times
its normal concentration; acetone and benzene were
absent; phenol and o-cresol were present at about
one-half their normal concentrations; and the suspended
solid was milky grey in color rather than black as
is normal. The presence of the cloudy suspended
phase caused a problem in setting the rotational
speed of the discs, because it was difficult to
198
-------
Table 33. Calculated Effectiveness of
the Dual Solvent Process
First extraction with n-butyl acetate at F /F = 0.097
S VV
Second extraction with isobutylene at F /F = 0.100
5 Vr
Feed After
Solute
Acetone
MEK
Phenol
Benzene
o-Cresol
n-Butyl
Acetate
Cone. First Extn.
38.0
217.0
13300.0
169.0
2107.0
0.0
34.3
126.0
308.0
30.5
25.4
6110.0
After Overall
Second Ext. % Removal
32.3
97.3
266.0
26.9
5.9
12.1
15.0
55.2
98.0
84.1
99.7
199
-------
determine at what point about 1% of the solvent
appeared to be entrained in the water phase.
The feed water to Run RS6 which was treated
with n-butyl acetate contained MEK (measured K.
= 4.56), phenol (Kd = 57.0, Appendix F), and o-
cresol (Kd = 206, Appendix F). Because of the
very high MEK concentration, two solvent flow rates
were used. In Run RS6A the solvent-to-water flow
ratio was about 0.1 to simulate the typical value
of F /F.. normally used when treating this type
o W
of waste water. In Run RS6B FS/FW was increased
to about 0.3 to provide information on how efficiently
MEK could be removed during the unusual condition
which caused the atypical waste water composition.
The settings of the independent variables and the
results of the extractions are shown in Table 34
for Run RS6A and in Table 35 for Run RS6B. The
measured solvent hold-up values were very low in
both these runs for an undetermined reason.
The steady-state material balance closed within
less than 6% for each solute in Run RS6B, but in
Run RS6A the material balance showed as much as
26% more solute entering with the feed water than
leaving with the product water and loaded solvent.
Inspection of the solvent-phase samples taken during
the approach to steady state showed that the loaded
solvent composition was still changing at the end of
the experiment. Run RS6A was made following Run
RS6B, and the usual practice of raising the main
interface to the top of the column and then lowering
it to the normal position was not followed. This
meant that the solvent phase in the column above the
200
-------
Table 34. Results from Run RS6A
Water flow rate
Solvent flow rate
Rotating disc diameter
Stator hole diameter
Rotational speed
3.21 gal/hr
0.369 gal/hr
1.50 inch
2.25 inch
1100 RPM
Analytical results (concentrations in ppm):
MEK
Phenol
o-Cresol
12220.
5883.
51.8
48400.
8751.
104.
98.8
66900.
892.
6.5
99.3
6600.
Feed water
Product water
Percent removal
Loaded solvent
n-Butyl acetate in product water = 15210 ppm,
Solvent hold-up too small to measure.
Measured column temperature = 24.2°C
201
-------
Table 35. Results from Run RS6B
Water flow rate = 3.21 gal/hr
Solvent flow rate = 1.11 gal/hr
Rotating disc diameter = 1.50 inch
Stator hole diameter = 2.25 inch
Rotational speed = 1100 RPM
Analytical results (concentrations in ppm):
MEK Phenol o-Cresol
Feed water 12220. 8751. 892.
Product water 2452. 77. 4.3
Percent removal 82.3 99.1 99.5
Loaded solvent 30600. 26200. 2880.
n-Butyl Acetate in product water = 15400 ppm.
Measured solvent hold-up = 0.00214
Measured column temperature = 23.4°C
202
-------
interface had. to be purged by the slow solvent flow
during Run RS6A. Incomplete purging accounts for the
lack of closure of the material balance for this
run. The flow rate of the total solvent phase
increased by only 1.9% in Run RS6A and 2.6% in Run
RS6B as it passed through the column because the
amount of solutes picked up almost exactly balanced
the amount of solvent dissolved and entrained in the
aqueous phase. Therefore, the analysis was calculated
on a weight fraction basis using total solvent- and
water-phase inlet flow rates.
The difficulty in experimentally establishing
the setting for the disc rotation speed because of
the turbid nature of the water-phase was previously
mentioned. The chosen setting of 1100 RPM was
probably higher than optimal, and because of setting
the rotation speed too high a considerable amount
of n-butyl acetate phase was entrained in the product
water. This fact is shown by the high concentration
of n-butyl acetate (> 15,000 ppm) in the product
water, which is more than twice the solubility of
the solvent in water. After settling for 12 hours,
the concentration of n-butyl acetate had fallen to
about 10,000 ppm, and after settling for 1 week,
the concentration had fallen to about 7,000 ppm. Even
after settling for a week the turbidity of the
product water was about the same as that of the feed
water.
Assuming that Glg = 0.2 as in previous experi-
ments with n-butyl acetate as solvent, the method
of Strand, et al . (1962) was used to estimate Pew
and Pe$, from which the values of NQW listed in
203
-------
Tables 36 and 37 were determined. The reason why
the removal of MEK in Run RS6A was higher than what
theory predicted as possible is not known, but it
may be associated with the high concentration of MEK
in both phases. The MEK removal would have had
to be 42.9% rather than the measured 51.8% to give
N = 5.5, as expected from the results on Run RS6B.
ow
By assuming G,g = 0.2 the solvent hold-up was
predicted as 0.008B for RS6A and 0.0272 for RS6B,
which are much larger than the measured values
listed in Tables 34 and 35. The maximum stable
drop diameter predicted with G^g = 0.2 was about
equal to the observed value.
When the data from Run RS6B were evaluated
in terms of a straight line plot of 1/N_,. vs 1/E,
0 W
N = 11 and N = 7.7 were estimated as the individual
w S
phase numbers of transfer units. From the data
from Run RS6A N was estimated to be about 11, but
W
N could not be estimated.
In Run RSZ the collected water which had been
treated by n-butyl acetate extraction in RS6 was
treated by extraction with isobutylene at a low
solvent-to-water flow ratio (F_/Flf = 0.102). The
S W
major solutes were n-butyl acetate (Kd = 168
Appendix E) and MEK (Krf = 2.49, Appendix E) with
smaller smounts of phenol (Kd = 0.7, Appendix E)
and o-cresol (Kd = 4.8, Appendix E). The water
treated by n-butyl acetate extraction in Run RS6A
was treated with isobutylene in Run RS7A, and the
water treated in RS6B was treated with isobutylene
204
-------
Table 36. Experimental Estimates of N
for Run RS6A
E
% Removal
N
MEK
0.46
51.8
oo
Phenol
5.76
98.8
12.5
o-Cresol
20.8
99.3
10.7
ow
assumed to be 0.2; Pew = 4.26
and Pe_ =22.1
s
E
%
N
Table 37. Experimental
for Run RS6B
MEK
1.39
Removal 79.9
5.5
Estimates
Phenol
17.3
99.1
10.2
of N
ow
o-Cresol
62.7
99.5
11.1
assumed to be 0.2; Pew = 4.36
and Pee =21.9
S
205
-------
in RS7B. The settings of independent variables and
the results of the treatment are shown in Tables
38 and 39. The removal efficiencies were very
similar in the two runs with a slightly better
result in RS7A. The results were also very similar
to those in Run RS4 (Table 32) where a simulated
waste water was used. The solvent hold-up measured
in these two experiments was about 5 times as large
as that measured in RS4. The material balance
closures were within 12% in both runs.
When this waste water was treated by n-butyl
acetate extraction in RS6, no appreciable improvement
in turbidity occurred. However, during the extrac-
tions using isobutylene in RS7, the milky grey sus-
pended phase was greatly reduced. The suspended
phase may have been particles of wax which were more
soluble in or possibly more readily coalesced with
the isobutylene droplets than with n-butyl acetate
droplets.
When the measured solvent hold-up was used to
estimate G-,g and the droplet diameter, very low
values for GIS (0.024 - 0.026) and d (0.0022 - 0.0024
inch) were calculated. With G^Q = 0.337, as was
found with the simulated water in RS4, values of d
agreed with those visually estimated, but the cal-
culated hold-up was much below the measured values.
Values of N were calculated for each solute
using GIQ = 0.337, which gave Pew = 3.55 and Pes = 32.5,
and using K, for each solute as though the solvent
were pure isobutylene. As expected this calculation
predicted that only n-butyl acetate could possibly
have been removed to the extent measured. This
206
-------
Table 38. Results from Run RS7A
Water flow rate = 3.21 gal/hr
Solvent flow rate = 0.553 gal/hr
Rotating disc diameter = 1.75 inch
Stator hole diameter = 2.00 inch
Rotational speed = 1450 RPM
Analytical results (concentrations in ppm):
Phenol MEK o-Cresol n-Butyl
Acetate
Peed
water
Product
water
Percent
removal
306.
227.
25.8
5573.
3597.
35.5
24.2
2.3
90.5
7133.
11.0
99.8
Measured solvent hold-up = 0.0362
Measured column temperature = 23.3»C
207
-------
Table 39. Results from Run RS7B
Water flow rate = 3.21 gal/hr
Solvent flow rate = 0.553 gal/hr
Rotating disc diameter = 1.75 inch
Stator hole diameter = 2.00 inch
Rotational speed = 1450 RPM
Analytical results (concentrations in ppm):
Phenol MEK o-Cresol n-Butyl
Acetate
Feed
water
Product
water
Percent
removal
229.
190.
17.0
2801.
1891.
32.5
18.0
2.8
84.4
6791.
15.
99.
2
8
Measured solvent hold-up = 0.0346
Measured column temperature = 23.3°C
208
-------
calculation gave NQW = 19.1 in Run RS7A and NQW = 17.3
in Run RS7B for n-butyl acetate. The numerical
calculation described for Run RS4 and in Appendix
C was used to estimate the removal efficiencies of
phenol and o-cresol predicted from the effect that
n-butyl acetate in the isobutylene phase has on the
values of K. for these two solutes. This calculation,
d
made assuming N = NW for n-butyl acetate, led to
the prediction that 27.4% of the phenol and 72.6% of
the o-cresol should have been removed. These results
are in reasonably good agreement with the experimental
values.
The chemical oxygen demand (COD) was determined
for the initial waste water used as feed in RS6 and
for the product solutions from RS7A and RS7B. In
Table 40 these results are compared to values of
theoretical oxygen demand (TOD) calculated from
the concentrations of MEK, phenol, o-cresol, and
n-butyl acetate. The agreement between COD and
TOD is remarkably good.
As was done previously in Runs RS3 and RS4, the
effectiveness of the dual solvent process was
estimated by calculating overall removal efficiencies.
The results for the combination RS6A and RS7A and
for the combination RS6B and RS7B are shown in Table
41. In these calculations the n-butyl acetate which
was present as a dispersed phase after the first
extraction was assumed to be completely removed with-
out affecting the removal of dissolved solutes in
the second extraction. This assumption was necessary
since no time would be provided for phase separation
between the two steps. The presence of a high
209
-------
Table 40. COD and TOD for Runs RS6 and RS7
COD TOD
(ppm) (ppm)
Feed to RS6 54,450 52,900
Product from RS7A 8,572 9,350
Product from RS7B 4,692 5,110
210
-------
Table 41. Overall Removals for the
Dual Solvent Process
First extraction with n-butyl acetate.
Second extraction with isobutylene.
Feed After After Overall
Cone. First Extn. Second Extn. % Removal
Runs RS6A and RS7A:
(Fg/I
MEK 12220.
Phenol 8751.
o-Cresol 892.
n-Butyl
Acetate 0.
TOD 52900.
?v = o.ioi)
5883.
104.
6.5
15210.
48150.
(Fg/Fw = 0.102)
3795.
77.2
0.6
11.0
9470.
68.9
99.1
99.9
82.1
guns RS6B and RS7B:
(Fg/Fw = 0.304)
(Fg/Fw = 0.102)
MEK 12220.
Phenol 8751.
o-Cresol 892.
n-Butyl
Acetate 0.
TOD 52900.
2452.
77.0
4.3
15400.
40120.
1655.
63.9
0.7
15.2
4230.
86.5
99.3
99.9
--
92.0
211
-------
TOD as well as the cost of lost n-butyl acetate can
be seen to provide the need for a second extraction
with the volatile solvent. The improvement in
final product by operating the first extraction with
a higher solvent flow rate is also illustrated by
comparing the results for the two combinations of runs
In Runs RS8 and RS9 a mixture of about 50 weight
percent n-butyl acetate and 50 weight percent
isobutylene was used to treat samples of lube oil
refining waste water. These experiments provide
information relating to a dual solvent process
with linked solvent cycles. The second step in
this dual solvent process utilizing isobutylene
extraction for n-butyl acetate recovery is similar to
isobutylene extraction in a dual solvent process with
separate solvent cycles. Therefore, the data from
Runs RS4 and RS7 apply directly, and the second step
in the process with linked solvent cycles was not
simulated.
In Run RS8 a mixture of 48.7 weight % n-butyl
acetate and 51.3 weight % isobutylene was used to
treat a sample of industrial lube oil refining
waste water. If we assume that the distribution
coefficient for each solute between water and the
solvent mixture was given by equation (7) as was
found to be valid experimentally for phenol, then
an approximate value for Kd can be determined for
each solute. The solutes which were present included
acetone (ICd = 0.83), MEK (Kd = 3.50), phenol (Kd
= 28.1), o-cresol (Kd = 100), and benzene (Kd
= 239).
212
-------
It was assumed that the physical properties
of the mixed solvent would be intermediate between the
properties of the two pure solvents. Therefore, the
column was set up with the smaller set of discs
(1.50 inch) which had been used for n-butyl acetate
extractions and with the stators having the smaller
holes (2.00 inch) which had been used for isobutylene
extraction. The initial flow settings were chosen
so that FS/FW was about 0.2; a dual solvent process
with linked solvent cycles with F/F., = 0.2 in the
s w
initial extraction step (50 weight % of each solvent)
and with F/F = 0.1 in the second extraction step
(pure volatile solvent) should be comparable to a
dual solvent process with separate solvent cycles
with FS/FW = 0.1 in each pure solvent extraction
step. The settings of the independent variables
and the results of this extraction are shown in
Table 42.
A problem which was encountered due to flow
line plugging during this run resulted in a gradual
decrease in both the solvent and the water flow rates.
For this reason a range of flow rates is given in
Table 42; the higher number is the initial setting,
and the lower number is the final setting. These
decreasing flows made calculation of the steady
state material balance inaccurate and caused the
solvent hold-up measurement to be invalid. However,
the measured removal efficiencies are believed
to be a good approximation of the efficiencies which
would have been measured at the median of the ranges of
flow rates. The values of E reported were calculated
at median flows. The COD of the feed and product
water streams was not determined.
213
-------
Table 42. Results from Run RS8
Water flow rate
Solvent flow rate
Solvent feed composition
Rotating disc diameter
Stator hole diameter
Rotational speed
3.92 - 4.16 gal/hr
1.055 - 1.200 gal/hr
48.7 weight % n-butyl
acetate in isobutylene
1.50 inch
2.00 inch
1090 RPM
Analytical results (concentrations in ppm):
Acetone
MEK
Phenol o-Cresol Benzene
Feed
water
Product
water
Percent
removal
E
24.
12.
50.
0.
6
1
8
17
108.
54.
49.
0.
9
2
72
17170.
1902.
88.9
5.76
2660.
124.
95.3
20.5
36.
9.
75.
49.
9
2
1
0
n-Butyl acetate in product water = 2364
ppm.
Measured column temperature = 21.2°C
Estimated maximum drop diameter =0.06 inch
214
-------
The comparison of results from Run RS8 with
the prediction from the method of Strand, et al. (1962)
was based on choosing G,Q from the observed maximum
stable drop size and from the observed hold-up
measured for the same system in Run RS9 (reported
below). Using G,Q = 0.17 as the chosen value resulted
in the values of Pe , Pe , and N for each solute
W 5 U W
shown in Table 43. The infinite value of N for
U W
acetone is undoubtedly the result of a large analyt-
ical error either for the feed or the product
water stream. For a reasonable value of ft = 0.5,
O W
only 15.2% removal would be expected. The values
of N«,, for MEK» phenol, and o-cresol fell on a
0 W
straight line when plotted as 1/N vs. 1/E
resulting in N,, = 4.4 and N = 3.3 as estimates
W 5
for the individual-phase numbers of transfer units.
As in previous extractions, the removal of benzene
was much less efficient than predicted from its
high value of E.
The appearance of the column during Run RS8
was similar to the appearance in Runs RS1A and
RS1B. In those previously described runs, we
found that the column operation was greatly improved
by converting to larger discs and to stators with
smaller holes. This modification in column geometry
was more successful in improving the contact between
solvent and water than could be achieved by simply
increasing the rotational speed. Apparently the
hydrodynamics of the mixed solvent system is more
nearly like extraction with pure isobutylene than
extraction with pure n-butyl acetate. Therefore,
the column might operate better using the 1.75-inch
discs and the stators with 2.00-inch holes.
215
-------
Table 43. Experimental Estimates of N for Run RS8
c ow
Acetone MEK Phenol o-Cresol Benzene
E 0.17 0.72 5.76 20.5 49.0
% Removal 50.8 49.2 88.9 95.3 75.1
N « 1.5 3.4 4.6 1.7
ow
F,0 assumed to be 0.17; PetT = 6.59 and Pe =33.8
18 w s
216
-------
In Run RS9 a mixture of 53.1 weight % n-butyl
acetate and 46.9 weight % isobutylene was used to
treat a prepared water stream which contained MEK
and phenol. The larger discs were used to determine
if the operation would be improved over that in Run
RS8, and the appearance of the column as well as
the rates of mass transfer showed that improvement
was indeed achieved. The droplets of mixed solvent
were smaller, and the hold-up was larger when using
the larger discs. However, no increase in the amount
of solvent entrained with the product was noted with
the larger discs.
The settings of the independent variables and
the results of the extraction with mixed solvent in
Run RS9 are shown in Table 44. Fe/F was set at
5 W
about 0.2 as in Run RS8. If we again assume that
the distribution coefficient for each solute between
water and the mixed solvent was given by equation
(7), then Kd = 3.59 for MEK and Kd = 30.6 for
phenol. The steady state material balance closed
within 8% for both solutes.
In the comparison of the results from Run RS9
with the prediction from the method of Strand, et al.
(1962), G,g calculated from the measured solvent
hold-up agreed with G,g estimated from the measured
maximum droplet diameter. Using this G,Q - 0.17
resulted in Pe = 6.54 and Pec = 25.9 and in N = 12.2
W o U W
for phenol. The observed 78.5% removal for MEK was
higher than would be predicted even with an infinite
value of NQW; this probably reflects a failing in
the method for estimating Kd for the mixed solvent.
217
-------
Table 44. Results from Run RS9
Water flow rate =4.66 gal/hr
Solvent flow rate = 1.364 gal/hr
Solvent feed composition =53.6 weight % n-butyl
acetate in isobutylene
Rotating disc diameter = 1.75 inch
Stator hole diameter = 2.00 inch
Rotational speed = 1090 RPM
Analytical results (concentrations in ppm):
MEK
424.
91.3
75.5
323.
0.78
Phenol
17320.
93.4
99.5
8762.
6.68
Feed water
Product water
Percent removal
Water in column
E
n-Butyl acetate in product water = 2593 ppm.
Measured solvent hold-up = 0.0238
Measured column temperature = 22.3°C
Estimated maximum drop diameter = 0.03 inch
218
-------
For a reasonable NAUJ = 6 with E = 0.78, 66.0%
o w
removal of MEK would be expected. The observed 99.5%
removal for phenol was the highest value measured for
the removal of phenol in any experiment reported for
this water.
Summary of Experiments on Lube Oil Refining Waste Water.
In Runs RS2, RS3, and RS6 n-butyl acetate was
used to treat simulated and actual samples of lube
oil refining waste water. The operation of the RDC
was similar in these runs except that the rotational
speed was higher in RS6, which resulted in some sol-
vent entrainment. The method of Strand, et al. (1962)
gave an adequate correlation of observed drop
sizes when G,g = 0.2 was used. By using this
monodisperse drop size model, the correlations dis-
cussed in Section V for mass transfer toward
stagnant drops, circulating drops, and drops in
a turbulent continuous phase (Calderbank and Moo-Young,
1961) were used to calculate values of N . These
W
theoretical estimates of N are compared in Table 45
W
to the experimental values which were determined by
plotting l/Nnil vs 1/E. The ratio of N,, predicted
0 W W
by the model for drops in a turbulent continuous
phase to the experimental value of N varied from
0.27 to 1.0, with 0.5 as found in Test Run RS13
being an average value. These are not sufficient
data for a clear choice of this model over those
for stagnant or circulating drops, but for the
purpose of estimating costs on a larger scale they
should be useful. Theoretical and experimental
estimates of NS are also listed in Table 45. The
ratio of N predicted for stagnant drops to N
219
-------
Table 45. Theoretical and Experimental Mass Transfer
Estimates for Extraction with n-Butyl Acetate
Run number RS2 RS3
N (RPM) 617 805
Pe 8.53 9.55
w
Pe 27.8 29.2
s
Experimental NW 2.5 7.3
N for turbulent
w transfer 1.1 2.0
N for stagnant
w drops 0.8 1.0
N for circulating
w drops 8.1 14.0
Experimental N 2.4 8.4
N for stagnant
s drops 5.6 15.0
N for circulating
s drops 17.7 34.9
RS6B
1100
4.35
21.9
11.0
11.3
4.0
66.0
7.7
29.0
54.0
220
-------
estimated from l/Nnil vs 1/E plots varied from 1.8
o w
to 3.7, with 2.6 as found in Test Run RS13 being
an average value. The fact that the experimental
estimates for N are even less than for stagnant
drops may be an artifact of the assumption of mono-
disperse drops. Variations in drop size and in the
intensity of mixing in the continuous phase surrounding
the drops can cause the ratio of individual-phase
mass transfer coefficients to be different in
different locations within the contactor. King (1964)
has described how a variation in the ratio of Individ-
ual-phase mass transfer coefficients over the total
interface can have this effect.
The regeneration of the loaded n-butyl acetate
from these runs was carried out in a glass, batch
distillation column at atmospheric pressure. The
separation between the phenolics and n-butyl acetate was
very easy to make. The separation between n-butyl
acetate and benzene was more difficult, but was
possible. The separation between MEK and n-butyl
acetate was of intermediate difficulty. These
qualitative results should also apply to continuous
distillations.
In Runs RSI, RS4, and RS7, isobutylene was used
to recover n-butyl acetate from actual and simulated
samples of lube oil refining waste water. The
hydrodynamic behavior of the column was greatly
improved by replacing the 1.50-inch discs and stators
with 2.25-inch holes with a set of 1.75-inch discs
and stators with 2.00-inch holes. Using the method
of Strand, et al. (1962) with GIB = 0.337 gave a
good representation of observed drop sizes and
221
-------
solvent hold-up. Based on the removal efficiency
for n-butyl acetate, the ratio of N from the turbulent
w
mass transfer model (Calderbank and Moo-Young, 1969)
to N estimated experimentally ranged from 0.06 to
W
0.13, with the lower value occurring more often.
The value of N could not be determined experimentally
because of interactions between solutes, but N = N
seemed to give a reasonable estimate.
The regeneration of isobutylene was carried
out continuously during these experiments, but the
very low production rate of recovered solutes made
quantitative evaluation difficult. The separation of
isobutylene from n-butyl acetate was very easy,
and MEK was the only solute that could be detected
in the regenerated isobutylene.
In Runs RS8 and RS9 mixtures of n-butyl acetate
and isobutylene were used to treat actual and simulat-
ed lube oil refining waste water. The operation of
the column was best when using the same set of
discs and stators used for pure isobutylene extraction,
and the hydrodynamics could be correlated by using
G,g = 0.17 for mixtures of solvent near 50% n-butyl
acetate. The ratio of N, predicted for turbulent
W
mass transfer to N(1 predicted from the l/Nnil vs. 1/E
W 0 W
plots varied from 0.5 to 0.8, with the larger value
corresponding to the higher removal efficiencies in
Run RS9. In RS8 where NS could be estimated, the
ratio of N predicted for stagnant drops to that
estimated experimentally was about 6.
The loaded mixed solvent was distilled continu-
ously during each experiment, but the temperature of
the Freon 114 heat source was not high enough to
222
-------
remove more than about half the isobutylene. The
remaining isobutylene and all the n-butyl acetate
were removed and distilled in the atmospheric
batch apparatus. About half the isobutylene was
thus lost. The isobutylene which was recovered
contained almost no (< 10 ppm) n-butyl acetate.
The batch distillation produced very pure n-butyl
acetate for recycle except for a small amount of
benzene which was difficult to separate.
In all three types of extraction steps when
treating actual and simulated lube oil refining
waste water, the removal of benzene is much lower
than predicted based on its high value of Kd<
Experiments on Eth.ylene Quench waste Water.
In Runs RS10 and RS11 samples of ethylene
quench waste water were treated with isobutylene
and isobutane, respectively. All the solutes present
in this polluted water were not identified; the
only individual pollutants whose concentrations were
measured were benzene, toluene, xylenes (o-, m-, and
P-xylene plus ethyl benzene), and phenol. The
COD of the feed and product waters was also measured.
Only processes with a single, volatile solvent were
simulated since the low concentration of phenol did
not appear to justify the added cost of a dual
solvent process.
In Run RS10 isobutylene was used to treat
ethylene quench water at a low solvent-to-water
flow ratio (F /F = 0.101). The principal solutes
were benzene (Kd = 407), toluene (Kd * 1690), and
223
-------
phenol (Kd = 0.7, Appendix E). The distribution
coefficient for toluene was estimated from K.
for benzene by correcting for differences in their
water solubilities and their solubility parameters.
The settings of the parameters and results are
shown in Table 46. The material balance closures
for each solute were within 12%.
In Run RS11 isobutane was used to treat
ethylene quench water at a low flow ratio (F /F =
5 W
0.097). The principal solutes were benzene (K^ - 338),
toluene (Kd ^ 1460), and phenol (Kd = 0.2, Appendix E).
The distribution coefficients for benzene and toluene
were estimated by correcting K. for benzene between
water and isobutylene. The settings of the parameters
and the.results are shown in Table 47. The material
balance closures for each solute were within 15%.
The major difference in results between Runs
RS10 and RS11 was an observation that the turbidity
of the feed water was reduced more dramatically by
isobutane extraction. The turbidity of feed
and product waters in Run RS11 was determined by
using a laboratory turbidimeter (Hach Chemical
Company, Model 1860A); a 72% reduction in turbidity
was measured. The turbidities of feed and product
waters in Run RS10 were not measured, but by
visual inspection no significant improvement
in turbidity occurred by isobutylene extraction.
The ability of isobutane to extract the dispersed
solids and liquids which caused the turbidity
was probably the reason for a greater COD reduction
with isobutane extraction (58%) than with isobutylene
extraction (35%).
224
-------
Table 46. Results from Run RSI0
Water Flow Rate =4.62 gal/hr
Isobutylene Flow Rate = 0.786 gal/hr
Rotating Disc Diameter = 1.75 inch
Stator Hole Diameter =2.00 inch
Rotational Speed = 1450 RPM
Analytical Results (concentrations in ppm):
Solute
Benzene
Toluene
Xylenes
Phenol
Feed
71.1
40.5
40.3
66.9
Product
2.9
2.3
<1
63.1
% Removal
95.9
94.3
>97
5.7
In Column
6.6
4.3
7.1
66.1
Measured Solvent Hold-up = 0.0217
Estimated Maximum Stable Drop Diameter =0.05 inch
Column Temperature = 20.8°C
Feed Product
Measured COD (ppm) 1880 1209
Calculated COD (ppm) 632 166
225
-------
Table 47. Results from Run RSll
Water Flow Rate =4.60 gal/hr
Isobutane Flow Rate = 0.786 gal/hr
Rotating Disc Diameter = 1.75 inch
Stator Hole Diameter = 2.00 inch
Rotational Speed = 1450 RPM
Analytical Results (concentrations in ppm):
Solute
Benzene
Toluene
Xylenes
Phenol
Feed
81.2
43.8
33.6
68.2
Product
2.4
1.6
<1
66.0
% Removal
97.0
96.3
>97
3.2
In Column
10.6
5.1
6.9
66.5
Measured Solvent Hold-up = 0.00961
Estimated Maximum Stable Drop Diameter = 0.05 inch
Column Temperature = 23.4°C
Feed Product
Measured COD (ppm) 1880 699
Calculated COD (ppm) 655 169
Measured Turbidity (FTU)* 142 40
* FTU = Formazin Turbidity Units
226
-------
The comparison of results from Runs RS10 and
RS11 with the predictions from the method of Strand,
et al . (1962) was based on choosing GIQ from the
observed maximum stable drop size. The Peclet
numbers and the values of Nmi are shown in Table
u w
48. The removal efficiencies and values of NQW
for benzene and toluene were slightly larger in
Run RS11 than in RS10; this was probably due to
a slightly better solvent dispersion. The removal
efficiency for phenol was better in RS10 because
of the higher value of Krf. The value of NQW for
toluene was lower than for benzene because the
diffusivity of toluene in water is lower; since
the resistance to mass transfer for these two solutes
lay almost entirely in the water phase, the higher
value of K. for toluene had little influence on NOW.
The ratio of N predicted by the model of Calderbank
and Moo-Young (1961) to NQW (« NW) determined
experimentally for benzene and toluene ranged from
0.51 to 0.60; no comparison of theory to experiment
could be made for Ng.
In both runs the volatile solvent was continu-
ously regenerated by distillation. The analysis
of regenerated solvent failed to show any trace
of benzene, toluene, or phenol. This indicated that
both volatile solvents could be regenerated with ease.
Experiments on Oxvchlorination Waste Water.
In Runs RS5 and RS12 samples of neutralized
oxychlorination waste water were treated with iso-
butylene and 2-ethyl hexanol, respectively. Additional
227
-------
Table 48. Experimental Estimates of N for
ow
Each Solute in Runs RS10 and RS11
G18
Pe
w
Pe
s
F /F
S W
E for Benzene
Benzene % Removal
N for Benzene
ow
E for Toluene
Toluene % Removal
N for Toluene
ow
E for Phenol
Phenol % Removal
N for Phenol
RS10
0.22
4.94
27.3
0.1010
41.1
95.9
5.1
171
94.3
4.3
0.070
5.7
00
RS11
0.20
5.04
28.8
0.0973
32.9
97.0
5.8
142
96.3
5.0
0.019
3.2
CO
ow
228
-------
experiments with this type of waste water are
described in Appendix I.
The sample of oxychlorination waste water
which was treated by isobutylene extraction in
Run RS5 had been neutralized by addition of solid
NaOH. As described in Section V it was later
determined that this method led to decomposition of
about 15% of the chloral to form chloroform and
sodium formate. Run RS5 had to be terminated
before steady state was reached because of a failure
of one of the shaft bearings, but it was estimated
that about 95% of the 3140 ppm of ethylene dichloride
in the feed and about 60% of the 1700 ppm chloroform
in the feed would be removed at an intermediate
flow ratio (F /F = 0.2). Essentially none of the
s w
ethanol , ethylene chlorohydrin, or undecomposed
chloral was removed during treatment with isobutylene.
In Run RS12 a sample of oxychlorination waste
water which had been neutralized by adding NaHC03
was extracted with 2-ethyl hexanol at a low flow ratio
(F/F = 0.106). The settings of independent
variables and the results are shown in Table 49.
Jn addition to the organic pollutants in the feed
waste water, 9.26 weight % NaCl remained from the
neutralization of the HC1. The presence of this
nigh a salt concentration made quantitative analysis
for this run difficult.
Measurements were made of the distribution coef-
ficient for chloral hydrate distributing between 2-ethyl
hexanol and both salt-free and salt-containing waste
Water. We found K. = 50 for chloral hydrate in
229
-------
Table 49. Results from Run RSI2
Water Flow Rate =4.32 gal/hr
Solvent Flow Rate = 0.551 gal/hr
Rotating Disc Diameter = 1.50 inch
Stator Hole Diameter = 2.25 inch
Rotational Speed = 800 RPM
Analytical Results (concentrations in ppm):
Solute Feed Product % Removal In
Column
Ethanol 286 265 7.3 303
Ethylene Dichloride 1505 <20 >99 503
Ethylene Chlorohydrin 1636 1292 21.0 1472
Chloral Hydrate 15220 7726 49.2 10610
2-Ethyl Hexanol in Product Water = 374 ppm
Column Temperature = 22.2°C
Measured Solvent Hold-up = 0.0156
230
-------
pure water and Kd = 146 for chloral hydrate in
10% NaCI solution. From these data the value of
Kd for chloral hydrate in RS12 was about 140.
The value of K, for ethanol distributing between
pure water and sec-octanol is about 0.83 (Krishnamurty
and Rao, 1954), so the value in RS12 was estimated
to be about 1. The value of K. for ethylene di-
chloride between pure water and isobutylene is 70.0,
so the value in RS12 was probably about 200 for this
polar solvent. A value for Kd for ethylene chloro-
hydrin was back calculated from the results on RS12
giving Kd * 5.
Using the method of Strand, et al. (1962),
G,g = 0.13 was estimated from the measured solvent
hold-up. As with runs using n-butyl acetate as
solvent, the hold-up appeared to increase just
below the mid-column bearing. A value of G^g = 0.20
gave a drop diameter closer to that estimated
visually, but there is not enough information available
from this one run to make a choice between these
two values of G^p-
Although the uncertainty in values for K^ makes
any comparison between theory and experiment meaning-
less for this run, several conclusions can be reached
by comparing one solute to the others and by comparing
theoretical estimates for this run with estimates
for other runs. By comparing the removal efficiencies
for the two solutes with large values of E, ethylene
dichloride and chloral hydrate, it is clear that
something caused a significant difference. This
result is also clear by comparing the experimental
values for N with calculated values for NAll
ow ow
231
-------
determined by calculating N from the turbulent mass
transfer model and N from the stagnant drop model
using 0.13 for G,fl. The experimental values of N
I o O W
for ethylene dichloride and chloral hydrate are 7.0
and 0.75, respectively, while the calculated values
are 6.4 and 5.5, respectively. The experiment
described in Appendix I clearly shows that chloral
hydrate undergoes a slow chemical reaction in
addition to simple mass transfer, which explains this
difference.
Another conclusion can be made by comparing N
W
calculated from the turbulent mass transfer model
with N calculated from the stagnant drop model.
In all previous runs, the calculated value for N
was 5 to 15 times larger than the calculated value
for N , while for this run the calculated values
W
are about equal. The explanation for this difference
is related to the high viscosity of 2-ethyl hexanol
(i.e., about 8 cp). When 2-ethyl hexanol is used
as solvent, the relative importance of the solvent-
phase resistance to mass transfer will be increased.
This will affect the removal of ethanol and ethylene
chlorohydrin most seriously.
232
-------
Experiments on Phenol-Formaldehyde Resin Manufacture
Wastewater
Table 50 reports results from Run RS15, made using
a mixed solvent of n-butyl acetate and isobutylene for
extraction of the wastewater from phenol-formaldehyde
resin manufacture. The disc and stator combination
used was that which had proven to be most successful
for mixed-solvent treatment of the lube-oil refining
wastewater. The flow ratio was F/F,, = 0.206, to
s w
allow comparison with a pure butyl-acetate run where
the flow ratio was on the order of 0.1. The three
solutes monitored were methanol (K , = 0.10, measured),
formaldehyde (Kd = 0.15, measured), and phenol (Kd = 6.15,
measured). All these reported K. values are for the
feed water concentration and for no other solutes
present. Values of Kd are, in fact, highly variable
with respect to solute concentration, and with respect
to the concentrations of other solutes in the system.
The observed % removal of phenol is very high under
these conditions, while the observed % removals of
methanol and formaldehyde are quite low.
Table 51 gives results from Run RS16, where pure
n-butyl acetate was used as solvent for treating the
same wastewater, this time with a flow ratio F/F = 0.12.
o W
The flow ratio could not be made lower because of the
need to maintain a sufficient density difference
between phases as phenol built up in the solvent phase.
The density of pure butyl acetate is 0.882 g/cm3; that
of pure phenol is 1.071 g/cm3. The disc and stator
combination is that used successfully in other experi-
ments with pure butyl acetate solvent. Under these
conditions K. for phenol at the feed concentration is
12.0. The observed % removal of phenol is lower than
233
-------
Table 50. Results from Run RS15
Wastewater from Phenol-Formaldehyde Resin Manufacture
Solvent: 48.2 wt % n-butyl acetate, 51.8 wt %
isobutylene
Water Flow Rate
Solvent Flow Rate
Rotating Disc Diameter
Stator Hole Diameter
Rotational Speed
Column Temperature
Analytical Results (concentrations in ppm):
= 2.41 gal/hr
= 0.677 gal/hr
= 1.75 i nch
= 2.00 inch
= 1200 RPM
= 22.7°C
Solute
Methanol
Formaldehyde
Phenol
Feed
12,000
17,370
48,270
Product
11 ,510
16,450
483
% Removal
4.1
5.3
99.0
234
-------
Table 51. Results from Run RS16
Wastewater from Phenol-Formaldehyde Resin Manufacture
Solvent: n-Butyl Acetate
Water Flow Rate = 2.18 gal/hr
Solvent Flow Rate = 0.295 gal/hr
Rotating Disc Diameter = 1.50 inch
Stator Hole Diameter = 2.25 inch
Rotational Speed = 1250 RPM
Column Temperature = 23.4°C
Analytical Results (concentrations in ppm):
jolute Feed Product % Removal
Methanol 12,000 7,608 36.6
Formaldehyde 17,370 10,370 40.3
Phenol 48,270 6,082 87.4
235
-------
in the mixed-solvent run (87.4% vs. 99.0%), but the
% removals of methanol and formaldehyde (37% and 40%,
respectively) are much higher than in the mixed-solvent
run.
It is interesting to explore the reasons for
the very different removals of the various solutes in
the two different runs. In the mixed-solvent run,
much of the phenol extraction occurs with a much higher
equilibrium distribution coefficient than that
prevailing at the feed concentration, since the dis-
tribution coefficient is much higher at lower concentra-
tions of phenol in the aqueous phase (see Appendix F).
Consequently there is a very high % removal of phenol.
For methanol and formaldehyde the extraction factors
are quite low, of the order of E = 0.02 to 0.03,
based upon the equilibrium distribution coefficients
for these components into the solvent (in the absence
of phenol). These values of the extraction factor
would correspond to expected removals in the range of
2 to 3%, somewhat lower than the values of 4 to 5%
observed. The somewhat larger actual percentage re-
movals probably result from the solvent power of the
phenol which builds up in the solvent. A mass balance
on phenol indicates that there is about 19% phenol
in the exit solvent. This would be sufficient to in-
crease the equilibrium distribution coefficients for
methanol and formaldehyde significantly.
The lower removal of phenol and the higher removals
of methanol and formaldehyde in the pure n-butyl-acetate
solvent run (RS16) can be rationalized on the basis
of the still higher build-up of phenol in the solvent
phase because of the lower flow ratio. A mass balance
236
-------
on phenol indicates that the phenol builds up in
the exit solvent phase to a level of about 25%,
while methanol and formaldehyde build up to levels
of 3 and 4%, respectively. The higher concentration
of phenol in the exit solvent will reduce the equili-
brium distribution coefficient of phenol at the
solvent-exit, water-inlet end of the column, thereby
causing a "pinch" situation in the operating diagram,
which serves to reduce the separation of phenol attain-
able. At the same time, the higher concentration of
phenol in the solvent phase should serve to increase
the equilibrium distribution coefficients for methanol
and formaldehyde substantially, and the presence of
significant amounts of methanol in the solvent may further
increase the distribution coefficient for formaldehyde,
and vice versa, since these two molecules can interact
strongly through hydrogen bonding:
H3C - 0 - H --- 0 = CH2
These results imply that one can trade-off between the
percent removals of phenol and of methanol and formal-
dehyde in extraction of this wastewater by adjusting
the solvent to water ratio and thereby altering the
extent to which the solutes build up in the solvent phase.
The very complex interactions between the solutes
and their high levels of concentration preclude any
reliable analysis of the mass-transfer behavior of
the extractor in these runs.
237
-------
Experiments on Hydrofiner Condensate Wastewater
Table 52 presents results of Run RS17, where methyl
isobutyl ketone (MIBK) was used as a solvent to treat
the hydrofiner condensate wastewater. The disc and
stator sizes correspond to those used for butyl
acetate runs with other waters. The flow ratio chosen
was FS/FW = 0.125. The one peak which was monitored
was that of phenol (measured Kd = 68), for which the
removal was better than 99.8%. The removal cannot
be determined more precisely since the limit of quan-
titative detection of the chromatographic analysis
method is about 1 ppm, and the level of phenol in
the effluent water was below that level. Other or-
ganic components were present at similarly low levels,
and the chromatographic method was not sufficiently
sensitive to monitor their removal.
The extraction with MIBK serves to leave consider-
able (15,700 ppm) MIBK in the product water. This
corresponds to about 42,500 ppm of COD; hence a COD
measurement for the exit water reflects primarily
the MIBK content and does not give information on the
degree of removal of other constitutents. Of course,
the large residual amount of MIBK in the exit water
could be removed efficiently by a subsequent extraction
with isobutylene or isobutane.
Table 53 shows results for Run RS18. where a
mixed solvent of MIBK and isobutylene was used to
extract the hydrofiner condensate at a phase flow
ratio of FS/FW = 0.209. The disc and stator combination
was the same as used previously for mixed solvents.
Again the removal of phenol is so large that the amount
of phenol left in the effluent water cannot be reliably
detected. The amount of MIBK left in the product water
238
-------
Table 52. Results from Run RS17
Wastewater: Hydro-finer Condensate
Solvent: Methyl Isobutyl Ketone
Water Flow Rate = 3.93 gal/hr
Solvent Flow Rate = 0.617 gal/hr
Rotating Disc Diameter = 1.50 inch
Stator Hole Diameter = 2.25 inch
Rotational Speed = 1250 RPM
Column Temperature = 24.5°C
Analytical Results (Concentrations in ppm):
Solute Feed Product % Removal
Phenol 400 <1 >99.8%
COD of feed water = 17,530 ppm
MIBK Concentration
in Product Water = 15,680 ppm
(42,500 ppm COD)
239
-------
Table 53. Results from Run RS18
Wastewater: Hydrofiner Condensate
Solvent: 49.5 wt % Methyl Isobutyl Ketone;
50.5 wt % isobutylene
Water Flow Rate = 2.50 gal/hr
Solvent Flow Rate = 0.753 gal/hr
Rotating Disc Diameter = 1.75 inch
Stator Hole Diameter = 2.00 inch
Rotational Speed = 1150 RPM
Column Temperature = 24.2°C
Analytical Results (concentrations in ppm)
Solute
Phenol
Feed
400
Product
% Removal
> 99.8
COD of Feed Water
COD of Product Water
MIBK Content of
Product Water
17,530 ppm
18,530 ppm
3150 ppm
(9000 ppm COD)
240
-------
is much less for the mixed-solvent case (3150 ppm, as
compared to 15,700 ppm for pure MIBK solvent), and
hence contributes less to the COD of the product water.
Subtracting an estimated 9000 ppm for the contribution
of the MIBK, the COD of remaining constituents of the
effluent water is about 9,500 ppm. This corresponds
to removal of about 46% of the COD from the consti-
tuents of the original water. As was indicated in
Section V, it is expected that most of the original
COD is attributable to HgS, which would be removed
by a stripping process. Again, the residual MIBK
could readily be removed by a subsequent extraction
with isobutane or isobutylene,
Experiments on Wastewater from Styrene Manufacture
Pure isobutylene was used as the solvent for
extraction of the wastewater from styrene manufacture,
since the components known to be present were all
aromatics, with very high distribution coefficients
into isobutylene. Table 54 reports experimental results
for the extraction of this water with isobutylene. The
combination of 1.75-inch discs and 2.00-inch stators
was employed, so as to give visually good dispersion.
Comparison of the feed concentrations with those
originally measured for this water (p. 78) shows an
appreciable loss of the various components during a
few months' storage.
The mass flow ratio of solvent to feed water
(FS/F ) was 0.107. For benzene (Krf = 407, measured),
the extraction factor (KdFs/Fw) was therefore 43.5.
Values of K, for ethylbenzene and styrene were not
measured, but would be expected to be greater than Kd
for benzene, because of the additional hydrocarbon
241
-------
groups on the molecule. Hence for all three solutes
the extraction factor was quite high, and may to a good
approximation be considered infinite for purposes of
interpreting the mass-transfer behavior.
The percentage removals observed for the three
components in this extraction run are all high, and
are in line with what would be expected by extrapolating
the mass transfer rates observed in previous runs
with other waters, for the case of a very large extrac-
tion factor. Thus, volatile-solvent extraction with
isobutylene serves t.o remove the dissolved aromatics
very effectively from this water.
The measured COD for the feed water is substantially
less than would correspond to complete oxidation of the
three aromatic species. This result is not surprising,
in view of the fact that these aromatics are known to
be refractory to both biological and chemical oxidation;
hence the measured COD does not reflect the total
amounts of dissolved organic species present. Similarly,
the percentage removal of COD cannot be looked upon
as a meaningful figure, except that it confirms that
the decontamination of the water by isobutylene
extraction is quite good. The refractiveness of the
dissolved organics to oxidation is an incentive for
treating the water by extraction, as opposed to biologi-
cal oxidation, in practice.
242
-------
Table 54. Results from Run RS19
Wastewater: Styrene Manufacture
Solvent: Isobutylene
Water Flow Rate = 3.00 gal/hr
Solvent Flow Rate = 0.543 gal/hr
Rotating Disc Diameter = 1.75 inch
Stator Hole Diameter = 2.00 inch
Rotational Speed = 1250 rpm
Analytical Results (concentrations in ppm)
jolute Feed Product % Removal
Benzene 290 10 97
Ethylbenzene 120 4 97
Styrene 15 <1 >93
COD of Feed Water = 530 ppm
COD of Product Water = 50 ppm
% Removal of COD = 91%
243
-------
SECTION IX
EFFECTS OF SCALE ON RDC DESIGN
The steps in the development of a new process
usually include operation on one or more small scale
units before a commercial process design can be
developed. This is certainly the case for a waste
water treatment process which is based on solvent
extraction. In this chapter we consider how the
data described in Section VIII for the RDC extractor
can be used to project the design of a commercial
process and to define the need for a pilot plant
program. The extraction of phenol and MEK from a
waste water using n-butyl acetate as solvent is
used as an example, but the discussion also applies
to other extraction processes.
Steps in Developing the Design of an RDC Extractor.
The initial step in developing the design of
an RDC extractor would probably be done on as small
a scale as possible. As an RDC is scaled down,
mechanical factors limit the minimum size. The 3-inch
diameter column used in this study is about as small
as can be used to obtain valid data for scale-up.
In a smaller unit the rotating shaft on which the
discs are mounted would take up too large a fraction
244
-------
of the volume of the column. Even on the 3-inch
RDC used in this study, it was necessary to install
a bearing midway up the 4-foot long column to
minimize shaft vibration. This bearing cut down
the cross-sectional area available for fluid flow
to the extent that a non-uniform hold-up was observed
in some experimental runs.
Another factor affecting the operation of the
RDC in this study was the volume of waste water
which was available for testing. Since we found that
about 45 minutes were required to reach steady
operation, the 5-gallon samples we had available
meant we had to use flow rates of 5 gal/hr or less.
In a 3-inch RDC this flow produces a water-phase
superficial velocity of 13.6 ft/hr, which is lower
than what would usually be used commercially.
Because the nature of this study was to investigate
a variety of waste waters, we accepted these
limitations and realized that further pilot plant
study would be required to develop the commercial
design.
A company with an existing or an anticipated
process which produces a waste water stream that
is amenable to treatment by solvent extraction
would probably study the extraction step in a pilot
plant. It would be desirable to use actual, fresh
waste water either from an existing process or from
a pilot plant. It would also be useful to operate
the solvent regeneration equipment in a continuous
manner to help fix the material balance for the
entire extraction process. The problem of having
a limited supply of waste water would not be as
245
-------
severe, so the pilot-scale RDC could be designed
and operated in a manner that would give the best
probability for successful scale-up to a full-scale
plant.
Considerations in Changing Scale of an RDC.
In Section V the equations describing an
RDC extractor were shown to suggest a rational basis
for scale-up. This basis included holding the
following quantities constant during a change in
seale:
1) The ratios d,/D, d /D, and H /D,
I o \f
2) The superficial velocities, Vd and V , and
3) The power per unit mass, P/M.
In addition consideration must be given to flooding
tendencies and to axial mixing, both of which will
change with changes in scale. We can use the results
discussed in Section VIII to estimate the optimum
values for these quantities on a commercial scale so
that the pilot plant can be operated at the best
possible set of conditions.
Consider as an example a waste water which
contains phenol and MEK. The results in Section VIII
showed that a good removal of phenol and a moderate
removal of MEK could be achieved with a low solvent-
to-water flow ratio (F/F = 0.1). We will consider
s w
a design for an RDC which operates with FS/FW = 0.1
using pure solvent feed and which achieves a 99%
removal of phenol. We will consider such designs
for two possible full-scale waste water flow rates,
10 gal/min and 100 gal/min. Within these constraints
we will consider a partial optimization to define
246
-------
an RDC extractor of minimum cost. Note that to
optimize the entire extraction process, the best
value for F /F would also have to be determined
s w
as was done in Section IV for several processes.
In this example we will consider only one value of
F /F , so the cost of solvent regeneration will
5 W
be fixed.
For the experiments with lube oil refining
waste water, we found that the column hydrodynamics
could be correlated by the equations of Strand, et al.
(1962) using G,g = 0.20. Assuming a column tempera-
ture of 22°C leads to the physical properties listed
in Table 55. The mass transfer estimates were made
by assuming that N was equal to N from the turbulent
W W
"lass transfer model divided by 0.5 and that N. was
o
equal to N from the stagnant drop model divided
by 2.6; these ratios were determined from experi-
mental data as described in Section VIII. The fraction
of flooding was estimated by the method of Logsdail,
et al. (1957).
The variables which were determined in the
optimization were VG and N (equivalent to optimizing
F7TT) . The ratios d./D, dg/D, and HC/D were held
at the values used in the experimental RDC (i.e.,
^/D = 0.5, ds/D = 0.75, and HC/D = 0.333). The
calculation procedure was as follows:
1. Assume a value for Vc> From the fixed
FS/FW and FW, calculate Vd and D.
2. Assume a value for the height of the column.
3. Calculate the removal efficiency for
phenol as a function of the disc rotational
speed, N. Locate the N corresponding to the
247
-------
Table 55. Physical Properties for RDC Design
Sample Calculation
Water-phase density = 0.9978 gm/cc
Water-phase viscosity = 0.955 cp
Solvent-phase density = 0.8788 gm/cc
Solvent-phase viscosity = 0.704 cp
Interfacial tension = 13.9 dyne/cm
Solute Diffusivities (105ft2/hr) in Water and
in n-Butyl Acetate, and K,'s:
Solute Dw Ds Kd
MEK 3.65 6.21 4.56
Phenol 3.47 6.04 57.0
248
-------
greatest phenol removal, which at the
same time corresponds to operating at
50% or less of the flooding velocities.
4. Iterate back to step 2 until a column
height is determined which gives 99% phenol
removal at the optimum value of N.
5. Repeat steps 1 to 4 for several values of
V so that a plot of column height vs. D
\f
may be prepared.
6. Use the plot of D vs. column height to
prepare a plot of column cost vs. V .
Determine the value of VG which gives the
minimum cost.
These calculations were made using a CDC 6400
computer with the program ROC, which is listed in
Appendix H.
This procedure will first be demonstrated for
a waste water flow of 10 gal/min. Figure 28 is
a plot of n (fraction of phenol not removed) vs. N
for several values of V between 15 and 60 ft/hr.
Consider the curve for V. = 15 ft/hr. At 100 RPM
t*
the removal efficiency is about 80%. As N increases,
the phenol removal improves because of better mass
transfer. However, at about 265 RPM a minimum
occurs, and at higher values of N axial mixing
becomes more severe so that the removal efficiency
drops. At 342 RPM the limit for 50% of flooding
Is reached. At larger values of N (dashed lines
1n Figure 28) the column is operating at greater
than 50% of flooding. Since the minimum in the
curve corresponds to 99% phenol removal, it is
clear that the iteration between steps 2 and 4
249
-------
o
CD
\ \
\ \
\ \
60 ft/hr\ ^5 ft/hr
0.004 -
200 300
N (RPM)
Figure 28. Effect of Rotational Speed on Phenol
Removal at 10 6PM Waste Water Flow
250
-------
above has already located the required column
height (18.8 ft). Considering next the curves
for larger values of Vc, we see that the constraint
requiring the column to operate at or below 50%
of flooding limits the maximum value of N before
the minimum in the curve is reached.
Following the procedure in step 5 above,
Figure 29 is a plot of D vs column height. All
points on this curve correspond to 99% phenol removal
at the optimum value of N. The curve in Figure
29 is used to determine the cost for a column using
the cost data of Clerk (1964). His costs are for
a fabricated RDC of carbon steel including discs,
stator rings, manways, nozzles, supports, and skirt,
but not including the motor. Clerk's data were
updated to Dec. 1973 costs by using the Marshall
and Stevens chemical equipment cost index. A plot
of purchased cost vs. V is shown in Figure 30.
\f
There are several points of interest on Figures
29 and 30. At high values of D in Figure 29 there
is a region where increasing Vc tends to decrease
both D and the column height. This result reflects
the importance of axial mixing at low values of Vc>
In this region the decrease in axial mixing is so
significant that 99% phenol removal can be achieved
in a shorter as well as smaller column when VQ is
increased. In Figure 30 there is a minimum in the
cost curve which occurs at a value of Vc of about
45 ft/hr. The cost for a column operating at Vc = 15
ft/hr is more than double the minimum cost.
251
-------
20
18
16
« 14
12
10
o
o
•5 8
UJ
o
Q
cc
0
0
Optimum
Cost
1
10 15 20 25 30
Extractor Diameter (in.)
Figure 29. Extractor Size for 10 GPM Waste Water Flow
252
-------
o
o
w
o
•*—
O
X
LJ
o
g2
0
Optimum
Cost
I
I
I
1
I
0 10
Figure 30
20 30 40 50 60
Vc (ft/hr)
Extractor Cost for 10 GPM
Waste Water Flow
253
-------
The removal efficiency of MEK from this hypo-
thetical waste water is only slightly affected over
the range of column designs considered. For all
cases when the removal efficiency for phenol was 99%,
the removal efficiency for MEK varied between 40.6
and 43.4%. The removal efficiency at the optimum
design for phenol removal was 43.3%.
A similar calculation was made for a waste water
flow of 100 gal/min. Figure 31 is a plot of n vs. N
for values of V between 15 and 90 ft/hr. Note that
at this higher waste water flow rate, flooding begins
to be the constraint limiting N at V of about 60
ft/hr instead of at 25 ft/hr as was found for 10
gal/min of waste water flow. Note also that the
maximum removal efficiency occurs at a lower value
of N in this larger column. Plots of column
height vs D and cost vs V are shown in Figures 32
and 33. Some difficulty was encountered at the
highest values of V for this case. At these
conditions the differences between the flooding
correlation of Logsdail, et al . (1957) and the
correlation for VV from Strand, et al . (1962) resulted
in no solution being possible for 50% of flooding
according to Logsdail, et al. A recent comparison
of flooding correlations for an RDC (Landau and
Houlihan, 1974) indicated the Logsdail, et al . ,
approach was better when comparing data for small
columns, but no comparison has been published
for larger columns. Despite this problem at very
high values of VG, it is clear from Figure 33 that
the optimum value of V is somewhat above 100 ft/hr.
254
-------
1.0
0.4
o
I O.I
0.04
0.01
V^ = 15 ft/hr
/
0
Figure 31.
50
/45 ft/hr
V/90 ft/hr
I I I
100 150
N (RPM)
200
Effect of Rotational Speed on Phenol
Removal at 100 GPM Waste Water Flow
255
-------
60
50
40
I
t_
o
t3 30
o
X
LU
or
20
10
1
1
1
1
I
0 20 40 60 80 100
Extractor Diameter (in.)
Figure 32. Extractor Size for 100 GPM
Waste Water Flow
256
-------
40
30
en
o
o
o
o
UJ
o
o
ce
20
10
0
0 20
40 60 80
Vc (ft/hr)
100
Figure 33. Extractor Cost for 100 GPM
Waste Water Flow
257
-------
Sizing a Pilot Plant RDC.
Once the optimum value of V has been determined,
then a pilot plant RDC can be sized. Consider
first a pilot scale RDC which will be used to
gather data for the 10 gal/min waste water treat-
ment. Assume that a 3-inch diameter column will
be used. At the optimum value of V = 45 ft/hr,
such an RDC would be able to treat 16.5 gal/hr
or 0.275 gal/min of waste water. At the optimum
operation of the large RDC, N = 260 RPM and F/TT =
76.3 ft-lbf/hr-lbm. The 3-inch RDC would give the
same P~7M when N = 566 RPM.
The height of the pilot scale RDC can be chosen
to achieve about the same removal efficiency as
for the large extractor. A 3-inch diameter RDC
operating at 566 RPM would have to be about 6.5
ft tall to give 99% phenol removal. The MEK
removal would be about 44%. By choosing the
pilot plant RDC in this manner, the loaded solvent
composition would be about like that in the full-
scale plant. This would make the operation of
regeneration equipment very similar to the actual
process.
The scale-up factor in terms of waste water
flow rates from the 3-inch RDC to the 18-inch RDC
would be about 36. A somewhat larger diameter pilot
plant RDC might be chosen to reduce this scale-up
factor. For the pilot plant RDC to be used to
determine the design of the large RDC which would
treat 100 gal/min, the diameter of the pilot
plant would probably be larger. If a 6-inch diameter
258
-------
column were operated at VG = 105 ft/hr, the pilot
plant could process 2.57 gal/min giving a scale-up
factor of 39.
At the optimum operation of the large RDC,
N = 118 RPM and P/M = 28.0 ft-lbf/hr-lbm. The
6-inch column would give the same P/M at N = 316
RPM. However, this design would operate at 58.7%
of flooding, so a slightly lower value of N might
be used. At 316 RPM, the phenol removal efficiency
for a 10-foot high column would be 99.2%, and the
MEK removal would be 43.9%.
It should be noted that the calculations made
in this chapter are only approximate. We have
assumed that the ratio of theoretical to experimental
mass transfer coefficients for each phase would
remain constant over a wide range of operation.
We also assumed that GIQ would be constant. To
compensate for these approximations, a pilot plant
RDC should be designed with a variable-speed
drive, and an experimental program should be designed
which includes the investigation of a range of
values for Vc. However, it is reassuring to note
that the calculations predict RDC extractors of
quite reasonable sizes (see Table 56).
259
-------
Table 56. Summary of RDC Designs
For 10 gal/min full-scale flow
Waste Water Flow
Column Diameter
Column Height
Rotational Speed
Pilot Plant Full-scale Plant
0.275 gal/min 10 gal/min
3 inches 18 inches
6.5 ft. 10.9 ft.
566 RPM 260 RPM
For 100 gal/min full-scale flow
Waste Water Flow
Column Diameter
Column Height
Rotational Speed
2.57 gal/min 100 gal/min
6 inches 38 inches
10 ft. 15.8 ft.
316 RPM 118 RPM
260
-------
SECTION X
SUMMARY AND DISCUSSION OF APPLICATION
In previous Sections we have discussed general
process considerations relative to treating a waste
water for the recovery of chemical pollutants and
have described an experimental program in which
solvent extraction was studied in a small pilot
plant. We studied a variety of potential and actual
pollutants with a major portion of the program
being directed toward the use of a volatile solvent.
In experiments conducted using an RDC extractor,
we also demonstrated several promising extraction
process configurations which utilize a less-volatile,
polar solvent as well as a volatile solvent.
In this chapter we will extend the experience
gained from the general process considerations and
from the experimental study to develop a strategy for
choosing a processing technique for a waste water
generated by the petroleum, petrochemical, or organic
chemical industry. Any particular waste water
stream will have its own characteristics which will
Probably require further research and process
development, but the strategy to be described should
be useful in directing such further study in the
m°st promising direction.
261
-------
Strategy of Process Selection.
The initial step in choosing a promising
processing method is to q u a n t i f y the major pol1utants
present in the waste water. We will direct our
attention primarily to individual waste water streams
which do not contain more than a few major pollutants.
The analytical techniques based on gas chromato-
graphy (Herz, 1972; Sugar and Conway, 1968) should
be most useful in identifying the pollutants. Gas
chromatography may also be among the most attractive
methods for quantitative analysis. If the waste
water stream comes from an existing plant, then a
sampling program should be able to determine the
typical range of concentrations for the major
pollutants. In the development of a new process, the
determination of typical concentrations of pollutants
in anticipated waste water streams should be a part
of the process development.
The next step is to determine the val ue o_f
the pollutants. The value of the individual pollu-
tants can be estimated by assuming a substantial
fraction recovery and sale at a price of approximately
one-half to one times the present market value.
When several pollutants are present, the value of
the mixture should also be estimated. If this mixture
can be recycled to the process, its value can be
estimated from the reduction in raw material needs
or the increase in product yield. When considering
recycle it is important to realize that all pollu-
tants will be recycled so that the product quality
262
-------
may be affected. The pollutant value can be determined
in units of $/thousand gallons of waste water.
The next step is to decide j_f recovery i_s_
promisi ng. If the value of pollutants is greater
than about $3/thousand gallons, then recovery for
pollutant value is a promising possibility. If the
value is between $1 and $3/thousand gallons, then
the possibility of recovery should be considered.
If the value of pollutants is below $l/thousand
gallons, then recovery will seldom be economically
justified. When a non-recovery technique is indi-
cated, biological oxidation or carbon adsorption
are likely techniques.
In cases where one or more of the major
pollutants is refractory to treatment by biological
oxidation or is toxic to the bacteria, then a process
which concentrates these pollutants may be the best
approach even though the pollutant value is below
Si/thousand gallons of waste water. Solvent extrac-
tion, particularly with a volatile solvent, should
be considered as an alternative waste water treat-
ment technique. The strategy for choosing between
solvent extraction and steam stripping as a method
for generating a concentrated pollutant stream
is the same as that described below for the case
of a recovery technique. The best solvent extrac-
tion or steam stripping process can then be compared
to carbon adsorption.
When the pollutant value indicates that a
recovery process is promising, steam stripping and
263
-------
solvent extraction should be the principal approaches
considered. In rare cases, other recovery processes
such as carbon adsorption or reverse osmosis could
be considered, but these techniques will generally
be more expensive than either steam stripping or
solvent extraction. Although we refer to the process
as steam stripping, reflux will be required to
produce an organic product with a low water content.
The process is actually distillation.
The next step is to choose a_ good solvent
extraction process so that the comparison between
extraction and stripping is valid. The determination
of a steam stripping process arrangement will not
be developed here, except to note several factors
which make steam stripping less favorable. The
lower the relative volatility of the organic pollutant,
the more difficult it is to remove by stripping.
Since with all but the lowest molecular weight
organic compounds an azeotrope will occur at some
concentration, the lower the organic content in
the azeotrope, the less highly purified is the
recovered pollutant stream from a steam stripping
process. With the higher molecular weight organic
compounds, a heterogeneous azeotrope will form.
There is an advantage in terms of the ease of
separation when the distillate separates into two
phases as it is condensed. When the waste water is
corrosive, steam stripping at elevated temperatures
will aggravate the problem, while solvent extraction
at ambient temperature will be favored.
264
-------
In trying to identify a good solvent extraction
process, first consider volatile sol vent extraction.
A value of Kd for each pollutant distributing between
isobutylene or isobutane and water should be estimated.
The data of Appendix E will be useful in making
this estimate. A simple experimental extraction
at atmospheric pressure using pentane or hexane may
also be useful in estimating Kd. If Kd is greater
than about 10, then volatile solvent extraction
should be a good choice. If Kd is between 2 and 10,
volatile solvent extraction may still be useful,
but other feasible techniques should also be considered
If K. is much less than 2, then volatile solvent
d
extraction will probably be uneconomical.
When a mixture of pollutants is present, all
pollutants which are extracted will be recovered as
a single mixture when using volatile solvent extrac-
tion. Separation of this mixture into its components
may be a practical consideration. When the pollutants
have a range of values of K., then it may be desirable
to extract only those components with high values of
Kd. The components with low values of Kd may not
be worth recovering, or another recovery method
might be considered.
If the value of Kd between water and isobutylene
or isobutane is low, then consider dual solvent
extraction. Values for Kd for each pollutant
distributing between water and a polar solvent should
be estimated. The polar solvents best suited for
the dual solvent process generally are those having
a water solubility between 0.5 and 2.0%. Some
examples are n-butyl acetate, methyl isobutyl ketone,
265
-------
n-hexanol , diisopropyl ether, and ethylene dichloride.
The data reported here or K. data from other sources
may be used in some cases, and experimental estimates
for K. may be made fairly simply in other cases. A
variety of polar solvents should be tested since
some pollutants may be very readily extracted by
one polar solvent and not well extracted by others
(e.g. chloral). By choosing polar solvents with
the above range of water solubilities, there is a
good chance that Kd for the polar solvent distributing
between water and a volatile solvent will be high.
Mixtures of polar solvents may also be considered.
Once the polar solvent is identified which
gives the highest value of Kd of those considered,
this value of Kd should be compared with the value
of Kd for volatile solvent extraction. If the value
of Kd for polar solvent extraction is no more than
about a factor of 2 higher than the value of K. for
volatile solvent extraction, then the dual solvent
process will not hold an advantage over volatile sol-
vent extraction. The data in Appendix E for aliphatic
aldehydes indicate that this class of organic
compounds have values of Kd for isobutylene extraction
which are only slightly less than for many polar
solvents. Typically dual solvent extraction will
be favored over volatile solvent extraction when
the values for Kd in both dual solvent extraction
steps are larger than 20 while the value for K.
d
for volatile solvent extraction is less than about 5.
This is the case for phenol extraction since for
extraction by n-butyl acetate Kd = 57.0 and for
266
-------
n-butyl acetate recovery by isobutylene K. = 168,
while K. = 0.70 for extraction of phenol using iso-
butylene.
When a mixture of pollutants is present, then
the relative volatility between the polar solvent and
each pollutant must be considered when the polar
solvent will be regenerated by distillation. If
the polar solvent is more volatile than all pollu-
tants or less volatile than all pollutants, then
a single mixture of pollutants will be produced with
a dual solvent process. However, if some pollutants
are more volatile than the polar solvent and others
are less volatile, then an additional distillation
would be required to regenerate the polar solvent.
It is much less likely that a dual solvent process
with an additional distillation step necessitated
by relative volatility considerations would be less
expensive than some other recovery technique.
The use of a polar solvent extraction step for
the removal of certain pollutants does not necessarily
dictate the use of volatile solvent extraction for
the recovery of the dissolved polar solvent. It
would be possible to use steam or inert gas stripping
(as in the Phenosolvan process described in Section III)
for polar solvent recovery. However, when the polar
solvent is chosen from solvents having a water solu-
bility of 0.5 to 2.0 weight %, then it is likely
that volatile solvent extraction will be a good
choice. In Section IV we suggested two flow scheme
variations for a dual solvent process. When the
solvent process with linked solvent cycles has
267
-------
advantages over the dual solvent process with separate
solvent cycles, then volatile solvent extraction
rather than a stripping technique would have to
be used for polar solvent recovery to accrue these
advantages.
We are not yet able to make many quantitative
generalizations on which of the two dual solvent
process arrangements would be best. However, when
the polar solvent is more dense than water (e.g.,
ethylene dichloride), a dual solvent process with
linked solvent cycles would probably not be used
because this would decrease the density difference
between mixed solvent and water. When the polar
solvent is slightly less dense than water and the
presence of the pollutant decreases this density
difference (e.g., extraction of phenol with n-butyl
acetate), then a dual solvent process with linked
solvent cycles will tend to be favored.
The final step is to compare the best
sol vent extraction process wi th steam stripping. In
Section IV this comparison was made for the recovery
of n-butyl acetate using either steam stripping or
volatile solvent extraction with isobutylene. This
solute has properties which favor both steam
stripping (by forming a heterogeneous azeotrope) and
volatile solvent extraction (a large Kd of 168).
When costs were compared for these processes they
were essentially equal, even though the assumptions
made with respect to the distillation column stage
efficiency were shown to favor the steam stripping
process. It is reasonable to expect that economic
considerations will favor volatile solvent extraction
in many cases.
268
-------
The comparison between the cost of pollutant
recovery by steam stripping and the cost of pollutant
recovery by dual solvent extraction is expected
to favor steam stripping in the majority of cases.
The reason why dual solvent extraction is favored
in some cases is apparent when details of steam
stripping processes are considered. The concentration
of organic in the liquid phase for a homogeneous azeo-
trope or in the organic-liquid phase for a hetero-
geneous azeotrope determines the maximum purity
of the recovered organic solute. For phenol, the
azeotrope contains only 9 weight % phenol (all these
quoted azeotropic concentrations are from Weast,
1970). For acetic acid, the azeotrope contains only
3 weight % organic. It is interesting to note
that these two solutes are often recovered by solvent
extraction. Many alcohols also show fairly low
organic contents in their water azeotropes. For
example, azeotropic mixtures with n-propanol , n-butanol,
cyclohexanol , and benzyl alcohol contain 72, 56,
20, and 9 weight % organic, respectively. However,
these alcohols tend to have higher relative volatili-
ties with respect to water than does phenol or
acetic acid. This makes it easier to reach the
azeotropic composition and easier to produce puri-
fied water. Organic acids and amines are other
classes of organics which form water azeotropes which
have low organic contents. Other specific organics
which show this property include furfural and pyridine.
The above discussion has been directed toward
Pollutants which are soluble in water. When dispersed
269
-------
orgam'cs are present, steam stripping is probably
not the best alternative to be compared with volatile
solvent extraction. Mechanical separation techniques,
such as using a coalescer or a centrifuge, are
likely ways to separate the dispersed organics.
However, if an appreciable quantity of dissolved
organics is also present, then solvent extraction
has the advantage of being able to remove both
dispersed and dissolved pollutants in one step.
Simply centrifuging a sample of the waste water and
measuring the remaining organic content would determine
if a treatment for dissolved pollutants is necessary.
Solvent extraction using a volatile solvent is expected
to be an economical approach to recovering a dispersed
organic pollutant in a number of cases.
Examples Illustrating the Strategy of Process Selection-
Consider a waste water which has been determined
to contain 2, weight %_ methanol as the major pollutant.
The pollutant value of about $3/thousand gallons of
water makes recovery a promising approach. Based on
the trends of Kd with decreasing molecular weight for
alcohols, neither volatile nor dual solvent extraction
is promising. Since methanol does not form an
azeotrope with water and is considerably more volatile
than water, steam stripping would be preferred.
Consider a waste water which has been determined
to contain 0.1 weight %_ crotonaldehyde. The pol-
lutant value if recycled to an n-butanol plant is
about Si/thousand gallons (based on full market
price of butanol). Although the pollutant value
is fairly low, at this concentration crotonaldehyde
270
-------
is a bacteriacide which makes biological oxidation
difficult. Recovery is presently accomplished by
steam stripping; however, recovery by volatile solvent
extraction may be a promising alternative. The value
of Kd between water and isobutylene is 2.48 (Appendix E),
and the value of K, for the best polar solvent,
octanol, is estimated to be about 8. With this
little difference in Kd for the volatile solvent
and Kd for the polar solvent, it is likely that
the volatile solvent extraction process would be
favored, unless there are other major pollutants
Present which could only be extracted with a polar
solvent.
Consider the lube oil refining waste water
which contained 2 weight % phenol, 0.2 weight % o-
cresol , and about 0.02 weight % MEK. The value of
Phenol and o-cresol is about $17/thousand gallons
(based on full market value since they could be re-
cycled to the lube oil treating system). The value of
the MEK is only $0.2/thousand gallons. This is a
°ase where we might consider recovering only phenol
and o-cresol and treating the MEK by biological
oxidation. With steam stripping all three components
would be recovered as a single mixture; the diffi-
culty of obtaining a phenol product of low water
content in such a case has been discussed above.
The value of K. for phenol using isobutylene is only
d
°-7, so volatile solvent extraction would not be
economical. However, using n-butyl acetate as the
Polar solvent in a dual solvent process would be
an effective method of recovering only phenol and
°-cresol. Since the volatility of MEK is greater
271
-------
than that of n-butyl acetate, it would not be separa-
ted during polar solvent regeneration. MEK would
build up to a steady-state level in both recirculating
solvents. This equilibrium level would be less than
0.1 weight % in each solvent, so that the presence
of MEK should not affect the extraction of the
p h e n o 1 i c s .
Finally consider the ethylene quench waste
water. The concentrations of the major identified
pollutants (benzene, toluene, xylenes, and phenol)
are all below 0.01%. Their economic value is much
less than $l/thousand gallons, but all except phenol
are refractory to biological oxidation. There is also
a considerable amount of dispersed organics present.
Steam stripping would probably produce a distillate
which contains a dispersion of organic droplets.
Tests in the pilot plant showed that volatile solvent
extraction using isobutane removed both the dispersed
and the dissolved organics. This would probably
be the best method of recovery, although this step
might just be used to concentrate the organic pollu-
tants for another disposal method.
272
-------
SECTION XI
NOMENCLATURE
Symbol Definition Units
A Cross-sectional area of column ft2
A Surface area of fully formed drop ft2
a Interfacial area per volume in
column ft2/ft3
C Area ratio - see Equation (B14)
K
D Column diameter ft
D., D Solute diffusivity - see subscripts
d c below ft2/hr
d. RDC disc diameter ft
d RDC stator hole diameter ft
s
d Drop diameter ft
P
E Extraction factor (= K, F /F )
P Mass transfer parameter - see
Appendix B
F , F. Flow rate - see subscripts below Ib/hr
s «
G Constant
g Conversion factor (4.18 x 108) lb -ft/
lbf-hr2
H Overall column height ft
H RDC compartment height ft
h Distance variable - Appendices B A C ft
273
-------
Kd Distribution coefficient (by weight)
K°? Distribution coefficient at infinite
a dilution
K"jix Distribution coefficient with mixed
solvent
K (K ) Overall mass transfer coefficient
ow oc based on water (or continuous)
phase concentrations ft/hr
Kdf Overall mass transfer coefficient
to drop during its formation
based on drop phase concentration ft/hr
k., k Individual-phase mass transfer
coefficient - see subscripts below ft/hr
MW. Molecular weight of i Ib/lb mol
m Defined in Equation (B8)
N RDC rotational speed RPM
N , N Individual-phase number of transfer
w s units - see subscripts below
N Overall number of transfer units
ow based on water phase concentrations
P RDC power input per compartment ft-lb^/hr
Pe,, Pe Peclet number based on column
c height - see subscripts below
Pe* Local Peclet number
Po Power number - see Equation (B9)
P/M Power per unit mass - see Equation
(Bll) ft-lbf/
hr-lbm
Q Ratio of experimental NQW to
theoretical N
u w
q. Roots of Equation C12
274
-------
R Mass transfer parameter - see
Appendix B
RDC Rotating Disc Contactor
Re RDC disc Reynold's number
V , V . Superficial velocity - see
c subscripts below ft/hr
V. Characteristic velocity - see
k Equation (81) ft/hr
V Characteristic velocity - see
n Equation (B16) ft/hr
V Slip velocity - see Equation (Bl) ft/hr
V Drop terminal velocity ft/hr
L«
V f, V,f Superficial velocity at flooding -
c see subscripts below ft/hr
W
X
Xin
wo
si
Coalescence coefficient - see
Equation (B16)
Solute weight ratio in water phase
Equilibrium solute weight ratio
based on water phase (= Y/K.)
Solute weight ratio at the interface
Solute weight ratio in inlet water
Solute weight fraction in outlet
water
Solute weight fraction in inlet
water
Solute weight fraction in inlet
sol vent
Solute weight ratio in solvent phase
275
-------
Y1 Solute weight ratio at the interface
Y. Solute weight ratio in inlet solvent
in
Z Dimensionless variable (= h/H)
e , e Axial dispersivity - see subscripts
w s below ft2/hr
Y • Activity coefficient of component i
1 i n p h a s e ' j
y , y . Viscosity - see subscripts below Ib/ft hr
n Dimensionless concentration - see
Equation (1)
Fractional hold-up of dispersed
phase
(j),. Fractional hold-up at flooding
p , p. Density - see subscripts below Ib/ft3
cr Interfacial tension lb/hr2
6f Time for drop formation - see
T Equation (B7) hr
Subscripts Meani ng
c Continuous phase
d Dispersed phase (except Kd)
w Water phase
s Solvent phase
276
-------
SECTION XII
REFERENCES
Akell, R. B., Chem. Eng. Progr., 62. (9), 50 (1966).
American Petroleum Institute, "Selected Values of
Physical and Thermodynamic Properties of Hydro-
carbons and Related Compounds," Project 44,
Carnegie Press, Pittsburgh, Pa. (1963).
American Society for Testing Materials, Standards, Part
23, "Water; Atmospheric Analysis," p. 246, Method
D, 1252-67 (1970).
Angelo, J. B., E. N. Lightfoot, and D. W. Howard, A.I.Ch.E
J., 12, 751 (1966).
Angelo, J. B., and E. N. Lightfoot, A.I.Ch.E. J., 14,
531 (1968). —
Bewley, B. R., U. S. Patent 3,467,721 (Sept. 16, 1969).
Beyaert, B. 0., L. Lapidus, and J. C. Elgin, A.I.Ch.E.
J., 7_, 46 (1961).
Beychok, M. R. "Aqueous Wastes From Petroleum and Petro-
chemical Plants," John Wiley & Sons, London (1967).
Beychok, M. R., Paper presented at ACS 168th National
Meeting, Atlantic City, N.J., September (1974).
Blanding, F. H., and J. C. Elgin, Trans. Am. Inst. Chem.
Engrs. , 3_8, 305 (1942).
Bollen, W. M,, Private communication (1974).
Brown, W. V., Chem. Eng. Prog., 59. (10), 65 (1963).
Brutvan, D. R., Ph.D. Dissertation, Rensselaer Polytechnic
Institute, Troy, N.Y. (1958).
277
-------
Cairns, E. J. and J. M. Prausnitz, Chem. Eng. Sci., 1_2,
20 (1960).
Calderbank, P. H., and M. B. Moo-Young, Chem. Eng. Sci . ,
j_6, 39 (1961).
Cavers, S. D., and J. E. Ewanchyna, Can. J. Chem. Eng.,
35., 113 (1957).
Chao, K. C., and J. D. Seader, A.I.Ch.E. J., 7_ (4),
598 (1961).
Clerk, J., Chem. Eng., 71_ (21), 232 (1964).
Colburn, A. P., Trans. Am. Inst. Chem. Engrs., 35, 211
(1939).
Daily, J. W., and R. E. Nece, Trans. Am. Soc. Mech.
Engrs., J. Basic Engr., D82(l), 217 (1960).
Donahue, D. J., and F. E. Bartell , J. Phys. Chem.,
56., 480 (1952).
Edmonds, R. G., and 6. F. Jenkins, Chem. Eng. Prog.,
50. (3), 111 (1954).
Elzinga, E. R., Jr., and J. T. Banchero, Chem. Eng.
Progr. Symp. Ser., 55. (29), 149 (1959).
Environmental Protection Agency, "Methods for Chemical
Analysis of Water and Wastes, 1971," p. 17, Storet
No. 00340 (1971).
Faith, W. L., D. B. Keyes, and R. L. Clark, "Industrial
Chemicals," 3rd Ed., John Wiley & Sons New York
(1965).
Fair, J. R., Gas-Liquid Contacting, in Perry, R. H., and
C. H. Chilton, "Chemical Engineers' Handbook," 5th
ed., McGraw-Hill Book Company, New York (1973).
Fleck, R. N., and J. M. Prausnitz, Ind. Eng. Chem.
Fundam. , 7_, 174 (1968).
Fox, R. D., Chem. Eng., 80 (18), 72 (1973).
Geankoplis, C. J., and A. N. Hixon, Ind. Eng. Chem.,
4£, 1141 (1950).
278
-------
Gier, T. E., and J. 0. Hougen, Ind. Eng. Chem., 45, 1362
(1953). ~~
Guthrie, K. M., Chem. Eng., 76_ (6), 114 (1969).
Handles, A. E., and T. Baron, A.I. Ch.E. J., 3_, 127 (1957)
Hanson, C., Chem. Eng., 7j± (18), 76 (1968).
Harriott, P., and H. Wiegandt, A.I.Ch.E. J., 10 (5),
755 (1964).
Hartland, S., and J. C. Mecklenburgh, Chem. Eng. Sci.,
21, 1209 (1966).
Hazelbeck, D. E., and C. J. Geankoplls, Ind. Eng. Chem.
Fundam., 2 (4), 310 (1963).
Heertjes, P. M., and L. H. deNie, Mass Transfer to Drops,
in C. Hanson (ed.), "Recent Advances in Liquid-
Liquid Extraction," Pergamon Press, Oxford (1971).
Helpinstill, J. G., and M. van Winkle, Ind. Eng. Chem.,
Process Des. Develop., 7_ (2), 213 (1968).
Henton, J. E., and S. D. Cavers, Ind. Eng. Chem. Fundam.,
9. (3), 384 (1970).
Henton, J. E., L. W. Fish, and S. D. Cavers, Ind. Eng.
Chem. Fundam., ]_2 (3), 365 (1973).
Herz, R. K., M. S. Dissertation, University of California,
Berkeley, 1972.
Hildebrand, J. H., J. M. Prausnitz, and R. L. Scott,
"Regular and Related Solutions," Van Nostrand
Reinhold Company, New York (1970).
Hinze, J. 0., A.I.Ch.E. J., 1, 289 (1955).
Hughmark, G. A., Ind. Eng. Chem. Fundam., 6 (3), 408
(1967).
Ingham, J., The Study of Longitudinal Mixing in Liquid-
Liquid Contactors," in C. Hanson (ed.), "Recent
Advances in Liquid-Liquid Extraction," Pergamon
Press, Oxford (1971).
279
-------
Jacques, G. L., and T. Vermeulen, Univ. Calif. Rad. Lab.
Report No. UCRL-8029 (1957).
Jones, H. R., "Environmental Control in the Organic and
Petrochemical Industries," Noyes Data Corporation,
Park Ridge, N.J. (1971).
Jones, R. H., Chem. Eng. Progr. Symp. Ser., 67 (107),
531 (1971). ~~
Kaiser, H. R., Sewage and Ind. Wastes, 27. 311 (1955).
Kiezyk, P. R., and D. MacKay, Can. J. Chem. Eng., 49,
747 (1971).
Kiezyk, P. R., and D. MacKay, Can. J. Chem. Eng., 51,
741 (1973). ~~
King, C. J., A.I.Ch.E. J., IjO, 671 (1964).
King, C. J., Ind. Eng. Chem. Fundam., 4 (2), 125 (1965).
King, N. B., U. S. Patent 3,527,699 (Sept. 8, 1970).
Klee, A. J., and R. E. Treybal, A.I.Ch.E. J., 2, 444
(1956).
Krishnamurty, V. V. G, and C. V. Rao, Trans. Indian
Inst. Chem. Engrs., 6_, 153 (1954).
Kronig, R., and J. C. Brink, Appl. Sci. Research,
A2, 142 (1950).
Landau, J., and R. Houlihan, Can. J. Chem. Eng., 52,
338 (1974).
Lapidus, L., "Digital Computation for Chemical
Engineers," McGraw-Hill Book Company, New York
(1962).
Lewis, W. L., J. Water Poll. Contr. Fedr., 40 (5),
869 (1968).
Logsdall, D. H., J. D. Thornton, and H. R. C. Pratt,
Trans. Instn. Chem. Engrs., 35, 301 (1957).
280
-------
Marks, G. W., J. Acoust. Soc. Am., 4T_ (1), 103 (1967).
Matsuura, T., and S. Sourirajan, J. Appl. Polymer Sci.,
T_7, 3683 (1973).
McSwain, C. V., and L. D. Durbin, Separation Sci., 1,
677 .(1966).
McKinney, R. E., J. Water Poll. Cont. Fedr., 39 (3),
346 (1967).
Minard, G. W. and A. I. Johnson, Chem. Engr. Progr.,
48 (2), 62 (1952).
Misek, T., Coll. Czech. Chem. Commun., 28, 426 (1963a).
Misek, T., Coll. Czech. Chem. Commun., £8_, 1631 (1963b).
Misek, T., Coll, Czech. Chem. Commun., 29, 2086 (1964).
Misek, T., Paper presented at Chisa II Congress,
Marienbad, Czechoslovakia (1965).
Misek, T., and V. Rod, Calculation of Contactors with
Longitudinal Mixing, in C. Hanson (ed.), "Recent
Advances in Liquid-Liquid Extraction," Pergamon
Press, Oxford (1971).
Misek, T., and B. Rozkos, Intern. Chem. Engr., 6 (1),
130 (1966).
Miyauchi, T., and T. Vermeulen, Ind. Eng. Chem. Fundam.,
2 (4), 304 (1963).
Miyauchi, T., H. Mitsutake, and I. Harase, A.I.Ch.E. J.,
]_2, 508 (1966).
Mugele, R. A., and H. D. Evans, Ind. Eng. Chem., 43,
1317 (1951).
Olney, R. B., A.I.Ch.E. J., 1_0, 827 (1964).
Peters, M. S., and K. D. Timmerhaus, "Plant Design and
Economics for Chemical Engineers," 2nd ed.,
McGraw-Hill Book Company, New York (1968).
Pfeil, E., H. Stache, and F. Lomker, Ann., 623, 74 (1959)
281
-------
Pierotti, G. J., C. A. Deal, and E. L. Derr, Ind. Eng.
Chem. , 51, 95 (1959).
Pratt, H. R. C., Ind. Eng. Chem., Process Des. Develop.,
14 (1), 74 (1975).
Reid, R. C., and T. K. Sherwood, "The Properties of
Gases and Liquids," 2nd ed., McGraw-Hill Book
Company, New York (1966).
Reman, G. H., and J. G. van der Vusse, Ge"nie Chim.,
74., 106 (1955).
Reman, G. H., Chem. Eng. Progr., 62_ (9), 56 (1966).
Rhodes, E. 0., U. S. Bur. Mines, I. C. 7490, (February,
1949).
Rod, V., British Chem. Eng., 9_ (5), 301 (1964).
Ruby, C. J., and J. C. Elgin, Chem. Eng. Progr. Symp.
Ser., 51 (16), 17 (1955).
Scheele, G. F., and B. J. Meister, A.I.Ch.E. J., 14,
9 (1968).
Scheibel, E. G., Ind. Eng. Chem., 4J5, 2007 (1954).
Sleicher, C. A., A.I.Ch.E. J., 5_, 145 (1959).
Sleicher, C. A., A.I.Ch.E. J., 6_, 529 (1960).
Steinour, H. S., Ind. Eng. Chem., 3J5, 618 (1944).
Stemerding, S., E. C. Lumb, and J. Lips, Chemie Ingr.
Tech. , 35., 844 (1963).
Stobaugh, R. B., Hydrocarbon Process., 4_5_ (1), 143
(1966).
Strand, C. P., R. B. Olney, and G. H. Ackerman, A.I.Ch.E
J., £ (2), 252 (1962).
Strausser, J. R., and R. S. Kurland, U. S. Patent
3,507,782 (April 21, 1970) .
282
-------
Sugar, J. W., and R. A. Conway, J. Water Poll. Contr.
Fedr., 40 (9), 1622 (1968).
Thornton, J. D., and H. R. C. Pratt, Trans. Instn.
Chem. Engrs., 31 , 289 (1953).
Thornton, J. D., Chem. Eng. Sci . , _5, 201 (1956).
Toropov, A. P., Zhur. Obshchei Khim., 2_6, 1285 (1956).
Treybal , R. E., "Liquid Extraction," 2nd. ed., McGraw-
Hill Book Company, New York (1963).
Treybal, R. E., Chem. Eng. Progr., 6_2 (9), 67 (1966).
Treybal, R. E., Liquid-Liquid Systems, in Perry, R. H.,
and C. H. Chilton, "Chemical Engineers' Handbook,"
5th ed., McGraw-Hill Book Company, New York (1973)
Tupholme, C. H. S., Ind. Eng. Chem., £5 (3), 303 (1933)
Union Carbide, "Chemicals and Plastics Physical Proper-
ties," 1973-74 ed., Union Carbide Corporation,
New York (1973).
Veldhuis, M. K., R. E. Berry, C. J. Wagner, Jr., E. D.
Lund, and W. L. Bryan, J. Food Sci., 3_7, 108
(1972).
Weast, R. C., Editor, "Handbook of Chemistry and
Physics," 51st ed., The Chemical Rubber Company,
Cleveland, Ohio (1970).
Weaver, D. G., and W. A. Biggs, Jr., Ind. Eng. Chem.,
51 (10), 773 (1961 ).
Weaver, R. E. C., L. Lapidus, and J. C. Elgin, A.I.Ch.E
J. , 5_, 533 (1959).
Weimer, R. F. and J. M. Prausnitz, Hydrocarbon
Process., 44 (9), 237 (1965).
Weller, R., H. Schuberth, and E. Leibnitz, J. Prakt.
Chemie, 2_1_ (4), 234 (1963).
Westerterp, K. R., and P. Landsman, Chem. Eng. Sci.,
17, 363 (1962).
283
-------
Westerterp, K. R., and W. H. Meyberg, Chem. Eng. Sci.,
]2., 373 (1962).
Wilburn, N. P., Ind. Eng. Chem., Fundam., 3 (3), 189
(1964).
Wilson, G. M., J. Am. Chem. Soc., 8J[, 127 (1964).
Witt, P. A., Jr., and M. C. Forbes, Chem. Eng. Progr.,
67 (10), 90 (1971).
Won, K. W. , Ph.D. Dissertation, University of California,
Berkeley (1974).
Won, K. W., and J. M. Prausnitz, A.I.Ch.E, J., 20,
?187 (1974). —
Won, K. W., and J. M. Prausnitz, J. Chem. Thermodynamics,
In press (1975).
Wurm, H.-J,, Gluckauf, 104 (12), 517 (1968). A reprint
of this article may~~b~e obtained by writing to
American Lurgi Corp., 235 East 42nd Street,
New York, N.Y. 10017.
Young, E. F., Chem. Eng., 64_ (2), 241 (1957).
Zenz, F. A., Petrol. Refiner, 36. (8), 147 (1957).
284
-------
Section XIII
Pub1icat ions
D. S. Abrams and J. M. Prausnitz, "Distribution of
Phenolic Solutes Between Water and Nonpolar
Organic Solutes". J. Chem. Thermo. 7_ 61 (1975)
J. P. Earhart, C. J. King, J. M. Prausnitz, and K.
W. Won, "Solvent Extraction as an Industrial
Waste Water Treatment Process". Presented at
the AIChE National Meeting, Pittsburgh, Pa.,
June 5, 1974.
J. P. Earhart, K. W. Won, J. M. Prausnitz, and C. J.
King, "Removal of Phenolics from Industrial
Wastewater by Dual-Solvent Extraction".
Presented at the AIChE-CVG Joint Meeting,
Munich, Germany, September 1974. Also in
Proceedings .
K. W. Won and J. M. Prausnitz, "Distribution
Coefficients at High Dilution for Organic
Solutes Between Water and Isobutane or
Isobutylene". AIChE Journal 2_0 1187 (1974).
K. W. Won and J. M. Prausnitz, "Distribution of
Phenolic Solutes Between Water and Polar
Organic Solvents". J. Chem. Thermo. 7
661 (1975).
285
-------
APPENDIX A
BASIS FOR COST ESTIMATES
The estimates of investment and operating costs
included in this report have been completed on
a consistent basis. This consistency allows a
valid comparison of alternatives (e.g., steam
stripping and solvent extraction), but the absolute
value of costs can only be compared with quoted
costs from other sources when the estimating methods
of the other authors are also clearly recorded. The
method used herein makes extensive use of the capital
cost estimating methods of Guthrie (1969) and the
operating cost estimates of Peters and Timmerhaus (1968).
To include the effect of inflation, all costs are
corrected to December, 1973, by using the Marshall & Stevens
chemical process industry cost index. Precision
is ± 30%.
Estimated Total Plant Investment
To estimate the capital cost of an installed
plant, the "module" technique of Guthrie was used.
A set of assumptions (listed in the text) was
established which allowed the heat and material
balances to be determined. On the basis of a second
set of assumptions (in lieu of a detailed optimization),
286
-------
the major pieces of on-site equipment (extraction and
distillation columns, heat exchangers, tanks, pumps,
etc.) were sized to perform the necessary stream
transformations. Then the cost diagrams of Guthrie
were used to determine the cost of each major item
of equipment. All equipment was assumed to be
constructed of carbon steel, but a 10% increase
in cost was arbitrarily added for equipment which
would be in contact with the waste water to account
for slightly more severe corrosion. For items of
equipment whose cost was not given by Guthrie, cost
data from other sources were used and were treated
to make them as consistent with Guthrie as possible.
Guthrie gives factors for each type of equipment
which when multiplied by the cost of the item of
equipment provides estimates for the auxiliary equipment
(piping, concrete, instruments, etc.) cost, the labor
costs for material erection and equipment installation,
and the indirect (freight, construction overhead,
engineering, etc.) costs. Typical total factors
range from 3.29 to 4.23 for process equipment. The
resulting installed equipment cost was then increased
by 18% for contingency plus contractor's fee. No
costs were included for site development or industrial
buildings. The off-site investment for steam
capacity, power generation, and cooling tower facilities
(Guthrie, 1969) was found to be negligible for the
cases considered. The working capital, which would
include solvent and recovered pollutant inventories,
was assumed to be negligible.
287
-------
In the text, the total of the above costs
is listed as the "Total Plant Investment." The
percentage of this total which is associated with
each item of on-site or off-site installed equipment
is also listed. The details of individual pieces
of equipment are discussed only in unusual situations.
Annual Operating Costs
The plant is assumed to operate for 8000 hours/
year. The following items are included in the operating
costs (Peters and Timmerhaus, 1968):
1. Chemicals - cost of solvent to make up for
losses in the product water and in the recovered
pollutant was included at the market price of
the solvent.
2. Utilities - cost of steam, cooling water, and
electricity were estimated by increasing the
prices given by Peters and Timmerhaus by 3%
per year for inflation. The following values were
used:
Steam: Exhaust $0.40/1000 Ib.
100 psig $0.80/1000 Ib.
Cooling Water $0.05/1000 gal.
Electricity $0.02/kw-hr.
3. Maintenance and Repairs - annual cost was taken
as 6% of Total Plant Investment.
4. Operating Supplies - annual cost was taken as
15% of Maintenance and Repairs.
5. Depreciation - annual cost was taken as 8% of
Total Plant Investment.
6. . Insurance and Taxes - annual cost was taken as
3% of Total Plant Investment.
288
-------
7. Return on Investment - cost was assumed to be
8% of Total Plant Investment.
The above costs can be summarized in a single equation
as follows:
Annual Operating Costs = Chemicals + Utilities
+ °-259
In addition to these operating costs, five more
items associated with operating labor and laboratory
charges should be listed (Peters and Timmerhaus, 1968).
These five items are not included in the costs
reported in this dissertation, not because they are
unimportant, but because an accurate method of estimating
them is not apparent. To estimate labor costs, man-
power requirements must be estimated, and these vary
greatly from company to company. Bollen (1974)
suggests that 1/4 shift would probably be sufficient
to operate this type of waste treatment process.
To estimate laboratory charges, the types of
waste water analyses, their cost, and their
frequency must be estimated. An arbitrary cost
of $1.00/hr seems reasonable. Based on these estimates,
the additional five items would be as follows:
8. Operating Labor - cost assumed to be for one-fourth
man at $4.00/hr. to give a constant value of
$8,000/year.
9. Supervisory and Clerical Labor -cost was taken as
15% of operating labor.
10. Plant Overhead - cost was taken as 60% of Operating,
Supervisory and Clerical Labor.
289
-------
11. Administration - cost was taken as 50% of Operating
Labor.
12. Laboratory - assumed to be $1.00/hr. mainly for
analyses to give a constant $8,000/year.
If included, these items would result in an additional
annual -cost of $26,700, or for a 100 GPM waste water
stream, $0.56 per thousand gallons treated. This is
certainly an appreciable additional cost.
In the text, the total operating cost is listed
usually as a value per thousand gallons of waste water
treated. The individual costs are also listed for
chemicals, utilities, and capital equipment costs
(Items 3, 4, 5, 6 and 7).
290
-------
APPENDIX B
HYDRODYNAMICS, AXIAL MIXING AND MASS TRANSFER IN
CONTINUOUS EXTRACTORS
Spray column extractors.
For spray column extractors, the design procedure
which was recommended by Treybal (1973) appears
to be the best available in terms of hydrodynamics
and mass transfer rates. Treybal's recommendation
relies substantially on the article by Hughrnark
(1967). Several recent articles which concern
axial mixing in spray column extractors were not
included in Treybal's review, and these are discussed
below. Most of the equations appear in Treybal's
review (1973), and many of them are not repeated
here.
The hydrodynamics of a spray column extractor
is largely determined by the means of generating
dispersed-phase droplets. The materials of con-
struction must be chosen so that the distributor
plate is preferentially wet by the continuous
phase when a flat plate with holes drilled through
it is to be used. For dispersing an organic
phase into water, stainless steel can be used;
for dispersing water into a continuous organic
phase, a plastic distributor should operate well.
When this precaution is taken, the correlation
of Scheele and Meister (1968) permits the calculation
291
-------
of drop volume for cases when the velocity of
dispersed phase flowing through the drilled holes
is less than the jetting velocity. The droplet
diameter may then be calculated by assuming that
the droplets are spherical. Use of this correlation
requires data on the interfacial tension between
solvent and water, the density of both phases,
and the viscosity of the continuous phase, as
well as a choice of orifice diameter and the dis-
charge velocity. The correlation was developed
using pure fluid phases, so the effect of simul-
taneous mass transfer is unknown (Treybal, 1973).
The error introduced by using pure-phase properties
will be most serious when the water-phase is
dispersed at the end of the column where pollutant
concentrations are highest.
For the purpose of developing a design pro-
cedure for describing flooding and hold-up, it
is useful to define a slip velocity, V , and a
characteristic velocity, V^, as follows (now
using terminology of continuous-phase and dispersed-
phase rather than solvent-phase and water-phase
since the equations apply for either phase being
dispersed):
v^ V,.
V = V. (1 - *) = -S- + —£— (Bl)
s k • * 1-4,
where = dispersed phase (fractional) hold-up.
Thornton (1956) found that V. was practically
constant at conditions near flooding and assumed
that if V is held constant and V. is increased
292
-------
by increasing , then 8Vd/3 = 0 at the flooding
point (Thornton and Pratt, 1953). Solving the
second equality in equation (Bl) for Vd and setting
the partial derivative to zero leads to the
following three equations (subscript f refers to
the flooding point):
Vcf - Vk (1 - 2 f) (1 - f)2 (B2)
Vdf = 2Vk(j>f2 (1 -
-------
a = (B5)
P
To calculate a, it is necessary to estimate t)>
at conditions well below the flooding velocities.
If V. were constant over a wide range of , then
equation (Bl) could be used. However, Treybal
(1973) recommends using a method developed by
several authors (Beyaert, et al . , 1961; Weaver,
et al., 1959) which is based on the finding that
the ratio of slip velocity, Vg, to the terminal
velocity of a single particle or droplet through
a quiet fluid, V., is a unique function of the
hold-up for all gas-solid, liquid-solid, and liquid-
liquid systems. This allows the correlation of
Zenz (1957) for the fluidization of solids to
be used to predict hold-up in liquid-liquid systems
since a correlation is available for predicting
the terminal velocity of a liquid droplet (Klee
and Treybal, 1956). Hughmark (1967) presents
an equivalent method which is slightly easier
to use.
Although the importance of axial mixing
in solvent extraction was first noted following
measurements of the continuous-phase concentration
profile in a spray column (Geankoplis and Hixon,
1950), there have been few quantitative studies
of axial dispersion in this type of contactor.
For this reason Treybal (1963), basing his consider-
ations on experimental measurements of continuous-
and dispersed-phase concentration profiles (Cavers
294
-------
and Ewanchyna, 1957; Gier and Hougen, 1953),
suggested that the continuous phase be considered
as completely mixed (Pec = 0) and that the dispersed
phase be considered to move in plug flow (Pe. = °°) .
For normal operation, the Pe^ = <=° assumption still
appears to be valid, but the experimental work of
Brutvan (1958), Hazlebeck and Geankoplis (1963),
and Henton and Cavers (1970) provides a basis
for making a better approximation of the value of
pv
The fact that the continuous phase often be-
haves as though it were completely mixed is the
main reason spray columns are seldom used commer-
cially. Only in cases where no more than a few
equilibrium stages of separation are required or
when the resistance to mass transfer in the contin-
uous phase is unimportant (e.g., water dispersed
when E is very large) does a spray column show
promise as a commercial extraction device. These
conditions are seldom encountered in a regenerable
solvent extraction process. For this reason we
will be satisfied if the literature only provides
a method of correcting our miniplant spray column
data for axial mixing. In the uncommon case
where a spray column can be commercially feasible,
a conservative estimate of its performance can
be made by assuming Pec = 0 for the large-scale
spray column.
Brutvan (1958) measured the continuous-phase
axial dispersivity, er, as glass beads dropped
U
through an ascending, continuous water phase by
295
-------
using a step function of tracer. He used four
diameters of glass beads from 0.12 to 0.24 inch
and three column diameters from 1 to 2 inches
and found that e decreased with increasing V ,
\f C
increased with increasing V,, decreased as particle
size increased and increased with increasing column
diameter, D. Hazlebeck and Geankoplis (1963) also
used a step-wise change in concentration of a
water soluble tracer to measure EC while methyl
isobutyl ketone droplets rose through a descending,
continuous water phase. They found that e increased
linearly with increasing V and was independent of
L»
V^; they used only one droplet size and one column
diameter of 1.41 inch. Henton and Cavers (1970)
measured the turbulent mixing contribution to e
in a system where methyl isobutyl ketone droplets
rose through a descending water phase by making
a steady injection of water soluble tracer and
measuring its concentration upstream (with respect
to the continuous phase) from the injection point.
They made extensive measurements in a 1.5-inch
diameter column and a limited number in a 3.0-inch
column. They found that ec was independent of
Vc (intermediate compared to previous studies),
decreased with increasing Vc (opposite to the
result of Brutvan), increased with increasing
droplet size (opposite to Brutvan), and increased
with increasing D (in agreement with Brutvan).
It is clear from this brief discussion that the
present understanding of axial dispersion in
spray columns is far from complete.
296
-------
Despite the lack of agreement among experi-
mental results, Henton, et al. (1973) have attempted
to develop several theoretical approaches based
principally on their own extensive data (Henton
and Cavers, 1970). They suggested an analogy
between the flow of a single-phase fluid through
a packed bed with a stationary frame of reference
and the two-phase flow in a spray column with the
frame of reference moving with the droplets.
Their treatment leads to the following relationship
between a local Peclet number, Pe * (based on
C
droplet diameter, d , and slip velocity, V ), and
r •>
the di spersed-phase hold-up, :
V d
Pec* = -^—^ = 2.36 1/3 (B6)
ec
The numerical constant was calculated by assuming
that the droplets on the average are located in
a lattice arrangement, but it is actually the
result of fitting their experimental data from
their 1.5 inch column since several lattice arrange-
ments were assumed and the one best fitting the
data was accepted.
Equation (B6) does successfully predict an
increase of e with increasing droplet diameter
{*
and with decreasing VG as found by Henton and
Cavers (1970), but it does not predict an increase
in backmixing when the column diameter is increased
since the hold-up is not a function of column
diameter when a single system is studied at the
same superficial velocities of the phases.
However, both Brutvan (1958) and Henton and Cavers
297
-------
(1970) determined experimentally that e increased
\*
with increasing column diameter. On the basis of
a multiple linear regression analysis of logarithms
of several dimension!ess variables, Henton, et al.,
(1973), using their data for 1.5 and 3.0 inch
columns, have estimated that Pe * increased linearly
-282
with (column diameter)" ' . However, from visual
observation of the differences in flow patterns,
they believe the increase in CQ in going from the
small column to the larger column is primarily
the result of large-scale mixing being present
only in the larger column.
In the present study we need an estimate of
e in a 1.0 inch diameter column operating with
\f
superficial velocities and droplet sizes similar
to those studied by Henton and Cavers (1970).
We have chosen to use equation (B6) to make that
estimate and have thus assumed that no large-scale
mixing was present in either our 1.0 inch column
or in the 1.5 inch column used to develop the
correlation. Had we used the factor (column
2 82
diameter)" to correct the prediction of equation
(12), the values of EC used to interpret our data
would have been about 1/3 as large.
Treybal (1963) recommends that for analysis
of mass transfer rates in a spray column, there are
three distinct operations: (1) drop formation and
release, (2) drop rise (or fall) through the
continuous phase, and (3) drop coalescence at
the main interface. In a recent review Heertijes
and deNie (1971) further divide the stages of
298
-------
mass transfer operations. The general approach
is to estimate the mass transfer during drop for-
mation by one of several theoretical relationships
and to assume that the product of the overall
dispersed-phase mass transfer coefficient, the
drop surface area, and the contact time is the same
during drop coalescence as during drop formation
(Treybal, 1973). Once these estimates have been
used to correct for the "end effects," then the
mass transfer during drop movement through the
partially backmixed continuous phase can be cal-
culated using the dispersion model.
Treybal (1973) shows that the various theories
for estimating the dispersed-phase mass transfer
coefficient during drop formation, k^^, are of the
form
?7> (B7)
where pd = density of dispersed phase,
6f = total time for drop formation (contact time),
D, = solute diffusivity in dispersed phase.
When this mass transfer coefficient is used to
calculate the total mass of solute transferred
during the time of drop formation, based on the
interfacial area of the fully formed drop,
A , and on the dispersed-phase concentration of the
feed solvent, then the constant, G13, calculated
from various theories ranges from 0.857 to 3.43,
and most of the available data are correlated
with G13 in the range from 1.3 to 1.8 (Treybal,
299
-------
1973). On the assumptions that the mechanism
for mass transfer in the continuous phase is the
same as for the drop phase and that the two mass
transfer resistances are additive, the following
equation predicts the product of overall dispersed-
phase mass transfer coefficient, drop surface area
and contact time when G^3 is chosen as the average
of experimental values:
1 KK A f D ft /TT ^
1 • l (true for volatile
solvent dispersed), the single drop equation predicts
300
-------
a value of kc up to three times larger (Hughmark,
1967) than the preferred value from the Ruby and
Elgin equation (1955).
Treybal (1973) recommends an equation based
on Hadamard-1ike internal drop circulation for
the calculation of k^ for circulating drops.He
also suggests that this expression be corrected
for the presence of a continuous-phase resistance
by using a result which Elzinga and Banchero (1959)
developed for heat transfer. However, several
apparent errors were introduced when Treybal (1973)
converted their heat transfer result to this mass
transfer situation.
In the table presented by Treybal (1973) a
dimension!ess parameter, F = k d /D , was suggested
w |J C
to determine the magnitude of the correction to simple
additivity of resistances. However, the correct
parameter to use when applying this correction
should have been F = k^p/K^d for the two-phase
mass transfer situation. The correct factor, F,
is proportional to the factor, R = kc/Kcjkd' used
by King (1964; 1965) when the usual approximation
for long exposure times is made to the equation
for kd (Treybal, 1963). Also, King (1965)
showed that the correction to additivity due to
an interaction of the two individual-phase mass
transfer coefficients results in an increase in
the overall mass transfer coefficient. As suggested
by Treybal, the Elzinga and Banchero (1959)
correction to k. should actually have been equal
to the overall mass transfer coefficient based on
the dispersed phase. The treatment used in this
301
-------
project assumes that simple additivity is a
sufficiently accurate approximation. This approach
is equivalent to using the equation of Kronig and
Brink (1950).
For drops larger than a critical size,
oscillations in shape occur as the drops fall
through a continuous phase. This critical size
can be predicted by the terminal velocity correla-
tion of Klee and Treybal (1956). Oscillations
not only result in a lower drop terminal velocity,
but they also increase the mass transfer coefficient.
Treybal (1973) recommends a calculation procedure
for k. in an oscillating drop which is based
on the surface stretch theory of Angelo, et al.
(1966; 1968). When coupled with the assumption of
additivity of mass transfer resistances and the
assumption that the same surface stretch mechanism
determines the magnitude of kc, this equation
leads to an estimate of K for oscillating drops.
O C
Rotating disc contactors. A typical RDC
(see Figure 10 in main text) consists of (1) a series
of thin discs attached to a central rotating shaft,
(2) a series of thin stator plates attached to
the vessel walls and located vertically at points
midway between the discs, (3) a variable speed
drive, and (4) inlet and outlet lines (shown for
solvent dispersed with a unique solvent inlet line).
The design procedure for an RDC which was de-
scribed by Treybal (1963) should more precisely
be called a scale-up procedure since the equations
contain "constants" which have been observed to
302
-------
vary over considerable ranges but which are generally
found to remain constant when changes involve
only the scale of the device. Since 1963, Misek
has published numerous contributions to the under-
standing of an RDC. Although the description to
follow is based on the treatment by Treybal , the
more recent results of Misek are also introduced
where applicable.
In considering the hydrodynamics of an RDC,
it must be realized that an additional variable
is introduced in comparison to the spray column
extractor. With an RDC the rotational speed of
the discs has a fundamental influence on the
performance, just as the distributor had on the
performance of the spray column. Photographic
studies of the drops formed at a spray column
distributor plate (Scheele and Meister, 1968)
have shown that the majority of interfacial area
results from drops which have a very narrow range
of diameters. Therefore, it is reasonable to
model the spray column in terms of a monodisperse
droplet size distribution. Photographic studies
of the droplet size distribution in an RDC
(Olney, 1964; Misek, 1963a) have shown that a
much wider range of droplet diameters exist.
Nevertheless, the following discussion illustrates
a moderately successful approach to modeling the
RDC in terms of an average drop diameter.
Daily and Nece (1960) showed that the torque
required to drive an enclosed rotating disc is
nearly a unique function of disc Reynolds number;
303
-------
they observed a slight influence of the ratio of
disc diameter to height of the enclosure. Earlier
Reman and van der Vusse (1955) developed an equiv-
alent correlation for predicting the power input
per compartment, P, for an RDC contactor. Their
results were given in the form of a graph of the
power number, Po, vs. the disc Reynolds number,
Re, defined as follows:
(B9)
r U -
Re =
pc "'V
di2 Npc
yc
(BIO)
where N = rotational speed of the discs
d. = diameter of the discs.
This correlation can be used to estimate the size
of drive required (the power requirement is usually
small for waste water treatment), but Equation (B9)
can also be used for scale-up when a power per
unit liquid mass, P/M, is defined as follows:
4P N3 d.5
]>7M = — ° - L- (Bll)
ir H D*
where H = distance between stators
D = column diameter.
One basis for scale-up is to hold P/M constant as
the dimensions are increased.
304
-------
Several authors (Olney, 1964; Misek, 1963a)
have studied the droplet size distribution in
the stirred region of an RDC with the conclusions
that, as predicted by the theory of Hinze (1955),
there is a maximum droplet size and that the
droplet size distribution follows the equation
of Mugele and Evans (1951). Strand, et al . (1962)
assumed that an average droplet size, d ,
could be used for the calculation of mass transfer
rates as well as for flooding and hold-up calcu-
lations, and that this average droplet diameter
was a constant fraction of the maximum drop diameter
predicted by the Hinze equation. This leads
to the following equation for d :
d = G18 (^-) ' (FT!)'0-4 (B12)
^ pc
where a = interfacial tension.
Misek (1963a) has presented a more detailed analysis
in which he considered that there are two regions,
one laminar and the other turbulent, and that the
average drop diameter is a different function
of N for these two regions. Olney (1964) discussed
the limitations of the assumption that a single
average drop diameter can be used for estimating
both mass transfer and hydrodynamics.
Strand, et al. (1962) have developed a
series of relationships which allow the constant
G,g to be evaluated from hold-up measurements.
305
-------
They assumed that the characteristic velocity,
V^i defined by equation (Bl) is equal to the terminal
drop velocity, Vt, multiplied by a constriction
ratio, CR, which is the minimum of (1) the area
between the disc and the column, (2) the area
within the stator hole, and (3) the area of the
frustum of a cone from the stator to the adjoining
disc, where each area is divided by the cross-
sectional area of the column to form a ratio.
This statement can be expressed as follows:
Vk " Vt CR
(B13)
C R = m i
1 - (d./D)
(ds/0)
di
D
•s - di
D
where d = diameter of hole in stator.
Vt can be calculated from CR and Vk by using
equation (Bl) and experimental hold-up data; then
the relationship between V. and d (Klee and Treybal
P
1956) discussed for spray column extractors is
used to determine d . Finally G,p is evaluated from
equation (B12). This procedure assumes that for
a given system at a given rotational speed, the
hold-up is defined by the following with a constant
306
-------
value of V^:
Vs = Vk (1 - 4>) (B15)
Misek (1963b) has had considerable success using
an alternative equation defining a different
characteristic velocity, V , as follows:
Vs = Vn (1 - 4.) exp (
-------
observations during mass transfer studies could
not be predicted by this method (Strand, et al . ,
1962). Similar difficulties are described with
respect to the present study in Section VIII.
Despite the limitations of any method which
models an RDC in terms of a monodisperse droplet
size distribution, the above correlation suggests
a rational basis for scale-up. Equation ((B12)
suggests that if P/M is held constant during scale-
up, then the average droplet size, d , (and hope-
fully the droplet size distribution) will remain
the same. Equations (B13) and (B14) suggest that
if the ratios of d /D, d./D, and H /D are held
j I \f
constant, then CR and V^ (since Vt is fixed by
d ) will remain constant. Finally if V- and V
p d c
are held constant, then V and will remain constant
during scale-up. As discussed later, the importance
of axial mixing does not remain constant during
this method of scale-up, and its effect must be
accounted for separately.
Flooding in an RDC is a much different
phenomenon than flooding in a spray column extrac-
tor or any other device which produces a narrow
range of droplet sizes. Flooding in an RDC
actually starts with the entrainment of the smallest
droplets along with the continuous phase. As
the value of V increases, a larger and larger
fraction of the dispersed phase is entrained
until finally the maximum stable droplet size
is entrained and the column is fully flooded.
308
-------
Nevertheless, several methods of estimating "the
flooding velocity" based on an average drop
diameter have been reported.
The flooding of an RDC can be estimated from
the results of any correlation of V. using equations
(B2), (B3), and (B4). Once G^g is evaluated from
one experimental measurement as described above,
then Vk can be determined. Logsdail, et al . (1957)
have given another correlation for V^ that is
useful for estimating the flooding point. The
relationship based on V in equation (816) has
also been analyzed by assuming 3V./9 = 0 at
the flooding point (Misek, 1963b).
In an RDC axial mixiji^ takes place in both
the dispersed and continuous phases. Much of the
work reported on axial mixing in RDC extractors
as well as all of the discussion to follow has
been developed in terms of the dispersion model.
The mechanism for axial mixing in the dispersed
phase is quite complex as it involves drop coales-
cence and redispersion, the distribution of drop
velocities due to their size distribution, and
radial variation in drop velocity due to vessel
geometry and rotating disc speed. Several approaches
to correlating the available data are discussed,
but these only provide an estimate of e.. Axial
mixing in the continuous phase is much better
understood, and e can be predicted with reasonable
accuracy.
A number of studies have dealt with the
measurement of e in the absence of a dispersed
phase. Westerterp and Landsman (1962) determined
309
-------
e with water as the continuous phase by using
a step injection of salt tracer, and found EC
could be correlated as the sum of two terms as
follows:
e
c 2C
GV (B17)
where G-| and 62 are constants.
Subsequent experiments (Westerterp and Meyberg,
1962) using a steady injection of salt tracer
and measuring its concentration upstream showed
that the first term in equation (B17) is due to
turbulent mixing (which is the only contribution
measured in the steady tracer injection) and
the second term is due to channeling (i.e., non-
uniform axial velocities). Strand, et al . (1962)
also determined the separate effects on the total
axial dispersion coefficient and correlated their
data as follows:
ec /d1 N
c = 0.5 + 0.09I n
VcHc \ Vc
(B18)
where H = height of one compartment.
This result shows the effect of column dimensions
and is of the same form as equation (B17), but it
implies that if d = dj, the turbulent mixing
term tends to zero which is unlikely. This objection
was later corrected, resulting in the recommen-
dation of Reman (1966) that the following expression
310
-------
be used:
—— = 0.5 + 0.012
(B19)
Other studies (Miyauchi, et al. 1966; Misek, 1965)
have led to similar correlations.
It should be pointed out that when using
the dispersion model we assume that all the various
mechanisms for mixing in the continuous phase
(and in the dispersed phase) can be estimated
by a single term of a form like that for molecular
diffusion (see Appendix C). One of the phenomena
this formulation can predict is the experimentally
observed jump in concentration at the continuous-
phase inlet. However, if we considered both
turbulent mixing and channeling in a more accurate
mathematical formulation, only turbulent mixing
would result in such a concentration jump.
Obviously there is an approximation involved
in treating both mechanisms as the sum of two
contributions to this single term.
Strand, et al (1962) also studied axial
mixing in the continuous phase when the dispersed
phase was in counter-current flow and found that
equation (B18) could be modified to account for
the dispersed phase by replacing V with V / (!-<(>).
L. v*
Making this same modification to equation (B19)
leads to the equation used in the present work
to estimate e , as follows:
311
-------
V!c
= 0.5 + 0.012 (1 - )
(B20)
Strand, et al . (1962) tentatively proposed
an equation similar to equation (B18) for estimating
ed as follows:
E ,
(B21)
They suggested that this expression would be best
at high rotational speeds where axial mixing of the
continuous phase is dominated by the turbulent
mixing effect and where the low inertia of the
small droplets of dispersed phase should cause
them to follow the continous phase fluctuations.
Stemerding, et al. (1963) found that ed did not
obey the form of equations (B17) through (B21), but
that ed was much larger than EC such that the
ratio Ej/e varied from about 100 to 1. At
flooding conditions, the droplet velocity is low
so ed/ec is approximately 1, whereas at normal
operation at about 80% of flooding, ed/e is about
10.
For the present study we have used equation
(B21) to estimate the dispersed phase axial mixing.
While we realize that this is very approximate,
there does not seem to be a better solution
available. As pointed out by Olney (1964), any
312
-------
attempt to develop a correlation of dispersed-
phase axial mixing while assuming a monodisperse
droplet size distribution has little hope for
success. It should be noted that the error
associated with using equation (B21) will be most
severe for solutes having a low extraction factor
and thus a major mass transfer resistance in the
dispersed, solvent phase.
The prediction of mass transfer rates in
an RDC, like that for the spray column, is very
approximate and requires experimental data to
judge between various models. Strand, et al .
(1962) suggested that the model for stagnant,
noncirculating drops and the model for fully
circulating drops should provide limits for the
expected mass transfer coefficients. They express
these limits as follows:
a. Stagnant drops: k. = 0.001 Vc (B22)
c s
kd
b. Circulating drops:
'4 D V ^1/2
2.D 0.00375
3d 1 + pd/pc
313
-------
Equation (B22) is empirical and was apparently first
introduced by Strand, et al. (1962). Equation (B23)
and the first term in equation (B25) are well
established for long exposure times, and the second
term in equation (B25) is due to circulation as
predicted by Handles and Baron (1957). Equation
(B24) is from penetration theory, and to be precise
should include a second term to account for diffusion
(Sherwood number equals 2). However, under normal
operation the magnitude of the second term is
usually less than the expected inaccuracy of the
first term, so neglecting it is not unreasonable.
Misek and Rozkos (1966) also considered a
third model for the continuous-phase mass transfer
coefficient after the theory of Calderbank and
Moo-Young (1961). Their expression for kc includes
the power per unit liquid mass, P/M, defined by
equation (Bll), and was expressed as follows:
kc = 0.13 (P7M)1/4 D2/3 (Pc/yc)5/12 (B26)
This correlation allows for turbulent mass transfer
and probably should be added to the terms for
diffusion and penetration. However, in their
original development Calderbank and Moo-Young
evaluated the coefficient while using this as the
only term, so it may be best to use it in the
form of equation (B26). In the study by Misek
and Rozkos (1966) it was concluded that equation
(B26) best fit their experimental data for the
314
-------
extraction of phenol from waste water. The expression
for kj used with k in equation (B26) could be
equation (B25) or the expression for circulating
drops discussed in conjunction with the spray
column.
To determine the overall mass transfer coef-
ficient in the RDC, the individual-phase mass transfer
coefficients were combined according to the assump-
tions of additivity of resistances. The use of this
assumption involves the approximation associated
with the interaction of individual resistances as
was discussed with respect to the spray column.
The correction to additivity due to this effect
was shown by King (1965) to result in an increase
in K of not more than about 20 percent above K
O C 0 C
from simple additivity. However, with the RDC
another effect is present due to the distribution
of drop sizes. Since the individual-phase mass
transfer coefficients depend on drop size in
such a way that R ( = k /K,k.) is not uniform over
the interface, a second correction to additivity
is needed. King (1964) has shown for several
mass transfer models that this second correction
results in a decrease in KQC which can be quite
large. Therefore, neglecting this second correction
(which can not be quantified easily) could result
in a design estimate of K from simple additivity
U L*
which is not conservative.
315
-------
APPENDIX C
DEVELOPMENT OF THE DISPERSION MODEL
Most of the information in the literature on
axial mixing in spray columns and in RDC extractors has
been analyzed in terms of the dispersion model. Many
of the studies have been discussed in Appendix B;
Vermeulen, et al. (1966) have reviewed the literature.
In this appendix the basic equations of the dispersion
model are developed in terms of quantities found useful
in the treatment of waste waters by solvent extraction.
The nomenclature used in this appendix differs
slightly from that of the remainder of this report
in that the subscript x refers to the aqueous phase and
the subscript y to the solvent phase. The equations
are developed for flows on a solute-free basis (i.e.,
F = Ib. solute-free water/hr) and for concentrations
J\
measured in weight ratios (i.e., X = Ib. solute/lb.
solute-free water). The use of weight ratios is
entirely equivalent to using weight fractions for the
dilute solutions encountered in most cases described
in this dissertation. However, in several cases
(notably the extraction of phenol with n-butyl acetate)
the concentration of solute in the solvent phase
can become large enough to introduce a significant
error. The use of weight ratios eliminates the problem
of a changing solvent-phase flow rate as long as the
two phases are immiscible. With the case of phenol
316
-------
being extracted by n-butyl acetate, K^ measured in
weight ratios shows less change with concentration
than does K. measured in weight fractions. The
effect of using weight ratios on the validity of
the assumption of a constant value of N is unknown.
U /\
Dispersion Model for Constant Kd .
Consider a section of column as shown in Figure
Cl . A is the column cross-sectional area, H the
total column height, and the solvent hold-up. The
terms for a material balance for the aqueous phase
over a differential slice can be expressed as follows:
a. Solute entering at h = h by convection and dispersion
FXX - px (1 - *) A ex «
b. Solute leaving at h = h + dh by convection and
di spersion
dh ) (FXX - px[l - *] A cx
c. Solute leaving x-phase by mass transfer
K a (X - X*) A dh
U /\
A material balance for the aqueous phase leads to the
fol 1 owi ng :
-Fx 3F+ Px'1 - *> A£X = Kox aA
-------
Fx'Xin FyYout
j :
h = 0
-Idealized droplet
interface
i_ _1_
n - n
h = h+dh
h = H
F Y F Y
rx» Aout ry1 T in
Figure Cl. Basis for Dispersion Model
318
-------
Assuming that the equilibrium may be described by
Y = K.X* and defining a dimensionless height, Z = h/H,
lead to the following:
IX
dZ
dX
Pe
ox
where
E ' KdFy/Fx
Nox = Kox a A H/Fx
Pe
FxH
x
x px(l - ) A
(X - X*)
(C4)
p...-
Equations (C3) and (C4) are second order, ordinary
differential equations in two dependent variables
(X and X*) and can be solved once two boundary condi-
tions are specified for each phase.
The usual boundary conditions express the
approximation that there is no axial mixing in the
inlet or outlet lines. This approximation is expressed
as follows:
319
-------
a. At h = 0 or 7 = 0:
FXX - px(l -
or
V -™ Y
Ain " *
FyYout
1 dX
" Fe~ HI
X
+ p d>A dY
FyYout
(C5)
or (C6)
d X*
b. At h = H or Z = 1 :
Fxxout
or (C7)
dZ
V + Py * A£y BT= FyYin
x* - x* + * (C8)
Xin * X +
A11 the necessary conditions have thus been met for a
mathematical solution to be developed.
It is convenient to define a dimensionless,
X-phase concentration as follows:
320
-------
X - X*
n = v Jy* (C9)
Ain " Ain
The solution to the equations will thus be of the
following form:
n(Z) = f(Z, E, NQX, Pex, Pey) (CIO)
When Z = 1, X takes on its value in the purified water
outlet stream giving the following equation:
n(D = f(E, Nox, Pex, Pey) (Cll)
When Z = 0, X takes on its value at the point within
the column where the water is discharged from the
inlet line.
Hartland and Mecklenburgh (1966) have shown
that the equations (C3) through (C9) may be solved
analytically in terms of the roots (q, , q^, q.,) of the
following equation:
q,3 + (Pey - Pex) q,.2 - (PexPey + PexNox + '^Si.) q.
Pex Pey Nox
These authors have set down the analytical solutions
for the general case and for a number of cases
where simplified equations are possible including
the following three cases most important in this
report:
1 . Pex and Pe finite; E = 1
2. Pex finite; Pe « °°; E ^ 1
3. Pe finite; Pew = ~; E = 1
x y
321
-------
The listing of a Fortran program developed to carry
out these calculations is reproduced at the conclusion
of this section.
It should be pointed out that in this develop-
ment we assume that E, N , Pe , and Pe are constant
O A A j
throughout the column, that there is radial homogeneity
at all levels, and that the idealized boundary condi-
tions are satisfied. A treatment where E is not
constant is discussed in the following section.
Mil burn (1964) has proposed other boundary conditions to
account for stagnant liquid in both end sections of the
extraction column.
322
-------
SUBROUTINE BETA
EET A < PEX • OEY ,E , NOX . ZE T A . K • Z • U . V )
C* EETA DETERMINES ZET/k WHEN PEX. PEV« E. ANO NOX ARE GIVEN. WHEN KM,
C**« NO CONCENTRATION PROFILE IS CALCULATED. WHEN Keg. THE PROFILE IS
C**« rALCU^ATEO AT 2(1) = zETACO.O), 2(2) * ZF.TA ( 0 . 1 ) « z ( .1) = ZETA(0.?>,
C#»# Z(4> * 2ETA(0.3)4 ZC5) » ZETAfO.5). 2(6) = ZETA(0.^),
C»«* Z(7> - ZETA(Oi8>, 7(8* = 7rTAfC.9), ANH Z(9) = 7crTA(I.O)
C»«* WHEN K = 3» ZETA«U(I». flN[? V(I) ARE CALCULATro FOR Ie)t!0l«
C*«* WHEN pEY .LT. C.O . CALCULATE FOR INFINITE PEY.
C*»* WHEN E »LT. 0.0 « CALC»-'LATF FOR INFINITE E.
C«»*
DlMF-NStON Z(9) tU< 1 01 ) ,V( 1OI )
REAL NOX.Kl,K2«K3
.O.O) GO TO 20
.0) GO TO 10
c***
f >***«»»* t * »4***##+-«#«»*» SECTION i STARTS HF.RF ****»*»*»»»**»****#***«*
C»«* CALCULATE ZETA ^O^ GE^FRAL CASE. FIRST SOLVE CUBIC.
P * pEY - PFX
1F(E.L.T.O.O) GO TO 1
Q = _(82)
EOT = EXPCQ3)
*
A IF«K,EO. 1 ) i^O TO 6
IFCK.EO.3) C,O TO 51
323
-------
Ztll = 1.0 - T = i. Q
IFfI.F.O.4 .OR. I.EO. GO TO S
J = .) 4- 1
X a Fl-OAT< I )/lO.O
Z(J) = 1.0 - + «2*)
1 4- y;3*
G3 = H3/J1.0 + OI/PEv)
EOZ e FXPC02)
c.Q-3 = EXPCQ3)
DENOH = H2-MEO3 - G3^»<1.0 - EOHJ/O? + H3*(G? - EO2)*().r - EO3)
1 /Q3 - (G3*EO2 - G?*EO3)*fl.O +• 1.0/PFX + l.n/PEV + J.O/NOX)
l'='«,c:O. 1 1 GO TO 15
te GO TO 16
Z<1) - l.O - (*(H3 - 1.OJ/O3
J = \
OO 1 3 I = 1 .9
IFU.EG.4 «OR. T.FO.*) GO TO 13
J - J * J
X = Fl.OATl ' )/lO.C
2 /r>eNOM
324
-------
16 DO IB 1 = 1 t 1O1
X = rLOATCt - I)/ 100.0
EO2X = EXPIO?»X>
TO?X = EVP(O3»X)
U * 1.0 - ((EQ3 - G3)*(H2 - EO2X)XQ2 + (G2 - EQ2)*(H3 - EQ3X)
1 X03 - (Gr»*nC2 - G2*EQ3)«* («2*EQ?>X - H2*E02)X02 + CG? - f Q2 > *( G3*EO3X
1 - H3*EQ3}/Ci3 - (G3*EG2 - G2*EO3)»(I.C - X + \ .OXPEY ) JXDENOM
IB CONTJNUE
Z-rlA = UUOM
c*«*
C
Ql = f -8 + m/2.0
OP - ( -B - DJ/2.0
c**# KiD'A- CALCULATE rnr. SOLUTION.
Hi = 1.0 - Ql
H2 = 1.0-
CO? s FXPtQ?)
IF(E.LT.O.O) GO TO 23
r>^NOM = Hl*«1.0 - FOi/E)/Ol/EOl - H2»C1.O - EO?/E S/O2/EO?
GO TO 24
23 DENO* = H1XQ1/FO1 - H?/O2/PO?
?4 ir(K,EC«l> GO TO ?6
TF«.E0.3> CO TO ?7
Z(l) = 1.0 - ((HI - l.O)/Ol/PQ4 - (M? - 1 .0)/O?/FQ?)/DrNOM
J c 1
DO 25 1=1.9
IF(I.EQ«* .OR. l.F.Q.A) GO TO 25
J = J + 1
X = rLOAT( I 1/10.0
ZU) = 1.0 - ((Hi - P-XP(O1*X) I/OIXEOI - IH? - EXP(Q2«X) >X02/E02)
1 /OCNOM
?•=? CONTINUE
P6 ?fTTA = Z<9) * 1.0 - (
-------
c***
C*« ****»*««***« »*fH *•»»»«* sTCTtON A STARTS HE" RE ********««•***«**«***«*«
O-** CALCULATE SOLUTION FOR E = J. WITH NO BACKMIXlNf, IN DISPEPSEO PHASE
3O PN « PFX •»•
^>E^40M = 1.0 •» t .O/PEy + 1 . 0/NOX - U.O - EPN)*NOX/PEX/PN/EPN
IF«.EQ. J ) f.O TO 35
IF(K.FC.3> GO TO 40
Zd) * 1.0 - U.O/PEy - (NOX/PEX + I.O)/PN/EPN)/DFNOM
J = 1
no 33 i=i%9
IFd.EO.A .OR. l.EO.ft) RO TO 33
J = J + 1
V s FLCATC 1 >/tO.O
z - (NOX/PEX •»• EXP JXPN/EPNJ/DENOM
33 COKTINUE:
3r=i ZFTA = Z(9) r l.O - ({I.O/PEX + l.Ol -
1 /OENOM
RETURN
4O DO 45 I *\ . 1O1
X = FLOAT (1 - D/100.C
t»(I) a l.O - ((l.O/P^X + X) - (NOX/PEX + EXP s (1.0 - X - NOx*(EXP(PN*(X - l.OM - 1 .01/P*FX/PN)/OF.NOM
Afi CONTTNUE
ZETA = U(1O1)
RETURN
326
-------
Dispersion Model For Varying K^
Several authors have considered the design of
extraction columns which exhibit axial mixing for
systems where K^ is not constant. Rod (1964)
described a graphical approach which is generally
applicable to any shape of the equilibrium curve.
Unfortunately his technique requires a time consuming
trial and error graphical solution for cases where
axial mixing is appreciable in both phases. McSwain
and Durbin (1966) developed a computerized solution
to the backflow model for systems where K. is not
constant. When the number of stages becomes very
large, their solution provides a suitable description
of a continuous extractor. However, these authors
made the incorrect assumption that the height of an
overall transfer unit (or the overall mass transfer
coefficient) was constant throughout the column.
Recently Pratt (1975) announced an upcoming paper in
which a new approach will be described. The technique
developed below is similar to the approach of
McSwain and Durbin (1966), but the values of the
two individual numbers of transfer units (rather than
the overall number of transfer units) were assumed to
be constant through the column.
Equations (C3) and (C4) derived for the case of
a constant K. were replaced by the following four
equations:
dX. . J_ d_X = _N (x _ xi} (C13)
dZ Pe dZ^ x
327
-------
dZ
(X
Pe, d3
y
- x1) -
,2
N/y
NXFX '
y
(Y
. Y) (C15)
Y1 = Kd1 X1 (C16)
In these equations the superscript i refers to
concentrations at the interface between phases.
Since we assumed that the value of Kj1 was known once
the value of X1 or Y1 was determined, the above
system of equations involved two new variables (Y1
and X1) and two new equations and was determinant
once the boundary conditions were specified. Boundary
conditions equivalent to those in the previous section
were written as follows:
a. At Z - 0:
*in • x -re- IT
A
(C17)
• o (cis)
b. At Z = 1
dX
3Y = 0 (C19)
Yin = Y + si- & (C20)
328
-------
To develop a numerical solution to this set
of equations, the column was divided into (N-l )
cylinders of equal length such that Z, = 0, Z« = h,
Z3 = 2h,..., ZN = 1 . At all points, from Z2 to ZfJ_1
equations (CIS) and (C14) were written in terms of
central finite differences. For example, at any point
Z.j for 2 < i < N-l, equation C13 became
'xi+i - *i-i) ' '"1+1 • 2xi * "1-1) , ., ,1.
- p - - N y \ A • - A . ;
2h Pev h^ x i i
X
When the terms were collected in a way which left
XJ on the right hand side, this equation became
- " + — L x * Nx + —^- x ^
x
-
2h Peh Peh 2h
At the points Z-| and ZN the central difference equa
tions were written in terms of two imaginary points
ZQ and Z- , and then the boundary equations were
used to eliminate these points from the system of
equations. This results in the following matrix of
equations:
A X = g (C21)
JJ Y_ jf (C22)
where the terms in the vectors X^, Y_, £, and f_ are
as follows:
329
-------
X. = concentration in X phase at Z.
Y. = concentration in Y phase at Z.
1 = NxV + Xin
i
91 ' NxXi
fi ' NyY1
N
I ^3 ^5 •*•) ' *
1=1,2 N-l
-------
where
B =
i
1 1
UN
vi
2h
2
Pe h2
y
= N +
Peh2
y
2
D~ U f-
v1 w] 0 0 .
U2 V2 W2 °
0 u3 v3 w3 .
w
N-l
UN VN
i = 2, 3, ..., N-l
i = 1, 2, 3,
N-l
2
w-
w.
Peyh<
2h
1
Peyh'
i =2, 3 N-l
Equations (C21) and (C22) involved four unknown vectors
X_, Y_, X.1 , and Y^1 . When equations (C15) and (C16)
were also written for each value of Z.., there were
four vector equations available in terms of these four
unknowns.
All the elements of matrices A and B are constant
= =
for the case where N , N , Pe and Pe are constant.
The inverses of these matrices were determined at the
331
-------
beginning of the calculation, which resulted in the
followi ng:
X. = A"1 a (C23)
Y = I"1 f (C24)
This completes the development of equations.
This set of equations was used for two types of
problem. One problem was encountered when two solutes
which interact strongly were present in the feed
water at fairly high concentrations. For example,
when n-butyl acetate at about 6000 ppm and phenol
at about 300 ppm were present in a feed waste water
which was treated by isobutylene extraction, the
presence of n-butyl acetate in the solvent led to
a much higher efficiency of phenol removal than was
expected based on the K^ for phenol between water and
pure isobutylene. The numerical approach was used
in this case by assuming that the presence of phenol
had little effect on the n-butyl acetate extraction
so that the concentration profile for n-butyl acetate
was predicted from the linear model. Equations (C15)
and (C16) were then used to determine Yg. from which
Kohenol was Determined. The solution then proceeded
as follows:
a. Set X_ and Y_ at initial values determined from
the 1i near model ,
b. Determine X_n and Y/1 from equations (C15) and
(C16),
c. Determine £ and f_ from their definitions,
d. Determine new values of X^ and Y_ from equa-
tions (C23) and (C24), and
332
-------
e. Return to step b., and iterate to convergence.
The results of these calculations are discussed in
Section VIII.
The second situation where this method of solu-
tion was useful was in determining the effect of a
variation in K. for one solute due to a change in
its own concentration. For example, when phenol was
present at about 3 pounds of phenol per 100 pounds
of feed water in a water stream which was treated by
n-butyl acetate extraction, the value of K^ changed
significantly through the column. The results of
using this numerical approach to analyze this case
are discussed below.
Several characteristics of this numerical
calculation were observed for both types of problem.
The accuracy of the numerical result as a function
of the grid size (h) was estimated by comparing
the converged numerical solution to the analytical
solution for the case of the constant K,. Some results
from such calculations are shown in Table Cl for a
case where the parameters were similar to those for
the extraction of phenol with recycled n-butyl
acetate in the miniplant. The concentration profile
at the end of the extractor where the solute
concentration was lowest (from which % removal was
calculated) was accurately predicted with ten column
segments. However, the concentration profile at the
other end of the extractor (as characterized by Y t)
was not very accurately predicted with fewer than
fifty column segments. No significant improvement in
accuracy was gained by using greater than fifty
333
-------
Table Cl. Results from Sample Numerical Calculation
Parameters:
Nx = 8.0
Pev =6.0
J\
=0.1
Ny = 8-°
Pe =25.0
Kd = 57.0
Xin = °*03 lb solute/lb feed water
Yin = °'02 lb solute/lb feed solvent
Results:
N - 1
10
20
50
Analytical
% Removal
95.80
95.75
95.74
95.73
out
0.3678
0.3233
0.3098
0.3072
334
-------
column segments, and percent removal was accurately
estimated with only ten segments.
In all cases tested, the direct substitution
iteration procedure which was described above converged
toward the final solution. However, the rate of
convergence was very slow and became even slower as
the final solution was approached. Several acceler-
ation methods were attempted with little, success;
usually the calculation then diverged. The most
successful acceleration method was to make at least
three direct substitution iterations followed by one
extrapolation toward the solution using the Wegstein
formula (Lapidus, 1962). This procedure usually
converged in about one-half the number of iterations
required for pure direct substitution. One procedure
which was more successful in reducing computation time
than any acceleration method was first to solve the
problem with ten column segments, then to interpolate
to obtain an initial solution for the calculation
using twenty or fifty column segments.
For almost all pollutants considered in this
dissertation, the value of K^ could be assumed to
be constant at the value taken on at infinite dilution.
The reasons for this observation are illustrated by
considering the overall driving force based on aqueous-
phase concentration,
AX = X - X* (C25)
where X is the aqueous-phase concentration at any
elevation in the column and X* is the concentration of
a hypothetical aqueous phase which would be in equilib-
rium with the solvent-phase concentration at the
335
-------
same elevation. For the case of a constant K,
X* = Y/Kd (C26)
at all points in the column. When K, varies with
concentration, the overall driving force given by
equation (C25) is still valid as long as equation (C26)
is used with Krf evaluated at X* and Y.
At the solvent inlet end of the extractor
(neglecting the concentration jump for the present
discussion), the solvent is nearly pure (Y * 0), so
that K^ = Kj. Three factors tend to make Y large at
the solvent outlet end of the extractor: (1) efficient
extraction, (2) low F /F , and (3) high concentration
of solute in the inlet water. These factors tend to
be mutually exclusive; for example, efficient extrac-
tion and low FS/FW imply a large Kd, which implies a
limited water solubility. Even if the feed water is
saturated with an easily extracted solute, Y at the
solvent outlet will seldom be larger than 0.05 Ib
solute per Ib solvent. At this concentration K, = K.3
is usually still a good approximation.
One case considered in this report where
the above argument did not hold was in the extraction
of phenol with n-butyl acetate. Since K^ = 57.0, 95%
removal was possible when FS/FW = 0.1. If the feed
water contained 0.03 Ib. phenol per Ib. water (about
one-third of saturation), then Y = 0.285 Ib. phenol
per Ib. solvent. The data of Won (1974) are shown
in Figure C2 and are compared to the empirical curve
Kd = 57.0 - 28.3Y (C27)
which appears to correlate the data up to about 50%
336
-------
phenol in the n-butyl acetate phase. At Y = 0.285
the value of K^ = 48.9, which was a significant
decrease from K^.
The numerical method of calculation was used for
the case of phenol extraction using n-butyl acetate as
solvent. Equation (C27) was used to predict K^ in
Equation (C16). The calculation predicted 96.59% re-
moval of phenol and Y = 0.2898 Ib. phenol per Ib. n-
butyl acetate for the conditions listed in Table C2.
Therefore, Kd = 48.80 at the solvent outlet, and
K. = 52.74. The results of calculations using the
linear model and an average K^ are shown in Table C2.
The results showed that using K°? gave a good estimation
of the removal efficiency, and that using the geometric
mean gave almost exactly the correct answer.
A number of calculations like that just described
were made for various values of the parameters. In
all cases the results showed that the geometric
mean value of K^ gave a better approximation than the
arithmetic mean. The results also showed that in all
cases except when the phenol concentration in the
feed water was greater than about 5 weight %, the
linear approximation using K^ = 57.0 gave a good
estimation of overall removal, but a slight improve-
ment was possible by using the geometric mean K.
337
-------
50
K
Kd = • = 57.0-28.3 Y
45
40
35
30
0
o Data (Won, 1974)
at 25°C
1
O.Oi 0.02
X (Ib. phenol/Ib. water)
Figure C2. Distribution of Phenol Between Water and
n-Butyl Acetate
338
-------
Table C2. Linear Approximation for K, Varying with
Solute Concentration
Parameters:
NX = 8.0 N = 8.0
Pex = 6.0 Pe = 25.0
Fy/Fx =0.1
Xin = Oi03 I*3' Phenol/lb. water
Yin = °*00 lb* Phenol/lb- n-butyl acetate
Numerical Calculation:
K, = 57.0 - 28.3Y
a
% Removal = 96.59
Linear Approximation:
Kd ^Kd basis^ % Removal
57.00 (Inlet solvent) 96.87
52.90 (Arithmetic mean) 96.61
52.74 (Geometric mean) 96.60
48.80 (Outlet solvent) 96.30
339
-------
APPENDIX D
ALTERNATIVE PROCESSES FOR VOLATILE SOLVENT DISTILLATION
In the volatile solvent extraction process using
a C* hydrocarbon as solvent, the majority of the total
cost is usually associated with the distillation
step for solvent regeneration. In this appendix
methods of minimizing this distillation cost are
discussed and illustrated, using the separation of
isobutylene from ethylene dichloride as an example.
The conclusions should also apply with" slight modi-
fication to the volatile solvent extraction of any
other pollutant of volatility comparable to that of
ethylene dichloride.
The loaded solvent from the extractor will be
a relatively dilute solution of pollutant in the
volatile solvent at about 80°F. After pumping to
the pressure of the^disti11ation column, the feed
stream will'thus be a subcooled liquid consisting
of a wide-boiling mixture. During the separation,
all the solvent must be boiled and then condensed.
Since this involves a "substantial amount of heat
supplied and removed, careful consideration must
be given to heat economy.
In Section IV we consider a process that uses
isobutylene (IB) to remove 95% of the ethylene
dichloride (EDC) from a 80°F waste water which
340
-------
contains 0.8% EDC and flows at 100 GPM. The concen-
tration of EDC in the loaded solvent depends on
the solvent-to-water flow ratio, F /F , while the
o W
concentrations of EDC in the regenerated solvent
(distillate) and in the recovered pollutant (bottom
product) do not depend on F /F but rather are fixed
by the required extraction efficiency and by the
required product purity (or allowed solvent loss),
respectively. For 95% EDC recovery and 1.5% IB
in the product EDC, these stream concentrations are
listed in the table below along with the total flow
of feed as a function of F /F .
Mole fraction IB in distillate = 0.9968
Mole fraction IB in bottom product = 0.0271
F /F Mole fraction IB in feed Feed Flow Rate
s w (Ib mol/hr)
1.50 0.9940 1335
0.70 0.9907 625
0.30 0.9826 270
0.08 0.9454 74.3
For each value of F /F.., there is an optimum
b W
distillation column design which results in minimum
cost. In this appendix we attempt to locate this
optimum by designing and determining alternative
costs for several levels of the specified variables
and for several process arrangements. The alternative
cost is the sum of the annual operating costs (steam,
cooling water, and pump power) and the total investment
(pumps, heat exchangers, distillation column and tanks)
341
-------
multiplied by 0.259, as is developed in Appendix A.
The following assumptions are made in determining the
design and cost estimates:
1. Murphree vapor stage efficiency is 0.75.
2. The reboiler provides one equilibrium stage.
3. The distillation column contains bubble
cap trays with 18-inch tray spacing and
provides a 3-minute liquid hold-up at the
bottom.
4. The overall heat transfer coefficient is
200 Btu/hr ft2 °F in the condenser and reboiler.
5. The condenser cooling water is heated from
80°F to 100°F.
6. The reflux tank provides a 10-minute liquid
hold-up.
7. The liquid leaving the extractor has a
pressure of 51.4 psia and a temperature of
80°F.
For the distillation column design, vapor-liquid
equilibria for the IB-EDC binary system are estimated
by the Chao-Seader (1961) method. The large increase
in temperature down the column and the fact that the
molar heat of vaporization of EDC is almost twice
that of IB makes heat effects important and rules
out the constant molar overflow assumption. The
molar flows tend to decrease down the column, resulting
in operating lines on a McCabe-Thiele diagram which
are concave upward in the rectifying section and
concave downward in the stripping section. Stream
enthalpies are estimated by using gas and liquid heat
342
-------
capacities and heats of vaporization for the pure
components and by estimating heats of mixing using
regular solution theory for the liquid phase and the
Red!ich-Kwong equation of state for the vapor phase.
First consider the case where FS/FW = 1.5, which
results in a loaded feed that is very dilute in EDC.
By assuming that the distillate is condensed at 120°F,
the column pressure is determined as 81.1 psia.
If the feed is pumped to this pressure and fed to the
column as a subcooled liquid, then no reflux is
needed and the feed enters the top plate. A stage-to-
stage calculation indicates that 5 actual stages are
required as shown in Figure Dl. Details on the top
two plates are illustrated in Figure D2. Operation
in this case results in large flows of liquid and
vapor throughout the column at near total reflux (L/V
varies from 1.0048 below the feed plate to 1.0025
above the reboiler). The reboiler operates at 275°F,
so all the heat required to boil the volatile solvent
must be supplied at a high temperature. The alternative
cost for this case is $129,600 per year with the break-
down shown in Table Dl. The steam used to drive the
reboiler accounts for 56% of the total alternative cost.
One alternative to the above operation is to
vaporize the feed before it is added to the column.
If the distillate is again condensed at 120°F, the
dew point of the feed is 122°F, so this vaporization
can be carried out using exhaust steam. By assuming
that reflux is added to the column at 40% above mini-
mum reflux, a stage-by-stage calculation shows 6
plates are required below the feed and 3 plates above.
343
-------
Loaded
Solvent
\~y
6X
Y
R -/Vwi—
Wfflt
1 '
*r 1
[ Regenerated
3 Solvent
1
1
— - -te D ~~* ^
Recovered
Pollutant
FIB
Figure Dl. Simplest Alternative
344
-------
X
0.98
IB
Q99
1.0
Distillate
Feed
(subcooled
liquid)
1.0
0.99
IB
0.98
Figure D2. Details of Simplest Alternative
Table Dl. Costs for Simplest Alternative
Alternative Investment = $132,000
Equipment Item
Condenser
Reboiler
Distillation Column
Solvent Recycle Pump
Reflux Tank
% of Total
35
29
26
4
6
Alternative Operating Cost = $129,600 /year
Cost Item % of Total
Capital Cost
Cooling Water
Steam (100 psig)
Power
26
17
56
1
345
-------
This mode of operation is illustrated in Figure D3
with details of the top few stages shown in Figure D4.
Although more total plates are required with a satu-
rated vapor feed than with a subcooled liquid feed,
the flows in the stripping section are much lower with
the vapor feed (V/F = 0.156 as compared to V/F = 1.17),
resulting in a less expensive distillation column.
The requirement of a small amount of reflux (R/D =
0.165) results in more total heat required in the
feed vaporizer and the reboiler, but the use of
cheaper exhaust steam in the feed vaporizer results
in a substantial savings. The alternative cost for
the case of a saturated vapor feed is $106,900 per
year with the breakdown shown in Table D2. The
alternative cost is reduced by about 18%.
To fix the design of the distillation column
with the saturated vapor feed, it is necessary to
assume the condensation temperature and the reflux
ratio. Holding the ratio of actual to minimum reflux
at 1.4, the effect on the alternative cost due to
changing the condensation temperature, and thus the
column pressure, was investigated. As the pressure
increases, the temperature driving force in the
condenser increases while the driving forces in the
feed vaporizer and in the reboiler decrease. Also,
the cost of the recycle pump and its power increase
as temperature and pressure increase. The overall
alternative cost decreases rapidly until the distillate
temperature reaches 120°F, and then the cost shows
a flat minimum between 120 and 140°F. The use of
120°F for cost comparisons seems reasonable.
346
-------
Loaded
Solvent
FV
9~
8
7
6
5
4
3
2
Regenerated
Solvent
Q
R
Recovered
"Pollutant
Figure D3. Alternative with Feed Vaporizer
347
-------
Feed
(saturated
vapor)
Figure D4. Details of Alternative with Feed Vaporizer
Table D2. Costs for Alternative with Feed Vaporizer
Alternative Investment = $127,900
Equipment Item % of Total
Condenser
Reboiler
Feed Vaporizer
Distillation Column
Pumps
Reflux Tank
Alternative Operating Cost
Cost Item
Capital Cost
Cooling Water
Steam (100 psig)
Steam (Exhaust)
Power
40
8
20
19
6
7
$106,900 /year
% of Total
31
24
10
34
1
348
-------
Another alternative arrangement utilizes a
subcooled liquid feed (and no reflux), but several
stages below the feed point the downflowing liquid
is drawn off and partially vaporized. To fix the
design in this arrangement, the condensation tempera-
ture is set at 120°F, the actual-to-minimum vapor
flow in the stripper is set at 1.2,.and the tempera-
ture of the side stream boiler is varied to locate
the optimum. As the side stream boiler is moved up
the column and operates at decreasing temperatures,
the number of stages and the flows of liquid and
vapor in the lower section of the column increase
causing an increase in the cost of this portion of
the column; however, the number of stages above the
side stream boiler decreases, resulting in a nearly
constant cost for the entire column. The lower
portion of the column is small in all cases and is
designed as a packed tower assuming its total height to
be the same as a plate tower with 18 inch plate spacing
As the side stream boiler temperature decreases, its
temperature driving force and cost decrease, but the
portion of the constant amount of total heat which
is supplied with higher cost steam in the reboiler
increases. An optimum alternative cost of $89,000 per
year occurs when the side stream boiler operates at
128°F resulting in 3 stages above and 4 stages below
the side stream boiler. This mode of operation is
illustrated in Figure D5 with details in Figure D6,
and a breakdown in costs is shown in Table D3. The
lower portion of the column is designed to operate
with V/F = 0.024 at 40% of flooding in a 14-inch
diameter column packed with 1-inch Raschig rings.
349
-------
Loaded
Solvent
\
6
5
Regenerated
Solvent
SSB
I
Q
R
Recovered
Pollutant
i i i I i i i i
Figure D5. Alternative with Side Stream Boiler
350
-------
0.80
0.85 IB 0.90
0.95
1.0
0.92
0.90
Figure D6. Details of Alternative with Side Stream Boiler
1.0
0.98
0.96
0.94
IB
Table D3. Costs for Alternative with Side Stream Boiler
Alternative Investment = $111,500
Equipment Item
% of Total
Condenser
Reboiler
Side Stream Boiler
Distillation Column
Solvent Recycle Pump
Reflux Tank
42
1
28
17
5
7
Alternative Operating Cost = $89,000 /year
Cost Item % of Total
Capital Cost
Cooling Water
Steam (100 psig)
Steam (Exhaust)
Power
32
25
2
40
1
351
-------
In comparing the vapor feed and side stream
boiler alternatives, the latter results in a 17% lower
alternative cost. This lower cost is the result of
operating without reflux and thus with smaller quantity
of heat supplied and removed, with a smaller condenser
and reflux tank, and with no reflux pump. Also,
a smaller fraction of the total heat supplied is in
the form of high pressure steam with a side stream
boiler.
Many of the conclusions discussed for the case
where F /F = 1.5 also hold for lower values of F /F .
o W J W
When F /F = 0.08, reflux is required even with a
subcooled liquid feed. The alternative costs for
subcooled liquid feed with one reboiler, for saturated
vapor feed, and for subcooled liquid feed with a side
stream boiler are $14,600/yr., $11 ,700/yr. ,• and
$10,400/yr., respectively. Although all items of equip-
ment are smaller because much less solvent must be
regenerated at lower values of F /F , the cost of the
distillation column decreases only moderately with
decreasing F /F and then makes up a larger fraction of
d W
the total cost. The optimum location of the side
stream boiler is approximately the same for all values
of FS/F between 1.5 and 0.08 (i.e., at a point where
the temperature is about 128°F). At values of F/F.,
o W
below about 0.15, the flow of purified water is large
enough so that it may replace the need for cooling
water in the condenser.
The general procedure recommended for the design
of a distillation system for separating any pollutant
of similar volatility to ethylene dichloride is to
352
-------
use a subcooled liquid feed and a side stream boiler.
Practical operation can be achieved by setting the
ratio of actual to minimum reflux at 1.4 (if any reflux
is needed), by condensing the C, hydrocarbon solvent
at 120°F, by setting the ratio of actual to minimum
vapor flow below the side stream boiler at 1.2, and
by locating the side stream boiler several stages
below the feed point. These four parameters can be
optimized for any pollutant at a later stage in
process development.
353
-------
APPENDIX E
DISTRIBUTION COEFFICIENTS FOR ORGANIC SOLUTES BETWEEN
ISOBUTYLENE OR ISOBUTANE AND WATER*
Introduction
This work presents distribution coefficients for
a variety of organic solutes between water and iso-
butane or isobutylene. Because of the solvent's
volatility, it is not possible to use standard
apparatus to measure distribution coefficients. A
special sampling device was constructed to facilitate
sampling of the volatile liquid mixture at pressures
ranging from 3 to 10 atmospheres.
Apporatus
E1>
The equilibrium celt, shown in Figure / is a cylindrical, stainless-
steel container about 30 cm long and 12 cm. outside diameter. The Inside
diameter Is about 8 cm and the inside depth is 25 cm. Above the top plate
Is a rotating holding magnet driven by a variable-speed motor. A stirrer
rotating inside the equilibrium cell is magnetically coupled to the variable-
speed motor. To allow visual observation of the immiscible mixtures under
* References and Nomenclature for this Appendix are
included at the end of the Appendix.
354
-------
SOLUTE
INLET
ROTATING
MAGNETS
THERMOWELL
HEAD
/-RING
"O"RING-
SOLVENT INLET
AND
AQUEOUS
SAMPLE
OUTLET
SAFETY 1 TO HE I SE
RUPTURE PRESSURE
DISC GAGE
—J THREE WAY
J BALL VALVE
TO
SAMPLING
DEVICE
-<3-d
MICROCAPILLARY
SAMPLE VALVE
INDALLOY
SAMPLE
TUBE
(SEE FIG. 3 FOR DETAIL)
-CELL BODY
TWO WAY
BALL VALVE
SCHEMATIC OF EQUILIBRIUM CELL ASSEMBLY
Figure El
355
-------
pressure, two Jerguson sight gages are boiled to the two opposite sides of
the cylinder. The equilibrium cell is completely immersed in a constant-
temperature water bath held at 25 + 0.05°C. A periscope is positioned
outside the front sight gage and a light source is located in the rear.
Sampling devices are attached to the side and to the top plate, which also
contains a high-pressure, sample-injection port. The pressure is measured
by a Heise gage and the temperature in the cell is measured by a liquid-
filled glass thermometer with a 0.1°C graduation. Liquefied gas is intro-
duced from an inverted cylinder Into the equilibrium cell driven by its own
vapor pressure augmented by helium gas at about I or 2 atm. All connections
ore 3.18mm and 6.35mm stainless-steel tubing. Ball valves are from Whitey
Research Company. Chemical analysis of both aqueous-phase and organic-
phase samples is achieved with a Perkin-Elmer 990 Gas Chroma tog raph.
Procedure
To start, about I liter of distilled water is introduced into the
equilibrium cell. Organic solute is introduced with a syringe through a
rubber septum located at the top of the cell. Liquefied isobutane (or iso-
butylene) is then introduced from an inverted gas cylinder.
Equilibration between the two phases is obtained by a stirring rod
with five propellers, two in each liquid phase and one at the liquid-liquid
interface. The two liquid phases are stirred for about ten hours and one
hour of settling (no stirring) is allowed before sampling.
356
-------
Sampling
The aqueous sample is removed at the bottom of the equilibrium
3
cell. After purging twice, a 50-cm sample is removed from the equilibrium
cell. From this large sample, six samples, 5 mlcroliters each, are intro-
duced in the chromatograph for chemical analysis. Reproducibtlity of this
analysis Is better than 2%.
Because of the high vapor pressure, it is much more difficult to
E2
obtain a sample of the organic phase for chemical analysis. Figure / shows
the special technique devised for sampling of a volatile liquid; this technique
is based on the special properties of Indalloy, as suggested by Fleck (1967),
based on the work of Nerheim (1964). Initially the flash chamber and the
sight gage are filled with solvent vapor and helium gas. After opening the
microcapillary sample valve, the pressure in the flash chamber is reduced by
opening the 3-way ball valve to the atmosphere. Liquid is allowed to Flow
into and through the Indalloy capillary sample tube until it fills the Pyrex-
tube sight gage.
The Indalloy capillary sample tube is swaged at both ends and
removed. This capillary tube, 12 centimeters long, 1.59mm outer diameter
and 0.254 mm internal diameter, is then swaged at intermediate points so as
to provide six encapsulated samples, each about one and a half centimeter
long, with a volume of approximately 0.5 microliter.
357
-------
6.35 mm O.D. Pressure
Reduction Tube
3-Way Ball
Valve
ft
2-WayBall
Valve
Solvent Vapor
Plus Helium
Organic Sample
Microcapillary
Sample Valve
Flash
Chamber
Pyrex Tube
Sight Gage
2-Way Ball
Valve
0.254mm ID.
Stainless
Steel Tube
Indalloy Capillary
Sample Tube
INDALLOY ENCAPSULATION SAMPLING DEVICE
Figure E2
358
-------
One at a time, each sample is placed into the induction oven as
E3 E4.
shown In Figures / and / The oven [Varian Aerograph Model 695],
operating at 250°C, melts the Indalioy (melting point 150°C) and completely
vaporizes the sample. Helium gas sweeps the vaporized sample into the
•
chromatograph. Reproducibility is better than 5%. Further details of this
analytical procedure are presented elsewhere (Won 1974).
Chemical Analysis
The chromatograph uses two hydrogen-flame-Ioriization detectors
and has linear temperature-programming capability. Two 3. 18mm O. D.
stainless-steel tubes, each 1 meter long, are packed with 80-100 mesh
3
Porapak Q (or T). Helium, the carrier gas, flows at 30 cm /min at ambient
temperature. The detector signals are recorded by a l.Omv Brown Electronik
recorder equipped with a disc-chart integrator. Temperatures of injector
block, detector manifold and induction oven are normally kept at 250°C.
Since the volatilities of solvent and solutes are widely different, tempera-
ture programming is normally employed, typically from 125°C to about 50°C
above the normal boiling point of the heaviest solute. To convert peak area
to mass, the response factor (mv sec/mg) of the volatile solvent is obtained
as a function of sample size by injecting a known amount of the sample with
a calibrated, gas-tight syringe made by Micro-Sampling Corporation. The
response factors of the solutes are obtained from chromatograms for mixtures
of known composition.
359
-------
Indalloy Sample
Capsule
co
a*
o
He
—OO-
i
Induction
Oven
Detector
Chroma tographic
Column
Oven
Temperature
Controller
Recorder
Integrator
Soap Bubble
Flow Meter
CHEMICAL ANALYSIS INSTRUMENTATION. GAS
CHROMATOGRAPH AND INDUCTION OVEN
Figure E3
-------
00
CTl
INDALLOY
SAMPLE
CAPSULE
"CfRING
SAMPLE
CARRIER
ROD
TO GAS CHROMATOGRAPH
(HELIUM AND
SAMPLE VAPOR)
CAP AND
COOLING FINS
ooooooooooooopoo
ooqoooooopooooooo
INDUCTION
COIL
TO
TEMPERATURE
CONTROLLER
CARRIER GAS
(HELIUM)
INDUCTION OVEN FOR MELTING OF INDALLOY SAMPLE CAPSULE
Figure E4
-------
Results
Distribution coefficients were measured for dilute solutions of
acetates, ketones, aldehydes and phenolics.' These coefficients, in units
- mole solute/1000 a solvent , • T LI n *. c/i
of s . t Jinnn. ar* shown in Tables El to E4.
mole solute/1000 g water
For polar solutes, the distribution coefficients in isobutylene are
larger than those in isobutane because of weak compiexing between the
double bond in isobutylene and the polar group of the solute. For a
homologous series, the distribution coefficient increases sharply as the
molecular weight of the solute rises.
Correlation of Results: Theory of Dilute Solutions
To interpret and correlate the experimental results, it is convenient
to introduce some simplifying assumptions which are applicable to dilute
solutions. Let c stand for concentration of the solute, let single prime
refer to the aqueous phase end let double prime refer to the organic phase.
Let Ag stand for the change in partial molar Gibbs energy of the
dilute solute when it is transferred at constant temperature and pressure from
the aqueous phase to the organic phase such that its concentration is the
some in both phases: c1 = c". The partial Gibbs eriergy of transfer is related
to the solute concentration and activity coefficients y by
362
-------
Table
Distribution Coefficients and Characteristic Volumes
fo-r Acetates. Distribution Between Water and Isobutylene
(and Isobutane) at 25°C.
Distribution Coefficient K
V*
Isobutylene Isobutane cm /nole
Methyl Acetate 2.56 1.47 80
Ethyl Acetate 10.2 5.R6 101
Butyl Acetate 168 107
Amyl Acetate 727 (400) 168T
f _ TOole of solute/1000 g solvent
~ roole of solute/1000 g water
A
•Estimated
( ) Extrapolated
363
-------
Table E2
Distribution Coefficients and Characteristic Volunes
for Ketones. Distribution between Water and Isobutylene
(and Isobutane) at 25°C.
Distribution Coefficient K
Isobutylene
Isobutane
cm /r.ole
Acetone
Butanone-2
Pentanone-3
Hcthyl Isobutyl
Ketone
Heptanone-2
0.63
2.49
13.4
41.5
222
0.33
1.35
(5.5)
24.4
U10)
74
92
112
134
157'
"^Estimated
( ) Interpolated or Extrapolated
364
-------
Table E3
Distribution Coefficients and Characteristic Volumes for
Aldehydes. Distribution between Water and Isobutylcne
(and Isobutane) at 25°C.
Distribution Coefficient K
Crotonaldehyde
Butyraldehyde
Furfuraldehyde
Valeraldehyde
Ilexyl aldehy d e
Isobutylene
2.48
8.05
1.44
(32.2)
130
Isobutane
1.37
4.36
(0.78)
17.2
(68.6)
V
cm^/nole
88+
92
92
112+
132+
'Estimated
C ) Interpolated or Extrapolated
365
-------
Table E4
Distribution Coefficients and Characteristic Volumes
for Phenolics. Distribution between Wuter and Isobutylene
(and IsobutaneJ at 25°C.
Distribution Coefficient K v*
Isobutylene Isobutane cm /nole
Phenol 0.7 0.2 98
o-Cresol 4.8 C1.28) 116
ja-Cresol 2.7 (0.7) 119
3,5-Xylenol (7) 2.14 139
C } Interpolated or Extrapolated
366
-------
Note that Ag is not the change In Gibbs energy when phases
and " are !n equilibrium.
From the definition of Gibbs energy
- g - g1 = Au-TAs + PAv (£2)
where Au = u" - u1
and As = I"1 - S1
Av = v" - v1
In a dilute solution, the partial molar entropy of a solute
is determined primarily by its concentration; inrermolecular forces between
solute and solvent are only of secondary importance. However, the partial
molar energy of the solute is strongly dependent on these forces. For the
process under consideration here (c1 = c"), we assume As = 0.
At the modest pressures considered here, P,Av is negligible
compared to Au. To obtain an expression for AU v/e use a concept pro-
vided by the perturbed-hard -sphere theory (Reed and Gubbins 1973). We
assume
u'-u° = k'q +U' (E3)
and
u" -u° = k'!q + U" (E4)
where superscript zero refers to the ideal-gas state at system temperature, q
is a (hard-sphere) size parameter for the solute and kf and k" are
367
-------
(temperature-dependent) constants of proportionality. The characteristic
energy U1 refers to (attractive) interactions between solute and water and
U" refers to similar interactions between solute and oraanic solvent.
(E3) (F4)
Equations / and / follow from the notion that in order to introduce
a solute molecule into the liquid solvent, it is first necessary to create
a vacancy (hole) of size q. The work required to make the hole in the
aqueous phase is k'q and that to make the hole in the organic phase is
k"q. Attractive forces between solute and water are responsible for
energy U1 and those between solute and organic solvent are responsible
for energy U".
Substitution gives for the Gibbs energy of transfer
Ag = -kq - AU (E5)
where k « k1 - k" and AU = U1 - U".
The distribution coefficient K is defined by
c" y1
v - ea. - eg ,_._.
K = ^r1 ~ rir— (E6)
eq 'eq
where subscript eq stands for equilibrium. Since both phases are dilute
with respect to solute, we assume that the activity coefficients are constants,
independent of concentration; that is, we assume
y' (at c') = T^ (at c^) and y " (at c") = yj (at c^) ( E 7 )
368
-------
(E5), (E6) (E7) (El)
Substitution of Equations / and / into Equation / gives the desired
result »„ _/kq\ , AU
*nK ~(W~) TTT (E8)
Among others, McGowan (1954), and Deno ana* Berkheimer (1960) have used
(E8)
Equation j , setting size parameter q equal to the'parachor.
Since the parachor is a poorly understood quantity we prefer to
*
base q on a parameter V reflecting the molecular volume. This
parameter is calculated from vapor-pressure and density data as outlined
subsequently. We have found that the experimental results are better
*
correlated with V than with any other commonly-available measure of
molecular size.
It can be argued that the energy needed to make a hole for intro-
ducing a solute is proportional not to the volume of a solute molecule but
to its surface area. Hence we propose to use for q the characteristic
*
volume V raised to a constant power m where m is somewhere
between 2/3 and unity. To fix m, we investigated the distribution co-
efficients of paraffins between water and n-butane (or n-heptane) shown in
E5 •
Table / tliese distribution coefficients were calculated from solubility data
for hydrocarbons in water (McAuIiffe, 1966) and from Henry's constant data
in hydrocarbons (Cook et_ aL , 1957; Aroyan et ajL , 1951; Hayduk et al.,
E5 (E8)
1970, 1973). As indicated by Figure / Equation / is obeyed when
369
-------
Table E5
Distribution Coefficients and Characteristic Volumes for n-Mkanes
Between Water and n-Butane (and n-IIcptane) at 25°C.
Distribution Coefficient K
n—Butane ri-Heptane
y f cm /nole
Methane
Ethane
Propane
Butane
Pentane
Hexane
Heptane
Octane
Hydrogen
70.6
300
1577
7040
3. 24x1 O4
1.5X105
5.33xl05
2.53X106
21.2
33.2
159
825
4000
1.67X104
9.05X104
3.4X105
1.7X106
8.7
35.1
51.4
71.4
91.0
111.7
132.7
154.4
178.4
23.6
370
-------
10'
0 40
V*
80 120
I06
O
v>
8
o»
O
O
tn
01
-------
* m * m
q = V with m = 0.7. When m = 0.7 a plot of log K versus V
gives a straight line. However, considerable curvature is observed when
m is set equal to unity. In all subsequent correlations, we use m = 0.7.
(EB)
The constants in Equation / were evaluated by least-squares; they are
shown in Table E6 .
The experimentally-determined distribution coefficients are
plotted on semilogarithmic coordinates as suggested by Equation (E8) ;
excellent straight lines are obtained as indicated by Figures E6 to E9 . The
*
volumes V used to prepare these figures are given In Tables El to E4.
Some additional volumes for organic solutes are given in Table E7 .
E6 E9
Figures /to/ provide a useful method for interpolation and
extrapolation toward providing good estimates of distribution coefficients
For those solutes where no experimental data are available.
The double bond in crotonaldehyde and the two double bonds
in furfuraldehyde tend to lower the distribution coefficient as shown in
E8 E3.
Figure / and in Table / However, if the double bond is well removed
from the polar functional group, its effect is reduced.
E9.
The effect of steric hindrance is evident in Figure / When
chain branching is close to the polar functional group, the distribution
coefficient tends to rise. However, branching well removed from the polar
functional group has little influence on the distribution coefficient.
372
-------
1000
100
cr
o
v»—
H-
o
O
c
o
=3
jQ
iZZ
If)
i5
10
,0
Isobutylene
Isobutane
i:
a>
o
<
o
3
-------
100
•S 10
O
O
O
Z3
-O
1.0
O.I
• Isobutylene
— x Isobutane
20
25
30
35
r*0.7
DISTRIBUTION COEFFICIENTS FOR
KETONES BETWEEN WATER AND C4
HYDROCARBONS AT 25°C
Figure E7
-374
-------
100
c:
Q>
O
5 ^
1 «
o S
S 2
o >>
0)
a
23
27
v*0.7
29
a>
•o
0>
15
X
DISTRIBUTION COEFFICIENTS
ALDEHYDES BETWEEN WATER
C4 HYDROCARBONS AT 25 °C
Figure E8
31
FOR
AND
375
-------
10
0>
'o
*•-
H-
O
CO
a>
O
o
c
a>
X
I
ro
i i L
I
25
30 35
*0.7
40
DISTRIBUTION COEFFICIENTS FOR
PHENOLICS BETWEEN WATER AND
C4 HYDROCARBONS AT 25°C
Figure E9
376
-------
Table E6
(E8)
Constants in Equation / with q = V
k-0.7
" RT
ISOBUTYLENE
Acetates 0.387 7.42
Aldehydes 0.41 7.62
Ketones 0.412 8.81
Phenolics 0.34 8.7,2
ISOBUTANE
Alkanes 0.41 0.713
Acetates 0.388 7.98
Aldehydes 0.403 8.08
Ketones 0.411 9.45
Phenolics 0.348 10.2
Alcohols'1" 0.41 11.27
^Estimated from correlation of Pierotti et al., Ind. Eng.
Chem...51 95 (1959)
377
-------
Table E7
Characteristic Volumes V for Soros
Other Organic Solutes (cm /mol)
Benzene 91.2
Toluene 112.6
m-Xylene 133.8
Aniline 102.6
Methanol 41.1
Ethanol 60.2
n-Propanol 78.7
n-Pentanol 118.1
n-Octanol 180.2
n-Decanol 228.8
Benzyl Alcohol 117.6
Ethylene Glycol 62.0
Resorcinol 103.2
Acetic Acid 61.2
n-Propionic Acid 82.3
n-Butyric Acid 103.7
n-Hexanoic Acid 147.7
n-Octanoic Acid 192.1
Benzole Acid 123.0
Fluoroiaethane 36.2
Chloromethane 47.9
Trichloromethane 80.6
Trichlorofluoromethane 87.8
Tetrachlororaethane 98.6
Liquid density data extrapolated with the Rackett equation.
378
-------
*
Evaluation of Size Parameter V
*
The molecular-size parameter V is the molar volume of a saturated
liquid in a standard state specified by Hildebrand's rule (Hildebrand, 1939).
The standard state chosen here Is that state where the vapor volume in
equilibrium with the saturated liquid is 100 liters per mole. This state is
useful because it corresponds to a temperature range where liquid density
data are most Frequently available.
*
The liquid volume V is determined from vapor-pressure and
liquid-density data. The vapor pressure P is represented as a function of
temperature T by the Antoine equation
Tf1 (E9)
where A, B and C are empirical constants. When the vapor volume is
100 liters per mole, the pressure is given by the ideal-gas law
P - RT(IOO)"1 (E10)
(E9) (F10)
Simultaneous solution of Equations / and / fixes the temperature.
The molar liquid volume V is given as a function of temperature
by an equation of the Rackett form (Rackett 1970):
379
-------
= D+E(1 -T/T
o
where T is the critical temperature and where D end E are empirical
(E9) (E10).
constants. When the temperature determined from Equations / and /
(Ell),
is substituted Into Equation / we obtain the parameter V The
compilation by Francis (J959) is particularly useful for obtaining constants
D and E from density data for organic liquids.
380
-------
Correlation of Characteristic Volume V*
To evaluate the characteristic volume V for a fluid, liquid-
density and vapor-pressure data are needed near the temperature where
the saturated vapor of the fluid is 100 liter/mole. For some fluids of
interest, such data are often not available. For the majority of polar
fluids, the characteristic volume V can be estimated with fair accuracy
based on that of the parent hydrocarbon containing the same number of
carbon atoms.
E10
Figure / shows the effect of carbon number on the character-
istic volume V for n-aikanes and primary n-alcohols. The characteristic
volumes for n-alkanes and primary n-aIcohols, when compared at equal
*
carbon number, differ by a constant. This constant (AV ) is a character-
*
istic group volume V of group g minus that of merhylene (or methyl) group
y
or hydrogen atom. For primary n-alkanols,
AVOH = VOH - VH
= 6.6 (ml/mole) (El 2)
*
AV may vary slightly when the molecule contains only one or tv/o carbon
atoms; however, for these small molecules, the necessary data are usually
available.
381
-------
400
300
CO
00
ro
o
E
j
> 200
o
en
-------
* . *
Table £8 shows AV for five common polar solvents. Using AV
9 9
* *
given In Table £8 V fora polar fluid is calculated from V for
parent hydrocarbon by
V* = V*+ AV* (E13)
ng n g v c. i o j
where subscript n refers to a molecule with carbon number n and
subscript g refers to a molecule with group g.
*
Table E9 shows the experimental and calculated V for
some typical fluids.
383
-------
Table E8
if
Relative Characteristic Group Volume, AV^For Equation (E1 3 )
Fluid
Aliphatic Alcohol
Aliphatic Ether
Aliphatic Ketone
Aliphatic Acetate Ester
Aliphatic Aldehyde
AV*
G
Definition
* *
V - V
OH VH
vo
* *
V - V
VCO VCH2
V* - V*
VC02 VCH2
V * - V*
vrnn VTH
ml/mole
6.6
9.7
1.4
11
1.0
384
-------
Table E9
*
Experimental and Calculated V for Seme Polar Fluids
Characteristic Volume , V (ml/mole)
Fluid Experimental Calculated
Phenol 98 97.8
m-Cresoi 119 119.2
3,5-Xylenol 139 140.2
Resorcinol 103.2 104.4
Benzylalcohol 117.6 119.2
Methyl Propionate 102 99.6
Ethyl Propionafe 121.4 122.7
Ethyl Propyl Ether 121.6 121.4
385
-------
Notation (Appendix E)
A, B,C,D,E = Empirical constants
_ f . g mole solute
c - Concentrat.on of solute, 1000gsolvent
g = Partial molar Gibbs energy,
mole
k — Proportionality constant
, .„ . concentration in organic phase
K = Distr.but.on coeffic.ent, concentraHon in aqueous phase
P — Vapor pressure
q = Molecular-size parameter
R = Gas constant
"s — Partial molar entropy, —r-
r/* mole
T = Absolute temperature
0 = Partial molar internal energy, —r-
n^ s
cal
U = Characteristic interaction energy, —r-
3
V - Molar liquid volume, —p
* cm3
V - Characteristic volume, |
3
v — Partial molar volume, —r-
mole
y = Activity coefficient, °Cf'V'^ ,.
' r concentration
386
-------
Subscripts
eq = Equilibrium
c = Critical
Superscripts
0 = Ideal-gas state
1 = Aqueous phase
11 = Organic phase
m = Exponent
387
-------
Literature Cited (Appendix E)
Aroyan, H. J. and Katz, D. L., "Low Temperature Vapor-Liquid
Equilibria", Ind. Eng. Chem. 43, 185 (1951).
Cook, M. W., Hanson, D. N. and Alder, B. J., "Solubility of Hydrogen
and Deuterium in Non Polar Solvents", J. Chem. Eng. Data 5, 1 (1957).
Deno, N. C. and Berkheimer, H. E., "Activity Coefficients as a Function
of Structure and Media", J. Chem. Eng. Data 5, 1 (1960).
Earhart, J. P., "Recovery of Organic Pollutants from Industrial Waste
water by Solvent Extraction", Ph. D. Thesis, University of California,
Berkeley (1974).
Earhart, J. P., Won, K. W. , Prausnitz, J. M. and King, C. J.,
"Removal of Phenolics from Industrial Wastewaters by Dual-Solvent
Extraction", presented at the A. I. Ch. E.-V.T. G. Meeting, Munich,
Germany, September 17-20 (1974a).
Earhart, J. P., Won, K. W., King, C. J., and Prausnitz, J. M.,
"Solvent Extraction as an IndustricI Wastewater Treatment Process",
presented at the 77th A. I.Ch. E. Meeting, Pittsburgh, Pennsylvania,
June 2-5 (1974b).
Fleck, R. N., "Ternary Fluid-Phase Equilibria at High Pressures with
One Normally Gaseous Component", Ph.D. Thesis, University of
California, Berkeley (1967).
Francis, A. W., "Pressure-Temperature-Density Relations of Pure Liquids",
Chem. Eng. Sci. JO, 37 (1959).
Hayduk, W. and Castaneda, R., "Solubilities of the Highly Soluble
Gases, Propane and Butane in Normal Paraffin and Polar Solvents",
Can. Journal of Chem. Eng. 51, 353 (1973).
Hayduk, W. and Cheng, S. C., "Solubilities of Ethane and Other Gases
in Normal Paraffin Solvents", Can. Journal of Chem. Eng. 48, 93 (1970).
Hildebrand, J. H., "Liquid Structure and Entropy of Vaporization",
J. Chem. Phys. 7, 233 (1939).
388
-------
McAuliffe, C., "Solubility in Wafer of Paraffin, Cycfoparaffin, Olefin,
Acetylene, Cycloolefin, and Aromatic Hydrocarbons", J. Phys. Chem.
70, 1267 (1966).
McGowan, J. C., "The Physical Toxicity of Chemicals. IV. Solubilities,
Partition Coefficients and Physical Toxicities", J. Appl. Chem. 4, 41 (1954).
Nerheim, A. B., "Indium Encapsulation Technique for Introducing Weighed
Samples in Gas Chromatography", Analytical Chem. 36, 1686 (1964).
Rackett, H. G., "Equation of State for Saturated Liquids", J. Chem. Eng.
Data J5, 514 (1970).
Reed, T. M. and Gubbins, K. E., "Applied Statistical Mechanics",
McGraw-Hill, Inc., New York (1973), Chapter 9, Sections 4 and 5.
Won, K. W., "Phase Equilibria for Extraction of Organic Solutes from
Industrial Waste waters", Ph. D. Thesis, University of California,
Berkeley (1974).
389
-------
APPENDIX F
DISTRIBUTION OF PHENOLIC SOLUTES BETWEEN POLAR ORGANIC
SOLVENTS AND WATER*
Introduction
Recently, Abrams and Prausnitz' ' presented distri-
bution-coefficient data for phenolics between water and
nonpolar hydrocarbon solvents and also gave a correlation
of experimental results based on the theory of associated
solutions. This Appendix presents distribution-coefficient
data for phenolics between water and polar organic solvents
Particular attention is given to the effect of solute
concentration on the distribution coefficient. The
experimental results are correlated with a theory simi-
lar to that used by Abrams but extended to allow for
solvation between the phenolic solute and the polar
organic solvent
For removal of phenolic solutes from water by ex-
traction, polar organic fluids are much better solvents
than nonpolar hydrocarbons; for example, the distribution
coefficient for phenol between water and butyl acetate
is much larger than that between water and benzene.
Unfortunately, however, compared to nonpolar solvents,
polar solvents are much more soluble in water. Therefore,
while a polar solvent is useful for extraction of
References and Nomenclature for this Appendix are
given at the end of the Appendix.
390
-------
phenolic solutes from water, the raffinate water must
be treated subsequently to remove the dissolved polar
solvent. Such removal is accomplished with high
efficiency by volatile-solvent extraction-.
Experimental
Distribution coefficients for phenol and for other phenolics were
determined by measuring equilibrium solute concentrations in the aqueous
phase and in the organic phase. Reagent-grade chemicals were used with-
out further purification. About 100 ml each of organic solvent and aqueous
solution were equilibrated in a 250 ml. Erlenmeyer flask, sealed with a
ground-glass top and stirred by a Teflon-coated magnetic stirring bar.
Equilibrium was attained after two hours of stirring. Both phases were
transferred into a separator/ funnel and allowed to settle for two hours prior
to separation. Both phases were removed and stored in 50 ml volumetric
flasks. Chemical analysis was performed with a Perkin-EImer model 990
gas chromatograph, equipped with a dual-flame ionization detector.
Two stainless-steel columns(l meter long, 3.18 mm O. D.), packed with
Porapak Qwere used with helium as carrier gas. Five-microliter aqueous
samples and one-half-microliter organic samples were used to obtain
chromatograms. For calibration, standard aqueous and organic samples
were prepared such that solute responses in the detector were similar to
those of the test samples. Fresh standard samples were prepared frequently
391
-------
because the solute response was not linear with solute concentration and
because the sensitivity of the detector varied slightly from day to day.
Typically five to six samples were analyzed for each distribution co-
efficient.
Effect of Solute Concentration
Distribution coefficients in butyl acetate and in methyl isoburyl
ketone were measured at high dilution for the following phenolic solutes:
phenol, m-cresol, 3,5-xylenoI, pyrocatechol, resorcinol, and o-chloro-
phenol. Results are given in Table Fl
w
The distribution coefficient K is defined by
„ w _ weight % solute in water-free solvent
weight % solute in water
The effect of solute concentration on the distribu
tion coefficient of phenol was measured for four
polar solvents: butyl acetate, methyl isobutyl ketone,
isopropyl ether and 1 ,2-dichloroethane; results are
given in Tables F2-F5 and Figures F1-F4. The effect
of solute concentration on the distribution coefficient
of resorcinol was measured with butyl acetate; results
are given in Table F6 and in Figure F5. Tables F2,
F3, F4 and F6 also report measured distribution coef-
ficients for the polar organic solvent between water
and itself.
392
-------
to
oo
o
c.
o
O
I 20X
13 X
-Q
Phenol, Experimental
Butyl Acetate, Experimental
Calculated
Predicted
10
10
,-3
240
220
o
o
a>
200 o
ZJ
180 ^
160
CD
c.
J>
o
a>
140 Q
to"
IO"2 10"' 10°
Aqueous Concentration of Phenol, Weight Percent
DISTRIBUTION COEFFICIENTS FOR PHENOL AND BUTYL ACETATE
BETWEEN WATER AND BUTYL ACETATE AT 25°C
10'
Figure Fl
-------
co
vo
120
100
• 80
o
*4—
vt_
O
c
O
60
J 40
CO
20
Isobutyl Ketone /
io-3
• Phenol, Experimental
x Methyl Isobutyl Ketone, Experimental
Calculated
Predicted
1
IO-2 JO'1 10°
Aqueous Concentration of Phenol, Weight Percent
DISTRIBUTION COEFFICIENTS FOR PHENOL AND METHYL ISOBUTYL
KETONE BETWEEN WATER AND METHYL ISOBUTYL KETONE AT 25°C
Figure F2
-------
GO
2 40
0>
JC
CL
30
c
Q>
O
O
c
O
20
10
CO
O
Phenol, Experimental
Phenol, Calculated
Isopropyl Ether, Predicted
I
10
-2
10"' 10°
Aqueous Concentration of Phenol, Weight Percent
O
in
160
O
140 I
ct>
120
O
100 |
O
•a
DISTRIBUTION COEFFICIENTS FOR PHENOL AND ISOPROPYL ETHER
BETWEEN WATER AND ISOPROPYL ETHER AT 25°C
Figure F3
-------
60
CO
VD
g 50
Q>
Q_
w.
£ 40
.1 30
o
O>
20
-O
£ 10
en
O
10
-3
Phenol, Experimental
Phenol, Calculated
Butyl Acetate, Predicted
10'
10
10
o
320
280
240
200
160
o
o
O
-^
CD
o
CD
CD
10
Aqueous Concentration of Phenol, Weight Percent
DISTRIBUTION COEFFICIENTS FOR PHENOL AND BUTYL ACETATE
BETWEEN WATER AND BUTYL ACETATE AT 45°C
Figure F4
-------
OJ
VD
-vj
03
O
12
10
8
-
s 6
o
1 4
k_
V)
0
10
-3
I0"d 10"' !0U
Aqueous Concentration of Resorcinot, Weight Percent
DISTRIBUTION COEFFICIENTS FOR RESORCINOL BETWEEN
WATER AND BUTYL ACETATE AT 25 °C
Figure F5
-------
TABLE Fl
Distribution Coefficients for Phenolics Between Water and Two Oraamc
Solvents at High Dilution at 25°C
Solute
Phenol
m-Cresol
3,5-Xylenol
Pyrocatechol
Distribution
Butyl Acetate
65
153
540
13.2
W
Coefficient. K
7
Methyl Isobutyl Ketone
110
264-
814
20.3
(o-Dihydraxy
Phenol)
Resorcinol
(m-DihydYoxy
Phenol)
o-Chlorophenol
9.9
287
15.2
490
398
-------
TABLE F2
25°C
45°C
Distribution Coefficients for Phenol
Between
Phenol Cone.
in Water, wt. %
0. 0025
0.007
0.0085
0.011
0.061
0.083
0.35
0.93
1.75
2.48
0.003
0.0125
0.13
0.52
0.8
1.3
2.45
Water and Butyl
and Butyl Acetate
Acetate
Distribution Coefficient, Kw
Phenol Butyl Acetate
64
61
55
58
58
56
42
32
24-
19
50
52
46
34
29
25
23
157
157
-
-
171
160
170
184
232
-
-
-
-
-
-
-
-
399
-------
TABLE F3
Distribution Coefficient for Phenol and Isopropyl Ether Between Wafer
and Isopropyl Ether at 25eC
w
Phenol Cone. Distribution Coefficient, K
in Water, wt. % Phenol Isopropyl Ether
0.03 32
0.013 34 99
0.066 34 108
0.089 32
0.36 31 117
0.52 34
1.32 27
1.55 25.
1.92 23
2.7 19
400
-------
TABLE F4
Distribution 'Coefficients for Phenol and Methyl Isobutyl Ketone
Bet-ween Water and Methyl Isobutyl Ketone at 25°C
w
Phenol Cone. Distribution Coefficient, K
In Water, wt. % Phenol M1B Ketone
0.0033 110 58
0.0041 108
0.0093 99
0.031 103 55
0.070 89 56
0.12 76
0.19 68 65
0.48 48
0.89 39
1.64 29 96
401
-------
TABLE F5
Distribution Coefficients for Phenol Between Water and
1,2-Dichloroethane at 25°C
Phenol Cone.
\v
in Water, wf. % Distribution Coefficient-, K
0.32 4.1
0.8 4.3
1.7 5.7
402
-------
TABLE F6
Distribution Coefficients For Resorcinol and Butyl Acetate Between
Water and Butyl Acetate at 25°C
w
Resorcinol Cone. Distribution Coefficient, K
in Water, wfr. % Resorcinol Butyl Acetate
0.0064 10
0.011 9
0.017 10
0.061 9 150
0.080 9
0.19 8 156
0.39 7
0.77 6
1.75 6
2.03 5
403
-------
The precision of the data is approximately + 5% for aqueous solute
concentrations above 200 ppm and approximately + 7% for aqueous concen-
trations below 200 ppm. The distribution coefficient data for phenol between
water and butyl acetate (and methyl isobutyl ketone) at 25°C agree well at
high solute concentrations with those reported by Narcsimhan et al.
F2
however, at low concentrations the results shown in Table/ for phenol at
25°C are about 20% below those reported by Kiezyk and Mackay . At
high dilution (aqueous concentration less than 500 ppm) the distribution
F3
coefficients for phenol shown in Table /are in fair agreement with those
of Kiezyk and Mackay^
Thegmodynamic Analysis
For a phenolic solute- (designated by subscript A) distributed
between an organic phase (designated by superscript o) and an aqueous
y
phase (designated by superscript a), the distribution coefficient K . is
defined by
X ° V a
K x = A = A (F2)
A Y a v o
XA 'A
For a polar solvent (designated by subscript B), the distribution
x
coefficient KR is similarly defined:
X °
„ x *B
/B
404
-------
where X is the mole fraction and y is the activity coefficient. The
standard state for each component in either phase is the pure liquid at
%f
system temperature and pressure. Distribution coefficient K is calcu-
w
lated From experimental K by
X _ W
A - KA
x „ w /M°
= V f^} (F5)
where M is the average molecular weight of the aqueous phase and M
is the average molecular weight of the organic phase on a water-free basis.
At equilibrium, the aqueous phase contains a small amount of
polar organic solvent. We consider the aqueous phase to be a ternary
solution with relatively low concentrations of solute and organic solvent
and an excess of water. For the activity coefficients of solute A and
organic solvent B in water W, we write two-suffix Margules equations:
ft / ft\ £• I o\ *" / \ a
^n-v=n l X 1 H-D (XI +(D,+D - D IX >
<*." /A AW \ W / ABV B / \ AW AB WBJ B
^^^ f~^ I I ^ f T * * ••" \ *" F ^ • » « • • «^ mr * %*
(F6)
a
W
- DWB( V)2 + DAB (XA° + (DWB * DAB ~ DAw)XA° V
(F7)
405
-------
(8)
where D designates the Margules coefficient. At 25°C, Tsonopoulos
reports that for phenol DAW is 3.9. Some other Margules coefficients in
dilute aqueous solution are given in Table F7 .
For yA° and yR°, we use a theory of associated solutions similar
(9)
to that 'given by Renon and Prausnitz , based on the model of Kretschmer
and Wiebe . The key assumptions are:
1. The solute in the organic phase exists in the form of linear,
hydrogen-bonded polymers designated by Aj for monomer, A^ for
dimer, etc. These polymers are in chemical equilibrium
A + A5^A (F8>
2. The equilibrium constant K , defined below, Is independent
of n. ,
0A
n + 1 n ,
n 1
where is the "true" volume fraction. Fora fixed solute, equilibrium
constant K is a function only of temperature. It is independent of the
organic solvent. Equilibrium constants < for phenolic solutes have been
correlated by Abrams and Prausnitz (1).
3. The molar volume of an n-mer is related to that of a monomer by
VA = nVA (F10)
n 1
406
-------
TABLE F7
(F6) (F7)
Mgrgules Constants For the Aqueous Phase [Equations / and / ] and
Solubilities
Butyl Acetate
in Solute-Free Water
fj
AB
-2.
Methyl Isobutyl Ketone -2.
Isopropyl Ether
-1.
'
15
5
37
of Polar
WB
7.0
5.85
6.34
Solvents at
25°C
v a f3^
x
AB
0.
2.
1.
94xlO~3
9x 10
8x 10
-3
-3
Subscript W refers to water, A to phenol, and B to polar solvent.
Superscript a refers to aqueous phase.
(1) Calculated from
(2) Calculated from
DWB = "
(3) Measured by authors
407
-------
4. Interactions between solute polymer A and polar organic
solvent B are described by the chemical equilibrium
A +B. ^ A B. (FIT)
n 1 * n 1 vr i i /
5. The equilibrium constant K , defined below, is independent cf
A B, n
K = -_ILJ - . - (F12)
5
n
A.
AB
Fora given solute, equilibrium constant tc is a function of
temperature and of the organic solvent.
2
Following a derivation outlined by Won , activity co
efficients y . and yft are given by
A
+
-------
In these equations, mole fractions X^ and X^ stand
for the overall (srochiometric) mole fractions of solute A and solvent B.
Similarly <£. and 0n stand for the overall* (stoichiometric) volume
fractions. The "true" volume fractions of the monomers. and <{>B
A1 Bl
ore found from simultaneous solution of the material-balance equations
,/6 -
*A, - Ke *A ( Fl 5
fn "•**
V '-Ke
(F16)
The "true" volume fraction of monomer in the pure state, designated by
, is given by
lim <£>A _ l+2< - „ . .., /ri7\
_n Ai ~ —, r_ (r\7)
(1.9)
The theory outlined above is similiar to that presented earlier
in the sense that both theories consider the solute to form linear hydrogen-
409
-------
bonded polymers which are In chemical equilibrium. However, whereas
the earlier theory, applicable to solutions in nonpolar solvents, describes
solute-solvent interactions with a van Laar term containing a physical
parameter P, the present treatment, applicable to solutions in polar
solvents, describes solute-sol vent interactions in terms of a chemical
equilibrium constant K . Similar chemical theories were proposed
previously for solutions containing alcohols ' .
We now consider some results obtained from the thermodynamic
F6 F7
analysis; these are given in Figures / and / The calculations shown
pertain to phenol near 25°C; we have used D.... = 3.9 and for
simplicity we have set X& = 0.
F6 F7
Figures / and /show how the distribution coefficient for phenol
varies with aqueous phenol concentration, solvation equilibrium constant,
K and molar volume ratio, y^A' wnere & stands for solvent and
A For phenol. For large K , the distribution coefficient falls rapidly as
the aqueous concentration of phenol rises. For intermediate values of K
(about 10), the distribution coefficient is insensitive to aqueous phenol
F6 F7
concentration. Although not shown in Figures /and/ for small values
of K t the distribution coefficient rises slightly with aqueous phenol
concentration, consistent with the physical theory (for £ < 0) discussed
earlier . The chemical theory and the physical theory become identical
410
-------
700-
Molar Densities of Solvent
and Phenol are the same
10 fc 10"
Aqueous Concentration of Phenol, Weight Percent
EFFECT OF SOLVATION EQUILIBRIUM CONSTANT, *-, AND THAT
OF AQUEOUS CONCENTRATION ON DISTRIBUTION COEFFICIENTS
FOR PHENOL AT 25°C
Figure F6
411
-------
900
Aqueous Concentration of Phenol, Weight Percent
EFFECT OF AQUEOUS PHENOL CONCENTRATION AND THAT OF
VOLUME RATIO (VB/VA) ON DISTRIBUTION COEFFICIENTS
FOR PHENOL AT 25°C
Figure F7
412
-------
in the limit, as < — 0 in the chemical theory and as £ -* 0 in the
physical theory.
F8.
A comparison of the two theories is given in Figure / Calcula-
tions were performed for phenol with three solvents. Using experimental
distribution-coefficient data at very low phenol concentrations, parameter
K (chemical theory) and parameter $ (physical theory) were determined
as briefly discussed in the next section. Using these parameters, distri-
bution coefficients were then calculated for phenol at higher concentrations.
F8
The results given in Figure/ show that a physical theory is
adequate only when the attraction between phenol and solvent (negative £)
is weak. If that attraction is strong (large ic ), the chemical theory is
required to represent the data over an appreciable concentration range. On
the other hand, the chemical theory is not suitable for physical solvents
which have little tendency to attract phenol. For such solvents /5 is positive
F8
and therefore the chemical theory does not apply. In Figure /, experimental
data for 1,2-dichIoroethane at high dilution were taken from Kiezyk and
Mackay^
To illustrate the applicability of the chemical theory, consider
the distribution coefficient for phenol between water and diethyl ketone.
Distribution coefficient data at high dilution, reported by Kiezyk and
413
-------
I03
o
c:
o>
x:
Q.
O)
o
c
o
I " ' I
METHYL ISOBUTYL KETONE
I F
^)3VA/RT = 4.18
/cs = 92.5
DI-ISOPROPYL ETHER
Experimental -^9VA/RT = 3.09'
Chemical Theory
Physical Theory
-/3VA/RT = 0.873-
1,2-DICHLOROETHANE
10
I
I
10
-5 |0-4 ,Q-3
Mole Fraction of Phenol in Aqueous Phase
CONCENTRATION DEPENDENCE OF DISTRIBUTION COEFFICIENTS
FOR PHENOL AT 25°C. PREDICTIONS BY CHEMICAL THEORY
AND BY PHYSICAL THEORY
10
-2
Figure F8
-------
Mackay , were used to find < . Distribution coefficients at higher
F9'
phenol concentrations were then predicted as shown in Figure / The
predictions by chemical theory are in good agreement with experiment ,
considering the large experimental scatter.
Distribution Coefficient at High Dilution
In the limit as X.-1- 0, the thermodynamic equations simplify
greatly. In that limit
exp
-»a]
i. A
^ wx lim v x _ ^ - - — - /,-in\
KA " X -* 0 A ~ rVT~ (F19)
r\ /» A w **
oo a
For phenol at 25°C, < =8.0, <*>* = 0.089 and y
(F6) e' Q Al '
Is obtained from Equation /with X^ = 0.
ooo
In the limit as X. -* 0, the activity coefficient y.
r\ **
according to the physical theory is
415
-------
o 120
o>
a.
,£ 100
80
o>
o
o
60
CO
40
Experimental (Kiezykand
Mackay, 1973)
Predicted
I
\
\
\
\
\
\
\
lO'3
IO-2 IO-1
Aqueous Concentration of Phenol, Weight Percent
10°
EXPERIMENTAL AND PREDICTED DISTRIBUTION COEFFICIENT FOR
PHENOL BETWEEN WATER AND DlETHYL KETONE AT 25°C
Figure F9
-------
oo o
y
A VI (F20)
_ B_ *
VA %
(F18) (F20)
Comparing Equations / and / we find
VA
(F21)
When
B
For small K s> and at high dilution, the physical theory (with
negative ft ) gives essentially the same results as the chemical theory,
as Indicated in Figure F8.
417
-------
Additional Solvents
F8
Table/ snows distribution coefficients for phenol between water
W
and thirteen polar organic solvents. The distribution coefficient K is
defined by Equation Fl .
Since the concentrations of phenol in both phases are very small, the
experimental distribution coefficients are essentially those corresponding
to high dilution. The precision of the data is approximately + 5%.
Thermodynamic Relations
A thermodynamic distribution coefficient K for phenol between
the aqueous phase (designated by superscript a) and an organic solvent
(designated by superscript s) is defined by
= 4- (F24)
where X is the mole fraction, and y is the activity coefficient of phenol,
jj
For dilute solutions, distribution coefficient K is related to
V/
distribution coefficient K by
418
-------
Table F8
Distribution Coefficients for Phenol at High Dilution
Between Water and Polar Solvent at 25°C
Solvent
n-Pentanol
n-Decanol
Methyl Isobutyl
Ketone
Octanone-2
Ethyl Acetate
n-Butyl Acetate
Di-Ethyl Ether
Aqueous Phenol
Cone. wt. %
0.016
0.029
0.003
0.015
0.018
0.003
0.016
Distribution
Coefficient, K
54
33
110
99
65
64
52
Purity of
Solvent
Boiling point
136-138°C (MCB)t
Melting point
5.5-6.5°C (MCB)
Boiling point
114-117°C (MCB)
Refractive Index
nD2°= 1.4151
(Ala'rich)
98 + Mole % (MCB;
Boiling point
124-126°C (MCB)
Analytical Reagent
DMsopropyl Ether 0.013
34
(Mallinckrodt)
Boiling point
67-69°C (MCB)
Stands for Matheson, Coleman and Bell
419
-------
Table F8 (continued)
Sol vent
Aqueous Phenol
Cone. wt. %
Distribution Purity of
Coefficient, Kx Solvent
Di-n-Butyl Ether 0.01
Nifroethane
0.14
16
14
Boiling point
140-142°C (MCB)
Boiling point
112-1H°C (MCB)
Valeraldehyde
0.035
Di-n-Propylamine 0.18
1,2-Dichloroethane 0.32
59
6.7
4.1
Boiling point
100-104°C (MCB)
Refractive Index
nD2°= 1.4051
(Aldrich)
Boiling point
82.7-84.2 °C (MCB)
420
-------
where M is the molecular weight of the solvent and M is the average
the
molecular weight of the aqueous phase. According to /chemical theory of
liquid solutions distribution coefficient K for phenol between water
and a polar organic solvent is related to the aqueous mole fraction of phenol
X , to phenol association equilibrium constant K , to solvation equi-
librium constant K , and to the molar volume ratio /V. /. V where
V / V
subscript A refers to phenol and B to the-solvent.
At high dilution, the relation is
"-
»AK
x-
where y°°a is the activity coefficient of phenol (referred to pure liquid
phenol) in the aqueous phase at high dilution. Near 25°C, y500 is 47.3
(8) K is 8.0 (1), the volume fraction of phenol monomer in pure liquid
F10
phenol, designated by <£* , is 0.089. Figure/ shows the effect of
Al
OoX f
salvation equilibrium constant K on K for several values of V /
* B/ V.
no / A
For comparison, Figure / also indicates the distribution coefficient obtained
with an inert non-polar solvent, where K = 0. When the solubility of
polar solvent In water is appreciable, a correction for y°° maybe
necessary as described above .
421
-------
1000
c
a
.-§ 800
o
o
"o
-------
F9
Table/ shows density, solubility in water, distribution coefficient
For phenol, solvation equilibrium constant and normal boiling point for
twenty-two polar solvents. To be useful as an extractant, the density of
the solvent should differ from that of water by about 0. 1 g/ml; the
solubility of the solvent in solute-free water should be small (less than
about 2 weight percent) and the distribution coefficient for phenol should
be large. For regeneration of the solvent from the phenolic solute, the
normal boiling point of the solvent should be well below that of phenol
(181.4°C).
Considering the four criteria described above, methyl isobutyl
ketone, butyl acetate and diisopropyl ether are potentially good solvents.
Aldehydes can not be recommended because of their odor and pentyl (or
rwfcqvier) alcohols may be difficult to regenerate from phenolics because ofv
their high boiling points.
Data Correlation
Solvation equilibrium constant < is found from experimental
wx 0 F26
values of K as indicated by Equation / These equilibrium constants
are plotted against the molar volume ratio ^R/W as shown in Figure FT 1«
' A
The data suggest that there is a linear relation for each of three types of
polar solvent. The data for isomeric decanols show considerable deviation,
probably because of branching near the hydroxyl group. Such branching
offers steric hindrance for solvation and lowers K .
s
423
-------
Table . po
Properties of Polar Solvents: Density, Solubility in Water, Distribution
oox
Coefficient for Phenol, K and Salvation Equilibrium Constant, K
J. J-4.
Solubility?" Distribution*"*" Salvation Normal
Density in Water Coefficient Equilibrium Boiling
Solvent (g/ml) (Weight %) of Phenol, K^ Constants Point (°C)
Benzyl
Alcohol
n-PentanoI
Methyl
Cyclo-
hexanol
n-HexanoI
n-Octanol
n - Decanol
Cyelo-
hexanone
Die thy 1
Ketone
Aceto-
phenone
Methyl
Isobutyl
Ketone
Octanone-2
Ethyl
Acetate
1.04
0.82
0.91-
0.94
0.82
0.83
0.83
0.95
0.82
1.03
0.8
0.82
0.90
4<17>
2.6
0.9
0.6*20)
0.054
(0.004)
5.0(3°>
4.7(20>
0.54
1.6
(0.1)
8.8
196 (7)
296
352 <*)
345 (7)
348 (7)
298,
434 (7)
554 (7)
514 '
611
704
318
27
45
55
54
63
60
63
82
77
96
126
45
205
138
155-
173
157
195
233
156
102
202
117
173
77
424
-------
F9
Table / (continued)
7 7
Solubility Distribution "*" Solvation Normal
Density in Water Coefficient Equilibrium Boiling
Solvent (g/m!) (Weight %) of Phenol, K°°X Constant, K Point (CC)
n-Butyl
Acetate
Benzyl
Acetate
n-Octyl
Acetate
Di-Ethyl
Ether
0.88
1.06
0.89
0.71
0.6
407)
(0.01)
75^°)
413
196 <7>
51V (7)
230
66
26
103
34
127
214
210
35
DMsopropyl
Ether 0.81 0.95 193 32 08
Di-n-Butyl
Ether
Valeralde-
hyde
Di-n-
Propylamine
Nirro-
e thane
1,2-Di-
0.77
0.82
0.74
1.04
1.26
(0.06)
_
2.5
4.7
0.9
106
282
37
57
19
20
39
5.3
4.5
1.1
142
103
no
115
84
chloro Ethane
' At 25°C, unless otherwise specified, Superscript number in parentheses
indicates temperature where the solubility Is measured. Number in
parentheses indicates estimated solubility.
ft
1 Experimental data from this research unless otherv/tse specified. Super-
script number in parentheses indicates data source.
425
-------
Fll
Figure /indicates that to a good approximation
VB
(FZ7)
s
where ic is the slope of the lines shown in Figure Fll.
**
It appears that the reduced salvation equilibrium constant K
depends only on the type of polar solvent. According to the chemical
theory of solutions the reduced solvation equilibrium constant <
is related to Ag , the standard Gibbs energy of complex formation
between phenol and polar solvent, by
r fcn*1 - " (F28)
where T is the coordination number, a constant, here taken as 10.
F10
Table / shows the reduced solvation equilibrium constant, and
the standard Gibbs energy change Ag for complex formation between
the hydroxyl group in phenol and the polar functional group in the solvent.
Large negative values of Ag indicate strong affinity between these groups.
F27
The simple relation given by Equation / holds for a homologous
series, that is, for a series where the polar functional group is freely ex-
F27
posed and not sterically hindered. For example, Equation / does not hold
for dialkyl ethers because the ether ^roup is increasingly shielded as
molecular size increases.
426
-------
140
120
100
c
o
o
• 80
JO
LU
c
O
o
_>
o
60
20
I
2
3
4
5
6
7
8
Pentanone-3
Methyl Isobutyl Ketone
Octanone-2
Ethyl Acetate
/7-Butyl Acetate
Benzyl Acetate
/7-Octyl Acetate
/7-Pentanol
9. Methyl Cyclohexanol
10. /7-Hexanol
II. /7-Octanol
-12. Decanol (Mixture
of Isomers)
Ketones
x
12
Alcohols
V. = Molar Volume of Phenol
at 25°C (87.9 ml/mole)
Vp= Molar Volume of Polar
Solvent at 25°C
I I I
0.4 0.8 1.2 1.6 2.0
Molar Volume Ratio VB/VA
2.4
EFFECT OF MOLAR VOLUME OF POLAR SOLVENT
ON THE SOLVATION EQUILIBRIUM CONSTANT *rs
FOR PHENOL
Figure Fll
427
-------
TABLE F10
Reduced Solvation Equilibrium Constant £ For Nine Types of Polar Solvents
Polar Solvent Type
Aliphatic and Aromatic Ketone
Naphthenic Ketone
Acetate
Aliphatic and Naphthenic
Alcohols
Aldehyde
Aromatic Alcohol
Nitroparaffin
Dialkyl Amine
Dichloroparaffin
5-
A
K
S
63
50
45
33
32
22
5.3
3.4
1.1
A 5 KCQI
Ag i —
mole
-1.15
-1.01
-0.95
-0.77
-0.75
-0.53
+0.31
+0.57
+1.24
Two chlorines attached to two different carbons
428
-------
Distribution Coefficients for Phenol Derivatives
At high dilution, the distribution coefficient K.00* for
phenolic derivative i between water and organic solvent, can be
ooX
related to that for phenol K . between water and the same solvent by
an expression of the form
An K.°°X = ^KA + v. £n K. (F29)
where subscript A refers to phenol, and subscript j refers to any group
in the meta or para position on the aromatic ring. The number of groups
• 00 X
of type j is designated by v. and K. is the "group distribution
coefficient" for group j.
Table FIT shows the logarithms of the group
distribution coefficients for methyl, hydroxyl and
chlorine groups calculated by Equation F29 from
experimental data given above. For comparison, Table
Fll also shows the logarithms of group distribution
°°x
coefficients K • for the methyl group between water and
six polar solvents (and three non-polar solvents)
calculated by analytical -sol ution-of-groups (ASOG)
theory (3). According to the experimental results and
ASOG theory, the group distribution coefficient
shows appreciable dependence on whether the solvent
is polar or non-polar; however, the dependence is
insignificant among six polar solvents and among
three non-polar solvents as shown in Table Fll.
429
-------
TABLE Fll
Calculation of Distribution Coefficients for Phenol Derivatives: Group
Contribution,
oox
for Methyl, Chlorine and Hydroxyl Groups
near 25°C
Solvent
Groups
f
Methyl (or Methylene) Chlorine Hydroxyl
Experimental Calculated...... Experi- Experi-
byASOG " mental mental
Polar
Solvent
Nonpolar
Solvent
Butyl Acetate
Amy! Acetate
Hexyl Acetate
Methyl Isobutyl
Ketone
Heptanone-2
Oc tan one -2
Average
Benzene
Isobutylene
Isobutane
Average
1.06
LOO
1.03
1.31
1.35
1.19
1.25
1.14
1.15
1.16
1.09
1.15
1.17
1.14
1.33
1.32
1.31
1.32
1.20 -1.89
1.21 -1.98
1.21 -1.95
T When the group is located at the ortho position, a steric correction is
necessary. Add to Equation 8:
oo.
In K , =0.3 for polar solvents
0.7 for nonpolar solvents
' T Parameters for ASOG method obtained from paraffin-water correlation
(2) and from paraffin-polar solvent data (9). The differences between
experimental and calculated Jto KQ^ show that there is a difference
between the methylene group In paraffins and that in phenolics."
430
-------
Distribution coefficients for phenol K°? between
J
water and non-polar solvents were reported above. To
estimate the distribution coefficient for phenol
between water and a polar solvent, obtain solvation
equilibrium constant KS for that solvent by Equation F27
and Table F10. With the solvation equilibrium constant
K: and the molar volume of the solvent VD, the distri-
5 D
bution coefficient for phenol can be estimated from
Figure FIT. For a phenolic derivative distributed
between water and a polar solvent, the distribution
coefficient at high dilution can be estimated from
that for phenol by Equation F29. Logarithms of group
distribution coefficients K .°°x are given in Table Fll.
431
-------
Nomenclature
C = concentration, weight percent
Ag = Gibbs energy for complex formation, -•""'•.
K = Distribution coefficient,
weight percent phenol in solvent, water-free basis
weight percent phenol in water
K = Distribution coefficient,
mole fraction phenol in solvent
mole fraction phenol in water
V = molar liquid volume, 1—
X = mole fraction
M = molecular weight
Greek Letters
T = coordination number
**
K = reduced salvation equilibrium constant
s
y ss activity coefficient
< = association equilibrium constant
C
< = salvation equilibrium constant
't£A * - volume fraction of monomeric phenol in pure liquid
Al
V. = numer of group }
Superscripts
co = at high dilution o = solvent phase a = aqueous phase
s = standard state r = reference sample
Subscripts
A = phenol i = phenol derivative B = solvent
432
-------
References
1. Abrams D. and Prausnitz, J. M., J. Chem. Thermodynamics,
in press.
2. Won, K. W., Ph.D. Thesis, 1974, University of
California, Berkeley.
3. Derr, E. L. and Deal, C. H.,-Proceedings of the
International Symposium on Distillation 1969,
Brighton, England, No. 32, p. 3:40, The Institution
of Chemical Engineers, 16 Belgrave Square, London.SWl.
4. Nagata, I., Z. Phys. Chem. (Leipzig), 1973, 254, 273.
5. Nitta, T. and Katayama, T., J. Chem. Eng., Japan,
1973, 6, 1 and 1974, 7., 1 .
6. Narasimhan, K. S., Reddy, C. C.,and Chari, K. S.,
J. Chem. Eng. Data, 1962, 7_, 340, 457.
7. Kiezyk, P. R., and Mackay, D., Canadian J. Chem.
Eng. , 1973, 5_1_, 741.
8. Tsonopoulos, C. and Prausnitz, J. M., Ind. Eng.
Chem. Fund. , 1971 , ]_0, 593.
9. Renon, H. and Prausnitz, J. M., Chem. Eng. Sci ., 1967,
22., 299, 1891 .
10. Kretschmer, C. B. and Wiebe, R. J., J. Am. Chem.
Soc. , 1949, 71, 3176.
433
-------
APPENDIX G
EXPERIMENTAL DATA FROM SPRAY COLUMN EXTRACTOR
The data which were directly measured at steady
state in each run conducted in the spray column
included the solvent and water flow rates, the temper-
atures of the two streams leaving the extractor, and
the concentrations of each solute in the feed and
product water. Since the difference between the
two temperature measurements was always less than
1°C, the average was taken as the column temperature.
These experimental data were then combined with
estimates for physical properties to compute experi-
mental mass transfer coefficients which were corrected
for axial mixing. Since these calculations were
repetitious, a computer program was developed and
used. In this appendix the calculational procedure
and computer programs are described; then the results
for all 37 runs conducted in the spray column are listed
Estimates of Physical Properties.
The aqueous phase density and viscosity at
the column temperature were taken as those listed
by Weast (1970) for pure water. The solvent-phase
density and viscosity were determined from data listed
434
-------
for pure solvents (API, 1963). The interfacial
tension was estimated by the method of Donahue and
Bartell (1952) as discussed in Section VII. The
diffusivities for each solute in water and in the
solvent were estimated by the method of Scheibel
(1954) assuming that the values at infinite dilution
would apply.
Computer Programs.
The majority of the spray column data reduction
was done in the Fortran program SPRAY using a CDC
6400 computer. The experimental data, which were
corrected for mass transfer during drop formation and
coalescence in SPRAY, were then reduced in terms
of the dispersion model using the Fortran program
EVAL. The nomenclature used in this appendix while
describing these programs is the same as that used
in the computer programs; a listing of these
programs follows this section.
Program SPRAY begins by solving the hydrodynamic
equations. The superficial velocities of each
phase, VD and VC, the column temperature, TEMP, and
the physical property data are read in. From the
total cross-sectional area of distributor plate
holes, a discharge velocity, VZERO, is calculated
and read in along with the hole diameter, DZERO.
This information is used in a trial and error
calculation to determine the drop diameter, DP.
The method of Scheele and Meister (1968) is used to
calculate DP since the discharge velocity in these
experiments was always less than the jetting velocity.
435
-------
The correlation of Minard and Johnson (1952) is
used to calculate the continuous-phase flooding
velocity, VCF, and the fraction of flooding, FF.
The single drop terminal velocity, VT, is determined
using the equations of Klee and Treybal (1956), and
the method of Hughmark (1967) is used to calculate
the hold-up, PHI, and the slip velocity, VS.
Program SPRAY continues by correcting the
experimental data for end effects. The contact time
for drop formation, CONTIME, and the interfacial
area of the fully formed drop, AP, are calculated,
which allows the mass transfer coefficient during
drop formation to be calculated using equation (B8).
The uncorrected aqueous-phase concentrations, XIN
and XOUT (measured directly), and the solvent-phase
concentrations, YIN (assumed to be zero) and YOUT
(calculated by a material balance), are corrected for
mass transfer during drop formation and drop coales-
cence. The corrected concentrations, CXIN, CXOUT,
CYIN, and CYOUT, are the values used in the dispersion
model to estimate mass transfer during drop rise.
The continuous-phase Peclet number, PEC, based on
the column height, COLHT, is calculated using
equation (B6).
The calculations in program SPRAY are completed
by determining theoretical estimates for the dispersed-
phase number of transfer units, ND (equivalent to
calculating the dispersed-phase mass transfer
coefficient, k^), and for the continuous-phase
number of transfer units, NC. The interfacial
area per unit of extractor volume, A, and the
436
-------
droplet rise (or fall) time, RT, are calculated.
Theoretical estimates are made for oscillating drops
using the theory of Angelo, et al . (1966; 1968) to
calculate NDOD and NCOD, and for circulating drops
using the equations of Ruby and Elgin (1955) to
calculate NCRE, the equations of Hughmark (1967) to
calculate NCHK, and the equation of Kronig and
Brink (1950) to calculate NDKB.
The final step in determining the experimental
values for the overall water-phase number of transfer
units was accomplished using the program EVAL and
subroutines FINDN and EETA. The subroutine EETA
was described in Appendix C; it is used to solve
the equations of the dispersion model. The contin-
uous-phase Peclet number, PEC, is read into EVAL; a
negative value of the dispersed-phase Peclet number,
PED, is read in to signal that the dispersed phase
travels through the extractor in plug flow (PED = °>).
A value of E, the extraction factor, and ZETA [n
in equation (1), calculated from corrected concen-
trations] is read into EVAL for each component. By
using the search subroutine FINDN, equation (3)
is solved implicitly to determine NOW.
A listing of the programs follows. At the
end of each program and subroutine listing, a typical
output is shown for Run SS12A.
437
-------
PROGRAM SPRAY
PROGRAM SPRAY (INPUT .OUTPUT)
C* <;DpAY ACCEPTS DATA FROM AN F.XPEP.I MENTAL RUN AND CALCULATES DROP
c*** D i AMF: T £ R « HOLD-UP. PEC-LET NUMBER. AND MASS TRANSFER COEFFICIENTS.
DIMENSI OM NAME {6)«DIrFD(6>.D!FFC<6)«E«6)
RCAL KO< S> 'KODF.KAPPa .NCOO.NOOD. NOl*' .NCH<. NCRE .ND« I^^Fir.^.F 10. l , I«)
1H1 .20X.33HANALYf.IS OF SPPAY COLUMN RUN tA'i.lOX,
16HPAGF 1///17X. 19HDISPFRSED PHASF IS . A7. //I f* , 2~HCONT t NUOU? PHASE
2 IS , A7.///HX.43H**** SUPERFICIAL VELOCITIES AND TE.MOEpATURr/23X.
-»27HVEUOC ITr DISPERSE'-, PHASF * .F6.?.6H FT/HR/23V , 2BHVELOC I TV COMT I
ANUOUJ PHASC e ,F6.2.fiH FT/HR/23X. 1 4HTFMPER ATUKEI = .Fa.l.
«! 10H
4 FOPM/^T ( 1 4X.?4M**«* PHYSICAL PROPcRTIES/23X.36HDENSITY DISPERSED PH
JASC r .F7.'5jBH GQAM/cC/^3X.a7HOENSITY CONTINUOUS PHASE = ,F7.P.
GRAMVCC/^3X.??2HINTF:RFACIAL TENSION = .F7.3.P.H DYNF/CM/?.1X .
SCf.il TY DIS°FRSn;n PHASE = .F7.4.3H CP/23X.
SCOSI TY CONTINUOUS PHASF = »F7.4,3H CP//)
5 FORMA T (Fl O.^.FIO.5)
ft Foo-iAT < i^x.?9H«»*« DISTRIBUTOR PLAT^ ST?INO/?'IV. . ISHHOLE DiAv^Tfrp
1= ,>r7. ?2M'i INGLE ORIFICE FLOW = .1PE9.3.8H CUFT/HR/?.TX, 1 f-HDROP DIAMETER
7 FORMA T ( UX,a9H**«* FLOOD!NG CHARACTER I ST 1 CS/?3X, 3 1 HCONT-PHASE FLOO
JOING VELOCITY = ,F6.?.6H FT/HR/P3X .23HFRACT I ON OF FLOODING = ,F6.f,
P FOPfUT { 1AX,3AH**»* HnLD-UP ESTIMATING PARAME TERS/23X , 1 OHP« * 0 . 1 5 =
I .F7.3/23X.21HOSCILLATICN FACTOR = «F7. 3X23X. 27HTRANS I T I ON DRO"3 DI
2METEO = .F7.6.5H I NC.M/23X , 20HTE!RM I NAL VELOCITY » .F7.2.
•» f-M rT/'HSJ/ija) .PHOPHAR = .Fn.7,c;H I NCH/?^X . I 1 HOP/OPBAR = .r7.?/23X,
4 7HVgAR = «F7.?4ftH FT/HRX/)
<9 FOt?MAT( 14X..?6H**** SOLUTION TO HOLD-UP CALCULAT I ONX23X , 1 PHMOLD-UP
»= ,F7. 5/?3X, I6HSL IP VELOCITY = .F7.2.6H FT/MR//)
10 FORMAT ( 1H1 .?Ox.33HANaLYSIS OF SPRAY COLUMN RUN .A5.1OX.
11 F--Or^AT ( AO.^X. a11" 10. 2)
12 FOrjMftTl 1«X,6?H»»** SoUUTt O1FFUS I V I T I CS . DISTRIBUTION COEFFICIENTS
1. ANf> E( 1 >/2HX« 15H1G**5 » SOFT/HR/ 1 7X . 6HSOLUTE . 5X iHHDD ( I ) . 5X .
I ) » 1 IX.^HKO t 1 ) .9X.4ME ( I ) />
Cl«>X,A8.F9.2.F1C.?«rif-,.2.F13.?J
14 FOHM/t f I irtO/14y,49H**»* TNO ErFtTCT CORRrrT I ON WITH SOLVENT
n/23x« IOHFS/FW = FO/^C = ,F6,3.ftH LPXLRX23X. 16H^WO=
F7.o«riH !N-.H/?1X. 15MOWIFICE FLOW = .1PF9.3.HH CUFT/HR/?3X
»l^HtONT^CT Tjwr = «F9.3«PH HI5/DROP/23X . 1 T.HSuRF AC.r A ft FA = .Fri.?.
i OH JOKT xr^ ^P/X 1 5X • 6nf OLUTF , qx , anr EEH . ^x . OHPROO . rix . 4 nrr-Fo . r-x ,
''X.''HOV~f?ALL/?3X. 1 IH(PPM) WATf R .4X , fHWATER . 3X . 7HSOL VENT .
i •; FORTM A T c i ox . v i o . i »
If rnpM^r ( I'tY.AR.^H MFAs.ro. t »3F9. 1 .F10.'-.)
17 fd^M/vT ( ? 'X.AHCO'J^.Fn. 1 . 3FO. I . F 1 n . r> )
tf» fC'JMxT I !HOxMX.47H*r««* TNO EFFECT COIJPrCTION WITH WATFR OISrT«Sr.O
438
-------
l/23<. 16HFS/FW = FC/FD = »F6.3«e>H LB/L3/a."!X , 1 6HOROP DIAMFITER * •
;> F7,c;,?;H INCHX 23X . 1 5nOR IF 1 C E FLOW = • 1 PK«> . 3 . BH CUPTXHR/?3X
3,jgHCONTACT TIMF = «e9.3«BH HR/r>ROP/?:>.X • '. r-HSURFACE ARtA = ,E9.3.
4 1 OH S ^r T/OPOP/X 1 ? X . 6MSOLUTF . OX . 4 H^F-ED • *X • A HPROO . "ix < 4Hrf" TO » RX «
fi 4HPf?OO. AX,7HOVERALL/?3x. 1 tH(PPM) WAT>-R,4X , V-HWATTR . 3X . 7HSOL VENT .
6 2X, 7H?>nLVIrNT«'iX«4H^ETA/)
19 FOPM-yT (K10.3)
?0 FOB«AT l lHO/14X.5?ftM«*** AXIAL MIXING PARAMCTr; RS/23X . 1 PHLOC AL PFC =
1 ,F7./,/?3X. 16HCOLUXN MF I GHT a «F6.3«3H FT/T3X «1 AHOVEP Al-L PfrC = .
?1 FORMAT ( 1H1 .?( y«33HAN/>LVSIS <•«•" SPRAY COLUMN RUN «Ar>»lOXt
I 6HPAGC 3//X/1«aX«?BH«*#* MASS TRANSFER FJiT IMATCS/23X . AHA = «F6.2«
2 10H SOcrT/ru'rT/23X « 1 2HRIS? TIMFT = .1PF1O.3.3H HR//I
?? frOnM^T ( 1H1 • ?OXt 33HANftLY?!S OP 5.PRAY COLUMN RUN .A'l.lCXi
I 6MPAGT 3////!«Xt?nH»*#* WARS TPANSF«iR rST IMATFS/23X » «HA = .F6«^•
3 10H «10FT/C'-JFTX23Xi IpH^ALl. TIM^ = »1PFin.3.3H HR/X)
?3 FORMftT ( |OXi47H»*»* OSCILLATING DROPS (5URFACF STRPTCH THEORY)/
1 83X,«HB = .F6.4. /-Ply ,«HW a t!PE9.3«4H /HR//1 SX • 6H5OLUTF . 6X « 2HNC •
2 -7X,2HND«8X« 1HQ, 7X «3HNOW«4X,5H 1 /NOW « AXt?Hl
5 FORMAT ' 1HO/1<,> .,3flH*«#* CIRCULATING DROPS - HK * HUGH.". APK/3OX <
= R'-'BY-GLG1N/39X, 17H---3 •-- KRCN I C— DP !NK/?2X. 1 PHHK! FACTORS -
. l/35X.P,HKAPoA = ,F7.4/1>5X .4HF = . FC.3/X I 5X , fcHF.Ri_UTE . 5X •
tl . 6X . I MR , 7X . 3HNCW . AX . 5H ! /U'OV.'X )
C***+***»*-»******** »**»** eNn Of ro^MAT STATF^rNTS #**»»#****»»««•**•»#**»
31 HEAR l .RUNNUM, NAMED «MAMCC, IDISP.VD.VC* TPMP.NCOVP>S
IFf Ipl SP.ro. CM GO TO ICO
C*** IF IOtSP=l« SOLVENT I c. OlSPFPSF.n. IF IOISP = ?« V'ATER IS D1SPFRSCD.
PRlNr ? «RUNNUM.NAMFO.NAMFC« VD« VC.TFMP
C*** READ AND PRINT PHYSICAL PROPERTY DATA { GM/CC • CV/CC. DYNUXCV . CP . CP)
READ 3«Dr;NSO»OENSC .SiGV.A, VIScDi VI f>CC
^ELDfN = AQSCDFNSn - DENSC )
PRINT A «Ot-S'SO«DEN!:>C • r. I GMA. VI SCO . V I SCC
C»»» READ TN DISTRIBUTOR PLATE HOLF OIAMFTMR AND DISCHARGE VELOCITY
C<»* (INCH) (FT/SEC)
C»«* TO B^fiASS D 1ST PI BUT OR CALCULATION. READ P7t RO = DP ANO VXFPO = O,0
02:RO.V2rt:>0
= 7R.ba*V7rRO»n?FRO*O7EPO
«NE. 0.0) Or. TO 32
OZE.PO = 0.0
GO TO -*0
C**» CALCULATIi DROPLET
-\T> A * 2.f"/7
I GMA>**ri. 3:1333
P a 4. 33(.V.r-05»-VISCC#v2roo»-O^Eno**3/'OELt>ifN
op = 2.0^ERO/I
1F(X,GT. 0.5) GO TO 35
F = i.OOOO - 0.7506*x + 0.4572«X*X
GO TO 39
439
-------
3S IFOPT(DENSC)
J 4 6,a7-it[3p**O.056*sORTJ9.*DELDcN*DP*DP*0/SIGVA
DPT = 9. 1 27E-Oa*5QRT( SIGVA/DFLDF.N/P)
IF(Dp.CT.DPT) GO TO 4/1
VT a 1 .4O3F.-4-OA*EXP < 0 . 5B*A(._OG ( DETLDEN ) 4- 0. 70*A1_OG (OP ) - 0.«5*ALOG(
1 DENSO - 0. 1 1*ALAG( VIPCC) )
GO TO ''S
Aft VT - 1 .31OF-4-O3*EXP( 0 , ?e*ALOG< DELDFN ) ••• O . 1 0«ALOG< V I SCC )
1 * 0. J6*AUOG(S IGMrt ) - O«55*ALOG(OTMSC> )
4S DPPAR = 1 .670E-O3* I Vl f.CC*VISCC/DFNSC/OF.LDEN)**C. 33333
A = (jP/OPHAR
WBA^ = ?vn.6^JVISCC*DFLDFN/DENSC/D!rN5,C)*«0. 33333
PRINT 8«RiOSCFACT • OPT • VT t DPBAR i A. VBAR
C»«* CALcUtATf: HOLD-UP AND SLIP VELOCITY
X = VD /VT
T^V.GT. O.OC) GO TO 46
H = X
GO TO 5O
46 1F(X,GT. Ci05) GO TO 48
H = -O.OO23 + 1 * 1 1 »X
GO To 50
4B H = _O.C?O5 •» 1.36*X + a«9fl»X*X
5O X = VOXH
V = x + VD - VC
PHI a (Y -' SOPT(Y*Y _ 4 .O* VD*X > ) X2 . 0/X
VS = VD/DHI -t VC/Ct.o - PHI)
PRINT PiDHl .V<;
PPINT lOiQUMNUM
C»** RF.AO fN AK'D PRINT
DO 64 I=1«NCOM°S
PtAO 1 1 «NAMF( I ) tOIFFof I » «OIFFCf I ) .KD< I )
C*** UPJINT DIFFUSIVITieS. DISTRIBUTION COFf- F I C 1 ETNTS* AND E(
PRINT «2
DO =55 I -1 .NCOWPS
E« I ) = FSOFW+KDt I )
SB PRINT !3*NAM£t I ) .DIF«--D( I ) .PIFFC ( I ) «KO( I ) «E ( I )
C*** CALCUl \Tfr- END EFFECT CORRECTION
CON-TtnE = 3.030E-04*nP*DP*DP/'SOFLOW
AP a ?. 183E-0?.*DP*DP
IF( tOt SP.FO.?.) GO TO 70
PRJfJT 14,F50Fv;.DP«SOrLOW.CONTlME« AP
DO 60 ' ='
440
-------
fJ?AD 15«XIM«XOUT.Y IN
YOUT = YIM + (XIN - XOUTIXFSOFW
= (XOUT - YINXKO< i >>X - YOUT
pY = 9. l?5*AP*SaptFFr>< t )*rONTlMF)*DYBOTxD»**3XM .0 -f
1 n-MSD-" SORTED I FFPX 1 )XDIFFC< 1 > )«KD« I ) XPrNSC >
(-.YIN - YIN + DY
CXOUT = XOOT «• OY*FSoFrW
p>Y = r>Y*DYTOP/DYBOT
CYOUT = YO'JT - DV
CXIN = XIN - DY*FSOF\ii
7F.TA = )
60 r>^|Nr 17,CXfN.CXOUTtcY1N«CYOuT»ZETA
c,o TO no
70 PRINT 1 O.FSOFW«DO«SOFLOWtCONTlME»AP
DO 7 1 1 = 1 « NCOMPS
pt:/\o i?, XIN, XOUT. Y IN
= YIN + (XIN - yOUT)XFSOFW
= (XOUT - YIN/KD< I n/ixiN - YIN/OHM
16. NAME 1 1 > »XIN,XOUT,YIN.YOUT,ZE:TA
= XIN - YOUT/KO»I>
~ XOUT - YIN/KD«I)
px = 9. 12b*AF *SORT( OTFfrD( I )*CONTIME)*OXTOP/DP**3/( 1 »0 4
t OEVSD-»SORT/OFNSC»
CXJW =• XIN - OX
CYOUT = YOUT - DX/FSQFW
OX « DX*DXBOT/DXTOP
CXQUr •» XOUT + DX
CYlrJ = YIN + OX/FSOFW
2E1/1 = (CXOUT - CYIN/KD( I > )/{CXlN - CYrN/KO(l)>
71 PRINT 1 V,CX|N.CXOUT»cYtNtCYOuT,ZETA
TE CONTIKUnUS-oHASF fJECt-ST
BO ("cfTArj 19.COLHT
PECU. = 2. 36«pw!«-» 0.33333
= VC*COLMTXffC/( 1 .0 - PH!)
PRINT 20.PECL .COLHT ,p^C
C»»* CALCULATE MASS TRANSFER ESTIMATES
A s -??.O«PHI/DP
pjT = rOL.HT*OHIXVO
IF ( Ipl SP.FO.2) GO TO Rl
PHI, NT ?1 .RUNNUM«A» RT
GO Tf> 02
R ] r.'q I ••; y r.g « RUNNU'-' « A « PT
l^st TOR OSCIULAT tNf, DROP^ - SURFACE STRETCH THEORY
B? C = o«'t
B = ^VCOLHTXVD
OO 0$ 1=1 .NCOMWS
NDOO = 1 .244»SORT< 1 .r|F.-Of>*DtFFD ( I )*W>*«?
NCOO = NOOO»SORT(O IFFC( 1 )XDlF«rO( I) ) *VDXVC
IF ( ![)! SP.FO.Z) GO TO PA
C = 1 .OXNCOD 4 1 .O/Ff \ )XNDOO
441
-------
MOW = 1 « 0/C
R r H)COD/E< I )/NDOO
o - i . o/e< i )
GO TO 85
HO C = l * G/NOOD + 1 . 0/E( I )/NCOD
NdW = 1 «0/C
R = Klt>09/Et 1 J/NC.OO
D = i . 0/e< 1 )
B"=> FJRINT 2< tNAMEf I ) . NCOo . NPOO.R «NOW. C . D
C*#* NO'V PpR CfRCUL-AT INC. DROPS
= FXP(/>|_Ofi(RE )/««0 + ALOG( VISCCXVISCD)/4.0 + ALOG(
|Vc.»VtSCC/Sir,M/i)/6. 0 )
F * 0»38l •»• 1.6l5*KApPA •»• 3.73#KAPPA*KAPPA - 1
ORE - 0.7?5*(1.0 - PH! )*RF*»O.S7
H = l.at-04*A«COLHT/Dp/vC
PRINT ZS.
oo eg 1=1
O = 1 . 342E •"•05*nP/D IFe-C ( I > «»0. 66667
SC = 3876.*VlSCC/DENsfXt)IFFC< I )
»*o.
XOLAM = 9.216f-0?*DlFFr>( I )*PT/OP/DP
NDKR - -ALCG(0.62'iO*EXP(-J ,ft^5»XOLAM) + C . 1 332»FXP f -O. CR«XOLAM )
I + O,O?587+EXP(-22.?*XOLAM) )
1F( Ipl SP.CO.2.) GO TO B7
c * i. o/NCcjr * » . o/e: ( i )/NOKB
NOW s 1 . 0/C
R r fjCRE/EC I )/ND<0
GO TO OS
87 C - i.O/NDXn + 1 . 0/E( I J/NCRE
NOW = I .0/C
R - NOKe/FU J/NCQE
HO PRINT ?6.NAME( 1 ) t NCH« « NCRF « NDKfl • R « NOW , C
GO TO 31
1OO CONTINUE
CND sPRAY
442
-------
A.*,ALYSli OF SPRAY1 COLUMN HUH SSiaA PAGE
PnASt IS
CONTIGUOUS PnASt IS
«««•« SUPt^KICI^L VELOCITIES AND TEMPLRATUWE
VELOCITY OISPERStlJ PHiSE = b2.?:> FT/HR
VfcLOClTY CONTINUOUS PhflSE - «?3.2fc "*'
Tti^t.' DECREES t
OcNbllY UUPEHS£» friASK 3 .'J'Sl^O GRAM/CC
utw5nv CONTINUOUS t'nasF. = ,«j'J7**o GKAH/CC
iNl t!-AT iou FACTO
lKANslTlOf4 UlniP OlttrtflTF.P 9 .15'.'>r-uC INCH
i^Al. VELOCITY •- l-}^7.2fa KT/l-tJ
5 2(14. fc I FT/"H
ru HOI U-UP C'*UCUL.ATION
UH = .0???^
SLIP VtL^CHY s ISM./tfl Kf/hH
443
-------
OF SPRAY COLUMN KUN SSj?A
PAGE 2
sot.uie
SOLUTt
ACETUNt
MEK
Cl?OTOu«L
N - r> LI A C t
JIM I)
t'4 . OU
IHSTRIHUTION COEFFICIENTS. AND
MH
DC (I) KU(1) E(I)
4 .<>f.-i)?. CUH/HR
AREA s 7. ,
-------
OF SPRAY COLUMN HUN SSl2<»
PAGE 3
««»« MASS IhAN-irEK EST!MflTt.S
As 1U.7J SIJFT/CUFT
St lint = 1.H2JC-03 MW
USClLLATjNo 090HS {SUi ftbb
I\-HIJ ACt 2.2/ii?
•«»«» ClKUJLAHmi 0,
NO
3.1542
3.01BH
3.018*)
2.7832
•(OPS - UK
Rt
« KOta
1.07'»'> 1.3705
.2721 2.1*0<>
,?7Jb 2.1*09
.0037 ?.?70*
= HilU'iMAMK
= HIJUY-CLGJN
I/NOW
,7297
1*671
.4*0*
1/E
1.1923
.3017
.3029
.00*5
hN FACTOl'i - RE = 781.
KAPPA * 1.3022
F = *.77o
SOLUTE
NOW
I/NOW
f F.K
ACt 3.
2.4(>23 .6102 4.Mil*
2.3*17 ,«SP*0 1
«^.3452 ,SH*I) 1
1.904S .53V3 .ij'jb
.4237 2.3602
1,0598 .9436
1.05*1
1.87*4
.533*
445
-------
PROGRAM EVAL
K-n AM EVA.l_f IN»UT .OUTPUT)
C« FVAL ACffl'TS EXPERIMENTAL DATA PROM EXTRACTING SOLUTES FROM WATER
C-"»* AND A'1'.l V7ES THEM IN .TRRWS OF THE" DIS>PERSION MOPEL
n i '-"••,''"• iOs- NAHc<6> • r < • 7<9> «ui 1 01 > »v < 101 1
DATA i\V 1
C*«* pPGIN FCHVAT STATEMt'NrV- **«**»*#»»**»*»•» »****** •*•»»*•»»*«****«****»»
( l r- )
\T ( An.2v.t--l r>. .lAH GIV
1INC- NOX = .P".:i// BOH *PROril.E Z(O.O) Z(C.l) 2(0.?) ZtO.3
2) ZfO,^) ZI0.7) Z(O.««) Z(0.9) 7 (1 . 0 ) X?X . (S^v. | xrn .orp.^)
11 POCMiT ( 10HO** »* FOR .AO.'-H E = .K8.A.I?H AND Z^TA r ,P6.^,1^>H GIV
1 INT fgOX = . A7//>
if- F^>^r1^T t ox. AOHVO foniTtON pns<;lBir FOP NOX .LT. ir-n.no/>
16 ^OFrt\T ( ftXO9HTRI ai_ AKI^ FRROR S^APCH O|n NOT CON'V^R^F/)
17 FCFM^T ( .?X.6HPLUG .Pr-,^.5/)
1C FOKr^Tf-'IH »5?NS1TIV|TY -- DX = CHANG? IN NOX CAUSFD BY A 1 P!rRCFN
IT DU("IU;'A.Si; IN X/SX.8M07FTA « «1PC9.?,8H . nt = .KP.P.10H . DPEX =
?.'J.9.7«10H . f PEY = .r-t5.;V)
C*«« ="Mr> O- rdrj:-'iAT 5TATl£Mr>gTS ****»»*»»»»**»»»*»#*«**»>***»••»««»*«••»««»
C*«* R^AI IN NljMflER OF
READ I .
C**« RCAn fW PCX. PFY, RUN NU^RFP* AND NUM^FR 0^ COMPONENTS
Rt'AD ? «P'r
C»*» Rfc'AD '^J *;
OO ,?0 I =1
?O W-'AO M ^JA
.LT.0.0) r-0 TO
PRINT 1 1
GO TO ,";;
1P.RMJM.PEX. AM! .NCOMPS
FOR N'OX FOR EACH COMPONENT
EA •-- r«NC)
ZA = 7ITTA(NT)
CALL f lUrN<=EX.PFV.eA , 7 A. NOX, 7 )
|F(S,'oX.GT. 1PO.01) GO TO ?0
PRINT 1 3 . N/ ; IF ( NO «£T (sir ) .7FTA (NC ) .NOX. (7(J) «J=1
Au<-ui ATE: PROFILE FOP ni ur> FLOW
IFf "(NC ) .F3. i.O) C.O TO SB
FA - 1 .0 - i ,o/r(NO
NOA r ALOGIEA/ZETA ( NC ) * J . 0/E ( NC ) ) /F A
PA = -f A«NOA
7A r -rXP(EA)/E(>JC )
J = 0
446
-------
no 23 i = 1 • M
IF
C**» rALr.ULATF 5EM&IT1VITY
MOA = 1.01*NOX
EA = P/ON
EA = 0.99*E'^C>
CALL f!rT/l(PF.X,PeY.EA.MOX.ZE« 1 «Z«U.V>
ONOE - (7ETA«NC> - 7£ ) /ON
PFA «
CALL FETA(PrA«PEY'EA,NOX.7Ft 1 «Z.U«V)
= (ZETA(NC) - 7iF
a 0.0
ONpPv = O.O
T.O.O) GO TO ?B
CALU CUT/\(PEX.DF"A.r A,NOX.ZF« 1 . Z.U.V)
DNOPY = «2ETA(NC) -
?P PRINT 1 B.>;ri2,DMOt: . r
r,o TO 30
C* »* PRIMT OUT •PFSULTS
po pqjWT 1« .NAMC(NC) .^ (KjC ) .Z^TAINC)
O. 100.01) PRlMT IS
iOi 1OO.OZ) PRINT 16
30 COMTjMUE
447
-------
SUBROUTINE FINDN
1 NFC r 1 MDN < PEX . PE Y . F . ZET A « NOX • Z )
c* FINON DETFRMINES NOX WH^N "FX. PEY. E. AND ZFTA APF GIVEN. IF PFY ts
C*#* .LT. 0.0 , THE SOLUTION If- FOR NO 9ACKMIX1NG IN TH^1 ORGANIC PHASF.
C»*» IF E yS LE?ri THAN O.O . THFN CALCULATION IS FOR INFINITE E.
C t** \F NO SOLUTION EXISTS FOR WHICH NOX .LT. JCO. « THE PROGRAM RETURNS
C*+* wlfH NOX = 100.01 IF THF PROGRAM OOFS NOT CONVERGE « NOX * 100. Of
Z(9) «U( 1 01 > ,V( 101 )
C« f« rj^sr DCTERM1ME THr MINIMUM ZETAIW WITH INFINITE MASS TRANSFER
IFtE.LT. 0.0) GO TO jO
IFfPgY.Gf:. 0.0) •"•0 TO 3
IFfE.'-O. 1.0) GO TO 2
DUM - EXP(E*PEX*( 1 .0/E - J.OM
1 ZcTAM = (E»OUM - DUM,/(E«F - HUM)
GO TO 5
ro T0 5
3 I^CE.EO. 1 .0) GO TO 4
OUM r F.XP«P«iX«PEY*n ,0/E - 1.0)/(PF.X * PEY/E ) »
r,O T^ 1
4 ZETAy = (PEX + PEY) / <2.0*(PEx ••• PEY) + PPx«PEY )
? I- (ZFf A.GT.ZtTAM) GO TO 1O
ft NOX -5 100.01
GO TO (SO
C*»* MO-V n-r,TW Tptftt. AND ERROP SFARCH FOR NOX
1O MA = 4. CO
CALL ECT4 (PEX,.DEY.c- .MA.ZETACAL, 1 .Z.U.VJ
IF( ARS< 7ETACAL - Z£TA>.LT. O.OOOCl) GO TO 16
NfJ = NA + O.I
TALL rETACPE>'.pEY,E.K|e«7ETAS. 1 .Z.U.V)
DrpM = 10.0»AI_OG< ZET^BXZFTACAL)
NOX r N'A - ALOC.
-------
ANALYSIS* Of- L'AIA F Of- f.X FfACT JON KUN KUI'tatk SSJ2A
Pfcy = 2.01/0 f-E Y = iMUt-. « COMPONENTS
ACEK'NE t = ,b4(>i. Af P ?t'1A = .64747 6JVJNC- NOX = .694
•H-C-HLI 2 « 0.0 I 2((. i> 2 2(0.P) 2(0.9) 2(1.0)
.67871 «e>62b'' .65160 .64747
0 .68449 .(
J rivn y -- l)X = CrW'fcl I|M MOX CAUStO ItY A 1 HEkCf.M OtCRtASL IN X
i^tlA = r.Sbt.-l'^ t bt - J,30L-(C< t UPtX - "'.(^f.-04 t O^tY = 0.
OK MtK E ~ ^.JlfiC AND 2MA = .^OtStJ ClVlfJC- MUX = 1.403
Itf 2(r.C) *(C•J) 2(u.?> 2(0.3) 2(n.^) 2(C.7) 210.6) 2(0,9)
^'J X11' «7?3E'T «b7ci;0 . fccl'Sl* «b/491 «496lJ .4 3 61' .4l4t4 .4007B
f*l l>0 ).('0(/C*0' ,91hi,t ••- . i«»J7S . ?7t)H* .6't3?jb ' «t3346 .4H413 .43^??!
»SH.S1TJV11Y -- IM = CM«(-'l;t. IN MOX CAUSLU bY A 1 HthCFM Ul.CRfc.ASfc IN X
• HMflLE Xln.D 2(t.*) /(l.?» 2(n.j) 2(O.S> 2(0.7) 2(0.8) 2(0.9) 7 < 1 .01
^ ) t t (' .7jYc9 .(>6i;ja .(^llf 1 .^('/»H/» .4H4f>U »4?3'J .'•Ol'^b .JB76) »3fi23t<
f"Ll(.' l»('plil'0 .SJlbib .oJt>71 »7f3( .3010'^ .29630
Kl I <-' 1.1(0^0 ,t(.S6l ./14?V .f.'Mfi. .h./hAf.'f .4?7n3 ..'7010
Sn'!VllY -- l'>. = CriANtc if i-iijX CAHSLU HY A 1 f'LKCF.M Ut-CKtASt. IN X
IVt.TA :: 1. (.<(>(-»( ? » l-t s b.l^it.OB i t'f'f y = 2.7 = 0.0383, a = 14.90 ft2/ft3
14.5%, RT = 6.73 seconds
K, Feed Product %
Water Water Removal
(ppm) (ppm)
3.51
3.08
0.70
4.80
17850
1450
2570
59.0
85.6
96.1
453
-------
Table G. Spray Column Data (Continued)
Spray column run # SS4B Temperature = 22.5 °C
Dispersed phase = Isobutylene
Continuous phase = Prepared waste water
Vd - 52.50 ft/hr, VG - 22.52 ft/hr, FS/FW = 1.381
p, = 0.5911 gm/cc, y, = 0.180 cp,
p = 0.9977 gm/cc, y = 0.947 cp, a = 41.5 dyne/cm
d = 0.1818 inch, = 0.0274, a = 10.86 ft2/ft3
Pe^ = 1.964, FF - 11.8%, RT = 6.58 seconds
Solute D, D
^ 9
(10 •ftVhr)
1. Phenol 23.7 3.51
2. o-Cresol 22.0 3.08
K, Feed Product %
Water Water Removal
(ppm) (ppm)
0.70 17850 4280 76.0
4.80 1450 68.0 95.3
Spray column run # SS5 Temperature = 21.2 °C
Dispersed phase = Isobutylene
Continuous phase = Prepared waste water
V, = 52.25 ft/hr, V = 23.31 ft/hr, F /F = 1.312
Q C S W
p, = 0.5927 gm/cc, y, = 0.182 cp,
Pc = 0.9979 gm/cc, yc = 0.976 cp, CT = 41-8 dyne/cm
d = 0.1825 inch, * = 0.0272, a = 10.73 ft2/ft3
Peo = 2.012, FF = 12.0%, RT = 6.56 seconds
C
Solute D, D K, Feed Product %
c 5 Water Water Removal
(10 'ftyhr) (ppm) (ppm)
1. Phenol 23.3 3.39 (0.7) 12900 2580 80.0
2. VinylAc 23.5 3.45 52.0 12000 97.0 99.2
Comments: The presence of vinyl acetate in the solvent
phase probably increased K for phenol.
454
-------
Table G. Spray Column Data (Continued)
Spray column run f SS6A Temperature = 21.7°C
Dispersed phase = Isobutylene
Continuous phase = Lube oil refining waste water
V = 21.56 ft/hr,
c
Fs/Fw = 2.535
Pe
= 92.24 ft/hr,
= 0.5913 gm/cc, y, = 0.182 cp,
= 0.9978 gm/cc, y = 0.965 cp,
= 0.1882 inch, = 0.0499,
RT = 6.81 seconds
a = 41.5 dyne/cm
a = 19.07 ft2/ft3
i ^ — «• • wx w | i j. — _L / • A » f JCVA w * u J. 0c;wwAjivAo
C
Solute D,
a
U05.ft2
1.
2.
3.
4.
5.
Acetone
MEK
Benzene
Phenol
o-Cresol
26.2
24.0
24.0
23.3
21.7
Dc
/hr)
4
3
3
3
3
.25
.61
.61
.44
.04
Kd
0
2
407
0
4
.63
.49
•
.70
.80
Feed
Water
(ppm)
37.0
232
170
23220
2040
Product
Water
(ppm)
16
12
7
4590
50
.0
.0
.0
.0
%
Removal
56
94
95
80
97
.8
.8
.9
.2
.5
Comments: 77 ppm of tert-butanol in product water.
455
-------
Table G. Spray Column Data (Continued)
Spray column run # SS6B Temperature = 21.7°C
Dispersed phase = Isobutylene
Continuous phase = Lube oil refining waste water
V, = 52.00 ft/hr, V^ = 21.56 ft/hr, F /F = 1.429
Cl C S W
pd = 0.5913 gm/cc, yd = 0.182 cp,
p = 0.9978 gm/cc, yc = 0.965 cp, a = 41.5 dyne/cm
d = 0.1818 inch, = 0.0271, a = 10.72 ft2/ft3
P
Pe = 1.867, FF = 11.6%, RT = 6.56 seconds
o
Solute D, D K, Feed Product %
_ - Water Water Removal
(10 'ftyhr) (ppm) (ppm)
1. Acetone 26.2 4.25 0.63 37.0 22.0 40.5
2. MEK 24.0 3.61 2.49 232 55.0 76.3
3. Benzene 24.0 3.61 407. 170 35.0 79.4
4. Phenol 23.3 3.44 0.70 23220 9550 58.9
5. o-Cresol 21.7 3.04 4.80 2040 328 83.9
Comments: 60 ppm of tert-butanol in product water.
456
-------
Table G. Spray Column Data (Continued)
Spray column run # SS7 Temperature = 22.5°C
Dispersed phase = Isobutylene
Continuous phase = Prepared waste water
Vc = 21.56 ft/hr, FS/FW = 1.867
0.5911 gm/cc, u, = 0.180 cp,
yc = 0.947 cp, a = 41.5 dyne/cm
4) = 0.0362, a = 14.12 ft2/ft3
Vd = 67.95 ft/hr,
Pc = 0.9977 gm/cc,
d = 0.1844 inch,
Pe^ = 2.091, FF = 13.8%,
c
Solute
(105-ft2/hr)
RT = 6.71 seconds
K, Feed Product %
Water Water Removal
(ppm) (ppm)
1. MEK 24.3
2. Crotonal. 24.3
3.69
3.70
2.49
2.48
1972
2094
440
482
77.7
77.0
Spray column run # SS8A Temperature = 21.8°C
Dispersed phase = Isobutylene
Continuous phase = Prepared waste water
V, = 36.06 ft/hr, V0 = 15.36 ft/hr, F /F = 1.393
*^ ^^ o Vf
Pd = 0.5919 gm/cc,
p = 0.9978 gm/cc,
i, = 0.181 cp,
i = 0.963 cp,
= 0.1789 inch, = 0.0183, a
= 41.6 dyne/cm
7.37 ft2/ft3
Peo = 1.151, FF = 8.1%,
c
Solute
RT
(105'ft2/hr)
1. Propionit. 25.7 3.93
2. n-BuOH 23.5 3.34
K
1.80
0.76
6.40 seconds
Feed Product %
Water Water Removal
(ppm) (ppm)
4092
6259
380
1344
90.7
78.5
457
-------
Table G. Spray Column Data (Continued)
Spray column run # SS8B Temperature = 21.8°C
Dispersed phase = Isobutylene
Continuous phase = Prepared waste water
V, = 67.95 ft/hr, V = 15.36 ft/hr, F /F = 2.624
U. C S Vr
Pd = 0.5919 gm/cc, y^ = 0.181 cp,
PC = 0.9978 gm/cc, yc = 0.963 cp, a = 41.6 dyne/cm
d^ = 0.1847 inch, = 0.0360, a = 14.03 ft2/ft3
P
Pe = 1.482, FF = 12.5%, RT = 6.67 seconds
C
Solute D, D K Feed Product %
Water Water Removal
C n ruauCJ-
(105'ftVhr) (ppm)
(ppm)
1. Propionit. 25.7 3.93 1.80 4092 161 96.1
2. n-BuOH 23.5 3.34 0.76 6259 742 88.1
Spray column run # SS9A Temperature = 22.7°C
Dispersed phase = Isobutylene
Continuous phase = Prepared waste water
Vd = 43.18 ft/hr, VQ = 51.27 ft/hr, Fg/Fw = 0.499
Pd =0.5908 gm/cc, Pd = 0.180 cp,
PC = 0.9976 gm/cc, yc = 0.962 cp, a = 41.5 dyne/cm
d = 0.1801 inch, - 0.0224, a = 8.97 ft2/ft3
Pec = 4.115, FF = 14.6%, RT = 6.54 seconds
Solute Dd D Kd Feed Product %
e - Water Water Removal
(103*ftVhr) (ppm) (ppm)
1. EDC 24.6 3.69 70.0 3014 225 92.5
2. IAA 22.2 3.10 3.53 2472 717 71.0
458
-------
Spray Column Data (Continued)
Spray column run # SS9B Temperature = 22.7°C
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
V, = 30.17 ft/hr, V = 51.27 ft/hr, F /Ft7 = 0.348
Q C S Vv
p, = 0.5908 gm/cc, u, = 0.180 cp,
p = 0.9976 gm/cc, u = 0.962 cp, a = 41.5 dyne/cm
d = 0.1773 inch, <(> = 0.0156, a = 6.33 ft2/ft3
Pec = 3.662, FF = 12.4%, RT = 6.51 seconds
Solute
1. EDC
2. IAA
D, D K, Feed Product %
_d ° Water Water Removal
(10 «ftVhr) (ppm) (ppm)
24.6 3.69 70.0 3014 659 78
22.2 3.10 3.53 2472 1169 52
.1
.7
Spray column run # SS9C Temperature = 22.7°C
Dispersed phase = Isobutylene
Continuous phase =Prepared Waste Water
V, = 17.17 ft/hr, V = 51.27 ft/hr, F /F = 0.198
Q C S vv
Pd = 0.5908 gm/cc, ud = 0.180 cp,
Pc = 0.9976 gm/cc, y = 0.962 cp, a = 41.5 dyne/cm
d = 0.1740 inch, * - 0.0089, a = 3.67 ft2/ft3
Pe = 3.072, FF = 9.7%, RT = 6.51 seconds
cis — -> • \J l £. f c C
Solute
1.
2.
EDC
IAA
(10
24
22
D
5
•
•
d
•ft2
6
2
— y , i -o i
Dc
/hr)
3.69
3.10
i Jt\ X
Kd
70.0
3.53
w •
-------
Table G. Spray Column Data (Continued)
Spray column run # SS10A Temperature = 23.4°C
Dispersed phase = n-Butane
Continuous phase = Prepared Waste Water
V, = 43.77 ft/hr, V = 51.27 ft/hr, F /F = 0.475
d C S W
Pd = 0.5549 gm/cc, yd = 0.197 cp,
Pc = 0.9974 gm/cc, y = 0.927 cp, a = 47.0 dyne/cm
d = 0.1317 inch, = 0.0217, a = 8.60 ft2/ft3
Pe = 3.852, FF = 14.1%, RT = 6.25 seconds
c
Solute D, D K, Feed Product %
a c Water Water Removal
(10 «ft /hr) (ppm) (ppm)
1. EDC 22.5 3.84 44.0 2890 735 74.6
2. IAA 20.4 3.22 1.44 2063 1185 42.6
Spray column run # SS10B Temperature = 23.4°C
Dispersed phase = n-Butane
Continuous phase = Prepared Waste Water
V, = 31.40 ft/hr, V - 51.27 ft/hr, F/F = 0.341
Q C S W
Pd = 0.5549 gm/cc, yd = 0.197 cp,
Pc = 0.9974 gm/cc, vc = 0.927 cp, a•- 47.0 dyne/cm
d = 0.1792 inch, = 0.0155, a = 6.24 ft2/ft3
Pe =3.468, FF = 12.0%, RT = 6.24 seconds
Solute D, D K, Feed Product %
52° Water Water Removal
{10 *ft /hr) (ppm) (ppm)
1. EDC 22.5 3.84 44.0 2890 1181 59.1
2. IAA 20.4 3.22 1.44 2063 1457 29.4
460
-------
Table G. Spray Column Data (Continued)
Spray column run # SS10C Temperature = 23.4°C
Dispersed phase = n-Butane
Continuous phase = Prepared Waste Water
Vd = 17.42 ft/hr, Vc = 51.27 ft/hr, Fg/Fw = 0.189
Pd - 0
PO • °
dp =o
Peo =
Solute
1. EDC
2. IAA
.5549 gm/cc, y, =
.9974 gm/cc, y =
.1759 inch, =
2.883, FF = 9.2%
Dd Dc
5 9
(105.ft2/hr)
22.5 3.84
20.4 3.22
0.197 cp,
0.927 cp, a = 47.0 dyne/cm
0.0086,
, RT =
Kd
44.0
1.44
a =
6.24
Feed
Water
(ppm)
2890
2063
3.53 ft2/ft3
seconds
Product %
Water Removal
(ppm)
1657 42
1668 19
.7
.1
Spray column run # SS11A Temperature = 21.4°C
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
Vd = 43.18 ft/hr, Vc = 36.80 ft/hr, Fg/Fw = 0.697
Pd = 0.5924 gm/cc, yd = 0.182 cp,
pc = 0.9979 gm/cc, yc = 0.972 cp, a = 41.7 dyne/cm
dp = 0.1806 inch, * = 0.0222, a = 8.87 ft2/ft3
Pec = 2.935, FF = 12.8%, RT = 6.49 seconds
Solute Dd DC Kd Feed Product %
5 2 Water Water Removal
(10 -ftVhr) (ppm) (ppm)
1. EDC 24.2 3.64 70.0 2892 370 87.2
461
-------
Table G. Spray Column Data (Continued)
Spray column run # SS11B Temperature = 21.4°C
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
Vd = 17
Pd - 0.
PC = o.
dp = 0.
Pec = 2
Solute
1. EDC
.17 ft/hr, V =
c
5924
9979
1744
.191,
(10
24
gm/cc , y , =
gm/cc, yc =
inch, =
FF = 8.0%,
Dd Dc
R 9
•ftVhr)
.2 3.64
36.80 ft/hr,
0.182
0.972
0.0088,
RT =
Kd
70.0
cp,
cp,
a =
6.46
Feed
Water
(ppm)
2892
F /F =
S W
a = 41.7
3.63 ft
seconds
0.277
dyne/cm
2/ft3
Product %
Water
(ppm)
1098
Removal
62.0
Spray column run # SS12A Temperature = 21.6°C
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
V, = 52.25 ft/hr, V^ = 23.26 ft/hr, F /F = 1.331
5 Vr
= 0.182 cp,
cr = 41.5 dyne/cm
P
p
d
d
c
P
Pe
=
=
=
c ~
0
0
0
•
•
•
2
5914
9979
1819
.017,
Solute
1
2
3
4
(10
. Acetone
•
•
MEK
Cro tonal.
. n-BuAc
gm/cc
gm/cc
inch
f
FF
Dd
5.
26
24
24
20
ft
.2
.0
.0
.4
2
I
, u
/ p
(j)
d
c ~"
=
= 11.9%
Dc
/hr)
4.
3.
3.
2.
22
87
88
71
0
/
= 0.970 cp,
a = 10.78 ft2/ft3
RT = 6.56 seconds
K
' Feed
Water
(ppm)
0.63
2.49
2.48
8.0
2058
4167
4422
2212
Product %
Water Removal
(ppm)
1307
1579
1620
620
36.5
62.1
63.4
72.0
462
-------
Table G. Spray Column Data (Continued)
Spray column run # SS12B Temperature = 21.6°C
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
Vd
pd
P
Pe
= 83.16 ft/hr,
= 0.5914 gm/cc,
= 0.9979 gm/cc,
= 0.1868 inch,
= 2.435,
Vc -
23.26 ft/hr,
F /F =
s w
2.189
Ud = 0.182 cp,
Uc = 0.970 cp,
<(> = 0.0447, a = 17.24 ft2/ft3
= 41.5 dyne/cm
G
Solute
Dd
c;
i. j. J.U
D
c
•?
. JT
>f I\A — o. /o seconds
Kd Feed Product
Water Water
(lO-'ff/hr)
1.
2.
3.
4.
Acetone
MEK
Crotonal.
n-BuAc
26
24
24
20
.2
.0
.0
.4
4.
3.
3.
2.
22
87
88
71
0
2
2
168
.63
.49
.48
.0
(ppm)
2058
4167
4422
2212
(ppm)
1090
1084
1152
431
%
Removal
47
74
74
80
.0
.0
.0
.5
Spray column run #
Dispersed phase =
Continuous phase =
Vd = 52.25 ft/hr,
Pd = 0.5922 gm/cc,
p = 0.9979 gm/cc,
C
d = 0.1822 inch,
SS13A Temperature = 21.6°C
Isobutylene
Prepared Waste Water
1.333
Vc = 23.26 ft/hr, Fs/Fw =
Pd = 0.182 cp,
Pc = 0.967 cp,
= 0.0272, a = 10.75 ft2/ft3
= 41.7 dyne/cm
Pe,, = 2.013, FF =
C
Solute D, D
d c
1.
2.
3.
4.
(
Acetone
Benzene
n-BuOH
n-BuAc
105
26
24
23
20
12.0%, RT
Kd
•ft2/hr)
.2
.0
.4
.4
4.
3.
3.
2.
24
60
44
73
0
407
0
168
.63
•
.76
•
RT = 6.56 seconds
Feed Product %
Water Water Removal
(ppm) (ppm)
2099
300.7
4174
2067
539
5.7
924
6.7
74.3
98.1
77.9
99.7
463
-------
Table G. Spray Column Data (Continued)
Spray column run # SS13B Temperature = 20.7°C
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
VJ = 26.86 ft/hr, VG =
= 0.5933 gm/cc, y, =
= 0.9981 gm/cc, u =
= 0.1774 inch, $ -
= 1.581, FF = 8.0%,
d
Pd
PC
P*c
Solute
F /F =
s w
0.686
23.26 ft/hr,
0.183 cp,
0.988 cp, a = 41.9 dyne/cm
a = 5.54 ft2/ft3
(105.ft2/hr)
0.0137,
RT =
K^
1. Acetone 25.9 4.15 0.63
2. Benzene 23.8 3.53 407.
3. n-BuOH 23.2 3.36 0.76
4. n-BuAc 20.3 2.67 168.
6.40 seconds
Feed Product %
Water Water Removal
(ppm) (ppm)
2072 1492 28.0
288.3 20.0 93.1
4122 2667 35.3
2110 197 90.7
464
-------
Table G. Spray Column Data (Continued)
Spray column run # SS14 Temperature = 21.6°C
Dispersed phase = Isobutylene
Continuous phase = Cresylic Acid Recovery Waste Water
V, = 60.59 ft/hr, V =20.14 ft/hr, F /F = 1.785
CL C 5 W
pd = 0.5922 gm/cc, y^ = 0.182 cp,
pc = 0.9979 gm/cc, yc = 0.967 cp, a = 41.7 dyne/cm
d = 0.1837 inch, = 0.0319, a = 12.50 ft2/ft3
Pe = 1.853, FF = 12.5%, RT = 6.63 seconds
C
Solute D, D K, Feed Product %
c , Water Water Removal
(10°.ft/hr) (ppm) (ppm)
1. Phenol 23.3 3.43 0.70 579 163 71.8
2. o-Cresol 21.7 3.01 4.80 307 31.2 89.8
3. m, p-Cresol
21.7 3.01 2.70 291 25.2 91.3
4. Xylenols 20.4 2.71 7.0 227 10.0 95.6
Comments: Feed Water COD = 4050;
Product Water COD = 1070.
465
-------
Table G. Spray Column Data (Continued)
Spray column run # SS15A Temperature = 21.3°C
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
V, = 52.25 ft/hr, V = 23.26 ft/hr, F /F = 1.334
Cl C S W
Pd = 0.5926 gm/cc, VU = 0.182 cp,
PC = 0.9979 gm/cc, y = 0.974cp, a = 41.7 dyne/cm
d = 0.1823 inch, 4> = 0.0272, a = 10.74 ft2/ft3
Pe = 2.011, FF = 12.0%, RT = 6.56 seconds
C
Solute D, D Kd Feed Product %
c Water Water Removal
(10 »ft /hr) (ppm) (ppm)
1. Acetone 26.1 4.21 0.63 2049 590 71.2
2. n-BuAc 20.4 2.70 168. 2113 11.8 99.4
Spray column run # SS15B Temperature = 21.3°C
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
V, = 26.86 ft/hr, Vo = 23.26 ft/hr, F /F = 0.686
Q C S W
p, = 0.5926 gm/cc, yd = 0.182 cp,
p = 0.9979 gm/cc, y = 0.974 cp, a = 41.7 dyne/cm
d = 0.1770 inch, = 0.0137, a = 5.56 ft2/ft3
Pe = 1.588, FF = 8.0%, RT = 6.42 seconds
c
Solute
1. Acetone
2. n-BuAc
(10
Dd
5 .
26
20
ft2
.1
.4
Dc
/hr)
4.
2.
21
70
K
d
0.63
168.
Feed
Water
(ppm)
2049
2113
Product
Water
(ppm)
1341
144
%
Removal
34
93
.6
.2
466
-------
Table G. Spray Column Data (Continued)
Spray column run # SS16 Temperature = 22.7°C
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
Vd = 52.25 ft/hr,
Pd = 0.5908 gm/cc,
p = 0.9976 gm/cc,
d = 0.1818 inch,
Pe = 2.026, FF = 11.9%
V = 23.26 ft/hr,
G
VFw =
1.330
= 0.180 cp,
a = 41.5 dyne/cm
yc = 0.942 cp,
* = 0.0273, a = 10.81 ft2/ft3
RT = 6.58 seconds
Solute
Dd
K. O
(10"
1.
2.
3.
4.
Acetone
MEK
Crotonal.
n-BuAc
26
24
24
20
•ft
.6
.3
.4
.7
Dc
Kd
/hr)
4
3
3
2
.37
.71
.72
.81
0
2
2
168
.63
.49
.48
*
Feed
Water
(ppm)
1986
3957
4247
2138
Product %
Water Removal
(ppm)
1093
1257
1373
508
45
68
67
76
.0
.2
.7
.2
Spray column run # SW1A Temperature = 20.0°C
Dispersed phase = Prepared Waste Water
Continuous phase = Isobutylene
Vc = 17.17 ft/hr,
VFw =
Vd = 51.27 ft/hr,
Pd = 0.9982 gm/cc, yd = 1.005 cp,
p^ = 0.5942 gm/cc, u = 0.184 cp,
c» c
d = 0.1947 inch, 4> = 0.0233,
Pe_ = 1.153, FF = 12.9%, RT = 5.73 seconds
0.199
a = 42.0 dyne/cm
a = 8.62 ft2/ft3
i^ -i. • j_~j ~j / j. ±. a.*. * ./ -Q / r\j.~ -J . / J ocowuua
C
Solute
1.
2.
EDC
IAA
(10
3
2
D
5
•
•
a
• ft2
50
94
Dc
/hr)
23.8
21.6
Kd
70.0
3.53
Feed
Water
(ppm)
2763
1920
Product
Water
(ppm)
1129
1327
%
Removal
59.
30.
1
9
467
-------
Table G. Spray Column Data (Continued)
Spray column run # SW1B Temperature = 20.0°C
Dispersed phase = Prepared Waste Water
Continuous phase = Isobutylene
V, = 51.27 ft/hr, V = 43.18 ft/hr, F /F = 0.501
Cl C S w
Pd = 0.9982 gm/cc, y, = 1.005 cp ,
p = 0.5942 gm/cc, y = 0.184 cp, a = 42.0 dyne/cm
d = 0.1947 inch, = 0.0236, a - 8.72 ft2/ft3
P
Pe = 2
c
Solute
1. EDC
2. IAA
.912, FF
Dd
^i
(105-ft
3.50
2.94
= 16.1%,
D
2/hr)
23.8
21.6
RT = 5.80
Kd Feed
Water
(ppm)
70.0 2763
3.53 1920
seconds
Product
Water
(ppm)
860
1076
%
Removal
68.9
44.0
Spray column run # SW2A Temperature = 18.6°C
Dispersed phase = Prepared Waste Water
Continuous phase = Isobutylene
V, = 36.80 ft/hr, V = 43.18 ft/hr, F /F = 0.700
Cl C* O rf
p, = 0.9985 gm/cc, y^ = 1.040 cp,
p = 0.5960 gm/cc, y = 0.186 cp, a = 42.3 dyne/cm
d = 0.1935 inch, $ = 0.0168, a = 6.24 ft2/ft3
P
Pe = 2.572, FF = 12.8%, RT = 5.74 seconds
Solute
1. EDC
Dd
(105-
3.37
3.17
2.51
D
C
ft2/hr)
23.4
22.7
19.8
Kd
70.0
0.76
168.
Feed
Water
(ppm)
3061
2065
2184
Product
Water
(ppm)
537
1546
456
%
Removal
82.5
25.1
79.1
468
-------
Table G. Spray Column Data (Continued)
Spray column run # SW2B Temperature = 18.6°C
Dispersed phase = Prepared Waste Water
Continuous phase = Isobutylene
V = 30.17 ft/hr,
c
F /F =
s w
V, = 36.80 ft/hr,
a
Pd = 0.9985 gm/cc, ud = 1.040 cp,
p = 0.5960 gm/cc, y = 0.186 cp,
C G
d = 0.1936 inch, 4> = 0.0167,
Pe_ = 1.793, FF = 11.5%, RT = 5.71 seconds
0.489
a = 42.3 dyne/cm
a = 6.20 ft2/ft3
\^r
Solute D^ DC
(105'ft2/hr)
1. EDC 3.37
2. n-BuOH 3.17
3. n-BuAc 2.51
23.4
22.7
19.8
K, Feed
Water
70.0 3061
0.76 2065
168. 2184
Product
Water
(ppm)
605
1581
547
%
Removal
80.2
23.4
75.0
Spray column run # SW2C Temperature = 18.6°C
Dispersed phase = Prepared Waste Water
Continuous phase = Isobutylene
Vd
pd
Pc
dp
Pec
= 27.23 ft/hr,
= 0
= 0
= 0
=
.9985
.5960
.1909
1.637,
Solute
1.
2.
3.
EDC
(10
3.
n-BuOH 3 .
n-BuAc 2 .
gm/cc
gm/cc
inch,
FF
Dd
V
1
,
i
zsz
Dc
\*
= 30.17 ft/hr,
C
y = i
yc = 0
= o
9.3%,
.040 cp
.186 cp
.0123,
RT =
Kd
vi
5«ft2/hr)
37
17
51
23
22
19
.4
.7
.8
70.0
0.76
168.
>
o
a -
5.71
Feed
Water
(ppm)
3061
2065
2184
F /F =
s' w
= 42.3
4.65 ft
seconds
0.661
dyne/cm
2/ft3
Product %
Water Removal
(ppm)
519
1507
473
83
27
78
.0
.0
.3
469
-------
Table G. Spray Column Data (Continued)
Spray column run # SW3A Temperature = 18.3°C
Dispersed phase = Prepared Waste Water
Continuous phase = Isobutylene
V, = 36.80 ft/hr, V = 43.18 ft/hr, F /F = 0.701
CL C S W
P, = 0.9985 gm/cc, y, = 1.048 cp,
p = 0.5963 gm/cc, y = 0.187 cp, a = 42.3 dyne/cm
d = 0.1936 inch, = 0.0168, a = 6.24 ft2/ft3
Pec = 2.571, FF = 12.8%, RT = 5.74 seconds
Solute D, D K, Feed Product %
c Water Water Removal
(10 'ft /hr) (ppm) (ppm)
1. EDC 3.34 23.3 70.0 3128 1529 51.1
Spray column run # SW3B Temperature = 18.3°C
Dispersed phase = Prepared Waste Water
Continuous phase = Isobutylene
V, = 36.80 ft/hr, V = 30.17 ft/hr, F /F = 0.490
d c s w
p, = 0.9985 gm/cc, yd = 1.048 cp,
p = 0.5963 gm/cc, y = 0.187 cp, a = 42.3 dyne/cm
d = 0.1936 inch, $ = 0.0167, a = 6.20 ft2/ft3
Pe = 1.793, FF = 11.5%, RT = 5.71 seconds
G
Solute D, D K, Feed Product %
d c Water Water Removal
(KT'ftyhr) (ppm) (ppm)
1. EDC 3.34 23.3 70.0 3128 1501 52.0
470
-------
Table G. Spray Column Data (Continued)
Spray column run # SW3C Temperature = 18.3°C
Dispersed phase = Prepared Waste Water
Continuous phase = Isobutylene
V, = 36.80 ft/hr, V = 17.17 ft/hr, F/F = 0.279
Q. C S Vr
Pd = 0.9985 gm/cc, yd = 1.048 cp,
p =0.5963 gm/cc, y = 0.187 cp, a = 42.3 dyne/cm
d = 0.1936 inch, $ = 0.0166, a = 6.16 ft2/ft3
Pe = 1.018, FF = 9.9%, RT = 5.67 seconds
C
Solute D, D K, Feed Product %
- , ° Water Water Removal
(10 -ftVhr) (ppm) (ppm)
1. EDC 3.34 23.3 70.0 3128 1535 50.9
Spray column run # SW3D Temperature = 18.3°C
Dispersed phase = Prepared Waste Water
Continuous phase = Isobutylene
V, = 27.23 ft/hr, V = 30.17 ft/hr, F /F = 0.662
U C S Vv
Pd = 0.9985 gm/cc, ud = 1.048 cp,
PC = 0.5963 gm/cc, PC = 0.187 cp, a = 42.3 dyne/cm
dp = 0.1910 inch, = 0.0123, a = 4.65 ft2/ft3
Pec = 1.637, FF = 9.3%, RT = 5.71 seconds
Solute Dd DC Kd Feed Product %
5o Water Water Removal
_ (10 -ftVhr) _ (ppm) (ppm)
1. EDC 3.34 23.3 70.0 3128 1406 55.1
471
-------
Table G. Spray Column Data (Continued)
Spray column run # SW3E Temperature = 18.3°C
Dispersed phase = Prepared Waste Water
Continuous phase = Isobutylene
V, = 27.23 ft/hr, V = 43.18 ft/hr, F /F = 0.947
CL C S W
Pd = 0.9985 gm/cc, y, = 1.048 cp,
p = 0.5963 gm/cc, y = 0.187 cp, a = 42.3 dyne/cm
d = 0.1910 inch, = 0.0124, a = 4.68 ft2/ft3
Pe^, = 2.347, FF = 10.5%, RT = 3.74 seconds
C
Solute D, D K, Feed Product %
j- 7 ° Water Water Removal
(10 'ftyhr) (ppm) (ppm)
1. EDC 3.34 23.3 70.0 3128 1395 55.4
472
-------
APPENDIX H
EXPERIMENTAL DATA FROM RDC EXTRACTOR
The data which were directly determined at
steady state in each run conducted in the RDC extrac-
tor included the solvent and water flow rates, the
diameters of the discs and stator holes, the compart-
ment heights and whole column height, the rotational
speed of the discs, the temperatures of the two
streams leaving the RDC, and the concentrations of
each solute in the feed and product water. In some
experiments the solvent hold-up and the solute
concentrations in the loaded solvent were also
measured. Since the difference between the two temper-
ature measurements was less than 1°C, the average
was taken as the column temperature. These experi-
mental data were then combined with estimates for
physical properties to compute experimental mass
transfer coefficients. Since these calculations
were lengthy and repetitious, a computer program
was developed and used. In this appendix the
calculational procedure and computer programs are
described, then the results for 15 runs conducted
in the RDC extractor are listed.
473
-------
Estimates of Physical Properties.
The aqueous phase density and viscosity at
the column temperature were taken as those listed
by Weast (1970) for pure water. The density and
viscosity for the volatile solvent phases were deter-
mined from data listed for the pure solvents (API,
1963). The interfacial tensions were estimated
by the method of Donahue and Bartell (1952) as
discussed in Section VII. The density and viscosity
for n-butyl acetate were from Toropov (1956).
The interfacial tension at 20°C was taken from
Logsdail, et al. (1957), and the correlation of
Donahue and Bartell (1952) was used to correct this
value to the column temperature. For mixtures of
n-butyl acetate and isobutylene, the density was
estimated by assuming no excess volume of mixing, the
viscosity was estimated from the molar average of
In y., and the interfacial tension was estimated
from the molar average of a^ , where y^ and a., are
for pure components. The density and viscosity of
2-ethyl hexanol were taken from Marks (1967). The
interfacial tension was assumed equal to that for
n-octanol as given by Weast (1970).
The diffusivities for each solute in water and
in the solvent were estimated by the method of
Scheibel (1954) assuming that the values at infinite
dilution would apply.
Computer Programs.
The majority of RDC extractor data reduction
was done in the Fortran program EXPTRDC and subpro-
grams RDC and POFUNC. Program EXPTRDC begins by
474
-------
reading in the superficial velocities of each phase,
VD and VC, the column temperature, TEMP, the rota-
tional speed, RPM, the measured hold-up, PHI, and
the column dimensions including disc diameter, DI,
stator hole diameter, DS, column diameter, DC, compart-
ment height, HC, and column height, COLHT. When
the hold-up was not measured or when Glg is assumed
for some other reason, the value read into PHI is
(1 + G18). The program senses that PHI is greater than
1 so that the correct value for PHI must be calculated
using 6lfi. All physical properties are next read in.
The subroutine RDC is called to make all design
calculations for the RDC extractor. Program EXPTRDC
then prints out the results.
Subroutine RDC uses data stored in various
common blocks to design the RDC. It first calculates
PHI or G18, depending upon which is specified and
which must be calculated, using the correlation of
Strand, et al. (1962). This step includes the
calculation of drop diameter, DP, slip velocity,
VS, and power per unit mass, POM. The function
subprogram POFUNC is used during this calculation
to determine the power number using the disc Reynolds
number, RE, after the correlation of Reman and
van der Vusse (1955).
Subroutine RDC continues by checking for flooding
both by the correlation of Logsdail, et al . (1957)
and by the method of Strand, et al . (1962). The
two Peclet numbers, PEC and PED, are calculated
using the equations developed in Appendix B. The
calculations in RDC are completed by determining
475
-------
theoretical estimates for the solvent-phase number
of transfer units, NS, for the water-phase number
of transfer units, NW, and for the overall water-phase
number of transfer units, MOW. The interfacial
area per unit of extractor volume, A, and the solvent
droplet rise time, R, are calculated. Theoretical
estimates for NOW are made for stagnant drops and
for circulating drops using the equations of Strand,
et al . (1962). Theoretical estimates are made for
NW using the model of Calderbank and Moo-Young (1961).
This estimate is then used with NS calculated for
stagnant drops to calculate a third estimate for NOW.
The final step in determining the experimental
values for the overall water-phase number of transfer
units was accomplished using the program EVAL and
subroutines FINDM and EETA. These programs were
listed in Appendix G. The values of PEC and RED
determined in RDC are read in. A value of E, the
extraction factor, and ZETA [n. in equation (1),
calculated using measured concentrations] is read into
EVAL for each component. By using the search
subroutine FINDN, equation (3) is solved implicitly
to determine NOW for each component.
A listing of the program follows. At the end
a typical output is shown for run RS13.
476
-------
PROGRAM EXPTRDC
exPTROC( I Mi"'DT .OUTPUT )
C* KXPTHOC ACCEPTS CATA FIV^M AN t XPFfP I r-TNTAL RUN ANH CALLS TOO THF.
c»»* TH.C: Cf.L.c.'jLATiON or Liuon DIAMF^TEP. PTCLF-'T NU^LJEHS* AND MA<;,S TRANSFER
c-f** eoTr'- tr ir NTT,.
covyp-j/rj r .v~N/r>i . r>r. . DC« Hc .COUNT , CQ
JOPt;/r>|i f.".r» • r>i;Nsc.nrLDf N , r, I',MA, vt^cn« vi SCC.MCOMPS.
" ) .OI<--FC ( f. ) .vr> ) t NSC I DC < 6 ) . NOV. I RC ( 6 ) « NW TJR9 ( 6 ) .
? WSnM>!>(6) .NPWTUP'M A )
REAL *«••»**»»
1 FO'}M/\T ( AH . IX. 1 ! «FH . A , r 1 0.4,Ft'> . 1 . r 1C. 1 • F | C> • 6 i HANAl_V.I S °F RC;C ^XTRATTrON QUN .A'.iTX.
] <,HoAGf' lX//|4X.a(lH»<«* FypcrjIMl MTAL COND I T | ON'.. /?3X .
2"»iJV! I OC 1 TV Ol SPf'WsL'O PHfl<-.L' -- .f'ft.^.f.H r TX>Jt vr^v . PfHvf I OT I T V CON
r>3X . ]4HTFVipf I-?.-, r^.^r' = .FA.].
,/iy . iCH(MTAr,u^E :>) /?::v.6HC.! o = .FT.
lAXi??H»*«v COLU'-'.'J DIV.FN'MnN <:./-. -IX,
??X . P3MST ATCR HOLC
r.!? = ,Fr..2.rH INCH/P3X.
I MCH/23X. 1 6t-TCLUMN HflGMT = .
I'ICH//)
7 rOOMAT ( IHI .20X.3M-!-~G:.r
lOHOl-n-l"1 - «FV. 6 ./,X. !2M(CALr.ULATrD)/?.''X.6MO] ." = .f7.<1,nx.
6 1l,,fK^TINAl-n)X//UX.?.?H»»»^ COLUMN ^ J MFNM O' :SX??X .
7 l6nr)ll>C OIAXfi'TI^ = .r-T..3,5•'->" IfCH/S.-JX . IgMCOLUVN DI AMrTEIH = .F';>. r-i'iM INCH/?;iX,
9 pnVu'-i'-'.VvlMttMT HLlCt'T = .Fn.r.LH INTH/r3X. K'-'COLUMN 'iriGMf = .
I f7 M.'J'I I'-lfM/X)
)'.V.?/IH»*»» fMYSiCAL prjorrfJT ies/?:'x,?6'<;if'N-.uTY oispr«<;r:) PH
.f7.'".f^X.27Hi>! NSM Y CCNT I NlK'-l )f. PHA'/1' = .FV.',,
op i'-'/r.cx?3x.e.-'MiNrr P-AC i AL TKNSION = •F7.:>tfi'
CS 1TY C>I?prtJ3C'' f'^Acr = ,F7.'..:(H CP/.'?\.
ITV COMT1 K'UO'JS PHf.Sr r .P7.1..T-I CP//>
r ( iox,;'9H*»*« FLoo.iirj", cHAr?/*c-rrRi sTicsxj-T-.-.
'Hi FLOPPING -- ,F7.6./?^x. I^.MVK (LOi-,rr>A u ) - .F-7.?.
^ 6H f-T/Ml-V.'^X.r-VMf-RACTION O'-' FLOOOIN:.; (L) •- . I'1':- . 4 . /?.~M*. .
-, i4HvK (STL-ANT) - ,r7..-'.isn r--T/Miv/r:N or rL.oor.MNr, d
flr ,f:r(./t./.'JX, lftH.-)k'Of rMAM-TKR =• .T7. 6. '-H I N( H/.-.IX , r?Hl W AN^. I T 1 ON
•^OP IJI A - .' V. 6, 'iH 1 NCM//3X. 1 7Hf'^.s! '-' I '• P MA'.l, • , r 7 . ,n ,
f, ,/lM I l-LHf /HU-LI»M/-.1V..-'|HOSC. II I AT ION FACTCU - .TV. I//)
477
-------
10 FORMAT ( 14X..T3H**** AY I AL MIXING CH APAC TFR I e,T 1C 5V??x ,
11 FORMAT ( IHI .?ox.36HANALY5is, OF ROC FXTRACTION PU>
j 6'-'p^r,^ ?//xi/»x,62H.>*»* SOLUTE DIFFUSIVITIES. DIM'
?ClFNTS. »NO £ ( 1 5/28X, 1 r-,H10*»<^ • SOFT/HP/17X.6HSOH.I-"
3 SX , 5HDC ( I ) . 1 1 X i 5H?.?.riO.?,F16.2,F13.?)
13 FORMAT < 1HO/1/X.28H**** MASS TRANSrrR r ST I MATF5V?3X , .
1 lOn SOFT/CUMV23X. i ?HRI p,F. TIME - . 1 PK 1 0 . T. 3H HP)
14 FORMAT < 1H3 4 14X. 1 7H*« STAGNANT DROPS/ 1 7X . 6HSOLUTF . <:> •
1 PX.3HNOW/J
15 FORMAT ( 16X.A8.2X, 3F1C..3)
16 FOPM^T I 1HO, 10X«20H»* CIRCULATING DROPS/ 1 ^x , 6HSOLUT r .
1 ?H)jS« 9X. 1HNOWX)
17 FORMAT! 1HO . HX. 1SH»* TUROuLTNT DROPf,/ l 7X . fcHSOLurr . .;-
18 FOPM^T ( !HJXl^Xt26H***« FTXPFR1 MENTAL RFMOV ALS/ 1 7x . ( -\'
2O FOC3MAT I IHXt 1SX. A8.F1 1 . 1 .r ! ? . 1 iF 12.n>
C«*» CUT Of FORMAT ^TATFMrnTr, t* «**#****•***»**»*»**««* * i
C»«* RCAr IN EXPST?! V^MT AL (-ON'MTJONf)
31 Rfc'ACl 1 «DUMK\)Mi JUVP i Vp . VC . TrMP,RPM,f>m .D I .OS-DC .HC .
IFfVp.EO. 0.0) GO TO l^o
CALCPHI = .r'ALStT.
IF(p(_(t ,GT, 1.01 CALc^KI = .TRU1?.
ir( J(jfp'CO. 11 r-n TO 36
c« t* PF:AO IN PHYSICAL PPOPERTY HATA (GMXCC.GM/CC .DvNir/r^
P"AO ? •DFNSfJ.tlCNSC .-S IGMA.V1 SCO.V! SCC
C*«* DfAo IN OlFFUSl V IT IES
10 3K 1=1 .NCC'HPS
c**» TALL
16 CALL PDClO.ri.OCD. O. 1 .0. 1 .0)
C<»» PRINT ODFRATING CONO I T I OST,
Ir(CALCPHl> GO TO ao
PRINT 6.RUNMUN'.Vr) . VC . TFMP , RPM . P H I . G 1 R . D I . OS . DC . HC .
GO Tp 41
40 ofviNT 7.fio(v.KK.JM,V3.VC.Trvp.PPM.PHI . r, 1 R , o I . Dr, . DC . HC . •
C.I *» nr?IrvT PHYSICAL PQOPFRr 1 r ;'.
a) "PINT 0'DF.NSO,r>"NSC tsic-vA, vir,cn .vi ro f\f, i -i .NCOK.PS
478
-------
6 PP1MT 1 -> «NAME< I ) • NWSfAGl I ) tNSSTAGl 1 ) .NOWSTAGI I )
DO 47 1=1 .NCOWPS
47 ptVjNT '.^.NAME( I ) ,MV'C|RC( I ) «N<;CIRC( 1 ) «NOWCIRC( I I
pc?tNT n
DO 4s ! ';1 .'JCOMPS
<:•«** RfTAn JM /'MO POINT. CXPfrRIMFNTAL REMOVAL EFFICIFNCIFS
PPJI-JT HI
DO r>o ' = ' •'•J-Q^pS
IF( J(j'*l'•?.<>• 1 ) GO TO 50
PKAD 1 '' «PPMr- ( 1 ) < PP'IP ( I )
ET'A«T> = PfJMPl I )/PPMr( I )
^O T'RINV ? O.NAM.1:* I ) . PPMr ( I ) ,PPMP( I ) . r.TA( I )
r,o TO 3)
ion co\'riM'i'
479
-------
SUBROUTINE RDC
INH: ROCIK^Y.FLOOD.MTTYPE; « ANW
c* RDC USES PATA sioReo IN COMMONS TO orr. IGN AN RDC EXTRACTOR
. 0 5 . Dr- , HC , CCLHT , CR
Of-NSC , DELDCN . S I GM A . V I SCO • V I SCC , NCOMPS .
Ol Ffn (6 ) ,fj ICFC( 6> «vD(6)
/VD . VC tfiPK . W^ . POM . PH I • G 1 8 , DP . DPT . OSCF ACT . A ,R
C.PHlF-.VK.FF?, ,VKL .FFL . E.T A ( 6 ) . fl ( 6 ) . NWf.T AG ( 6 ) •
1 NS!STf.G(6) .WO'A'f.TAC. («,) . NWCIRCtfe ) t TJSC IPC ( 6 ) .NOWC1RC (6) .NWTUQR(6> ,
REAL KO»N'/r-T/.G«NSSTAG»NO'STAC> ,sjwC I Rr tNSC I Rr «NOWC IRC .NWTI IQH .Nr-,TURP .
1 MOwTLlRU.NOX
OlNfMS ION if (O ) tVIA< 1 0 j ) ,Ufl( 1 0 J )
LOGICAL. TLOOO
1 rOHM/,T ( lHO,4£iH**** *»#« *«•*» PH I LOOP DID NOT CONvFRGE •««*//)
^LOOD = .FAl.SC.
AA = VO/VC
RHIF = •« *?
AB = ( OS/DC )*»2
AC = ((US + O\ )XDC) *sORT( ( (OS - DI)XDC)»«? + C )**? )
IS = AM INI ( AA.AB. AC)
IF'KEv«fo'l> GO TO IQ
Ci ** jr KTy.NF.l. CALCULATp PAR A«.'FTTERP P^P-NDING ON PHYSICAL PrjOiTWTirS
P = ( I .O?9r^O^*D^NSf;*rrNSC *SIGMA**3/VI 5rCC*«a/pK:LOr~N>** C. 1 «">
DPT s 9. 1?7E-0?*SORT( SI r^'A/DFLDEIN/P)
AD = I . AA^t->f A«ExP( 0.f;B*ALOG(r)FLDCN> - C. 4?5#ALOG ( OF.NSC ) - 0.11 +
I *Lf)G< VI SCO )
C-f** CALCULATE G10 OR PHI, D^ . VS. AND VK
10 PH = 1 O.Vn'DItD^RRM^DrN'Sr./VlSCC
POM = 191O. e-POFUNC (Rf)»RPM*»?*OI «*r./HC/DC»#2
AA = f:XP(O.ft»ALO&( ) * 1 ? . 0
[F(PuI tf-T. 'l ,0) GO TO 1?
VS = VD/PHI -i vc/(t.o -
IF(PnI .GC. PHIF) GO TO 1
v< = vr.xt i .0 - PHI )
VT = VK/CR
T/AO) /0.7)
) DP = DoT
G10 = DP/A/V
'•.o TO ao
i r GIB = PHI - j . o
O" = G IP* A A
Iff Op. GT.D^T ) GO TO i3
VT 3 AO*nP*»C>.7
GO TO I ^
13 VT a 1 .31 0r + 03«eXP( 0,?P«ALOG(DFLDFN) -f O. 1 D
I 4. 0. 1 O«Al. OG( S 1 GM/\ ) - O.SS^ALOC^D'TNSC) )
I A VK = V1*CR
IF(Vk' «Lf . VK'F) GO TO «n
PHI - ((VK + VD - VC) - =,ORT((VK 4 VO - VC)**,"5 - A . P* VK*VO ) ) /? , O/
1 VK
DO 15 1=1.10
480
-------
FP = vc*PHt -4 VO»M.A - PHI) - VK*(l.T - PHI>**?*PHI
OFOP = VC - VO - VK*» 3.0*PHI*PHI - 4.O»PH| + l.O)
HELP = FP/DFPP
IF(AQS CALCULATE FLOODING PARAMETERS
?0 AR = (1.0 - 2.0*PHIFi *< | .0 - PHIF)*»2
FFS = VC/V*( HC/D I
V = NOWTUR9< I > = O.O
?1 t t I t = AA«KO« I )
A = 7?.0*"H!/pp
AA = A«COLHT/\/C/l?. 0
AB = A*COLHT/\/O/l 2 . 0
P - (C)LHT#PHI/Vr>/ 1? .0
C«»* )F MTTVPF .EO. 0, CALCULATE NOW ( I ) FOR ALL THtJF.F VOOCLS
C-««» 1^ MTTY^r. ,EQ. 1. CALcULATf NOW ( 1 ) FOR STAGNANT DROP.S ANO CALL TFTA
IF(MrTYPr .CO. ?. .OP. viTTvpr .^n. 3) GO TO 25
C*»» CALCULATIONS FCR S>T AGNiANT
OO 32 1-1 .NCCMPS
V.*AA
1) = 1 ,0/( 1 ,0/NVi>c,TAO( 1 ) + 1 ,0/f( I )/Nc,f,TAG( I ))
IF(MTTYPjr .J_-Q. i, GO TO 35
Cv4« if- MTiYPL .to. 2. tALc.ULATf NOV.' ( 1 ) FOR CIRC PROPS ANO CALL FFTA
?5 ir{VTTYre .CQ. 3) GO TO :in
C»"» CALCULATIONS FOR CIRCULATING r>-"?OPS
no r? i - i .NCOVP^
N'Vf I p c: { I ) = 1 ,?^6F-Op»SORT (OlFFC ( I >*V^/Or>) *AA
N?ctuc c I > - <7.i06c-o<:'*r>irrD{ 1 1 /DP + n.7r>r;-c;.T»vs/
1 (1.° + VI<^fC>/VI SCC) >« AH
?7 NO'.cC|f}C < I ) = 1 ,0/( 1 .AXNWC I PCI 1 ) + 1 .0/f( M /NSC I PC ( I) >
^(MTTYPF ,fO. ?) GO TO 35
C*»« CALCULATIONS FOP TU^n, iLFNT HRO^S
3O DO 3£ I = 1.MCOMPC,
- ?.330e-0/)«POM«*0..">1?»D!FFC<
cc) «*<: . 4)667* AA/ANW
NSTUI71K1) - 7,e9M-0A»OIFF3( I )*AU/DP/ANS
481
-------
12 NOwTiiR'3 ( I ) = 1 . 0/ f 1 . n/NWTUP!1 ( I ) + 1 . 0/r ( I) /NSTUWt1 ( I ) )
IFCMTTYPE .CO. O) RETUPN
35 DO 3(, ! =1 .NCOMP'--,
NOX - N!OWST/ + tNloWCIPC(I) -t-
FA = E ( n
CALL FETA
-------
ANALYSIS OF f-DC EXTRACTION BUN K513 PAGE 1
ii [ AL COMMIT I ONi
Vtt.UCiTY DISPERSED PHASE = 1.610 f-'I/HR
vtuocirv cfj^T luoous PHASL = is.c^s f
-'f. = 21,0 l)f.0»tt.b C
= .i>12352 (CALCULATED)
.2000 IFSTIMATEO)
'lf^ DIMENSIONS
t)Ji>C ulAMKTt.k = l.'aO INCH
J.OO
= 1.00
COLUMN Ht. IIJHT = IflM'_ rt.'iv = ,O^I.^^4
TN^?O J T I0'l l.ihOp Li-lA = .Ii3t)b3 I'-Ch
AXUL
483
-------
AI\ OK I»OC LXrw/kCflON RUN H3i3 PAGE 2
•><••>• ..Ol-Ure JlrhoSl/ITICSi JlSruI'tUf ION TOtKF ICjENTSi AND t(I>
v.>Lt.iri_ UCM 11 ocd) Kij(i) Ed)
t ' •••'.!. b.9/ j.^0 11 . «•<) l,2b
0-'..r". VJL 'j.b^ 3.0? 200.00 22.95
A = Al . / 1 SiJKT/CoH
SouUlt iJrt NS NOW
M r ^,;. L i. i '41 i <•. f> H (-,
i-. T /iCi. 1.U1 I3.b^n 1.0^,9
1 -' ^ A^c 1 . 1« ) 1^. 7«io 1 .1 TJ
1.1*1 U.f.a? 1.13'
Sul '.i ,'L N'v NS KOvi
HC ACE- I'.i.Hnf, 3-V.^O? 7.3P7
FT ACe I't./Cr' J3.109 10.8*3
0-Ci.'t.n.;L \J.l.^(t J K;. I't.r-; iMLiv 1 i>i Kt.i-'fJV*LS
bt'l 0 fc
!•: !. .'. C ^.
M " ( K
I-i'" (r..L
0 - C i ' :. .? v.. L
f-ttr ui'f
i-^ /..-..
r 7 (i . :••
•S / ;. . r
•3'^c'r'., <•
P-VOO ('I'M
1 P 1 . P
H 1 .'•
o .., t 3
f'bS.f
Eti
. 7 •) 73 6
• 3 0 7 ? «'
. 1 if- 1 &
. n <. b 3 o
484
-------
ANALYSIS 0> U.VTA FOR EXTRACTION *'-JN NUKdtK «S13
»'tX = tf.lti/1 ^F./ = 29.03:>7 A COMPONENTS
F..IK ;•!: AtK t = .-+100 AND ZETA = .79736 GIVING NOX = .337
ii.h /. l(i.u) 2(11. j) 2(u.,>> 2(0.3) 2(0.5) 2(0.7) ZI0.8) 2(0,9) 2(1.0)
.9?S26 .3es?4 .BOol .B22:>9 .80544 ,7<*7."?o
.62202 .797J6
»st.i-.s! T I vl i Y -- DX = CHAMJL IN 'jox cauSflo BY A l ^tfcCLNT ULCRLASE IN x
- 1
»-; r ;,cE t = l.^boo AND ZETA s .3022^ olvi^C NOx = 2.S20
/(n.l1) XlO.i) /(u.3) i(0.3) /(0.5) i(0.7) ^(0.8) /(0,<5) 2(1,0)
.'J4711 .4240* .Jb90l' •32374 ,30l«ilO .48451 .42143 .36069 .30222
1 I V 1 Ty -- L>X = CIAK'JL ifJ U0< CAUStO ^Y A 1 ^EHCRM Ct-CKtASE IN X
^ > Uc. = .<» 2(0.3) 2(0.3) ^(0.7) 2(0.8) i(0.9) 2(1.0)
,t'i;^'' .b'/'t-.id .t):>ni9 .4^068 ,3inbO .20570 ,16715 ,13068 .\2b\f
.'56235 .3/797 .249^6 .200@1 .16013 ,1261d
r I vi I'Y •- i'X = LMftNijt. If: fjox CauS^lO HY A l HtHCRNT UtCPf.ASE IN X
- e.21c-02 • Ut. = ^,4:SE-03 « ^r'EX = 6.94E-03 • O^EY = 8.3QE-04
«,»« ft)(v u-O'KSOL t = iii'.'VriQO AND 2tTA = .04336 GIVING NOX a 4,339
PHDMU. .tio.'i) /(U.J) /. (').?] 2(0.3) 2(0. S) 2(0.7) 2(0. P) 2(0,9) 2(1.0)
I': A! i) ./T/\-n ,l>4jo«J .-. j.J^J .?49TS .164U .089<5b .06706 .05169 .
.0622H ,
NSlTtvj i Y •••• i'X = Lri».r.'JL Ii| NOX CAUScL) Hy A 1 ^t«CE^T UcC«EASE IN X
-.-- i.H'.f.-lK' , In - 2.2nC-o3 ., D"IX = l.Ob'C-02 i i^EY = 3.12E-04
485
-------
List of Experimental Data.
In the Table H to follow, the experimental data
for runs conducted in the RDC extractor are listed.
The table entries include the measured phase velocities,
the column temperature, the dimensions of the column
and its internals, the disc rotational speed, and
the feed and product aqueous-phase concentrations.
FS/FW is the ratio of solvent mass flow rate to
water mass flow rate; when this quantity is multi-
plied by Kd, the extraction factor, E, for each
solute results. Also included in the table are
the estimates of physical properties and the following
quantities calculated in the computer program:
G,g = value assumed to give correct drop size
^ai/. = calculated solvent hold-up
ca i c
^mn^c = measured solvent hold-up
me a s
d = average drop diameter
a = interfacial area per column volume
Pec = water-phase Peclet number
Ped = solvent-phase Peclet number
FF = percentage of flooding according to
Logsdail, et al. (1957)
POM = power per unit mass
486
-------
Table H. RDC Extractor Data
RDC extractor run # RS1A
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
Temperature = 22.0°C
Vd
d.
= 3.704 ft/hr,
= 1.50 inch,
= 0.5917 gm/cc,
= 0.9978 gm/cc,
= 0.337,
= 0.0417 inch,
Fs/Fw = 0.3538
= 6.210 ft/hr,
H = 23.00 inch
= 0.181 cp, N = 1430.0 RPM
d = 2.25 inch,
s
y = 0.955 cp,
= °-0078' W
a = 13.43 ft /ft
a = 41.6 dyne/cm
= 30.52
FF
= 8.3%,
Solute
1.
2.
3.
MEK
DEK
n-BuAc
n
(10
24
22
d
5
•
•
20.
POM
1
5
6
D
ft2
3
3
= 586.6 ft-lbf/hr-lbm, Pec =
c
/hr)
.65
.22
2.76
K^
d
2.49
13.4
168.
Feed
Water
(ppm)
2109
4297
4393
Product
Water
(ppm)
639
183
77.3
= 1.91
i
D6
1
Removal
69
95
98
.7
.7
.2
RDC extractor run # RS1B
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
Temperature = 22.0°C
Vd
di
V
18
FF
= 1.226 ft/hr,
= 1.50 inch,
= 0.5917 gm/cc,
= 0.9978 gm/cc,
= 0.337, calc = 0.0026,
F/F = 0.1171
s "*
= 6.210 ft/hr,
H = 23.00 inch
= 0.181 cp, N = 1430.0 RPM
a = 41.6 dyne/cm
d = 2.25 inch,
s
u.
u = 0.955 cp,
G
meas
= 0.0417 inch,
= 4.8%, POM
a = 4.42 ftVft ,
586.6 ft-lbf/hr-lbr
Pe, = 30.58
Pec • 1'897
Dd
Solute (105
1.
2.
3.
MEK
DEK
n-BuAc
24.
22.
20.
D
c
ft2/hr)
1
5
6
3.
3.
2.
65
22
76
Kd
2.
13.
168.
Feed Product
Water Water
(ppm) (ppm)
49
4
2109
4297
4393
1340
888
288
Removal
36.
79.
93.
5
3
4
487
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS2 Temperature = 22.0°C
Dispersed phase = n-Butyl Acetate
Continuous phase = Prepared Waste Water
V, = 1.909 ft/hr, V,, = 14.054 f t/hr , F/F = 0.1196
Q C o W
d. = 1.50 inch, d = 2.25 inch, H = 23.00 inch
.L 3
p. = 0.8788 gm/cc, Pd = 0.704 cp, N = 617.0 RPM
pc
= 0.
G18 = °«
d
P
FF
= 0.
= 15
Solute
1.
2.
3.
MEK
DEK
9978
200,
0269
.6%,
D
(10
6.
5.
Phenol 6 .
gm/cc
' G
$ -, = 0
ycalc
inch,
POM
D
d
5 ft2
21 3
79 3
04 3
a =
= 92.
c
/hr)
.65
.22
.47
= 0.955 cp,
.0120,
32
3
K
4
16
57
.17
*meas
ft2/ft3
ft-lbf/hr-lb
*
•
•
56
2
0
Feed
Water
(ppm)
2213
4314
6143
G = 13.9 dyne/cm
= — —
' PeA =
d
m' ec ~
Product
Water
(ppm)
1413
1485
1073
27.
79
8.530
%
Removal
36
65
82
.2
.6
.5
Comment: Product water concentrations calculated by
material balance from loaded solvent analysis.
488
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS3 Temperature = 22.0°C
Dispersed phase = n-Butyl Acetate
Continuous phase = Prepared Waste Water
Vd
d
V = 15.388 ft/hr,
c
d =2.25 inch,
s
18
d
P
FF
= 1.705 ft/hr,
= 1.50 inch,
= 0.8788 gm/cc,
= 0.9978 gm/cc,
= 0.200,
= 0.0213 inch,
= 17.3%, POM = 165.7 ft-lbf/hr-lb
F /F = 0.0974
s w
H = 29.75 inch
yd = 0.704 cp, N = 805.0 RPM
y = 0.955 cp, a = 13.9 dyne/cm
= 0.0130,
= 0.0630
Imeas
a = 43.97 ftVft , Ped = 29.24
m'
Pe = 9.554
c
Dd
Solute (10
1.
2.
3.
4.
5.
Acetone
MEK
Benzene
Phenol
o-Cresol
6
6
6
6
r-
Dc
Kd
b ftVhr)
.77
.21
.21
.04
f i
4.
3.
3.
3.
•-%
30
65
65
47
f\ ir
1.
4.
61.
57.
*\ /\ f
05
56
5
0
Feed
Water
(ppm)
38.0
217
169
13300
n i m
Product
Water
(ppm)
34.3
126
30.5
308
TC A
%
Removal
9
41
82
97
OR
.7
.9
.0
.7
R
489
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS4 Temperature = 26.8°C
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
Vd = 2.152 ft/hr, Vc = 12.622 ft/hr, FS/FW = 0.0996
d. =1.75 inch, d =2.00 inch, H = 28.31 inch
•J- 5
Pd = 0.5856 gm/cc, yd = 0.175 cp, N = 1430.0 RPM
Pc = 0.9966 gm/cc, y = 0.855 cp, a = 40.6 dyne/cm
3,p = 0.337, , = 0.00658, <(> = 0.00658
lo Tcalc ~ meas
d = 0.0324 inch, a = 14.64 ft /ft , Pe, = 31.99
P d
FF =13.8%, POM = 1070.1 ft-lb^/hr-lb , Pe = 4.922
r m c
Dd
c;
Solute (10° ft
1.
2.
3.
4,
5.
6.
Acetone
MEK
Benzene
n-BuAc
Phenol
o-Cresol
27.7
25.4
25.4
21.6
24.7
22.9
Dc
•)
Kd
/hr)
4
4
4
3
3
3
.88
.14
.15
.13
.94
.46
0.
2.
407.
168.
0.
4.
63
49
70
80
Feed
Water
(ppm)
29.
124
68.
5457
605
72.
Product
Water
9
5
8
(ppm)
28
83
60
10
522
17
.2
.3
.4
.8
.0
%
Removal
5.7
22.8
11.8
99.8
13.7
76.6
Comment: Feed water is the product water from run RS3.
490
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS6A Temperature = 24.2°C
Dispersed phase = n-Butyl Acetate
Continuous phase = Lube Oil Refining Waste Water
Vd
d.
18
d
P
FF
V =8.743 ft/hr,
=2.25 inch,
= 1.005 ft/hr,
= 1.50 inch, d
o
= 0.8765 gm/cc, yd = 0.687 cp,
= 0.9972 gm/cc, yc = 0.907 cp,
= 0.200, *calc = 0.0088,
= 0.0161 inch, a = 39.14 ft'/ff,
= 28.0%, POM = 316.2 ft-lbf/hr-lbm
F /P = 0.1010
H = 29.00 inch
N = 1100.0 RPM
a = 13.5 dyne/cm
= 22.13
= 4.256
D
Solute (10
1.
2.
3.
MEK
Phenol
6.
6.
d
Dc
K, Feed Product %
Water Water Removal
5 ftVhr)
41
23
3.
3.
87
69
(ppm) (ppm)
4.56
57.0
12216
8751
5883
104
51
98
.8
.8
0-Cresol
5.
79
3.
24
206.
892
6.5
99
.3
RDC extractor run # RS6B Temperature = 23.4°C
Dispersed phase = n-Butyl Acetate
Continuous phase = Lube Oil Refining Waste Water
Vd
d.
P'
"18
%
FF
= 3.023 ft/hr,
= 1.50 inch,
= 0.8774 gm/cc,
= 0.9974 gm/cc,
= 0.200,
= 0.0162 inch,
= 45.6%,
D
Vc = 8.743 ft/hr,
FS/FW = 0.3042
Solute
(10-
1. MEK 6.36
2. Phenol 6.19
3. o-Cresol
5.75
i
/cc
/cc
*ca
ch,
POM
D
9
ft~_
3
3
d = 2.25 inch, H
s
, yd = 0.690 cp,
, y = 0.932 cp,
C
, = 0.0272, cj>
a = 121.1 ft2/ft"
= 29.19 inch
N = 1100.0 RPM
cr = 13.7 dyne/cm
= 0.00214
, Pe, = 21.88
a
= 323.0 ft-lb./hr-lb , Pe =4.359
t m c
K, Feed
c a Water
/hr) (ppm)
.76 4.56 12216
.58 57.0 8751
Product %
Water Removal
(ppm)
2452 82.3
77.0 99.1
3.14 206.
892
4.3
99.5
491
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS7A Temperature = 23.3°C
Dispersed phase = Isobutylene
Continuous phase = Treated Lube Oil Refining Waste Water
Vd
d.
pc
G18
d
P
FF
= 1.506 ft/hr,
= 1.75 inch,
= 0.5900 gm/cc,
= 0.9975 gm/cc,
= 0.337,
= 0.0319 inch,
= 8.743 ft/hr,
H = 29.31 inch
F /F = 0.1019
s w
d =2.00 inch,
s
Ud = 0.179 cp,
y = 0.926 cp,
c
mea<
a = 10.51 ftVft
N = 1450.0 RPM
a = 41.3 dyne/cm
= 0.0362
Ped = 32.76
= 10.6%,
Solute
POM = 1136.1 ft-lbf/hr-lbm/ Pec = 3.573
1. MEK
2. n-Bi
3. Pher
4. o-Cresol
22.2
Dd
do5
24.
20.
23.
Dc
Kd
ffVhr)
5
9
8
3.
2.
3.
78
86
60
2
168
0
.49
,
.70
Feed
Water
(ppm)
5573
7133
306
Product
Water
(ppm)
3597
11.0
227
Remo va 1
35
99
25
.5
.8
.8
3.16
4.80
24.2
2.3
90.5
Comment: Feed water is the product water from run RS6A.
492
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS7B Temperature = 23.3°C
Dispersed phase = Isobutylene
Continuous phase = Treated Lube Oil Refining Waste Water
Vd = 1.506 ft/hr, Vc = 8.743 ft/hr, FS/FW = 0.1019
d. = 1.75 inch, d = 2.00 inch, H = 28.88 inch
J» b
Pd = 0.5900 gm/cc, yd = 0.179 cp, N = 1450.0 RPM
p = 0.9975 gm/cc, y = 0.926 cp, a = 41.3 dyne/cm
Glg = 0.337, Acalc = 0.0047, A = 0.0346
d = 0.0319 inch, a = 10.51 ft /ft , Ped =
FF =10.6%, POM = 1136.1 f t-lb^/hr-lb. Pe^
I lu c
D, D K, Feed Product
^ c Water Water
Solute (10 ft /
1.
2.
3.
4.
MEK
n-BuAc
Phenol
24
20
23
.5
.9
.8
3.
2.
3.
hr)
78
86
60
32.28
= 3.521
Removal
(ppm) (ppm)
2
168
0
.49
m
.70
2801
6791
229
1891
15.2
190
32
99
17
.5
.8
.0
o-Cresol
22
.2
3.
16
4
.80
18.0
2.8
84
.4
Comment: Feed water is the product water from run RS6B.
493
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS8 Temperature = 21.2°C
Dispersed phase = 48.7% n-Butyl Acetate in Isobutylene
Continuous phase = Lube Oil Refining Waste Water
vd
di
= 3.072 ft/hr,
= 1.50 inch,
= 0.7325 gm/cc,
= 0.9979 gm/cc,
= 0.171,
V = 11.003 ft/hr,
C
F /F = 0.2049
s w
d = 2.00 inch,
S
yd = 0.279 cp,
y = 0.973 cp,
= 0.0157,
H = 29.00 inch
N = 1090.0 RPM
a = 33.1 dyne/cm
meai
d
P
FF
= 0.0233
= 14.1%,
Dd
Solute (103
1.
2.
3.
4.
5.
Acetone 17
MEK 15
Phenol 15
o-Cresol
14
Benzene 15
inch,
POM
D
A
a = 48.47 ft: /ft , Ped =
= 327.8 ft-lbf/hr-lbm, Pec
K, Feed Product
c Water Water
ftVhr)
.0
.6
.2
.1
.6
4
3
3
2
3
.20
.57
.40
.99
.57
0.
3.
28.
100.
239.
83
50
1
(ppm)
24.6
108
17170
2660
36.9
33.77
= 6.590
Remo va 1
(ppm)
12
54
1902
124
9
.1
.9
.2
50
49
88
95
75
.8
.2
.9
.3
.1
RDC extractor run # RS9 Temperature = 22.3°C
Dispersed phase = 53.1% n-Butyl Acetate in Isobutylene
Continuous phase = Prepared Waste Water
V
= 3.715 ft/hr,
di -
pd =
pc -
G18 -
d~ =
P
FF =
Solute
1.75 inch,
0.7438 gm/cc,
0.9977 gm/cc,
0.171,
CdJ.
0.0186 inch,
22.5%, POM
Dd Dc
5 2
(10s ft"1/
S
7c = 12.692 ft/hr,
=2.00 inch, H = 28.88 inch
F /F = 0.2182
O Wr
yd = 0.292 cp,
y = 0.948 cp,
C
= 0.0238,
a = 92.31 ftVff,
POM = 542.2 ft-lbf/hr-lbm,
N = 1090.0 RPM
o = 31.7 dyne/cm
= 0.0238
Pe, = 25.90
a
Pec = 6.544
1. MEK
2. Phenol 14.6
'd
5
0
Kd
r ftVhr)
i.O
1.6
3.68
3.50
3.59
30.6
Feed
Water
(ppm)
424
17320
Product
Water
(ppm)
91.3
93.4
Removal
75.5
99.5
494
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS10 Temperature = 20.8°C
Dispersed phase = Isobutylene
Continuous phase = Ethylene Quench Waste Water
Vd = 2.141 ft/hr, Vc = 12.583 ft/hr, FS/FW = °-10l°
d. = 1.75 inch, d = 2.00 inch, H = 28.81 inch
J- 5
Pd = 0.5923 gm/cc, yd = 0.183 cp, N = 1450.0 RPM
Pc = 0.9980 gm/cc, u = 0.983 cp, a = 41.8 dyne/cm
G18 = °'220' * " °'0093' * = 0'0217
d = 0.0208 inch, a = 32.10 ft /ft , Ped = 27.33
FF = 16.0%, POM = 1154.9 ft-lb-/hr-lb , Pe = 4.944
r in c
Dd Dc Kd
Solute (105 ft2/hr)
1.
2.
3.
4.
Benzene 23.8 3.54 407.
Toluene 22.2 3.12 1690.
Xylenes 21.1 2.83
Phenol 23.1 3.36 0.70
Feed
Water
(ppm)
71.1
40.5
40.3
66.9
Product
Water
(ppm)
2.9
2.3
63.1
Removal
95.9
94.3
>97
5.7
Comment: Feed COD = 1880 ppm; Product COD = 1209 ppm.
495
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS11 Temperature = 23.4°C
Dispersed phase = Isobutane
Continuous phase = Ethylene Quench Waste Water
V, = 2.198 ft/hr, V = 12.528 ft/hr, F/FTf = 0.0973
d c s w
d.^ =1.75 inch, d =2.00 inch, H = 29.50 inch
Pd = 0.5530 gm/cc, ud = 0.201 cp, N = 1450.0 RPM
p = 0.9974 gm/cc, y = 0.924 cp, a = 47.7 dyne/cm
G18 - 0'200' *calc = °-0090' *meas = °-0096
d = 0.0207 inch, a = 31.41 ftVft , Pe- = 28.83
P d
FF =12.3%, POM = 1135.4 ft-lbf/hr-lb , Pe = 5.041
Dd Dc
Solute (105 ft2/hr)
1.
2.
3.
4.
Benzene
Toluene
Xylenes
Phenol
21.9
20.4
19.3
21.2
Kd
Feed
Water
(ppm)
3.79 338.
3.35 1460.
3.03
3.58
—
0.
20
81.
43.
33.
68.
2
8
6
2
Product
Water
(ppm)
2.
1.
<1
66.
4
6
0
Removal
97
96
>97
3
.0
.3
.2
Comment: Feed COD = 1880 ppm; Product COD = 699 ppm.
Turbidity reduced by 72%.
496
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS12
Temperature = 22.2°C
Dispersed phase = 2-Ethyl Hexanol
Continuous phase = Neutralized Oxychlorination Waste
Water
vd
di
pd
pc
G18
d
P
FF
= 1.501 ft/hr,
=1.50 inch, d(
= 0.8311 gm/cc,
= 0.9977 gm/cc,
= 0.200,
= 0.0159 inch,
Vc = 11.766 ft/hr,
Fs/Fw = 0.1063
=2.25 inch, H = 30.00 inch
y
= 8.038 cp,
= 0.964 cp,
= 0.0112, «
a = 50.94 ft'
;meai
N = 800.0 RPM
a = 8.5 dyne/cm
= 0.0156
Ped = 29.85
= 26.1%,
Solute
POM = 164.7 ft-lb./hr-lb , Pe =7.686
r m c
2. EDC
3. ECH
4. Chloral Hyd.
0.49
Dd Dc
(105 ft2/hr)
1 0.64 4.99
0.55 3.73
0.58 4.11
Kd
1.0
200.
5.0
Feed
Water
(ppm)
286
1505
1636
Product
Water
(ppm)
265
20
1292
%
Removal
7.3
98.7
21.0
3.08 140.
15220 7726
49.2
Comment: Water phase contained 9.26 wt. % NaCl.
497
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS13 Temperature = 21.6°C
Dispersed phase = n-Butyl Acetate
Continuous phase = Prepared Waste Water
Vd = 1.650 ft/hr, Vc = 13.049 ft/hr, FS/FW = 0.1114
d. = 1.50 inch, d = 2.25 inch, H = 29.19 inch
J. S
p, = 0.8792 gm/cc, y, = 0.708 cp, N = 800.0 RPM
Pc = 0.9979 gm/cc, y = 0.964 cp, o = 13.9 dyne/cm
G18 = 0.200, «calc = S.0124, 6 = 0.0428
d = 0.0213 inch, a = 41.71 ft2/ft, Ped = 29.04
FF =24.0%, POM = 164.7 ft-lb./hr-lb , Pe = 8.187
i ITT C
Solute
1.
2.
3.
MeAc
EtAc
i-PrAc
D
11°
6.
5.
5.
d
5
45
97
60
ft
3
3
3
Dc
2/hr)
.93
.40
.04
3
11
34
Kd
.64
.2
.1
Feed
Water
(ppm)
227
270
676
Product
Water
(ppm)
181
81.6
85.3
%
Removal
20
69
87
.3
.8
.4
4. o-Cresol
5.57 3.02 206. 5622 255. 95.5
498
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS14
Dispersed phase = Isobutylene
Continuous phase = Prepared Waste Water
Temperature = 25.3°C
Vd
di
= 9.986 ft/hr,
V = 8.607 ft/hr,
\*r
FS/FW = 0.6867
= 1.75 inch, d = 2.00 inch, H = 29.00 inch
PC
G
18
FF =
0.5869 gm/cc,
0.9970 gm/cc,
0.337,
Ud = 0.177 cp,
y = 0.883 cp,
= 0.0270
N = 1200 RPM
a = 40.9 dyne/cm
vcalc
0.0392 inch, a = 49.62 ft*/ft'
19.4%,
Pe, = 36.90
a
POM = 670.08 ft-lbf/hr-lb ,
Pe = 4.246
c
Solute
D
(10
K
Crotonaldehyde
25.35
ft/hr)
3.34
Feed
Water
(ppm)
2.48 5572
Product
Water
(ppm)
936
Removal
83.2
RDC extractor run # RS15 Temperature = 22.7 °C
Dispersed phase = 48.2% n-Butyl Acetate in Isobutylene
Continuous phase = Phenolic Resin Plant Water
Vd
di
PC
G18
d
P
FF
1.844 ft/hr,
1.75 inch, d
0.7294 gm/cc,
0.9976 gm/cc,
0.0171 , c1
0.0171 inch,
V = 6.583 ft/hr,
C
FS/FW = 0.2059
=2.00 inch,
yd = 0.285 cp,
yc = 0.940 cp,
= 0.0115,
a = 48.56 ft2/ft3,
H = 28.18 inch
N = 1200 RPM
a = 32.1 dyne/cm
= 12.9 %,
POM = 681.8 ft-lbf/hr-lb ,
Ped = 24.30
Pe = 3.163
c
D
Solute
(10
ftyhr)
Methanol 14.18 4.68
Formaldehyde
17.12
Phenol
6.12
7.36
3.53
Kd
0.10
0.15
6.15
Feed
Water
(ppm)
12000
17370
48270
Product
Water
(ppm)
11510
16450
483
Removal
4.1
5.3
99.0
499
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run # RS16 Temperature = 23.4 °C
Dispersed phase = n-Butyl Acetate
Continuous phase = Phenolic Resin Plant Water
Vd
d.
FF
= 0.804 ft/hr,
= 1.50 inch, d
0.8744 gm/cc,
= 0.9974 gm/cc,
= 0.200 ,
0.0144 inch,
= 26.5 %,
Vc = .936 ft/hr,
F /F = 0.1188
s/ w
= 2.25 inch, H = 30.50 inch
calc
= 0.0075
= 0.690 cp,
= 0.932 cp,
N = 1250 RPM
0 = 13.7 dyne/cm
a = 37.34 ft2/ft3
Pe, = 20.58
a
POM = 428.1 ft-lbf/hr-lbm,
Pe = 2.753
c
Solute
D
(10
K
ftVhr)
Methanol 14.31 4.69
Formaldehyde
Feed
Water
(ppm)
0.10 12000
Product
Water
(ppm)
7608
Removal
36.6
Phenol
17.
6.
30
19
7.
3.
44
58
0
12
.15
.0
17370
48270
10370
6082
40.
87.
3
4
RDC extractor run # RSI7
Temperature = 24.5 °C
F /F = 0.1249
s w
di
Dispersed phase = Methyl Isobutyl Ketone
Continuous phase = Hydrofiner Waste Water
Vd = 1.682 ft/hrf VQ = 10.717 ft/hr,
= 1.50 inch, d = 2.25 inch, H = 30.25 inch
S
= 0.7914 gm/cc, yd = 0.542 cp, N = 1250 RPM
= 0.9972 gm/cc, yc = 0.89R cp, a = 9.90 dyne/cm
= 0.200 , 99.8
Feed COD = 17530 pprft
MIBK Concentration in Product Water = 15680 ppm
(COD 42500 ppm)
500
-------
Table H. RDC Extractor Data (Continued)
RDC extractor run *RS18 Temperature = 24.2 °C
Dispersed phase = 49.5% Methyl Isobutyl Ketone in Isobutylene
Continuous phase = Hydrofiner Waste Water
Vd = 2.052 ft/hr, VQ = 6.808 ft/hr, FS/FW = °-2091
di = 1.75 inch, d = 2.00 inch, H = 30.38 inch
pd = 0.7916 gm/cc, yd = 0.265 cp, N = 1150 RPM
p_ = 0.9972 gm/cc, y = 0.907 cp, 0 = 29.7 dyne/cm
C C
G, B = 0.171 , , = 0.0150
J_ o CciXC 99
Comment: Feed COD = 17530 ppm; Product COD = 18580 ppm
MIBK Concentration in Product Water = 3150 ppm
(COD 9000 ppm)
RDC extractor run #RS19 Temperature = 21.8°C
Dispersed phase = Isobutylene
Continuous phase = Styrene Waste Water
Vd = 1.480 ft/hr, Vc = 8.170 ft/hr, FS/FW = 0.1072
d. = 1.75 inch, d = 2.00 inch, H = 30.00 inch
••" S
pd = °'5926 gm/cc, u^ = 0.181 cp, N = 1250 RPM
pc = 0.9978 gm/cc, yc = 0.959 cp, a = 41.7 dyne/cm
Gig = 0.337, <(>„.,.,_ = 0.0043
dp = 0.0376 inch, a = 8.27 ft2/ft3, Ped = 36.06
FF = 7.93%, POM = 766.11 ft-lb,/hr-lb , Pe =3.811
r me
D<3 Dc Kd Feed Product %
5 ? Water Water Removal
Solute (10J ftVhr) (ppm) (ppm)
1. Benzene 24.00 3.62 407. 290 10 96.6
2. Ethyl-
benzene 25.17 2.90 120 4 96.7
3. Styrene 26.19 2.99 15 <1 >93
501
-------
APPENDIX I
LABORATORY EXTRACTIONS OF OXYCHLORINATION HASTE WATER
One of the samples of industrial waste water
which was studied in this work cane from an oxychlo-
rination plant. The most serious pollutant present in
this waste water was found to be chloral hydrate,
which was present at a concentration of about 1.5
weight %. This waste water also contained between
1.5 and 5.8 weight % HC1, between 1500 and 3460
ppm ethylene dichloride, and less than 500 ppm of
several other organic compounds.
The measurement of the equilibrium distri-
bution coefficient for chloral hydrate distributing
between salt-free water and 2-ethyl hexanol
indicated that this was an excellent solvent. K.
varied from 50 with 500 ppm in the aqueous phase to 17
with 15,000 ppm in the aqueous phase. However, when
a sample of neutralized oxychlorination waste water
was contacted at a low flow ratio with 2-ethyl hexanol
in the RDC extractor (Run RS12), the removal efficiency
was found to be only 49% rather than about 98% expected
for a solute with such a large value of Kd. In this
appendix several experiments are discussed which show
that this low removal is probably due to a slow chemical
reaction which accompanies the transfer of chloral from
the water phase to the solvent phase.
502
-------
Initial Experiment.
An aqueous solution containing ethyl acetate,
n-butyl acetate, and chloral hydrate was prepared.
An accurately measured quantity (359 gm) of this
solution was added to a 500-ml separatory funnel. A
50-gm sample of 2-ethyl hexanol was added, and shaking
of the mixture was begun immediately. After a
precisely measured time interval, about 30 ml of
the dispersion was removed into a 60 ml separatory
funnel, and shaking of the 500-ml separatory was
continued. After 40 seconds to allow the phases to
separate in the 60-ml separatory funnel, about 10 ml
of aqueous phase was drained into a sample bottle.
Three samples were collected from three 60-ml
separatory funnels before the shaking of the 500-ml
separatory funnel was terminated. These samples
were analyzed for the three solutes using the gas
chromatograph, and then the contents of each 60-ml
separatory funnel and of each sample bottle (only
a few microliters were used for the GC analysis)
were added back into the 500-ml separatory funnel.
After 24 hours with occasional shaking, the aqueous
phase from the equilibrated mixture was analyzed using
the GC.
Results and Discussion on Initial Experiment.
In Table II the results are shown from the
analyses of the initial aqueous phase, of the un-
equilibrated samples taken at 1, 4, and 10 minutes
after shaking began, and of the sample taken after
24 hours which is assumed to be at equilibrium.
503
-------
Table II. Results from Extraction of a Prepared
Water Solution using 2-Ethyl Hexanol
Initially pure solvent added to 500-ml separatory
funnel = 50.0 grams
Prepared water solution added to 500-ml separatory
funnel = 359. grams
Water phase analyses (concentrations in ppm by weight)
Time after beginning Ethyl n-Butyl Chloral
to shake funnel (min) Acetate Acetate Hydrate
Initial 1,802 4,410 15,000
1 896 264 10,350
4 883 262 8,750
10 809 235 5,220
860 244 2,360
Apparent values of K,;
a
1
4
10
7.26
7.47
8.81
7.85
113
114
128
123
3.2
5.1
13.5
38.5
504
-------
It appears that both ethyl acetate and n-butyl
acetate had essentially equilibrated within the
first minute of shaking. However, even after 10
minutes of shaking, the concentration of chloral
hydrate is still changing with time.
Another way of presenting the results is to
calculate an apparent value of K. based on a material
balance. From the known quantities of the two
phases, K. can be calculated from the initial and
final concentrations by assuming equilibrium is
achieved. These results shown at the bottom of
Table II again indicate a slow approach to equilib-
rium in the case of chloral hydrate. The final
value of K. for chloral hydrate after a long time
(24 hours) agreed well with the separately determined
equilibrium measurements of 38.
A possible explanation for the observed slow
rate of mass transfer for chloral can be postulated
based on the extensive work of Jensen and coworkers.
In a study of the kinetics of the reversible reaction
between chloral and various primary and secondary
alcohols, Jensen, et al. (1970) found that the
reaction in a nonpolar solvent was catalized by
acetic acid. The product of reaction is the hemi-
acetal , as summarized by the following:
/H /H
CC1--C = 0 + RON —»• CC1.-C — OH
6 J "OR
In the presence of water, chloral hydrate rather
than chloral would likely be the reactant, but the
505
-------
hemiacetal could still be the final product.
Whereas chloral hydrate would be expected to favor
the aqueous phase strongly, it is likely that the
hemiacetal of chloral and 2-ethyl hexanol would
favor the alcohol phase. The observed high value of
Kd could well be much more the result of the reaction
to form the hemiacetal than of the transfer of
chloral itself into the alcohol phase.
Additional Experiments.
Since the previous reports had shown carboxylic
acids to be good catalysts for increasing the rate
of reaction of chloral with various alcohols, we
made several extractions of chloral using n-octanol
to which several percent of various organic acids
had been added. The procedure was the same as for
the previous experiment, except the ethyl acetate and
n-butyl acetate were not present in the initial aqueous
phase and a sample taken after 60 minutes was the
final sample analyzed.
The carboxylic acids chosen for testing were
hexanoic, benzoic, and p-toluic. These acids are
expected to show high values of Kd between water and
octanol so that they should not be extracted back
into the waste water to a large extent.
Additional Results and Discussion.
The results of the three additional experiments
are reported in Table 12 in the form of apparent
values of K .. For comparison the previous results
from the extraction of chloral hydrate with pure
2-ethyl hexanol are included in Table 12. The
506
-------
carboxylic acids are seen to have a significant
catalytic effect, but not so great as to make this
a very promising process approach. In a recycle
process with solvent regeneration by distillation,
other problems such as esterification of the car-
boxylic acid would have to be considered.
507
-------
Table 12. Results from Extractions of Prepared
Water Solutions Which Contained Chloral Hydrate
Initial water solutions contained 15,000 ppm of
chloral hydrate.
Apparent values of K,:
Time after
beginning to
shake funnel
(min) Run A Run B Run C Run D
1 3.2
4 5.1
10 13.5
60
5.4
9.5
19.2
__
7.2
13.3
25.7
36.1
7.0
13.2
25.4
36.8
38.5
Run A used pure 2-ethyl hexanol as solvent.
Run B used 1 wt. % hexanoic acid in n-octanol.
Run C used' 2 wt. % benzoic acid in n-octanol.
Run D used 2 wt. % p-toluic acid in n-octanol.
503
-------
APPENDIX J. METRIC CONVERSION TABLE
9.480 xlO"4 BTU = 1 joule
1.8 °F or °R = 1 °C or 1 c
0.03281 ft = 1 cm
0.2642 gal = 1 liter
2.642 x 10"4 gal = 1 cm3
1.341 HP = 1 kilowatt
0.3937 in = 1 cm
0.002205 Ib = 1 gram
14.696 psi (lb/in2) = 1 a tin
14.223 psi (lb/in2) = 1 kg./cm2
14.504 psi (lb/in2 = 1 bar
509
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing}
1. REPORT NO.
EPA-600/2-76-220
4. TITLE AND SUBTITLE
Extraction of Chemical Pollutants from
Industrial Wastewaters with Volatile Solvents
3. RECIPIENT'S ACCESSION-NO.
5. REPORT DATE
December 1976 (Issuing Date)
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Jonathan P. Earhart, Kwang W. Won,
C. Judson King, and John M. Prausnitz
8. PERFORMING ORGANIZATION REPORT NO
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Department of Chemical Engineering
University of California
Berkeley, CA 94720
10. PROGRAM ELEMENT NO.
1BB610
11. CONTRACT/GRANT NO.
R801030
12. SPONSORING AGENCY NAME AND ADDRESS
Robert S. Kerr Environmental Research Lab.
Office of Research and Development
U.S. Environmental Protection Agency
<\da, Oklahoma 74820
- Ada, OK
13. TYPE OF REPORT AND PERIOD COVERED
Final 1971-1975
14. SPONSORING AGENCY CODE
EPA/600/15
15. SUPPLEMENTARY NOTES
16. ABSTRACT
Solvent extraction with volatile solvents was studied as a method for
treating wastewaters from petroleum refineries and petrochemical plants.
Extraction is most attractive when the loading of organics is high,
when substances are present which pose difficulties for biological
oxidation, and/or when the chemical value of recovered organics is high.
Volatile solvents (isobutylene and isobutane) were given particular
attention, since they are easily regenerated and since they have a low
solubility in the effluent water. Equilibrium distribution coefficients
were determined for numerous systems of water, various solvents and
various solutes, and correlations of these coefficients were developed.
A miniplant extraction facility was used to demonstrate the capabilitie
of extraction for treatment of seven different industrial wastewaters,
and for the purpose of analyzing the underlying mass transfer and axial
mixing behavior. The scale-up of the extraction system and the logic
of selecting extraction processes for wastewater treatment are explored
at length.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
*Water Pollution
*Solvent Extraction
Solvents
Extractors
Wastewater
b.lDENTIFIERS/OPEN ENDED TERMS
Selected wastewaters
from refining & petro-
chemical processes.
Volatile solvent extrfcc-
tion
Dual solvent extraction
Rotating disc contactor
considerations
COSATl Held/Group
07A
for scale-up
3. DISTRIBUTION STATEMENT
Release to Pub!i c
19. SECURITY CLASS (This Report)
unclas si fi ed
21. NO. OF PAGES
530
20. SECURITY CLASS.(Iluspyse)
unclassTM ecr
22. PRICE
EPA Form 2220-1 (9-73)
510
•K U.S. GOVERNMENT PRINTING OfFICE: 1977- 757-056/5545
-------
|