imam. .I
2 v- .I' ’3“)qu
5‘

.
J

9» in.

3:51.. V f
u.) inn.
..

.1 . .r
12...: .1“. 1
.9

. o
afifiafi
ghafilfl
2 .
hu‘vadfiu

sfifluafim
‘ n.

.l ’0‘
.il..‘i..aa.
I . .

I r g

:' n"
. .;§,J

31.5»: “,4 .. .
.3 .vrzla;
S v.

‘5 .
42L

‘ ‘1,
.2?!
-1? in 1A

l»
,.

‘
t'f‘n

gfi§§§§§ .1 m . , , H ‘ Wig;

y .7. hair?

 

 

me

TE unwensm LIBRARIES

iiiiiiiii\iiiiiiuiMilli

3 1293 01020 5056 r

 

 

 

This is to certify that the
thesis entitled
Carbon Dioxide Extraction of Aqueous
Benzaldehyde Solutions
presented by

Neil Douglas Barnes

has been accepted towards fulfillment
of the requirements for

 

 

M. S . degree in Chemical Engineering

MAC.

Major professor

Date Noveinher 4, 1994

O-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

LIBRARY
Mlch'gan State
Unlverslty

 

 

 

PLACE II RETURN BOXto romovo thin ohookout from your rooord.
TO AVOID FINES Mum on or bdoro duo duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Io An Afflnnoflvo AotioNEquol Opportunity lnotltwon
Wis-no.1

CARBON DIOXIDE EXTRACTION OF AQUEOUS
BENZALDEHYDE SOLUTIONS

By

Neil Douglas Barnes

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements

for the degree of

MASTER OF SCIENCE

Department of Chemical Engineering

1 994

ABSTRACT

CARBON DIOXIDE EXTRACTION OF AQUEOUS
BEN ZALDEHYDE SOLUTIONS

By

Neil Douglas Barnes

An apparatus was designed and constructed to measure the phase equilibria of the
benzaldehyde-water—carbon dioxide ternary system. Values for the benzaldehyde
distribution coefficient determined using this apparatus at a pressure of 1,500 psig and at
temperatures of 20 °C and 35 °C were 69 and 52 (mole units) respectively.

A pilot-scale, liquid-liquid extraction column 5 feet high x 1 inch I.D. capable of
operating at pressures up to 2,000 psig was designed and constructed. This apparatus was
used to extract benzaldehyde from a dilute aqueous solution. Inlet and outlet stream
benzaldehyde concentrations were measured using ultraviolet spectroscopy and gas phase
chromatography. The measured values were used to perform a benzaldehyde balance for
each experimental run of the column. Benzaldehyde recoveries ranged from 81.4% to
105.6%.

The measured inlet and outlet stream benzaldehyde concentrations were used to
calculate scale-up parameters for the column. The number of equilibrium stages in the 5
foot tall column was calculated as about 1.0, the number of transfer units by film-transfer
theory as 2.5, and the number of stages by the Kremser method as 1.0. The height
equivalent to a theoretical stage was determined to be 5 feet and the height of a transfer

unit was calculated to be about 2 feet.

To my parents for giving me every opportunity to make something of myself
and

to my sons, Matthew and Justin, who fill me with pride every time I see them.

iii

ACKNOWLEDGEMENTS

The author wishes to express his appreciation to the following people for their
contributions and support:

Bharath Rangarajan and Ramkumar Subramanian, Graduate Students, Chemical
Engineering, for their assistance, counsel, and perspective;

Carol Barnes for her patience and support;

Julie Caywood, Faith Peterson, and the rest of the Office Staff in the Department
of Chemical Engineering for their help with administrative details which was not taken for
granted;

and last, but most importantly, to Susan Jones for her help with a long list of things

that helped me see this project through to completion.

iv

TABLE OF CONTENTS

Page

LIST OF TABLES .................................................................................................. viii

LIST OF FIGURES ................................................................................................... x
CHAPTER

1. INTRODUCTION ....................................................................................... 1

2. BACKGROUND ......................................................................................... 6

Liquid-Liquid Extraction ..................................................................... 7

Literature Review ................................................................................ 7

Distribution Coefficient ....................................................................... 9

Supercritical Fluid Extraction ............................................................ 13

3. LIQUID-LIQUID EXTRACTION ............................................................. 21

Solvent Selection ............................................................................... 21

Dispersed Phase ................................................................................ 25

Column Capacity ............................................................................... 25

Interfacial Tension ............................................................................. 33

Packing ............................................................................................. 35

Extractor Internals ............................................................................. 37

Extraction Column Sizing ................................................................. 38

Determination of the Number of Transfer Stages ............................... 40

Determination of the Number of Film Transfer Units ........................ 44

Height of a Film Transfer Unit .......................................................... 50

Number of Overall Film Transfer Units ............................................. 50

Simplified Integration ....................................................................... 54

Extraction Factor ............................................................................... 55

4. BENZALDEHYDE DISTRIBUTION COEFFICIENT

MEASUREMENT .................................................................................... 57
Prediction of Distribution .................................................................. 58
Equipment Design ............................................................................. 61
Experimental Procedure .................................................................... 65
Results and Discussion ...................................................................... 67

5. BENZALDEHYDE EXTRACTION EQUIPMENT DESIGN .................... 73
Feed Section ...................................................................................... 73
Extraction Column ............................................................................ 78
Receiver Section ............................................................................... 81

6. BENZALDEHYDE EXTRACTION EXPERIMENTAL METHOD ........... 84
Preparation ........................................................................................ 88
Operating Procedure .......................................................................... 91
Chemical Analysis ............................................................................ 94
Material Balance Calculations ........................................................... 95

7. BENZALDEHYDE EXTRACTION RESULTS ....................................... 101
Number of Stages and the Height of a Transfer Unit ....................... 104

8. CONCLUSIONS AND RECOMMENDATIONS ..................................... 109
Conclusions ..................................................................................... 109
Recommendations ........................................................................... 1 10

APPENDICES

A. PHYSICAL PROPERTIES ...................................................................... 112
Carbon Dioxide ............................................................................... 112
Water ............................................................................................... 114
Benzaldehyde .................................................................................. 1 14
Isopentyl Alcohol ............................................................................ 114
n-Propanol ....................................................................................... 1 15

vi

B. ETHANOL PHASE GAS CHROMATOGRAGHY .................................. l 16
C. AQUEOUS PHASE GAS CHROMATOGRAPHY .................................. 124
D. ULTRAVIOLET SPECTROPHOTOMETRY ........................................... 135
E. SAMPLE CALCULATIONS ................................................................... 138
E. NOTATION ............................................................................................. 141
REFERENCES CITED .......................................................................................... 144

vii

Table
2-1
2-2

2—4
3-1
3-2

4-1
6-1
7-1
7-2
7-3
7-4

A-l
A-2
A-3

A-S
B-l
B-2

02
03
C-4

LIST OF TABLES

Page
Typical Liquid-Liquid Extractions ....................................................... 6
Distribution and Activity Coefficients for Alcohols and Esters ......... 14
Critical Conditions for Various Solvents ........................................... 17
Comparison of Properties for a Gas, SCF, and Liquid ....................... 18
Curve-Fit Parameters for the Correlation of Rao and Rao ................. 32
Interfacial Tension for Pure Water with Various Gases ...................... 35
Equilibrium Data ............................................................................... 43
Distribution Coefficient Measurement Data ...................................... 68
Flooding Correlation for Prestart-up Conditions ................................ 89
Extraction Column Material Balances ............................................. 102
Extraction Column Fluid Mechanics ............................................... 103
Equilibrium Stage Comparison ........................................................ 105
NTU and HTU Based on Film-Transfer Theory ............................... 107
Equilibrium Stages Calculated by the Kremser Method ................... 108
Physical Properties of Carbon Dioxide ............................................ 112
Physical Properties of Water ............................................................ 114
Physical Properties of Benzaldehyde ............................................... 1 14
Physical Properties of Isopentyl Alcohol ......................................... 114
Physical Properties of n-Propanol .................................................... 115
Ethanol Phase Calibration Standards ............................................... 117
Ethanol Phase G. C. Calibration Data .............................................. 118
G. C. Calibration Mixtures of n-Pr0panol/Water ............................. 125
G. C. Calibration Mixtures of Benzaldehyde/Water ......................... 125
G. C. Calibration Mixtures of Benzaldehyde/Water ......................... 126
n-Propanol/Water G. C. Calibration Data ........................................ 127

viii

C-S Benzaldehyde/Water G. C. Calibration Data .................................... 128
C-6 Benzaldehyde/Water G. C. Calibration Data .................................... 129
D-l Benzaldehyde/Water U. V. Calibration Data .................................... 136

ix

Figure

1-1

2-2

2-3

2-4
2-5

3-1
3-2
3-3
3-4
3-5
3-6

3-8
3-9
4-1
4-2
5—1
5-2

5-4

LIST OF FIGURES

Page
Chemical Structures for Benzaldehyde, Mandelonitrile,
and Amygdalin .................................................................................... 2
Liquid-Liquid Extraction Schematic Diagram ..................................... 8
Distribution Coefficient versus Carbon Atoms forll
n-Primary Alcohols ........................................................................... 11
Distribution Coefficient versus Carbon Atoms for
Selected Esters .................................................................................. 12
Pressure-Temperature Diagram for a Pure Component ...................... 16
Reduced Density vs. Reduced Pressure Near the Critical Point ......... 18
Solubility of Naphthalene in Ethylene ............................................... 20
Scheibel Extractor ............................................................................. 22
Crawford-Wilke Correlation .............................................................. 28
Flooding Correlations of Dell and Pratt ............................................. 30
Treybal Interfacial Tension ................................................................ 34
Graphical Representation of Extraction Column Stages .................... 39
Countercurrent, Stagewise Extraction Schematic .............................. 41
Concentration Profile Near a Liquid-Liquid Interface ....................... 44
Over-all and Individual Film Driving Forces ..................................... 47
Extraction with Continuous Countercurrent Contact ......................... 47
Equilibrium Relationship ................................................................... 58
Benzaldehyde Distribution Coefficient Measurement Schematic ....... 62
Benzaldehyde Extraction Column Feed Section Schematic ............... 74
Extraction Column Detail .................................................................. 75
Extraction Column Receiver Section Detail ...................................... 76
Benzaldehyde Extraction Column Flow Schematic ........................... 77

X

6-1
6-2
6-3

6-5
7-1
B-l
B—2
B-3
B-4

C-2
C-3

C-5
D-l
D-2

Liquid Redistributing Rings .............................................................. 80

One Equilibrium Stage Flow Simulation ........................................... 84
Two Equilibrium Stage Flow Simulation ........................................... 86
Benzaldehyde Extraction Column Flow Schematic ........................... 92
Sample Chromatogram for Ethanol Phase ......................................... 96
Run 3 Material Balance Diagram ...................................................... 98
Graphical Determination of Equilibrium Stages .............................. 106
IPA Concentration versus G. C. response ........................................ 120
Benzaldehyde Concentration versus G.C. Response ........................ 121
G. C. Column Operating Conditions for the Ethanol Phase ............. 122
Sample Chromatogram for Ethanol Phase ....................................... 123
n-Propanol Concentration versus G. C. Response ............................ 130
Benzaldehyde Concentration versus G. C. Response ....................... 131
Benzaldehyde Concentration versus G. C. Response ....................... 132
G. C. Column Operating Conditions for Aqueous Phase ................. 133
Sample Chromatogram for Aqueous Phase G. C. Analysis .............. 134
UV. Spectra for Benzaldehyde ........................................................ 135
Benzaldehyde/Water U. V. Calibration Curve .................................. 137

xi

 

 

w

SI

It:

CHAPTER 1

INTRODUCTION

The state of Michigan is one of the largest cherry growing states in the United
States. Each year, millions of pounds of cherries are processed to separate the fruit
from the pits. While the fruit is sold commercially, the cherry pits currently have no
commercial value and, in fact, are a liability since there is a cost associated with their
disposal.

The Cherry Marketing Institute has undertaken to finance research to find
profitable uses for the cherry pits. Research at Michigan State University has shown
that benzaldehyde, the chemical commonly referred to as cherry flavoring, is
recoverable from the cherry kernel via hydrolysis (Grethlein et a1, 1990) and
mechanical crushing which produces an oil containing benzaldehyde. The quantities
recovered from these two methods are 0.25-0.4 percent of the kernel weight as
benzaldehyde for the hydrolysis method (Gribb and Lira, 1992) and 32-36 percent of
the kernel weight as an oil containing predominately triglyceride with some
benzaldehyde for mechanical crushing (Briggs, 1991). It has also been shown that not
only is free benzaldehyde present in the kernel but that two other chemicals,
mandelonitrile and amygdalin, are also present. Both mandelonitrile and amygdalin
will react during hydrolysis to form benzaldehyde and by-products. The chemical
structures of benzaldehyde, mandelonitrile, and amygdalin are shown in Figure 1-1.
Although Grethlein’s study of cherry pit hydrolysis shows that benzaldehyde is present
in significant quantity, methods to recover the benzaldehyde from the dilute hydrolysis
broth were not considered.

Since benzaldehyde can be synthesized at a relatively low cost, further
investigation of the hydrolysis method is justified only if an economical method of

recovery from the dilute hydrolysis broth can be developed. It was thought at the

O OH
I l l
@c—H @C-CN
|
H
Benzaldehyde M andelonitrile
110 °\
on
no “'2
Mo Amygdalin
0" \
cucu

Figure 1-1. Chemical Structures for Benzaldehyde, Mandelonitrile, and Amygdalin.

beginning of this project that if any processing techniques used to recover the
benzaldehyde employed only naturally occurring materials, such as those typically
used in food processing, the value of this “natural” benzaldehyde would be sufficient
to justify the development of the required recovery process. In fact, the value of
“natural” benzaldehyde is close to $ 100 per pound or approximately one hundred times
the value of the synthesized material. This large increase in value could justify
processing methods not normally considered because of their high cost.

The low solubility of benzaldehyde in water and inhibition of the hydrolysis by
benzaldehyde concentrations greater than 40 parts per million (Grethlein et a], 1990)
mean that recovery from a dilute solution must be performed if the hydrolysis method
is to be developed to a commercial scale. Three methods that can be considered for

recovery of benzaldehyde from a dilute solution are steam distillation, adsorption, and

liquid-liquid extraction.

Steam distillation involves passing steam through the benzaldehyde/water broth
and volatilizing the benzaldehyde preferentially to the water. Steam can be added until
the benzaldehyde is concentrated into the overhead stream to its solubility limit and
then the overhead product can be collected as two phases (Cruess, 1958). This is an
inefficient technique because of the large amount of water which must be heated for the
recovery of a small amount of product. Also, it is possible that benzaldehyde will react
with oxygen present during the distillation to form benzoic acid resulting in an
appreciable yield loss.

An alternative to steam distillation is adsorption. In this process, the hydrolysis
broth is passed through an adsorbent bed where the benzaldehyde is retained and the
water passes through. The benzaldehyde is recovered by passing a new solvent with a
high affinity for benzaldehyde through the bed and desorbing the benzaldehyde from
the solid. Finally, the benzaldehyde must be recovered from the solvent. It is this final
step that generally determines the economic feasibility of the adsorption method since
the method usually used for separating the recovery solvent and the desired product is
distillation. Distillation has a relatively high capital cost compared to most unit
operations because it is energy intensive.

Recovery of the benzaldehyde from the dilute hydrolysis broth is also possible
using liquid-liquid extraction with either an organic solvent or liquefied gas as the
extraction solvent. Liquid-liquid extraction involves the contacting of a solution with
an immiscible solvent in which one or more of the original solution components is
soluble. Two liquid phases are formed after the original solution is contacted with the
new solvent. These two immiscible phases separate because of a difference in
densities. If the desired component of the original solution is more soluble in the new
solvent than in the initial solvent, the desired component will be extracted from the

original solvent. If organic solvents which are liquids at room temperature are used as

 

F.
l'C
th
as

C3

sir
flu
6x:
the
PFC

its

4

extraction solvents, distillation is usually required to separate the extraction solvent
and the solute (desired product). Again, distillation has a relatively high capital cost
compared to most unit operations because it is energy intensive.

The use of carbon dioxide as the extraction solvent for benzaldehyde is of
interest for four reasons: 1) benzaldehyde is completely soluble in liquid carbon
dioxide (Francis, 1954), 2) Related experiments indicate that the distribution
coefficient is on the order of 800-1000 which means that efficient extraction of the
benzaldehyde is possible in a small column, 3) carbon dioxide will not extract enzymes
necessary for hydrolysis which may allow recycle of the spent hydrolysis broth, and 4)
carbon dioxide extraction is currently used in food and flavor processing commercially
because it meets the strict requirements for extremely low (essentially nondetectable)
residual solvent concentrations and because it fulfills the all “natural” requirement for
increased product value.

In order to use carbon dioxide as a solvent in an extraction process it must be
compressed to a high enough pressure to liquefy it at the chosen operating temperature.
For economic reasons, as well as practical ones, this temperature is usually around
room temperature. As a general rule, the cost of liquefying carbon dioxide is high and
therefore its use as an extraction solvent is limited to high value added products such
as “natural” benzaldehyde. It is interesting to note that the critical temperature of
carbon dioxide is only slightly above room temperature at 31 degrees Celsius.
Therefore, at a temperature only a small amount higher than room temperature and at
similar pressures as those used for liquefying carbon dioxide, it becomes a supercritical
fluid. This is of particular interest when considering carbon dioxide for liquid-liquid
extraction because the diffusivity, viscosity and density are significantly different in
the supercritical fluid state than in the normal liquid state. It is possible that if the
processing conditions can be carefully controlled so that the carbon dioxide is kept near

its critical point, a significant increase in the solubility of benzaldehyde might occur

 

re
cc
cc
pr
se

de

be
be
Tl
de

USI

5

being greater than the solubility in liquid carbon dioxide. This increase in solubility
would reduce the cost of carbon dioxide extraction by decreasing the amount of carbon
dioxide used to perform the extraction and, therefore, would reduce the size of the
extraction column required for the desired recovery.

Because of the advantages of using liquid-liquid extraction with carbon dioxide
as the extraction solvent either in its liquid or supercritical-fluid state versus steam
distillation or adsorption, a pilot scale extraction apparatus was built to investigate the
recovery of benzaldehyde from a dilute hydrolysis broth. Successful operation of this
column could then be used to predict the feasibility of building a commercial-scale
column to recover benzaldehyde from hydrolysis broth that would come from
processing the fifteen million pounds of cherry pits that are currently disposed of every
season in Michigan. Prior to building the extraction apparatus, it was necessary to
develop the analytical techniques required to determine the amount of each component
in the streams entering and exiting the extraction column in order to determine its
performance. It was also necessary to measure the distribution coefficient of
benzaldehyde for carbon dioxide and water experimentally since this value could not
be found in the literature nor could sufficient data on activity coefficients or solubility
be found that could be used to predict the distribution coefficient with any reliability.
The theory underlying the principles of liquid-liquid extraction, an explanation of the
design of the equipment used in this project, a description of the analytical methods

used, and the results of the experiments performed are presented in this report.

 

 

t mll,N].“ilNI‘

.IIDWK

Many industrially important solutions of liquids form constant boiling mixtures
(azeotropes) or have components with such close boiling points that separations of
these liquids by ordinary distillation is not practical. Many other compounds,
particularly in the foods and flavorings industry, are heat sensitive or are not volatile
(e.g. penicillin) and can not be purified by distillation. Various other methods of
separation have been applied to such systems including liquid-liquid extraction.
Compared to distillation as a means of separation, liquid-liquid extraction is a more
recent operation. It has reached industrial significance only since 1930 (Orberg and

Jones, 1963). Some examples of liquid-liquid extractions of commercial significance

are given in Table 2-1.

CHAPTER 2

BACKGROUND

Table 2-1. Typical Liquid-Liquid Extractions

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Original Solvent Solute New Solvent
Water Acetic Acid Ethyl Acetate
Water Adipic Acid Diethyl Ether
Reforrnate Aromatics Diethylene Glycol
Water Benzoic Acid Carbon Tetrachloride
Water Diethylamine Toluene
LPG H2S MEA
Naphtha Mercaptans NaOH Solution
Water MEK Toluene
Lubricating Oil N aphthenes Furfural
Water Phenol Chlorobenzene

 

 

Lit

im:

 

 

sol

sol

wil

sol

ori

ha

m:

C0

St

in

C3

Liquid-Liquid Extraction

Liquid-liquid extraction involves the contacting of a solution with an
immiscible solvent in which one or more of the original solution components is
soluble. Two liquid phases are formed after the addition of the new solvent. If the new
solvent has a different density than the original solvent, the two phases will separate
with the less dense phase rising to the top. If the desired component of the original
solution is more soluble in the new solvent, this component will be extracted from the
original solution through contact with the new solvent.

Extraction may be carried out as a continuous or a batch process. Economics
generally favor continuous contacting because of high throughput capacities, ease of
handling the feed and shorter down times. There are several types of contactors which
may be used for continuous extraction: spray columns, packed columns, sieve tray
columns, and mechanically agitated columns of which the two most common are the
Scheibel column (Scheibel, 1948) and the Karr column. The spray, packed, and sieve
tray columns have the advantage of not using any internal moving parts and have a low
capital cost relative to the mechanically agitated type. However, mechanical agitation
can provide more surface area for mass transfer between the two phases resulting in a
more efficient extraction. A schematic diagram of a liquid-liquid extraction apparatus
is shown in Figure 2-1. This particular apparatus was used in pilot scale tests by
Schultz and Randall (1974) for the extraction of essences from liquids such as apple,
orange and pear juices in pilot scale tests. An interesting feature of this extraction
system design is the modifications that allow the use of liquid carbon dioxide as the

extraction solvent.

Literature Review
Solvent properties of liquid carbon dioxide were described by Gore (1861) who

published information on the solubility of naphthalene, iodine, and similar compounds.

 

 

 

 

up;

FEED

CAFflNATE

amnion
ouvt

 

 

 

        
 

\
txrucnou
comma

4
v

 

'vvvvrvv '

v vv

 

  

EVAPOIAIOR

 
 

 

co,

  

CO: SUPPLY

’ . WATER

\ CO:
CONDENSER

t

"ESSUIE
VESSEl WALL

 

 

 

 

(I
”‘3‘?

mo, cucuunou
rum

Figure 2-1. Liquid-Liquid Extraction Schematic Diagram.

I

 

Low
pate
6556
on t
the

yiel
as 2
tor

iru

Di

cil
or,
of
be
h(
SC

81

hi

9

Lowry et a1. (1927) measured the solubility of water in liquid carbon dioxide. Early
patents that specifically mention carbon dioxide as an extraction solvent for food
essences were obtained by Horvath (1939) and Brandt (1942). Francis (1954) worked
on the separation of dicyclic hydrocarbons from aliphatics and monocyclics and listed
the solubilities of 261 compounds in liquid carbon dioxide. The late 1960’s and 1970’s
yielded a plethora of patents and publications describing uses of liquid carbon dioxide
as a solvent; Sivetz (1963) cites usages of liquefied gases for extracting coffee aroma
constituents, Schultz (1966) reported on the use of liquid carbon dioxide for extracting
fruit aroma constituents, and spice extracts and equipment are described by Pekhov et

a1. (1969) and others.

Distribution Coefficient

The liquid carbon dioxide extraction of a juice or aqueous essence such as those
cited above will follow general extraction principles. The maximum amount of an
organic solute that theoretically can be extracted by liquid carbon dioxide in one stage
of mixing and settling is determined by the distribution (partition) of the solute
between the liquid carbon dioxide and the water. For oxygenated species in a
homologous series, as the molecular weight of the solute increases a greater part of the
solute will concentrate in the liquid carbon dioxide phase. Thus, fewer extraction
stages are needed to extract heavier compounds from an aqueous solution (Schultz and
Randall, 1974). It may be expected by inference that this phenomenon occurs in other
homologous series of compounds because of the increasing percentage of the structure
which is aliphatic.

The ability of liquid carbon dioxide to extract a solute from aqueous solution is
determined by the equilibrium distribution of the solute between the liquid carbon
dioxide and the water. For extremely dilute solutions Henry’s Law is applicable and

the distribution coefficient, m, of any compound, i, between the extract phase (solvent

 

 

 

OI

6X

3F
qt

SC

I6

th
dE

ht
C8

di

CO

be

10

or liquid carbon dioxide phase) and the raffinate phase (aqueous) phase can be

expressed as:

y i
m1 = — 2-1
x.
1
where yi mole fraction of component i in the extract phase
xi mole fraction of component i in the raffinate phase

The general significance of the distribution coefficient is that values greater than
approximately five or ten indicate a high recovery of solute is possible with the
quantity of extract probably being less than the raffinate. Distribution coefficients less
than one indicate the solute is incompletely extracted unless very large amounts of
solvent are used. This affects the economics of the liquid-liquid extraction since
recovery of the solute from the extract must always be taken into account and this will,
more than likely, be more expensive with a larger amount of solvent.

Given the significance of the distribution coefficient, methods or rules-of-
thumb for predicting it are very valuable for making preliminary determinations of the
desirability of using liquid-liquid extraction versus other separation techniques. For
dilute solutions, Henry’s Law is applicable. Schultz and Randall (1970) found that
there is a correlation between distribution coefficient and the number of carbon atoms
in a molecule for similar species (Figures 2-2 and 2-3). It may be that this relationship
holds for other species as well and, therefore, estimates of the distribution coefficient
can be made for any solute if data for similar compounds are known. In the absence of
distribution coefficient data, as is the case with liquid carbon dioxide systems in
general, activity coefficients can provide some basis for estimating how well a
compound might be extracted. Schultz and Randall (1970) found that a trend exists

between the activity of an organic compound in water and the equilibrium distribution

11

Distribution Coefficient

0.2 -

 

 

‘ 1 l I l J
o 12 s 4 e e 7
NumberofCarbon Atomsin Molecule

Figure 2-2. Distribution Coefficient versus Carbon Atoms for n-Primary Alcohols

12

   

 

 

800 -
600 -
400 -
‘5
.2 200 ~—
.2
‘6
o
o .00 _ 0 ACOIOICS
8 e Other n-eetere
E: 60 ._ .8 Breached ester:
.23 4o _
Q
m b
n I I I I I ' I I
O I 2 3 4 5 6 7
Numberot Carbon Atomsin Molecule

Figure 2-3. Distribution Coefficient versus Carbon Atoms for Selected Esters

13

of the compound between liquid carbon dioxide and water. See Table 2-2. Although
no formal correlation was attempted by these authors for the alcohols and esters
studied, in all cases large activity coefficients in water correspond to large distribution
coefficients.

When working with liquid carbon dioxide it is advantageous to keep costs down
by operating the extraction equipment at ambient temperature to eliminate the need for
expensive low-temperature cooling capability even though decreasing the temperature
of operation facilitates keeping the carbon dioxide in its liquid state. At ambient
temperatures, however, the equipment must be designed to operate at a minimum
pressure equal to the vapor pressure of carbon dioxide which is 882 psig at 22.4 °C.
Once this criterion is met, the maximum operating pressure will normally be specified
at approximately 1,500 psig or more to allow for fluctuations. When this is realized it
is useful to note that the extraction equipment is able to operate above the critical point
of carbon dioxide. This means that an examination of the advantages of performing a
supercritical fluid (SCF) extraction versus a liquid carbon dioxide extraction is
worthwhile and this was done for the separation of benzaldehyde from the hydrolysis

broth under investigation in this project.

Supercritical Fluid Extraction

Over the last decade there has been substantial interest in the use of supercritical
fluids. According to Akgerman et a1. (1991) three factors have contributed to the
recent attention: (1) the environmental problems associated with common industrial
solvents (mostly chlorinated hydrocarbons), (2) the increasing cost of energy-intensive
separation techniques (for example distillation), and (3) the inability of traditional
techniques to provide the necessary separations needed for emerging new industries
(microelectronics, biotechnology, etc.). In addition, more stringent pollution-control

legislation and the availability of inexpensive, nontoxic supercritical solvents such as

14

Table 2-2. Distribution and Activity Coefficients for Alcohols and Esters

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Compound CDiS‘FiPNiE“ I: 1 Y: 1
oeff101ent , K (experimental) (calculated)
16 °C 25 °C 25 °C
Methanol 0.40 --- 1.53
n-Propanol 0.66 l 1.7 11.40
n-Butanol 1.8 48.2 ---
n-Pentanol 4.5 169.0 ---
n-Hexanol 15 .0 --- 644.0
n-Heptanol 31.0 --- 2740.0
i-Propanol 0.35 6.54 ---
i-Butanol 1.7 --- 44.3
t-Butanol 0.82 10.1 «-
s-Butanol 1.23 --- 31.2
i-Pentanol 5.0 --- 157.0
Ethyl formate 16.0 52.0 ---
Ethyl acetate 42.0 65.0 77.6
Isopropyl acetate 80.0 --- ---
Methyl Butyrate 120.0 400.0 394.0
Butyl acetate 350.0 —-- 1270.0
Propyl propionate 370.0 1 100.0 1210.0
Isoamyl acetate 850.0 --- 5400 (est)

 

 

 

1 Activity coefficients by Pierotti et al., 1959
2 Distribution coefficients by Schultz et a1. Between water and liquid C02

 

 

 

car
wa
be
to

fill

in
an
$11

lis

be

CC

W
S L

SU

Si
Si
IC

15

carbon dioxide have prompted research into supercritical fluid extraction as a means of
waste water treatment. The focus of supercritical fluid use as a clean-up technique has
been in three major areas (Akgerman et al., 1977): 1) extraction of organic
contaminants from water, 2) extraction of organics from contaminated soil, and 3)
extraction of organics from adsorbents.

The critical point of a pure substance is defined as the highest temperature and
pressure at which a substance can exist in vapor-liquid equilibrium. At temperatures
and pressures above this point a single homogeneous fluid is formed and it is said to be
supercritical. A schematic diagram of this behavior is shown in Figure 2-4. Table 2-3
lists the critical temperatures and pressures for a number of gases and liquids.

Supercritical fluids exhibit some physical properties that are intermediate
between those of a gas and those of a liquid as shown in Table 2-4. It is this unique
combination of physical properties, at least in part, that frequently garners attention to
the use of supercritical fluids as extraction solvents. From a theoretical viewpoint, it
would appear that the decreased viscosity and increased diffusion coefficient of
supercritical fluids relative to liquids would indicate better mass transfer in
supercritical fluids and, therefore, more efficient extraction. In addition, the solvent
power of a supercritical fluid can be related to its density in the critical region. This
statement can be clarified by considering the density behavior of a pure component as
shown in Figure 2-5 (McHugh and Krukonis, 1986, p. 5). According to these authors,
for a reduced temperature, TR , in the range of 0.9 to 1.2 and at reduced pressures, PR ,
greater than 1.0, the reduced density, pR , of the solvent can change from a value of
about 0.1, a gas-like density, to about 2.5, a liquid-like density. As the reduced density
becomes liquid-like the supercritical fluid begins to act as a liquid solvent. This
increase in density corresponds to an increase in the solvent power of the supercritical
fluid. Conversely, a decrease in density accompanying a decrease in pressure results

in a dramatic decrease in the solubility of the solute. This is particularly useful for

16

 

7

SUPERCRHWCAL

%//

PRESSURE

 

 

 

 

 

TEMPERATURE

Figure 2-4. Pressure-Temperature Diagram for a Pure Component

17

Table 2-3. Critical Conditions for Various Solvents

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Critical Critical
Solvent Temperature Pressure

(°C) (atm)
Nitrogen Dioxide 157.8 100
Nitrogen ~147 33.5
Oxygen -118.4 50.1
Hydrogen -239.9 12.8
Helium -267.9 2.26
Carbon Dioxide 31.1 72.8
Ethane 32.3 48.2
Ethylene 9.3 49.7
Propane 96.7 41 .9
Pr0pylene 91 .9 45 .6
Cyclohexane 280.3 40.2
Isopropanol 235.2 47.0
Benzene 289.0 48.3
Toluene 318.6 40.6
p-Xylene 343.1 34.7
Chlorotrifluoromethane 28.9 38.7
Trichlorofluoromethane 198.1 43 .5
Ammonia 132.5 111.3
Water 374.2 217.6

 

 

 

 

 

Weast, 1969-1970

 

 

18

Table 2-4. Comparison of PrOperties for a Gas, SCF, and Liquid

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Property Gas SCF Liquid
Density (kg/m3) 1 700 1000
viscosity (118/I112) ‘ ‘ ' 11015 “510:4.1- - 1073 '
Diffusion Coefficient (cmzls) 10'1 10'4 10'5
I I I I I I II] I r I I I TI
.. T -
I 0.9
2.0 . : I I0 -
| l
1 I
I l
U
of" ' l
t! I l
0: l0 ~ I I .
o. I l
l I
' I
l I
I
I
O l 1 I 1 1 III 1 1 I 1 1 11
0.1 1.0 10.0

-P
Paw»,

Figure 2-5. Reduced Density vs. Reduced Pressure Near the Critical Point

 

SC

8C

I13

qu
SO

an

IE

lit
50
0C
A(

SU;

61"

l9

separation of the supercritical fluid and the solute after the actual extraction is
accomplished.

The solvent power of supercritical fluids is typified by the solubility of solid
naphthalene in supercritical ethylene as shown in Figure 2-6. At a temperature of
12 °C (TR = 1.01) the solubility of naphthalene in supercritical ethylene increases
quite dramatically as the pressure is increased to 50 atmospheres and higher. The
solubility reaches a maximum of about 1.5 mole percent at pressures greater than 90
atmospheres. At pressures below 50 atmospheres the naphthalene solubility is quite
low as would be expected for the solubility of a solid in a gas. The solubility behavior
along the 12 °C isotherm can be interpreted by considering the reduced-density
isotherm at 1.0 shown previously in Figure 2-5. Note that the naphthalene isotherm has
the same general characteristic shape as the reduced-density isotherm at 1.0. From this
example, it can readily be seen why the solvent behavior of a supercritical fluid is
related (to a first approximation) to the solvent density behavior in the critical region
(McHugh and Krukonis, 1986, p 5). Considering the 35 °C isotherm in Figure 2-6, it
can be seen that the solubility behavior is not as sensitive to pressure in the region near
50 atmospheres as it was for the 12 °C isotherm. However, at pressures greater than
100 atmospheres where ethylene exhibits liquid-like densities, the 35 °C isotherm
reaches a higher limiting solubility value of about 5 mole percent.

The properties of gas-like diffusivity and viscosity, zero surface tension, and
liquid-like density combined with the pressure dependent solvent power of a
supercritical fluid make its use as an extraction solvent very attractive. However, on
occasion these properties may offer no advantage relative to conventional solvents.
According to McHugh and Krukonis (1986, p 10), it is necessary to evaluate
supercritical fluid technology on a case-by-case basis since the advantage gained by the
mass transfer properties of a supercritical fluid are relative to the entire system under

evaluation.

MOLE FRACTION OF NAPHTHALENE

2O

 

 

 

 

01000 I I I I I I
35°C

0.0100 - 12 °C "

0.0010 - -

0.000! I I I I I I

O 50 100 150 200 250 300 350
PRESSUREIotm)

Figure 2-6. Solubility of Naphthalene in Ethylene

CHAPTER 3

LIQUID-LIQUID EXTRACTION

Liquid-Liquid extraction is the separation of the components of one liquid
solution by contact with another insoluble liquid. Two liquid phases form after the two
liquids are mixed together, therefore, the densities of the two phases must be different
so that the phases will separate. The solution which is to be contacted is the feed and
the liquid with which the feed is to be contacted is the solvent. The components
dissolved in the feed are the solutes. The solvent-rich product of the extraction
operation is the extract and the solute depleted feed is the raffinate. Ideally, the
valuable components of the initial solution will preferentially partition into the solvent
thus being extracted from the feed.

There are many types of liquid-liquid extraction devices. The various devices
are generally classified into stage and continuous-contact (differential) types. The
stage type includes mixer-settlers and sieve-tray (perforated plate) towers. The
continuous contact type includes spray towers, packed towers, centrifugal extractors,
and mechanically agitated, countercurrent extractors. The later type includes the
rotating-disk contactor, Mixco Lightnin CM contactor, Scheibel extractor (Figure 3— 1 ),
and Karr column.

There are many factors involved in the selection and design of a liquid-liquid
extraction device such as those described above. The following discussion examines

these factors from both the operational and mass transfer points of view.

Solvent Selection
There are several important factors governing the selection of the solvent used
for extraction. The following criteria are recommended by Treybal (Treybal, 1980, p

488):

21

22

 

 

 

 

Figure 3-1. Scheibel Extractor

l. Selectivity. The effectiveness of solvent B for separating a solution of solvent

A and solute C into its components is the selectivity, B. At equilibrium:

t

_ (wtfractionAinR) (wtfractionCinE) _ (wtfractionAinR)xa
(wtfractionAinE) (wtfractionCin R) (wtfractionAin E) xR

 

 

13

where E is the extract stream
R is the raffinate stream
For useful extraction operations the selectivity must exceed unity. If the selectivity is

unity, no separation is possible.

i
2. Distribution Coefi‘icient. This is the ratio xE/xR in the above equation. While

23

it is not necessary for the distribution coefficient to be greater than unity, large values
are desirable since less solvent will then be required for the extraction.

3. Insolubility of solvent. The extraction solvent should be as completely
insoluble in the original solvent as possible. Ideally, the two solvents are completely
immiscible. Immiscible solvents will have higher selectivities than solvents that
exhibit partial solubility.

4. Recoverability. It is usually preferred to separate the solvent from the solute
after the extraction and recover the solute for reuse. This is most often accomplished
by distillation therefore the relative volatility of the solute and solvent should be
considered. Whether it is the solvent or the solute, that substance which is present in
the extract in the lesser quantity should be the most volatile. The cost of separating and
recovering the solute can significantly influence the cost of the extraction operation.

5. Density. A difference in the saturated liquid densities is necessary for the
phases to separate. The larger the difference the better. The density difference for
equilibrium phases decreases with increasing solute concentration and this must be
carefully considered during the solvent selection.

6. Interfacial tension. There exists an interfacial tension between two immiscible
liquid phases. Quite often the interfacial tension between the two liquid solvents is
altered markedly by the presence of the solute. While high values of interfacial tension
tend to increase the difficulty of dispersing one phase in another, a large value of
interfacial tension will promote the coalescence of the dispersed phase droplets. On
the other hand, low values of interfacial tension may produce an emulsion of the
dispersed phase. The tiny droplets of such an emulsion do not separate easily from the
continuous phase which makes operation of the extractor difficult.

7. Chemical Reactivity. The solvent should be inert toward the other components
of the system and towards the materials of construction.

8. Viscosity. If the solvent is the continuous phase, a high solvent viscosity will

 

W
to
so

[IE

C0

P“

dis

31

OI
ma
IllC

6X1

fen
for
1w
0f 1
Carl

fiat

24

decrease the rate of rise (or fall) of the dispersed phase for a given pressure drop. This
increases the contact time between phases and the total mass of solute transferred.
However, flooding correlations such as those of Crawford and Wilke (Crawford and
Wilke, 1951) or Rao and Rao (Rao and Rao, 1958) and others show that a high viscosity
continuous phase lowers the column capacity by lowering the flooding velocity. If the
solvent is the dispersed phase, a high viscosity will result in a lower rate of mass
transfer relative to a lower viscosity and the rate of solute diffusion will be reduced.
However, it has been demonstrated that dispersed phase viscosity has little affect on
column capacity (Breckenfeld, 1942).

9. Vapor pressure and freezing point. These should be as low as possible to
promote ease of handling.

10. Safety. The solvent should be nontoxic and nonflammable.

Since no solvent meets all of the optimum criteria, the selectivity and
distribution coefficient usually play a significant role in selecting the solvent used for
a particular extraction. However, the recoverability of the solvent can have a high
enough cost so as to make the choice of solvent favor the one with a lower selectivity
or distribution coefficient. For pharmaceuticals and food applications, solvent toxicity
may be the overriding criterion. Regardless of which solvent property is deemed the
most important, the physical properties of the solvent will affect which type of
extractor is selected.

Because the purpose of this research was to extract a food flavoring from a
fermentation broth, the solvent toxicity was of primary importance. In addition, even
for a solvent with acceptable toxicity, the solvent recoverability in terms of very low
levels of residual solvent in the product was also important. Carbon dioxide met both
of these criteria. The high cost of an extraction process designed to utilize liquid
carbon dioxide was thought to be justified by the high value of the recovered cherry

flavoring (benzaldehyde).

25

Once carbon dioxide was selected as the solvent, its properties in relation to the
water/benzaldehyde feed mixture, the cost of operation, and the relative capital cost of
each type of extractor were used to select the continuous, packed column as the type of

extractor to use for this project.

Dispersed Phase

When extraction is carried out in a packed column, the more dense solvent is
introduced near the top and the less dense solvent is introduced near the bottom. In
order for mass transfer of the solute to occur, the phases must contact each other. This
occurs in such a way that one phase is dispersed in the other in the form of droplets.
The smaller the droplets the greater the interfacial surface area and the greater the rate
of mass transfer. The phase which is not dispersed is the continuous phase. Either of
the solvents can be the dispersed phase.

The choice of which phase is the continuous one and which is the dispersed one
is based on mass transfer. Usually, the solvent with the higher mass transfer coefficient
should be selected as the continuous phase since the path length for solute travel is
larger in the continuous phase. Since the higher mass transfer coefficient is related to
lower viscosity, the continuous phase should also be the phase with the lower viscosity.

For this research, carbon dioxide was selected as the continuous phase since it
has a lower viscosity than water. With carbon dioxide as the continuous phase, water
must be introduced at the top of the column because its density is greater than that of

carbon dioxide.

Column Capacity
Extraction is often accomplished in a column filled with packing similar to the
types used in distillation. During proper operation, the dispersed phase droplets fall

(or rise) through the interstitial voids in the packing immersed in the continuous phase.

26

If the flow rate of the dispersed phase is increased, the droplets will occupy a larger
fraction of the of the void space. Further increases in the dispersed phase flow will
eventually result in the void space becoming filled with droplets that coalesce to
completely fill the void space in the packing. This phenomenon is flooding and the
superficial velocity at which flooding occurs is the flooding velocity. Flooding can
also occur if the dispersed phase flow is held constant and the flow rate of the
continuous phase is increased. Eventually, the continuous phase flow will reach a
velocity sufficient to prevent the dispersed phase droplets from moving down (or up)
the column. Thus, the flooding of the column is dependent upon the flow rates of both
phases.

The physical properties of each of the solvents are also related to column
flooding. Consider the following. hypothetical experiments. The flow rates of both
phases are held constant at those corresponding to normal operation of the column for
each of the following experiments. The more dense, dispersed phase is assumed to be
descending through the less dense continuous phase (the situation for this research):

1. If the density of the dispersed phase could be increased independently of any
other properties, the terminal velocity of the droplets would increase and there would
result a corresponding increase in the flooding velocity;

2. If the viscosity of the continuous phase were increased independently of any
other properties, the rate of descent of the dispersed phase droplets would decrease and
there would be a corresponding decrease in the flooding velocity;

3. If the surface tension of the dispersed phase were varied independently of any
other properties, the droplet size would change resulting in a change in the rate of
descent of the droplets thus changing the flooding velocity;

4. If the void space between the individual pieces of packing is decreased it will
take fewer droplets to fill the smaller space. This would correspond to a lower flooding

velocity.

27

Thus, it would appear that the flooding velocity is affected by density, viscosity,

surface tension, and the packing free volume as well as the superficial velocity of both

phases.

Investigations of the flooding rates in columns have been carried out and

reported in the literature. Although very useful, many of these studies and the resulting

correlations suffer from the deficiency of having observed only pure component, binary

systems. The correlations resulting from three of these studies are presented below:

1. Modified Crawford-Wilke (Crawford, 1951) correlation:

pc
AP

0'

92 (1)02 (F32)

1.5
versus

PC 2
(VD+VC)

apuc 3-2

 

Surface area of packing (ft2/ft3)

Packing factor, dimensionless

Continuous phase superficial velocity (ft/hr)
Distributed phase superficial velocity (ft/hr)

Void fraction of packing, dimensionless

Continuous phase viscosity (cps)
Density of the continuous phase, (lb/ft3)

Density difference between phases (lb/ft3)

Interfacial tension (dyne/cm)

Generally, values for the constants in the left-hand expression are known or are

easily measurable. Calculating this value and locating its intercept with the curve in

Figure 3-1 allows the value for the right-hand expression to be read from the curve.

This value corresponds to the combined phase flow rates that will flood the column

(vcu + V,” 1'

28

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I 1 All

 

10

0-
I.
4:-
II-
I.

Y Y I

an to ea :00 eoo'eoo iooo

lirl-I‘I-i- -—l"“"’”

Figure 3-2. Crawford-Wilke Correlation

 

pl

29

since the values for pc, pc, and ap have been determined previously. When the
velocity of one phase is specified the other can be calculated so that the combined flows
will be less than the combination which will result in flooding.

The original Crawford and Wilke correlation was based on eight different

binary liquid systems. The properties varied as follows:

packing void fractions maximum of 0.74
liquid viscosity 0.58 to 7.8 cps
liquid density difference 9.4 to 37.2 lb/ft3
interfacial tension 8.9 to 44.8 dyne/cm

The modified correlation was based on data from six different tower packings
with void fractions, 8, having values up to 0.94. All of these flooding data were based
purely on hydraulic flow data with no mass transfer of solute occurring.

2. Rao and Rao (Rao, 1957) correlation:

 

where ap Surface area of packing (ftZ/ft3)
g Gravitational acceleration, ft/hr2
VC Continuous phase superficial velocity (ft/hr)
VD Distributed phase superficial velocity (ft/hr)
8 Void fraction of packing, dimensionless
pC Density of the continuous phase, (lb/ft3)

pD Density of the discontinuous phase, (lb/ft3)

and

floc

C011

1169

WE

Jmiefiwxfiflwvnnoé + .

3O

Ap Density difference between phases (lb/ft3)
G Interfacial tension (dyne/cm)

C Curve-fit parameter

n Curve-fit parameter

The authors correlated their data from over 50 runs with different liquid systems
and packings. Interfacial tensions ranged from 10 dyne/cm to 39 dyne/cm. Their
flooding data were fitted by the dimensionless expression Equation 3-3 when the
constant, C, and the value of the slope of the curve, n, were taken to be 0.66 and
negative 0.25 respectively. The original development for the dimensionless equation

was done by Dell and Pratt. Such a plot is shown in Figure 3-3.

)‘4

)tt-gg-

Va
Vc

I + 0-eae(

 

102030601002004006001000 30006000

V 0 Po
g€

be:

 

Figure 3—3. Flooding Correlation of Dell and Pratt

31

The solvent physical properties and packing parameters for any system under
consideration are generally known. For a known continuous phase superficial velocity,
VC, the expression on the right-hand side of the expression can be calculated. Locating
this value on the curve and reading the value for the left-hand expression and setting it
equal to the left-hand expression allows a value for the dispersed phase superficial
velocity, VD, to be calculated. This dispersed phase velocity will result in a flooding
condition in the column when combined with the previously specified continuous
phase velocity.

3. Venkataraman and Laddha (Venkataraman, 1960) correlation:

 

 

p 0.25 V 0.5 Vza p a 87 n
90 VC ge AP pCVC
where ap Surface area of packing (ft2/ft3)

g Gravitational acceleration, ft/hr2

VC Continuous phase superficial velocity (ft/hr)

VD Distributed phase superficial velocity (ft/hr)

8 Void fraction of packing, dimensionless

pC Density of the continuous phase, (lb/ft3)

pD Density of the discontinuous phase, (lb/ft3)

Ap Density difference between phases (lb/ft3)

Interfacial tension (lb/hr)

C Curve-fit parameter

n Curve-fit parameter

al
51

be

fr.

311

pa

alJ

32

Table 3-1. Curve-Fit Parameters for the Correlation of Rao and Rao

 

 

 

 

 

 

Average

Type of Packing C n Deviation
%
Raschig rings 0.894 -0.078 7.6
Ber] saddles 0.882 -0.052 8.8
Lessing rings 0.853 -0.046 3.7
Spheres 0.839 —0.029 5.0

 

 

 

 

 

 

The parameters C and n were obtained by the method of least squares for the
packings given in Table 3-1.

The parameter n varies significantly with packing type. The authors found that
after a statistical analysis of their experimental data and those reported by others, with
all the types and sizes of packing, different tower diameters, and different liquid
systems taken into account, the values of C and n as a general correlation are found to
be 0.812 and negative 0.048 respectively. With these values of C and n the average
deviation of all the points is 12.6%, and the maximum deviations are +32% and -29%
from those predicted by the correlation. The accuracy for the specific values of C and
n for a particular packing are significantly better than those for the general correlation.

These three correlations were selected for review primarily because their scope
encompassed the range of variables encountered in this study (i.e. packing size, surface
tension, and density). All three of these correlations claim to predict flooding in
packed columns with the qualification that the operating velocity should be

approximately one half the flooding velocity.

 

Inter

pack
finch
easfl;
anul
cant

vdue

isno
corre
transl
cdcui
agnn
Mfluu
qud
The]

CSllm;

ngI

Pitssu

Where

33

Interfacial Tension

The interfacial tension between the two liquid phases affects the operation of a
packed column by determining the size of the dispersed phase droplets. If the
interfacial tension is very low, tiny droplets will be formed which will not separate
easily from the continuous phase. If the interfacial tension is sufficiently low, an
emulsion of the dispersed phase could be formed. High values of interfacial tension
can cause difficulty in dispersing one phase in the other while. On the other hand, high
values of interfacial tension promote coalescence of the dispersed phase for separation.

That interfacial tension is important in liquid extraction is rather obvious. What
is not obvious is how to calculate interfacial tension values for use in the above
correlations for column capacity or to estimate the interfacial surface area for mass
transfer calculations. The interfacial tension between the two liquid phases cannot be
calculated as the difference between the individual phase surface tensions measured
against air. Since the interfacial tension is strongly dependent on the concentration of
solute present in a ternary system, values of the surface tension for a variety of binary
liquid systems reported in the literature can provide only useful approximations at best.
The Treybal correlation for interfacial tension shown in Figure 3-4 is useful for
estimating interfacial tension in the absence of pure component, binary data.

Another useful method for estimating interfacial tension is that of Massoudi and
King (1974). For pure water with various gases the interfacial tension as a function of

pressure is given by:

o=GO+BP+CP2+DP3 3.5

where Go is the interfacial tension of water with the gas at atmospheric

pressure.

where

lnterteciei Teneion ~ dyne/cm

34

 

 

em W em: m 0.000 on 0.02 w I.“ I” 0.20 M)
II“... a; x«-

x AB is the mole fraction of solvent A dissolved in solvent B,

XBA is the mole fraction of solvent B dissolved in solvent A,

xC A is the mole fraction of solute C dissolved in solvent A, and

XCB is the mole fraction of solute C dissolved in solvent B.

Figure 3—4. Treybal Interfacial Tension

Va

ex}

PaI

56C]

for

35

Values for the coefficients determined by fitting the polynomial equation to

experimental data are given in Table 3-2.

Packing
Packing performs four important functions in continuous extractors:

1. Increases the flow path length for the dispersed phase. This increases the
contact time between the phases producing a greater amount of solute transfer. For a
given steady state flow rate, the relationship between increased path length and
packing void volume is the holdup;

2. Maintains a uniform distribution of flow of both phases across the entire cross
section of the column;

3. Distorts the dispersed phase droplets which increases the surface area available

for mass transfer and also causes mixing within the droplet again increasing mass

Table 3-2. Interfacial Tension for Pure Water with Various Gases

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Gas B C 2 D 3
dyn/cm atm dyn/cm atm dyn/cm atm
02 -0.0779 0.000104
N2 -0.0835 0.000194
Ar -0.0840 0.000194
CO -0.1041 0.000239
CH4 -O.1547 0.000456
C2H4 -0.6353 0.003160
C2H6 -0.4376 -0.001570
C3H8 -0.9681 -0.0589
N20 -0.6231 0.00287 -0.000040
C02 -0.7789 0.00543 -0.000042

 

11311

the

pat
the
for
for
pat
pat

car

Be
are

are

oh

P21

111.

36

transfer. This droplet distortion also occurs as continuous coalescence and breakup of
the drops (Coulson and Richardson, 1978);

4. Prevents back mixing of the phases.

The role of packing in liquid-liquid extraction is somewhat different than for
packing in gas-liquid contacting (distillation or absorption). In gas-liquid contacting,
the liquid must wet the packing so as to provide the maximum possible interfacial area
for mass transfer. In liquid-liquid extraction however, interfacial area occurs in the
form of dispersed phase droplets and it is undesirable for the dispersed phase to wet the
packing since this generally results in the dispersed phase running in rivulets down the
packing if it is the more dense phase or an increase in holdup and decreased column
capacity if the dispersed phase is the less dense phase.

Packings for liquid-liquid extraction are almost exclusively randomly packed.
Berl saddles, Rashig rings, Pall rings, Intalox saddles, and Lessing rings, among others,
are used. A wide variety of materials of construction for these manufactured packings
are used including different metals, ceramics and polymers.

Packing size selection should follow two guidelines:

1. The packing should not be larger than 1/8 times the column diameter to avoid
channeling and to reduce dispersed phase contacting at the wall. For this research, the
packing selected was 1/8 inch, perforated-316 SS squares;

2. The minimum packing size should be such that the mean void height is not less
than the mean drop diameter. This critical packing size for a pair of liquids is given by

(Gayler, Roberts, and Pratt, 1953):

0.5

dcm = anti] 3-6

37

where 0' interfacial surface tension, M /T2
Ap density difference between the two phases, M /L3
g acceleration due to gravity, L/T2

Extractor Internals

Although the most common manner of operating an extractor is with the light
phase dispersed, the following discussion of the extractor internals assumes that the
light phase is the continuous phase since this is the desirable method for this project.
Nevertheless, the description of the internals applies to the dispersion of either phase.
(If the light phase is the dispersed phase, then the following equipment item
descriptions will be valid if they are exchanged end-for-end in the column where
appropriate.)

At the top of the column it is necessary to disperse the heavy phase. For small
columns, an arrangement of nozzles utilizing the discharge pressure of the feed pump
is used to introduce dispersed phase streams through orifices typically 0.125-0.25
inches in diameter. For this research, the feed nozzle orifice diameter was 1/32 inches.
The exact diameter is chosen to cause the formation of small droplets with the intent
of maximizing the interfacial surface area for mass transfer. For larger columns, a
disperser plate is used. In this case, a sparger is used to fill the disperser plate with the
heavy phase. The difference in density between the two phases causes the heavy phase
to fall through downcomers in the plate arranged to distribute the heavy phase as
uniformly as possible across the cross section of the column. The less dense
continuous phase rises up to the bottom of the disperser plate and passes through holes
arranged uniformly across the cross section of the plate.

The light phase is introduced into the side of the column near the bottom

through a standard side fitting. The distance of the light phase injection point from the

bot

bot

usu
the

bot

int:

Ex“

ind
U51
cor
con
the

C011

Slag
Selt
€Xlr
nun

9011:

38

bottom is sufficient to allow room for the separation of the two phases in the very
bottom of the column.

At the bottom of the bed it is necessary to separate the two phases. This is
usually done by a flow control valve which regulates the flow of the heavy phase out
the bottom of the column while maintaining a small level of the heavy phase in the
bottom of the column. This small level of the heavy phase insures that none of the light
phase escapes out the bottom opening. The control valve is adjusted to maintain the

interface between the two phases at the desired level.

Extraction Column Sizing

As previously discussed, flooding curves have been developed correlating the
solvent physical properties and physical characteristics of the column and packing to
the maximum phase velocities which result in a flooding condition in the column.
Knowledge of the flooding velocities is used to calculate the required column diameter
for a given throughput. This is done by adjusting the column diameter to give the
individual phase velocities which correspond to 50% of the flooding velocities.
Usually, it is necessary to build a pilot scale column and run it to determine which
correlation(s) fit the data or to generate a flooding curve unique to the system under
consideration. The second criteria required for the sizing of an extraction column is
the column’s height. Several methods used to predict the height of an extraction
column are described below.

1. One way to calculate column height is to estimate the number of theoretical
stages (NTS) required for a given separation. Each stage can be viewed as a mixer/
settler in which the two phases are well contacted and then separated cleanly into
extract and raffinate streams. Material balances can be done for each stage and the
number of theoretical stages calculated. The NTS can be determined from data

collected from a pilot column giving a measurable separation with a known height of

39

packing. The Height Equivalent of a Theoretical Stage (HETS) is the height of the
packing for the experimental column divided by the NTS required for the separation.
This procedure can be done graphically by plotting the ratio of the solvent flows as a
line on rectangular plot of the raffinate composition (abscissa) versus the extract
composition (ordinate). This line is called the operating line. The number of
equilibrium stages is determined by connecting the operating line with the equilibrium
curve for the system by a series of steps as shown in Figure 3-5.

2. A second method for determining the height of an extraction column is to view
the solute concentration in each phase as changing differentially with height. This
method is based on the diffusion of solutes across a thin film at the interface separating
the two bulk phases. The value for the number of transfer units (NTU) when multiplied

by the height of a transfer unit, HTU, gives the height of the tower:

 

10

 

8
\

 

 

 

 

 

 

 

.2; /
g /
is t.

a /

 

\

\\

("a
/
«V

 

 

 

 

 

 

 

 

 

 

0 0.5 1.0 1.5 240 2.5 3.0 3.5

lb Solute per 1000 to Water

Figure 3-5. Graphical Representation of Extraction Column Stages

40

H = NTU XHTU 3-7

3. A third method for calculating the column height for the special case when the
two solvents are mutually insoluble and the distribution coefficient is independent of
concentration relies on the extraction factor, U, first described by Kremser (1930). The
extraction factor is related to the number of theoretical stages required for a specified
separation and to the column height. Whereas the previous two methods rely on
experimental column testing, the Kremser extraction factor method does not and can
be used as a predictive method.

Each of the three methods is described in detail below.

Determination of the Number of 'Ii'ansfer Stages (NTS)

The concept of the number of transfer stages required for a given extraction may
be visualized by considering the development of a graphical method for determining
the number of transfer stages required for a series of discrete mixing and separating
vessels. These vessels are arranged for countercurrent contacting utilizing two
completely immiscible solvents. The conditions for this type of flow can be
represented as shown in Figure 3-6. Each circle represents a mixer and a separator.
The initial solution F of the solute B dissolved in solvent A is fed to the first unit and
leaves as raffinate R1. This stream passes through the units and leaves from the nth
unit as stream Rn. Fresh solvent enters the nth unit and passes in the reverse direction
through the units, leaving as extract E1. Since the two solvents are immiscible, the
original solvent in the raffinate streams remains constant at A units and the added fresh
solvent in the extract streams also remains constant at S units.

Material balances for the solute can be written in several ways. For the first

stage as:

38!!

bee

And

41

Xf‘ X1 X2 X3 Xn-1 Xn

OHIO-0-
E 4 E n E n + 1
Y1 Y2 Y3 Y4 Yn Yn e 1

Figure 3-6. Countercurrent, Stagewise Extraction Schematic

AXf+SY2=AXl+SYl 3-8

where solute concentrations in the raffinate and extract streams, X and Y, are expressed
as mass units per unit of solvent. Material balances for the solute for the nth stage can

be expressed as:

AX +SYn+1 = AXn+SYn . 3-9

n-l

And for the whole unit as:

AXf+SYn+1 = AXn+SY1 3—10

42

11+

A
Y 1=§(xn—xf)+Y1 3-11

Equation 3-11 is in the form of an equation for a straight line of slope A/ S . It
is known as the operating line and is shown in Figure 3-5. It is seen to pass through
the points (Xf, Y1) and (Xn, YD”). Also drawn in Figure 3-5 is the equilibrium relation
Yn versus X“. The number of stages required to pass from Xf to Xn is found by drawing
in steps between the operating line and the equilibrium line; in Figure 3-5 four stages
are required and (Xn, Yn+1) corresponds to (X4, Y5). It should be noted that the
operating line connects the compositions of the raffinate stream leaving each unit and
the solvent stream entering each unit.

This graphical development of the number of stages required for an extraction
is frequently used to estimate the number of stages required for a continuous extraction
because of its ease of use. In most cases, the data necessary to generate the equilibrium
curve can be generated in the laboratory. Consider the following example (Treybal,
1951):

An aqueous process effluent contains 3.4 lb of a valuable organic
compound per 1,000 lb of water. Since the organic material has a boiling

point only slightly higher than water, an extraction of this organic

appears to be the least expensive method for recovery of the solute. This

organic is highly soluble in aromatic solvents that are almost immiscible

in water. Therefore, p-diethyl benzene (DEB) will be used as the

solvent. Tests of mixtures prepared in the laboratory give the

distribution coefficients shown in Table 3-3.

The data from Table 3-3 can easily be used to construct an
equilibrium diagram on rectangular coordinates as in Figure 3-5. To

recover 95% of the solute would give a concentration of 0.17 lb of solute

43

Table 3-3. Equilibrium Data

 

 

 

 

 

 

 

 

Solute in Solute in
Water DEB m
lb/l ,000 lb lb/1,000 lb
2.1 21.1 10.05
1.7 14.2 8.35
1.2 8.7 7.25
0.7 4.3 6.14
0.2 1.2 6.00

 

 

 

 

per 1,000 lb of water in the raffinate. Assuming the DEB is solute free
means that the concentration at the bottom of the extractor can be
established. This concentration will lie on the X-axis of the diagram.
Since DEB and water are almost immiscible and the amount of solute
transferred is small compared to the total mass flow rates of the solvents,
the operating line of this column will be essentially a straight line. Next,
the slope of the operating line can be fixed by choosing the DEB flow
rate so as to avoid a pinch against the equilibrium curve. Further, since
a packed column has been selected, the DEB flow rate will be selected
so that no more than five theoretical stages of mass transfer are required.
The equilibrium stages are stepped off and are shown as the dashed line
in Figure 3-5. For this extraction, 4.3 theoretical stages are necessary

for 95% recovery.

Dehn'

can be
hquid
dnxin
dnius
theph
1924)
conce
sohne
extrac
the si

condn

44

Determination of the Number of Film Transfer Units

The transfer of solute in liquid-liquid extraction in a continuous packed column
can be explained by applying the two-film theory of diffusion. For the two immiscible
liquids in contact, there will exist concentration gradients in both phases which act as
driving forces for the transfer of solute from one phase to the other. The resistance to
diffusion can be represented by an effective film thickness near the interface between
the phases. Assuming that equilibrium is established at the interface, (Whitman, p 147,
1924) the transfer of solute on the raffinate side of the interface is driven by the
concentration difference CR — CRi where the subscript i refers to the concentration of
solute that exists at the interface between the two phases. The transfer of solute on the
extract side is driven by the concentration difference CEi — CE. Figure 3-7 represents
the situation just described (Coulson and Richardson, 1968). Under steady-state

conditions, the rate of transfer of solute through the extract phase equals the rate of

 

 

 

 

 

 

 

Raffinate (:51
Film
C
0 E

s e
2 CR 95’
a H
0
U
c
0
U

CR1

Extract
Film
Distance

Figure 3-7. Concentration Profile Near a Liquid-Liquid Interface

transfer

express

where

lianSfer‘

(IN

01

 

45

transfer through the raffinate phase. The rate of transfer across these films, N, can be

expressed in the following way:

N = kRS (CR—C = kES (CEi—C 3-12

Ri) E)

where CR, CE are the concentrations of solute in the raffinate and extract

phases, gmol /cm3

CRi, CEi are the concentrations of solute in the raffinate and extract

. 3
at the interface between the phases, gmol / cm

kR, kE are the transfer coefficients for raffinate and extract films,

gmol

 

2 3
sec -cm .gmol/cm

S cross-sectional area perpendicular to the direction of mass

2
transfer, cm

N rate of mass transfer, gmol /sec

For continuous extraction, it is useful to consider the differential rate of mass

transfer, dN , which will occur through a differential area, dS (Treybal, 1951). Then:

(IN = des (CR—Cm) = kEdS (CIBi —CE) 3-13

01'

dN

deSACR = kEdSACE 3-14

point
equihl
3-8.

contin

di

nhere

Iltste:
theOry
Treyb;

averag

Where

46

The equilibrium distribution curve will contain the point (CEi, cm) and the
point represented by the main body concentrations (CE,CR) will be below the
equilibrium curve (Treybal, 1951, p 119). These concentrations are shown in Figure
3-8.

A solute balance over the differential height dH for the extraction with

continuous countercurrent contact shown in Figure 3-9 (Treybal, 1951, p 242) is:

d(RxR) = d(ExE) 3-15
where R is the raffinate phase flow rate, gmol/sec
E is the extract phase flow rate, gmol/sec
xR is the solute mole fraction in the raffinate
xE is the solute mole fraction in the extract

At steady state, the solute from the raffinate was previously described by the two-film
theory to be transferred across the interface between the two phases (Equation 3-13).

Treybal (1951, p 242) defines the average solute concentration, CRM gmol/sec, as the

average of the values of CR and CRi- Substitution into Equation 3-13 yields:

 

dN = deSCRM (xR — xRi) 3-16
where xR is the mole fraction of solute in the raffinate
XRi is the mole fraction of solute in the raffinate interfacial film
(CR ' CRi) .
CRM = 2 where CR refers to the total concentratron of all

substances present.

47

.9.

 

flail/brim»
dd’fn‘bu'fi'on

nun-«-

 

C‘
n

 

 

 

CourimtinnofdistnhrledstbsbminplmeE
O
V b
op-a 8

 

 

/Fc;-

0: Ca; Ca
Concentration of distributed Wm
in phase R

 

 

Figure 3-8. Over-all and Individual Film Driving Forces

 

 

 

 

 

 

 

R; E,
X‘! Xe:
Raffimie

Figure 3-9. Extraction with Continuous Countercurrent Contact

C01

imn

incl

free

a cm
[OWE
trans

IrlCOI

48

Combining Equations 3-15 and 3-16 for the raffinate phase:

dN = d(RxR) = deSCRM (xR —xRi) 3-17

Use of this equation is limited to cases where the solvents are completely
immiscible or to dilute solutions since the diffusion upon which Equation 3-13 is based
includes only that of the solute and not the solvents.

The total raffinate R varies from one end of the tower to the other but the solute-

free raffinate, R (1 — xR) , remains constant. Consequently,

 

 

x Rdx
R J R 3-18

d(RxR) = R(1—xR)d[1__x

R

The mass transfer coefficient kR includes a term (1 — x which varies

R) N
throughout the tower. The quantity kR (1 — xR) m is more likely to be constant
(Treybal, 1951, p 243). Because it is difficult to quantify the interfacial area for mass
transfer, S cmz, it is convenient to define the interfacial area per unit of tower volume,
a cm2/cm3, so that d8 = aAdH where A is the tower cross-sectional area and H is the
tower height. The interfacial transfer area, a, is usually combined with the film-
transfer coefficient kR or kg and treated as one entity being either kRa or kEa.
Incorporating these terms in Equation 3-17 yields:

(l—x Rde

R)IM
(l-XR)

 

= kRa(1—xR)iMACRM (xR—xRi)dH 3-19

wher

1951
than;
Chang
integ'
thisi:
phase
llllj

migh

49

 

 

 

(l—xR)ideR kRa(l—xR)iMACRMdH
= 3-20
(1 —xR) (xR—xRi) R
(1—x.)—(1—x )
where (1 —xR). = R‘ R
M ln[(1—xRi)/(1—xR)]
A tower cross-sectional area, cm2

Since the terms (1 - xR) M and (1 — xR) are usually nearly unity (Treybal,
1951, p 243), the left-hand portion of Equation 3-20 is essentially the concentration
change de experienced per unit of concentration difference (xR — xRi) causing the
change and represents a measure of the difficulty of the extraction. When solved by
integrating over the change in concentration from the bottom of the tower to the top,
this is designated as the number of transfer units based on mass transfer in the raffinate
phase, NTUR. When the NTUR’s is multiplied by the experimentally determined factor

HTU, the height of a transfer unit, the result is the height of the tower, H:

 

 

 

le H
(1-xR). de
NTUR= j ‘M =j d" = H 3.21
(1 —xR) (xR —xRi) 0 HTUR HTUR
XR

In similar fashion, the concentration differences in terms of the extract phase

might have been used which would have resulted in:

 

 

 

H
NTUE = j ' = j = 3-22
0

50

Height of a Film Transfer Unit
Consideration of the right-hand side of Equation 3-20 and Equation 3—21 shows

that HTUR and the mass transfer coefficient are related:

R

HTU = 3-23
R

kRa(1—xR)iMACR

 

3V

(CRM1—CRM2)

where CR = 2 . The subscripts 1 and 2 refer to the

 

concentrations of all substances present at the raffinate (feed)
inlet and outlet respectively.

In similar fashion, the height of a transfer unit, HTUE, based on the concentration

difference between the bulk extract and the extract interfacial concentration is:

HTUE = E 3-24
kEa (1 -— XE) MACE

1 av

 

Number of Overall Film Transfer Units

In most practical situations, it is impossible to determine the interfacial
concentrations accurately. Therefore, true film driving forces, ACR and ACE
(Equation 3-14), can not be obtained. If the equilibrium-distribution curve is a straight
line so that at all concentrations encountered CEi is proportional to cm then:

= mC . 3-25

C RI

Ei

where m is the distribution coefficient (slope of the equilibrium line)

expressed in the same units as the concentrations.

6X11

Sim

If m.

manr

Where

and

51

Further, a concentration Cl3 can be defined as the concentration of solute in the
extract that would be in equilibrium with CR:

mC

3—26

u e * e e e e e e e e
Similarly CR rs defined as a concentratron in equilibrium wrth CE:
mCR

3-27

If the solute concentrations are expressed in mole fractions, then:

3-28

The complete mass-transfer process in both phases can now be represented in a
manner analogous to Equation 3-13 and Equation 3-16:

N = KEa(CE—C

E) = KEaACOE 3-29
where K5 is an over-all mass transfer coefficient based on an over-all
concentration gradient in the extract phase.
and

N = KRa(CR—CR) = KRaACOR

3-30

wher

anaf

Whe:

relaI

base

52

where KR is an over-all mass transfer coefficient based on an over-all

concentration gradient in the raffinate phase.
The number of film transfer units based on the over-all concentration gradients

are (Treybal, 1951):

' (l—xR) de
OM
NTUOR = 1 . 3-31
R (l—xR) (x—xR)

 

X

 

 

II dXR H
x (l-x;) HTUOR
RI(1-xR)1n ——

(1 —XR)

(1—x;) — (l—xR)

In[(I—x;)/(I—xR)]

 

where (1 —xR)OM

The height of a transfer unit based on the over-all concentration gradient is

related to the over—all mass transfer coefficient:

_ R
OR —
KRa(1—xR)OMCRwA

 

HTU 3-32

The equations for the number of transfer units and the height of a transfer unit

based on the over-all concentration gradient in the extract phase are:

I08

011

1m

53

 

_ 3-33
NTUOE - It:

 

 

 

HTUO - 3-34

In cases where the principal resistance to diffusion lies in the raffinate phase (m
large), Equation 3-31 is used for the extraction column design. In cases where the

principal resistance is in the extract phase (III small), Equation 3—33 is used for the

(1 —xR)OM
design. To evaluate NTUOR either the quantity or

*

(1-xR)(x-XR)

 

l

(I-x;)
(l—xR)ln[—(—1—_—x——]

R)

 

is calculated and plotted against xR. The area under the

resulting curve between the limits x

R and xRz rs the required value.

It should be kept in mind that, in general, the mass-transfer coefficients depend
on complex functions of diffusivity, viscosity, density, interfacial tension, and
turbulence. Similarly, the mass-transfer area of the droplets depends on complex
functions of viscosity, interfacial tension, density difference, extractor geometry,
agitation intensity, agitator design, flow rates, and interfacial rag deposits. Only

limited success has been achieved in correlating extractor performance with these basic

print

Sim]

aver

than

Sub;

subs

the

Sub

54

principles (Perry and Green, Eds., 1984).

Simplified Integration
If (1 — x;) and (1 — xR) differ by no more than a factor of two, an arithmetic
average rather than a logarithmic average for (1 - xR) OM incurs an error of no more

than 1.5% (Wiegand, p 679, 1940). Thus,

(1—x;) + (l—xR)

 

 

(1—xR)OM = 2 3-35
Substitution into Equation 3-31 leads to:
XR' d 1 x
X - R
NTUOR — j R, +% 1n 1 ’ 3-36
— x
XR (X — XR) RI

For dilute solutions, (1 — xR) and (1 — XE) are nearly unity and R and E are
substantially constant (Treybal, 1950). A material balance over the lower portion of

the tower in Figure 3-9 (p 3-48) then becomes approximately:

R(xR-—xR2) = E(xE—xE2) 3-37

*
Substitution of m xR for XE and rearrangement leads to:

3-38

This ma‘

(Treybal

lnasi

Extr

€Xll
fact

the

“'1
Im

C0

55

This may be substituted into the integral of Equation 3-36 and the integral evaluated

(Treybal, 1950, p 246):

 

 

 

 

 

 

xx] d i {x 3:»
X 1 R R
R,“ : In (1——) l m al.3— 3_39
(X _ ) R mE XE mE
xR R xR 1-—— x __2
2 mE _. \Rz m / .1
In a similar manner,
XE]
de 1 mE “3,—me2 mE
I—r——= mEln 1——E— —— +—R— 3-40
x (XE-XE) 1__ XE, me2
52 R

Extraction Factor

A predictive method for determining the number of stages required for an
extraction tower was developed by Kremser (1930). Kremser defined the extraction
factor; U, as the slope of the equilibrium line (distribution coefficient), m, divided by

the slope of the operating line, R/E:
U = m (E) 3-41

where E is the fresh solvent rate to the column and R is the rate of solvent flow in the
feed. Both E and R must be expressed in the same units used for expressing the solute

concentrations in the distribution coefficient.

The height of the extraction column, H, and the number of theoretical stages, n,

can '

and

Whr

ant
are
$01

inc

56

can be calculated using one of the Kremser equations, Equation 3-43 or Equation 3-44

(Coulson, 1968):

 

"‘0'x U“1 -U

H = __._._...._L = ._.._._l_..._._ 3-42
X _yn+l Un+ —l
0 m

 

and
1 1 1
Rel—I Ill-aha]
n = 3-43
logU

where x0 mole fraction of solute in the feed

xn mole fraction of solute in the raffinate leaving stage n

yy +1 mole fraction of solute in the extract

In order to apply the Kremser formulae, the equilibrium line must be straight
and its intercept zero, and the operating line must also be straight. These restrictions
are met when the two solvents selected for the process are mutually insoluble and the
solute concentrations are dilute enough so that the distribution coefficient is

independent of concentration, respectively.

sepa
cont
hnnl
he s
carb
wan:
spec
lon'
chox

phas

alet
to th

"L h

thec
[bed
as th
ofgC

COHCI

CHAPTER 4
BENZALDEHYDE DISTRIBUTION COEFFICIENT MEASUREMENT

In liquid-liquid extractions, at least one component in the original mixture to be
separated must have a higher selectivity in the extraction solvent than the other
components. The two solvents should be mutually immiscible 0r exhibit only a very
limited mutual solubility. However, the transferred component, or solute, which is to
be separated is soluble in both liquid phases. For the ternary system benzaldehyde/
carbon dioxide/water the solute, benzaldehyde, is soluble in both carbon dioxide and
water. If the solute has a different solubility in each of the solvents at the conditions
specified then a partitioning of the solute will occur. Since the two solvents exhibit a
low mutual solubility at temperatures and pressures near the critical point of carbon
dioxide, a two phase mixture will exist with a different concentration of solute in each
phase.

There exists a definite concentration relationship between the two liquid phases
at equilibrium. The ratio of the concentration of benzaldehyde in carbon dioxide XE
to the concentration of benzaldehyde in water xR is called the distribution coefficient,

m, for this temperature:

Solubility in a normal liquid increases with increasing temperature; however,
the change in solubility with temperature may not be the same in both solvents. Thus,
the distribution coefficient for a solute between two solvents may increase or decrease
as the temperature rises. The distribution coefficient also depends on the total amount
of solute present in a given amount of both solvents. As shown in Figure 4-1, the

concentration of solute in the extract phase can be plotted against the concentration of

57

58

 

 

 

Concentration of Solute in Extract

 

 

 

 

 

xR

Concentration of Solute in Raffinate

Figure 4-1. Equilibrium Relationship

solute in the raffinate phase. The resulting line is the equilibrium line and its slope is
represented by the distribution coefficient, m. At low solute concentrations the slope
of the equilibrium curve is essentially constant and, hence, the distribution coefficient

is constant.

Prediction of Distribution

In the discussion that follows, B is considered the extracting solvent for
removing a distributed substance C from A-C solutions. Provided that the same
standard state is chosen for each substance, the activities of the substance in each phase

at equilibrium are equal (Treybal, 1951):

a =a 4-2

59

where subscript CA refers to the activity of solute C in the C-A binary
subscript CB refers to the activity of solute C in the C-B binary
and so on.

From the definition of the activity coefficient:

a etc. 4-3

a a

CA = YCAXCA CB : YCBXCB AA = YAAXAA

where YCA is the activity coefficient of C in the C-A binary and Va; is the activity of
C in the C-B binary.

Substituting the appropriate expressions from Equation 4-3 into Equation 4-2
and rearranging provides an alternate definition for the distribution coefficient:

7
m=£i 4-4

YCB
This relationship between the distribution coefficient and the activity coefficients is
useful for predicting values of the distribution coefficient from literature values for
activity coefficients and vapor-liquid equilibrium data as well as from integrated forms
of the Gibbs-Duhem equation such as the Margules and van Laar equations.
Activity constants for binary systems can be estimated fairly accurately with the
van Laar equations (Perry, 5th ed., p 13-9):

A
ln 7] = ‘2 2 4-5

[ Arzxr]
1+
A21x2

 

 

A.’ w-—'_ A.‘--- *-

60

and

 

2 4-6

 

These two-suffix equations have been extended to cover the ternary systems such as

those found in liquid-liquid extraction (Wohl, 1946):
2 2

A A A A A
2 BA 2 CA BA CA AC
x A [—] +x A [—] +x x [—][—)[A +A —A [—H
B AB C AC B C AB AC CB
AAB AAC AAB AAC ACA

logtrA = 2 4-7
A B C
A AB A AC

The expression for log‘yB can be obtained by substituting subscripts B for A, C for B,

 

and A for C throughout the equation. Thus in the equation for logyB , xB in Equation

4-8 is changed to xc, A AB to A A AC to A etc. Similarly, in that for logyC ,

BC’ BA’

x in Equation 4-8 is changed to xB , ABA to A etc. Wohl has shown that these

C AC’

equations are limited to those cases where:

A A A
CI =[ CH IIJ 4_8
ABC AAC ABA

In the absence of published values for the van Laar constants, reasonable estimates for

 

them can be made using mutual solubility data following the procedure of Treybal
(1951, p 60).

Unfortunately, values for the van Laar or Margules constants for the binaries
benzaldehyde-water, benzaldehyde-carbon dioxide, and carbon dioxide-water could
not be found in the literature. Mutual solubility data was unavailable as well.

Therefore, it was undertaken to experimentally determine the value of the distribution

61

coefficient for benzaldehyde in the carbon dioxide-water system.

Equipment Design

The process equipment for the measurement of the benzaldehyde distribution
coefficient consists of three sections: the pumping or feed section, the equilibrating
section, and the sampling section. The equipment schematic for the entire process is
shown in Figure 4-2.

In the feed section the carbon dioxide and aqueous feed containing
benzaldehyde were metered into the phase separating cell. A 50 pound cylinder
equipped with a drop tube was used as the source of liquefied carbon dioxide. The
carbon dioxide was passed through a 1/4" FPT check valve to prevent contamination
of the cylinder and a 1/4" FPT, 1 micron filter to prevent contaminants from fouling the
downstream metering pump. The check valve and filter were supplied by Forberg
Scientific with part numbers 4F-C4L-1/3-SF and 4F-F4L-l-SS respectively. The
carbon dioxide was then passed through a 35' long piece of 1/8" stainless steel tubing
shaped into a coil approximately 3" in diameter. The coil was placed in an ice water
bath to lower the vapor pressure of the carbon dioxide to prevent cavitation at the pump
inlet. The 1/8" coil was connected as closely as possible to the suction side of one head
of a LCD Analytical duplex pump model 2396 having a capacity of 49 to 920 ml/hr.
The pump flow rate was controlled by a micrometer type adjustment on each head. For
the aqueous feed, a 1 liter glass flask with an outlet tangential to the bottom was placed
on top of a Mettler PM 4600 Deltrange weigh scale with a capacity of 4 kg. A short
piece of flexible plastic tubing connected the feed flask to a length of 1/8" stainless
steel tubing fastened to the second head of the duplex pump to allow accurate weighing
of the feed on a continuous basis. The outlets of both pump heads were connected to
1/ 16" stainless steel tubing joined together by a cross (not shown in Figure 4-2). The

third port of the cross (again, not shown in Figure 4-2) was connected to a High

62

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Extract LCV
(Heated) h
——> Vent
Extract
NJ Receiver _'__.'J
F— C02 Flow Meter
Vent
A
Raffinate LCV
(Heated) v Benzaldehyde
Solution
Ice Water Bath Feed Tank
I —»I><I——
" 0 0 O

_

Vent Q5 Weigh Scale

Raffrnate
Receiver
Agitator/
Heater
F
Phase F 1011
Separation Cell
'1
Sight Level £33m
g Gauge Bath
0 n—«vaw—
Temperature C02
j Equilibrating “El Cylmdef
'—’ C011 00 Separation Cell
Water Bath

 

 

 

 

 

 

Figure 4-2. Benzaldehyde Distribution Coefficient Measurement Schematic

63

Pressure Equipment Company rupture disk housing model 15-63AF1 containing an
OSECO, Inc. 1/4A, standard, 316 stainless steel frangible with a burst pressure of
3,000 psig at 72 °F.

The equilibrium section of the distribution coefficient measurement equipment
begins with the fourth port of the above cross connected to a short length of 1/16" SS
tubing. This leads to a 20' length of 1/8" SS tubing coiled to fit on the bottom of a 24"L
x 18"W x 16"H fiberglass water bath. The 1/8" coil served as an equilibrating chamber
for both temperature and mass transfer. Water temperature in the bath was maintained
by an Isotemp Immersion Circulator model 730 from Fisher Scientific Company. The
equilibrating coil is connected to a separating cell rated for 5,000 psig made by the
Michigan State University machine shop. The internal volume was measured to be 65
ml by taring the empty cell on a balance, filling it with room temperature water and
reweighing it to determine the weight of water added. The separating cell has a side
port for introduction of the two-phase feed mixture, a bottom port for discharge of the
aqueous phase and a top port for discharge of the carbon dioxide phase. In addition,
two 1/2" diameter sight glasses were part of the separating cell. One was located on
the side and was used to introduce light into the cell via a fiber optic cable that could
be immersed in water without harm. The other sight glass served as a sight level gauge
so that the interface between the two phases could be controlled at a point roughly in
the middle of the cell. The fiber optic light was a model 9745—00 from Cole-Partner
Instrument Company.

The sampling section consisted of a short length of 1/8" tubing running from the
bottom port of the separating cell, out of the water bath and terminating in a 1/8" SS
flow control valve series lOVRMM2812 manufactured by Autoclave Engineers. This
two-way angle valve has a micrometering valve stem with a Cv of 0.004 and an orifice
diameter of 0.062". The valve was heated to prevent freezing caused by the sudden

decrease in solubility and rapid expansion of the carbon dioxide dissolved in the water

64

resulting from the pressure drop across the valve. The heater was made by sandwiching
the valve body between two 3/ " thick aluminum blocks roughly cut to the same height
and width as the valve. The top of each block was drilled with a hole of the correct
diameter and depth to allow the insertion of a Vulcan Electric Company T. B. cartridge
heater. The two cartridges were electrically wired in parallel and one was inserted into
each block. The wires connecting the cartridges terminated in a standard 120V two-
prong plug inserted into a Staco Energy Products Company model 3PN1010 variable
transformer. The discharge of the raffinate (water phase) valve was connected to 1/8"
SS tubing inserted into the opening of a 500 ml Erlenmeyer flask which served as the
raffinate receiver. The inlet tubing extended through the top opening to about half the
depth of the flask. The top port of the separating cell was connected to a 1/8" SS tubing
tee. One of the two remaining ports of the tee was connected to a Heise model 901A
digital electronic pressure gauge with a range of 0-10,000 psi g and the other to the inlet
side of a 1/8" Autoclave Engineers micrometering valve identical to the raffinate valve.
It was fitted with a cartridge heater of the same design as described above. The
discharge of the heated extract valve (carbon dioxide phase) was connected to 1/8" SS
tubing inserted into a rubber stopper fitted into the opening of the extract receiver. The
extract receiver was a 20 mm OD. x 14" L glass tube with a 7 mm glass side arm vent
extending radially from the side and positioned 2" from the top. The 1/8" SS tubing
from the separating cell extended through the stopper to within approximately 1/2 " of
the bottom of the test tube. The test tube was filled with approximately 75 ml of
dehydrated, 200-proof ethanol to trap the benzaldehyde as it emerged from the end of
the drop line carried by the carbon dioxide. The ethanol trap was immersed in an ice
water bath to lower the temperature of the ethanol to minimize evaporation losses.
Plastic tubing connected the side arm vent to a Precision Scientific Wet Test Meter
model number 63126 to measure the flow of carbon dioxide. The wet test meter was

filled with Amoco 5-N F light weight mineral oil instead of water to reduce error caused

65

by the solubility of carbon dioxide in the water normally used in the meter.

Experimental Procedure

1. The aqueous feed mixture was prepared by mixing weighed quantities of
benzaldehyde and water in a 1 liter glass flask with a side arm outlet port located so as
to be flush with the bottom of the flask to allow complete drainage. The clean, dry flask
was tared on a Mettler PM 4600 Deltarange balance having a maximum capacity of 4
kg. To this was added approximately 500 ml of room temperature, reverse osmosis
purified water and the weight recorded. Approximately 1 gram of reagent grade
benzaldehyde from Aldrich was pipetted into the 1 liter flask and the weight recorded.
A magnetic stirrer was added, the flask stoppered and the mixture stirred for 12-18
hours at room temperature. All of the feed mixtures had concentrations of 2000 i 200
ppm benzaldehyde.

2. After the aqueous phase was finished mixing, the water bath circulating pump
and temperature controller were turned on and allowed to reach steady state. The
carbon dioxide and the extract trap ice baths were filled with ice and water and allowed
to cool. The discharge valve heaters were also turned on to a low setting and allowed
to heat up.

3. While waiting for the various temperature controlled devices to reach steady
state, the feed mixture was analyzed to verify the concentration of benzaldehyde
present. The gas chromatographic method is described fully in Appendix B. The
benzaldehyde concentrations varied markedly from the prepared concentrations. The
consistent decrease in concentration was thought to be due to the reaction of
benzaldehyde and oxygen to form benzoic acid. Two or three injections of the feed
mixture were made and the average of the analyses used for the actual feed
concentration.

4. The aqueous phase 1 liter flask was placed on the weigh scale, the weight

66

recorded and the plastic tubing between the flask and the metering pump was connected
taking care to eliminate any entrapped air.

5. The carbon dioxide flow meter reading was noted and recorded.

6. After checking the raffinate and extract discharge valves to be sure that both
were closed, the micrometer adjustments on both heads of the feed pump were adjusted
to give roughly equivalent flows of each phase and the carbon dioxide cylinder valve
was opened. It is important to note that the carbon dioxide phase began to rapidly fill
the separation cell as soon as the cylinder valve was opened since the pressure in the
cell was atmospheric pressure versus the 838 psig equilibrium vapor pressure of the
carbon dioxide in the room temperature cylinder. However, the small volume of the
separating cell reached equilibrium with the cylinder pressure very quickly.

7. The feed pump was started allowing both the aqueous and carbon dioxide
phases to flow into the separating cell.

8. The feed into the cell was allowed to continue uninterrupted until either the
pressure reached 1500 i 100 psig or the water level reached the midpoint of the sight
level glass. If the pressure reached the desired value first, the extract valve was opened
very slowly to control the pressure in the desired range. When the aqueous phase level
reached the middle of the sight glass, the raffinate valve was slowly opened to maintain
the level as closely as possible to that point. If the aqueous phase level reached the
middle of the sight glass first, the raffinate valve was opened and then when the
pressure reached the desired setpoint the extract valve was opened.

9. The flow of each phase was continued for approximately two hours. This
corresponded to 6-8 turnovers of the cell contents of each phase. After this number of
turnovers it was assumed that a steady state for mass transfer had been reached. Both
the extract and raffinate receivers were changed over to clean, empty ones and the
experiment continued as before.

10. After another 1.5-2 hours the experiment was terminated by shutting off the

67

feed metering pump and closing both the raffinate and extract valves as quickly as
possible. Samples of both the raffinate and extract were taken and prepared for
analysis.

11. The carbon dioxide cylinder valve was closed.

12. The raffinate valve was opened slightly and the remainder of the aqueous phase
allowed to collect in the raffinate receiver. As soon as all of the raffinate was drained
from the cell, the raffinate valve was closed carefully to avoid damaging the valve stem.

13. The extract valve was opened to allow the remainder of the carbon dioxide to
bubble through the ethanol trap.

14. The final weight of the aqueous feed flask was noted and recorded. A sample
of the final feed mixture was taken and prepared for analysis. The flask was
disconnected from the feed pump and the unused feed discarded.

15. Shutdown of the experiment was completed by emptying the ice baths, shutting

off the valve heaters, and shutting off the water bath circulating pump and heater.

Results and Discussion

A summary of the experimental results found in this investigation are shown in
Table 4-1 shown on the following pages. Runs 1-3 can be neglected because the
analytical method used to determine the concentration of benzaldehyde in water was
found to have an unacceptable level of error which resulted in unreliable mass balance
calculations. After run 3 the analytical techniques used to determine benzaldehyde in
both water and the ethanol solutions were changed to the internal standard methods
described in Appendix B. In addition, material balance calculations after run 3 used an
average value for the benzaldehyde concentration in the aqueous feed. This was done
because the benzaldehyde concentration in the aqueous feed decreased significantly
with time due to oxidation. To complicate things further, the low solubility of

benzaldehyde in water made it difficult to determine when a feed mixture had achieved

Table 4-1. Distribution Coefficient Measurement Data

68

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Run Run Time, Feed Conc., Benz. Water Distribution
Number minutes ppm Benz. Recovery, % Recovery,% Coefficientl

l 108 3170 69.8 - 15.0
2 108 3626 66.5 96.7 15.8
3 184 3650 62.6 101.8 12.6
4 165 1854 83.9 97.3 11.7
5* 150 1878 96.0 102.4 19.2
6 180 1908 96.3 100.3 11.9
7 185 1868 90.0 99.0 14.9
8 151 1947 90.0 102.0 7.2
9 154 1869 99.4 100.9 9.4
10 150 1917 91.0 96.2 12.2
11 150 1883 94.0 94.7 11.2
12 406 1837 97.6 102.8 16.7
13 242 1716 97.7 100.2 26.2
14* 250 1757 83.0 100.0 19.2
15 272 1932 95.9 100.7 21.5
16 240 1982 96.6 100.3 20.5
17 243 2149 92.3 101.9 19.0
18 191 2630 93.2 100.3 30.2
19 197 1911 94.1 100.0 23.9
20 226 1924 90.1 100.0 20.0
21 253 1952 91.3 99.6 21.6
22 255 2052 98.8 99.4 21.1
23 241 2019 93.1 99.3 21.3
24 242 2080 99.5 99.9 28.2
25 240 2025 96.0 97.6 28.3

 

 

 

 

 

 

 

 

1 Weight Distribution Coefficient

* Ethyl Acetate solute

 

 

69

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4-1. (Continued)
Run Distribution Temperature COZIWater Feed Rate Water
Number Coefficient2 °C gm/gm gm/min Turnovers

1 36.7 30 .712 3.606

2 38.3 " .678 3.770

3 30.9 " .974 1.831

4 28.7 " .477 3.322

5* 47.0 " .602 3.557

6 29.1 " 1.232 3.108

7 36.4 " 1.186 2.956

8 17.5 24 1.386 3.277

9 23.0 37 1.267 3.190

10 29.7 " 1.285 3.147

11 27.3 5 25 1.029 2.891

12 40.9 " .897 .952

13 64.0 " 1.082 2.834 5.79
14* 46.8 35 1.319 3.136 6.22
15 52.7 " .991 2.716 5.52
16 50.2 " 1.015 1.858 4.82
17 46.5 " .951 4.073 3.40
18 73.8 " 1.016 4.763 7.21
19 58.3 " .787 4.413 10.0
20 49.1 " .662 4.026 10.7
21 52.9 " 1.241 4.413 8.35
22 51.5 " 1.165 3.865 8.54
23 5 52.0 " .897 3.408 7.40
24 68.9 20 1.508 4.507 7.52
25 69.2 21 1.406 4.440 8.14

 

 

 

 

 

 

 

 

2 Molar Distribution Coefficient

* Ethyl Acetate solute

 

 

7O

equilibrium and the benzaldehyde was completely dissolved. In order to minimize
these two problems, a sample of the feed was taken at the beginning and end of each
run and the average of the two analysis used in the mass balance calculations and,
therefore, to determine the percentage of benzaldehyde recovery shown in Table 4-1.

Run 5 differed from the previous runs in a very important respect. The solute
was changed from benzaldehyde to ethyl acetate so that a calibration of the
experimental apparatus and method could be made. For the liquid carbon dioxide
extraction of ethyl acetate from water at room temperature and 918 psig Schultz (1970)
reported a distribution coefficient of 42. Although the conditions used for run 5 were
different at 30 °C and 2,000 psi the experimentally determined value of 47 agreed well
with Schultz's value.

Given the fairly good agreement of the experimentally determined distribution
coefficient with the literature value, the experiments to determine the benzaldehyde
distribution coefficient were continued as before. Runs 6 and 7 were conducted under
identical conditions but produced a distribution coefficients of 29.1 and 36.4
respectively. Although this was not unusually bad agreement, it was decided that the
two values should be reproducible to at least within ten percent and preferably within
five percent. Experiments 9 and 10 were run at identical conditions as were
experiments 11 and 12. The only difference between these two pairs of experiments
was that they were run at 37 °C and 25 °C respectively. Again, the deviation between
the distribution coefficients for each of these pairs was found to be greater than the
acceptable upper limit of ten percent.

After reviewing the experimental procedure it was concluded that the run times
were too short to allow equilibrium to be reached. The method was modified to change
both the raffinate and extract receivers after an amount equal to 7-10 cell volumes for
each phase had been pumped through the separating cell. It was felt that after this

approximate volume of material had been pumped through the cell that the mass

71

transfer would be at equilibrium and, perhaps more importantly, the cell contents of
each phase would be of a uniform composition equal to the equilibrium concentration.

Run 14 used the modified method but again used ethyl acetate as the solute. The
distribution coefficient was determined to be 47, again showing the good agreement
with the literature value of 42 as before.

Runs 15-18 were conducted at the same conditions except for the total feed rate
of the combined aqueous feed and carbon dioxide. The combined flows were varied to
determine if there was a corresponding change in the distribution coefficient. It was felt
that if the system was at equilibrium an increase or decrease in total flow rate to the
separating cell should not have resulted in a change in the distribution coefficient.
However, changing the total flow into the cell did result in an unacceptable change in
the distribution coefficient.

Prior to run 21 it was decided that the lack of mixing in the separating cell was
still the problem causing the lack of a reproducible distribution coefficient. A small
agitator was finally built that was not only small enough to fit in the separating cell but
that would not get "hung up" and stop spinning. A 1/32" hole was drilled through the
teflon coating of a small magnetic stirring bar. A piece of wire in the shape of a "t" was
inserted into the hole in the stirring bar so the top of the tee would be near the top of
the separating cell and would contact only the top (carbon dioxide) phase. A water
proof magnetic stirrer was immersed in the water bath and the separating cell placed
on top of it. The speed of rotation was kept at approximately 40-60 rpm. At these low
speeds the stirring bar itself caused agitation in the bottom (aqueous) phase and the
cross bar of the t-shaped wire caused agitation in the top phase with only a slight upset
of the interface between the two phases. Thus, each phase of the cell was now well
mixed and the two phases separated cleanly.

Runs 21-23 were run at identical conditions and were the first series of three

consecutive runs to produce essentially the same value for the distribution coefficient

72

at 52.9, 51.5, and 52.0 respectively. The maximum deviation between them is 2.7%.

The temperature for runs 24 and 25 was decreased from 35 °C to 20 °C (room
temperature) and the same conditions as for runs 21-23 were repeated. The runs gave
virtually identical values of 68.9 and 69.2 respectively.

The work to determine the distribution coefficient was terminated at the end of
run 25. It was decided that the value of the distribution coefficient for benzaldehyde
in the benzaldehyde/carbon dioxide/water ternary was 52 at 35 °C and 69 at 20 °C. For
subsequent work using these values of the distribution coefficient it was assumed that
these distribution coefficients would remain constant in the benzaldehyde

concentration range of 0-2,000 ppm.

CHAPTER 5

BEN ZALDEHYDE EXTRACTION EQUIPMENT DESIGN

The benzaldehyde extraction equipment description is divided into three
segments; the pumping or feed section, the extraction column, and the receiver section.
Schematic drawings for each section are shown in Figures 5-1, 5-2, and 5-3

respectively. A flow schematic for the entire process is shown in Figure 5-4.

Feed Section (Figure 5-1)

In this section the aqueous feed and the carbon dioxide solvent are introduced
into the system. A 20 liter Nalgene tank served as the aqueous feed vessel. It was
equipped with a Lightnin Mixing Company Labmaster TSM 2510 agitator for mixing
the benzaldehyde/water feed solution. The bottom outlet of the feed tank was
connected to the inlets of both heads of a LDC Analytical model 2396 metering pump.
The LDC 2396 is capable of delivering 49-920 ml/hr against a maximum discharge
pressure of 6,000 psig. Separate micrometer-type adjusters control the delivery of each
head independently. The discharge nozzles of each head were joined together by a 1/
16" tubing cross. The third port of the cross was connected to a High Pressure
Equipment Company model 15-63AF1 rupture disk housing. An OSECO, Inc. 1/4A,
standard, 316 SS rupture disk with a burst pressure rating of 3,000 psig at 72 °F was
installed in the rupture disk housing to protect the apparatus from exceeding its
maximum pressure rating. The fourth port was connected to the top feed port isolating
valve on the extraction column.

Carbon dioxide was supplied to the system from a 50 pound cylinder equipped
with a siphon tube. From the cylinder valve the carbon dioxide flowed through a check
valve (part number 4F-C4L-1/3-SS) to prevent backflow into the cylinder and a 1

micron filter (part number 4F-F4L-1-SS) to remove small particulates from the line to

73

74

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Lighnin’ .
Agitator 3,000 psrg
Vent
[ I
To Extraction
i [X] . Column
Aqueous Feed
Distributor
cs
20 Liter Benzaldehyde
Aqueous Feed Solution
Tank Feed Pump
3.000 psig
Vent
GP . ._, I.
IN ' F ‘ fi'Xl—> Extraction
1 p81 1” Column
Haskel
Gas Booster
X 3,000
psig
50 lb. vent
C02 f
Cylinder :I
' ‘3
75 psig Air Supply N p 4 1 Liter, 1500 mg
10011 C02 Reservoir

Figure 5-1. Benzaldehyde Extraction Column Feed Section Schematic

 

 

 

 

 

 

 

 

 

 

_l_.l_
1 1:11:11
6”
l a 9”
j/ ‘ 1
Feed Distributor Nozzle

1/4” FNPT
1/ ” FNPT

Pro-Pak 0.16”, 316 SS
Protruded Metal Packing Z

Custom Fabricated ’

 

  

 

 

 

 

 

 

 

 

72”
Liquid Redistrubutor Ring
3/4” I.D.x 15/16”O.D. x 3/32” Dia.
O-ring
1/4” FNPT
1/4” FNPT /
9”
6”
l_m_l1
‘L— l/ ” Taper Seal

Figure 5-2. Extraction Column Detail

 

 

76

 

 

 

 

Extract C02 Flow
(9 (D O
¢ +Vent
E ' '
F
50 Heated
” BPR p [
3,000 psig i— r

Ice Bath

 

 

 

 

 

 

 

 

 

 

U\ U

\— 500 ml Ethanol Trap

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sight
U Level Heated Raffinate C02 Flow
Gauge LCV
T {X} {E i p G G C) ——> Vent
lOu \ l I
1/2”W Silicone-Coated
Electric Tracing 4 Liter
Extract
Receiver

Figure 5-3. Extraction Column Receiver Section Detail

77

 

 

 

 

 

M 75 psig Air Supply

 

 

 

 

 

 

 

Benzaldehyde Receiver
Solution
Feed Pump Vent
? Mr W
Extract C02 Flow
0) C) CD

   

 

 

 

 

 

 

 

 

 

From Feed Tank Vent 4_ ¢ ‘_.
Extraction
Column
C02 Pump Raffinate C02 Flow
0) O O

 

 

Vent <— ¢ <—

CO Feed Pump
111 F 2 + > LI——r

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_I1I
Sight
U Level
Gauge
Heated LCV
X F
1011
q; [Si—>Vent
C02
Cylinder . R affi nate
1500 P513 . Receiver
C02 Reservorr

 

 

 

 

 

 

 

 

 

Figure 5-4. Benzaldehyde Extraction Column Flow Schematic

78

protect the pumps downstream. Both the check valve and the filter were supplied by
Forberg Scientific, Inc. The downstream side of the filter was connected by 1/4" SS
tubing to the suction side of a Haskel, Inc. Gas Booster model AG- 152. This pump was
used to maintain a reservoir of carbon dioxide at between 1,300-1,500 psig.

The reservoir was a 1 liter, 316 SS vessel, model number OC-ll manufactured
by High Pressure Equipment Company. The reservoir supplied liquid carbon dioxide
to the suction side of an LDC Analytical model 2396 metering pump at a sufficient
pressure to prevent cavitation and insure accurate, reproducible delivery of carbon
dioxide into the system. The inlet and outlet ports of the reservoir were connected by
short lengths of 1/ " SS tubing to 1/ " block valves manufactured by High Pressure
Equipment Company (model 10-1 1AF4). A pressure gauge was connected between the
outlet of the reservoir and the block valve to monitor the pressure. In a configuration
similar to that of the aqueous phase feed pump discharge, the discharge nozzles of the
carbon dioxide metering pump were connected to two ports of a 1/16" cross. The third
port of the cross was connected to a rupture disk assembly identical to the one
protecting the aqueous phase metering pump. The fourth port of the cross was

connected to the isolating valve on the lower feed port of the extraction column.

Extraction Column (Figure 5-2)

A model TOC-72, 316 SS Tubular Series Reactor with standard AF4 taper seal
connections at each end was purchased from High Pressure Equipment Company and
modified to serve as an extraction column by adding four 1/4" FPT connections. Two
of the additional ports are located on the same side of the column 6" from each end and
are used for the top and bottom solvent feeds. The 6" measurement is taken from the
end of the O-ring plug that fits into the end of the column and not the end of the column
itself or the end of the end caps. In this way, the 6" dimension is the actual height of

the separating section for the two phases. The remaining two ports are on the side of

79

the column directly opposite of the feed ports and are 6" from each end. These ports
were used to connect a sight level gauge at either the top or bottom of the column to
monitor the interface between the two phases in the column. In other words, the
additional ports added to the column are located in a symmetrical fashion about the
axial middle of the column so as to allow the column to be run with either phase as the
continuous phase. The column is 72"L with a 1" ID. and a 2" OD. The end
connections with the 1/4" female (AF4) taper seal fittings use Buna N O-rings (part
number B-113) for a pressure seal.

The external configuration of the column included a Jergusen 12-R-32 reflex
sight level gauge with a maximum pressure rating of 4,000 psig at 72 °F. The height
of the view port of this cell is approximately 5". The sight gauge was mounted at the
bottom of the column to measure the interface between the continuous (carbon dioxide)
phase and the dispersed (water) phase. The gauge was mounted so that the bottom of
the view port aligned with the bottom of the column so as to monitor the interface level
in the bottom 5" of the column.

Both of the column solvent feed ports and the top and bottom outlet ports were
connected by stainless steel tubing to High Pressure Equipment Company block valves
which could be used to isolate the contents of the column and prevent the loss of the
liquid carbon dioxide between successive runs. The top solvent feed port used for the
aqueous feed and both of the outlet valves were 1/ " SS model 10-11AF4 valves. The
bottom solvent feed port used for the carbon dioxide feed was a 1/8" SS model 15-
11AF2 valve. The remaining top port was connected by 1/16" SS tubing to a Heise
digital pressure indicator model 901A with a range of 0-10,000 psig and to a High
Pressure Equipment Company 1/ " rupture disk holder model 15-61AF1 containing an
OSECO 1/4A, standard rupture disk with a burst pressure of 3,000 psig at 72 °F.

Three items made up the extraction column internals; the column packing, the

liquid redistributors, and the aqueous feed distributor. The packing was 0.16"L, 316

80

SS Pro-Pak protruded metal packing purchased from Scientific Development
Company. The packing was added to the column dry by slowly pouring the packing
through a funnel inserted into the top of the column with the column in the vertical
position. Four liquid redistributing rings were installed in the column spaced 16" apart
and beginning 16" from the bottom of the column. The design of the redistributing
rings is shown in Figure 5-5. The redistributors were made from aluminum while the
O-ring seals were Buna N rubber. The aqueous feed distributor was a 4" long piece of
1/ " SS tubing plugged at one end by a TIG weld. A 1/8" diameter hole was drilled
into one side of the tube centered exactly 1/2" from the plugged end to serve as the
spray nozzle. The liquid redistributing rings and the feed distributor were fabricated
in the Michigan State University engineering machine shops. The distributor tube was
installed in the column by inserting it through a Swagelock 1/4" tube x 1/ " MPI‘

adapter fitting that had been bored out so that the feed distributor would slide

 

 

 

 

1 PHI
; 5\5 ET},
A»

Figure 5-5. Liquid Redistributing Rings.

81

completely through it. After the modified fitting had been tightened into the top feed
port, the distributor tube was marked at the end that would remain outside the column
so that the discharge nozzle location could be verified after installation. Care was
taken to be sure that the discharge nozzle pointed straight downward and that the
distributor tube was pushed through the modified adapter fitting until the plugged end
came to rest against the far wall of the extraction column. By doing this, the precise
distance of the nozzle from the plugged end of the tube insured that the nozzle was in

the exact center of the column.

Receiver Section (Figure 5-3)

The carbon dioxide phase flowed out the top of the extraction column through
the isolating valve and then through a 50 micron model 4Z-F4L-50-SS filter purchased
from Forberg Scientific. The filter discharge was connected by l/ " SS tubing to a
Tescom Corporation model 26-1724-24 back pressure regulator to control the
operating pressure in the extraction column. Because of the cooling caused by the
Joule-Thomson expansion of the carbon dioxide across the back pressure regulator, an
aluminum-block heater was designed to heat the regulator. Two holes sized to fit a
Vulcan Electric Company T.B. cartridge heater (approximately 1/4" OD. x 2"L) were
drilled into the side of a 3"W x 3-3/4"L x 3/4"D aluminum block to a depth sufficient
to allow the cartridges to center on the centerline of the block. Fiberglass insulation
was packed into the holes behind the cartridges to hold them in place. The cartridges
were electrically wired in parallel to a standard 120V, two-prong plug. The heaters
were plugged into a Staco Energy Products variable autotransformer type 3PN1010 to
regulate the temperature of the aluminum block. The back pressure regulator was
placed on top of the aluminum block and covered with fiberglass insulation. An Omega
type K thermocouple connected to an Omega CN9000A temperature indicating

controller was placed underneath the aluminum block heater to monitor temperature.

82

The flow of carbon dioxide vapor discharged from the back pressure regulator was split
into two streams by a 1/4" plastic "Y". Both branches of the "Y" went to identical
ethanol traps to collect the benzaldehyde extract. The ethanol traps were test tubes
14"L with an outer diameter of 2-1/8". A 7 mm O.D. side arm was located 2" from the
top to provide a vent for the carbon dioxide bubbling through the trap. An ACE Glass
Company gas sparger was extended to 18" in length and inserted into a hole bored
through a number 101/2 rubber stopper. The stopper was pressed into the end of the test
tube to provide a leak free seal. The glass frit at the end of the sparger was adjusted to
be 1/2-1" from the bottom of the test tube. Approximately 500 ml of dry ethanol was
poured into the test tube to trap the benzaldehyde. The entire trap was then immersed
in an ice water bath to minimize evaporation of the ethanol during the run. The vents
from the two traps were connected together and the combined flow was then directed
into 3 Precision Scientific Wet Test Flow meter catalog number 63126. The Wet Test
Meter was vented to the atmosphere.

The bottom isolating valve of the column was connected by 1/4" SS tubing to a
10 micron Parker 4F-F4L-10-SS filter. The raffinate then flowed through 1/8" SS
tubing to a control valve to control the interface level between the two phases in the
bottom of the column. The flow control valve was a series lOVRMM2812
manufactured by Autoclave Engineers. This two-way angle valve has a micrometering
valve stem with a Cv of 0.004 and an orifice diameter of 0.062". Because of cooling
caused by the expansion of carbon dioxide dissolved in the raffinate, an aluminum
block heater was used to heat the valve to prevent freezing. The heater consisted of two
3/4" thick aluminum blocks slightly larger than the valve body. Two holes sized to fit
Vulcan T.B. cartridge heaters were drilled into the top of the blocks. Mounting holes
were drilled into the sides of both blocks at one top corner and at the bottom corner
diagonally opposite to the top hole. The blocks were placed on either side of the flow

control valve with threaded rod inserted through the corner holes and then secured with

83

hex nuts. As before, the cartridge heaters were wired in parallel and plugged into a
Staco Energy Products model 3PN1010 variable autotransforrner. The flow control
valve discharge was connected by 1/4" SS tubing inserted through a hole bored into a
rubber stopper. The stopper was forced into the top of a 4 L Erlynmeyer flask which
served as the raffinate receiver. The raffinate tubing extended about half way into the
flask to allow separation of the raffinate and carbon dioxide. The Erlenmeyer flask was
modified by adding a glass stopcock to the bottom so that the entire contents of the
flask could be drained without holdup. Carbon dioxide was vented out the top of the
flask through a vent connected by Tygon tubing to a wet test meter identical to the one

described above. The meter was vented to the atmosphere.

CHAPTER 6
BENZALDEHYDE EXTRACTION EXPERIMENTAL METHOD

Prior to the first run of the extraction column an estimate of the carbon dioxide
feed rate required to extract benzaldehyde from a dilute aqueous solution was made.
This estimate was used as the initial carbon dioxide feed flow to the column. The
approach used was to assume that there were either one or two equilibrium stages in
the column and perform a material balance around each stage. The benzaldehyde
concentration in the feed was taken as 200 ppm and it was desired to recover 50% of
the benzaldehyde from the feed. This would result in a decrease in the benzaldehyde
concentration from 200 ppm to 100 ppm in the raffinate. A schematic diagram of the
single stage model is shown in Figure 6-1. The material balance for the single
equilibrium stage is:

IN IN _ our our
onw +ychoz _ xIQw +yIQcoz 6'1

In order to simplify the material balance, it was assumed that carbon dioxide and water
are mutually insoluble. This is a reasonable assumption considering that Wiebe (1977)

determined that at 25 °C and 125 atmospheres of pressure the equilibrium solubility of

 

 

 

 

 

 

 

QOUT QIN
Y] = ? ppm 4 C02 H C02 yo = 0 ppm
One
IN g Equilibrium
Qw = ZOO—.— Stage Qour
_ mm W x = 100 m
x0 — 200 ppm + :D 1 PP

 

Figure 6-1. One Equilibrium Stage Flow Simulation

84

85

carbon dioxide in water is close to 6.3% by weight while the solubility of water in
carbon dioxide is very low at only 0.14% by weight. This mutual insolubility combined

with the dilute concentrations of benzaldehyde mean that the flow of both phases is

constant:
IN __ our IN _ our
Qw ‘ Qw and Qc02 " Qcoz 6‘2

Substituting into Equation 6-1, noting that the concentration of benzaldehyde in the

fresh solvent stream, yo, is zero, and rearranging yields:

(xO—XI)QW = yiQCOZ 6'3

Since there are two unknowns, y] and QC02’ remaining in the above expression another
relationship must be found before Equation 6-3 can be solved. The equilibrium
distribution coefficient for the ternary system benzaldehyde/carbon dioxide/water
relates the concentration of benzaldehyde in the raffinate, X], to the concentration of

benzaldehyde in the extract, yl. The distribution coefficient reported previously was

 

52. Therefore, y1 = 52xl and equation 6-3 becomes
(X0 '— X 1) Qw Q 6 4
52X 1 — C02 -

Solving this equation for the flow rate of carbon dioxide given the values shown in
Figure 6-1 yields a value of 3.8 g C02/min.

The same technique can be used to estimate an initial carbon dioxide flow rate
assuming that there are two equilibrium stages in the column as shown in Figure 6-2.
Again making the assumption that the two phases are completely immiscible and the

solute concentration is dilute enough so that the flow of each phase is constant, an

86

 

 

 

 

 

 

 

 

 

 

 

 

Qcoz = 7
Y Y0=0
‘ 2 Y1 _ ppm
Equilibrium Equilibrium
Stage 1 Stage 2
= 2 X =
x0 00ppm> x] _ . 2 100ppmg >
= 200——-
Qw min
Figure 6-2. Two Equilibrium Stage Flow Simulation
overall material balance is given by:
y2QC02 : (XO-XZ)QW 6-5

The distribution coefficient provides a relationship between the streams leaving the

second equilibrium stage:

Y; = mxl 6-6
Substituting Equation 6-6 into Equation 6-5 yields:

xIchoz = (XI-x2)Qw 6'7
A material balance around the first equilibrium stage is given by:

(x1 'x2)Qw = leCOZ 6'8

Again, the distribution coefficient can be used to eliminate an unknown variable:

87

y 1 : mxg 6-9
Substituting this expression into Equation 6-8 gives:

(XI—x2)QW z x2ch02 6'10

Solving for X]:

Q(:02 m x 2
1 = — + x2
Qw
Finally, Equation 6-11 can be substituted into Equation 6-7 for x1 yielding an equation

with only one unknown variable:

Qcozmxz
QCO2m|:——+x2 = (xo—x2)QW 6-12
Qw
Rearranging:
2 2
Q(:02m x2
————+QCO2mx2— (xo—x2)QW = 0 6-13

Qw

Substituting the known values from Figure 6-2 for the appropriate variables, Qw = 200
g/min, x0 = 200 ppm and x2 = 100 ppm, and recalling that the distribution
coefficient, m, is 52 and solving the quadratic expression yields a value for the carbon
dioxide flow rate, QC02’ of 2.4 g COZ/min.

The estimated initial carbon dioxide flow rate of 3.8 g COzlmin. or 2.4 g COZI
min. using the assumption of one or two equilibrium stages respectively were found to
be within the C02 pump capacity of 0-15 g COzlmin. at 1,500 psig.

The next step in preparing to run the extraction column was to examine the

88

estimated flow rates of both the feed (200 g/min) and the solvent (2.4-3.8 g/min) with
the flooding correlations of Crawford and Wilke, Rao and Rao, and Venkataraman and
Laddha to insure that these conditions would not flood the column. A summary of the
calculations is shown in Table 6-1.

It is apparent from Table 6-1 that there is disagreement between the three
correlations as to what total flow rate results in column flooding. For design purposes,
the authors recommend that 50% of the calculated flooding value should not be
exceeded. Since the packing used in the benzaldehyde project was different from any
of those used in the published flooding correlations this conservative approach seemed
appropriate. Accordingly, the initial values for the carbon dioxide and aqueous feed
flow rates were chosen to be in the range of 4.0-5.0 g/min. and 10-15 g/min.
respectively.

Selecting the carbon dioxide flow rate in the 4.0-5.0 g/min. range corresponds
to a continuous phase retention time of 2.9-2.3 hours. Since it was assumed that a
minimum of three complete turnovers of the carbon dioxide were necessary to achieve
steady-state operation this range of retention times corresponds to a minimum

operating time of 8.6-6.9 hours.

Preparation

1. Between 10-15 liters of aqueous feed solution were prepared prior to starting a
run of the extraction column. The benzaldehyde and water were allowed to mix with
gentle stirring overnight to insure that the benzaldehyde was completely dissolved.

2. About 1-2 hours prior to each run, the gas chromatograph used to analyze for
benzaldehyde in ethanol was turned on and allowed to equilibrate.

3. The ultraviolet-visible light spectrophotometer used for determining the
concentration of benzaldehyde in the feed and raffinate streams was turned on and

allowed to equilibrate.

89

Table 6-1. Flooding Correlation For Prestart-up Conditions

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Continuous Discontinuous C02 % Flooding
Phase Phase Retention
Flow Rate Flow Rate Time C &W R &R V &L
g/mrn g/mrn hr
2.4 200 4.77 70 91 127
" 150 " 54 80 1 13
" 100 " 38 67 95
" 75 " 30 60 84
" 50 " 21 51 71
" 25 " 12 39 55
" 10 " 7 29 40
3.8 200 3 .02 74 93 126
" 150 " 58 83 1 12
" 100 " 41 70 95
" 75 " 33 62 84
I. 50 " 24 54 72
" 25 " 14 42 56
" 10 " 8 31 42

 

 

 

 

 

 

 

 

C&W = Crawford and Wilke (1951)
R&R = Rao and Rao (1958)
V&L = Venkataraman and Laddha ( 1960)

The following values for the correlation parameters were used:

pC = 62.4 lb/ft3 e = 094
p1D = 53.781b/ft3 F = 693

(1) = 1.0 in y = 21.6 dyne/cm
“C = 0.068 cps RST = 0.3

 

 

90

Table 6-1. Continued

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Continuous Discontinuous C02 % Flooding
Phase Phase Retention
Flow Rate Flow Rate Time C &W R &R V &L
g/mrn g/mrn hr
7.5 200 1.53 82 99 125
" 150 " 65 88 l 12
" 100 " 47 76 96
" 75 " 38 68 86
" 50 " 28 59 75
" 25 " 18 47 60
" 10 " l 1 37 47
15.0 200 0.76 87 107 136
" 150 " 68 95 121
" 100 " 5 6 84 98
" 75 " 46 76 90
" 50 " 35 67 79
" 25 " 24 55 65
" 10 " 15 45 53

 

 

 

 

 

 

 

 

C&W = Crawford and Wilke (1951)
R&R = Rao and Rao (1958)
V&L = Venkataraman and Laddha (1960)

 

 

91

4. The variable power controller used to supply electric current to the heating
cartridges in the extraction column backpressure regulator heating block was turned on
at a setting of approximately 55% and allowed to equilibrate for about an hour prior to
running the extraction column. The thermocouple measuring the bottom surface of the
backpressure regulator was not allowed to exceed 120°C to prevent damage to the
regulator.

5. The variable power controller used to supply power to the silicone-coated
electric tracing on the metal tubing between the raffinate level control valve and the
raffinate receiver was turned on an hour before running the column to preheat.

6. The variable power controller used to supply power to the electric heating
cartridges on the water phase level control valve was turned on and allowed to
equilibrate for an hour prior to column start-up. Preheating this valve was found to be
very important in preventing freezing of the raffinate line early during the run.

7. The two ethanol traps were rinsed clean, filled with fresh ethanol, and installed
on the discharge of the continuous phase backpressure regulator in series.

8. The current readings of the carbon dioxide flow meters were recorded.

Operating Procedure

Refer to Figure 6-3 in reference to the following discussion. The valve numbers
described below are shown in Figure 6-3 as circled numbers next to the appropriate
valve.

1. The carbon dioxide cylinder pressure was checked to be sure that there was
sufficient carbon dioxide for the run. The final carbon dioxide inlet block valve into
the column was checked to be sure that it was closed. The cylinder valve was opened
to supply liquid carbon dioxide through the cylinder dip-tube into the Haskel gas
booster pump.

2. The blockvalve (1) supplying air to the Haskel gas booster was Opened. The air

92

 

75 psig Air Supply

 

 

Benzaldehyde
Solution

 

Feed Pump I: 1 W Vent

Extract C02 Flow

   

 

 

From Feed Tank

  

 

Haskel
Gas
‘Booster

 

 

 

 

 

 

luF

 

 

 

 

 

 

 

 

 

Extraction
Column

C02 Feed
7 Pump

 

 

 

 

 

Cylinder

1500 psig
C02 Reservoir

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Extract
Receiver

 

Vent ‘—

 

@030
l ‘—

 

 

Raffinate C02 Flow

 

Vent <—

©®®

 

 

 

 

 

 

1

Heated LCV

 

 

 

 

 

Raffinate
Receiver

 

 

Figure 6-3. Benzaldehyde Extraction Column Flow Schematic

93

supply pressure was 75 psig.

3. The blockvalve (2) between the discharge of the Haskel gas booster and the high
pressure carbon dioxide reservoir was opened. The reservoir pressure was maintained
at 1,500 psig to provide a continuous, uninterrupted flow of liquid carbon dioxide into
the metering pump.

4. The raffinate phase level in the sight level gauge was checked to determine if
there was a level visible. The level was maintained at 1/2 to 2/3 the height of the glass.

5. The aqueous phase feed tank was connected to the suction side of the feed
metering pump. The manual valves between the feed pump and the suction of the feed
metering pump were opened. The micrometering adjusters on the pump were set to the
desired percentage of maximum flow.

6. The blockvalve (3) between the aqueous feed tank and the column was opened.
The internal check valves in the pump heads were relied upon to prevent backflow from
the column into the pump.

7. The pressure in the carbon dioxide reservoir was monitored. When it reached
1,500 psig the blockvalve (4) between the reservoir and the carbon dioxide feed
metering pump was opened.

8. The blockvalve (5) supplying carbon dioxide to the column was opened. The
reservoir has sufficient capacity to fill the column and will do so when the reservoir
pressure is maintained at 1,500 psig.

9. The discharge blockvalve (6) on the top of the column was opened slowly.
Opening this valve to quickly with the column pressure regulator not set pr0perly will
result in a sudden discharge of carbon dioxide through the regulator. This sudden
depressurization of the column could be sufficient to blow the ethanol out of the cold
traps into the wet test flow meters.

10. The raffinate level control valve was checked to be sure that it was closed. The

blockvalve between the bottom of the column and the heated level control valve was

94

opened.

11. The micrometering adjustments on the carbon dioxide pump were set at the
desired settings. The pump was turned on.

12. The aqueous phase feed pump was turned on.

13. The raffinate level in the sight glass was monitored. The level control valve
was adjusted to maintain the level in the sight gauge.

14. The column pressure was monitored and small adjustments were made to the
backpressure regulator to maintain the column pressure at the desired setpoint. The
regulator is oversized by a small amount so that when it opened the column pressure
decreased by 50-80 psig. The rate of carbon dioxide flow into the column was slow
enough so that the pressure took about 10-15 minutes to recover enough for the

regulator to open again.

Chemical Analysis

In order to be able to perform a material balance around the column it was
necessary to analyze three of the column steams for benzaldehyde; 1) the aqueous feed,
2) the raffinate, and 3) the extract. The solvent feed was cylinder carbon dioxide and
did not contain any benzaldehyde. The analytical methods used for these analysis are
described below:

1. The aqueous feed was analyzed for benzaldehyde using Ultraviolet
Spectrophotometry (u.v.). Benzaldehyde exhibits a strong absorption in the ultraviolet
region of the electromagnetic spectrum at 241 nanometers. Concentrations in the range
of 1-10 ppm could be measured. Calibration curves and other details of the u.v. method
are presented in Appendix D.

For the assay of the column aqueous feed, serial dilutions of ten volumes of
reverse-osmosis purified water to one volume of sample were done to dilute the

benzaldehyde to give an on-scale reading on the u.v. spectrophotometer. Since the feed

95

solution was prepared to give benzaldehyde concentrations in the range of 100-700
ppm only one dilution was required to decease the concentration to a level which would
yield an absorption readable on the u.v.

The feed tank was sampled three or four times during the course of a run to
monitor the decrease in the benzaldehyde concentration due mostly to oxidation to
benzoic acid and also partly due to evaporation. The benzaldehyde concentration did
decrease by as much as much as 10% during the runs.

2. The raffinate streams were also analyzed for benzaldehyde by ultraviolet
spectrophotometry. Dilutions of 1:1 or 2:1 were used.

3. Atmospheric pressure on the downstream side of the backpressure regulator
controlling the column pressure resulted in vaporization of the carbon dioxide. This
gas stream was bubbled through ethanol to trap the entrained benzaldehyde. In order
to be able to perform a material balance around the column, the ethanol trap contents
were analyzed to determine the amount of benzaldehyde dissolved therein. Gas
chromatography using a Perkin-Elmer model 8500 equipped with a flame ionization
detector was used for the gas chromatography. The column was an Alltech Associates,
Inc. Econo-Cap capillary column, catalog number 19653, 10 m long x 0.53 mm (b, with
a 1.2 um thick carbowax stationary phase.

Isopentyl alcohol (IPA) was used as an internal standard. Because the
benzaldehyde concentration was less than one percent, the ethanol peak coming off the
column had a long shoulder extending to more than five minutes although the baseline
was quite close to being re-established at this time. The IPA peak eluted at very close
to five minutes. In the sample Chromatogram shown in Figure 6-4, the IPA peak eluted
at 4.96 minutes. The benzaldehyde peak appears at 7.6 minutes. Further details about

the ethanol phase chromatography are described in Appendix B.

Material Balance Calculations

Material balance calculations were performed for each of the extraction column

96

om.m

'0
Us
‘1'

on.“

howhmm hxxcwmc A mx¢wm m

I

mmo_.m : «mm_.m om.“
44mm.x : cum..n cm.v
uxum.msnu a «sum.asmu on..
x «was on. :mmq km
s "onecunoJao ammunnoz m noxemz
moxvoxmm swam v 23m

 

nzmmM
l"11 H m

m m :

 

 

 

3302325 Ema $2
s: 8% 3.:

zcm gt a" u m :

anan02 m noxewz
an Eaves” 3 md moxcoxom no «m 4. 23

Figure 6—4. Sample Chromatagram for Ethanol Phase.

97

runs. The amount of benzaldehyde fed to the column was compared to the amount of
benzaldehyde that left the column in the raffinate and extract streams and the amount
remaining in the column when the run was terminated. The amount of benzaldehyde
fed into the column over the course of the run was calculated by analyzing the feed
several times during the run. An average of the feed concentrations was then
multiplied by the decrease in weight of the feed vessel. The benzaldehyde leaving the
extraction column in the raffinate was calculated by averaging the concentrations of
benzaldehyde in samples taken at regular intervals throughout the run and multiplying
the average times the total amount of raffinate collected. The benzaldehyde in the
extract stream was determined by analyzing a single, representative sample of the
ethanol cold-trap solution for the concentration of benzaldehyde and multiplying this
value times the amount of ethanol remaining in the trap at the end of the run. Finally,
the benzaldehyde remaining in the column at the end of the run was estimated by
multiplying the carbon dioxide volume remaining in the column times the extract
benzaldehyde concentration. The amount of benzaldehyde in the raffinate remaining
in the column at the end of the run was assumed to be negligible.
The complete material balance calculations for Run 3 are given below. The
measured parameters are shown in Figure 6-5:
1. The carbon dioxide flow rates for the extract and raffinate streams were
measured with wet test flow meters. For Run 3 the flows were 1032.2 L and 277.6 L
respectively. The volumetric rates were converted to mass flows according to the

following equations;

 

 

1032.2LC02( gm°1K )(1 atm)(flg_C2_2_) = 1845.1 gC02 6-14
0.08205 atm L 293 K gmol

 

 

)=496.l gC02 6-15

277.6L( gmolK )(latm)(44gC02

0.08205 atm L 293 K gmol

98

 

 

8183 g feed 1845.1 g C02

175 ppm benzaldehyde 450 ppm benzaldehyde
8120 g raffinate
18 ppm benzaldehyde 2341-7- g C02
496,1 g €02 0 ppm benzaldehyde

Figure 6-5. Run 3 Material Balance Diagram

The total carbon dioxide fed to the column was 1309.8 L (2341.2 g);

2. Benzaldehyde fed to the column;

175 g benz.

f d )2 1.432gbenz. 6-16
g ee

8183g feed(

3. Benzaldehyde leaving the column in the raffinate;

18 g benz.

 

] = 0.15 gbenz. 6-17

8120 g raffinate 6
10 g raffinate

4. The amount of benzaldehyde in the extract stream was determined by first by

99

analyzing the ethanol trap solution for benzaldehyde. For Run 3 this value was 1800

g benz./106 g ethanol. The total amount of trap solution was 460.74 g and the

benzaldehyde leaving the column in the extract was;

1800 g benz.

 

460.74 g extract [ ) = 0.83 g benz.

10 g extract

5. The amount of carbon dioxide remaining in the column at the end of each run

was 840 ml. The concentration of benzaldehyde in this holdup was estimated to be the

same as the average benzaldehyde concentration in the extract;

0.83 g benz. _ 450 g benz.

 

and the accumulation of benzaldehyde in the carbon dioxide holdup;

 

840 ml (0.85 g C02 ) [450 g benz.

6 = 0.33 g benz.
“‘1 10 gC02

6. The benzaldehyde balance is given by;
in = out + accumulation

7. For Run 3 Equation 6-22 becomes;

1.432 g benz. = 0.15 g benz. +0.83 g benz. + 0.33 g benz. = 1.31 g benz.

The benzaldehyde recovery is then

 

(1.31 g benz.

)100 = 91%
1.432 gbenz.

6-19

6-20

6-21

6-22

6-23

100

The carbon dioxide solubility in the raffinate was also calculated to determine
if the experimental values were close to the 6.3% at 25 C and 125 atm (1837 psi)

predicted by Wiebe (1977, p 476):

 

( 496.1 gC02

)100 = 6.1% 6-24
8120 g raffinate

which is reasonably good agreement.

CHAPTER 7

BENZALDEHYDE EXTRACTION EXPERIMENTAL RESULTS

The extraction column was run a total of seven times. Run 1 and Run 2 were
test runs to check the operation of the equipment and the results for these runs are not
reported.

The operating conditions for Runs 3 and 4 were kept identical to check the
reproducibility of the entire system. The results of these runs were, in fact, identical
within reasonable limits.

For Run 5, the operating parameters were kept the same as for the previous two
experiments except that the feed concentration was increased by a factor of two. There
was no significant change in the concentration of benzaldehyde (solute) in the raffinate
although the concentration of benzaldehyde in the extract did increase as expected.

For Run 6, the concentration of benzaldehyde in the feed was again increased
by a factor of two over the previous run with all other parameters held the same.

For the last run in this project, Run 7, the aqueous feed rate was decreased and
the carbon dioxide feed rate was increased such that the total feed to the column was
essentially the same as for all of the previous runs. This decrease in the solvent ratio
(i.e. decrease in slope of the operating line) was also a decrease in the carbon dioxide
retention time.

Table 7-1 focuses on the extraction column benzaldehyde material balances.
Considering the very low concentrations of benzaldehyde used for these experiments,
the benzaldehyde recoveries were very good with four of the five runs having 95% or
greater recovery.

Table 7-2 examines the column fluid mechanics that are an important part of the
information required for scale-up. The percent of flooding as predicted by the

correlations of Venkataraman and Laddha, Rao and Rao, and Crawford and Wilke are

101

102

 

 

 

 

 

 

 

 

 

 

 

mg: in. gm. 3 wd wood mod wad comm wbd— 020 b
v.5 3 A; man. 89m 35% w~.m M: .m wwvm egg :45 o
oéo com. 3. van; wovd Rm 3 .m acvm o9: we; m
9mm: mmm. mi. mom. in; 40m 36 wmvm 9.: 33 v
03 mwm. m3. omw. N9: Om.m wwé 3mm 3.2 mm; m
w
w N
6S @5203 m w ODE SEE w 58% noon .
@8983“ N00 035th ~8be com 0:5 8mm 30,—. 8mm 0 oZ
. .Ncom wEHEoQO .8 E SE
.Ncom E .Ncom E .535 E .Ncom
Bop Co 82m Bum Now 8:28 BE

 

 

 

 

 

 

 

 

 

mecca—om 13.832 55.30 .353be A.» oth.

 

103

 

4:0 o—th. HOW UDmD OmOH—u mm DEflw on: 0.3 mEOUN—Dtoo MGMGOOE Gnu MOM tom: wuouDEwuflm Eomwsnnm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.23: 6,634 93 5888838, mo 5:39:00 wEnooE 5:200 :38“me u 48>
.AwmaC and can and mo cocflotoo wEuooE 5:200 :ouombxm n mgwm
.88: .853 23 28380 We 839:5 @683 528 Susim u 33
as 53. 32 2:... E q: 3 h
:3 3m 3N 2m «.3 “.2 G o
:3 mom an M: 3m _.~_ mm m
#6 Ra «3 in gm _.2 mm v
8.0 gm mg 2m 3; q: 8 m
0K. 292:5. .3 13> m3 >30 .6 “38:2
Baa—tam cog—Bum 9320:
N00 and
5 ~00 NOD wEwooE mo “condom 83>?

 

 

 

 

 

 

mvmflfln—OQE Hyman—rm nus—:90 EGmuoflhuNm .Nnb 07—3—5

 

104

of particular interest. Conservative flow rates for both phases were used for the initial
column runs to avoid flooding. No evidence of flooding was observed since the
column could be run for extended periods of time without difficulty. Further runs at
higher total flow rates are required in order to determine which of the correlations is
the most accurate. The carbon dioxide retention time and turnovers are also included
in Table 7-2. The turnovers in particular were important since this parameter was used
as a rough prediction of when the column would reach equilibrium. It was believed
that a minimum of 2-3 turnovers were required before the system was at equilibrium.
In practice, samples of raffinate taken at fifteen minute intervals after the start of the
run showed that the raffinate benzaldehyde concentration reached a constant value
within a half hour or within less than one turnover. The concentration of carbon
dioxide in the raffinate phase was also recorded. This value is between 6.0-6.5%(wt)
for all of the runs which agrees very well with the 6.3% (wt) value for carbon dioxide
solubility in water at the same temperature and pressure (25°C and 1,500 psig)

determined by Wiebe (1977).

Number of Stages and the Height of a Transfer Unit

Calculations were performed for the three methods described in Chapter 3, p 39
for determining the number of stages in the column: 1) the number of theoretical stages,
NTS, which compares the continuous column to a staged contactor, 2) the number of
transfer units, NTU, which is based on film-transfer theory, and 3) the method of
Kremser;

1. Table 7-3 first lists the feed, extract and raffinate concentrations as measured
for Runs 3-7. Next a material balance is performed for one equilibrium stage (Equation
3-8) using the inlet concentrations for the feed and carbon dioxide streams actually
used in each experiment. The extract stream concentration calculated in this manner

is listed in the last column. It can be seen that for all of the runs except Run 5, the actual

105

Table 7-3. Equilibrium Stage Comparison

 

 

 

 

 

 

 

 

Average Solute Concentration, ppm One
Run Theoretical

Number Feed Extract Raffinate Stage Extract
Concentration

3 167 450 18 522

4 149 469 18 522

5 294 783 22 638

6 709 1561 65 1885

7 646 1083 63 1827

 

 

 

 

 

 

 

benzaldehyde concentration in the extract is slightly less than that calculated for the
case where one equilibrium stage was assumed (column five). The implication is that
in the 5 ft high experimental column there is close to one theoretical stage.

A graphical determination of the number of transfer stages performed following
the discussion in Chapter 3 is shown in Figure 7-1. Because of the steep slope of the
equilibrium curve it is somewhat difficult to see, but there is less than one stage for all
of the runs except Run 5 which has very close to one stage. The dashed lines in Figure
7-1 show the step representing an equilibrium stage for Run 5 and Run 6;

2. Table 7-4 displays the values for the number of transfer units, NTU, and the
height of a transfer unit, HTU, based on film-transfer theory. Since the principal
resistance to diffusion was in the raffinate phase (large distribution coefficient),
Equation 3-31, was used to calculate the NTU values. Further, since the benzaldehyde
concentrations were dilute and the mutual solubilities of water and carbon dioxide are

low, the integral of Equation 3-31 was evaluated using Equation 3-40. Examples of

Solute Concentration in Extract, ppm

2600

2400

2200

2000

1800

1600

l 400

1200

1000

800

600

400-

200

106

 

 

 

 

 

 

 

_— 'Equilibrium Curve

__ K=52

;

P 6

T

—_ 7

f.

- 5

2.. 4

i 3

3%l[llllllllllllllllllllllJlllll
100 200 300 400 500 600 700

Solute Concentration in Raffmate, ppm

Figure 7-1. Graphical Determination of Equilibrium Stages

800

107

Table 7-4. NTU and HTU Based on Film-Transfer Theory

 

 

 

 

 

 

 

Run Number NTU H?
3 2.52 1.98
4 2.38 2.10
5 2.90 1.72
6 2.68 1.87
7 2.48 2.02

 

 

 

 

 

these calculations are shown in Appendix E. The height of a transfer unit was
calculated by dividing the height of the experimental column, 5 ft, by the number of
transfer units;

3. A determination of the extraction factor and the number of stages using the
Kremser method is shown in Table 7-5. Examples of these calculations are also shown

in Appendix E.

108

Table 7-5. Equilibrium Stages Calculated by the Kremser Method

 

 

 

 

 

 

 

Run Number Extraction Factor Number of Stages
U n
3 14.9 1.08
4 15.6 0.95
5 15.7 1.45
6 15.9 1.17
7 25.4 0.84

 

 

 

 

 

CHAPTER 8

CONCLUSIONS AND RECOMMENDATIONS

Conclusions

The flooding correlation of Venkataraman and Laddha appears to be to
conservative for the benzaldehyde/carbon dioxide/water system. More runs are
required to determine if the correlations of Rao and Rao or Crawford and Wilke are
suitable for use in predicting flooding with this system. This is a very important
information required for any scale-up.

The limited number of runs conducted so far all used low values for the ratio of
the aqueous phase flow to the carbon dioxide phase flow (i.e. the slope of the operating
lines, R/E) ranging from 2.05-3.5. These initial values were conservative because the
equipment was new and unproven with respect to where flooding would occur. The
very high distribution coefficient of 52 indicates that much higher ratios of the two
solvents can be used.

As the ratio of aqueous phase to carbon dioxide is increased, the 6.5% (wt)
solubility of carbon dioxide in water must be compensated for. The carbon dioxide
flows in Runs 3-7 were greater than 5 g/min because at an aqueous feed of about 17 g/
min. such as those used in Runs 3-6, the weight of carbon dioxide in the raffinate will
be approximately 1.2 g/min which is almost 25% of the carbon dioxide feed. There is
a practical aspect of using carbon dioxide flows less than 5 g/min that must also be
considered. The backpressure regulator which controls the column pressure by
regulating the flow of carbon dioxide is slightly oversized. When operating, it would
open too far and allow the column pressure to decrease by approximately 80 psig. It
would take about 15 minutes for the pressure to recover before the valve opened again.
Lower carbon dioxide flows will make this problem worse. If possible, a smaller

backpressure regulator should be purchased and installed.

109

110

The material balance for one equilibrium stage for Runs 3-7 clearly indicate
there is close to one stage in the 5 ft high column (Table 7-3) as does the Kremser
method (Table 7-5). Film—transfer theory yields a different value of about 2.5 transfer
units. However, the film-transfer theory calculations are suspect in regards to
accuracy. The integral in the defining equation for the number of transfer units,
Equation 3-31, was simplified so that it could be solved analytically, Equation 3-40.
This simplification assumes that the two solvents are insoluble. Although the
solubility of carbon dioxide in water is relatively low (6.5% wt) it may still not be low
enough for the simplification to be used. The integral should be solved numerically to
determine if the number of transfer units is closer to one as suggested by the other

methods.

Recommendations
The following work and/or changes to continue this project are as follows:

1. Install a weigh cell or scale to monitor the flow of carbon dioxide. Currently,
only the benzaldehyde and water flows can be monitored and used for material balance
calculations.

2. Make more runs of the column to investigate where flooding occurs to
determine whether or not one of the correlations by Rao and Rao or Crawford and
Wilke is applicable. This is necessary to provide information for scale-up of the
column to commercial size.

3. Make more runs with different ratios of aqueous feed to carbon dioxide.
Compare the number of theoretical stages, the number of transfer units, and the
Kremser method stages calculated with this additional data to the numbers already
obtained to verify their accuracy.

4. Make additional runs using different feed distributor designs to determine if a

more efficient design is possible.

111

5. Consider purchasing a smaller backpressure regulator to control the column
pressure and regulate the carbon dioxide flow out the top of the column. The existing

regulator is slightly oversized.

APPENDIX A

PHYSICAL PROPERTIES

Table A-1. Physical Properties of Carbon Dioxide

 

 

 

 

 

 

 

 

 

 

 

 

 

U.S. Units1 Metric Units]

International Symbol C02 C02
Molecular Weight 44.01 44.01
Vapor Pressure

at 70 F (21.1 C) 838 psig 5778 kPa, gage

at 32 F (21.1 C) 491 psig 3385 kPa

at 2 F (-16.7 C) 302 psig 2082 kPa

at -20 F (-28.9 C) 200 psig 1379 kPa

at -69.9 F (-56.6 C) 60.4 psig 416 kPa

at -109.3 F (-78.5 C) 0 psig 0 kPa
Density of the gas

at 70 F (21.1 C) 0.1144 lb/ft3 1.833 kg/m3

at 32 F (0 C) 0.1234 1b/ft3 1.977 kg/m3
Specific gravity of the gas

at 70 F (21.1 C), 1 atm (air = 1) 1.522 1.522

at 32 F (0 C), 1 atm 1.524 1.524
Specific Volume of the gas

at 70 F (21.1 C), 1 atm (air = 1) 8.741 ft3/1b 0.5457 m3/kg

at 32 F (0 C), 1 atm 8.104 ft3/lb 0.5059 m3/kg
Density of liquid, saturated

at 70 F (21.1 C) 47.6 lb/ft3 762 kg/m3

at 32 F (21.1 C) 58.0 113/03 929 kg/m3

at 2 F (-16.7 C) 63.3 lb/ft3 1014 kg/m3

at -20 F (-28.9 C) 66.8 lb/ft3 1070 kg/m3

at -69.9 F (-56.6 C) 73.5 1b/ft3 1177 kg/m3
Critical temperature (1 atm) 87.9 F 31.1 C
Critical pressure 1070.6 psia 7382 kPa, abs

 

112

 

113

Table A-1. Continued

 

 

 

 

 

 

 

U.S. Units1 Metric Units1

Critical density 29.2 113/113 468 kg/m3
Latent heat of vaporization

at 32 F (21.1 C) 100.8 Btu/lb 234.5 kJ/kg

at 2 F (-16.7 C) 119.0 Btu/lb 276.8 kJ/kg

at -20 F (-28.9 C) 129.6 Btu/lb 301.4 kJ/kg
Specific heat of gas

at 77 F (25 C)

CI) 0.203 Btu/1b F 0.850 kJ/kg C

Cv 0.157 Btu/lb F 0.657 kJ/kg C
Ratio of specific heats 1.304 1.304
Solubility in water, vol/vol

at 68 F (20 C) 0.90 0.90
Solubility in water, (wt/wt)2

at 25 C, 125 atm 0.063 0.063
Viscosity of saturated liquid

at 2 F (-l6.7 C) 0.287 lb/ft hr 0.119 g/cm sec

 

 

 

 

1 Above values from: Compressed Gas Association, H_andbookfiof Compressed Gases

2 Wiebe, 1977

 

 

114

'lhble A-2. Physical Properties of Water

 

U.S. Units Metric Units

 

Solubility in carbon dioxide
at 25 C, 125 atm2 0.0014 0.0014

 

 

 

 

2Wiebe, 1977

 

'Ihble A-3. Physical Properties of Benzaldehyde

 

 

 

 

 

 

 

 

 

 

 

U.S. Units3 Metric Units3
Formula C7H60
M. W. 106.13 106.13
Viscosity of liquid at 25 C 1.39 g/cm sec
Boiling Point 178 C
Specific gravity 1.04510’4 1.04510’4
3 Weast, 1970

 

Table A-4. Physical Properties of Isopentyl Alcohol (Isoamyl Alcohol)

 

 

 

 

 

 

U.S. Units3 Metric Units3
Formula C5H120
M. W. 88.15 88.15
Viscosity of liquid at 10 C 6.20 g/cm sec
Boiling Point 128.5 C
Specific gravity 0.8092 0.8092

 

 

 

 

3Weast, 1970

 

 

 

 

115

Table A-5. Physical Properties of n-Propanol

 

 

 

 

 

 

 

 

U.S. Units3 Metric Units3

Formula C3H80
M. W. 60.11 60.11
Viscosity of liquid

at 20 C 2.256 g/cm sec

at 30 C 1.72 g/cm sec
Boiling Point 97.4 C
Specific gravity at 20 C 0.8035 0.8035

 

 

3Weast, 1970

 

 

APPENDIX B

ETHANOL PHASE GAS CHROMATOGRAPHY

An internal standard method using isopentyl alcohol (IPA) as the standard was
used to determine the concentration of benzaldehyde in the aqueous phase. A Perkin-
Elmer model 8500 equipped with a hot-wire thermal conductivity detector was used for
the gas chromatography. The column was an Alltech Associates, Inc. Econo-Cap
capillary column, catalog number 19653, 10 m long x 0.53 mm 0, with a 1.2 pm thick
carbowax stationary phase.

Calibration mixtures of isopentyl alcohol and benzaldehyde were prepared
according to Table B-1. The column was calibrated by injecting each of the sample
mixtures multiple times and recording the number of area counts detected per unit of
sample mixture injection volume. The data are presented in Table B-2. Calibration
curves were prepared for both the internal standard, isopropyl alcohol, and
benzaldehyde and are shown in Figure B-1 and Figure B-2.

The column operating conditions including the temperature versus time profile
are shown in Figure B-3.

A sample chromatogram is shown in Figure B-4.

116

117

Table B-1. Ethanol Phase Calibration Standards

 

 

 

 

 

 

 

 

 

Sample Ethanol IPA Benz. IPA Benz.
Description g pl 111 Conc. Conc.
PPm (Wt) PPm (Wt)
1 14.7316 20 15 1100 1068
2 13.9332 30 25 1744 1882
3 13.6305 50 40 2971 3078
4 13.1170 75 55 4631 4398
5 12.1774 85 65 5654 5599
6 11.2367 100 85 7209 7935
7 8.3474 100 100 9704 12,567

 

 

 

 

 

 

 

 

Specific Gravity IPA = 0.81 at 20 °C
Specific Gravity of Benzaldehyde = 1.049 at 20 °C

 

 

118

Table B-2. Ethanol Phase G. C. Calibration Data

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sample 13033:): IPA Benz.
Description 111 AC*/ul AC*/},Ll
1 .3 16.67 17.28

1 .3 16.85 18.01

1 .3 16.89 17.86

1 .3 16.37 17.58

2 .3 26.44 30.66

2 .3 25.36 29.93

2 .3 27.33 31.89

2 .3 27.28 31.92

3 .3 41.22 45.76

3 .3 48.27 52.29

3 .3 48.16 53.23

3 .3 45.37 49.54

4 .3 74.13 74.42

4 .3 73.69 72.65

4 .3 70.85 70.81

4 .3 71.88 72.53

 

 

 

 

 

 

* AC = Area Counts

 

 

Table B-2. Continued

119

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sample 131106133): IPA Benz.
Description 11.1 AC*/},tl AC*/},tl
5 .3 89.90 93.91
5 .3 90.03 102.17
5 .3 88.15 91.83
5 .3 90.30 ’ 93.33
6 .3 106.74 122.35
6 .3 111.27 127.32
6 .3 102.71 118.56
6 .3 108.13 125.20
7 .3 152.05 206.48
7 .3 151.20 205.30
7 .3 136.65 185.03
7 .3 120.11 162.45

 

 

 

* AC = Area Counts

 

 

120

_

oooor
_

 

«magnum 0.0 883 :owmbaooaou «an: .Tm 0.5me

Q5 Ema .aougcooaou <n=

ooom 0000 000%

_________F_

P

 

tnlov ‘esuodsau '3 '9

000m?

121

 

8.530% .U .0 389» 5355800 33022823 .~-m Bummm

03 can .couabcoocoo oPEoEfiaom

coco? ooom o
_ _ _ _ _ _ r _ _ _ o

 

III/av ‘asuodsau '3 '9

122

SECTION 1 GO CONTROL

1
OUEN TEMP (BEG C) 68
ISO TIME (MIN) 4.0 0.5
RAMP RATE (BEG C/MIN) 30.8

HNB 1 RANGE OFF HHB 1 POLARITY B-A
FIB 2 SENS LON

BET ZERO ON
INITIAL BET 2

IHJ 1 TEMP OFF
INJ 2 TEMP_ 250
DET 1 TEMP 250
n57 2 TEMP 270

FLON 1 1 flL/MIN FLON 2 1 ML/HIN PRESSURE 3 1.3 PSIG
CARRIER GAS 1 NE CARRIER GAS 2 HE

EQUILIB TIME 8.8 MIN
TOTAL RUN TIHE 9.1 MIN

SECTION 3 DATA HANDLING
‘ DATA ACQUISITION” « REPORT

START TIME 8.00 MIN CALC TYPE 2

ENB TIME 9.16 MIN AREA/HT CALC AREA
PRINT TOL 0.0000

NIBTH 3 . OUTPUT »

SKIM SENS 10 ' SCREEN YES

BASELINE CORR U-U PRINTER NO
EXT BEU N0

BET 1 AREA SENS 200
BET 2 AREA SENS 200
BET 1 BASE SENS 4
BET 2 BASE SENS 4
PEAK IDENTIFICATION QUANTITATION

UNRETB PEAK TINE 0.00 MIN SCALING FACTOR 25.3389
AREA/NT REJECT 1.0808

REF PK: TIME 3.36 MIN
TIHE TOL 0.18 HIN

COMPONENT LIST .
RT RF STD AMT NQNE GRP

3.36 1.0000 199.1080 IBA 0
6.22 1.0800 0.0000 BENZ 8

Figure B-3. G. C. Column Operating Conditions for Ethanol Phase

123

Powhmm »I\qmm¢ A quwa m

mmo«.m D mmm~.m om.“
vvwm.m 3 mumm.n wm.v
mumm.m~mm 3 memm.«vmm on.“
N iwmi Om Immq km
N "ZOHFQJDUJCU DmHuHQOE W QDIFME

t vtqm

azmmM

'3'!

W@\#@\mm km"

'3
II.

o
[M

 

 

'0
fl \
v U
an
ID
”3'
I

 

mm; +

20m *1 0H 0 m :

canoe—mnnom SE SE amigo: m nothz
.8»: Eng 3.: at 235: 3 md mmxvoxmm mm "m 4 23.

Figure B—4. Sample Chromatogram for Ethanol Phase

APPENDIX C
AQUEOUS PHASE GAS CHROMATOGRAPHY

An internal standard method using n-propanol as the standard was used to
determine the concentration of benzaldehyde in the aqueous phase. A Perkin-Elmer
model 8500 equipped with a hot-wire thermal conductivity detector was used for the
gas chromatography. The column was an Alltech Associates, Inc. Econo-Cap capillary
column, catalog number 19653, 10 m long x 0.53 mm d) , with a 1.2 pm carbowax
stationary phase.

Calibration mixtures of n-propanol and benzaldehyde were prepared as shown
in Table C-1, Table C-2, and Table C—3. The column was calibrated by injecting each
of the mixtures multiple times and recording the number of area counts detected per
unit of sample mixture injected into the gas chromatograph. The results are presented
in Table C-4, Table C-5, and Table C-6. Calibration curves were prepared for both the
aqueous phase internal standard, n-propanol, and benzaldehyde and are shown in
Figure C-l, Figure C-2, and Figure C-3 respectively.

The column operating conditions including the temperature versus time profile
are shown in Figure C-4.

A sample chromatogram is shown in Figure C-5.

124

Table C-l. G. C. Calibration Mixtures of n-PropanollWater

 

 

 

 

 

 

 

 

Sample Water n-Propanol n-Propanol
Description g g ppm
4 13.9088 .0318 2286
5 15.8686 .0400 2521
6 14.8453 .0447 3011
7 16.4039 .0580 3536
8 15.9350 .0645 4048

 

 

 

 

Table C-2. G. C. Calibration Mixtures of Benzaldehyde/Water

 

 

 

 

 

 

 

 

Sample Water Benzaldehyde Benzaldehyde
Description g g ppm
1 99.47 .0136 137
2 99.47 .0220 221
3 99.38 .0325 327
R1 99.45 .0115 116
R2 99.49 .0275 276

 

 

 

 

 

 

126

Table C-3. G. C. Calibration Mixtures of Benzaldehyde/Water

 

 

 

 

 

 

 

Sample Water Benzaldehyde Benzaldehyde
Description g g ppm
F1 18.0231 .0272 1509
F2 13.5357 .0253 1869
F3 15.6002 .0332 2128
F4 11.7451 .0315 2682

 

 

 

 

 

127

'Ihble C-4. n-Propanol/Water G.C. Calibration Data

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sample Sample Volume n-Propanol
Description 111 AC/ 111
4 .5 31.0372
4 .5 31.9292
4 .5 30.1992
4 .4 32.3858
4 .3 32.1330
5 .4 35.1278
5 .4 35.7135
5 .3 36.156
6 .3 42.664
6 .3 43.2277
6 .3 42.7687
7 .3 49.654
7 .3 49.9863
7 .3 50.1897
8 .3 57.0787
8 .3 56.8027
8 .3 56.5543

 

 

 

 

 

AC = Area Counts

 

 

128

Table C-S. Benzaldehyde/Water G.C. Calibration Data

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sample Sample Volume Benzaldehyde

Description [.11 AC/ul
1 .5 2.4842

1 .5 2.4152

1 .5 2.4510

2 .5 4.0940

2 .5 4.1094

2 .5 4.0060

3 .5 6.1258

3 .5 6.0394

3 .5 6.2208
R1 .5 2.1376
R1 .5 2.0526
R1 .5 2.0264
R2 .5 4.8794
R2 .5 4.9504
R2 .5 4.8908

 

 

 

 

129

Table C-6. Benzaldehyde/Water G.C. Calibration Data

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sample Sample Volume Benzaldehyde
Description 111 AC/ul
F1 .5 26.9498
F1 .3 27.5753
F1 .3 27.4340
F2 .3 34.2380
F2 .3 34.0917
F2 .2 34.3615
F3 .3 39.3780
F3 .2 39.3450
F3 .2 39.5530
F4 .2 47.6050
F4 .2 48.7700
F4 .2 49.6200

 

 

 

 

130

Scanned .U .0 380.» couabcuocoo 6585-: .TU 8.3mm

can .couabcooaou Efiqocmé

comm . ooom

_________...____

co

m

N

 

ooom
0N

 

m/ov ‘esuodsau '3 '9

131

omaoamom 0.0 «85> nova—E880 onxnovfincom .NO Esmfi

can .couabcoocou onaaocancom
0mm com omw
. . w _

_______L_,____

 

 

III/3v ‘asuodseu '3 '9

132

ooom

0mcoam0m .0 .0 0:82, 80880250 00200385 .mO 0.5me

comm

Ema .aouabe0ocoo 00300383

ooom

_ _ _ _ .._.. _

coo?

 

 

 

IHIITIIIIIIIIlllllllllljlll

mm

mm

ov

om

m/3v 'asuodsau '3 '3

133

SECTION 1 GO CONTROL
1
OUEN TEHP (DEG C) 78
ISO TIME (NIH) 0.8
RAH? RATE (DEG C/HIN) 36.6

HHD 1 RANGE OFF
FID 2 SENS LON

HUD 1 POLARITY B-R

nET ZERO ON
INITIAL BET 2
INT 1 TEHP OFF
INJ 2 TEHP 250
OET 1 TEHP 259
SET 2 TENP 279
FLOH 1 1 HL/HIN PLON 2 1 HL/HIN PRESSURE 3 1.3 PSIG
CORRIER ans 1 HE CORRTER 69$ 2 HE
EOUILIB TIHE 9.9 NIN
TOTAL RUN TIHE 5.3'HIN
SECTION 3 ONTO HANDLING
DQTQ acaUISITION REPORT
STRRT TTNE 0.99 "IN anc TYPE 2
END TIHE 5.33 HIN RREn/NT caLc AREA
PRINT TOL 9.9099
HIBTH 5 OUTPUT
SKIN SENS 1 SCREEN YES
BASELINE CORR u-u PRINTER NO
EXT DEU NO
BET 1 ORER SENS 299
nET 2 AREA SENS 299
SET 1 BOSE SENS 4
nET 2 BOSE SENS 4
PEAK IDENTIFIchION QUANTITATION
UNRETn PEAK TTNE 0.99 HIN SOPLINO FACTOR 1.9099
AREn/HT REJECT 9.2599 '
REF PK: TTNE 9.00 NTN
TIHE TOL 9.10 NTN

Figure 04. G. C. Column Operating Conditions for Aqueous Phase

134

pomhmm ._.I\qmm¢ A mnvEmm m

mmma.md Humfi.a mm.¢
mmmm.w¢ 3 omwv.a mv.m
o¢¢m.mm mmmm.o SH.“

.x ammq um ammo Pm

...sn "ZDHFEJDUJGU DwHuHQDE .v QOIFME

Haxemxmm mmunfi m 22m

 

 

m¢.m
m 3
SE.“ m
26m a“ a
OEEOESEB 8% mm ananoz 0 901.5:
Baas: can 9. 25:83 3 no H oxvaxmm mm ... N a 23.

Figure 05. Sample Chromatogram for Aqueous Phase G. C. Analysis

APPENDIX D
ULTRAVIOLET SPECTROPHOTOMETRY

Ultraviolet spectrophotometry was used to analyze for benzaldehyde in water.
Benzaldehyde exhibits a strong absorption in the ultraviolet region of the
electromagnetic spectrum at 241 nanometers as shown in Figure D-l (Simons, W. W.,
Ed, spectra 2163).

Two calibration standards of 10 ppm (wt) and 210 ppm (wt) were prepared and
used to calibrate the u.v. spectrophotometer. Aliquots of each of these standards were
diluted and run on the u.v. to measure their absorbance. The results are presented in
Table D—l. The results for both standards are plotted in the calibration curve shown in

Figure D—2.

2163

BENZALDEHYDE
C7H60

Mol. Wt. 100.12

9.9. 179°C (111.)
Solvent: Cyclohexane

Max. am Cell mm Cone.

338 27.3 10 13

   

Figure D-l. U.V. Spectra for Benzaldehyde

135

136

Table D-l. Benzaldehyde/Water U. V. Calibration Data

 

Standard

Sample

 

 

 

 

 

 

 

 

 

 

 

 

 

Concentration D122: (on Concentration AbRsélrclfiirge
PPm (Wt) ppm (wt)
10 1.5:] 6.67 .735
10 2:1 5.0 .606
10 3:1 3.3 .406
10 4:1 2.5 .306
10 10:1 1.0 .125
210 40:1 5.25 .596
210 40:1 5.25 .606
210 40:1 5.25 .616
210 50:1 4.2 .489
210 80:] 2.63 .306
210 80:1 2.63 .300

 

 

 

 

 

137

 

PEG couefifiu cacaoeofioeooam .>.D usagnznocficom dd 8&5

8.328%? .> .D

 

(1m) mdd ‘uopenuaouog Spfiqapfezuag

APPENDIX E

SAMPLE CALCULATIONS

Number of Transfer Units

The defining equation for the number of transfer units developed from film-

transfer theory is given by Equation 3-31:

E-l

 

 

 

Because the benzaldehyde concentrations were very dilute and the water and carbon

dioxide solvents are relatively insoluble (0.014 g HzO/g C02 and 6.4 g C02/g H20 at

25 C, 1500 psig respectively) Equation 3-31 can be simplified using the assumption

described in Chapter 3: Simplified Integration:

 

 

 

 

 

 

x / XE \
RI dx xR ——‘E
x (XR_XR) 1-— sz
R2 mE xR —
__ K 2 m} _

I38

139

where xR mole fraction benzaldehyde in feed
I
sz mole fraction benzaldehyde in raffinate
sz mole fraction benz. in carbon dioxide feed
m is the distribution coefficient = 52
R water flow rate, gmol/min
E carbon dioxide flow rate, gmol/min

For Run 3 the above parameters were:

 

 

 

 

 

 

 

 

mo] fraction benz. in H20 feed 2.84 x 10'5
mol fraction benz. in C02 feed 0

mol fraction benz. in extract 1.87 x 10'3
mol fraction benz. in raffinate 3.06 x 10'6
water flow rate 0.95 gmol/min
carbon dioxide flow rate 0.21 gmol/min

 

 

 

 

 

Then
0.951%”?01
— = mm = 0.087
(52)0.21g"T°
rmn
and

1-3. =1—0.087 = 0.913
mE

fl

 

= 1.096

E-3

E-4

140

— =0 E-6

 

 

 

 

 

( XE \
2
1 R XRT‘Tn— R 284x10"5
ln (1——) —— +—— =(1.096)1n[0.913['——:5]+0.087]
hi “IE XE. “IE 3.06x10
m5 _ \sz I) j

= 2.52

The height of a transfer unit, HTUOR is given by Equation 3-21

 

APPENDIX F
NOTATION

Mass flow of feed solvent alone in feed (moi/h)

Column (tower) cross-sectional area (m2)

Interfacial area per unit volume of tower (m2/m3)

Surface area of packing (mz/m3)

Concentration of solute in extract phase (kmol/m3)

Concentration of solute in extract liquid film interface (kmol/m3)
Concentration of solute in raffinate phase (kmol/m3)

Equilibrium concentration of solute in the raffinate, (kmol/m3
Concentration of solute in raffinate liquid film interface (kmol/m3)
Logarithmic mean of (CR — C11) 1 and (CR — C11) 2

Extract

Packing factor

Gravitational acceleration, 32.2 ft/s2

Extraction column height, (m)

Height of a transfer unit based on the extract phase mass transfer
coefficient
Height of an overall transfer unit based on raffinate phase mass
transfer resistances (m)
Height of an overall transfer unit based on extract phase mass
transfer resistances(m)
Height of a transfer unit based on the raffinate phase mass transfer
coefficient

Height of a film transfer unit, (m)

Extract liquid-film mass transfer coefficient (m/s)

141

142

Raffinate liquid-film mass transfer coefficient (m/s)

Overall mass transfer coefficient based on extract phase (Tn/s)
Overall mass transfer coefficient based on raffinate phase (m/s)
Extract volumetric flow rate per unit area (m3/m2 s)

Raffinate volumetric flow rate per unit area (m3/m2 s)

Distribution coefficient in weight or mole fractions

Solute transfer rate per unit area (kmol/m2 3)

Number of transfer units based on extract mass transfer coefficient
Number of transfer units based on raffinate mass transfer coefficient
Number of film transfer units

Pressure, Pa

Reduced pressure

Mass flow rate, g/min

Raffinate

Mass flow rate of extraction solvent alone in feed (mol/h)
Temperature, (°C )

Reduced temperature

Extraction factor

Continuous phase superficial velocity (m/s)

Distributed phase superficial velocity (m/s)

Weight fraction solute in extract

Weight fraction solute in feed

Weight fraction solute in raffinate

Mass fraction solute in aqueous phase into first equilibrium stage
Mass fraction solute in aqueous phase out of first equilibrium stage
Mass fraction solute in aqueous phase out of second equilibrium

stage

143

Equilibrium mole fraction solute in extract

Mole fraction of solute in extraction column feed

Mole fraction of component i in extract

Equilibrium mole fraction solute in raffinate

Mass fraction solute in solvent phase into first equilibrium stage
Mass fraction solute in solvent phase out of first equilibrium stage
Mass fraction solute in solvent phase out of second equilibrium stage
Mole fraction of component i in raffinate

Height of packing (m)

Selectivity

Activity coefficient

Activity coefficient at reference state

Void fraction of packing

Continuous phase viscosity (N s/mz)

Discontinuous phase viscosity (N s/mz)

Density (kg/m3)

Density of the continuous phase, (kg/m3)

Density of the dispersed phase

Density difference between phases (kg/m3)

Interfacial tension (J/mz)

Interfacial tension of water in equilibrium with its pure vapor at

25 °C (71.98 dyne/cm)

REFERENCES CITED
Akgerman, A., R. K. Roop, R. K. Hess, S. D. Yeo, “Supercritical Extractions in

Environmental Control”, pp 479-509, In: Summritical Fluid Technology. Reviews in
Modern Theog and Applications, T. J. Bruno, and J. F. Ely, Eds., CRC Press, Inc., Boco

Raton, FL, (1991).
Breckenfeld, R. R., and CR. Wilke, Chem. Eng. Prog., 66, p 305, (1942).

Briggs, D., “French Oil Mills machinery Company Pressing”, Memo to the Cherry
Marketing Institute, (October 25, 1991).

Compressed Gas Association, Handbook of Compressed Gases, 3rd ed., Van Nostrand
Reinhold, New York, (1990).

Coulson, J. M., and J. F. Richardson, Chemical Engineering, Volume 2, Unit Operations,
2nd ed., Pergammon Press, New York, (1968).

Crawford, J. W., and C. R. Wilke, “Limiting Flows in Packed Extraction Columns”,
Chem. Eng. Prog., 47, (8), 423, (1951).

Cruess, W. V., Commercial Fruit and Vegetable Products, 4th ed., McGraw—Hill, New
York, pp 738—743, (1958).

Donahue, D. J., and F. E. Bartel, “The Boundary Tension at Water-Organic Liquid
Interfaces”, Jour. Phys. Chem., 56, p 480, (1952).

Gaylor, R. N., and H. R. C. Pratt, “Transactions of Institution of Chemical Engineers”,
London, 35, p 267, (1957).

Gaylor, R. N., N. W. Roberts, and H. R. C. Pratt, “Transactions of Institution of Chemical
Engineers”, London, 57, (1953).

Gill, T., A. H. C. Chen, and H. Grethlein, “Feasibility Study Using Cherry Pits as a
Natural Benzaldehyde Source”, Report No. 2 to the Cherry Marketing Institute, (October
31, 1990).

Gribb, A., and C. T. Lira, “Benzaldehyde Recovery Process Desi gn”, Report to the Cherry
Marketing Institute, (August 12, 1992).

Gore, G., “On the Properties of Liquid Carbonic Acid”, Proc. Roy. Soc., London, 1 1, p 85.

144

145

Hannay, J. B., and J. Hogarth, “On the Solubility of Solids in Gases”, Proc. R. Soc.,
London, 30, 324, (1879).

Holmes, M. J., and M. VanWinkle, “Prediction of Ternary Vapor-Liquid Equilibria from
Binary Data”, Ind. Eng. Chem., 62, (1), 21, (1970).

Kremser, Natl. Petrol. News, 22, (21), p 42, (1930).

Lira, C. T., “Physical Chemistry of Supercritical Fluids - A Tutorial”, In: Sup_ercritical

Fluid Extraction and Chromatography: Techniques and Applications, B. A. Charpentier,
and M. R. Sevenants, Eds., ACS Symposium Series 366, Washington D. G, 1-25, (1987).

Massoudi, R., and A. D. King, Jr., “Effect of Pressure on Surface Tension of Water.
Adsorption of Low Molecular Weight Gases on Water at 25°.”, Jour. Phys. Chem., 78,
(22), P 2262, (1974).

McHugh, M. A., and V. J. Krukonis, Supercritical Fluid Extraction Principles and
Practices, Butterworth, Stoneham, MA, (1986).

Nair, M. G., and A. Chandra, “HPLC Analyses of Kernel and its Cake from Montmorency
Cherry Pits”, Report to the Department of Horticulture, Bioactive Natural Products
Laboratory, Michigan State University, (October 29, 1991).

Null, H. R., and D. A. Palmer, “Predicting Phase Equilibria”, Chem. Eng. Prog., 65, (9),
47, (September, 1969).

Oberg, A., and S. C. Jones, “Liquid-Liquid Extraction”, Chemical Engineering, 70, part 2,
(15), P 119, (July 22, 1963).

Paulitis, M. E., et. al., “Phase Equilibria Related to Supercritical-Fluid Solvent
Extraction”, Ber. Bunsenges. Phys. Chem, 88, pp 869-875, (1984).

Perry, R. H., and C. H. Chilton, Eds., Chemical Engineers’ Handbook, 5th ed., McGraw-
Hill, New York, (1973).

Rao, M. R., and C. V. Rao, “Flooding Rates in Packed Extraction Towers”, Chem. Eng.
Sci., 9, pp 170-175, (1958)

Scheibel, E. G., “Fractional Liquid Extraction”, Chem. Eng. Prog., 44, p 681, (1948).

Schultz, W. G., and J. M. Randall, “Liquid Carbon Dioxide for Selective Aroma
Extraction”, Food Technology, 24, pp 1282-1286, (November 1970).

146

Schultz, W. G., et. al., “Pilot Plant Extraction with Liquid C02”, Food Technology, p 32,
(June 1974).

Simons, W. W., Ed., The Sadtler Handbook of Ultraviolet Spectra, Sadtler Research
Laboratories, Divison of Bio-Rad Laboratories, Inc., (1979).

Treybal, R. E., Liguid Extraction, lst ed., McGraw-Hill, New York, (1951).

Treybal, R. E., Liguid Extraction, 2nd ed., McGraw-Hill, New York, Chap. 4, p 132,
(1963).

Venkataraman, G., and G. S. Laddha, “Limiting Velocities, Holdup, and Pressure Drop at
Flooding in Packed Extraction Columns”, A. I. Ch. E. Jour., 6, 3, (1960).

Weast, R. C., Ed., CRC HandbookLof Chemistry and Physics, 50th ed., The Chemical
Rubber Company, Cleveland, Ohio, (1967-1970).

Weibe, R., “The Binary System Carbon Dioxide-Water Under Pressure”, Fenilizer
Research Division, Bureau of Plant Industry, U. S. Department of Agriculture,
Washington, D. C., (1977)

Whitman, W. G., Chem. Met. Eng., 29, p 147, (1923).

Wiegand, J. H., Trans. Am. Inst. Chem. Engrs., 36, p 679, (1940).

Wohl, K., “Thermodynamic Evaluation of Binary and Ternary Liquid Systems”, Trans.
Am. Inst. Chem. Engrs., 42, 215, (1946)

Woods, W. S., et. al., “Carbon Dioxide Solubility in Water”, Ind. Eng. Chem., 1, 1, (1956).

HICHIGRN S A

111111111111111“