“It: L'u
. as'fi‘h '4
R1

1

.u. .
‘93.,
.9.

no... u It.
{Nil 21:...

. .flWU...
, , . . gflhiib
. u I." u.-’. ‘Vll
; 1,‘ A V. . . . .
. . . 9.». .5 «Km»... . .. . .. ‘.

. .. .. .Jt'.) Futuwfi” V . . . . .
NMhMudwmqunv. Inna... ‘ .
IJWJW A A 113. 533;“!wa

'v.«y‘)\lk
~.vry A '

 

\.

‘ 11197.!
Aall v! .1:
'1!!!ch

 

 

 

THESIS

_——

IN!INHIHHHIHHUHIHHIHHUN"!!!UIUJUJIHIHM

31293 01565 9307

L'BPAPY
Mic - gar: date
University

 

 

 

 

This is to certify that the

dissertation entitled

Optimization of Batch Antisolvent Crystallization

presented by

Satu Marketta Uusi-Penttila

has been accepted towards fulfillment
of the requirements for

 

Ph. D. degree in Chemical Engineering

,1,
Major professor

Date €/ 3/717

MS U is an Affirmative Action/Equal Opportunity Institution 0.12771

PLACE IN RETURN BOX to remove thlo checkout from your record.
TO AVOID FINES return on or botoro date duo.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU to An M'Irrnotlvo ActiorVEquol Opportunity Intuition

 

OPTIMIZATION OF BATCH ANTISOLVENT CRYSTALLIZATION

By

Satu Marketta Uusi-Penttila

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Chemical Engineering

1997

ABSTRACT

OPTIMIZATION OF BATCH ANT ISOLVENT CRYSTALLIZATION

By

Satu Marketta Uusi-Penttilé

Batch antisolvent crystallization is a commonly used crystallization method in the pharmaceutical
industry, which produces very pure crystals with a narrow particle size distribution and high
yields. It is also an effective method to crystallize heat sensitive materials, since the

crystallization can be achieved at low temperatures.

The objectives of this research were to study the effect of the antisolvent addition rate on various
crystallization parameters. including operational parameters and product specification
parameters, and to find an operating procedure that produces a desired particle size distribution.

The monitoring of the system is done in situ using attenuated total reflection Fourier transform

inframd (ATR Fl'lR) spectroscopy.

Since antisolvent crystallization involves two solvents. the polarity behavior of binary systems is
addressed first. This involved the introduction of a spectroscopic method to estimate polarities of
pure solvents. The same method was applied to binary mixtures. It was shown that even small
amounts of antisolvent will cause significant nonideality in the polarity of the system. This should

be accounted for in antisolvent crystallization.

The chosen crystallization was l-lysine monohydrochloride purification using water as a solvent
and ethanol as an antisolvent. The solubility data for I-lysine monohydrochloride in water.
ethanol, and mixtures of water and ethanol were determined. The addition of ethanol decreased

the solubility of l—lysine monohydrochloride in water significantly.

The growth kinetics of l-lysine monohydrochloride were estimated from nucleation cell and
seeded laboratory scale experiments. These data were used to predict nucleation rates. It was

shown that the nucleation rate is very high throughout the crystallization.

The effect of the antisolvent addition rate on bulk supersaturation and particle size was studied.
ATR FTIR spectroscopy was used to study the bulk supersaturation. it was shown that the bulk
supersaturation was not a function of antisolvent concentration. Sieving showed that the particle

size was strongly dependent on the antisolvent addition rate.

This dissertation employed a spectroscopic method for studying solvent polarity behavior. The
feasibility of ATR FT IR spectroscopy for monitoring batch antisolvent crystallization was

demonstrated. Also, an operating scheme for producing the desired particle size was presented.

ACKNOWLEDGMENTS

I would like to thank my advisor, Dr. Kris Berglund, for his enthusiasm and the relaxed way in
which he guided me through my research project. Tack as mycketl

I would like to thank the members of my dissertation committee. Dr. Daina Briedis. Dr. Alec
Scranton, and Dr. Gary Blanchard for their interest toward my research.

I acknowledge the financial support from the Fulbright Commission in Finland, Nests 0y
Foundation, Helsinki University of Technology, Helsinki University of Technology Foundation,
Finnish Cultural Foundation. and us. Department of Agriculture.

I would like to thank everybody in the lab. A special thank you goes to Sanjay Yedur who is a
true friend. His help in sorting out my research related (and unrelated) disaaers is greatly
appreciated. Dilum Dunuwila deserves my thanks for guiding me through the mysteries of the
inframd spectrometer. I would also like to thank Dr. Hasan Alizadeh for the valuable discussions
and practical advice.

IamgratefultoBenandMarleBohnhorstforproofreadingmydissertationandshowingmeby
theirown example howto boa good bridge player.

I would like to thank my parents, Sulo and Pirkko Uusi-Penttila, and my brother, Jyrki, and my
sister, Pauliina. for all their support. I would also like to thank my host parents, Frank and Sue
Boley. I would have to write another thesis to list all the things they did for me throughout the
years.

The biggest thank you goes to my husband Rik who survived this whole ordeal. His support and

belief in me were a great help. Our endless walks, talks and bridge sessions were great
medication tomystress.

iv

TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................... vii
LIST OF FIGURES ....................................................................................................... ix
LIST OF SYMBOLS ...................................................................................................... xi

CHAPTER 1
INTRODUCTION ..........................................................................................................
1.1 Crystallization from solution ........................................................................
1.2 Research objectives ....................................................................................
1.3 Batch crystallization ....................................................................................
1.31 Background...
1 ..3 2 Controlled operation of. batch. crystallrzers ...................................
1.4 Batch antisolvent crystallization ..................................................................
1.4.1 Introduction
1.4.2. Choice of solvents ......................................................................
1.4.3 Amino acid salt recrystallization ...................................................
1.5 Polarity changes in binary systems .............................................................

CVNO’O’bNNN-‘d

-l

CHAPTER2
SPECTROSCOPICALLY DETERMINED DIELECTRIC CONSTANTS FOR VARIOUS
ESTERS ....................................................................................................................... 11
2.1 Introduction 11
2.2 Expenmental 12
2.2.1 Chemicals 12
2.2.2 Instrumentation 12
2. 2. 3 Sample preparation ..................................................................... 13
2.3.1 Absorption versus emissron ......................................................... 13
2. 3. 2 Onsager and "De-bye". functions ...................................................... 14
2.4 Discussion... 16
2.5 Conclusrons 22

CHAPTER 3
POLARITY IN BINARY SYSTEMS ................................................................................ 23
3.1 introduction 23
3.2 Experimental .............................................................................................. 23
3.2.1 Chemicals 23
3.2.2 lnstmmentatron 24
3.2.3 Sample preparation ..................................................................... 24
3.3 Results ....................................................................................................... 24
3.4 Discussion .................................................................................................. 25
3.5 Conclusions ................................................................................................ 31

CHAPTER 4

SOLUBILITY OF L-LYSINE MONOHYDROCHLORIDE .................................................... 32
4.1 Introduction .................................................................................................... 32
4.2 Solubility of l-lysine monohydrochloride in water ............................................. 33
4.3 Solubility of I-lysine monohydrochloride in ethanol .......................................... 33
4.4 Solubility of l-Iysine monohydrochloride in ethanol-water mixtures .................. 35
4.5 Discussion ........................................ . ............................................................. 35
4.6 Conclusions .................................................................................................... 37
CHAPTER 5
IN SITU MONITORING OF ETHANOL AND L-LYSINE MONOHYDROCHLORIDE
CONCENTRATION IN A BENCH SCALE CRYSTALLIZER WITH ATR FTIR ................... 38
5.1 Introduction .................................................................................................... 38
5.2 Experimental .................................................................................................. 39
5.2.1 Materials ......................................................................................... 39
5.2.2 Instrumentation ............................................................................... 39
5.3 Attenuated total reflection Fourier transform infrared spectroscopy ................. 40
5.3.1 Introduction to ATR FTIR spectroscopy ........................................... 40
5.3.2 Derivative spectroscopy .................................................................. 45
5.4 Calibration of infrared spectra and component concentrations ........................ 47
5.4.1 Calibration curve for ethanol concentration ..................................... 47
5.4.2 Calibration curve for I-Iysine monohydrochloride concentration ....... 49
5.5 Conclusions .................................................................................................... 51
CHAPTER 6
EXPERIMENTAL ARRANGEMENT AND CRYSTALLIZATION PROCEDURE ................. 52
8.1 Introduction .................................................................................................... 52
6.2. Materials ........................................................................................................ 52
6.3 Instrumentation ............................................................................................... 54
6.4 Crystallization ................................................................................................. 54
6.5 Sieving ........................................................................................................... 55
CHAPTER 7
CRYSTALLIZATION KINETICS FOR L-LYSINE MONOHYDROCHLORIDE ..................... 57
7.1 Introduction .................................................................................................... 57
7.2 Crystal growth ................................................................................................. 58
7.2.1 Growth rate dependency on crystal size .......................................... 58
7.2.2 Overall growth rate from seeded experiments ................................. 60
7.3 Nucleation ...................................................................................................... 63
7.4 Conclusions .................................................................................................... 67
CHAPTER 8
INFLUENCE OF ETHANOL ADDITION RATE ON PARTICLE SIZE DISTRIBUTION... 71
8.1 Introduction ................................................................................................. 71
8.2 Crystal mass ............................................................................................... 72
8.3 Supersaturation .......................................................................................... 74
8.4 Weight based mean crystal size .................................................................. 76
8.5 Crystal size distribution ............................................................................... 78
8.6 Conclusions ................................................................................................ 80
CHAPTER 9
CONCLUSIONS ............................................................................................................ 81
APPENDIX .................................................................................................................... 84
LIST OF REFERENCES ............................................................................................... 103

vi

IJST OF TABLES

Table 2.1. The emission wavelengths for Nile Red in the various solvents of known
dielectric constants. The values for the Onsager function, f(D), are calculated from
Equation 2.1; for the Debye function, d0), from Equation 2.2, and the dielectric constants
have been obtained from CRC Handbook of Chemistry and Physics [26]. The comparisons
in the last column are: ‘ from Smyth and Walls [47]. and ° from Stolarova et al. [48] .......... 20

Table 2.2. The emission wavelengths for Nile Red in various eaers of unknown dielectric
constants. The values for the Onsager and Debye functions using Figure 2.2, and the
respective dielectric constants calculated from Equations 2.1 and 2.2. The comparisons in

the last column are: “ from Smyth and Walls [47], and ” from Stolarova et al. [48] ............... 21
Table 3.1. Nile Red emission maxima for diethyl succinate-ethanol binary mixture as a
function of the mole fraction of diethyl succinate at 25 °C. The accuracy is :I:1 nm .............. 26
Table 3.2. Nile Red emission maxima for diethyl fumarate-ethanol binary mixture as a
function of the mole fraction of diethyl fumarate at 25 °C. The accuracy is :I:1 nm ............... 27
Table 3.3. Nile Red emission maxima for diethyl maleate-ethanol binary mixture as a
function of the mole fraction of diethyl maleate at 25 °C. The accuracy is :1 nm ................. 28
Table 3.4. Nile Red emission maxima ethyl lactate-water binary mixture as a function of
the mole fraction of ethyl lactate at 25 °C. The accuracy is 11 nm ....................................... 29
Table 3.5. Nile Red emission maxima for ethanol-water binary mixture as a function of the
mole fraction of ethanol at 25 °C. The accuracy is :I:1 nm .................................................... 30
Table 5.1. Conelation coefficients for multivariate analysis of the band at 1411 cm",
i-Iysine monohydrochloride concentration and ethanol concentration ................................... 49
Table 6.1. Two sieve series used for determining the product crystal size distribution ......... 56

Table A.1. Data for Figure 2.1. The Nile Red absorption and emission maxima in various
esters and alcohols. The emission spectra were excited at absorption maxima ................... 84

Table A.2. Data for Figure 4.1. Solubility of l-lysine monohydrochloride in water as a
function of temperature ....................................................................................................... 85

Table A3. Data for Figure 4.2. Solubility of l-lysine monohydrochloride in various binary
mixtures of water and ethanol at 30 2t 1 °C .......................................................................... 86

Table A.4. Data for Figure 5.2. The penetration depth of the evanescent wave for pure
water and pure ethanol as a function of wavenumber. The data were calculated using
Equation 5.4 ........................................................................................................................ 87

vii

Table A.5. Data for Figure 5.5. Relative absorbance as a function of ethanol weight
percentage .......................................................................................................................... 88

Table A.6. Data without ethanol for Figure 5.8. Difference of the second derivative peaks
at 1411 cm" and 1432 cm" as a function of the weight fraction of l-lysine
monohydrochloride .............................................................................................................. 89

Table AJ. Data with ethanol for Figure 5.6. Difference of the second derivative peaks at
1411 cm'1 and 1432 cm'1 as a function of the weight fraction of l-Iysine
monohydrochloride .............................................................................................................. 90

Table A.8. Data without Ethanol for Figure 7.2. Spherical equivalent diameter for single
crystals in a seeded nucleation cell ...................................................................................... 91

Table A.9. Data with Ethanol for Figure 7.2. Spherical equivalent diameter for single
crystals in a seeded nucleation cell ...................................................................................... 92

Table A.10. Data for Figure 7.3. Mass of I-lysine monohydrochloride crystals retained on
the sieves after 2, 5, and 10 minutes of crystallization. The experiment was done at 30 °C
and the ethanol addition rate was 20 mllmin. The average seed size was 550 um and the
mass of seeds was 10 g ..................................................................................................... 93

Table A.11. Data for Figure 7.4. Estimate for the overall growth rate as a function of
ethanol addition rate from seeded experiments ................................................................... 94

Table A.12. Data for Figure 7.5. Crystal size distribution of the fines after 30 minutes of
crystallization. Antisolvent was added at 1 mllmin and the seeds for the seeded

experiments were the 500-800 urn sieve fraction ................................................................ 95
Table A.13. Data for Figure 7.6. Predicted nucleation behavior during crystallization .......... 96
Table A.14. Data for Figure 7.6. Predicted nucleation behavior during crystallization .......... 97

Table A.15. Data for Figure 7.7. Experimental and predicted yield for unseeded
experiments using the predicted nucleation and growth rates .............................................. 98

Table A.16. Data for Figure 8.1. The effect of ethanol addition on the increase in crystal
mass during the crystallization ............................................................................................. 99

Table A.17. Data for Figure 8. 2. Relative bulk supersaturation as a function of added
ethanol for various ethanol addition rates ............................................................................ 100

Table A.18. Data for Figure 8.3. Weight mean size of crystals as a function of ethanol
addition rate at 30 °C. The initial solution contained 42 w-% of l—Iysine monohydrochloride
in water ............................................................................................................................... 101

Table A.19. Data for Figure 8.4. Experimental and predicted cumulative crystal size
distributions for 5 mllmin and 30 mllmin antisolvent addition rates ...................................... 102

viii

LIST OF FIGURES

Figure 1.1. Feedback diagram to demonstrate the complex interaction between crystal

size distribution and factors that form the crystal size distribution [44] ................................ 3
Figure 1.2. Supersaturation profiles for natural and controlled cooling (MuIIin [34]) ............. 5
Figure 1.3. Solubility of adipic acid in different solvents (Myerson [36]) .............................. 8
' Figure 1.4. The solubility of various amino acids in aqueous 1-propanol solutions [42] ....... 9

Figure 2.1. Nile Red emission maxima as a function of Nile Red absorption maxima in
various esters (o) and alcohols (o). The emission spectra were excited at absorption
maxima ............................................................................................................................... 15
Figure 2.2. f(D) and do) for solvents of known dielectlic constant as a function of the
emission wavelength of Nile Red (alcohols (o) and esters (x) for f(D). and alcohols (o) and
esters (+) for 41(0)) ............................................................................................................... 18
Figure 2.3. Dielectric constant as a function of Nile Red emission maximum ...................... 19

Figure 3.1. The emission maxima of Nile Red in differing mole fractions of diethyl
succinate in ethanol. The straight line represents the ideal behavior of the system .............. 26

Figure 3.2. The emission maxima of Nile Red in differing mole fractions of diethyl
fumarate in ethanol. The straight line represents the ideal behavior of the system ............... 27

Figure 3.3. The emission maxima of Nile Red in differing mole fractions of diethyl
maleate in ethanol. The straight line represents the ideal behavior of the system ................ 28

Figure 3.4. The emission maxima of Nile Red in differing mole fractions of ethyl lactate in
water. The straight line represents the ideal behavior of the system .................................... 29

Figure 3.5. The emission maxima of Nile Red in differing mole fractions of ethanol in
water. The straight line represents the ideal behavior of the system .................................... 30

Figure 4.1. Solubility of I-Iysine monohydrochloride in water as a function of temperature. 34

Figure 4.2. Solubility of l-Iysine monohydrochloride in various binary mixtures of water
and ethanol at 303 :t: 1 K ..................................................................................................... 36

Figure 5.1. Total internal reflection ..................................................................................... 41

Figure 5.2. Penetration depth of the evanescent wave for pure water and pure ethanol as
a function of wavelength ...................................................................................................... 43

ix

Figure 5.3. Infrared spectra of water, aqueous l-Iysine monohydrochloride, and ethanol at
30 °C ................................................................................................................................... 44

Figure 5.4. First and second derivatives for two overlapping Gaussian bands (Cahill [5]).... 46
Figure 5.5. Relative absorbance, A,, as a function ethanol weight percentage .................... 48

Figure 5.6. Difference of the second derivative peaks at 1411 cm'1 and 1432 cm'1 as a
function of the weight fraction of I-Iysine monohydrochloride ............................................... 50

Figure 6.1. The experimental arrangement: A temperature controlled antisolvent
reservoir, B pump; C jacketed crystallizer with a marine type impeller, D Dipper-210 deep
immersion probe; E FTIR spectrometer, F computer ........................................................... 53

Figure 7.1. The temperature controlled nucleation cell. a rod with parent crystal glued to
it, b glass slide and the rod it is glued to, c temperature element, d chamber for constant
temperature water, e chamber for saturated solution. [48] ................................................... 59

Figure 7.2. Spherical equivalent diameter for single crystals in a seeded nucleation cell.
The experiments were done at 30 °C. The ethanol/water ratio was 1.221 on volume basis... 61

Figure 7.3. Mass of l-lysine monohydrochloride crystals retained on the sieves after 2, 5,
and 10 minutes of crystallization. The experiment was done at 30 °C and the ethanol
addition rate was 20 mllmin. The average seed size was 550 um and the mass of seeds
was 10 g ............................................................................................................................. 82

Figure 7.4. Estimate for the overall growth rate of l-lysine monohydrochloride as a
function of ethanol addition rate from seeded experiments .................................................. 64

Figure 7.5. Crystal size distribution of the fines after 30 minutes of I-lysine
monohydrochloride crystallization at 30 °C. Antisolvent was added at 1 mllmin and the
seeds for the seeded experiments were the 500-600 um sieve fraction ............................... 68

Figure 7.6. Predicted nucleation behavior during crystallization .......................................... 89

Figure 7.7. Experimental and predicted yield for unseeded experiments using the
predicted nucleation and growth rates .................................................................................. 70

Figure 8.1. The effect of ethanol addition on the increase in crystal mass during the
Hysine monohydrochloride crystallization at 30 °C .............................................................. 73

Figure 8.2. Relative bulk supersaturation for I-lysine monohydrochloride purification as a
function of added ethanol for various ethanol addition rates ................................................ 75

Figure 8.3. Weight mean size of I-lysine monohydrochloride crystals as a function of
ethanol addition rate at 30 °C. The initial solution contained 42 w-% of l-lysine
monohydrochloride in water ................................................................................................. 77

Figure 8.4. Experimental and predicted cumulative crystal size distributions for 5 mllmin
and 30 mllmin antisolvent addition rates. The l-lysine monohydrochloride purification was
done at 30 °C and the initial solution contained 42 w—% of l-lysine monohydrochloride in
water ................................................................................................................................... 79

LIST OF SYMBOLS

absorbance, [-]

absorbance at k, {-1

spherical equivalent area, [mm’]

relative absorbance, [-]

absorptivity, [dm3 I g cm]

nucleation rate, [#I m3 s]

kinetic order of nucleation, [-]

pathlength. {cm}

coefficient of variation, {-1 .

concentration, [9/ dm’]

conesntration of solute in solution, [kg/m3]

concentration of solute in solution at equilibrium, [kg/m3]
supersaturation, Ac = cL - cL .q, [kglm’]

static dielectric constant, {-1

difference in second derivative peaks, {-1

penetration depth, [um]

Onsager function, {-1 ,

Gaussian probability function, {-1

growth rate, [um/min]

kinetic order of growth, {-1

intensity of the second derivative peak at A, {-1

exponent, {-1

growth rate constant. {-1

nucleation rate constant, {-1

volume shape factor, k. a 1:18, {-1

size, length, [m]

mean size of a sieve out i, [urn]

weight based mean size, [um]

particle size at 16% on the cumulative undersize or oversize plot, [um]
particle size at 84% on the cumulative undersize or oversize plot, [um]
length of a crystal. [um] ‘
mass of ethanol, [9] g

l‘“ moment of the crystal size distribution, [mfi

mass of l-Iysine monohydrochloride, [g]

equilibrium mass of I-lysine monohydrochloride in solution, [9]
initial mass of l-Iysine monohydrochloride, [9]

mass of solids, [9]

suspension density, [kg/m3]

mass of water, [kg]

number of reflection points in contact with the sample, [-]
population density, [#1 m‘]

population density based on total volume of the crystallizer, [#1 m‘]
reflective index of the denser material, {-1

xi

assasisgeee §§£§$<“*”““pe

D

12

refractive index of the lea dense material, {-1

volumetric ethanol flowrate, [mllmin]

radius of a sphere, characteristic length of a crystal, [um]
correlation coefficient, {-1

supersaturation, {-1

temperature, [°C]

transmittance, [~]

time, [min]

working volume of crystallizer, [ml]

volume of water, [ml]

mass of crystals in suspension. [kg crystals/kg of solvent]
width of a crystal, [um]

mass of crystals on sieve i, [9]

weight-% of I~Iysine monohydrochloride, [w-%]

refractive index of sample, [-]

refractive index of ATR crystal, {-1

incident angle of infrared radiation, [-]
absorption wavelength, [nm]

wavelength in the ATR crystal, [nm]
emission wavelength, [nm]

density of l-lysine monohydrochloride, [glml]
density of ethanol, [glml]

density of water, [glml]

standard deviation of the particle size, [um]
relative bulk supersaturation. [-]

Debye function. {-1

xii

Chapter 1

INTRODUCTION

1.1 Crystallization from solution

A wide variety of products in chemical, food, and pharmaceutical industries are manufactured by
crystallization from solution. Crystallization can be used to separate or purify a product or an
intermediate, and to obtain a solid product with a desired particle size distribution and crystal
habit. In many applications the product specifications are very strict. In spite of that the control of
crystallizers is still rare, and for most industrial crystallizations a trial-and-error approach is used

to meet the product specifications.

Crystallization from solution is especially useful in the pharmaceutical industry where almost all
the products are crystallized or precipitated at least once during the manufacturing. Furthermore,
at room temperature, the solid state is the most stable form of the majority of pharmaceutical
products used today [43]. Crystallization also determines the particle properties of the product
that can have a major influence on the pharmaceutical characteristics of the drug [43].
Properties, like bulk density, particle size distribution, surface area, crystal form, and crystal
shape, all have an effect on the bioavailability of the active component; and consequently, they
directly affect the dosage of the drug. The particle size distribution can also be critical in

downstream processing, including steps like tableting, filtration and pumping.

2
Crystallization is, however, a complicated operation. To achieve a desired crystal size

distribution the driving force of crystallization, i.e. supersaturation, has to be controlled.
Supersaturation control is very difficult since the factors that influence the crystal size distribution
also have feedback effects [44], [59], as demonstrated in Figure 1.1. Therefore, the mass
balance, population balance, and growth and nucleation kinetic equations have to be solved

simultaneously to obtain the crystal size distribution.

1.2 Research objectives

The goal of this research is to develop an operating strategy for a batch antisolvent
crystallization system to obtain a desired crystal size distribution. Attenuated total reflection
Fourier transform infrared spectroscopy will be used to monitor the liquid phase in situ during the
crystallization. The infrared data will be used to study the effect of antisolvent addition rate on
various parameters. The model system will be l-lysine monohydrochloride purification from
aqueous solution using ethanol as the antisolvent. This study will provide better control of

product quality and reduce batch-to-batch variations in process performance.

1.3 Batch crystallization
1.3.1 Background

The bulk production of pharmaceuticals handles much smaller quantities of product than most
other chemical industries, such as inorganic chemicals or fertilizers. Therefore, pharmaceutical
production is usually done in batch mode. The main advantage of batch crystallization over

continuous crystallization is the simpler equipment. The same crystallizer can be used for a

Growth rate

 

L
A Growth Growth rate

Kinetics

 

 

 

 

 

 

 

V
Feed Mass Super-
Balance saturation

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l
_, Nucleation Nucleation Population CSD _
Kinetics rate Balance fl
Crystal
Area

 

 

 

Figure 1.1. Feedback diagram to demonstrate the complex interaction between crystal size
distribution and factors that form the crystal size distribution [44].

4
variety of different products and the crystallizer is easy to clean between batches. Cleaning is

necessary to prevent contamination from batch to batch. Other advantages of batch
crystallization include low level of maintenance, and particular suitability for difficult processes,
like processing of toxic materials or exclusion of contaminants. Batch crystallizers can also
produce a narrower crystal size distribution (Wey [59]). However, the analysis of a batch process
is considerably more complex than that of a continuous system due to the dynamic nature of the
batch process. Also batch-to—batch fluctuations pose a problem in batch processing, which can
cause considerable variation in the crystallization process resulting in final product divergence.
Such variation is unacceptable in the pharmaceutical industry. Reworking a batch that does not
meet the specifications is time consuming and expensive, and it opens the opportunity for

additional contamination.

1.3.2 Controlled operation of batch crystallizers

Batch crystallizers are generally operated under one of the following three modes: cooling
crystallization, evaporative crystallization, or antisolvent crystallization (Mullin [34]). The aim of
the controlled operation of a crystallizer is to obtain a desired crystal size distribution. In cooling

and evaporative crystallization this is achieved by controlling the level of supersaturation in the

crystallizer.

Operating a batch cooling crystallizer under a natural cooling profile without any temperature
control is known to produce a supersaturation peak in the crystallizer. This peak causes a burst
of nucleation that leads to excessive formation of very small particles (fines) and small average
crystal sizes. These small crystals, in tum, lead to fouling problems, reduced product yields, and
problems in downstream product handling (Mullin [34]). To avoid this problem, the crystallizer is
operated according to an established cooling curve, is. the cooling rate is small initially and is
gradually increased towards the end of the batch. This approach keeps the level of

supersaturation constant in the crystallizer and prevents excessive nucleation. The

Supersaturation

 

 

 

‘ - . , Natural cooling

 

 

Time

Figure 1.2. Supersaturation profiles for natural and controlled cooling (Mullin [34]).

8
supersaturation profiles obtained from uncontrolled (natural) and controlled cooling are shown in

Figure 1.2.

A similar strategy is used for evaporative batch crystallization processes. In this case, the
controlled operations involve operating the crystallizer initially at a lower evaporation rate and
then gradually increasing it. The problems occurring due to natural evaporation are the same as

for natural cooling described eariier.

1.4 Batch antisolvent crystallization

1 .4.‘I Introduction

A batch crystallization operation commonly uwd in the pharmaceutical and biochemical industry
is antisolvent crystallization. Numerous amino acids, including proline, l-asparagine, and
l-alanine, and pharmaceuticals, including antibiotics, are crystallized using antisolvent
crystallization (Kirwan and Orella [25]). In antisolvent crystallization a solute is crystallized from
solution by the addition of another substance (a soluble solid, liquid or gas). The added
substance effectively reduces the solubility of the solute in the original solvent, and thus
increases the supersaturation. This type of crystallization is known by a variety of terms. When
the added substance is another liquid and the solute is organic, like in pharmaceutical
crystallizations, the term antisolvent crystallization is used. Also terms like sailing-out, diluent
crystallization, and watering-out are used. Setting-out and diluent crystallization usually refer to

inorganic solutes, and watering—out can be used when the antisolvent is water.

The advantages of antisolvent crystallization are many. The major advantage is the possibility to
perform the crystallization at low temperature which is essential when heat sensitive solutes, like

amino acids, are crystallized. This method also produces narrow crystal size distributions.

7
Further advantages that can be obtained by the right choice of antisolvent are high purity crystals

(Karpinski [23]) and high yield. The disadvantage of the technique is that, usually, a separation

unit is required to recover the added antisolvent.

1.4.2. Choice of solvents

ln crystallization the choice of solvent is important. The solvent can have a major effect on the
growth rate and the crystal habit, and therefore also on the crystal size distribution (Myerson et
al. [37], Davey [10]). Also, the choice of solvent can affect the solubility of the solute
significantly. Figure 1.3 demonstrates the effect of various solvents on the solubility of adipic

acid in various solvents (Myerson [36]).

Choosing the antisolvent should also be done carefully. The antisolvent should be miscible with
the original solvent over the ranges of concentrations encountered and the solute should be
relatively insoluble in it. Also, the final solvent-antisolvent mixture must be readily separable.
Figure 1.4 shows the effect of addition of iso-propanol on the solubilities of some aqueous
solutions of amino acids (Kirwan and Orella [25]). These data demonarate that the right choice
of solvent can decrease the solubility of the solute by orders of magnitude, and indicates the

possibility of high solute yields.

1 .4.3 Amino acid salt recrystallization

The system that was chosen for this research is the purification of l-lysine monohydrochloride
from aqueous solution using ethanol as an antisolvent. X-ray crystallography was used to confirm
that ethanol does not change the crystal structure of l-lysine monohydrochloride. L-lysine
(2,6-diaminohexanoic acid) is an essential amino acid that has been shown to affect the growth
of rats (Budavari [3]). It is often the limiting amino acid in animal nutrition, and thus it is added to

animal feed (Kirk-Othmer [24]). It is also an essential part of pre- and post-operational nutrition

SOLUBILITY (mole fraction)

0.201

IDEAL
0.15-

0.10d

005‘ IN ET HANOL/

IN ISOPROPANOL

IN WATER
0.“) ——-—; I l l 1
20.0 25.0 30.0 35.0 40.0 45.0

TEMPERATURE, °c

 

Figure 1.3. Solubility of adipic acid in different solvents (Myerson [36]).

 

10 l l 1 u

e
d

 

Relative Solubility X/X°

 

 

 

 

 

 

.01
9- .

.001 0 9W” A
O alanine ‘.
|'_'| asparaglnemp ‘
A isoleucine
O phenylalanine f

.0001 - . l '
0.0 0.2 0.4 0.0 0.8

Mole Fraction Alcohol in Solution

Figure 1.4. The solubility of various amino acids in aqueous 1-propan0l solutions [42].

10
for humans. This system was chosen as a model of a pharmaceutical crystallization. L-lysine

monohydrochloride is easy to crystallize, and it is freely soluble in water and very slightly soluble

in ethanol.

1.5 Polarity changes in binary mm:

In the case of batch antisolvent crystallization, there is very little published information on
operating strategies; even though the technique is commonly used for the production of
pharmaceuticals and amino acids. Most of the related research is done for continuous
crystallization (for example, Bator [2], Mina-Mankarios and Pinder [31]). There are a few batch
studies, but they are done using inorganic solutes (Budz et al. [4], Jones and Teodossiev [22],
Jones et al. [21], Karpinski and Nyvlt [23], Mullin et al. [35], and Tavare and Chivate [54]). There
are two more closely related studies. 68095 and Laguérie [15] have studied the antisolvent
crystallization of D-xylose. They add only 1 w-% of antisolvent, and consequently assume that
the change in solubility is linear. This makes it difficult to generalize their results. Ny'vlt’s study
[40], on the other hand, is'theoretical. Both of these studies assume that the salvation of the
solute is a simple competition between the two solvents. However, when the two solvents are of
significantly different polarity, the salvation is affected by nonlinear polarity changes due to either
dielectric enrichment effect or hydrogen bonding. There is a vast amount of literature available
on this topic (Ghoneim and Suppan [16], Midwinter and Suppan [30], Nitsche and Suppan [39],
Reichardt [45], Suppan [50, 51, 52]). These solvent effects cause the local polarity around the
solute molecule to differ significame from the bulk polarity of the mixture. This makes the
system more complex than that assumed by Gabas and Laguérie [15] and Ny'vlt [40], and the
optimal operating conditions will not be achieved using these approaches. Thus, instead of
assuming ideal behavior of solvents, the nonlinear polarity changes can be exploited in

developing the operating strategy for antisolvent crystallization.

Chapter 2
SPECTROSCOPICALLY DETERMINED DIELECTRIC CONSTANTS FOR VARIOUS ESTERS

2.1 Introduction

ThepolarityofasolventcarlbedefinedbytheOnsagerfunctionortheDebyefunction. Theyare
both functions of the static dielectric constant of a solvent (Ghoneim and Suppan [16], Suppan
[52]). Polarity has also been shown to be related to the solvatochromic shifts of the absorption
and fluorescence spectra (Dutta et al. [14], Suppan [51]). Using a polarity sensitive
solvatochromic probe molecule. such as Nile Red [14], the Onsager function and the Debye
function can be consisted with the peak shifts In the emission maxima of Nile Red. These
correlations an then be used to estimate dielectric constants for various less known solvents,
like dibasic esters.

The emission maxima of the probe molecule Nile Red are measured in different solvents of
known dielectric constants, and relationships between the emission maxima and the Onsager
and Debye functions are established. Both the Onsager function and the Debye function will give
a linear relationship when correlated with either the maximum absorption wavelength of Nile Red
or the maximum emission llllavelength of Nile Red. These plots an be uwd to estimate
dielectric constants for less known solvents.

This chapter follows closely the article 'Spectroscopically determined dielectric constants for
various esters” by Uusi-Penttila et al. [57].

11

12
2.2 Experimental procedure

2.2.1 Chemicals

The following solvents were used to determine the linear relationship bstllveen the emission
maxima of Nile Red aid the two polarity functions: methanol (absolute) from Mallinckrodt;
ethanol (anhydrous) from Quantum Chemical Corporation; n—pmpanol (anhydrous). n-octanol
(994-96), and ethyl acetate (absolute) from J.T. Baker; n-butanol (994-96). n—pentanol (99+%),
n—hexanol (98%), ethyl acetoaoetate (99%), ethyl female (97%), propyl formats (97%). butyI
acetate (99%), methyl propionate (99%), and methyl butyrate (99%) from Aldrich. Nile Red was
alsofromAldrich.

ThedidedficMmdaaflfinedfafltefdmnglessmsdvmzdmahyl
succinate (984-96), dimethyl glutarate (98+%) and dimethyl adipate (984%) that were graciously
supplied by Du Pont; diethyl succinate (99%), diethyl maleate (97%), diethyl fumarate (98%),
diethyl l—tartrate (99%), dibutyl maleate (97%), dibutyl l-tartrate (99%), ethyl l-lactate (98%), and
triethylcitlate(99%)fromAldrich; dlbutyl ltacorlaerromLancastel: dlbutyl fumarate(90%)frorn
Kodak;anddirnethylmaleate(8akergrade)frornJ.T. Baker. All chemicalswereusedwithout
further purification.

2.2.2 Instrumentation

TheeqmmlonusedfuabsorptimnnasuranentsmsaPaHmElmerLanbdaMW-Ws
Spactmphotoneter,alddlefluoescalceexpenmntsmperfonnedonaSPEx
FLUOROLOG1681022 r'n Spectrometer. Quartz cuvettes were used for both absorption and
anissimexpefinentsfiheaccumcyfobmhflieabsamlmmdonisdmspemmmflnm.

1 3
2.2.3 Sample preparation

Theeaerspossessastrongcharacteristicabsorption below350nmthatsaturatetheabsorption
instrument. Therefore, Nile Red that absorbs and emits at considerably higher wavelengths has
used as a solvatochromic probe. According to Days and Berger [11] and Dutta et al. [14], Nile
Red is very sensitive to the polarity of the medium in which it is located. Nile Red has also been
found to be very soluble and strongly fluorescent in organic solvents (Gm and Fowler
[17])-

Nile Red is only needed intraceamountsforthe measurements. Thus, a method described In
StreetandAcreel491lIlasusedtoensuretheaccuracyoftl'le NIleRedconcentration.A10‘3M
stocksolutionotNlleRedmsprepamdpydlsoolvingdoosasgoerleRedlnzomlor
spectroscopicgrademematol.rhestocksolutionwaspipettedlmovlals. 50ulofstocksolutlon
was used for absorption sample vials and 10 pl for emission sample vials. Thosewlll yield final
Nile Red concentrations of 10“5 M forabsorption samples and 211045 M for the emission samples.
Thesolventwasallouadtoairdry, leavingtheappropriateamountolele Redinthevial, and
the Nile Redvasredissolvedinsmlofachosensolvent.

2.3 Results

2.3.1 Abeorpdon versus emission

As expected from the abundant literature on this subject (Suppan [53, 54], Ghoneim and Suppan
[16]. Onsager [41]), no geraral polarity scale can be established on the basis of solvatochromic
shifts of a probe. However. within categories of solvents. such as alcohols or esters, the
absorption and emission peak shifts follow a trend. These trends can be correlated with the

14
polarity of the solvent that is reflected by the static dielectric constant of that solvent. D, through

either the Onsager function or the Debye function (Suppan [53, 54], Onsager [41]).

Onsager function RD) = 2 (0-1) I (20+1) 2.1

Debye function rp(D) = (D-1) I (0+2) 2.2

Fluaescuwespeamswpyvasdnsenhaemmmabsapdmspedmscopybecamdits
enhanced sensitivity, i.e. larger peak shifts (Ghoneim and Suppan [16]). This isdemonstrated in
Figure 2.1 wheretheabsorption peaksforthechosenestersand alcoholsfallwithintherangeof
519nm and 557nm (38nm) whereas the emission peaks cover the range from 583nm to
640nm (57nm). Thelinewcorrelaticn betmn disorwonand emission peaksis given in
Equation2.3.Asanbemnfromtheequation,emissiongivesdaout1.5tirnesla'gerpeak
shiftsthanabeorption.

w = -225 1' 1.57 ' W (I2 = 0.88) 2.3

2.3.2 Onsager and Debye functions

Therelationshipsbetvaentl'lehmfunctionsandtheNile Redemission maximallere
established. Since some of the esters studied here contain hydroxyl groups both esters and
alcohols were included in the correlations. Using only the esters fornthe correlations would
considerably underestimate the values of the dielectric constants for esters containing hydroxyl
groups, whereasusing onlythealcohol datathelowerdielectricconstantscould notbeestimated
using fills method. Combining the two groups gave the most consistent results for both the high
andthe lowend ofthedielectn’c constants.

15

 

 

limb“ 3 " 225 +1.57 lfigorpfion l

640‘ 320.88 5 0

 

 

If'IIII

5301

I I I I

U T T r I j V U

610 -

I I I

Emission wavelength, [nm]

6004

590-»

 

 

 

 

Absorption wavelengdl, [nm]

Figure 2.1. Nile Red emission maxima as a function of Nile Red absorption maxima in various
esters (o) and alcohols (o). The emission spectra were excited at absorption maxima.

16
Thebediinearwndafimsmobtainedwhenflnexdhfimwaspeflumedmflwwavdmgth

of the maximum Nile Red absorption peak for each sample. Nile Red emission spectra were
recordedforpuresolventsofknowndielectricconstants. TheOnsagerarld Debyefunctionswere
calculated from Equations 2.1 and 2.2 and plotted as a function of the Nile Red emission
maxima. The plots are shown in Figure 2.2 and the numerical data are presented in Table 2.1.
The linearequationsweregenerated in EXCELbyaleaasquarestreatmentofthedata.

«0) = -1 .45 + 0.0037 * M (r2 = 0.95) 2.4

41(0) = -2.55 + 0.0055 * 11......»" (r2 = 0.94) 2.5

The linear relationships in Figure 2.2 were then used to predict the dielectric cormms for
chemically similar substances of unknown dielectric constants. The emission spectrum of Nile
Redinasdvanofmmamdieieanccmstarnvasmcadedbyexdfingatdleabsuptim
maximum. Figure 2.2 was used to determine the values for the Onsager and the Debye functions
and the dielectric constant of the solvent was calculated from Equations 2.1 and 2.2. The
resulting dielectric constants for selected solvents are presented in Figure 2.3 and in Table 2.2.

D = (f(D) + 2) I [2 (1 - l70)] 2-1

o = [1 + 2 “on I [1 - 41(0)] 2.2

2.4 Discussion

Despite a comprehensive literature search, direct memurement of dielectric constants for ethyl

lactate, triethyl citrate and many dibasic esters, that are considered as environmentally benign

17
substitutes for various chlorinated solvents, were not found. The spectroscopic approach was

found to provide an easy, experinental method to estimate static dielectric constants of less
common solvents.

It can be seen from Figure 2.3 that both the Onsager function and the Debye function give very
similar results. Thus, for the systems studied here neither model is superior. The scatter in
Figure2.3alsoshowsthatbecauseoftheexpomntialnatureofthetvinmodels,thesmaller
values of the dielectric constant are more reliable than the larger values. For dielectric constants
up to 15 this method tends to underestimate the value of the dielectric constant up to 15 %. For
higher values of dielectric constants this method tends to overestimate the values, about 25 % at
dielectricconstantofzs,andtheerrerincreasestowardsthehighervaluesofdielectric
W

The accuracy of the estimated dielectric constants was compared with other related results from
literature. An earlier publication by Smyth and Walls [47] gives electric moments for dilute
solutions of ethyl formate, ethyl acetate, diethyl maleate, diethyl fumarate, and diethyl succinate
in benzene. Extrapolation of their results to obtain the dielectric constants for pure solvents
agrees well with the results presented here. Further comparison was done with the results of
Stolarova et al. [48]. They have published values for various solvent polarity parameters
including results on diethyl maleate and ethyl lactate. The function they have used is similar to
the two functions used in this thesis. Thus. the dielectric constants they have used are easily
reproduced. Their value for diethyl maleate is in good agreement with theresults presented here.
The value for ethyl lactate differs some from the estimate here. This was expected based on the
bigger errors in the estimation of the larger values of dielectric constants. These values from
Smyth and Walls [47] and Stolarova et al. [48] are also presented in Tables 2.1 and 2.2.

18

 

 

 

 

 

 

 

 

   

 

 

 

 

 

 

1.0
, Onsagerfunction
. f(D)=-1.45+0.0037*zm . e
, r280.95
o
0.9-i-
C P
33 L
a 0.8"
l-
g l
a E
gou-
Q .
c
o
. Debyefunctlon‘
0.6-“- ¢(D)3-2.66+0.m55'1m
P r2=0.94
0.5 AJ-s:....4'444.:as.4%4...¥..¥a:.141
580 590 600 610 620 630 640 650
Emission wavelength, [nm]

Figure 2.2. f(D) and 41(0) for solvents of known dielectric constant as a function of the emission
wavelength of Nile Red (alcohols (o) and esters (x) for f(D), and alcohols (o) and esters (+) for
40(0))-

Dielectric constant

19

 

 

 

 

 

 

35
: e
30-- lesolventswilhknownD's 7‘
I i xmlculated using Onsager function
~ locelcuated using Me funcb'on
25¢
. e
_ x
o
20-: e
' B
' e
I ll
15» e
I 0 °
10 .: R E e
I a fl 8 C
i
l .. e
5-- O
O ‘ i ‘ ‘ ‘ i ‘ ‘ ‘ ‘ i ‘ ‘ 1‘ ‘ ‘ ‘ i ‘ i
580 590 600 610 620 630 640
Emisslon wavelength, [nm]

Figure 2.3. Dielectric constant as a function of Nile Red emission maximum.

20

Table 2.1. The emission maxima for Nile Red in the various solvents of know dielectric
constants. The values for the Onsager function, f(D), are calculated from Equation 2.1; for the
Debye function, tp(D), from Equation 2.2. and the dielectric constants have been obtained from
CRC Handbook of Chemistry and Physics [26]. The comparisons in the last column are: ‘ from
Smyth and Walls [47], and b from Stolarova et al. [491.

 

 

 

SOLVENT EMISSION ran) 41(0) D 15“
methanol 640 nm 0.955 0.913 32.63 32f
ethanol 635 nm 0.940 0.886 24.30 24.5”
n-pmpsnol 633 nm 0.927 0.994 20.10 20.3"
n-butanol 631 nm 0.919 0.949 17.90 17.4"
n-pentanol 632 nm 0.996 0.911 13.90 13.9”
n—hexanol 629 nm 0.991 0.904 13.30 13.4“
n-octanol 626 nm 0.961 0.756 10.30 10.4“
ethyl formats 615 nm 0.904 0.672 7.16 7.2', 7.2"
propyl romeo 611 nm 0.919 0.691 7.72 7.7”
ethyl acetate 599 nm 0.770 0.626 6.02 6.3‘. 9.0"
ethyl acetoacetate 614 nm 0.903 0.924 15.00

butyl acetate 594 nm 0.729 0.572 5.01 5.0”
methyl propionate 599 nm 0.750 0.600 5.50 5.5”
methyl butyrate 596 nm 0.754 0.605 5.90

ethyl butyrate 594 nm 0.732 0.577 5.10 5.1”

 

 

 

 

 

 

 

21

Table 2.2. The emission maxima for Nile Red in various were of unknown dielectric constants.
The values for the Onsager and Debye functions using Figure 2.2, and the respective dielectric
constants calculated from Equations 2.1 and 2.2. The comparisons in the last column are: ' from
Smyth and Walls [47], and ° from Stolarova et al. [49].

 

 

ESTER EMISSION «0) 41(0) D(f(D)) D(¢(D)) D"
ethyl lactate 636 nm 0.92 0.95 19 19 13.11r
dimethyl succinate 606 nm 0.91 0.99 7.4 7.5 7.9‘
dimethyl maleate 621 nm 0.97 0.77 11 11

dimethyl glutarate 605 nm 0.91 0.69 7.3 7.5

dimethyl adipate 602 nm 0.90 0.67 6.9 7.0

diethyl succinate 599 nm 0.79 0.65 6.5 9.6

diethyl maleate 607 nm 0.91 0.69 7.6 7.9 10', 9.5"
diethyl fumarate 599 nm 0.79 0.64 6.4 6.5 9.5‘
diethyl mutate 640 nm 0.94 0.99 24 22

dibutyl fumarate 607 nm 0.91 0.69 7.6 7.9

dibutyl l-mrtrate 633 nm 0.91 0.94 16 17

dibutyl itaconate 599 nm 0.79 0.64 6.4 9.5

dibutyl maleate 611 nm 0.93 0.72 9.3 8.6

triethyl citrate 619 nm 0.99 0.76 10 11

 

 

 

 

 

 

 

 

 

2.5 Conclusions

Nile Red emission was recorded in various solvents of known polarity, and a linear relationship
wasfoundwhentheOnsagerfunctionandtheDebyefunctionwereplottedasafunctionofthe
emission maxima of Nile Red. This relationship was used to estimate dielectric constants for
some dibasic esters, triethyl citrate, and ethyl lactate. The results agree are" with related
publications. Thus,theapprcachprovidesaconvenientexperimental methodtoestimatestatic
dielectricconstantsofsomelesscommon organic solvents.

Chapter 3
POLARITY lN BINARY SYSTEMS

3.1 introduction

ChafiuZWmexpefimufldmeflndtoesfinotedideancmnaantsfalessknwn
sdventsThesmleappmachufillbeusedheretodetaminepdantydlangesin binalysystems.
The binary systems will be studied using fluorescence swctroscOpy. A polarity sensitive
solvatochromic probe, Nile Red, will be used to indicate changes in the polarity of the system.

This chapter follows loosely a manuscript “Dielectric enrichment in binary systems containing

environmentally benign esters“ by Uusi-Penttila et al. [58].

3.2 Experimental

3.2.1 Chemicals

Ethanol (anhydrous) was from Quantum Chemical Corporation. Diethyl succinate (99%), diethyl
maleate (97%), diethyl fumarate (98%), ethyl l-lactate (98%), and Nile Red were purchased from
Aldrich. HPLC grade water from Fisher Scientific Company was used in the eXperiments. All
chemicals were used without further purification.

23

24
3.2.2 instrumentation

The equipment used for abwrption measuremts was a Perkin-Elmer Lambda 3A UV-Vls
Spectmphotometa.andmefluorescenceexpennentswsrepenonnedmaSPEx
FLUOROLOG 1681 0.22 m Spectrometer. Quartz cuvettes were used in all the experiments.

Theaccuracyofboththeabsorptionandemissionspectraverefl nm.

3.2.3 Sample preparation

AtraceamountofNileRedwasneededfortheexpeliments,l.e.10‘MofNileRedfor
absorph'on experiments and 2*10‘M for emisson experiments. For greater accuracy, Nile red X
was diluted according to a procedure described in Chapter 2.2.3. Binary mixtures were prepared
on volume basis dissolving the Nile Red first into the solvent that it was more soluble in.

3.3 Results

In Uusi-Penttila at al. [57] the absorption and emission maxima of Nile Red in pure solvents have
beendetermined. lnthesamepaperfliedielectncconstantshavebeencalculatedforpure
solvents using both the Onsager model and Debye model [39], [53]. The best linear correlation
betweentheernissionmaximaandthetwomodelswereobtainedwhenthesamplesmre
excited at their respective absorption maxima.

The same approach was used for binary systems. The emission spectra of Nile Red, excited at
absorption maxima, were taken for five binary systems: diethyl succinate-ethanol, diethyl
fumarateethanol, diethyl maleate-ethanol, ethyl lactate-water, and ethanol-water. The resultsare
presented in Figures 3.1-3.5 and in Tables 3.1-3.5. The dots represent the experimental

25
emission maxima of Nile Red and the straight line indicates the linear addition of polarities for

ideal systems.

The experimental values presented in Figures 3.1-3.5 differ considerably from the ideal linear
relation of polarities. According to Midwinter and Suppm [30], this can be caused by either of two
effects: the dielectric enrichment or hydrogen bonding. They have determined that the dielectric
enrichrrent causes a red shift in the absorption and emission maxima. i.e. a shift to higher
wavelength. and hydrogen bonding causes a blue shift, l.e. a shift to lower wavelength. They
havealsoconcludedthatinthecaseofhydrogen bondingtherernaybeathreshold value below
whichtheadditionofthesecond solventdoesnoteffectthehydrogen bonding enough tochange
the spectrum. ‘ITleseeffects have to be accounted forwhenworking with binary systems.

The first three systems in Figures 3.1-3.3 demonstrate the dielectric enrichment effect. They all
show a slight shift of the Nile Red emission maxima toward higher wavelengths. However, the
ethyl lactate-water system in Figure 3.4 and the ethanol-water system in Figure 3.5 deviate from
ideality toward lower wavelengths. The dielectric constant of ethanol is 24 [26], and in Chapter 2
the dielectric constant for ethyl lactate was estimated to be 19. Figures 3.4 and 3.5 clmny show
maneeHMlactate—eatasystemmauggadifiaenceinmedieleaflcconstantsis
significantly less ideal than the ethanol-water system. However, in both cases. the less polar
solventhasaconsiderableeffectonthehydrogen bondingofwater, andeven small amountsof

ethylladateoretharolMilcauselargecharigesinthelocalpolmtyofmesystem.

3.4 Discussion

The polarity behavior of binary mixtures including two solvents of considerably different dielectric
constants is nonideal. The polarities of the pure solvents in a binary mixture are not additive
since the polarity change is not linear. This can be accounted for either by dielectric enrichment

26
Table 3.1. Nile Red emission maxima for diethyl succinate-ethanol binary mixture as a function

of the mole fraction of diethyl succinate at 25 °C. The accuracy is :1 nm.

 

 

 

 

 

 

 

 

Mole fraction of Nile Red emission Mole fraction of Nile Red emission
diethyl succinate maximum, [rim] diethyl succinate maximum, [nm]
0.000 635 0.355 629
0.039 635 0.594 624
0.084 634 0.767 618
0.196 632 1.000 599

640 -

 

635
630
625
620
61 5
61 O

605

Nile Red emission maximum, [nm]

595

 

 

 

0.4

0.6

Mole fraction of diethyl succinate, {-1

Figure 3.1. The emission maxima of Nile Red in differing mole fractions of diethyl succinate in
ethanol. The straight line represents the ideal behavior of the system.

27

Table 3.2. Nile Red emission maxima for diethyl fumarate-ethanol binary mixture as a function
of the mole fraction of diethyl fumarate at 25 °C. The accuracy is :1 nm.

 

 

 

 

 

 

 

 

Mole fraction of Nile Red emission Mole fraction of Nile Red emission
diethyl fumarate maximum, [nm] diethyl fumarate maximum, [nm]
0.000 635 0.358 630
0.040 634 0.598 625
0.085 633 0.770 618
0.199 632 1.000 598
640 -

 

635
630
625
620
61 5
61 0
605

 

Nile Red emission maximum, [nm]

595

 

 

5% a a a a l a a a 4 l L a a l l a a a #1 a a J a
I l ' 1 I |

0.0 0.2 0.4 0.6 0.8 1 .0
Mole fraction of diethyl fumarate, [-]

Figure 3.2. The emission maxima of Nile Red in differing mole fractions of diethyl fumarate in
ethanol. The straight line represents the ideal behavior of the system.

28

Table 3.3. Nile Red emission maxima for diethyl maleate-ethanol binary mixture as a function of
the mole fraction of diethyl maleate at 25 °C. The accuracy is 1:1 nm.

 

Mole fraction of Nile Red emission Mole fraction of Nile Red emission

diethyl maleate maximum, [nm] diethyl maleate maximum, [nm]

 

0.000 635 0.361 633
0.040 636 0.601 629
0.086 636 0.772 624
0.201 635 1 .000 607

 

 

 

 

 

 

 

635

630

625

620

615

610

Nile Red emission maximum, [nm]

605

 

 

l L 1 a a l
r T

0.0 0.2 0.4 0.6 0.9 1.0
Mole fraction of diethyl maleate, [.1

P
l-

 

! a a
I

Figure 3.3. The emission maxima of Nile Red in differing mole fractions of diethyl maleate in
ethanol. The straight line represents the ideal behavior of the system.

29

Table 3.4. Nile Red emission maxima ethyl lactate-water binary mixture as a function of the
mole fraction of ethyl lactate at 25 °C. The accuracy is :l:1 nm.

 

 

 

 

 

 

 

 

 

 

 

Mole fraction of ethyl lactate, [-]

Mole fraction of Nile Red emission Mole fraction of Nile Red emission
ethyl lactate maximum, [nm] ethyl lactate maximum, [nm]
0.000 670 0.193 649
0.017 661 0.389 645
0.038 659 0.589 642
0.096 653 1.000 636
680
675
E, 670
5'
g 665
it
E 660 .
r:
.2
g 655
g e
650
'8 e
a:
g 645 e
z e
640
6:55.“ ‘r"‘L%‘ .-.....2 ts
0.0 0.2 0.4 0.6 0.8

1.0

Figure 3.4. The emission maxima of Nile Red in differing mole fractions of ethyl lactate in water.
The straight line represents the ideal behavior of the system.

30

Table 3.5. Nile Red emission maxima for ethanolwvater binary mixture as a function of the mole
fraction of ethanol at 25 °C. The accuracy is 1:1 nm.

 

 

Mole fraction of Nile Red emission Mole fraction of Nile Red emission

ethanol maximum, [nm] _ ethanol maximum, [nm]
0.000 677 0.227 650
0.032 666 0.407 649
0.068 662 0.540 645
0.1 1 2 660 0.726 641
0.164 655 1.000 636
0.439 652

 

 

 

 

 

 

 

Nile Red emission maximum, [nm]

 

 

 

 

630 . a a a a = a a a a : a a a a : 4 a a a = A a
0.0 0.2 0.4 0.6 0.8 1.0

Mole fraction of ethanol, {-1

Figure 3.5. The emission maxima of Nile Red in differing mole fractions of ethanol in water. The
straight line represents the ideal behavior of the system.

31
effect or by hydrogen bonding. The dielectric enrichment effect can be seen from Figures 3.1-3.3

where a small amount of the less polar solvent causes a much smaller change in the system
polarity than would be expected. Similarly, adding a small amount of the more polar solvent into
the less polar solvent causes a dramatic change in the system polarity. However, as can be seen
from Figures 3.4 and 3.5, the hydrogen bonding causes an opposite effect. The addition of the
less polar solvent decreases the polarity of the system considerably; whereas, adding the more
polar solvent into the less polar solvent hardly changes the polarity.

3.5 Conclusions

Five binary systems were studied using fluorescence spectroscopy. The changes in the emission
maxima of Nile Red were plotted as a function of the mole fraction of the less polar solvent. The

systems behaved nonideally. This nonideality was explained by dielectric enrichment and by

hydrogen bonding.

Chapter 4

SOLUBILITY OF L-LYSINE MONOHYDROCHLORIDE

4.1 introduction

No published solubility data were found for l-lysine monohydrochloride. Therefore, the solubility
data for l-lysine monohydrochloride in two solvents, water and ethanol, was determined within a
narrow temperature range. The solubility in mixtures of water and ethanol will also be presented.
Furthermore, the implications of these data on the purification of l-lysine monohydrochloride by

recrystallization will be addremd.

Orella and Kinlvan [42] have studied various other amino acids. Their results were presented in
Chapter 1 (Figure 1.4). They reported that adding alcohol to aqueous amino acid mixtures can
have a significant effect on the solubility. They have concluded that the effect depends on the
polarity of the side chains of the amino acid. From the previous chapter it can be seen that
adding ethanol to water affects the polarity of the system greatly. Thus, the l-lysine

monohydrochloride solubility would also be expected to decrease dramatically.

This chapter is based on parts of a manuscript titled 'In situ monitoring of antisolvent

crystallization using attenuated total reflection Fourier transform infrared spectroscopy' by Uusi-
Penttila and Berglund [55].

32

33
4.2 Solubility of l-lysine monohydrochloride in water

The solubility of l-lysine monohydrochloride in water as a function of temperature was
determined first. The procedure for solubility measurement was taken from Myerson and Ginde
[38]. Distilled water and an excess amount of l-lysine monohydrochloride were stined in a
temperature controlled sealed vessel for 24 hours. The solids were filtered, dried and weighted,
and the equilibrium concentration of l-lysine monohydrochloride was calculated. This prowdure
was repeated for different temperatures and a solubility curve was obtained for a narrow

temperature range and is presented in Figure 4.1.

Wm = 0.46 T+ 29.6 (20 °C 5 T5 45 °C) 4.1

where Wm is w-% of l-lysine monohydrochloride, and T is temperature.

As can be seen from Figure 4.1, the l-lysine monohydrochloride solubility in water is only slightly
dependent on the temperature. Within a range of 25°C (from 20°C to 45°C) the solubility
increases only from about 40 weight-% up to about 50 weight-%. In cooling crystallization,
therefore, the yield of l-lysine monohydrochloride would be very small.

4.3 Solubility of l-lysine monohydrochloride in ethanol

Similar control experiments were done to determine l-lysine monohydrochloride solubility in
ethanol. Within the temperature range from 15 “C to 50 °C the solubility varied between 0 and 2
weight-96. Therefore, the I-lysine monohydrochloride solubility in ethanol was assumed to be

negligible in the crystallization experiments.

52

 

sor-

IrIIII'IIII

II—fIIII

Weight-9t» of llmhc in water

 

 

p
38 a a L_L_L a a a a l a a a a I a a a a l a a a a I a a as
l I I I T

 

 

15 20 25 30 35 40 45
Temperature, [‘C]

Figure 4.1. Solubility of I-lysine monohydrochloride in water as a function of temperature.

35
4.4 Solubility of l-lysine monohydrochloride in ethanol-water mixtures

The previous procedure was also used for ternary l-lysine monohydrochloride-ethanol-water
systems. The solubility of l-lysine monohydrochloride in various binary mixtures of water and
ethanol was determined at 303 11 K Figure 4.2 shows the solubility curve. During the actual
antisolvent crystallization experiments the amount of water in the crystallizer was chosen to be
kept constant. Therefore, this solubility curve was normalized based on the amount of water in

the crystallizer. The following equation was fitted from the experimental data:

fl _ 0.33

_ m — 0.02 4.2
mw 0.44 + —2—

where m is the mass of l-lysine monohydrochloride, mm is the mass of water. and m. is the mass

of ethanol.

4.5 Discussion

Figure 4.1 shows that l—lysine monohydrochloride solubility in water is not very temperature
sensitive. Therefore, a cooling crystallization is not efficient for l-lysine monohydrochloride
purification. It should also be noted that an amino acid should not be exposed to very high
temperatures. However, based on the large decrease of the l-lysine monohydrochloride solubility
in the presence of ethanol, an altemative method for I-lysine monohydrochloride purification is
proposed: antisolvent crystallization. This method easily produces yields of 90 w-%. The other

advantage is the possibility of doing the crystallization at a low temperature.

0.8

 

Solubility in water = 0.729
‘—

llmhc in solvent, [kg llmhclltg water]

Solubility in ethanol
is about 1 w-%

 

 

Ethanollwater, [kg ethanollltg water]

Figure 4.2. Solubility of l—lysine monohydrochloride in various binary mixtures of water and
ethanol at 303 i 1 K

 

37
4.6 Conclusions

Solubility data for l—lysine monohydrochloride in water, in ethanol, and in mixtures of ethanol and
water were determined. It was concluded that due to the minimal temperature dependence of
l-lysine monohydrochloride solubility in water, cooling crystallization is not an effective way to
purify l—lysine monohydrochloride. However, adding ethanol into an I-lysine monohydrochloride-
water system decreases the solubility of l-lysine monohydrochloride in water significantly. This

effect can be exploited in the purification of l-lysine monohydrochloride.

Chapter 5
IN SITU MONITORING OF ETHANOL AND L-LYSlNE MONOHYDROCHLORIDE

CONCENTRATION IN A BENCH SCALE CRYSTALLIZER WITH ATR FTIR

5.1 Introduction

Antisolvent crystallization is a commonly used crystallization method in pharmaceutical industry.
Therefore, purificationofl-lysine monohydrochloride usingethanolasan antisolventwaschcsen
to simulate a pharmaceutical crystallization process. The process was monitored udng
attenuated total reflection Fourier transform infrared (ATR FTIR) spectroscopy. The unique
configuration of the OIPPER' 210 deep immersion probe made it possible to monitor the liquid
phase in situ in the crystallizer. The applicability of this method for monitoring batch cooling
crystallization has been demonstrated by Dunuwila et al. [13] and by Dunuwila [12]. However. the
peak shift approach that they used was not applicable for this system. Therefore, the infrared
peakintensitychangeswereusedto monitorthesystem.

This chapter presents the calibration curves that correlate component concentrations with
infrared peak intensities. The infrared spectrum of a three component system is very complex.
Son'leofthechalacteristicpeaksfortl'lesolute, solventandantisolventaresoclosetoeach
otherthattheyappearasonepeakintheinfrared spectrum. Therefore, derivativespectroscopy
was Md to resolve the overlapping infrared peaks into the individual component peaks. Special
carewastakentofind peaksthatarenot influenced bytheothercomponentsinthecrystallizer.

39
This chapter is based on parts of a manuscript titled “In situ monitoring of antisolvent

crystallization using attenuated total reflection Fourier transform infrared spectroscopy“ by Uusi-
Penttila and Berglund [55].

5.2 Experimental

5.2.1 Materials

The antisolvent crystallization system of choice for the current experiments was recrystallization
ofl-lysinemonohydrochloride. Thesolventwaswaterandtheantisolverltwasethanol. L-lysine
monohydrochloride (USP grade) was purchased from Kyowa Hakko Kogyo Company. Ethanol
(anhydrous) was obtained from Quantum Chemical Corporation. Distilled water was used in all
experiments. The ethanol was used without further purification and l-lysine monohydrochloride
was recrystdlized before use.

5.2.2 instrumentation

A one liter, jacketed crystallizer is kept at a constant temperature (30°C) for the calibration
measurements. The spectrometer is a Perkin-Elmer 1750 Fourier transform infrared
spectrometer, and the ATR element used is a DIPPER° 210 deep immersion probe with an
AMTIR’ ATR crystal from Axiom Analytical.

40
5.3 Attenuated total reflection Fourier transform infrared spectroscopy

5.3.1 Introduction to ATR FTIR spectroscopy

in attenuated total reflection Fourier transform infrared (ATR FTIR) spectroscopy the inframd
radiation is directed into the interface between the reflectance element and the sample in such
an angle that all the radiation is reflected back. in order for this to be possible, the reflectance
element has to be of optically denser material than the sample is, and the incident angle of the
infrared radiation has to be larger than the ratio of the reflective index of the sample over the
refractive index of the reflectance element. According to Colthup [9] and Mirabella [32], despite
the total reflection of the radiation at the interface, there is an evanescent wave that persists into

the sample as illustrated in Figure 5.1.

Two most common variables measured with infrared spectroscopy are the transmittance, T, and

the absorbance, A [19].

T=10"”’° 5.1

and

A=-logT=abc 5.2

where a is absorptivity, b is pathlength, and c is concentration.

In attenuated total reflection spectroscopy an effective pathlength is used. This pathlength is

defined as the product of the number of reflection points in contact with the sample on the

surface of the reflection element, N, and the depth of penetration, dp, (Coetzee [8]).

b=~¢ 53

41

 
  

 

reflectance n,
element
sample n2

o'=:c—- penetration

Flgure 5.1. Total internal reflection.

42
The wavelength dependent penetration depth of the evanescent wave is obtained according to

lngle and Crouch [19] and Miiller and Abraham-Fuchs [33] from

{2”[5i"20;(as/nc)2]}1'2

where 1c is the wavelength in the ATR crystal, 0 is the incident angle of the infrared radiation, 71s

 

up = 5.4

is the refrective index of sample, and Us is the refractive index of the ATR crystal.

The penetration depth as a function of the wavelength for both pure water and pure ethanol was
calculated using Equation 5.4. The results are presented in Figure 5.2. It can be seen that there
is hardly any difference from solvent to solvent. Therefore, the differences in the spectra cannot
be attributed to the change in the penetration depth because of the change of solvent. Also, the
penetration depth is so small that there is no interference from the crystals. Therefore, it can be
safely assumed that the spectrum reflects the liquid phase conditions. This attribute is why this

method is excellent also for heavy slurn‘es.

According to Colthup [9] and Mirabella [32], the evanescent wave has the same wavelength as
the infrared radiation and decays exponentially in the optically less dense medium. The wave
interacts with the reflected radiation at the interface by either reducing it or increasing it, and
thus, producing an infrared spectrum of the sample. Figure 5.3 shows the transmission spectra

for ethanol, water and aqueous l-lysine monohydrochloride.

43

3.5

 

3.0 -'

2.5 5

    

2.0 r

1.5‘

Iilir‘TITVVI

Penetration depth. [urn]

1.0‘

Irfii

 

0.5 .1

 

 

 

00 anLnlaaaalnnLJlnaanlna
. r ' 1 '

700 1200 1700 2200 2700 3200 3700

Wavenumber, [cm“]

all.
.1

Figure 5.2. Penebetiondepthoftheevaanwaveforpurewaterandpureeflrandasa
function of wavenumber.

 

 

anon zone 1500 1030 1&0
W. [om-1]

Wmonohyrroclioflde

 

W, [em-1]

 

 

 

 

Wow-rumba. [urn-1]

Figure 5.3. Infrared spectra of water, aqueous l-lysine monohydrochloride. and ethanol at 30 °C.

45
5.3.2 Derivative spectroscopy

It can be seen from Figure 5.3 that several of the peaks within the region from 1300 cm'1 to
1700crrt'1 are overlapping. The ethanol peaks at 1042 crn'1 or 1088 crn'1 can be used to
determine the ethanol concentration. To determine the l-lysine monohydrochloride concentration

in the solution the overlapping peaks have to be resolved. Derivative spectroscopy was used.

Derivative spectroscopy enhances the fine structures of the spectrum [5], [8], [18], [27], [28]. For
two peaks that are so close to each other that they appear as one peak in the infrared spectrum,
the second derivative of the spectnrm can be used to resolve these peaks into two separate
peaks. Figure 5.4 demonstrates this. The first derivative crosses the zero baseline at the position

of the peak maximum of the original spectnrm. The second derivative has a minimum at the
peak position.

Cahill [5, 8] points out that due to the linear nature of derivation operation, the peak intensities of
the second derivative spectrum follow Beer's law just like the intensities of the original spectrum.
Therefore, the peak intensities of the second derivative spectra can be used to determine the
solution concentrations. The main problem with this data treatment method is the fact that the

signal-to—noise ratio has to be low for this method to work.

 

occafiomfi

3.0 5 0232.3“. «or

8.3 x. 2:25..

tau

50

30

10

Figure 5.4. First and second derivatives for two overlapping

Gaussian bands (Cahill [5]).

47
5.4 Calibration of infrared spectra and component concentrations

5.4.1 Calibration curve for ethanol concentration

Figure 5.33howedthetransmisdonspectraoftln pure components in the crystallizer. Eitherone
ofthe two main peaks in the ethanol spectrum, i.e. 1042cm" or roeacm". can be used to
detennimflwethanlconcmhflm.3eceuseofflnbdtasendt~hyfleC-Ostretching
vibration peak at 1042 cm‘1 is chosen. Using Equation 5.2 the transmission spectrum was
WiMoanabeapfimspecuummdflieefltmdconcenUahonunscaiculated.

To prevent disturbance due to baseline fluctuations from experiment to experiment, a relative
abeorbanceA,ispreferred.Therefore.thefollowing formulaisusedto relatetheethanol
concentrationandtheinfraredmeesurement

Ar ‘-' Area /Amo 5-5

whereAroa istheabsorbanceat 1042 cm", andAmo istheabsorbance at 1110 cm".

The calibration curve for ethanol concentration is presented in Figure 5.5. In the crystallization
experiments, the amount of ethanol varies between 0 and 61 w-%. Therefore. a linear correlation
isusedforthereiativeabeorbanceandthemaaofethanol.

 

A,.—. 1 - 0.32._’"o_ 5.6
m,+ mw
orsolvingforthemassofethanol
(1 - A,) mw
”'°= A,- 0.69 5-7

wherem.isthemaesofethanol,andm.isthemassofwater.

Relative absorbance, A r. [-l

48

 

1.00

0.90 1-

0.80 -r-

 

0.70

 

o with l-lysine monohydrochloride
e no l-lysine monohydrochloride

 

 

Ethanol, m. [(m. + m.), [w-%]

Figure 5.5. Relative absorbance, A,, as a function of ethanol weight percentage.

49
5.4.2 Calibration curve for l-lysine monohydrochloride concentration

The determination of the concentration of l-lysine monohydrochloride is more complicated.
L4ysinemonohydrochloridehasavaietyoffunctionalgroupsavailable. However,ascanbe
seen from Figure 5.3. the l-lysine monohydrochloride peaks fall in the same region as peaks
from ethanol. To overcome this problem derivative spectroscopy was chosen to resolve the

WWW Pm

Foraccuratesdtneconcenhationmeasurementsitisimportanttoflndapeakthatisnotaffected
byheprwnceofflntwoedvents.Theaeconddefivaflmspachumonysine
nmcnyomarrcndewaeaudred.auacemoxyratemm1411cn"msnaafleaedoym
www.mismsvenfledudngmudvaimemdyfismecwuafiombamm
ma1411un",uyammamydmcnrcndecormueumweumaconcenmmm
calculated.ThecorrelationcoefflcientsarestmninTable5.1.ltisclearthatthereisno
correlationbetweentheethanolconcentrationandthebandat1411cm".Therefore.thispeak
waschosenfaflredetaminaflmofflielwmmondwdmchbndeconcenhafiminflie
crystallizer.

Table 5.1. Conelation coefficients for multivariate analysis of the band at 1411 cm", l-lysine
monohydrochloride concentration and ethanol concentration.

 

 

 

 

 

Ethanol concentration and Ethanol‘concentration L-lysine monohydrochloride
l-lysine monohydrochloride and the band 1411 cm" concentration and
concentration the band 1411 cm'1
-o.51 -o.4s 0.93

 

 

 

 

 

 

 

18
o [I
16 l . o F
i e noethmol . o
r - ° 0
14 ; 0 ethanol ! .
J i 00
12 5. —Linear(all) Jl .:
£10 0 o
a 0
g ° 0 '
6 Q 0
4
O
2
o- H: *i“‘L% J‘i*+“i‘*tt**4 : .L
o 5 10 15 20 25 30 35 40
L-lysine monohydrochloride. rnL ItrnL '0 me). [W-tt]
Figure 5.8. Difference of the second derivative peaks at 1411 crn‘1 and 1432 crn‘1 as a function

ofthe vaight fraction of l-lysine monohydrochloride.

51
To eliminate the baseline fluctuations the following difference. D. was used for the calibration

D = [1411 - 11432 5.8
wlm Imr is the intensity of the second derivative at 1411 cm". and 11.32 is the intensity of the
second derivative at 1432 cm“.

Figure 5.8 shows the calibration curve for I-lysine monohydrochloride concentration. The
fdlmvingconelationcanbeobbinedforthedifferenceandl-Iysine monohydrochloride

 

 

mL
D = 43.4 5-9
(mL + m, + mw)
andsolvingfortheamountofl-lysine monohydrochloride
D m + m
mt = ( ’ W) 5.10

43.4—0
MieremListhemassofl-Iyslnemonohydrechloride,nuistiremassofethanolandmwisthe

massofwater.

5.5 Conclusions

In this chapter the calibration curves for ethanol concentration and l-Iysine monohydrochloride
conwnhafimminflamdspedrawaegenemted.0envafivespedmscopyvasusedto
calibrate l-lysine monohydrochloride concentration. Within the operating range both mllbration
curvescanbeflttedtoalinearapproximation.Thesecurveswereused in Chapters7and8to
determine the effect of the antisolvent addition rate on various parameters.

Chapter 8
EXPERIMENTAL ARRANGEMENT AND CRYSTALLIZATION PROCEDURE

8.1 introduction

Thisdtamaslmflieexpenmentdmngmiunfunnnhodngfltewnflcafimdmsim
monohydrochloride using attenuatedtotal reflection Fouriertransform influed spectroscopy. The
promdure for the crystallization and the product twatment will also be explained.

6.2. Materials

The antisolvent crystallization system of choice for the current experiments was recrystallization
ofI-lysine monohydrochloride. Thesolventwaswaterandtheantisoiventwasethanol. L-Iysine
monohydrochloride (USP grade) was purchased from Kyowa Hakko Kogyo Company. Ethanol
(anhydrous) was bought from Quantum Chemical Corporation. Distilled water was used in all
experiments. The ethanol was used without further purification. The crystdlized l—lysine
monohydrochloride was reused in the next batches.

52

 

IEIrrnrt

_ . Spectrometer
A
LBJ o
IE]

 

 

 

 

 

 

 

 

—l I

Figure 8.1. The experimental arrangement: A temperature controlled antisolvent
weir. 3 pump; c jacketed crystallizer with a marine type impeller; D Dipper-210
dup immersion probe; E FTIR spectrometer, F computer.

6.3 instrumentation

Figure 6.1 shows the experimental arrangement. A one liter, jacketed crystallizer is used for the
crystallization. It is kept at a constant temperature (30 11 °C) with Brinkmann's RC6 LAUDA

water bath. The crystallizer is equipped with a marine type impeller.

The antisolvent reservoir is placed in a waterbath and kept at the same temperature as the
crystallizer. A peristaltic pump is used to add the antisolvent at rates of 5 mllmin or more. A
syringe pump is used for smaller addition rates. The capacity of the syringe pump was one fifth
of the total amount of the added antisolvent. Therefore, for the small addition rates the
antisolvent addition had to be stopmd momentarily five times during the crystallization to fill up
the syringe. This did not have significant effect on the results.

The spectrometer is a Perkin-Elmer 1750 Fourier transform inframd spectrometer. The ATR
element is a DlPPER° 210 deep immersion probe with an AMTIR’ ATR crystal by Axiom
Analytical [1]. AMT lR’ is a mixture of Arsenic, Selenium and Germanium glass. Its refractive
index is 2.5 at 1000 cm“, and the spectral range is 11,000-750 cm" [7]. The DIPPER’ 210 is
dipped straight into the solution in the crystallizer and a program is run to take the spectra of the

solution at desired time intervals.

8.4 Crystallization

The crystallization is started by preparing a saturated aqueous I-lysine monohydrochloride
solution at 30 °C. From Chapter 4, the solubility of l-Iysine monohydrochloride at 30 °C is
42 w-%, is equivalent to 109 g of l-lysine monohydrochloride and 150 ml of distilled water.

Ethanol at 30 °C is added at a constant addition rate using either a peristaltic pump or a syringe

55
pump. To obtain good yields the crystallization is continued until 300 ml of ethanol has been

added.

The product is filtered using a water jet pump. The filter used is a Lab Glass fritted disc funnel
with a pore size of 10—15 pm. The crystals are washed twice with 150 ml of ethanol and dried in

an oven overnight at 50 °C.

6.5 Sieving

The crystal size distribution of the dried product is determined by sieving. The procedure for the
sieving and the length of the sieving time are according to lrani and Callis [20] and ASTM
Standard No. 4478 [29]. The sieving time is long enough when the weight of the sieve fraction
on a sieve during the sieving does not change more than 1 w-% within the specified period of

time.

The crystals are sieved using two sets of sieves. First, the larger particle sizes (> 0.425 mm) are
divided into eight sieve cuts. The adequate time is 25 minutes using a W.S. Tyler Model RX-86
sieve shaker and a W.S. Tyler (20 cm diameter) sieve set. The bottom fraction (< 0.425 mm) is
sieved for another 20 minutes using a Scientific Industries Vortex-Genie 2 shaker and W.S. Tyler
(5 cm diameter) sieve set. This anangement divides the product further into six fractions. The

sieve cuts for the sieve series are presented in Table 6.1.

Table 6.1. Two sieve series used for determining the product crystal size distribution.

 

 

200mdiameter sieves 5cmdiameter sieves

1.18 mm 0.355 mm
1.00 mm 0.250 mm
0.85 mm 0.180 mm
0.71 mm 0.125 mm
0.60 mm 0.090 mm
0.50 mm

0.425 mm

 

 

 

 

Chapter 7
L-LYSINE MONOHYDROCHLORIDE KINETICS

7.1 introduction

Determining the growth and nucleation kinetics for a batch experiment is tedious and time
consuming. Usually this is done from continuous experiments. The most common correlations

used for growth and nucleation kinetics are empirical power-law expressions (Wey [59]).

G=kgAc° 7.1

so = k" Ac” (mr)' 7.2

where G is the growth rate, kg is the growth rate constant, Ac is the supersaturation, g is the
growth order, 80 is the nucleation rate, kN is the nucleation rate constant, 0 is the nucleation

kinetic order, m7 is the suspension density, and j is the exponent of suspension density.

This chapter shows the results of nucleation cell experiments. These experiments were used to
determine whether the growth rate of an individual l-lysine monohydrochloride crystal is size-
dependent or size-independent. Then seeded experiments were used to estimate the overall
growth rate in the crystallizer. Finally, the nucleation rates were iterated from the population

balance using the moment equations.

57

58
7.2 Crystal growth

7.2.1 Growth rate dependency on crystal size

Two sets of nucleation cell experiments were concluded to determine whether the growth rate of
l-lysine monohydrochloride crystals is size-dependent or not. The first set of experiments was
done using only water as the solvent, and the second set of experiments had a mixture of water
and ethanol as the solvent. The experimental anangement is from Shanks and Berglund [46] as

presented in Figure 7.1. The nucleation cell is temperature controlled at 30 :l: 1 °C.

A supersaturated solution of l-lysine monohydrochloride is made by adding an excess amount of
I-Iysine monohydrochloride into the solvent in a beaker. The solution is heated until all I-Iysine
monohydrochloride is dissolved. and then cooled slowly down to 30 °C. The nucleation cell is
filled with supersaturated solution. A parent crystal that has been glued to a rod is placed into the
supersaturated solution. The nucleation cell is closed and the parent crystal is dragged over a
glass plate. This causes nucleation. The growth of nuclei is followed by taking photographs of the
nucleation cell through a microscope. The growth rate of the I-Iysine monohydrochloride crystals

is then determined from the photographs.

To determine the growth rate of the crystals a spherical equivalent diameter was used.

A,,,,=wr=zzr2 . 7.3

where Ag, is the area of the rectangular crystal and a sphere of an equal size, w is the width of

the crystal, I is the length of the crystal, and r is the radius of a sphere.

59

 

 

 

top view

 

 

 

 

 

 

 

 

 

 

side view

Figure 7.1. The temperature controlled nucleation cell. a rod with parent crystal glued to it. b
glass slide and the rod it is glued to, c temperature element, d chamber for constant temperature
water, e chamber for saturated solution. [46]

60
This equation is solved for the radius of the sphere, r, which is taken as size of the crystal. Figure

7.2 shows the spherical equivalent size of the crystals grown in the nucleation cell experiments
as a function of time. For the ethanol-water experiments nine different crystals from three
experiments were measured, and for the water experiments eleven crystals from two
experiments were measured. All the measured crystals follow a linear trend. The slope of this
trend is the growth rate. Figure 7.2 also clearly demonstrates that there is no difference in the
individual growth rates from one experiment to the other. Therefore, the presence of ethanol
does not seem to effect the growth kinetics significantly. Also, since the slope is constant, the

growth rate of l-lysine monohydrochloride crystals is size-independent.

7.2.2 Overall growth ran from seeded experiments

The overall growth rate for the l-lysine monohydrochloride crystals in the batch experiments was
estimated from nine seeded experiments. The crystallizations were run at three different lengths
of time for three different antisolvent addition rates. Figure 7.3 shows the actual crystal mass
retained on each sieve at 20 mllmin ethanol addition rate after 2, 5, and 10 minutes of
crystallization. The sieve cuts that were smaller than 550 um were ignored and the weight mean

size of the grown seed crystals was determined using the following equation.

me = 2(Lr Wr) / Z M 7.4

where L,- is the mean size of a sieve out i in um, and w,- is the mass of crystals on sieve i in

grams.

The weight mean sizes were plotted as a function of time and the following overall growth rates

were obtained: 7.1 pmlmin for 1 mllmin ethanol addition rate, 10 uni/min for 5 mllmin, and

61

 

 

 

 

 

 

 

3.5
i
r X

3.0 -- X
"E" .
E. ; X
g 2.5-- x
E:
0
§ 1 0 °
5 2.0+ °
w . 0 x0
E o 0
'6 o x o
e- " O
i ‘* at: °
3 I 0 x
8- t 8 g o owithethanol
0 o
g 1'0 T x a 8 o o o xwithoutethanol
g . x" I <9 o x L
a. _ o o X
00 _ §E o X

05 9:

o
0.0 ' a a a e e % a a a a L ‘L j a n L a :
0:00 0:30 1:00 1:30

Time, [inmin]

Figure 7.2. Spherical equivalent diameter for single crystals in a seeded nucleation cell. The
experiments were done at 30 °C and the ethanol/water ratio was 1.2:1 on volume basis.

62

 

\\\\\\\\\

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\k

 

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

 

V\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

 

\\\\\\\\\\\\\\\k

V\\\\\\\\\\\\\\

V\\\\\\\\\\\\\\\\\\\\

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\.

 

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

 

 

I2 min

 

05min

I10 min

 

 

V\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

 

.\\\\\\\\\\\\k
I

\\\\x
I

v
_

 

 

 

18

16-

14‘

12‘

i
d
d
d

10‘
8
6
4
2
0

.2 see:

 

32 77 153 215 303390463550655735925109013m1400

0

Averagesizeofdtesievecutfitm]

minutes of crystallization. The experiment was done at 30 °C and the ethanol addition rate was

Figure 7.3. Mass of I-lysine monohydrochloride crystals retained on the sieves after 2, 5, and 10
20 mllmin. The average seed size was 550 um and the mass of seeds was 10 g.

63
31 tun/min for 20 mllmin addition. These rates are plotted in Figure 7.4 and the following

estimate for the overall growth rate as a function of ethanol addition rate is obtained
G=1.28‘Q.+4.9 7.5
where the growth rate, G, is expressed in urn/min, and ethanol addition rate, Q... in mllmin.

As will be shown in Chapter 8, in the antisolvent crystallization of l-lysine monohydrochloride the
supersaturation is only a function of the concentration of ethanol in the crystallizer. Rather than
presenting the growth rate as a function of supersaturation. all the constants and the variables
that depend on the ethanol addition rate are grouped together. Thus, the overall growth rate is
presented as a function of the ethanol addition rate only.

7.3 Nucleation

To estimate the nucleation rate for the i-lysine monohydrochloride purification, the crystal size
distributions from the seeded experiments were compared with the unseeded experiments.
Figure 7.5 shows the difference in the mass of nuclei in the seeded experiments and unseeded
experiments. it can be seen that the amount of small crystals is significantly smaller in seeded
experiments. This smaller amount can be attributed to 0stwald ripening. In that, particles in a
suspension with different particle sizes dissolve at different speeds depending on the particle
size. The small particles dissolve and deposit on the bigger particles. This preferential dissolution
decreases the amount of fines in the product and increases the average crystal size of the

product (Myerson [36]). Thus, the nucleation rate is estimated from unseeded experiments.

 

 

G=1.28'Q.+4.9

 

 

15-

Growth rate, [um/min]

104

 

54

 

 

 

1 l j I L L l
1 r I r I 1 '

0 2 4 6 8 10 12 14 16 18 20
Ethanol addition rate, [mllmin]

Figure 7.4. Estimate for the overall growth rate of I-lysine monohydrochloride as a function of
ethanol addition rate from seeded experiments.

65
Using the previously obtained overall growth rate and experimental suspension density. the

nucleation rate can be estimated from the population balance [44], [59]

sow) + 6(GnV) _ 0
at EL _

7.6

 

During the l-lysine monohydrochloride purification the amount of solvent (water) is kept constant
and only antisolvent (ethanol) will be added into the crystallizer. Therefore, this set up is a
semibatch system, and thus the working volume of the system varies with time. According to
Randolph and Larson [44] and Wey [59], in this case the population balance can be expressed
on the basis of the total operating volume of the crystallizer rather than the working volume.
However, since l-Iysine monohydrochloride is practically insoluble in ethanol, ethanol was chosen
be treated as an inert that does not effect the volume in the crystallizer. Thus, only the volume of

water will be considered in these calculations.
:1 = nV 7.7

Now, the population balance will be

:9: . LEG) -
5t é’L

0 7.8

This equation will be solved using the definition of moments

m,- = I: fiUdL 7.9

66
Multiplying by L’ and integrating over dL [59], the population balance will be expmd as

dt + fr. aL(Gn)dL _ 0 7.10

and for size-independent growth the first four moments will be

at = HOG = 3,, 7.11
93‘“. = meG 7.12
dm2
— = 2 6 7.13
dt mt
_dm3 = 3m2c; 7.14
dt

Using the definition of a derivative and the crystal suspension density. mr. from Wey [59] and
Randolph and Larson [44]

mr = kvpcm3 7.15

the nucleation rate in Equation 7.11 can be iterated by estimating the nucleation rate behavior
during the crystallization. Figure 7.6 shows the estimated nucleation behavior during the
crystallization for various ethanol addition rates. The initial nucleation rates are of the same

order of magnitude as the nucleation rates published in Mina-Mankarios and Pinder [31]. This is

67
the only related article to publish any kinetic data. Figure 7.7 shows the results for 1 mllmin

ethanol addition rate.

7.4 Conclusions

Nucleation cell experiments demonstrated that the growth rate of l-lysine monohydrochloride is
sizeJMepeMan.TheexpennentsalwshovedmmmeaddmmofethanddoesMhavea
significant effect on the growth rate. Seeded experiments were used to obtain an estimate for the
overall growth rate in the crystallizer. The growth rate was further used with the population
balance to iterate a prediction for the nucleation rate.

68

 

 

 

 

 

30
25 J: r
_ ; Iseeded(27.6g)
i 9 g
t Dunseeded (47.7 g) I
E 201_
£7 .
8 I-
g h
u 15 -
b
o
I
I
a I-
2 10 di-

01
L'

 

 

 

 

 

 

 

0 31.5 76.5 152.5 215 312.5 30 4625
Particle size, [pm]

Figure 7.5. Crystal size distribution of the fines after 30 minutes of l-lysine monohydrochloride
crystallization at 30 °C. Antisolvent was added at 1 mllmin and the seeds for the seeded
experiments were the 500-600 pm sieve fraction.

69

 

 

 

 

 

1.8813
x . .
x le1mllmin l
1.6E+13 x X :o1mllmin ;
foSml/min i
lx30mllmini
5:: 1.4E+13 - ‘
”I
E. o
i r:
",2 1.ZE+13 U r:
8
5
E 1.0E+13
.2
U
3
2 8.0E+12
6.0E+12 ‘° - - .......................
4.0E+12r44‘4i"s'% 14: . :~--§n--
0 50 100 150 200 250
Ethanol, [ml]

Figure 7.8. Predicted nucleation behavior during crystallization.

 

70

 

 

 

 

 

 

60
I . t
.- 0. o
l. o.
50" O. ...
.08. e.
o o
, 8
H 40-“- 008:
.3 t 00 e
e: ' co
5 r o
E
u 30.: ..
h
o t -,
re (_ .
I e
. e e
g 201*- .‘ . . ” .
P . . ' o 1mllmin
o
. ._‘ , 0. . e 1mllmin
- . e . ‘ _P .
10 r. . . redrcted
. o
r .’o
. I
o A a L : a 4 1 l 4 +4 1 e L a : A a 1 IL 1 L
0:00 1:00 2:00 3:00 4:00 5:00
Time, [h:min]

Figure 7.7. Experimental and predicted yield for unseeded experiments using the predicted
nucleation and growth rates.

Chapter 8

INFLUENCE OF ETHANOL ADDITION RATE ON PARTICLE SIZE DISTRIBUTION

8.1 Introduction

There are very few studies dealing with the fundamentals of antisolvent crystallization, even
thoughitisacommonlyusedmethodforbothpurificationandseparationinthefoodand
pharmaceutical industries. In spite of the importance of this process. current methods for
monitoring the influence of the antisolvent addition rate on the final crystal size distribution are
lacking. In Uusi-Penttila and Berglund [56] it has been shown that the dependence of the crystal
size distribution on the antisolvent addition rate is significant.

In this chapter the effect of the antisolvent addition rate on the crystal mass. supersaturation,
weight based average crystal size, and crystal size distribution is addressed. Attenuated total
reflection Fourier transform infrared spectroscopy was used for the in situ monitoring of the

system.

This chapter is based on the article titled “Spectroscopic monitoring of environmentally benign
anti-solvent crystallization” by Uusi-Penttila and Berglund [56] and on a manuscript titled “In sifu
monitoring of antisolvent crystallization using attenuated total reflection Fourier transform
infrared spectroscopy" by Uusi-Penttila and Berglund [55].

71

8.2 Crystal mass

The antisolvent addition decreases the solubility of the solute in the solvent significantly. This
indicates that the yields obtained using this crystallization method should be very high. The yield
increasesfrombarely10%uptoabout90%astheamountofethanolincreasasfrom0.1t02
grams per gram of water. The 90 % yield is much better than yields achieved with other
methods. such as cooling crystallization.

Using the calibration curves presented in Chapter 5, the composition of the liquid phase in the
crystallizercanbemonitoredinsitu.Themassofethanol inthecrystallizerisobtained from
Equation 5.4

 

m, = 5.4

and the mass of dissolved I-lysine monohydrochloride from Equation 5.7

D(m, + m)
43.4 - D

 

mL = 5.7

Thus the ane monohydrochloride mass balance can be used to determine the amount of

solids in the crystallizer at any time during the crystallization.

m. = m, - mL 8.1

wmre mu, is the initial mass of l-lysine monohydrochloride, and m is the mass of l-lysine

monohydrochloride at time t. Figure 8.1 demonstrates the increase in the crystal mass, ms,

during crystallization experiments at different antisolvent addition rates.

73

 

 

 

 

 

 

 

 

o1 mllmin
e1 mllmin a
_ I 5 mllmin
3?
E. x 30 mllmin
2’
5
I
a
0
I-
o
2 8 g
.2 e
>- . 8 3 o o
O . O
O 8 8 . o O
e8398.8
9309'
1:30 2:00 2:30 3:00

Time, [hzmin]

Figure 8.1. The effect of ethanol addition on the increase in crystal mass during the l-lysine
monohydrochloride crystallization at 30 °C.

74
Since the trends for the increasing mass are linear, the following equation was obtained to

predict the amount of solids in the crystallizer.

m3=0.917 kychQt 8.2

where Irv is the volume straps factor assuming spherical particles, pc is the density of I-Iysine

monohydrochloride. Q. is the antisolvent addition rate, and t is time.

8.3 Supersaturation

ATR FTIR data can be used to monitor the level of bulk supersaturation in the crystallizer. It has
beensuggestedthattheoptimal operatingmodeforabatchcrystallizermuldbetousea
constant level of supersaturation [34]. This is especially true for antisolvent crystallization where
controlling the excessive nucleation due to the antisolvent addition is one of the biggest

problems.

The relative bulk supersaturation in the crystallizer during an experiment was calculated from
infrared and solubility data using the following equation

C - C m - m
a' = ———L L” = ———L L” 8.3
CL.” mL.oq

where mt is obtained from Equation 5.7 and mm, is solved from the solubility equation 4.1 .

75

 

 

 

 

or ['1

 

 

 

 

 

O. ['1

 

 

0.0 l 4 a L i a a a a i a a a a 11 a e a : a a a a J. |
0 50 100 150 200 250 300
Ethanol, [ml]

 

 

—x— 30 mllmin

 

 

 

or ['1

I l

 

d.

0.0 I a a a a a a a I
150 200
Ethanol, [ml]

Figure 8.2. Relative bulk supersaturation for l-lysine monohydrochloride purification as a
function of added ethanol for various ethanol addition rates.

76

“QatTAva
0.18+$Q°t
Avvw

 

mm, = 0.13 8.4

 

where p. is the density of ethanol, Q. is the volumetric ethanol flowrate, t is the time, pi, is the
density of water, and V, is the volume of water. The profiles for the relative bulk supersaturation
using various ethanol addition rates are shown in Figure 8.2.

As can be seen from Figure 8.2, monitoring the bulk supersaturation exposed an interesting
feature of the l-lysine monohydrochloride purification system. The effect of the ethanol addition
issostmrgflutthebulksupematurafimdependsmlymflieethandconcenhafimandhardly
on the ethanol addition rate. However. it should be kept in mind that the local supersaturation at
the point of the ethanol addition can be very different from the bulk supersaturation.

8.4 Weight based mean crystal size

The weight based mean crystal size, Lem, is determined from the sieved crystal size distributions

using Equation 7.4.

Lm=Z(Lin)/2Wi 7-4

In Figure 8.3 the weight based mean crystal size is plotted as a function of ethanol addition rate.
The following correlation is found from the data

L..." = 800 exp(-0.022 Q.) + 40 8.5

where Lem is the weight based mean size in mm, and Q. is the volumetric ethanol addition rate in
mllmin.

77

 

1000

L...,=000*exp(-0.022*Q.)+40

Weight mean size. [ um]

 

 

n+1 lglJlLllll
T I

0 50 100 150 200 250 300
Ethanol addition rate, [mllmin]

Figure 8.3. Weight mean size of crystals as a function of ethanol addition rate at 30 °C. The
initial solution contained 42 w-% of l-lysine monohydrochloride in water.

78
8.5 Crystal size distribution

Antisolvent crystallization produces very narrow crystal size distributions. The crystal size
distributions for sixteen experiments were determined by sieving and further analyzed
statistically according to the procedure in Mullin [34]. Since a 2‘“ sieve series was used. an
arithmetic mean was used for a mean size of each sieve cut. The standard deviation for the
distribution was mlculated from

= Lens - L1896

8.6
2

 

wl'lere Lent is the particle size at 84% on the cumulative undersize percentage plot. and Lists is
the particle size at 16% on the cumulative undersize percentage plot.

The coefficient of variation was calculated from the standard deviation and weight mean size

-.i.
..me

CV 8.7

An average coefficient of variation for sixteen experiments was 0.526. This is remarkably close
to 0.52, the coefficient of variation for a Gaussian distribution. The main argument against using
the Gaussian distribution to predict particle size distributions is the symmetry of the distribution.
Due to the symmetry. the distribution predicts negative particle sizes. Therefore. the
experimental l-lysine monohydrochloride data was also fitted using log-normal and gamma
distributions. These distributions overcome the problem of negative particle sizes. However. the
best fit was obtained using the Gaussian distribution, and this distribution was used to fit the
crystal size distributions for I-lysine monohydrochloride purification.

79

 

   
 

 

   

 

 

 

 

 

1.0
r
0.8 «r-
5 i
E 0.6 «tr-
J: .
I
‘ I
I
e r
s .
g 0.4 -
‘5
E .
a o 5 mllmln
. e 30 mllmin
0-2 ‘ —5 mllmin Gaussian
[ —30 mllmin Gaussian
0.0 a a a a a : a j a a : A a a : a a f .e
0 200 400 600 800 1000 1200 1400
Particle size, [urn]

Figure 8.4. Experimental and predicted cumulative crystal size distributions for 5 mllmin and
30 mllmin antisolvent addition rates. The l-Iysine monohydrochloride purification was done at
30 °C and the initial solution contained 42 w-% of l-Iysine monohydrochloride in water.

Equation 8.8 gives the Gaussian distribution.

1 _ (L - me)2

.../“'2." °"" 2.2

f(L) = 8.8

This equation is used to predict the crystal size distributions for 5 mllmin and 30 mllmin
antisolvent addition rates. The weight based mean size is obtained from Equation 8.5, and the
coeflicientofvariafionisassumedtobeO.52.TheresultsarepreserfiedinFigure8.4.
Cmpaiwnmexpenmdamflnwsmufisapuoadtuedidsfiewmulafivepafidesize
distributions welL

8.8 Conclusions

Theefiectdflremfisdvudaddifionratemflreaystaflizafionwasimeflgated.ltwasshown
matflieueigMbasednnanpafidesizemdflteaystdnassmsbmglyafiededbyfln
antisolvent addition rate. The bulk supersaturation was only affected by the antisolvent
concentrationinthecrystallizerandnotbytheantisolventadditionrate.Theshapeofthecrystal
size distribution was Gaussian using this crystallization method, and it was not affected by
varyingantisolventadditionrates.

Chapter 9
CONCLUSIONS

This thesis is basic study of a batch antisolvent crystallization. Antisolvent crystallization is an
excellent method for crystallization when a narrow particle size distribution and high yields are
desired. It is also effective for heat sensitive materials, since the crystallization can be achieved
at low temperatures. The objective was to find an operating procedure that produces a desired
particle size distribution.

Since antisolvent crystallization is a three component system, the interactions between the
soiventarfltheantisolventmstudiedfirst. Itwasshownthattheinfluenoeoftheantisolvent
addition can be much more dramatic than expected based on an ideal system. Therefore. a new
method for estimating the polarity of a binary mixture was presented.

Chapter2introducedthemethodwherethechangcsintheemission maximumofapolanty
sensitive solvatochromic probe molecule were related to polarity changes of common solvents. it
was also shown how to use this method to estimate dielectric constants for less knovm solvents.

lnChapter3thesameapproachwasusedtostudythepolaritybehaviorofvariousblnary
mixtures. The nonideality in the ethanol-water system was due to hydrogen bonding effects. The
results lndiwted that a small amount of ethanol caused a significant change in the system
polarity, and thus, a very sharp decrease in solubility.

81

82
No solubility data were found in the literature for ane monohydrochloride in water or ethanol.

Therefore. the solubility of l-lysine monohydrochloride in water, in ethanol and in mixtures of
ethanol and water was determined. The l-lysine monohydrochloride solubility in water was found
to be only slightly temperature dependent, and l-lysine monohydrochloride was practically
insoluble in ethanol. The solubility in mixtures of ethanol and water decreased considerably with
the ethanol concentration. This was exploited in the crystallization.

Attenuated total reflection Fourier transform spectroscopy was used for in situ monitoring of the
crystallization. In Chapter 5 the calibration curves for both ethanol concentration and l-lysine
monohydrochlorideconcentrationwaregenelated. Thwewerethen used in Chapters7and8to
obtain acmrate information about the crystallization.

Thekineticsofthesystemwerestudied usinganucleation cell andseeded batchexperlments.
The nucleation cell experiments confirmed that the growth rate of i-lysine monohydrochloride is
dze4Mependan.Theseexpenmntsdsoindiwedflutedtmddoesnodiaigeumgth
behaviorsigniflcmtly. Thegrowthofsaedcrystalsinseadedbatch experimentswasthen usedto
estimate overall growth rates for I-lysine monohydrochloride crystals. Estimates for the
nucleation rates were iterated using the population balance. However, since the nucleation rate
is very sensitive to the operating conditions, the numerical values presented in this thesis only
reflecttheorderofmagnitudeofthe nucleation rate.

Chapter 7 discussed the effect of the antisolvent addition rate on the product specifications. This
was studied by sieving the crystallized product. A relationship beMen the weight based average
particle size and the arusolvent addition rate has developed. This can be used to predict the
antisolventaddition ratetoobtainadesired averagesizefortheparticlesize distribution. The
experiments confirmed that the crystal size distributions obtained in the antisolvent crystallization
are narrow, and not affected by the antisolvent addition rate. Chapter 7 addressed also the effect
of the antisolvent addition rate on parameters related to the operation of the crystallizer. The

83
ATR FT iR data was used to analyze this. A linear relationship was found between the change in

the crystal mass during the crystallization and the antisolvent addition rate. The bulk
supersaturation was obtained from the in situ measurements as well. It was shown that the bulk
supersaturation was only a function of antisolvent concentration and not of antisolvent addition
rate. Therefore, controlling the level of bulk supersaturation is impossible when the antisolvent is

added at constant flowrate into the crystallizer.

The study of the bulk supersaturation introduced the main problem of the constant antisolvent
addition rate approach. Since the level of bulk supersaturation cannot be controlled, and much
less the level of local supersaturation at the point of antisolvent addition. the addition of the
antisolvent causes excessive nucleation during the crystallization. This nucleation in turn
produces a large number of very fine particles which leads to a less uniform crystal size and
possible problems in the downstream processing. This same problem occurs also in other batch
crystallization processes, like cooling and evaporative crystallizations. For cooling crystallization
it has been found that the desimd mode of operation is that of constant supersaturation,
effectively controiing the excessive nucleation and thus producing better crystals. For antisolvent
crystallization there are two different possibilities to control the level of bulk supersaturation:
seeding and adding solvent into the crystallizer along with the antisolvent. As was shown in
Chapter 6, seeding effectively reduces the amount of nuclei in the crystallizer. Also, the quality
of the grown seed crystals is very good. The other approach of also adding solvent into the
crystallizer will decrease the amount of nucleation, but it also decreases the yields. Both of these

approaches would need more experimental work to be exploited further.

As a result of this research, the antisolvent approach is already used in industry for small scale
I-lysine monohydrochloride purification. The process has been successfully scaled-up to five
times the original laboratory scale. Compared to the previously used cooling crystallization this
method produces significantly higher yields and better quality crystals. Also, since the crystal

size distribution is very narrow, the milling step is no longer needed.

APPENDIX

APPENDIX

Table A.1. Data for Figure 2.1. The Nile Red absorption and emission maxima in various esters
and alcohols. The emission spectra were excited at absorption maxima.

 

 

 

 

 

 

SOLVENT ABSORPTION EMISSION
[nmL [nm]
methanol 557 640
ethanol 549 635
n-propanol 546 833
n-butanol 547 631
n-pentanol 548 632
n-hexanol 544 629
n-octanol 541 828
ethyl fonnate 528 815
propyl forrnate 526 61 1
ethyl acetate 521 596
ethyl acetoacetate 535 614 .
butyl acetate 521 584
methyl propionate 521 588
methyl butyrate 520 586
ethyl butyrate 519 584
ethyl lactate 547 838
dimethyl succinate 533 608
dimethyl maleate 538 621
dimethyl giutarate 525 605
dimethyl adipate 530 802
diethyl succinate 531 599
diethyl maleate 534 607
diethyl fumarate 532 598
diethyl l-tartrate 550 640
dibutyl fumarate 528 607
dibutyl l-tartrate 545 633
dibutyl itaconate 528 598
dibutyl maleate 530 81-1
triethyl citrate 535 619

 

85

APPENDIX

Table A1. Data for Figure 4.1. Solubility of I-Iysine monohydrochloride in water as a function of
temperature.

 

 

Temperature L-lysine monohydrochloride concentration
[°C] [w-%]
20.0 39.4
25.0 41 .3
25.0 40.1
25.8 40.1
27.0 42.1
27.3 42.4
27.7 42.4
29.0 43.0
30.5 43.8
30.8 44.4
30.9 42.4
31.0 44. 6
33.0 45.3
33.6 45.1
35.1 44.6
44.9 50.2
45.2 50.3

 

 

 

 

APPENDIX

Table A3. Data for Figure 4.2. Solubility of l-lysine monohydrochloride in various binary
mixtures of water and ethanol at 30 :i: 1 °C.

 

 

mass of ethanol I mess of water l-lysine monohydrochloride in solution
[gethanol IQwaler] [_9llmhc [Dealer]
0.00 0.73
0.22 0.42
0.27 0.45
0.28 0.35
0.81 0.22
0.29 0.35
0.42 0.29
0.42 0.29
1.62 0.08
0.75 0.17
0.91 0.10
0.75 0.15
0.90 0.11
0.84 0.11
0.90 0.10
0.90 0.12
0.53 0.18
1.58 0.04
2.83 0.02
2.83 0.01
8.83 0.00
4.73 0.01
0.11 0.48
0.21 0.36
1.05 0.07
2.88 0.02
1.77 0.05
2.17 0.03

 

 

 

 

87

APPENDIX

Table A.4. Data for Figure 5.2. The penetration depth of the evanescent wave for pure water and
pure ethanol as a function of wavenumber. The data were calculated using Equation 5.4.

 

 

 

 

 

wavenumber penetration depth in water penetration depth in ethanol
lcm“l rum] [run]
700 3.408 3.482
800 2.981 3.047
900 2.849 2.709
1000 2.384 2.438
1100 2.168 2.218
1200 1.987 2.031
1300 1.834 1.875
1400 1.703 1.741
1500 1.590 1 .825
1800 1.490 1.524
1700 1 .403 1 .434
1800 1.325 1 .354
1900 1 .255 1.283
2000 1.192 1.219
2100 1 .135 1.161
2200 1 .084 1.108
2300 1.037 1.080
2400 0.994 1.018
2500 0.954 0.975
2800 0.917 0.938
2700 0.883 0.903
2800 0.852 0.871
2900 0.822 0.841
3000 0.795 0.813
3100 0.789 0.788
3200 0.745 0.782 '
3300 0.723 0.739
3400 0.701 0.717
3500 0.681 0.896
3800 0.682 0.677
3700 0.844 0.859
3800 0.827 0.841
3900 0.811 0.625
4000 0.598 0.609

 

 

88

APPENDIX

Table A8. Data for Figure 5.5. Relative absorbance as a function of ethanol weight percentage.

 

 

 

 

 

 

 

 

w—% of ethanol LOG(A10.2 / Arm) llmhc w-% of ethanol LOGQQLI Amo) llmhc
0.0 1.003891 yes 34.6 0.851830 yes
0.0 1 .002844 yes 35.9 0.844773 yes
0.0 1.003247 yes 38.1 0.827419 yes
0.0 1.003134 yes 37.0 0.862060 no
0.0 0.996155 no 37.0 0.856584 no
0.0 1.006393 no 37.4 0.839118 yes
0.0 0.997089 no 38.5 0.818246 yes
0.0 1.007596 yes 38.5 0.814135 yes
0.0 1 .003628 yes 40.2 0.829758 yes
1 .7 0.989849 yes 40.8 0.809942 yes
5.9 0.970487 yes 40.8 0.838384 yes
6.9 0.966018 yes 41.3 0.813880 yes
9.2 0.958335 yes 42.7 0.822030 yes
11.1 0.940956 yes 42.7 0.818690 yes
13.0 0.938648 yes 42.9 0.802030 yes

13.5 0.929001 yes 43.6 0.817957 yes
15.0 0.980044 yes 43.9 0.795135 yes
15.6 0.928817 yes 44.0 0.838722 no
15.7 0.925488 yes 44.0 0.834154 no
15.8 0.916796 yes 44.9 0.795578 yes
18.4 0.937143 no 48.1 0.810716 yes
18.4 0.931872 no 48.7 0.790222 yes
18.3 0.915245 yes 48.8 0.783885 yes
20.0 0.898118 yes 48.3 0.798029 yes
23.0 0.895278 yes 48.3 0.801989 yes
23.9 0.881808 yes 48.4 0.779543 yes
23.9 0.880889 yes 48.4 0.778475 yes
26.1 0.903208 yes 49.5 0.820253 no
28.1 0.887121 yes 50.5 0.813970 no
27.2 0.878280 yes 50.6 0.795974 yes
27.2 0.877580 yes 52.8 0.785550 yes
27.3 0.864954 yes 52.8 0.783068 yes
28.2 0.892743 no 52.8 0.789995 yes
28.2 0.887394 no 52.8 0.785631 yes
30.5 0.849937 yes 54.1 0.806806 no
30.9 0.861286 yes 54.1 0.802808 no
32.0 0.842387 yes 100.0 0.706624 no
33.4 0.837980 yes 100.0 0.706218 no
34.2 0.859437 yes 100.0 0.706698 no
34.3 0.849958 yes 100.0 0.706339 no

 

 

89

APPENDIX

Table A8. Data without ethanol for Figure 5.6. Difference of the second derivative peaks at
1411 cm" and 1432 cm" as a function of the weight fraction of l-lysine monohydrochloride.

 

 

w-% of llmhc D
20.1 7.5455
26.0 10.8680
29.4 13.4487
33.4 15.1594
35.6 16.2807

0.0 0.0885
5.0 2.6185
10.0 4.4791
15.0 5.8755

 

 

 

 

APPENDIX

Table A.1. Data with ethanol for Figure 5.6. Difference of the second derivative peaks at
1411 cm" and 1432 cm‘1 as a function of the weight fraction of I-lysine monohydrochloride.

 

 

 

 

 

 

 

w-% of llmhc D tat-96 of llmhc D
0.0 1.2978 19.8 8.6312
0.0 1.5019 20.9 9.2622
0.0 0.7365 20.9 8.7713
0.0 0.8436 21 .2 9.2780
0.0 1 .5876 22.1 9.9542
0.0 1.9218 23.4 10.3646
0.0 1.8485 24.2 10.5069
0.0 2.0384 24.2 9.9752
0.0 2.1293 24.9 10.9160
0.0 0.5892 26.0 11.1057
0.0 0.7167 26.7 11.4068
0.0 0.9904 28.2 12.1259
0.0 1.1170 28.7 13.434
0.0 1.3452 28.7 11.9348
0.0 1.3440 28.7 12.5894
0.0 1.3796 29.7 14.5051
0.0 0.8691 32.1 13.7347
0.0 1.0637 33.8 14.5476

16.4 6.2071 34.1 16.7578
17.0 7.5141 34.7 14.3204
17.1 7.9027 35.6 14.9418
17.9 8.1602 37.7 15.8416
18.4 7.1804 40.1 17.0472
18.4 7.1439 40.1 16.5054
18.8 8.3451 40.1 17.0968
18.9 8.3531

 

91

APPENDIX

Table A.8. Data without Ethanol for Figure 7.2. Spherical equivalent diameter for single crystals
in a seeded nucleation cell.

 

 

Time Spherical equivalent Ttme Spherical equivalent
diameter diameter
[hzmin] [mm] [hzmin] [mm]
0:00 0.529 0:15 0.857
0:00 0.357 0220 0.668
0:00 0.265 0:20 1 .074
0:00 0.219 0:20 1.066
0:00 0.403 0:20 1.007
0:00 0.263 0:20 1 .074
0:00 0.224 0:20 0.668
0:00 0.155 0:30 0.748
0:07 0.772 0:30 0.814
0:07 0.656 0:30 1.267
0:07 0.610 0:30 1.253
0:07 0.508 0:30 1.215
0:10 0.529 0:40 1.391
0:10 0.535 0:40 1.456
0:10 0.852 0:45 1.336
0:10 0.553 0:55 1.674
0:10 0.658 1:00 1.858
0:10 0.575 1:11 2.438
0:15 0.994 1:20 2.679
0:15 0.990 1:31 3.018
0:15 0.818 1:42 3.103

 

 

 

 

 

 

92

APPENDIX

Table A.9. Data with Ethanol for Figure 7.2. Spherical equivalent diameter for single crystals in a
seeded nucleation cell.

 

 

Time Spherical equivalent Time Spherical equivalent
diameter diameter
[hzmin] [mm] [hzmin] [mm]
0:00 0.239 0:25 1.128
0:00 0.359 0:25 0.888
0:00 0.350 0:25 1.175
0:00 0.359 0:30 0.930
0:05 0.409 0:30 1 .278
0:05 0.52 0:30 1.056
0:05 0.479 0:30 1 .276
0:05 0.535 0:34 0.944
0:10 0.535 0:34 1.356
0:10 0.659 0:39 1.043
0:10 0.628 0:39 1 .497
0:10 0.705 0:39 1 .206
0:10 0.667 0:39 1 .438
0:10 0.757 0:42 1 .014
0:15 0.648 0:42 1 .495
0:15 0.814 0:42 1.664
0:15 0.728 0:42 1.576
0:15 0.867 0:42 1.391
0:19 0.829 0:52 1.894
0:19 0.997 0:52 1.755
0:20 0.731 0:52 1.525
0:20 0.977 1 :02 2.073
0:20 0.857 1 :02 1 .924
0:20 1.016 1 :02 1 .672
0:25 0.818 1:12 2.212
0:25 1.116 1:12 2.054
0:25 0.947 1 :12 1.756

 

 

 

 

 

 

93

APPENDIX

Table mo. Data for Figure 7.3. Mass of l-lysine monohydrochloride crystals retained on the
sieves after 2. 5. and 10 minutes of crystallization. The experiment was done at 30 °C and the

ethanol addition rate was 20 mllmin. The average seed size was 550 pm and the mass of seeds
was 10 g.

 

 

Particle size, [pm] Mass. [9]
2 min 5 min 10 min
31 .5 0.46 0.18 0.36
76.5 0.64 0.65 1.16
152.5 0.99 2.62 3.51
215.5 1.04 8.79 8.61
302.5 1.15 13.58 13.01
390.0 0.90 8.32 12.22
462.5 1.11 3.57 5.53
550.0 2.75 3.96 3.91
655.0 6.29 3.84 4.23
735.0 6.72 17.01 17.65
925.0 0.98 7.69 17.67
1090.0 0.96 5.29 15.55
1300.0 0.96 5.29 2.08

 

 

 

 

 

 

APPENDIX

Table A.11. Data for Figure 7.4. Estimate for the overall growth rate as a function of ethanol
addition rate from seeded experiments.

 

 

 

Time Weight based mean particle size
[min] {um}
1 mllmin 5 mllmin 20 mllmin
0 463 550 550
2 708
5 796
10 884
1 5 634 844
30 804 931
45 1028
60 896
growth rate 7.1 10 31

 

 

 

 

 

 

95

APPENDIX

Table A.12. Data for Figure 7.5. Crystal size distribution of the fines after 30 minutes of
crystallization. Antisolvent was added at 1 mllmin and the seeds for the seeded experiments
were the 500-600 pm sieve fraction.

 

 

 

 

Average size Seeded Unseeded
{um} [91 [9]

0.0 0.00 3.93
31.5 1.22 12.34

76.5 1 .56 24.72
152.5 2.75 6.48
215.0 4.92 0.20
302.5 8.1 1 0.00
390.0 4.49 0.00
462.5 4.51 0.00
27.6 47.7

 

 

 

 

APPENDIX

Table A.13. Data for Figure 7.6. Predicted nucleation behavior during crystallization.

 

 

 

 

Ethanol Nucleation rate Ethanol Nucleation rate
[ml] 1 mllmin 1 mllmin [ml] 1 mllmin 1 mllmin
0.00 7.18E+12 7.05E+12 153.45 5.58E+12 5.48E+12
4.95 6.63E+12 6.50E+12 158.40 5.57E+12 5.47E+12
9.90 6.40E+12 6.28E+12 163.35 5.57E+12 5.46E+12

14.85 6.28E+12 6.16E+12 168.30 5.56E+12

19.80 6.19E+12 6.07E+12 173.25 5.55E+12 5.44E+12
24.75 6.12E+12 6.00E+12 178.20 5.54E+12 5.44E+12
29.70 6.06E+12 5.95E+12 183.15 5.53E+12 5.43E+12
34.65 6.01 E+12 5.90E+12 188.10 5.53E+12 5.42E+12
39.60 5.97E+12 5.86E+12 193.05 5.52E+12 5.42E+12
44.55 5.94E+12 5.83E+12 198.00 5.51 E+12 5.41 E+12
49.50 5.91 E+12 5.80E+12 202.95 5.51 E+12 5.40E+12
54.45 5.88E+12 5.77E+12 207.90 5.50E+12 5.40E+12
59.40 5.85E+12 5.74E+12 212.85 5.49E+12 5.39E+12
64.35 5.83E+12 5.72E+12 217.80 5.49E+12 5.38E+12
69.30 5.81 E+12 5.70E+12 222.75 5.48E+12 5.38E+12
74.25 5.79E+12 5.68E+12 227.70 5.47E+12 5.37E+12
79.20 5.77E+12 5.66E+12 232.65 5.47E+12 5.36E+12
84.15 5.75E+12 5.64E+12 237.60 5.46E+12 5.36E+12
89.10 5.74E+12 5.63E+12 242.55 5.46E+12 5.35E+12
94.05 5.72E+12 5.61 E+12 247.50 5.45E+12 5.35E+12
99.00 5.71 E+12 5.60E+12 252.45 5.45E+12 5.34E+12
103.95 5.69E+12 5.59E+12 257.40 5.44E+12 5.34E+12
108.90 5.68E+12 5.57E+12 262.35 5.44E+12 5.33E+12
113.85 5.67E+12 5.56E+12 267.30 5.43E+12 5.33E+12
118.80 5.66E+12 5.55E+12 272.25 5.43E+12 5.32E+12
123.75 5.64E+12 5.54E+12 277.20 5.42E+12 5.32E+12
128.70 5.63E+12 5.53E+12 282.15 5.42E+12 5.31 E+12
133.65 5.62E+12 5.52E+12 287.10 5.41E+12 5.31E+12
138.60 5.61E+12 5.51E+12 292.05 5.41E+12 5.30E+12
143.55 5.60E+12 5.50E+12 297.00 5.40E+12 5.30E+12
148.50 5.59E+12 5.49E+12 301.95 5.30E+12

 

 

 

 

 

 

97

APPENDIX

Table A.14. Data for Figure 7.6. Predicted nucleation behavior during crystallization.

 

 

 

Ethanol Nucleation rate Ethanol Nucleation rate
[ml] 5 mllmin [ml] 30 mllmin
0.00 1 .39E+13 0.00 1 .76E+13

25.75 1.28E+13 19.00 1.80E+13
51.42 1.24E+13 37.50 1.74E+13

77.08 1.22E+13 56.00 1.70E+13

102.75 1.20E+13 74.50 1.68E+13

128.33 1.18E+13 93.00 1.66E+13

153.92 1.17E+13 111.50 1.65E+13

179.58 1.17E+13 130.00 1.63E+13

205.25 1.16E+13 148.50 1.62E-l-13

230.92 1.15E+13

256.58 1 .14E+13

282.25 1.14E+13

307.92 1.13E+13

 

 

 

 

 

98

APPENDIX

Table A15. Data for Figure 7.7. Experimental and predicted yield for unseeded experiments
using the predicted nucleation and growth rates.

 

 

 

 

 

 

 

Time Yield (x) Yield (0) Predicted Time Yield (x) Yield (0) Predicted
mzminzsl [w-%] [w-%] [w-%] [hzminzsj [w-‘liil [Viv-96] [VI-96]
00:00:00 0.0 0.0 0.0 02:33:27 27.5 30.1 29.3
00:04:57 0.0 0.0 1.1 02:38:24 27.4 29.8 30. 2
00:09:54 0.0 1.4 2.1 02:43:21 30.0 33.8 31.1
00:14:51 0.0 0.3 3.1 02:48:18 31.8 35.2 32.0
00:19:48 3.2 0.0 4.1 02:53:15 33.0 37.7 32.9
00:24:45 2.5 2.3 5.1 02:58:12 35.2 38.0 33.8
00:29:42 6.6 4.5 6.1 03:03:09 33.7 38.0 34.7
00:34:39 6.5 4.5 7.1 03:08:06 34.4 40.2 35.6
00:39:36 9.2 6.4 8.0 03:13:03 38.1 39.7 36.5
00:44:33 11.5 5.8 9.0 03:18:00 40.6 41.8 37.4
00:49:30 9.9 9.2 9.9 03:22:57 39.8 42. 2 38.3
00:54:27 13.3 8.2 10.9 03:27:54 38.4 44.7 39.1
00:59:24 12. 2 10.5 11.8 03:32:51 41.0 42.2 40.0
01:0421 12.3 14.5 12.8 03:37:48 40.2 44.6 40.9
01:09:18 14.4 13.6 13.7 03:42:45 43.0 44.0 41.8
01:14:15 14.4 12.8 14.7 03:47:42 46.3 46.1 42.7
01:19:12 15.3 14.8 15.6 03:52:39 43. 8 47.1 43.6
01:24:09 16.3 16.0 16.5 03:57:36 46.6 47.5 44.5
01:29:06 17.0 15.3 17.4 04:02:33 46.7 49.5 45.3
01:34:03 15.3 17.4 18.4 04:07:30 50.2 50. 3 46.2
01:39:00 18.1 16.2 19.3 04:12:27 47.8 51.3 47.1
01:43:57 19.8 18.0 20.2 04:17:24 46.4 51.4 48.0
01:48:54 21.8 19.3 21.1 04:22:21 47.3 50. 9 48.9
01:53:51 18. 8 20.3 22.0 04:27:18 49.0 51.3 49. 7
01:58:48 21.6 23.1 23.0 04:32:15 49.7 52.0 50. 6
02:03:45 22.5 24. 3 23.9 04:37:12 50.0 53.0 51.5
02:08:42 24.1 23.8 24.8 04:42:09 51.5 54.6 52.4
02:13:39 20.8 26.2 25.7 04:47:06 51.9 55.4 53.2
02:18:36 21.8 25.4 26.6 04:52:03 52.1 55.5 54.1
02:23:33 26.4 26. 2 27. 5 04:57:00 49.6 55.3 55.0
02:28:30 27.1 27. 4 28. 4 05:01 :57 55.9 56.5 55.9

 

 

 

 

 

APPENDIX

Table A16. Data for Figure 8.1. The effect of ethanol addition on the increase in crystal mass
during the crystallization

 

 

 

Time Yield, [w-%] Yield, [w-%]] Time Yield, [w-‘iiill Time Yield, [w—‘Xil
[h2minzs] 1 mllmin 1 mllmin [hzminzs] 5 mllmin [h2min2s] 30 mllmin
00:00:00 0.0 0.0 00:00:00 0.0 00:00:00 0.0
00:04:57 0.0 0.0 00:05:09 6.4 00:00:38 3.7
00:09:54 0.0 1.4 00:10:17 9.7 00:01 :15 6.0
00:14:51 0.0 0.3 00:15:25 14.8 00:01:52 11.6
00:19:48 3.2 0.0 00:20:33 25.5 00:02:29 15.2
00:24:45 2.5 2.3 00:25:40 41.2 00:03:06 15.4
00:29:42 6.6 4.5 00:30:47 48.2 00:03:43 18.2
00:34:39 6.5 4.5 00:35:55 54.8 00:04:20 21.6
00:39:36 92 6.4 00:41 :03 57.3 00204257 32.8
00:44:33 11.5 5.8 00:46:11 60.6 00:05:34 55.4
00:49:30 9.9 9.2 00:51 :19 63.1
00:54:27 13.3 8.2 00:56:27 65.3
00:59:24 12.2 10.5 01:01:35 67.4
01:04:21 12.3 14.5 01:06:43 69.0
01:09:18 14.4 13.6 01:11:51 71.1
01:14:15 14.4 12.8 01:16:59 72.5
01219212 15.3 14.8 01:22:07 273.1
01224209 16.3 16.0 01:27:15 72.5
01:29:06 17.0 15.3 01:32:23 72.3
01:34:03 15.3 17.4 01:37:31 72.4
01:39:00 18.1 16.2 01:42:39 71.2
01:43:57 19.8 18.0 01:47:47 71.5
01:48:54 21.8 19.3 01:52:55 72.4
01:53:51 18.8 20.3 01:58:03 71.1
01258248 21.6 23.1 02:03:11 71.2
02:03:45 22.5 24.3 02208218 71 .9
02:08:42 24.1 23.8 02:13:25 71.3
02:13:39 20.8 26.2 02218233 71.8
02:18:36 21.8 25.4 02:23:41 70.0
02:23:33 26.4 26.2 02:28:49 68.9
02:28:30 27.1 27.4 02:33:57 70.1
02:33:27 27.5 30.1 02239204 68.0
02:38:24 27.4 29.8 02:44:12 69.5
02:43:21 30.0 33.8 02:49:20 69.0
02:48:18 31.8 35.2 02:54:28 68.3
02253215 33.0 37.7 02:59:35 70.1
02:58:12 35.2 38.0

 

 

 

 

 

 

 

 

100

APPENDIX

Table A.11. Data for Figure 8.2. Relative bulk supersaturation as a function of added ethanol for
various ethanol addition rates.

 

 

 

 

 

 

 

Ethanol Supersaturation
[ml] 1 mllmin 5 mllmin 10 mllmin 30 mllmin
0 0.631 0.631 0.631 0.631
10 0.957 0.946 0.947
20 1.236 1.210 1.214
30 1.475 1 .431 1 .438 1 .428
40 1 .678 1 .614 1 .624
50 1 .851 1 .764 1.777
60 1.997 1.885 1.902 1.877
70 2.118 1.980 2.001
80 2.219 2.051 2.077
90 2.300 2.102 2.132 2.088
100 2.364 2.135 2.170
110 2.413 2.151 2.191
120 2.447 2.152 2.197 2.131
130 2.469 2.140 2.190
140 2.480 2.115 2.170
150 2.480 2.079 2.140 2.050
160 2.471 2.032 2.099
170 2.453 1 .976 2.049
180 2.426 1.912 1.990 1.875
190 2.393 1 .839 1 .923
200 2.352 1.759 1.849
210 2.305 1.672 1.768 1.627
220 2.253 1.579 1.681
230 2.194 1 .480 1 .588
240 2.131 1.375 1.490 1.322
250 2.063 1 .266 1 .387
260 1.990 1.151 1.279
270 1.914 1.032 1.166 0.969
280 1 .833 0.909 1 .049
290 1.749 0.782 0.929
300 1 .661 0.651 0.805 0.579

 

 

101

APPENDIX

Table A.18. Data for Figure 8.3. Weight mean size of crystals as a function of ethanol addition
rate at 30 °C. The initial solution contained 42 w-% of l-lysine monohydrochloride in water.

 

 

Ethanol addition rate Weight based mean size

[ml/min] [NM]
1 962

1 821

1 754

1 71 7

2 560

5 874

5 752
20 352
30 526
50 378
300 44

 

 

 

 

102

APPENDIX

Table A.19. Data for Figure 8.4. Experimental and predicted cumulative crystal size distributions
for 5 mllmin and 30 mllmin antisolvent addition rates.

 

 

 

 

 

Average size Cumulative weight fraction on sieve

[WI] 5 mllmin 30 mllmin
0.0 0.000 0.000
31.5 0.000 0.000
76.5 0.004 0.01 1
152.5 0.014 0.032
215.0 0.040 0.086
302.5 0.095 0.213
390.0 0.171 0.385
462.5 0.287 0.567
550.0 0.37 0.671
655.0 0.491 0.786
735.0 0.592 0.873
925.0 0.767 0.973
1090.0 0.871 0.986
1300.0 0.973 0.995
1400.0 0.999 0.999

 

 

LIST OF REFERENCES

LIST OF REFERENCES

[1]. Axiom Analytical. Operator's Manual, DPR-200 Series Deep Immersion Probes, Revision D,
California, 1992.

[2]. Bator, E., Continuous crystallization of magnesium asparaginate by salting out, Hungarian
Journal of Industrial Chemistry Veszpram, 15 (1987), 141-148.

[3]. Budavari, 8., (Ed.), The Merck Index, 11th Ed., Merck and 00., NewJersey, 1989.

[4]. Budz, J., Karpinski, P.H., Mydlarz, J., and Nyvlt, J., Salting-out precipitation of
cocarboxylase hydrochloride from aqueous solution by addition of acetone, Ind. Eng. Chem.
Prod. Res. Dev.. 25 (1986), 657-664.

[5]. Cahill, J.E., Derivative spectroscopy: understanding its application, American laboratory. 11
(1979), No. 11, 79-85.

[6]. Cahill, J.E., and Padera, F.G., Derivative analysis of uvlvisibie spectra, Amen’can
Laboratory, 12 (1980), No. 4, 101-112.

[7]. Coates. J.P., The industrial applications of infrared internal reflection spectroscopy, In
Internal reflecb'on spectroscOpy: theory and applications, Mirabella. F.M., Jr., (Ed.), Marcel
Deldter, NewYork, 1993.

[8]. Coetzee, J.F., and Ritchie, C.D. (Ed.), Solute-Solvent Interactions. Vol. 2, Marcel Dekker,
New York, 1976.

[9]. Colthup, N.B., Daly. L.H., and Wrberley, S.E., Introduction to Infrared and Raman
Spectroscopy. 3rd Ed., Academic Press, Boston, 1990.

[10]. Davey, R.J., The role of the solvent in crystal growth from solution, J. Cryst. Growth, 76
(1986). 637-644.

[11]. Deye, J.F., and Berger, T.A., Nile red as a solvatochromic dye for measuring solvent
strength in normal liquids and mixtures of normal liquids with supercritical and near critical fluids,
Anal. Chem, 82 (1990). 615-622.

[12]. Dunuwila, 0.0., An investigation of the feasibility of using in situ ATR FTIR spectroscopy in
the measurement of crystallization phenomena for research and development of batch
crystallization processes, Ph.D. Dissertation, Michigan State University, 1996.

[13]. Dunuwila, D.D., Carroll ll, L.B., and Berglund, K.A., An investigation of the applicability of
attenuated total reflection infrared spectroscopy for measurement of solubility and
supersaturation of aqueous citric acid solutions, J. Cryst. Growth, 137 (1994), 561 -568.

[14]. Dutta, A.K., Kamada, K., and Ohta. K., Spectroscopic studies of nile red in organic solvents
and polymers, J. of Photochem. and Photobr’ology, A: Chemistry. 93 (1996), 57-64.

103

104

[15]. Gabas, N.. and Laguerie, 0., Batch crystallization of d-xylose by programmed cooling or by
programmed adding of ethanol, Chem. Eng. Sci, 47 (1992), No. 12, 3148-3152.

[16]. Ghoneim, N.. and Suppan, P., Solvatochromic shifts of non-dipolar molecules in polar
solvents. Spectrochimica Acta, 51A (1995), No. 6, 1043-1050.

[17]. Greenspan, P., and Fowler. S.D., Spectrofluorometric studies of the lipid probe, nile red, J.
of Lipid Res. 26 (1985), 781-789.

[18]. Hawkes, S., Maddams, W.F., Mead, W.L., and Southon, M.J., The measurement of
derivative i.r. spectra - II. Experimental measurements, Spectrochimica Acta, 38A (1982), No. 4,
445-457.

[19]. lngle, JD, Jr., and Crouch, S.R., Spectrochemical analysis, Prentice-Hall, New Jersey,
1988.

[20]. Irani, RR, and Callis, C.F., Particle size: measurement, interpretation, and application.
John Wiley, New York, 1963.

[21]. Jones, AG., Budz, J., and Mullin, J.W., Batch crystallization and solid-liquid separation of
potassium sulphate, Chem. Eng. Sci., 29 (1974), 1075-1087.

[22]. Jones, A.G., and Teodossiev. N.M., Microcomputer programming of dosage rate during
batch precipitation, Cryst. Res. Technol.. 23 (1988), No. 8. 957-966.

[23]. Karpinski, P.H., and Nyvlt, J., Metastable zone Mdth in salting-out crystallization, J. Cryst.
Res. Technol.. 18 (1983), No. 7, 959-965.

[24]. Kirk-Othmer, Encyclopedia of Chemical Technology, 4th Ed., John \Niley, New York, 1992.

[25]. Kirwan. D.J., and Orella, C.J., Crystallization of biological molecules, in Handbook of
Industrial Crystallization, Myerson, AS. (Ed.), Butterworth—Heinemann. Maryland, 1993.

[26] Lide, D.R., (Ed.).; CRC Handbook of Chemistry and Physics, 73rd Ed., CRC Press, Florida,
1992. 9-52.

[27]. Maddams, W.F., and Mead, W.L., The measurement of derivative i.r. spectra - I.
Background studies, Spectrochr'mica Acta, 38A (1982), No. 4, 437-444.

[28]. Maddams, W..F, andSouthon. M..,J Themeasurementofderivativeir. spectra-III. The
effect of band width and band shape on resolution enhancement by derivative spectroscopy.
SpectrochimicaActa, 38A(1982), No. 4, 459-466.

[29]. Manual on testing siean methods. ASTM 447B, ASTM, 1985.

[30]. Midwinter, J., and Supmn, P., Spectral shifts in solvent mixtures, Spectrochim. Acta, 25A
(1969). 953-958.

[31]. Mina-Mankarios, G., and Pinder. K.L., Salting out crystallization of sodium sulfate, Can. J.
Chem. Eng, 69 (1991). 308-324.

[32]. Mirabella, F.M., Jr., Principles, theory, and practice of internal reflection spectroscopy, In
Internal reflection spectroscopy: theory and applications, Mirabella, F.M., Jr., (Ed.), Marcel
Dekker, NewYork, 1993.

105

[33]. Muller. G.J., and Abraham-Fuchs, K, Matrix dependence in single and multilayer IRS
spectra, In Internal reflection spectroscopy: theory and applications, Mirabella, F.M., Jr., (Ed.),
Marcel Dekker, New York, 1993.

[34]. Mullin, J.W., Crystallization, 3rd Ed.. Buttemorth-l-leinemann, Great Britain, 1993.

[35]. Mullin, J.W., Teodossiev, N.. and Schnel, 0., Potassium sulphate precipitation from
aqueous solutionbysalting-oulwithacetone, Chem. Eng. Process.,26(1989),93-99.

[36]. Myerson, AS., Solutions and solution properties. In Handbook of Industrial Crystallization,
Myerson, AS. (Ed.), Butterworth—l-leinemann, Maryland. 1993.

[37]. Myerson, AS., Decker. SE, and Welping, F., Solvent selection and batch crystallization.
Ind. Eng. Chem. ProcessDes. Dev.. 25 (1986), 925-929.

[38]. Myerson, AS.. and Ginde, R., Crystals. crystal growth, and nucleation, In Handbook of
Industrial Crystallization, Myerson, AS. (Ed.), Butterworth-Heinemann, Maryland, 1993.

[39]. Nitsche. KS, and Suppan, P., Solvatochromic shifts and polarity of solvent mixtures.
Chimr'a. 36 (1982), No. 9. 346-348.

[40]. Nyvlt, .r.. Batch setting-out crystallization, Chem. Eng. Processing, 31 (1992), 39-42.

[41]. Onsager, L., Electric moments of molecules in liquids, J. Am. Chem. Soc, 58 (1936), No. 8,
1486-1493.

[42]. Orella, G.J., and Kirwan. D.J., The solubility of amino acids in mixtures of water and
aliphatic alcohols. Biotechnology Progess, 5 (1989), No. 3, 89-91.

[43]. Paul. EL. and Roses. C.B., Challenges for chemical engineers in the pharmaceutical
industry, Chem. Eng. Prog., 86 (1990), No. 12. 17-25.

[44]. Randolph, AD, and Larson, MA, Theory of Particulate Processes. 2nd Ed., Academic
Press, Califomia, 1988.

[45]. Reichardt, C., Solvents and Solvent Effects in Organic Chemistry, 2nd Ed., VCH,
Darmstadt.Germany. 1990.

[46]. Shanks. B.H., and Berglund, KA, Contact nucleation from aqueous sucrose solutions,
AIChE Journal, 31 (1985), No. 1, 152-154.

[47]. Smyth, C.P., Walls, W.S., Electric moment and molecular structure. I. The ethyl esters of
mono- and dicarboxylic acids, J. Am. Chem. Soc., 53 (1931 ), 527-539.

[48]. Stolarova, M., Buchtova, M., and Bekérek, V., Solvent polarity parameters and their mutual
interrelations, Acta Univ. palac. Olomucensrls, Facultas rerum Natur. Chem. )O(III, 79 (1984).
77-88.

[49] Street, K.W., Jr., and Acres, W.E., Jr., Estimation of the Effective Dielectric Constant of
Cyciodextrin Cavities Based on the Fluoreswnce Properties of Pyrene-3-Carboxaldehyde,
Applied Spectroscopy. 42 (1988), No. 7. 1315-1318.

[50]. Suppan, P., Excited-state dipole moments from absorptionlfluorescence solvatochromic
ratios, Chem. Phys. Lett., 94 (1983). No. 3, 272-275.

[ ._._-___-

 

106

[51]. Suppan, P., Local polarity of solvent mixtures in the field of electronically excited molecules
and exciplexes, J. Chem. Soc. Farad. Trans. 1, 83 (1987), 495-509.

[52]. Suppan, P., Solvatochromic shifts: The influence of the medium on the energy of electronic
states, J. of Photochem. and Photobiology, A: Chemistry, 50 (1990), 293-330.

[53]. Suppan, P., Solvent effects on the energy of electronic transitions: experimental
observations and applications to structural problems of excited molecules. J. Chem. Soc. A:
Inorg. Phys. ‘I’heor., 1987, 3125-3133.

[54]. Tavare, N.S., and Chivate, MR, 080 Analysis from a batch dilution crystallizer, J. Chem.
Eng. Japan, 13 (1980), No. 5, 371-379.

[55]. Uusi-Penttila, M., and Berglund, K.A, In situ monitoring of antisolvent crystallization using
attenuated total reflection infrared spectroscopy, Parts I and II, (manuscript).

[56]. Uusi-Penttila, M., and Berglund, KA, Spectroscopic Monitoring of Environmentally Benign
Anti-Solvent Crystallization, J. of Cryst. Growth, 166 (1996), 967-970.

[57]. Uusi-Penttila, M., Richards, R.J., Torgerson, BA, and Berglund. K.A, Spectroscopically
Determined Dielectric Constants for Various Esters, Ind. Eng. Chem. Res, 36 (1997), 510-512.

[58]. Uusi-Penttila, M., Torgerson, BA, and Berglund, K.A., Dielectric Enrichment in Binary
Systems Containing Environmentally Benign Esters. (manuscript).

[59]. Way, J.S., Batch crystallization, In Handbook of lndustn'al Crystallization. Myerson, AS.
(Ed.), Butterworth—Heinemann, Maryland, 1993.

"lllllilillu