.1 :
.1... 1.. . x:

1 ..3:u.-

‘;9. .....r

a. .4 r.

.3 .

$.-

rl...i.‘ :
2‘ 5...,

.
9 EC .9.
iﬁﬁkﬂ. .i.

41.7%“.

1
:Jﬂll.
$3.91
:3 3.... {5.1
. (My.

{3.3.
I

:33.
ﬁg"?

at.

s...
x 3.5:...) I:
ﬁstﬂﬁ?! 3
s mﬂmmaﬁi.
‘2 I.
«a 91“.... .3
II... X.‘.
ﬁreﬁght;

y:

at «.1..-
.
.... .
’ . .
5.0.5.... .5
:2...
z r

3.... .
I .90
. ‘-

a}:

in

.529“?

. haiku... .
Lu).

v.2). 1....

an»
Jae... .

3!.
9...“...
3.3.

4

1%

 

 

 

 

-
z; s. z.
1...:

 

 

 

:1 . x. r‘cux. .)
2;. .5.
- . 1.1

THESIs
-”/

__ 1.: (T

f l'
r" \-

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the

dissertation entitled

OPTIMIZATION OF CRYSTALLIZATION PROCESSES
USING VIBRATIONAL SPECTROSCOPIES

presented by
Fang Wang

has been accepted towards fulﬁllment
of the requirements for

Ph.D. degree in Chem smymhemical Engin

 

$461.24“?

Major professor

Date 3/20”
7 /

MS U is an Afﬁrmative Action/Equal Opportunity Institution 0- 12771

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6’01 cJCIRC/DateDue.p65-p.15

 

OPTIMIZATION OF CRYSTALLIZATION PROCESSES USING
VIBRATION AL SPECTROSCOPIES

By

Fang Wang

A DISSERTATION

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

Doctor of Philosophy
Department of Chemistry
and

Department of Chemical Engineering

2000

ABSTRACT

OPTIMIZATION OF CRYSTALLIZATION PROCESSES USING
VIBRATIONAL SPECTROSCOPIES

By
Fang Wang

Control of crystallization directly affects the physical and chemical properties of the
resulting crystals. Current studies demonstrate that optimal crystallization conditions can

be determined using in situ vibrational spectroscopies.

The ﬁrst study discusses optimization of pH swing crystallization of nicotinic
acid(HNic). HNic is produced by adding hydrochloric acid (HCl) to a sodium nicotinate
(NaNic) aqueous solution. In situ Attenuated Total Reﬂection-Fourier Transform
Infrared (ATR-FTIR) spectroscopy allowed us to measure solubility and supersaturation
including the metastable limit of HNic. The Partial Least Squares regression (PLS)
method was employed for IR data analysis. The solubilities of HNic vary signiﬁcantly
with pH and the concentration of NaNic. Metastable zone width is dependent on the
process parameters including the concentration of HCl, the addition rate of HCl and the
agitation speed. A set of optimized parameters was obtained to produce crystals of HNic
with larger mean size through spontaneous(primary) crystallization. The metastable zone
deﬁnes the operation range for the controlled crystallization of HNic. We seeded HNic
within the operation range to greatly increase mean crystal size of HNic. The degree of
secondary nucleation is demonstrated to be dependent on the level of relative
supersaturation at which seeding is performed. Throughout the post seeding acid

addition, crystallization was controlled by keeping relative supersaturation at a constant

level, which was realized by feedback control of relative supersaturation. It was found
that keeping the solution at low relative supersaturation throughout seeding and post
seeding acid addition processes promotes crystal growth and suppresses secondary
nucleation to the largest degree. Therefore, we are able to obtain HNic crystals with
larger mean size with the optimized process parameters for both seeded and unseeded

crystallization processes.

The second study covers an investigation of solvent-mediated polymorphic
transformation of progesterone using in situ Raman spectroscopy. Many analytical
techniques, such as Differential Scanning Calorimetry (DSC), X-Ray Diffraction (XRD),
Infrared Spectroscopy (IR) and Raman spectroscopy can differentiate between crystalline
polymorphs of the same chemical entity. While all of these techniques are routinely
applied to off-line analysis of materials, only Raman instrumentation technology
currently exists for in situ monitoring of solid phase behavior. We employed Raman
spectroscopy to demonstrate the solvent-mediated polymorphic phase transformation of
progesterone. In situ Raman analysis showed that the appearance of Form I progesterone
is always preceded by the formation of Form II. Phase transformation rates increase
monotonically as temperature increases, which indicates that the polymorphic system is
monotropic. Form I is thermodynamically more stable than Form 11, while Form H is
kinetically favored over Form 1. The results support Ostwald’s law of stages and also
lead to an in-depth understanding of the polymorphic transformation process. The in situ
capabilities of Raman spectroscopy allowed us to deﬁne the processing parameters

required to control the morphology of progesterone.

ACKNOWLEDGMENTS

First of all, I would like to thank my advisor, Dr. Kris A. Berglund, for all the
opportunities he has created for me. Without his guidance and help, I would not have
come to this point today. I am grateful for the unique opportunity to pursue the dual
degree in Chemistry and Chemical Engineering. It helps me understand and solve
problems from more than one perspective. I would like to thank all the other professors
who are or have been my committee members, Dr. Blanchard, Dr. Briedis, Dr. Leroi and
Dr. Crouch, for all their time, advice and support, especially Dr. Blanchard for his timely

support.

I am thankful to Dr. John A. Wachter in Phannacia Corp. for opening a new sight for me
in the pharmaceutical industry. Last summer has been one of most important stages for
me among these four years’ endeavor. Without your valuable advice I would not have
survived the last step in graduate school. I would like to acknowledge Dr. Tim C. Frank
in Dow Chemical Company for helping me understand and solve problems from the real
industrial point of view. I also appreciate the support from ASI Applied Systems and
Kaiser Optical Inc. for their instrument donation, which has saved me so much time that

I am able to ﬁnish timely.

I would like to say “thank you” to every member in Berglund group including Dilum,
Jennifer, Lili, Hasan, Dale, Mike, Anna, Parminder, Sri, Charles, Carina, Sanjay and

Adam. Each one of you has helped to made my time easier in many ways. I would also

iv

like to thank all my friends for their encouragement and friendship, including Hongying,
Liu Qiang, Chen J ing, Shi Yi and Liu Bangzhi. Graduate school in the U. S. would have

seemed a lot longer without the pleasant time we had together.

I am indebted to my family for all their care, support and encouragement. Hearing my
parents’ voice has been the most comforting thing throughout my stay in this distant
country. I am thankful to my husband, Yonggang, for all the sacriﬁce you made for me
over these years. Thank you for all the endless discussions with me. It is your critisizm

that helps me grow and your comfort that keeps me going.

TABLE OF CONTENTS

List of Tables ......................................................................................... x
List of Figures ........................................................................................ xi
Chapter 1
INTRODUCTION ................................................................................... 1
1.1 Background for pH Swing Crystallization Study .............................. 2
1.1.1 Crystallization .............................................................. 3
1.1.1.1 Supersaturation ................................................... 3
1.1.1.2 Nucleation ......................................................... 4
1.1.1.3 Crystal growth ..................................................... 6
1.1.2 Attenuated Total Reﬂectance Fourier Transform Infrared
Spectroscopy (ATR-FTIR) ............................................... 7
1.1.3 Partial Least Squares Regression Method (PLS) ..................... 10
1.2 Background for In Situ Monitoring of Polymorphic Transformation of
Progesterone ........................................................................ 13
References ................................................................................... 1 5
Chapter 2

MONITORING pH SWING CRYSTALLIZATION OF NICOTINIC ACID BY THE
USE OF ATTENUATED TOTAL REFLECTION F OURIER TRANSFORM

INFRARED SPECTROMETRY
......................................................................................................... 17
2. 1 Introduction ........................................................................ 1 7
2.2 Experimental Section ............................................................. 19
2.3 Results and Discussion ........................................................... 21
2.3.1 The Structures of HNic and NaNic in Aqueous Solutions with
Different pH and IR Peak Assignments ................................ 21

vi

2.3.2 Calibrations of NaNic and HNic + NaNic in Aqueous Solutions

....................................................................................... 25
2.3.3 Monitoring the Crystallization Process ................................ 29
2.4 Conclusions ........................................................................ 33
2.5 Acknowledgments ................................................................. 33
References ................................................................................... 34
Chapter 3
OPTIMIZATION OF THE PROCESS PARAMETERS FOR SPONTANEOUS pH
SWING CRYSTALLIZATION OF NICOTINIC ACID
......................................................................................................... 35
3. 1 Introduction ......................................................................... 3 5
3.2 Experimental Section ............................................................. 38
3.2.1 Apparatus .................................................................. 38
3.2.2 Procedures ................................................................. 39
3.3 Results .............................................................................. 41
3.4 Discussion .......................................................................... 49
3.4.1 Effect of Acid Addition Rate — The Apparent Order of NucleatiosnO
3.4.2 Effect of Acid Concentration ........................................... 52
3.4.3 Effect of Agitation Speed ................................................ 53
3.5 Conclusions ........................................................................ 54
3.6 Acknowledgments .................................................................. 55
References ................................................................................... 56

vii

Chapter 4

FEEDBACK CONTROL OF SEEDED SEMI-BATCH pH SWING
CRYSTALLIZATION OF NICOTINIC ACID

......................................................................................................... 57
4. 1 Introduction ......................................................................... 5 7

4.2 Experimental Section .............................................................. 59

4.2.1 Apparatus .................................................................. 59

4.2.2 Interface between the LabMax® and the ReactIR 1000. . . .. . ...60
4.2.3 Procedures .................................................................. 61

4.2.3.1 Seeding at Different Levels of Supersaturation. . . . . . ...61

4.2.3.2 Seeding with Post Addition of HCl ........................... 63
4.3 Results and Discussion ............................................................ 64
4.3.1 Seeding at Different Levels of Supersaturation ....................... 64

4.3.2 Seeding with Post-seeding Acid Addition Controlled by
Supersaturation ........................................................... 75
4.4 Conclusions ......................................................................... 87
References ................................................................................... 88

Chapter 5

AN INVESTIGATION OF SOLVENT-MEDIATED POLYMORPHIC
TRANSFORMATION OF PROGESTERONE USING IN SI TU RAMAN
SPECTROSCOPY

.......................................................................................................... 89
5. 1 Introduction ......................................................................... 90
5.2 Experimental Section .............................................................. 92
5.3 Results and Discussion ........................................................... 94
5.4 Conclusions ....................................................................... 107

viii

5.5 Acknowledgments ............................................................... 107

References .................................................................................. 1 09
Chapter 6
CONCLUSIONS .................................................................................. 1 1 1
APPENDIX ......................................................................................... 1 14

Table 2.1

Table 2.2

Table 3.1

Table 3.2

Table 3.3

Table 3.4

Table 3.5

Table 4.1

Table 4.2

Table 4.3

LIST OF TABLES

The working range of the diamond ATR crystal.

....................................................................................... 19
The IR peak assignments for nicotinic acid and sodium nicotinate in
aqueous solutions.
....................................................................................... 25
Metastable limit comparison for different addition rates of 2 M HCl.
....................................................................................... 45
Metastable limit comparison for different acid concentrations.
....................................................................................... 45
Metastable limit comparison for different agitation speeds.
....................................................................................... 46
Metastable limit comparison for different addition rates of 0.5 M HCl.
....................................................................................... 49

The rate of decrease of the concentration of sodium nicotinate at different
acid addition rates and their corresponding metastable zone width.

....................................................................................... 51
The number of crystals of each size and the mean crystal size of the
product from 60% relative supersaturation seeding of nicotinic acid, the
seed concentration is 1.5 g seeds/kg solution.
....................................................................................... 70
The number of crystals of each size and the mean crystal size of the
product from 30% relative supersaturation seeding of nicotinic acid, the
seed concentration is 1.5 g seeds/kg solution.
....................................................................................... 72
The mean crystal size from three experiments which are controlled by
different levels of relative supersaturation.
....................................................................................... 86

Figure 1.1

Figure 1.2

Figure 1.3

Figure 1.4

Figure 1.5

Figure 2.1

Figure 2.2

Figure 2.3

Figure 2.4

Figure 2.5

LIST OF FIGURES

Types of nucleation.

......................................................................................... 4
Degrees of supersaturation related to different nucleation processes.
......................................................................................... 5
Schematic showing total internal reﬂection.
......................................................................................... 8

ATR-IR probe immersed in a slurry with the number of reﬂection points,
N = 6.
.......................................................................................... 9

Factors of the data set in Principle Component Regression.
........................................................................................ 1 1

Schematic diagram of the crystallization experiment setup. A. water
jacketed acid container, B. pump, C. water-j acketed crystallizer, D.
diamond ATR probe, E. spectrometer, F. computer, M. Marine type
impeller, P. pH meter, T. temperature control.

........................................................................................ 20

IR spectrum of nicotinic acid in water. The concentration of nicotinic acid
is 2.72 x 10‘3 mole fraction with pH of 3.22. The temperature is 30.0 i:
0.1 °C.

........................................................................................ 23

IR spectrum of sodium nicotinate in water. The concentration of sodium
nicotinate is 6.04 x 10'3 mole ﬁaction with pH of 9.33. The temperature is
30.0 i 0.1°C.

........................................................................................ 24

IR spectrum of nicotinic acid aﬁer subtraction of the spectrum of water.
The concentration of nicotinic acid is 2.72 x 10'3 mole ﬁ'action with pH of
3.22. The temperature is 30.0 i 0.1°C.

........................................................................................ 26

IR spectrum of nicotinic acid in deuterated water (D20). The
concentration of nicotinic acid is 2.70 x 10’3 mole fraction. The
temperature is 30.0 i 0.1°C. D20 spectrum is not subtracted.

........................................................................................ 27

Figure 2.6

Figure 2.7

Figure 3.1

Figure 3.2

Figure 3.3

Figure 3.4

Figure 3.5

Figure 4.1

Solubility of nicotinic acid in sodium nicotinate solutions of various pH at
T = 30.0 i 0.1°C.
........................................................................................ 30

Crystallization of nicotinic acid while 1 M HCl is being added at the rate
of 1 mL/min to sodium nicotinate solution at 30.0 i 0.1 °C with the
starting concentration of sodium nicotinate of 8.14 x 10'3 mole fraction.
Region I: Undersaturated nicotinic acid in sodium nicotinate aqueous
solution; Region II: Supersaturated solution with the degree of
supersaturation increasing, no crystal forming; Region III: Rapid
desupersaturation, nucleation and crystal grth occur; Region IV:
Return to saturation, nucleation and crystal grth stop.

........................................................................................ 32

The concentration of nicotinic acid vs. the concentration of sodium
nicotinate during the metastable zone measurement of the spontaneous
crystallization of nicotinic acid at 30 i 01°C, with agitation speed of 1.1
m/s and acid addition rate of 60 g/hr of 2 M HCl.

........................................................................................ 42

The metastable zone of the crystallization of nicotinic acid measured
under the conditions that the solution is agitated at 1.1 m/s and 2 M HCl is
added at the rate of 60 g/hr at 30°C.

........................................................................................ 43

Comparison of the ﬁrst metastable limit measurement with different acid
addition rates, 1 g/hr and 60 g/hr. The acid concentration is 2 M HCl and
the agitation speed is 1.1 m/s. The temperature is 30 i 0.1°C.

........................................................................................ 44

Inﬂuence of the acid concentration on crystal size: (a): 2 M HCl, addition
rate of 1 g/hr and 1.1 m/s agitation speed, The average size of the product
is about 25 mm. (b) 0.5 M HCl, addition rate of 1 gm and 1.1 m/s
agitation speed, (b.1), (b.2) and (b.3) are the crystals after sieving. The
average size of (b.l) is 152 pm, that of (b.2) is 214 pm and that of (b.3) is
305 um.

........................................................................................ 47

The crystal size distribution of the product from the experiment with the
addition of 0.5 M HCl at the rate of 1 g/hr and agitation speed of 1.1 m/s.
........................................................................................ 48

Concentration proﬁle of seeded crystallization of nicotinic acid at 60%
relative supersaturation seeding and the seed concentration of 1.5 g

xii

Figure 4.2

Figure 4.3

Figure 4.4

Figure 4.5

Figure 4.6

Figure 4.7

Figure 4.8

Figure 4.9

seeds/kg of solution added at 30 mins after the relative supersaturation
reached 60%. T = 30 i 0.1°C.
........................................................................................ 65

Relative supersaturation proﬁle and HCl addition proﬁle of seeded
crystallization of nicotinic acid at 60% relative supersaturation seeding
and the seed concentration of 1.5 g seeds/kg of solution added at 30 mins
after the relative supersaturation reached 60%. T = 30 :1: 0.1°C.

........................................................................................ 66

The turbidity proﬁle for 60% relative supersaturation seeding
crystallization of nicotinic acid. The seed concentration is 1.5 g seeds/kg
solution. T = 30 i 0.1°C.

........................................................................................ 68

The crystal size distribution of the products from 60% relative
supersaturation seeding crystallization of nicotinic acid. The seed
concentration is 1.5 g seeds/kg solution. The average size of the seeds is
152 pm.

........................................................................................ 69

Relative supersaturation proﬁles comparison for 30% and 60% relative
supersaturation seeding crystallization experiments of nicotinic acid. The
seed concentration is 1.5 g seeds/kg solution.

........................................................................................ 71

The comparison of the crystal size distributions resulted from seedings at
three levels of relative supersaturation, 30%, 45% and 60%. The seed
concentration is 1.5 g seeds/kg solution.

........................................................................................ 73

The mean crystal size dependence on the level of the relative
supersaturation at which seeding is performed. The seed concentration is
1.5 g seeds/kg of solution. The average size of the seeds is 152 pm. The
line between the datum points is meant only to be a guide to the eye.

........................................................................................ 74

Concentration proﬁles of sodium nicotinate and nicotinic acid and the
solubility and metastable limit of nicotinic acid for crystallization
controlled at about 30% relative supersaturation with post seeding acid
addition.

........................................................................................ 76

Relative supersaturation proﬁle and HCl addition proﬁle for crystallization
controlled at about 30% relative supersaturation with post seeding acid

xiii

Figure 4.10

Figure 4.11

Figure 4.12

Figure 4.13

Figure 4.14

Figure 5.1

Figure 5.2

addition. The seeds are added at 30% relative supersaturation. The seed
concentration is 1.5 g seeds/kg solution.
........................................................................................ 77

Concentration proﬁles of sodium nicotinate and nicotinic acid and the
solubility and metastable limit of nicotinic acid for the crystallization
controlled at about 0% relative supersaturation with post seeding acid
addition. The initial seeding is performed at just above saturation.

........................................................................................ 80

Relative supersaturation proﬁle and HCl addition proﬁle for crystallization
controlled at about 0% relative supersaturation with post seeding acid
addition. The seeds are added at just above 0% relative supersaturation.
The seed concentration is 1.5 g seeds/kg solution.

........................................................................................ 81

Concentration proﬁles of sodium nicotinate and nicotinic acid and the
solubility and metastable limit of nicotinic acid for the crystallization
controlled about 0% relative supersaturation with post seeding acid
addition. The initial seeding was performed at 30% relative
supersaturation.

........................................................................................ 83

Relative supersaturation proﬁle and HCl addition proﬁle for crystallization
controlled at about 0% relative supersaturation with post seeding acid
addition. The seeds are added at 30% relative supersaturation. The seed
concentration is 1.5 g seeds/kg solution.

........................................................................................ 84

The crystal size distribution from three controlled supersaturation
experiments. The three cases are: I. Seeding at 0% relative
supersaturation, and controlled about 0% relative supersaturation during
post acid addition; H. Seeding at 30% relative supersaturation, and
controlled about 0% relative supersaturation during post seeding acid
addition; 111. Seeding at 30% relative supersaturation, and controlled at
about 30% relative supersaturation during post acid addition.
........................................................................................ 85
The X-ray diffraction patterns of progesterone polymorphs: (A) Form I
and (B) Form II.
........................................................................................ 95

The DSC thermodiagrams of progesterone: (A) Form I and (B) Form 11.

The heating rate was 5 °C/min.
........................................................................................ 96

xiv

Figure 5.3

Figure 5.4

Figure 5.5

Figure 5.6

Figure 5.7

Figure 5.8

Raman spectra of solid state progesterone: (A) Form I and (B) Form 11.
......................................................................................... 98

Molecular structure of progesterone.
......................................................................................... 99

Calibration curve for the relative concentration of Form I in the mixture of
Form I and Form H of progesterone slurried in water.
....................................................................................... 100

Raman spectra showing peak shift during the phase transformation process
at 45°C.
....................................................................................... 102

Concentration proﬁles of Form I and Form 11 of progesterone throughout
the phase transformation process at 45°C. The lines are drawn through the
data as a guide to the eye.

....................................................................................... 103

The temperature dependence of phase transformation rate of
progesterone.The line is drawn through the data as a guide to the eye.
....................................................................................... 104

XV

Chapter 1

INTRODUCTION

Crystallization is one of the most widely used separation processes in the food,
pharmaceutical, and chemical industries. Control of the crystallization processes will
enable the production of a crystalline product with the desired physical and chemical
properties. Crystal size distribution (CSD) is one of the most important targets for
control. A controlled crystallization process yields crystals with a favorable CSD
signiﬁcantly reducing the cost of downstream processes such as ﬁltration and drying.
Crystals with uniform size improve the ﬂow, handling and packaging in the later steps of
the production processes. Polymorphism is another signiﬁcant property that draws a lot
of attention, particularly in the pharmaceutical crystallization processes. Polymorphs are
different crystal forms of the same compound. Different polymorphs have different
physical and chemical properties including dissolution rates and bioavailability. The
polymorph with desired physical and chemical properties can be obtained through the

optimization of the crystallization process parameters.

This work consists of two studies. The ﬁrst study focuses on improvement of CSD
through the optimization of pH swing crystallization. The second study aims at obtaining
the desired polymorph by in situ monitoring of the polymorphic transformation process

and subsequent optimization of the process parameters.

1.1 Background for pH Swing Crystallization Study

In order to obtain crystals with desired CSD, it is necessary to understand the critical
factor that controls the ﬁnal CSD from a batch or semi-batch crystallization process.
Supersaturation is the driving force for crystallization to occur. Once enough
supersaturation is generated, the solute molecules aggregate to form nuclei, the
aggregation process is called nucleation. Further integration of molecules to the nuclei is
deﬁned as crystal growth. The degree of supersaturation governs the rates of nucleation
and crystal growth. The competition between nucleation and crystal growth determines
the ﬁnal CSD. Therefore, the in situ control of crystallization requires measurement and

subsequent control of supersaturation.

A number of studies have been conducted to measure supersaturation in conventional
crystallizations such as cooling (temperature control), evaporation of solvent (solvent
control) and addition of antisolvent (solvent control). 1’ 2’ 3’ 4’ 5’ 6 However, very little work
has been focused on the measurement and control necessary for pH swing crystallization
although it is a common practice in the chemical industry. In a pH swing crystallization,
the solubility difference between a protonated and a deprotonated species is exploited to
create supersaturation by the adjustment of pH. We performed in situ measurement and
control of supersaturation during pH swing crystallization of nicotinic acid using
Attenuated Total Reﬂectance-Fourier Transform Infrared Spectroscopy (ATR-FTIR)

coupled with the Partial Least Squares regression (PLS) method.

1.1.1 Crystallization

Crystallization proceeds by the formation of nuclei, followed by crystal growth. It is a
complex process that is governed by both thermodynamics and kinetics of the interface
between the crystalline phase and solution phase. The onset of crystallization is the result
of the driving force caused by the thermodynamic equilibrium between a supersaturated
solution and a saturated solution. The kinetics of mass transfer from the bulk solution

phase to the solid surface determines the nucleation rate and crystal grth rate.

I. 1. 1. 1 Supersaturation

Supersaturation needs to be generated in order to achieve nucleation and crystal growth.
Supersaturation is deﬁned as the state of solution that has a concentration greater than the
equilibrium saturation concentration at a given temperature. The fundamental driving
force for crystallization is the difference between chemical potential of the given
substance in the solution phase and in the crystalline state. Under a supersaturated
condition, the chemical potential difference provides the driving force for the phase
transition from solution to solid crystals. The following equation expresses the

relationship between chemical potential and supersaturation.8

Ap=RTln(S) (1.1)
where 411 is the chemical potential difference between a supersaturated solution and a

saturated solution, R is the gas constant, and T is the absolute temperature.

. . a . . .
Supersaturation, S, rs deﬁned as S = —7, where a rs the actrvrty of the supersaturated
a

solution and a' is the activity of the saturated solution. Typically the ratio of activity

coefﬁcients is assumed to be unity; therefore, we can replace activity with concentration.
Under the assumption, supersaturation can be approximated to the ratio of the

concentration of the supersaturated solution to the concentration of the saturated solution.

1.1.1.2 Nucleation

Alter supersaturation is generated, crystal formation from the solution takes place. There
are two steps involved in the formation of a crystal. Nucleation ﬁrst takes place,
followed by crystal growth. If solid foreign particles are present in a supersaturated
solution, nucleation can be facilitated by the so-called heteronuclei; the nucleation based
on the heteronuclei is deﬁned as heterogeneous nucleation. Otherwise, homogeneous
nucleation will take place as a result of a large chemical potential driving force without
the presence of foreign particles. It is generally accepted that true homogeneous
nucleation is a not a common event. Both homogeneous and heterogeneous nucleation
take place in the absence of crystals of its own type. Secondary nucleation occurs in the
presence of growing crystals. Secondary nucleation takes place at a lower

supersaturation than primary nucleation. This is summarized in Figures 1.1 and 1.2.9

 

——-[ homogeneous I
J1 primary I—

aneleatimJ- ——l heterogeneous}

1f secondary I

 

 

 

 

 

 

 

 

 

 

Figure 1.1. Types of nucleation

x/ /’
II /
I /
/
c: o /
2 8° 9
*5 9&6 .90
3:: 9 c)“
5 ,{Q'ﬁ 9° _
° «20 qﬁl \\9
‘3 6 ”<3
8 009 X“
x o ‘0
x’ / 9 sow»
’ /
/
/
/
/ /

 

 

pH, or temperature, etc.

Figure 1.2. Degrees of supersaturation related to different nucleation processes.

Although no single nucleation theory is accepted for predicting nucleation rate, several
correlations based on the power law model have satisfactorily explained the experimental

data,10

B=k~ Ac;W 02)
Where B is the nucleation rate, k N is the nucleation rate constant and AC W is the

metastable zone width. The exponent n is the apparent order of nucleation. During the

process of nucleus formation, there is an equilibrium between the growth of nuclei and
the dissolving of nuclei. Some of the nuclei reach a critical size, which is dependent on
the change of the free energy of the particle, and they become stable nuclei. Post

nucleation growth is made possible in the presence of stable nuclei.

1.1.1.3 Crystal Growth

Various theories have been proposed to represent crystal growth, including the Two-
Dimensional Growth, the BCF Model, and the Diffusion Layer Model. These models
provide a theoretical basis to correlate the experimental data with the kinetic parameters.

The basic expression relating crystal grth and supersaturation is10

G=kgSg (1.3)

which employs a linear crystal growth velocity correlating the grth rate (G) with the
supersaturation (S). k8 is deﬁned as the grth rate constant and the exponent g is the

order of crystal growth.

Comparing growth kinetics with nucleation kinetics helps us understand the effect of
supersaturation on the ﬁnal crystal size distribution. In most systems, the order of
nucleation, n, is greater than g, the order of crystal growth.11 Consequently, higher
supersaturation favors nucleation and lower supersaturation promotes crystal growth.
Therefore, supersaturation becomes the critical parameter for controlling the kinetics of
crystal growth and nucleation. The purpose of obtaining large mean crystal size in
industrial crystallization processes requires maintaining the supersaturation at low levels,
or close to the solubility, so as to suppress nucleation and promote crystal growth. This
requires real time measurement and control of supersaturation. Attenuated Total
Reﬂection Fourier Transform Infrared spectroscopy has been demonstrated to be a viable

technique for this purposef”5

1.1.2 Attenuated Total Reﬂectance Fourier Transform Infrared Spectroscopy (A T R-

FTIR)

Fourier Transform infrared (FTIR) spectroscopy is a widely used spectroscopic method
for both quantitative and qualitative measurements. Infrared spectroscopy is classiﬁed as
vibrational spectroscopy in that the absorption of infrared radiation is associated with the
molecular vibrations. Therefore, FTIR is able to probe the molecular changes in the
sample. The spectra recorded through FTIR spectroscopy require short collection times,
e.g. on the order of seconds, which is a very important feature for the purpose of in situ

control of supersaturation within the time frame of nucleation and crystal growth.

Total internal reﬂectance as a general-purpose sampling technique has paralleled the
growth and development of FTIR. In total internal reﬂectance, also called attenuated
total reﬂectance (ATR), the infrared radiation is directed into the interface between the
reﬂectance element (ATR crystal), and the sample in such an angle that all the radiation
is reﬂected. To achieve internal reﬂection, the reﬂectance element must be of higher
optical density than the sample is, or the refractive index of the ATR crystal must be
greater than that of the sample, r7, > 0,. Also, the incident angle of the radiation has to be
greater than the critical angle, which is equal to sin"(ry,/ 77C). According to Cothup12 and
Mirabella13 , despite the total reﬂection of radiation at the interface, radiation penetrates a
small distance into the sample. This penetrating radiation is called the evanescent wave,

as shown in Figure 1.3.

     
  

'lc
ATR crystal

 

Evanescent

Intensity

sample wave

I >
Tls | Distance from the interface

 

 

Figure 1.3. Schematic showing total internal reﬂection.

A sample is capable of absorbing the radiation from the evanescent wave at speciﬁc
wavenumbers. Therefore, the reﬂected beam is attenuated at the characteristic
wavenumbers corresponding to the absorption bands, which appear in the spectrum of the

sample.

According to Beer’s Law, A = abC, where A is absorbance, a is absorptivity, b is optical
pathlength, and C is concentration, a calibration curve of A versus C can be constructed if
the optical pathlength at a speciﬁc wavenumber is constant for samples of different
concentrations and the absorptivity is constant. In an ATR-FTIR conﬁguration, an
effective optical pathlength is used. It is deﬁned as the product of the number of

reﬂection points in contact with the sample on the surface of reﬂection element, N, and

the depth of penetration, dp, as shown in Figure 1.4.14 The wavelength dependent

penetration depth of the evanescent wave is given as 15’ 1°

dP: . 2 1C 7 1/2 (14)
{Zelsm 6-(77./rr.)‘]}

 

where Ac is the wavelength of the IR radiation in the ATR crystal, and 0 is the incident

angle.

It was shown thath is nearly constant for species with different refractive index.17 Also,
Ndp is so small, on the order of microns, that the sample thickness in current research
always exceeds the effective optical pathlength. Therefore, variations in thickness do not
affect the effective optical pathlength. The ATR absorption is proportional to

concentration, thus useful quantitative measurements can be made with the ATR-FTIR

 

 

 

 

 

 

 

 

 

 

technique.
m ,_IRSQHLQL,
. | —————————————————————————— I
Probenmnersed___>: 6 l ‘___:_ ATR C1359]
msolutron I ____________________________ I
_o ——'o — o
O O _
o o O o O

 

 

 

0 Crystal in slurry, N = 6
Figure 1.4. ATR-IR probe immersed in a slurry with the number of reﬂection points,

N=6.

The combination of FTIR and ATR makes the instrument more efﬁcient and more
accessible to different samples. ATR extends the role of FTIR into new areas of
application, and the improved performance of FTIR has helped attenuated total reﬂection

to become one of the most ﬂexible sampling techniques.

1.1.3 Partial Least Squares Regression (PLS) Method

In this study, pH swing crystallization is carried out through a reaction which converts
the salt to its corresponding acid form, in the presence of water. This situation presents a
challenge to determine the concentrations of both the salt and the acid accurately because
of the similarity in their structures. The conventional data analysis method such as
univariate calibration based on peak area or a single point peak height can no longer

predict the unknown concentration of a sample accurately.

More reliable data analysis techniques based on multivariate calibration, including
Principle Component Regression (PCR) and Partial Least Squares Regression (PLS)
methods,18’ '9 have been developed and are being widely utilized for quantitative analysis.
Typically, we have two data matrices X and Y to be correlated with each other. In the
current work, the elements in matrix X are the 1R absorbances at all the wavenumbers
within the speciﬁed peak region which correspond to certain vibrations of the sample.
The concentrations of the standard samples form another matrix Y. The goal of the data
treatment is to build a mathematical model, on the basis of the two matrices X and Y, to
predict the concentrations in matrix Y. Both PCR and PLS are factor-based modeling

techniques.

10

In PCR, a factor is deﬁned as a vector that describes the maximum amount of variation
in the sample. Mathematically, the ﬁrst factor will be the eigenvector with the largest
eigenvalue for the covariance matrix of the data set, X. If one factor cannot account for
all of the variation, a second or even more eigenvectors need to be identiﬁed to describe
the maximum amount of the residual variance. This can be explained graphically in
Figure 1.5. The signiﬁcance of the factors is that they best describe the variation within
the sample with the minimum amount of data. Once the factors are determined, the
original data set, X, will be re-expressed as a score matrix T, which is the projection of
the original data set onto the eigenvectors, W. Then the regression coefﬁcients matrix,
Q, can be determined through multi-linear regression of Y matrix onto the score matrix.
The prediction for the concentration of unknown samples can be performed using

eigenvectors, W, and the regression coefﬁcients matrix, Q.

 
   

F actorl

X2

Figure 1.5. Factors of the data set in Principle Component Regression.

PLS differs from PCR in that it simultaneously estimates factors for both X and Y
matrices. W, consisting of the factors of matrix X, describes the variation in Y better
than the factors in PCR. This is achieved by starting with the columns of Y matrix to
calculate the factors for X and using iterative substitution to determine all of the factors
in both X and Y matrices. The factor matrix of X in PLS will deviate from the direction
of the factors in PCR by being slightly rotated toward the Y vectors. It is this slight
rotation that uses all the information available including both X and Y matrices to
determine the inner relationship coefﬁcient b between T and U, the score matrices of X
and Y, respectively. After the absorbance matrix, Z, from an unknown sample is
measured, the score matrix Tm, is determined using W, the factor matrix of X. Utilizing
the coefficient b, we can obtain the score matrix Uun, based on Y matrix. This new score
matrix Um, is transformed into estimated concentration using Q, the factor matrix of Y.”’

20 This can be represented by the following equations.

X = T.W (1.5)
11
Y = U.Q (1.6)

Xun 19TH —b—> U... —9—> Y

12

1.2 Background for In Situ Monitoring of Polymorphic Transformation of
Progesterone

The pharmaceutical industry is frequently confronted with the different physiochemical

properties of polymorphs of the same compound, particularly with solid, oral dosage drug

products.” 22

The characteristics affected by polymorphism include solubility,
dissolution rate, stability, hygroscopicity and solid state reactions. The effect of
polymorphism on bioavailability is the most important consequence if the bioavailability
is mediated via dissolution.” 23 In situ monitoring of the solvent mediated polymorphic
transformation is therefore of great signiﬁcance in understanding thermodynamics of the

polymorphic systems and mechanisms of the polymorphic transformation, and

consequently determining the optimized crystallization parameters to produce the desired

polymorphs.

In polymorphic studies, Raman spectroscopy, coupled with an immersible ﬁber optic
probe, has tremendous advantages over the traditional techniques, including X-ray
powder diffraction (XRD), Differential Scanning Calorimetry (DSC), solid state NMR
and Infrared spectroscopy (IR). The ﬁber optic probe in the Raman technique provides
the capability of in situ or on-line monitoring of the formation of crystals and the
transformation from one polymorph to another. DSC, XRD and NMR can perform off-
line analysis, which can distinguish the polymorph of the end product,24 but cannot
provide insight about transformation processes since the transformation may continue

during drying and sampling.

13

Infrared Spectroscopy can be used to in situ monitor some processes. However aqueous
solutions may cause problems in IR measurement.12 If water is used as either a reagent
or a byproduct or a solvent in a process, strong infrared absorption of water can mask
some of the useful information from the interested species. In contrast, water is a poor
Raman scatterer, which often makes Raman preferable to IR when water is present in the

5 This can be explained by the different mechanisms of interaction of

processes.2
electromagnetic radiation with matter in the two spectroscopic techniques. Dipole
moment changes during molecular vibrations cause IR absorption while changes in
polarizability associated with molecular vibrations result in Raman scattering. Usually
molecules with polar functional groups and asymmetric vibrational modes have stronger
absorption in IR and weaker absorption in Raman.15 Raman scatterring also favors solid
phase in a slurry other than the solution phase as the ATR-FT IR technique. The higher
density of the solid phase results in stronger Raman scattering than the solution phase
does. This allows us to monitor the disappearance of one solid phase and the appearance
of the other solid phase. In addition, we can use Raman spectroscopy to perform remote
sensing since the ﬁber optic probe can be made as long as a couple of hundred meters.26
By comparison, Attenuated Total Reﬂectance (ATR) probe in ATR-FTIR technique can
only reach a couple of meters.27 Remote sensing is especially useful in hazardous
environments where operators can work safely by staying far from the reactor. This

study employed Raman spectroscopy coupled with a ﬁber optic probe to study the

polymorphic transformation of progesterone.

l4

References

1. Garside, J .; Gibilara, L. B; Tavare, N. S. Chem. Eng. Sci. 1982, 37, 1625.

2. Tavare, N. S.; Garside, J. Chem. Eng. Res. Des. 1986, 64, 109.

3. Qiu, Y.; Rasmusson, A. C. AIChE J. 1990, 36, 665.

4. Dunuwila, D. D.; Berglund K. A. J. Cryst. Growth, 1997, 179, 185.

5. Uusi-Penttila, M. S.; Berglund, K. A. J. Cryst. Growth, 1996, I66, 967.

6. Schwartz, A. M.; Berglund, K. A.; J. Cryst. Growth, 1999, 203, 599.

7. Wang, F.; Berglund, K. A. Ind. Eng. Chem. Res. 2000, 39, 2101.

8. Mullin, J. W. Crystallization, 3rd ed.; Butterworth-Heinamann Ltd.: Oxford, 1993.

9. Mersmann, S. Crystallization Technology, 3rd ed., Butterworth—Heinamann Ltd.:
Oxford, 1995.

10. Myerson, A. S. Handbook of Industrial Crystallization, Butterworth-Heinamann Ltd.:
Boston, 1993.

11. Sb'hnel, O.; Gardise, J. Precipitation: Basic Principles and Industrial Applications,
Butterworth-Heinamann Ltd.: Oxford, 1992.

12. Cothup, N. b.; Daly, L. H.; Wiberly, S. E. Introduction to Infrared and Raman
Spectroscopy, 3rd ed., Academic Press: Boston, 1990.

13. Mirabella, F. M., Jr. Internal Reﬂection Spectroscopy, Theory and Applicaitions,
Mercel Dekker, Inc.: New York, 1993.

14. Coates, J. P. The Industrial Application of Infrared Internal Reflection Spectroscopy,
In Internal Reﬂection Spectroscopy, Theory and Applicaitions, Mercel Dekker, Inc.: New
York, 1993.

15. Ingle, J. D., Jr.; Crouch, S. R. Spectrochemical Analysis, Prentice-Hall: New Jersey,
1988.

16. Muller, F. J .; Abraham-Fuchs, K. Matrix Dependence in Single and Multilayer IRS
Spectra, In Internal Reﬂection Spectroscopy, Theory and Applicaitions, Mercel Dekker,
Inc.: New York, 1993.

17. Uusi-Penttila, S. M. Optimization of Batch Antisolvent Crystallization, PhD
Dissertation, Michigan State University, East Lansing, Michigan, 1997.

15

 

18. Beebe, K. R.; Pell, R. J .; Seasholtz, M. B. Chemometrics, A Practical Guide, John
Wiley & Sons, Inc.: New York, 1998.

19. Beebe, K. R.; Kowalski, B. R. Anal. Chem. 1987, 59, 1007A.

20. Esbensen, K.l Schonkopf, S.; Mictgaard, T. Multivariate analysis in practice;
Computer-aided modelling AS: Trondheim, Norway, 1996.

21. Yu, L.; Reutzel, S. M.; Stephenson, G. Pharm. Sci. T echnol. T o. 1998, 1(3), 118.
22. Giron, D. Ihermochim. Acta, 1995, 248, 1.
23. Bym, S. R. Solid-State Chemistry of Drugs, Academic Press: New York. 1982.

24. Raghavan, K.; Dwivedi, A.; Campbell, G. C.; Nemeth, G.; Hussain, M. A. J. Pharm.
Biomed. Anal. 1994, 12(6), 777.

25. Pelletier, M. J. Analytical Applications of Raman Spectroscopy, Blackwell Science
Ltd: Oxford, 1999.

26. Kaiser Optical Systems, Inc. HoloLab Series 5000 Operations Manual, Kaiser
Optical Systems, Inc.: Ann Arbor, Michigan, 1997.

27. ASI Applied Systems, ReactIR 1000 User ’3 Guide; 3rd ed., ASI Applied Systems:
Millersville, Maryland, 1997.

16

Chapter 2

MONITORING pH SWING CRYSTALLIZATION OF NICOTINIC
ACID BY THE USE OF ATTENUATED TOTAL REFLECTION
FOURIER TRANSFORM INFRARED SPECTROMETRY*

*Industrial & Engineering Chemistry Research, 2000, 39, 2101-2104.

The crystallization of nicotinic acid is achieved by adding hydrochloric acid (HCl) to a
sodium nicotinate aqueous solution. An Attenuated Total Reﬂection Fourier Transform
Infrared (ATR-FTIR) spectrometer is used to monitor the crystallization process of
nicotinic acid. The Partial Least Squares regression (PLS) method is employed for IR
data analysis. The solubility of nicotinic acid varies with the pH and the concentration of
sodium nicotinate. The ability to measure the concentration and solubility of nicotinic
acid in situ in aqueous mixtures of nicotinic acid and sodium nicotinate provides the
possibility of controlling the degree of supersaturation of the solution and in turn

optimizing the crystallization conditions for nicotinic acid.

2.1 Introduction
Crystallization is one of the most widely used separation processes in the food,
pharmaceutical, and chemical industries. Control of the crystal size distribution (CSD)

signiﬁcantly impacts the cost of downstream processes such as ﬁltration and drying.

Supersaturation is the driving force for both nucleation and crystal growth. It controls the

rate of crystallization and the resulting CSD. Therefore, the control of a crystallization

17

process requires in situ measurement and subsequent control of supersaturation.
Signiﬁcant effort has been expended on the study of more conventional techniques of
supersaturation generation, including temperature control or cooling of solutions,1
solvent control such as evaporation of solvents and the addition of anti-solvents.2
However, far less work has been directed at the measurement and control necessary for

pH swing crystallization, which is a commonly encountered case in industrial

separations.

In pH swing crystallization, solubility differences between protonated and deprotonated
species are exploited to create supersaturation by the adjustment of pH. The chemical
system for the current study is nicotinic acid (HNic)-sodium nicotinate (NaNic)-water
(H20). At 30 oC, NaNic has a solubility of about 60 g/ 100 g of H20 at pH = 9.53, while
only 1.9 g of HNic can be dissolved in 100 g of H20 at pH = 3.22 at the same
temperature. The crystallization of HNic is achieved by the addition of hydrochloric acid
to NaNic solution to decrease pH. It is proposed to monitor the crystallization process by
the use of Attenuated Total Reﬂection Fourier Transform Infrared Spectroscopy (ATR-
FTIR). IR spectra provide information about the molecular structure and chemical nature
of the system under study. Any change in the system due to reaction will be reﬂected in
the IR spectra. Differences in spectra are of two types, peak intensity changes and peak
position shifts. Peak intensity changes are caused by concentration variation, and the
peak position shiﬁs are the result of the structural changes or intermolecular interactions.
The multivariate data analysis technique of partial least squares regression (PLS) can be

employed to analyze the spectral data obtained. This study demonstrates the feasibility of

18

 

using ATR-FTIR combined with the PLS data analysis method to monitor the

crystallization of HNic in situ.

2.2 Experimental Section

The experimental setup for monitoring the crystallization of nicotinic acid by ATR-FTIR
is shown in Figure 2.1. A 500 ml, water-jacketed crystallizer was kept at a constant
temperature (30.0 i- 0.1 °C) with a Brinkrnann RC6 LAUDA water circulator. The
crystallizer was equipped with a marine type impeller. A Cole-Parmer Masterﬂex tubing

pump was used to control the acid addition rates.

A dry air purged ASI-REACTIR 1000 Spectrometer was used for the ATR-FTIR
measurements. An MCT detector cooled with liquid nitrogen was employed to detect the
reﬂected light. A 0.25mm thick diamond wafer was used as a multiple reﬂection ATR
element. The working range of this material is shown in Table 2.1. 98% Nicotirric acid,
98% sodium nicotinate and hydrochloric acid of volumetric standard were purchased
from Aldrich Chemical Company. They were used without further puriﬁcation. Distilled
water was used in all experiments. In the solubility measurement, each datum point was

extracted from the spectrum that was collected when both pH and the concentration of

HNic had been stabilized.

Table 2.1. The working range of the diamond ATR crystal.3

 

 

pH Temperature Pressure Optical Range, cm'I
Range, °C
1-14 -80-250 -10 torr-100 psi 4400-2150 1950-650

 

 

 

@J or

1 . ASI
pH meter , ..._._. REACTIR
1000

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 2.1. Schematic diagram of the crystallization experiment setup. A. water jacketed
acid container, B. pump, C. water-jacketed crystallizer, D. diamond ATR probe, E.
spectrometer, F. computer, M. Marine type impeller, P. pH meter, T. temperature control.

2.3 Results and Discussion

2.3.1 The structures of HNic and NaNic in Aqueous Solutions with Diﬁ’erent pH and
IR Peak Assignments.

According to Nakomoto et a1. 4 and Wojcik et al.,5 the structure of HNic in aqueous
solutions of different acidities changes according to several equilibria. HNic in the solid
phase exists as the neutral molecule instead of the zwitterion. The equilibrium is shown

as follows when HNic is dissolved in water,

91.1.
N 0 N ,0
O c” —‘ @/ 6
\OH 1:10 K0

Zwitterion

The equilibria when HNic is dissolved in acidic aqueous solution are given below,

(all

O C//0 H“ O C//0

\OH \OH

 

 

 

 

21

 

The equilibria are as follows when HNic is dissolved in a basic solution or NaNic is

dissolved in a neutral aqueous solution.

0 - .O
©CK... 0 K. +
H
(9.. O

/0 OH- /
@Kf ©KO° +

HNic molecules form zwitterions when they dissolve in pure water. The quantity of

 

 

 

 

zwitterions is ahnost 10 times as much as that of the nonionic molecules.4 This explains
the absence of strong absorbance around 1700 cm'1 in Figure 2.2, which is characteristic
of carbonyl (C=0) stretching and is a common peak for un-ionized carboxylic acids.
Figures 2.2 and 2.3 present the IR spectra of HNic and NaNic dissolved in water,
respectively. Similar to amino acids, HNic can be considered as a dibasic acid. The
concentrations of the deprotonated nicotinate ions, zwitterions and protonated ions

depend on the pH and the dissociation constants of the protonated ions and zwitterions.

The spectrum of HNic does not provide much information in the presence of H20 except
for one peak around 1390 cm". One reason is that the concentration of an aqueous HNic
solution is low with its solubility only about 2%. The other reason is that some of the
peaks expected such as the vibrations of the pyridine ring are overlapped with or masked

by the strong peak at 1640 cm’1 from water.

22

 

 

 

 

 

 

0,6 . 1640
Q)
g 0.4
.D
H
O
(I)
.D
<
0.2 /
O 0 L . I n l . l n
l 800 1600 1400 1200 1000

Wavenumber (cm‘l)

Figure 2.2. IR spectrum of nicotinic acid in water. The concentration of nicotinic acid is
2.72 x 10"3 mole fraction with pH of 3.22. The temperature is 30.0 i 0.1 °C.

23

 

0.8

1640
1605
3
g 1390
'2 1556 R
8
.D
.<

 

 

 

 

 

 

 

0.0

1800 -1600 .1400 -1200 .1000

Wavenumber (cm")

Figure 2.3. IR spectrum of sodium nicotinate in water. The concentration of sodium
nicotinate is 6.04 x 10'3 mole fraction with pH of 9.33. The temperature is 30.0 i 01°C.

24

This effect is veriﬁed by the subtraction of the water spectrum from the spectrum of
HNic as shown in Figure 2.4 and it is also conﬁrmed by the spectrum of HNic in a D20

solution in Figure 2.5.

Because the solubility of NaNic is quite high, it is much easier to correlate the peak
positions with the molecular vibrations. According to the structure transformation of
HNic from un-ionized form in solid phase to ionized form in solution phase and
comparison between the spectra of HNic zwitterion and NaNic in aqueous solution
including H20 and D20 solutions,"’7 the following peak assignments were made and
shown in Table 2.2. The absorbance around 1390 cm'1 is the characteristic frequency for
ionized carboxyl group, which demonstrates that HNic molecules exist mainly as dipolar

ions in aqueous solutions.

Table 2.2. The IR peak assignments for nicotinic acid and sodium nicotinate in aqueous
solutions.

 

 

Vibrational Mode HNic peak N aNic peak
position (cm'l) position (cm'l)

Symmetric stretching of C00' 1390 1390
and pyridine ring deformation

Asymmetric extremely weak 1556

stretching of C00'
Pyridine Ring deformation 1586 1605
N-H deformation 1640

 

2.3.2. Calibrations of NaNic and HNic + NaNic in Aqueous Solutions
Because the chemical system used is a mixture of NaNic and HNic in water, it is

necessary to build calibrations for both NaNic and HNic in water in order to predict the

25

 

 

0.035

1390

I

0.030

I

0.025

0.020 L .
- 1586

I

0.015

Absorbance

1 640

 

 

I
m.

0.010

0.005 '-

b

0.000 r

I
rial

 

 

I. l n l 1 IQAI l n l

1800 1700 1600 1500 1400 1300 1200 1100 1000

 

Wavenumber (cm'l)

Figure 2.4. IR spectrum of nicotinic acid after subtraction of the spectrum of water. The
concentration of nicotinic acid is 2.72 x 10'3 mole fraction with pH of 3.22. The
temperature is 30.0 i 01°C.

26

 

 

 

 

 

 

 

 

0.12
1390
0.09 -- 1610
o
o
E
a
.D
a...
O
.8
< 0.06 ~
0.03
0
1900 1800 1700 1600 1500 1400 1300

wavenumber (cm'l)

Figure 2.5. IR spectrum of nicotinic acid in deuterated water (D20). The concentration
of nicotinic acid is 2.70 x 10'3 mole fraction. The temperature is 30.0 i 0.1°C. D20

spectrum is not subtracted.

27

composition of the system at any point. However, from the information in Table 2.2, it is
observed that HNic only exhibits three peaks, two of which are masked by the broad peak
around 1640 cm'1 from water. Subtraction of water from the spectra of HNic was
attempted, however it did not give a good calibration since the subtraction of the H20
spectrum from the HNic spectrum cannot eliminate the interaction between these two
compounds in solution. The only peak not obscured by water is around 1390 em], but
NaNic also has this peak at the same position, which implies that this peak cannot be
used to calibrate either HNic or NaNic independently in the mixture. However, this peak
region (1358 — 1416 cm") is very useful to calibrate the total amount of HNic and NaNic
present in the solutions. The contributions from HNic and NaNic are counted in the PLS
method and are validated by testing samples. For NaNic, there is a prominent peak
centered about 1556 cm", which can be used to monitor the concentration of NaNic in
the mixture. Now that both the concentrations of NaNic and HNic + NaNic can be
determined, the concentration of HNic can be obtained through the mass balance. The
calibrations of NaNic and HNic + NaNic were built using the peak regions 1510 — 1570

cm'1 and 1358 — 1416 cm'l, respectively.

To extract precise information for both HNic and NaNic, the Partial Least Squares
regression (PLS) technique is used to correlate the data from the IR spectra with the
concentrations of solutions being analyzed. PLS starts with two matrices X and Y. In the
current work, the elements in matrix X are the absorbances at all the wavenumbers within
the speciﬁed peak region which correspond to certain vibrations of the molecules such as

NaNic and HNic. The concentrations of the standard samples form another matrix Y.

28

The goal of the data treatment is to build a mathematical model, on the basis of the two
matrices X and Y, to predict the concentrations in matrix Y. The process involves
manipulating the two matrices to ﬁnd the eigenvectors of matrix X which are in the
directions of maximum covarianceg’9 The number of eigenvectors which corresponds to
the maximum covariance is also termed as the number of factors. The eigenvectors form
a new set of mutually independent variables, which are combined with the regression
coefficients to predict the concentrations of the samples in matrix Y. PLS calibrations for
NaNic and NaNic + HNic are built by the cross validation called the " leave-one-out"
method, in which one standard is left out and a calibration model is constructed using the
remaining standards. The concentration of the sample that is left out is then predicted
using this model. This process is repeated for every standard used in the calibration. The
optimal number of factors is determined as the one that gives a minimum predicted
residual error sum of squares G’RESS). The average percent errors for both of the

"leave-one-out" calibrations of NaNic and NaNic + HNic are less than 1%.

2. 3.3 Monitoring the Crystallization Process

Once the calibrations are built, it is possible to monitor the crystallization process of
HNic in situ. In order to better understand the crystallization of HNic, the solubility of
HNic in NaNic solution was obtained ﬁrst. It is found that the solubility of HNic varies
with the concentration of NaNic or with the pH value of the solution, as shown in Figure
2.6. This result can be explained by the ionic strength effect10 and the equilibrium
between nicotinate anions, protons and zwitterions of HNic. The crystallization process

of HNic was monitored and then compared with the solubility curve of HNic, which is

29

 

9° :5 z“

m o N

T I I
up!

9°
m
1

9°
.5
I

 

 

 

Solubility of HNic (mole fraction X 1000)
:3
r

 

3.0-
2.8-
’ I
2.6~
24. I L I n I 1 I n I L I n I n
20 25 30 35 4.0 45 50 55
pH

Figure 2.6. Solubility of nicotinic acid in sodium nicotinate solutions of various pH at T
= 30.0 :t 0.1 °C.

30

shown in Figure 2.7. As indicated in the ﬁgure, when a small amount of HCl is added
into NaNic aqueous solution, the concentration of HNic is below its solubility, which
corresponds to the stable undersaturated solution in region I. Once the concentration of
HNic exceeds the solubility curve, the solution enters the metastable zone, in regions 11
and III, which is deﬁned by the solubility and the highest supersaturation, called
metastable limit. When the concentration reaches this limit, spontaneous crystallization
occurs in region 111 and results in a large number of tiny nuclei. The concentration of
HNic returns to the solubility limit when nucleation and crystal growth end, as shown in
region IV. The ability to monitor the crystallization of HNic in situ is of great
signiﬁcance because it provides the possibility of controlling the degree of
supersaturation. As can be seen from Figure 2.7, the degree of supersaturation is very
high in region II, which is the favorable condition for nucleation to occur. Therefore,
numerous nuclei were formed and the supersaturation was all consumed by nucleation,
which is unfavorable for crystal growth. The undesired crystal size will lead to serious
problems in the later stage operations such as ﬁltration and drying. Ideally, the
supersaturation should be relatively high at ﬁrst so that enough nuclei can be produced.
After the desired amount of nuclei are formed, the supersaturation should be kept as close
to the solubility curve as possible so that the nuclei can grow further and form larger
crystals.ll Future work will be focused on controlling the crystallization process in order

to obtain crystals with a desired crystal size distribution.

31

 

4.6

 

. C(HNic) its
4.4 1 .

A The solubility of HNic ~
4.2 ’ .

 

 

3.8 '

3.6 0 .o’ ‘ o

000
. A

3.4 $45.2. ‘

O
“.O.’ O.. ‘Q
A

C(HNic) (mole fraction x 1000 )

3.2 ,

 

 

 

 

 

 

4 4.5 5 5.5 6 6.5 7 7.5 8 8.5
C(NaNic) (mole fraction x 1000)

Figure 2.7. Crystallization of nicotinic acid while 1 M HCl is being added at the rate of
1 mL/min to sodium nicotinate solution at 30.0 i 01°C with the starting concentration of
sodium nicotinate of 8.14 x 10'3 mole fraction. Region I: Undersaturated nicotinic acid in
sodium nicotinate aqueous solution; Region II: Supersaturated solution with the degree
of supersaturation increasing, no crystal forming; Region III: Rapid desupersaturation,
nucleation and crystal grth occur; Region IV: Return to saturation, nucleation and
crystal grth stop.

32

2.4 Conclusions

It was demonstrated that pH swing crystallization of nicotinic acid can be monitored in
situ with ATR-FTIR. The PLS method correlates the spectral information with the
concentrations of samples. The solubility of nicotinic acid varies as the concentration of
sodium nicotinate and pH of the solution change. The solubility curve of nicotinic acid
and the measurement of supersaturation in situ make it possible to control and optimize

the crystallization process of nicotinic acid.

2.5 Acknowledgments

The authors wish to express thanks to the Dow Chemical Company for the ﬁnancial

support of this project and also appreciate support provided by ASI Applied Systems.

33

References

1. Dunuwila, D. D. 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 process; Ph.D. Dissertation, Michigan State
University, 1996.

2. Uusi-Pentila, S. M. Optimization of batch anti-solvent crystallization; Ph.D.
dissertation, Michigan State University, 1997.

3. A81 Applied Systems, ReactIR 1000 User ’s Guide; 3rd Edition, ASI Applied Systems:
Millersville, MD,1997.

4. Nakamoto, K.; Morimoto, Y.; Martell, A. E. Infrared Spectra of Aqueous Solutions.
IH. Ethylenediarninetetraacetic Acid, N—Hydroxyethylenediaminetriacetic Acid and
Diethylenetriaminepentaacetic Acid. J. Amer. Chem. Soc. 1963, 85, 309.

5. Wojcik, J. F.; Stock, T. H. Aqueous Infrared Studies of the Pyridine Carboxylic
Acids. J. Phys. Chem. 1969, 73, 7, 2153.

6. Colthup, N. B.; Daly, L. H.; Wiberley, S. E. Introduction to infrared and raman
spectroscopy; 3rd Edition, Academic press: Boston, 1990.

7. Cohn, E. J. and Edsall, J. G. Proteins, amino acids and peptides as ions and dipolar
ions; Waverly press: New York, 1943.

8. Beebe, K. R.; Kowalski, B. R. An Introduction to Multivariate Calibration and
Analysis. Anal. Chem. 1987, 59, 17, 1017A.

9. Esbensen, K.; Schdnkopf, S.; Mictgaard, T. Multivariate Analysis in Practice;
Computer-aided modelling AS: Trondheim, Norway, 1996.

10. Green, R. W.; Tong, H. K. The Constitution of the Pyridine Monocarboxylic Acids
in their Isoelectric Forms. J. Amer. Chem. Soc. 1956, 78, 4896.

11. Mullin, J. W. Crystallization; 3rd Edition, Butterworth-Heinemann Ltd: Oxford,
1993.

34

Chapter 3

OPTIMIZATION OF THE PROCESS PARAMETERS FOR
SPONTANEOUS pH SWING CRYSTALLIZATION OF NICOTINIC
ACID'

*Submitted to Crystal Growth & Design

Nicotinic acid (HNic) is crystallized by adding hydrochloric acid (HCl) to sodium
nicotinate (NaNic) aqueous solution in a semi-batch crystallization process. The
crystallization process is monitored in situ with an Attenuated Total Reﬂectance Fourier
Transform Infrared Spectrometer (ATR-FTIR). The in situ measurement of the
metastable limit and solubility enables us to deﬁne the metastable (operating) zone for
the crystallization of nicotinic acid. The inﬂuence of the acid addition rate, acid
concentration and agitation speed on the metastable zone width have been studied. The
metastable zone width’s dependence on the acid addition rate is used to obtain the
apparent order of nucleation of HNic. The effectiveness of micromixing is correlated
with and explained by the metastable zone width. The most favorable parameters that
will facilitate the formation of larger crystals without seeding are determined to be Slow
addition (1 g/hr) of diluted HCl (0.5 M) with agitation at just complete suspension speed,

NJS, which is equivalent to 1.1 m/S stirrer tip speed.

3.1 Introduction
pH swing crystallization is one of the most commonly used types of crystallization in the
chemical industry. In a pH swing crystallization, a solution of the deprotonated form is

mixed with an acid. The protonated species is formed by a neutralization reaction

35

resulting in the concentration of the acid exceeding its solubility. This requires that the
deprotonated form have a higher solubility than the protonated form has. pH swing
crystallization can also be carried out in the opposite direction if the protonated form has
a higher solubility than the deprotonated form. However, this is less commonly
encountered. pH swing crystallization exploits the generally large solubility difference
between the deprotonated and protonated forms to crystallize the less soluble species.
Crystallization through pH swing is more energy efﬁcient than cooling crystallization,
since heating and cooling are not necessary. It is also advantageous over antisolvent
crystallization because there is no need to perform separation of the solvents in the post

crystallization processes.

AS in most industrial crystallization processes, the ﬁnal crystal size distribution (CSD) is
usually one of the most important properties of the product because the ﬁnal CSD
directly affects the cost of the post crystallization processes such as ﬁltration, drying, and
packaging. In order to improve the CSD, it is necessary to measure the level of
supersaturation at which spontaneous(primary) crystallization takes place, which is also
called the metastable limit. Mullin et al.1 determined the metastable zone for cooling
crystallization of an ammonium sulphate aqueous solution using maximum allowable
undercooling temperature to express the maximum allowable supersaturation. The use of
temperature to represent the metastable limit avoids measurement of the real
concentration. However, it is not possible to use any other parameters to replace the real
concentration of the crystallizing species in pH swing crystallization because pH change

is not sensitive enough for a buffered solution, which is true in pH swing crystallization.

36

Measurement of the concentration of the crystallizing species is more challenging in pH
swing crystallization than in cooling and anti-solvent crystallization. Due to the nature of
supersaturation generation in pH swing crystallization, the presence of reactants during
the crystallization of the product makes it more difﬁcult to extract accurate information
about the crystallizing species. This problem does not exist in cooling crystallization,
and is less serious in anti-solvent crystallization. This challenge has led to far less work
directed toward the in situ monitoring of pH swing crystallization than that on the more
traditional ways of crystallization. There is only some theoretical modeling work done
by Lindberg and Rasmuson2 that approximates absolute concentration as supersaturation
neglecting the solubility change as reaction continues. Aslund and Rasmuson3 presented
some experimental results on the inﬂuence of process variables on the crystal Size
distribution in a semibatch reaction crystallization process of benzoic acid. However,
they correlated the crystal size distribution with the processing parameters without
measuring the changing supersaturation versus the changing concentration of the

reactant, which is the real driving force of crystallization.

Based on the method we developed earlier,4 this chapter presents the in situ experimental
ATR-FTIR measurement of the supersaturation and solubility of HNic in aqueous
nicotinate solutions with the presence of sodium and chloride ions, under the conditions
of the actual crystallization process. With the knowledge of the metastable zone width,
analysis is provided for the inﬂuence of the process parameters, including the acid

concentration, acid addition rate and the agitation speed, on the ﬁnal crystal size

37

 

distribution. Products with signiﬁcantly improved crystal size distribution are obtained

without tedious trial and error.

3.2 Experimental Section

3.2.] Apparatus

A LabMax® system manufactured by Mettler-Toledo was used to carry out the
crystallization experiment. The LabMax’s physical system consists of a 1-liter glass
reactor, a propeller, an F TS thermostat and a computer. The LabMax has two electronic
units, the thermostating unit and the controller unit. The thermostating unit measures and
controls the stirrer and a temperature sensor inside the reactor. The controller unit
measures and controls external sensors and devices including a pH electrode, a
ProMinent® gamma G/4b metering pump by ProMinent Fluid Controls, Inc. and a
turbidity probe by Dow Chemical Company. The ProMinent® pump is used to control
the addition of HCl. The turbidity probe allows the determination of the onset of

spontaneous crystallization. The gain of the turbidity probe was pre-determined to be 5.

The glass reactor has a 100 mm internal diameter. Agitation is performed by an upﬂow
propeller which has a diameter of 50 mm and is positioned with a clearance of 25 mm
from the bottom of the crystallizer as recommended in the literatures’ 6’ 7 During the
experiment, an ATR-FTIR probe is immersed in the solution or slurry to in situ monitor
C(HNic) and C(NaNic). With the presence of an ATR-FTIR probe, a turbidity probe, a
pH electrode and a temperature sensor, it is not necessary to add any bafﬂes to the

crystallizer to enhance mixing.

38

 

3. 2.2 Procedures

First, NJS, the just complete suspension stirrer speed was determined experimentally.
When the slurry is completely suspended, all particles are in motion and no particles
remain on the reactor bottom for more than 1-2 seconds.7 Therefore, the entire surfaces
of the particles are exposed to the ﬂuid to ensure that maximum surface area is available
for further nucleation and growth. A mixture of 3.6 grams of 455 um mean sized crystals
and 600 mL saturated HNic aqueous solution was prepared, which has a Similar density
as the slurries in actual experiments. The reactor was well illuminated, and careful
observation was made of the particles around the bottom of the reactor. No particles
remained stationary on the base for more than 1-2 seconds when the agitation speed
reached 420 rpm, which corresponds to a tip speed of 1.1 m/S. It is easier to correlate the
tip speed with secondary nucleation than using the revolutions per minute, due to the

different dimensions of different types of stirrers.

With the agitation speed at 1.1 m/s, the experiments were carried out at 30 at 01°C by
feeding HCl with the ProMinent pump to about 520 mL of 1M NaNic aqueous solution.
One-half molar and 2 M HCl were made by dilution of the 32 wt% HCl, and were
standardized with NaOH. NaOH was titrated against the primary standard, potassium
hydrogen phthalate (KHP). NaNic was purchased from Aldrich Chemical Company and
used as received. The ReactIR 1000 system was used to monitor C(HNic) and C(NaNic)
in situ throughout the experiments. The ﬁrst metastable limit was determined when the

turbidity probe’s signal reached 0.03 and then increased rapidly. The addition of HCl

39

was stopped once the relative intensity of the Signal ﬁom the turbidity probe increased to
0.03. Enough time (about 2 hrs) was allowed for the spontaneous crystallization to take
place so that C(HNic) could attain a constant value, which is the solubility of HNic at a
speciﬁc conentration of NaNic. The Slurry was ﬁltered with the ﬁltrate being kept about
30°C. The ﬁltrate was put back to the LabMax® reactor, and the addition of HCl was
restarted until the turbidity probe showed an increase in the signal again. The procedure
was repeated to obtain multiple metastable limits and solubility data so that an apparent
metastable zone can be deﬁned. The apparent metastable zone is not a fundamental
property of the chemical system, but depends on the process parameters employed.8’ 9
The experiment was repeated three times. The average values for metastable limits and
solubility are reported, respectively, because three values of each would replicate. The
metastable limit was also measured with another agitation speed of 750 rpm (a tip speed
of 1.9 m/s). Different acid addition rates, 1 g/hr, 2 g/hr and 60 g/hr, were examined for

their effect on the metastable limit.

The product from each experiment was ﬁltered, dried and observed under an optical
microscope at the magniﬁcation of 10. The product from the experiment with the
addition rate of 1 g/hr of 0.5 M HCl and with agitation speed of 1.1 m/s was sieved for
size distribution determination, but all the other combinations of the process parameters
resulted in bulky solids that are aggregates of very small crystals that could not be sieved
effectively. The crystal size distribution of the larger crystals was measured through

sieving with a set of Scienceware Mini-Sieves from Fisher using a Tyler RX-86 sieve

40

shaker. Sieving was completed after about 1 hr or until the change of the weight above

each sieve was less than 5%.

3.3 Results
Figure 3.1 shows the result of metastable zone measurement at 30°C with the addition of

2M HCl at the rate of 60 g/hr and with the agitation speed at 1.1 m/s.

The measurement starts with pure 1M NaNic solution. As 2M HCl is added, NaNic is
converted to HNic. After about 80 grams of 2M HCl is added, the ﬁrst spontaneous
crystallization takes place. Averaging the three measurements, the best ﬁt lines, which
deﬁne the metastable zone, are plotted in Figure 3.2.

The solubility, C ‘(HNic), can be ﬁtted into a linear equation as follows,

C(HNic) = 0.18C(NaNic) .+ 2.3 (3.1)
The metastable limit, M(HNic), can be expressed with another linear equation below,

M(HNic) = 0.21 C(NaNic) + 2.9 (3.2)

Similar measurements were performed varying the process parameters including the acid
addition rate, acid concentration and agitation speed. Firstly, the acid addition rate is
changed to 1 g/hr while the acid concentration is kept at 2M and agitation speed at 1.1
m/s.

Figure 3.3 shows the concentration proﬁle of the spontaneous crystallization at different

acid addition rates.

41

C(HNic)><103, (mole fraction)
N w A u.

 

(,3 ~ 221,113.11;
0 5 10 15 20

C(NaNic)><103, (mole fraction)

Figure 3.1. The concentration of nicotinic acid vs. the concentration of NaNic during the
metastable zone measurement of the spontaneous crystallization of nicotinic acid at 30 i
01°C, with agitation speed of 1.1 m/s and acid addition rate of 60 g/hr of 2 M HCl.

42

'
E ° Metastable limit |
l - Solubility i i

C(HNic)><103, (mole fraction)

4 qu”
l P”?
l |
3 '~ 1
l
' l
2 1 _k, 7,, . .W- . . a- 3‘, _ _, i .. -_f, ,i, _.-._l . .-__ __ ____. . ml
2 4 6 8 10

C(NaNic)><103, (mole fraction)

Figure 3.2. The metastable zone of the crystallization of nicotinic acid measured under
the conditions that the solution is agitated at 1.1 m/s and 2 M HCl is added at the rate of

60 g/hr at 30°C.

43

 

 

 

 

6yw_———e-+eww~v—w*ep+-r~w
.7: 51 ‘
s l
d: l i
.2 4 l“ l
O l
g l
an 31 l
2 l 1* l
x l l 1
3.? 21' l l
g l L
e: 11
l . ’
0 1-”, ~ - ‘ tel
8 10 12 14 16

C(NaNic)>< 103, (mole fraction)

Figure 3.3. Comparison of the ﬁrst metastable limit measurement with different acid
addition rates, 1 gm and 60 g/hr. The acid concentration is 2 M HCl and the agitation
speed is 1.1 m/s. The temperature is 30 i 01°C.

44

If the width between the metastable limit and the solubility measured with the acid
addition rate at 60 g/hr is taken as unity, the width between the metastable limit and

solubility with the acid addition rate at 1 g/hr is equivalent to 0.51.

The result is summarized in Table 3.1 as follows.

Table 3.1. Metastable limit comparison with different addition rates of 2 M HCl.

 

 

Acid addition C(HCl) Agitation tip speed, Metastable zone
rate, (glhr) (mole/l) (m/s) width
1 2 1.1 0.51
60 2 1.1 1

 

The products from both of the two measurements are powder-like particles. There is no
signiﬁcant difference in size as observed under the optical microscope. The particles

have similar size to those shown in Figure 3.4a.

In the second set of measurements, the acid concentration was varied with the addition
rate at 1 g/hr and agitation speed at 1.1 m/s. Similar proﬁles are obtained to Figure 3.3.
The results are summarized in Table 3.2, again with the metastable zone width
normalized against that measured with the addition of 2 M HCl at the rate of 60 g/hr and
agitation speed at 1.1 m/s.

Table 3.2. Metastable limit comparison with different acid concentration.

 

C(HCl) Acid addition Agitation tip speed, Metastable zone

 

(mole/l) rate, (g/hr) (m/s) width
0.5 1 1.1 0.50
2 l 1 .1 0.51

 

45

The products from the two experiments are signiﬁcantly different. Figure 3.4 presents

the photos of the two products taken with the optical microscope.

The experiment with the addition of 0.5 M HCl at the rate of 1 g/hr and agitation speed of
1.1 m/s produced crystalline materials of larger size as compared to the one with the
addition of 2 M HCl at the same addition rate and agitation speed. The crystal size

distribution for the case with addition of 0.5 M HCl is shown in Figure 3.5.

In the third set of measurements, the agitation speed was varied while the acid
concentration was held constant at 0.5 M and the acid addition rate was kept at l g/hr.

The metastable zone widths are compared in Table 3.3.

Table 3.3. Metastable limit comparison with different agitation speed.

 

Agitation tip speed, C(HCl) Acid addition Metastable zone

 

(nr/s) (mole/l) rate, (g/hr) width
1.1 0.5 1 0.50
1.9 0.5 1 0.29

 

The products from both of the two measurements are powder-like particles. There is no
signiﬁcant difference in size between them as observed under the optical microscope.

The particles have similar sizes to those shown in Figure 3.43.

The results ﬁ‘om the three sets of measurement suggest that the product with the

following process parameters, 0.5 M HCl added at 1 g/hr and agitation speed of 1.1 m/s,

results in the largest mean size crystal size distribution. However, the slow acid addition

46

 

Figure 3.4. Inﬂuence of the acid concentration on crystal size: (a): 2 M HCl, addition
rate of 1 gm and 1.1 m/s agitation speed, The average size of the product is about 25 pm.
(b) 0.5 M HCl, addition rate of 1 g/hr and 1.1 m/s agitation speed, (b.1), (b.2) and (b3)
are the crystals after sieving. The average size of (b.1) is 152 pm, that of (b.2) is 214 um
and that of (b.3) is 305 um.

47

 

 

 

 

;
o l
E 90 I
"“3 75 l
5 1
a 60 l
l
a 45 |
E 30 ‘
‘3 i
:5 ,
g 15 l
u 0 -———~—~—.—————~- . — . ».—l
0 100 200 300 400

Crystal size, um

Figure 3.5. The crystal size distribution of the product from the experiment with the
addition of 0.5 M HCl at the rate of 1 g/hr and agitation speed of 1.1 m/s.

48

 

rate made it very time consuming to conduct the crystallization. The next set of
measurements was designed to study the possibility of increasing the acid addition rate
without sacriﬁcing the quality of the product. The other acid addition rates examined
were 2 g/hr and 60 g/hr while C(HCl) and agitation speed were kept at 0.5 M and 1.1 m/s,

respectively. The metastable zone width results are summarized in Table 3.4.

Table 3.4. Metastable limit comparison with different addition rates of 0.5 M HCl.

 

Acid addition rate, Agitation speed, C(HCl) Metastable

 

(g/hr) (m/s) (mole/l) zone width
1 1.1 0.5 0.50
2 1 .1 0.5 0.66
60 1 .1 0.5 0.90

 

Both of the acid addition rates at 2 g/hr and 60 g/hr resulted in powder-like particles.
There was no signiﬁcant difference in size between them under the optical microscope.
The particles had similar sizes as those shown in Figure 3.4a. This result indicates that
the acid addition rate of 1 g/hr is the only rate, with combination of 0.5 M C(HCl) and

agitation speed of 1.1 m/s, with which the resultant crystals can grow into larger sizes.

3.4 Discussion

As compared to other systems such as benzoic acid,3 this particular crystallizing system,
HNic, has a ﬁnal crystal size distribution that cannot be directly correlated to the process
parameters. Massive nucleation takes place once some nuclei are formed when
micromixing is not effective or the intensity of agitation is too high. The metastable zone
width helps us understand the interaction of the nucleation kinetics, micromixing and

agitation intensity on the ﬁnal CSD of the product.

49

3.4.1 Eﬂect of Acid Addition Rate —- the Apparent Order of Nucleation
Utilizing the result shown in Table 3.1, we can estimate the nucleation kinetics with a
similar method of those given in the literatureg’9 The method developed by Nyvlt9 is

applied for cooling crystallization utilizing the approximation as follows,

811.31%

 

t

 

where B is the nucleation rate, is the solubility’s dependence on temperature, r,is

dT . . . .

7, the rate of coollng. Thrs method 1S based on the assumption that, at least at the
t

initial stage of nucleation (on attaining the metastable limit), the mass nucleation rate or

the consumption rate of supersaturation equals the generation rate of supersaturation. We

can derive a similar equation to express the nucleation rate for pH swing crystallization of

HNic as follows,

dCMAX _ dC(NaNic) _
dt dt

B:

 

rm, (3.4)

where C MM is the concentration of HNic at the metastable limit. rpH is the generation rate
of supersaturation in the pH swing crystallization. The nucleation rate can also be
expressed in the form of a power law when heterogeneous nucleation is considered to be

the predominant form of primary nucleation}IO

B=k~ rag“, (3.5)

50

where k N is the nucleation rate constant and AC M M is the metastable zone width. The
exponent n is the apparent order of nucleation, which is not a fundamental property of a
system.

Combining equations (3.4) and (3.5), we can calculate the apparent order of nucleation
using the metastable zone width measured at different rates at which supersaturation is

generated by the acid-base reaction.

1
10g10( /sz]

 

(3.6)

From the data listed in Table 3.5, n is determined to be 6.3.

Table 3.5. The rate of decrease of concentration of sodium nicotinate at different acid
addition rates and their corresponding metastable zone width.

 

 

Addition rate of 2 M HCl dC(NaNic) Metastable zone width
(g/hr) d:
(mole fraction/hr)
1 0.07 x 10'3 0.51
60 4.82 x 10‘3 1

 

According to the apparent order of nucleation, 6.3, we can estimate that the nucleation
rate for the case with the acid addition rate of 60 g/hr is 70 times as high as that for the
case with the acid addition rate of 1 g/hr. A large nucleation rate results in very small
particle Size, which is observed experimentally. However, for the acid addition rate of 1
g/hr, the ﬁnal crystal size should have been larger, which is contradictory to the
experimental result. This can be due to the inadequate micromixing, which may cause

high local supersaturation and consequently high nucleation rate, while the measured

51

bulk supersaturation is not a good indicator for the nucleation rate. This is further
conﬁrmed by the acid concentration’s effect on the metastable limit and ﬁnal product, as

shown in Table 3.2 and Figure 3.4.

3.4.2 Effect of Acid Concentration

The results in Table 3.2 showed almost no difference in the metastable limit or
metastable zone width between the additions of 0.5 M and 2 M HCl. However, their
resultant products have very signiﬁcant difference in terms of size. This further conﬁrms
that the addition of 2 M HCl generates highly supersaturated solutions in the reaction
zone and micromixing is not effective in dispersing the high supersaturation before
nucleation starts.” 1’12 Micromixing is the mixing process by which two different liquids
contact and mix at the molecular level so that the segregated areas of uniform, but
mutually different composition disappear; in other words, the mixture attains complete

homogeneity on the molecular level.13

The time it takes for acid-base reactions to go to
completion is generally considered to be shorter than the time of micromixing. The
effectiveness of micromixing is determined by the micromixing time scale and the

induction time scale.3’ ‘2

If the induction time is shorter than micromixing time,
nucleation will take place in the reaction zone before the molecular diffusion facilitated
micromixing dilutes the local concentrated solution. This is likely what happened in the
experiment with the addition of 2 M HCl, which is so concentrated that it generates
relatively high supersaturation in the reaction zone and results in a very short induction

time. Conversely, the addition of 0.5 M HCl does not cause the generation of high local

supersaturation and the induction time is longer than the micromixing time. Therefore,

52

nucleation happens in the bulk solution after the local supersaturation is dispersed into or

diluted by the bulk environment.

3. 4.3 Effect of Agitation Speed

Contradictions exist in the literature as to the effect of agitation speed on the mean size,
as summarized by Rasmuson et a1.3 Some of the studies did not consider the necessity of
employing agitation speed at least at the just complete suspension speed, N13, which
ensures the homogeneity of mixtures macroscopically. With agitation speed less than
N35, it is impossible to predict the mass transfer process for the unsuspended particles,
which may cause some of the contradictions in the effect of agitation speed on the ﬁnal
particle size. In this study, the lowest agitation speed applied is NS, 1.1 m/s. The higher
speed of 1.9 m/s initiates crystallization when the relative metastable limit is 0.29, much
lower than that at 1.1 m/s. This result indicates that the micromixing at 1.9 m/s results in
faster redistribution of supersaturation. However, the ﬁnal products are powder-like
particles similar to that in Figure 3.4a. This appearance can be explained by attrition,
breakage or fragmentation, which becomes signiﬁcant for the needle-shaped crystals of

HNic upon excessive intensive agitation.

The most favorable condition which yields the largest crystals is the addition of 0.5 M
HCl at the rate of 1 g/hr with agitation speed at 1.1 m/s . Attempts were made to increase
the acid addition rate to Zng and 60g/hr to shorten the length of time needed with lg/hr
addition rate. The results show that neither of the two rates can compete with 1g/hr when

the ﬁnal products are compared. This is again due to the effect of micromixing on

53

nucleation as can be explained by the metastable zone width. Therefore, it can be
concluded that 1 g/hr addition of 0.5 M HCl to l M NaNic aqueous solution stirred at 1.1
m/s is the best combination of the process parameters examined in current study to yield
crystals through spontaneous crystallization with favorable CSD and ﬁltration and drying
properties. Further improvement can be made by seeding crystallization, which is the

focus of the next chapter. 14

3.5 Conclusions

The metastable zone for the semi-batch crystallization of HNic was measured using in
situ ATR-FTIR spectroscopy when 2 M HCl was added to 1 M NaNic aqueous solution
at the rate of 60 g/hr with agitation speed at 1.1 m/s. Process variables inﬂuence the
metastable zone width and the product crystal size distribution signiﬁcantly. With the
knowledge of the metastable zone width from in situ measurement, the order of
nucleation and the relative nucleation rate can be estimated. In situ results of metastable
zone width also clarify our understanding of the effect of micromixing on the crystal size
distribution. It is found that 2 M HCl can not be effectively mixed with NaNic at 1.1 m/s
no matter whether the addition rate is l g/hr or 60 g/hr. When the acid concentration
decreases to 0.5 M, the addition rate of 1 g/hr and agitation speed at 1.1 m/s will yield
larger crystals with reduced times for ﬁltration and drying. Excessive intensive agitation
at 1.9 m/s causes fragmentation of particles and in turn unfavorable crystal size

distribution.

54

3.6 Acknowledgment
The ﬁnancial support from the Dow Chemical Company for this study is gratefully
acknowledged. The authors would also like to express appreciation to ASI Applied

Systems and Mettler-Toledo for the contribution of the LabMax® instrument.

55

References

l.

2.

9.

Mullin, J. W.; Charkraborty, M.; Mehta, K. J. of Appl. Chem. 1970, 20, 367-371.

Lindberg, M.; Rasmuson, A. 0, Chemical Engineering Science 2000, 55, 1735-1746

. Aslund, A. L.; Rasmuson, A. C. AIChE J. 1992, 38(3), 328-342.
. Wang, F.; Berglund, K. A. Ind. Eng. Chem. Res. 2000, 39, 2101-2104,
. He. Y.; Takahashi. K. J. Chem. Eng. Japan 1995, 28(6), 786-789.

. Shimizu, K.; Nomura, T.; Takahashi, K. J. Cryst. Growth 1998, 191, 178-184.

Hamby, N.; Edwards, M. F .; Nienow, A. W. Mixing in the Process Industries, 2nd
Edition, Butterworth-Heinemann Ltd: Oxford, 1992.

Myerson, A. S. In Handbook of Industrial Crystallization; Myerson, A. S. Eds.;
Butterworth-Heinemann Ltd: Oxford, 1992; Chapter 2, pp 43-52.

Nyvlt, J. J. Cryst. Growth 1968, 3/4, 377-383.

10. Mullin, J. W. Crystallization, 3rd Edition, Butterworth-Heinemann Ltd: Oxford,

1993.

11. Barata, P. A.; Serrano, M. L. J. Cryst. Growth 1996, 163, 426-43.

12. Phillips, R.; Rohani, S.; Baldyga, J. AIChE J. 1999, 45(1), 82-92.

13. Sohnel, 0.; Garside, J. Precipitation: Basic Principles and Industrial Applications,

Butterworth-Heinemann Ltd: Oxford, 1992.

14. Wang, F.; Berglund, K. A., submitted to Crystal Growth & Design

56

Chapter 4

FEEDBACK CONTROL OF SEEDED SEMI-BATCH
pH SWING CRYSTALLIZAION 0F NICOTINIC ACID*

*Submitted to Crystal Growth & Design.

Optimization of the process parameters for semi-batch pH swing crystallization was
demonstrated in the preceding chapter. Due to the slow addition rate of diluted
hydrochloric acid, it is extremely time-consuming to produce the large mean size crystals
by spontaneous nucleation. The length of time needed to induce crystallization can be
signiﬁcantly reduced by external seeding. Seeding at different levels of supersaturation
demonstrates increased supersaturation increases secondary nucleation. The level of
supersaturation was in situ controlled at constant level through the automatic adjustment
of post addition of HCl. It is found that keeping the solution at low supersaturation, or
close to the equilibrium solubility promotes crystal growth and suppresses secondary

nucleation to the largest degree.

4.1 Introduction

As in most industrial crystallization processes, production of a speciﬁed crystal size
distribution (CSD) is one of the primary goals in semi-batch pH swing crystallization.
Signiﬁcant cost reductions can be achieved by creating CSDs that have favorable
ﬁltration and drying properties. Supersaturation, as the driving force for crystallization,
determines the relative rate of nucleation and crystal growth and, as such, controls the
resulting CSD."2 Therefore, control of crystallization processes requires in situ

measurement and subsequent control of supersaturation. Compared to the numerous off-

57

line analytical techniques for the measurement of solubility and supersaturation,3 ATR-
FTIR provides a unique sampling conﬁguration allowing in situ measurement of solution

phase concentration in slurry without phase separationfﬁ

The ability of in situ
measurement of supersaturation is critical for the in situ control of semi-batch pH swing
crystallization. It has been demonstrated that maintenance of constant supersaturation
during a crystallization process leads to improved crystal size distribution for batch
cooling crystallization and semi-batch anti-solvent crystallizationé’7 However, there
have been few studies focused on the control of semi-batch pH swing crystallization.
The primary reason may be due to the presence of multi-species in this type of
crystallization. Based on the technique we developed earlier,5’8 the study of in situ

control of seeded pH swing crystallization of nicotinic acid (HNic) is presented in this

chapter.

In the preceding chapter, we have demonstrated that production of crystals with favorable
CSD through spontaneous crystallization is extremely difﬁcult because of the high
supersaturation and excessive nucleation that often occurs. Even though we have
identiﬁed the process parameters which facilitate the production of favorable CSD, the
productivity is extremely limited by the slow nucleation and crystal grth kinetics. One
solution to overcome this difﬁculty is to use external seeding. Seeding the supersaturated
solution not only avoids the time-consuming primary nucleation process, it also provides
control over the product size and size distribution, and consequently consistency from
batch to batch. In this chapter, we demonstrate the importance of the level of

supersaturation’s effect on secondary nucleation. The in situ supersaturation value is

58

used to control the rate of post-seeding acid addition so that secondary nucleation is
continuously minimized, and thereafter the ﬁnal crystal size distribution can be

optimized.

4.2 Experimental Section

4.2.1 Apparatus

A LabMax® system manufactured by Mettler-Toledo was used to carry out the
crystallization experiments. The LabMax®’s physical system consists of a l-liter glass
reactor, a propeller, an FTS thermostat and a computer. The LabMax® has two electronic
control units, the thermostat unit and the controller unit. The thermostat unit measures
and controls the stirrer and the temperature sensor inside the reactor. The controller unit
measures and controls external sensors and devices including a pH electrode, a
ProMinent® gamma G/4b metering pump by ProMinent Fluid Controls, Inc. and a
turbidity probe by Dow Chemical Company. The ProMinent® pump is used to control

the addition of HCl. The gain of the turbidity was pre-deterrnined to be 5.

The glass reactor has a 100 mm internal diameter. Agitation is performed by an upﬂow
propeller which has a diameter of 50 mm and is positioned with a clearance of 25 mm
from the bottom of the crystallizer. The ratio of the propeller diameter to the reactor
diameter and the ratio of the clearance from the bottom to the reactor diameter are chosen

as recommended in the literature to ensure the best macromixing.5’ 6’ 7

During the
experiments, an ATR-FTIR probe was immersed in the solution or slurry to monitor the

concentrations of nicotinic acid (HNic) and sodium nicotinate (NaNic) in situ. With the

59

presence of an ATR-FTIR probe, 3 turbidity probe , a pH electrode and a temperature

sensor, it is not necessary to add any bafﬂes to enhance mixing.

4. 2.2 Interface between the LabMax® and the ReactIR 1000

The LabMax®-ReactI_R interface is a collection of programs and a serial cable which
provides communications capabilities between the ASI Applied Systems ReactIR
spectrometer and the Mettler LabMax® reactor. The interface transmits LabMax® status
and values from the LabMax® to the ReactIR, and the ReactIR values from the ReactIR
to the LabMax®. The interface also provides synchronization to allow a ReactIR
experiment to start automatically when the LabMax® starts executing a recipe. The
hardware requirement for the interface is a null modem serial cable that connects the
LabMax® and the ReactIR computers through their RS-232-ports. The software required
for the LabMax® is MaxSeril.exe written by Mettler Toledo. This program reads values
from the LabMax® and sends them to the ReactIR system. It also accepts values from the
serial port and uses DDE (Dynamic Data Exchange) to transmit them to the LabMax®
program. The software required for the ReactIR is CommRW.exe written by ASI
Applied Systems. The program reads data from the LabMax®, and also sends the
ReactIR data to the LabMax® via the serial port. The interface automatically sends
values for a speciﬁc set of variables to the LabMax®. These variables include:
React[R_Ready, Conc_A, Conc_B. Any ReactIR variables which are deﬁned with the
above names and proﬁled within the ReactIR software will be sent to the LabMax®.
These variables may also be used to terminate a stage within the LabMax® recipe. We

deﬁned the relative supersaturation as ReactIR_Ready, and employed this variable to

60

terminate or start the addition of HCl when the supersaturation deviates from the set
points. In doing so, supersaturation is controlled within the desired range to realize the

control of ﬁnal crystal size distribution.

4. 2.3 Procedures

With the agitation tip speed set to 1.1 m/s, as determined in the preceding chapter,8 the
experiments were carried out at 30 i 01°C by feeding HCl with the ProMinent pump to
about 520 mL of 1M NaNic aqueous solution. One-half molar and 2M HCl were made
by dilution of the 32wt% HCl, and were standardized with NaOH. NaOH was titrated
against the primary standard potassium hydrogen phthalate (KHP). NaNic was purchased
from Aldrich Chemical Company and used as received. The ReactIR 1000 system was

used to monitor concentrations of HNic and NaNic in situ throughout the experiments.

4. 2.3 . 1 Seeding at Diﬂerent Levels of Supersaturation

In this set of measurements, we adopt the same expression for supersaturation as in
Chapter 3. The relative supersaturation is deﬁned as a fraction of the unit metastable
zone width, which is measured with addition of 2M HCl at the rate of lg/min and
agitation tip speed of 1.1 m/S. We have shown in the previous chapter that the equation
for the solubility of HNic, C(HNic), as a function of C(NaNic) is,

C(HNic) = 0.18C(NaNic) + 2.3 (4.1)
And the metastable limit of HNic, M(HNic), as a function of C (NaNic) is

M(HNic) = 0.21C(NaNic) + 2.9 (4.2)
The relative supersaturation compared to the metastable limit at any point of the reaction

or crystallization is deﬁned as,

61

S = C (HNic)-C ‘(HNic)
R M(HNic)—C(HNic)

 

(4.3)

For the seeding experiments, the seeds were prepared by dissolving HNic in ethanol at an
elevated temperature. The solution is allowed to cool down at room temperature without
being agitated, and the solvent was allowed to evaporate at room temperature and
pressure. The recrystallized HNic was sieved with a set of Scienceware Mini-Sieves
from Fisher using a Tyler RX-86 sieve shaker. Sieving was completed until the change
of the weight above each sieve was less than 5%. The crystals with an average size of
152 um, sieve cut between 127 um and 177 um, were used for seeding experiments.
Five hundred and twenty milliliters of 1 M aqueous NaNic was prepared as the starting
solution. Sixty grams of 2 M HCl were added to the NaNic solution at the rate of 6
g/min. Then the addition of 2 M HCl was slowed to 1 g/min until relative supersaturation
reached the speciﬁed value, which is 30% or 45% or 60%. It is worth mentioning that
the termination of acid addition is controlled by the value of the relative supersaturation
transmitted from the ReactIR to the LabMax®. At least 30 minutes were allowed to
ensure that no spontaneous crystallization occurred at the speciﬁed level of relative
supersaturation before seeds were added to the supersaturated solution. C(HNic)
measured by the ReactIR started to decrease until SR approached 0%, which indicated
that crystal growth of the added seeds and secondary nucleation had consumed all the
supersaturation and crystallization was completed. At this point, the slurry was ﬁltered
and dried overnight. Sieve analysis was performed with a set of Scienceware Mini-
Sieves from Fisher using a Tyler RX-86 sieve shaker. Sieving was completed until the

change of the weight above each sieve was less than 5%.

62

4. 2.3.2 Seeding with Post Addition of HCl

This set of experiments started with 520 mL of 1 M aqueous NaNic. The seed
concentration was 1.5 g seeds/kg solution. The procedures up to the addition of seeds
are the same as those mentioned in section 4.2.3.1. Once seeds were added into the
crystallizer, HCl was changed from 2 M to 0.5 M. In the ﬁrst experiment, relative
supersaturation was controlled at about 30% i 10%. In other words, the addition of 0.5
M was prompted automatically once SR went below 20%, and the addition of the acid was
terminated once SR surpassed 40%. During the stages when 0.5 M HCl was added, the
addition rate was at 6 g/hr. This rate was estimated from the previous data of seeding
without post-seeding acid addition. The second experiment was to add the seeds at just
above saturation or 0% relative supersaturation and then control the solution at 0%
relative supersaturation during the post addition of 0.5 M HCl. The last experiment
carried out was to introduce the seeds at 30% relative supersaturation and then control

relative supersaturation at about 0%.

The products from each experiment were ﬁltered and dried overnight. Then they were
observed under an optical microscope at the magniﬁcation of 10. The crystal size
distribution was measured by Sieve analysis with the Scienceware Mini-Sieves from
Fisher using a Tyler RX-86 sieve shaker. Sieving was completed until the change of the

weight above each sieve was less than 5%.

63

4.3 Results and Discussion

4. 3 . 1 Seeding at Diﬂerent Levels of Supersaturation

The experiment started with pure 1 M NaNic aqueous solution. As 2 M HCl was added,
NaNic was converted to HNic. The relative supersaturation reached 60% after about 73 g
2 M HCl was added. Upon addition of 1 g of the 152 um HNic seed, C(HNic) started to
decrease as crystal grth on the seeds took place. At the same time, secondary
nucleation also consruned some of the supersaturation. Secondary nucleation and crystal
growth stopped upon depletion of the supersaturation. Figure 4.1 presents the proﬁles of
C(NaNic), C(HNic) and its relative position with respect to its solubility and metastable
limit. The acid addition proﬁle and the supersaturation proﬁle for 60%-SR seeding are
shown in Figure 4.2. The acid addition was set at a faster rate, 6 g/min, at the beginning
of the experiment. After 60 g of 2 M HCl were added, the relative supersaturation value
from the ReactIR indicated that the solution is close to saturation with respect to C(HNic)
although it was still undersaturated. At this point, the acid addition rate was slowed to l
g/min to ensure that no local supersaturated region was generated in the solution so that
early nucleation could be avoided. As the acid was continuously added to convert NaNic
to HNic, the relative supersaturation of HNic increased. The acid addition was
terminated by the LabMax® once the value of the relative supersaturation transmitted
from the ReactIR reached 60%. After this point, seeds were introduced to the solution
and the relative supersaturation started to decrease. Finally the relative supersaturation
was reduced to the value of 0%. Throughout the whole crystallization process, the solid

phase was

 

 

 

 

 

 

 

 

 

 

 

 

 

 

IV
5
uoaaaonna‘ﬁonnoaﬂﬂuouoou -14 '8
t: .. E
g .ftfttrxxxxxxxxx x xxrxxx %
g i - 12 a.
O M
-—- o
C v—
E 3 .x o o o x
g : .0.... 0.00.0000 0.0. O. .I 10 g?
x 2 'x lASolnbilityofHNic (Zn
3 x a Metastabb limit ofHNic g
g 1 i C _ “ 8
o t‘I H IH " (“N“)
E 0C(NaNi:)
O I l l 1 F r 6
0 0.5 1 1.5 2 2.5 3 3.5

Time since the start of acid addition, hrs

Figure 4.1. Concentration proﬁle of seeded crystallization of nicotinic acid at 60%
relative supersaturation seeding. The seeds were added at 30 mins after the relative
supersaturation reached 60% with seed concentration of 1.5g seeds/kg of solution. T =
30 :t 0.1°C. Region 1: Starting with sodium nicotinate aqueous solution of 1 M, which is
equivalent to 16.5 X 10'3 mole fraction, the concentration of nicotinic acid approaches its
solubility as sodium nicotinate is converted to nicotinic acid by HCl. The solution is in
the undersaturated state. Region II: Supersaturated solution with the degree of the
relative supersaturation increasing until it reaches 60%, no nucleation or crystal growth
takes place. Region III: Addition of 1 g of 152 um seeds of nicotinic acid.
Supersaturation decreases as it is being consumed by crystal growth and secondary
nucleation until it drops to saturation. Region IV: The solution equilibrates at saturation
and crystal grth and secondary nucleation stop.

65

100

 

 

 

 

 

1
O ' ; XXXXXXXXXXxxxxxxxxxxxx
at ' 80
a X 0000000000000000000000 QB
‘é -1 -. g:
5 x ‘ 60 -°e’
:13 0 '8
Q, _2 .x _
a <> 0
I
.“2’ :3 - 40
*5 _ .
£2 3 *9 i . .
<> ' X Relative supersaturation
ff, 1 - 20
'4 f 0 HCl added
-5E I l l l l l O
0 0.5 1 1.5 2 2.5 3 3.5

Time since the start of the acid addition, (hrs)

Figure 4.2. Relative supersaturation proﬁle and HCl addition proﬁle of seeded
crystallization of nicotinic acid at 60% relative supersaturation seeding and the seed
concentration of 1.5 g seeds/kg of solution added at 30 mins after the supersaturation

reached 60%. T = 30 i 01°C.

66

monitored by a turbidity probe, which also facilitated the determination of the endpoint of
the crystallization process. Figure 4.3 presents the turbidity proﬁle as the crystallization
was carried out. At the beginning of the experiment, there was no solid in the system so
the turbidity intensity stayed constant. Upon introduction of the seeds to the
supersaturated solution, the turbidity of the system increased rapidly at ﬁrst, then
continued to rise more slowly as the crystals grew. At the end of the crystallization, the
signal reached a plateau, which indicates the solid phase does not change any more or the
crystallization was complete. With the measurement of both the solution phase and solid

phase, we were able to ensure the completion of the crystallization process.

The crystal size distribution of the ﬁnal product from the 60%-SR seeding is shown in
Figure 4.4. The results show that about 20 wt% of the particles are smaller than the seeds
added. Fiﬁty-three weight percent of them had the same average size as the seeds. Only
27 wt% of the overall particles were larger than the seeds added to the supersaturated
solution. This result indicates that secondary nucleation is predominant over crystal
grth for seeding at 60%-SR. This can be explained more explicitly by

comparison of the number of particles of each size with the number of seeds added. We

can calculate the number of crystals of each size using the equation given by Mullin,l

 

M M
n=——-

m _ MM" (44)

where n is the number of crystals of certain size; M is the mass of the crystals, m is the
mass of each crystal; f, is the volume shape factor which is 10 for needle shaped crystals

as given in the literature;9 p is the density of HNic which has a

67

0.7

0.6
5:; 0.5
$3
a
.a
o 0.4
.2
E!
g 0.3
0.2
0.1

 

 

 

 

_ o. o ."o‘...
‘0
' O
U
_ O
I
O
O
"' 0
O
O
0— L r J I
O 0.5 l 1.5 2 2.5

Time since the start of the acid addition, hrs '

3.5

Figure 4.3. The turbidity proﬁle for 60% relative supersaturation seeding crystallization
of nicotinic acid. The seed concentration is 1.5 g seeds/kg solution. T = 30 :t 01°C.

68

 

 

 

 

 

 

100 a
(D
.S
8 80
'O
S:
:3
T3 60
a
On
2 40
a
E
U 20

0

0 100 200 300 400

Crystal size, urn

Figure 4.4. The crystal size distribution of the products from 60% relative
supersaturation seeding crystallization of nicotinic acid. The seed concentration is 1.5 g
seeds/kg solution. The average size of the seeds is 152 um.

69

value of 1.47 g/cm3, as given by J.T. Baker;10 d is the equivalent diameter that is
approximated as the average of the two sieves apertures between which the sieving has
been carried out. Accordingly, the number of seeds added is determined to be about
19,400. Table 4.1 shows the number of crystals of each size at the end of the 60%-SR

seeding crystallization.

Table 4.1. The number of crystals of each size and the mean crystal size of the product
from 60% relative supersaturation seeding of HNic, the seed concentration is 1.5 g
seeds/kg solution.

 

Crystal Size Number of wt% of Total Number of Mean Crystal

 

(pm Crystals Mass Seeds Added Size (pm)
64 90,300 20
152 18,000 53 19,400
214 2,600 22
305 200 5
156

 

Comparatively, there are 93,700 new crystals formed as a result of secondary nucleation.

96% of them are smaller than the added seeds, which made the mean crystal Size (156

pm) of the product only a little larger than the size of the seeds (152 pm).

The result of 60%-SR seeding indicates that 60% relative supersaturation is too high for
crystal grth to be predominant and consequently we are not able to suppress
nucleation. For this reason, seeding was performed at lower levels of relative
supersaturation, 45% and 30%. For illustration purposes, the supersaturation proﬁles are
compared for only 60%-SR and 30%-SR seedings, as shown in Figure 4.5. It can be
observed from Figure 4.5 that the desupersaturation rates for the two cases are

signiﬁcantly different. This is due to the

70

 

 

 

 

 

 

 

0.8
.2 0.6 - §
g A
«.5 0.4 ' g
I: O A &0 0
o 0.2 1' co (1)0990 °
.5 oM°oooo o cocooo o o
to _2 ° ‘4“? 0.3“ °<9 00 °°°°Q9°<90°°°°°°°59°°°>°°°
5 0 0
§ 0

-O.2 '5
3 A
5 o
‘3 -0.4 '0 -
.5 0 6 _ 1° Supersaturation of HNic for 30%-SR seeding
cc - .
T) “A Supersaturation of HNic for 60%-SR seeding
‘3‘ -O.8 4 ~

-1 l I I
0 2 4 6 8

Time since the start of the acid addition, (hrs)

Figure 4.5. The relative supersaturation proﬁle comparison for 30%-SR and 60%-SR
seeding crystallization experiments of HNic. The seed concentration is 1.5 g seeds/kg
solution.

71

kinetic difference in secondary nucleation and crystal grth caused by different level of
supersaturation. Comparison of the crystal size distribution for the three measurements is
shown in Figure 4.6. It is found that the weight percentage of the crystals larger than the
added seeds increases from 27 wt% to 66 wt% as the level of relative supersaturation
decreases from 60% to 30%. This can be more explicitly illustrated by Figure 4.7, which
demonstrates the dependence of the mean crystal size on the level of relative
supersaturation at which seeding is performed. As we decrease the level of relative
supersaturation at which seeding is performed, the mean crystal size increases, which
suggests that crystal growth has become more and more dominant over secondary
nucleation as the relative supersaturation for seeding is lowered. Table 4.2 presents the
number of crystals of each size for the 30%-SR seeding measurement. In comparison
with the result in Table 4.1, there are only 13,000 new crystals formed as the result of
secondary nucleation, seven times less than those formed in 60%-SR seeding experiment
(93,700). The number of crystals that are larger than the seeds (6300) more than doubles
those from 60%-SR seeding case (2800). Therefore we can conclude that seeding at 30%
relative supersaturation produces crystals with signiﬁcantly improved size distribution
and mean size as compared to 60% relative supersaturation seeding.

Table 4.2. The number of crystals of each size and the mean crystal size of the product

from 30% relative supersaturation seeding, the seed concentration is 1.5 g seeds/kg
solution.

 

 

Crystal Size Number of wt% of Total Number of Mean Crystal
(pug Crystals Crystal Mass seeds added Size (pm)
64 16,700 4
152 9,400 30 19,400
214 5,800 52
305 500 12
455 20 2
204

 

72

 

..----------o

100'

   
 

 

Q)

.E

a 80 -

Q)

'U

5

1% 6o-

8

Q)

0..

40- Tr“ . r
E I l —[3- seedrngat6O°/o-SR
:3 L

g 20 l -A—seedingat45%—SR
c.)

 

l - -<> - seedingat30%—SR

I l I J

 

 

0 100 200 300 400 500 600
Crystal size, (um)

Figure 4.6. The comparison of the crystal size distributions resulted from seedings at
three levels of relative supersaturation, 30%, 45% and 60%. The seed concentration is
1.5 g seeds/kg solution.

73

 

210
200 '
190 '
180 '
170 '
160 '
150 ‘
140 "
130 '
120 ' ' '

25% 35% 45% 55% 65%

Relative supersaturation at which seeding is performed

Mean crystal size, (pm)

 

 

 

Figure 4.7. The mean crystal size dependence on the level of relative supersaturation at
which seeding is performed. The seed concentration is 1.5 g seeds/kg of solution. The

average size of the seeds is 152 pm. The line between the datum points is meant only to
be a guide to the eye.

74

4. 3.2 Seeding with Post-seeding Acid Addition Controlled by Relative Supersaturation

It has been frequently suggested 7’1" 12

that constant supersaturation has to be maintained
in order to suppress nucleation and maximize the product mean size. In this section, we
demonstrate that relative supersaturation can be controlled at a constant level
dynamically with the adjustment of acid addition after seeds are introduced. As it is
found in the previous section, seeding at 30% relative supersaturation can produce
crystals of larger mean size. We carried out the ﬁrst experiment with seeding at 30%-SR
and continued to control the relative supersaturation at about 30% during the post acid
addition process. The concentration proﬁles are presented in Figure 4.8. The relative
supersaturation proﬁle and the acid addition proﬁle are shown as Figure 4.9. As shown
in the two ﬁgures, pure NaNic solution was used to start the experiment. During the ﬁrst
10 min, 2 M HCl was added rapidly to bring the solution to saturation, then the addition
rate was slowed until the speciﬁed relative supersaturation level, 30% was achieved.
Upon the addition of seed crystals to the solution, the relative supersaturation started to

decrease. One half molar HCl addition started when the level of relative supersaturation

dropped below the lower limit 20%.

The addition rate of 0.5 M HCl was set at 6 g/hr, which was estimated from the
desupersaturation proﬁle of HNic in Figure 4.5. It can be estimated from the proﬁle that
it takes about 3 hrs for the relative supersaturation to drop from 30% to 10%, which
corresponds to the supersaturation difference caused by about 4 g of 2 M HCl in the

similar solution. Therefore, if we add the acid at a similar rate as it is consumed, 4 g

75

 

  

 

 

 

 

 

 

 

t: I:
.2 .2
23 \___ 3
g 5 . ""~’ ------- Q.-. ----- .-. ------- .‘ 14 ‘g
2 t 2
E 4 OQDOOOOOOOOOOOOOGUO-POOQQO 0000000000000 12 E
2; at l I 2;
._. ”(xxx x ﬂ
’3 3 . O X XXXXXXXXXXXXXXXXXXXX XXXXXXX . 10 :8
g 2 _ IT 0 C(HNic) l I - s R:
U o ‘ Solubility l 5

1 . 1 - - - -Metastablelimitj ‘ 6

1 x C(NaNic)
0 L-e I ' ' ' I 4
-1 1 3 5 7 9 11 13

Time since the start of acid addition, hrs

Figure 4.8. Concentration proﬁles of sodium nicotinate and nicotinic acid and the
solubility and metastable limit of nicotinic acid for crystallization controlled at about
30% relative supersaturation with post seeding acid addition.

76

 

 

 

 

 

 

 

 

2 200

1 D
C: x x x X x XX xx . 160
.2 0 manual: xxx x xXxxxx X xXxxx X XXxxxxxxx A
1'6 20 39
r... . 1 ..
z -1 - a
"‘ 1:
“g. -2 - - 80 2
”3 U
E -3 l E
a ' 40
0 _4 . , -——-- .
a: l X Relative Supersaturation . .

-5 - I — HCl added J 0

—6 I I I I I l ‘40

-1 1 3 5 7 9 11 13

Time since the start of the acid addition, (hrs)

Figure 4.9. Relative supersaturation proﬁle and HCl addition proﬁle for crystallization
controlled at about 30% relative supersaturation with post seeding acid addition. The
seeds are added at 30% relative supersaturation. The seed concentration is 1.5 g seeds/kg
solution.

77

HCl/3 hrs, there should not be accumulation of supersaturation and no signiﬁcant
nucleation as a result. As it has been demonstrated in Chapter 3, even when adding 2 M
HCl at an extremely slow rate (1 g/hr), micromixing is still not effective in dispersing
local high supersaturation to the bulk solution before nucleation occurs. Due to this
concern, we chose to add 0.5 M HCl instead of 2 M HCl during the post seeding
crystallization process. When we convert the addition rate for 2 M HCl (4 g/3 hr) to the
rate for 0.5 M HCl, 6 g/hr is adopted considering the dilution effect of more solvent

introduced when 0.5 M HCl is added.

As 0.5 M HCl was added to the slurry mixture, it can be seen from Figure 4.8 that
C(NaNic) decreased; C(HNic) remained almost constant. The relative supersaturation in
Figure 4.9 follows the same trend as C(HNic). This result indicates that there is no
signiﬁcant accumulation of supersaturation for the ﬁrst 30 g HCl added. However, after
the continuous addition of 30 g HCl, the overall addition rate was slowed down due to the
unstable value of supersaturation. This suggests that there might be some accumulation
of supersaturation after a signiﬁcant amount of HCl was added. The ﬁnal crystal size
distribution in Figure 4.14 shows that the size of 85% of the particles are smaller or equal
to that of the seeds, which indicates that secondary nucleation is predominant during the
post addition of HCl. The mean crystal size (107 pm) is much smaller than that from
30%-SR seeding without post addition of acid. It is implied that the generation of 30%
relative supersaturation with the presence of crystals can no longer sustain crystal growth,

but only enhance secondary nucleation. This is consistent with the suggestion that the so-

78

called “metastable limit for secondary nucleation” is much lower than the metastable

limit for primary nucleation.l

In order to suppress secondary nucleation in the post acid addition process, an additional
experiment was performed with minimized relative supersaturation, 0%-SR, or the
solution kept about saturation throughout the crystallization process. The concentration
proﬁles for HNic and NaNic are shown in Figure 4.10. The relative supersaturation and
acid addition proﬁles are presented as Figure 4.11. Initially, 2 M HCl was added to 1 M
NaNic aqueous solution to bring the solution to just above saturation with respect to
HNic. One gram of 152 um seeds was added when the relative supersaturation exceeds 0
It can be seen from Figure 4.11 that the relative supersaturation decreased a little which
indicates that the added seeds consumed that little amount of supersaturation created
earlier. Then 0.5 M HCl was added to facilitate post seeding growth. Initially 30 g of 0.5
M HCl were added continuously at the rate of 6 g/hr, and there was no accumulation of
supersaturation. The addition of HCl was slowed due to the unstable value of relative
supersaturation, which indicates that there might be some accumulation of
supersaturation at this point. The ﬁnal crystal size distribution in Figure 4.14 suggests
that crystal growth is predominant for the 0% relative supersaturation controlled
crystallization even though a small percentage of nucleation (17%) still occurs, which is

almost impossible to overcome in a realistic seeding crystallization process.

79

 

 

 

 

 

 

 

 

7 18
E 6 - 1 16 E
a *5
3 5 - ~ 14 .1;
2 2
O O
E 4 - - 12 E
13?. a."
X 3 ' O [ 77‘ XX 1 10 :
’3 0 ‘ - - - - Metastable1 ’3
.... O i . .._.
E 2 ' o l o C(HNic) g ' 8 %
O 1 _ 3 ! Solubility o g

o [ x _ C(NaNic)
O a I I I I 4
-1 1 3 5 7 9

Time since the start of acid addition, (hrs)

Figure 4.10. Concentration proﬁles of sodium nicotinate and nicotinic acid and the
solubility and metastable limit of nicotinic acid for the crystallization controlled at about
0% relative supersaturation with post seeding acid addition. The initial seeding is
performed at just above saturation.

80

 

 

 

 

 

 

 

1 200
‘5 O ' WXWVXxxxﬂxmnnxxxni N - 160
E G?)
a -1 - X ‘1
g - 120 3
o.) x :8
2* -2 1 x 3
U

E -3 - "
g ”7’ _ . __-

4 3 )1: Relative supersaturation . 40

' L— HCl added

-5 ' J ‘ ' O

-1 1 3 5 7 9

Time since the start of the acid addition, (hrs)

Figure 4.11. Relative supersaturation proﬁle and HCl addition proﬁle for crystallization
controlled at about 0% relative supersaturation with post seeding acid addition. The
seeds are added at just above 0% relative supersaturation. The seed concentration is 1.5 g
seeds/kg solution.

81

The last controlled crystallization was conducted by seeding at 30% relative
supersaturation and followed by control of relative supersaturation around 0% during the
post addition of acid. The concentration proﬁles for HNic and NaNic are shown in
Figure 4.12. The relative supersaturation and acid addition proﬁles are presented as
Figure 4.13. Initially, 2 M HCl was added to 1 M NaNic aqueous solution to bring the
solution to 30% supersaturation with respect to HNic. One gram of 152 um seeds was
added once the relative supersaturation exceeded 30%. The seeds were allowed an hour
to grow until the supersaturation dropped to 0%. It can be seen from Figure 4.12 that the
relative supersaturation decreased after the addition of seeds, which indicates that the
seeds consumed the supersaturation created earlier. One half molar HCl was added to
facilitate post seeding grth once the relative supersaturation dropped to saturation (SR
= 0). The same acid addition rate, 6 g/hr of 0.5 M HCl was applied for the ﬁrst 10 g of
acid. There was no accumulation of supersaturation. The addition of HCl was increased
to 12 g/hr for the next 10 g of acid, and then further increased to 24 g/hr for the last 10 g
of acid added. There was no signiﬁcant accumulation of supersaturation in the post-
seeding acid addition process. The ﬁnal crystal size distribution shown in Figure 4.14
suggests that crystal growth is still predominant for the 30% relative supersaturation
seeding and 0% relative supersaturation controlled crystallization. However, secondary
nucleation seems more pronounced in this case than the 0% relative supersaturation
seeding and 0% relative supersaturation controlled crystallization. This can be explained
by two facts, the ﬁrst is that 30% relative supersaturation initial seeding results in higher
degree of secondary nucleation than the 0% relative supersaturation seeding. The second

reason is that faster addition of acid during the later stage of the crystallization may also

82

 

 
    

 

 

 

 

 

 

7 18
3"
g 6 .. ﬂ " 16 E
g E
d: 5 ' " 14 2
s a
a 4 - r i 12 m.
m“ ° C
o . .—
I 3 ~ 8' .,.... xxx x”(xx"XXXXXXXXXXxxxx - 10 ”XS
’8 ° _ ._
E o . Solubility %
2 b o ' . . ‘ 8 E,
{j o . - - - -Metastab1ellmlt U
1 _ o l o C(HNic) . . 6
0 Il x C(NaNb) l
O a I I I I I 4
-0.5 0.5 1.5 2.5 3.5 4.5 5.5

Time since the start of acid addition, hrs

Figure 4.12. Concentration proﬁles of sodium nicotinate and nicotinic acid and the
solubility and metastable limit of nicotinic acid for the crystallization controlled about
0% relative supersaturation with post seeding acid addition. The initial seeding was
performed at 30% relative supersaturation.

83

 

 

 

 

 

 

 

1 200
O ' xx x x Xx Xx
xxxagxxxxxXx x XXX”! :1: . 160

c:
.9 x
E '1 ' x 3‘9
2 -2 - * a
53 x '5
x ‘ 8° m
a '3 " x

4 * Relative supersaturation - 40

' lg -— HCl added

-5 I I I I I O

-1 O 1 2 3 4 5 6

Time since the start of the acid addition, (hrs)

Figure 4.13. Relative supersaturation proﬁle and HCl addition proﬁle for crystallization
controlled at about 0% relative supersaturation with post seeding acid addition. The
seeds are added at 30% relative supersaturation. The seed concentration is 1.5 g seeds/kg

solution.

84

 

     

 

 

 

 

 

 

 

105
3
g 90 -
r: I
E 75 - 4:
*a
d . I
g 60 I
Q
g 45 - /
1‘3 I _e_ OO/O-SR
"g 30 II - -A- - 30%-0%-SR
c3 15 III —<>— 30%-SR
O I I I
o 100 200 300 400 500

Crystal size, (um)

Figure 4.14. The crystal size distribution from three controlled supersaturation
experiments. The three cases are: I. Seeding at 0% relative supersaturation, and
controlled about 0% relative supersaturation during post acid addition; 11. Seeding at
30% relative supersaturation, and controlled about 0% relative supersaturation during
post seeding acid addition; 111. Seeding at 30% relative supersaturation, and controlled at
about 30% relative supersaturation during post acid addition.

85

contribute to more secondary nuclei in the 30%-0% relative supersaturation controlled
case. From the crystal size distributions of the three controlled crystallization processes,
we can conclude that 30% relative supersaturation control during the post-seeding acid
addition is not favorable for crystal grth to occur; instead it promotes secondary
nucleation and results in ﬁne particles with an average size smaller than the seeds added.
Zero percent relative supersaturation seeding and controlled crystallization results in
signiﬁcantly improved crystal size distribution. Thirty percent relative supersaturation
seeding and 0% relative supersaturation control produce crystals with an acceptable mean
crystal size, as shown in Table 4.3. However, it also causes a higher degree of secondary
nucleation than the 0% relative supersaturation seeding case. The presence of smaller
crystals as a result of secondary nucleation will further impede the rate of crystal grth

‘3 Therefore, it is recommended that 0% relative

due to the size-dependent growth rate.
supersaturation seeding and control is the most appropriate method to maximize the mean
crystal size.

Table 4.3. The mean crystal size from three experiments which are controlled by
different levels of supersaturation.

 

 

Level of relative supersaturation controlled Mean crystal size (pm)
30% seeding and then controlled at about 30% 107

30% seeding and then controlled at about 0% 212

0% seeding and then controlled at about 0% 223

 

86

4.4 Conclusions

The ability to monitor the relative supersaturation in situ with ReactIR and the interface
between the LabMax® and the ReactIR have allowed us to control the pH swing
crystallization of HNic. The level of relative supersaturation at which seeding is
performed signiﬁcantly inﬂuences the ﬁnal crystal size distribution. It was found that the
lower relative supersaturation seeding yields a much more improved mean crystal size
than high level relative supersaturation seeding. It is also demonstrated that feedback
control of relative supersaturation has enabled us to maintain the relative supersaturation
at a constant level during the post seeding acid addition process. It can be concluded that
seeding at just above saturation and then maintaining the solution about saturation

produces crystals with the maximum mean size.

87

References

l.

10.

11.

12.

13.

Mullin, J. W. Crystallization, 3rd Edition, Butterworth-Heinemann Ltd: Oxford,
1993.

Randolph, A. D. K.; Larson, M. A. Theory of Particulate Processes, Academic Press,
New York, USA., 1988.

Zimmerman, H. K. Chemical Reviews, 1952, 51, 25-65.

Dunuwila, D. D.; Berglund, K. A. J. Cryst. Growth, 1997, 179(1-2), 185-193.

. Wang, F.; Berglund, K. A.. Ind. Eng. Chem. Res. 2000, 39, 2101-2104.

Jones, A. G. K. Mullin, J. W. Chem. Eng. Sci., 1974, 29, 105-118.
Jones, A. G.; Teodossiev, N. M. Cryst. Res. T echnol, 1988, 23, 957-966.
Wang, F; Berglund, K. A. Chapter 3

Myerson, A. S. In Handbook of Industrial Crystallization; Myerson, A. S. Eds.;
Butterworth-Heinemann Ltd: Oxford, 1992; Chapter2, pp 41.

http://www.jtbaker.com/msds/n3360.htm
Larson, M. A.; Garside, J. The Chemical Engineer, 1973, June, 318-328.

Tavare, N. S.;Garside, J .; Chivate, M. R. Ind. Eng. Chem. Process Des. Dev, 1980,
19, 653-665.

Garside, J .; Davey, R. J. Chem Eng Comm, 1980, 4, 393-424.

88

Chapter 5

AN INVESTIGATION OF SOLVENT-MEDIATED POLYMORPHIC
TRANSFORMATION OF PROGESTERONE USING IN SITU
RAMAN SPECTROSCOPY

.Accepted by Organic Process Research & Development, May, 2000.

Many analytical techniques, such as Differential Scanning Calorimetry (DSC), X-Ray
Diffraction (XRD), Infrared Spectroscopy (IR) and Raman spectroscopy can be used to
differentiate between crystalline polymorphs of the same chemical entity. While all of
these techniques are routinely applied to off-line analysis of materials, Raman
spectroscopy has the advantage over these other techniques in that Raman technology
currently exists for in situ monitoring of the solid phase behavior within a mixed
suspension of liquid and solid. In this paper, we present our results ﬁ'om an in situ
Raman study demonstrating the solvent—mediated polymorphic phase transformation of
progesterone. In situ Raman analysis has shown that the appearance of Form I
progesterone is always preceded by the formation of Form 11 progesterone. Phase
transformation rates were found to increase monotonically as the temperature increases,
which indicates that the polymorphic system is monotropic. Form I was found to be
thermodynamically more stable than Form 11, while Form II was found to be kinetically
favored over Form 1. The results from this study are consistent with Ostwald's law of
stages and lead to an in-depth understanding of the polymorphic transformation process
of progesterone. The in situ monitoring capabilities of Raman spectroscopy have allowed
us to deﬁne the processing parameters required to control the morphology of crystalline

progesterone.

89

5.1 Introduction

The pharmaceutical industry is frequently confronted with the presence of multiple
crystal polymorphs of the same chemical entity. The presence of multiple polymorphs of
the active pharmaceutical ingredient is particularly challenging with solid, oral dosage
drug products." 2 The characteristics affected by the polymorphism include solubility,
dissolution rate, stability, hygroscopicity and solid state reactivity. The effect of
polymorphism on bioavailability is the most important consequence if the bioavailability
is mediated via dissolution.2’3 In situ monitoring of the solvent mediated polymorphic
transformation is therefore of great signiﬁcance in understanding the thermodynamics of
the polymorphic systems and the mechanisms of polymorphic transformation. In situ
monitoring offers tremendous potential for efﬁciently determining the crystallization

parameters required to obtain a desired crystal form.

In polymorphic studies, Raman spectroscopy, coupled with an immersible ﬁber optic
probe, has tremendous advantages over the traditional techniques, such as X-ray powder
diffraction (XRD), differential scanning calorimetry (DSC), solid state NMR and infrared
spectroscopy (IR). The ﬁber optic probe in Raman technique provides the capability of
in situ or on-line monitoring of the formation of crystals and the transformation from one
polymorph to another. DSC, XRD and NMR can perform off-line analysis, which can
distinguish the polymorph of the end product,4 but cannot reliably provide insight about
transformation processes since a polymorphic transformation may continue to occur

during sample ﬁltration, drying and sampling preparation.

90

Infrared Spectroscopy (IR) can be used to monitor some processes in situ. However
aqueous solutions may cause problems in IR measurements.5 If water is used as a
reagent, a byproduct or a solvent in a process, strong infrared absorption of water will
mask some of the useful information from the species of interest. In contrast, water is a
poor Raman scatterer, which often makes Raman preferable to IR when water is present
in the processes.6 This phenomenon can be explained by the different mechanisms of
interaction of electromagnetic radiation with matter in the two spectroscopic techniques.
Dipole moment changes during molecular vibrations cause IR absorption while changes
in polarizability associated with molecular vibrations result in Raman scattering.
Usually, molecules with polar functional groups and asymmetric vibrational modes have
stronger absorption in IR and weaker absorption in Raman.7 Raman scatterring also
favors solid phase in a slurry other than the solution phase as the ATR-FTIR technique.
The higher density of the solid phase results in stronger Raman scattering than the
solution phase does. This allows us to monitor the disappearance of one solid phase and
the appearance of the other solid phase. In addition, Raman spectroscopy can be used to
perform remote sensing because the ﬁber optic probe can be made as long as a couple of
hundred meters.8 By comparison, the probe in Attenuated Total Reﬂectance (ATR)-

FTIR can reach no further than 2-3 meters.9

Remote sensing is especially useful in
hazardous environments, where operators can work safely by staying far from the reactor.
This study employed Raman spectroscopy coupled with a ﬁber optic probe to study the
polymorphic transformation of progesterone at a laboratory scale of less than 1 liter. The

results from this study will be applied to a production scale process of several thousand

liters.

91

Progesterone has ﬁve known polymorphs.10 For the range of conditions considered in
this study, only Form 1 and Form 11 will be produced. Raman spectroscopy was empoyed
to differentiate the two forms of progesterone. In order to verify our interpretation of the
Raman spectra, several samples were also characterized by DSC and XRD and compared
to the known reference standards of these two polymorphs. Raman Spectroscopy was
also employed to monitor the solvent-mediated phase transformation of progesterone

from Form II to Form I in situ.

5.2 Experimental Section

The crude progesterone for this study was prepared at Pharmacia Corporation.
Recrystallizations of progesterone were carried out in a l-liter automatically controlled
LabMax® reactor at different temperatures ranging from 5-45°C. The process which we
analyzed involved 2 grams of progesterone dissolved in 25 mL of an organic solvent.
This solution was then added to 500 mL of distilled water, which was kept at a constant
temperature in the LabMax® reactor. Due to the heat released by mixing water with the
progesterone solution and crystallization, a jacket supplied with a cooling ﬂuid and close
temperature control provided by LabMax® were necessary to keep the temperature within
1°C of the set-point. Following the completion of the addition, the system was allowed
to stir isothermally for several hours. During this post-addition stir period, the
polymorphic transformation from Form II (kinetically favored) to Form I

(thermodynamically favored) was observed semi-continuously through in situ monitoring

92

with a HoloLab Series 5000 Raman spectrometer coupled to a remote ﬁber-optic probe

equipped with an immersion optic.

Details of the HoloLab 5000 have been described previously1 1’12. Brieﬂy, the 250 mW
output from a 785 nm stabilized diode laser is transmitted through a ﬁber-optic cable to
the remote probe. The laser light is focused into slurry using the immersion optic.
Backseattered Raman light is collected by the same optics and transmitted back to the
spectrometer for analysis. This spectrometer is equipped with a "HoloPlex" holographic
transmission grating and a thermal-electrically-cooled charge coupled device (CCD) for
detection. The HoloPlex grating allows simultaneous full spectral coverage (100 cm"l to

3450 cm'l) at high spectral resolution (4 cm") with no moving parts.

Solid state Raman analysis of pure Form I and Form II was carried out in aluminum
trays. It is important to note that sample preparation for this Raman analysis was
minimal. Unlike grinding and compression required for IR and XRD analysis (which
may alter the polymorph of the material being analyzed), the sample preparation for
Raman analysis consists simply of illuminating the crystalline sample in the vicinity of
the focal point of the incident laser from the ﬁber optic probe. For analysis of prepared
mixtures of these two polymorphs, the sample preparation was only slightly more
complex. Varying relative amounts of Form I and Form II were stirred in water and
Raman spectra was collected of the aqueous slurry. This slurry technique was used in
order to distribute the Form I and Form 11 materials homogeneously. Because of the very

limited solubility of progesterone in water, there was no concern of polymorphic

93

transformation during the 3-5 minutes required to collect any individual Raman spectrum.

Spectra were collected for a total of 11 samples of varying relative compositions of Form

I and Form 11.

The thermal properties of the polymorphs were characterized on a differential scanning
calorimeter (TA instrument Model DSC 2920 modulated DSC) and a thermogravimetric

analyzer (TA instrument Model HiRes TGA 2950).

X-Ray powder diffraction patterns of progesterone polymorphs were recorded with a
Rigaku Miniﬂex X-ray difﬁactometer. The instrument operates using the Cu kai
emission with a nickel ﬁlter at 1.50451. For each sample, ﬁne powders were packed in

an aluminum sample plate with a 2 mm indent or placed on a zero-background sample

holder.

5.3 Results and Discussion

Crystallization of progesterone at various temperatures or various time length of mixing
may yield different polymorphs. The XRD spectra (Figure 5.1(A) and (B)) show
signiﬁcant differences between the diffraction patterns of the two polymorphs. At a
heating rate of 5 °C/min, the DSC thermodiagrams of Form I and Form 11 (Figure 5.2 (A)
and (B)) present sharp endotherms with melting points at 129.1°C and 121.2°C,
respectively. TGA of both forms indicated no weight loss. The melting points of Form I

and Form II are comparable with their literature values.13

94

120‘

100‘

O
O
A

IrnonsitﬂCountuxlltO‘S)
P

00‘

 

 

 

20‘

 

 

 

I)
2-Thota(dcg)

 

 

Intanatty(Countc)

L

 

 

 

 

 

 

 

 

 

u v v v Y ff 1 ‘7 Y Tj V T r I I v v v 1 Y 1 v v Y v’f Y 1‘7 frﬁ Y I

5.0 ' 10 ' 15 an :n as 40
2-Theta(deg)

Figure 5.1. The X-ray diffraction patterns of progesterone polymorphs: (A) Form I and
(B) Form 11.

95

 

 

 

 

 

 

F’
j
.

100 110 120 130 140
Temperature (°C)

 

 

 

Figure 5.2. The DSC thermodiagrams of progesterone: (A) Form 1 and (B) Form 11. The

heating rate was 5 °C/min.

96

The Raman spectra of the solid state Form I and Form 11 are presented in Figure 5.3.
Both XRD and DSC analysis had previously established the solid phase identity of these
samples. There are several spectral regions where peak shifts are present which
differentiate between Form I and Form II of progesterone. No attempts were made to
assign all the peaks over the entire spectral region. The region where the most signiﬁcant
difference exists between the Raman spectra of the Form I and the Form 11 material can
be explained by conjugation theory.14 As shown in the structure of progesterone (Figure
5.4), the conjugation between the C=O at position 3 and C=C at position 4 results in the
delocalization of the it electrons and reduces the double-bond character of both bonds.
This causes the absorption at lower wavenumbers than those of the unconjugated
systems. As a result of this interpretation, 1617 cm"1 is assigned to C=C stretching and
1662 cm'1 to C=O stretching for Form I while 1616 cm'1 is from C=C stretching and
1667 cm'1 from C=O stretching for Form 11. The unconjugated C=O bond at position 20
contributes to the absorption at 1698 cm'1 for Form I and 1706 cm'1 for Form 11. These
Raman peak shifts indicate that the two crystalline polymorphs may have different modes
of packing. T wo ends of the molecule, functional groups around positions 3 and 20,
which are less rigid than the rings, are likely affected the most by the different packing

environments.

The peak with the largest observable shiﬁ during the transformation between the two
polymorphs (1662 cm'1 for Form I and 1667 cm'1 for Form 11) was used to set up a
calibration (Figure 5.5) to correlate the concentration of Form I with the peak position.

Eight wavenumber shift between 1698 cm"1 and 1706 cm'1 cannot be seen in the spectra

97

 

 

1662

1
1667

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l l l
0 1000 2000 3000

Raman Shift (cm “)

Figure 5.3. Raman spectra of solid state progesterone: (A) Form I and (B) Form 11.

98

20

O

Figure 5.4. Molecular structure of progesterone.

99

110

 

\O
O

\r
o
O

Conc. of Form I (wt%)
b.) LII
O O
O

p—s
O

 

 

L.
o

 

1661 1663 1665 1667

Peak Position (cm'l)

Figure 5.5. Calibration curve for the relative concentration of Form I in the mixture of
Form I and Form 11 of progesterone slurried in water.

100

of slurries during the phase transformation because the organic solvent has strong
scattering around 1700 cm'1 region, which obscured the peak shiﬁ. The spectra of eleven
mixtures of varying relative amounts of Form I and Form 11 show signiﬁcant peak shift
from pure Form I to pure Form 11. The calibration curve in Figure 5.5 clearly indicates
that we can adequately predict the relative amount of Form I to Form H simply by the

position of the peak in the region from 1662 to 1667 cm'1

. While more sophisticated
peak deconvolution techniques exist, this straightforward relationship between peak

position and solid phase composition was adequate for the purposes of this study.

Crystallizations were monitored using Raman spectroscopy over a range of temperatures
from 5 to 45°C. Each crystallization was carried out isothermally and allowed to stir for
several hours after the onset of the crystallization. During this process, the solid phase of
the system was monitored semi-continously with in situ Raman spectroscopy. The data
collected from one of these crystallizations, carried out at 45°C, are shown in Figures 5.6
and 5.7. As shown in Figure 5.6, the two bold spectra at 1667 cm'1 and 1662 cm'l, which
bracket all of the other spectra, were collected at the beginning and the end of the
crystallization, respectively. The initial peak position at 1667 cm'1 indicates that Form II
formed at the beginning of the crystallization. With additional post crystallization stir
time, the peak shifted gradually to 1662 cm'1,which indicates that Form 11 was gradually
transforming to Form 1. This process of polymorphic transformation is depicted in Figure
5.7 in which the concentration proﬁles, calculated from the calibration curve in Figure

5.5, are plotted against time. The polymorphic transformation from Form 11 to Form I

101

 

 

 

 

 

<
Form I $ “39," Form II
.é‘
E
O
E
a)
.3
E ,
a, _/
m ,
l I l
1660 1670 1680
Raman Shift (cm'l)

Figure 5.6. Raman spectra showing peak shift during the phase transformation process
at 45°C .

102

 

 

 

 

95 4
1—1 75 ~
5 -a-Fonn1
g: 55 -
“5' -9-Fonn II
a 35 ~
:
G
U 15 ~

-5 l 1 l

0 20 40 60 80

Transformation time (min)

Figure 5.7. Concentration proﬁles of Form I and Form 11 of progesterone throughout the

phase transformation process at 45°C. The lines are drawn through the data as a guide to
the eye.

103

 

L2

0
OO

Transformation Rate (hr 4)
.0 O
h 0\

O
N

 

 

 

0 10 20 30 40 50
Tem peratu re (°C)

Figure 5.8. The temperature dependence of phase transformation rate of progesterone.
The line is drawn through the data as a guide to the eye.

104

was monitored in situ at different temperatures ranging from 5 to 45°C. The
transformation rates were found to increase as the temperature increases, as shown in
Figure 5.8. This trend suggests that this polymorphic system is monotropic and Form I is

2 This was

the thermodynamically stable form while Form II is the metastable form.
conﬁrmed by the thermal analysis result that showed the melting enthalpy of Form I is
greater than that of Form II.'1 The melting enthalpy rule of Burger states that if the high
melting form has the higher melting enthalpy, they are monotrophs. If the high melting

form has the lower melting enthalpy, the two forms are enantiotrophs.l6

Therefore, in
situ Raman techniques and thermal analysis are consistent in predicting that Form I and
Form 11 are monotrophs. The phase transformation phenomenon observed by Raman
spectroscopy is consistent with Ostwald’s law of stages, which states that the formation

of a metastable phase will precede the appearance of a thermodynamically stable phase

once the supersaturation has spontaneously decreased.l7

According to Burger,l6 if Form 1 and Form H are monotrophs and Form I is the stable
form, the solubility of Form I will be lower than that of Form 11. This is also conﬁrmed
by classical nucleation theory, which states that the solid phase modiﬁcation that is ﬁrst
formed under given conditions has a higher equilibrium solubility.18 This is because the
nucleation rate is the maximum for those modiﬁcations having the lowest surface energy,
which is inversely proportional to the equilibrium solubility.l8 The solubility difference
between Form 1 and Form 11 provides additional insight about the phase transformation
process. During the crystallization process, supersaturation is removed with respect to

Form 11 as Form II crystallizes out from the solution as the initial solid phase. At a

105

certain point, the concentration decreases to the solubility of Form H. As the solubility of
Form H is higher than that of Form 1, the solution is supersaturated with respect to Form
1. The solubility difference can be so large that the solubility of Form II lies in the
unstable or labile region with respect to Form 1, where spontaneous nucleation of Form I
occurs. Even if this solubility difference is not so large, it can be reasonably assumed
that, although Form 11 is the major solid phase in the mixture, some nuclei of Form I may
also be formed.19 These initial nuclei set the starting point for the transformation process.
As the growth of the nuclei consumes the supersaturation of Form 1, the solution goes
into undersaturated zone with respect to Form II, which makes it possible for Form H to
dissolve and produce continuous supersaturation for Form I to grow. The transformation
will be complete when all of Form 11 dissolves and the solution is saturated with respect

to Form I.

The advantage of Raman over off-line techniques lies in its ability of in situ monitoring
the progress of the phase transformation between polymorphs. By contrast, an off-line
analytical technique measures the system several steps removed from the crystallizer.
These steps include drying, storage and sampling that have all been carried out on the
material. During these extraneous steps, samples may undergo phase transformation,
leading to conclusions from the off-line analysis that might be inaccurate and misleading.
Based on the powerful insight provided by Raman spectroscopy coupled with an in situ
immersion probe, instead of using off-line analytical techniques, we are able to efﬁciently
measure the transformation rate of progesterone from Form 11 to Form I over a wide

range of temperatures in this solvent system. As such, we now fully understand the

106

kinetics of polymorphic transformation of progesterone over a wide range of process
temperatures. Consequently, we are now able to specify a process for consistently and

reliably producing either of the two known polymorphs.

5.4 Conclusions

It is demonstrated that Raman spectroscopy can distinguish between Form I and Form II
progesterone crystals. The principal advantage of the Raman technique over other
methods is being able to analyze samples in their native state or without the need for
sample preparation. More signiﬁcantly, Raman spectroscopy is proved to be able to
monitor the solvent-mediated phase transformation process in situ. Transformation rates
were found to increase as temperature increases, which indicated that Form I and Form 11
of progesterone are monotrophs. The observations obtained through in situ monitoring
with Raman spectroscopy are consistent with Ostwald's rule of stages. Because this
system is monotropic, it is expected thermodynamically that Form I has a lower solubility
than Form H. The solubility difference and crystallization theory were utilized to explain
the mechanism of the phase transformation process. The data collected from in situ
monitoring of this system has allowed us to estimate the rate of polymorphic

transformation for this monotropic conversion over a wide range of process temperatures.

5.5 Acknowledgments

The authors would like to express appreciation to Kaiser Optical Systems for providing

the Raman instrument, HoloLab series 5000 and to ASI Applied Systems for providing

107

the automatically controlled LabMax® reactor. We also wish to thank Pharmacia

Corporation for permission to publish this work.

108

References

(1) Yu, L.; Reutzel, S. M.; Stephenson, G. Pharm. Sci. T echnol. T o. 1998, 1(3), 118.
(2) Giron, D. Thermochim. Acta, 1995, 248, 1.

(3) Bym, S. R. Solid-State Chemistry of Drugs, Academic Press: New York. 1982.

(4) Raghavan, K.; Dwivedi, A.; Campbell, G. C.; Nemeth, G.; Hussain, M. A. J. Pharm.
Biomed. Anal. 1994, 12(6), 777.

(5) Colthup, N. B.; Daly, L. H.; Wiberley, S. E. Introduction to Infrared and Raman
Spectroscopy; 3rd Edition, Academic press: Boston, 1990.

(6) Pelletier, M. J. Analytical Applications of Raman Spectroscopy, Blackwell Science
Ltd: Oxford, 1999.

(7) Ingle, J. D. and Crouch, S. R. Spectrochemical Analysis, Prentice Hall: Englewood
Cliffs, New Jersey, 1988.

(8) Kaiser Optical Systems, Inc. HoloLab Series 5 000 Operations Manual, Kaiser Optical
Systems, Inc.: Ann Arbor, Michigan, 1997.

(9) ASI Applied Systems, ReactIR 1000 User 's Guide; 3rd Edition, ASI Applied

Systems: Millersville, Maryland, 1997.

(10) Brandstatter-Kuhnert, M.; Koﬂer, A. 0esterr. Apoth.-Ztg. 1959, 13, 297.

(11) Battey, D.E.: Slater, J .B.: Wludyka, R. S.; Owen, H.: Pallister, D.M.; Morris, M.D.;
Appl. Spectrosc., 1993, 47, 1913.

(12) Owen, H.; Battey, D.E.; Pelletier, M.J.; Slater, J .B.; Proc. SPIE, 1995, 24(6), 260.
(13) Muramatsu, M.; Iwahashi, M.; Ushio, T. J. Pharm. Sci. 1979, 68(2), 175.

(14) Silverstein, R. M.; Bassler, C. G.; Morrill, T. C. Spectrometric Identiﬁcation of
Organic Compounds, 5th Edition, John Wiley & Sons: New York, 1991.

(15) Mullin, J. W. Crystallization, 3rd Edition, Butterworth-Heinemann Ltd: Oxford,
1993.

(16) Burger, A.; Rarnberger R. Mikrochim. Acta 11. 1979, 2(3-4), 259.

(17) Ostwald, W. Z. Physik, Chem. (Leipzig), 1897, 22, 289.

109

(l8) SOhnel, O.; Garside, J. Precipitation: Basic Principles and Industrial Applications,
Butterworth-Heinemann Ltd: Oxford, 1992.

(19) Cardew, P. T.; Davey, R. J. Proc. R. Sgc. Lond. A, 1985, 198, 415.

110

 

Chapter 6

CONCLUSIONS

This dissertation focuses on two studies including in situ monitoring and control of pH
swing crystallization and in situ monitoring of polymorphic transformation. pH swing
crystallization is more energy efﬁcient than cooling crystallization in that heating and
cooling are not necessary. It is also advantageous over antisolvent crystallization in that
separation of solvents is not involved in the post-crystallization processes. The objective
of the pH swing crystallization study was to develop a control strategy that enables the
production of a crystalline product with the desired crystal size distribution. The study of
polymorphic transformation aimed at obtaining the desired polymorph by in situ

monitoring of the polymorphic transformation process.

The basic structural properties of the crystallizing species, HNic, were studied in Chapter
2. It was found that the transformation of the molecular structure upon dissolution in
water results in signiﬁcant peak shiﬁs in its infrared spectrum. The calibrations for the
solution concentrations of both NaNic and HNic were built accordingly on their
corresponding IR absorbance regions. PLS calibration models were employed to extract
accurate information on both species from the aqueous mixture. The solubility of HNic
was found varying with the pH and with a change in C(NaNic). Attenuated Total
Reﬂection Fourier Transform Infrared Spectroscopy (ATR-FT IR) was used for

monitoring of pH swing crystallization of HNic in situ.

lll

Using the method developed in Chapter 2, the metastable zone was measured in situ with
ATR-FTIR in Chapter 3. The width of the metastable zone was found to be dependent on
the acid concentration, acid addition rate and the agitation speed. Nucleation kinetics
were determined by the relationship between the metastable zone width and the acid
addition rate. In addition, the effect of micromixing was correlated with the metastable
zone width. Crystalline products with the improved crystal size distribution were

obtained under the optimized process conditions for spontaneous(primary) crystallization.

Due to the time-consuming nucleation in spontaneous(primary) crystallization, seeding
was adopted to reduce the length of time to induce crystallization in Chapter 4. Seeding
at different levels of relative supersaturation demonstrated that increased relative
supersaturation increases secondary nucleation. With the communication between the
LabMax and the ReactIR 1000, the feedback control of the relative supersaturation was
realized to keep the solution constantly at low relative supersaturation in the post-seeding
process, which promoted crystal grth and suppressed secondary nucleation.
Consequently products with a large mean size and improved crystal size distribution were
obtained using far less amount of time as compared to spontaneous(primary)

crystallization.

The solvent-mediated polymorphic transformation of progesterone was monitored in situ
using Raman spectroscopy in Chapter 5. The Raman analysis showed that the
appearance of Form I progesterone is always preceded by the formation of Form 11

progesterone. Phase transformation rates were found to increase monotonically as the

112

temperature increases. Form I was found to be thermodynamically more stable than
Form 11, while Form 11 was kinetically favored over Form 1. It was concluded that Form
I and Form 11 of progesterone are monotropic. The solubility difference between Form I
and Form 11 was used to explain the polymorphic transformation process. The in situ
monitoring capability of Raman spectroscopy allowed us to deﬁne an operating

procedure that produces the desired polymorph.

113

 

APPENDIX

114

 

Table A.l. Data for Figure 3.2. The metastable zone of the crystallization of nicotinic
acid measured under the conditions that the solution is agitated at 1.1 m/s and 2 M HCl is
added at the rate of 60 g/hr at 30°C.

 

 

 

Metastable Limit Solubility
C(NaNic) C(HNic) C(NaNic) C(HNic)
(g. f.‘ x103) (m. f. x103) (m. f. x 103) (m. 1‘. X103)
9.28 5.23 9.51 4.06
8.28 4.81 8.35 3.86
7.43 4.51 7.41 3.73
6.42 4.43 6.39 3.46
5.44 4.16 5.55 3.31
4.60 3.89 4.67 3.07
3.89 3.68 3.80 3.04
3.16 3.57 3.26 2.92

 

‘ m. f.: mole fraction.

115

 

Table A.2. Data for Figure 3.3. Comparison of the ﬁrst metastable limit measurement
with different acid addition rates, 1 g/hr and 60 g/hr. The acid concentration is 2 M HCl
and the agitation speed is 1.1 m/s. The temperature is 30°C.

 

C(HNic) Profile with the acid addition C(HNic) Proﬁle with the acid

 

rate of 1 g/hr addition rate of 60j/hr
C(NaNic) C(HNic) C(NaNic) C(HNic)
(m. f. x 103) (m. f. x 103) (m. f. x 103) (m. f. x 103)
16.67 -0.01 16.62 0.00
16.63 0.04 16.59 0.00
15.83 0.80 16.54 0.00
15.18 1.32 16.11 0.40
14.51 1.72 15.60 0.83
14.31 1.88 15.18 1.18
13.77 2.24 14.73 1.55
13.37 2.53 14.35 1.83
12.92 2.81 13.91 2.14
12.57 3.05 13.44 2.51
12.29 3.24 13.03 2.79
11.96 3.43 12.69 3.01
11.62 3.65 12.28 3.32
11.08 4.08 11.89 3.57
10.95 4.13 11.56 3.81
10.62 4.34 11.18 4.08
10.24 4.59 11.20 4.05
10.07 4.68 11.15 4.10
9.67 4.97 11.11 4.13
9.52 5.06 11.00 4.21
9.28 5.21 11.01 4.25
9.36 5.13 10.75 4.48
9.30 5.11 10.77 4.44
9.35 4.98 10.73 4.46
9.30 4.95 10.64 4.54
9.33 4.68 10.58 4.57
9.38 4.40 10.55 4.63
9.44 4.17 10.55 4.57
9.51 4.08 10.50 4.46
9.49 4.09 10.50 4.36
9.44 4.12 10.49 4.26

 

116

Table A.3. Data for Figure 3.5. The crystal size distribution of the product from the
experiment with the addition of 0.5 M HCl at the rate of 1 g/hr and agitation speed of 1.1
m/s.

 

 

Size of the Sieve Aperture Cumulative Percent Undersize
(11m) (wt%)
0 0
127 34.9
177 54.0
250 96.7

359 100

 

117

Table AA. Data for Figure 4.1. Concentration proﬁle of the seeded crystallization of
nicotinic acid at 60% relative supersaturation seeding.

 

 

Time Since the Solubility of Metastable C(NaNic) C(HNic)
Start of Acid HNic limit of HNic (m. r. x 103) (m. 1'. x 10’)
Addition (hr) (m. r. x 103) (m. r. x 10’)

0.00 5.30 6.43 16.56 -004
0.02 5.28 6.42 16.49 0.03
0.03 5.21 6.33 16.08 0.46
0.05 5.09 6.19 15.43 1.00
0.07 4.98 6.05 14.80 1.47
0.10 4.78 5.81 13.68 2.29
0.12 4.68 5.69 13.13 2.68
0.15 4.51 5.49 12.16 3.34
0.18 4.33 5.27 11.14 4.05
0.22 4.31 5.25 11.05 4.10
0.25 4.27 5.20 10.83 4.27
0.28 4.25 5.18 10.71 4.33
0.32 4.22 5.14 10.54 4.46
0.37 4.19 5.10 10.38 4.55
0.43 4.17 5.09 10.29 4.61
0.50 4.15 5.05 10.13 4.73
0.57 4.14 5.05 10.13 4.73
0.62 4.14 5.04 10.10 4.75
0.67 4.14 5.04 10.08 4.71
0.73 4.15 5.05 10.13 4.60
0.80 4.15 5.05 10.14 4.55
0.83 4.14 5.05 10.11 4.55
0.90 4.15 5.05 10.14 4.48
0.97 4.14 5.05 10.11 4.46
1.00 4.14 5.05 10.13 4.43
1.08 4.14 5.05 10.13 4.38
1.25 4.14 5.05 10.10 4.37
1.42 4.15 5.06 10.17 4.29
1.58 4.17 5.08 10.26 4.18
1.75 4.16 5.07 10.20 4.23
1.92 4.16 5.07 10.20 4.24
2.08 4.16 5.07 10.20 4.23
2.25 4.16 5.07 10.22 4.20
2.58 4.16 5.07 10.21 4.22
2.92 4.16 5.07 10.23 4.19

 

118

Table A.5. Data for Figure 4.3. Turbidity proﬁle for 60% relative supersaturation seeding
crysallization.

 

 

Time since the Start of Acid Addition Turbidity
1hr) (Relative Intensity)
0.00 0.02
0.10 0.02
0.20 0.02
0.30 0.02
0.40 0.02
0.50 0.02
0.60 0.02
0.62 0.02
0.63 0.02
0.65 0.13
0.67 0.16
0.70 0.21
0.73 0.26
0.77 0.28
0.80 0.31
0.83 0.33
0.87 0.34
0.90 0.39
0.93 0.41
0.97 0.43
1.00 0.42
1.08 0.44
1.25 0.55
1.42 0.55
1.58 0.55
1.75 0.58
1.92 0.50
2.08 0.54
2.25 0.53
2.42 0.53
2.58 0.53
2.75 0.50

2.92 0.50

 

119

 

Table A.6. Data for Figures 4.6 and 4.7. The comparison of the crystal mean size and the
crystal size distribution resultant from seeding at three levels of relative supersaturation,

30%, 45% and 60%.

 

 

 

 

Size of the Sieve Aperture Cumulative Percent Undersize
(11m) (wt%)
30%-SR 45%— SR 60%- SR
0 0 0 0
127 4 18 20
177 34 41 73
250 86 89 94
359 98 100 100
550 100
Crystal Mean Size (pm) 204 185 156

 

120

Table A.7. Data for Figure 4.14. The comparison of the mean crystal size and the crystal
size distribution resultant from three controlled relative supersaturation experiments. I.
Seeding at 0%— SR and control about 0%- SR; H. Seeding at 30%- SR and control about
0%- SR; III. Seeding at 30%- SR and control about 30%- SR.

 

 

 

Size of the Sieve Aperture Cumulative Percent Undersize
(pm) (wt%)
0%- SR 30%-0%- SR 30%- SR

0 0 0 0

127 3 11 67

177 18 22 85

250 77 78 96

359 98 99 100

550 100 100

121

Table A.8. Data for Figure 5.5. The concentration of Form I progesterone vs. the peak
position around 1660 -1667 cm".

 

 

wt% of Form I Peak Position

1cm")
0 1667.2
10 1666.7
20 1666.2
30 1665.9
40 1665.2
50 1664.6
60 1663.9
70 1663.2
80 1662.8
90 1662.3

100 1661.9

 

122

 

Table A.9. Data for Figure 5.7. Concentration proﬁles of Form I and Form 11 of
progesterone throughout the phase transformation at 45°C.

 

 

Transformation time C(Form I) C(Form II)
(min) (wt%) (wt%)
5 6 94
11 21 79
22 49 51
32 69 31
43 83 17
54 100 0
66 100 0

 

123

 

Table A.10. Temperature dependence of phase transformation rate of progesterone.

 

 

Temperature (°C) Transformation Rate (hr'l)
5 0.03
1 5 0.05
25 0.19
35 0.45
45 1.00

 

124