Crystallization of paraxylene with scraped surface heat exchangers

WTHD - No. 189

Crystallization of paraxylene with

scraped surface heat exchangers

September 1988 R. de Goede ^Mfy

TR diss

1665

U Delft

Faculty of Mechanical Engineering and Marine Engineering

I

, ■

-FACULTY OF MECHANICAL ENGINEERING AND

MARINE ENGINEERING

Delft University of Technology

The Netherlands

CRYSTALLIZATION OF PARAXYLENE WITH

SCRAPED SURFACE HEAT EXCHANGERS

by

R. de Goede

WTHD no. 189

September 1988

Laboratory for Process Equipment

Leeghwaterstraat 44, 2628 CA Delft

TRdiss

1665

CIP-GEGEVENS KONINKLIJKE BIBLIOTHEEK, DEN HAAG

Goede, R. de

Crystallization of paraxylene with scraped surface heat exchangers /

by R. de Goede. - Delft: Laboratory for Process Equipment, Faculty of Mechanical En­ gineering and Marine Engineering, Delft University of Technology.- (WTHD; no.189) Proefschrift Delft. - Met lit. opg. - Met samenvatting in het Nederlands.

ISBN 90-370-0018-5

SISO 542.3 UDC 532.78:54(043.3) Trefw. paraxyleen; kristallisatie.

Copyright © 1988, Faculteit der Werktuigbouwkunde en Maritieme Techniek, Technische Universiteit Delft

Alle rechten voorbehouden.

Niets uit dit rapport mag op enigerlei wijze worden verveelvoudigd of openbaar gemaakt zonder schriftelijke toestemming van de auteur.

Gebruik of toepassing van de gegevens, methoden en/of resultaten enz., die in dit rapport voorkomen, geschiedt geheel op eigen risico. De Technische Universiteit Delft, Faculteit der Werktuigbouwkunde en Maritieme Techniek, aanvaardt geen enkele aansprake­ lijkheid voor schade, welke uit gebruik of toepassing mocht voortvloeien.

Any use or application of data, methods and/or results etc., occurring in this report will be at user's own risk. The Delft University of Technology, Faculty of Mechanical Engineer­ ing and Marine Engineering, accepts no liability for damages suffered from the use or ap­ plication.

A CKNOWLEDGEMENTS

I hereby would like to express my gratitude to all those who contributed to this thesis and to the execution of the underlying project. In particular:

All employees of the mechanical- and electronical workshops of the Laboratory for Process Equipment, TU-Delft, who took care of the construction and maintenance of the major part of the equipment;

ing. J.H.P. Rotteveel, for coordinating the work mentioned in the foregoing item; All students who delivered their contributions in the scope of their M.Sc-work, litera­ ture research and other assignments;

dr. ir. G. Hakvoort from the Laboratory for Physical Chemistry (TUD) for his valu­ able input to the D.S.C.- work;

dr. H. van Koningsveld and J.C. Jansen for their effort put into the determination of the crystal structure of paraxylene;

drs. C. F. Woensdregt (RU-Utrecht) and ing. W.G. Marchee for their help with the determination of the morphology and crystal growth kinetics of paraxylene crystals; Prof. dr. ir. J. de Graauw for his stimulating discussions about heat transfer modell­ ing;

Mrs. M.J.J. Tetteroo-La Croix and Mrs. P.W.M. van Hagen for typing out the manuscript;

Mr. W. Hoogstad and Mr. B. Sodderland, for drawing the major part of the graphs and illustrations;

ir. J.E. Hille and dra. J.C. Golunski, from Kinetics Technology International, Zoeter-meer, for their contributions to the realization of the final layout;

dr. ir. L.J.M.J. Blomen, also from K.T.I., for his support;

Finally, I would like to thank Esso Nederland B.V. for sponsoring this project. The sequence of the acknowledgements is random.

TABLE OF CONTENTS SUMMARY I SAMENVATTING III Chapter 1 INTRODUCTION 1 1.1 USE OF PARAXYLENE 1 1.2 SOURCES OF PARAXYLENE 2 1.3 METHODS FOR SEPARATION OF PARAXYLENE FROM

A C8-AROMATIC MIXTURE 4 1.4 CRYSTALLIZATION PROCESSES FOR ISOLATION OF

PARAXYLENE 5 1.4.1 Processes Using Direct Contact Cooling 5

1.4.2 Processes Using Indirect Cooling 6 1.5 PURIFICATION OF THE PRODUCT 7

1.5.1 Washing Columns 7 1.5.2 Distillative Freezing 8 1.6 THE PARAXYLENE UNrT USED BY EXXON 11

1.7 FORMULATION OF THE PROBLEM 14

1.8 SCOPE OF INVESTIGATIONS 15 1.9 SET-UP OF INVESTIGATIONS AND OF THIS THESIS 16

1.10 REFERENCES 18 Chapter 2

SOLID-LIQUID EQUILIBRIA IN MIXTURES OF C8-AROMATICS AND

TOLUENE AT ATMOSPHERIC PRESSURE 19

2.1. INTRODUCTION 19 2.2. OBJECTIVE AND SCOPE OF THE INVESTIGATION 20

2.3. SET-UP OF THE INVESTIGATIONS 20 2.4. THEORETICAL DESCRIPTION OF SOLID-LIQUID EQUILIBRIA 21

2.4.1 General Approach 21 2.4.2 Non-ideality Of The Liquid Phase 26

2.5. EXPERIMENTAL 29 2.5.1. Sampling And Analysis Of Crystals At

The Plant 29 2.5.2. Analysis Of Solid And Liquid Phases In Equilibrium 29

2.6. RESULTS AND DISCUSSION 38 2.6.1. Analysis Of Plant Product 38 2.6.3. Equilibrium Measurements On Laboratory Scale 39

2.6.3. D.S.C. Measurements 41

2.7. CONCLUSIONS 47 2.8. REFERENCES 48 2.9 LIST OF SYMBOLS 50

Chapter 3

STRUCTURE AND CRYSTAL GROWTH PHENOMENA OF

PARAXYLENE CRYSTALS 51

3.1 INTRODUCTION 51 3.2 SCOPE AND OBJECTIVES OF THE INVESTIGATIONS 51

3.3 SET-UP OF THE INVESTIGATIONS 52 3.4 DETERMINATION OF THE CRYSTAL STRUCTURE 52

3.4.1 Experimental 52 3.4.1.1 Single-crystal Structure Determination 52

3.4.1.2 Powder Diffraction Analysis 53

3.4.2 Results And Discussion 53 3.4.2.1 Molecular Structure And Crystal Packing 53

3.4.2.2. Comparison Between Paraxylene And

a - Toluene 58 3.4.2.2 Conformation Of Cell Dimensions By

Powder Diffraction 60 3.5 THE MORPHOLOGY OF PARAXYLENE CRYSTALS 61

3.5.1 Theoretical Background 61 3.5.2 Prediction Of The Morphology From X-ray Data 62

3.5.3 Experimental 65 3.5.4 Results And Discussion 65

3.6 CRYSTAL GROWTH KINETICS 66 3.6.1 Theoretical Modelling Of Crystal Growth Kinetics 66

3.6.2 Experimental 75 3.6.3 Results And Discussion 77

3.7 CONCLUSIONS 88 3.8 REFERENCES 88 3.9 LIST OF SYMBOLS 90

Chapter 4

HEAT TRANSFER PROPERTIES OF A SCRAPED SURFACE HEAT

EXCHANGER UNDER NON-CRYSTALLIZING CONDITIONS 91

4.1 INTRODUCTION 91 4.2 SCOPE OF INVESTIGATIONS 91

4.3 SET-UP OF THE INVESTIGATIONS 92

4.4 LITERATURE SURVEY 92

4.5.1 Turbulent Pipe Flow 94 4.5.2 The Influence Of The Scraper Action On The Heat

Transfer Coefficient 96 4.5.3 The Influence Of The Scraper Action On The Flow

Pattern 100 4.6 EXPERIMENTAL 102

4.6.1 Visualization Of The Flow Pattern 102 4.6.2 Measurement Of Heat Transfer Coefficient 103

4.7 DATA DEDUCTION 105 4.8 RESULTS AND DISCUSSION 106

4.8.1 Visualization Of The Flow Pattern 106

4.8.2. Heat Transfer Coefficients 107

4.9 CONCLUSIONS 109 4.10 REFERENCES 109 4.11 LIST OF SYMBOLS 110 Chapter 5

HEAT TRANSFER PROPERTIES OF A SCRAPED SURFACE HEAT EXCHANGER UNDER BOTH CRYSTALLIZING AND

NON-CRYSTALLIZING CONDITIONS ON PILOT-PLANT SCALE 111

5.1 INTRODUCTION 111

5.2 SCOPE 111 5.3 SET-UP OF THE INVESTIGATIONS 112

5.4 LITERATURE 112 5.5 THEORY 113

5.5.1 Coupled Heat- And Mass Transfer At The Inner Side 113 5.5.1.1 Temperature- And Concentration Profile In

Axial Direction 114 5.5.1.2 Temperature- And Concentration Profiles In

Radial Direction 116 5.5.1.3 Growth Phenomena Of The Crystal Layer 117

5.5.1.4 Heat And Mass Flux To The Wall 118 5.5.1.5 The Heat Transfer Coefficient At The Inner

Side 119 5.5.2 Heat Transfer At The Coolant Side 120

5.6 EXPERIMENTAL 120 5.6.1 Chemicals 120 5.6.2 Equipment 120 5.6.3 Procedure 123 5.7 DATA DEDUCTION 124 5.8 RESULTS AND DISCUSSION 125

5.8.1 Measurements Under Non-crystallizing Circumstances 125 5.8.2 Measurements Under Crystallizing Circumstances 127 5.8.3 The Heat Transfer Coefficient At The Coolant Side 136

5.9 CONCLUSIONS 136 5.10 REFERENCES 137 5.11 LIST OF SYMBOLS 137

Chapter 6

FINAL DISCUSSION 139

6.1 INTRODUCTION 139 6.2 CHOICE OF PROCESS 139 6 3 THE SIGNIFICANCE OF THE RESULTS FOR THE

EXXON-PROCESS 140 6.3.1 Performance Of The Scraped Surface Heat Exchangers 140

6.3.2. Overall-performance Of The Plant 141 6.3.2.1. The Population Balance Concept 141

6.3.2.2. Estimations Based On Foregoing Results 142

6.3.2.3. Experimental Results 144 6.3.3 Recommendations 146 ER INVESTIGATIONS 147 147 148 6.4 6.5 6.6

SUGGESTIONS FOR FUR" REFERENCES LIST OF SYMBOLS APPENDIX I PHYSICAL PROPERTIES AI.1 AI.2 AI.3 Al.4 C8-AROMATICS METHYLENECHLORIDE FREON 22 REFERENCES 149 149 150 150 150 APPENDIX II

EXPERIMENTAL DATA OF SOLID-LIQUID EQUILIBRIA 151

AII.1 PLANT PRODUCT AN ALYSIS 151 AII.2 ANALYSIS OF SOLID AND LIQUID IN EQUILIBRATION 152

AII.2.1 Binary Mixtures 152 AII.2.2 Multicomponent Mixtures 153

AII.3 D.S.C.-MEASUREMENTS 154 APPENDIX III

EXPERIMENTAL DATA OF CRYSTAL STRUCTURE DETERMINATION 157 APPENDIX IV

SUPPLEMENT TO CHAPTER 4

AIV.1 EXPERIMENTAL DATA OF HEAT TRANSFER

MEASUREMENTS ON BENCH SCALE 159

AIV.2 THE PHOENICS PROGRAM 160 AIV.2.1. General Description 160 ATV.2.2. Application To The Scraped Surface Heat Exchanger 161

APPENDIX V

EXPERIMENTAL DATA OF HEAT TRANSFER

SUMMARY

In this thesis, a study on several aspects of crystallization of paraxylene with a scraped sur­ face heat exchanger process is presented. The main objectives are to obtain useful tools to increase the purity of the final product and to reach a more efficient way of processing. The mechanisms leading to contamination of the product can be divided into two classes: (i) in­ corporation of impurities in the crystal body, and (ii) attachment of impurities to the crys­ tal surface. Class (i) mechanisms involve the formation of solid solutions because of ther-modynamic and /or kinetic reasons. Class (ii) mainly involves incomplete solid-liquid separation. Therefore, the specific surface area strongly determines the contribution of the second mechanism. Hence, control of the crystal size distribution is required in order to optimize the product quality for a given combination of crystallization- and purification steps. The underlying crystallization process is inefficient in the sence that incrustration at the wall of the scraped surface heat exchangers severely disturbs heat transfer. The presence of a solid layer introduces an additional resistance against heat transfer. There­ fore, a lower coolant temperature is required than without incrustration. The set-up and outcome of the separate investigations on the aspects mentioned above are summarized as follows:

In chapter 1, a brief description of the most common crystallization processes for paraxylene is presented. They all have in common that the product purity does not fulfill the market requirements. Two alternative methods to upgrade the product have been discussed: (i) washing columns, and (ii) distillative freezing. Washing columns are preferred because of their ability to remove impurities from the crystal matrix by recrystallization.

In chapter 2, a study is presented on the incorporation of impurities in the crystal matrix. First, samples were taken at the plant and analyzed with GC. The results showed an impurity concentration of 0,3%. The concentration ratio's of the impurities in the crystal matrix appeared to differ from those in the liquid phase, indicating a dif­ ferent tendency towards incorporation of the individual components in the crystal matrix. Additional measurements were carried out by slow equilibration of solid- and liquid phases on laboratory scale in order to eliminate kinetic effects. No stable solid solutions of paraxylene with the other C8-isomers were found. However, toluene and paraxylene have proven to form stable solid solutions. Measurements with D.S.C. in­ dicated that metastable solid solution formation of paraxylene with orthoxylene and metaxylene frequently occurs upon cooling rates in the order of 4°C/min. For design purposes phase equilibria can be described assuming ideal behaviour of the liquid phase and formation of a pure solid phase. A more accurate way of calculating solubilities involves activity coefficients estimated from heats of mixing in the liquid phase.

In chapter 3, more attention has been paid to the crystalline state and its growth properties. Determination of the crystal structure at 180 K revealed a striking simi­ larity with the structure of toluene crystals, which explains the ease with which these

Crystallization of Paraxylene with Scraped Surface Heat Exchangers

two compounds form solid solutions. Additional powder diffraction data showed that no phase transition in the solid phase occurs between 180 K and the melting point. The crystallographic data have been used to predict the morphology of paraxylene crystals. The predicted morphology has been compared with the morphology of paraxylene crystals grown from different solvents. From these experiments, it ap­ peared that the main part of the crystals indeed have the predicted morphology, but a second type also occurs with a S-face as the predominant one. No correlation be­ tween conditions and morphology appeared from the experiments. Measurements of crystal growth kinetics showed that paraxylene crystals probably grow according to the spiral growth mechanism. When growing from the melt, instabilities like dendrite formation have been observed at an undercooling beyond 2°C.The growth of para­ xylene crystals growing from solution depends on surface integration as well as on volume diffusion under the given circumstances on laboratory scale.The growth rate appeared to be length-independent, but the spiral growth mechanism may lead to growth rate dispersion. Extrapolation of the results towards process conditions learns that growth rates of 5*10' - 10" m/s are to be expected at-45°C.

In chapter 4, a study is described comprising heat transfer properties. Because the wall temperature is the lowest temperature encountered in the process, it is supposed to have a strong impact on the nucleation rate. To obtain well-defined conditions, a bench-scale model of the scraped surface heat exchanger was built to determine the heat transfer coefficient as a function of volumetric throughput and scraper rotation­ al speed. A theoretical model based on penetration theory has been developed and the agreement between theory and experiment was satisfactory provided that each scraper action is counted for two surface renewal events: the first one because of directly wiping off the thermal boundary layer and the second one due to the genera­ tion of a vortex, that causes an additional surface renewal event to occur. A significant dependence of the heat transfer coefficient on the scraper rotational speed has been observed.

In chapter 5, the outcome of investigations on heat transfer properties on pilot plant scale is presented.The model presented in chapter 4 turned out to be inapplicable here. The discrepancy is probably due to the wall roughness of the pilot plant heat ex­ changer. The amount of incrustration appeared to be independent of flow rate and scraper rotational speed. The undercooling at the phase boundary turned out to be limited to only 3°C .Lowering the wall temperature results in further growth of the crystal layer, thereby counteracting the effect of an increasing driving force by an in­ crease in resistance. An analysis of the temperature- and concentration profiles near the wall shows that constitutional undercooling occurs. This phenomenon explains the observations if it is assumed that the scraper only removes the growth irregularities from the surface but not the entire crystal layer.

In chapter 6, the results of the foregoing chapters are related with the performance of the real process. Probably, the majority of the crystals rise at the wall of the scraped surface heat exchangers, due to scraping off the growth irregularities. For this reason, the scraper rotational speed is assumed to play a crucial role in the overall-nucleation rate.

Ill Crystallization ofParaxylene with Scraped Surface Heat Exchangers

SAMENVATTING

In dit proefschrift is een onderzoek beschreven naar verschillende aspecten van kristal-lisatie van paraxyleen met schrapende warmtewisselaars. De voornaamste doelstellingen zijn het verhogen van de produktzuiverheid en het bereiken van een meer efficiënte proces­ voering. De mechanismen die leiden tot verontreiniging van het eindprodukt kunnen wor­ den onderverdeeld in twee klassen: (i) inbouw van verontreinigingen in het kristal, en (ii) verontreinigingen aan het kristaloppervlak door aanhangende moederloog. Klasse (i) mechanismen omvat de vorming van vaste oplossingen door thermodynamische dan wel kinetische oorzaken. Klasse (ii) omvat hoofdzakelijk onvolledige vast/vloei-stofscheiding en dus wordt de bijdrage van mechanisme (ii) voornamelijk bepaald door het specifieke kristaloppervlak. Daarom is het noodzakelijk de kristalgrootteverdeling voor een gegeven combinatie van kristallisatieproces en zuiveringsstap in de hand te kunnen houden. Het onderzochte kristallisatieproces is inefficiënt in die zin dat aankorsting op de wanden van de schrapende warmtewisselaars de warmteoverdracht nadelig beïnvloedt. De aanwezig­ heid van een kristallaag op de wand geeft een extra weerstand tegen warmteoverdracht waardoor een lagere temperatuur van het koelmiddel nodig is dan wanneer er geen aankorsting op zou treden. De aanpak en resultaten van de verschillende deelonderzoeken zijn in de volgende punten samengevat:

In hoofdstuk 1 is een korte beschrijving gegeven van de meest bekende kristal-lisatieprocessen voor paraxyleen. In alle gevallen blijkt de zuiverheid van het produkt niet aan de eisen te voldoen. Twee alternatieve zuiveringstechnieken zijn beschreven nl. (i) waskolommen en (ii) destillatieve kristallisatie. Waskolommen zijn aan te bevelen omdat deze het produkt omkristalliseren en zodoende ook verontreinigin­ gen in het kristal verwijderen.

In hoofdstuk 2 is een onderzoek beschreven naar inbouw van verontreinigingen in het kristal. Ten eerste werden monsters genomen van kristallen uit het werkelijke proces waarvan de samenstelling werd bepaald met GC. De totale concentratie verontreinigingen bleek gelijk te zijn aan 0,3%- De onderlinge verhouding van de concentraties van de individuele verontreinigingen bleek af te wijken van die in de vloeistoffase, hetgeen wijst op een verschillende affiniteit voor inbouw van de af­ zonderlijke componenten. Op laboratoriumschaal werden aanvullende metingen verricht door langzame evenwichtsinstelling tussen vaste- en vloeibare fase te be­ werkstelligen om zodoende kinetische invloeden uit te sluiten. Hieruit bleek dat de C8-aromaten onderling geen vaste oplossingen vormen. Paraxyleen bleek echter wel een stabiele vaste oplossing te vormen met tolueen. D.S.C.- metingen gaven als resul­ taat dat de vorming van metastabiele vaste oplossingen veelvuldig optreedt bij een afkoelsnelheid rond 4°C/min. Voor ontwerpdoeleinden kunnen de vast-vloeistof-evenwichten worden beschreven onder de aanname van ideaal gedrag in de vloei­ stoffase en de vorming van een zuivere vaste fase. Een nauwkeuriger benadering is mogelijk door activiteitscoefficienten te schatten op basis van mengwarmtes in de vloeibare fase.

Crystallization ofParaxylene with IV Scraped Surface Heat Exchangers

In hoofdstuk 3 is aandacht besteed aan de kristallijne toestand en de groeifenomenen daarvan. Uit de opheldering van de kristalstructuur bij 180 K bleek een sterke over­ eenkomst te bestaan met de structuur van tolueenkristallen. Deze waarneming verklaart de vorming van vaste oplossingen van deze twee componenten. Uit poeder­ diffractie metingen bleek dat er tussen 180 K en het smeltpunt geen faseovergang optreedt. Vanuit de kristallografische gegevens werd de morfologie voorspeld en de voorspelde morfologie werd vergeleken met die van paraxyleenkristallen gegroeid onder verschillende omstandigheden. Van het grootste gedeelte van de kristallen bleek de morfologie overeen te komen met de voorspelling. Daarnaast komt echter een tweede type voor met een S-vlak als het voornaamste vlak. Er is geen verband gevonden tussen morfologie en experimentele omstandigheden. Uit metingen van de groeikinetiek bleek dat paraxyleenkristallen waarschijnlijk voornamelijk groeien vol­ gens het spiraalgroeimechanisme. Bij groei uit de smelt werden groeiïnstabiliteiten zoals dendrietgroei waargenomen bij onderkoelingen groter dan 2°C. Bij groei vanuit oplossingen blijkt dat onder de gegeven omstandigheden van het laboratoriumexperi­ ment de groeisnelheid afhangt van zowel oppervlakteintegratie als volumediffusie. De groeisnelheid bleek lengte-onafhankelijk te zijn. Het spiraalgroeimechanisme kan echter aanleiding geven tot groeidispersie. Extrapolatie van de resultaten naar procesomstandigheden geeft een verwachtingswaarde van de groeisnelheid bij -45°C van5*10"9-10"rm/s.

In hoofdstuk 4 is een onderzoek beschreven naar warmteoverdracht onder niet-kris-talliserende omstandigheden. Omdat de temperatuur aan de wand van de schrapende warmtewisselaars de laagste is die in het proces voorkomt, wordt deze verondersteld een grote invloed te hebben op de totale kiemvormingssnelheid. Om onder goed gedefinieerde omstandigheden te kunnen werken werd een laboratoriummodel van een schrapende warmtewisselaar gebouwd. Hiermee werden warmteoverdrachts-coëfficienten gemeten als functie van doorzet en schrapertoerental. Een theoretisch model werd ontwikkeld op basis van de penetratietheorie. De overeenkomst tussen model en experiment was bevredigend mits aan iedere schraperactie twee grenslaag­ verversingen werden toegeschreven: een door het afvegen van de thermische grens­ laag en een door de vorming van een wervel, die een extra verversing bewerkstelligt. Een duidelijke verhoging van de warmteover-drachtscoëfficient door de schraperac­ tie werd waargenomen.

In hoofdstuk 5 zijn de resultaten van onderzoek naar warmteoverdracht met een proeffabriek gepresenteerd. Het model, ontwikkeld in hoofdstuk 4, bleek hier niet overeen te stemmen met de experimentele resultaten. De afwijking werd toe-ge­ schreven aan de ruwheid van de wand van de warmtewisselaar. De mate van aankorst-ing bleek niet te beïnvloeden te zijn door schrapertoerental of doorzet. De haalbare onderkoeling aan het grensvlak vast/vloeistof bleek beperkt te zijn tot 3°C. Verlag­ ing van de wandtemperatuur heeft slechts tot gevolg dat de kristallaag verder aan­ groeit en zodoende het effect van de grotere drijvende kracht teniet doet door het vergroten van de weerstand. Uit een modelmatige beschrijving van temperatuur- en concentratieprofielen aan de wand blijkt dat constitutionele onderkoeling optreedt. Dit verschijnsel verklaart alle waarnemingen indien aangenomen wordt dat de

V Crystallization ofParaxylene with Scraped Surface Heat Exchangers

schraper alleen de onregelmatigheden van het oppervlak afschraapt maar niet de hele kristallaag.

In hoofdstuk 6 zijn de resultaten uit de voorafgaande hoofdstukken in verband gebracht met het gedrag van het werkelijke proces. Waarschijnlijk ontstaat het grootste gedeelte van de kristallen op de wanden van de schrapende warmtewis­ selaars door het afschrapen van onregelmatigheden. Daarom wordt een sterke in­ vloed verwacht van het schrapertoerental op de kiemvormingssnelheid.

Chapter 1

INTRODUCTION

1.1 USE OF PARAXYLENE [1]

After the second world war, paraxylene has become important as the basic material for the production of the polyester PET, polyethyleneterephtalate. The first step is liquid phase air oxydation of paraxylene to terephtalic acid using a Cobalt catalyst:

C H

3-0)^

3

OX

H O - C ^ Q } > - C - O H (l.l)

This reaction is carried out in acetic acid or dilute nitric acid as a solvent. The terephtalic acid produced by this reaction is of technical grade quality. Further purification to polymer-grade terephtalic acid or esterification with methanol yielding polymer-polymer-grade dimethyl-terephtalate is the subsequent step in the process. Reaction of these products with ethylene glycol leads to the monomer according to eq. (1.2) or (1.3):

O O (i.2) H O - C H ^ > - C - O H + 2 HO-CH2-CH2-OH O O — H O - C H2- C H2- 0 - C H ^ - C - 0 - C H2- C H2- O H + 2 H20 C H3- o - C - @ - C - 0 - C H3 + 2 HO-CH2-CH2 OH H O O 0 - C H2- C H2- 0 - C H Q ^ C 0-CH2-CH2-0» — n (1.3)

o o

Polycondensation of the monomer yields PET:

O O

n * H O - C H2- C H2- 0 - C - < Q ) - C - 0 - C H2- C H2- O H

(1.4)

2 Crystallization of Paraxylene with Scraped Surface Heat Exchangers

Until 1965, all PET was produced from dimethylterephtalate.The route via polymer-grade terephtalic acid, however, has proven to be advantageous in a number of aspects: e.g. no methanol is being formed as a by-product and the product itself has a better quality.

The mechanical properties of PET depend strongly on the mean molecular weight and the fraction of crystalline material. PET is widely applied for the manufacture of fibers which, blended with cotton or wool, are used in the textile industry. Owing to the high strength of the material it is also used in heavy yarns for application in the car industry e.g. for seat belts, V-belts and tire cord. Besides high strength favourable properties of PET are its resis­ tance against sunlight, mineral acids (except for concentrated H2SO4) and water. Hydro­ lysis is only noticeable above 100°C: after one week the material looses 20% of its strength under these conditions. Besides fiber manufacturing, PET is also applied in films and resins.

A brief overview of the most important process routes for the production of paraxylene is presented in the next section.

1.2

SOURCES OF PARAXYLENE [1]

The main source of paraxylene is petroleum. After reformation of the fraction with boil­ ing range 65-175°C, the light- and heavy components are removed by distillation and the heart-cut of the reformate is subjected to an extraction process, where the aromat.es are separated from the non-aromatics. A general scheme for the further proces*::'''? of the aromatics is presented in figure 1.1.

light components r isomeruaMon heart t u t reformate extraction aromatics benzene toluene non aromatics ^ 9 Cg aromatics P-xylene separation paraxyle

T

Introduction 3

Benzene and toluene are recovered as top-products in two subsequent distillation steps.

The thir<i distillation yields a mixture of the C8-aromatics ortho-, meta-, paraxylene and

ethylbeni>e:ie at the top and a bottom product that contains the major part of the or-thoxylene and the heavier aromatics. This part of the oror-thoxylene is recovered by another distillation step.

Then, the C8-aromatic mixture is fed to a unit where paraxylene is separated from its isomers (see section 1.3). The remaining part of the mixture is sent to an isomerization sec­ tion. Elevated temperatures and the use of a catalyst allow the composition to shift towards equilibrium and paraxylene is thus produced from its isomers. During the isomerization process, other components than C8-aromatics are unavoidably formed. The lighter com­ ponents are distilled off and the remaining mixture that consists of C8-aromatics and heavier components is recycled to the third distillation column.

The production capacity of a refinery plant for xylenes can be increased by the addition of a transalkylation process. Commercial routes for this reaction are:

(i) disproportionation of toluene:

2 <O^CH

= Q ♦ CH

-<§> (..5)

C P U

and:

(ii) reaction of a mixture of toluene and C9-aromatics:

f ^

, ♦ ® - C H , ■= 2 C H 4 ^

3 2

The reactions (1.5) and (1.6) reach an equilibrium that depends on the applied conditions. Although the world-wide production of C8-aromatics by transalkylation is much less than the amount produced by the physical separation train as presented in figure 1.1, it provides the manufacturer with the ability of a greater flexibility in fulfilling the market require­ ments.

In 1980 the relative contribution of the C8-aromatics produced in the U.S. from the petroleum reformate fraction by distillation, as shown in fig. 1.1, amounted 94.5%. Another 0.9% was produced by transalkylation. The remaining 4.6% was produced by pyrolysis of gasoline and from coke oven light oil, but because of their minor importance, no further attention will be paid here to these routes.

4 Crystallization of Paraxylene with Scraped Surface Heat Exchangers

13 M E T H O D S FOR SEPARATION O F PARAXYLENE F R O M A C8-AROMATIC MIXTURE

Isolation of p-xylene from the other C8-aromatics is not possible by distillation because the boiling points of the components are too close.

However, coolcrystallization has proven to be a successful separation technique and this is still the most widely applied method.

In the concentration range(20 - 25%) that is usually found in the C8-aromatic mixture as it occurs in the separation train described in the foregoing section, paraxylene is the first component that crystallizes upon cooling. Crystallization starts at -30 to -35 C and the lower temperature limit is determined by cocrystallization of one of the other components.

Usually this occurs when the paraxylene-metaxylene eutectic is reached at a temperature of approximately -60°C. At this temperature, the remaining concentration of paraxylene is about 10%, so the recovery per pass is approximately 60%.

The modern trend in isolation of paraxylene from a C8-aromatic mixture is by means of molecular sieve adsorption. The advantage of this process is a higher paraxylene recovery per pass (90-95%) compared with crystallization. This means that by applying molecular sieve adsorption, more paraxylene can be produced when other streams such as the feed stream to the p-xylene separation unit, the feed stream to the isoformer section and the recycle streams to the destination columns are kept constant. For this reason, this process is economically more attractive than a crystallization process.

The most widely applied molecular sieve adsorption process is the Parex process [2]. This process is based on the principle of continuous selective adsorption in the liquid phase, employing a fixed bed of solid adsorbent. The adsorbent is made from zeolite material and all components are allowed to enter the pore structure. So, the separation technique is not based on shape selectively but solely on small differences in affinity to the adsorbent. Paraxylene has the strongest affinity to the adsorbent and is thus preferentially absorbed. The affinity of the desorbent liquid has to lie between those of paraxylene and the other feed components. When the affinity is too low, it will take a lot of effort to remove the paraxylene from the adsorbent. If the affinity is too high, the feed mixture is not capable of removing the desorbent. Furthermore, the volatily of the desorbent should differ sufficient­ ly from that of paraxylene to allow for separation of the paraxylene-desorbent mixture by distillation. Paradiethylbenzene has proven to be a suitable desorbent.

Although the molecular sieve adsorption processes offer some advantages compared with the conventional crystallization technology, a definitive choice between these two ways of processing can only be made when possiblities for improvement of crystallization technol­ ogy have sufficiently been investigated. An attempt in this direction has been reported in this thesis. For this reason, a brief review of the most common crystallization processes for paraxylene is presented in the next section.

Introduction _S

1.4 CRYSTALLIZATION PROCESSES FOR ISOLATION OF

PARAXYLENE

Crystallization of paraxylene is usually carried out in two stages. In the first stage, the C8-aromatic mixture is cooled at -60°C, the temperature of the paraxylene-metaxylene eutec-tic. At this temperature, the liquid phase still contains 8 -10% paraxylene. After separation of the crystals from the mother liquor, the latter is first sent to a heat exchanger to precool the feed stream and then to the isomerization reactor. The crystal cake is remelted and sub­ jected to a second crystallization stage for further purification. A general scheme for this

two-stage process is presented in figure 1.2.

2 stage mother liquor

e 1 stage crystallizer

1 stage mother liquor

to isomerization

Figure 1.2 General outline of a two-stage crystallization process.

In principle, the recovery per pass can be increased by addition of carbon tetrachloride, that forms an equimolar solid compound with paraxylene but not with the other isomers [3,4]. The eutectic formed by metaxylene and the solid compound occurs at a temperature of = -75°C, the equilibrium mole fraction of paraxylene in the liquid phase being 1%. A disadvantage is that additional separation steps for the recovery of carbontetrachloride are required. Although a process based on the principle of solid compound formation with car­ bontetrachloride has been proposed [3] no industrial applications have been described in the literature.

The processes developed and applied on industrial scale for crystallization of paraxylene, can roughly be divided into two classes with respect to the method of heat removal: (i) those with direct contact cooling and (ii) processing applying indirect cooling.

1.4.1 PROCESSES USING DIRECT CONTACT COOLING

Application of direct contact cooling is advantageous because the absence of cold spots prevents incrustration to occur on the walls. Here, the coolant is directly brought into con­ tact with the aromatic mixture. Evaporation of the coolant results in cooling of the C8-aromatic mixture and crystallization of paraxylene. Advantages of direct contact cooling are high heat transfer rates and no crystal damaging due to a scraper mechanism. A disad­ vantage is that the coolant has to be handled very carefully with respect to contamination. Expensive compressors have to be used because the coolant has to remain free from oil. When low pressures are applied, leakage of incondensable gasses may occur.

solid-liquid separator crystals solid-liquid separator product

6 Crystallization ofParaxylene with Scraped Surface Heat Exchangers

The Chevron process [10] is the most widely used direct cooling process and CO2 is used as the direct contact coolant. A high pressure and a low pressure crystallizer are used in series to control gradual evaporation of CO2, thus avoiding high supersaturations. The crys­ tals are separated from the mother liquid by two centrifuge stages. The Maruzen process [11] consists of two stages. The first stage consists of a number of crystallizers in series to cool the mixture gradually to the temperature of the metaxylene - paraxylene eutectic using ethylene as a direct contact coolant. Solid-liquid separation is carried out with centrifuges and the crystal cake is melted and freed from ethylene by stripping. In the second stage, the mixture is cooled with scraped surface heat exchangers. Melting of the second-stage cake that results from the final centrifuge step gives paraxylene with a purity of 99.5%.

A one-stage process using an immiscible direct contact coolant in a column crystallizer has been developed by I.F.P. [12], but this process has not been commercialized. A purity of 99.8% was reported.

The use of CCIF3 as a direct contact coolant has proven to be advantageous in a number of ways [13], but the material is too expensive for use on an industrial scale.

1.4.2 PROCESSES USING INDIRECT COOLING

Indirect cooling implies that the removal of heat takes place through the wall of the crys­ tallizer or heat exchangers. Because of the relatively high undercooling at the walls these tend to incrustrate. Because of the low thermal conductivity of solid paraxylene, the presence of a crystal layer on a heat exchanging wall has a strong negative influence on the heat transfer coefficient. Scraper mechanisms are installed in order to prevent scaling by periodically cleaning the wall, but these scrapers are believed to be responsible for the for­ mation of large amounts of small crystals by breaking larger crystals on the wall.

This class of processes can be divided into two subclasses:

- processes using scraped drums crystallizers. Here the crystallizer consists of a drum with a coolant jacket at the outside. This drum is equipped with a scraper mechanism in order to reduce scaling;

- processes using scraped surface heat exchangers. In this case the crystallizer is an adiabatic stirred drum. The removal of heat is performed by scraped surface heat ex­ changers in an external recycle loop.

The most widely applied processes using scraped drums are the Amoco process [5] and the Arco process [6]. Both processes consist of two stages. In the first stage, ethylene is used as a coolant and the temperature of the mixture is lowered stepwise to nearly the para-xylene/metaxylene eutectic by using two or more crystallizers in series. The cake that results after centrifugation of the first stage slurry is melted to free occluded liquid entities and impurities in the crystals. The resulting liquid mixture is then fed to the second stage crys­ tallizer. Because of the increased paraxylene concentration, the crystallization tempera­ ture is higher than in the first stage. For this reason evaporating propane is used as the second stage coolant. After centrifugation of the second stage slurry, the cake is melted to give a paraxylene product with a purity of 99.1 % in the Amoco process. In the Arco-process, the second stage cake is additionally washed with toluene and the resulting

toluene-Introduction 7 paraxylene mixture is separated by distillation. Due to the toluene washing step, the purity of the product is 99,7%. The same degree of purity turned out to be attainable by washing the first stage cake with toluene [6]. A comparable singe stage process, using a toluene washing step has been patented by Mobil Oil [7]. They claim a purity of 99.5%.

The Krupp-process [8] also consists of two stages. Here, scraped surface heat exchangers are used for the removal of heat and the crystallizer is an adiabatic, stirred drum. The first stage slurry is separated by centrifuges and the cake is washed with second-stage mother liquid. The resulting slurry is then fed to the second stage. The second-stage cake is washed with pure product, resulting in a purity of 99.5%. The Phillips-process [9] also uses two stages with scraped surface heat exchangers, but the second stage slurry is separated in a pulsed wash column to produce paraxylene with a purity of 99.5%. The EXXON-process, being subject to the presented research work, also belongs to the subclass of processes using scraped surface heat exchangers. This process will be treated in more detail in section 1.6.

1.5 PURIFICATION OF THE PRODUCT

When solid-liquid separation is carried out by centrifuges, as in most of the processes described in the foregoing section, the maximum attainable purity is about 99.5%. Much better results are achieved when the product is upgraded by using washing columns. This type of purifiers will be treated briefly in the first part of this section. In the second part, distillation freezing will be presented as an alternative way of upgrading the product. 1.5.1 WASHING COLUMNS

In a washing column, the crystals are washed countercurrently with pure product which gradually becomes more contaminated as it proceeds through the column. Owing to the temperature gradient in axial direction, washing is not the only mechanism that purifies the crystals. Recrystallization occurs in the case that the crystals themselves contain impurities. Then, they can be regarded as solid solutions and the melting point decreases with increas­

ing impurity concentration. At a given temperature, relatively pure crystals will grow but relatively impure crystals will dissolve. Crystals with relatively impure regions will be purified by a mechanism called sweating, because these parts of the crystals will melt while the relatively pure parts will grow. Due to the volume increase upon melting, a part of the enclosed liquid will flow outward through pores and cleaves on microscale and gets into contact with the mother liquid.

An example of a washing column is the one used in the TNO-Thijssen process [14]. A schematic representation of this column is given in figure 1.3.

8 Crystallization ofParaxylene with Scraped Surface Heat Exchangers

si

Figure 1.3 The TNO-Thijssen column.

The feed stream enters the column in the upper part. The mother liquid is rejected via fil­ ters (3) that are mounted in the wall of pipes (2) running in axial direction. The crystal suspension is reslurried by a pure product recycle stream in section 4. In section 5 the crys­ tals are melted to give the final product. This column has proven to be capable of produc­ ing paraxylene with a purity of 99.95%.

1.5.2 DISTILLATIVE FREEZING

Although no applications of distillative freezing for crystallization of paraxylene have been found in the literature, the process developed by Calyxes [ 15] looks very promising because very high purities can be obtained at relatively low energy costs. This process makes use of the fact that at low pressure two-phase equilibria occur where a solid and a vapour phase coexist. So a solid-liquid separation step can be avoided by bringing the mixture into the solid-vapour region. This process is suitable for the separation of a mixture of at least two volatile components. This mixture may be a liquid as well as a crystal slurry. A necessary condition for the application of this process is that the solid phase consists only of a pure component because recrystallization does not occur.

Two situations are distinguished here: (i) the crystallizing component is less volatile than the non-crystallizing component, and (ii) the volatility of the crystallizing component is equal to or higher than the volatility of the non-crystallizing component.

Phase diagrams of both types of systems at a constant pressure higher than the triple point pressure of the crystallizing component are presented in figure 1.4.

Introduction ₉

Figure 1.4. A schematic representation of phase diagrams of binary mixtures with

the crystallizing component (B) being less volatile (a) or more volatile (b) than the non-crystallizing component A.

When the pressure is lowered beneath the triple point pressure of the crystallizing com­ ponent, a three phase equilibrium occurs at a certain temperature where vapour, liquid and solid B coexist. This is illustrated in the graphs presented in figure 1.5.

Cooling of a liquid feed below the triple point temperature and decreasing the pressure results in the formation of a solid and a vapour phase, according to the three-phase equi­

libria V-L-SB in figure 1.5. In the case of the crystallizing component being less volatile, as

represented in figure 1.5a, the phase transformation that occurs upon expansion is L —► V + SB and in the case of the crystallizing component being more volatile: L + SB —► V. Completion of the phase transformation results in dry crystals and a vapour phase contain­ ing all the impurities. Cooling of the vapour phase leads to a liquid in case (a) and to a slur­ ry in case (b).

10 Crystallization ofParaxylene with Scraped Surface Heat Exchangers

x . x .

Figure 1.5. The systems presented in figure 1.4, at a pressure below the triple point pressure of the crystallizing component.

In this process, the phases are all transported through the distillative freezer in the same direction and so the contact pattern is parallel. The phase transformation occurs adiabati-cally, so the three phase mixture cools down upon evaporation of the liquid. The pressure at which the liquid phase has disappeared is referred to as the "drying up pressure" and the corresponding temperature as the "drying up temperature". A block diagram of this process has been presented in figure 1.6:

Feed Distillative freezer Dry s Melter Pro iucr Vapour * ^ olid Condensor Cond( Melter Conce ntrare nsare

Introduction II

The distillative freezing process is rather energy conserving and capable of producing crys­ tals with a very high purity. Test runs with a paraxylene-metaxylene mixture resulted in paraxylene being 99.98% pure. The only drawback is that the impurity concentration in the feed may not be too high (< 10%), because otherwise extremely low drying up pressures and temperatures are required. Recycling of a part of the solid product has proven to be a major improvement of the process [16]. The recycled solid phase serves as a heat carrier in two ways: (i) when these relatively cold crystals are mixed with the feed, they cool down the feed and (ii) the solid phase delivers heat required for evaporation of the liquid in the distillative freezer. Drying up pressure and temperature are significantly higher when solid recycling is being applied.

From the foregoing section it will become clear that direct application of the distillative freezing process to C8-aromatic mixtures occuring in the refinery process is not possible because of the low paraxylene content. However, distillative freezing could be used in com­ bination with a conventional crystallization process. A first stage cake might, after melting, serve as the feed for the distillative freezing process.

1.6 THE PARAXYLENE UNIT USED BY EXXON

This process will be described in more detail because it is subject to the research work described in this thesis. A simplified flowscheme is presented in figure 1.7.

VI

®

first-stage filtrate T = -S7°C. 25% paraxylene second-stage filtrate T = -10°C, 56% paraxylene

first-stage filtrate < & " centrifuges C2H4 Qw= 1550kW paraxylene product

12 Crystallization of Paraxylene with Scraped Surface Heat Exchangers

Crystallization is performed in one stage that consists of two steps: the first step operates at -45°C, the second one at -60°C. Both steps consist of a crystallizer and a recycle loop with scraped surface heat exchangers. The crystallizers operate adiabatically. After the second crystallizer the crystals are separated from the liquid phase and washed in three centrifuge stages. The basic idea behind this stepwise operation is that the crystal cake in one particular stage is washed and reslurried with a liquid mixture that contains more paraxylene than the filtrate that has just been removed in that stage. In this way the over­ all paraxylene content of the slurry is increased stepwise and so is the purity of the result­ ing crystal cake, because the adhering and occluded mother liquid also contains more paraxylene. The cake resulting from the third stage has a purity > 99% and is removed as product after melting. The filtrate from the third stage is recycled to reslurry the cake from the second and first stage centrifuges. Second stage filtrate is mixed up with the feed be­ cause the paraxylene concentration of this stream is higher than the equilibrium composi­ tion at -60°C. This makes recycling useful because there is still paraxylene to recover from this stream. The filtrate from the first stage centrifuges is partly used to cool the feed stream from = -20 to -35 C and is then fed to the isomerization section. The remaining filtrate stream from the first stage, which contains more paraxylene, is recycled to the first crystal­ lizer.

The heat exchanger cascade

The removal of heat is performed by a cascade of 36 scraped surface heat exchangers. The cascade is divided into three parallel benches, which on their turn consist of three parallel sections. Each section contains four scraped surface heat exchangers in series. The distribu­ tion of the recycle flow over the heat exchanger cascade is schematically represented in figure 1.8.

sechon

Figure 1.8 Schematic representation of the configuration of the heat exchanger cascades.

Introduction 13

The scraped surface heat exchangers

A scraped surface heat exchanger consists of two coaxial cilinders. Along the axis of the cilinders a shaft is mounted with scraper blades; this shaft rotates with a fixed frequency in order to clean the wall periodically. The coolant flows through the annular space between the cylinders. A schematic cross section with the relevant dimensions is presented in figure 1.9.

Figure 1.9 A schematic cross section of a scraped surface heat exchanger.

The length of the coolant jacket is 12 m. These heat exchangers are designed for a through­ put of 500 gallons/min (= 114 m /h), but because a crystal layer is being formed at the wall under operation conditions the pressure drop increases and the flow reduces to 300 gal­ lons/min.

The scraper blades are spring loaded, which means that they are pressed against the heat exchanger wall by springs in normal direction. The scraper blades are made of teflon. This construction is shown in figure 1.10.

14 Crystallization of Paraxylene with Scraped Surface Heal Exchangers

scraper blade

«rape-shaft

scraper blade

Figure 1.10 The blade construction.

The length of the blades is 50 cm and they are mounted in pairs on opposite sides of the shaft. Each pair of blades is orthogonal to the next.. This is illustrated in figure 1.11.

Figure 1.11 The configuration of the scraper blades.

The scaper mechanism has proven to be incapable of preventing scaling completely. There­ fore, parts of the cascade have to be taken out of function regularly to have the crystal layer melted up.

1.7

FORMULATION OF THE PROBLEM

The presence of impurities that can take part in the reaction sequence by which poly-ethyleneterephtalate is produced has a strong negative impact on the quality of the final product. The most likely impurities in paraxylene are other C8-aromatics and small amounts of toluene and C9-aromatics. The acids formed from these impurities during the oxidation of paraxylene all form monomers with ethyleneglycol resulting in an inferior product when these "impurity monomer" molecules are incorporated in the polymer molecule. Toluene and ethylbenzene only have one alkyl group, so the resulting esters will

Introduction IS cause termination of the polymerization reaction. Ortho- and metaxylene have two func­ tional groups, but the angle between those groups is 60 and 120° respectively. Incorpora­ tion of a monomer that originates from those components would introduce an angle in the polymer molecule.

It is therefore evident that for the production of polyethyleneteraphtalate very high demands are put upon the purity of the paraxylene to be used. When paraxylene is being produced by crystallization impurities can be incorporated into the final product because they are by their nature attracted to the crystal surface and become enclosed in the crystal matrix. Further contamination may occur because of incomplete solid-liquid separation. The adhering mother liquid contains all the impurities and thus decreases the purity of the final product. This defines the aim of the present investigation; all mechanisms that may contribute to impurification of the crystals have to be investigated. Their contribution has to be minimized in order to increase the purity of the paraxylene produced.

1.8 SCOPE OF INVESTIGATIONS

Impurities can get into the final product because they are enclosed in the crystal matrix or by the adhering mother liquid that remains after separation of the crystals from the liquid phase.

Impurities can be incorporated in the crystal because of thermodynamic and kinetic reasons:

1. Formation of solid solutions or mixed crystals.

In this case a solid phase containing paraxylene as well as impurities is thermodynamical-ly more stable than pure solid paraxylene. If this phenomenon occurs, the onthermodynamical-ly way of preventing the crystals from being contaminated is to remove these impurities prior to crys­ tallization of paraxylene. If the formation of a solid solution is inevitable, the crystals have to be purified by fractional crystallization.

2. Kinetic incorporation of impurities.

Impurities that are adsorbed on the surface of a growing crystal may be "trapped" in the crystal. The magnitude of this effect depends on the growth rate of the crystal. At first sight impurification increases with increasing growth rate, but the reverse effect may also occur, depending on the affinity of the impurity towards the crystal surface and on its adsorption rate.

Kinetic impurification may also occur due to entrapment of liquid entities into the crystal. This phenomenon takes place when crystals are exposed to very high supersaturations; the large concentration gradients found in the scraped surface heat exchangers often lead to growth instabilities like dendrite formation which may result in occlusion of portions of the liquid phase into the crystal. This phenomenon is called constitutional undercooling. The amount of impurities originating from mother liquid adhering to the crystal surface depends on the crystal size distribution. The specific surface area of the crystals is inverse­ ly proportional to their diameter and so is the amount of mother liquid attached to the crys­ tals. When solid-liquid separation is carried out by centrifuges followed by a simple washing

16 Crystallization ofParaxylene with Scraped Surface Heat Exchangers

step to remove the attached mother liquid, large crystals are preferred. The relative amount of mother liquid that remains decreases with increasing crystal size and so does the amount of product recycle to wash the crystals. On the other hand, when a washing column is ap­ plied in which sweating and recrystallization occurs, small crystals are preferred because a large specific area for mass transport is required. This already explains why centrifuges as a final step are not quite adequate owing to conflicting requirements. An optimization is therefore needed and a proper control of the crystal size distribution helps to reach a max­ imum purity of the final product.

The maximum solid phase recovery of a crystallization process is determined by ther­ modynamics that relate solubility with temperature. The size distribution of the crystalline mass, however, is determined by a competition between nucleation and crystal growth. When the nucleation rate is high, many crystals will be formed, but because of the limited amount of solid phase to produce, they will remain small. When relatively few crystals are formed they can grow larger because each crystal has more supersaturation to consume. Because crystals can only grow as long as they are exposed to supersaturation, the shape of the size distribution is determined by the residence time distribution in the crystallizer. This residence time distribution is a matter of hydrodynamics.

1.9 SET-UP OF INVESTIGATIONS AND OF THIS THESIS

Both mechanisms that might lead to incorporation of impurities in the crystal have been investigated. For design purposes phase behaviour is normally described assuming ideal behaviour in the liquid phase and the formation of pure solid phases. In the literature no evidence has been found against these assumptions. Practical experiences, however, have shown that a small increase in the toluene concentration in the feed to the crystallizer has a strong negative influence on the quality of the paraxylene produced [17]. Furthermore unpublished experimental data [18] suggested the formation of solid solution in binary mix­ tures of xylenes.

For the reason mentioned above, determination of solid-liquid equilibria in C8-aromatic mixtures and toluene have been a subject of this investigation. The experimental part con­ sisted of (i) analysis by gas chromatography of both phases that are supposed to be in equi­ librium over a certain temperature range, and (ii) a calorimetric (D.S.C.) method has been used. This subject has been described in chapter 2 of this thesis.

Knowledge of the crystal structure of paraxylene is required to understand the morphol­ ogy of the crystals. Introductory studies of the crystal shape revealed that the observed mor­ phology does not match with the crystallographic data found in the literature. Therefore, determination of the structure has been performed using X-ray diffraction.

The influence of different solvents on the morphology and the growth rate of the crystals and thus on their shape has been investigated on laboratory scale. Paraxylene crystals were grown from mixtures containing 70 mol% paraxylene and 30% of a second component being benzene, toluene, orthoxylene, metaxylene or ethylbenzene. The determination of the crystal structure and growth phenomena have been reported in chapter 3 of this thesis.

Introduction 17

The supersaturation at the heat exchanger walls is directly related to the imposed tempera­ ture gradients and to the hydrodynamics in the boundary layer. Heat transfer phenomena in the scraped surface heat exchangers are therefore playing an important role in the result­ ing crystal size distribution. Removal of heat requires a temperature gradient across the laminar part of the flow, so the wall temperature of the heat exchanger is always lower than the bulk temperature. The degree of undercooling increases with decreasing distance from the wall, so nucleation is likely to occur at or near the wall. Because nucleation rate strong­ ly depends on the supersaturation, the wall temperature is believed to have a major in­ fluence on the amount of nuclei produced in the scraped surface heat exchangers. For a prescribed heat flux and bulk temperature, the wall temperature can only be controlled by hydrodynamic variables: a change in heat transfer coefficient results in a change in tempera­ ture difference between the bulk and the wall, because these parameters are inversely proportional to each other according to:

Qw = a ( T B - Tw). (1.7)

The influences of flow rate and scraper rotational speed on the resistance against heat trans­ fer by the hydrodynamic boundary layer have been investigated experimentally on a bench-scale model of a scraped surface heat exchanger under non-crystallizing circumstances. An attempt has been made to derive a mathematical model that takes both effects into account. This subject has been described in chapter 4 of this thesis.

Another problem encountered in coolcrystallization is the formation of a crystal layer on the heat exchanger surfaces and its influence on heat transfer properties. The scrapers have proven to be incapable of preventing the formation of such a layer, often called scaling, and the presence of this crystal layer on the wall has a dramatic influence on the overall heat transfer coefficient because of the low thermal conductivity of solid paraxylene.

Chapter 5 deals with heat transfer measurements on pilot plant scale in the absence and presence of a crystal layer. The thickness of the crystal layer has been determined from heat transfer coefficients at the inner tube side that consist of two resistances: a resistance due to the thermal boundary layer and a resistance due to the crystal layer. The resistance due to the thermal boundary layer has been measured under non-crystallizing conditions which makes calculation of the resistance of the layer, and so its thickness, possible from the measured value of the heat transfer coefficient under crystallizing conditions. The thick­ ness of the crystal layer has been determined as a function of flow rate, rotational speed and wall temperature.

In chapter 6, the significance of the results of the investigations described in the foregoing chapters for the EXXON-process will be outlined. Crystal size distributions obtained from measurements at the site are the basic information for the interpretation of the results from pilot-plant and bench scale experiments. Furthermore, recommendations for better pro­ cessing and suggestions for further investigations are given.

18 Crystallization of Paraxylene with Scraped Surface Heat Exchangers

1.10 REFERENCES

[1] Kirk - Othmer, Encyclopedia of Chemical Technology. John Wiley and Sons, 3rd. ed. New York 1984.

[2] Thornton, D.P., Hydrocarbon Proc. 49 (1970) pp. 151-155.

[3] Egan, C.J., R.V. Luthy, Ind. Eng. Chem. 47 (1955) pp. 250-253.

[4] Garazi, A.M., P.M. Haure, D.G. Löffler, Ind. Eng. Chem. 23 (1984) pp. 849-851.

[5] U.S. Pat. 3.177.265., (6-4-'65), G.C. Lammers, Standard Oil Company, Indiana.

[6] R.J. Desiderio et al., Hydrocarbon Processing, Aug. 74, pp. 81-83.

[7] Ned. Pat. 7.101.250. (5-8-71), G.H. Boelsma, Mobil Oil Cooperation, New York.

[8] H. Ritzer, Erdöl und Kohle, 26 (1973) 327-331.

[9] McKay, D.L., G.H. Dale, D.G. Tabler, Chem. Eng. Progr. 62 (1966) 104-112.

[10] U.S. Pat. 3.467.724. (16-9-'69), S.A. Laurich, Chevron Research Company.

[11] Hatanaka, Y., T. Nakamura, Oil GasJ. 70 (47) 60 (1972).

[12] R. Lafay, Chimie et Industrie - Genie Chimique 99 (11), 1968, pp. 1555-1559.

[13] Duncan, A.G., R.H. Phillips, Trans. Instn. Chem. Engrs. 54 (1976) pp. 153-159.

[14] Arkenbout, G.J., A. van Kuyk, L.H.J.M. Schneiders in: Industrial Crystallization 1984, ed. S.J. Jancic and E.J. de Jong. Elseviers Science Publishers B.V., Amsterdam 1984.

[15] U.S. Pat 4.433.588. (28-2-'84), Cheng et al.

[16] U.S. Pat 4.578.093. (25-5-'86), Cheng et al.

[17] Private communincation with Esso Nederland B.V.

Chapter 2

SOLID-LIQUID EQUILIBRIA IN MIXTURES OF

C8-AROMATICS AND TOLUENE AT ATMOSPHERIC

PRESSURE

2.1. INTRODUCTION

A typical C8-aromatic mixture that is fed to the paraxylene separation section consists of 50% metaxylene, 15% ethylbenzene, 11% orthoxylene, 23% paraxylene and 1% toluene. For development and design purposes, it is generally assumed that the C8-aromatics be­ have like an ideal mixture and that the solid phases consist only of pure components. The limitations of the process as they are imposed by phase equilibria are shown in a calculated ternary (o-, m- and p-xylene) temperature projection diagram. As far as the crystallization of p-xylene is concerned, non-xylene compounds (ethylbenzene and toluene) are con­ sidered to behave similarly to ortho and metaxylene [1]. Ideality of the liquid phase jus­ tifies for this assumption because in an ideal mixture the crystallization temperature is only a function of the concentration of the crystallizing component and does not depend on the nature of the other components present. A phase diagram, based on the assumptions men­ tioned above is presented in figure 2.1.

Figure 2.1. Temperature projection diagram of the ternary xylene mixture assum­

20 Crystallization of Paraxylene with Scraped Surface Heat Exchangers

At the corners, the melting points of the pure components are given. On the edges, eutec­ tic data of the binary mixtures are presented. From these eutectic points, three-phase lines (a liquid and two solid phases) of the ternary system connect points of lower temperatures at increasing concentrations of the third components. These three-phase lines intersect at the ternary eutectic point which is nonvariant. The dashed lines are isotherms, which rep­ resent the boundaries of two-phase solid-liquid regions. On one isotherm, a series of liq­ uid compositions is laying which are, at the corresponding temperature, in equilibrium with the solid compound at the corner. The solid line through the p-xylene corner and the point indicated as "feed" represents the mass balance of p-xylene during the crystallization process. While crystallizing p-xylene, the composition of the liquid phase shifts along this line, from the feed point to the metaxylene-paraxylene eutectic line.

2.2. OBJECTIVE AND SCOPE OF THE INVESTIGATION

Haddon and Johnson [2] and Porter and Johnson [3] have reported solubility data of p-xylene in C8-aromatic mixtures. Significant deviations from ideal behaviour were observed in the temperature range of technical importance (-40 to -65°C). The explanation for these deviations were that the assumptions of an ideal liquid phase and pure solid phases are not correct. Experimental data of heats of mixing in the liquid phase have been reported in the literature [4-6] and the contribution of these effects to the deviation from ideal behaviour will be estimated in section 2.4. Furthermore, the assumption of pure solid p-xylene might not be valid. So far, no experimental proof of the purity of the solid phases has been reported in the literature. Moreover, unpublished results from measurements of solid-liq­ uid equilibria of para-metaxylene and para-ortho-xylene binary mixtures performed else­ where in charge of EXXON indicate strong miscibility in the solid phases. Furthermore, relatively much toluene (= 0.1%) is found in the paraxylene crystals in spite of the low con­ centration (« 1%) in the feed. The amount of toluene in the solid phase seems to depend strongly on the amount in the liquid phase. Because the formation of solid solutions would have a negative influence on the purity of the product and hence requires additional purification of the p-xylene produced by crystallization, this study was initiated to reveal whether miscibility in the solid phases may occur and, if so, to what extent.

2 3 . SET-UP OF THE INVESTIGATIONS

The experiments described in this chapter are divided into three groups:

(i) First, samples of the solid phase were taken at the plant and analyzed by gas-liquid chromatography in order to get an impression of the relative contribution of the other components than paraxylene to the impurification of the crystals. Because these ex­ periments don't provide information about the question whether impurification oc­ curs because of thermodynamic reasons or due to kinetic events, additional expe­ riments, under better defined conditions, are required.

(ii) In order to eliminate kinetic effects, laboratory scale experiments were performed by slowly growing paraxylene crystals out of mixtures with different compositions. After equilibration, both phases were analyzed with gas-liquid chromatography. This me­ thod is disadvantageous because of being time-consuming. Besides, the validity of the results is questionable, because the time required for equilibration is unknown and

Solid-Liquid Equilibria in Mixtures of 21 C8-Aromatks and Toluene at Atmospheric Pressure

the accuracy of the analysis of the solid phase suffers from the presence of liquid at­ tached to the solid phase, due to incomplete solid-liquid separation.

(iii) In order to overcome the difficulties as mentioned in item (ii) and to be able to deter­ mine the effect of a controlled cooling rate, additional experiments were carried out using D.S.C. ( = Differential Scanning Calorimetry) [7]. This technique is relatively fast, no analysis of the phases is required and cooling- and heating rates can be programmed. The principles of this technique are described in section 2.5.3. Because D.S.C.-curves of multicomponent systems are difficult to interpret, binary mixtures of ortho- and paraxylene and meta- and paraxylene were chosen as indicative systems. Measurements on binary mixtures of p-xylene with ethylbenzene and toluene were not possible because such measurements require cooling at -150°C, which is not feasible with the D.S.C.-equipment described in this chapter.

2.4. THEORETICAL DESCRIPTION OF SOLID-LIQUID EQUILIBRIA

2.4.1 GENERAL APPROACH

When a solid and a liquid phase are in equilibrium the chemical potential of each com­ ponent i in the liquid phase has to be equal to that in the solid phase according to:

y.i,L = u.i,S (2.1)

In equation 2.1 the subscripts L and S denote the liquid and solid phase. In principle, the chemical potential is a function of temperature, pressure and composition. Here, only solid-liquid equilibria at atmospheric pressure are considered. Therefore, the chemical poten­ tial may be expressed as a function of temperature and composition only:

Wj = Nj*(T) +RTlnaij (2.2)

In equation 2.2 jiij*(T) represents the chemical potential of the pure component in state j and ai j is the activity of component i in phase j under the given circumstances. Combina­ tion of equations 2.1 and 2.2 yields:

Hi,L*(T) + RT In ailL = ni,S*(T) + RT In ai.s (2.3)

or:

ai.L

RT In -^- = - Ap.*s-»L(T) (2.4)

where Ap.*s-*L(T) represents the difference in chemical potential between pure solid and pure liquid component i at temperature T. For the right-hand side of equation 2.4 we may write:

22 Crystallization ofParaxylene with Scraped Surface Heat Exchangers

In equation 2.5, Tref is an arbitrary reference temperature. When the melting temperature Tm is chosen to be the reference temperature, then ALL'S-»L(Tref) = 0. The term under the

integral sign can be rewritten to:

a(An*s-L(T))

= - ASS-»L(T) (2.6) d i

where ASS— L ( T ) is the entropy change per mole for the phase transition at temperature T.

Combination of equations 2.4,2.5 and 2.6 results in:

T

R T l n — = fASs-L(T)dT (2.7)

a'.s _J

From the equality of the chemical potentials of the liquid and solid phase at the melting temperature it follows that:

ASs-»L(Tm) = (2.8)

Tm

The entropy difference ASs— L(T) can be expressed as:

r 3 ( A S S - L ( T ) ) A S ^ L ( T ) = ASs-.L(Trcf) + f g g f g ^ d T (2.9) With: Tref aASs^L(T) ACP,S->L dT T

with ACP,S-»L being the difference in heat capacities of the pure solid and liquid phase at

temperature T.

Equations 2.8, 2.9 and 2.10 may be combined to yield:

4 f c

,

* i ! I = »

f

^ CUD