UNIVERSITE DE MONTREAL

METHANOL-TO-OLEFINS IN A FLUIDIZED BED:

EXPERIMENTAL AND MODELLING STUDY

MADS KAARSHOLM

DEPARTEMENT DE GENIE CHIMIQUE

ECOLE POLYTECHNIQUE DE MONTREAL

THESE PRESENTEE EN VUE DE L'OBTENTION

DU DIPLOME DE PHILOSOPHIAE DOCTOR

(GENIE CHIMIQUE)

MARS 2009

© Mads Kaarsholm, 2009.

1*1 Library and
Archives Canada
Bibliotheque et
Archives Canada

Published Heritage Direction du

Branch Patrimoine de I'edition
395 Wellington Street 395, rue Wellington
Ottawa ON K1A0N4 Ottawa ON K1A0N4
Canada Canada

Your file Votre reference

ISBN: 978-0-494-49419-6
Our file Notre reference
ISBN: 978-0-494-49419-6

NOTICE: AVIS:
The author has granted a non- L'auteur a accorde une licence non exclusive
exclusive license allowing Library permettant a la Bibliotheque et Archives
and Archives Canada to reproduce, Canada de reproduire, publier, archiver,
publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public
communicate to the public by par telecommunication ou par Plntemet, prefer,
telecommunication or on the Internet, distribuer et vendre des theses partout dans
loan, distribute and sell theses le monde, a des fins commerciales ou autres,
worldwide, for commercial or non- sur support microforme, papier, electronique
commercial purposes, in microform, et/ou autres formats.
paper, electronic and/or any other
formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur

ownership and moral rights in et des droits moraux qui protege cette these.
this thesis. Neither the thesis Ni la these ni des extraits substantiels de
nor substantial extracts from it celle-ci ne doivent etre imprimes ou autrement
may be printed or otherwise reproduits sans son autorisation.
reproduced without the author's
permission.

In compliance with the Canadian Conformement a la loi canadienne

Privacy Act some supporting sur la protection de la vie privee,
forms may have been removed quelques formulaires secondaires
from this thesis. ont ete enleves de cette these.

While these forms may be included Bien que ces formulaires

in the document page count, aient inclus dans la pagination,
their removal does not represent il n'y aura aucun contenu manquant.
any loss of content from the
thesis.

Canada
 UNIVERSITE DE MONTREAL

ECOLE POLYTECHNIOUE DE MONTREAL

Cette these intitulee:

METHANOL-TO-OLEFINS IN A FLUIDIZED BED:

EXPERIMENTAL AND MODELLING STUDY

presentee par: KAARSHOLM Mads

en vue l'obtention du diplome de: Philosophiae Doctor
a ete dument acceptee par le jury d'examen constitue:

M. LEGROS Robert, Ph.D., president

M. PATIENCE Gregory S„ Ph.D., membre et directeur de recherche
M. CHAOUKI Jamal, Ph.D., membre et codirecteur de recherche
M. FRADETTE Louis, Ph.D., membre
M. LATTNER James R, Ph.D., membre
 IV

ACKNOWLEDGEMENTS

I would like to express my sincere gratitude to the people that have helped and guided

me in this work and without whom it would not have been possible. First of all I would

like to thank my supervisors at Ecole Polytechnique de Montreal Professor Gregory S.

Patience and Professor Jamal Chaouki for their guidance, encouragement and fruitful

discussion both while I was in Montreal and during my time at Haldor Topsoe in

Denmark. I would also like to thank my supervisors at Haldor Topsoe Roberta Cenni,

Finn Joensen and Jesper Nerlov for guidance and inspiration.

My thanks go to my friends and family - especially to Kamilla Hansen for support and

understanding during the last four years. I would also like to thank my fellow Ph.D

students at Ecole Polytechnique de Montreal AH Shekari and Jean-Philippe Laviolette

for their friendship, help and interesting discussion during my time in Montreal. My

thanks also go to Banafsheh Rafii and Julia Duchastel-Legare for their valuable help in

the laboratories.

I would also like to thank the technical staff both at Ecole Polytechnique de Montreal

and at Haldor Topsoe. A special thank go to Daniel Dumas, Jean Huard and Robert

Delisle for support to the experimental setups. I am also thankful to Kim Ravn and
 V

Svend-Erik Mikkelsen for their help with the fixed bed experiments at Haldor Topsoe

and to Peter Rahbek for his help with spray drying of the catalyst

Finally I would like to thank Haldor TopS0e A/S and the Danish Ministry of Science

Technology and Innovation for financial support making this work possible.
 VI

RESUME

Dans cette these, la transformation du methanol en defines (MTO) a ete etudiee

principalement dans un lit fluidise et aussi dans un lit fixe. Cette reaction peut se diviser

en 2 parties : la deshydratation du methanol en ether dimethylique (DME) et la

formation d'hydrocarbures a partir du melange methanol/DME. Dans les conditions

habituellement utilisees pour le procede MTO - approximativement entre 450 et 500°C,

la premiere reaction est tres rapide et generalement considered a l'equilibre en tout

temps. A des temperatures plus basses, typiquement inferieures a 300°C, la formation

d'hydrocarbures est pratiquement nulle et la reaction de deshydratation du methanol en

DME peut ne plus etre a l'equilibre. C'est le cas de plusieurs procedes industriels ou la

deshydratation du methanol se fait dans un reacteur concu a cette fin, pour limiter

1'augmentation de la temperature inherente au procede MTO. Le catalyseur utilise dans

cette etude est base sur de la zeolite ZSM-5 et modifie avec du phosphore pour reduire la

formation de methane. Le travail realise a ete divise en trois etapes : d'abord, la

deshydratation du methanol en DME a ete etudiee dans un lit fluidise; ensuite, la

reaction MTO a ete analysee dans un lit fixe ; enfin, des etudes cinetiques ont ete

menees dans un lit fluidise et un modele representant 1'ensemble de ces reactions

complexes a ete propose.

La transformation du methanol en DME a ete etudiee dans un reacteur a lit fluidise en

quartz, de 46 mm de diametre, pour examiner la cinetique de la reaction et les effets du

transfert de matiere sur celle-ci. L'analyse des produits a ete effectuee par un
 vii

spectrometre de masse (SM) et en partie par chromatographic en phase gazeuse (CG).

Le SM fournit des mesures en ligne et en temps reel, ce qui permet de detecter

facilement les variations des conditions de reaction et de determiner les regimes

permanents. Les experiences ont ete realisees a partir du methanol dime dans de l'argon,

a quatre concentrations distinctes : 5, 15, 30 et 33 mol% et a trois

temperatures differentes : 250, 275 et 325°C. A 325°C, - dans certains cas et

dependamment du rapport debit massique et masse de catalyseur (WHSV) - la reaction

de deshydratation du methanol se poursuit au-dela de la formation du DME et produit

d'autres hydrocarbures. Ces experiences-la ont ete rejetees. Lorences et al., 2006 ont

montre que Phydrodynamique des lits fluidises peut etre assimilee a celle des reacteurs

parfaitement melanges en serie (CSTR) quand la vitesse superficielle de gaz est faible et

que la dispersion croit avec l'augmentation de la vitesse. Les mesures de Distribution du

Temps de Sejour (DTS), realisees dans le cadre de ce travail, ont permis de caracteriser

le lit fiuidise comme etant equivalent a 6 reacteurs CSTR, lorsque la vitesse superficielle

est egale a 10 fois la vitesse minimale de fluidisation - umf. Afin d'obtenir des WHSV

relativement elevees, meme a faibles vitesses superficielles, nous avons varie la masse

de catalyseur dans le reacteur de 25 a 200 g. Les conversions experimentales ont ete de

l'ordre de 30 a 99+% par rapport a l'equilibre.

Pour l'analyse cinetique, deux reacteurs modeles differents ont ete utilises dans le but de

modeliser l'hydrodynamique du lit fiuidise : un modele a deux phases - phase emulsion

et phase bulles - et une serie de 6 reacteurs parfaitement melanges. Un modele cinetique

 viii

de la deshydratation du methanol en DME a ete propose et compare a celui de Bercic et

Levee (1992).

Plusieurs series de parametres cinetiques ont ete trouvees en couplant les deux modeles

hydrodynamiques avec les deux modeles cinetiques et par optimisation des donnees

experimentales aux trois plus faibles concentrations de methanol.

L'analyse des resultats a montre que le modele cinetique propose avec le modele des n-

CSTR en serie permet d'obtenir les predictions les plus proches des donnees

experimentales et des predictions excellentes pour les plus grandes concentrations de

methanol. Le modele hydrodynamique n-reacteurs parfaitement melanges, couple au

modele cinetique propose ou a celui emprunte dans la litterature, donne les meilleurs

resultats, quoique le modele du lit fluidise n'ait pas ete rejete.

Cette etude a montre qu'une experience cinetique dans un lit fluidise a petite echelle a

des avantages pour les reactions exothermiques, car elle elimine les incertitudes liees

aux points chauds et aux gradients de temperature.

Pour utiliser les n-CSTR en serie comme reacteur modele, il faut maintenir une vitesse

superficielle de gaz relativement faible. Par contre, a cause des contraintes sur les

vitesses superficielles et la masse de catalyseur, les conversions a des temperatures

donnees peuvent etre limitees.

Le procede MTO a ete etudie dans plusieurs lits fixes de faible diametre. Les

mecanismes de la reaction principale et de la deactivation ont ete analyses pour

differentes WHSV, compositions a l'entree et longueurs de reacteur (3, 6 et 8 mm).

 ix

La chromatographic en phase gazeuse a ete utilisee pour determiner la distribution des

produits de la reaction. Les composes a faibles nombres de carbone ont ete analyses

individuellement, alors que les fractions de C4 aromatiques et de C5+ non aromatiques

ont ete regroupees. Ces experiences suggerent que les C5+ non aromatiques sont des

produits intermediaries de la reaction MTO, conduisant a la formation de propylene et

de butylene (et non d'ethylene) par des reactions de craquage supplementaires. La

production d'ethylene se fait, quant a elle, par division des aromatiques (principalement

des xylenes et des trimethylbenzenes), ce qui concorde avec les travaux de Svelle et al.

(2006). Le methane, qui est predominant a de faibles conversions du methanol/DME et

a des taux eleves de WHSV, doit probablement se former directement a partir du

methanol ou du DME et non des hydrocarbures superieurs.

La deactivation du catalyseur est tres sensible a la WHSV et croit avec celle-ci. Par

consequent, la capacite du reacteur a transformer le methanol depend du flux a l'entree.

La composition de ce flux a egalement une influence significative sur la deactivation du

catalyseur et la composition des produits. Des pressions partielles basses de methanol

favorisent la production d'olefines, mais elles reduisent la capacite du reacteur a

transformer le methanol. Ceci contredit les travaux de Chen et al. (2000) montrant que la

deactivation du catalyseur est fonction de la quantite de methanol converti et de la

temperature et non de WHSV.

La troisieme partie de la these traite de la reaction MTO dans un lit fluidise. Les effets

de la composition de 1'alimentation, de la temperature et de la WHSV ont ete analyses

dans le meme reacteur que celui de la deshydratation du methanol. En alimentant en 1-

 X

hexene d'une part et en methanol d'autre part, nous avons propose un mecanisme

probable de reaction. La composition des produits de reaction ne change pas de facon

significative avec une alimentation en methanol. Dans le cas de 1'alimentation en

hexene, les differences majeures sont une production significativement plus basse de

methane et une plus grande formation d'ethylene. Ces observations confirment les

resultats obtenus dans un lit fixe, selon lesquels le methane se forme directement a

partir du methanol et/ou du DME, et l'ethylene est produit par craquage secondaire.

Un modele cinetique a ete propose en se basant sur le mecanisme «lot

d'hydrocarbures », ou les olefines sont produites par des reactions reversibles a partir

d'un lot d'hydrocarbures superieurs.

Tel qu'observe dans les experiences anterieures realisees dans des lits fluidises et fixes,

le methane est suppose se former a partir du methanol. Les aromatiques sont considerees

comme etant C9H12 et les paraffines, tout comme le propane, sont produits directement a

partir du propylene.

Pour la reaction MTO en lit fluidise, le modele hydrodynamique a deux phases a ete

choisi. En effet, comme la hauteur du lit etait generalement beaucoup plus elevee que

dans les experiences de deshydratation du methanol, la taille des bulles etait plus grande.

De plus, ce modele a deux phases n'avait pas ete rejete lors de l'etude de deshydratation

du methanol.
 xi

Le modele predit bien la production d'olefines legeres, mais moins bien celle des

paraffines et des fractions C6., ceci s'expliquant par le fait que les aromatiques, les

paraffines et les olefmes sont agglomeres et qu'ils presentent des mecanismes de

reaction differents.
 xii

ABSTRACT

In this thesis the Methanol To Olefin (MTO) reaction has been studied in fixed bed and

fluid bed with emphasis on the fluid bed. The reaction from methanol to olefins can be

divided into two parts: dehydration of methanol to dimethyl ether (DME) and reaction of

methanol/DME mixture to hydrocarbons, including paraffins, olefins and aromatics.

Under reaction conditions for the MTO process app. 450-500°C the methanol to DME

reaction is very fast and it is usually considered to be in equilibrium at any time. At

lower temperatures, below about 300°C, hydrocarbon formation is essentially absent and

the methanol/DME reaction can not be considered in equilibrium. This is the case for

Lurgi methanol to propylene (MTP) process where the dehydration of methanol takes

place in a separate reactor to limit the temperature increase during the MTO reaction.

The MTO reaction is catalysed by an acidic catalyst and in this study the used catalyst is

based on a ZSM-5 zeolite and was modified with phosphorous to reduce methane

formation. The catalyst used in this study was based on a ZSM-5 zeolite and was

modified with phosphorous to reduce methane formation. The work has been divided

into three parts: first the methanol to DME reaction is studied in fluidized bed. Second

the MTO process is investigated in fixed bed and last kinetic studies of the MTO

reaction is conducted in fluidized bed and a model for the complex reaction is proposed.
 xiii

The methanol to DME reaction has been studied in a 4.6 mm inner diameter quartz

fluidized bed reactor to investigate the reaction kinetics and effect of mass transfer on

the reaction. Analysis of the products was conducted by mass spectrometer (MS) and in

part by gas chromatograph (GC). The MS provides real time measurements and changes

in reaction conditions are therefore easy to detect and steady state conditions can be

determined. The experiments were performed with methanol diluted in argon at four

different concentrations (5, 15, 30 and 33 mol%) and at three temperatures (250, 275 and

325°C). At 325°C the reactions in some cases, depending on the weight hourly space

velocity, WHSV, proceeded beyond DME formation, forming hydrocarbons and these

results were omitted. Lorences et al., 2006 have shown that the hydrodynamics of the

fluid bed can be approximated by CSTR's in series at low gas velocities and with

increasing gas velocity the dispersion increases. Residence time distribution (RTD)

measurements showed that the fluid bed could be modelled as 6 CSTR's in series at a

flow of approximately 10 times umf. To get at larger WHSV range while still keeping the

gas velocity low, the catalyst inventory in the fluid bed was varied in the range 25 -

200g. Conversion to equilibrium in the experiments was in the range 30 - 99+%.

Two different reactor models have been used for the fluid bed during the kinetic

analysis. A two phase fluid bed model with interphase mass transfer and a n-CSTR in

series model. A kinetic model for the methanol to DME reaction have been proposed

and compared to the literature model of Bercic and Levee, 1992.

The kinetic parameters were found by coupling the two kinetic models with the two

reactor models and optimizing them to the experimental data at the three lowest
 xiv

methanol concentrations. The analysis showed that the proposed kinetics with the n-

CSTR's in series model gave the best fit to the experimental data and that the predictions

of the data at the highest methanol concentrations was excellent. The n-CSTR reactor

model gave the best result for both the proposed model and the literature model although

the fluid bed model was not significantly worse.

The study showed that kinetic experiment in a small scale fluid bed has advantages for

exothermic reactions since uncertainties regarding hotspots or temperature gradients can

be eliminated. To be able to use the n-CSTR's in series reactor model the gas velocities

have to be kept relatively low. Given the restraints on gas velocities and catalyst

inventory the conversion range at a given temperature can be limited.

The methanol to olefin reaction has been studied in small diameter fixed bed reactors

with different diameters. By changing the WHSV, feed composition and reactor size (3,

6 and 8 mm) the reaction pathways and deactivation patterns were studied. GC

measurement was used to determine the product distribution. The small components was

analysed separately while the C4, aromatics and C5+ (without aromatics) fraction are

lumped. These studies suggest that non-aromatic C5+ hydrocarbons are intermediates in

the MTO reaction, forming propylene and butylene (but not ethylene) by secondary

cracking reactions. Ethylene production paralleled the aromatics formation suggesting

that ethylene are formed by splitting off from aromatics (mainly xylenes and

trimethylbenzenes) which is in accordance with labelling studies of Svelle et al., 2006.

Methane predominated the products at low oxygenate conversion and high WHSV
 XV

which suggest that methane is formed from methanol and/or DME and not from

reactions by higher hydrocarbons.

Deactivation of the catalyst was highly dependent on the space velocity with increasing

deactivation as the WHSV was increased. As a result the methanol capacity of the

catalyst was dependent on the feed rate. The feed composition also has a high influence

on the deactivation and product distribution. Low partial pressure of methanol favours

olefin production but at the same time decreases the methanol capacity. This contradicts

the findings of Chen et al., 2000 who found that deactivation was based on methanol

converted and temperature and not WHSV.

The third part on the thesis treats the MTO reaction in fluidized bed. The effect of feed

composition, temperature and WHSV was investigated in the same reactor setup in

which the methanol dehydration reaction was studied. Feeding 1-hexene in argon and

co-feeding 1-hexene with methanol in argon was used to investigate the reaction

mechanism. The product distribution obtained by feeding 1-hexene did not change

significantly from a methanol feed. The main difference being the methane production

which was considerably lower when feeding 1-hexene confirming the observations in

fixed bed, that methane originates directly from methanol and/or DME. Ethylene was

also found to be a major part of the product distribution from 1-hexene suggesting that

ethylene is a part of the secondary cracking reactions.

A kinetic model has been proposed which is based on the hydrocarbon pool mechanism

where the olefins are produced through reversible reactions with a larger hydrocarbon
 xvi

species. Methane is considered to be formed from DME in accordance to the findings

from fixed and fluidized bed. Aromatics are considered as C9H12 and paraffins as

propane both are formed directly from propylene.

For the MTO reaction in a fluidized bed, the two phase model was chosen as the

hydrodynamic model. This was because the bed height was generally higher than in the

experiments with methanol dehydration resulting in larger bubbles and since the gas

velocity spanned a larger range. Further, the two phase model did not perform

significantly worse than the n-CSTR model for the methanol dehydration. The model

predicts the light olefins fractions well but some problems were observed with the

paraffin and C6+ fractions. The latter because aromatic, paraffins and olefins are lumped

together which have different reaction patterns.

 XV11

CONDENSE EN FRAN^AIS
Dans cette these, la reaction transformant le methanol en olefines (MTO) a ete etudiee

principalement dans un lit fluidise et aussi dans un lit fixe. Cette reaction peut etre

divisee en deux sous-ensembles de reactions : la deshydratation du methanol en ether

dimethylique (DME) et les reactions du melange de methanol/DME conduisant a la

formation d'hydrocarbures. Dans les conditions habituellement utilisees pour la

transformation MTO, approximativement 450 a 500°C, la reaction de deshydratation du

methanol est tres rapide et peut etre consideree a l'equilibre. A des temperatures plus

faibles, typiquement inferieures a 300°C, la formation d'hydrocarbures est pratiquement

nulle et la reaction de deshydratation peut ne plus etre a l'equilibre. C'est le cas de

plusieurs precedes industriels ou cette reaction a lieu dans un reacteur qui lui est

specifique, afin de limiter 1'augmentation de temperature inherente a la transformation

MTO. Le catalyseur utilise pour ces reactions est un melange de 10% de ZSM-5, avec

un rapport de Si/Al de 140, qui a ete fixe dans une matrice de Si/Al, constitute de

Catapal B, de Levasil 100s/30% et de kaolin. Ce melange a ete seche par atomisation

puis calcine a 550°C pendant 4h. Le diametre moyen des particules obtenues etait de 100

urn. Cette poudre a d'abord ete melangee dans une solution de (NH3)2HP04, ensuite

sechee puis calcinee afin d'obtenir un catalyseur contenant 1.5% de phosphore. Ce

dernier a ete ajoute pour reduire la formation de methane.

Le travail realise se divise en trois etapes : en premier lieu, la reaction de deshydratation

du methanol a ete etudiee en utilisant un lit fluidise. La reaction MTO a ete investiguee
 xviii

dans un lit fixe dans une deuxieme etape, et dans un lit fluidise dans une etape finale.

Ainsi, un modele cinetique representant 1'ensemble de ces reactions complexes a ete

propose.

La reaction de deshydratation du methanol a ete etudiee dans un reacteur fluidise en

quartz de 46mm de diametre. Ce reacteur a permis d'investiguer aussi bien les effets de

transfert de matiere que la reaction chimique intrinseque. Les analyses des

concentrations des produits de la reaction ont ete realisees par un spectrometre de masse

(SM) et en partie par chromatographic en phase gazeuse (GC). Comme le SM permet

des mesures en ligne, les changements dans les conditions de reaction sont facilement

detectables et les regimes permanents peuvent etre determines. Les experiences ont ete

realisees avec du methanol dilue dans de l'argon a quatre concentrations distinctes : 5,

15, 30 et 33 mol% et a trois temperatures differentes : 250, 275 et 325°C. A 325°C,

dependamment du rapport debit massique et de la quantite de catalyseur (WHSV), la

reaction de deshydratation du methanol peut conduire a la formation d'autres

hydrocarbures que le DME. Ces cas-la ont ete ignores. Lorences et coll. (2006) ont

montre que l'hydrodynamique des lits fluidises peut etre assimilee a celle des reacteurs

parfaitement melanges en serie a faibles vitesses superficielles. Les mesures de

Distribution du Temps de Sejour (DTS), realisees dans le cadre de ce travail, ont permis

de caracteriser le lit fluidise comme etant equivalent a 6 reacteurs parfaitement melanges

en serie, lorsque la vitesse superficielle est egale a 10 fois la vitesse minimale de

fluidisation. Afin d'obtenir des WHSV relativement elevees meme a faibles vitesses
 XIX

superficielles, la quantite de catalyseur a ete variee de 25 a 200g. Les conversions

experimentales obtenues ont ete de l'ordre de 30 a 99+% par rapport a l'equilibre.

Afin d'analyser la cinetique de reaction, deux reacteurs modeles differents ont ete

utilises pour modeliser rhydrodynamique du lit fluidise. Le premier est un modele a

deux-phases (phase emulsion et phase bulles). La reaction a lieu seulement dans la phase

emulsion. Le gaz est considere en ecoulement piston, alors que la phase emulsion est

parfaitement melangee. Le coefficient de transfert de matiere entre les deux phases est

donne par la correlation de Sit et Grace (1981). Le coefficient de diffusion moleculaire

du methanol a ete assume egal a 0.4cm2/s. Le diametre des bulles a ete calcule en

fonction de la hauteur du lit, en utilisant la correlation de Mori et Wen (1975). Le

deuxieme modele hydrodynamique considere est une serie de 6 reacteurs parfaitement

melanges, obtenu par DTS. Un modele cinetique de la reaction de deshydratation du

methanol en DME a ete propose en considerant i) aussi bien la reaction principale que la

reaction inverse, ii) l'absorption de l'eau et du methanol (equation 1).

r
kCM
cc '
C2 Tf
_ \ ^M^eq J
(l + KMCM+KwCwf

Ce modele a ete compare a celui de Bercic et Levee (1992) (Equation 2)

_ kK2M{c2M-CwCEIKeq)
r
MeOH ~ I \4 (2)
(\ + 2{KMCMf5+KwCw)
 XX

Plusieurs series de parametres cinetiques ont ete trouvees en couplant les deux modeles

hydrodynamiques avec les deux modeles cinetiques et par optimisation des donnees

experimentales obtenues aux trois plus faibles concentrations de methanol. L'analyse

des resultats a montre que le modele cinetique propose (equation 1), obtenu en utilisant

n-reacteurs parfaitement melanges, permet d'obtenir les predictions les plus proches des

donnees experimentales. De plus, 1'extrapolation de ce modele aux donnees obtenues

pour les plus grandes concentrations de methanol etait excellente. Le modele

hydrodynamique n-reacteurs parfaitement melanges, couple soit au modele cinetique

propose soit au modele cinetique de la litterature, donne de meilleurs resultats (valeur de

R respectivement egale a 0.934 et 0.910), meme si le modele a deux phases ne peut pas

etre rejete (valeur de R respectivement egale a 0.918 et 0.9006).

Cette etude a aussi montre qu'une experience cinetique dans un lit fluidise a petite

echelle a des avantages pour les reactions exothermiques, car elle elimine les

incertitudes liees aux points chauds et aux gradients de temperature. Pour utiliser les n-

reacteurs parfaitement melanges comme reacteur modele, il faut maintenir une vitesse

superficielle relativement faible. Par contre, avec les contraintes sur les vitesses

superficielles et la masse de catalyseur, les conversions a des temperatures donnees

peuvent etre limitees.

La reaction MTO a ete etudiee dans plusieurs lits fixes de faible diametre. Les

mecanismes de la reaction principale et de la deactivation ont ete analyses pour

differentes WHSV, compositions a Fentree et longueurs de reacteur (3, 6 et 8 mm). La

chromatographic en phase gazeuse a ete utilisee pour determiner la composition des

 xxi

produits de la reaction. Les composes a faibles nombres de carbone ont ete analyses

individuellement, alors que les fractions de C4 et de Cs+ (non aromatiques) ont ete

mesurees comme un tout. Ces experiences ont suggere que les hydrocarbures non

aromatiques C5+ sont des intermediaires de la reaction MTO conduisant a la formation

de propylene et de butylene (mais pas d'ethylene) par un craquage supplemental. La

production d'ethylene se produit, quant a elle, par divisions des aromatiques

(principalement xylenes et trimethylbenzenes), ce qui Concorde avec les travaux de

Svelle et coll. (2006). Par contre, le methane qui est predominant a de faibles

conversions du methanol/DME et a des taux eleves de WHSV doit certainement se

former directement a partir du methanol ou du DME et non a partir des hydrocarbures

superieurs.

La desactivation du catalyseur est tres sensible a la WHSV et croit avec celle-ci. Par

consequent, la capacite du catalyseur a transformer le methanol depend du flux a

l'entree du reacteur. La composition de ce flux influence beaucoup la desactivation du

catalyseur et la composition des produits. De faibles pressions partielles de methanol

favorisent la production d'olefmes, mais elles reduisent la capacite du catalyseur a

transformer le methanol. Ceci est en contradiction avec les travaux de Chen et coll.

(2000), qui ont montre que la desactivation du catalyseur est fonction de la quantite de

methanol converti et de la temperature et non de WHSV.

Par des inspections visuelles, le catalyseur desactive est constitue essentiellement de 3

zones : une zone a l'entree, faiblement desactivee, ou la reaction de deshydratation du

methanol a lieu; une zone intermediaire noire (presence de coke), ou la production

 xxii

d'olefmes est preponderante; et une zone post-reaction, ou la presence du coke est plutot

limitee, tres probablement a cause des reactions de craquage secondaire d'oxygenes qui

consomment du carbone.

La troisieme partie de la these traite de la reaction MTO dans un lit fluidise. Les effets

de la composition de 1'alimentation, de la temperature et de la WHSV ont ete analyses

dans le meme reacteur que celui de la reaction de deshydratation (voir premiere partie du

travail). En alimentant un melange 1-hexene/argon d'une part et un melange 1-

hexene/methanol/argon d'autre part, nous avons propose un mecanisme probable de

reaction. La distribution des produits de reaction n'a pas change de facon significative

avec une alimentation en methanol. Avec l'hexene, les differences majeures sont la plus

faible formation de methane et la plus grande formation d'ethylene. Ceci confirme

1'etude realisee en lit fixe concluant que i)le methane est produit principalement a partir

du methanol ou DME, ii)l'ethylene est produit par craquage secondaire.

Plusieurs cinetiques de cette reaction ont ete proposees dans la litterature. Cependant, la

plupart d'entre elles se basent sur l'utilisation d'un catalyseur de type zeolite SAPO et

ne prennent pas en compte la formation d'hydrocarbures superieurs, car les pores de ce

catalyseur ne permettent pas le passage des grosses molecules. Les modeles bases sur

des catalyseurs de type ZSM-5 ont ete developpes principalement pour la reaction

methanol-gazoline, qui combinent les defines legeres. Quelques rares modeles, tels que

celui de Schoenfelder et coll. (1994), traitent les olefmes legeres individuellement, mais

ne tiennent pas compte du craquage secondaire des hydrocarbures lourds. Un modele

 xxiii

cinetique a ete propose en se basant sur le mecanisme «lot d'hydrocarbures », ou les

olefines sont produites par de reactions reversibles a partir d'un lot d'hydrocarbures

superieurs. La figure 1 montre un schema du modele propose. Comme l'indiquent les

conclusions anterieures, le methane est suppose se former a partir du methanol/DME.

Les aromatiques sont considered comme etant C9H12 et les paraffines, tout comme le

propane, sont produits directement a partir du propylene.

\ /
2 MeQH *=* DME + Hfi—^ C/ ^ ^ C4

1 / \
CH4 + C + H^) C§+ C§
Figure 1: Schema du modele cinetique propose

Pour la reaction MTO en lit fluidise, le modele hydrodynamique a deux phases a ete

choisi. En effet, comme la hauteur du lit etait generalement beaucoup plus elevee que

dans les experiences realisees dans la premiere partie de ce travail, la taille des bulles

etait done plus grande. Par ailleurs, ce modele n'a pas ete rejete lors de l'etude de

deshydratation du methanol. Les valeurs des constantes cinetiques du modele ont ete

estimees en ajustant les predictions du modele (hydrodynamique + cinetique) propose

avec les donnees experimentales. Le modele predit bien la formation d'olefmes legeres,

mais moins bien la production des paraffines et des fractions C6+.

 xxiv

TABLE OF CONTENTS
ACKNOWLEDGEMENTS iv
RESUME vi
ABSTRACT xii
CONDENSE EN FRANCAIS xvii
TABLE OF CONTENTS xxiv
LIST OF TABLES xxvii
LIST OF FIGURES xxix
LIST OF FIGURES xxix
LIST OF SYMBOLS xxxvi
Chapter 1 Introduction 1
Chapter 2 Literature study 8
2.1 Introduction 8
2.2 Zeolites 9
2.2.1 ZSM-5 9
2.2.2 SAPO-34 11
2.3 Mechanisms 12
2.4 Reaction conditions 17
2.4.1 Temperature effect 17
2.4.2 Pressure effect 18
2.4.3 Catalyst deactivation 19
2.4.4 Effect of water 20
2.4.5 Gas residence time 21
2.5 Kinetic 22
2.6 Processes 32
2.6.1 Industrial Processes 37
[Link] Lurgi Process 38
[Link] UOP/Hydro Process 39
[Link] Mobil Process 40
Chapter 3 Objective and methodology 42
3.1 Objective 43
3.2 Methodology 44
3.2.1 Catalyst 44
3.2.2 Fixed bed experiments 44
3.2.3 Fluid bed experiments 45
3.2.4 Kinetic model 47
 XXV

Chapter 4 MeOH to DME in Bubbling Fluid Bed: Experimental and Modelling 48

4.1 Presentation of the article 48
4.2 MeOH to DME in Bubbling Fluidized Bed: Experimental and Modelling 49
4.2.1 Abstract 49
4.2.2 Introduction 50
4.2.3 Experimental 54
[Link] Equipment 54
[Link] Catalyst 55
[Link] Experimental conditions 56
4.2.4 Experimental results 57
[Link] RTD 57
[Link] Methanol to DME 60
4.2.5 Modeling 64
[Link] Kinetics 64
[Link] Fluid bed model 66
4.2.6 Modelling result and discussion 70
4.2.7 Conclusion 75
4.2.8 Nomenclature 76
4.2.9 Literature Cited 78
Chapter 5 Phosphorous modified ZSM-5: Deactivation and product distribution for
MTO* 83
5.1 Presentation of the article 83
5.2 Phosphorous modified ZSM-5: Deactivation and product distribution for MTO
84
5.2.1 Abstract 84
5.2.2 Introduction 84
5.2.3 Experimental 87
[Link] Catalyst 87
[Link] Reactor setup 87
[Link] Procedure 89
5.2.4 Results 89
5.2.5 Conclusions 98
5.2.6 References 99
Chapter 6 Trends in the MTO reaction by MS 103
6.1 Introduction 103
6.2 Experimental 103
6.3 Results and discussion 104
6.4 Conclusion 107
Chapter 7 Kinetic modelling of MTO reaction over ZSM-5 in fluid bed 109
7.1 Presentation of the article 109
7.2 Kinetic modelling of MTO reaction over ZSM-5 in fluid bed 110
7.2.1 Abstract 110
7.2.2 Introduction Ill
 xxvi

7.2.3 Experimental 114

[Link] Equipment 114
[Link] Catalyst 115
[Link] Experimental conditions 116
7.2.4 Experimental data 117
7.2.5 Fluid bed model 125
7.2.6 Kinetic model 128
7.2.7 Modelling results 132
7.2.8 Conclusion 137
7.2.9 Nomenclature 138
7.2.10 References 141
Chapter 8 General discussion 146
8.1 Elevation of the catalyst with glass beads 148
8.2 Kinetic and Modelling of the MTO reaction 150
Chapter 9 Conclusion and recommendations 154
9.1 General conclusions 154
9.2 Recommendation 157
9.2.1 Kinetics in fluidized bed 157
9.2.2 MTO reaction kinetic 157
9.2.3 Equilibrium inside the catalyst? 158
References. 159
Appendix A - Catalyst preparation / 73
Appendix B - Fluidized bed reactors used. 176
 xxvu

LIST OF TABLES

Table 2-1: Product distribution of the MTO reaction in fluid bed at 482°C and 102

kPa methanol partial pressure (Tabak and Yurchak, 1990) 33

Table 2-2: Properties of the H-ZSM-5 and the catalyst used in the 34

Table 4-1: Particle properties of the catalyst 56

Table 4-2: n-CSTR model parameters for the RTD experiments done at uo

0.052m/s 58

Table 4-3: Experimental data obtained- all flows are listed at STP conditions 61

Table 4-4: Kinetic models including the reverse term 64

Table 4-5: Hydrodynamic correlations - the correlation for bubble diameter is in

cm and not in meters 69

Table 4-6: Best fit parameters for the kinetic model Equation 4-5 and the level of

fit to the experimental data 70

Table 4-7: Best fit parameters for model of Bercic and Levee (1992) and the level

of fit to the experimental data 70

Table 4-8: Fixed bed experiments in a 4 mm inner diameter quartz reactor 74

Table 5-1: Product distribution after 2 hours on stream in the 3 and 6 mm diameter

reactors 91

Table 7-1: Particle properties of the catalyst 116

Table 7-2: Carbon deposition on the MTO catalyst 118

 xxviii

Table 7-3: Experimental results from the fluid bed experiments - product

distribution is given in carbon % * 100ml H2 was co-feed with

methanol and argon 119

Table 7-4: Hydrodynamic correlations 128

Table 7-5: Kinetic model with optimized parameters for k0 and activation energy

k* is calculate with a reference temperature of 500°C 132

Table B-l: Dimensions of the 7.79 cm inner diameter reactor 177

 XXIX

LIST OF FIGURES

Figure 1: Schema du modele cinetique propose xxiii

Figure 2-1: Methanol to hydrocarbon reaction path at 371°C (Chang, 1984) 9

Figure 2-2: Structure of the MFI framework

(International Zeolite Association, 2006) 10

Figure 2-3: Schematic drawing of an acid site on a ZSM-5 11

Figure 2-4: Structure of the CHA framework (International Zeolite Association,

2006) 12

Figure 2-5: Schematic drawing of the oxonium ylide mechanism (Stocker, 1999)... 14

Figure 2-6: Reaction for the free radical mechanism (Chang et al., 1989) 15

Figure 2-7: The hydrocarbon pool mechanism (Dahl and Kolboe, 1994) 16

Figure 2-8: Ethylene formation by reaction of methanol withmethylbenzenes.

(Aguayo et al., 2005) 16

Figure 2-9: Product distribution of the MTO reactionat high, low and atmospheric

pressure (Chang, 1984) 18

Figure 2-10: Evolution with time on stream of the composition of the lumps of the

kinetic scheme, in three successive reaction-regeneration cycles in fixed

bed. Reaction conditions: temperature, 773 K, space time, 0.093 (kg of

catalyst) h (kg of methanol)-1; water content in the feed, 50 wt%. Points:

experimental results. Solid lines: calculated (Gayubo et al, 2003) 28

 XXX

Figure 2-11: Scheme and kinetic model for the MTO-reaction by Schoenfelder et al.

(1994) 29

Figure 2-12: Reaction scheme for the MTO reaction by Bos et al. (1995) 30

Figure 2-13: Evolution of the mass fractions of each lump, until steady state is

reached at three different temperatures, space time 0.040 gcath/gMeOH

and average solid residence time 0.866h_1 (Ortega et al., 1998) 34

Figure 2-14: Contours of the relative production rate of light olefins for different

residence times in the reactor and regenerator. Temperature 420°C

space time 0.021 gcath/gMeOH (Gayubo et al. 2000) 35

Figure 2-15: Comparison between measured (symbols) and predicted (lines) product

distribution of the CFB reactor at a temperature of 500°C (Schoenfelder

etal, 1994) 37

Figure 2-16: Flow diagram of Lurgi MTP® Process (Stocker 2003) 39

Figure 2-17: Simplified flow diagram of the UOP/Hydro MTO Process (Houdek

and Andersen 2005) 40

Figure 4-1: Drawing of the glass fluidized bed with glass beads, catalyst and

thermocouple 54

Figure 4-2: Schematic diagram of the experimental setup 55

Figure 4-3: RTD experimental results over the fluid bed with a gas velocity of 5.4

cm/s. A) reactor loaded with 540g glass beads B) Reactor loaded with

540g glass beads and 50g of catalyst C) Reactor loaded with 220gof
 XXXI

catalyst. Experimental data represented by dots and n-CSTR model

with a line 59

Figure 4-4: MS results from experiments done at 275°C with 25 g of catalyst 62

Figure 4-5: Conversion of methanol to DME and water based on equilibrium

conversion as function of catalyst amount at five different gas velocities

(experiment 20 - 39). Lines are the model predictions using kinetic

model equation 4-5 and 6 CSTR's in series 63

Figure 4-6: Schematic of the two phase fluid bed model 68

Figure 4-7: Model predictions vs. experimental with 6 CSTR's in series. - Model

Eq. 4-5; • Model of Bercic and Levee (1992) 72

Figure 4-8: Effect of the molar diffusion coefficient with the use of the best fit

parameters for 6-CSTR's in series with the two phase model and the

kinetic model of Equation 4-5. The experimental points are at 275°C

and a superficial gas velocity of 7.79 cm/s 73

Figure 5-1: Experimental setup 88

Figure 5-2: MTO product distribution at 500°C in the 6 mm inner diameter reactor

with 10% methanol in nitrogen feed: (a) WHSV 0.22 h"1 (dashed lines

due to missing data), (b) WHSV 0.43 h"1 90

Figure 5-3: Hydrocarbon distribution as a function of oxygenate conversion. Based

on experiments in Table 1 at 500°C 92

 XXX11

Figure 5-4: Hydrocarbon distribution as a function of oxygenate conversion. Based

on experiment at 500°C in 8 mm inner diameter reactor with pure

methanol feed. WHSV 2.37 h"1 93

Figure 5-5: Pore size distribution of fresh, coked and regenerated catalyst 96

Figure 5-6: Catalyst coking at t = 0, 10, 20, 50, 110, 170, 1310 minutes in the

6 mm reactor, 10% MeOH in nitrogen, T=500°C, WHSV 0.43

g(MeOH)/gcat/h (gas flows from the top down through the bed) 97

Figure 6-1: MS traces during the MTO reaction with temperature increments of

50°C from 250°C to 550°C. A) Methanol, B) DME, C) Hydrocarbons

and D) Methane 105

Figure 6-2: Product distribution from the MTO in the temperature interval 450 -

550°C with several different feed compositions 106

Figure 6-3: Comparison between ethene and other hydrocarbon concentration

for the MTO reaction 107

Figure 7-1: Drawing of the glass fluid bed reactor - glass beads are used to elevate

the catalyst for catalyst amounts below lOOg 114

Figure 7-2: Schematic diagram of the experimental setup 115

Figure 7-3: Comparison between experimental data and equilibrium composition

of the C2 - C6 olefins in the temperature range 400 - 550°C a)

ethylene and propylene b)C4, C5,C6 121

 xxxiii

Figure 7-4: Product distribution vs. residence time with methanol/Argon feed

(molar ratio 1:1) over ZSM-5 at 550°C. A: conversion , • : ethylene,

• : propylene, •: propane, A : C4, • : C5, O: C6+ 122

Figure 7-5: Product distribution vs. residence time with methanol/Argon feed

(molar ratio 1:1) over ZSM-5 at 500°C. A: conversion , • : ethylene,

• : propylene, •: propane, • : C4, • : C5, O: C6+ 123

Figure 7-6: Product distribution vs. residence time with methanol/Argon feed

(molar ratio 1:1) over ZSM-5 at 450°C. A: conversion , • : ethylene,

• : propylene, •: propane, • : C4, • : C5, O: C6+ 124

Figure 7-7: Product distribution vs. residence time with pure methanol feed over

ZSM-5 at 500°C. A: conversion , • : ethylene,*: propylene,

•: propane, • : C4, • : C5, O: C6+ 124

Figure 7-8: Schematic drawing of the fluid bed model 126

Figure 7-9: Product distribution of reaction over ZSM-5 with three different feed

compositions and a total inlet flow rate of 1.57 L/min at 500°C 1)

Methanol/Argon molar ratio 1:2 2) Methanol/1-hexene/Argon molar

ratio 9:1:20 and 3) 1-hexene/Argon molar ratio 1:2 129

Figure 7-10: Schematic drawing of the kinetic model 131

Figure 7-11: Comparison of the calculated product distribution to the measured

data 134

Figure 7-12: Measured (points) and calculated (lines) data at 550°C with

methanol/argon (ratio 1:1) feed as function of residence time in the fluid

 xxxiv

bed. A(-"): conversion: • ( — ) : ethylene,B(—): propylene,

A(--): C4) • ( — ) : C5, O ( - ) : C6+. 135

Figure 7-13: Measured (points) and calculated (lines) data at 500°C with

methanol/argon feed (ratio 1:1) as function of residence time in the

fluid bed. A ( - ) : conversion: • ( — ) : ethylene, •(—): propylene,

A C - ) : C4, • ( — ) : C5, O ( - ) : C6+. 135

Figure 7-14: Measured (points) and calculated (lines) data at 450°C with

methanol/argon feed (ratio 1:1) as function of residence time in the

fluid bed. A(-—): conversion: • ( — ) : ethylene,B(—): propylene,

A(--): C4, • ( — ) : C5, O ( - ) : C6+. 136

Figure 7-15: Measured (points) and calculated (lines) data at 500°C with methanol

feed ratio 1:1 as function of residence time in the fluid bed. A(—):

conversion: • ( — ) : ethylene,B(—): propylene, A(—): C4> • ( — ) : C5,

0(—):C 6 + . 137

Figure 8-1: Particle size distribution of the MTO catalyst, sand and glass beads used

- measured was preformed n a Horiba LA-950 150

Figure 8-2: Calculated values with the use of the model of Schoenfelder et al. (1994)

versus the obtained experimental data from fluidized bed 151

Figure A-l: Methane formation with time at different phosphor loadings of the fluid

bed catalyst 174

Figure A-2: picture of the spray dried particles 175

Figure B-l: Drawing of the 3" inner diameter fluid bed Legend: 1) feed inlet, 2)

windbox, 3) distribution grid 4) solid outlet, 5) fluid bed region, 6)

sampling/measuring ports, 7) freeboard, 8) electrical heating, 9)

solid inlet, 10) disengagement zone 177

Figure B-2: Schematic diagram of the 3" fluidized bed setup 179

Figure B-3: 4.1cm inner diameter fluidized bed reactor 181

 LIST OF SYMBOLS

Chapter 2

A, B, C, D lump of oxygenates , :CH2, light olefins, and the remaining components

a Remaining catalyst activity due to coke deposition

c weight percentage of coke on the catalyst, kg/( 100kgcat)

Cj Concentration of component j ,

dp particle diameter, urn

E(t) Residence-time distribution function

k; Kinetic constant for component i, h"1

kdi Kinetic constant for deactivation for component i, h"1

k;° Kinetic constant for component i for zero coke concentration, h"1

N Number of tanks in C STR model

P Pressure, bar

Q Volumetric flow rate, m 3 /s

r. Reaction rate for component i,

rjo Reaction rate of fresh catalyst for component i,

t Time

tm Mean residence time, s

Vj Volume of tank j , m

x; Molar fraction of component i, mol/mol

Xi Weight fraction of component i, kg/kg

 xxxviii

Pe = Peclet number (Pe = UgL/D)

r, = Reaction rate of component i (kmol/kg-s)

t-/2 . = Halftime (s)

tm - Mean residence time (s)

Umf = Minimum fiuidization velocity (m/s)

Umb = Minimum bubbling velocity (m/s)

Ub = Bubble velocity (m/s)

Uo = Superficial gas velocity (m/s)

ve = Volume of emulsion (m )

w = Catalyst weight (kg)

Q = Gas flow (ml/min)

X = Total conversion

X = Mole fraction

y - Mole fraction of gas

z - Height in bed (m)

Greek symbols

8 = Bubble phase volumetric fraction

smf = Void fraction at minimum fiuidization

u = Gas viscosity (kg/m-s)

a2 •= Variance

pp = Particle density (kg/m3)

Ptapped ~ Tapped density (kg/m3)

 xxxvii

a Remaining catalyst activity due to irreversible deactivation

P Catalyst activity paramenter

Chapter 4

At . = Cross-sectional area of the fluid bed (m2)

c = Molar concentration (kmol/m3)

© = Molar diffusion coefficient (m2/s)

D
< = Reactor diameter (m)
• ) -

dp = Mean particle diameter (m)

db = Bubble diameter (m)

dbo = Initial bubble diameter at the distributor (m)

dbm = Maximum bubble diameter (m)

Ea = Activation energy (kJ/mol)

F = Fine fraction

Fb = Flow rate in bubble phase (kmol/s)

Fe - Flow rate in emulsion phase (kmol/s)

g = gravity (m/s2)

Keq = Equilibrium constant

k = Reaction rate constant (m6/kmol-s)

Kbe = Bubble to emulsion mass transfer coefficient (s"1)

kbe = Bubble to emulsion mass transfer coefficient (m/s)

N = Number of CSTR's
 •2
=
pmf Minimum fluidization density (kg/m )

pg = Gas density (kg/m )

x = Residence time (s)

Subscripts

exp = Experimental

calc = Calculated

n = CSTR number

i = species

be = Bubble to emulsion

b = Bubble

e =Emulsion

Chapter 7

At = Cross-sectional area of the fluid bed (m z )

C = Molar concentration (kmol/m 3 )

C^ = Hydrocarbon pool species (Decene)

D = Molar diffusion coefficient (m 2 /s)

Dt = Reactor diameter (m)

dp = Particle diameter (m)

db = Bubble diameter (m)

dt,o - Initial bubble diameter at the distributor (m)

=
dbm Maximum bubble diameter (m)
 xl

Ea = Activation energy (kJ/mol)

F = Fine fraction

F = Flow rate (kmol/s)

g = gravity (m/s 2 )

h = Height in bed (m)

Keq = Equilibrium constant

k; = Reaction rate constant (m /kmol-s)

Kbe — Bubble to emulsion mass transfer coefficient (s~ )

kbe = Bubble to emulsion mass transfer coefficient (m/s)

N = Total number of components in the system

rj = Reaction rate of component i (kmol/kg-s)

T = Temperature (K)

umf = Minimum fluidization velocity (m/s)

umb = Minimum bubbling velocity (m/s)

Ub = Bubble velocity (m/s)

uo = Superficial gas velocity (m/s)

Ve = Volume of emulsion (m )

W = Catalyst weight (kg)

y = Mole fraction

Greek symbols
T = Gas expansion due to reaction

5 = Bubble phase volumetric fraction

£mf = Void fraction at minimum fluidization

Scat = Void fraction of the catalyst

(A = Gas viscosity (kg/m-s)

Pp = Particle density (kg/m3)

Pb = Bulk density (kg/m3)

Ptapped = Tapped density (kg/m3)

Pmf = Minimum fluidization density (kg/m3)

•5

Pg = Gas density (kg/m )

Subscripts

e = Emulsion Phase

b = Bubble Phase

i = Component index

n = CSTR in series

exp = experimental data

cal = calculated value

 1

Chapter 1
Introduction

In the recent years the demand for the light olefins ethylene and propylene has increased

with the largest growth in propylene. Propylene is mainly produced from steam cracking

units which produces about twice as much ethylene compared to propylene. With the

increases propylene/ethylene ratio in the global demand there is a need to produce

propylene from other sources. Fluidized catalytic cracking units are currently filling this

gab in production but it is expected that it will not be able to cover the demand and

alternative sources are needed (Andersen, 2003; Houdek and Andersen, 2005). A large

part of the worlds natural gas resources are located in regions where the need is low or

non existent and transport to the gas market is expensive. These remote or stranded

natural gas fields therefore offer low cost gas for on-site chemical production. One way

to utilize the stranded natural gas is to produce methanol in large scale plants 5,000 -

10,000 MTD. Methanol from these plants offers a cheap feedstock for a Methanol To

Olefin (MTO) plant. Olefin production from a MTO facility are comparative with

naphtha crackers when the crude oil price is around 16$ per barrel (Houdek and

Andersen, 2005). With the recent increase in oil price the MTO process is competitive

with standard olefin technology and a good way to use stranded natural gas resources

when coupled with a mega-methanol plant.

 2

Several companies already offer MTO technology for large scale olefin production.

Lurgi has the Methanol To Propylene (MTP) process which converts methanol to

propylene with a selectivity of approximately 70%. The process is based on fixed bed

reactors in series with a ZSM-5 zeolite as the active catalyst. Several plants are

planned/under construction with the majority located in China (Lurgi, 2008).

UOP/Hydro offers a MTO technology based on fluidized bed reactors with a SAPO-34

zoelite as the catalyst, which is continually regenerated to reduce deactivation due to

coking. The first commercial plant is scheduled to come online in 2012 and consists of

the MTO technology coupled with an olefin cracking unit to increase the light olefin

yield (UOP, 2008). Exxonmobil have conducted extensive research in the process and

have demonstrated a fluid bed MTO process.

Although research in the methanol to olefin reaction has been ongoing since the late

1970's, there are still several aspects that are poorly understood. The fundamental

reaction to form the first C-C bond is still disputed and several different reaction

mechanisms have been proposed over the years. The kinetics of the reaction are

complicated with a large amount of products including most hydrocarbons from methane

to polymethylated-benzenes. Depending on the type of zoelite, the product spectrum can

vary between a narrow range of Ci - C5 for the SAPO zeolite to Ci - C10 for the ZSM-5.

The product distribution depends on the pore structure of the zeolites with narrow pores

leading to a narrow product spectrum. The formation of the hydrocarbons is taking place

inside cages in the zeolite which are able to accommodate molecules larger than what
 3

can be transported out of the zeolite pores. Molecules unable to diffuse out of the pores

become trapped and lead to catalyst deactivation. The deactivation rate of the catalyst

depends on the pore structure and pore diameter. The SAPO catalysts which have

narrow pores deactivate rapidly while the ZSM-5 activity remains for a longer period of

time. For all commercial purposes of the MTO reaction both types of catalysts need to

be regenerated at regular intervals to maintain the desired activity. In the present study

the focus have been on the ZSM-5 zeolite.

Kinetic studies are normally conducted in fixed beds or Berty reactors since the

hydrodynamics of the reactor is well understood and decoupling of the reaction kinetics

from the reactor hydrodynamic is well established. In this work, a fluidized bed is

employed for the kinetic study for the following reasons:

• The MTO reaction is exothermic and heat transfer limitations in a fixed bed can

result in hot spots or radial and axial gradients creating uncertainty of the actual

reaction temperature. In a fluidized bed solid circulation will ensure a uniform

temperature profile in the bed, due to high heat transfer of the solids

• Deactivation of the catalyst in a fluidized bed is homogeneous throughout the

bed while in a fixed bed a deactivation front will progress through the bed

resulting in different levels of catalyst activity in the reactor

• MTO Fluid bed catalyst with a broad particle distribution and an average

particle size around lOOum can be use directly whereas fixed bed experiments

will require larger particles to avoid undesired pressure drop over the bed.
 4

Possible effects of altering the particle size to fit fixed bed experiments are also

avoided.

Using a fluidized bed to conduct kinetic experiments raises the concern about

decoupling the hydrodynamic from the kinetics since gas-phase fluid bed

hydrodynamics are more complicated than the fixed bed and not as well established.

Lorences et al. 2006 showed that at gas velocities above umf the fluid bed can be

modelled well with both the dispersion model as well as the n-CSTR in series model. At

the beginning of the bubbling fluidization regime the bed characteristics deviate more

and more from plug flow conditions. Several two and three phase models with varying

complexity have been developed over the years to describe the hydrodynamics of the

fluidized bed. Due to the complex kinetics of the MTO reaction it is desired to minimize

the complexity of the hydrodynamic model while still being able to separate the two. A

simple two phase fluidized bed model including an emulsion and a bubble phase is

therefore well suited for this purpose. A study of the hydrodynamics in the fluid bed is

therefore of importance to be able to validate the model.

The methanol to dimethylether (DME) reaction is used in the study of the

hydrodynamics of the fluid bed. In contrast to the MTO reaction, the methanol to DME

reaction is equimolar and the reaction is much less complicated and the kinetics are

established in the literature (Bercic and Levee, 1992, Mollavali et al. 2008). The fluid

bed model is evaluated through methanol to DME experiments and residence time
 5

distribution measurements (RTD) to characterize the flow. The RTD measurements are

preformed by switching between air and argon and monitor the argon response as

function of time. The main effect of these experiments is to establish the mass transfer

between the bubble and emulsion phase in the hydrodynamic model which is one of the

main uncertainties.

Deactivation of the catalyst has a large influence on the MTO process and the catalyst

has to be either continuously or periodically regenerated to maintain the activity. The

product distribution is dependent on the deactivation of the catalyst and the continuous

regeneration in a circulation fluidized bed reactor is therefore beneficial for steady-state

operation. The deactivation rate of the catalyst is critical; several parameters affect the

rate and include: temperature, methanol and water partial pressure. Fixed bed

experiments are excellent to study these effects on deactivation. As the catalyst

deactivates the number of active sites decreases. By gradually reducing the number of

active sites on the catalyst some insight into the reaction mechanism can be obtained as

more intermediate species are present in the product distribution.

The general mechanism of the MTO reaction has been disputed for several years but in

the recent years the hydrocarbon pool mechanism proposed by Dahl and Kolboe (1993)

has gained widespread acceptance. The hydrocarbon pool consists of (CH2)n located

inside the cage of the zeolite which reacts to the lower olefins. The light olefins are

formed by reaction of methanol with polymethylated benzenes which then splits off
 6

olefins after rearrangement under formation to a lower order polymethylated benzene.

From carbon isotopic labeling experiments BJ0rgen et al (2007) found that the

mechanism between SAPO and ZSM-5 differ regarding the polymethylated benzenes

that are active in the formation of the light olefins. They also proposed that the reaction

consists of two parallel mechanisms one consisting of a modified hydrocarbon

mechanism for benzene and ethylene the other with alkene methylations and

interconversion of propene and higher alkenes.

As mentioned above the mechanism is highly complicated but the majority of the kinetic

reaction schemes proposed over the years have consisted of a few lumped species which

capture main trends in the reaction. Only a few kinetic models include the light olefin

species as individual components and these kinetics are for the SAPO type catalysts and

neglect the higher hydrocarbon species. It is well known that cracking of higher olefins

over ZSM-5 is an important reaction mechanism and including it in the kinetic reactions

can prove to be important. In this work a kinetic model based on the hydrocarbon pool

mechanism is developed for the MTO reaction over ZSM-5 which includes the light

hydrocarbons as separate components while also including the cracking and

oligomerization reactions for the higher olefins.

Gas chromatograph (GC) and mass spectrometer (MS) are used to analyse the fluid bed

MTO experiment. The former gives the product distribution of the hydrocarbons with

separate quantification of the light species and lumps based on carbon number for the
 7

C4+ hydrocarbons. The MS is able to show trends in the reaction but due to a large

overlap between the hydrocarbons precise quantification is difficult. For the methanol to

DME reaction, however, the MS is well suited to determine the product distribution.
 8

Chapter 2
Literature study
2.1 Introduction
The Methanol-to-Olefin (MTO) process over a zeolite catalyst has been of great interest

since the late 1970's when Chang and Silvestri (1977) first proposed a mechanism for

the reaction. It was discovered by accident by two different groups at Mobil. One of the

groups was trying to form other oxygen compounds from methanol over a ZSM-5

catalyst while the other was reacting methanol with iso-butane. Both groups produced

unwanted hydrocarbons in the gasoline boiling range. Since then, a lot of research has

been conducted centred on the methanol to gasoline (MTG) and later the methanol to

olefin reaction. Figure 2-1 presents the product distribution of methanol conversion over

a ZSM-5 catalyst at 371°C. It shows that the olefin concentration is highly dependent on

the space time. Other factors, such as temperature, pressure, Si/Al ratio and deactivation

also influence the reaction and will be discussed in the following section.
 9

TO 1—r-TTTTTy 1 n i i t i | - 1 i-' f i n |
i—1 1—r T- i n nf—™—1 r r r t r-i
J.
80 , Methanol
.--•-••-Water

I 40
a"
Paraffins
| 30 (and Cg+ Olefins)

&- 20
Aromatic*
10

0 1 1 » « < i

1G"4 icr 3
10" 2
icr 1
i 10
1/LHSV, M
Figure 2-1: Methanol to hydrocarbon reaction path at 37FC (Chang, 1984)
2.2 Zeolites

The MTO reaction is an acid catalyzed reaction and several zeolites are therefore of

interest. The two most used zeolites for the MTO reaction are ZSM-5 and SAPO-34 but

several others have also been studied, including: MeAPSO-44, SAPO-18, SAPO-44,

ALP0 4 -5, ALPO4-I4, and ZKU-4 (Stocker, 1999). To be able to categorize zeolites, the

International Zeolite Association (IZA) has sorted zeolites with framework type codes.

The framework type code of ZSM-5 and SAPO-34 is MFI and CHA (International

Zeolite Association, 2006). The zeolite used in this project is a ZSM-5 with a Si/Al ratio

of 140. The focus will therefore be on this specific zeolite but a brief description of the

SAPO-34 zeolite will also be given.

2.2.1 ZSM-5
ZSM-5 has a 3-dimensional channel system with pore sizes of 5.1x5.5 and 5.3x5.6A in

two directions. The channels consist of 10-membered rings with one channel going
 10

straight and parallel to [010] and the other having a sinusoidal structure and running

parallel to [001] (Kokotailo, 1978). Cages with larger cross section than the channels

are formed where channels meet in both directions. In these cages, larger molecules can

form but are not able to pass through the channels and are trapped inside the zeolite,

leading to blockage of the pore system. A drawing of the MFI structure is shown in

Figure 2-2.

Figure 2-2: Structure of the MFI framework (International Zeolite Association, 2006)

ZSM-5 consists of Silica and Alumina in tetrahedral coordination with oxygen. The

introduction of alumina into a tetrahedral framework results in Brondsted acid sites

where a proton is loosely bound to the framework as shown in Figure 2-3. This form of

ZSM-5 is also called H-ZSM-5 (Chang, 1983). In the MTO reaction, the product

distribution over a ZSM-5 is wide, giving products in the range of Ci to

polymethylatated benzenes with the main products being propylene and butylenes. It is
 11

catalyzed by the acid sites. The reaction rate and product distribution are influenced by

the numbers of acid sites in the zeolite. Since it is the amount of alumina that governs

the acidity of the ZSM-5 zeolite, the acid strength can be represented by the Si/Al ratio.

Chang et al. (1984) investigated the effect of different Si/Al ratios between 17 and 835

at 500°C. For every ratio, it was found that all the catalysts had a maximum olefin

production at a space time where the conversion of oxygenates was complete. As the

Si/Al ratio is increased, the contact time for complete conversion of oxygenates

increases. The maximum production of olefins was found to be dependent on the Si/Al

ratio with an optimum around 250.

H+
0\ /0\ /0\ /O
x
/
/
Si Al Si
+
'+4\ / 3 \ / + 4 \

Figure 2-3: Schematic drawing of an acid site on a ZSM-5

2.2.2 SAPO-34

SAPO-34 is a silica-aluminophosphate zeolite with a CHA framework structure. It has a

3-dimensional pore structure with 3.8A diameter channels (International Zeolite

Association, 2006). A schematic drawing of the CHA framework is given in Figure 2-4.

As in the case of the MFI structure, the CHA structure also has internal cages that allow

formation of larger molecules but since the channels are narrower than in the case of the
 12

MFI, only smaller molecules are able to pass through the channels. The SAPO-34 gives

a narrower product spectrum then the ZSM-5 but the coking rate is higher due to

entrapment of larger molecules (Stocker, 1999).

Figure 2-4: Structure of the CHA framework (International Zeolite Association, 2006)

2.3 Mechanisms

Many mechanisms have been proposed, mainly concerning the formation of the first C-

C bond. There is a general consensus of the dehydration of methanol to DME and the

reaction of light olefins to paraffins, higher olefins and aromatics are known to be

formed from hydrocarbon chemistry in acidic media, where the reaction proceeds via a

classical carbenium mechanism with hydrocarbon transfer. A short description of the

different mechanisms from the first C-C bond are given below.
 13

In the oxonium ylide mechanism, dimethyl ether interacts with a Bronsted acid site on

the catalyst to form a dimethyl oxonium ion. The oxonium ion then reacts with DME to

form a trimethyl oxonium ion that is then deprotonated by an acid site to form a

dimethyl oxonium methyl ylide on the surface. Next the dimethyl oxonium ion either

undergoes a Stevens rearrangement or an intermolecular methylation leading to a

methylethyl ether and ethyldimethyl oxonium ion, respectively. Ethylene is then formed

by P-elimination of either of the two components. A schematic drawing is shown in

Figure 2-5. Many studies have been preformed to verify the existence of the oxonium

ions and the zeolites ability to abstract a proton from the oxonium ions to form the ylides

but no definitive answers have been found (Stocker, 1999).

 14

GHgQHg
CHgDCHa -*- CH3 Ct%

o
I
CM,

CHgOCHjCHa CH CH

• ?
+ CHgDH CHjtCHj

Figure 2-5: Schematic drawing of the oxonium ylide mechanism (Stocker, 1999)

The carbene mechanism involves a surface associated carbene and is based on a-

elimination of water from methanol. The resulting carbene then either reacts with

another carbene to olefins or reacts with methanol or DME by concurrent sp3 insertion.

It is not yet clear if the carbene is formed by cooperative action of acid and basic sties on

the catalyst or by decomposition on the surface (Stocker, 1999).

The free radical mechanism involves the participation of free radicals in the

conversion of methanol to hydrocarbons. Clarke et al. (1986) found that DME could be a

source of methyl radicals as they identified free radicals in the reaction of DME over
 15

ZSM-5. Chang et al. (1989) suggested a reaction path for the free radical mechanism

which is shown in Figure 2-6.

CH*PH + Z< til • CH:if> - I \ U.,0

CH^O Z + R* —* 'fI!.jD- Z + RH
•[Link] Z + Zfl *• 7C (7/j
1 +ZO*
2-0- CHf
H R + ZG* • JJ. + ZOH

Figure 2-6: Reaction for the free radical mechanism (Chang et al., 1989)

The mechanisms proposed for the methanol to hydrocarbon reaction may be classified in

two groups (Dahl and Kolboe, 1994).

Consecutive type mechanisms:

2C, >C2H4+H20
C2H4 + Q >C3H6
C3H6 +Cl >C4HS....

Parallel-type mechanisms:

-^C2H4
C, ^C3H6

Based on experiments with co-feeding 13C-methanol, ethanol and water over a SAPO-34

catalyst, Dahl and Kolboe (1993) found that only a small part of the propylene formed

was made from addition of methanol to ethanol. This favours the parallel-type

mechanism and on this basis they proposed a modified parallel mechanism based on

(CH2)n known as the "Hydrocarbon pool mechanism" depicted in Figure 2-7.

 16

QM4
A-

CHiOH

[. \„ i saturated hydrocarbons
CiH% coke

Figure 2-7: The hydrocarbon pool mechanism (Dahl and Kolboe, 1994)

Recent studies by Svelle et al. (2006) BJ0rgen et al. (2004) and Haw et al. (2003) have

suggested polymethylated benzenes play an important role in the in the formation of the

light olefins. The polymethylated benzenes are formed in the cages of the catalyst and

act as the reaction sites for the MTO process. It is believed that methanol reacts with

polymethylated benzenes which then splits off olefins after rearrangements under

formation of a lower order polymethylated benzene which then again can be methylated

by methanol. A schematic drawing is given in Figure 2-8

(CH 3 ) S
CHjOH
OLOH

^
(CH,)4 CHA

PA

CJHL (CH 3 ) 4

Figure 2-8: Ethylene formation by reaction of methanol withmethylbenzenes. (Aguayo et al., 2005)
 17

2.4 Reaction conditions

The MTO reaction is highly dependent on the operation conditions and acidity of the

zeolite. Several investigations of the MTO and the related MTG reactions have been

done over the last thirty years in connection to the reaction conditions. The following

section describes the effect of the main variables that influence the reaction and include:

temperature, pressure, water co-feed, residence time and catalyst coking.

2.4.1 Temperature effect

The MTO reaction is very temperature dependent with respect to the product

distribution. At low temperatures of about 300 - 400°C, a high level of aromatic and

heavy hydrocarbons are formed that lie in the gasoline boiling range. This is the typical

temperature range for the MTG reaction. For the MTO reaction, the typical operation

temperature is 400 - 500°C. The selectivity of the olefins generally increases with

temperature. Above 500°C, the formation of methane, CO and H2 begins to increase

significantly (Keil, 1999; Chang, 1984). Dewaele et al. (1999) studied the temperature

effect on the product distribution of olefins over an H-ZSM-5 catalyst. They found that

the ethylene to propylene ratio increases with temperature (from 0.05 at 375°C and up to

0.35 at 475 °C) while the ratio between the higher olefins (C3-C5) remains unchanged.

Shoenfelder et al. (1994) conducted experiments in a riser reactor at 400, 450 and 500°C

with a catalyst containing 25% H-ZSM-5. They found that ethylene production

decreased with increasing temperature while propylene increased slightly. Further, the

overall selectivity towards olefins was increased with temperature.

 18

2.4.2 Pressure effect

Several authors have investigated the effect of pressure on the MTO reaction (Chang,

1984, Dewaele et al, 1999). By reducing the partial pressure of methanol, the reaction of

olefins to higher hydrocarbons is slowed considerably. Figure 2-9 illustrates the effect of

pressure observed during the studies in the late 70'ties at 370°C. The effect of pressure

is significant and it is desirable to run the reaction at a low pressure to optimize the

olefin production.

REACTION COORDINATE »-

Figure 2-9: Product distribution of the MTO reactionat high, low and atmospheric pressure (Chang, 1984)
 19

2.4.3 Catalyst deactivation

Deactivation of the ZSM-5 based catalyst is relatively slow compared to other catalysts

like the SAPO-34, which is due to the ZSM-5 larger pore sizes. The deactivation of the

catalyst can happen by three different mechanisms (Stocker, 1999):

• Deposition of carbon residue both on the outer surface and inside the catalyst

pores,

• Irreversible activity loss due to the effect of steam on the zeolite structure, and

• Structural changes due to high temperature during the regeneration of the

catalyst.

The rate of carbon deposition on the catalyst depends on the acidity of the catalyst and

the operating conditions (temperature, space velocity, feed composition).

A large part of the catalyst coking is due to the formation of higher hydrocarbons that

are unable to escape the catalyst since they are larger than the catalyst pores. It was

found that the acidity (Si/Al ratio) of the catalyst has a significant influence on the

coking rate. By decreasing the Si/Al ratio, the production of higher hydrocarbons

increases and thus increases the coking rate.

Irreversible activity loss occurs because of the dealumination of the catalyst caused by

steam. The level of dealumination is affected by the water content during the reaction
 20

together with the reaction temperature. As the water content and/or temperature increase,

the dealumination becomes increasingly important (Gayubo et al., 2003).

The dealumination caused by the high temperature during regeneration can be avoided

by regeneration of the catalyst below the calcination temperature (Gayubo et al., 2004a).

2.4.4 Effect of water

Water plays an important role in the MTO process: it reduces the deactivation rate due to

coke deposition; it changes the product distribution, increasing the selectivity to light

olefins; and, it increases the dealumination rate, which leads to irreversible deactivation

of the catalyst. The importance of water has been investigated by several authors

(Gayubo et al., 2004a; Aguayo et al., 2005; Wu and Anthony, 2001; Moller at al. 1999;

Campbell et al., 1996).

Campbell et al. (1996) showed that dealumination of the ZSM-5 catalyst during the

MTG reaction is caused by the high water vapour pressure. Moller at al. (1999)

investigated the effect of co-feeding water in a jet loop reactor and found that water

reduced the conversion of methanol probably because it reduces the number of active

sites in the catalyst.

Wu and Anthony (2001) investigated the effect of co-feeding water with methanol at

different water/methanol ratios at 400°C in a fixed bed reactor on SAPO-34. They found

that increasing the mole fraction of water in the feed to 0.74-0.8 increased by eight times
 21

the amount of methanol that the catalyst could process before methanol breakthrough

was observed. Further, the selectivity to olefins was increased. The reason for the

increased lifetime and higher olefin production is attributed to water occupying the

strong acid sites in the catalyst thus preventing olefin oligomerization and coking on the

sites.

The effect of water in the temperature interval 400-500°C on the deactivation was

studied by Gayubo et al. (2004a). Pure methanol and methanol/water with a 1:1 mass

ratio was fed to a fixed bed reactor and after 2h on stream, the catalyst was regenerated.

After two reaction/ regeneration cycles, the acidity of the catalyst was measured by NH3

TPD. They found that irreversible deactivation becomes important above 450°C when

water is co-fed at a mass ratio of 1:1. Irreversible deactivation became evident with pure

methanol feed above 500°C.

2.4.5 Gas residence time

Residence time influences the product distribution of the MTO process. At high gas

residence time, the main products are higher hydrocarbons in the gasoline boiling range.

As the residence time decreases, the product distribution changes favouring the

production of light olefins (see Figure 2-1) (Chang, 1984). At very low residence time

the main product is ethylene or propylene. Dessau (1986) conducted experiments on

ZSM-5 with a very low aluminium content (Si/Al=1670). The experiments were done at

reduced partial pressure of methanol with nitrogen as the diluent. Results showed that

propylene was the main product and that ethylene was formed by secondary reactions.
 22

Chu and Chang (1984) conducted experiments with a ZSM-5 catalyst with higher

aluminium content. At 500°C and low methanol partial pressure, they found that the

main product was ethylene. The residence time during these experiments were, however,

higher than the ones used by Dessau (1986).

2.5 Kinetic

Kinetic models are important in the design of chemical reactors and thus a lot of work

has been devoted to develop suitable models for the MTO/MTG reaction. The models

can be divided into two categories:

Lumped models that group similar components into one including oxygenates,

olefins, heavy hydrocarbons and light gases.

Detailed models that are based on elemental reactions where all species are

accounted for.

The lumped model is more common since they are easier to use than the detailed models

and are often accurate enough for most purposes. In the MTO/MTG process where the

mechanisms for methanol conversion to hydrocarbons are not fully understood, a

detailed model may be equally inaccurate as lumped models.

Several kinetics/reaction paths have been proposed for the MTO and MTG process since

the late seventies when Chang and Silvestri (1977) proposed the first model (Equation 2-

1). The methanol/DME equilibrium is reached very fast and, in the initial studies, the
 23

two oxygenates are treated as a single kinetic species. Methanol and DME are then

converted to light olefins and finally the olefins react to paraffins, higher olefins and

aromatics.

SCffiOH = = ± CHsOCHs ^ ^ ~
+H3O
n= n= pt'rafftn* (v2 _1)
t'a — t>B —* artitnattcs '
fyvlopum f fitiB
Cft~ olefins

Many models have been developed since Chang and Silvestri (1977) proposed their

model. Keil (1999) gives a summary of the model development of both lumped and

detailed models. Chen and Reagan (1979) found that oxygenates disappearance was

autocatalytic over ZSM-5 and proposed the following reactions:

A—*->B
A + B^^^B (2-2)
B—^C

Where A are oxygenates, B olefins and C aromatics and paraffins. Chang (1980)

expanded the model to include ' CH2. It was based on the following conditions and

constraints:

- Methanol and DME are always at equilibrium and can be treated as a single

kinetic species.

Generation of the reactive intermediate is first order in oxygenates.

Consumption of the reactive intermediate is first order in oxygenates.

Olefins can be treated as a single kinetic species.

 24

Disappearance of olefins is first order in olefins.

The conditions led to the following set of reactions:

A —k±->B

B+ C^^^C

Where A is oxygenates, B ('CH2) and C olefins and D aromatics and paraffins.

The kinetics first given by Chang (1980) has been modified by Anthony (1981) and is

given in Equation 2-4 below.

-dA
= klA + k2AB
dt
-dB
= kxA-k2AB-k3C (2-4)
dt
-dC
= k2AB-k4C
dt

Similar models have been proposed by Sedran et al. (1990), Schipper and Krambeck

(1986) and Gayubo et al. (1996). Sedran et al. (1990) proposed three different models

that take the light olefins into account as separate species. In one case the formation of

the olefins is divided into three reactions in a consecutive type mechanism. Oxygenates

react to ethylene and the higher olefins are formed by reaction between lighter olefins

and oxygenates. The model of Schipper and Krambeck (1986) was used in the MTG

process of Mobil. Here, methanol and DME are considered to be in equilibrium and

react to form light olefins. The light olefins can polymerize to form heavy hydrocarbons

and the heavy hydrocarbons can react with oxygenates and light olefins to produce

additional heavy hydrocarbons. Gayubo et al. (1996) proposed two alternative models.
 25

One is a modification of Chen's (1980) model (Equation 3) where the third step is

substituted by B + d (Ci= ethylene + propylene) —>• C2 (higher olefins). In the second

model, oxygenates are absorbed on the acid sites and an absorbed oxygenate reacts with

a non-absorbed oxygenate to form the light olefin. Based on experiments in a fixed bed

reactor at 300-375°C for different contact times, Gayubo et al. (1996) found that the

fitting of the models of Chen and Reagan (1979) and Schipper and Krambeck (1986)

was superior to the others.

Novella et al. (1988) modelled the process with methanol, DME and three lumps:

gaseous olefins, liquid hydrocarbons and gaseous paraffins. DME reacts to olefins and

olefins react to both liquid hydrocarbons and paraffins (two reactions). They considered

the reaction to heavy hydrocarbons and paraffins to be both second order in olefins.

They introduced a temperature dependent induction time on the reaction from olefins to

heavy hydrocarbons and paraffins that represent the necessary space time for the

formation of olefins from the dehydration of methanol and DME. The model shows

good agreement to the experimental results in the 320-420°C temperature interval.

Schipper and Krambeck (1986) introduced a catalyst activity parameter p which

accounted for both activity loss due to coking (a) and permanent deactivation (a).

P = a-a (2-5)
 26

The reaction rate, r;( of the different lumps/components are then calculated as the product

of the reaction rate with a fresh catalyst, r;o, and the catalyst activity parameter given by

Equation 2-5.

rt=P-rm (2-6)

The permanent deactivation kinetic is:

^ = -Ka-ah (2-7)
at
The activity loss due to coking of the catalyst is:

^- = -kd(a-a)d (2-8)
at
The total deactivation, p, can then be found to be:

^ = a[- kd (a • a)d - Kaah~2 {a • a)] (2-9)

dt

Benito et al. (1996) studied catalytic deactivation and found that a composition

dependent deactivation kinetic expression was necessary to characterize of the MTG

process. In their work, activity loss due to coking was calculated by:
^ = [Z(«)V (2-10)
Where ka; is the kinetic constant for deactivation by coke deposition for lump i and a is

defined as fl = -f

Gayubo and co-workers presented a kinetic model for the MTG process with eight

reactions involving six components/lumps including coke and water (Gayubo et al.,

2001; Gayubo et al., 2004b). The effect of water was included in the reaction kinetic as a
 27

term in the denominator in the kinetic equations of the form (l+k w Xw). By including

this term, agreement between the model prediction and experiments was improved with

pure methanol feed as well as with feed consisting of methanol/water mixtures.

The recovered activity in the regeneration of the catalyst was also investigated and the

following empirical equation was proposed:

a0 = 1 - (1 - a)exp[- (b,tc + b2t2c )j (2-11)

Where ao is the initial activity after the regeneration, tc is the combustion time during

regeneration. The parameters a, h\ and hi were found to be:

a-5.00(±1.43)-10" 3
bx =1.38(±0.1l)l0- 1
b2 =6.13(±0.50)-10"4

Gayubo et al. (2003) included the activity loss by coke deposition and irreversible

deactivation for the MTO process, the reaction kinetics were based on their previous

work mentioned above. The activity was defined similar to Equation 2-5. Several

models for the irreversible deactivation were studied. The most suitable model was

given by:

- ^ = kdirXla (2-12)

The model was experimentally verified in three successive reaction-regeneration cycles

with feed consisting of 50% MeOH and 50% water. Results form the simulation is

shown in Figure 2-10.

 28

tlmn on strum, It timo on stream, h time on stream, h

Figure 2-10: Evolution with time on stream of the composition of the lumps of the kinetic scheme, in
three successive reaction-regeneration cycles in fixed bed. Reaction conditions: temperature, 773 K, space
time, 0.093 (kg of catalyst) h (kg of methanol) '; water content in the feed, 50 wt%. Points: experimental
results. Solid lines: calculated (Gayubo et al, 2003)

Most models describe the trends in the MTG/MTO process but fail to characterize the

individual olefin components, which are important in the MTO process. Schoenfelder et

al. (1994) developed a lump model given in Figure 2-11 for the MTO process. The

catalyst used was based on a ZSM-5 zeolite. Based on experiments run in a Berty reactor

at temperatures between 400 - 500°C, the kinetic constants were found at 5 different

temperatures. The parameters found were in agreement with the Arrhenius law. The

model was used in conjunction with a one-dimensional circulating fluidized bed model

in the calculations of the MTO process in. The model was able to predict the product

distribution fairly well on fresh catalyst but is unable to characterize the effect of time on

stream since coke deposition on the catalyst was not included.

 29

2 CH3OH *> CH3OCH3 + H2O

A 2 A — ^ B + 2W

B CH2 = CH2 C
5 CH 4l CO,
A+B-^-^C +W

C
3F
CH2 = CH-CH 3
~ H2, H 2 0
A+C D+W

3£ 3B — ->2E
W
C "<4
—»E
D CH2=CH-CH2-CH3
1 1 • • 3D ^ >4E
\7 XT
paraffins (C2H6, C3H8, C4H10), aromatics -^3F+W
2A
Figure 2-11: Scheme and kinetic model for the MTO-reaction by Schoenfelder et al. (1994)

Bos et al. (1995) proposed a kinetic model for a SAPO-34 based catalyst. The model

was developed on fixed bed experiments conducted over a very short time span to avoid

coking. To include the coking effect the experiments were done on samples containing

3.7, 8.9 and 12.3 wt % coke. The reaction scheme developed is given in Figure 2-12.

From experimental data with butylenes as feed gas and model predictions, it was

concluded that the reaction from ethane to coke should be excluded form the reaction

scheme (ki3=0). The final model consisted of 6 components lumps plus coke described

by 12 reactions. The final model is given in equations 2-13-2-16.

 30

MEOH
1
-** m # m
-*~ nMmm s t
4s *^-4l
-2—•*• propen» J i , * «*«
4 •* 12
propane 11

•** mtmm
•*- sura OS

•** coke

Figure 2-12: Reaction scheme for the MTO reaction by Bos et al. (1995)

r
i — ^iXMeOH^ i ~ 1 7 (2-13)
r8 — ksxCi=FxMeOHF (2-14)

rt=k,xCi=P / = 9 -11 (2-15)

Four different correlations were investigated for the coke dependency of the reaction.

They found that an exponential dependency (Eq. 2-16) best represented the coking

effect.

ki{c) = k*e-a'c (2-16)

where a; is an empirical deactivation constant and c is weight percentage of coke.

Detailed kinetic models have been proposed that describe the production of each

component in the process (Park and Froment, 2001a; Park and Froment, 2001b; Mihail

et al., 1983a; Michail et al., 1983b). Mihail et al. (1983a) developed their model using

the following approach:

 31

• Stoichiometric matrix of all the components found in the experimental results

• Choice of a set of linear independent reactions

• Reactions are deleted or supplemented in accordance to data in the literature

• Kinetic parameters are fitted to experimental data

In the stoichiometric matrix, they included 16 components (up to C5) with the rank of 3

(C, H, O), from that 13 independent reactions were chosen. By deleting and

supplemented these reactions, the final model comprised of 27 reactions. Mihail et al.

(1983b) expanded the model to higher hydrocarbons (olefins and paraffins up to C7 and

methylated benzene up to C12) which resulted in a reaction network with 53 reactions.

Maria and Muntean (1987) simplified the olefin model of Mihail et al. (1983a). With the

use of numerical methods and experimental data, the reaction model was reduced by

eliminating reactions with limited impact on the final result. This reduced of the kinetic

model to only include 14 reactions.

Park and Froment (2001a) developed a detailed kinetic model based on elementary steps

for the MTO process over an H-ZSM-5 catalyst. The elementary steps to describe the

formation of the primary olefins production included 8 different mechanisms due to the

large number of possible reactions for the formation of the first C-C bond. The total

number of parameters in the reaction network amounted to 500. The number of

parameters was reduced with the use of the single-event concept and the Evans-Polanyi
 32

relation and thermodynamic constraints. This reduced the number of parameters to 33.

The parameters were obtained by optimizing the model to experimental data with the use

of sequential quadratic programming and the Levenberg-Marquardt routine. Through

this, the best mechanism for the primary product formation was found to be between

oxonium methyl ylide and dimethyloxonium ion (Park and Froment, 2001b).

2.6 Processes

The MTO reaction has mainly been investigated in fixed beds (Chu and Chang, 1984;

Park and Froment; 2001b; Gayubo et al., 2003; Stocker, 1999) but due to the high

exothermicity of the reaction and coking rate of the ZSM-5 and SAPO catalysts, a fixed

bed may be less suited compared to a fluid bed. In the case of fast deactivation for an

industrial process, the regeneration should take place in the same reactor with successive

reaction-regeneration cycles or circulation of the catalyst between a reaction and a

regeneration reactor (Gayubo et al, 2000). The exothermic reaction can be a problem in

a fixed bed which will operate under non isothermal conditions resulting in hot spots.

Several researchers have reduced this problem in fixed bed by using a dehydration

reactor before the MTO reactor hence reducing both the energy production and coking in

the MTO reactor (Sapre, 1997; Tabak and Yurchak, 1990).

A fluidized bed or circulating fluid bed (CFB) will be a more obvious choice due to the

advantages it offers with respect to high heat transfer between gas and solid combined

with rapid solid mixing, insuring isothermal conditions. Furthermore, a CFB offers
 33

continually regeneration of the catalyst while avoiding the need for several reactors in

series.

Tabak and Yurchak, (1990) give the product distribution of fluid bed experiments

performed by Exxon-Mobil in their pilot plant. The operation conditions were 482°C

with partial pressure of methanol of 100 kPa in the feed and the reaction was run to

almost complete methanol conversion. A breakdown of the product distribution in their

experiments is shown in Table 2-1. They consider the fluid bed reactor to be the best

suited for the MTO reaction since it allows for steady-state operation at maximum olefin

production at complete methanol conversion and makes it easy to remove the reaction

heat with the use of heating coils in the bed.

Table 2-1: Product distribution of the MTO reaction in fluid bed at 482°C and 102 kPa methanol partial
pressure (Tabak and Yurchak, 1990)
Product Ci C2 C3 C4 C2= C3= C4= C5-C11 oxygenates

Wt % hydrocarbons 1.4 0.3 2.3 3.9 5.0 31.8 19.6 35.7 0.3

Ortega et al. (1998) have studied the lumped product distribution (oxygenates, light

olefins and other hydrocarbons) over a HZSM-5 containing catalyst. The properties of

the catalyst are given in Table 2-2. The experiments were performed in a 30 mm internal

diameter fluid bed with solid circulation.

 34

Table 2-2: Properties of the H-ZSM-5 and the catalyst used in the
experiments of Ortega et al. (1998)
HZSM-5
zeolite catalyst*
Sl/Al ratio 24
Bron^r pd/Lmft is J alio 2.9
ci\&tall»iltv 97%
f R^tol -.17? ,«H1 6,3
pai tide slzt-. mm 0.3-0.5
appaient densin g«».m-3 0;94 L21
BET Mirtat'* aiea. m 2 -g -1 420 124
poie \olunu-, fm 3 , g _ 1 0.65 0.43
rnia npote \oltnii'* rm' 1 g _1 0,17
(^QHiof diatnetei * 0 7 tunl
poi»> lolunu- distribution of thr-
eat alyst vol 4.
4 --3 io- 3 «m2 si
10~ to lU- mn 14 7
10--to2um 77 2
zeolite aclditv
rn^a^ntc-mi-nts NHa fert-butylamlne
tutal dddlty tmniolofbasc-1'12-' 0,51 0„46
temperature peaks In the TPD 422 °C 304 °C
" Ci'mpositiuu zeolite-, 21 v.t %* bentonlte, 30 wt %• alumina,
45 wt '<>

4#a
. a MeOH+OME •
» lighi ofsifos
®M 420
~~.:~r»^.

:
0,6 •

fl,4
L 400

380

r
Or ^—t&-"~S8—&
0.2 3BQ
1

340
tt1234S6?ii10 It t2
time, h

Figure 2-13: Evolution of the mass fractions of each lump, until steady state is reached at three different
temperatures, space time 0.040 gcath/gMeOH and average solid residence time 0.866h_1 (Ortega et al., 1998)

The reaction was carried out between 380 and 420°C at space times between 0.022 to

0.0040 gcath/gMeOH- The product distribution as a function of time is shown in Figure

 35

2-13. The authors attribute the initial approach to steady state to increasing catalyst mass

in the fluid bed while the approach to equilibrium when the temperature is increased is

due to increased catalyst deactivation.

Gayubo et al. (2000) continued the investigation of the effect of residence time on

reaction and regeneration operations on the process together with the space time and

temperature. The regeneration was conducted at 550°C. By optimizing the light olefin

production based on catalyst residence time, it was found that the time on stream during

the reaction and regeneration is dependent on each other. Figure 2-14 shows a contour

plot of the results. The production rate is highest when the residence time in the reactor

is about 3 times longer than in the regenerator. Experiments conducted with increasing

methanol flow rate at temperatures between 380 and 420°C showed that the light olefin

production was substantially increased with decreased residence time and that the effect

of temperature was highest at low residence times.

06 06 1 1.£ 14 16 II "J
*i. h

Figure 2-14: Contours of the relative production rate of light olefins for different residence times in the
reactor and regenerator. Temperature 420°C space time 0.021 gcath/gMeOH (Gayubo et al. 2000)

Schoenfelder et al. (1994) investigated the MTO reaction in a CFB consisting of a riser

made of six tube sections with an inner diameter of 30 mm and a height of 11m, a
 36

stripper and a regenerator. They used catalyst consisting of 25% H-ZSM-5 with a Si/Al

ratio of 400 the mean Sauter diameter of the fluid bed catalyst was lOOum. To derive the

kinetics, experiments were conducted in a Berty reactor system. The riser was modelled

using four zones; bottom zone described by a bubble phase and a suspension phase, a

splash zone modelled as a CSTR, a recirculation zone described by a lean phase and a

dense phase and an exit zone modelled as a CSTR. Experiments in the CFB were

conducted at 400, 450 and 500°C and at different space velocities. The product

distribution obtained consisted mainly of olefins and at 500°C the olefin yield reached

97% and went through a maximum propylene production. Figure 2-15 compares the

product distribution of the experiments at 500°C against the predicted yield from their

model. The model predicts the overall behaviour of the system but some deviations are

evident especially at low WHSV"1 and it does not catch the maximum in propylene yield

seen from the experimental results.

 37

Figure 2-15: Comparison between measured (symbols) and predicted (lines) product distribution of the
CFB reactor at a temperature of 500°C (Schoenfelder et al, 1994)

2.6.1 Industrial Processes

There are currently three companies that offer/have the technology for large scale

production of methanol to olefin (MTO). They include UOP/Hydro, Lurgi and Exxon-

Mobil. The strategy of the three companies are quite different: Lurgi offers fixed bed

technology using a ZSM-5 zeolite in there catalyst; UOP/Hydro offers a fluid bed/riser

process where the catalyst is continuously regenerated, the zeolite used is SAPO-34;

and, Exxon-Mobil has many patents on MTO processes, they are concentrated around

high velocity riser reactors with catalyst regeneration in a fluid bed, like UOP/Hydro,

they also use the SAPO-34 zeolite catalyst. Further description of the processes is given

below.
 38

[Link] Lurgi Process

For several years Lurgi and Statoil have run a demonstration unit in Norway. This has

led to the first contract on an industrial scale MTP unit that is to be built in Iran for the

Fanavaran Petrochemical Company. The plant will have an annual capacity of 100,000

MT of propylene (Lurgi, 2005).

The Lurgi MTP process relies on fixed beds where methanol is first partially converted

to DME. From the pre-converter a split stream is sent to the first fixed bed reactor of

three which is run in series. The remaining feed from the pre-converter is sent to the

second and third reactor. After purification and separation, the unwanted olefins are re-

circulated to the first reactor together with some water. A schematic diagram of the

process is given in Figure 2-16. The final product consists of approximately 70%

propylene, 20% gasoline and some LPG and fuel gas. The process is based on a ZSM-5

zeolite from Siid-chemie and the process is operated at slightly elevated pressure (about

0.3 - 0.6 barg) and in a temperature range of 420 - 490°C (Ondrey, 2005). Due to

coking, the fixed bed reactors have to be regenerated and one extra reactor is needed for

continuous production.
 39

Figure 2-16: Flow diagram of Lurgi MTP® Process (StQcker 2003)

[Link] UOP/Hydro Process

UOP and Norsk Hydro have put over 10 years of development into the MTO process

and have been operating a 1 ton methanol a day demo unit in Norway since 1995. The

technology is based on fluid bed technology with continuous catalyst regeneration in a

separate fluid bed. A schematic drawing of the process is given in Figure 2-17. The

catalyst is based on a SAPO-34 catalyst designated MTO-100, with a pore size of 3.8A.

The operating temperature is 350-550°C and the operating pressure is 1-3 barg. The

olefin yield is around 80% where about 10% is butylenes. The ethylene to propylene

ratio can be varied from 0.75 to 1.5 depending on process conditions, with the highest

total olefin yield at a ratio of 1. Based on the demonstration unit, catalyst tests and
 40

process calculations, the process can be scaled up to 1000 kt olefin a year (Houdek and

Andersen 2005;Andersen, 2003).

UOP HYDRO MTO Process

Quench Cam tin C. C
Reacttn' Regeisetaloi "fairer De-C3 De-C; SpKttet- P'-fj <jpl,im I>e-C4

TaU&a:

Reges 'Cat

[Link]) el
1\ at*r

Figure 2-17: Simplified flow diagram of the UOP/Hydro MTO Process (Houdek and Andersen 2005)

[Link] Mobil Process

Exxon-Mobil has a long history in MTO and has completed a lot of research in the field

over the last three decades. They have run a 4000 t/year fluid-bed demonstration plant at

Wesseling (Germany), which had previously been used to demonstrate their methanol to

gasoline (MTG) process (Stocker, 1999).

In recent years, Exxon-Mobil has concentrated on process design and optimization,

which is reflected in the patent literature where several different processes for the MTO

reaction have been patented with the use of both ZSM-5 and SAPO based catalysts

(Chisholm et al. 2003, Beech et al. 2004, Coute et al. 2004, Lattner et al. 2005).
 41

Chisholm et al. (2003) patented a process where the MTO reaction is carried out in a

riser reactor; the gas/solid mixture from the riser is then separated in a fluid bed with

internal cyclones by means of gravity and the cyclones. Part of the catalyst is re-

circulated to the riser while another part is sent to a stripping zone where the absorbed

hydrocarbons are stripped with an inert gas. The stripped catalyst is then sent to a fluid

bed where it is regenerated and then re-circulated to the riser. The space velocity in the

riser is preferably above 4 m/s and the WHSV in the range of 20 h"1 to 500 h"1. The

partial pressure of the feed is preferably between 0.2 to 5 bars and the temperature

between 350°C and 550°C and a feed stock conversion of 75-95%.

The reason to recycle some of the catalyst without regenerating it is because the coke

loading should be 1-2 carbon atoms pr. acid site for optimal operation. The catalyst

should be regenerated at 550 - 700°C and at a pressure between 2.07 and 4.14 bars. The

oxygen containing stream consists of 0.01 to 5% of oxygen and the residence time is

most preferably between one and 100 minutes.

Beech et al. (2004) have patented four different riser reactor configurations with external

regeneration. They all include a feed inlet at the bottom of the riser and exit through the

top which is inside a vessel. The separation of the gas and solids are done by both

gravity and cyclones. In one case the riser reactors are located near the side of the

separation vessel while in the other cases the riser reactors are centred in the vessel. A

part of the catalyst is sent to regeneration in a fluid bed and then reintroduced into the

vessel. The remaining catalyst is led to the bottom of the riser reactors. To control the

catalyst flow, solid flow regulators are used.

 42

Chapter 3
Objective and methodology
As described in the literature study, a significant amount of research has been conducted

on the MTO process. In general, the light hydrocarbons have been lumped together both

experimentally and in the kinetic analyses, which limit the utility of this research with

respect to optimizing the olefin product mix (that is, ethylene versus propylene or

butylene). A model capable of predicting the light olefin yield is crucial for both the

design of down-stream product separation but also for the product value. Some models

include the light olefins as separate species, like the model of Schoenfelder et al. (1994),

but don't take the higher olefins produced in the ZMS-5 zeolite into account. The model

proposed by Bos et al. (1995) includes deactivation from coking but it only considers

ethylene and propylene as the light olefins since it is based on the SAPO-34 zeolite. For

the ZSM-5 zeolite, the catalyst system used in this study, propylene and butylenes are

the main products.

Kinetic studies in fluidized bed systems represent significant experimental and

modelling challenges. Although the bed temperature is uniform and the solids are

completely backmixed, the gas phase hydrodynamics are complex: bubbles form at the

grid and rise through a gas-solid emulsion at elevated velocities. The challenge will be to

establish experimental conditions that minimize these phenomena.

 43

3.1 Objective
The objectives of this thesis can be formulated as:

• Investigate the mechanism of the MTO process in order to,

• Develop a kinetic model for the methanol to propylene process in a fluid bed

with respect to conditions such as temperature, space velocity and feed

composition. The model must include the main olefins: ethylene, propylene and

C4 as separate components.

• Determine the kinetic parameters of the proposed reaction kinetic for the MTO

reaction

To reach the objectives the following tasks have been preformed

Kinetic study of the methanol to DME reaction in fluid bed

The kinetic of the methanol to DME reaction was studied in the fluid bed. The reaction

kinetics over a ZSM-5 based catalyst were studied in the temperature range of 250 -

325°C where and at a wide range of WHS V and feed compositions. A kinetic model was

derived and this model was coupled together with hydrodynamic models to characterize

the fluidized bed.

 44

Catalyst investigation in fixed bed

The catalyst will be studied in fixed bed to evaluate the effect of residence time,

temperature and feed composition on the product distribution and deactivation. Special

focus was put on the effect of space velocity on the deactivation rate of the catalyst.

Kinetic study of the MTO reaction in fluidized beds

Kinetic experiments were conducted in a fluidized bed to investigate the effect on

temperature, WHSV and feed composition. 1 -hexene was feed to the reactor to examine

the reaction network. The goal was to gain a better understanding of the reaction

pathways and with this understanding develop a kinetic model for the MTO reaction

which takes into account the main olefins, ethylene, propylene and C4 as separate

components.

3.2 Methodology
3.2.1 Catalyst
The catalytic system used in this work was based on a ZSM-5 from Zeolyst designated

CBV 28014 with a Si/Al ratio of 140. The ZSM-5 zeolite was imbedded in a

silica/alumina matrix during spray drying and impregnated with 1.5 wt % phosphorous

to reduce methane formation. A description of the catalyst is given in Appendix A.

3.2.2 Fixed bed experiments

The experiments were conducted in a 9 mm inner diameter quartz reactor into which a 3

or 6 mm inner diameter quartz reactor could be inserted. The vessel was heated

electrically and the temperature was regulated with a PID controller. The reference
 45

temperature was measured at the wall and a second thermocouple monitored the

temperature just below the catalyst bed. The feed stream was prepared by passing

nitrogen through two bubble flasks containing MeOH: the first was at room temperature

(approx. 22°C) and the second at 16°C to ensure that nitrogen was saturated by methanol

yielding 10% methanol in the feed. The product stream was analyzed by FID on a

Hewlett Packard 5890 series-II GC. Equipped with a pre-column (Porapak Q 80/100

mesh size - 0.5 m x 2.16 mm ID x 1/8" OD packed column in stainless steel) followed

by a capillary column: CP-PoraPLOT A1203, KC1 - 10 m x 530 um x 5 urn. For some

of the experiments a GC with a TCD measured the CO/CO2 and other light gases.

3.2.3 Fluid bed experiments

Methanol to DME

Methanol to DME experiments were preformed in a 4.6 cm inner diameter glass

fluidized bed a description of the system is given in Chapter 4.2.3. Other fluidized bed

reactors have also been used, mainly for the MTO reaction, but methanol decomposition

in the distributor was found to be significant. Further description of the other fluidized

bed reactors are given in Appendix B. The feed stream consists of 5, 15, 30 and 33

mol% methanol and in argon which was preheated before it enter the fluid bed. The

superficial gas velocity in the fluid bed was between 0.5 and 8.0 cm/s. The velocity was

kept low to ensure that catalyst stays in the fluid bed and to avoid changing the

hydrodynamic significantly. To have a higher span of WHSV's the catalyst inventory

was varied between 25 and 200g. Due to poor heating just above the distributor plate in

the fluid bed the catalyst was elevated into the heating zone. Glass beads with a
 46

minimum fluidisation velocity above the maximum gas velocity were used to elevate the

catalyst into the heating zone.

Product distribution was monitored online on a MS with reference measurements by GC

to help calibrate and determine uncertainties on the MS measurements. Using a MS has

several advantages including fast response to changes in product distribution and

determination of by products.

MTO experiments

To establish when deactivation begins to influence the product distribution during

operation long term experiments (one day) have been conducted in order to be able to

design experiments where deactivation do not influence the results. The effect of

regeneration was investigated and the coke content of the catalyst was measured in a

thermal gravimetric analyser (TGA)

To gain insight into the kinetics of the MTO reaction, experiments at different flow rates

(1.3 - 10 cm/s) and temperatures (400 - 550°C) were conducted and the catalyst bed

height was varied to obtain a larger range of residence times. The effect of water and

dilution with argon were investigated by using different feed compositions: pure

methanol, methanol/water and methanol/argon mixtures. 1-Hexene was also used both

as feed and co-feed with methanol to gain insight into the reaction mechanism. By only

feeding 1-hexene the reaction pattern of the larger olefin can be investigated which will

be important in the development of a kinetic for the MTO reaction.

The product distribution was analysed by GC and trends in the reaction monitored on

MS.
 47

3.2.4 Kinetic model

Based on the fluid bed experiments and literature knowledge a kinetic model for the

reaction is proposed. The model is based on the hydrocarbon pool mechanism with a

high molecular weight molecule, Cx+ as the hydrocarbon pool species. The model is

implemented together with a fluid bed model with a bubble and emulsion phase where

all catalyst is considered to be in the emulsion phase.

 48

Chapter 4
MeOH to DME in Bubbling Fluid Bed:
^
Experimental and Modelling

4.1 Presentation of the article

The article looks at the modelling of a bubbling fluidized bed with a relatively well

known reaction methanol to DME over a ZSM-5 containing catalyst. Experimental data

for the methanol to DME reaction is obtained in a small fluid bed reactor (4.6 cm ID) at

250°C - 325°C with different feed mixtures of methanol and argon. The data is analysed

with the use of MS and GC. A reaction kinetic model is proposed for the methanol

dehydration reaction and its performance is compared to the literature model of Bercic

and Levee (1992). The fluidized bed is modelled with both a detailed fluid bed model

and a CSTR in series model. Coupling the proposed reaction kinetic with the CSTR in

series model gives a superior performance in the prediction the conversion in the

fluidized bed.

The fluid bed is shown to be a viable reactor type for kinetic measurements of the

exothermic dehydration reaction.

This article was submitted to Canadian Journal of Chemical Engineering

 49

4.2 M e O H to D M E in Bubbling Fluidized Bed: Experimental and

Modelling
Mads Kaarsholm , Finn Joensen1^ Roberta Cenni1^ Jamal Chaouki, Gregory

S. Patience

Department of chemical engineering, Ecole Polytechnique de Montreal, Canada

Haldor Tops0e A/S, Denmark

4.2.1 Abstract

Methanol dehydration over a ZSM-5 containing catalyst in a fluidized bed reactor at 250

- 325°C gave conversion of 30 - 100% to the equilibrium. Side reactions were

negligible at low temperatures while hydrocarbon formation was more significant at

325°C. Online gas analysis by MS provided real time measurements that allowed for fast

determination of steady state conditions.

Residence time distribution (RTD) measurements of the flow pattern are performed and

are shown to mimic plug flow at low catalyst amount with increases dispersion as

catalyst is added. At the highest catalyst loading the RTD measurements can be

modelled by a n-CSTR in series model with 6 CSTR's. Both a detailed fluid bed model

and the CSTR in series model have been used to characterise the hydrodynamics.

Dehydration reaction kinetics are proposed that include the reverse reaction and is

compared to a literature model. Coupling n-CSTR model with the proposed kinetics

* Corresponding auther. Tel: + 45 45 27 22 55

E-mail address: mkaa@[Link]
 50

gives a superior fit and good predictions at other reaction conditions. The fluid bed is

shown to be a viable reactor type for kinetic measurements of the exothermic reaction

and other exothermic reactions where hotspots or temperature gradients are a concern in

fixed bed could benefit from fluid bed kinetic measurements.

Keywords: Methanol dehydration, DME, Fluidized bed, n-CSTR in series, Reaction

kinetics

4.2.2 Introduction

The dehydration of methanol over an acidic catalyst is an important reaction for the

production of dimethyl ether (DME). DME is considered as one of the best alternatives

to diesel fuel with decreased NOx, CO and hydrocarbon emissions with both lower

particle emissions and global warming potential than conventional fuels (Semelsberger

et al., 2006; Gray and Webster, 2001). The reaction is also important as the first step to

produce olefins (MTO) or gasoline (MTG) (Keil, 1999).

Several authors have investigated the dehydration reaction over different catalysts e.g.

alumina, silica-alumina or acidic iron exchange resins. Most kinetic expressions are

based on either the Langmuir-Hinshelwood or the Eley-Rideal mechanisms and takes the

adsorption of water and methanol into account but leaves out the contribution of DME

since the absorption constant is insignificant compared to the others. A summary of

models can be found in Bercic and Levee, 1992 and Mollavali et al, 2008 who both

studied the intrinsic reaction kinetics. Almost all of the kinetic expressions assume

irreversible reaction and only a few incorporate the reversible reaction to account for the
 51

decrease in reaction rate close to equilibrium which is critical for most industrial

applications.

Often reaction rate kinetics is measured in reactors under differential conditions. The

main advantage of the fixed bed is that the flow pattern can be assumed to be ideal-plug

flow and a simple well described reactor model can therefore be used to model the

reaction together with the reaction kinetics. Fixed beds do have some limitations that

must be considered. In reactions where deactivation is important the catalyst

deactivation is likely to progress down the bed resulting in a non uniform activity down

the bed (Kaarsholm et al., 2007). For highly exothermal reactions temperature gradients

and/or hot spots are problematic and the exact reaction temperature can be difficult to

measure precisely. For intrinsic measurements the catalyst particles have to be small to

avoid transport gradients but this leads to higher pressure drop over the bed. Fluidized

bed reactors have the advantage of isothermal operation due to rapid heat transfer by

solid circulations, limiting temperature uncertainties considerable compared to fixed

bed. Due to the solids back-mixing the catalyst have a uniform state e.g. the deactivation

of the catalyst is uniform as is the temperature. The greatest drawback of the fluidized

bed is the gas flow which is less than ideal due to the presence of bubbles. The effect of

bubbles has been investigated for several decades and a large number of mathematical

models have been proposed to model the hydrodynamic of the fluidized bed (Kunii and

Levenspiel, 1991).

Over the years many models for the bubbling fluidized bed have been proposed

including single phase models with axial dispersion which have been shown to work
 52

well at low gas velocities (Lorences et al. 2006) and multiphase models. Horio and Wen

(1977) divides multiphase models into three groups: level 1 is described as two phase

models with several adjustable parameters that are not directly related to the bubble side

this include the Van Deemter model (Van Deemter, 1961) and the model of May (1959).

Level 2 models relate the bed parameters to the bubble size that is treated as a constant

or a fitted parameter. A model that is included in this group is the model of Kunii and

Levenspiel (1968) where the bubble diameter is measured by frequency probes. The

third level of the bubbling models include the bubble diameter as a function of height

and may also include grid region and freeboard effects. Many of the models proposed in

the recent years are included in this level and include the models of Werther and Hartge,

(2004), Christensen et al. (2008) and Radmanesh et al (2006). Hetsroni (1982) shows

that two phase systems with plug flow in the bubble phase and either plug flow or a well

mixed reactor model for the emulsion with mass transfer between the phases can give

reasonable predictions of fluidized bed reactor performance.

The mass transfer between the bubble and emulsion phases is a key parameter for the

fluid bed reactor models with regard to both overall reactor effectiveness and in reactor

scale-up. Two different models for the interphase mass transfer coefficient between the

bubble and emulsion phase are usually used either it is calculated by a correlation of Sit

and Grace (1981) or the interphase mass transfer between the bubble and wake and

between wake and bubble is calculated separately to get the overall interphase mass

transfer (Kunii and Levenspiel, 1991). Sit and Grace (1981) developed a correlation for

the interphase mass transfer coefficient between the bubble and emulsion phase based on
 53

studies of bubble interaction and coalescing in a 2D experimental setup which was then

extended to three dimensions.

Shaikh and Batran (2007) have further expanded the mass transfer term by including

film resistance on the bubble side alongside the mass transfer coefficient between the

emulsion and bubble phase. They conclude that this term is important when the reactant

is in a mixture or for fast reactions.

In this article, the methanol to DME reaction in a gas-solid bubbling bed will be

investigated at a range of gas velocities and bed heights. The reaction was studied in a

fluid bed reactor for several reasons. The catalyst used is a fluid bed catalyst and to

avoid a high pressure drop over a fixed bed the catalyst have to be pelletized and

crushed to increase the particle size which could change the surface properties of the

catalyst. Hot spots or a reaction front with a resulting temperature gradient down the

fixed bed is also a concern due to the exothermic reaction. These problems are avoided

in a fluid bed and thus better controls of the experiments are possible. Low gas flows are

used in the fluidized bed to operate as close to ideal gas flow as possible and RTD

measurements are conducted to characterise the flow. The kinetics of the reaction are

investigated and both a two phase fluid bed model and an n-CSTR's in series model is

used. The goal is to look at the feasibility of doing the kinetic study in a fluid bed.
 54

4.2.3 Experimental

[Link] Equipment
The experiments were carried out in a 46 mm inner diameter quartz fluid bed (Figure

4-1) with a 66 mm inner diameter disengagement zone.

Thermocouple

Figure 4-1: Drawing of the glass fluidized bed with glass beads, catalyst and thermocouple

The gas is distributed through a quartz frit. The reactor temperature is maintained by two

electrical band heaters, one located in the fluidized bed zone and one in the

disengagement zone. A thermocouple is inserted in the middle of the reaction zone and
 55

the band heaters are controlled by temperature measurements on the outside wall of the

reactor. Liquid methanol is fed from a dual piston pump and the gas is metered by one of

two gas inlet lines both controlled by a Brooks mass flow controller (MFC). The inlet

stream is preheated to 150°C to ensure liquid feed is entirely vaporized. The effluent gas

was analysed by a Hiden mass spectrometer QIC-20 (MS) and a Varian CP3800 GC. A

diagram of the system is shown in Figure 4-2

_/SC\

Jc^ -bJ-

rm

P-M- 1 iJfi—nr\—I— -hsh

Ws FrT^i Trim

Figure 4-2: Schematic diagram of the experimental setup

[Link] Catalyst

The catalyst is composed of 10% CBV28014 (Zeolyst) imbedded in a Si/Al matrix

consisting of Catapal B, Levasil 100s/30% and kaolin which was spray dried then

calcined in air at 550°C for 4 h. The powder was contacted with a (NFL^HPCv solution

and then dried and calcined so that the resulting catalyst contained 1.5% phosphorous.

The particle size distribution was measured on a Horiba LA-950 and the mean particle

diameter was measured to 108 um. The minimum fluidization velocity at ambient
 56

temperature and pressure was measured to 0.005 m/s in a 76 mm perspex column in

which the tapped particle density and density at minimum fluidization were also

measured. The catalyst properties are listed in Table 4-1.

Table 4-1: Particle properties of the: catalyst

dp <45 0.4 200 < d p > 229 2.9
45 <dp< 68 4.7 229 < d p < 262 1.6
68 <dp< 89 16.6 262 < d p < 350 1.3
89 <dp< 102 15.5 dp 108 urn
102 <dp< 117 16.7 Umf 0.0051 m/s
117 <dp< 133 15.3 Pp 1270 kg/m3
133 <dp< 153 11.9 Ptapped 855 kg/m3
153 <dp< 175 8.1 Pmf 744 kg/m3
175 <dp< 200 5.0 £mf 0.405

[Link] Experimental conditions

Glass beads with a particle diameter of 500 urn were placed at the bottom of the reactor

and the catalyst was loaded on top to ensure that the catalyst was adjacent to the band

heaters. No mixing of the two solids occurred since the maximum gas velocity used was

insufficient fluidize the glass beads. The feed gas consisted of 5, 15, 30, 33 mol%

methanol in argon. The superficial gas velocity ranged from 0.45 - 8.4 cm/s and the

experiments were conducted with 4 different catalyst loadings 25, 50, 100 and 200g.

Experiments ended when the product distribution was stable over a period of a few

minutes on the MS.

The MS was calibrated with different known mixtures of water, DME, methanol and

argon in the same range as the expected product distribution and the exit composition

was also measured on a GC to verify the MS. The retention times of the components
 57

have been validated from binary gas mixtures of argon and the gas of interest except in

the case of methanol which was done from a methanol air mixture

RTD measurements were performed on the reactor loaded with 540g of glass beads, 50g

of catalyst on top of the glass beads and a third experiment without glass beads and with

220g of catalyst. The experiments were carried out with argon as the tracer gas by

switching between air and argon which were controlled by two MFC's. To minimize

perturbations the change in gas flow is done by a multiport valve and the gas that is not

sent to the reactor is purged. Pressure differences between the two inlet lines are hereby

kept at a minimum and MFC delay is avoided. Switching the multiport valve gives a

heavyside unit step function which first derivative is the Dirac delta pulse. The change in

exit gas compositions was recorded at the exit of the reactor by MS at a frequency of 4.4

Hz

4.2.4 Experimental results

[Link] RTD

The E-curve from the RTD measurements is shown in Figure 4-3. There is almost no

difference in the results form RTD measurements A and B in Figure 4-3 and the effect

of adding 50g of catalyst to the 540g of glass beads already in the reactor does not

change the flow significantly. Using 220g of catalyst gives a noticeable change in the

flow compared to the using only the glass beads (Figure 4-3 C). An increase in the mean

residence time is observed which is due to a lower volume of catalyst compared to glass

beads. The broadening of the peak shape when 220g of catalyst is used is due to more

dispersion and less plug flow behaviour as the amount of catalyst is increased.
 58

A n-CSTR model is used to describe the system showing a similar result for the

experiments with glass beads and glass beads plus catalyst. With 220g of catalyst in the

bed only half the number of CSTR's is needed to describe the system. The result of the

n-CSTR model is given in Table 4-2. By subtracting the result from reactor with glass

beads from the data with 220g of catalyst and accounting for the volume of glass beads

the Pe number for the bed can be found to 10 which give 6 n-CSTR in series (Pe=2(n-

1)). The axial dispersion can be obtained from the Pe number and gives a value of

0.0008 m2/s which correspond to the values found by Lorences et al. 2006. The RTD

experiments done at ten times the minimum fluidization velocity shows that the reactor

can be approximated by a plug flow with increased dispersion at increased catalyst

amount.

Table 4-2: n-CSTR model parameters for the RTD experiments done at UQ 0.052m/s
Catalyst amount c 2 (s) tm(s) N Pe
540g glass beads 573 26.8 28 54
540g glass beads + 50g catalyst 428 26.7 27 52
220g catalyst 618 28.0 15 28
 59

-2 3
Si

-o 2

5
' n ' f ^ 1 1 • — " "|"~ '•'•'" " '•>" ' l » W f

0.2 0.4 0.6 0.8 1.0 1.2

Time, min

0.4 0.6 O.J 1.0 1.2

Time, min

0.4 0.6 0.8 1.2

Time, min

Figure 4-3: RTD experimental results over the fluid bed with a gas velocity of 5.4 cm/s. A) reactor loaded
with 540g glass beads B) Reactor loaded with 540g glass beads and 50g of catalyst C) Reactor loaded
with 220g of catalyst. Experimental data represented by dots and n-CSTR model with a line.
 60

[Link] Methanol to DME

The experimental results from the MeOH/DME reaction are given in Table 4-3. The

table lists the experimental data with the actual temperature, equilibrium constant, flows,

catalyst amount and exit concentrations. A problem with a temperature probe has

resulted in a temperature offset in the temperature of experiment 20-46. Based on the

uncertainties in the experiments the maximum uncertainty is estimated to be ±2% on the

methanol conversion. Methanol conversion is defined as the conversion to equilibrium

while the total conversion is the actual conversion of methanol.

MS measurements was set up to measure not only Ar, MeOH, DME and H2O fractions

but also other hydrocarbon fraction to make sure possible side reactions were recorded.

Only a few experiments at 325°C are included in the table as MS traces revealed

hydrocarbon production in a majority of the data and they have therefore been omitted

for analysis. MS results of the experiments 20 - 24 are shown in Figure 4-4. At 30

minutes the gas velocity was close to umf and the DME and H2O concentrations were

high while methanol concentration was close to equilibrium. After 35 min the gas

velocity was increased to 0.89 cm/s and the DME and water concentrations dropped

while methanol increased. The same trend is observed for the subsequent feed rate

increases. The new product distribution was established within 2-5 minutes depending

on flow rate and the change is clearly defined. No propene formation was detected but a

low level of methane is present in all the experiments.

 61

Table 4-3: Experimental data obtained - all flows are listed at STP conditions.
Exp. no. Temp. Cat. QMeOH QAT U0 Keq y Me0H yDME MeOH Conv Total conv.
°C g ml/min ml/min cm/s % ± 2% %
1 250 50 16.8 334 0.67 16.0 0.007 0.020 95.8 85.2
2 250 50 50.5 947 1.92 16.0 0.010 0.021 91.2 81.1
3 250 50 124 2340 4.73 16.0 0.012 0.019 86.3 76.7
4 250 200 50.5 947 1.92 16.0 0.007 0.022 96.1 85.5
5 250 200 124 2334 4.73 16.0 0.008 0.021 93.9 83.5
6 275 50 50.5 947 2.01 12.8 0.009 0.021 94.1 82.6
7 275 50 124 2340 4.95 12.8 0.010 0.020 90.4 79.3
8 250 50 39.3 214 0.49 16.0 0.026 0.064 93.4 83.0
9 250 50 152 858 1.94 16.0 0.050 0.050 74.6 66.4
10 250 50 382 2160 4.88 16.0 0.075 0.038 56.2 50.0
11 275 50 39.3 214 0.51 12.8 0.026 0.065 94.9 83.3
12 275 50 152 858 2.03 12.8 0.034 0.058 87.8 77.1
13 275 50 382 2160 5.12 12.8 0.044 0.053 80.4 70.5
14 250 50 157 355 0.98 16.0 0.089 0.109 79.8 70.9
15 250 50 309 710 1.96 16.0 0.150 0.076 56.6 50.4
16 250 50 792 1840 5.05 16.0 0.209 0.046 34.3 30.5
17 275 50 157 355 1.03 12.8 0.066 0.120 89.4 78.4
18 275 50 309 710 2.05 12.8 0.081 0.111 83.5 73.2
19 275 50 792 1840 5.29 12.8 0.121 0.090 68.2 59.8
20 278 25 78.5 154 0.47 12.5 0.060 0.139 93.9 82.3
21 277 25 157 307 0.94 12.6 0.090 0.125 83.9 73.5
22 278 25 309 613 1.86 12.5 0.122 0.107 72.7 63.7
23 279 25 791 1580 4.82 12.4 0.183 0.075 51.5 45.2
24 277 25 1250 2560 7.79 12.6 0.224 0.053 36.6 31.9
25 280 50 78.5 154 0.47 12.3 0.059 0.140 94.2 82.6
26 277 50 157 307 0.94 12.6 0.075 0.132 88.9 78.0
27 278 50 309 613 1.86 12.5 0.088 0.123 84.1 73.6
28 279 50 791 1580 4.82 12.4 0.138 0.098 67. 58.7
29 278 50 1250 2560 7.71 12.5 0.174 0.077 53.6 47.0
30 278 100 78.5 154 0.47 12.5 0.046 0.146 98.7 86.3
31 279 100 157 307 0.94 12.4 0.053 0.143 96.4 84.3
32 279 100 309 613 1.87 12.4 0.064 0.135 92.3 80.8
33 282 100 791 1580 4.84 12.1 0.093 0.120 82.5 72.2
34 284 100 1250 2560 7.79 11.9 0.115 0.107 74.5 65.0
35 287 200 78.5 154 0.48 11.6 0.044 0.147 99.8 87.0
36 288 200 157 307 0.96 11.5 0.046 0.146 99.1 86.4
37 288 200 309 613 1.90 11.5 0.049 0.143 98.0 85.3
38 289 200 791 1580 4.90 11.4 0.062 0.136 93.5 81.5
39 290 200 1250 2560 7.88 11.3 0.073 0.128 88.9 77.7
40 325 25 78.5 154 0.51 8.66 0.044 0.147 101.8 87.0
41 325 25 157 307 1.01 8.66 0.044 0.148 101.8 87.1
42 325 25 309 613 2.02 8.66 0.054 0.140 98.1 83.8
43 325 25 791 1580 5.22 8.66 0.073 0.130 91.3 78.0
44 325 25 1250 2560 8.37 8.66 0.087 0.121 86.0 73.6
45 325 50 791 1580 5.22 8.66 0.061 0.136 95.6 81.8
46 325 50 1250 2560 8.37 8.66 0.069 0.130 92.5 79.1
 62

The conversion of methanol from the GC - relative to the MeOH/DME equilibrium -

was 96.4% in experiment 31 and 88.9% in experiment 39 which was in agreement with

the MS. From the GC data the level of methane was found to be 0.25% on carbon basis

which is considered to be negligible.

30 40 50 60 70
Time (min)
Figure 4-4: MS results from experiments done at 275 °C with 25 g of catalyst

Due to a slightly declining background level of water in the MS trace, the water fraction

was based on the DME which is valid due to the reaction stoichiometry. A graphical

representation of the methanol conversion of experiment 20-39 is shown in

Figure 4-5. The data all follows the same trend with higher conversion as the catalyst

amount increases and decreased conversion when the gas velocity is increased. With

200g of catalyst the conversion at the three lowest gas velocities are almost the same and
 63

within full conversion when the uncertainty is taken into account. The increase in

velocity compared to catalyst amount also correspond to each other e.g. going from at

gas velocity of 0.94g and 50g of catalyst to 1.86cm/s and lOOg of catalyst result in

approximately the same conversion. At higher gas velocities the same trend is seen with

slightly higher conversions by doubling the catalyst amount while increasing the

velocity by 60%. At low gas velocity close to umf conversions are high with only a slight

rise in conversion as the catalyst amount is increased. The difference between the data

points at 25g and 50g at this velocity is smaller that what would be expected from the

other gas velocities. The used experimental parameters are also shown to give a wide

conversion range from 30 to almost 100% conversion.

100 -,

t '"*
80 A
O
'w
>
60 A

oc
PJ • 0.47 cm/s
•£ 40 A 0.94 cm/s
a*
0 1.86 cm/s
A 4.82 cm/s
• 7.79 cm/s
20 —i
50 100 150 200

Catalyst amount (g)

Figure 4-5: Conversion of methanol to DME and water based on equilibrium conversion as function of
catalyst amount at five different gas velocities (experiment 20 - 39). Lines are the model predictions using
kinetic model equation 4-5 and 6 CSTR's in series.
 64

4.2.5 Modeling

[Link] Kinetics
Mollavali et al. (2008) recently studied the methanol to DME reaction (Equation 4-1)

and summarized the published kinetic models with reaction orders of lA, 1 and 2.

2CH3OH <=> CH3OCH3 +H20 (4-1)

Calculations of the reaction order by the half-lives method from the fluid bed data gives

a reaction order in the range 1 . 7 5 - 2 and therefore a second order reaction has been

assumed in this work.

Due to the high conversion levels the reverse reaction also has to be included in the

kinetic expression. Literature models that include the equilibrium term are given in

Table 4-4. The absorption term of DME is neglected in all of the models because it is

much smaller than the absorption of methanol and water (Bercic and Levee, 1992).

Table 4-4: Kinetic models including the reverse term.

Model Equation Ref.

k(CM-CwCD/(KeqCM))
' MeOH
Mollavali et al. (2008)
' {l + KMCM+CJKw)

_k(c2MICw-CDIKeq)
-r, Lu et al. (2004)
[l + KMCM +KWCW)

kK2M(c2M-CwCEIKeq)
-r. MeOH Bercic and Levee (1992)
(l + 2(KMCMf5+KwCw)4

A simple model for the for the methanol to DME reaction is proposed where methanol is

absorbed on a active site on the catalyst (Equation 4-2), two absorbed methanol species
 65

then react to DME and an absorbed water molecule (Equation 4-3) this reaction is

considered to be in the rate determined. Finally the water is desorbed (Equation 4-4).

CH3OH + S <-• CH3OHS (4-2)

CH3OHS + CH3OHS <-> CH3OCH3 + H2OS + S (4-3)

H2OS ^ H 2 0 + S (4-4)

The reaction rate of methanol from Equation 4-2 to 4-4 is given in Equation 4-5 where k

follows an Arrhenius relationship. This kinetic expression will be compared to the model

of Bercic and Levee (1992) where the model parameters have been re-estimated to

account for the different temperature interval and catalyst composition.

f
cc^

+ K
(l MCM+KWCw)

Hetsroni (1983) divides first order kinetics into three categories slow, intermediate and

fast reactions in fluidized beds based on their reaction rate constant. Slow reactions are

controlled by reaction kinetics while intermediate reactions can be controlled by both

reaction rate or mass transfer between bubble and emulsion. Fast reactions are

characterized by k values above 5 s" and slow reactions have k values below 0.5 s" . For

a first order reaction ty= ln(2)/k and a fast reaction in this view is a reaction with a half-

life below 0.14 s and a slow reaction with a half-life above 1.4 s. Using the same criteria

for second order reactions, we categorise the methanol dehydration reaction as either
 66

fast, intermediate or slow. The average half-life from the experiments can be calculated

by Equation 4-6 ignoring the reverse reaction,

<»=—-r— (4-6)
— 1
l-X

Where X is the total conversion of methanol and t is the residence time in the bed. For

experiment 43, the average half-life is 0.11 s which is below the criteria for a fast

reaction and it would be assumed that the mass transport between the emulsion and

bubble will play a role in the modelling of this data. For all the experiments below

300°C the half-life is above 0.5 s and the reactions rates are considered to be in the

intermediate region. The experimental data with conversion to equilibrium above 95%

all have an average half-life above 1.4 s and slow reaction rates which would also be

expected for second order reactions close to equilibrium where the reaction rate is

reduced.

[Link] Fluid bed model

The fluid bed has been modelled with a two region model with a bubble phase and an

emulsion phase as well as an n-CSTR's in series model with 6 CSTR-s in series. The

number of CSTR's in series has been considered to be constant over the range of gas

velocities and catalyst inventory. Improving the n-CSTR's in series model with the

amount of CSTR's as function of gas velocity and catalyst amount is possible but have

not been included in this study. The two phase model is well described in the literature

and has recently been used by Werther and Hartge (2004) to model industrial fluidized

bed reactors and by Abba et al. (2003) where it was used as part of a comprehensive
 67

model to characterise fluidization from bubbling conditions to fast fluidization.

Although the general model is of the same form, different assumptions are made to the

hydrodynamic. In this work where the following assumptions are made.

• Gas flow only in the axial direction - dispersion in the radial direction is not

considered (with the exception of interphase mass transfer)

• Interphase mass transfer between bubble and emulsion phase

• No catalyst in the bubble phase

• The activity of the catalyst is considered constant e.g. no deactivation due to

short reaction time compared to the deactivation time of the catalyst

The mass balance of the bubble and emulsion phases of the model can be written as

Bubble phase:

-ub^ = Kbe(c.b-CiJ (4-7)

az
Emulsion phase:

u LL
mf^ = 7r^:Kbe(Ci!b-C,e)+(l-£mf)pp(-ri) (4-8)
az (1-0)

The model has been implemented in Fortran as two parallel CSTR in series with mass

transfer between the two. The implemented equations are given below and a schematic

drawing is shown in Figure 4-6.

 68

Bubble phase Emulsion phase

N
»-be

ub *•• ;; . ,'Ujjjb

CAb *•.'-.: cAs

-1-6 •

CAi IUQ

Figure 4-6: Schematic of the two phase fluid bed model

The CSTR volume is based on equal catalyst amount e.g. constant emulsion volume,

which makes the CSTR's volume in the bubble phase dependent on the bubble velocity

S
^bXb,i,n ~ ^bXb,i,n-\ V K,\p x, —p x ) (4-9)
i c e be\r g b,i,n r g e,l,n I

c
F X F X
e e,i,n= e e,i,n-X + J ^ K K
be(PgXb,i,r, ~ PgXe,i,n )+ ^H"/ ) (4-10)

Optimization of parameters have been done with the use of the Simplex method and the

error calculations have been evaluated from Equation 4-11

 69

/ i \ i,n,cal /,n,exp /
n
R:=I- (4-11)
/ i V i,n,exp ;,exp /

The hydrodynamic correlations used in this work are given in Table 4-5. The mass

transfer coefficient correlations used in this work is the one from Sit and Grace (1981).

The bubble diameter has been calculated form the correlation of Mori and Wen (1975).

Since the used particles are Geldart A particles the minimum bubbling velocity is used

in the correlations for the gas in the emulsion phase.

Table 4-5: Hydrodynamic correlations - the correlation for bubble diameter is in cm and not in meters
Variable Correlation Ref.
Mass U
77 A
Sit and Grace
r _ mb
transfer K
+ A -be
. —•
be - „
du (1981)
coefficient v ^ b J

d
Bubble b=dbm-{dbm-dbo)GW Mori and
diameter v A j Wen (1975)
db0 = 0.00376(w0 - umb f, dbm = 0.652(4 (Mo - umb )T
Bubble u
velocity b =uo-umb + 0J\\(gdb)^
Minimum
M d3PPgipP-pg)g Wen and Yu
fluidization mf (33,7)2 + 0.0408 •33.7
d p (1966)
velocity f
umb _ 2300p»A»fi™" exp(0.716F)
.0.126 ,,0.523 Abrahamsen
Minimum
and Geldart
dpgm"ifir-p,t
934
bubbling *mf (1980)
Bubble g _ U
0 U
mb
fraction U U
b- mb
 70

4.2.6 Modelling result and discussion

All parameters in the fluid bed model besides the molecular diffusion coefficient are

given by the experimental conditions or the hydrodynamic correlations. The molecular

diffusion coefficient varies between 0.2-0.6 cm /s for the binary mixtures and for the

modelling of the fluid bed an average value of 0.4 cm2/s has been used. It has further

been found that slight changes in the molecular diffusion coefficient - within the range

of the binary mixtures - do not changes the modelling results significantly. The models

parameters were fit to the data in experiments 1 to 19. The parameters and level of fit to

the experimental data for the fluid bed model and 6 CSTR's in series are given in Table

4-6 and Table 4-7.

Table 4-6: Best fit parameters for the kinetic model Equation 4-5 and the level of fit to the experimental
data.
Reactor model k Ea Kw Km R R2
m 6 /(kgskmol) 3 3
kcal/mol m /kmol m /kmol E x p 1-19 all exp
Fluid bed 21900 2AJ5 47000 14000 0.918 0.901
6 CSTR's 21400 22.1 47000 15000 0.934 0.940

Table 4-7: Best fit parameters for model of Bercic and Levee (1992) and the level of fit to the
experimental data.
Reactor model k Ea Kw Km R R2
kmol/(kgs) kcal/mol m 3 /kmol m 3 /kmol E x p 1-19 all exp
Fliud bed 0.619 331) 590 48 0.906 0.884
6 CSTR's 0.677 32.2 480 30 0.910 0.925

The heat of adsorption has not been included into the kinetics of the model from

Equation 4-5 since the data was not sufficient to obtain reliable results and they have

therefore been considered as temperature-independent. For the model of Barcic and

Levee (1992) their values for the heat of adsorption for methanol and water was used.

The lack of temperature dependency in the absorption term for the proposed model
 71

might explain some of the difference in the activation energy seen between the two

models since the absorption has a significant influence on the reaction. The kinetic

model of Equation 4-5 shows the best result when coupled with 6 CSTR's in series

compared to the other kinetic and reactor model. Modelling the rest of the data set

(experiment 20-46) shows a very good fit to the data derived from the first 19

experiments. In Figure 4-5 the data from experiment 20-39 is plotted together with the

model prediction. The overall agreement is very good and only at high conversion does

the model over predict the actual conversion. The ability for the CSTR's in series model

to predict the reaction with different catalyst inventories even better than the fluid bed

model indicates that the mass transfer between the bubble and emulsion phase in the

fluid bed is not predominant under the given experimental conditions, which is also

indicated by the high conversions levels which would not be expected if mass transfer

between the emulsion and bubble was limiting the reaction.

 72

100

80

•a
IS 60 -
^3
#• //
« • yS
O 40-

20

0 -r i i i i i

20 40 60 80 100
Experimental data
Figure 4-7: Model predictions vs. experimental with 6 CSTR's in series. • Model Eq. 4-5; • Model of
Bercic and Levee (1992)

A parity plot of all the data compared to the model estimations with the use of the

kinetic model Equation 4-5 and the model of Barcic and Levee (1992) are given in

Figure 4-7. The model predictions are not that different at high conversions but at low

the model given by Equation 4-5 are slightly better. The model shows that the absorption

term - which is the different between the two models - have a large influence on the

reaction. Overall the predicted conversions are within 5% of the measured values.

The influence of the mass transfer coefficient in the two phase model have been

investigated by using the best fit parameters from the CSTR's in series model. In Figure

4-8 the effect of changing the mass transfer coefficient is depicted. The effect at low

catalyst inventory is limited and increases as the catalyst amount is increased, which is a
 73

result of the increasing bubble size along the bed. The deviation when 200g of catalyst is

used is significant and even with a molar diffusion coefficient ten times the estimated

the prediction is off by 5%. Some deviation is expected since the best fit parameters

from the CSTR's in series model is used the best fit parameters is however similar and

does not change the trends in the figure significantly.

100

c 80
O
'w
m
>
c 60
o
"o
c • Experimental
(Q
6 CSTR's
£ 40
CD D = 0.4 cm2/s
D = 0.8 cm2/s
D = 4 cm2/s
20
50 100 150 200

Catalyst amount (g)

Figure 4-8: Effect of the molar diffusion coefficient with the use of the best fit parameters for 6-CSTR's
in series with the two phase model and the kinetic model of Equation 4-5. The experimental points are at
275°C and a superficial gas velocity of 7.79 cm/s

The conversions measured in the present work are relatively high for a standard kinetic

study but due to the nature of the fluid bed the operational range is limited. If the gas

velocity is increased further the hydrodynamic of the bed could change leading to

different hydrodynamics in the kinetic dataset which will have to be accounted for.
 74

Investigations of the limitations in the superficial gas velocity in which the n-CSTR's in

series model can be use will therefore be useful but have not been pursued in this work.

Decreasing the catalyst inventory further will lead to a very shallow bed and possibility

of by-pass has to be considered. Changing the temperature is the best option to

investigate a large range of conversions, and lowering the temperature further than the

250°C done in this work would help explore the kinetic at low conversions.

Data from fixed bed experiments in a 4 mm inner diameter quartz reactor together with

predictions of the conversion using Equation 4-5 is given in Table 4-8. The predicted

conversions is lower then the measured with increasing difference as the reactions

temperature is increased. The higher difference between model prediction and

experimental data is most likely due to temperature gradients in the bed, caused by the

exothermic reaction, which increases with increased reaction temperature and thus

higher reaction rate.

Table 4-8: Fixed bed experiments in a 4 mm inner diameter quartz reactor.

em p. Cat. QMeOH Keq YlVleOH YDME MeOH Conv. Model prediction
<C 9 ml/min ml/min cm/s % ± 2%
250 0.1 1.3 12 3.39 16.0 0.066 0.017 39 50.2
250 0.1 2.8 25 7.05 16.0 0.070 0.015 34 34.1
275 0.1 1.3 12 3.55 12.8 0.027 0.036 83 74.6
275 0.1 2.8 25 7.39 12.8 0.034 0.033 75 58.7
275 0.1 5.6 50 14.8 12.8 0.057 0.022 50 42.8
300 0.1 1.3 12 3.71 10.4 0.016 0.042 97 91.5
300 0.1 2.8 25 7.73 10.4 0.019 0.040 93 80.2
300 0.1 5.6 50 15.5 10.4 0.026 0.037 85 66.0
300 0.1 1.3 12 3.71 10.4 0.014 0.043 99.4 91.5

Using the fluid bed for kinetic measurements of the methanol to DME reactions have in

this work been shown to give good results. The isothermal temperature profile of the
 75

fluid bed ensures that the temperature is easily controlled and well known. Since a small

temperature uncertainty can easily result in poor estimation of the reaction rate or

activation energy it is essential that the temperature is known and hot spots and gradients

are avoided. This is difficult in fixed bed when investigating exothermic reactions since

hot spots and temperature gradients will affect the result (Fogler, 2005) and one is forced

to dilute the catalyst with inert and run experiments at very low conversions to reduce

the effect. Diluting the catalyst introduces the possibility of bypassing and it has been

shown to influence the observed conversion especially at conversions above 0.4 (Berger

et al. 2002).

4.2.7 Conclusion

The reaction kinetics of the exothermic dehydration of methanol to DME have been

investigated in a fluidized bed at low gas velocities. A kinetic expression for the

methanol to DME reaction has been proposed and the kinetic parameter has been

estimated with both a two phase fluid bed model and an n-CSTR's in series as the

reactor model. The kinetic model coupled with the n-CSTR model was shown to give

superior result compared to the two phase fluid bed model. This was observed with both

the proposed model and the literature model of Barcic and Levee (1992). The fluid bed

was shown to be a good reactor type for the kinetic modelling of the methanol to DME

reaction due to the isothermal conditions and the conversions that could be obtained at

relatively low gas velocities. Other exothermic reactions where hotspots or temperature

gradients are of concern during kinetic measurements could benefit from the fluid bed
 76

technology. Due to restraints on the catalyst inventory and gas velocities the conversion

range at a given temperature is can be limited.

4.2.8 Nomenclature

At = Cross-sectional area of the fluid bed (m2)

C = Molar concentration (kmol/m )

<D = Molar diffusion coefficient (m 2 /s)

Dt = Reactor diameter (m)

dp = Mean particle diameter (m)

db = Bubble diameter (m)

dt,o = Initial bubble diameter at the distributor (m)

dbm = Maximum bubble diameter (m)

Ea = Activation energy (kJ/mol)

F = Fine fraction

Fb = Flow rate in bubble phase (kmol/s)

Fe = Flow rate in emulsion phase (kmol/s)

g = gravity (m/s )

Keq = Equilibrium constant

k = Reaction rate constant (m /kmols)

Kbe = Bubble to emulsion mass transfer coefficient (s"1)

kbe = Bubble to emulsion mass transfer coefficient (m/s)

N = Number of CSTR's
Pe = Peclet number (Pe = U g L/D)

r; = Reaction rate of component i (kmol/kg-s)

ti/2 = Halftime (s)

tm = Mean residence time (s)

=
umf Minimum fiuidization velocity (m/s)

umb = Minimum bubbling velocity (m/s)

Ub = Bubble velocity (m/s)

uo = Superficial gas velocity (m/s)

Ve = Volume of emulsion (m 3 )

W = Catalyst weight (kg)

Q = Gas flow (ml/min)

X = Total conversion

x = Mole fraction

y = Mole fraction of gas

z = Height in bed (m)

Greek symbols

8 = Bubble phase volumetric fraction

8mf = Void fraction at minimum fiuidization

(j, = Gas viscosity (kg/m-s)

a2 = Variance

pp = Particle density (kg/m 3 )

ptapped = Tapped density (kg/m 3 )

 78

pmf = Minimum fluidization density (kg/m3)

pg = Gas density (kg/m3)

x = Residence time (s)

Subscripts

exp = Experimental

calc = Calculated

n = CSTR number

i = species

be = Bubble to emulsion

b = Bubble

e =Emulsion

4.2.9 Literature Cited

Abba, I. A., J. R. Grace, H. T. Bi and M. L. Thompson, "Spanning the flow regimes:

Generic Fluidized-bed reactor model," AIChE Journal. 49, 1838 - 1848 (2003).

Abrahamsen, A. R., and D. Geldart, "Behaviour of gas-fluidized beds of fine powders

Part 1. Homogeneous expansion," Powder Technology. 26, 35 - 46 (1980).

Bercic, G. and J. Levee, "Intrinsic and Global Reaction Rate of Methanol Dehydration

over Y-A1 2 0 3 Pellets," Ind. Eng. Chem. Res. 31, 1035-1040 (1992).
 79

Berger, R. J, J. P. Ramirez, F. Kapteijn and J. A. Moulijn, "Catalyst performance testing:

bed dilution revisited" Chem. Eng. Sci. 57, 4912 - 4932 (2002)

Christensen, D., D. Vervloet, J. Nijenhuis, B.G.M. van Wachem, J.R. van Ommen and

M.-O. Coppens, "Insights in distributed secondary gas injection in a bubbling fluidized

bed via discrete particle simulations" Powder Technology. 183, 454 - 466 (2008)

Fogler, H. S. "Elements of chemical reaction engineering" Prentice Hall, 4th edition,

2005

Gray, C. and G. Webster, "A Study of Dimethyl Ether (DME) as an Alternative Fuel for

Diesel Engine Applications" Transport Canada Publication No.: TP 13788E, May 2001

Hetsroni, G., "Handbook of Multiphase Systems" Hemisphere-McGraw Hill, 1982

Horio, M. and Wen, C. Y. "An assessment of fluidized-bed modelling" AIChE

symposium series, 73, 9 - 21 (1977)

Kaarsholm, M., F. Joensen, J. Nerlov, R. Cenni, J. Chaouki and G. S. Patience,

"Phosphorous modified ZSM-5: Deactivation and product distribution for MTO" Chem.

Eng. Sci., 62, 5527 - 5532 (2007)

 80

Keil, F. J. "Methanol-to-hydrocarbons: process technology" Microporous and

Mesoporous Materials, 29, 49-66 (1999)

Kunii, D. and O. Levenspiel, "Fluidization Engineering," Butterworth-Heinemann series

in chemical engineering, Newton, MA (1991).

Kunii, D. and O. Levenspiel, "Bubbling bed model. Model for the flow of gas through a

fluidized bed" I&EC Fundamentals. 7(3), 446 - 452 (1968)

Lorences, M. J., J.-P. Laviolette, G. S. Patience, M. Alonso and F. V. Diez, "Fluid bed

gas RTD : Effect of fines and internals," Power Technology. 168, 1 - 9 (2006).

Lu, W.-Z., L.-H. Teng and W.D. Xiao, "Simulation and experiment study of dimethyl

ether synthesis from syngas in a fluidized-bed reactor" Chem. Eng. Sci. 59, 5455 - 5464

(2004)

May, W. G., "Fluidized-bed reactor studies" Chemical Engineering Progress, 55(12), 49

-56(1959)
 81

Mollavali, M., F. Yaripour, H. Atashi, and S. Sahebdelfar, "Intrinsic Kinetic Study of

Dimethyl Ether Synthesis from Methanol on y-Al 2 0 3 Catalysts," Ind. Eng. Chem. Res.

47, 2365-3273 (2008).

Mori, S. and Y. Wen, "Estimation of bubble diameter in gaseous fluidized beds," AIChE

Journal. 21(1), 109 - 115 (1975).

Radmanesh, R., J. Chaouki, and C. Guy, "Biomass Gasification in a Bubbling Fluidized

Bed Reactor: Experiments and Modeling" AIChE Journal. 52(12), 4258 - 4272 (2006)

Semelsberger, T. A., R. L. Borup, and H. L. Greene, "Dimethyl ether (DME) as an

alternative fuel" Journal of Power Sources. 156,497-511 (2006)

Shaikh, A. A. and H. Batran, "On bubble-side transport limitations in catalytic fluid-bed

reactors" Chem. Eng. Res. Des. 85, 1215 - 1218 (2007).

Sit, S. P., and J. R. Grace, "Effect of bubble interaction on interphase mass transfer in

gas fluidized beds," Chem. Eng. Sci. 36, 327 - 335 (1981).

Van Deemter, J. J., "Mixinf and contacting in gas-fluidized beds" Chem. Eng. Sci.

13(3), 143-154(1961)
 82

Wen, C.Y., and Y. H. Yu, "Mechanics of fluidization," Chemical Engineering Progress

Symposium Series. 62, 100-111 (1966).

Werther, J., and E.-U. Hartge, "Modeling of industrial fluidized-bed reactors," Ind. Eng.

Chem. Res. 43, 5593 - 5604 (2004).

 83

Chapter 5
Phosphorous modified ZSM-5: Deactivation and
product distribution for MTO*

5.1 Presentation of the article

The objective of the second article is to investigate the deactivation and product

distribution of the phosphorous modified ZMS-5 catalyst in small diameter fixed bed

reactors. Both pure methanol and 10% methanol in nitrogen at different feed rates is feed

to the reactor showing a substantial change in the amount of methanol converted per

catalyst volume before deactivation dependent of feed rate and composition. Secondary

cracking reactions of higher olefins are found to be and important route to the lower

olefins. For better understanding of the deactivation and change in product distribution

the deactivation of the catalyst have been followed during the reaction. Due to the colour

change of the catalyst during reaction the coke front down the bed could be followed and

three zones identified.

* Chemical Engineering Science Volume 62, Issues 18-20, September-October 2007,

Pages 5527-5532
 84

5.2 Phosphorous modified ZSM-5: Deactivation and product

distribution for M T O
Mads Kaarsholm, Finn Joensen1^ Jesper Nerlov*, Roberta Cenni^, Jamal
Chaouki, Gregory S. Patience
Department of Chemical Engineering, Ecole Polytechnique de Montreal, Canada
'Haldor Topsoe A/S, Denmark

5.2.1 Abstract

The product distribution and deactivation in the MTO process over a phosphorous

modified catalyst containing 10% H-ZSM-5 was studied in small diameter fixed bed

reactors. These studies suggest that methane is formed directly from methanol and/or

dimethyl ether and that non-aromatic C5+ hydrocarbons are intermediates in the MTO

reaction, forming light olefins (C3= and C4=, but not ethylene) by secondary cracking

reactions. Based on photographs of the catalyst taken during the course of the reaction,

three distinct coking patterns were observed that might be attributable to the different

reactions. Deactivation time of the catalyst is highly dependent on the contact time,

doubling the feed rate decreased the deactivation time by a factor of ten and lowered the

olefin production. Changing the feed form pure methanol to 10% methanol in nitrogen

reduced methanol capacity of the catalyst considerably, but a slight increase in

propylene selectivity was also observed.

Keywords; Zeolites; Catalyst selectivity; ZSM-5; Methanol-to-olefins

5.2.2 Introduction

The methanol to hydrocarbon process over acid zeolites has received significant

academic and industrial attention since its discovery in the late 1970's (Chang, 1984). In

the last two decades the main focus of this research has been on the methanol-to-olefin
 85

(MTO) part of the reaction due to the increasing demand for light olefins (Chen et al.,

2005). This has become even more pronounced with the high price of crude oil. The

MTO process is an acid catalyzed reaction of which SAPO-34 and ZSM-5 are the most

common catalysts (Stocker, 1999).

Several kinetics/reaction paths have been proposed for the MTO and MTG process.

Chang & Silvestri (1977) proposed the first model.

2Offs0ff = = ± CHsOCHs ^U
+HaO
__ _ paraffins
Gp — €^r ——» aromaiies
aydoparaffins
C$+ olefins

Equilibrium between MeOH and DME is reached very rapidly. This mixture then reacts

to form light olefins followed by paraffins and aromatics. In early kinetic studies, the

two oxygenates were treated as a single species (Chang, 1983; Keil, 1999). In addition,

the kinetic rates were derived for fresh catalyst and deactivation was ignored. Aguayo et

al. (2005) modelled the reaction kinetics and deactivation for both ZSM-5 and SAPO-18

catalysts and accounted for both the deactivation due to coking and irreversible activity

loss that occurs with each regeneration cycle. However, their model lumped the light

olefins into one component and is therefore unsuitable for differentiating between

ethylene and propylene yield.

Detailed models have been proposed by Mihail et al. (1983) and Park and Froment

(2001a,b) who characterized the process based on elementary reactions. Mihail et al.
 86

(1983) considered 27 reactions for the C1-C5 fractions and an additional 26 for higher

hydrocarbons. Dehertog and Froment (1991) examined the effect of phosphorus on H-

ZSM-5 catalysts. They showed that the space velocity had to be reduced by 75% in

order to achieve the same conversion compared to the non-modified zeolite. The

phosphorus modification was shown to increase the lower olefin yield at low

temperatures; at high temperature (480°C), the effect of phosphorus was not as

pronounced. Other investigations of P-modified ZSM-5 showed a similar decrease in

conversion and increase in selectivity of up to 70 % at complete methanol conversion

(Froment et al., 1992).

Chen et al. (2000) investigated the effect of space velocity over a SAPO-34 catalyst.

They found that the coking rate was lower with high methanol feed rates under partial

methanol conversion conditions. This was also observed for ZSM-5 by Benito et al.

(1996). The increased coking rate was largely attributed to the increased conversion of

oxygenates and, to a small extent, to the space velocity. Chen et al. (2000) found that the

coke content was proportional to the amount of hydrocarbon formed (per mass of

catalyst) and independent of space velocity.

In the present study, the effect of space velocity, temperature and reactor diameter on the

activity profile of the catalyst is reported. These parameters are important in the design

of MTO reactors for both fixed and fluid bed processes.

 87

5.2.3 Experimental

[Link] Catalyst

The catalyst was made of 10% CBV28014 (from Zeolyst) imbedded in a Si/Al matrix

consisting of Catapal B, Levasil 100s/30% and kaoline which was spray dried and

calcined in air at 550°C for 4 h. The average particle size of the resulting catalyst was

100 urn. The powder was contacted with a (NH3)2HP04 solution and then dried and

calcined so that the resulting catalyst contained 1.5% phosphorous. It was made into

tablets, crushed and sieved into a fraction of 400 - 600 um.

[Link] Reactor setup

The experimental setup is shown in Figure 5-1. The experiments were conducted in an 8

mm stainless steal reactor and a 9 mm inner diameter quartz reactor into which a 3 or 6

mm inner diameter quartz reactor could be inserted. The outer diameters of the two

insertion tubes were 5.0 and 8.0 mm resulting in an annulus between the inner and outer

tubes of 2 and 0.5 mm, respectively. The small insertion tube expanded to an outer

diameter of 8.2 mm approximately 3 cm above the catalytic bed reducing the length of

the larger annulus to a minimum. The vessel was heated electrically and the temperature

was regulated with a PID controller. The distance between the 9 mm quartz tube and the

heater was 1.75 mm. The reference temperature for control was measured at the wall of

the heater. The temperature at the reactor exit was monitored by a thermocouple placed

in a thermowell 10 mm below the bed. Additional thermocouples were inserted in the

catalyst bed. When the feed gas was switched from nitrogen to a 10% methanol in
 88

nitrogen mixture, the bed temperature rose 3°C. The pressure in the reactor was

maintained between 0-0.1 barg for all experiments.

N,
c*a cwj

II h©
\
MeOH

> vent

Figure 5-1: Experimental setup

The feed stream was prepared by passing nitrogen through two saturation evaporators

containing MeOH: the first was kept at room temperature and the second at 16°C to

maintain a 9:1 nitrogen to methanol ratio. A Brooks 5850TR flow controller was used

for the nitrogen feed gas. The product stream was analyzed by FID on a Hewlett Packard

5890 series-II GC equipped with a pre-column (Porapak Q 80/100 mesh size - 0.5

mx2.16 mm IDxl/8" OD packed column in stainless steel) followed by a capillary

column: CP-PoraPLOT A1203, KC1 - 10 m x 530 urn x 5 (am. A GC with a TCD

measured the CO/CO2 and other light gases for some experiments.
 89

[Link] Procedure

The 3 and 6 mm ID reactors were loaded with 0.185 g and 0.53g, respectively, to obtain

equal bed lengths. Experiments with 100% MeOH were conducted in the 8 mm reactor

with 1 g of catalyst. The reactor was heated to a given set-point in nitrogen before

feeding reaction gases. The flow rate was varied between WHSV 0.22 - 2.4 h-1 (based

on total catalyst mass). Product gases were analyzed on-line and the experiments were

terminated when only equilibrium mixtures of MeOH and DME were detected.

5.2.4 Results

The experiments showed that the space velocity had a significant influence on the

deactivation rate and product distribution. Figure 5-2 shows the product distribution at

two space velocities and 500°C in the 6 mm reactor. The carbon mass balance was based

on C1-C9 hydrocarbons only. Production of higher hydrocarbons, CO and CO2 in the

product gas and carbon deposition on the catalyst were neglected. The figure shows that

the deactivation rate depends on the methanol feed rate: Methanol breakthrough

occurred after 40 h at a WHSV of 0.22 h"1 (Figure 5-2a) compared to less than 5 h at a

WHSV of 0.43 h"1 (Figure 5-2b). By increasing the WHSV to 0.86 h"1, the C5+ fraction

increased and methanol breakthrough was observed after only a few minutes. The

experiments show that the olefin yield decreases with increasing flow rate, which is

contrary to some experiments reported in the literature over pure zeolite (Park and

Froment, 2001b; Dessau, 1986).

 90

o Methane
A Ethylene
• Propylene
/ x Methanol
£ xDME
SK oC4
V X*
• C5+

20 40 60 80 100 120
Time on Stream (hours)

uu - o Methane
B
X A Ethylene
as 40
X • Propylene
O (
• • • • x Methanol
• X
ro • xDME
•^ 3 0 - oC4
C i• • X
o • • X • C5+
CD X
L_ B X
4—* K
§ 20 • *
o • •
c 0 •0 •0 a •
O O •
0 o O 0 s 8 i O
* 10 H X o 0
i 0 • o
A A •
A
X
X
i A
A A I i
A A
0 *! $ —*— - T * — I I

5 10 15 20
Time on Stream (hours)

Figure 5-2: MTO product distribution at 500°C in the 6 mm inner diameter reactor with 10% methanol in
nitrogen feed: (a) WHSV 0.22 h"1 (dashed lines due to missing data), (b) WHSV 0.43 h"1
 91

The product distribution of the experiments conducted in the 3 and 6 mm reactor after 2

hours on stream is given in Table 5-lTable 5-1. As expected both the yield of light

olefins and methane increase with temperatures. However, at 500°C and partial

methanol conversion, increased space velocities resulted in a decrease in olefin

selectivity. At the highest space velocity (WHSV of 1.43 h"1) in the 3 mm reactor,

methane became the dominant hydrocarbon product, accounting for almost half of the

total.

Table 5-1: Product distribution after 2 hours on stream in the 3 and 6 mm diameter reactors.
Reactor diameter, mm 3 3 3 6 6 6 6 6
Temperature 500 500 500 500 500 500 450 400
WHSV, h"1 1.47 0.87 0.43 0.86 0.43 0.22 0.23 0.25
Contact time, s 0.18 0.31 0.63 0.32 0.63 1.25 1.24 1.24

Product distribution, %
Methane 7.2 7.8 7.6 5.3 12 3.6 2 1.2
Ethane 0.3 0.3 0.4 0.2 0.8 0.2 0.1 0.1
Ethylene 0.9 3.1 7.3 2 7.5 8.3 5.4 2.7
Propane 0.1 0.1 0.6 0.1 0.4 0.9 0.6 0.1
Propylene 2 11 32 8.1 28 40 39 16
Methanol 16 22 1 14 0.6 0.1 0 17
DME 69 27 0.1 16 0 0 0.3 28
C4 1.5 7.8 17 6 15 21 24 11
c5+ 2.9 21 33 48 35 26 29 24
Sum 99.9 100.1 99 99.7 99.3 100.1 100.4 100.1

Aromatics 0.5 4.6 11.9 7.6 17 13.1 5.8 1.5

Aromatics (in C5+ fraction %) 18.7 21.7 35.4 16 48.1 50.5 19.8 6.3
% MeOH/DME conversion 15 51 98.9 70 99.4 99.9 99.7 55

The first six experiments reported span the regime of partial to essentially full

MeOH/DME conversion. The fact that methane predominated at the highest space

velocity, where only 15 % of MeOH/DME was converted into hydrocarbons, leads us to

 92

propose that methane forms parallel to the equilibrium reaction from either MeOH or

DME, or from both.

60
o Methane (6mm)
50 & Ethylene (6mm) ,•*' ~~~
c B Propylene (6mm)
140 • C5+ (6mm)
u.
c
o o Methane (3mm)
"is 30 H A Ethylene (3mm)
c
a>
o • Propylene (3mm) ,-» ,,--'"''
§ 20
O • C5+ (3mm) ''' „---'"
x
-
w 10 & „„.--"--
::
i-'-'- - "" .A i 4
20 40 60 80 100
% MeOH/DME Conversion
Figure 5-3: Hydrocarbon distribution as a function of oxygenate conversion. Based on experiments in
Table 1 at 500°C

At 50-70% MeOH/DME conversion, the C5+ fraction predominated, whereas at close to

full conversion propylene was the most abundant hydrocarbon. In Figure 5-3, the

product distributions shown in Table 5-1 (500°C data) are plotted against the

MeOH/DME conversion. The plot demonstrates that: (1) Methane is formed very early

in the reaction, before the higher hydrocarbons; (2) At higher conversion, methane levels

are largely constant; (3) As reaction progresses, the C5+ yield reaches a maximum and

subsequently declines while the formation of light olefins accelerates. The latter

observation may be rationalized in terms of higher aliphatic (and possibly naphthenic)

hydrocarbons undergoing secondary cracking reactions, thus increasing the amount of

light products. The data in Table 5-1 show that the C5+ fraction at low conversion is low
 93

in aromatics and, therefore, relatively labile with respect to secondary cleavage

reactions. The sequence of reactions described above may also explain both the coking

pattern (vide infra) and the fact that, in experiments conducted at low space velocity

with complete MeOH/DME conversion, C5+ yields tend to increase as breakthrough

approaches. This peak in C5+ yields at close to full conversion is apparent from Figure

5-2a and b. It is also evident from Figure 5-4, showing the product distribution as a

function of MeOH/DME conversion in an integral experiment conducted with a feed of

100% methanol.

40

35 0 Methane
A Ethylene
"v 30 • Propylene '•-*.
o "*-•-•-•
_Q • C5+
c5 2 5
+ Aromatics
CO
2 0
0
Complete
E
~ 15
CD Conversion
O

o 10
O
LU -.-.,--0-
:—¥-

20 40 60 80 100 % Conversion
(Time on Stream) •
Figure 5-4: Hydrocarbon distribution as a function of oxygenate conversion. Based on experiment at
500°C in 8 mm inner diameter reactor with pure methanol feed. WHSV 2.37 h"1.

Figure 5-4 also shows that the ethylene yield closely parallels that of the aromatics

(predominantly xylenes and trimethylbenzenes). This trend agrees with recent data

published by Svelle et al. (2006) who concluded that ethylene is formed from xylene and

trimethylbenzene intermediates rather than by secondary alkene cracking reactions. Also

 94

apparent from Figure 5-4 is that methane formation gradually increases as breakthrough

is approached and subsequently remains virtually constant. This is in full accordance

with the suggestions presented above, namely that methane is formed directly from

MeOH and DME: In the early stages of the MTO reaction MeOH and DME are

completely converted in a narrow zone close to the reactor entrance. As coking

progresses the zone where MeOH and DME are present gradually expands through the

catalyst bed and the methane formation increases. After breakthrough, i.e., when MeOH

and DME extends the full catalyst bed, methane formation reaches a maximum and stays

constant thereafter.

We considered channeling as a possible mechanism to account for the lower olefin

production at higher flow rates. The average particle diameter to tube ratio in the 8, 6

and 3 mm inner diameter reactors was 16, 12 and 6, respectively, and these values are

greater than the standard rule of thumb of 10 for the two larger reactors. In the case of

the 3 mm reactor, some channeling might occur over a limited length of the bed.

However, channeling is considered to be negligible when the bed length to tube diameter

is greater than 10, which is the case of the 3 mm reactor. After unloading the catalyst

from the reactor, radial gradients of colour at various distances from the bed entrance

were not seen. Finally, all recorded data of the exit composition showed that the

MeOH/DME equilibrium was established. Thus, we conclude that channeling effects

were insignificant.
 95

In Figure 5-2a breakthrough occurred after less than 50 hours. In a comparative

experiment using 100% MeOH as a feed and maintaining almost the same contact time

(1.13 sec) the time until breakthrough was 10 hours. Taking into account the total

methanol (100% vs 10%) feed and reaction time (10 h vs. < 50 h), the capacity of the

catalyst to react is at least twice as high when the feed is 100% MeOH. The increased

methanol capacity with pure methanol may be attributable to the higher water vapour

pressure that reduces the coking rate (Gayubo et al., 2004). The product distribution was

slightly different: with diluted methanol; notably propylene yields were higher whereas

the C4 fraction was lower at the low methanol concentration. Whereas there is virtually

no difference in ethylene yields supporting the two different reaction pathways for

ethylene and propylene. With pure MeOH feed methane formation was significantly

reduced, probably due to the higher water partial pressure

With respect to coking, nitrogen QBET measurements clearly demonstrate a reduction in

pore volume of the deactivated catalyst over the entire range, as shown in Figure 5-5.

The total pore volume of the fresh, coked and regenerated samples were 0.207, 0.0973

and 0.202 cm3/g. The lower volume maybe attributable to two mechanisms: accumulated

surface carbon deposition and blocking of the pore entrances to the pores. The

regenerated catalyst was treated with air at 500°C and, in the region of 20 to 1000 A, its

pore volume distribution is identical to that of the fresh catalyst. However, in the rage of

13-20 A there appears to be a minor difference between the two. The pore volume of the

coked catalyst was unexpectedly higher in this region.

 96

Pore Diameter [A]

Figure 5-5: Pore size distribution of fresh, coked and regenerated catalyst.

Figure 5-6 demonstrates the coking pattern along the reactor: very quickly the colour

changes from beige to gray (less than 10 minutes) and three distinct regions become

evident as shown in Figure 5-6c: The middle of the bed is darkest gray; the entrance

region (upper box) is slightly brighter; and, in the exit region of the bed (lower box),

there is a gradient of colour from dark to light gray going down the reactor. These

observations may be rationalized as follows: At the entrance, MeOH and DME reach

equilibrium and C-C bond formation occurs to a limited extent. In the middle section,

olefin production dominates together with methylbenzenes, the reactive intermediates of

the MTO and MTG reactions (Arstad and Kolboe, 2001; Svelle et al, 2006). Chen et al.

(2000) found that the oxygenate to olefin reaction was the main source of coke. In the

last section, coking continues but only to a limited extent and most likely due to
 97

secondary cracking reactions as all oxygenates would have been consumed. Coking by

hydrocarbon decomposition at high temperatures due to significant temperature increase

from the exothermic reaction could be ruled out since temperature measurements

indicated the bed was isothermal.

a b c d e f g
Figure 5-6: Catalyst coking at t = 0, 10, 20, 50, 110,170, 1310 minutes in the 6 mm reactor, 10% MeOH
in nitrogen, T=500°C, WHSV 0.43 g(MeOH)/gcat/h (gas flows from the top down through the bed).

The coking pattern shown In Figure 5-6 is representative of the product distribution

given in Figure 5-2b. The product distribution was almost constant during the first 3

hours corresponding to the coking pattern in Figure 5-6b-f. The product distribution

appears to be independent of the coke deposition. Methanol breakthrough is observed a

short time after Figure 5-6f, which indicates that the strong acid sites are inaccessible.
 98

5.2.5 Conclusions
Experiments conducted at high space velocities in the regime of partial to essentially full

methanol and dimethyl ether conversion suggest that methane is formed directly from

methanol and/or dimethyl ether. The largely non-aromatic C$+ intermediates

predominate up until 100% MeOH/DME conversion. At this point, propylene yield is

highest due to the cleavage of the C54- fraction. Increasing the feed flow rate lowered the

olefin production even at complete oxygenate conversion. The formation of ethylene

closely parallels the aromatics, suggesting that it is by splitting off from xylene and

trimethylbenzene intermediates in accordance with the labeling studies made by Svelle

et al. (2006). Thus, ethylene is not a product of secondary cracking reactions.

In the region of full oxygenate conversion, catalyst deactivation depended on contact

time: doubling the feed rate (from a WHSV of 0.22 h"1 to 0.43 h"1) decreased the

deactivation time by a factor of ten. As a result, the methanol capacity of the catalyst

increased significantly as the feed rate was reduced.

The product distribution and catalyst methanol capacity (kg of MeOH converted per kg

of catalyst) are highly dependent on the feed composition. At reduced MeOH partial

pressure (10% methanol in nitrogen), the olefin production was higher compared to pure

methanol feed. At the same time, the methanol capacity of the catalyst was decreased

significantly. Clearly, increasing the methanol partial pressure increases the catalyst

methanol capacity, while increasing the methanol feed by increasing the space velocity
 99

decreases the methanol capacity. This is in contrast to the finding of Chen at al. 2000 for

the SAPO-34 catalyst where they concluded that the coke deposition was based on the

amount of methanol converted and temperature but not on WHSV.

Based on photographs of catalyst taken at intervals during an experiment, coking rates

are low at the reactor entrance where the MeOH/DME reaches equilibrium. The main

coking is due to the oxygenate to olefin and intermediates reaction. Following the main

reaction zone, coking rates are low as only secondary cracking reactions occur and

because the water level has reached its maximum.

5.2.6 References

Aguayo, A. T., Gayubo, A. G., Vivanco, R., Alonso, A., Bilbao, J., (2005). Initiation

step and reactive intermediates in the transformation of methanol into olefins over

SAPO-18 catalyst. Industrial & Engineering Chemistry Research, 44 (19), 7279-7286.

Arstad, B., Kolboe, S., (2001). The Reactivity of Molecules Trapped within the SAPO-

34 Cavities in the Methanol-to-Hydrocarbons Reaction. Journal of the American

Chemical Society, 123, 8137-8138.

Benito, P. L., Gayubo, A. G., Aguayo, A. T., Castilla, M., Bilbao, J., (1996).

Concentration-dependent kinetic model for catalyst deactivation in the MTG process,

Industrial & Engineering Chemistry Research, 35, 81-89.

Chang, C. D., Silvestri, A. J., (1977). The conversion of methanol and other O

compounds to hydrocarbons over zeolite catalysts. Journal of Catalysis, 47, 249-259.

Chang, C. D., (1983). Hydrocarbons from methanol. Catalysis Reviews - Science and

Engineering, 25, 1-118.

Chang, C. D., (1984). Methanol Conversion to Light Olefins. Catalysis Reviews -

Science and Engineering, 26 (3&4), 323-345.

Chen, D., Rebo, H. P., Gronvold, A., Moljord, K., Holmen, A., (2000). Methanol

conversion to light olefins over SAPO-34: Kinetic modeling of coke formation.

Microporous and Mesoporous Materials, 35-36 121-135.

Chen, J. Q., Bozzano, A., Glover, B., Fuglerud, T., Kvisle, S., (2005). Recent

advancements in ethylene and propylene production using the UOP/Hydro MTO

process. Catalysis Today, 106, 103-107.

 101

Dehertog, W. J. H., Froment, G. F., (1991). Production of light alkenes from methanol

on ZSM-5 catalysis. Applied Catalysis, 71, 153-165.

Dessau, R.M., (1986). On the H-ZSM-5 catalyzed formation of ethylene from

methanol or higher olefins. Journal of Catalysis, 99, 111-116.

Froment, G. F., Dehertog, W. J. H., Marchi, A. J., (1992). Zeolite catalysis in the

conversion of methanol into olefins. Catalysis 9, 1-64.

Gayubo, A. G., Aguayo, A. T., Atutxa, A., Prieto, R., Bilbao, J., (2004). Role of

Reaction-Medium Water on the Acidity Deterioration of a HZSM-5 Zeolite. Industrial &

Engineering Chemistry Research, 43, 5042-5048.

Keil, F. J., (1999). Methanol-to-Hydrocarbons: Process Technology. Microporous

And Mesoporous Materials, 29, 49-66.

Mihail, R., Straja, S., Maria, G., Musca, G., Pop, G., (1983). Kinetic Model for

Methanol Conversion to Hydrocarbons. Chemical Engineering Science, 38, 1581-1591.

Park, T-Y., Froment, G. F., (2001a). Kinetic Modeling of the Methanol to Olefins

Process. 1. Model Formulation. Industrial & Engineering Chemistry Research, 40, 4172-

4186.
 102

Park, T-Y., Froment, G. F., (2001b). Kinetic Modeling of the Methanol to Olefins

Process. 2. Experimental Results, Model Discrimination, and Parameter stimation.

Industrial & Engineering Chemistry Research, 40, 4187-4196.

Stocker, M., (1999). Methanol-to-hydrocarbons: catalytic materials and their

Behaviour. Microporous and Mesoporous Materials, 29, 3-48.

Svelle, S., Joensen, F., Nerlov, J., Olsbye, U., Lillerud, K.P., Kolboe, S. Bjorgen, M.,

(2006). Conversion of methanol in to hydrocarbons over zeolite H-ZSM-5: Ethene

formation is mechanistically separated from the formation of higher alkenes. Journal of

the American Chemical Society. 128 (46), 14770-14771.

 103

Chapter 6
Trends in the MTO reaction by MS

6.1 Introduction
This chapter will shortly discuss the mass spectrometry (MS) as a viable measurement

technique to follow trends in the MTO reaction as well as show trends in the reaction as

function of temperature in a fluidized bed reactor. Due to the wide product distribution,

product analysis by GC is time consuming and thus identifying optimal catalyst

compositions, reactor control and kinetic analysis is essential to accelerate the

development of the technology. MS is coupled together with GC measurements to be

able detect rapid changes in the reaction kinetics while at the same time allowing for a

quantitative analysis in periods of steady state. The focus of this work is twofold: (1)

investigate the feasibility of MS to determine key product distribution changes and (2)

improve the understanding of the MTO reaction mechanism over a ZSM-5 catalyst

6.2 Experimental
The catalyst for the study of the MTO reaction is a phosphorous modified ZSM-5 zeolite

(Si/Al ratio of 140) with a mean diameter of 108 \im. Experiments were conducted in a

46 mm inner diameter fluidized bed and the reaction temperature was varied between

250-550°C with a WHSV between 0.2 - 0.8h_1. The feed consisted of pure methanol,

methanol/water (molar ratio 1:1) and methanol/Ar (molar ratio 1:1) mixtures to assess

the effect of methanol partial pressure and water. Analysis of the product distribution
 104

were made with a Varian CP3800 GC (FID) and online measurements by a Hiden MS

QIC-20 were used to follow the trends in the reaction.

6.3 Results and discussion

Using MS to follow the product distribution of the MTO reaction poses several

challenges due to the high level of overlap between the hydrocarbons. The majority of

the components have therefore been lumped together with only a few separated

individual species. The components that can be separated are water, argon, methane,

methanol and DME. All other hydrocarbons overlap each other to such an extent that

separation is only viable for component classes as olefin and paraffins.

Compared to the GC, the MS has the advantage that these products or product lumps can

be followed in real time compared to the GC where it can take up to an hour to analyze a

single sample. To obtain a better understanding of the MTO reaction, an experiment

was carried out where the temperature was increased in steps of 50°C from 250°C to

550°C. Figure 6-1 A and B show the conversion of methanol and DME from 250°C to

550°C. It is clear that almost complete conversion is obtained above 450°C. Figure 6-1

C and D shows the production of olefins and methane. Reaction from MeOH and DME

to hydrocarbons starts between 300 and 350°C and has a maximum around 500°C. At

temperatures above 450°C the methane production increases significantly, especially

from 500 to 550°C and at the same time a drop in olefin production is observed. The

drop in DME and the increase in olefins in the step from 300 - 350°C before it levels off
 105

are due to an overshoot by a few degrees in the temperature control. Generally, the

trends are easy to follow on the MS and could be a valuable technique when following

the conversion of MeOH/DME during a reaction/regeneration cycle where it could be a

fast way to determine when regeneration is needed. This is most likely more suitable for

the SAPO system where the deactivation is fast and time consuming GC measurements

will be too slow to be practical. Especially for circulating fluidized beds the MS could

help controlling regeneration to obtain optimum operation conditions.

250-300-C
300-350°C
350-400-C
400-450"C
450-500°C
500-550°C

20 30 40
Time (min) Time (min)

ay-^aiepKi wmm?>*i+*i* -M***^j

._ - «^*^t*. i*ff r 1 « 250-300X
0.4- • 300-350X
• 350-400"C
0.2- • 400-450X
. 450-500X
• 500-550,,C
0- 1 1 i i i i
20 30 40 20 30 40
Time (min) Time (min)

Figure 6-1: MS traces during the MTO reaction with temperature increments of 50CC from 250°C to
550°C. A) Methanol, B) DME, C) Hydrocarbons and D) Methane.

Experiments with different feed compositions and WHSV in the temperature range 450

- 550°C showed clear trends in the product distribution (Figure 6-2). The lower olefins

- ethene and propene - increases with temperature as the higher olefins are broken down.
 106

The methane concentration also starts to increase significantly from 500 - 550°C. The

influence of feed composition on product distribution decreases with temperature with

less spread in the individual components concentrations. This could indicate that the

reaction progresses further towards an equilibrium at the highest temperature and that

the equilibrium of the hydrocarbons is relatively independent of the feed concentration

e.g. water, argon and partial pressure.

450 500 550

Temperature (°C)

Figure 6-2: Product distribution from the MTO in the temperature interval 450 - 550°C with several
different feed compositions

Comparing the concentrations of ethylene to the other hydrocarbons some trends is

shown in Figure 6-3. At ethene concentrations below 5%, oxygenate conversions is

below 95% and the trends in these data can not be directly compared to the

concentrations above 5%. As the ethene concentration increases, the concentration of the

higher hydrocarbons decrease with the largest hydrocarbons decreasing the most, this
 107

trend is mostly coupled to the reaction temperature. Propene seems to be relatively

independent of ethene and no definite trend can be observed. The scattering in the

propene and, to a degree, the C4 fraction relates to the changes in feed composition and

WHSV. The high amount of experimental data gives a good insight into the MTO

reaction and form a basis for modelling the reaction.

50 -
0 C2= V. C3=
• C2= V. C3H8
40- ° ° 0 0 cP0 • C2= V. C4
0
0 0 ° 6° a 0 • C2= V. C5
° °oo ° ° 0
• C2= v. C6+
0
0 r? °
c
30 - 8°
o
Si
i_ 5 »*% • •! •
m 20 - O
8• • • • • • •t ••
O 9
* «d?*
• •
•
{J • * • • •j % •
10-
% ®
0
^ 11 %

0- m- ,t^\'K\ 1 1
10 20 30
C2= (Carbon %)

Figure 6-3: Comparison between ethene and other hydrocarbon concentration for the MTO reaction

6.4 Conclusion
The MS have been shown to be able a valuable instrument to follow key components in

the MTO reaction with respect to oxygenate conversion and the undesired by-product

methane in real time. The MS can be a good analysis tool for the control of the

regeneration of the MTO reaction. It does, however, have limited use in following the
 108

main products due to the overlap in the fragments. MTO experiments in the temperature

range 450 - 550°C have shown some clear trends in the product distribution both with

respect to temperature dependency and relation between the individual components that

will help in the determination of a general reaction mechanism for the reaction.
 109

Chapter 7
Kinetic modelling of MTO reaction
over ZSM-5 influidbed*

7.1 Presentation of the article

The following article present the work related to the methanol to olefins reaction in

fluidized bed. The objective of the article is to investigate the MTO reaction kinetic and

develop a kinetic model.

Experimental data for the MTO reaction with different feed compositions of methanol

either pure or co-feed with argon or water is presented. Based on these experimental data

and experiments with 1-hexene feed a kinetic model is proposed based on the

hydrocarbon pool mechanism and implemented into a fluid bed model. The fluid bed

model is based on a two-phase model with a bubble and emulsion phase. The model is

shown to be able to model the system fairly well.

This article was submitted to I&EC Research

 110

7.2 Kinetic modelling of M T O reaction over ZSM-5 in fluid bed

Mads Kaarsholm , Finn Joensen*, Roberta Cennr*, Jamal Chaouki, Gregory

S. Patience
Department of chemical engineering, Ecole Polytechnique de Montreal, Canada

'Haldor Topsoe A/S, Denmark

7.2.1 Abstract

Reaction of methanol to olefins has been studied in a small scale fluid bed reactor over a

phosphorus modified ZSM-5 containing catalyst. Increasing the temperature from 400°C

- 550°C showed significantly changes in the product distribution with increased light

fractions while changes were not as pronounced when changing feed rate and

compositions. Feeding 1-hexene resulted in approximately the same product distribution

as with methanol suggesting that the reaction path of the olefins is coupled to

equilibrium reactions. A model based on the hydrocarbon pool mechanism is proposed

where the olefins are produced through equilibrium reactions with a larger hydrocarbon

species in the catalyst pores. The model predicts well the product distribution of the

olefinic species in the investigated temperature interval. Further experimental work is

required to adequately characterise the paraffin and C6+ fractions.

Corresponding auther. Tel: + 45 45 27 22 55

E-mail address: mkaa@[Link]
 Ill

7.2.2 Introduction

The Methanol To Olefin (MTO) reaction over acid zeolites was first discovered in the

late 1970's by researchers at Mobil when trying to convert methanol over ZSM-5 to

others oxygenated compounds . Large effort have since been taken to optimize the

reaction with respect to both olefin and gasoline production. Stocker gave an excellent

review of the different catalyst with respect to pore size, zeolite types and effects of

zeolite modifications with different metals.

ZSM-5 and SAPO-34 are among the most intensively studied catalysts for methanol

conversion. The main difference between the two catalysts is the pore and cage

structure: SAPO-34 has narrow pores with a diameter of 3.8A and large cages, larger

aromatic compounds are therefore allowed to form inside the cages but the narrow pores

inhibits aromatic intermediates from exiting the cages. ZSM-5 has larger pores with

diameters of 5.1 - 5.6A where larger molecules like aromatics can escape. The SAPO-34

zeolite therefore has a narrow product distribution of C1-C5 hydrocarbons while the

ZSM-5 zeolite gives products in the range of C1-C10. Due to the accumulation of large

aromatic species inside the cages of SAPO-34 the catalyst deactivates much faster than

the ZSM-52.

Studies of the MTO reaction carried out in fixed bed reactors3"6 have drawbacks of radial

and axial gradients and a moving coking front7. A fluidized bed would be more ideal

reactor for exothermic reactions with catalyst deactivation due to uniform temperature

and possibility of continuous regeneration. Further, coking will be uniform due to solid

circulation. Several investigations of the MTO reaction in fluidized bed have also been
 112

presented with both SAPO-34 and ZSM-5 zeolites8"11 showing good selectivity.

Constant catalyst activity and product distribution can also be obtained due the

continuous regeneration.

The kinetics of the MTO reaction have been the focus of many studies over the years.

Stocker summarizes the work on the pathway for the first C-C bond which has been

disputed for years. Keil10 gives a review of the kinetic modelling of the reaction which

has mainly been in terms of lumped models. Detailed models have also been

developed12"14. Recently the group of Bilbao at Universidad del Pais Vasco have

collected extensive amount of experimental data and modelled the MTO/MTG reaction

kinetics and deactivation on both ZSM-5 and SAPO catalysts5"6,8'15"17. This work is

reasonable successful at characterising the deactivation both as function of time on

stream and regeneration. The modelling of the product distribution has mainly been

focused on lumped fractions ignoring for the individual components. However, the

individual olefinc species have been considered for kinetic studies over SAPO-34, and a
1n

model consistent with the hydrocarbon pool mechanism was proposed . Part and

Froment developed a kinetic model based on elementary steps of carbenium ion

chemistry in which ethylene and propylene are the main products from methanol and

DME and with equilibrium reactions between C4 to Cg olefins. Schoenfelder et al.9

developed a kinetic model for MTO over ZSM-5 that predicts the separate olefins

ethylene, propylene and butylenes while lumping the paraffins and higher olefins. The

propylene and butylenes are formed by reaction with lower olefins and methanol. The

model predicts the experimental data well, but higher hydrocarbons than butylenes were
 113

not quantified in the experimental work and, therefore, been incorporated with the

aromatics fraction. Previous work on kinetic models generally neglect cracking of

olefins or exclude propylene and ethylene cracking/oligorimerisatrion reactions - though

Bos et al19 included direct reactions from both butylene and propylene to ethylene.

Buchanan et al.20 performed experiments of olefin cracking over ZSM-5 at 510°C with

C5= - C8= olefins and showed that C5= cracks to ethylene, propylene and some butylenes

while the higher olefins predominantly cracks to C3= - C5= olefins. These reactions have

been shown to be relatively fast at this temperature and it appears necessary to include

them in the reaction pool in order to model MTO correctly. Zhou at al.21 reacted

ethylene, propylene and n-butylene over SAPO-34 at 450°C in a small scale fixed bed

reactor with WHSV ranging from 1 to 424 h"1. They showed that the resulting product

distribution was close to equilibrium and proposed a kinetic model where all olefins are

in equilibrium with a carbenium ion lump. The model was able to reproduce their results

fairly well.

In this work experimental data from the MTO reaction over a phosphorus-modified

ZSM-5 catalyst conducted in a small scale fluid bed will be presented. Based on

available literature studies and 1-hexene experiments, a kinetic model is proposed. The

model is based on the hydrocarbon pool mechanism where the olefins are formed

through reversible reactions with a large hydrocarbon species. The fluid bed is modelled

as a two phase system with a bubble and an emulsion phase with reaction only in the

emulsion phase.
 114

7.2.3 Experimental

Thermocouple

Figure 7-1: Drawing of the glass fluid bed reactor - glass beads are used to elevate the catalyst for
catalyst amounts below lOOg

[Link] Equipment
MTO experiments were carried out in a 4.6 cm inner diameter glass fluid bed (Figure

7-1) with a 6.6 cm inner diameter disengagement zone. The gas was distributed through
 115

a quartz frit. The reactor was heated by two electrical band heaters: one in the reaction

zone and one in the disengagement zone. Both were controlled based on thermocouples

mounted on the exterior surface of the quartz reactor. A thermocouple, inserted in the

middle of the reaction zone recorded the temperature. Liquid feed are delivered via a

dual piston pump with feed steps of 0.01 ml/min and gas was fed from one of two lines

both controlled by Brooks mass flow controllers with a maximum flow rate of 2NL/min.

The reactor feed composition may be varied by a Valco 8-port valve. The feed was

preheated in a sand bath to 300°C before entering the fluid bed. All flows and reactor

temperatures were controlled by a computer. The exit gas was analysed by a Hiden MS

QIC-20 and a Varian CP3800 GC. A diagram of the system is shown in Figure 7-2.

jfc—txt-
V

Vent

PKI
m
rf?
;r4E553-
MS
-ec

ra
jcyLc><!~

SSI KM. FBTOft ( W

__|
I
,i TO nmjuu
I PT 0-?00
I PSI K55255H5555
e- S3

Figure 7-2: Schematic diagram of the experimental setup

[Link] Catalyst
The catalyst was made of 10% CBV28014 (Zeolyst) imbedded in a Si/Al matrix

consisting of Catapal B, Levasil 100s/30% and kaolin which was spray dried and
 116

calcined in air at 550°C for 4 h. The catalyst was contacted with an aqueous

(NH3)2HPC>4 solution and then dried and calcined. The resulting catalyst contained 1.5%

phosphorus and had an average particle size of 108 um. The phosphorous-modification

has been done to decrease methane formation. The introduction of Catapal, Levasil and

kaolin increase methane formation and impregnation for the fluid bed catalyst

significantly decrease the methane formation from these components.

The particle size distribution was measured on a Horiba LA-950. The minimum

fluidization velocity at ambient temperature and pressure was 0.00507 m/s in a 3"

transparent fluid bed where the tapped particle density and density at minimum

fluidization, were also measured. The catalyst properties are listed in Table 7-1.

Table 7-1: Particle properties of the catalyst

dp <45 0.4 200 < d p > 229 2.9
45 <dp< 68 4.7 229 < d p < 262 1.6
68 <d p < 89 16.6 262 < d P < 350 1.3
89 <dp< 102 15.5 dp 108 (am
102 <dp< 117 16.7 Umf 0.005 m/s
117 <dp< 133 15.3 Pp 1270 kg/m3
133 <dp< 153 11.9 Ptapped 855 kg/m3
153 <dp< 175 8.1 Pmf 744 kg/m3
175 <dp< 200 5.0 Gmf 0.405

[Link] Experimental conditions

The fluid bed was loaded with 20 to 330g of catalyst. The heating elements were

positioned 4 centimetres above the distributor and the temperature control of the

catalytic bed was therefore not optimal for inventories less than lOOg. With larger

inventories the temperature just above the distributor was measured to ensure that the

temperature gradient over the bed was low (between 3-7°C). For experiments conducted
 117

with below lOOg of catalyst glass beads were used to elevate the bed. The gas velocity

was at all times below umf of the glass beads thereby ensuring the catalyst remained

above the beads. The feed consisted of pure methanol and mixtures of methanol with

water or argon. The total flow rates range from 0.5 - 3.6 NL/min corresponding to

superficial gas velocities between 1.3 - 10 cm/s at operating conditions. All experiments

were conducted between 400 - 550°C.

7.2.4 Experimental data

Prior to studying the MTO kinetics, the stability of the catalyst with regard to the

product distribution was tested in order to establish the influence of coking on activity.

An experiment carried out at WHSV of 2.4h_1 with pure methanol at 550°C showed that

the product distribution was stable for the first 15 hours in the fluid bed. Other

experiments were performed at 450, 500 and 550°C with a WHSV of 0.4h_1 with pure

methanol feed for 23 hours. The catalyst tested at 550°C was regenerated in air for 14

hours and the experiment was repeated. The product distribution before and after

regeneration was similar indicating that irreversible deactivation was minimal. Samples

of the catalyst were taken from the experiments and the carbon content analysed by

TGA (Thermogravimetric analysis) measurements the maximum carbon was 4.8 wt %

after 23 hours at 550°C. The results are given in Table 7-2. The selectivity to carbon is

also higher at 550°C where approximately 1.5% of the feed is converted to carbon. All

experiments reported hereafter were conducted over a period of 4-5 hours where the
 118

effect of reversible/irreversible deactivation was negligible. Three GC traces were taken

for each experiment.

Table 7-2: Carbon deposition on the MTO catalyst

Temperature TOS (h) Wt % carbon selectivity to Carbon %
550 4 0.77 1.38
550 8 1.47 1.33
550 22 4.80 1.67
500 4 0.04 0.07
500 8 0.55 0.49
500 22 1.84 0.61
450 4 0.46 0.82
450 8 1.56 1.41
450 23 1.34 0.42
550* 4 0.74 1.32
550* 8 1.47 1.33
550* 22.5 4.40 1.46

The experimental data are given in Table 7-3. In experiment No. 20, 100 ml/min H2

were added to the feed gas to investigate the effect of H2 on the reaction. No change in

the product distribution was observed with the addition of hydrogen indicating that

hydrogen does not influence the reaction mechanism by hydrogenation of the olefins.
Table 7-3: Experimental results from the fluid bed experiments - product distribution is given in carbon % * 100ml H2 was co-feed with methanol and argon
Exp
no. Temp. MeOH (I) ml/min Ar (g) ml/min Water (I) ml/min Cat, g Methane Ethene Ethane Propene Propane DME MeOH C4 C5 C6+
1 402 1.84 139.5 0 220 0.43 8.36 0.15 20.35 2.96 3.69 2.49 24.65 16.49 20.43
2 450 1.84 139.5 0 220 0.81 7.75 0.24 29.28 3.36 0.62 0.89 26.03 16.13 14.89
3 500 1.84 139.5 0 220 2.15 12.84 0.67 34.85 3.47 0.00 0.30 23.70 12.16 9.87
4 546 1.84 139.5 0 220 3.93 19.09 1.14 35.49 2.62 0.00 0.10 20.22 6.55 10.87
5 550 0.92 558 0 220 4.22 18.20 0.94 39.02 2.46 0.00 0.09 20.16 9.13 5.78
6 450 0.92 558 0 220 1.04 7.54 0.19 34.11 2.37 0.21 1.40 24.71 15.45 12.99
7 451 3.68 139.5 0 220 0.80 6.09 0.17 30.74 2.04 3.65 3.06 23.58 14.99 14.88
8 450 0.92 0 0.92 220 0.96 6.75 0.17 37.37 1.90 1.36 5.09 22.64 14.24 9.53
9 553 1.84 0 1.84 220 6.12 13.42 0.73 38.27 1.63 0.78 1.87 19.78 11.98 5.42
10 552 1.84 0 0 330 2.91 19.22 1.22 38.45 4.21 0.00 0.07 21.16 8.05 4.71
11 549 1.84 0 1.84 330 2.60 21.71 0.51 40.88 1.76 0.19 0.20 18.48 9.24 4.43
12 550 1.84 1144 0 330 2.85 21.20 0.51 40.17 2.91 0.00 0.04 20.37 8.04 3.91
13 450 1.84 1144 0 330 0.80 9.02 0.16 31.43 3.43 0.09 0.15 27.73 15.35 11.85
14 499 1.84 0 0 330 1.34 11.21 0.46 32.71 4.63 0.01 0.13 27.01 13.97 8.52
15 499 1.84 1144 0 330 1.23 11.96 0.30 38.05 2.78 0.02 0.16 24.04 12.86 8.62
16 502 1.84 0 0 220 2.00 14.14 0.71 34.93 5.52 0.00 0.04 25.92 10.88 5.86
17 500 1.84 1144 0 220 1.17 12.64 0.23 39.51 2.66 0.00 0.15 22.16 10.87 10.60
18 501 0.92 558 0 220 1.62 15.22 0.56 36.57 3.57 0.00 0.11 22.88 10.15 9.32
19 500 0.92 0 0.92 220 1.43 13.26 0.37 41.07 2.62 0.18 0.45 22.40 12.03 6.20
20* 500 0.92 558 0 220 1.53 13.28 0.41 37.66 3.17 0.00 0.31 23.18 11.53 8.94
21 500 0.92 558 0 220 1.58 14 0.54 38.35 3.44 0 0.63 23.7 11.39 6.35
22 554 0.92 0 0 220 4.89 22.06 1.71 39.50 2.59 0.00 0.01 19.76 6.04 3.44
24 501 0.92 558 0 110 2.95 15.22 0.34 37.27 1.83 0.42 2.72 20.07 11.55 7.63
25 500 0.92 0 0 110 3.23 15.25 0.58 40.29 2.78 0.00 0.75 22.98 8.73 5.40
26 500 1.68 2288 0 20 0.72 2.63 0.06 20.41 0.31 8.29 25.65 13.69 10.90 17.36
27 453 2.50 0 0 50 0.54 5.64 0.06 26.60 1.35 2.10 6.76 21.53 15.47 19.88
28 501 2.50 0 0 50 0.91 7.14 0.11 35.40 1.34 0.70 3.82 21.78 16.00 12.78
29 550 2.50 0 0 50 3.29 6.98 0.26 31.95 1.06 2.13 5.46 19.68 14.62 14.56
31 500 1.68 0 0 50 0.97 8.21 0.14 33.89 1.78 0.48 2.38 22.46 15.16 14.54
32 501 0.84 0 0 50 0.93 9.33 0.19 32.74 2.32 0.26 1.40 22.49 14.55 15.78
34 502 1.68 2288 0 50 0.84 5.85 0.07 36.23 0.66 2.78 3.11 21.50 15.42 13.54
35 450 0.92 558 0 110 0.95 7.82 0.14 30.29 1.91 0.09 8.24 23.01 13.74 13.81
36 550 0.92 558 0 110 2.56 18.69 0.40 36.89 1.75 0.00 3.13 18.80 8.43 9.34
 120

Feeding 1 -hexene instead of methanol to the catalyst results in a product distribution

similar to the one obtained from methanol feed, indicating that cracking and

oligomerization are an important part of the reaction scheme. This is also in agreement

with the literature, where investigations on the reaction of C3= - Cs= and hexane over

ZSM-5 have shown a product spectrums of C2 - C10 together with parafms and

aromatic ' " . Recently Zhou et al. have shown that feeding ethylene, propylene or n-

butylene over SAPO-34 gives approximately the same product distribution which is

close to the thermodynamic equilibrium. As SAPO-34 and ZSM-5 are both acidic

catalysts the reaction mechanism of the light olefins interconversion is likely to be

similar and a similar pattern should be expected for ZSM-5.

 121

50 - • Ethene
A
• Propene
- Ethene EQ
_•
40 - - -
• Propene EQ
•

I 1
•
•
1
9 ••
•• •
° 30- |
C _
0 • • „ "
.Q
i -
^* 1
(0 20 - • • * •
O * 1
*
| ^
10 -
•
J. -•

•
•

0 - •

380 430 480 530 580

Temperature (C)
40 -t
B

30 A

20 H
CO
o
10

380 420 460 500 540 580

Temperature (C)
Figure 7-3: Comparison between experimental data and equilibrium composition of the C2 - C6 olefins
in the temperature range 400 - 550CC a) ethylene and propylene b)C4, C5,C6
 122

Due to the results obtained in the 1-hexene experiment, the work of Norval et al.25-26

who have studied the equilibrium of the MTG process over ZSM-5 and the recent work

of Zhou et al.21 it is of interest to establish whether or not the olefins reach equilibrium

concentrations. The equilibrium distribution between C2 - C6 olefins have been

calculated with HCS Chemistry 5 in the temperature range 400 - 550°C where all

isomers of the olefins have been considered in the calculations. Figure 7-3 shows that

the product distribution and the equilibrium concentration. The experimental data is far

from equilibrium concentrations, especially propylene which is higher than the

equilibrium concentration, while the C4 and C5 are lower. On the other hand, the trends

between the equilibrium calculations and experimental data are in agreement.

50 1 T 100

40 H 98

6"
5 s 30 H + 96 c
o
O "w
L.
>
8 20 H A + 94 £
o
o
10 D 92
Q
O

•
90

Residence time (s)

Figure 7-4: Product distribution vs. residence time with methanol/Argon feed (molar ratio 1:1) over ZSM-
5 at 550°C. A: conversion , • : ethylene,*: propylene, •: propane, • : C4? • : C5, O: C6+
 123

50 -, x 100

40 - • + 98
n
A
SS 3 0 - + 96
c O
o
•£ i >
w __ * + 94 c
O 20 - o
i
O

10 - £ + 92
*
••
90
2.5 3.5 4.5 5.5
Residence time (s)
Figure 7-5: Product distribution vs. residence time with methanol/Argon feed (molar ratio 1:1) over ZSM-
5 at 500°C. A: conversion , • : ethylene,•: propylene, •: propane, • : C4> • : C5, O: C6+

Figure 7-4 - 7-6 show the product distribution as function of time with an equimol

methanol/argon feed at residence times between 2.5 and 6 seconds. The product

distribution is relatively constant and full conversion is obtained after 4 seconds

regardless of temperature. The product distribution at each temperature is relatively

insensitive to the residence time in the interval investigated. Larger changes are evident

with changes in temperature: higher temperatures result in an increase in the ethylene

and propylene and a decrease in the higher fractions. The change is largest for ethylene

and the C6+ fractions. Figure 7-7 shows the product distribution as function of residence

time with pure methanol feed. At short residence times, the concentration of higher

hydrocarbons increases suggesting that they are intermediates. This was also found in a
 124

previous fixed bed study where an increased Cs+ fraction was observed as the catalyst

began to deactivate indicating that they are intermediates .

50 - A 100
4

40 -
95
•
3* 30 - . • c
C
A
A.
o
o A
A
!_
JQ >
O20- O
D • O
c3 o 85
10 - *
4> •
•
• •
0- 1 1 1 80
2.5 3.5 4.5 5.5 6.5
Residence time (s)

Figure 7-6: Product distribution vs. residence time with methanol/Argon feed (molar ratio 1:1) over ZSM-
5 at 450°C. A: conversion , • : ethylene,^: propylene, •: propane, • : C4; • : C5, O: C6+

50 T 100

40 + 98

5 s 30 H 96 c
o
*
'w
$ A.
>
20 4 + 94 C
o
O
10 H + 92

• •
90
2 4 6 10
Residence time (s)

Figure 7-7: Product distribution vs. residence time with pure methanol feed over ZSM-5 at 500°C. A:
conversion , • : ethylene,^: propylene, •: propane, A: C4j • : C5, O: C6+
 125

7.2.5 Fluid bed model

The fluid bed was modelled assuming gas rises predominately in the bubble and the

reaction takes place in the catalyst, in the emulsion phase, with interphaes mass transfer

between the two. This two phase model is well documented in the literature27. The

model was used recently by Werther and Hartge28 to characterize an industrial fluidized

bed reactors and by Abba et al.29 to describe fluidization from bubbling conditions to

fast fluidization. For modelling of the fluid bed the following assumptions have been

made:

• Gasflowonly in the axial direction - dispersion in the radial direction is not

considered (with the exception of interphase mass transfer)

• No catalyst in the bubble phase

• The activity of the catalyst is considered constant e.g. no deactivation due to

the short reaction time compared to the deactivation time of the catalyst

The mass balance around the bubble and emulsion phases for the model can be written
as:
Bubble phase:

^^=-«Jcw-c„>r-i_. (7-i)
dh 1 - Ye,c:^

Emulsion phase:

d{{l S ,Cj
' b = ^K-Cj-T-^
v +W(-n) (7-2)
dh " "•' '•" \-Y „,
e,Cr

Where Tis the gas expansion due to reaction, which considered to go to the bubble

phase by formation of bubbles in the emulsion phase. The implemented equations are
 126

given below and a schematic drawing is shown in Figure 7-8. The equations were

discretized using a block element approach with the number of blocks as a parameter the

implemented equations are given in Equation 7-3 and 7-4.

0 = -Fb,nyb,i,n + Fb,„-iyb,i,„-i -j^KKbe{cb,i,„ - c ^ J + X W - r , ) ) - - (7-3)

J e.i.n

•y e,C*x,n

0 = -F„y.j* +F
e,^yb,i,„-i +T^KKbe{cbMn-CeJ-fj{W„(-ri))- *'•>* +W„(-r,) (7-4)

Ci

Bubble phase Emulsion phase

Kbe

x 1
e,i
II

Wb -• limb

Cb,i + •• o e,i

-K>-

C >o u 0

Figure 7-8: Schematic drawing of the fluid bed model.

 127

Optimization of parameters has been done with the use of the simplex method and the

error calculations have been evaluated from Equation 7-5

N . .

/ i /,w,exp V i,n,cal /,«,exp /

R- = 1 - ^~N (7-5)
/ irincxn (,exp /
n

The hydrodynamic correlations used to in the fluid bed calculations are given in Table

7-4. The mass transfer coefficient correlations used in this work is the one from Sit and

Grace . The bubble diameter has been calculated form the correlation of Mori and
-i i

Wen . The used catalyst is Geldart A particles and the minimum bubbling velocity is

therefore used in the correlations for the gas in the emulsion phase. The molecular

diffusion coefficient is not known and has been estimated to be 0.4 cm2/s, based on the

average binary diffusion coefficient between mixtures of methanol, DME and water.

Since a porous plate distributor and relatively low gas velocities are used in the work the

initial bubble size is small, below 5 mm in the first centimetre of the bed. The mass

transfer coefficient in the first part of the bed is therefore high and no limitations on the

mass transfer are expected. General values of the mass transfer coefficient are in the

range of 3 - 15 s"1 at the exit of the bed with the lowest value at the highest gas velocity

and catalyst inventory.

 128

Table 7-4: Hydrodynamic correlations

Variable Correlation Ref
Mass ADe
*mi fu,^ Sit and
transfer k,be = + mf b A-be
. —
Grace30
v

coefficient 'b J

Bubble d
b=dbm-{dbm-dbo)Q*V
(cm) Mori and
v A j
diameter Wen 31
db0 = 0.00376(w0 - umb ) 2 , t/im = 0.652(4 (wo - ««* ))°'4
Bubble . _.,/ , xi/
M =M
velocity * ° " u™» + 0J1 l^d» >
Minimum d
fluidization umf =
V (33.7)2+0.0408
lPg(ps-Pg)s -33.7 Wen and
d p M z Yu32
velocity
Minimum umb = 2 3 0 0 / Q / " ^ exp(0.716F) Abrahamsen
and
bubbling Umf d°/g^(pp-pjm Geldart 33
Bubble g _ U0 U
mb
fraction

7.2.6 Kinetic model

Several kinetic models have been proposed for the MTO reaction, mostly over ZSM-5

and SAPO zeolites5'8"9'14'17"19'34"36. The main difference in the models based on ZSM-5

and SAPO is the addition of C6+ compounds in the models with the ZSM-5 zeolite.

Further it has been shown through 13C experiments that the catalytically active reaction

intermediates in ZSM-5 and SAPO-34 are somewhat different37'38. Transient 12

C/13C

methanol conversion studies on ZSM-5 showed that, upon switching from CH30H to
13
CH30H, the 13C content of ethylene closely follows that of the aromatics and that the

higher olefins to a considerable extent are formed from methylation and interconversion.
 129

50 n
• Methanol

40 H
• Methanol + Hexene
• Hexene n
" 30 H
Hltl
o
CO 20
o
10

^^

/
/ " / / / ^
Figure 7-9: Product distribution of reaction over ZSM-5 with three different feed compositions and a total
inlet flow rate of 1.57 L/min at 500°C 1) Methanol/Argon molar ratio 1:2 2) Methanol/1 -hexene/Argon
molar ratio 9:1:20 and 3) 1-hexene/Argon molar ratio 1:2

The experiments with 1-hexene in argon feed over ZSM-5, previously described, have

shown that the product distribution is quite similar to the product distribution with pure

methanol feed. Figure 7-9 shows the product distribution of experiments made with

methanol and 1-hexene diluted in Argon feed over ZSM-5 catalyst with constant inlet

flow rate. The product distributions are similar, although there are some changes in the

individual olefin fractions. Feeding 1-hexene yields slightly more propylene, butylenes

and hexenes compared to feeding methanol and less of the other olefins. Methane is

formed from methanol/DME and not from the olefin reactions which also have been

found from fixed bed experiments25. Since no methane is formed, formation of pentenes

most likely happens from cracking of higher olefins formed by oligomerization and not
 130

from direct cracking of 1-hexene. Propane and aromatic compounds are also observed in

the exit gas with 1-hexene but to a lesser degree than with methanol.

Based on the product distribution of the experiments with methanol and 1-hexene feed

the product distribution looks like it is highly dependent on the cracking and

oligomerization reactions and the exact route from methanol to hydrocarbons might

therefore not be crucial to the product distribution. A route from higher olefins to

ethylene and propylene should also be included in the kinetics, which in most cases is

not present in the current available models.

The kinetics of the MTO reaction are complicated, involving oxygenates, paraffins,

olefins and aromatics, with the main components being the light olefins ethylene and

propylene. The kinetic model is developed to enable the prediction of the main fractions

of the MTO process. Ethylene and propylene are taken as individual components while

the olefins and paraffins in the C4 and C5 fractions are lumped together.

The kinetic scheme that has been used to model the reaction is based on fast reaction

between methanol and DME to give an equilibrium composition. Methanol and DME

react to form a hydrocarbon pool component (Cx+), which in this work has been taken as

decene. Cx+ is considered to react to each of the C2-C6 olefins through reversible

reactions. The Cx+ is treated as a component related to the catalyst and therefore only

present in the emulsion phase.

 131

CJ^4 CjH6 — CgHi2 + 3 CjHff

\ /

2 MeOH *=* DME + HjD—+ C/ *=* C4

/ \
Cr% *r C "r /TJC/ C^ + Lj
Figure 7-10: Schematic drawing of the kinetic model

Reactions for the formation of aromatics have been taken into account and are assumed

to form from propylene and the excess hydrogen goes to the formation of paraffins. The

aromatic compound in this work is taken as trimethylbenzene and the paraffin as

propane. Methane is assumed to be formed directly from methanol and the extra carbon

present in the reaction from methanol to methane is assumed to go to carbon formation

since CO and CO2 was not measured in amounts that could justify reaction to these

components. All reactions with the exception of the MeOH/DME equilibrium have been

regarded as first order reactions and the influence of water have also been taken into

account the reactions are listed in Table 7-5 a schematic drawing is shown in Figure

7-10.

The influence of water in the kinetics is taken into account in terms of equation 7-6; the

constants in Kw have been found by optimization. In the calculation Z is in the range 0.8

- 1, the effect large but it does improve the predictions especially when water is co-feed.

1 •12.1
Z = where Kw =3.6exp (7-6)
(l + KwCw) R T Toy
t
 132

The temperature dependency of the reactions is taken into account with the Arrhenius

equation with a reference temperature of 500°C.

E„ 1__L
k = k* exp (7-7)
R T T*

Table 7-5: Kinetic model with optimized parameters for k0 and activation energy k* is calculate with a
reference temperature of 500°C
No Reaction Reaction rate ko k* Ea
(m3/(kg-s) (kcal/mol)
1 2CHiOH <-> C2H5OH Considered in
Equilibrium
2 l0CH,OH^C;+l0H2O — r2 — k2L,MeOHZ 1.7 2.83-10"' 6.28
3 5C2H5OH^Cx+5H20 — r3 = k3CDMEZ 9 4.43-10"4 15.2
4 C+x -> 5C2H4 — r4 = ft4Cc+Z 2.70-10° 0.508 20.2
5 SC2H4 -> c; — r5 =kiLCiHZ 14.6 1.63-10"° 21.0
6 C^fC.H, 2.76-10* 1.56 11.46
~~r6 = k(Pc*
7 fc,H6->c; — r7 — k7LCiH6Z 65.0 2.05-10"4 19.41
8 C+x -> 2.5C,HS ~ rs =
k%Lc+Z 87.1 0.940 6.94

9 2.5C4HS -> c ; — r9 = kgCc^HijZ 0.23 2.31-10"° 14.12

10 C 2C H =
0.73 0.467 0.69
x ~> i W ~ 1o ^IO^C*
11 2C5Hl0 -> C+ — r u = kllLCiHmZ 1.33-10* 1.03-10"° 3.92
12 C+ -^1C H — rl2 — knL*c+z 2.52 0.893 1.59

13 _r —v f 7 13.5 2.44-10"" 5.23

3 ^ 6 - " 12 ^ W —
'13 ""B^Cyy^
14 6C3H6 —> C9Hl2 + 3C3/f8 — r14 = kuLc^H6Z 1.11-10"* 4.23-10"° 1.48
15 2CH30H ->CH4+C + 2H20 3.46 1.11-10"'' 8.80
~ ^15 —k-X'-S-'MeOH^

7.2.7 Modelling results

The parameters in the model were estimated by minimizing Equation 7-5 with simplex.
*

The k and Ea values were optimized simultaneous with a reference temperature of

500°C. Since the C6+ fraction was measured as a lump fraction and the aromatic have
 133

not been measured specifically the C6+ and aromatic fractions in the model account for

the total measured C^+ fraction. The methanol reaction to DME is considered to be in

equilibrium due to high reaction rate at the reaction temperature. The results are

presented in Table 7-5. The rate constants from decene to olefins at 500°C are generally

4-5 orders of magnitude higher than the reverse reaction. Furthermore, the reaction from

methanol to decene is approximately 10 times faster than from hexene to decene.

Calculations of the 1-hexene experiments with the obtained kinetics shows only 11% 1-

hexene conversion versus the 89% conversion observed during the experiment. It is

evident that the model does not accurately describe the cracking of the higher olefinic

species and further investigation into this part of the kinetic model is needed for a more

accurate description.

The activation energy values of the Cx+ to olefin reactions decreases with an increase in

olefin mass. This agrees with experimental findings where increasing temperatures leads

to higher concentrations of the lower olefins. The values of the activation energies are

higher than the values reported by Gayubo et al.17 which were in the range of 55 - 70

kj/mol for ethylene and propylene and 20-25 kj/mol for the higher olefins. Their work

was based on SAPO catalyst and the difference in the active sites on the catalyst

compared to the ZSM-5 might influence the activation energies to some extent.

Looking at the difference in the forward and backward reactions from the hydrocarbon

pool species to the olefin, it is clear that the reverse reaction is much slower. Zhou et

al.21 have reported reaction rates for similar reactions in a SAPO catalyst at 450°C and

they reported that the reaction rates were all of the same order of magnitude. The large
 134

difference in reaction rates could be due to a large fraction of higher components in the

SAPO catalyst remain in the pores compared to the ZSM-5.

0 10 20 30 40 50 60
Measured (Carbon %)
Figure 7-11: Comparison of the calculated product distribution to the measured data.

The calculated and experimental data are compared in Figure 7-11. The model capture

the overall trends quite well but requires improvements to simulate the species

concentrations. The model is poor in representing the trends for the C6+, paraffin

fractions and oxygenate conversion.

 135

50 - " ' " " " " " ~ " ••> . . . . . . . . .A . . . . ----------- ---"A -rT 100

A
40 - • •
• - --95
*»8
C
S^ 30 - O
£
O --90 "w
i >
* c
• • o
--85 o
10 -
8^~___ _ 0 n
-- — — _ _ _ O
Q

o- , i
. 80
2 3 4 5
Residence time (s)
Figure 7-12: Measured (points) and calculated (lines) data at 550°C with methanol/argon (ratio 1:1) feed
as function of residence time in the fluid bed. A(- •): conversion: • ( — ) : ethylene,B(—): propylene,
A(»-): C4, • ( - ): C5, O ( - ) : C6+.

50 -i - A T-100

40 H
-. +95
C
30 H o
o + 90 'w
1"
« 20 >
O c
o
o
10 H

Residence time (s)

Figure 7-13: Measured (points) and calculated (lines) data at 500°C with methanol/argon feed (ratio 1:1)
as function of residence time in the fluid bed. A( •• ••): conversion: • ( — ) : ethylene, •(—): propylene,
A ( - - ) : C4, • ( ): C5, O ( - ) : C6+.
 136

40 n x 100
A

30 H + 95

c
£ i" o
2 20 + 90
>
o c
~~Q o
O
10 A + 85

T" 80
5 6

Residence time (s)

Figure 7-14: Measured (points) and calculated (lines) data at 450°C with methanol/argon feed (ratio 1:1)
as function of residence time in the fluid bed. A(—•): conversion: • ( — ) : ethylene,B(—): propylene,
A ( - ) : C4> • ( -): C5, 0(-~): C6+.

Figure 7-12 - Figure 7-14 compare the experimental data and the calculations with

methanol/argon feed at three different temperatures. Overall, the calculations of the

olefins are in agreement with the experimental data with the exception of the C6+

fraction. This could be because the C6+ fraction lumps both aromatics and large olefins.

The mechanism for these fractions most certainly are different and therefore require

separate kinetics. However, they only represent a small fraction of the overall product.

Figure 7-15 shows the calculated and measured data at 500°C with a feed consisting of

methanol. The same trends as for the methanol/argon feed are observed. Overall, the

model is able to fit the experimental data fairly well from 450°C to 550°C especially

with respect to the main olefin species.

 137

T-100

o
0)
>
c
o
o

Residence time (s)

Figure 7-15: Measured (points) and calculated (lines) data at 500°C with methanol feed ratio 1:1 as
function of residence time in the fluid bed. A( -••): conversion: • ( — ) : ethylene,>(—): propylene, A(—•):
C4, • ( ): C5, 0(—): C6+.

7.2.8 Conclusion

Experimental data for the MTO reaction in fluid bed show that methanol and 1-hexene

feed yield approximately the same product distribution. Methane formation is almost

non-existent with 1-hexene feed compared to methanol feed indicating that methane is

predominately formed from oxygenates. When oxygenates are converted to

hydrocarbons, they follow the same reaction mechanism as with 1-hexene feed and the

olefin composition approaches an equilibrium composition. Equilibrium calculations,

however, show that much less light olefins and more higher olefins should be produced.

This could be due to the narrow pores in the ZSM-5 which could allow easier passage
 138

for the light olefins out of the catalyst thus shifting the product distribution towards

lighter products.

A model for the MTO reaction over ZSM-5 is proposed. The model is based on the

hydrocarbon pool mechanism and consists of 15 reactions where all olefins are formed

through equilibrium reactions with a large hydrocarbon species. The model predicts the

experimental data fairly well for the olefins but require improvements with respect to the

paraffin's and C6+ fractions.

7.2.9 Nomenclature
At = Cross-sectional area of the fluid bed (m2)

C - Molar concentration (kmol/m3)

c+ X
= Hydrocarbon pool species (Decene)

D = Molar diffusion coefficient (m /s)

D, = Reactor diameter (m)

dp = Particle diameter (m)

db = Bubble diameter (m)

dbo = Initial bubble diameter at the distributor (m)

dbm = Maximum bubble diameter (m)

Ea = Activation energy (kJ/mol)

F = Fine fraction

F = Flow rate (kmol/s)

g = gravity (m/s2)

h = Height in bed (m)

Keq = Equilibrium constant

ki = Reaction rate constant (m6/kmol-s)

Kbe = Bubble to emulsion mass transfer coefficient (s 1 )

kbe = Bubble to emulsion mass transfer coefficient (m/s)

N = Total number of components in the system

r; = Reaction rate of component i (kmol/kg-s)

R = Ideal gas constant (kcal/mol-K)

T = Temperature (K)

Umf = Minimum fluidization velocity (m/s)

Umb = Minimum bubbling velocity (m/s)

Ub = Bubble velocity (m/s)

Uo = Superficial gas velocity (m/s)

-5
Ve = Volume of emulsion (m )

w = Catalyst weight (kg)

y = Mole fraction

Greek symbols

T = Gas expansion due to reaction

8 = Bubble phase volumetric fraction

=
£mf Void fraction at minimum fluidization

Scat = Void fraction of the catalyst

\i = Gas viscosity (kg/m-s)

pp = Particle density (kg/m 3 )

Pb = Bulk density (kg/m 3 )

ptapped = Tapped density (kg/m 3 )

pmf = Minimum fluidization density (kg/m 3 )

pg = Gas density (kg/m )

Subscripts

e = Emulsion Phase

b = Bubble Phase

i = Component index

n = CSTR in series

exp = experimental data

cal = calculated value

 141

7.2.10 References

(1) Chang, C. D.; Silvestri, A. J. The conversion of methanol and other O-compounds to

hydrocarbons over zeolite catalysts. Journal of Catalysis. 1977, 47, 249

(2) Stocker, M. Methanol-to-hydrocarbons: catalytic materials and their Behaviour.

Microporous and Mesoporous Materials. 1999, 29, 3

(3) Chang, C. D.; Chu, C. T-W. Methanol conversion to Olefins over ZSM-5. [Link]

Distribution. Journal of Catalysis. 1984, 86, 297

(4) Park, T-Y.; Froment, G. F. Kinetic Modeling of the Methanol to Olefins Process. 2.

Experimental Results, Model Discrimination, and Parameter Estimation. Ind. Eng.

[Link]. 2001,40,4187

(5) Gayubo, A. G.; Anuayo, A. T.; Olazar, M.; Vivanco, R.; Bilbao, J. Kinetics of the

irreversible deactivation of the HZSM-5 catalyst in the MTO process. Chem. Eng. Sci.

2003, 58, 5239

(6) Mier, D. Aguayo, A. T. Atutxa, A. Gayubo, A. G. Bilbao, J. Kinetic study of the

simultaneous cracking of paraffins and methanol on HZSM-5 zeolite catalyst.

International Journal of Chemical Reactor Engineering, 2007, 5, A60

(7) Kaarsholm, M.; Joensen, F.; Nerlov, J.; Cenni, R.; Chaouki, J.; Patience, G. S.

Phosphorus modified ZSM-5: Deactivation and product distribution for MTO. Chem..

Eng. Sci. 2007, 62, 5527

(8) Gayubo, A. G.; Aguayo A. T.; Alonso, A.; Bilbao J. Kinetic Modeling of the

Methanol-
 142

to-Olefins Process on a Silicoaluminophosphate (SAPO-18) Catalyst by Considering

Deactivation and the Formation of Individual Olefins. Ind. Eng. Chem. Res. 2007, 46,

1981

(9) Schoenfelder, H.; Hinderer, J.; Werther, J.; Keil, F. J. Methanol to olefins -

predictions of the performance of a circulating fluidized-bed reactor on the basis of

kinetic experiments in a fixed-bed reactor. Chem. Eng. Sci. 1994, 49, 5377

(10) Keil, F. J. Methanol-to-hydrocarbons: process technology. Microporous And

Mesoporous Materials. 1999, 29, 49

(11) Chen, J. Q.; Bozzano, A.; Glover, B.; Fuglerud, T.; Kvisle, J. Recent advancements

in ethylene and propylene production using the UOP/Hydro MTO process. Catalysis

Today. 2005, 106, 103

(12) Mihail, R.; Straja, S.; Maria, G.; Musca, G.; Pop, G. Kinetic Model for Methanol

Conversion to Hydrocarbons. Chem. Eng. Sci. 1983, 38. 1581

(13) Park, T-Y.; Froment, G. F. Kinetic Modeling of the Methanol to Olefins Process. 1.

Model Formulation. Ind. Eng. Chem. Res. 2001, 40, 4172

(14) Park, T-Y.; Froment, G. F. Kinetic Modeling of the Methanol to Olefins Process. 2.

Experimental Results, Model Discrimination, and Parameter Estimation. Ind. Eng.

Chem. Res. 2001, 40, 4187

(15) Gayubo, A. G.; Aguayo, A. T.; Castilla, M.; Olazar, M.; Bilbao, J. Catalyst

reactivation kinetics for methanol transformation into hydrocarbons. Expressions for

designing reaction-regeneration cycles in isothermal and adiabatic fixed bed reactor.

Chem. Eng. Sci. 2001, 56, 5059

 143

(16) Gayubo, A. G.; Aguayo, A. T.; Castilla, M.; Moran, A. L.; Bilbao, J. Role of water

in the kinetic modelling of methanol transformation into hydrocarbons on HZSM-5

zeolite. Chemical Engineering Communications. 2004, 191, 944

(17) Gayubo, A. G.; Aguayo, A. T.; Sanchez del Campo. A. E.; Tarrio, A. M.; Bilbao, J.

Kinetic modelling of methanol transformation into olefins on a SAPO-34 catalyst. Ind.

Eng. Chem. Res. 2000, 39, 292

(18) Park, T-Y.; Froment, G. F. Analysis of fundamental reaction rates in the methanol-

to-olefins process on ZSM-5 as a basis for reactor design and operation. Ind. Eng. Chem.

Res. 2004, 43, 682

(19) Bos, R. A. N.; Tromp, P. J. J.; Akse, H. N. Conversion of methanol to lower olefins.

Kinetic modelling, reactor simulation, and selection. Ind. Eng. Chem. Res. 1995, 34,

3808

(20) Buchanan, J. S.; Santiesteban, J. G.; Haag, W. O. Mechanistic considerations in

acid-catalyzed

cracking of olefins. Journal of Catalysis. 1996, 158, 279

(21) Zhou, H.; Wang, Y.; Wei, F.; Wang, D.; Wang, Z. Kinetics of the reactions of the

light alkenes over SAPO-34. Applied Catalysis A: General. 2008, 348, 135

(22) Tabak, S. A.; Krambeck, F. J.; Garwood, W. E. Conversion of propylene and

butylene over ZSM-5 catalyst. AIChE Journal. 1986, 32(9), 9 1526

(23) Lukyanov, D. B.; Shtral, V. I.; Khadzhiev, S. N. A Kinetic model for the hexane

cracking reaction over H-ZSM-5. Journal of Catalysis. 1994, 146, 87

 144

(24) Norval, G. W.; Phillips, M. J.; Virk, K. S.; Simons, R. V. Olefin conversion over

zeolite H-ZSM-5. Canadian Journal of Chemical Engineering. 1987, 67, 521

(25) Norval, G. W.; Phillips, M. J.; Missen, R. W.; Smith, W. R. Identification and

application of partial chemical equilibria in reactor modeling. AIChE Journal. 1992, 38,

1288

(26) Norval, G. W.; Phillips, M. J.; Missen, R. W.; Smith, W. R. Calculated Equilibria

for the alkene and alcohol aromatization processes. Applied Catalysis. 1989, 54, 37

(27) Kunii, D.; Levenspiel, O. Fluidization Engineering. Butterworth-Heinemann series

in chemical engineering, Newton, MA (1991).

(28) Werther, J.; Hartge, E.-U. Modeling of industrial fluidized-bed reactors. Ind. Eng.

Chem. Res. 2004, 43, 5593

(29) Abba, I. A.; Grace, J. R.; Bi, H. T.; Thompson, M. L. Spanning the flow regimes:

Generic Fluidized-bed reactor model. AIChE Journal. 2003, 49, 1838

(30) Sit, S. P.; Grace, J. R. Effect of bubble interaction on interphase mass transfer in

gas fluidized beds. Chem. Eng. Sci. 1981, 36, 327

(31) Mori, S.; Wen, Y. Estimation of bubble diameter in gaseous fluidized beds. AIChE

Journal 1975, 21(1), 109

(32) Wen, C.Y.; Yu Y. H. Mechanics of fluidization. Chemical Engineering Progress

Symposium Series. 1966, 62, 100

(33) Abrahamsen, A. R.; Geldart, D. Behaviour of gas-fluidized beds of fine powders

Part 1. Homogeneous expansion. Powder Technology. 1980, 26, 35

 145

(34) Alwahabi, S. M.; Froment, G. F. Conceptual reactor design for the methanol-to-

olefins process on SAPO-34. Ind. Eng. Chem. Res. 2004, 43, 5112

(35) Iordache, O. M.; Maria, G. C ; Pop, G. L. Lumping analysis for the methanol

conversion to olefins kinetic model. Ind. Eng. Chem. Res. 1988, 27, 2218

(36) Soundararajan, S.; Dalai, A. K.; Berruti, F. Modeling of methanol to olefins (MTO)

process in a circulation fluidized bed reactor. Fuel. 2001, 80, 1187

(37) BJ0rgen, M.; Svelle, S.; Joensen, F.; Nerlov, J.; Kolboe, S.; Bonino, F.; Palumbo,

L.; Bordiga, S.; Olsbye, U. Conversion of methanol to hydrocarbons over zeolite H-

ZSM-5; On the origin of the olefinic species. Journal of Catalysis. 2007, 249 195

(38) Svelle,S.; Joensen, F.; Nerlov, J.: 01sbye,U.; Lillerud, K-P.; Kolboe, S.; Bjergen,

M. Conversion of Methanol into Hydrocarbons over Zeolite H-ZSM-5: Ethene

Formation Is Mechanistically Separated from the Formation of Higher Alkenes. J. Am.

Chem. Soc. 2006, 128, 14770

 146

Chapter 8
General discussion
As described in Chapter 2 considerable effort have been devoted to identifying reaction

pathways and developing kinetic models capable of describing the complex reaction but

so far there remain significant shortcomings. The two main zeolite types used in the

MTO reaction ZSM-5 and SAPO follow distinct reaction pathways, deactivation time

and product. On could argue that a SAPO based catalyst would be more suitable for the

work describe herein since it is most suitable for fluidized bed operation, however, there

are several reasons to examine the ZSM-5 system. The coking rate for ZSM-5 is much

slower, which makes it easier to study the steady state operation since the catalyst does

not have to be continuously regenerated or changed. The broader product distribution of

the ZSM-5 zeolite was also considered to be of interest. Furthermore, since both fixed

bed and fluidized bed studies were completed in this study the ZSM-5 was chosen to be

the best choice of zeolite.

The spray dried catalyst used in the study is an experimental catalyst that has yet to be

optimized for the MTO reaction. The introduction of kaolin, levasil and especially

catapal B resulted in an increased methane production compared to what was observed

for the pure zeolite. The methane production rate was reduced by doping it with

phosphorous but the level of methane is still above what is found for the pure zeolite.

The shape of the catalyst differs from a normal fluid bed catalyst by having a relatively
 147

large fraction which has blow-holes, as can be seen in Figure A-2. It has, however, been

considered as a spherical catalyst during calculations.

In the fixed bed experiments we concluded that methane was predominately produced

from methanol and/or DME. This was later confirmed during the fluid bed experiments

where it was shown during experiments with 1-hexene as feed that methane only form in

a limited amount from the cracking/oligorimazation reactions. Methane formation is, to

a large degree, formed from the support material in the spray dried catalyst and not on

the zeolite and the amount of methane in a commercial catalyst would be expected to be

lower. The fact that methane is more related to the support material than the zoelite also

supports the idea that methane formation is not directly coupled to the methanol to olefin

reaction.

The experiment with 1-hexene showed that ethylene was produced at relatively high

concentrations, which would not have been expected from the fixed bed experiments and

the latest literature (Bjorgen et al. 2007) on the subject, who concluded that ethylene is

predominately formed from xylene and trimethylbenzene and not from secondary

cracking reactions. Some aromatics were present in the products from 1-hexene feed but

since methanol is not present to support the methylazation of the benzene ring,

formation of ethylene from xylene and trimethylbenzene as the primary source is

unlikely in this case.

 148

Feeding methanol and 1-hexene resulted in almost the same product distribution and it

was assumed it was because the product distribution reached chemical equilibrium, has

also been proposed by others (Norval et al, 1989; Zhou et al, 2008). However

equilibrium calculations showed that the experimental results had more light olefins

compared to the equilibrium. It could be argued, that equilibrium is established inside

the catalyst and the small molecules would have an advantage getting out of the small

pores of the zeolite, hence shifting the product distribution to the lighter fractions.

Three different fluidized bed setups were used for the MTO reaction a 3" fluidized bed

in stainless steel equipped with either a sintered metal distributor or a perforated plate

distributor and a 4.1 cm inner diameter fluidized bed in stainless steel with a sintered

metal distributor in addition to the quartz reactor. It was found that methanol

decomposition on the sintered metal distributor gradually increased the inlet pressure to

the point where the reactor had to be shut down. Changing the distributor for the 3"

fluidized bed reactor resulted in operation problems where perturbations in the inlet

pressure could cause backflow of catalyst into the windbox and feeding line. This was

especially problematic during feed changes between methanol feed and inert gas during

startup and shutdown. For a description of the two fluidized bed reactors and the

problems occurred refer to Appendix B.

8.1 Elevation of the catalyst with glass beads

Due to a gap between the distributor and the heating zone the catalyst had to be elevated

when a small amount of catalyst was used, otherwise the reaction temperature could not
 149

be reached. Elevating the catalyst was facilitated by adding sand/glass beads with a

larger particle diameter and higher density on top of the distributor and then placing the

catalyst on top. This configuration was tested in a 3" transparent cold flow fluid bed.

Sand with an average diameter of 250um and a density of 2650 kg/m3 was placed in the

bottom with catalyst on top. The catalyst was fluidized on top of the sand from the

minimum fluidization and up to the minimum fluidization of the sand. At this point the

sand began to fluidize which caused the two kinds of particles to mix. Reducing the gas

velocity below the minimum fluidization of the sand particles caused the sand and

catalyst to separate with catalyst moving to the top of the bed. The majority of the

separation was done after approximately 5 min and reached a steady level after 30 min.

The separation was never complete since a bit of catalyst could be observed close to the

bottom of the reactor and at the wall although it was not much. This can be attributed to

the fact that there was an overlap in the particle size distribution even though the large

density difference between the particles should help the separation. Only the principle

was tested to verify that it could be used. To avoid the problem with overlapping in the

particle size glass beads with an average size of 500um was used for the experiments in

the fluid bed. The particle size distribution of the catalyst, sand and glass beads is

depicted in Figure 8-1. The overlap between the sand and catalyst is evident and

changing from sand to glass beads reduced the overlap significantly.

 150

30

20 H

10 A

MTO Catalyst
— -Sand
- - • Glass beads

10 100 1000 10000

Diameter (|jrn)

Figure 8-1: Particle size distribution of the MTO catalyst, sand and glass beads used - measured was
preformed n a Horiba LA-950

8.2 Kinetic and Modelling of the MTO reaction

As described in Chapter 2 much work has been done on the modelling of the MTO

reaction over both SAPO and ZSM-5 catalysts. Only a limited number of models include

the lower olefins as separate species, while still being relatively simple. All the models

based on SAPO only include hydrocarbons up to C5 and the majority of models based on

ZSM-5 lumps the olefins together. The model of Schoenfelder et al. (1994) include both

separate light olefins and higher carbons and it was therefore used to model our

experimental data. The model is also based on a ZSM-5 catalyst but with a higher Si/Al

ratio than the zeolite used in this work. The model was implemented in the two phase
 151

model described in Chapter 7 and since the model parameters are based on another

catalyst the parameters was re-estimated to fit the fluid bed data. Figure 8-2 depicted the

model prediction vs. the experimental data. Agreement between the experimental data

and model is poor which may be due to the experimental data used in the development

of the model had very low amounts of hydrocarbons higher than C4, which is not the

case with the catalyst in this study.

c
o
«
o
"O

0 20 40 60

Experimental (Carbon %)
Figure 8-2: Calculated values with the use of the model of Schoenfelder et al. (1994) versus the obtained
experimental data from fluidized bed.

The model of Park and Froment (2004) include the cracking and oligorimazation

reactions, which allows 1-hexene to form other olefins but not ethylene and propylene.

A hydrocarbon pool mechanism with a high molecular weight hydrocarbon as the

carbon pool component was found to model the MTO reaction well. A similar scheme
 152

has recently been used by Zhou et al (2008) to characterize the MTO reaction over

SAPO-34. Some of the main differences between the two models are the fact that over

ZSM-5 hydrocarbons larger than C5 have to be included both as olefins and aromatics.

Further, it was found that methane is formed mainly from the methanol and/or DME and

not from the hydrocarbon pool. The main purpose of the model is to be able to

characterize the light olefin fractions, to obtain this goal while still maintaining a limited

amount of reactions, lumped fractions for the paraffins and C6+ have been used. The

model is also in line with work in olefin interconversion where Tsunoda and Sekiguchi

(2008) recently have published a reaction scheme for the OMEGA process with butane

as feed stock in which butane through reversible reactions react to all C2 to C8 olefins.

In Chapter 4, both a two phase model as well as a n-CSTR in series model were used to

characterize the MeOH to DME reaction in a fluidized bed. The n-CSTR was superior.

For the MTO reaction in a fluidized bed, the two phase model was chosen because the

bed height was generally higher than in the experiments with methanol dehydration and

because the gas velocity spanned a larger range. Further, the two phase model did not

perform significantly worse than the n-CSTR model for the methanol dehydration.

Using the n-CSTR model for the MTO reaction instead of the two phase model would

not change the results significantly.

Optimum production of propylene was found to be at residence times slightly higher

than what is needed to obtain full conversion of oxygenates. This is evident from the
 153

investigations in fixed beds in Chapter 5. The proposed kinetic model follows the same

trend with a decrease in the hexene concentration and increase in propylene at residence

times over that of full conversion. Optimum propylene yield from the model is 38.5% on

carbon basis at 550°C. The model slightly under predicts the maximum propylene

production from the fluid bed experiments which have a maximum value of 41%.
 154

Chapter 9
Conclusion and recommendations

9.1 General conclusions

The main objective of the work has been to develop a kinetic model for the methanol to

light olefins reaction which includes each of the light fractions of ethylene, propylene

and C4. A fluidized bed was identified to be the most suitable reactor type for the

investigation. To be able to develop the kinetic model, the study was divided into three

sections 1) evaluate the methanol dehydration kinetics in a fluidized bed. 2) investigate

the reaction mechanism and deactivation of the MTO reaction in a fixed bed and 3)

kinetic study of the MTO reaction in a fluidized bed reactor.

The dehydration of methanol to DME was studied in a 4.6 cm fluidized bed reactor at

low gas velocities. A kinetic expression for the methanol to DME reaction has been

proposed and the kinetic parameters were estimated with both a two phase fluid bed

model and an n-CSTR's in series reactor model. The kinetic model coupled with the n-

CSTR model was shown to fit the experiments better compared to the two phase fluid

bed model. The new model was superior to the literature model of Barcic and Levee

(1992). The fluidized bed was shown to be a good reactor type for the kinetic modelling

of the methanol to DME reaction due to the isothermal conditions and the conversions

that could be obtained at relatively low gas velocities. Low gas velocities are needed to

operate at conditions where the n-CSTR in series model can be used which puts some

limitations on the conversion range obtainable at a given temperature. Varying the

 155

catalyst inventory could increase the conversion range. Using the fluidized bed to

investigate the kinetics was shown to be feasible and reactions where hotspots or

temperature gradients are of concern, like the MTO reaction, could benefit from this

reactor technology.

Fixed bed experiments of the MTO reaction suggested that methane was formed directly

from methanol and/or dimethyl ether. At partial conversion largely non-aromatic C5+

intermediates dominated the product distribution. At residence time a bit longer that

what would result in 100% conversion of the oxygenates propylene yield is highest,

which is attributed to cleavage of the C5+ fraction. The formation of ethylene closely

parallels the aromatics, suggesting that it is by splitting off from xylene and

trimethylbenzene intermediates, in accordance with the labeling studies made by Svelle

et al. (2006), and not a product of secondary cracking reaction.

The product distribution and methanol capacity of the catalyst are highly dependent on

the feed composition with higher olefin production at lower partial pressure of methanol

feed. At the same time, methanol capacity is reduced.

Catalyst deactivation is highly dependent on the residence time doubling the feed rate

(from a WHSV of 0.22 h"1 to 0.43 h"1) decreased the deactivation time by a factor often.

The methanol capacity is therefore also dependent on the feed rate.

Decreasing the partial pressure and increasing the space velocity of methanol feed both

decrease the methanol capacity of the catalyst by increasing the deactivation. This is in

contrast to the findings of Chen at al. 2000 for the SAPO-34 catalyst where they
 156

concluded that coke deposition was based on the amount of methanol converted and

temperature but not on WHSV.

Catalyst coking was examined based on photographs of catalyst taken at intervals during

an experiment. Three distinct zones were identified: At the entrance the coking rate is

low which is attributed to the methanol dehydration reaction, the main coking zone due

to oxygenate to olefin and intermediates reaction follows shortly after the entrance zone.

Following the main reaction zone, coking rates are low as only secondary cracking

reactions takes place and full oxygenate conversion have been reached.

The experimental study of the reaction paths of the MTO reaction conducted in a

fluidized bed showed that methanol and 1-hexene feed yielded approximately the same

product distribution. This result confirms that cracking and oligorimerisatrion reactions

are an important part of the reaction network. Methane formation from the 1-hexene

reaction was almost non-existent confirming the findings from the fixed bed experiments

that methane is predominately formed from oxygenates. Ethylene production from the 1-

hexene experiment was not significantly lower than from methanol and cracking of

higher olefins to ethylene is therefore a part of the proposed reaction kinetics - this is in

contrast to the findings of the fixed bed where it was concluded that ethylene

predominantly is formed from splitting of polymethylated benzenes. When oxygenates

are converted to hydrocarbons, they follow the same reaction mechanism as with 1-

hexene feed and the olefin composition approaches an equilibrium composition.

Equilibrium calculations, however, show that much less light olefins and more higher
 157

olefins should be produced. Restrictions from the narrow pores in the ZSM-5 which

could allow easier passage for the light olefins might shift the product distribution

towards lighter products, if one considers equilibrium inside the zeolite.

A reaction kinetic model is proposed for the MTO reaction over ZSM-5 which includes

15 reactions. All olefins are considered to be formed through reversible reactions with a

large hydrocarbon species. The model characterizes the experimental data fairly well for

all the light olefins, which was the main objective but requires improvements with

respect to the paraffins and C6+ fractions.

9.2 Recommendation

9.2.1 Kinetics in fluidized bed

Kinetic measurements in a small scale fluidized bed have shown good results in this

work. Expanding the work to include an in-depth analysis of the effect of the

hydrodynamic at low gas velocities with exothermic reactions where the reaction

kinetics are well known would be desirable. Determination of basis operation interval

where the isothermal conditions are maintained and changes in the bed hydrodynamic

are minimal would be highly attractive for future studies of exothermic reaction.

9.2.2 MTO reaction kinetic

While the proposed model are able to predict the product distribution of the light olefins

improvements of the paraffin and C6+ fractions are still needed. From the fixed bed study

it was shown that the aromatic and heavy olefins follows separate reaction pathways.
 158

Modeling of the 1-hexene experiments was unsuccessful and splitting the aromatic

fraction from the C6+ would be desired.

9.2.3 Equilibrium inside the catalyst?

The experimental data from the MTO reaction in fluidized bed suggested that the

reaction is close to equilibrium. Calculations of the equilibrium showed that the

experimental data followed the same trends of the equilibrium but the fraction of the

light olefins was much larger than the equilibrium this was especially the case for the

propylene. Considering equilibrium concentration inside the cages of the zeolite it would

be interesting to determine if transport restrictions out of the zeolite pore system could

account for the observed difference between the experimentally obtained product

distribution and the equilibrium.

 159

References

Abba, I. A., Grace, J. R., Bi, H. T., & Thompson, M. L. (2003). Spanning the flow

regimes: Generic Fluidized-bed reactor model. AIChE Journal, 49, 1838-1848.

Abrahamsen, A. R., & Geldart, D. (1980). Behaviour of gas-fluidized beds of fine

powders Part 1. Homogeneous expansion. Powder Technology, 26, 35-46.

Aguayo, A. T., Gayubo, A. G., Vivanco, R., Alonso, A., & Bilbao, J. (2005). Initiation

step and reactive intermediates in the transformation of methanol into olefins

over SAPO-18 catalyst. Industrial & Engineering Chemistry Research, 44 (19),

7279-7286.

Alwahabi, S. M., & Froment, G. F. (2004). Conceptual reactor design for the methanol-

to-olefins process on SAPO-34. Industrial & Engineering Chemistry Research,

43,5112-5122

Andersen, J. (2003). Mto: meeting the needs for ethylene and propylene production,

Paris. ERTC Petrochemical Conference.

Anthony, R. G. (1981). A kinetic model for methanol conversion to hydrocarbons.

Chemical Engineering Science, 36, 789.

Arstad, B., & Kolboe, S., (2001). The Reactivity of Molecules Trapped within the

SAPO-34 Cavities in the Methanol-to-Hydrocarbons Reaction. Journal of the

American Chemical Society, 123, 8137-8138.

Bercic, G., & Levee, J. (1992). Intrinsic and Global Reaction Rate of Methanol
 160

Dehydration over Y-AI2O3 Pellets. Industrial & Engineering Chemistry

Research, 31, 1035-1040.

Beech, J. H., Nicoletti, M. P., Lattner, J. R., Alexion, D. G., & Paik, P. K. (2004).

Method and system for regeneration catalyst from a plutilaty of hydrocarbon

conversion apparatuses. Exxonmobil chemical company. US 20040064007A1

Benito, P. L., Gayubo, A. G., Aguayo, A. T., Castilla, M., & Bilbao, J. (1996).

Concentration-dependent kinetic model for catalyst deactivation in the MTG

process. Industrial & Engineering Chemistry Research, 35, 81-89.

Berger, R. J, Ramirez, J. P., Kapteijn, F., & Moulijn, J. A. (2002). Catalyst performance

testing: bed dilution revisited. Chemical Engineering Science, 57, 4912-4932.

Bjorgen, M., Olsbye, U., Svelle, S., & Kolboe, S. (2004). Conversion of methanol to

hydrocarbons: the reactions of the hepta-methylbenzenium cation over zeolite

H-beta. Catalysis Letters, 93, 37-40.

Bjorgen, M., Svelle, S., Joensen, F., Nerlov, J., Kolboe, S., Bonino, F., Palumbo, L.,

Bordiga, S., & Olsbye, U. (2007) Conversion of methanol to hydrocarbons

over zeolite H-ZSM-5; On the origin of the olefinic species. Journal of

Catalysis, 249, 195-207

Bos, R. A. N., Tromp, P. J. J., & Akse H. N. (1995). Conversion of methanol to lower

olefins. Kinetic modelling, reactor simulation, and selection. Industrial &

Engineering Chemistry Research, 34, 3808-3816.

Buchanan, J. S., Santiesteban, J. G., & Haag, W. O. (1996). Mechanistic considerations

in acid-catalyzed cracking of olefins. Journal of Catalysis, 158, 279-287

 161

Campbell, S. M., Bibby, D. M., Coddington, J. M., & Howe, R. F. (1996).

Dealumination of HZSM-5 Zeolites. Journal of Catalysis, 161, 350-358.

Chang, C. D., & Silvestri, A. J. (1977). The conversion of methanol and other O-

compounds to hydrocarbons over zeolite catalysts.- Journal of Catalysis, 47,

249-259.

Chang, C. D. (1980). A kinetic model for methanol conversion to hydrocarbons.

Chemical Engineering Science, 35, 619-622.

Chang, C. D. (1983). Hydrocarbons form methanol. Catalysis Review Science and

Engineering, 25, 1-118.

Chang, C. D. (1984). Methanol Conversion to Light Olefins. Catalysis Review Science

and Engineering, 26 (3&4), 323-345.

Chang, C. D., Chu, C. T-W., & Socha, R. F. (1984). Methanol conversion to Olefins

over ZSM-5. I. Effect of temperature and zeolite Si02/Al203. Journal of

Catalysis, 86, 289-296.

Chang, C. D., Hellring, S. D., & Pearson, J. A. (1989). On the existence and role of free

radicals in methanol conversion to hydrocarbons over HZSM-5. Journal of

Catalysis, 115, 282-285.

Chen, D., Rebo, H. P., Gronvold, A., Moljord, K., & Holmen, A., (2000).

Methanol conversion to light olefins over SAPO-34: Kinetic modeling of coke

formation. Microporous and Mesoporous Materials, 35-36, 121-135.

Chen, J. Q., Bozzano, A., Glover, B., Fuglerud, T., & Kvisle, S. (2005). Recent
 162

advancements in ethylene and propylene production using the UOP/Hydro

MTO process. Catalysis Today, 106, 103-107.

Chen, N.Y., & Reagan, W.J., (1979). Evidence of autocatalysis in methanol to

Hydrocarbon reactions over zeolite catalysts. Journal of catalysis, 59, 123-129.

Christensen, D., Vervloet, D., Nijenhuis, J., van Wachem, B.G.M., van Ommen J.R., &

Coppens, M.-O. (2008). Insights in distributed secondary gas injection in a

bubbling fluidized bed via discrete particle simulations. Powder Technology,

183,454-466.

Chisholm, P. N., Coute, N. P., Lattner, J. R., Xu, T., Smith, J. S., & Kuechler, K. H.

(2003). Convertingoxygenates to olefins over a catalyst comprising acidic

molecular sieve of controlled carbon atom to acid site ratio. Exxonmobil

chemical company, US 20030125598A1

Chu, C. T-W., & Chang, C. D. (1984). Methanol conversion to Olefins over ZSM-5.

[Link] Distribution. Journal of Catalysis, 86, 297-300.

Clarke, J. K. A., Darcy, R., Hegarty, B. F., O'Donoghue, E., Amir-Ebrahimi, V., &

Rooney, J. J. (1986). Free radicals in dimethyl ether on H-ZSM-5 zeolite. A

novel dimension of heterogeneous catalysis. Journal of the Chemical Society,

Chemical Communications, 425-426.

Coute, N. P., Kuechler, K. H., Chisholm, P. N., Vaughn, S. N., Lattner, J. R., &

Kuechler, W. L. (2004). Process for making olefins. Exxonmobil chemical

company, US 6,673,978B2
 163

Dahl, I. M., & Kolboe, S. (1993). On the reaction mechanism for propene formation in

the MTo reaction over SAPO-34. Catalysis Letters, 20, 329-336.

Dahl, I. M., & Kolboe, S. (1994). On the reaction mechanism for hydrocarbon formation

from methanol over SAPO-34. Journal ofCaralysis, 149, 458-464.

Dehertog, W. J. H., & Froment, G. F., (1991). Production of light alkenes from

methanol on ZSM-5 catalysis. Applied Catalysis, 71, 153-165.

Dessau, R.M. (1986). On the H-ZSM-5 catalyzed formation of ethylene from

methanol or higher olefins. Journal of Catalysis, 99, 111-116.

Dewaele, O., Geers, V. L., Froment, G. F., & Marin, G. B. (1999). The conversion of

methanol to olefins: A transient kinetic study. Chemical Engineering Science,

54, 4385-4395.

Fogler, H. S. (2005). Elements of chemical reaction engineering. Prentice Hall, 4

edition

Froment, G. F., Dehertog, W. J. H., & Marchi, A. J. (1992). Zeolite catalysis in the

conversion of methanol into olefins. Catalysis 9, 1-64.

Gayubo, A. G., Benito, P. L., Aguayo, A. T., Aguirre, I., & Bilbao, J. (1996). Analysis

of kinetic models of the methanol-to-gasoline (MTG) process in an integral

reactor. Chemical Engineering Journal, 63,45-51.

Gayubo, A. G., Aguayo, A. T., Sanchez del Campo. A. E., Tarrio, A. M., & Bilbao, J.

(2000). Kinetic modelling of methanol transformation into olefins on a SAPO-

34 catalyst. Industrial & Engineering Chemistry Research. 39, 292-300

 164

Gayubo, A. G., Ortega, J. M., Aguayo, A. T., Arandes, J. M., & Bilbao, J. (2000). MTG

fluidized bed reactor-regenerator unit with catalyst circulation: process

simulation and operation of an experimental setup. Chemical Engineering

Science, 55, 3223-3235.

Gayubo, A. G., Aguayo, A. T., Castilla, M., Olazar, M., & Bilbao, J. (2001). Catalyst

reactivation kinetics for methanol transformation into hydrocarbons.

Expressions for designing reaction-regeneration cycles in isothermal and

adiabatic fixed bed reactor. Chemical Engineering Science, 56, 5059-5071.

Gayubo, A. G., Anuayo, A. T., Olazar, M., Vivanco, R., & Bilbao, J. (2003). Kinetics of

the irreversible deactivation of the HZSM-5 catalyst in the MTO process.

Chemical Engineering Science, 58, 5239-5249.

Gayubo, A. G., Aguayo, A. T., Atutxa, A., Prieto, R., & Bilbao, J. (2004a). Role of

Reaction-Medium Water on the acidity deterioration of a HZSM-5 Zeolite.

Industrial & Engineering Chemistry Research, 43, 5042-5048.

Gayubo, A. G., Aguayo, A. T., Castilla, M., Moran, A. L., & Bilbao, J. (2004b). Role of

water in the kinetic modelling of methanol transformation into hydrocarbons

on HZSM-5 zeolite. Chemical Engineering Communications, 191, 944-967.

Gayubo, A. G., Aguayo A. T., Alonso, A., & Bilbao J. (2007). Kinetic Modeling of the

Methanol-to-Olefms Process on a Silicoaluminophosphate (SAPO-18) Catalyst

by Considering Deactivation and the Formation of Individual Olefins..

Industrial & Engineering Chemistry Research. 46, 1981-1989.

 165

Gray, C , & Webster, G. (2001). A Study of Dimethyl Ether (DME) as an Alternative

Fuel for Diesel Engine Applications. Transport Canada Publication No.: TP

13788E, May 2001

Haw, J. F., Song, W., Marcus, D. M., & Nicholas, J. B. (2003). The mechanism of

methanol to hydrocarbon catalysis. Accounts of Chemical Research, 36, 317-

326.

Hetsroni, G., (1982). "Handbook of Multiphase Systems" Hemisphere-McGraw Hill.

Horio, M., & Wen, C. Y. (1977). An assessment of fluidized-bed modelling. AIChE

symposium series, 73, 9-21.

Houdek, J., & Andersen, J. (2005). On-purpose propylene - technology development.

Kuala Lumpur, ARTC 8th annual meeting.

International Zeolite Association, (2006). [Link]

Iordache, O. M., Maria, G. C , & Pop, G. L. (1988). Lumping analysis for the methanol

conversion to olefins kinetic model. Industrial & Engineering Chemistry

Research, 27,2218-2224

Kaarsholm, M., Joensen, F., Nerlov, J., Cenni, R., Chaouki, J., & Patience, G. S. (2007).

Phosphorous modified ZSM-5: Deactivation and product distribution for MTO

Chemical Engineering Science, 62, 5527-5532.

Keil, F. J. (1999). Methanol-to-hydrocarbons: process technology. Microporous

And Mesoporous Materials, 29, 49-66.

Kokotailo, G. T., Lawton, S. L., Olson, D. H., & Meier, W. M. (1978). Structue of

Synthetic zeolite ZSM-5. Nature, 272, 437-438.

 166

Kunii, D. & Levenspiel, O. (1991). Fluidization Engineering. Butterworth-Heinemann

series in chemical engineering, Newton, MA

Kunii, D. and Levenspiel, O. (1968). Bubbling bed model. Model for the flow of gas

through a fluidized bed. I&EC Fundamentals, 7(3), 446-452.

Lattner, J. R., Smith, J. S., & Davuluri, R. P. (2005). Circulating fluid bed reactor

system. Exxonmobil chemical company, US 20050087477A1

Lorences, M. J., Laviolette, J.-P., Patience, G. S., Alonso M., & Diez, F. V. (2006).

Fluid bed gas RTD: Effect of fines and internals. Power Technology, 168, 1-9.

Lorences, M. J., Patience, G. S., Diez, F. V., & Coca, J. (2003). Butane oxidation to

maleic anhydride: kinetic modeling and byproducts. Industrial & Engineering

Chemistry Research, 42, 6730-6742.

Lu, W.-Z., Teng, L.-H., & Xiao, W.D. (2004). Simulation and experiment study of

dimethyl ether synthesis from syngas in a fluidized-bed reactor. Chemical

Engineering Science, 59, 5455-5464.

Lukyanov, D. B., Shtral, V. I., & Khadzhiev, S. N. (1994). A Kinetic model for the

hexane cracking reaction over H-ZSM-5. Journal of Catalysis, 146, 87-92

Lurgi (2005). Lurgi successful in the middle east and in south east asia. Press release

March 7, [Link]

Lurgi (2008). New Project Development Agreement for integrated polypropylene

complex. Press release September 9, [Link]

 167

Maria, G., & Muntean, O. (1987). Model reduction and kinetic parameters identification

for the methanol conversion to olefins. Chemical Engineering Science, 42,

1451-1460.

May, W. G. (1959). Fluidized-bed reactor studies. Chemical Engineering Progress,

55(12), 49-56.

Mier, D., Aguayo, A. T., Atutxa, A., Gayubo, A. G., & Bilbao, J. (2007). Kinetic study

of the simultaneous cracking of paraffins and methanol on HZSM-5 zeolite

catalyst. International Journal of Chemical Reactor Engineering, 5, A60

Mihail, R., Straja, S., Maria, G., Musca, G., & Pop, G. (1983a). Kinetic Model for

Methanol Conversion to Olefins. Industrial and engineering chemistry process

design and development, 22, 532-538.

Mihail, R., Straja, S., Maria, G., Musca, G., & Pop, G. (1983b). Kinetic Model for

Methanol Conversion to Hydrocarbons. Chemical Engineering Science, 38,

1581-1591.

Mollavali, M., Yaripour, F., Atashi, H., & Sahebdelfar, S. (2008). Intrinsic Kinetic

Study of Dimethyl Ether Synthesis from Methanol on Y-AI2O3 Catalysts. .

Industrial & Engineering Chemistry Research, 47, 2365-3273.

Moller, K. P., Bohringer, W., Schnitzler, A. E., van Steen, E., & O'Connor, C. T.

(1999). The use of a jet loop reactor to study the effect of crystal size and the

co-feeding of olefins and water on the conversion of methanol over HZSM-5.

Microporous and Mesoporous Materials, 29, 127-144

 168

Mori, S. & Wen, Y. (1975). Estimation of bubble diameter in gaseous fluidized beds.

AIChE Journal, 21(1), 109-115.

Novella, E. C , Alonso, M. A., Zamorano, M. A. U., & Granados D. P. S. (1988).

Conversion del metanol en hidrocarburos Ci-Cio sobre zeolita tipo ZSM-5 :

Modelo cinetico simplificado. Anales De Quimica, 85, 292-297.

Norval, G. W., Phillips, M. J., Virk, K. S., & Simons, R. V. (1987). Olefin conversion

over zeolite H-ZSM-5. Canadian Journal of Chemical Engineering, 67, 521-

523

Norval, G. W., Phillips, M. J., Missen, R. W., & Smith, W. R.(1989) Calculated

Equilibria for the alkene and alcohol aromatization processes. Applied

Catalysis, 54, 37-52

Norval, G. W., Phillips, M. J., Missen, R. W., & Smith, W. R. (1992). Identification and

application of partial chemical equilibria in reactor modeling. AIChE Journal,

38, 1288-1293

Ondrey, G. (2005). Methanol-to Propylene process slated for its commercial debut.

Chementator, [Link]. April, 5

Ortega, J. M., Gayubo, A. G., Anuayo, A. T., Olazar, M., & Bilbao, J. (1998). MTG

process in a fluidized bed with catalyst circulation: operation and simulation of

an experimental unit. Industrial & Engineering Chemistry Research, 37, 4222-

4230.

Park, T-Y., & Froment, G. F. (2001a). Kinetic Modeling of the Methanol to Olefins
 169

Process. 1. Model Formulation. Industrial & Engineering Chemistry Research,

40, 4172-4186.

Park, T-Y., & Froment, G. F. (2001b). Kinetic Modeling of the Methanol to Olefins

Process. 2. Experimental Results, Model Discrimination, and Parameter

Estimation. Industrial & Engineering Chemistry Research, 40, 4187-4196.

Park, T-Y., & Froment, G. F. (2004). Analysis of fundamental reaction rates in the

methanol-to-olefins process on ZSM-5 as a basis for reactor design and

operation. Industrial & Engineering Chemistry Research. 43, 682 -689

Radmanesh, R., Chaouki, J., & Guy, C. (2006). Biomass Gasification in a Bubbling

Fluidized Bed Reactor: Experiments and Modeling. AIChE Journal, 52(12),

4258-4272.

Sapre, A. V. (1997). Catalyst deactivation kinetics from variable space-velocity

Experiments. Chemical Engineering Science, 52, 4615-4623.

Schipper P. H., & Krambeck, F. J. (1986). A reactor design simulation with reversible

and irreversible catalyst deactivation Chemical Engineering Science, 41, 1013-

1019.

Schoenfelder, H., Hinderer, J., Werther, J., & Keil, F. J. (1994). Methanol to olefins -

predictions of the performance of a circulating fluidized-bed reactor on the

basis of kinetic experiments in a fixed-bed reactor. Chemical Engineering

Science, 49, 5377-5390.

 170

Sedran, U., Mahay, A., & De Lasa, H. L. (1990). Modelling methanol conversion to

hydrocarbons: alternative kinetic models. Chemical Engineering Journal, 45,

33-42.

Semelsberger, T. A., Borup, R. L., & Greene, H. L. (2006). Dimethyl ether (DME) as an

alternative fuel. Journal of Power Sources, 156, 497-511.

Shaikh, A. A. & Batran, H. (2007). On bubble-side transport limitations in catalytic

fluid-bed reactors. Chemical Engineering Research and Design, 85, 1215-

1218.

Sit, S. P., & Grace, J. R. (1981). Effect of bubble interaction on interphase mass transfer

in gas fluidized beds. Chemical Engineering Science, 36, 327-335.

Soundararajan, S., Dalai, A. K., & Berruti, F. (2001). Modeling of methanol to olefins

(MTO) process in a circulation fluidized bed reactor. Fuel, 80, 1187-1197

Stocker, M. (1999). Methanol-to-hydrocarbons: catalytic materials and their

Behaviour. Microporous and Mesoporous Materials, 29, 3-48.

Stocker, M. (2003). Methanol to olefins - present scenario and future perspectives.

SINTEF Materials and Chemistry Oslo.

Svelle, S., BJ0rgen, B., Kolboe, S., Kuck, D., Letzel, M., Olsbye, U., Sekiguchi, O., &

Uggerud, E. (2006). Intermediates in the methanol-to-hydrocarbons (MTH)

reaction: a gasphase study of the unimolecular reactivity of multiply

methylated benzenium cations. Catalysis Letters, 109,25-35.

Svelle, S., Joensen, F., Nerlov, J., Olsbye, U., Lillerud, K.P., Kolboe, S., &
 171

BJ0rgen, M., (2006). Conversion of methanol in to hydrocarbons over zeolite

H-ZSM-5: Ethene formation is mechanistically separated from the formation of

higher alkenes. Journal of the American Chemical Society. 128 (46), 14770-

14771.

Tabak, S. A., Krambeck, F. J., Garwood, W. E. (1986). Conversion of propylene and

butylene over ZSM-5 catalyst. AIChE Journal, 32(9), 1526-1531

Tabak, S. A., & Yurchak, S. (1990). Conversion of methanol over ZSM-5 to fuels and

Chemicals. Catalysis Today, 6, 307-327.

Tsunoda, T., & Sekiguchi, M. (2008) The omega process for propylene production by

olefin interconversion. Catalysis Surveys from Asia, 12, 1-5

UOP (2008). UOP Technology Licensed to EuroChem to Convert Methanol to Olefins

in New Nigerian Petrochemicals Plant. Press release January 30,

[Link]

Van Deemter, J. J. (1961). Mixing and contacting in gas-fluidized beds. Chemical

Engineering Science, 13(3), 143-154.

Wen, C.Y., & Yu, Y. H. (1966). Mechanics of fluidization. Chemical Engineering

Progress Symposium Series, 62, 100-111.

Werther, J., & Hartge, E.-U. (2004). Modeling of industrial fluidized-bed reactors.

Industrial & Engineering Chemistry Research, 43, 5593-5604.

Wu, X., & Anthony, R. G. (2001). Effect of feed composition on methanol conversion

light olefins over SAPO-34. Applied Catalysis A: General, 218, 241-250.

 172

Zhou, H., Wang, Y., Wei, F., Wang, D., & Wang, W. (2008) Kinetics of the reactions of

the light alkenes over SAPO-34. Applied Catalysis A: General, 348, 135-141
 173

Appendix A

Catalyst preparation
The catalyst was prepared for the fluid bed by dissolving Catapal B in a weak solution of

nitric acid with stirring until a gel forms. After, Levasil SI00/30% was added during

stirring together with water to control viscosity. After a couple of hours, kaoline and the

zeolite (CBV 28014 from Zeolyst, Si/Al =140) is introduced to the mixture. The slurry is

stirred to make it homogenous before spray drying. The dry matter content before it is

spray dried is 27 - 29% and with a pH-value of 4.0. The inlet pressure to the spray drier

is between 6 and 8 bar, and the inlet temperature 280°C with an outlet temperature of

110°C. After spray drying the catalyst was calcined at 550°C for four hours

To test the catalyst in a fixed bed it was made into tablets, crushed and sieved to a

particle size of 500 - 700 urn before testing in a 8 mm inner diameter fixed bed at 500°C

with a WHSV of 23.5 gMeOH/gcat'h. The catalyst produced a high amount of methane. To

reduce the methane production the catalyst was doped with phosphorous by wetting the

catalyst with an aqueous solution of (NH3)2HP04. The catalyst was then dried and

calcined (at 550°C for four hours). The phosphorous loading tested was 0.5, 0.75, 1.0,

1.5 and 5 wt%. An additional test was done where (NH3)2HP04 was introduced into the

slurry mixture before it was spray dried (0.5 wt%). The catalysts were tested in the fixed

bed reactor and the methane formation decreased with phosphorous loading and a
 174

positive effect on the catalyst lifetime was also observed. The results for the methane

production are given in Figure A-l. The catalyst with 5 wt% phosphorous deactivated

very rapidly and results are therefore not shown. It was found that a phosphorous

loading of the catalyst of 1.5 wt% was optimal for the catalysts tested. The catalyst

particle size range is between 40 and 250 urn with a mean diameter of 108 um. The

particles are shown in Figure A-2, it is seen that the majority of the particles have a

mushroom-like shape and are not spherical as would have been desired for fluidized bed

operation.

20 n

15 H

10-T
— o% p
a 0.5% P (slurry)
O
— 0.5% P
- « - 0.75% P
5H
+-1% P
-*•-- 1.5% P
-•— Pure zeolite

—i
20 25

Time (h)
Figure A-l: Methane formation with time at different phosphor loadings of the fluid bed catalyst.
Temperature 500°C and WHSV 23.5 gMeoH/gcat-h
 175

Figure A-2: Picture of the spray dried particles

 176

Appendix B

Fluidized bed reactors used

A significant amount of work in several different fluidized bed reactor setups has been

conducted. This Appendix will summarize problems that occurred with the different

reactors and the reason for conducting the MTO reaction in the 4.6 cm inner diameter

quartz fluidized bed reactor.

Three different reactors have been used a 3" inner diameter fluidized bed reactor setup in

stainless steel, a 4.1 cm inner diameter fluidized bed in stainless steel and a 4.6 cm inner

diameter fluidized bed in quartz. In the following a short description of the two setups -

that had to be abandoned mainly due to methanol decomposition problems in the feeding

section - will be given.

3" inner diameter fluidized bed

A fluid bed reactor system with a 3" inner diameter fluid bed in stainless steel was

constructed including a feed section for both liquid and gas feed, a quench section to

collect higher hydrocarbons and water and a gas analysis line to a GC. A diagram of the

entire system is given in Figure B-2. A control system was setup for the reactor to

monitor and control the reaction with shutdown procedures if pressure or temperature

limits were exceeded. A drawing of the fluid bed is given in Figure B-l and the

dimensions of the different sections are listed in Table B-l. The reactor is heated in three
 177

zones - the windbox, bed section and the freeboard - with each zone heated by a

separate heater.

Figure B-l: Drawing of the 3" inner diameter fluid bed Legend: 1) feed inlet, 2) windbox, 3) distribution
grid 4) solid outlet, 5) fluid bed region, 6) sampling/measuring ports, 7) freeboard, 8) electrical heating, 9)
solid inlet, 10) disengagement zone

Table B-l: Dimensions of the 7.79 cm inner diameter reactor.

# Type Nominal Schedule Material Wall Inner Outside
diameter thickness diameter diameter
(cm) (cm) (cm) (cm)
1A Pipe 15,24 40 SS316L 0.71 15.41 16.83
2A Pipe 7,62 40 SS310 0.55 7.79 8.89
3A Pipe 7,62 40 SS310 0.55 7.79 8.89
 178

The fluid bed was initially equipped with a sintered metal distributor and MTO

experiments showed promising results with a stable product distribution but the inlet

pressure was increasing with time until it was shutdown either manually or by the

control system. The pressure drop was reduced during regeneration and it was found to

be caused by coking of the sintered metal distributor. The initial experiments were done

by preheating the windbox to about 300°C. To reduce coking the inlet temperature to the

fluid bed was reduced to a minimum by reducing the windbox temperature to 110-

130°C. The coking on the sintered metal distributor was not reduced significantly mainly

because it was heated by the catalyst at the reaction temperature of 500°C. Since it was

not possible to use the sintered metal distributor it was changed to a perforated plate

distributor with 3/64" holes. Changing the distributor eliminated the pressure build-up in

the feeding section. Another problem occurred, since the catalyst particles are much

smaller than the distributor holes weeping of the particles became a problem. Weeping

of particles was not a problem during normal operation since the feed gas kept the

catalyst from entering the windbox. It became a significant problem if perturbations

occurred in the feed flow and during start-up and shutdown where a valve switch

between gas and liquid feed is needed. It was found that a small change in the inlet

pressure caused the pressure above the catalyst bed to be higher than the inlet pressure as

a result catalyst was pushed into the windbox and further down into the feeding section.

The regular shutdowns each time catalyst was pushed into the windbox and the cleaning

needed of the feeding section made it impractical to conduct the MTO reaction in this

reactor and it was therefore abandoned.

Figure B-2: Schematic diagram of the 3" fluidized bed setup -J
 180

4.1 cm inner diameter stainless steel fluidized bed reactor

At the same time as the 3" fluidized bed setup was ready for experimental work a newly

acquired 4.1 cm inner diameter stainless steel fluidized bed setup became available and

experimental work in this reactor was also conducted. A quartz fluidized bed with two

external electrical heaters was also supplied with this setup. A schematic drawing of the

stainless steel fluidized bed setup is given in Figure B-3 for schematic diagrams of the

entire setup refer to Figure 4-2 where the only change is the fluidized bed setup. The

fluidized bed is immersed in a fluidized bed sand bath which heat the bed and feed

section of the reactor to the given reaction temperature. The gas distributor is made of

sintered metal. The coking problem in the feeding section of this setup was even more

pronounced than in the 3" fluidized bed with coking not only on the sintered metal

distributor but also in the windbox and the feed line to the windbox. After several hours

of MTO experiments in the reactor the feed line became blocked by coke and large

lumps of coke was flushed out of the windbox during the subsequent cleaning. Due to

the high decomposition rate of methanol in the stainless steel fluidized bed reactors it

was decided to conduct the experiments in the quartz reactor where these problems were

not encountered. A description of the quartz reactor is given in Chapter 4.2.3.

 181

Multipoint
thermocouple

_Dis engagement zone

8 cm inner diamenter

Freeboard /
-Feed

Bed.

Wind box Fluidized sand

ath

Figure B-3: 4.1cm inner diameterfluidizedbed reactor