Fischer Tropsch

INGENIERO QUMICO
Universidad de Valladolid
INFORME DE PRCTICAS
ASIGNATURA: EXPERIMENTACIN EN INGENIERA QUMICA II

CURSO: 2011/12
TRABAJO BIBLIOGRFICO
GRUPO: B
PAREJA: 7
TIEMPO EMPLEADO
ALUMNO
EN INFORME
MIGUEL NGEL CALVO CARRASCAL
10 h
HCTOR CANTALAPIEDRA SIGENZA
10 h
FECHA: 24/10/2011
INDICE
1. Definicin de reactor de membrana ....................................................................... 2
2. Tipos de reactores de membrana .......................................................................... 3
3. Usos o aplicaciones de los reactores de membrana .............................................. 4

a. Reformado de una corriente de etanol ....................................................... 4
b. Transesterificacin catalizada de aceite de canola ..................................... 7
c. Hidrlisis enzimtica de la lignocelulosa ..................................................... 9
4. Bibliografa ........................................................................................................... 12
5. Anexo .................................................................................................................. 15
USO O EMPLEO DE REACTORES DE MEMBRANA EN PROCESOS DE

PRODUCCIN DE BIOCOMBUSTIBLES
1. DEFINICIN
Son aquellos reactores que combinan la reaccin y la separacin en una sola unidad;
la membrana selectivamente remueve una (o ms) de las especies reactantes o
productos.
Estos reactores han sido comnmente usados para incrementar el rendimiento y la
selectividad de reacciones enzimticas y catalticas influyendo a travs de la
membrana sobre la concentracin de una o ms especies intermedias, removindolas
selectivamente o ayudando a mantenerlas en una concentracin baja mediante el
filtrado continuo de productos, evitando la posibilidad de que dichos compuestos
envenenen o desactiven el catalizador y para proveer una interfase controlada entre
dos o ms reactantes.
2. TIPOS DE REACTORES DE MEMBRANA
Reactor de membrana densa Pd-Ag
Es un tipo destacado de membrana inorgnica. El paladio y sus aleaciones son
permeables al hidrgeno, en su forma de biocombustible para producir energa.
Existen numerosos estudios y casos documentados del uso de este tipo de
membranas.[11-13]
stos se emplean en diversas reacciones para producir, y a la vez separar los
productos generados en esa reaccin. Especialmente tiles son las membranas Pd-Ag
densas, ya que permiten conseguir directamente un producto ultrapuro sin ser
necesario el uso de algn otro mtodo de purificacin [14-18]
Estos reactores tambin se pueden emplear en el reformado de metanol y etanol de
forma simultnea
Reactor de membrana orgnica
La membrana de stos es de carbono normalmente polimrico, tienen gran inters ya
que son estables a temperaturas elevadas y resisten bastante bien los ataques
qumicos, sobre todo medios muy cidos y medios bsicos [21-23]
Reactor de membrana inorgnica
Este tipo de reactores se utiliza mayoritariamente en reacciones en fase lquida con
disolvente inorgnico. Debido a la excelente estabilidad trmica de la membrana, estos
reactores pueden ser empleados en reacciones a altas temperaturas. Estas
membranas son muy resistentes al ataque qumico y existen una gran cantidad de
materiales que pueden ser utilizados en la fabricacin de estas membranas. Para
entornos muy hostiles son el nico tipo de membrana que se puede utilizar [21-23]
3. USOS O APLICACIONES DE LOS REACTORES DE MEMBRANA

Existen muchos usos en los que se puede utilizar reactores de membrana frente a
reactores tradicionales.
La mayora de los usos de los reactores de membrana se centra en la obtencin de
biocombustibles, stos son combustibles limpios obtenidos a partir de aceites
vegetales y grasas animales, por lo que constituyen una fuente renovable de energa
con una fuerte proyeccin hacia el futuro
Esta importancia se debe a que los combustibles tradicionales a partir de fsiles estn
en rpida desaparicin y los biocombustibles tienen la ventaja de no poseer en su
composicin con sulfuros, hidrocarburos aromticos, metales o residuos de crudos lo
que la hace una energa menos contaminante que sus homlogas.
Algunos reactores centrados en la obtencin de biocombustibles se exponen a
continuacin:
Reformado de una corriente de etanol en un reactor de membrana densa Pd-Ag
El objetivo del reformado del etanol es obtener una corriente de hidrgeno gaseoso
para su uso como biocombustible.
Existen mquinas capaces de dar energa al obtener hidrgeno, los sistemas de celda
polimrica son aparatos electroqumicos capaces de obtener energa elctrica de la
oxidacin del hidrgeno con oxgeno atmosfrico. Por eso es importante el desarrollo
de mtodos eficaces de obtencin de hidrgeno con la mayor pureza posible.
La obtencin de hidrgeno puede darse por dos vas mayoritariamente.
A partir de alcoholes (metanol, etanol) o hidrocarburos (Gas natural, gas propano,
gasolinas). Debido a que el uso de combustibles fsiles incrementa el calentamiento
global y aumentan las emisiones de CO2 a la atmsfera, el uso de biocombustibles,
etanol en este caso, contribuye a reducir el efecto invernadero.
El proceso de reformado de etanol proveniente de biomasa se lleva a cabo en un
reactor de membrana densa Pd-Ag con el objetivo de conseguir una corriente continua
de hidrgeno de pureza muy elevada y con unas conversiones de la reaccin
superiores a las que se podra obtener por un equilibrio termodinmico convencional
El estudio del proceso se llevar a cabo comparando los resultados obtenidos con el
reactor de membrana utilizado frente a un reactor convencional [1]
La reaccin que se lleva a cabo es catalizada y endotrmica, y su conversin se
incrementa con la temperatura [1-7]:
C2H5OH+H2O2CO+4H2
Pero como por las leyes ambientales vigentes se procura no verter CO al aire, se
emplea vapor de agua para convertirlo a CO2, por lo que la reaccin es:
C2H5OH+3H2O2CO2+6H2
La conversin obtenida as como la selectividad de los productos depende en gran
medida del catalizador usado, as como de la temperatura de operacin. Por ejemplo,
un exceso de agua en la alimentacin incrementa la selectividad del hidrgeno y
reduce a la formacin de coque, que puede envenenar los catalizadores cuando se
trabaja a bajas temperaturas.
El hecho de usar un tipo de reactor u otro permite obtener mismas conversiones a

distintas temperaturas, es decir, el reactor tradicional [8-10] alcanza una conversin fija
a una temperatura que el reactor de membrana es capaz de obtener a una
temperatura bastante menor. Todo esto se ampliar en el estudio comparativo entre
diferentes variables.
La optimizacin del rendimiento del proceso de reformado de etanol no haba sido
posible ya que no exista informacin acerca de expresiones cinticas tiles.
Pasado un tiempo se presentaron por primera vez expresiones cinticas para dicho
proceso en un reactor tradicional [2].
Dichas expresiones sern utilizadas para simular el comportamiento de ambos
reactores.
Los resultados obtenidos acerca de la conversin de etanol y la produccin de oxgeno
para ambos rectores sern comparados teniendo en cuenta diferentes variables de
proceso: temperatura, presin, relacin molar de la alimentacin)
Las expresiones que permiten determinar la conversin de etanol, la produccin de
hidrgeno y su recuperacin son:
(1)
(2)
(3)
Sealar por otra parte, que los resultados que se mostrarn a continuacin provienen
del estudio del proceso de reformado basado en un modelo terico, hecho que no
haba sido desarrollado anteriormente.
Resultados y discusin
En primer lugar se procede a validar el modelo terico utilizado para comparar ambos
reactores y se emplea la grfica 1. Usando los datos experimentales proporcionados
[2]. para reactores tradicionales se pudo validar dicho modelo para el rango de
temperaturas desde 673 K a 873 K y una relacin molar entre agua y etanol de 0 a 17
kg catalizador/(mol/s). Una vez realizada la experimentacin se demuestra que el
modelo terico y los resultados experimentales muestran una gran concordancia para
todo el rango de parmetros investigados.
Sin embargo la comparacin entre reactores de membrana y tradicionales no es
posible por carecer de datos de el reformado de etanol en reactores de membrana.
Fig 1. Conversin de etanol en funcin alimentacin
Factores que influyen en el funcionamiento del reactor de membrana

1
Atendiendo a la temperatura
Segn la grfica 2 existe una relacin directa entre la conversin de etanol y la

relacin molar de alimentacin para los dos reactores
Fig 2. Conversin de etanol en funcin de alimentacin
Lo primero que se observa es que la conversin se incrementa para ambos reactores

con la temperatura debido a la elevada actividad cataltica. Adems para todo el rango
de relaciones molares de la alimentacin y la temperatura la conversin de etanol es,
como se puede esperar, superior en reactores de membrana. Esto se debe al paso de
hidrgeno a travs de la membrana que desplaza la reaccin hacia los productos.
Profundizando ms se puede indicar que la diferencia de conversin entre ambos
reactores se incrementa con la temperatura, ya que existe una dependencia entre el
flujo de hidrgeno a travs de la membrana y la temperatura, que provoca que a ms
temperatura el flujo aumenta, y con l la conversin. Tambin se puede ver un
incremento en la diferencia entre ambas conversiones a medida que aumenta la
relacin molar
2
Atendiendo a la presin
En la grfica 3 se observan tres zonas distintas segn la relacin molar de

alimentacin.
Fig. 3. Conversin de etanol en funcin de la alimentacin
En la primera zona, para W/F<3kgcat/(mol/s).

Incrementando la presin en el reactor de 1 a 8 bares la conversin en un reactor
tradicional y de membrana aumenta, lo que puede deberse a la baja relacin de W/F lo
que incrementa la adsorcin de las especies sobre la superficie cataltica a medida
que aumenta la presin, lo cual contribuye a aumentar la conversin del etanol
En la segunda zona, para 3<W/F<15 kgcat/(mol/s).
El comportamiento de la conversin frente a la presin se maximiza para bajas
presiones a medida que W/F aumenta. A altas presiones el control termodinmico es
superior y provoca conversiones menores
En la tercera zona, para W/F>14 kgcat/(mol/s).
nicamente la termodinmica ejerce control en el proceso, y a ms temperatura ese
control es mayor, luego la conversin decrece en reactores tradicionales al aumentar
la presin. Y en reactores de membrana la conversin aumenta al aumentar la presin
La presin tiene dos efectos contrarios en la reaccin: uno positivo acerca del paso de
de producto a travs de la membrana y otro negativo o positivo segn el reactor y las
condiciones del proceso
Transesterificacin catalizada de aceite de canola
La reaccin principal que tiene lugar es la transformacin del aceite de canola gracias
a metanol y a un catalizador acido o bsico en biocombustible FAME (fatty acid methyl
esters)
El biodiesel que se obtiene de esta forma est mas oxigenado y por tanto es un mejor
lubricante que los tradicionales y permite incrementar la vida de los motores y hace
posible combustiones ms completas.
Los mtodos convencionales de transesterificacin se llevan a cabo en reacciones con
dos fases lquidas distintas e inmiscibles, lo que da como resultado una transferencia
de masa limitada [19-20].
De una forma ms especfica para el caso que se trata aqu, el canola oil y el metanol
son inmiscibles, esto hace que para obtener rendimientos elevados en la operacin
sea necesaria la utilizacin de un reactor de membrana, concretamente de membrana
de carbono orgnico.
El proceso de reaccin consiste en la produccin de biodiesel usando un reactor de
membrana de carbono con dos fases y flujo semi-batch, con ayuda de un catalizador
cido-basico.
Un esquema de la operacin a utilizar podra ser esta:
Fig 4. Esquema del proceso
Esto se llevar a cabo para diferentes concentraciones catalticas, flujos de

alimentacin y temperaturas con el objetivo de establecer un rango de mximo
rendimiento y estabilidad
Estrategias de operacion
Se parte de una carga de aceite de canola dentro del reactor de membrana. Una
mezcla de metanol y acido sulfrico mas catalizador previamente mezclados se
introducen con flujo constante dentro del reactor con ayuda de una bomba.
A continuacin se conecta un intercambiador de calor para alcanzar las temperaturas
de operacin de 60 65 70C
Una vez conectado el intercambiador, se observ que la reaccin se estabilizaba a los
30 minutos para 60C, 40 min para 65C y 45min para los 70C
El experimento analizado tuvo lugar con una presin controlada de 138kpa y de
duracin seis horas cada uno de ellos.
La miscibilidad es un factor a tener en cuenta en este proceso ya que son conocidos
los problemas de miscibilidad entre el aceite de canola y el metanol, siendo la
temperatura un factor con poco efecto sobre este aspecto
A su vez el biodiesel obtenido (FAME) es completamente miscible en metanol para un
amplio rango de temperaturas.
Como se indico anteriormente la inmiscibilidad de los compuestos de alimentacin es
clave en los procesos de transferencia de materia que tendrn lugar y es justamente el
factor determinante para la transesterificacin en un reactor de membrana
El principio de funcionamiento de la membrana se explica en la siguiente figura
Fig. 5. Funcionamiento de la membrana
Debido a la inmiscibilidad de la alimentacin y a varias fuerzas superficiales el aceite

de canola existe en el reactor en forma de emulsin como gotas suspendidas en el
metanol.
Las membranas tienen poros de reducido dimetro que no permiten que las gotas de
aceite las atraviesen porque son demasiado grandes para los poros, por otro lao el
FAME es soluble en el metanol y debido a su pequeo tamao molecular pasa a
travs de la membrana llevando tras de s etanol, glicerol y catalizador
La reaccin principal que se da es reversible, aunque el equilibrio tiende en la
direccin de los productos FAME y glicerol.
Para incrementar la produccin de biocombustible el FAME debe ser removido durante

la reaccin para desplazar la reaccin ms hacia los productos.
Resultados y discusin
Los resultados obtenidos de la reaccin evidencian que durante sta el aceite de
canola no apareci en la zona con filtrado
La tcnica utilizada para determinar los componentes en ambas zonas, filtrado y
retenido, es una cromatografa de liquidos de alto rendimiento.
Se observa que en la zona del retenido el componente mayoritario es el aceite de
canola y algunas trazas de biodiesel y otros compuestos. Por el lado del filtrado se
observa la ausencia del pico correspondiente al aceite de canola, indicando su
ausencia y la alta pureza de biodiesel en la zona de filtrado.
El efecto de la temperatura sobre la reaccin produce un aumento de la conversin a
medida que aumenta la temperatura, y variando la temperatura de 60 90C se produce
una gran variacin del tiempo de reaccin [24-25]
Por otro lado un aumento de la concentracin de catalizador provoca un aumento de la
conversin y hace la reaccin ms sensible a la temperatura
En cuanto al efecto de flujo de alimentacin de metanol acido, un aumento de ste
provoca un aumento de la conversin
El efecto que tiene utilizar catalizador bsico es destacable respecto al acido ya que se
obtienen conversiones mucho mayores para el primer caso. Con pocas cantidades de
catalizador bsico se observan conversiones muy elevadas [26]
Las membranas de carbono utilizadas en este proceso poseen una gran resistencia
qumica a cidos y bases. Y puede decirse que 10 meses despus de los
experimentos realizados no se apreciaban signos de deterioro
Como conclusin del proceso indicar que la puificacion del biodiesel obtenido a partir
del aceite de canola constituye uno de los retos ms importantes en la produccin de
biodiesel, las cuales se ven aliviadas en gran medida por el uso de reactores de
membrana
Como resea final informar de que la concentracin del biodiesel en el filtrado no era
constante en el tiempo, al principio era muy alta y despus disminua, probablemente
debido a efectos de ensuciamiento
Hidrlisis enzimtica de la lignocelulosa

La lignocelulosa es el principal componente de la pared celular de las plantas, esta
biomasa producida por la fotosntesis es la fuente de carbono renovable ms
prometedora para solucionar los problemas actuales de energa. Se han desarrollado
diversos mtodos que mejoran la hidrlisis de la lignocelulosa para poder utilizar su
potencial como biocombustible
Un problema importante de la hidrlisis enzimtica de la lignocelulosa es la inhibicin
de los productos de reaccin. Por ello, es importante el diseo de reactores para
minimizar este problema.[27-28]
Claves en la operacin
Una solucin podra ser usar reactores de membrana que mediante el uso de enzimas
no inmovilizadas en la membrana del reactor y retirando continuamente los productos
de reaccin se consigue evitar en gran medida los problemas de inhibicin.
Fig 6. Rendimiento frente a tiempo de hidrlisis
La grfica nos [29] permite observar que para tiempos de hidrlisis elevados la
inhibicin es baja, ya que el rendimiento de la operacin (ordenada) aumenta con el
tiempo segn la relacin molar de la concentracin de inhibidor y la concentracin de
glucosa.
No obstante, a tiempos de reaccin bajos el rendimiento es menor por lo cual los
mtodos para operar en continuo con elevado rendimiento, como aqu se habla, son
muy apreciados.
El empleo de un reactor de membrana para este proceso, permite una alimentacin
continua de substrato y una retirada continua de producto sin prdida de enzimas[3335] E incluye[30]:
a) El uso de enzimas celulticas por largos periodos de tiempo al ser retenidas en
el sistema.
b) La obtencin de una mayor conversin de los productos al reducir su tasa de
inhibicin.
c) La obtencin de productos de hidrlisis puros, libres de contaminantes como
enzimas, sustrato sin convertir y otros residuos e impurezas del proceso que de
otra manera disminuiran el rendimiento de la hidrlisis.
d) La posibilidad de mantener una corriente de concentracin constante de
productos sin aadir nuevas enzimas.
Por otra parte, el usar reactores de membrana nos es muy adecuado en cuanto al
proceso de retirada de glucosa ya que a concentraciones muy bajas revierte en una
concentracin final de etanol baja en la mezcla de fermentacin y provoca unos costes
elevados de destilacin para la recuperacin de etanol.
Para que la operacin se mantenga con unas concentraciones de etanol finales
aceptables el volumen de reaccin debe ser constante; lo cual, obliga a mantener una
operacin continua y simultnea de alimentacin y filtrado en el reactor. Tambin
obliga a una carga de alimentacin no demasiado elevada ya que puede haber
problemas con la resistencia de la membrana y con la polarizacin de las enzimas, lo
cual revierte en una reduccin del flujo a travs de la membrana.
La alta viscosidad de la alimentacin de lignocelulosa entorpece la mezcla perfecta y
la transferencia de materia y es un problema grave a altas concentraciones de
alimentacin.
10
Por estas razones dadas, no se puede hidrolizar de forma continua y a gran escala la
lignocelulosa.
El empleo de reactores de membrana produce un beneficio en cuanto a conversin ya
que comparndolo con procesos batch, la diferencia de conversiones es superior a un
40%. En la mayora de los casos, el grado final de conversin ronda el 70-90%[35-36]c.
Fig 7. Formacin de producto frente a rendimiento
Esta grfica muestra las diferentes formas de aproximacin a un rendimiento elevado

segn tres curvas de inhibicin y comparando todo ello con la formacin de productos.
La recta superior sera la ms favorable para la obtencin de productos con elevado
rendimiento y corresponde a la ausencia de inhibicin. Mientras que la curva
representa la inhibicin de los productos y solo permite una formacin de productos
elevada para bajos rendimiento. Sin embargo, la ltima recta es una representacin de
la inhibicin clsica y sirve como referencia para las otras dos.
Adems, un reactor de membrana es cinco veces ms efectivo por unidad msica de
enzima que un proceso batch que es a su vez ms eficaz que el resto de procesos
tradicionales (continuos, tubular).[31]
Estrategias de operacin en reactores de membrana:
Existen varias configuraciones a la hora de operar en las cuales el reactor de
membrana es el dispositivo principal. En algunos casos, se utiliza un reactor continuo
de tanque agitado como mezclador auxiliar, como un volumen adicional de reaccin o
como un tanque de descarga de material no convertido. Tambin pueden encontrarse
mdulos de ultrafiltracin de membrana plana y tubular.
Estrategias de retirada de productos.
Los productos pueden ser retirados del sistema mediante filtracin mediante un
incremento de la presin. La retirada del inhibidor a travs de la membrana hace
necesaria la adicin de ms lquido para mantener el volumen de reaccin constante lo
cual implica operar en continuo.
Retencin de enzimas:
Las enzimas suelen aadirse al principio de la reaccin y se reutiliza a medida que
esta avanza y estas deben ser retenidas por la membrana a la vez que sta es
permeable a los productos.[31] [37]
Las enzimas que no estn firmemente unidas al substrato se recuperan mediante una
combinacin de unidades membranosas.
11
Factores que influyen en el funcionamiento del reactor de membrana

En general para una reaccin enzimtica con inhibicin de productos existen ciertos
factores de los que depende el proceso:
a)
b)
c)
d)
e)
f)
Cantidad de producto formado

ndice de la inhibicin de productos de celulasas.
Tasa de retirada de productos
Composicin de la alimentacin
Tasa de desactivacin de las celulasas.
Polarizacin de las enzimas y ensuciamiento.
Tambin existen factores ms especficos de los que depende el proceso:

a) Reaccin lenta
b) Rpida inhibicin de los productos.
c) Rpido ensuciamiento de la membrana debido a la viscosidad de la mezcla de
alimentacin
d) El tamao de productos intermedios de reaccin son suficientes para atravesar
la membrana.
4. BIBLIOGRAFA
[1] K Liguras D, Kondarides DI, Verykios XE. Production of hydrogen for fuel cells by
steam reforming of ethanol over supported noble metal catalysts. Appl Catal 2003;43:
34554.
[2] Sahoo DR, Shilpi Vajpai, Sanjay Patel, Pant KK. Kinetic modelling of steam
reforming of ethanol for the production of hydrogen over Co/Al2O3 catalyst. Chem Eng
J 2007;125(3):13947.
[3] Freni S, Mondello N, Cavallaro S, Cacciola G, Pardon VN, Sobyanin VA. Hydrogen
production by steam reforming of ethanol: a two process. React Kinet Catal Lett
2000;71:14352.
[4] Srinivas D, Satyanarayana CVV, Potdar HS, Ratnasamy P. Structural studies on
NiOCeO2ZrO2 catalysts for steam reforming of ethanol. Appl Catal 2003;246:323
34.
[5] Llorca J, Homs N, Sales J, de la Piscina PR. Efficient production of hydrogen over
supported cobalt catalysts from ethanol steam reforming. J Catal 2002;209:30617.
[6] Haga F, Nakajima T, Yamashita K, Mishima S. Effect of crystallite size on the
catalysis of alumina-supported cobalt catalyst for steam reforming of ethanol. React
Kinet Catal Lett 1998;63:2539.
[7] Kaddouri A, Mazzocchia C. A study of the influence of the synthesis conditions upon
the catalytic properties of Co/SiO2 or Co/Al2O3 catalysts used for ethanol steam
reforming. Catal Commun 2004;5:33945.
12
[8] Shu J, Grandjean BPA, Van Neste A, Kalaguine S. Catalytic palladium-based

membrane reactors: a review. Can J Chem Eng 1991;69:103660.
[9] Kikuchi E. Membrane reactor application to hydrogen production. Catal Today
2000;56:97101.
[10] Armor JN. Applications of catalytic inorganic membrane reactors to refinery
products. J Membr Sci 1998;147: 21733.
[11] Paturzo L, Basile A, Drioli E. High temperature membrane reactors and integrated
membrane operations. Rev Chem Eng 2002;18(6):51151.
[12] Tosti S, Basile A, Bettinali L, Borgognoni F, Chiaravalloti F, Gallucci F. Long-term
tests of PdAg thin wall permeator tube. J Membr Sci 2006;284:3937.
[13] Wieland S, Melin T, Lamm A. Membrane reactors for hydrogen production. Chem
Eng Sci 2002;57:15716.
[14] Amandusson H, Ekedahl LG, Dannetun H. Alcohol dehydrogenation over Pd
versus PdAg membranes. Appl Catal A: Gen 2001;217:15764.
[15] Basile A, Gallucci F, Paturzo L. A dense Pd/Ag membrane reactor for methanol
steam reforming: experimental study. Catal Today 2005;104:24450.
[16] Basile A, Gallucci F, Paturzo L. Hydrogen production from methanol by oxidative
steam reforming carried out in a membrane reactor. Catal Today 2005;104:2519.
[17] Basile A, Tosti S, Capannelli G, Vitulli G, Iulianelli A, Gallucci F, et al. Co-current
and counter-current modes for methanol steam reforming membrane reactor:
experimental study. Catal Today 2006;118:23745.
[18] Gallucci F, Basile A, Tosti S, Iulianelli A, Drioli E. Methanol steam reforming and
ethanol steam reforming in membrane reactors: an experimental study. Int J Hydrogen
Energy 2007;32(9):120110.
[19]Boocock, D.G.B., Konar, S.K., Mao, V., Lee, C., Buligan, S., 1998. Fast formation
of high-purity methyl esters from vegetable oils. J. Am. Oil hem. Soc. 75, 11671172.
[20]Boocock, D.G.B., Konar, S.K., Mao, V., Sidi, H., 1996. Fast one-phase oil-rich
processes for preparation of vegetable oil methyl esters. Biomass Bioenergy 11, 43
50.
[21]Saracco, G., Neomagus, H.W.J.P., Versteeg, G.F., van Swaaij, W.P.M., 1999.
High-temperature membrane reactors: potential and problems. Chem. Eng. Sci. 54,
19972017.
[22]Saracco, G., Specchia, V., 1994. Catalytic inorganic-membrane reactors: present
experience and future opportunities. Cat. Rev.: Sci. Eng. 36, 305384.
[23]Saracco, G., Versteeg, G.F., Van Swaajj, W.P.M., 1994. Current hurdles to the
success of high temperature membrane reactors. J. Membr. Sci. 95, 105123.
13
[24]Liu, J., 2004. Biodiesel Production from Canola Oil Using a Membrane Reactor.
M.A.Sc. Thesis, Department of Chemical Engineering, University of Ottawa, Canada0
[25]Liu, K., 1994. Preparation of fatty acid methyl esters for gas-chromatographic
analysis of lipids in biological materials. J. Am. Oil Chem. Soc. 71, 11791187.
[26]Freedman, B., Pryde, E.H., Mounts, T.L., 1984. Variables affecting the yields of
fatty esters from transesterified vegetable oils. J. Am. Oil Chem. Soc. 61, 16381643
[27]Gan et al., 2003 Q. Gan, S.J. Allen and G. Taylor, Kinetic dynamics in
heterogeneous enzymatic hydrolysis of cellulose: an overview, an experimental study
and mathematical modeling. Process Biochem, 38 (2003), pp. 10031018.
[28]Gusakov et al., 1987 A.V. Gusakov, A.P. Sinitsyn and A.A. Klyosov, Factors
affecting the enzymatic hydrolysis of cellulose in batch and continuous reactors :
computer simulation and experiment. Biotechnol Bioeng, 40 (1987), pp. 663671.
[29]Andri et al., 2010b P. Andri, P.A. Jensen, A.S. Meyer and K. Dam-Johansen
Reactor design for minimizing product inhibition during enzymatic lignocellulose
hydrolysis: I. Significance and mechanism of glucose inhibition on cellulolytic enzymes
(2010).
[30]Cantarel et al., 2009 B.L. Cantarel, P.M. Coutinho, C. Rancurel, T. Bernard, V.
Lombard and B. Henrissat, The carbohydrate-active enzymes database (CAZy): an
expert resource for glycogenomics. Nucleic Acids Res, 37 (2009), pp. D233D238.
[31]Hahn-Hgerdal et al., 1981 B. Hahn-Hgerdal, E. Andersson, M. Lpez-Leiva and
B. Mattiasson, Membrane biotechnology, co-immobilization, and aqueous two-phase
systems: alternatives in bioconversion of cellulose. Biotechnol Bioeng Symp, 11
(1981), pp. 651661.
[32]Blafi-Bak et al., 2006 K. Blafi-Bak, A. Koutinas, N. Nemestthy, L. Gubicza
and C. Webb, Continuous enzymatic cellulose hydrolysis in a tubular membrane
bioreactor. Enzyme Microb Technol, 38 (2006), pp. 155161.
[33]Hong et al., 1981 J. Hong, G.T. Tsao and P.C. Wankat, Membrane reactor for
enzymatic hydrolysis of cellobiose. Biotechnol Bioeng, 23 (1981), pp. 15011506.
[34]Yang et al., 2006 S. Yang, W. Ding and H. Chen, Enzymatic hydrolysis of rice straw
in a tubular reactor coupled with UF membrane. Process Biochem, 41 (2006), pp.
721
[35]Alfani et al., 1983 F. Alfani, M. Cantarella and V. Scardi, Use of a
membrane reactor for studying enzymatic hydrolysis of cellulose. J Membr Sci, 16
(1983), pp. 407416.
[36]Henley et al., 1980 R.G. Henley, R.Y.K. Yang and P.F. Greenfield, Enzymatic
saccharification of cellulose in membrane reactors . Enzyme Microb Technol, 2
(1980), pp. 206208.
[37]Knutsen & Davis, 2004 J.S. Knutsen and R.H. Davis, Cellulase retention and sugar
removal by membrane ultrafiltration during lignocellulosic biomass hydrolysis. Appl
Biochem Biotechnol, 113116 (2004), pp. 585599.
14
5.- ANEXO
15
Biotechnology Advances 28 (2010) 407425
Contents lists available at ScienceDirect
Biotechnology Advances
j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / b i o t e c h a d v
Research review paper
Reactor design for minimizing product inhibition during enzymatic

lignocellulose hydrolysis
II. Quantication of inhibition and suitability of membrane reactors
Pavle Andri, Anne S. Meyer , Peter A. Jensen, Kim Dam-Johansen
Department of Chemical and Biochemical Engineering, Technical University of Denmark, DK-2800 Kgs. Lyngby, Denmark
a r t i c l e
i n f o
Article history:
Received 14 October 2009
Received in revised form 26 January 2010
Accepted 13 February 2010
Available online 19 February 2010
Keywords:
Membrane reactors
Glucose removal
Enzyme inhibition
Hydrolysis kinetics
a b s t r a c t
Product inhibition of cellulolytic enzymes affects the efciency of the biocatalytic conversion of
lignocellulosic biomass to ethanol and other valuable products. New strategies that focus on reactor designs
encompassing product removal, notably glucose removal, during enzymatic cellulose conversion are
required for alleviation of glucose product inhibition. Supported by numerous calculations this review
assesses the quantitative aspects of glucose product inhibition on enzyme-catalyzed cellulose degradation
rates. The signicance of glucose product inhibition on dimensioning of different ideal reactor types, i.e.
batch, continuous stirred, and plug-ow, is illustrated quantitatively by modeling different extents of
cellulose conversion at different reaction conditions. The main operational challenges of membrane reactors
for lignocellulose conversion are highlighted. Key membrane reactor features, including system set-up,
dilution rate, glucose output prole, and the problem of cellobiose are examined to illustrate the quantitative
signicance of the glucose product inhibition and the total glucose concentration on the cellulolytic
conversion rate. Comprehensive overviews of the available literature data for glucose removal by
membranes and for cellulose enzyme stability in membrane reactors are given. The treatise clearly shows
that membrane reactors allowing continuous, complete, glucose removal during enzymatic cellulose
hydrolysis, can provide for both higher cellulose hydrolysis rates and higher enzyme usage efciency
(kgproduct/kgenzyme). Current membrane reactor designs are however not feasible for large scale operations.
The report emphasizes that the industrial realization of cellulosic ethanol requires more focus on the
operational feasibility within the different hydrolysis reactor designs, notably for membrane reactors, to
achieve efcient enzyme-catalyzed cellulose degradation.
2010 Elsevier Inc. All rights reserved.
Contents
1.
2.
3.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1.
Inuence of product inhibition on enzyme-catalyzed rates . . . . . . . . . . . . . . . . . . .
1.2.
Dimensioning of ideal continuous reactors for enzymatic degradation of (ligno)cellulose . . . .
1.3.
Glucose formation rates at different lignocellulose dry matter contents (DM%) in a batch reactor
Design of membrane reactors for hydrolysis products removal . . . . . . . . . . . . . . . . . . . .
2.1.
Membrane bioreactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.
Membrane bioreactors for lignocellulose hydrolysis: key issues . . . . . . . . . . . . . . . .
2.3.
Quantitative effects of product removal on the cellulolytic hydrolysis rates and extent of cellulose
Membrane reactors operation strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.
Product removal strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.
System set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.
Enzyme retention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.
Specic design features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Corresponding author. Tel.: +45 4525 2800.

E-mail address: am@kt.dtu.dk (A.S. Meyer).
0734-9750/$ see front matter 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.biotechadv.2010.02.005
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
conversion
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
408
408
410
411
412
412
412
413
414
414
414
415
416
408
P. Andri et al. / Biotechnology Advances 28 (2010) 407425
4.
Key factors inuencing membrane reactor performance for enzymatic cellulose hydrolysis . . . .
4.1.
Glucose output prole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.
The effect of dilution rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.
The problem of cellobiose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.
Factors affecting the membrane ux . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.
Fouling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.
Reaction slurry properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.
Molecular cut-off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.
Concentration polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.
Membrane reactors for glucose removal during (ligno)cellulose hydrolysis: operational challenges
6.1.
The glucose concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2.
Fed-batch operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3.
Continuous operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4.
Enzyme activity retainment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.
Membrane reactors and hydrolysis kinetic studies . . . . . . . . . . . . . . . . . . . . . . .
7.1.
General kinetic studies using membrane reactors . . . . . . . . . . . . . . . . . . . .
7.2.
Product inhibition studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.
Other techniques for glucose removal . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.1.
Two-phase systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.
Dialysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.3.
Simultaneous hydrolysis and fermentation (SSF) and removal of ethanol . . . . . . . . .
9.
Membrane bioreactor design for glucose removal: conclusions and recommendations . . . . . .
9.1.
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.2.
Advantages and challenges of membrane reactors . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1. Introduction
Product inhibition of cellulases by cellobiose and glucose has long
been known to signicantly retard the rates of enzyme-catalyzed
cellulose hydrolysis (Gan et al., 2003; Gusakov et al., 1987). This
inhibition constitutes a main obstacle for achieving efcient enzymatic degradation of cellulose and high glucose yields in current
lignocellulose-to-ethanol processing schemes (Andri et al., 2010a;
Bla-Bak et al., 2006; Xiao et al., 2004). The product inhibition of
cellulolytic enzymes also affects the efciency of other processes
involving conversion of lignocellulosic biomass to valuable products.
Alleviation of this product inhibition, notably the inhibition by the
hydrolysis end-product glucose, is therefore a key prerequisite for
achieving cost-efcient conversion of lignocellulosic biomass to
biofuels notably bioethanol and biobutanol and other valuable
products such as platform biochemicals. A number of glucose tolerant
fungal -glucosidases, produced by various Aspergillus spp. and e.g.
Humicola insolens, have been identied relatively recently (Decker
et al., 2001; Sonia et al., 2008), but the prospects of developing and
using glucose tolerant enzymes seem to receive surprisingly limited
attention in the commercial enzyme development for biomass
utilization. Rather, the industrial focus has mainly been on reducing
the enzyme costs by improving the efciency of known enzymes,
identifying new, more active enzymes, creating optimal enzyme
mixtures for selected pre-treated substrates, and on minimizing the
enzyme production costs (Merino and Cherry, 2007; Rosgaard et al.,
2007b). A careful analysis of the mechanisms and kinetics of the
product inhibition induced by glucose and cellobiose on microbial
cellulases and -glucosidase has substantiated that reactor designs
which involve continuous or semi-continuous product removal
notably glucose removal must be at the core of future-directed
design strategies for lignocellulose-to-ethanol processes (Andri et al.,
2010b).
Simultaneous saccharifaction and fermentation (SSF), with or
without separate fermentation of pentose monosaccharides, is
considered a main technology scenario in current biomass-to-ethanol
processes (Hahn-Hgerdal et al., 2006; Lynd et al., 2008). Although
alleviation of product inhibition is a rationale for SSF, it seems to have
been overlooked that the efciency of this technology is restricted by
the inhibition that the ethanol exerts on the cellulolytic enzymes
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
416
417
417
417
418
418
418
418
418
419
419
419
419
420
421
421
421
421
421
422
422
423
423
423
424
(Bezerra and Dias, 2005). Hence, a certain degree of separate

enzymatic hydrolysis of the cellulosic biomass appears to be the
most feasible approach for accomplishing the enzymatic degradation
of cellulose to glucose in future large scale cellulose-to-ethanol
processes and in other lignocellulosic biomass upgrading processes as
well.
The purpose of this review is to examine the quantitative effects of
product removal on lignocellulose hydrolysis efciency, i.e. the
inuence of glucose removal on the rates and extents of conversion
in enzymatic cellulose hydrolysis, and to discuss the key reactor
design issues, operational features, and the overall advantages and
disadvantages of membrane reactors for glucose product removal
during cellulolytic enzyme hydrolysis. By highlighting the immense
potential as well as the challenges that lie ahead in the development
of reactor systems that reduce the product inhibition of cellulases, our
objective is to provide an improved knowledge-base for rationally
designing reactor systems for efcient enzymatic cellulose hydrolysis.
The present review is tightly connected to another report which
examines the reaction mechanisms and product inhibition kinetics on
enzymatic cellulose hydrolysis in relation to the particular complexities of enzyme-catalyzed cellulose hydrolysis (Andri et al., 2010b).
1.1. Inuence of product inhibition on enzyme-catalyzed rates
The effects of inhibitors especially their inuence on the initial
reaction rate have been extensively studied in classical enzyme
kinetics and enzymology. The evaluation of enzyme inhibition has for
example for a long time been one of the major methods used in
pharmacological research to analyze and quantify the action of drugs
and in drugs development (Levenspiel, 1993). It is of course also well
known that product inhibition can hinder the obtainment of high
yields and high converison rates in industrial enzyme technology
(Riebel and Bommarius, 2004; Frieden and Walter, 1963; Fullbrook,
1996). However, apart from a few important cases (e.g. lactose
hydrolysis), the negative effects of product inhibition has surprisingly
rarely led to drastic changes in processing regimes and reactor design
in large scale industrial enzyme reactions. If product inhibition had
been more in focus it is our presumption that signicantly fewer
simple batch reactors and batch reactions would be in place in
industrial enzyme technology.
The particular inhibition effect that product inhibition exerts can

be assessed by comparing the quantitative inuence of a product
inhibitor (P) and a classic (non-product) inhibitor (I) on biocatalytic
product formation rates in a batch reactor. Based on a previous
analysis (Andri et al., 2010a) we employed non-competitive
inhibition kinetics to quantify the product inhibition in multienzymatic lignocellulose hydrolysis (Fig. 1). In this case, where
two inhibited enzymatic reactions have the same kinetic properties
(KM, kcat) and (hypothetically) have the same reactants and enzyme
concentrations, and are analyzed under the same conditions, the
presence of I affects the (uninhibited) rates through the constant
quantity 1 + I/KI while P acts through the term 1 + P/KIP (the same
inhibition strength, KIP = KI, both non-competitive). Although the two
terms are essentially similar, the important difference is that 1 + P/KIP
increases as the reaction progresses because P increases, while the
term 1 + I/KI remains constant throughout the enzymatic reaction.
Thus, the presence of I in the reaction medium reduces the rates
virtually instantaneously, followed by a reduction of an equal portion
of the reaction rate throughout the reaction (Fig. 1). P on the other
hand diminishes the enzyme-catalyzed rates to a lesser extent during
the initial stages of the reaction, namely where P is low and from
where it follows that also P/KIP and hence 1 + P/KIP are low. However,
the factor 1 + P/KIP then increasingly affects the reaction rate as the
reaction progresses because of the product concentration increment
(Fig. 1). It is a poor consolation that the only curb on this is that the
fraction of free, soluble enzymes (f cl) that may be inhibited by
classical non-product inhibitors, is constant during the course of
reaction, whereas the fraction of theoretically available free enzyme
that is product-inhibited (f pr) will increase as the reaction progresses:
cl
= 1
Ecl
1

= 1
I
E0
1+
KI
f pr = 1
Epr
1

= 1
P
E0
1+
KIP
In these equations Ecl and Epr represent the concentrations of free

enzyme when classical non-product and product inhibitors are
present, respectively, E0 is the total initial enzyme concentration, and
Fig. 1. The effect of the non-competitive classical

non-competitive product inhibitor
dG
dt
dG
dt
nkcat
E0 S
KM + S 1 +
nkcat
E0 S
KM + S 1 +

P
KIP

I
KI
and
409
E is the concentration of the theoretically available free enzyme in the

uninhibited state:
E
E0
0
=

Ecl =
S
I
I
1+
1+
1+
KM
KI
KI
E
pr
E
E
0
=

=
S
P
P
1+
1+
1+
KM
KIP
KIP
These two types of inhibitors will exhibit the same effect on the
catalysis rate at the point where P = I (in Fig. 1 this point is 0.17 g/g or
0.01 M). Since the product concentration increases during the
enzyme-catalyzed reaction starting from virtually 0 and then
gradually approaching a maximal concentration at the maximum
extent of conversion the enzyme will experience a range of product
concentrations in succession that are (usually) higher than the
concentration of I. The efciencies of enzyme-catalyzed reactions at
high extents of substrate conversion are thus signicantly affected in
both batch and continuous processes when sensitive to product inhibition. Since high product concentrations are required in the
prospected large scale lignocellulosic conversion processes it must
be anticipated that the product inhibition will signicantly retard the
hydrolysis reaction rates.
It is important to note that the progressive feature of product
inhibition is the reason why it is often neglected in initial rate enzyme
kinetics inhibition studies. Nevertheless, the signicant inuence of
the product concentration increment in product inhibition is exactly
the reason why design of reactors that involve continuous or semicontinuous removal of the products from the enzyme-catalyzed
reaction during the reaction must be considered in industrial-scale
biomass processing demanding high conversion degrees.
Cellobiose exerts the strongest inhibition effect on cellulase
activity with typical KI ranges between 0.01 and 6 g/L (Andri et al.,
2010b). However, this inhibition is usually alleviated by adjusting the
dosing of -glucosidase (EC 3.2.1.21) so that the cellobiose is rapidly
hydrolysed to glucose. Unfortunately, glucose also inhibits cellulase
activities, with reported overall KI ranges varying widely from 0.1 to
70 g/L the variation depending mainly on the experimental
conditions. Glucose also exerts a strong inhibition on the activity of
-glucosidases with reported KI values typically ranging from 0.1 to
0.8 g/L (Andri et al., 2010b).
The effect that glucose exerts as a product of cellulose degradation
on the lignocellulose enzymatic conversion may formally be classied
as medium when the molar ratio KM/KIP = 0.14, or very strong, with
mass ratio KM/KIP = 58 where IP indicates glucose as product
inhibitor. These effect estimates are based on parameters published
previously (Andri et al., 2010a) and a classication given by Riebel
and Bommarius (2004). Even for the medium effect (molar KM/KIP =
0.14) the effect of the inhibition on the hydrolysis rate and the glucose
yield is considerable (Fig. 2). When KM/KIP = 10, the reaction is almost
halted, requiring a massively extended reaction time to increase the
yields (Fig. 2).
The physical meaning of the inhibition constant KI or KIP may be
interpreted in a similar fashion as the Michaelis constant KM.
The KM designates the initial substrate concentration (S0) at which
the initial reaction rate (v0) is exactly equal to of the maximal rate
Vmax:
on the enzymatic
reaction model simulation of a batch reaction. The product formation rates are given
as the percentage of the initial uninhibited rate (t = 0; I, P (yield) = 0). Model
parameters and constants: k c at = 12 h 1 , K M = 0.9 mM, K IP = K I = 6.4 mM;
S0 = 0.14 mM, E0 = 0.01 mM, MS = 73566 g/mol, MP = MI = 180 g/mol; I = 0.01 M.
Figure legend:
I classical inhibitor
P product inhibitor
No
inhibition.
v0 =
kcat E0 S0
k E S
V
= cat 0 0 = max
KM + S0
S0 + S0
2
Correspondingly, the KI or KIP can be dened as the concentration

of inhibitor (present initially), e.g. glucose concentration, which
410
catalyzed degradation of lignocellulosic biomass is however complicated by:

a) The required action of multiple enzymatic activities for the
hydrolysis,
b) The dynamically changing heterogeneous reaction system consisting of a solid substrate, i.e. the (ligno)cellulose, and a liquid
phase of soluble intermediate and nal products,
c) The product inhibition on the enzymatic reactions,
d) That the current complicated kinetic models of the reactions only
partially describe the events, and that signicant confusion
regarding the inhibition kinetics exists (Andri et al., 2010b).
Fig. 2. Effect of supposed glucose inhibition power on glucose yield. Model simulation.
Model parameters and constants are given in Fig. 1, except KI which is varied according
to a desired molar KM/KI ratio. The real molar KM/KI ratio is equal to 0.14 (Andri et al.,
2010a). Figure legend:
KM/KI = 10
KM/KI = 1
KM/KI = 0.14
KM/
KM/KI = 0.01
KM/KI = 0.
KI = 0.1
reduces the initial catalyzed rate of uninhibited enzyme (vP,0) to one

half meaning that KI = I or KIP = P0:
v0 =

dP
kcat E0 S0

=
dt 0
KM + S0 1 +
P0
KIP
=
KM
kcat E0 S0

+ S0 1 +
P0
P0
=
vP;0
2
Because vP,0 is given as:

vP;0 =
kcat E0 S0
KM + S0
The numerical value of the inhibition constant KI or KIP may thus be

regarded as fundamental since it provides the quantitative information about the effect of the inhibitor on the initial catalyzed rate (at a
certain initial substrate concentration S0 and initial total enzyme
concentration E0). In turn, the value is also fundamental for the
extended rates which are moreover of industrial relevance. Herein,
the inhibition constants for the classical-non-product and product
inhibitor might not be directly compared due to the progressive
nature of the product inhibition which particularly diminishes the
extended rates. Furthermore, provided that the reported inhibition
constants for glucose on cellulases and -glucosidases are roughly
b10 g/L and in many cases b1 g/L (Andri et al., 2010b), it is indeed
clear that the glucose as a product displays a profound effect on the
enzyme-catalyzed rate of cellulose hydrolysis. The values of these
inhibition constants are particularly signicant considering that current cellulose hydrolysate goals are with glucose at least N100 g/L
(Andri et al., 2010b).
1.2. Dimensioning of ideal continuous reactors for enzymatic
degradation of (ligno)cellulose
Classical chemical conversion of large quantities of raw material is
almost always more feasible and economical when done in continuous reactor regimes rather than in batch reactors (Levenspiel, 1999).
The advantages of continuous reactor systems are also expected to
become apparent in the near-future large-scale production of
lignocellulose-based biofuels. For continuous biocatalytic reactions
in general, the continuous stirred-tank reactor (CSTR) type is a suitable reactor conguration for reactions subject to substrate inhibition,
since the design allows minimization of the substrate concentration.
In contrast, plug-ow reactors (PFR) are considered as more advantageous for reactions subject to product inhibition since they allow
for minimization of the reactor volume for high extents of conversions
as compared to CSTRs (Riebel and Bommarius, 2004). Enzyme-
We have recently reported that simplied MichaelisMenten

based inhibition models actually reasonably well describe the glucose
inhibition of enzyme-catalyzed (Trichoderma reesei cellulases (Celluclast 1.5L) + A. niger -glucosidase (Novozym 188)) degradation of
hot-water pre-treated wheat straw at 2% (w/w) DM in a batch reactor
lab-scale system (Andri et al., 2010a). The supplementation of glucosidase was done because of the low -glucosidase activity in the
T. reesei cellulase product, which is due to the Celluclast 1.5L production process (Rosgaard et al., 2006). However, the -glucosidase
supplementation also served to prevent cellobiose build-up during
the conversion. The progress curves of experiments with different
levels of glucose added to the enzymatic cellulose hydrolysis reactions
were modeled best with the non-competitive MichaelisMenten
inhibition model (Andri et al., 2010a). This model can be used to
compare the required dimensions, i.e. volume and/or length, of ideal
hypothetical continuous reactors of the CSTR and PFR types,
respectively, to obtain a given conversion of (ligno)cellulose, e.g.
30% conversion.
For a given conversion (we have chosen 30% conversion as a goal),
the required reaction time in a batch reactor (tBR = 6 h) and residence
time in continuous reactor (CSTR = 15.6 h, PFR = 6 h), can be obtained
from design equations (Table 1). To demonstrate the correlation
between cellulosic conversion requirements and reactor dimensions,
the appropriate ideal reactor dimensions and productivities were
calculated from 3 different scales of ow rates corresponding to
typical lab, intermediate and pilot/larger scale, respectively: Continuous conversion of lignocellulose requires ideal reactors of large sizes
even to obtain low yields, e.g. only 0.3 gglucose/gglucose potential, and
resulting low glucose concentrations, 3.6 g/L (Table 1). For instance, at
the largest scale, the required CSTR volume of 15.6 m3 or a PFR of 2 m
with L/D = 1 (and a related linear velocity 10 6 m/s) give an
impression of the effect of the reaction rate being b1 g/(L h) (Table 1).
All other things being equal, the size of the equipment will
obviously increase profoundly with increased desired product concentrations as the higher product concentration will demand for
higher conversion degrees (Table 2). In cellulosic processing this must
however be achieved at very low rates (Fig. 1). For instance, an
increase in the desired conversion degree from 15 to 80% will require
an increase in reactor dimensions of 100 times for a CSTR and 40 times
for a PFR. In a hypothetical case where no inhibition by glucose is
occurring, and with the same desired conversion degree increase from
15 to 80%, the required increase in reactor dimensions would only
be 26 times for a CSTR and 10 times for a PFR. Under the kinetic
conditions employed, the PFR volume will always be lower than the
volume of a CSTR. For lignocellulose conversion this volume is 3075%
lower and 1060% lower with glucose inhibition and (hypothetically)
without glucose inhibition, respectively. The direct inuence of
inhibition on reactor dimensioning is seen from Table 2. The presence
of inhibition requires hydrolysis reactors that have a 210 times
(CSTR) or 1.56 times (PFR) higher volume, for cellulose conversion
degrees of 15 and 80%, respectively, than in the absence of inhibition.
Thus, the product inhibition by glucose directly increases the capital
costs of the hydrolysis reactors in the envisaged large-scale
production of bioethanol or biochemicals from cellulosic biomass.
411
Table 1
Comparison of ideal reactor sizes for conversion of lignocelluloses (hydrothermally pre-treated wheat straw) based on experimental results in batch reactor: 30% cellulose
conversion, 3.6 g/L glucose outlet/nal concentration, r = 1.1v = 0.23 g/(L h) (Andri et al., 2010a). The CSTR and PFR were calculated from the design equations based on the
experimentally determined r from the batch reactor (at t = 6 h). The predetermined ratio of the reactors' height to diameter (H/D) for BR/CSTR and reactor's length to diameter (L/D)
for PFR is equal to 1. Reaction conditions: pH 5, 50 C, enzyme dosage 8 FPU/gDM and 13 CBU/gDM; substrate 2% DM content (48% cellulose), S0 = 10 g/L.
Ideal reactor type
Batch
(BR)
Design equation and reaction/residence time [h]
Continuous stirred
(CSTR)
S
tBR = dS = 6 h
S0
CSTR =
S0 S
v
Plug-ow
(PFR)
S
= 15:6 h
PFR = dS = 6 hr
S0
Flow ratea [L/h]
1 10 2
1
1000
1 10 2
1
1000
Volumeb [L]
0.06 c
6
6000
0.16
15.6
15 600
0.06
6
6000
Glucose productivity [g/(L h)]
0.6 d
0.23
0.6
Diameter (DBR/CSTR) or length (LPFR) [m]
0.043
0.2
2
0.058
0.27
2.7
0.043
0.2
2
Suspension velocity [m/s]
1.2 10 6
9.1 10 6
9.1 10 5
a
b
c
d
Fixed.
From reactors design equation.
Approximate size of BR used in inhibition study (Andri et al., 2010a).
Productivity in batch mode.
1.3. Glucose formation rates at different lignocellulose dry matter

contents (DM%) in a batch reactor
When using the MichaelisMenten model incorporating noncompetitive product inhibition (Andri et al., 2010a) to predict the
glucose concentration levels and the glucose formation rates on a broad
range of lignocellulose dry matter contents, ranging from 2 to 40 DM%, it
becomes apparent that the glucose formation rate decreases rapidly as
the reaction progresses (Fig. 3). It also becomes evident that this
decrease is pronounced at all substrate dry matter levels (Fig. 3). Apart
from being related to the rate decrease due to the substrate
consumption, because the reaction rate is a function of [S], the main
part of the rate decrease is caused by the inhibitory effect exerted by the
glucose formed. For example, when more than 20 g/L of glucose have
been released as a result of the enzymatic hydrolysis of cellulose to
glucose, the estimated reaction rate fall-off for a 10% w/w DM substrate
reaction will be similar to that of a 40% w/w DM substrate reaction,
namely, 95% (Fig. 3). These values correspond to a volumetric

productivity decrease to 0.20.45 kgglucose/(m3reactor volume h). However,
the extent of conversion obtained for the 40% DM reaction is
signicantly smaller than that obtained for the 10% DM substrate: 6%
vs. 35%, respectively. Hence, all other things being equal, it appears that
the absolute level of glucose or related glucose yield plays a key role
in decreasing the hydrolysis rate. Since it is the relative ratio of inhibitor:
enzyme(s) that is decisive for the inhibition and not the absolute
glucose concentration, the quantitative data are in effect a consequence
of the increase in the glucose:enzyme ratio, as in the given model
simulation the enzyme concentration per L mixture was kept constant.
Since the glucose:enzyme ratio then constantly increases during regular
enzymatic degradation of cellulose to glucose, the continuous product
removal is obviously a main prerequisite to keep conversion rates high.
With the currently employed enzyme dosage levels in (experimental)
lignocellulosic conversion processes that do vary widely, but which
are generally in the enzyme:substrate range of 510% by weight a
Table 2
Comparison of ideal continuous reactor sizes for conversion of lignocelluloses at
different conversion degrees and inuence of glucose inhibition on reactor dimension,
for the given ow rate of 1000 L/h. Model simulation. Model parameters and constants
are given in Fig. 1. For a special case where inhibition is excluded from simulation, KIP
(for glucose) = 0. The reaction conditions and other data are given in Table 1. rcalc
glucose formation rate obtained from non-competitive kinetic model (Andri et al.,
2010a).
Conversion
[%]
Glucose
[g/L]
rcalc
[g/(L h)]
Reactor volume [m3]

Continuous stirred (CSTR)
Plug-ow (PFR)
0.51
0.27
0.03
2.9
11.5
293
1.9
6.5
75
Without glucose inhibition

14
1.5
1.17
29
3.1
0.98
81
8.8
0.28
1.3
3.2
31
1.2
2.7
12.8
With glucose inhibition

14
1.5
29
3.1
81
8.8
Fig. 3. Inuence of actual glucose concentration on glucose formation rate at different

DM % levels model simulation of batch reaction. Model parameters and constants are
given in Fig. 1, except S0 which was varied according to the DM % level. Glucose
formation rate is given in % of initial value and initial enzyme concentration,
E0 = 0.01 mM is assumed constant. Figure legend:
2%DM
2.5%DM
5%
DM
10%DM
20%DM
40%DM.
412
rough rule of thumb is, therefore, that the glucose should be removed to
at least below 10 g/L in order to drastically regain a heavily inhibited
glucose production rate. It is important to note that this extent of
removal will be independent of the employed substrate concentration. If
the remaining glucose concentration is much above a glucose level of
10 g/L even with glucose removal, the relative ratio of glucose:enzyme
will usually be so high that the positive effect of the glucose removal will
be insignicant (Fig. 3). These relations may explain why many product
removal studies have failed to obtain prominent effects on the
cellulolytic hydrolysis rates (see discussion below).
2. Design of membrane reactors for hydrolysis products removal
In-situ product removal by integration of the biocatalysis reactor
with a separation unit (reactionseparation hybrids) has shown
promising results with product inhibited or equilibrium limited
enzyme-catalyzed conversions (Ahmed et al., 2001; Gan et al., 2002).
On this background, the introduction of membrane (bio)reactors seems
to be one of the obvious approaches to accomplish simultaneous in-situ
removal of glucose during enzymatic hydrolysis of lignocellulose.
2.1. Membrane bioreactors
A membrane (bio-)reactor is a multifunction reactor that combines the reaction with a separation, namely in this case product
removal by membrane separation, in one integrated unit, i.e. in-situ
removal, or alternatively in two or more separate units. In practice,
the hitherto used membrane bioreactors in enzyme technology have
mainly employed ultra- and nanoltration for the separation (Drioli,
2004; Pinelo et al., 2009).
Ultraltration membrane reactors were rst used in conjunction
with development of novel enzyme immobilization techniques.
However, immobilized enzymes are not suitable for insoluble,
polymeric substrates, and this will include lignocellulose, due to the
necessary enzyme adsorption on the macromoleculer substrate
particles that becomes severely mass transfer limited with immobilized enzymes (Alfani et al., 1983). The use of free, un-immobilized,
enzymes conned in membrane reactors avoids some of these
problems, and still allows continuous product removal (HahnHgerdal et al., 1981). Membrane reactors have been investigated for
use in inorganic catalytic reactions for a very long time (Sun and Khang,
1988), but have also been employed for a range of very different
biocatalysis based reactions. These applications include e.g. classical
production of citric acid by fermentation in which the product has to be
removed to maintain high production rates (Chekhova et al., 2000);
selective production of physiologically active chitosan oligosaccharides
by continuous hydrolysis of chitosan (Kuroiwa et al., 2009); production
of whey hydrolysates with low contents of phenylalanine (CabreraPadilla et al., 2009), and continuous production of pure and sterile
glucose solutions from tapioca starch powder (Sarbatly et al., 2007).
Membrane reactors also nd use in the pharmaceutical industry, e.g.
for production of S-ibuprofen (Cauwenberg et al., 1999) as well as in
waste water treatment (Meng et al., 2009). However, apart from a few
seminal studies discussed below, there is a surprising scarcity of data
on membrane reactor performance for enzymatic conversion of
lignocellulose in potential lignocellulose-to-ethanol processes.
2.2. Membrane bioreactors for lignocellulose hydrolysis: key issues
The molecular weight of glucose is 180 g/mol while the molecular
weight of most of the currently used fungal cellulases used for lignocellulose hydrolysis range from 35,000 to 65,000 g/mol (Cantarel
et al., 2009). Several studies using various fungal cellulase systems
and different cellulose substrates have conrmed that it is possible,
via membrane technology, to retain the enzymes present in the
system while allowing the transfer of low-molecular weight reaction
products such as glucose through a membrane (Alfani et al., 1983;

Bla-Bak et al., 2006; Ghose and Kostick, 1970). Due to the
possibility of complete rejection of the long polysaccharide chain
(or rather the solid lignocellulosic substrate particles) and the
biocatalyst, and zero rejection of the main reaction products passing
through the membrane, it can be assumed that the concentration of
the products, i.e. glucose, in the reactor is equal to the concentration of
the products in the permeate. Although some differences may exist
due to concentration polarization affecting the ux (Pinelo et al.,
2009), employment of a membrane reactor principally enables a
design conguration involving continuous feeding of substrate and
removal of product without enzyme loss (Bla-Bak et al., 2006; Gan
et al., 2002; Hong et al., 1981; Yang et al., 2006). The major advantages
of using membrane reactors encompassing product removal during
enzymatic hydrolysis of lignocellulosic materials include:
(a) the use of the cellulolytic enzymes for long periods of time, via
retention in the system;
(b) the obtainment of a higher degree of conversion due to the
reduced product inhibition;
(c) the obtainment of pure hydrolysis products, i.e. free of contaminants such as enzymes, unconverted substrate or other highmolecular weight substances that can harm the processing steps
downstream from the hydrolysis; and
(d) the possibility of maintaining a stream with constant product
concentration without supplying additional enzymes during
extended hydrolysis (i.e. fed-batch or continuous hydrolysis).
On the other hand, a major drawback of using membrane reactors for
glucose removal during bioconversion of lignocellulose is the relatively
low concentration of the product glucose obtained in the permeate, and
the possible leaching of cellobiose. However, the latter will only take
place in case there is not sufcient -glucosidase relative to cellulases in
the enzyme mixture employed. In case of lignocellulose-to-ethanol the
low glucose concentration in the permeate will result in a low nal
percentage of ethanol in the fermentation mixture and related high
distillation costs for the ethanol recovery (Andri et al., 2010b).
Moreover, in a membrane reactor, simultaneous permeation and
dilution is required to keep the reaction volume constant. A key issue
with respect to membrane reactors is that their operational feasibility is
currently not t for high solids biomass loadings; this is mainly due to
the unresolved problems of membrane fouling in current membrane
reactor designs, lack of robustness of the membranes, mixing problems
etc. Moreover, during extended enzymatic reaction, some loss of
enzyme activity may result from thermal or other inactivation, and
this in effect may decrease the nal product concentration in membrane
reactor operations unless the dilution rate is accordingly controlled
(Ishihara et al., 1991). In addition, there may be problems with enzyme
concentration polarization i.e. the build-up of an enzyme boundary
layer near the membrane. If this phenomenon occurs it may reduce both
the ux through the membrane and deplete the enzyme in the bulk
solution (Hong et al., 1981). The enzyme concentration polarization
may be pronounced at elevated feed ow rates and pressures (Hong et
al., 1981), but may thus be controlled via optimization of the particular
reaction. Just as in other hydrolysis reactor types, the high viscosity of
the lignocellulosic biomass mixture is a particular challenge, especially
at high solids loadings. The high viscosity is an obstacle for obtaining
favorable mixing and mass transfer conditions to promote the
enzymatic reactions and the product removal. Another main issue in
relation to lignocellulose conversion is obviously the build-up of
unreacted lignocellulose in the reactor. This unreacted substrate may
notably include lignin and particularly recalcitrant cellulose (see
discussion further below). The problem of unreacted lignocellulose
substrate build-up may in fact be the most signicant problem to
overcome in practical large-scale and/or continuous lignocellulose
processing encompassing membrane bioreactors for the enzymatic
hydrolysis step.
2.3. Quantitative effects of product removal on the cellulolytic hydrolysis

rates and extent of cellulose conversion
The positive effect of the product removal on the enzymecatalyzed hydrolysis rate, on the extent of hydrolysis (degree of
conversion), as well as on the hydrolysis yields is well documented
(Table 3). In general, the extents of conversion of cellulose achieved
by using membrane reactors encompassing product removal have
413
been up to 40% higher than comparable conversions achieved in batch

processes in many cases the nal degree of conversion has been
between 70 and 90%, which is higher than the results typically
obtained in batch reactions (Table 3). Conversion degrees N90% have
been reported only in a few cases, and this high conversion has mainly
been obtained with: (a) operation at very low substrate concentrations (Alfani et al., 1982; Henley et al., 1980); (b) relatively high
enzyme dosage and reaction time (Kinoshita et al., 1986) or, (c) with a
Table 3
Overview of the studies of glucose removal by UF membranes.
Enzyme feeding
Substrate feeding
MF unit/buffer
replacement
Removal method
Dilution rate
[h 1]
Cut-off
[kDa]
Vreactor
[L]
Gmax
[g/L]
Source
Start
Fed-batch
Start
Continuous1
Fed-batch2
Continuous
No/yes
In-situ, continuous
In-situ, intermittent
Separate, continuous
0.38
N/A
N/A9
30
7531
6332
N/A
Ghose and Kostick (1970)
0.61.9
0.310
0.20.5
0.61.3
0.060.6
N/A11
0.060.312
0213
0.120.1614
10
30
10
10
20
4/5
10
0.2
5.518
0.22 + 0.119
0.22 + 0.0320
0.260.3321,24
7
0.065
0.250.325
0.05
1026
0.1
10
N/A
0.060.25
0.4415
T
[C]
4.25.260
4.85
4.8
4.8
4.8
50
Avicel
No/yes
Start
Start
Start
Start
Start
Start
Start
Continuous
Start/continuous
Start
Start/fed-batch
Fed-batch
Fed-batch
Start/continuous
No/yes
No/yes
No/yes
No/yes
No/yes
Yes/yes6
No/yes
In situ, continuous
Separate/continuous
In situ, continuous
In situ, continuous
In-situ/continuous
Separate, intermittent
In situ, continuous
Start
Start
No/yes
Start
Start
Start
Start/fed-batch
Start4
Continuous5
Yes/yes
No/yes
No/yes7
In situ, intermittent
and continuous
Separate, intermittent
Separate, continuous
In situ, continuous
Enzyme source
Substrate type
pH
T. viride37
Solca Floc50
Solca Floc51
Solca Floc52
Cellobiose
Solca Floc53
T. viride38
T. viride39
T. reesei38,40l
A. phoenicis41
T. viride42
Sweet almonds43
T. reesei38
A. niger44
Sporotrichum
cellulophilum45
T. viride46
T. viride
T. reesei47
N/A48
T. reesei45
T. reesei49
50
33
Henley et al. (1980)

Hong et al. (1981)
Klei et al. (1981)
Alfani et al. (1982)
Ohlson et al. (1984)
Kinoshita et al. (1986)
Ishihara et al. (1991)
Lee and Kim (1993)
127
1.9
3.5
0.08
9.534
18
N/A35
1112,34
2513
9.5 34
5016
10
N/A17
0.0522,28
0.0629
0.2730
2736 60
3534
N/A
Knutsen and Davis (2004)

Yang et al. (2006)
Bla-Bak et al. (2006)
DMmax
[% w/v]
E0
[% w/v]
treaction
[h]
Xmax
[%]
90
50
50
50
10
1063
0.04564
0.265
1
0.07773
0.000274
0.02275,76
81
243
0.20.584
2033
200
77
71
92
9133
80
Henley et al. (1980)

Hong et al. (1981)
Klei et al. (1981)
4.8
45
0.1166
0.003377
25
5.2
Alfani et al. (1982)
Sallow54
4.8
40
1067
178
20
9485
KC Floc
55.561
37
1067
279
120
N 90
4.7
45
0.5
5062
40
45
50
50
568
2.569
1570
18.571
2.5
0.180
0.0181
3.282
3.483
588
49
82
7286
53
N 6070
44
50
207059
4.862
4.7
4.7
4.8
4.8
192
240
48
48
16836408
25
25
55
Steamed hardwood
Hardwood kraft pulp
-cellulose
-cellulose56
Corn stover57
Rice straw58
Solca Floc53
Mavicell59
23
45
67
24
46
68
46 h batch, 20% w/v.

48 h batch.
3
Cellobiose.
4
0.5 h batch.
5
9 h batch, circulation.
6
Lignin removal.
7
Not clear.
8
1 mL/min.
9
25, tres based.
10
37 mL/min.
11
max 2.5, 4 L/h.
12
Experimental.
13
Modeling.
14
Membrane 0.0177 m2, 1 L.
15
6.6 L/(m2h), 0.018 m2.
16
Vaccum lter 2025 m.
17
0.1 m = 1000 kDa.
18
4 L CSTR, 1.5 L UF cell.
19
CSTR + UF cell.
20
CSTR + hollow-ber cartridge.
21
Without UF cell.
22
50 g slurry.
250300 rpm.
810 rpm, magnetic bar.
25
500 rpm, 4-blade propeller.
26
Flat-blade impeller.
27
5090 rpm.
28
Shaking.
29
No stirring.
30
Recirculation.
31
7.5 w/v%.
32
6.3 w/v%.
33
3-stage reactor.
34
Reducing sugar.
35
3.5%, 19 h.
36
No fed-batch.
37
QM 9123.
38
SP122.
39
Isolated cellobiase, Miles labs.
40
Powder.
41
QM329 alumina immobilized.
42
B.D.H. Italia.
43
Cellobiase, BBR.
44
Novozym 188.
Enzyme powder extracted.

Meicelase.
47
Extracted Sigma.
48
Iogen.
49
Celluclast 1.5 L.
50
SW 40A, b37 m.
51
SW 40A, b25 m.
52
BW200, 3035 m.
53
BW 200.
54
Salix Q082 alkali pre-treated.
55
Shirakamba.
56
180 mm length.
57
Sulfuric acid pre-treated.
58
Steam exploded.
59
Untreated and heat treated.
60
CSTR studies.
61
Enzyme optimum.
62
From (Lee and Fan, 1983).
63
Total 30% initial.
64
0.45 g/L.
65
2 g/L.
66
1.1 g/L, kinetic study max 2.9.
Gan et al. (2002)
Ghose and Kostick (1970)
Lee and Kim (1993)

Gan et al. (2002)
Knutsen and Davis (2004)
Yang et al. (2006)
Bla-Bak et al. (2006)
100 g/L dry wt.

50 g/L.
69
25 g/L.
70
w/w %.
71
185 g/L dry wt.
72
10% suspension in cellulases.
73
0.77 g/L.
74
2 mg/L.
75
0.22 g/L, 1115 IUCx/L, 5000 IUCx/g.
76
112 IUC1/L, 500 IUC1/g.
77
33 mg/L.
78
10 g/L both enzymes.
79
20 g/L.
80
1 g/L.
81
0.1 g/L.
82
20 FPU/gcellulose.
83
20 FPU/gstraw.
84
tres.
85
1 day operation.
86
Optimal D prole.
87
60 in BR control.
88
v/w %.
414
3-stage hydrolysis of cellobiose (Hong et al., 1981). In several

instances, however, the conversion was lower b70%, and reaching
the limit has typically been explained to be a result of the build-up of
recalcitrant substrate (Bla-Bak et al., 2006).
Ohlson et al., (1984) reported that when the product was
continuously removed the initial rate of enzymatic hydrolysis of
unwashed and washed lignocellulose substrate increased by 4 to 7
times, respectively, as compared to what was obtained in comparable
batch processes. The corresponding extents of conversion on the
unwashed and washed substrate were 70% and 95% as compared to
40% conversion obtained in a batch mode (Ohlson et al., 1984). Henley
et al. (1980) achieved high conversion of cellulosic material as a result of
the diminishing inuence of inhibitory end-products as these were
immediately removed when combined CSTR-HFC (hollow-ber cartridge) or CSTR-UF (ultraltration unit) systems were used. The
amounts removed were 8991% (residence time 1417 min) and
87.591.5% (2327 min), respectively. The comparable reactions in a
batch reactor (BR) had 50% conversion in 20 min or 6266% in a
continuous stirred reactor (CSTR) (12.314.5 min). Klei et al., (1981)
found that the operation of a reactor with immobilized -glucosidases in
continuous mode gave 60% higher enzyme efciency (gproduct/genzyme)
than in a comparable batch hydrolysis, and ascribed the effect to be a
result of the reduced product inhibition. Similarly, Kinoshita et al.
(1986) reported that membrane reactor operation in a semi-continuous
system was ve times more effective per unit weight of enzyme than the
batch reaction. Hahn-Hgerdal et al. (1981) described a batch hydrolysis
(cellulase SP122, 40 C) with product removal at a dilution rate of 0.09
0.28 h 1 which gave 4 times higher reaction degree than that obtained
in a comparable classical batch reactor run. Ghose and Kostick (1970)
found that the continuous membrane reactor was very effective in
separation of sugars from a digest, composed of a dense suspension of
unhydrolyzed cellulosic material in cellulolytic enzymes, reaching 77%
conversion, compared to 60% in a CSTR. Gan et al. (2002) thus found that
the semi-continuous and continuous removal of inhibitory reaction
products i.e. glucose markedly increased the extents of conversion
(51 and 53%, respectively) compared to what was obtained in a batch
reaction (35%).
membrane reactor is the central feature of every set-up; in some

instances there is a stirred reactor (STR) which provides an auxiliary
mixing unit, additional volume for reaction, and/or a possibility to
discharge the unconverted material (Fig. 4). The separation units most
frequently encountered are the common ultraltration modules,
including tubular and at-sheet membrane modules, while special
types of ultraltration, e.g. incorporating adsorption surface or with
immobilized -glucosidases in the shell, and microltration units
have been used less often. The application of widely available
cellulose-based membranes is infrequently seen due to suspected
cellulose-degrading effect of the cellulases.
3.1. Product removal strategies
The inhibitory product can be intermittently, i.e. semi-continuously, or continuously removed from the reaction system primarily in the
form of a permeate. Principally, the product can be removed via
pressure by means of simple dead-end ltration, which, however, also
partially removes the aqueous medium necessary for the continuous
glucose release by the enzyme-catalyzed reaction, and in this way
concentrates the reactants, including the unreacted substrate, in the
retained reaction medium. If removal of the product inhibitor from the
reaction environment is accomplished by membrane separation in this
way, it is necessary to adequately and semi-continuously supplement
new liquid to maintain a constant reaction volume. With time this
supplementation will dilute the glucose level in the permeate stream.
In one of the earliest works, Ghose and Kostick (1970) used a
continuous saccharication membrane reactor system consisting of
an agitated reservoir, in fact a substrate slurry supply vessel, which
contained an aqueous suspension of the predigested (46 h) substrate,
the membrane cell reactor, and a permeate reservoir. The units were
connected to pressurized cylinders for maintenance of desired
pressures (Ghose and Kostick, 1970). The equilibrium conditions for
the ow and reaction rate were attained soon after the ow of the
substrate slurry into the membrane reactor was at the same rate as
the aqueous solution of the reaction products was removed from the
membrane cell (Fig. 5, 2a and 2b).
3. Membrane reactors operation strategies
3.2. System set-up
The operation of integrated membrane reactors for the simultaneous cellulose hydrolysis and removal of inhibitory product can be
performed using several different congurations (Figs. 4 and 5). The
In membrane reactors the enzymes are usually added in the

beginning of the reaction and re-used during its course, but the substrate
can be added at the start of the reaction, i.e. fed semi-continuously as a
Fig. 4. Schematic representation of the process ow sheet for simultaneous lignocellulose hydrolysis and removal of produced glucose. STR (stirred reactor) is needed for achieving
additional conversion due to available volume and/or for unconvertible fraction discharge.
415
Fig. 5. Strategies for operation of membrane reactors: 1. Integrated reaction vessel (stirred reactor) with UF membrane (Alfani et al., 1982; Alfani et al., 1983; Gan et al., 2002; Ghose
and Kostick, 1970; Hong et al., 1981; Kinoshita et al., 1986; Lee and Kim, 1993; Ohlson et al., 1984); 2. Separate reaction vessel with: (2a) UF membrane (Henley et al., 1980; model
system 2 from Ghose and Kostick (1970); combined with PBR in Yang et al. (2006)) , (2b) ordinary (model system 1 from Ghose and Kostick (1970) and special membrane reactor
(with adsorption of substrate and enzymes from Bla-Bak et al., 2006), (2c) MF and UF for recovery of bounded and soluble enzymes, respectively (Knutsen and Davis, 2004),
(2d) UF and MF membrane for recovery of enzymes and removal of lignin reach fraction, respectively (Ishihara et al., 1991); 3. Separate reaction vessel with UF membrane with shell
immobilized -glucosidases (BG), in (3a) 1-stage, and (3b) 2-stages (Klei et al., 1981). PBR packed bed reactor.
fed-batch or rarely continuously. Lee and Kim (1993) used a cellulose

hydrolysis reactor system which consisted of a water jacketed stirred
tank connected to a cellulose and buffer feed tank for maintenance of a
constant reaction volume. The enzymatic hydrolysis reactor was
equipped with a polysulfone membrane for the ultraltration of the
reaction mixture which was collected in a fraction collector (Fig. 5, 1). In
the work by Kinoshita et al., 1986 air was supplied from the compressor
to the reservoir with the buffer solution which was sent to the
intermittently substrate-fed ultraltration membrane reactor in which
thermotolerant cellulases were retained, while the ltrate was collected
(Fig. 5, 1). Hong et al., 1981 used a system where the reacting volume
inside a membrane reactor for cellobiose hydrolysis to glucose was
replaced by a substrate solution continuously fed from a pressurized
reservoir, while the permeate was automatically collected (Fig. 5, 1).
Alternatively, the reaction may be accomplished in the same vessel as
the separation (in situ) or separately, i.e. in the stirred vessel or unstirred
bed, packed with substrate. In the case of separate units, the reaction
mixture (stirred reactor) or the ltrate (packed bed reactor) have to be
transported to the separation unit and then recycled back.
The cellulose hydrolysis-reducing sugars removal system designed by
Yang et al. (2006) consisted of a tubular reactor, in which the substrate
was retained with a porous lter at the bottom and buffer entered at the
top through a distributor, and the separate hollow-ber module
ultraltration polysulfone membrane, through which the permeate was
transported and removed (Fig. 5, 2a). To keep the volume constant in the
tubular reactor, all remaining buffer was recycled back from the UF
membrane and the make-up buffer was continuously supplied from the
reservoir (Yang et al., 2006). To improve the efciency of the batch and
continuous hydrolysis, Henley et al. (1980) incorporated an UF membrane
stirred cell (UF) or hollow-ber cartridge (HFC) into the CSTR-UF and
CSTR-HFC system, respectively (Henley et al., 1980) (Fig. 5, 2a).
In some of the few reports describing lignocellulose hydrolysis in
membrane reactors, an additional microltration unit has exceptionally been used to retain the unconverted lignin-rich solid fraction due
to the present tightly bound enzymes (Knutsen and Davis, 2004) or
has been employed to remove the unconverted substrate from the
reactor. These set-ups result in slightly complex process layouts for
the hydrolysis (Fig. 5, 2c and 2d).
Ishihara et al. (1991) accomplished a semi-continuous hydrolysis
reaction by using a continuously stirred reservoir tank, connected to a
suction lter unit for the removal of the lignin-rich residue and an
ultraltration membrane unit (tubular module), through which the
ltrate was pumped in order to separate the hydrolysis products from
the ltrate containing cellulases (Figs. 5, 2d). The concentration of the
lignocellullose substrate in the reactor was maintained almost
constant by addition of fresh substrate at appropriate intervals; the
lter and ultraltration units were operated intermittently, while the
enzymes were added at the start, recovered in the UF module and
recycled back into the reactor (Ishihara et al., 1991).
3.3. Enzyme retention
The retention of enzymes in the reaction system, either in situ or in
separate units, is as a rule accomplished by the membrane which is at
the same time permeable for the products. In order to retain the
enzymes in the reacting system, the combination of MF and UF units
may be used to recover the enzymes that are not rmly bound to the
substrate. Knutsen and Davis (2004) did a batch saccharication using
416
shaked asks and simple, intermittent solidliquid large-pore vacuum

ltration and ultraltration after 4, 12, 24 and 96 h, with replenishment of buffer removed as permeate after each separation. The
permeate removal was done to eliminate inhibitory products such as
glucose and cellobiose, while retaining the corn stover and the
cellulose enzymes in the reactor. In a semi-batch hydrolysis, vacuum
was used to remove the hydrolysis products as ltrate while the
soluble enzymes were recovered by ultraltration (both ltrations
after 4 and 8 days) and added together with the fresh pre-treated corn
stover slurry to the residual solids and tightly bound enzymes
(Knutsen and Davis, 2004)(Fig. 5, 2c). The cellulosic material was
continuously added and the soluble hydrolysis products were
continuously withdrawn from the system with immobilized glucosidase on the shell side of the hollow-ber reactor (Klei et al.,
1981) (Fig. 5, 3a). The cellulose slurry from the reservoir was
converted by Klei et al. (1981) to soluble low-molecular weight
mono and oligosaccharides that permeated the shell side of the ber
walls of the hollow-ber cartridge (Romicon), that retained the
enzymes and unreacted cellulose within its lumens to be recycled
back to the reservoir. In a different set-up, the soluble reducing sugars
were transferred to the shell of second hollow-ber cartridge
(Amicon), which contained immobilized -glucosidase (Fig. 5, 3b)
(Klei et al., 1981). In this set-up, the saccharied product could cross
the lumens of the second cartridge and could be led back to the
reservoir (batch) or be withdrawn separately (continuous) (Klei et al.,
1981) (Fig. 5, 3b).
the rate of product removal;

lignocelluloses, cellulases and buffer feeding rates;
cellulases deactivation rate;
concentration polarization and fouling.
As the above may be general for the enzymatic reaction with

product inhibition, with lignocellulose enzymatic conversion there
might be several specic and distinguishing issues such as the
inherently slow reaction rate, relatively fast inhibition rate and
pronounced membrane fouling with sticky cellulosecellulase reaction mixture. Furthermore, as opposite to classical membrane bioreactor systems in which the macromolecular substrates have much
more different sizes than its products, the intermediate product of
cellulose hydrolysis cellobiose has similar molecular dimensions
as the product glucose and is typically withdrawn together with it as
the permeate without being converted to glucose.
The change in the cellulose (S), cellobiose (C) and glucose (G)
concentration, respectively, in the membrane reactor can be represented as follows:

dS
F
dS
= S
dt
dt consumed in reaction 1
VR

dC
dC
dC
= DCout + 1:11
dt
dt produced in reaction 1
dt consumed in reaction 2

dG
dG
= DGout + 1:05
dt
3.4. Specic design features

The specic mechanical design features in connection to the
membranes are usually employed in order to reduce the negative
effect that the hydrolysis reaction mixture has on membrane ux.
Ohlson et al., 1984 investigated batch and semi-continuous hydrolysis
with continuous product removal and replacement of permeate with
buffer, in a stirred polyamide membrane reactor with 2 mm margin
between membrane and propeller to prevent fouling (Fig. 5, 1). A
stirred reactor integrated with a at-sheet polysulfone membrane and
supplied with in-situ electrokinetic membrane cleaning to prevent
continuous accumulation of enzyme molecules and substrate particles
at the surface, was used by Gan et al. (2002) in a batch reaction and in
operation with intermittent and continuous removal of products, with
replenishment of buffer lost in the permeate (Fig. 5, 1). An electrical
backpulse drastically increased the ux level immediately after the
impulse, but the elevated ux could not be sustained (Gan et al.,
2002). A batch hydrolysis, followed by operation in a continuous
mode with simultaneous reducing sugars removal, was performed by
Bla-Bak et al. (2006): They used a tubular membrane module
consisting of a stainless steel tube covered by a non-woven textile
layer, providing a hairy surface for immobilization of cellulose
particles and cellulases by adsorption to reduce membrane fouling
and diffusion resistances, and improve membrane selectivity
(Fig. 5, 2b). In a batch mode, the reaction mixture containing the
substrate and the enzymes was rst circulated (80 mL/min) between
the agitated vessel and the membrane module to allow fractional
adsorption on the membrane surface, while in continuous mode the
permeate outlet was opened and the permeate containing glucose
was collected (Bla-Bak et al., 2006).
4. Key factors inuencing membrane reactor performance for
enzymatic cellulose hydrolysis
The performance of membrane biocatalytic reactors for conversion
of lignocellulose may in general, depend on the following factors:
the rate of product formation;
cellulases product inhibition rate;
Reaction 1 denotes cellulose hydrolysis to cellobiose which is

further converted to glucose in reaction 2, as may be illustrated by the
following model, adopted from Philippidis et al. (1993) and Wyman
(1996), a model which was originally developed for SSF:

dC
dt

=
produced in reaction 1
k1 Ec SS et
1K1L L
1 + C = K1C + G = K1G KE + EC

k2 EG Eg* C
dG
1K2L L
=
KM 1 + G = K2G + C
FS designates the substrate feeding rate (kg s 1), VR is the reaction volume (m3), D is the dilution rate (h 1), Ss is the concentration
of the available cellulose surface (m2 m 3) which is related to the
cellulose concentration S (kg m 3) by the specic area (m2 kg 1), C
is the cellobiose concentration (kg m 3), G is the glucose concentration (kg m 3), EC and EG are the concentration of the cellulases
and -glucosidases in the solution, respectively (kg m 3). The k1
(kg m 2 h 1) and k2 (kg CBU 1 h 1) are specic rates of reactions
1 and 2, respectively, KE is the equilibrium constant for cellulose
adsorption to cellulose (kg m 3), K1L and K2L are constants for the
cellulase and -glucosidase adsorption to lignin (m3 kg 1), respectively, L is the concentration of lignin (kg m 3), K1C, K1G and K2G are
cellobiose (C) and glucose (1G and 2G) inhibition constants for
cellulases and -glucosidases (kg m 3), E*g is -glucosidase specic
activity (CBU kg 1), KM is the MichaelisMenten constant for
cellobiose (kg m 3), and is the specic rate of decrease of effective
cellulose specic surface area (h 1) which denes the quality of the
cellulosic substrate, but can as well in general designate the
exponential decrease of the enzymatic activity with time. The above
models highlight the signicance of the product inhibition constants
for the rate of the cellobiose and glucose formation, respectively, and
are some of the few models, that include the unproductive
adsorption of the enzymes to the lignin.
4.1. Glucose output prole

Depending on the operational mode, the glucose output prole
(Fig. 6, 12) from the membrane bioreactor typically demonstrates a
soft peak in both batch and fed-batch after some time of reaction
usually this glucose output peak appears in the order of hours into the
reaction (Kinoshita et al., 1986; Knutsen and Davis, 2004; Lee and Kim,
1993; Ohlson et al., 1984.; Yang et al., 2006) (Fig. 6, 1ac). In the case of
continuous operation, the glucose output level reaches a plateau after
initial dynamic phases (Alfani et al., 1982; Ghose and Kostick, 1970;
Hong et al., 1981; Klei et al., 1981) (Fig. 6, 2ab). The different cases
from Fig. 6 are explained later in the paper. In several instances, the
permeate glucose concentration is given as the level of total sugars
produced, and this level typically increases steadily during the reaction
(Gan et al., 2002; Ishihara et al., 1991). The glucose concentration
prole depends on the E/S ratio, the ratio of -glucosidases to
cellulases and the kind of pre-treatment employed (Frenneson et al.,
1985). However, the glucose concentration prole rst and foremost
depend on the reaction rate and the ux through the membrane. The
glucose maximum will be achieved when the glucose production rate
is equal to the removal rate (Alfani et al., 1982) (Fig. 6, 1a). The
maximal value will be directly proportional to the reaction rate and
inversely to D (Frenneson et al., 1985., Alfani et al., 1982) (Fig. 6, 3). The
time it takes to reach the glucose peak will be shorter if the residence
time ( = V / Q) increases or the dilution rate (D = Q / V) decreases.
Clearly, higher dilution rates leads to lower glucose output concentrations. As an example, Yang et al., (2006) obtained 220 g/L of product
at the end of hydrolysis (527 g/L of the total permeate, 280440 g per
g of dry biomass) when converting 125185 g/L of substrate, and the
product curves showed typical peak after 24 h, reaching 1535 g/L.
However, the end glucose concentrations were lower than the
corresponding glucose concentration achieved at the end of the
batch hydrolysis (2031 g/L) (Yang et al., 2006).
4.2. The effect of dilution rate
The operation of enzymatic degradation of lignocellulose at high
glucose concentration will cause a decrease in the hydrolytic activity
of the enzyme system due to the product inhibition, and thus, it is
typically seen that the increase in dilution (removal) rate results in an
increased conversion rate up to a certain level. Ohlson et al. (1984)
showed that the reducing sugar concentration prole had a typical
peak after 24 h for all dilutions rate examined (Table 3). They also
showed that an increased dilution rate, in their case up to 1 h 1, led to
an increased conversion rate, owing to the elevated rate of the
continuous removal of inhibitory products, i.e. both sugars and lowmolecular byproducts resulting from the biomass pre-treatment. Yang
et al., (2006) observed an increase in the hydrolysis rate and glucose
417
concentration at the low dilution rates, 0.0570.075 h 1, but obtained

higher conversion rates and lower glucose concentrations at the
higher dilution rates, namely 0.0750.25 h 1. Kinoshita et al. (1986)
also noticed that the conversion rate increased with dilution rate from
0.06 to 0.3 h 1, but also found that the trade-off was a decrease in the
end-product concentration. Obviously, the exact break-even point
for the conversion rate and the dilution rate will depend on the
reaction features, including substrate and enzyme concentrations in
the reactor unit.
4.3. The problem of cellobiose
When both the substrate and product have similar molecular size and
if they both pass the membrane, it is necessary to match the membrane
transport rate with the reaction rate to ensure that as the enzyme
reaches the substrate, it is converted, and the product is transported to
the other side (Giorno and Drioli, 2000). An increase in dilution rates
over a critical point may however result in a decrease in the
lignocellulose hydrolysis conversion rate even when the inhibition is
reduced. This phenomenon is due to the simultaneous removal of
soluble oligosaccharides, mainly cellobiose, i.e. -glucosidase substrates.
In a modeling study with a membrane reactor Lee and Kim (1993)
have demonstrated the importance of retaining the cellobiose at a
certain level in order to increase the glucose production rate,
especially when having a cellulase mixture with low BG activity.
Based on hydrolysis kinetic models proposed by (Lee and Fan, 1983),
Lee and Kim, (1993) also determined the optimal prole of the
dilution rates to maximize the glucose production from a membrane
bioreactor system (Fig. 6, 1c). The bioreactor system consisted of: 1. a
rst batch operation, 2. operation at maximum dilution rate, 3. a
second batch operation, and 4. operation at approximately constant
dilution rate. The optimal prole was dependent on the balance
between increase in glucose formation from the cellobiose and a
reduction due to the cellulase inhibition by cellobiose and glucose
(Lee and Kim, 1993). Mainly, when the latter became higher than
former reactor operation mode was switched from batch to maximum
dilution rate in order to reduce the inhibition and maximal dilution
was maintained until the gain from higher cellobiose concentration
became larger. Thus, the cellulose conversion to glucose was
increased by roughly 7% and the glucose concentration from the
collected permeate was up to 2 times higher when the membrane
reactors were operated with the described optimal prole of dilution
rates, than with the constant dilution rate (Lee and Kim, 1993). When
inspecting Table 3, it is interesting to note, that the product
concentration in many cases has been given as the amount of
reducing sugars and that no -glucosidase was added. This raises the
suspicion that a signicant amount of the product obtained could be
cellobiose.
Fig. 6. Typical glucose concentration prole in the permeate from membrane bioreactor in a: (1a) batch (Lee and Kim, 1993; Ohlson et al, 1984, Yang et al., 2006); (1b) fed-batch
(Knutsen and Davis, 2004; Kinoshita et al., 1986); (1c) batch with optimal dilution rate prole (Lee and Kim, 1993); (2a), (2b) continuous operation mode (Alfani et al., 1982; Ghose
and Kostick, 1970; Hong et al., 1981; Klei et al., 1981); and (3) the typical change of glucose permeate prole with increase in dilution rate in the batch mode (Kinoshita et al., 1986;
Lee and Kim, 1993; Ohlson et al., 1984; Yang et al., 2006). The relative position of the curves is due only to the overview and it is not related to the absolute value of the glucose
concentration.
418
5. Factors affecting the membrane ux
5.4. Concentration polarization
Each of the separate compounds from the hydrolysis reaction

mixture lignocelluloses particles, macromolecular cellulases and
product glucose can cause a drastic reduction of ux through the
membrane compared to the ux of pure water. This ux reduction is
due to the membrane fouling, enzyme concentration polarization and/
or high permeate viscosity. Furthermore, when the cellulosic substrate
and enzymes are mixed together, the ux through the membrane can
be more reduced than with the individual components.
The lignocellulose and enzyme particles present in the substrate

solution can cause strong concentration polarization in the membrane
bioreactor resulting in a notable ux decline.
In one study, the permeate ow rate through a UF membrane (=
the ux) declined as a result of the presence of agitated cellulases
(0.04% (w/v)) or soluble cellulose (0.1% (w/v) ZMC), and this decline
was ascribed to be most likely due to concentration polarization
(Alfani et al., 1983). In contrast, in the same study, the ux did not
decrease when crystalline cellulose (0.3% (w/v) Avicel) was added,
most likely because the added cellulose remained suspended, i.e. it
was not soluble, in the bulk solution (Alfani et al., 1983).
The phenomenon of this concentration (c) build-up of a solute, e.g.
in the case of lignocellulose notably the enzyme(s), the soluble
cellulose, or the soluble products cellobiose and glucose, in the
boundary layer may be represented by the following general nonsteady-state differential mass balance:
5.1. Fouling
With dead-end ltration and crossow microltration, Mores et al.
(2001) found that the water ux ( 14 000 L/(m2 h)) through a
polysulfone 0.2 m (P = 0.7 bar) membrane was appreciably reduced by pre-treated yellow poplar (5% (w/v), 1500 L/(m2 h); 0.2%
(w/v), 1300 L/(m2 h), respectively), cellulases (0.3% (w/v), 300 and
400 L/(m2 h) , respectively) and their mixture (100 and 360 L/(m2 h) ,
respectively). The authors speculated that the cellulaselignocellulose
mixture showed increased fouling due to the stickiness of enzymebounded particles, especially since the ux was dramatically recovered by backushing of the membrane (11 20011 300 L/(m2 h)).
5.2. Reaction slurry properties

The permeation rate during the ultraltration is affected principally by solids concentration of the reaction mixture, but also by
apparent viscosity and density of the slurry system, and the total
soluble sugars (Ghose and Kostick, 1970). Within 20 h of reaction
time, Ohlson et al. (1984) found that the membrane fouling was
responsible for the 30% and 55% initial UF membrane ux decline, for
the washed and unwashed substrate slurry, respectively. Mores et al.,
2001 found a 30% reduction of the permeate ux through the UF
membrane (P = 3.7 bar) in the presence of 8% (w/v) solution of
glucose, owing to its high viscosity. When increasing the glucose
concentration from 5 to 10% (w/v), Ghose and Kostick, 1970 saw
roughly 15% (425360 L/(m2 h)) and 30% (3525 L/(m2 h)) reduction
in the UF membrane ux, for 30 and 10 kDa membranes, respectively.
5.3. Molecular cut-off
The selection of the membrane with respect to molecular cut-off
can have a profound inuence on ux and glucose removal rates
(dilution), the permeation of the cellulases and the response to transmembrane pressure change. In general, the highest UF ux rates and
increased glucose removal rates were seen with the higher cut-off
values, in this case 1030 kDa, 517 L/(m2 h) (Ghose and Kostick,
1970). On the other hand, increasing the cut-off to more than 30
50 kDa, can result in leaking of the cellulases through the membrane
(Kinoshita et al., 1986; Mores et al., 2001). Furthermore, in order to
increase the permeate velocity, it is usually necessary to increase the
trans-membrane pressure and the response will be dependant on the
membrane fouling. However, the permeate ow rate may be
inuenced by the history of the applied pressure, e.g. due to membrane compaction (Alfani et al., 1983). This problem poses a
signicant operational challenge, because a consistent relationship
velocitypressure is then difcult to obtain (Hong et al., 1981).
Kinoshita et al. (1986) have shown that with 0.550 kDa membranes,
the smaller the membrane pore size, the higher is the pressure
required to maintain a suitable ow rate. Table 3 highlights that most
studies were conducted using membrane sizes from 10 to 30 kDa,
and that lower (4 kDa) or higher (50 kDa) cut-off values were only
employed occasionally.
c
2 c
c
= DS 2 Jv
t
x
x
in which Jv is the solvent ux (m3 m 2 h 1) and DS is the solute
diffusion coefcient (m2 h 1).
If the solute molecules are completely retained by the membrane,
the convective ow of the solute molecules towards the membrane
surface will be equal to the diffusive transport back to the bulk phase,
at steady-state conditions. The concentration polarization will be
more pronounced at higher dilution rates, i.e. at higher product
removal rates or at higher permeate ux. Unfortunately, the negative
effect of increased dilution rate is not counterbalanced by the benet
of the corresponding reduction in the concentration of the reaction
product in the reactor (Alfani et al., 1982). An illustration of this was
problem was reported in a modeling study with a xed boundary
polarization layer thickness of 25 m: In this case a doubling of the
dilution rate from 0.8 to 1.6 h 1 decreased the fractional bulk
concentration of enzyme from 0.97 to 0.12 (Hong et al., 1981). Since
the cellulose is insoluble and requires the enzymes to be adsorbed in
order for the reaction to take place, concentration polarization will
decrease the effective concentration of enzyme and correspondingly
the conversion rate.
The enzyme concentration polarization level mainly depends on
solute accumulation due to membrane rejection, which is very strong
for enzymes, and back diffusion, which is slow for enzymes, into the
bulk solution. It can therefore be reduced and partially avoided by
vigorous stirring in the region next to the membrane surface to
decrease the polarization layer or by increasing the trans-membrane
pressure for a higher permeate ux. However, the high local shear rate
can deactivate the enzymes that are rejected and accumulated in a
polarization lm region immediately next to the membrane surface.
This deactivation is accelerated by the exchange of the deactivated
enzyme in the polarization layer where shear from stirring is high
with the enzyme in the bulk solution through the diffusion and
convection (Alfani et al., 1982; Hong et al., 1981). Thus, the conversion
rate is affected both by the decrease of bulk concentration of enzyme
and at the same time by the shear deactivation of the accumulated
enzyme. Due to simultaneous effect of 2 distinct phenomena (product
inhibition and concentration polarization), Alfani et al. (1982) found
that the cellulose conversion rate was increased for small and medium
values of the dilution rate (0.170.33 h 1) and reached a maximum
(0.38 h 1). This increase in the conversion rate for the relatively small
dilution rates was mainly due to the lower product accumulation in
the reactor, at which there is a negligible enzyme inactivation by
accumulation in the laminar sub-layer on the membrane surface that
can be avoided by agitation of the reactor contents. Hong et al. (1981)
determined that, when operating CSTR membrane cell for cellobiose
hydrolysis at a low dilution rate (0.58 h 1) the initially loaded

enzyme participates 100% to the overall conversion and only 25%
when dilution rate is increased to critical, 1.3 h 1, after 10 h.
6. Membrane reactors for glucose removal during (ligno)cellulose
hydrolysis: operational challenges
6.1. The glucose concentration
The overall benet of removing the enzyme inhibiting product
(here glucose) can be heavily diminished if its concentration is not
greatly reduced. It can be seen that, on the contrary to conversion, the
nal outlet concentration of glucose is in general ranges between 0.1
and 75 g/L and typically bellow 35 g/L (Table 3). Knutsen and Davis
(2004) showed that the resulting glucose concentrations (727 g/L)
after corn stover hydrolysis with intermittent removal were lower
than in the control batch experiment (maximal 70 g/L) but did not
result in an increased nal cellulose conversion (60%). This may be
due to the reduced concentration still being too high to signicantly
inuence the reaction rate (see Fig. 3). The authors speculated that the
lower product concentration and thus enzyme inhibition could
possibly be obtained by more frequent ltration or washing. This
would however result in a too dilute product stream. Yang et al.
(2006) presented a several-fold increase in the hydrolysis rate when
increasing the substrate concentration, but each of the individual
hydrolysis rates declined drastically during the rst several hours.
This was suggested to be a result of a gradual increase in the nonproductively adsorbed enzymes, transformation of cellulose into a less
digestible form and incomplete elimination of the product inhibition.
In another study, by Gan et al. (2002), the reaction rate improvement
obtained immediately after product removal was equivalent to an
increase in the conversion rate of 410 times, but the average reaction
rate after 20 h was not signicantly affected. This led the authors to
conclude that relief of the product inhibition through the continuous
product removal is limited due to the inherently slow cellulose
hydrolysis kinetics (Gan et al., 2002). We disagree, with this
conclusion, and would rather suggest that the overall result may be
a consequence of insufcient or delayed product removal (re
Fig. 3). A relatively small concentration of glucose strongly inhibited
the activity of -glucosidase such that a two-fold increase in glucose
concentration, from 0.27 to 0.66 g/L, caused a specic cellobiose
hydrolysis rate reduction of almost 40% (Alfani et al., 1983). Therefore,
glucose concentration in the reaction medium has to be kept as low as
possible to limit the extent of inhibition. This can be achieved only in a
reactor working at relatively high dilution rates and encompassing
continuous removal of glucose. This however, yields low glucose
concentration in the permeate and requires drastic concentration (up
to 100 g/L) in order to perform the subsequent fermentation to
ethanol economically (Alfani et al., 1983).
6.2. Fed-batch operation
It appears that the fed-batch feeding of the cellulosic substrate can
have a positive effect on the conversion rate and especially, on the
specic enzyme efciency or consumption allowing more substrate to
be converted with the same loading of enzymes, while at the same
time the viscosity of the reaction slurry may be reduced. Ishihara et al.
(1991) reduced the specic enzyme consumption from 32 FPase IU
per g reducing sugars, produced by fed-batch substrateenzyme
hydrolysis of steamed hardwood and hardwood kraft pulp to 27 and
7.4 FPU per gram of reducing sugar produced for steamed hardwood
and hardwood kraft pulp, respectively. The fed-batch feeding was
accomplished by a semi-continuous feeding of substrate with
intermittent product removal. When Ohlson et al. (1984) ran the
membrane reactor semi-continuously for 20 h, the fresh substrate
was intermittently fed while the dry matter concentration was kept
419
constant based on estimation of products in the permeate. This

operation, which was repeated each day for 3 days with the same
batch of enzymes, gave a nal conversion of 80% (94%, 81% and 60%
after each day), but also very high cellulase efciency of roughly 25 g
reducing sugars per g of enzyme used compared to classical batch
hydrolysis, which by a rough estimate gives 5 g reducing sugars/g
enzyme (Ohlson et al., 1984). In the work of Knutsen and Davis
(2004), the hydrolysis was promoted by the fed-batch addition of
fresh substrate and an increase in the hydrolysis rate was seen after
each ltration and substrate addition (Fig. 6, 1b). This effect was
explained by presence of fresh amorphous cellulose, rather than by
reduced inhibition due to the removal or increased substrate
concentration. Furthermore, the semi-batch hydrolysis with thorough
washing of material with water or extra addition of enzymes along
with fresh substrate to replace deactivated enzyme, after each
ltration step, resulted in a marked increase in conversion (Knutsen
and Davis, 2004). On the contrary, Kinoshita et al. (1986) found that
the concentration in the membrane bioreactor gradually decreased
during the reaction although with transient increase due to the fedbatch feeding of substrate, following the rst-order decay kinetics.
Finally, unless very dry, pre-treated substrate is added, a negative
dilution effect which lowers the actual glucose concentration, will
result because of the simultaneous addition of water with the pretreated substrate. This phenomenon has been observed in practice in
fed-batch reactions with pre-treated straw (Rosgaard et al., 2007a).
6.3. Continuous operation
The continuous mode of operation assumes continuous feeding of
reactants and continuous removal of the products (glucose) and
unconverted biomass. In the related literature, the term continuous
hydrolysis is slightly ambiguous, since, sometimes, it really refers to
continuous product removal, and not to the overall mode of reactor's
operation. By operating around 13 h in the continuous mode, BlaBak et al. (2006) found that conversion was practically maintained at
the same level as achieved after 9 h batch hydrolysis, but the authors
did not state explicitly whether the continuous mode assumed
continuous feeding of the fresh substrate and buffer into the reservoir
vessel.
The reported dry matter levels of cellulosic material in exploratory
enzyme-catalyzed lignocellulose hydrolysis reactions have generally
been b10%, and has only been higher, e.g. 1519% in exceptional cases
(Table 3). These levels are considerably lower than what is required
for a glucose-rich hydrolysate for subsequent fermentation to produce
bioethanol (Andri et al., 2010b). The advantages of continuous
operation of biocatalytic reactors for hydrolysis of (ligno)cellulose has
been advocated already 40 years ago (Ghose, 1969), but has still only
been little studied, and only exceptionally with substrate levels higher
than 2.55 DM% (Table 3).
The inherently slow reaction of the enzyme-catalyzed cellulolytic
hydrolysis requires relatively large volumes for the given conversion
or in practice low ow rates of the reaction mixture. Another main
problem related to continuous operation at high DM% is the inefcient
mixing of the viscous slurry. Other problems include the fouling of the
membranes and the difculty in discharging the unconverted residue
that accumulates in the system. Ghose and Kostick (1970) converted a
membrane cell into a continuous reactor which maintained steady
hydrolysis of 10% (w/w) cellulose and simultaneous removal of
products at a high conversion level, namely 77%, mainly because of the
rapid transport of the reaction products through the membrane lm
(0.074 g/min). This resulted in a relatively high average glucose content, 7.5%, in the efuent.
The membrane reactors in the experimental set-ups reported until
now (Table 3) have chiey been in mL scale, from 50 to 300 mL, and
have rarely included larger reactors of e.g. 5.510 L. Thus, to date
there is a surprising scarcity of reports on continuous operation of
420
membrane reactors for lignocellulose hydrolysis. Considering the

signicant importance of operational features to overcome the
possible restrictions outlined above, the design of large scale membrane reactors for lignocellulose hydrolysis still deserves signicant
research efforts.
6.4. Enzyme activity retainment
Maintenance of a high level of cellulase activity is essential for the
efcient operation of the membrane reactor, due to the possible reuse of the enzymes in several successive reaction cycles. Although
within the membrane reactors with product removal the enzyme
activity loss as a result of the product inhibition is signicantly
reduced, the other well known reasons for the enzyme deactivation,
i.e. mainly unproductive adsorption of enzymes to lignocellulose and
heat inactivation in these reactors also can arise. Table 4 gives an
overview of the activity studies performed in relation to the operation

of the biocatalytic membrane reactors. In general, it appears that
cellulolytic enzymes can relatively long maintain a reasonably high
level of activity. The presence of substrate generally stabilizes the
activities of CBH and EG more than the activity of -glucosidases,
which act in the aqueous phase. The (ir)reversible adsorption of
enzymes to the non-convertible fraction seems to be one of the major
causes of the activity reduction, especially if the substrate in question
is lignocellulose, except in (Knutsen and Davis, 2004), where the
tightly bound enzymes were actually capable of efciently hydrolyzing the freshly added ligocellulosic material. The shear deactivation
may also, albeit to a lower extent, be responsible for activity decay due
to the passage through the hollow bers (Klei et al., 1981) or tubes
(Ishihara et al., 1991; Yang et al., 2006) or in the stirred cells because
of localization of enzymes in the thin layer due to the concentration
polarization, especially in the case -glucosidases (Hong et al., 1981).
Table 4
Overview of enzyme deactivation studies in membrane reactors. CBH is cellobiohydrolase; EG is endoglucanase; BG is -glucosidase.
Activity Activity loss Period Substrate Comment

type
[%]
[day]
presence
Source
CBH1
EG4
BG
5667
4450
0.2
No
None of the enzyme found in efuent; maybe inadequacies in enzyme assay procedure; some
other inactivation
Ghose and Kostick

(1970)
CBH
EG
BG
035
1060
No
Thermal deactivation negligible; shearstress and concentration polarization related deactivation

in membrane reactor
Hong et al. (1981)
CBH
EG
BG
Yes10
29
Deactivation due to passage through hollow bers or adsorption onto substrate; soluble BG easily
Klei et al. (1981)
shear-deactivated; 40% of BG activity lost on immobilization, but stability better than in the soluble form
CBH
EG
BG
10
10
No
Low thermal deactivation; remarkable mechanical stability at 500 rpm
CBH
EG
BG
906,7
606,7
Yes10
Reversible/non-reversible adsorption and denaturation
CBH3
EG2
BG
20508
5158
45808
Yes
Thermotolerant; at 30 C negligible deactivation; inactivation in the reactor at temperatures 3060 C

follows 1. order decay the activity of proteases from enzyme preparation probable reason
CBH1
EG2
BG
85
75
65
15
No
Signicant loss after 67 days; proteolytic modication T, pH; in reactor irreversible adsorption on
insoluble residue and UF tubular unit physical deactivation
CBH1
EG2
BG5
55
10
30
No
Thermal deactivation
Lee and Kim (1993)
CBH
EG
BG
Yes10
Shear deactivation seems insignicant
Gan et al. (2002)
Small9
CBH1
EG
BG
50
30
Yes10
Adsorption to non-convertible biomass fraction; shearstress deactivation negligible in tubular

reactor; no thermal deactivation in 1 day without substrate
Yang et al. (2006)
CBH
EG
BG
0.9
Yes10
Slow permeation of enzymes though the membrane
N/A
Bla-Bak et al.
(2006)
Filter paper activity.

Activity on CMC.
3
Activity on KC Floc.
4
Cx activity.
5
Activity on pNPG.
6
Soluble activity.
7
Semi-continous operation.
8
5060 C.
9
90 rpm.
10
In the reactor.
2
421
The thermal inactivation does not emerge as the particularly

important factor (Hong et al., 1981; Ohlson et al.; 1984, Yang et al.,
2006), although quite a few studies were performed at temperatures
lower than 50 C (Alfani et al., 1982; Gan et al., 2002; Ishihara et al.,
1991; Kinoshita et al., 1986; Knutsen and Davis, 2004; Ohlson et al.,
1984), including one case, as explicitly stated, with thermotolerant
enzymes (Kinoshita et al., 1986).
7. Membrane reactors and hydrolysis kinetic studies
Apart from representing a promising concept of potential realization in real industrial processes, it appears that integrated reactors
with product removal such as membrane bioreactors can be valuable
tools for understanding mechanisms and performing various kinetic
studies of enzymatic reactions, including lignocellulose enzymatic
hydrolysis. Specically, membrane reactors can be used for obtaining
kinetic information on reaction order towards substrates, the rate of
enzyme thermal inactivation and the extent of product inhibition
(Alfani et al., 1983; Frenneson et al., 1985).
7.1. General kinetic studies using membrane reactors
With respect to thermal inactivation of cellulases and hydrolysis
reaction orders, the advantage lies in eliminating the strong product
inhibitory effect which can disguise the true results under the studied
conditions. Further, it is known that the enzyme stability is increased
by the substrate presumably because the cellulases remain bound
to the residual solids (Knutsen and Davis, 2004) especially when
the substrate is present in high levels. The batch reactors seem
inappropriate for this purpose because S changes with the course of
the reaction making it necessary to extrapolate the measured activity
at zero time (leading to errors). In membrane reactors it is possible to
perform studies on cellulase activity due to the substrate feeding
option and the selectivity of the membrane. These options allow for
keeping the substrate and enzyme concentrations constant at the
same time and thus determination of enzyme activity or stability at
prolonged times and different temperatures (Alfani et al., 1983). As
equally important is that the activity/stability studies can be performed at operationally encountered substrate levels and not just at
typical assay conditions which are the most likely performed without
the substrate presence. Although not kinetically modeled, comparisons of special membrane reactor designs and membrane constructions have also allowed the experimental assessment of the
consequences of e.g. simultaneous immobilization of the enzymes
and the substrate onto the membrane surface with continuous
product removal (Bla-Bak et al., 2006).
In order to obtain more information on cellulose hydrolysis kinetics
during continuous product removal Alfani et al. (1982) used an
ultraltration cell for removal of glucose, continuously fed with buffer
solution, for studies on microcrystalline cellulose. To maintain a
constant substrate concentration in the reactor, the conversion was
kept at a low level using very low substrate and enzyme concentrations. With longer reaction times a negative derivative in the glucose
kinetic curve could be due to the enzyme deactivation and/or substrate
consumption by cellulose hydrolysis.
7.2. Product inhibition studies
In the case of product inhibition, the reaction rate and inhibitor
concentration are related. Experiments with constant product inhibitor
concentration(s) can obviously not be accomplished in batch reactors. In
membrane reactors, inhibitor levels can be selected and the extents of
product inhibition can be tested at a wide range of concentrations
and reaction times by varying the dilution rate and enzyme concentration all at a constant substrate concentration. Theoretically,
even a completely uninhibited reaction could be accomplished if the
Fig. 7. Inuence of added glucose on glucose formation rate in a study where glucose is
externally added (040 g/L) prior to reaction (2%DM); the formation rates are
calculated from the non-competitive inhibition model (Andri et al., 2010a). Figure
legend:
0 g/L
10 g/L
20 g/L
40 g/L.
product removal was initiated immediately and if it was then run

continuously and securing a complete product removal.
Because the reaction mixtures of enzymatic lignocellulose hydrolysis experiments are viscous and heterogeneous, most of the
inhibition studies have assessed the inhibition in batch reactions
with addition of the inhibitor (e.g. glucose) at concentrations ranging
from high to relatively low substrate levels, i.e. low S/G and E/G ratios
(Figs. 7 and 8). During genuine operations, however, without added
glucose, the S/G ratio (as well as the E/G ratio) will be quite high for
most of the reaction time, due to lower level of present glucose.
Hence, the initial addition of glucose in inhibition studies in effect
exaggerates the product inhibition (Fig. 7).
Inhibition can be studied in membrane reactors in the operational
range of S/G and E/G ratios by varying the dilution rate. The removal
rather than addition of inhibitor is very helpful with respect to avoiding
the unwanted reversible reaction (transglycosylation) which occurs at
low E/G ratios and which may give false information on the real
inhibitory inuence of the studied products. As the enzymes are in
practice dosed per weight of substrate (cellulose or dry matter), low E/G
will mean usually low S/G ratios. We have recently observed the
transglycosylation effect at S/G b 0.1% (w/v)/(g/L) and E/G b 0.25 g/g
(Andri et al., 2010b). Many of the studies on inhibition of cellulolytic
enzymes and especially on -glucosidases have been performed at
very low S/G ratios (and related E/G) (Fig. 8) which in turn might have
induced transglycosylation effect. The attention should furthermore be
paid to reactions with genuine lignocellulosic substrates, as the S/G in
reality may be diminished due to the presence of a non-convertible
substrate fraction (Fig. 8).
8. Other techniques for glucose removal
Although the ultraltration has extensively been used as the most
convenient glucose removal technique, a few other techniques must
be mentioned: (a) two-phase systems; (b) addition of another
enzyme (glucose-oxidase) which oxidizes glucose to gluconic acid;
(c) simultaneous saccharication and fermentation (SSF), where the
glucose is instantly fermented to ethanol; or (d) glucose crystallization. Among these, the SSF and aqueous two-phase system, reported
by Tjerneld et al. (1985), are the most commonly encountered.
8.1. Two-phase systems
This technique is based on partitioning of the enzyme and substrate
to the bottom phase while the product can be extracted in the top phase.
The enzymes can then be recycled and process run in a semi-continuous
mode. The phases with extractants are biocompatible in order not to
denature or inhibit the enzymes. The reduction of the specic enzyme
422
Fig. 8. Summary of inhibition studies from Andri et al., 2010a. The dashed lines designate the beginning of transglycosylsation effect observed in Andri et al., 2010a. The marked
studies have employed lignocellulosic substrates the dashed-dotted arrows point at real cellulosic substrate-to-added glucose ratio. Frenneson et al., 1985 have used a membrane
reactor. Figure legend:
Cellulases
Cellulases + -glucosidases
-glucosidases.
consumption is reported (17 FPU per gram reducing sugar) in these

systems as compared to the batch hydrolysis (Tjerneld et al., 1985).
8.2. Dialysis
The dialysis separation technique is based on different diffusion
coefcients of glucose and cellulases, so the membrane is permeable
only for the low-molecular weight compound while the larger ones
(enzymes and substrate) are retained within the membrane. Only the
glucose thus diffuses through the thin dialysis membrane (Fig. 9) and
there is no need for the liquid (buffer) replacement which lowers the
glucose concentration, as seen with the MF or UF membranes. However,
the glucose must diffuse into the surrounding dialysate buffer whose
large volume ensures sufcient driving force for the separation and
unfavorably lowers the nal glucose concentration (Fig. 10). In
addition, due to the differences in the chemical potential across the
dialysis membrane, caused by the presence of the biomass substrate and
osmotic pressure, the dialysate buffer penetrated into the dialysis
membrane reactor, lowering the reactants and products concentration
inside the reaction mixture (Fig. 11).
We have recently used dialysis to remove the glucose from the
reaction mixture consisting of hydrothermally pre-treated wheat
straw and commercial enzyme mixture (Andri et al., 2010a). When
glucose was removed by in-situ dialysis during the enzymatic
hydrolysis, the rate of enzyme-catalyzed glucose release during 48
Fig. 9. The principle of hydrolysis of lignocelluloses with in-situ removal of glucose by

dialysis.
72 h of reaction recovered from 20 to 40% to become 70% of the rate

recorded during 624 h of reaction (Table 5). This indicated the
importance of the alleviation of the product inhibition, and the
potential workability of in-situ dialysis as a principle for product
removal during enzymatic lignocellullose hydrolysis.
Apart from the obvious problems, i.e. mechanical stability, mixing
in the membrane, employment of high DM%, the future dialytic
process for removal of glucose from lignocelluloses mixture with
cellulase, will need to address: (a) prevention of the dialysate in-ux
to the membrane (for example by applying pressure from the internal
side); (b) reduction of the amount of dialysate buffer needed for the
separation, in order to obtain higher glucose concentrations, without
drastically reducing the driving force and thus membrane ux, and/or
(c) energy efcient post-concentration step.
8.3. Simultaneous hydrolysis and fermentation (SSF) and removal of
ethanol
The concept of in-situ removal of ethanol during the SSF reaction
may encompass simultaneous removal of cellobiose, glucose and
Fig. 10. Dialysis of model solutions. The performance of membrane with 10 g/L glucose
solution. The glucose is collected in a buffer reservoir surrounding the membrane; the
concentration is measured and used to calculate the glucose concentration in a dialysis
membrane, which was furthermore measured at the start and end of dialysis. Starting
volume was 25 mL. The green and dark blue solid lines are the ts of the calculated
glucose concentration in dialysis membrane and in the reservoir, respectively. Figure
legend:
glucose in dialysis membrane_exp
glucose in dialysis membrane_calc
glucose in reservoir_exp.

Table 5
The effect of glucose removal by in-situ dialysis on glucose yield (g/g) from a reaction
mixture with pre-treated wheat straw during treatment with a commercial cellulase
mixture (Andri et al., 2010a). The dialysis was applied from 24 to 48 h.
Control hydrolysis
In-situ glucose removal
24 h
48 h
72 h
96 h
0.54
0.54
0.65
0.71
0.70
0.93
0.76
0.94
ethanol, all of which are inhibitory for the enzymes. Therefore, this
concept from the theoretical point of view represents the most
desired processing option, as minimizes all inhibitory effects on the
rate of the enzyme- and yeast-cell-catalyzed reactions. The concept is
further based on producing fermentations broths with concentrated
ethanol that is less energy demanding to further purify by distillation
to the product grade than the ordinary distillation, which is in turn
typically done after the fermentation of the hydrolyzate (Cardona and
Sanchez, 2007). Furthermore, the ethanol may increase volatility
when found in the fermentation broth containing also the enzymes
(Roychoudhury et al., 1986). One attractive option within this concept
is to evaporate the waterethanol mixture from the SSF vessel by
rapid vacuum shortly applied in cycles while the substrate can be
semi-continuously fed (Roychoudhury et al., 1992). Another interesting solution is to remove the ethanol by the membranes; in a SSF
process coupled with the pervaporation membranes or the membrane
distillation modules (Cardona and Sanchez, 2007). It is our impression
that this potentially attractive subject has not received a major
attention in the literature. In general, the search for the feasible
integrated reactionseparation technology for the lignocellulose
degradation is currently ongoing and the summary of these concepts
e.g. vacuum fermentation, fermentation with gas stripping, fermentation coupled with pervaporation, extractive fermentation, etc, is
presented in Cardona and Sanchez (2007).
The SSF coupled with ethanol removal is conceptually advantageous conguration to SHF due to the reduced cellobiose, glucose and
ethanol inhibition, but on the other hand, represents a heavy
compromise of two optimal reaction conditions (pH and temperature) and includes another (vulnerable) biocatalyst i.e. fermenting
microorganism like yeast cells, in the reaction vessel containing highly
heterogeneous and viscous mixture (lignocellulose, cellulases, yeasts,
nutrients, buffers, sugars, etc.). Furthermore, if it is desired to
423
manufacture some other products from glucose hydrolyzate such as

biochemicals, biobutanol or all together in a biorenery, the SSF
concept may become inappropriate. We highly acknowledge the SSF
and connected separation technologies, but give an advantage to the
SHF concept due to the larger variety of products that is possible to
generate from glucose and uncompromised optimal conditions, as
well as due to the simpler reactor operation and interpretation of the
experimental data.
9. Membrane bioreactor design for glucose removal: conclusions
and recommendations
9.1. Summary
A number of reactor improvements for achieving optimal membrane reactor performance for lignocellulose product removal reactions have been suggested in the literature. For the production of
concentrated glucose syrups from lignocellulose, Ghose and Kostick
(1970) already advised the use of a relatively large reactor volume
and a comparatively small separation system to attain a high initial
saccharication rate with a dense reaction slurry. This approach is of
course only justiable if it is possible to remove the products as fast as
they are formed, so that the product removal results in a glucose
concentration below at least 10 g/L (Fig. 3). This will also allow re-use
of the enzymes. A model system was proposed with an STR for nely
milled cellulose with a concentrated cellulase preparation. The inputs
consisted of substrate and small amounts of dilute cellulases combined with a thin-channel membrane separation cell (Ghose and
Kostick, 1970). Essentially, the idea is to have a high initial rate to
build-up glucose fast, since the controlling factor of the model to
operate successfully is the initial rate of the product formation and not
the ux of product removal.
For the prevention of concentration polarization and the deactivation by shear forces, Hong et al. (1981) proposed operation of
membrane reactors for enzymatic cellulose hydrolysis to be done at
dilution rates lower than critical. Similarly, Kinoshita et al. (1986)
recommended adjustment of the ow rate from high initial level to
lower, in accordance with the decrease in the activity in reactor, in
order to obtain a constant concentration of products.
For partial recovery of enzymes Knutsen and Davis (2004)
reasoned that use of ultraltration appeared unnecessary suggesting
that sufcient (even -glucosidase) activity was adsorbed onto
lignocellulosic particles to convert the fresh added material, and
concluding that the use of more than 10 FPU/gcellulose is unnecessary if
long-term conversion is the goal. Finally, Bla-Bak et al. (2006)
have suggested to immobilize -glucosidases in the lumen (permeate) side of the module for the improvement of the conversion in the
membrane reactor, with respect to expected considerable amount of
cellobiose in the product solution.
9.2. Advantages and challenges of membrane reactors
Contemplation of the available data lead us to recommend the
following: although the operational feasibility of membrane reactors is
currently questionable at high solids loadings, membrane reactors
principally seem advantageous for large-scale enzyme-catalyzed
production of hydrolysates for bioethanol manufacture or for platform
biochemicals based on lignocellulosic material.
Advantages:
Fig. 11. Dialysis of model solutions, Buffer back-ux into the membrane with 2% DM
real lignocellulosic substrate (pre-treated wheat straw). Starting volume was 25 mL.
The solid line is the t of the measured amount of buffer which penetrated into the
dialysis membrane. Figure legend: weight of buffer penetrated [g] buffer back-ux
[g/(min cm2)].
Membrane reactors can provide for increased cellulolytic hydrolysis

reaction rates because glucose product removal, and thus reduced
inhibition, can be accomplished.
Membrane reactors allow the continuous re-use of enzymes.
The possibility of cellulases re-utilization paves the way for higher
enzyme usage efciency (kgproduct/kgenzyme) and, thus, allows
424
(economically) viable employment of enzyme dosages that are

higher than in the conventional batch or continuous reactors
(enzyme single-usage) which in turn further increases the conversion rates.
Fed-batch feeding to secure low viscosity and high reaction rates
may in principle be accomplished without altering the reaction
volume as long as the dilution rate for product removal is adjusted
accordingly, while maintaining a high cellulase partitioning tendency (insoluble vs. liquid phase) as the reaction is progressing.
Membrane reactors with glucose product removal enable simultaneous removal of other inhibitory lower-molecular compounds, e.g.
resulting from substrate pre-treatment that can eventually reduce
the biocatalyst performance.
However, a number of important challenges must be recognized to
accomplish successful operation.
These challenges are:
The outlet (permeate) glucose concentrations is generally too low.
To fully alleviate product inhibition by glucose product removal, the
glucose levels in the reactor and hence in the permeate, must be
very low, typically b10 g/L to have an effect.
Membrane fouling is pronounced a high substrate (dry matter)
levels, and currently only low to medium dry matter levels are
possible to process.
Discharge of the non-convertible fraction which accumulates during
extended reaction, especially with fed-batch operation is difcult.
Removal of the lignin fraction prior to hydrolysis might be helpful.
Scale-up of the membranes as well as the membrane reactors.
The nal selection of the appropriate system, i.e. integrated/
separate units; conventional reactors without removal, etc., depends
on the capital and operational cost of the ltration units, cost of slurry
transport, and biocatalyst and raw material and the market price of
the nal product. However, it seems certain that a serious focus on
reactor design for biocatalytic conversion of cellulose to glucose is
required for successful industrial realization of cellulosic ethanol.
References
Ahmed F, Stein A, Lye GJ. In-situ product removal to enhance the yield of biocatalytic
reactions with competing equilibria: -glucosidase catalysed synthesis of
disaccharides. J Chem Technol Biotechnol 2001;76:9717.
Alfani F, Albanesi D, Cantarella M, Scardi V, Vetromile A. Kinetics of enzymatic
saccharication of cellulose in a at-membrane reactor. Biomass 1982;2:24553.
Alfani F, Cantarella M, Scardi V. Use of a membrane reactor for studying enzymatic
hydrolysis of cellulose. J Membr Sci 1983;16:40716.
Andri P, Meyer AS, Jensen PA, Dam-Johansen K. Effect and modeling of glucose
inhibition and in situ glucose removal during enzymatic hydrolysis of pretreated
wheat straw. Appl Biochem Biotechnol 2010a;160:28097.
Andri P, Jensen PA, Meyer AS, Dam-Johansen K. Reactor design for minimizing product
inhibition during enzymatic lignocellulose hydrolysis: I. Signicance and mechanism of glucose inhibition on cellulolytic enzymes; 2010b. doi:10.1016/j.
biotechadv.2010.01.003.
Bla-Bak K, Koutinas A, Nemestthy N, Gubicza L, Webb C. Continuous enzymatic
cellulose hydrolysis in a tubular membrane bioreactor. Enzyme Microb Technol
2006;38:15561.
Bezerra RMF, Dias AA. Enzymatic kinetic of cellulose hydrolysis. Inhibition by ethanol
and cellobiose. Appl Biochem Biotechol 2005;126:4959.
Cabrera-Padilla RY, Pinto GA, Giordano RLC, Giordano RC. A new conception of
enzymatic membrane reactor for the production of whey hydrolysates with low
contents of phenylalanine. Process Biochem 2009;44:26976.
Cantarel BL, Coutinho PM, Rancurel C, Bernard T, Lombard V, Henrissat B. The
carbohydrate-active enzymes database (CAZy): an expert resource for glycogenomics. Nucleic Acids Res 2009;37:D2338.
Cardona CA, Sanchez OJ. Fuel ethanol production: process design trends and integration
opportunities. Bioresour Technol 2007;98:241557.
Cauwenberg V, Vergossen P, Stankiewicz A, Kierkels H. Integration of reaction and
separation in manufacturing of pharmaceuticals: membrane-mediated production
of S-Ibuprofen. Chem Eng Sci 1999;54:14737.
Chekhova E, Barton P-I, Gorak A. Optimal operation processes of discrete-continuous
biochemical processes. Comput Chem Eng 2000;24:116773.
Decker CH, Visser J, Schreier P. -Glucosidase multiplicity from Aspergillus tubingensis
CBS 643.92: purication and characterization of four -glucosidases and their
differentiation with respect to substrate specicity, glucose inhibitin and acid

tolerance. Appl Microbiol Biotechnol 2001;55:15763.
Drioli E. Membrane reactors. Chem Eng Process 2004;43:11012.
Frenneson I, Trgrdh G, Hahn-Hgerdal B. An ultraltration membrane reactor for
obtaining experimental reaction rates at dened concentrations of inhibiting
sugars during enzymatic saccharication of alkali-pretreated sallow: formulation
of a simple empirical rate equation. Biotechnol Bioeng 1985;27:132834.
Frieden E, Walter C. Prevalence and signicance of the product inhibition of enzymes.
Nature 1963;198:8347.
Fullbrook PD. Practical limits and prospects (kinetics). In: Godfrey T, West S, editors.
Industrial Enzymology 2nd Ed. London UK: MacMillan Press Ltd; 1996. p. 50540.
Gan Q, Allen SJ, Taylor G. Design and operation of an integrated membrane reactor for
enzymatic cellulose hydrolysis. Biochem Eng J 2002;12:2239.
Gan Q, Allen SJ, Taylor G. Kinetic dynamics in heterogeneous enzymatic hydrolysis of
cellulose: an overview, an experimental study and mathematical modeling. Process
Biochem 2003;38:100318.
Ghose TK. Continuous enzymatic saccharication of cellulose with culture ltrates of
Trichoderma viride QM 6a. Biotechnol Bioeng 1969;11:23961.
Ghose TK, Kostick JA. A model for continuous enzymatic saccharication of cellulose
with simultaneous removal of glucose syrup. Biotechnol Bioeng 1970;12:92146.
Giorno L, Drioli E. Biocatalytic membrane reactors: applications and perspectives.
Trends Biotechnol 2000;18:33949.
Gusakov AV, Sinitsyn AP, Klyosov AA. Factors affecting the enzymatic hydrolysis of
cellulose in batch and continuous reactors: computer simulation and experiment.
Biotechnol Bioeng 1987;40:66371.
Hahn-Hgerdal B, Andersson E, Lpez-Leiva M, Mattiasson B. Membrane biotechnology,
co-immobilization, and aqueous two-phase systems: alternatives in bioconversion
of cellulose. Biotechnol Bioeng Symp 1981;11:65161.
Hahn-Hgerdal B, Galbe M, Gorwa-Grauslund MF, Liden G, Zacchi G. Bio-ethanol
the fuel of tomorrow from the residues of today. Trends Biotechnol 2006;24:
54956.
Henley RG, Yang RYK, Greeneld PF. Enzymatic saccharication of cellulose in
membrane reactors. Enzyme Microb Technol 1980;2:2068.
Hong J, Tsao GT, Wankat PC. Membrane reactor for enzymatic hydrolysis of cellobiose.
Ishihara M, Uemura S, Hayashi N, Shimizu K. Semicontinuous enzymatic hydrolysis of
lignocelluloses. Biotechnol Bioeng 1991;37:94854.
Kinoshita S, Chua JW, Kato N, Yoshida T, Taguchi H. Hydrolysis of cellulose by cellulases
of Sporotrichum cellulophilum in an ultralter membrane reactor. Enzyme Microb
Technol 1986;8:6915.
Klei HE, Sundstrom DW, Coughlin RW, Ziolkowski K. Hollow-ber enzyme reactors in
cellulose hydrolysis. Biotechnology and Bioengineering Symposium 1981;11:
593601.
Knutsen JS, Davis RH. Cellulase retention and sugar removal by membrane
ultraltration during lignocellulosic biomass hydrolysis. Appl Biochem Biotechnol
2004;113116:58599.
Kuroiwa T, Izuta H, Nabetani H, Nakajima M, Sato S, Mukataka S, et al. Selective and
stable production of physiologically active chitosan oligosaccharides using an
enzymatic membrane bioreactor. Process Biochem 2009;44:2837.
Lee Y-H, Fan LT. Kinetic studies of enzymatic hydrolysis of insoluble cellulose: (II).
Analysis of extended hydrolysis times. Biotechnol Bioeng 1983;25:93966.
Lee S-G, Kim H-S. Optimal operating policy of the ultraltration membrane bioreactor
for enzymatic hydrolysis of cellulose. Biotechnol Bioeng 1993;42:73746.
Levenspiel O. The chemical reactor omnibook. Cortvallis, Oregon: OSU Book Stores, Inc.;
1993.
Levenspiel O. Chemical reaction engineering. Third edition. John Wiley & Sons Inc;
1999.
Lynd LR, Laser MS, Bransby D, Dale BE, Davison B, Hamilton R, et al. How biotech can
transform biofuels. Nat Biotechnol 2008;26:16972.
Meng F, Chae S-R, Drews A, Kraume M, Shin H-S, Yang F. Recent advances in membrane
bioreactors (MBR): membrane fouling and membrane material. Water Res
2009;43:1489512.
Merino ST, Cherry J. Progress and challenges in enzyme development for biomass
utilization. Adv Biochem Engin/Biotechnol 2007;108:95-120.
Mores WD, Knutsen JS, Davis RH. Cellulase recovery via membrane ltration. Appl
Biochem Biotechnol 2001;9193:297309.
Ohlson I, Trgrdh G, Hahn-Hgerdal B. Enzymatic hydrolysis of sodium-hydroxidepretreated sallow in an ultraltration membrane reactor. Biotechnol Bioeng
1984;26:64753.
Philippidis GP, Smith TK, Wyman CE. Study of the enzymatic hydrolysis for production
of fuel ethanol by the simultaneous saccharication and fermentation process.
Pinelo M, Jonsson G, Meyer AS. Review: Membrane technology for purication of
enzymatically produced oligosaccharides: molecular features affecting performance. Sep Pur 2009;70:1-11.
Riebel BR, Bommarius AS. Biocatalysis, enzyme reaction engineering. Weinheim:
Wiley-VCH; 2004. p. 91-134.
Rosgaard L, Pedersen S, Cherry JR, Harris P, Meyer AS. Efciency of new fungal cellulase
systems in boosting enzymatic degradation of barley straw lignocellulose.
Biotechnol Progr 2006;22:4938.
Rosgaard L, Andri P, Dam-Johansen K, Pedersen S, Meyer AS. Effects of substrate
loading on enzymatic hydrolysis and viscosity of pretreated barley straw. Appl
Biochem Biotechnol 2007a;143:2740.
Rosgaard L, Pedersen S, Langston J, Akerhielm D, Cherry JR, Meyer AS. Evaluation of
minimal Trichoderma reesei cellulase mixtures on differently pretreated barley
straw substrate. Biotechnol Progr 2007b;23:12706.

Roychoudhury PK, Ghose TK, Ghosh P, Chotani GK. Vapor liquid equilibrium behaviour
of aqueous ethanol solution during vacuum coupled simultaneous saccharication
and fermentation. Biotechnol Bioeng 1986;23.
Roychoudhury PK, Ghose TK, Ghosh P. Operational strategies in vacuum-coupled SSF
for conversion of lignocellulose to ethanol. Enzyme Microb Technol 1992;14:
5815.
Sarbatly R, Krishnaiah D, England R. Continuous recycle enzymatic membrane reactor
system for in situ production of pure and sterile glucose solution. J Appl Sci 2007;7:
10638.
Sonia KG, Chadha BS, Badhan AK, Saini HS, Bhat MK. Identication of glucose tolerant
acid active -glucosidases from thermophilic and thermotolerant fungi. World
Microbiol Biotechnol 2008;24:599604.
Sun Y-M, Khang S-J. Catalytic membrane for simultaneous chemical reaction and
separation applied to a dehydrogenation reaction. Ind Eng Chem Res 1988;27:
113642.
425
Tjerneld F, Persson I, Albertsson P-. Enzymatic hydrolysis of cellulose in aqueous twophase systems. II. Semicontinous conversion of a model substrate, Solca Floc BW
200. Biotechnol Bioeng 1985;27:104450.
Wyman C. Cellulose bioconversion technology: the simultaneous saccharication and
fermentation process. Model formulation. Handbook on bioethanol: production
and utilization. Taylor & Francis; 1996. p. 2656.
Xiao Z, Zhang X, Gregg DJ, Saddler JN. Effects of sugar inhibition on cellulases and glucosidase during enzymatic hydrolysis of softwood substrates. Appl Biochem
Biotechol 2004;113116:111526.
Yang S, Ding W, Chen H. Enzymatic hydrolysis of rice straw in a tubular reactor coupled
with UF membrane. Process Biochem 2006;41:7215.
ARTICLE IN PRESS
I N T E R N AT I O N A L J O U R N A L O F H Y D R O G E N E N E R G Y
33 (2008) 644 651
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/ijhydene
Ethanol steam reforming in a dense PdAg membrane

reactor: A modelling work. Comparison with the
traditional system
F. Galluccia, M. De Falcob, S. Tostic, L. Marrellib, A. Basilea,
a
CNR-ITM, c/o University of Calabria, Via Pietro Bucci, Cubo 17C, 87030 Rende (CS), Italy
Universita` di Roma La Sapienza, Dip. Ing. Chimica e Materiali, Via Eudossiana 18, 00184 Roma, Italy
c
ENEA, Dipartimento Fusione, Tecnologie e Presidio Nucleare, C.R. ENEA Frascati, Via E. Fermi 45, 00044 Frascati (RM), Italy
b
ar t ic l e i n f o
abs tra ct
Article history:
The ethanol steam-reforming reaction for the production of synthesis gas has been studied
Received 2 February 2007
theoretically. A mathematical model has been formulated for a traditional reactor packed
Received in revised form
with a Co-based catalyst and then applied to a membrane reactor (MR) in which the
27 September 2007
hydrogen production is increased by removing the hydrogen produced from the reaction
Accepted 18 October 2007
mixture through a highly selective (100%) palladium-based membrane. Our simulation
Available online 28 November 2007
results show that with MR it is possible to obtain both higher conversions of ethanol and
Keywords:
Ethanol steam reforming
Membrane reactor
Pd membranes
Hydrogen production
higher hydrogen selectivities compared to those obtained in a traditional reactor operating

at the same experimental conditions. The theoretical analysis provides a set of parameters
allowing to maximize the (pure) hydrogen production and/or ethanol conversion if adopted
in an experimental device.
& 2007 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights
reserved.
Modelling
1.
Introduction
Polymeric fuel cell systems are electrochemical devices able

to generate electrical power by the electrochemical oxidation
of hydrogen with atmospheric oxygen. Many studies about
hydrogen production for fuel cells deal with the use of two
types of carbon compounds: the oxygen-containing compound, such as methanol or ethanol, and hydrocarbons, such
as natural gas, propane gas, gasoline, etc. [18].
The hydrogen production from hydrocarbons and alcohols
has been widely studied and applied: actually, in large size
plants 410 000 Nm3 =h, the hydrogen is produced by the
hydrocarbons conversion into syngas followed by the water
gas shift reaction, where part of the CO is converted into CO2
and hydrogen. Mainly, the methane reforming reaction and
then the gasification of coal are the processes used for
producing hydrogen. However, the use of fossil fuels increases the global warming as a consequence of the large
releases of carbon dioxide. Therefore, the hydrogen production starting from biofuels, such as the ethanol obtained from
biomasses, could contribute to stabilize the carbon dioxide
concentration and to reduce the greenhouse effect.
Many authors [914] studied the ethanol steam reforming
using traditional reactors. It is an endothermic catalysed
reaction whose conversion increases with the temperature:
C2 H5 OH H2 O 3 2CO 4H2 ,
DH298 K 239:47 kJ=mol.
If the water gas shift reaction is used for converting the CO,
the complete reaction is
C2 H5 OH 3H2 O 3 2CO2 6H2 ,
DH298 K 157:09 kJ=mol.
Corresponding author. Tel./fax: +39 0984 492013.
E-mail address: a.basile@itm.cnr.it (A. Basile).

0360-3199/$ - see front matter & 2007 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2007.10.039
ARTICLE IN PRESS
Nomenclature
A
CTs1
D
Ea
JH2
Kj
kr , kw ,
MR
N0i
membrane area, m2
total surface concentration of site 1, mol=m2 ,
membrane diameter, m
apparent activation energy, kJ/mol
hydrogen flux through the membrane,
mol=m2 s
adsorption constant for the species j, dimensionless
kd rate constants for the reactions 1, 2 and 3,
depending on the equation
membrane reactor, dimensionless
flow rate of the component i, mol/s
Usually, catalysts based on Rh, Ru, Pd, Pt, Ni, Co and Cu are
used on supports of Al2O3, SiO2, MgO and La2O3. The reaction
conversion and selectivity of the products (mainly H2, CO, CO2
and in minor part, CH4, CH3CHO and C2H4) depend on the
catalyst used as well as on the operating temperature. A large
excess of water in the feed stream increases the hydrogen
selectivity and reduces the coke formation that may poison
the catalyst particularly when the reaction is operated at low
temperature.
Using a traditional reactor, for example Liguras et al. [9]
obtained the complete ethanol conversion at 800 C with a
5%-Ru on alumina catalyst. In order to increase the ethanol
conversion at lower temperature, the use of membrane
reactors for carrying out the steam reforming is proposed in
this work. In fact, a membrane reactor is a device in which a
reaction and a selective separation take place simultaneously:
in this way, the continuous removal of one of the product (i.e.
hydrogen) gives reaction conversion beyond the thermodynamic equilibrium which is an upper limit to be considered in
a traditional reactor (shift effect) [1517]. In practice, when a
dehydrogenation reaction is performed, the use of a membrane selectively permeable to the hydrogen permits to build
a reactor producing and separating pure hydrogen with high
reaction conversions. The palladium and its alloys are
permeable to the hydrogen and are used for manufacturing
permeators and membrane reactors aimed at separating and
producing hydrogen: several studies report the use of these
membranes [1820]. Especially, the use of dense PdAg
membranes permits to maximize the shift effect and produce
directly ultra pure hydrogen without using any other purification unit.
The application of dense PdAg membrane reactors was
also suggested for carrying out methanol and ethanol steam
reforming reactions [2125]. The kinetic expressions for a Cubased catalyst is available in literature for the methanol
steam reforming. The use of PdAg membrane reactor was
also considered from a simulation viewpoint giving the
optimal set of operating parameters for maximising the COfree hydrogen recovery [26,27]. Such an optimisation was not
possible for the ethanol steam reforming since no reasonable
kinetic expression were available in the specialised literature.
In a recent paper, Sahoo et al. [28] presented, for the first
time, useful kinetic expressions for the ethanol steam
33 (2008) 644 651
645
P0
pj
pre-exponential factor, mol m=s m2 Pa0:5
plumen
H2
pshell
H2
partial pressure of hydrogen in the lumen side, Pa
R
ri
T
TR
V
W=F
W
ni;j
partial pressure for the species j, Pa
partial pressure of hydrogen in the shell side, Pa

gas constant, J/(K mol)
rate equation for the reaction i, mol=s m2
absolute temperature, K
traditional reactor, dimensionless
reactor volume, m3
time factor, kgcat =mol s
catalyst weight, kg
stoichiometric coefficient of the component i in
the reaction j dimensionless
reforming reaction carried-out on Co-based catalyst in a

traditional system.
In the present work, the rate expressions presented by
Sahoo et al. were used for simulating the behaviour of both a
traditional system (TR) and a dense PdAg membrane reactor
(MR). The performances in terms of ethanol conversion and
hydrogen production of both the TR and the MR have been
compared taking into account the effect of the operating
conditions (temperature, pressure, sweep gas mode and the
water/ethanol feed molar ratio).
2.
Theoretical model of ethanol steam
reforming
The model has been developed using all the thermodynamic
values for the species involved in the reaction and the kinetic
expressions; hydrogen removal through a self supported
PdAg membrane having a 23% Ag content was calculated
using experimental permeabilities [26].
The equation considered in the ethanol reaction system
carried out on a Co-based catalyst are as follows:
C2 H5 OH 3H2 O 3 2CO2 6H2 ,
DH298 K 157:09 kJ=mol,
(2)
CO H2 O 3 CO2 H2 ,
DH298 K 41:19 kJ=mol,
(3)
C2 H5 OH 3 CO CH4 H2 ,
DH298 K 33:18 kJ=mol.
Reactions (2) and (4) are both reversible and endothermic

reactions and proceed under volume increase, suggesting
that the highest ethanol conversions are obtained at high
temperature and low pressures. Reaction (3) is the (exothermic) water gas shift reaction which proceeds simultaneously
with the ethanol steam reforming, and with no volume
change.
ARTICLE IN PRESS
646
The rate equations are taken from Sahoo et al. [28], as well
as kinetic and adsorption parameters:
r1
0
1 "
pCH3 CH2 OH
A 1
kr KCH3 CH2 O1 @
1=2
pH2
!#
p4H2 p2CO2
Kr p2H2 O pCH3 CH2 OH
CTS1 2 ,
DEN
(5)
"
kw KHCOO1 pCO2 1
r2
!#
pH2 O pCO
Kw pH2 pCO2
CTS1 2 ,
DEN
p3H2 p2CO2
kd KCH3 CHO1
! "
1
p2H2 O
r3
p2H2 O pCO pCH4

Kd
(6)
!#
p3H2 p2CO2
CTS1 2 ,
DEN
(7)
where ri is the rate equation for the reaction i, pj the partial

pressure for the species j, Kj the adsorption constant for the
species j, kr , kw , kd the rate constants for the reactions 1, 2 and
3, respectively.
A mass balance for a differential reactor volume along the
z-axis of both the TR (Fig. 1) and the MR (Fig. 2) has been
written for the reaction zone and for the permeating zone (for
the MR only). The following common hypotheses have been
considered: plug flow, kinetic control of the reaction system,
isothermicity, constant catalyst bed density through the
reactor, validity of the Richardsons equation for the hydrogen
permeation.
The mass balances are:
Reaction zone:
3
dNi W X
pD
.
v r Ji
V j1 ij i
A
dz
DEN 1 KCO2 1 pCO2 KCO1 pCO KCH4 1 pCH4

1=2
Kr
KCH3 CHO1 p5H2 p2CO2
KHO1 pH2 O
p3H2 O
1=2
pH2
1=2
pH2
(9)
kw
KHCHO1
kw

,
Kw
KOH1 KCO1
KCH3 CHO1
z 0.
Reaction zone (lumen side):
kd
kd
(10)
N01 ;
N02
N03
N04
N05
N06

KCH4 1 KCO1
(11)
CH3 CH2 OH flow rate at the inlet side of the lumen;

mN01 ;
H2 O
flow rate at the inlet side of the lumen;
107 N01 ;
H2
107 N01 ;
CO2
107 N01 ;
CO
107 N01 ;
CH4
N03 0 ) ri ! 1,
where m is the H2 O=CH3 CH2 OH molar feed ratio and the sum
of N0i is the feed molar flow rate F.
Permeation zone (shell side):
N07 N0sweep
gas ;
N08 0;
Catalyst Pellets (lumen side)

Feed
Products
Stainless
Steel Tube
(13)
Boundary conditions:
kr
kr
,
KCH3 CHO1 KH2 1
Kd
dNi
pD
.
Ji
A
dz
KCH3 CH2 O1 pCH3 CH2 OH1
KCH3 CH2 O1
(12)
Permeation zone:
1=2
KHCOO1 pH2 pCO2 KH2 pH2
33 (2008) 644 651
Glass Spheres
z
sweep gas flow rate at the inlet side of the shell;

H2 flow rate at the inlet side of the shell in co-current mode:
The flux through the membrane is calculated by the

Richardson equation
Ea
0:5
0:5 pshell
JH2 P0 e RT plumen
H2
H2 ,
(14)
Fig. 1 Sketch of a traditional reactor.
Sweep Gas Inlet

(shell side)
Permeate Outlet
Catalyst Pellets
(lumen side)
Retentate Stream
Feed
Pd /Ag Membrane Tube
Thermocouple
z
Fig. 2 Sketch of a membrane reactor.
Stainless
Steel Tube
Graphite
Gasket
ARTICLE IN PRESS
where the values of the apparent activation energy and preexponential factor are 29.16 kJ/mol and 1:12 105 mol m=
s m2 Pa0:5 , respectively.
A non reactive sweep gas (nitrogen or steam) has been
considered in the simulations. The sweep gas ratio, defined as
the sweep gas molar rate/ethanol feed molar rate, has been
set at 0.1 except for the last simulations in which the effect of
the sweep ratio was studied.
The set of differential equations has been solved using a IV
order RungeKutta method with variable step in order to
overcome the instabilities in the first part of the reactor. The
code was written by using the Salford Plato3 FORTRAN95
platform.
In order to compare the results for the two reactors, the
ethanol conversion, the hydrogen production and the hydrogen recovery are defined as follows:
CH3 CH2 OH conversion; %
CH3 CH2 OHIN CH3 CH2 OHOUT
100,
CH3 CH2 OHIN
Results and discussion
3.1.
Validation of the model
3.2.
Comparison between TR and MR
In the following, the rate kinetic expression have been

considered to be valid also for the MR. The comparison
between the TR and the MR is made considering different
operating parameters such as temperature, pressure, sweep
gas flow rate and so on.
3.2.1.
(16)
H2;outshell
100,
H2;outlumen H2;outshell
3.
whole range of parameters investigated. In fact, the maximum deviation between the model and the experimental
results in terms of ethanol conversion is 5.8% at 773 K and
W=F 15 kgcat =mol s.
Unfortunately, to the best of our knowledge, there are no
data in literature regarding the use of membrane reactors for
the ethanol steam reforming on Co-based catalyst. For this
reason, a direct comparison of the modelling results and
membrane reactor experimental results is not possible.
15
H
H2;outshell
H2 produced; % 2;outlumen
100,
6 CH3 CH2 OHIN
H2 recovery; %
(17)
The model was validated by using the experimental results

reported by Sahoo et al. [28] for the TR. In Fig. 3 it can be seen
that the model was validated in the range of temperature
673873 K and for the time factor W=F in the range
0217 kgcat =mol s. The experimental and modelling results
are related to the ethanol conversion at a reaction pressure of
1 bar. Our theoretical model and the experimental results
reported by Sahoo et al. [28] show a good agreement in the
Effect of the temperature
Fig. 4 reports a direct comparison, in terms of ethanol

conversion versus W=F at different temperatures, between
the TR and the MR. The first information that can be
highlighted is that the conversion increases for both the TR
and the MR with the temperature increase due to the higher
catalytic activity. Moreover, it can be seen that, in the whole
range of W=F investigated and for each temperature the
ethanol conversion in the MR is higher than the one in the TR.
The difference between the two reactors is mainly due to the
hydrogen removal through the PdAg membrane which shifts
the reaction system towards the products. More in detail, the
difference in terms of conversion between the TR and the MR
increases by increasing the temperature. This effect can be
explained by considering the temperature dependence of the
hydrogen flux through the dense membrane. In fact, the
hydrogen permeates through the dense PdAg membrane by
following the Richardson equation in which the temperature
dependence is an Arrhenius like equation that is to say that
100
MR , 873K
Ethanol conversion, %
Ethanol Conversion, %
100
80
60
40
673 K
773 K
873 K
Model
20
MR , 773K
80
MR , 673K
60
TR , 873K
TR , 773K
40
TR , 673K
20
TR
MR
0
0
0
0
6
8
10
12
W/F, kg cat/(mol/s)
647
33 (2008) 644 651
14
16
18
Fig. 3 Comparison between experimental results and

theoretical ones for TR. p 1 bar.
6
8
10
12
W/F, kg cat/(mol/s)
14
16
18
Fig. 4 Ethanol conversion vs W=F for MR and TR at different

temperatures. Theoretical results. plumen 1 bar,
pshell 1 bar.
ARTICLE IN PRESS
648
the hydrogen flux increases by increasing the temperature.

The increase of hydrogen flux through the membrane results
in a higher ethanol conversion. For this reason the temperature has a higher positive effect in the MR than in the TR. The
second effect which can be highlighted is that, the difference
in terms of conversion between the MR and the TR increases
by increasing the W=F. This can be explained considering
that, for a fixed feed flow rate, the increase of the W=F results
in an increase not only of the catalyst weight but also an
increase of the membrane area. This results in an increase of
the hydrogen flux through the membrane and in a higher
conversion.
From the figure it can also be seen that the MR is able to give
the same conversion of the TR but at lower temperature. In
fact, the MR operated at 673 K gives the same conversion of
the TR operated at 773 K.
3.2.2.
Effect of the pressure
The effect of the reaction pressure on the reaction system is

shown in Figs. 5 and 6. In particular, in Fig. 5 the conversion
100
8 bar
4 bar
2 bar
1 bar
80
1 bar
60
2 bar
4 bar
8 bar
40
20
TR
MR
0
0
6
8
10
12
W/F, kg cat/(mol/s)
14
16
18

pressures. Theoretical results. T 673 K, pshell 1 bar.
100
1 bar
1 bar
2 bar
80
4 bar
60
8 bar
40
20
TR
MR
0
0
4 5
9 10 11 12 13 14 15 16 17 18
W/F, kg cat/(mol/s)

33 (2008) 644 651
versus W=F at 673 K for both the TR and the MR is reported at

different pressures (in the range 18 bar). Concerning the
pressure effect on the TR performances, it can be noted in
Fig. 5 that, at 673 K, there are three zones of W=F. In the first
zone, corresponding at W=Fo3 kgcat =mol s in which by
increasing the reaction pressure from 1 to 8 bar, the conversion in the TR increases. This effect could be explained
considering the low value of W=F that can correspond at a
kinetic limiting zone in which an increase of the pressure
favours the adsorption of the species on the catalyst surface
resulting in a higher activity and thus in a higher ethanol
conversion. In the second zone, corresponding to a W=F
values between 3 and 15 kgcat =mol s, the conversion versus
pressure behaviour have a maximum which corresponds at
lower pressures when the W=F increases. This because, by
increasing the W=F value the system becomes more and more
controlled by the thermodynamics which imposes a decrease
of the conversion by increasing the pressure. In fact, the
reaction system proceeds with a mole number increase and
so it is stoichiometrically favoured by low pressures.
In the third zone, i.e. W=F414 kgcat =mol s, the thermodynamics is the only controlling step. In fact, in a TR the
conversion continually decreases by increasing the pressure.
Fig. 6 shows the same behaviour of Fig. 5 but at 873 K. It can
be seen that there are always three zones of W=F but in this
case the first zone corresponds to W=Fo2 kgcat =mol s while
in the second zone, i.e. W=F values between 1.8 and
15 kgcat =mol s, the maximum conversion values versus the
pressure are shifted towards lower values of W=F. Finally, in
the third zone the difference between the conversion at 1 bar
and the conversion at 8 bar is higher than the one at 673 K.
These considerations suggest that the thermodynamic control of the reaction system is stronger at high temperatures.
Considering the MR, Fig. 5 shows that the conversion
increases by increasing the pressure in the whole range of
W=F investigated. It has to be considered that the pressure
has two opposite effects on the reaction system. The first one
is the same described for the TR: the pressure can have a
positive or a negative effect on the reaction system depending
on the W=F value and on the temperature. The second effect
is always a positive effect on the ethanol conversion and it is
related to the permeation of hydrogen through the membrane. In fact, by increasing the reaction pressure the
hydrogen partial pressure square root difference between
lumen side and shell side of the reactor increases resulting in
a higher hydrogen recovery from the reaction zone. This
positively affects the reaction system.
In our simulations, the combination of the two effects
described above results in an overall positive effect of the
pressure on the reaction system. In particular, at 8 bar and
W=F 17 kgcat =mol s, the conversion in the MR is 95.3%, while
it is 44.5% in the TR.
The effect of the reaction pressure on the hydrogen
recovery versus the W=F is reported in Fig. 7. It can be seen
that in the whole range of W=F investigated the hydrogen
recovery (as COx -free stream) increases by increasing the
reaction pressure and, in particular, it is 96.3% at
17 kgcat =mol s and 8 bar, while 43.9% at 1 bar. The comparison between Figs. 5 and 7 demonstrates that, by using an MR,
it is worth working at relatively high pressure in order to
ARTICLE IN PRESS
100
Hydrogen recovery, %
4 bar
80
2 bar
60
40
1 bar
20
MR
90
90
TR
80
70
60
60
50
50
40
40
30
30
6
8
10
12
W/F, kg cat/(mol/s)
14
16
18
Fig. 7 Hydrogen recovery vs W=F for MR at different

Hydrogen produced, %
100
8 bar
4 bar
2 bar
80
1bar
60
40
20
0
0
10
W/F, kg cat/(mol/s)
15
20
Fig. 8 Hydrogen produced vs W=F for MR at different

increase both the ethanol conversion and the hydrogen

recovery. The increase of these two variables results in a
higher hydrogen production decreasing the hydrogen loss as
by products (i.e. methane).
Fig. 8 reports the results in terms of hydrogen produced,
calculated as the total hydrogen produced (lumen side shell
side) divided by the total hydrogen produced if the ethanol is
completely converted in carbon dioxide and hydrogen, versus
the W=F at various pressures. The figure confirms that the
hydrogen production increases by increasing the reaction
pressure. It should be said that, the difference between the
hydrogen produced at 4 and at 2 bar is higher than the ones
between 8 and 4 bar. This means that, a deeper investigation
is necessary in order to determine the best reaction pressure
depending on the hydrogen production and the lifetime of the
dense membrane (not considered in this work).
3.2.3.
20
0
Effect of the feed gas molar ratio
Some consideration about the effect of the water/ethanol feed

molar ratio on the performances of the reactors are presented
in this work. As shown in Fig. 9, the effect of the feed molar
80
70
20
0
100
Hydrogen recovery, %
8 bar
100
649
33 (2008) 644 651
3 4 5 6 7 8 9 10 11 12
Steam/ethanol molar ratio, -
Fig. 9 Ethanol conversion vs feed molar ratio for MR.

Theoretical results. T 673 K, plumen 1 bar, pshell 1 bar.
ratio on the TR performances is positive since the only

important parameter for the TR, the ethanol conversion,
monotonously increases by increasing the feed ratio. In fact,
the ethanol conversion in the TR is 27.9% at the feed molar
ratio 1 while it increases up to 86% at feed molar ratio 11.
Using the MR, it should be pointed out that there are two
important parameters to be considered: the ethanol conversion and the hydrogen recovery. In Fig. 9 it is evident that the
ethanol conversion shows the same behaviour as the one in
TR while the hydrogen recovery shows a maximum feed
molar ratio around 2 and afterwards it slightly decreases by
increasing the feed molar ratio. On the one hand, the effect of
the feed molar ratio on the conversion is quite obvious since,
by keeping constant the total molar feed rate, an increase in
the feed molar ratio means an increase on the time factor for
the ethanol and so an increase of the ethanol conversion. On
the other hand, by increasing the feed molar ratio part of the
excess of water dilutes the hydrogen produced in the reaction
zone resulting in a lower hydrogen partial pressure difference
between the lumen side and the shell side of the reactor and
so in a lower hydrogen recovery.
For this reason it is evident that, also for finding the optimal
feed molar ratio, a deeper investigation is needed. In
particular, the optimisation must be done considering also
the carbon formation on the catalyst surface. This means
that, the optimal feed molar ratio must be as high as the
minimum feed ratio which avoids the carbon formation on
the catalyst.
3.2.4.
Effect of the sweep gas flow rate
An important parameter which influences the performances

of the MR is the sweep gas flow rate because, by increasing
the sweep gas flow rate the hydrogen partial pressure in the
shell side decreases resulting in an increase of the hydrogen
recovery. In this work, the sweep gas flow rate was changed as
a multiple of the ethanol feed flow rate (and indicated as
sweep ratio) in the range 120. Fig. 10 sketches the ethanol
conversion versus the W=F at different sweep ratio; the curves
represent the ethanol conversion at sweep ratio 1, 1.5, 4, 8 and
20, respectively. It can be seen that, by increasing the sweep
ratio, as expected, the ethanol conversion increases in the
whole range of W=F investigated. In particular at W=F
ARTICLE IN PRESS
650
100
Ethanol Conversion, %
Sweep ratio
20
80
1
60
40
20
0
0
6
8
10
12
W/F, kg cat/(mol/s)
14
16
18
Fig. 10 Ethanol conversion vs W=F for MR at different

sweep ratio. Theoretical results. T 673 K, plumen 1 bar,
pshell 1 bar.
17 kgcat =mol s the ethanol conversion increases from 79.4%

at sweep ratio 1 up to 98% at sweep ratio 20. It should be
noted that the highest increase in the ethanol conversion
occurs in the sweep ratio range 18 where the conversion
increases from 79.4% to 94%. Afterwards the conversion
reaches a plateau value and an increase of the sweep ratio
just makes the hydrogen containing stream in the shell side
much diluted without any benefit in terms of reactor
performances.
4.
Conclusions
In this paper the ethanol steam-reforming reaction for

producing hydrogen has been, for the first time, investigated
from a theoretical point of view. The performance of the MR
was compared with the TR ones indicating that the ethanol
conversion can be increased by using the MR with respect to
the TR. Moreover a COx -free hydrogen stream can be directly
produced in the MR and it can be directly fed to a proton
exchange membrane fuel cell system.
A set of parameters which influence the reactor performances were studied. The results indicate that the ethanol
steam reforming reaction can be carried out in the MR at
relatively high temperature and high reaction pressure with a
high water/ethanol feed ratio in order to maximise the
ethanol conversion and the COx -free hydrogen production.
In fact, the best results are obtained at T 673 K, p 8 bar and
H2 O=C2 H5 OH 3 where the ethanol conversion is 95.3%, the
hydrogen production 90% and the hydrogen recovery 94%.
An optimization of the MR must be done in order to find the
optimum set of operative parameters.
R E F E R E N C E S
[1] de Wild PJ, Verhaak MJFM. Catalytic production of hydrogen

from methanol. Catal Today 2000;60:310.
33 (2008) 644 651
[2] Wiese W, Emonts B, Peters R. Methanol steam reforming

in a fuel cell drive system. J Power Source 1999;84:
18793.
[3] Breen JP, Burch R, Coleman HM. Metal-catalysed steam
reforming of ethanol in the production of hydrogen for
fuel cell applications. Appl Catal B: Environ 2002;39:
6574.
[4] Cavallaro S, Chiodo V, Vita A, Freni S. Hydrogen production by
auto thermal reforming of ethanol on RhAl2O3 catalyst.
J Power Source 2003;123:106.
[5] Maggio G, Freni S, Cavallaro S. Light alcohols/methane
fuelled molten carbonate fuel cells: a comparative study.
J Power Source 1998;74:1723.
[6] Heinzel A, Vogel B, Hubner P. Reforming of natural gashydrogen generation for small scale stationary fuel cell
systems. J Power Source 2002;105:2027.
[7] Shen JP, Song C. Influence of preparation method performance of Cu/Zn-based catalysts for low-temperature
steam reforming and oxidative steam reforming of
methanol for H2 production for fuel cells. Catal Today
2002;77:8998.
[8] Lindstrom B, Pettersson LJ. Hydrogen generation by steam
reforming of methanol over copper-based catalysts for fuel
cell applications. Int J Hydrogen Energy 2001;26:92333.
[9] K Liguras D, Kondarides DI, Verykios XE. Production of
hydrogen for fuel cells by steam reforming of ethanol over
supported noble metal catalysts. Appl Catal 2003;43:
34554.
[10] Freni S, Mondello N, Cavallaro S, Cacciola G, Pardon VN,
Sobyanin VA. Hydrogen production by steam reforming
of ethanol: a two process. React Kinet Catal Lett
2000;71:14352.
[11] Srinivas D, Satyanarayana CVV, Potdar HS, Ratnasamy P.
Structural studies on NiOCeO2ZrO2 catalysts for steam
reforming of ethanol. Appl Catal 2003;246:32334.
[12] Llorca J, Homs N, Sales J, de la Piscina PR. Efficient production
of hydrogen over supported cobalt catalysts from ethanol
steam reforming. J Catal 2002;209:30617.
[13] Haga F, Nakajima T, Yamashita K, Mishima S. Effect of
crystallite size on the catalysis of alumina-supported cobalt
catalyst for steam reforming of ethanol. React Kinet Catal
Lett 1998;63:2539.
[14] Kaddouri A, Mazzocchia C. A study of the influence of the
synthesis conditions upon the catalytic properties of Co/SiO2
or Co/Al2O3 catalysts used for ethanol steam reforming. Catal
Commun 2004;5:33945.
[15] Shu J, Grandjean BPA, Van Neste A, Kalaguine S. Catalytic
palladium-based membrane reactors: a review. Can J Chem
Eng 1991;69:103660.
[16] Kikuchi E. Membrane reactor application to hydrogen
production. Catal Today 2000;56:97101.
[17] Armor JN. Applications of catalytic inorganic membrane
reactors to refinery products. J Membr Sci 1998;147:
21733.
[18] Paturzo L, Basile A, Drioli E. High temperature membrane
reactors and integrated membrane operations. Rev Chem
Eng 2002;18(6):51151.
[19] Tosti S, Basile A, Bettinali L, Borgognoni F, Chiaravalloti F,
Gallucci F. Long-term tests of PdAg thin wall permeator
tube. J Membr Sci 2006;284:3937.
[20] Wieland S, Melin T, Lamm A. Membrane reactors for
hydrogen production. Chem Eng Sci 2002;57:15716.
[21] Amandusson H, Ekedahl LG, Dannetun H. Alcohol dehydrogenation over Pd versus PdAg membranes. Appl Catal A: Gen
2001;217:15764.
[22] Basile A, Gallucci F, Paturzo L. A dense Pd/Ag membrane
reactor for methanol steam reforming: experimental study.
Catal Today 2005;104:24450.
ARTICLE IN PRESS
[23] Basile A, Gallucci F, Paturzo L. Hydrogen production from

methanol by oxidative steam reforming carried out in a
membrane reactor. Catal Today 2005;104:2519.
[24] Basile A, Tosti S, Capannelli G, Vitulli G, Iulianelli A, Gallucci
F, et al. Co-current and counter-current modes for methanol
steam reforming membrane reactor: experimental study.
Catal Today 2006;118:23745.
[25] Gallucci F, Basile A, Tosti S, Iulianelli A, Drioli E. Methanol
steam reforming and ethanol steam reforming in membrane
reactors: an experimental study. Int J Hydrogen Energy
2007;32(9):120110.
33 (2008) 644 651
651
[26] Gallucci F, Paturzo L, Basile A. Hydrogen recovery from

methanol steam reforming in a dense membrane
reactor: simulation study. Ind Eng Chem Res 2004;43:
242032.
[27] Gallucci F, Basile A. Co-current and counter-current modes
for methanol steam reforming membrane reactor. Int J
Hydrogen Energy 2006;31:22439.
[28] Sahoo DR, Shilpi Vajpai, Sanjay Patel, Pant KK. Kinetic
modelling of steam reforming of ethanol for the production
of hydrogen over Co/Al2O3 catalyst. Chem Eng J
2007;125(3):13947.
Bioresource Technology 98 (2007) 639647
Biodiesel production using a membrane reactor

M.A. Dube *, A.Y. Tremblay, J. Liu
Department of Chemical Engineering, University of Ottawa, Ottawa, Ont., Canada K1N 6N5
Received 31 May 2005; received in revised form 5 January 2006; accepted 3 February 2006
Available online 31 March 2006
Abstract
The immiscibility of canola oil in methanol provides a mass-transfer challenge in the early stages of the transesterication of canola oil
in the production of fatty acid methyl esters (FAME or biodiesel). To overcome or rather, exploit this situation, a two-phase membrane
reactor was developed to produce FAME from canola oil and methanol. The transesterication of canola oil was performed via both
acid- or base-catalysis. Runs were performed in the membrane reactor in semi-batch mode at 60, 65 and 70 C and at dierent catalyst
concentrations and feed ow rates. Increases in temperature, catalyst concentration and feedstock (methanol/oil) ow rate signicantly
increased the conversion of oil to biodiesel. The novel reactor enabled the separation of reaction products (FAME/glycerol in methanol)
from the original canola oil feed. The two-phase membrane reactor was particularly useful in removing unreacted canola oil from the
FAME product yielding high purity biodiesel and shifting the reaction equilibrium to the product side.
2006 Elsevier Ltd. All rights reserved.
Keywords: Biodiesel; Methanol; Acid-catalyzed transesterication; Base-catalyzed transesterication; Two-phase membrane reactor
1. Introduction
Biodiesel is a clean-burning diesel fuel produced from
vegetable oils, animal fats, or grease. Its chemical structure
is that of fatty acid alkyl esters (FAAE). Commercially,
biodiesel is produced by transesterication. This reaction
consists of transforming triglyceride (TG) into FAAE, in
the presence of an alcohol (e.g. methanol, ethanol) and a
catalyst (e.g. alkali, acid, enzyme) with glycerol as a major
by-product. The reaction scheme is shown in Fig. 1.
In Fig. 1, X represents the alkyl group of the alcohol
(e.g. CH3 for methanol) while R is a carbon chain typically
on the order of 1120 carbon atoms long. The conversion
of TG to FAAE comprises three consecutive reversible
reactions with diglyceride (DG) and monoglyceride (MG)
as intermediate products. Biodiesel is being used increasingly in public transportation in Europe, Japan and North
America.
*
Corresponding author. Tel.: +1 613 562 5800x6108; fax: +1 613 562

5172.
E-mail address: Marc.Dube@uottawa.ca (M.A. Dube).
0960-8524/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2006.02.019
Due to diminishing petroleum reserves and the environmental consequences of exhaust gases from petroleumderived fuels, such as gasoline and diesel, biodiesel has
attracted attention during the past decade as a renewable
and environmentally friendly fuel. Because biodiesel is
made entirely from vegetable oil or animal fats, it is renewable, environmentally benign (biodegradable), and does
not contain any sulphur, aromatic hydrocarbons, metals
or crude oil residues.
Like petroleum diesel, biodiesel operates in compressionignition engines such as those used in farm equipment, and private and commercial vehicles. Essentially no
engine modications are required, and biodiesel maintains
the payload capacity and range of diesel. Because biodiesel
is oxygenated, it is a better lubricant than diesel fuel,
increasing the life of engines, and is combusted more completely. Indeed, many countries are introducing biodiesel
blends to replace the lubricating eect of sulfur compounds
in low-sulfur diesel fuels (Anastopoulos et al., 2001; Dmytryshyn et al., 2004). The higher ash point of biodiesel
makes it a safer fuel to use, handle and store. With its
relatively low emission prole, it is an ideal fuel for use
M.A. Dube et al. / Bioresource Technology 98 (2007) 639647
640
H2C-O-H
H 2C-O-R
Catalyst
HC-O-R
H 2C-O-R
3X-OH
HC-O-H
3RCOOX
H 2C-O-H
Fig. 1. Stoichiometric reaction for conversion of triglyceride to fatty acid

alkyl esters.
in sensitive environments, such as marine areas, national

parks and forests, and heavily polluted cities.
Existing technology to produce biodiesel continues
to require government subsidies in order to be protable.
One reason is the high cost of virgin vegetable oil, the
source of TG (Coltrain, 2002). In order to reduce production costs and make it competitive with petroleum diesel,
low cost feedstock, such as non-edible oils, waste frying oils
and animal fats, could be used as raw materials (Zhang
et al., 2003a). However, the relatively higher amounts of
free fatty acids (FFA) and water in this feedstock results
in the production of soap in the presence of alkali catalyst.
Thus, additional steps to remove any water and either
the FFA or soap from the reaction mixture are required.
In fact, commercial base-catalyzed processes often employ
an acid-catalyzed pre-esterication reactor to remove
excess FFAs.
Alkali-catalyzed transesterication proceeds much faster
than that catalyzed by an acid and is the one most used
commercially (Freedman et al., 1984). Considerable
research has been done on biodiesel made from virgin
vegetable oils (e.g. soybean oil, sunower oil, rapeseed oil)
using alkali catalysts. The majority of biodiesel today is
produced by alkali-catalyzed (e.g. NaOH, KOH) transesterication with methanol, which results in a relatively short
reaction time. However, as mentioned above, the vegetable
oil and alcohol must be substantially anhydrous and have a
low FFA content because the presence of FFA promotes
soap formation. The soap formed partially consumes the
catalyst, lowers the yield of esters and renders the downstream separation of the products dicult (Freedman
et al., 1984).
Freedman et al. (1985) reported a 98% yield of biodiesel
in 1 h using alkali catalysts such as sodium hydroxide or
sodium methoxide with alcohols such as methanol, ethanol, and 1-butanol. Freedman et al. (1984) investigated
the eect of the molar ratio of the alcohol to oil, type of
catalyst (base vs. acid), temperature and degree of renement of the oil on the yield of biodiesel. For the alkali-catalyzed reaction, the eect of alcohol to oil ratio was found
to be the most important variable aecting the yield, while
temperature had a signicant eect on the initial reaction
rate. It also was mentioned in their study that acid catalysts
would be more eective when the degree of renement of
oil was low, and for oils that had a high FFA content.
Several studies have been done on the production of biodiesel from waste oils or animal fats (Nye et al., 1983;
Watanabe et al., 2001) describing the feasibility of making
quality biodiesel from this feedstock while identifying the

problems with the FFAs present in the raw materials.
Experiments in our laboratory (McBride, 1999; Ripmeester, 1998; Zheng, 2003) showed that greater than 99.5%
conversion of waste frying oil to biodiesel could be
achieved within a reasonable time period (3 h at 70 C)
through transesterication with an excess of methanol
(methanol:oil molar ratio was >50:1) catalyzed by sulphuric acid. Despite the slower reaction rate, this approach
has several advantages over the base-catalyzed method
(Canakci and Van Gerpen, 1999; Zhang et al., 2003a,b):
it employs a one-step process as opposed to a two-step
process; it can handle feedstock with a high FFA content;
downstream separation of the biodiesel is straightforward;
and a high quality glycerol by-product is produced.
Aside from the slow reaction rate, another drawback of
the acid-catalyzed process is the requirement for the reactor
to withstand an acidic environment. Nonetheless, our economic assessment carried out on four dierent continuous
processes with dierent types of oil (virgin vs. waste) and
catalysts (acid vs. alkali) showed that although the alkalicatalyzed process using virgin oil had the lowest capital
investment cost, the cost of using virgin oil led to a higher
total manufacturing cost (Zhang et al., 2003b). When waste
frying oil was used in the alkali-catalyzed process, a pretreatment unit was required to reduce the content of the
FFA. Thus, the cost associated with the pre-treatment unit
oset the cost savings due to the use of waste frying oil. In
contrast, the acid-catalyzed process using waste frying oil
did not require a pre-treatment step and had the lowest total
manufacturing cost, and the highest protability. Yet
another drawback to the acid-catalyzed process, is that high
alcohol to oil ratios are necessary to promote the conversion
of oil to FAAE (Freedman et al., 1985). These higher
amounts of alcohol increase the reactor size but recycling
can mitigate some of the associated increases in cost. Nonetheless, the issue of separating these substantial amounts of
alcohol from the FAAE becomes important.
Another issue that plagues biodiesel production is the
removal of residual TG and glycerol from the biodiesel
product. One approach is to drive the reaction as close to
complete conversion of the TG as possible. However, the
transesterication of TG is an equilibrium reaction, and
there are thus, limits to this approach. Other approaches
employ multiple water washing steps of the product
stream, which can give rise to a challenging waste treatment problem in the wastewater stream. Karaosmanoglu
et al. (1996) studied three dierent separation methods
and found that hot water washing at 50 C was the best
way to separate and purify the biodiesel product. Bam
et al. (1995) suggested that glycerine could be added as a
solvent to wash impurities. Hayafuji et al. (1999) studied
the use of a solid absorbent, such as activated clay, activated carbon, activated carbon bre, etc. to purify the
resultant esters.
Miscibility is an important factor in biodiesel production. The conventional transesterication method results
in a two-phase reaction which is, as a result, mass-transfer

limited. More specically, the vegetable oils and methanol
are not miscible. The approach of many existing commercial enterprises has been largely focused on steps
to enhance reaction rate by attempting to overcome this
immiscibility. For example, the addition of a co-solvent
to generate a homogeneous reaction mixture can greatly
enhance the reaction rate (Boocock et al., 1996, 1998).
While this enhances reaction rate signicantly, the co-solvent must eventually be separated from the biodiesel and
this requires additional processing. Considering that reaction rate may not necessarily restrict process protability
(Zhang et al., 2003b), transesterication is an equilibrium
reaction and downstream processing of the biodiesel is of
utmost concern, another approach can be envisaged. This
approach involves the use of membrane separation technologies, which exploit the immiscibility of the oil and
methanol.
In this paper, we report on the production of biodiesel
using a semi-batch two-phase carbon membrane reactor
using either an acid- or base-catalyzed transesterication
of canola oil. Carbon membranes are of particular interest
as they can be stable at high temperatures and resist chemical attack.
2. Background
Membrane based separations are well-established technologies in water purication, protein separations and
gas separations. To date, commercial applications of membrane technologies are limited to separations involving
aqueous solutions and relatively inert gases. The use of
membranes to treat non-aqueous uids is an emerging area
in membrane technologies (Kim et al., 2002; White and
Nitsch, 2000).
Synthetic membranes also have been used in reactor systems to remove the products of reactions as they are being
formed. Product-inhibited and equilibrium reaction systems are very common in commercially important reactions. Removing reaction products as they are formed
drives reactions to very high conversions that cannot be
reached in conventional reactors. To date, much eort
has been spent on membrane reactors involving easily separated mixtures such as hydrogen and methane (Armor,
1998; Hsieh, 1991, 1996; Saracco et al., 1999; Saracco
and Specchia, 1994; Zaman and Chakma, 1994).
The membrane can be either organic in nature (i.e.,
polymeric) or inorganic. Inorganic membranes are more
suitable for use with organic solvents and, due to their
excellent thermal stability, they can be used at high reaction temperatures (Saracco et al., 1994, 1999).
Membrane reactors can have several advantages over
conventional reactors (Armor, 1989):
(a) an integration of reaction and separation into a single
process, thus reducing separation costs and recycle
requirements,
641
(b) an enhancement of thermodynamically limited or

product-inhibited reactions resulting in higher conversions per pass,
(c) a controlled contact of incompatible reactants and
(d) an elimination of undesired side reactions.
3. Experimental
3.1. Materials
Methanol (99.85% Reag. Grade containing <0.1%
water) was supplied by (Commercial Alcohols Inc., Brampton, ON, Canada) and the canola oil by (No Name, Toronto, ON, Canada, purchased at a local foodstore). Fatty
acid methyl esters (FAME or biodiesel) used for miscibility
testing was produced from the acid-catalyzed transesterication of waste oils from a previous study (Zheng, 2003).
Sulfuric acid (9598%, Reagent Grade) and tetrahydrofuran (99.95%, Chromatography Grade) were supplied by
(EMD Chemicals Inc., Gibbstown, NJ, USA).
3.2. Membrane reactor design and experimental design
A 300 mL membrane reactor system was constructed and
is shown schematically in Fig. 2. A carbon membrane (Koch
Membrane Systems, Inc., Wilmington, DE, USA) was used
in the reactor. The pore size of the membrane was 0.05 lm.
The inside and outside diameters of the membrane were 6
and 8 mm, respectively. The length of carbon membrane
tube was 1200 mm giving a surface area of 0.022 m2 for
the entire membrane. A controlled volume pump (Milton
Roy Company, Ivyland, PA, USA) was used to feed the
oil and methanol/catalyst mixtures to the system while a
seal-less centrifugal canned motor pump (Labcor Inc. Concord, ON) was used to circulate the mixture at a speed of
15.2 mL/min. A heat exchanger (Neslab Instruments, Inc.,
Portsmouth, NH, USA) coupled with LabViewTM software
was used to control the reaction temperature.
Experiments were carried out at 60, 65 and 70 C in a
300 mL membrane reactor for 6 h, 0.5, 2, 4 and 6 wt.% concentrations of sulfuric acid catalyst were investigated (see
Table 1). Canola oil (100 g) was used in each run. Pressure
was controlled at 138 kPa between the permeation side and
reaction side of the membrane. All experiments and sample
analyses were carried out in random order to minimize any
potential experimental errors. Several replicate runs also
were performed (see Table 1). Additional experiments were
conducted to verify the eect of methanol feed ow rate
and the use of a base catalyst.
3.3. Membrane reactor experiments procedure
The methanol and sulfuric acid were pre-mixed and
charged into the membrane reaction system prior to each
reaction. Canola oil (100 g) was charged into the membrane reactor, the membrane reactor was sealed and the
642
Fig. 2. Schematic diagram of separative membrane reactor.
Table 1
Experimental conditions
Temperature (C)
Catalyst concentration
(wt.%)
# of replicates
60
65
70
60
65
70
60
65
70
60
65
70
0.5
0.5
0.5
2
2
2
4
4
4
6
6
6
2
2
2
3
3
2
2
2
2
4
4
4
circulation pump was started. After a 10 min circulation

time, methanol and acid catalyst were charged continuously
into the membrane reactor with the feed pump at a ow rate
of 6.1 mL/min. The heat exchanger was switched on to
achieve the reaction temperature (60, 65 and 70 C). A thermocouple was used to monitor the reaction temperature. A
stable reaction temperature (0.1 C) was achieved within
30 min for 60 C, 40 min for 65 C and 45 min for 70 C
of starting the heat exchanger. Pressure was controlled at
138 kPa. The permeate product was collected in a
2000 mL ask. All experiments were conducted for 6 h.
Additional experiments were conducted to observe the
eect of methanol/acid catalyst feed ow rate on conver-
sion for both acid- and base-catalyzed transesterications.

These ow rates were 2.5, 3.2 and 6.1 mL/min.
The permeate product collected during the entire experiment time was mixed with an equivalent volume of reverse
osmosis water (produced from tap water) and shaken by
hand for about 5 min. This step served to stop any further
reaction in the samples by promoting a phase separation of
the glycerol phase containing most of the catalyst from the
FAME phase. The mixture was allowed to settle for 24 h
and ltered using a 0.5 lm membrane lter (Nalge Company, New York, NY, USA). The upper layer of the resulting two-phase mixture was transferred to a separatory
funnel and washed with 1 L of reverse osmosis water.
The resulting mixture was allowed to settle for 24 h, after
which the upper layer was analyzed using high performance liquid chromatography (HPLC) according to the
method used by Dube et al. (2004). Any unreacted oil in
the retentate stream was also analyzed by HPLC. The
retentate solution was neutralized by sodium hydroxide
solution before analysis by HPLC. The HPLC analysis
revealed that the purication method was eective and no
residual acid was found in the samples.
3.4. High performance liquid chromatography (HPLC)
analysis
A Waters Corp. HPLC system consisting of an HPLC
pump, a controller, a dierential refractometer and autosampler was used to analyze the contents of the permeate
The fractional conversion of oil to FAME, based on the

amount of oil remaining in the reactor, was taken to represent the actual conversion. The oil to FAME conversion at
time t was calculated from
and retentate streams. Waters Millennium 32TM software

(Waters) was utilized for analysis. The columns used were
two 300 7.5 mm Phenogel columns of 3 lm particles
pore size (Phenomenex, Torrance, CA, USA)
and 100 A
connected in series. The mobile phase was tetrahydrofuran
(THF) at a ow rate of 0.5 mL/min at 23 C.
THF was used to make a 20 mg/g solution of the sample. Two grams of the solution was injected into the autosampler vials. Prior to analysis, the solutions were ltered
through a 0.5 lm polytetrauoroethylene (PTFE) syringe
lter.
The HPLC analysis was conducted according to the
method shown by Dube et al. (2004) and Darnoko et al.
(2000). A calibration curve was generated from ve
standards: triolein (TG), diolein (DG), monoolein (MG),
methyl oleate (FAME), glycerol. The injection masses were
plotted against the peak area. Each standard was injected
three times at ve dierent concentrations. The calibration curves of the standard solutions showed good linearity. The retention times of the standards are shown in
Table 2. Fig. 3 shows a typical chromatogram of a mixture of standards (note: sample concentrations were
0.548 mg/mL TG, 0.654 mg/mL DG, 0.602 mg/mL MG,
0.642 mg/mL FAME and 0.584 mg/mL glycerol (injection
volume was 2 lL)).
Retention time
(min)
Relative retention
time
Triolein (TG)
Diolein (DG)
Monoolein (MG)
Methyl oleate (FAME)
Glycerol
24.57
25.45
27.12
28.68
30.95
1
1.04
1.10
1.17
1.26
M oilt0 M oiltt
M oilt0
where X was the fractional conversion, Moil(t = 0) was the

original of mass of oil (or TG equivalents in order to account for the presence of any DG or MG) in the reactor.
Moil(t = t) was the mass of TG left in the reactor after the
6 h reaction time.
4. Results and discussion
4.1. Principle of membrane reactor operation
It is known that methanol is only slightly miscible in
canola oil and that temperature has only a slight eect
on this miscibility (Liu, 2004). For all practical purposes,
one could say that the two phases are immiscible. On the
other hand, FAME is entirely miscible in methanol over
a broad range of temperatures. At normal reaction temperatures (e.g. 60 C), FAME and methanol are miscible.
The immiscibility of oil and methanol is at the root of
the mass transfer issues for biodiesel production. However,
in the operation of a membrane reactor, the formation of a
two-phase (emulsied) system is exactly what is necessary.
Thus, the transesterication of TG to FAME is ideally suited to operation in a membrane reactor.
The principle of membrane reactor operation is depicted
in Fig. 4. Due to the immiscibility of canola oil and methanol, and due to various surface forces, the canola oil will
exist in the form of an emulsion; i.e. droplets suspended in
methanol. One can thus envisage the transesterication
occurring at the surface of the canola oil droplets. In the
Table 2
Retention time of standards
Standard
643
80
Diolein
Monoolein
70
Triolein
60
Methyl oleate
Glycerol
mV
50
40
30
20
10
0
22
23
24
25
26
27
28
29
30
Elution time (min)
Fig. 3. HPLC chromatogram of mixture of standards.
31
32
33
644
was used for the determination of compounds in both the

permeate and retentate. A typical HPLC chromatogram
of the retentate is shown in Fig. 5. It is seen that the main
component in the retentate is TG (retention time = 25 min)
or canola oil. Trace amounts of DG (retention time = 26
min) and FAME (retention time = 29 min) also are
evident.
Fig. 6 shows a typical chromatogram of the permeate.
The complete absence of a peak at 25 min indicates that
very high purity FAME was produced by the membrane
reactor.
At the reaction conditions in this study, as mentioned
previously, methanol is only slightly miscible in canola
oil. At the same time, FAME and methanol are miscible.
These physical characteristics are what permit the membrane reactor to separate the FAME from the oil.
4.2. Eect of temperature
Fig. 4. Separation of oil and FAME by a separative membrane.
mV
presence of a permeable membrane, the oil droplets are too

large to pass through the pores of the membrane. On the
other hand, the FAME is soluble in the methanol and,
due to its small molecular size, will pass through the membrane pores along with the methanol, glycerol and catalyst.
The main reaction for producing FAME was shown earlier in Fig. 1. The reaction is reversible, although the equilibrium lies towards the production of FAME and glycerol.
In order to increase the production of FAME, FAME or
glycerol should be removed during the reaction in order
to shift the equilibrium to the product side.
A microporous carbon membrane reactor can selectively
permeate FAME, methanol and glycerol during the transesterication from the reaction zone. The molecule of canola
oil is trapped in droplets forming an emulsion. The droplets cannot pass through the pores of the membrane
because they are larger than the pore size of the carbon
membrane. Results showed that during the reaction,
canola oil did not appear in the permeate side. HPLC
Liu (1994) noted that heating was required for faster

reaction and the reaction time may vary from a few minutes to several hours for a temperature range of 6090 C
for acidcatalyzed transesterication. From our experiments, three dierent reaction temperatures, 60, 65 and
70 C, were selected. Fig. 7 shows the conversion versus
temperature data as a function of acid concentration. At
each acid concentration, an increase in nal conversion
was evident as temperature was increased.
4.3. Eect of catalyst concentration
The catalyst concentration was found to aect the conversion of canola oil to FAME. It is evident from Fig. 7
that an increase in acid concentration served to increase
the conversion of TG to FAME. Based on Fig. 7, one
can see that between 0.5 and 2 wt.% acid concentration
the conversion increased substantially at higher temperatures, but the conversions of 2, 4 and 6 wt.% were not very
dierent (<10% conversion). Thus, concentrations of acid
1500
1400
1300
1200
1100
1000
900
800
700
600
500
400
300
200
100
0
22
23
24
25
26
27
28
29
Elution time (min)
30
31
32
33
Fig. 5. HPLC chromatogram of the phase-separated oil phase of the emulsion in the reactor after 6 h of operation (reaction at 65 C, 2 wt.% acid catalyst).
mV
1500
1400
1300
1200
1100
1000
900
800
700
600
500
400
300
200
100
0
22
23
24
25
26
27
28
29
Elution time (min)
30
645
31
32
33
Fig. 6. HPLC chromatogram of permeate after washing (reaction at 65 C, 2 wt.% acid catalyst).
Table 3
Eect of ow rate on conversion
1
0.9
0.8
Conversion at 6 h
0.7
Expt.
Flow rate
(mL/min)
Temperature
(C)
Conversion via
acid-catalyst (%)
Conversion via
base-catalyst (%)
1
2
3
2.5
3.2
6.1
65
65
65
35
48
64
95
96
96
0.6
0.5
0.4
0.3
0.50%
2%
0.2
4%
0.1
6%
0
58
60
62
64
66
Temperature (C)
68
70
72
Fig. 7. Eect of reaction temperature and acid-catalyst concentration

(linear t for 0.05, 2 and 4 wt.%, t not plotted for 6 wt.%).
beyond 2 wt.% are not necessary at 70 C. In addition, the

reaction was more sensitive to temperature at high acid
concentration.
the acid-catalyzed case shows that the base catalyst provided a much higher conversion, than that of acid catalyst.
Freedman et al. (1984) studied the eect of the type of catalyst on the reaction. It was found that 98% conversion was
observed at 1 wt.% sodium hydroxide. They also found
that greater than 90% of the oil was converted to methyl
esters at 1 wt.% sulphuric acid. In our base-catalyzed
experiments, small amounts of soap were detected in the
wash waters. These were not found in the acid-catalyzed
runs. One possible reason was that the canola oil may have
contained signicant amounts of FFA that were converted
to soaps rather than FAME by the base catalyst. This
would have implications for the use of an acid catalyst
which, despite the slower reaction rate, may provide both
a technological and economic advantage for the use of
lower cost waste feedstock, which contain higher levels of
FFA (Zhang et al., 2003a,b).
4.4. Eect of ow rate

4.6. Membrane material resistance to degradation
The methanol/acid catalyst feed ow rate was set to 2.5,
3.2 and 6.1 mL/min for three separate experiments at
2 wt.% acid concentrations (see Table 3). A signicant
increase in conversion was observed as the ow rate was
increased.
4.5. Eect of base catalyst
The use of a 1 wt.% NaOH catalyst concentration was
tested at dierent ow rates (see Table 3). Comparison to
An important consideration when dealing with high

acid or base catalyst concentration is the life of the
carbon membrane used in the reactor. The carbon membrane was able to resist the high acid and base environments in our experiments. FAME also presents very
strong solvent qualities. After 10 months of operation
and contact with methanol/acid or methanol/base solution,
no tangible evidence of degradation of the membrane was
observed.
646
5. Conclusions
The purication of FAME from a canola oil/methanol/
catalyst reaction mixture poses some important challenges
in the production of biodiesel. This work has demonstrated
that a membrane reactor can be used to alleviate many of
these diculties and successfully carry out the transesterication of triglycerides to FAME.
In the present semi-batch process, a large methanol:oil
ratio was employed. From visual observations, the concentration of FAME in the permeate was not constant as the
reaction progressed. Initially, the permeate was quite
concentrated but as the reaction proceeded, the permeate
concentration decreased. In a continuous process, oil and
methanol would be fed to the reactor at a xed ratio resulting in the continuous production of a concentrated permeate. The present work illustrates that oil and methanol can
readily co-exist in the reactor at a volume ratio of 1:2 without plugging the membrane pores. This allows for the reaction to be carried out in an emulsion where oil and reacted
products can be continuously separated in order to produce a TG-free FAME.
In most commercial processes, as the reaction progresses, the formed FAME will eventually behave as a
mutual solvent for the TG and alcohol phases. Noureddini
and Zhu (1997) have discussed the benets of the formation of a homogeneous alcohol/TG/FAME phase as
FAME is formed in the reaction. As discussed in Section
4.1, maintaining a two-phase system in the membrane reactor inhibits the transfer of TG and non-reacting lipids to
the product stream. One of the benets of producing a
TG-free FAME is a simplication of the often onerous
downstream purication of FAME. This, of course, leads
to the production of high quality FAME. The membrane
reactor allows a phase barrier which limits the presence
of TG and non-reacting lipids in the product. This is highly
desirable in maintaining quality assurance in the production of biodiesel. Maintaining a phase barrier prohibits
the transfer of highly hydrophobic molecules to the product. This provides a limiting barrier in the production of
biodiesel. This parallels the advantages of using distillation
in maintaining product quality in the petroleum processing
industries.
Acknowledgements
The authors thank the support from the Premiers
Research Excellence Award of Ontario and the University
of Ottawa.
References
Anastopoulos, G., Lois, E., Karonis, D., Zanikos, F., Kalligeros, S., 2001.
A preliminary evaluation of esters of monocarboxylic fatty acid on the
lubrication properties of diesel fuel. Ind. Eng. Chem. Res. 40, 452456.
Armor, J.N., 1989. Catalysis with permselective inorganic membranes.
Appl. Cat. 49, 125.
Armor, J.N., 1998. Applications of catalytic inorganic membrane reactors

to renery products. J. Membr. Sci. 147, 217233.
Bam, N., Drown, D.C., Korous, R., Homan, D.S., Johnson, T.G.,
Washam, J.M., 1995. Method for Purifying Alcohol Esters. US Patent
No. 5,424,467, June 13.
Boocock, D.G.B., Konar, S.K., Mao, V., Lee, C., Buligan, S., 1998. Fast
formation of high-purity methyl esters from vegetable oils. J. Am. Oil
Chem. Soc. 75, 11671172.
Boocock, D.G.B., Konar, S.K., Mao, V., Sidi, H., 1996. Fast one-phase
oil-rich processes for preparation of vegetable oil methyl esters.
Biomass Bioenergy 11, 4350.
Canakci, K., Van Gerpen, J., 1999. Biodiesel production via acid catalysis.
Trans. ASAE 42, 12101230.
Coltrain, D., 2002. Biodiesel: is it worth considering. In: 2002 Risk and
Prot Conference, Manhattan, Kansas.
Darnoko, D., Cheryan, M., Perkins, E.G., 2000. Analysis of vegetable oil
transesterication products by gel permeation chromatography. J. Liq.
Chrom. Rel. Technol. 23, 23272335.
Dmytryshyn, S.L., Dalai, A.K., Chaudhari, S.T., Mishra, H.K., Reaney,
M.J., 2004. Synthesis and characterization of vegetable oil derived
esters: evaluation for their diesel additive properties. Biores. Tech. 92,
5564.
Dube, M.A., Zheng, S., McLean, D.D., Kates, M., 2004. A comparison of
attenuated total reectance-FTIR spectroscopy and GPC for monitoring biodiesel production. J. Am. Oil Soc. Chem. 81, 599603.
Freedman, B., Pryde, E.H., Mounts, T.L., 1984. Variables aecting the
yields of fatty esters from transesteried vegetable oils. J. Am. Oil
Chem. Soc. 61, 16381643.
Freedman, B., Buttereld, R.O., Pryde, E.H., 1985. Transesterication
kinetics of soybean oil. J. Am. Oil Chem. Soc. 63, 13751380.
Hayafuji, S., Shimidzu, T., Oh, S., Zaima, H., 1999. Method and
Apparatus for Producing Diesel Fuel Oil from Waste Edible Oil. US
Patent No. 5,972,057, October 26.
Hsieh, H.P., 1991. Inorganic membrane reactors. Cat. Rev.: Sci. Eng. 33,
170.
Hsieh, H.P., 1996. Inorganic Membranes for Separation and Reaction.
Elsevier Science, Amsterdam.
Karaosmanoglu, F., Cigizoglu, K.B., Tuter, M., Ertekin, S., 1996.
Investigation of the rening step of biodiesel production. Energy
Fuels 10, 890895.
Kim, I.-C., Kim, J.-H., Lee, K.-H., Tak, T.-M., 2002. Preparation
and properties of soluble copolysulfoneimide and performance of
solvent-resistant ultraltration membrane. J. Appl. Polym. Sci. 85,
10241030.
Liu, J., 2004. Biodiesel Production from Canola Oil Using a Membrane
Reactor. M.A.Sc. Thesis, Department of Chemical Engineering,
University of Ottawa, Canada.
Liu, K., 1994. Preparation of fatty acid methyl esters for gas-chromatographic analysis of lipids in biological materials. J. Am. Oil Chem. Soc.
71, 11791187.
McBride, N., 1999. Modeling the Production of Biodiesel from Waste
Frying Oil. B.A.Sc. Thesis, Department of Chemical Engineering,
Noureddini, H., Zhu, D., 1997. Kinetics of transesterication of soybean
oil. J. Am. Oil Chem. Soc. 74, 14571463.
Nye, M.J., Williamson, T.W., Deshpande, S., Schrader, J.H., Snively,
W.H., 1983. Conversion of used frying oil to diesel fuel by transesterication: preliminary test. J. Am. Oil Chem. Soc. 60, 1598
1601.
Ripmeester, W., 1998. Modeling the Production of Biodiesel from Waste
Frying Oil. B.A.Sc. Thesis, Department of Chemical Engineering,
Saracco, G., Neomagus, H.W.J.P., Versteeg, G.F., van Swaaij, W.P.M.,
1999. High-temperature membrane reactors: potential and problems.
Chem. Eng. Sci. 54, 19972017.
Saracco, G., Specchia, V., 1994. Catalytic inorganic-membrane reactors:
present experience and future opportunities. Cat. Rev.: Sci. Eng. 36,
305384.

Saracco, G., Versteeg, G.F., Van Swaajj, W.P.M., 1994. Current hurdles
to the success of high temperature membrane reactors. J. Membr. Sci.
95, 105123.
Watanabe, Y., Shimada, Y., Sugihara, A., Tominaga, Y., 2001. Enzymatic
conversion of waste edible oil to biodiesel fuel in a xed-bed
bioreactor. J. Am. Oil Chem. Soc. 78, 703707.
White, L.S., Nitsch, A.R., 2000. Solvent recovery from lube oil ltrates
with a polyimide membrane. J. Membr. Sci. 179, 267274.
Zaman, J., Chakma, A., 1994. Inorganic membrane reactors. J. Membr.
Sci. 92, 128.
647
Zhang, Y., Dube, M.A., McLean, D.D., Kates, M., 2003a. Biodiesel
production from waste cooking oil. 1: Process design and technological
assessment. Biores. Tech. 89, 116.
Zhang, Y., Dube, M.A., McLean, D.D., Kates, M., 2003b. Biodiesel
production from waste cooking oil. 2: Economic assessment and
sensitivity analysis. Biores. Tech. 90, 229240.
Zheng, S., 2003. Biodiesel Production from Waste Frying Oil: Conversion
Monitoring and Modeling. M.A.Sc. Thesis, Department of Chemical
Engineering, University of Ottawa, Canada.
New Biotechnology Volume 25S September 2009
native, environmentally correct, to the replacement of synthetic

thermoplastic. As P(3HB) is an intracellular product, its production may reach high yields through high biomass concentrations.
C. necator is the main microorganism studied to P(3HB) production. This microorganism produces the biopolymer in two phases:
growth and production phase, limited by an essential element. The
exponential growth phase demands a precise control of the bioreactor conditions, mainly the oxygen transference. Aiming to study
a strategy to reach high cellular density, a linear phase was included
between the exponential and production phase to promote a better growth control. Three cultures (A, B and C) were performed
initially with a batch process with the initial carbon source (S) concentration of 40 g L1 to produce approximately 20 gcel L1 (based
on YX/S = 0.5 g g1 ). After that, the linear growth phase of the residual biomass started. The linear growth phase was characterized by
limitation of carbon source during the cell growth, keeping adequate concentration of nitrogen source to avoid the production
phase. At this period, a continuous feeding of 3.0 gSubstrate L1 h1
was established in order to obtain a dXR/dt = 1.5 gXR L1 h1 and
dS/dt = 0 in a reactor (stock solution of glucose/fructose 500 g L1 ),
until reach a total biomass of 30 g L1 . For culture A only the
growth phase (exponential + linear) was studied. For cultures B and
C, the P(3HB) production phase was established after the linear
growth phase. The beginning of nitrogen source limitation (production phase) was marked by the increase of O2 percentage in
the medium. At this moment, carbon source was added to bioreactor to increase the concentration to 40 gSubstrate L1 . The linear
growth phase duration was longer in culture B than in culture C.
The specic XR growth rate (XR ) was investigated in all cultures.
The results showed that the linear growth strategy is easier to maintain growth and reach the desired biomass concentration than the
culture when only the exponential growth was established. It was
also possible to observe that it was necessary to maintain XR values above 0.08 h1 , to allow the P(3HB) synthesis to occur with
greater productivity.
doi:10.1016/j.nbt.2009.06.236
2.6.134
Screening of yeast producer of fructosyltransferase
E. Setsuko Kamimura 1, , G.S. Silva 1 , L.M. Covre 1 , S. Campos
Hossri 1 , A.M. Pinheiro Santos 2
1
Universidade de So Paulo, Pirassununga/SP, Brazil

Universidade Federal de Pernambuco, Recife, PE, Brazil
In the last decades the interest in fructooligosaccharides (FOS)

are increasing due to with their favorable functional properties
such as low calorie and non-carcinogenic sweetener and the main
advantage is a consecutive improvement of intestinal microbial
(bidobacteria), relief of constipation, decrease of total cholesterol and lipid in the serum. Then the researches efforts on the
screening and selection of microorganisms that are able to produce
enzymes with transfructosilation activity for further industrial
application for production of fructooligosaccharides. According
to the literature the Pichia genus has not been studied about the
ABSTRACTS
fructosyltransferase production, then the aim of this work was

the selection of yeasts from Pichia genus that are able to produce fructosyltransferase and compare with Kluyveromyces genus
that are more studied about FOS production. It was applied the
23 experimental design and response surface methodology, the
independent variables are yeast (Pichia pastoris, P. guilliermondi,
Kluyveromyces lactis and K. fragilis), temperature (30 and 35 o C), aeration and dependent variable are the transfructosylation activity
(TA) and hydrolytic activity (HA). It was carried out the fermentation in shaker according to two experimental design 23 , for
Pichia genus and other for Kluyveromyces genus. The results showed
through the fermentation that K. lactis (2.79 U/mL) and K. fragilis
(2.27 U/mL) low values of transfructosylation activity compared
with more higher values got by Pichia pastoris (35.63 U/mL) and P.
guilliermondi (17.17 U/mL). The Pichia genus studied in this work
showed a potential for FOS production.
doi:10.1016/j.nbt.2009.06.237
2.6.135
Membrane reactors for bioethanol production of enzymatically pretreated wheat straw
M.C. Sgua , S.M. Paixo, L. Baeta-Hall, B. Ribeiro, J. Pereira, A.M.
Anselmo, J.C. Duarte
INETI, Lisboa, Portugal
Lignocelluloses from agricultural, industrial and forest residues

account for the majority of the total biomass present in the world
and represent the most prospective feedstock for ethanol production and other value-added products. Biofuels from plant biomass
wastes represent one of the most promising alternative sources
of energy. The economic prospects for ethanol production from
waste biomass is tied not only to petroleum prices but is also
related to costs of fermentable sugars making lignocellulosic wastes
more interesting than crops harvested specically for bioethanol
production.
However, cellulose is a highly recalcitrant substrate for enzymatic degradation because of its physical properties and its
bioconversion is a multi-step process requiring a multi-enzyme
complex for efcient bioconversion into fermentable sugars.
Reduction in cost of bioethanol may also be achieved by efcient
technologies for saccharication which includes the use of better enzyme cocktails and cheaper and more efcient enzymatic
hydrolysis of the lignocellulosic material, which is one crucial part
in the commercialization of lignocellulose based process.
Wheat straw, from the wheat grain production, is an abundant
and low-cost lignocellulosic material which is commonly used as
feedstuffs. In this work, wheat straw biomass pretreated at 205 C
and 215 C (auto-hydrolysis), without hemicellulosic fraction and
containing 55 g/l and 61 g/l of polysaccharides, respectively, was
used for enzymatic hydrolysis using mixtures of commercial
cellulase enzymes (Celluclast 1.5 l) at different conditions. The
inuence of different parameters such as temperature (3560 C),
time (17 days) and enzyme loading (1015 FPU/g polysaccharides) during the bioconversion process were assessed to evaluate
www.elsevier.com/locate/nbt S243
New Biotechnology Volume 25S September 2009
ABSTRACTS
process effectiveness. The optimization criterion is the fermentable

sugar yields, which were analyzed by HPLC.
Zymomonas mobilis and several Saccharomyces cerevisiae strains
were compared for their ability for ethanol production on a MBR
type biorector. A sustainability assessment for the whole processes
was also carried out.
doi:10.1016/j.nbt.2009.06.238
2.6.137
Improving yield and quality of recombinant proteins in
Escherichia coli by stress minimisation and mutant selection
Y. Sevastsyanovich , S. Alfasi, L. Grifths, T. Overton, J. Cole
University of Birmingham, Birmingham, United Kingdom
2.6.136
Bioprocess development to produce rabies virus glycoprotein in bioreactor using insect S2 cells
A. Tonso , D.C. Ventini, P.B. Vieira, R.M. Astray, K.N. Greco, S.A.C.
Jorge, C.A. Pereira
Escola Politcnica da Universidade de So Paulo, So Paulo, Brazil
Drosophila melanogaster Schneider 2 (S2) cells have become increasingly utilized over the past few years for the expression of
heterologous proteins. High levels of protein expression with pharmacological and biotechnological interest can be achieved using
Drosophila Expression System procedure. Insect cell cultures are
easier to handle than mammalian cells, being capable to multiply in monolayers or in suspension at temperatures ranging from
22 to 30 C. The recombinant rabies virus glycoprotein (RVGP) is
an interesting biotechnological product as it is responsible for the
induction of protective immune response against rabies infection.
The aim of this work is the development of a bioprocess using a
bioreactor to produce RVGP. Two cell constructions were tested,
where the gene vectors were constructed in order to be under the
control of metalotionein or actin promoter. The rst is inducible
by CuSO4 and the second is constitutive. The cells were grown
in different conditions in two bioreactors using SF-900II medium
and varying temperature, agitation and aeration. Kinetic studies
of S2 cells were made by analyzing cell growth, RVGP expression,
consumption of glucose and glutamine and production of lactate. Storage tests were performed using different cryoprotectors
(threalose, sucrose and glycerol) in a buffered solution containing
protease inhibitor (PMSF). Cell lyses and RVGP extraction were
made either by mechanic stress (homogenization) followed by
detergent treatment, or just by detergent treatment. Accordingly
to runs with different cell constructions and conditions, the best
results of cell growth and RVGP expression were obtained with
inducible cells in the following culture conditions: at 28 C, 90 rpm
agitation, pitched blade impellers, dissolved oxygen of 10% of air
saturation and 4 gas sparging aeration. The solution with 20% of
glycerol led to the best storage performances and the homogenization associated with detergent was the best lysis-extraction
protocol. The bioreactor cultivation of S2 cells expressing RVGP
is a very promising way to obtain great quantities of the protein,
showing the possibility of performing a well controlled and reproducible process. The storage, lyses and extraction studies are the
preliminary steps to the development of an efcient preservation
and purication procedures.
doi:10.1016/j.nbt.2009.06.239
S244
www.elsevier.com/locate/nbt
The general stress response is strongly induced during recombinant protein production in bacterial hosts. This problem is
particularly severe when T7 promoter-based expression systems
are used, such as Novagens pET vectors that require E. coli BL21 or
its derivatives as production hosts. In a typical production protocol, high concentrations of IPTG are used for induction, resulting
in rapid but transient burst of recombinant gene expression. Due
to the limited folding capacity of a cell, misfolded protein species
begin to accumulate and are deposited into inclusion bodies.
This together with depletion of cellular recourses due to excessive foreign protein synthesis results in growth arrest, induction
of RpoH-dependent stress response, ppGpp-dependent stringent
response, RpoS-dependent stationary phase response, and nally
overgrowth of the culture by plasmid-free, non-productive bacteria. We showed using recombinant E. coli BL21*(DE3) producing
CheYGFP fusion protein that growth arrest is accompanied
by >95% loss of cultivability of the productive population, so
plasmid-free bacteria are able to grow without competition and
by 24 h post-induction comprise up to 8090% of the culture. We
then demonstrated that even challenging recombinant proteins,
such as gonococcal cytochromes could be accumulated successfully by redesigning fermentation conditions that minimised stress
response. These include decreasing the inducer concentration and
keeping low constant temperature throughout the fermentation
in order to avoid both drastic changes in growth rate and stress
from temperature shift upon induction. This method has been
successfully reproduced in shake asks, batch and fed-batch fermentations, in which cell densities up to 30 g dry mass per litre of
culture have been achieved. Using this approach, culture growth
and productivity were sustained for at least 70 h and CheYGFP
yields were typically 2530% of total cellular protein. Unlike
the original production protocol, the majority of this protein (up
to 90%) was found in the soluble cell fraction and showed signicantly higher uorescence indicating correct protein folding.
Flow cytometric analysis also showed that virtually all bacteria in
the nal population accumulated recombinant protein of good
conformational quality and in high yield. Western blotting analysis suggested that the major reason for success was decreased
accumulation of the T7 RNA polymerase. We have used different
techniques, including uorescence-activated cell sorting, for isolation of mutants that are resistant to IPTG-induced stress, but also
accumulate high levels of good quality recombinant protein when
the normal induction protocol is followed. The molecular basis of
this improvement is currently being investigated.
doi:10.1016/j.nbt.2009.06.240

Fischer Tropsch

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fischer Tropsch

Uploaded by

Copyright:

Available Formats

INGENIERO QUMICO

ASIGNATURA: EXPERIMENTACIN EN INGENIERA QUMICA II

MIGUEL NGEL CALVO CARRASCAL

HCTOR CANTALAPIEDRA SIGENZA

1. Definicin de reactor de membrana ....................................................................... 2

2. Tipos de reactores de membrana .......................................................................... 3

3. Usos o aplicaciones de los reactores de membrana .............................................. 4

USO O EMPLEO DE REACTORES DE MEMBRANA EN PROCESOS DE

3. USOS O APLICACIONES DE LOS REACTORES DE MEMBRANA

El hecho de usar un tipo de reactor u otro permite obtener mismas conversiones a

Fig 1. Conversin de etanol en funcin alimentacin

Factores que influyen en el funcionamiento del reactor de membrana

Segn la grfica 2 existe una relacin directa entre la conversin de etanol y la

Fig 2. Conversin de etanol en funcin de alimentacin

Lo primero que se observa es que la conversin se incrementa para ambos reactores

En la grfica 3 se observan tres zonas distintas segn la relacin molar de

Fig. 3. Conversin de etanol en funcin de la alimentacin

En la primera zona, para W/F<3kgcat/(mol/s).

Fig 4. Esquema del proceso

Esto se llevar a cabo para diferentes concentraciones catalticas, flujos de

Fig. 5. Funcionamiento de la membrana

Debido a la inmiscibilidad de la alimentacin y a varias fuerzas superficiales el aceite

Para incrementar la produccin de biocombustible el FAME debe ser removido durante

Hidrlisis enzimtica de la lignocelulosa

Fig 6. Rendimiento frente a tiempo de hidrlisis

Fig 7. Formacin de producto frente a rendimiento

Esta grfica muestra las diferentes formas de aproximacin a un rendimiento elevado

Factores que influyen en el funcionamiento del reactor de membrana

Cantidad de producto formado

Tambin existen factores ms especficos de los que depende el proceso:

[8] Shu J, Grandjean BPA, Van Neste A, Kalaguine S. Catalytic palladium-based

Biotechnology Advances 28 (2010) 407425

Contents lists available at ScienceDirect

Research review paper

Reactor design for minimizing product inhibition during enzymatic

Corresponding author. Tel.: +45 4525 2800.

P. Andri et al. / Biotechnology Advances 28 (2010) 407425

(Bezerra and Dias, 2005). Hence, a certain degree of separate

P. Andri et al. / Biotechnology Advances 28 (2010) 407425

The particular inhibition effect that product inhibition exerts can

In these equations Ecl and Epr represent the concentrations of free

Fig. 1. The effect of the non-competitive classical

E is the concentration of the theoretically available free enzyme in the

Correspondingly, the KI or KIP can be dened as the concentration

P. Andri et al. / Biotechnology Advances 28 (2010) 407425

catalyzed degradation of lignocellulosic biomass is however complicated by:

reduces the initial catalyzed rate of uninhibited enzyme (vP,0) to one

Because vP,0 is given as:

The numerical value of the inhibition constant KI or KIP may thus be

We have recently reported that simplied MichaelisMenten

P. Andri et al. / Biotechnology Advances 28 (2010) 407425

Flow ratea [L/h]

Glucose productivity [g/(L h)]

Diameter (DBR/CSTR) or length (LPFR) [m]

Suspension velocity [m/s]

1.3. Glucose formation rates at different lignocellulose dry matter

namely, 95% (Fig. 3). These values correspond to a volumetric

Reactor volume [m3]

Without glucose inhibition

With glucose inhibition

Fig. 3. Inuence of actual glucose concentration on glucose formation rate at different

P. Andri et al. / Biotechnology Advances 28 (2010) 407425

products such as glucose through a membrane (Alfani et al., 1983;