ww.sciencedirect.com

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 8

Available online at w

ScienceDirect

journal homepage: www.elsevier .com/locate/he

Directing selectivity of ethanol steam reformingin membrane reactors

Maria Anna Murmura b, Michael Patrascu a, Maria Cristina Annesini b,Vincenzo Palma c, Concetta Ruocco c, Moshe Sheintuch a,*

a Technion-Israel Institute of Technology, Department of Chemical Engineering, Haifa 32000, Israelb University of Rome “Sapienza”, Department of Chemical Engineering, Materials and Environment,

Via Eudossiana 18, 00184 Rome, Italyc University of Salerno, Dipartimento di Ingegneria Industriale, Via Giovanni Paolo II 132, 84084 Fisciano, SA, Italy

a r t i c l e i n f o

Article history:

Received 21 January 2015

Received in revised form

26 February 2015

Accepted 5 March 2015

Available online 30 March 2015

Keywords:

Catalytic membrane reactor

Reforming

Hydrogen

Ethanol

Modeling

* Corresponding author. Tel.: þ972 4 8202823E-mail address: cermsll@tx.technion.ac.il

http://dx.doi.org/10.1016/j.ijhydene.2015.03.00360-3199/Copyright © 2015, Hydrogen Ener

a b s t r a c t

In a systemofparallel reversible reactions, separating the correspondingproduct can enhance

a desired reaction. If the same product is produced in several reactions, its concurrent sepa-

ration enhances the reaction with the higher stoichiometry. Here we demonstrate this effect

by separating hydrogen from ethanol during steam reforming in a Pd membrane reactor

packed with a Pt/NieCeO2 catalyst. Interest in ethanol SR stems from the need to produce

ultra-pure H2 from renewables (the energy will be supplied by solar-heated molten salt).

For the conditions tested full conversion (of reaction (1) below) has been achieved with

H2, CO2, CH4 and CO as products. These products can be represented by three reactions

(W ¼ H2O): 1. EtOH 4 CH4 þ CO þ H2 (ethanol decomposition), 2. CH4 þ 2W 4 CO2 þ 4H2

(methane steam reforming, MSR) and 3. CO þ W 4 CO2 þ H2 (water gas shift, WGS).

Separating H2 directs the selectivity towards CO and CO2, resulting also in increased ratio

between CO and CH4 mole fraction.

In general, increasing temperature (613e753 K), pressure (6e10 bar) and introducing

sweep flow (0.5 NL/min N2 for a similar feed rate) led to better separation, to an increase in

selectivity towards CO and CO2 and in hydrogen yield. Increasing pressure and introducing

sweep flow also increased hydrogen recovery. A further increase of sweep gas flow rate (to

1 NL/min) did not result in an appreciable improvement.

The results of this work show that the combination of theNi/Pt catalyst and the Pd

membrane for hydrogen removal produce veryhighvalues of hydrogenyield, despite the low

steam to ethanol ratio and the moderate pressure levels. In particular about 4.5 mol H2/

mol EtOHwere produced at 753 K, feed and sweep flow rates of 0.5 NL/min each in the entire

pressure range examined. A one-dimensional model based on the kinetics of a limited but

realistic set of reactions has been developed to simulate the behavior of the reactor. With a

literature kinetics model and only two adjustable parameters, the permeance and heat

transfer coefficient, a very good agreement with experimental data has been obtained. The

results indicate strong permeance inhibition compared to pure hydrogen measurements.

Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights

reserved.

.(M. Sheintuch).13gy Publications, LLC. Publ

ished by Elsevier Ltd. All rights reserved.

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 85838

Introduction

The worldwide hydrogen demand is increasing, and its use in

Fuel Cells (FCs) is expected to make hydrogen one of the

widely-used fuels in a future energy system, mainly thanks to

its low environmental impact when used in FC systems. On

the other hand, the extensive introduction of hydrogen as

energy vector is limited by the actual lack (and costs) of a

reliable hydrogen distribution infrastructure. This barrier can

be surmounted by the development of systems for decen-

tralized hydrogen production (i.e. close to the end-user). Fuel-

flexibility and the possibility to power the process with re-

newables (solar and biomass) are two additional pre-

conditions for sustainable hydrogen production. The above

considerations are at the basis of the CoMETHy project

(Compact Multifuel-Energy to Hydrogen converter). CoMET-

Hy's general objective is to support the intensification of

hydrogen production processes, by developing a membrane

reformer that can operate with various fuels and that even-

tually will be heated by solar-heated molten salt [1]. An

electrically-heated lab membrane reformer was tested in the

Technion with methane SR [2], and this work extends that

study to ethanol SR using the same catalyst, membrane and

system.

Hydrogen production from biomass derived oxygenates

has attracted interest for its potential application in fuel cells.

Bio-fuels for the production of hydrogen could bring signifi-

cant environmental benefits, as the carbon dioxide produced

is consumed in biomass production giving a CO2 neutral en-

ergy supply. Steam reforming of bioethanol has been widely

investigated over supported transition and noble metal cata-

lysts [3,4].

Hydrogen production by SR of ethanol often encounters

several problems, which are addressed in the CoMETHy

project:

- High-purity hydrogen is required for its application in fuel

cells: the most promising FCs, based on PEM, require low

levels of CO; hydrogen separation by a Pd membrane

should satisfy this condition.

- Steam reforming is highly endothermic and requires heat

supply; in this project the reformer is heated by solar

energy.

- The reaction is accompanied by the formation of various

by-products, which greatly affect the selective production

of hydrogen; as we show in this article, hydrogen separa-

tion will shift the product distribution in the desired di-

rection, i.e., toward the reaction that produces most

hydrogen.

- The reaction is limited by equilibrium; hydrogen separa-

tion will shift the equilibrium conversion.

The ethanol steam reforming (ESR) reaction is:

C2H5OH þ H2O $ 2CO þ 4H2 (1)

The conversion of CO to CO2 is considered through the

wateregas shift (WGS) reaction:

CO þ H2O $ CO2 þ H2 (2)

The overall ESR reaction, which refers to complete con-

version of CO to CO2, is therefore:

C2H5OH þ 3H2O $ 2CO2 þ 6H2 (3)

Aside from these desired reactions, several others may

take place [5].

C2H5OH $ C2H4O þ H2 (4)

C2H4O $ CH4 þ CO (5)

C2H4O þ H2O $ 2CO þ 3H2 (6)

C2H5OH $ C2H4 þ H2O (7)

2C2H5OH $ C3H6O þ CO þ 3H2 (8)

C2H5OH þ 2H2 $ 2CH4 þ H2O (9)

C2H5OH $ 0.5CO2 þ 1.5CH4 (10)

CH4 þ H2O $ CO þ 3H2 (11)

2CO $ CO2 þ C (12)

CO2 þ 4H2 $ CH4 þ 2H2O (13)

Achieving the desired product is usually obtained by a

proper choice of catalyst and support (see below) and oper-

ating conditions. Selectivity may be further directed through

the use of membrane reactors, which shift the equilibrium in

the direction of the selectively permeating component

(hydrogen in this case).

Recently, catalysts containing more than one active spe-

cies have also been investigated because of their significantly

different catalytic properties with respect to either of the

parentmetals [6,7]. A synergic effect of Pt addition to Ni-based

catalysts was shown to improve the activity of the non-noble

metal towards hydrogenation reactions, resulting in

decreased coke formation rates with respect to the ones

observed over monometallic Ni-catalysts [8]. When Pt and Ni

are supported on CeO2, by depositing the non-noble metal on

the support surface earlier than the noble one, as Pt is directly

available at the gasesolid interface, ethanol adsorption is

favored and a better agreement with thermodynamic data in a

regular fixed bed was observed [9].

Table 1 e Summary of experimental conditions and results previously reported in literature.

Ref T [K] P [bar] S/E Sweep to feedflow rate ratio

Inlet flow rate[molEtOH/kgcatmin]

EtOH conversion [%] H2 yield

[10] 598e673 5e14 6 0 1.67 98.6e100 3.2e3.7

[11] 573e673 1.3 3e9 0.5e1.4 0.83 16e100 0.6e3.3

[12] 673e723 1.5e2 8.4e13 0.9e3.7 0.04e0.64 100 0.2e5

[13] 673e723 1.5e2 8.4e13 0.9e3.7 0.04e1 100 0.3e3.6

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 8 5839

Conversion and product distribution in ESR in membrane

reactors (MRs) depend on various operating conditions:

- Increasing temperature will increase hydrogen partial

pressure and its permeating flux, leading to higher con-

version and selectivity.

- While increasing pressure has detrimental effect on con-

version in a closed system at equilibrium, it will have an

advantageous effect on conversion when significant

hydrogen separation is achieved, as follows from

stoichiometry.

- At low temperatures, increasing the S/C ratio (to avoid

coking) leads to an increase in H2 and CO2 selectivity, while

decreasing CO and CH4 selectivity. Hydrogen recovery de-

creases due to dilution effects with increasing S/C ratios.

Table 1 summarizes the main operating parameters and

results of ethanol steam reforming in membrane reactors

previously reported in literature. Domı̀nguez et al. [10] used a

PdeAg membrane for ESR on cobalt talc (Co3[Si2O5]2(OH)2).

Studies conducted on a PdeAgmembrane reformer with a Ru-

based catalyst at 1.3 bar [11] showed that increasing temper-

ature shifts the reactions toward the production of CO2 and

CH4. In a study over a metal foil membrane, using Ru, Pt, Ni

based catalyst [12,13] positive effect of increasing the reten-

tate pressure was observed. In these last studies the effect of

co-current vs. counter-current sweep flow was also investi-

gated. None of these works included the development of a

model to describe the experimental results obtained.

In the present work, experimental data collected on low-

temperature (�753 K) ethanol steam reforming in a mem-

brane reactor, equippedwith a high permeance Pdmembrane,

are presented and analyzed, with a particular focus on the

operating conditions which maximize selectivity towards

carbon dioxide and hydrogen. Full ethanol conversion is

achieved under conditions studied and the problem is reduced

to competition between carbondioxide and methane produc-

tion. High hydrogen yields and recoveries have been obtained

even when working with stoichiometric steam to ethanol ra-

tios (S/E¼ 3), moderate pressures andwith low ratios of sweep

gas to feed gas flow rates. Previous works (Table 1) used high

S/E ratios (6e13) and obtained high conversion due to dilution.

No deactivation was detected even when running with a

stoichiometric S/C ratio. This work, in a high permeance

membrane, also presents better selectivities and hydrogen

separation, and higher throughputs, than those presented in

previous works. Furthermore, the work developed and tests a

simple model that described well the results and can be used

for further optimization; it shows that results are insensitive

to catalytic activity.

The structure of the paper is as follows: thermodynamic

equilibrium analysis of a limited but realistic set of reactions is

presented (Section Thermodynamic analysis), followed by a

description of the experimental system and results obtained

(Section Experimental activity). A one-dimensional model,

developed to simulate the behavior of the reactor, is also

presented and analyzed in Section Membrane reactor

modeling.

Thermodynamic analysis

Thermodynamic analysis will provide a limit to the system

operation. Kinetic tests previously carried out on the catalyst

formulation employed in the present work have shown that

the system may be fully described by considering the

following reactions: ethanol dehydrogenation (4), acetalde-

hyde decomposition (5), modified ethanol decomposition (10),

acetaldehyde steam reforming (6), wateregas shift (2), and

CO2 methanation (13) [5]. Since no acetaldehyde formation

was noticed in the course of the present work, reactions (4)

and (5) can be combined and (6) may be disregarded. Of the

four remaining reactions only three are linearly independent,

the systemmay therefore be studied by considering only three

reactions. Table 2 summarizes the reactions considered, along

with their DH0 and DG0. The same three reactions can be

represented in other linear-dependent ways, leading to same

results.

The equilibrium conditions of the three reactions consid-

ered yield:

KEDðTÞ ¼ pCOpCH4pH2

pEtOH¼ P2

TOT

yCOyCH4yH2

yEtOH(14)

KMSRðTÞ ¼pCO2

p4H2

p2H2O

pCH4

¼ P2TOT

yCO2y4H2

y2H2O

yCH4

(15)

KWGSðTÞ ¼ pH2pCO2

pCOpH2O¼ yH2

yCO2

yCOyH2O(16)

A preliminary thermodynamic analysis maps hydrogen

partial pressure in a closed system as a function of total

pressure and temperature (Fig. 1, S/C ¼ 3) while considering

the three reactions mentioned above. Hydrogen separation

occurs if its partial pressure in the reactor side exceeds that in

the permeate side. Increasing pressure leads to a less than

linear increase ofhydrogenpartial pressure, becausehydrogen

concentration decreaseswhile the total pressure increases. To

carry out low-temperature ESR efficiently, taking advantage of

the entire membrane surface, it is therefore essential to

consider the use of sweep gas in the permeate side.

Table 2 e Reactions accounted for in ethanol steam reforming.

Reaction name Reaction DH0 [kJ/mol] DG0 [kJ/mol]

1 Ethanol decomposition (ED) C2H5OH $ CH4 þ CO þ H2 �49.4 �355.6

2 Methane steam reforming (MSR) CH4 þ 2H2O(g) $ CO2 þ 4H2 164.6 113.3

3 Water gas shift (WGS) CO þ H2O(g) $ CO2 þ H2 �41.2 �28.8

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 85840

Assuming reaction (1) goes to completion the equilibrium

selectivity ratio (s.r.) between CO and CH4 can be expressed as:

s:r:≡yCO=yCH4 ¼ KMSRðTÞ=ðKWGSðTÞP2TOTÞ$ðyH2O=y3H2Þ. If H2 is

removed during reaction it will direct the selectivity towards

CO and CO2 but the ratio yCO/yCH4 will increase by increasing

the conversion of MSR to a greater extent than the conversion

of WGS reactions. This is evident by the (yH2)-3 dependence in

the above expression. Recall that yH2O also varies, and in a

stoichiometric feed (as in this work), yH2O ~ yCO þ 2yCH4.

Considering that a different subset of reactions may take

place in correspondence of the inlet section (e.g.,

EtOH þ 3H2O $ 2CO2 þ 6H2, which goes to complete conver-

sion) suggests that the hydrogen driving force will allow for

some transport across the membrane even at temperatures

lower than 700 K in the absence of sweep.

Experimental activity

Experimental setup

Experimental activity, designed to test ethanol steam

reforming in a laboratory scale membrane reactor, has been

carried out in the Chemical Reactor Engineering and Envi-

ronmental Catalysis Laboratory of the Technion. The system

employed is the one used for methane SR [2] after some

modifications. The 40 cm long membrane reactor (Fig. 2a)

consists of a tube-and-shell structure. Reactions take place in

Fig. 1 e Equilibrium partial pressure of hydrogen (in bar) as

a function of temperature and pressure for feed of S/E ¼ 3.

Gray area indicates feasible conditions for non-sweep

operation assuming equilibrium compositions.

the shell side, packed with a structured foam catalyst to

expedite radial heat transfer. Hydrogen permeates through

the Pd membrane (the inner reactor wall, having a 14 mm

outer diameter) into the tube-side, where a sweep gas may

flow (counter-currently). Reactor characteristics have been

summarized in Table 3.

In the present work, nitrogen was used as a sweep gas;

however, it is likely that steam will be used in the pilot scale

application. This will allow an easy separation of hydrogen

from the permeate stream.

The stainless-steel reactor is heated from the outside

(40 mm outer diameter) by four concentric electric heaters,

shown as transparent bodies in Fig. 2a. Each heater is

controlled independently to maintain a set temperature,

measured on the outer wall of the reactor, in the midpoint of

the heater body. A fifth heating element is placed at the

entrance of the reactor. This last heating element acts as a

vaporizer and its temperature is set to 300 �C throughout all

the experiments. Four additional thermocouples are located

inside the membrane, in the same axial position as the outer

control thermocouples. Pipeline elements were sealed using

Hamlet® Let-Lok Tube Fittings (Double Ferrule).

The effluent from the reactor is cooled, to allow the

condensation of unreacted ethanol and steam, and sent to a

silica gel column, where any residual vapor is absorbed. The

dried effluent is then sent to an analyzing system, comprised

of IR analyzers for CH4, CO2, and CO (Edinburgh Instruments

Ltd.) and a gas chromatograph (Trace GC Ultra, Thermo-

Scientific). Either the permeate or the retentatemay be sent to

the analyzing system. The liquid product has been collected

and analyzed at the end of each experiment, showing that it

only contained unreacted water. This indicates that full

ethanol conversion has been achieved and that no liquid by-

products have been formed. This result is in line with the

observations of Palma et al. [5], who showed that with the

catalyst tested, acetaldehyde is present only for very short

contact times and ethanol is fully converted almost

immediately.

The reactor is packed with structured foam catalyst, spe-

cifically prepared in the framework of the CoMETHy project

from cooperative work of the Fraunhofer Institute for Ceramic

Technologies and Systems, IKTS (Germany) and the ProCEED-

lab of the University of Salerno (Italy).

The structured catalyst is made of a CeO2-washcoated SiC

open cell foam (see Fig. 2b), prepared by IKTS and activated

with the bimetallic (Pt(3)Ni(10)) formulation by the University of

Salerno. This formulation was specifically selected on the

basis of the considerationsmentioned in Section Introduction.

Ni and Pt catalysts promote ethanol steam reforming when

supported on ceria, which is expected to diminish deactiva-

tion due to coking and promote catalyst dispersion [14]. SiC

open cell foams have been chosen as the structured carried for

Fig. 2 e a) Detailed drawing of the membrane reactor; b) Structured foam.

Table 3 e Reactor characteristics.

Bed and membranelength [m]

Reactordiameter [m]

Inner reactor diameter(membrane outer diameter) [m]

Mass of catalyts þ structuralsupport [g]

0.4 34 � 10�3 14 � 10�3 200

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 8 5841

their good heat transfer characteristics, and to minimize

pressure drops [15].

The membrane tube (Hysep©-technology), provided by the

Energy Research Centre of the Netherlands (ECN), is made of a

4e5 mm palladium layer, deposited by electroless plating on

the outside of a porous alumina tube provided with two thin

porous alumina layers to decrease the outer pore dimension

to less than 0.15 mm. The membrane is capped with ECN

compression seals on both sides. Membrane permeance was

measured by the producer in pure hydrogen at various tem-

peratures and pressures. The permeance varied with tem-

perature according to an Arrhenius-type law (Em ¼ 11 kJ/mol,

Am ¼ 5.6 mol/(m2 s bar0.5), see Eq. (21)) and hydrogen flux

through the membrane followed Sievert's law. The experi-

mental conditions tested in the present work have been

summarized in Table 4.

Catalyst preparation

Starting from the washcoated foam, the active species depo-

sition (Pt and Ni) was performed at University of Salerno by

multiple cycles of wet impregnation with an aqueous solution

of the metal precursor (C4H6O4Ni∙4H2O or PtCl4, both from

Aldrich), at 60 �C for 1 h; the sample was then dried at 120 �Cfor 2 h and calcined for 1 h at 400 �C for Ni deposition and at

550 �C for Pt deposition. The two different temperatures were

selected on the basis of thermo-gravimetric analysis results.

This impregnation-drying-calcination procedure was

Table 4 e Experimental conditions.

T [K] Retentatepressure [bar]

Permeatepressure [bar]

613e753 6e10 1

repeated the number of times necessary to reach the desired

amount of metal (3 wt% for Pt and 10 wt% for Ni), considering

the solubility of the precursor salt. After the above cycles, the

sample was calcined at 600 �C for 3 h, to stabilize it. The whole

procedure (impregnation-drying-calcination steps and final

calcination) is repeated for each active species to be deposited

on the washcoated foam.

Experimental results

In all experiments conducted here complete conversion of

ethanol was achieved with CO, CO2, CH4, and H2 as the only

products. Permeate purity has beenmeasured, and impurities,

mainly methane and carbon dioxide, were always found in

concentrations < 0.5% ( ~ 0.2% CH4, ~ 0.3% CO2). These mea-

surements could be partly affected by the presence of residues

in the pipeline, which was used to measure the retentate

current as well. A summary of the experimental conditions

and main results obtained has been reporte in Table 5.

Experimental results have been expressed in terms of

hydrogen recovery and yield:

Hrec ¼ H2 permeatedtotal H2 produced

Hyield ¼ total H2 producedethanol in feed

(17)

where the maximum values obtainable are equal to 1 and 6,

respectively; and in terms of selectivity towards carbon-

containing species:

Si ¼ moles of i produced2ðmoles of ethanol convertedÞ100 i ¼ CO2;CO;CH4 (18)

Total feedflowrate [NL/min]

Sweep flowrate[NL/min]

S/E

0.5e1 0e1 3

Table 5 e Summary of experimental conditions and results in present work.

T [K] P [bar] S/E Sweep to feedflow rate ratio

Inlet flow rate[molEtOH/kgcatmin]

EtOH conversion [%] H2 yield

613e753 6e10 3 0e2 0.03e0.05 100 0.3e4.8

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 85842

If complete ethanol conversion is assumed, the selectivity

of each species may be determined by:

Si ¼ yi

yCO2þ yCO þ yCH4

100 i ¼ CO2;CO;CH4 (19)

In the figures that follow, product selectivity has been

compared to its value at equilibrium calculated based on feed

composition and outlet temperature.

Selectivity to CO was found to be small under all condi-

tions. Better separation of hydrogen, achieved by increasing

temperature, decreasing flow rate and increasing sweep gas

flow rates, will shift the reaction toward CO2 and suppress

selectivity to methane.

Hydrogen separation may lead to faster coking by reactions

like (12). In this study catalytic activity seems to have remained

stable during the 200 h of operation with ESR.

Influence of temperatureIn the absence of sweep flow, hydrogen separation is small and

theproductdistribution is similar to that atequilibrium(Fig. 3a).

Increasing temperature leads to small increase in selectivity

towards CO and CO2 and a decrease in selectivity towards CH4,

mainly due to higher methane steam reforming and reverse

WGS conversions. This situation also leads to an increase in

hydrogen yield (Fig. 4a). Deviations from equilibrium become

significant only above 703 K, but in this range conversion to CO

cannot be ignored. Hydrogen recovery decreases with

increasing temperature, despite increasing membrane per-

meance (Fig. 4a). This occurs because hydrogen production in-

creases, but its flow across themembrane is limited by the high

partial pressure of hydrogen in the permeate.

When a sweep flow rate of 1 NL/min is used (Fig. 3b),

hydrogen separation is more significant, and selectivity to-

wards CO2 and CO increases with increasing temperature

reaching 80% and 4%, respectively, at 750 K, whereas

Fig. 3 e Product selectivity as a function of temperature at P ¼ 8

1 NL/min (b), and its comparison with equilibrium values.

selectivity towards CH4 decreases to 15%. The positive effect

of hydrogen separation is evident in Fig. 4b: in the entire

temperature range hydrogen recovery is very close to 1, and

the yield increases steadily with temperature. Under these

conditions, the partial pressure of hydrogen in the permeate is

reduced, allowing for efficient separation and significant de-

viation from equilibrium.

Influence of pressureIncreasing pressure may have contrasting effects on the

hydrogen-producing reactions, and therefore on the overall

product distribution: the ethanol steam reforming process

results in an increase in the number of moles, low pressures

therefore shift the equilibrium conversion towards the prod-

ucts; however, a high retentate to permeate pressure gradient

improves hydrogen separation. As separation is promoted the

equilibrium of the reaction moves further towards the prod-

ucts. Fig. 5 shows the influence of pressure on the selectivity

of carbon-containing species at 753 K, with a feed flow rate of

0.5 NL/min, and for two sweep gas flow rates. In the absence of

sweep flow (Fig. 5a), increasing pressure leads to an increase

in CO2 selectivity and a decrease in CH4 selectivity, indicating

that permeation has a more significant effect than thermo-

dynamics. Under these conditions, hydrogen recovery indeed

increases with pressure, as may be seen by the results shown

in Fig. 6a. The same may be said for the total hydrogen yield

(Fig. 6a), which is favored by the removal of hydrogen itself. A

similar trend in CO2 and CH4 selectivities is observed when

working with sweep gas, even though the effect of pressure

appears to be much less significant. Fig. 6b shows that under

these conditions, hydrogen recovery is very close to 1 even

when working at 6 bar, indicating that hydrogen removal has

already reached its maximum value. The total hydrogen yield

is also unaffected by pressure (Fig. 6b), which indicates that

the system has reached its maximum performance and

bar, F ¼ 0.5 NL/min, without (a) and with a sweep flow of

Fig. 4 e Hydrogen recovery and yield as a function of temperature at P ¼ 8 bar, F ¼ 0.5 NL/min, without (a) and with a sweep

flow of 1 NL/min (b).

Fig. 5 e Product selectivity as a function of pressure at Tw ¼ 753 K, F ¼ 0.5 NL/min, without (a) and with a sweep flow of

0.5 NL/min (b), and comparison with equilibrium values.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 8 5843

further hydrogen production may only be obtained by

increasing the operating temperature or sweep flow.

Influence of sweep gas to feed flow rates ratioThe effect of improving separation, without acting on other

phenomena, may be better understood by studying the

Fig. 6 e Hydrogen recovery and yield as a function of pressure a

flow of 0.5 NL/min (b).

influence of sweep gas flow rate. Increasing sweep flow rate

(Fig. 7) leads to better selectivity towards CO2 and CO and

decreases that of CH4, due to H2 separation, which favorsWGS

and methane reforming to carbon dioxide. In the presence of

sweep gas, CO2 production exceeds CH4 production. The MSR

reaction is more heavily influenced by hydrogen removal

t Tw ¼ 753 K, F ¼ 0.5 NL/min, without (a) and with a sweep

Fig. 7 e Product selectivity as a function of sweep flow rates at P ¼ 8 bar, Tw ¼ 703 K feed flow rate 0.5 NL/min (a) and 1 NL/

min (b) and comparison with equilibrium values.

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 85844

compared to the WGS reaction, as the stoichiometric ratio

between hydrogen and methane in MSR is four, whereas only

1 mol of H2 is produced for every mole of CO in WGS. The di-

rection of selectivity towards CO is better observed in Fig. 8,

where the exit ratio of CO andCH4mole fraction (which equals

that of the respective selectivity ratio) is compared to the

closed system equilibrium values. As sweep flow (and conse-

quently the permeate flow) increases, the MSR is enhanced

compared toWGS,which in turn increases CO content relative

to CH4 by as much as an order of magnitude. The influence of

sweep gas is more pronounced when a lower feed flow rate is

used, because the higher residence time allows a better

hydrogen separation.

Membrane reactor modeling

The behavior of the membrane reactor has been described

with a one-dimensional pseudo-homogeneous model, with

axial dispersion of mass and heat, in which the following

main assumptions have been made:

Fig. 8 e Selectivity ratio between CO and CH4 directed by H2

removal. Experimental measurements of retentate

composition compared to the closed system equilibrium

values. P ¼ 8 bar, S/E ¼ 3.

- Ideal gas behavior;

- Negligible axial pressure drop;

- Negligible radial gradients

� radial temperature gradients were measured to be small;

� effects of concentration polarization were estimated to

be small, based on an analysis carried out for methane

steam reforming in the same reactor [2], using a correc-

tion reported in Ref. [21]. The correction depends on the

reactor's radius, hydrogen diffusivity and membrane

permeance, the conclusions reached in Ref. [2] are

therefore extendable to the present work;

- Hydrogen permeation through Pd-based membranes in-

volves the steps of dissociative adsorption, subsurface

penetration, diffusion through the membrane and the in-

verse steps on the other side.When the rate-limiting step is

hydrogen atom diffusion through the membrane, as in the

case of Pd-based membranes down to 1 mm [16], perme-

ation in pure hydrogen follows Sievert's law (eq. (20)).

When considering a gaseousmixture, the same law applies

with a membrane permeance that depends on co-

adsorption of reactants and products:

JH2¼ Qm

� ffiffiffiffiffiffiffiffiffiffiPyH2

p �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPSWySW

H2

q �(20)

where membrane permeance is evaluated as:

Qm ¼ QAm exp

��Em

RT

�(21)

andQ is a parameter accounting for permeance inhibition due

to competitive adsorption of reactants and products.

Model equations

The material balance equations are adapted from Ref. [2],

applying them to the components and reactions of interest:

εrgvyi

vt¼ �1

AO

�vFvz

yi þ vyi

vzF

�þ Deffrg

v2yi

vz2þX3j¼1

ai;jrjð1� εÞrs; isH2

(22)

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 8 5845

εrgvyH2

vt¼ �1

A

�vFvz

yH2þ vyH2

vzF

�þ Deffrg

v2yH2

vz2

Table 6 e Pre-exponential factors and activation energiesof the kinetic constants.

Reaction k0 Ea [kJ/mol]

Modified ethanol

decomposition

7.94 � 104 mol/(gcat s bar) 104

Water gas shift 2.77 � 108 mol/(gcat s) 74

CO2 methanation 7.65 � 10�1 mol/(gcat s bar4) 8

O

þX3j¼1

aH2 ;jrjð1� εÞrs �pdm

AOQm

� ffiffiffiffiffiffiffiffiffiffiPyH2

p �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPSWySW

H2

q �

(23)

rgvySW

H2

vt¼ 1AI

vFSW

vzySWH þ vySW

H2

vzFSW

!þpdm

AIQm

� ffiffiffiffiffiffiffiffiffiffiPyH2

p �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPSWySW

H2

q �(24)

Effective dispersivity has been considered to be the same

for all species, and equal to 1.7 � 10�4 m2/s.

The following boundary conditions apply:

Fjz¼0 ¼ F0; FSW��z¼L

¼ FSW0 ;

yCH4

���z¼0

¼ yCO2

���z¼0

¼ yCO

��z¼0

¼ yH2

���z¼0

¼ 0;

yEtOH

��z¼0

¼ 11þ S=E

; yH2O

���z¼0

¼ S=E1þ S=E

; ySWH2

���z¼L

¼ 0

vyi

vz

����z¼L

¼ 0

The heat balance is:

�rcpeff

vTvt

¼ kax:effv2Tvz2

� F0

A

�c0pg � cp;H2

FH2;perm

F0

�vTvz

þ rsð1� εÞX3j¼1

��DHj

rj � hw

pDA

ðT� TwallÞ (25)

where hw is wall to catalyst heat transfer coefficient, Tw is the

controlled wall temperature, and FH2 ;perm is the hydrogen

permeation rate. The term:

FH2 ;permcp;H2

vTvz

accounts for heat loss from the gas flow associated with

hydrogen permeation.

The boundary conditions are:

NgCpg

�T0 � Tjz¼0

¼ �kax;effvTvz

����z¼0

vTvz

����z¼L

¼ 0

Although, axial mass dispersion is usually ignored in

membrane reactor studies, the incorporation of axial con-

ductivity is essential for this non-isothermal system.

A non-stationarymodel has been developed to simplify the

numerical solution procedure of this boundary-value prob-

lem. The system of partial differential equations has been

solved using the PDE Toolbox of Matlab®.

Model parameters

The kinetic expressions used in the model for the ethanol

decomposition (Eqn. (10)), wateregas-shift (Eqn. (2)) and

methanation (Eqn. (13)) reactions are those reported by Palma

et al. [5], obtained by studying the kinetic activity of the

catalyst used in the present work in an ethanol steam

reforming reactor operating at different contact times be-

tween temperatures of 523 and 873 K. The rate equations have

been re-written here for clarity:

rED ¼ kED

�pEtOH � 1

kEDp0:5CO2

P1:5CH4

�(26)

rWGS ¼ kWGS

KadsCOKadsH2O

�pCOpH2O � 1

KWGSpCO2

pH2

��1þ KadsCOpCO þ KadsH2O

pH2O þ KadsCO2pCO2

�2 (27)

rCO2M ¼ kCO2

M

�pCO2

p4H2

� 1KCO2

MpCH4

p2H2O

�(28)

The adsorption constants KCO, KCO2, and KH2O have been

considered constant and equal to their average value in the

range 300e500 �C, as suggested by Palma et al. [5] in their ki-

netic studies. The pre-exponential factors and activation en-

ergies of the kinetic constants are reported in Table 6 [5].

The effective axial conductivity has been evaluated

through the following correlation:

kax;eff ¼ 13ksð1� εÞ (29)

where ks is the intrinsic conductivity of the solid, equal to

50 W/mK and ε is the bed porosity, equal to 0.85.

The wall heat transfer coefficient has been evaluated

through a ChiltoneColburn type correlation:

jH ¼ Nuw

Pr1=2Re¼ aReb�1 (30)

A thermal characterization of the solid foam has been

previously carried out, and the value of the parameter b has

been determined to be 0.29 [17]. The value of a has been ob-

tained in the course of the present work by fitting experi-

mental data and is equal to 24.1.

A complete list of parameters used in the model, along

with their meaning and value, is reported in the List of Sym-

bols section at the end of the paper.

Model results

Fitting the experimental results presented here (conditions

summarized in Table 2) with themodel described above, using

thermodynamic and kinetic parameters from the literature,

and only two adjustable parameters, has shown good agree-

ment (Fig. 9). However, the estimated permeance at 723 K is

about 12 Nm3h�1m�2bar�0.5. This value should be compared

with the permeance of the membrane used in the present

work that was measured by the producer (ECN) to be about

75 Nm3h�1 m�2 bar�0.5 at 723 K in pure hydrogen (Em ¼ 11 kJ/

mol and Am ¼ 5.6 mol/(m2sbar0.5)). Hydrogen flow in pure

hydrogen and mixtures was found to follow Sievert's law,

indicating that the limiting step in its permeation is the

diffusion of atoms through the bulk of themembrane. This is a

Fig. 9 e Comparison between experimental (points) and

calculated (lines) molar fractions on dry basis as a function

of temperature at P ¼ 9 bar, 0.5 NL/min feed flow rate,

sweep gas flow rate 0 NL/min (a), 0.5 NL/min (b), and 1 NL/

min (c).

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 85846

significant inhibition (Q ¼ 0.15.) and we try to understand its

sources. No attempt was made to fit the data using an inhi-

bitionQ function that depends on composition, as that would

involve adjusting too many parameters. Furthermore, gas

composition in proximity of the entrance section is not known

precisely.

The decrease in apparent permeancemay be due to several

reasons:

(i) In wide reactors the incorporation of high-permeance

membranes leads to the hydrogen concentration at

the membrane wall to be much lower than in the bulk,

while the concentration of other components is higher.

This phenomenon is known as concentration-

polarization. In narrow reactors, such as the one used

in the present study, the effect of concentration-

polarization is not negligible, but cannot account for

the large decline (see below);

(ii) Co-adsorption of reactants and products on the Pd-

surface may lead to inhibition: While there are many

studies of the inhibiting effects of CO [18] the effects of

methane adsorption is contradictory, while computa-

tional (DFT) studies show negligible effects due to CO2,

CH4 or H2O adsorption [19]. The relatively small con-

centrations of CO make this explanation questionable

as well;

(iii) Surface reactions (such as reverse-WGS) may cause

adsorption of CO on the Pd surface.

An inhibition effect similar to the one noticed here was

previously observed in a study ofmethane steam reforming in

the same membrane reactor [2] and an attempt was made to

discriminate between various sources of inhibition. It was

argued that the strong inhibition cannot be explained only by

concentration polarization or by competitive adsorption by

gas-phase species. The former effect was ruled out by

approximate criteria and corrections to the 1-D model ac-

counting for transversal effects, which were developed for

hydrogen separation without [20] or with [21] reaction. The

latter effect was ruled out using partial information on

adsorption properties from the literature: CO inhibition is

strong but its concentration is small. On the other hand, the

inhibition can be accounted for by competitive adsorption by

species produced on the surface. This issue is hardly

addressed in the literature, but some indications do exist: Li

et al. [22] found that both CO and H2O reduce hydrogen

permeation through a Pd/stainless steel membrane at 653 K.

Permeation tests with binary H2/CO and H2/H2O mixtures of

the same concentrations of CO and H2O showed that steam

had a stronger inhibition effect than CO on hydrogen perme-

ation. Hulme et al. [23] found significant decline of hydrogen

permeability through a PdeAg/VeNi alloy membrane in

presence of CO2 at 473e673 K (60% decline with 30% CO2 and

stronger at lower temperatures). These results can only be

explained by higher CO surface concentrations than those

expected from gas phase composition.

Fig. 10 shows the calculated temperature profile along the

reactor axis when the wall temperature has been set at 753 K

and the reactor operates at 8 bar with a 1 NL/min feed flow

rate and 1 NL/min sweep gas flow rate. The inlet temperature

was equal to 573 K: the calculated temperature increases

rapidly in proximity of the inlet, due to the combined effect of

the exothermicity of the ethanol decomposition reaction and

the heat transfer from the reactor wall.

Temperature tends to decrease towards the final part of the

reactor. This behavior may be attributed to the fact that when

the sweep gas is sent counter-currently with respect to the

reactor flow, the hydrogen partial pressure in the permeate

side is close to zero in correspondence of the exit section of

the reactor. Hydrogen permeance through the membrane

shifts the carbon dioxide methanation reaction backwards,

absorbing heat and causing a reduction in temperature.

Fig. 10 e Calculated (continuous) and measured (points)

temperature along reactor axis for Tw ¼ 753 K, P ¼ 8 bar,

1 NL/min feed flow rate, sweep gas flow rate 1 NL/min.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 8 5847

Conclusions

The present work showed that in a system of competing

reversible reactions, the desired one can be accelerated by

separating one of its products, hydrogen in this case. This

novel idea has the potential of suggesting new designs for

improving selectivity.

It was shown experimentally that pure hydrogen can be

produced by ethanol steam reforming in an integrated cata-

lytic membrane reactor, with complete conversion, good

selectivity, and with no appreciable catalyst deactivation. As

explained in the Introduction amembrane reformer heated by

solar energy through molten salt represents an interesting

process for renewable hydrogen production, which can be

converted to electricity via FC. The possibility of working with

low amounts of water and relatively low pressures reduces

the energetic cost of conducting ethanol steam reforming.

Using steam instead of nitrogen as sweep gas will also make

the overall process easier, by allowing separation of the

permeate side current by simple condensation of water.

Process optimization can proceed with the mathematical

model that was developed; the model uses only two adjust-

able parameters: a permeance inhibition factor, which was

found to be significant, and the ChiltoneColburn coefficient of

the wall heat transfer coefficient. Understanding the drop in

permeance will allow to reduce the membrane (and reactor)

size and cost.

Acknowledgments

This work has been conducted within the framework of the

CoMETHy (Compact Multifuel-Energy to Hydrogen converter)

project, funded by the European Commission under the Fuel

Cells and Hydrogen Joint Undertaking (GA No. 279075). The

authors wish to acknowledge the Fraunhofer Institute for

Ceramic Technologies and Systems, IKTS (Germany), for

providing the structured foam, and the Energy Research

Centre of the Netherlands (ECN), for providing themembrane.

The experimental equipment was partially supported by

Technion Grand Energy Program (TGEP).

List of Symbols

a coefficient of ChiltoneColburn correlation (24.1, from

data fitting), �A0 annular section (1.1 � 10�3), m2

AI membrane cross section (6.24 � 10�5), m2

Am membrane permeance pre-exponential factor (6.6),

mol/m2/s/bar0.5

b exponent of ChiltoneColburn correlation (0.29), �dm outer membrane diameter (0.014), m

dw outer reactor diameter (0.04), m

Deff effective dispersivity (1.7 � 10�4), m2/s

Ea rate constant activation energy, kJ/mol

Em membrane permeance activation energy (11), kJ/mol

F gas flow rate (evaluated from model), mol/s

FH2 ;perm hydrogen permeation rate (evaluated from model),

mol/s

jH ChiltoneColburn coefficient (evaluated from model), �kax,eff effective axial conductivity (2.5), W/m/K

kg gas conductivity (6.9 � 10�2), W/m/K

ks intrinsic solid conductivity (50), W/m/K

k0 pre-exponential factor of the rate constants, mol/

gcat s bar

KadsCO CO constant (0.37), atm�1

KadsCO2CO2 adsorption constant (1.701), atm�1

KadsH2O H2O adsorption constant (80073.2), atm�1

KCO2M CO2 methanation reaction equilibrium constant

KED ethanol decomposition reaction equilibrium

constant

KWGS water gas shift reaction equilibrium constant

Ng specific gas flow rate (evaluated frommodel), mol/m2/s

Nuw wall Nusselt number (evaluated from model), �P pressure, bar

Pr Prandtl number (Pr ¼ cp,gmg/kg), �Qm membrane permeance (evaluated from model), mol/

m2/s/bar0.5

rj j-th reaction rate (evaluated from model), mol/gcat/s

Re Reynolds number for a packed bed (Re ¼ rgudw/mg), �S/E steam:ethanol molar ratio at inlet, �T temperature, K

u gas superficial velocity (evaluated from model), m/s

y molar fraction (evaluated from model), �

Greek letters

ε bed porosity (0.85), �rg gas phase molar density (evaluated from model), mol/

m3

rs catalyst density (1.9), g/mL

a stoichiometric coefficient, �Q inhibition parameter (0.15, from data fitting), �

Subscripts

i i-th component

j j-th component

w wall

m membrane

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 0 ( 2 0 1 5 ) 5 8 3 7e5 8 4 85848

s solid

g gas

Superscripts

SW sweep side

r e f e r e n c e s

[1] Giaconia A, Turchetti L, Monteleone G, Morico B,Iaquaniello G, Shabtai K, et al. Development of a solar-powered, fuel-flexible compact steam reformer: theCoMETHy project. Chem Eng Trans 2013;35:433e8.

[2] Patrascu M, Sheintuch M. On-site pure hydrogen productionby methane steam reforming in high flux membrane reactor:experimental validation, model predictions and membraneinhibition. Chem Eng J 2015;262:862e74.

[3] Frusteri F, Freni S, Spadaro L, Chiodo V, Bonura G, Donato S,et al. H2 production for MC fuel cell by steam reforming ofethanol over MgO supported Pd, Rh, Ni and Co catalysts.Catal Commun 2004;5:611e5.

[4] Mattos LV, Noronha FB. Hydrogen production for fuel cellapplications by ethanol partial oxidation on Pt/CeO2

catalysts: the effect of the reaction conditions and reactionmechanism. J Catal 2005;233:453e63.

[5] Palma V, Castaldo F, Ciambelli P, Iaquaniello G. CeO2-supported Pt/Ni catalyst for the renewable and clean H2

production via ethanol steam reforming. App Cat B2014;145:73e84.

[6] Furtado AC, Gonc C, Alonso CG, Cantao MP, Fernandes-Machado NRC. Bimetallic catalysts performance duringethanol steam reforming: influence of support materials. IntJ Hydrogen Energy 2009;34:7189e96.

[7] Vizcaino AJ, Carrero A, Calles JA. Ethanol steam reforming onMg- and Ca-modified CueNi/SBA-15 catalysts. Cat Today2009;146:63e70.

[8] Orucu E, Gocalimer F, Aksoylu AE, Onsan ZI. Ethanol steamreforming for hydrogen production over bimetallic PteNi/Al2O3. Catal Lett 2008;120:198e203.

[9] Palma V, Castaldo F, Ciambelli P, Iaquanilello G, Capitani G.On the activity of bimetallic catalysts for ethanol steamreforming. Int J Hydrogen Energy 2013;38:6633e45.

[10] Domı̀nguez M, Taboada E, Molins E, Llorca J. Ethanol steamreforming at very low temperature over cobalt talc in amembrane reactor. Cat Today 2012;193:101e6.

[11] Basile A, Gallucci F, Iulianelli A, Tosti S. CO-free hydrogenproduction by ethanol steam reforming in a PdeAgmembrane reactor. Fuel Cells 2008:62e8.

[12] Tosti S, Basile A, Borgognoni F, Capaldo V, Cordiner S, DiCave S, et al. Low temperature ethanol steam reforming in aPdeAg membrane reactorepart 1 Ru-based catalyst. J MembrSci 2008:250e7.

[13] Tosti S, Basile A, Borgognoni F, Capaldo V, Cordiner S, DiCave S, et al. Low-temperature ethanol steam reforming in aPdeAg membrane reactorepart 2. Pt-based and Ni-basedcatalysts and general comparison. J Membr Sci2008;308:258e63.

[14] Fatsikostas AN, Verykios XE. Reaction network of steamreforming of ethanol over Ni-based catalysts. J Catal2004;225:439e52.

[15] Turchetti L, Monteleone G, Annesini MC. Comparativeanalysis of heat transport in SiC solid foam and Al2O3

granular packings for fixed-bed reactors. Chem Eng Trans2011;24:1405e10.

[16] Ward TL, Dao T. Model of hydrogen permeation behavior inpalladium membranes. J Membr Sci 1999;153:211e31.

[17] Colonna C. Caratterizzazione e modellizzazione deimeccanismi di trasporto di calore e quantit�a di motoattraverso supporti catalitici di schiuma solida [Unpublishedmaster's thesis]. Department of Chemical Engineering,Materials and Environment. University of Rome “Sapienza”;2012.

[18] Israni SH, Harold MP. Methanol steam reforming in single-fiber packed bed PdeAg membrane reactor experiments andmodeling. J Membr Sci 2011;369:375e87.

[19] Abir H, Sheintuch M. Modeling H2 transport through a Pd orPd/Ag membrane, and its inhibition by co-adsorbates, fromfirst principles. J Membr Sci 2014;466:58e69.

[20] Nekhamkina O, Sheintuch M. Effective approximations forconcentration-polarization in Pd-membrane and separators.Chem Eng J 2015;260:835e45.

[21] Sheintuch M. Pure hydrogen production in a membranereformer demonstration, macro-scale modeling and atomic-scale modeling. Chem Eng J 2014. http://dx.doi.org/10.1016/j.cej.2014.11.100.

[22] Li A, Liang W, Hughes R. The effect of carbon monoxide andsteam on the hydrogen permeability of a Pd/stainless steelmembrane. J Membr Sci 2000;165:135e41.

[23] Hulme J, Komaki M, Nishimura C, Gwak J. The effects of gasmixtures on hydrogen permeation through PdeAg/VeNialloy composite membrane. Curr Appl Phys 2011;11:972e5.