07 - Modelling of Membrane Reactors

Modelling of membrane reactors
Martin van Sint Annaland & Fausto Gallucci

Chemical Process Intensification
DEMCAMER/CARENA Workshop
January 30th, 2013 @ Eindhoven TU/e
Overview lecture
Modelling membrane permeation
Porous membranes
Dense (supported) membranes for H2
Dense membranes for O2

Different reactor concepts
Packed bed membrane reactors
Detailed particle model for heterogeneous catalysts
Coupling of SDMs with particle models
Fluidized bed membrane reactors
Phenomenological two-phase model
CFD-based models: discrete particle models
Multiphase Reactors Group, SPI
11-2-2013
Page 1
Modelling of membrane permeation

Porous membranes
Membrane flux:
(Dusty-Gas-Model):
Ni
1
4
K0
RT
3
8 RT
M
B0
p
g
pi ,m
ri
ln
ri
ri
Parameters K0 and B0
depend on the pore
size/structure distributions
11-2-2013
Page 2

Porous membranes
Fick model: only diffusion
Ni
eff
i
p xi
RT r
1
Bosanquet:
Blancks law:
Dieff
1
Dieff,m
Dieff,m
1
1 xi
1
Dieff,K
n
Dieff, j x j
j 1
j i
Effective molecular diffusion:
Dieff, j
Effective Knudsen diffusion:
Dieff, K
11-2-2013
(harmonic averaging of
molecular and Knudsen diffusion )
Page 3
(Bosanquet + Blanck laws

strictly only valid for dilute mixtures)
Di , j
4
8RT
K0
3
Mi
( = porosity)
( = tortuosity)
Ficks law: What is wrong?

Ficks law (applied to film model):
Flux with respect
to the mixture:
Film model:
Ji
ci ,0
Di
dci
dz
[mol/m2s]
Diffusivity
[m2/s]
ci ,
Flux with respect

to the interface:
Ni
k ci ,0 ci ,
[mol/m2s]
Mass transfer
coefficient [m/s]
z
Film model: k
11-2-2013
Page 4

Binary equimolar diffusion
Constant total concentration
(i.e. isobaric & isothermal)
ctot
c1 c2
constant
Flux of component 1
with respect to the mixture:
J1
dc1
D1
dz
Flux of component 2
with respect to the mixture:
J2
D2
dc2
dz
Total net flux:
Jtot
Ji
J1 J2
i 1
Life is nice
and simple?
d
c1 c2
dz
provided that there is only 1 binary diffusivity D,

independent of composition
11-2-2013
Page 5

Consider two ideal gasses separated by
a porous membrane
pure Ar
high flow rate
pure He
high flow rate
He
Ar
298 K
105 Pa
298 K
105 Pa
Is the He flux equal to the Ar flux

(i.e. NHe = NAr)?
11-2-2013
Page 6

Consider two ideal gasses separated by
a porous membrane
pure He
high flow rate
He
Ar
298 K
105 Pa
298 K
105 Pa
Observation:
NHe 3 NAr
Much higher He flux despite identical concentration gradients!
pure Ar
high flow rate
Interaction with
membrane
Explanation:
Movement of mixture with respect to membrane!
Friction He-membrane < Friction Ar-membrane
pressure gradient
Resulting in viscous flow from right to left: retarding He, accelerating Ar
11-2-2013
Page 7

Porous membranes
Extended Fick model: diffusion + dArcy
Extended
Fick:
Ni
dArcy:
Ntot
eff
i
Extended Fick model:
p xi
RT r
xi N tot
B0 p
ctot
r
Ni
11-2-2013
accounts for net total molar flow rate
B0 p p
RT r
1
xi
eff
Di p
RT
r
Page 8
viscous flow
B0 xi p p
r

Porous membranes
Extended Fick model for tubular membrane:
R2
r
Membrane tube
dr
R1
pA(r=R1) = pA,1
pA(r=R2) = pA,2
R1
rm
R2
R1
Steady state component continuity equation for cylindrical membrane:
1 d
rN i
r dr
rNi
Permeation flux = flux at r = R2:
11-2-2013
Page 9
constant
R2 N i ,2
Ni ,2
c
R2
c
rm

Porous membranes
Component continuity:
Ni
c
r
Substitution of extended Fick model:
Ni
1
xi
eff
Di p
RT
r
1
xi
eff
Di p
RT
r
c
r
Ni
B0 xi p p
r
Integration over the tube thickness:

R2
dxi
eff
Di p
dr
dr
R1
R2
B0 xi p dp
dr
dr
R1
11-2-2013
Page 10
R2
cRT
dr
r
R1
B0 xi p p
r

Porous membranes
R2
R2
dx
Dieff p i dr
dr
R1
diffusion
important when
concentration gradient
is negligible
xi
xi ,2
eff
i
dxi
p xi ,2
xi
B0 xi
xi ,1
eff
i
cRT
dr
r
R1
viscous flow
important when
pressure drop
is negligible
R2
B0 xi p dp
dr
dr
R1
p2
p1
xi ,1
11-2-2013
dr
pdp cRT
r
R1
B0 xi
Page 11
R2
1
2
p22
p12
cRT ln
R2
R1

Porous membranes
eff
i
Note:
p xi ,2
1
2
xi
eff
i
Ni ,2
c
R2
B0 xi
xi ,1
p22
p12
xi
B0 p
N i ,2
2
2
2
1
1
2
1
2
p2
p1
pi ,2
1
Dieff
RT
pi ,1
R2
cRT ln
R1
p2
p1
p pi ,2
R2
cRT ln
R1
pi ,m
B0 p
ri
ln
ri
m
m
11-2-2013
Page 12
pi ,1

Consider three inert ideal gases in two-bulb system
A
N2, CO2
at t=0 connected
by narrow capillary
N2, H2
xH0 2 ,A
xH0 2 ,B
0.50
xN0 2 ,A
0.50
xN0 2 ,B
0.50
0
xCO
2 ,A
0
xCO
2 ,B
0.50
Does N2 flow: a) from A to B?
b) from B to A?
c) not at all?
d) or does it do a), b) and c)?
11-2-2013
Page 13
(isobaric &
isothermal)

A
N2, CO2
N2, H2
at t=0 connected by
narrow capillary
(isobaric &
isothermal)
Reverse
diffusion
Taylor & Krishna (1993)

Fig. 5.4, p. 109
11-2-2013
Page 14

A
N2, CO2
N2, H2
at t=0 connected by
narrow capillary
Observation:
H2 and CO2 change monotonically to equilibrium
concentrations as expected
N2 initially diffuses first without concentration gradient and
later even against its concentration gradient from A to B,
only later bulbs go gradually back to equal compositions
Explanation:
H2 moves from B to A, and CO2 from A to B
N2 molecules more friction with CO2 than H2
and CO2 seems to drag N2 along from A to B
11-2-2013
Page 15
(isobaric &
isothermal)
Reverse
diffusion
cannot be explained
with Ficks law!

Generalized Maxwell-Stefan equations:
Derivation by considering binary interactions between
molecules of different type:
xi N j
In terms of fluxes with

respect to interface Ni:
di
x j Ni
u2
ctot ij
j 1
di
ci
T ,P
ci Fi
c j Fj
i
j 1
xi
with:
RT
n 1
T ,P
ij
xi
ln
j 1
11-2-2013
u1
Generalized driving forces:

ctot RTd i
Page 16
xj
xj
T ,P ,
2
2
di = driving force for

component i
F 1 x1x2 u1 u2

Generalized Maxwell-Stefan equations:
n
Rewriting in (n-1) matrix notation using
Ji
( Ji
Ni
xi Ntot )
i 1
In n-1 matrix notation:

ctot d
J
Bii
Bij
B J
ctot B
xi
in
n
j 1
j i
1
xi
ij
(v) = vector with n-1 components

[M] = square matrix of rank n-1
xj
ij
B11
B12
...
B1,n
d2
...
B21
...
B22
...
...
...
B2,n 1
...
dn
1
in
11-2-2013
d1
Page 17
[B ]
1
Bn
1,1
Bn
1,2
... Bn
1,n 1

Porous membranes
Dusty Gas Model (DGM):
Based on the generalized Stefan-Maxwell equations for diffusion
in multi-component mixtures
Consider membrane as one of the components in the mixture
n
j 1
j i
xi N j
x j Ni
eff
i, j
pD
Ni
pDieff, K
1 xi
RT r
xi
pRT
B0 p
Dieff,K
see e.g. Veldsink et al., Chemical Engineering Journal 57 (1995) 115-125
11-2-2013
Page 18
p
r

Dense Pd-membranes for H2
(Pd, Pd/Ag, Pd/Au based membranes)
High pressure
Mechanism involves
a series of steps:
Low pressure
3
1) adsorption
2) dissociation
3) diffusion
4) re-association
5) desorption
1 2
N H2
Km
ret n
H2
perm n
H2
(0.5 < n < 1)

(n = 0.5: Sieverts law)
Pd-membrane
H2
H2O
CO2

Bulk diffusion control:
H
H
H
HH
HH
H2 ( g )
dcH
DH,Pd
dx
For ideal (dilute) systems:
DH,Pd
ret
H
DH,Pd
CH
and
KH
11-2-2013
Page 20
H(ads)
K H pH0.52
is constant
DH,Pd K H,Pd
perm
H
KH
mole ,H 2
N H2
Dissociation and association

at equilibrium:
HH
Pd
ret 0.5
H2
perm 0.5
H2

Bulk diffusion control:
Pure H2 streams
at permeate and retentate
(i.e. no sweep gas)
N H2
QH
ret n
H2
perm n
H2
QH = 1.54 10-4 mol m-2 s-1 Pa-n

n = 0.730
(P(support) 0.1 bar)
Jurriaan Boon et al. (ECN)

11-2-2013
Page 21
http://www.hysep.com
Modelling membrane permeation

Supported Pd-membranes for H2
Without sweep (i.e. only H2 is present)
Pd
H H
H
H
H
H H
H H
H
H
H
Bulk diffusion Pd-layer:

DGM support:
11-2-2013
Page 22

With sweep
Pd
H H
H
H
H H
H H
concentration
polarisation
Pd
permeation
11-2-2013
Page 23
resistance in support
concentration
polarisation

With sweep: Dusty Gas Model (DGM):
Pressure drop H2-support:
Diffusion resistance H2-stagnant N2:

11-2-2013
Page 24

With sweep: Comparison of different mass transfer resistances:
low sweep gas flow rate
high sweep gas flow rate

11-2-2013
Page 25

air
N2

O2 permeation flux through dense perovskites
O2
O2
MIEC
= Mixed Ion and
Electron Conducting
O2
2O2- + 4e-
O2-
2e-
O2/O2- + ?
N O2
CH4, CO, H2
N O2
Wagner equation:
CO, CO2, H2, H2O
Dv 0
4Vm m n
ret n
O2
11-2-2013
perm n
O2
Page 26
Dv
4Vm m
ln pOperm
2
( )d ln pO2
ln pOret2
( ) = non-stoichiometry:
n
p
0 O2
Note n is typically small (and <0)!!

Compressing air not very effective!
Partial pressure at permeate side
important!

O2 permeation flux through dense perovskites
(LaCa)(CoFe)O3- : Flux studied experimentally for reducing atmospheres
Greatly enhanced in presence of CO and H2 (left)
Highly dependent on thickness (right), indicating bulk diffusion
(T = 1223 K, v =
200 ml/min (STP), dilution with N2)
11-2-2013
Page 27

O2 permeation flux through dense perovskite (LaCa)(CoFe)O3 CO2 inhibits permeation rate (left)
Rate determined by local equilibrium of CO combustion,
proportional to CO/CO2 ratio and thus xO2 (right)
(T = 1223 K, v =
200 ml/min (STP), dilution with N2)
11-2-2013
Page 28

O2 permeation flux through dense perovskite (LaCa)(CoFe)O3 Permeation expressions derived to describe results
N O2
Eact
RT
exp
xO2
ret n
O2
perm n
O2
xCO2
xCO K eq
n = -0.153
Eact = 260 kJ/mol,
C = 2.01108 (cm4/cm2/min, STP)
Zhang et al., J. Membrane Science 291 (2007) 19-32
11-2-2013
Page 29

O2 permeation flux through dense perovskite (LaCa)(CoFe)O3 Model can also accurately predict O2 permeation rate for H2 as
reducing gas assuming local equilibrium of H2 combustion
Zhang et al., J. Membrane Science 291 (2007) 19-32

11-2-2013
Page 30

Basic classification
Packed bed membrane reactors:
Immobilized catalyst
Fluidized bed membrane reactors:
Particles kept in suspension by fluidum
11-2-2013 Page 31

Packed bed reactors:
Single adiabatic bed
Multi-tubular reactor
used for reactions

with small heat effect
used for reactions

with a large heat effect
11-2-2013 Page 32

Packed bed membrane reactor configurations
Many different possible configurations:
(a)
A, B
Gas addition/gas extraction
Flat vs. tubular membranes

Dead end membranes vs. sweep
Co-current/countercurrent A+B
sweeping
D +E
C +B
(b)
Catalyst as particles inside/outside
membrane tubes Catalytic membrane
Catalytically active membranes
A
A+B
C+D
(c)
Supported/unsupported
B
etc. etc.
Example:
multi-tubular module
with sweep gas
Reactants in
Sweep gas out
Catalyst inside
tubular membranes
Reaction Zone
A
Gas Phase
A, B
Gas Phase
(d)
Products out
Catalytic Membrane
Sweep gas in
11-2-2013 Page 33
Catalytic membrane

Phenomenon of fluidization
11-2-2013 Page 34

Phenomenon of fluidization
Particles suspended
in a gas stream
behaves as a fluidum
Bubbles (= voids) are formed,

which create excellent mixing
in the fluidized bed
11-2-2013 Page 35

Fluidization at large scale: Granulation and coating
11-2-2013 Page 36

Fluidization at large scale: FCC and olefin polymerization
Ethylene/propylene
polymerization
FCC (Fluid Catalytic Cracking)

11-2-2013 Page 37

Fluidized bed membrane reactor configurations
Immersed
membrane tubes
Flat membranes
integrated in the walls
Retentate
Retentate
Permeate
Sweep in
Sweep out
Zona de
Reaction
Reaccin
Permeate
zone
Zona de
Regeneracin
Regene500z
Feed
Feed
11-2-2013 Page 38
ration zone

Packed bed and fluidized bed membrane reactors:
Packed bed
membrane reactor
O2
O2
Fluidized bed
membrane reactor
Gas outlet
Product gas outlet
Cyclone
Selective
O2 dozing
Selective
H2 extraction
Membrane tubes
Heat exchange
surface
P, W
H2 W
H2
Simple reactor construction

Possibly large temperature and
concentration gradients (in case of
fast & strongly exothermic reactions)
Reactor instability
11-2-2013 Page 39
Gas inlet
11-2-2013
Isothermal reaction conditions

Catalyst attrition
Gas back-mixing / gas by-passing

Modelling of packed bed membrane reactors
First steps:
Determine what is rate limiting
Determine the key performance quantities

Determine required model accuracy in relation to available data:
e.g. 1D/2D model? Isothermal/adiabatic? Phenomenological vs. CFD?
Modelling membrane permeation:

Flux constant? Permeation equation?
Extended multi-component DGM?
Modelling catalyst:
simplify
whenever
possible!!!
Internal/external mass transfer limitations?

Simplify with effectiveness factors?
Pseudo-homogeneous models?
Heterogeneous models with detailed particle models?
11-2-2013
Page 40

Detailed particle model
Component mass balances (n-1) using
Maxwell-Stefan description for molar diffusion fluxes:
1
2
r
n
2
r r
g
g
with:
ntot
nreact
Sr
Sr ,i
Mi 1
i, j j
j 1
ntot
Homogeneous energy balance:
C p ,s
T
t
1
2
r
r2 r
Effect of particle
porosity/tortuosity via:
Deff
11-2-2013
Page 41
eff , s
Di
T
r
Sh
nreact
Sh
rj hr , j
s
j 1

Detailed particle model
Boundary conditions:
rp
nc 1
r
T
r
nc 1
0
d
g eff ,i , k
0
k 1
eff , s
T
r
r
g
kg
i tot
s ,ik
k ,bulk
k 1
Tbulk
+ constitutive equations for the flux correction factors etc.
Yields local production rates (source terms)

To be coupled with reactor model
11-2-2013
Page 42

Heterogeneous reactor model:
Catalyst in
annular space!
Total continuity equation gas phase:

g
g
Superficial velocity!
g ug
m , tot am
with am =
4d i
d 02 di2
Component continuity equations gas phase (i = 1..nc-1):

g
g
ug
Dax
ni as
m ,i
ni computed from
particle model
Gas phase energy balance:

g
C p,g
T
t
C p,g
ug
T
z
nc
T
z
am
ni as H i
i 1
nc
a Hi
m ,i m
a T Tw
w w
i 1
+ boundary conditions (Danckwerts type)

11-2-2013
Page 43

Constitutive equations:
Pressure drop (differential Ergun equation):
p
z
150
1
g ug
d
2
p
1.75
3
g
2
1
u
g g
g
3
g
dp
Membrane flux equation(s),

e.g. in case of O2 addition via porous membranes:
M O2
m ,O2
RT
4
K0
3
8RT
M
B0
pO2 ,m
ri
ri
ln
ri
+ Reaction kinetics, transport parameters, physical properties
11-2-2013
Page 44

Combining OCM + SMR in a single dual-function
catalyst particle
H
H
C
H
O O
H
H
CCH
O
Outer shell of particle

Heat generation
Ethylene production
C
H HO
O
H
CHOHH
H
H
Catalyst particle
H
C
H
H
11-2-2013
Page 45
Inner core of particle

Heat consumption
energy integration
Synthesis gas production

Novel reactor concept: PBMR for OCM/SRM
with dual function catalyst
Distributive O2 feeding & autothermal operation!
porous membrane
O2
impermeable
O2
CO, CO2,
C2H6, C2H4
CH4 + H2O
O2
O2
O2 concentration
inert
dual function catalyst
OCM
mole
R
r [m]
Use strong internal

diffusion limitations to
avoid O2 to the SRM
SRM
Ensure that C2 flux out of particle is much

larger than C2 flux toward particle core
(minimize C2 losses)
Tune CH4 conversion in OCM and SRM
(autothermal operation)

Extension to 2D:
Determination of extent & effect of concentration polarization
Continuity + Navier-Stokes eqns.
Note: membrane permeation is implemented

via (Danckwerts) boundary conditions

Modelling of fluidized bed membrane reactors
First steps (same as before!):
Determine what is rate limiting
Determine the key performance quantities
Determine required model accuracy in relation to available data:
e.g. Phenomenological vs. CFD?
Phenomenological model:
Phenomenological = two-phase model (bubble-emulsion phase)
Use semi-empirical correlations for bubble and solids behaviour
Bubble size (and thus all other key quantities) function of axial position
Extent of gas phase mixing in emulsion phase via number of CSTRs
in series
CFD model:
At what scale: particle scale vs. industrial scale?
Computational resources? Patience?
simplify
whenever
possible!!!

Phenomenological two-phase model:
Two phases: emulsion and bubble phase
Axial dispersion in emulsion and bubble
phases via number of CSTRs in series
(to be determined from experiments)
(typically: emulsion phase = 1 CSTR;
bubble phase large nr of CSTRs)
Emulsion phase remains at incipient

fluidization conditions despite reaction or
membrane permeation
instantaneous
transfer with bubble phase
Only reaction in emulsion phase
(relatively low reaction rates)
Isothermal and isobaric conditions

Model equations
Ratio of membrane
addition/extraction to/from
bubble and emulsion phases
according to bubble hold-up
Transfer Q is required to
maintain emulsion phase
at incipient conditions
SF(x) = Heaviside function:

x<0: SF(x) = 0, else SF(x) = x
11-2-2013
Page 51

Constitutive equations
Effect of internals
not accounted for
(correlations not available)!
11-2-2013
Page 52
11-2-2013
Page 53

CFD: Multi-scale modelling strategy
Larger geometry, larger scale phenomena
DNS
Discrete
Particle
Continuum
Discrete Bubble
Phenomenological
Fluid-particle
interaction
Particleparticle
interaction
11-2-2013
Particle-particle
interaction;
Bubble behavior
Page 54
Large scale
motion
Industrial size
Black box
Industrial size

Discrete Particle Model
drag
Fcontact ,a
gravity
gas flow
Intermediate-level model
Eulerian grid size is larger than the particle size
Incorporation of detailed particle interaction models
Computation limitation: 104~106
11-2-2013
Page 55

Gas phase conservation equations (3D-DPM model)
Volume-averaged Navier-Stokes equations
+ continuity equation
inter-phase momentum transfer
ug
0
Sp
+ momentum equation
ug
g
g
11-2-2013
ug ug
ug
Page 56
ug T
2
3
1
Vcell
Va
a cell
p Sp
r ra
g g
ug I
u g va

Particle motion equations (DPM model)
+ translational momentum
dva
ma
dt
Va p
Vp
u g va
ma g Fc ,a
(Cundall & Strack (1979))
+ rotational momentum
Npart: 104~106
d a
Ia
dt
180
2
g s
2
p g
Ta
18
Computation resource
3
g s
1 1.5
d
2
p
0.31
(Beetstra et al., 2007)

11-2-2013
Page 57
g s
g
Re s [
d p2
1
g
g s
3 s
1 10
8.4 Re s 0.343 ]
Res 0.5

Flat membrane walls experiments:
Effect of gas addition/extraction
on bubble phase
extraction
extraction
addition
addition
Gas extraction results in increase of equivalent bubble size

Gas addition results in decrease of equivalent bubble size
11-2-2013
Page 58

Flat membrane walls: Effect of gas addition/extraction
40 %
0%
+40 %
Correlation bubble
behaviour and
emulsion phase
circulation patterns!

Flat membrane walls: Effect of gas addition/extraction
40 %
0%
11-2-2013
Page 60
+40 %

Flat membrane walls: DPM simulations
DPM predicts same results as experiments!

11-2-2013
Page 61

Submerged membranes
Development of DPM/IBM
(Immersed Boundary Model)
11-2-2013
Page 62
Concluding remarks
Detailed reactor and particle models (and combinations thereof)

may be required to adequately account for all relevant effects
Make models as simple as possible but not simpler!
Use models to enhance understanding, but experimentation

remains a critical component to validate the models!
11-2-2013
Page 63

07 - Modelling of Membrane Reactors

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

07 - Modelling of Membrane Reactors

Enviado por

Direitos autorais:

Formatos disponíveis

Modelling of membrane reactors

Martin van Sint Annaland & Fausto Gallucci

Dense membranes for O2

Modelling of membrane permeation

Multiphase Reactors Group, SPI

Modelling of membrane permeation

Effective molecular diffusion:

Effective Knudsen diffusion:

Multiphase Reactors Group, SPI

(Bosanquet + Blanck laws

Ficks law: What is wrong?

Flux with respect

Multiphase Reactors Group, SPI

Ficks law: What is wrong?

Total net flux:

provided that there is only 1 binary diffusivity D,

Ficks law: What is wrong?

Is the He flux equal to the Ar flux

Multiphase Reactors Group, SPI

Ficks law: What is wrong?

Multiphase Reactors Group, SPI

Modelling of membrane permeation

Extended Fick model:

Multiphase Reactors Group, SPI

accounts for net total molar flow rate

Modelling of membrane permeation

Steady state component continuity equation for cylindrical membrane:

Permeation flux = flux at r = R2:

Multiphase Reactors Group, SPI

Modelling of membrane permeation

Substitution of extended Fick model:

Integration over the tube thickness:

Multiphase Reactors Group, SPI

Modelling of membrane permeation

Multiphase Reactors Group, SPI

Modelling of membrane permeation

Multiphase Reactors Group, SPI

Ficks law: What is wrong?

Does N2 flow: a) from A to B?

Ficks law: What is wrong?

Taylor & Krishna (1993)

Ficks law: What is wrong?

Ficks law: What is wrong?

In terms of fluxes with

Multiphase Reactors Group, SPI

Generalized driving forces:

di = driving force for

Ficks law: What is wrong?

Rewriting in (n-1) matrix notation using

In n-1 matrix notation:

Multiphase Reactors Group, SPI

(v) = vector with n-1 components

Modelling of membrane permeation

see e.g. Veldsink et al., Chemical Engineering Journal 57 (1995) 115-125

Multiphase Reactors Group, SPI

Modelling of membrane permeation

(0.5 < n < 1)

Modelling of membrane permeation

For ideal (dilute) systems:

Multiphase Reactors Group, SPI

Dissociation and association

Modelling of membrane permeation

QH = 1.54 10-4 mol m-2 s-1 Pa-n

Jurriaan Boon et al. (ECN)