Figure 1. Experimental facilities in reactor engineering group.
Programme design in brief
Our objective is to develop unifying concepts in multicomponent diffusion and multiphase hydrodynamics. We cover both areas of separations and reaction engineering. We use the Maxwell-Stefan formulation to develop a unified theory of diffusion within fluid phases, within macroporous solids, within microporous materials such as zeolites, and in complex three-phase (vapour-liquid-liquid) systems. In the area of zeolite diffusion, our work stresses the close inter-relationship between diffusion and sorption. In ordered zeolite structures, size and configurational entropy effects are exploited to separate isomers of alkanes. In the area of (heterogeneous) azeotropic distillation we show that the Maxwell-Stefan theory can provide new ways to separate azeotropic mixtures using a single column arrangement. Distillation boundaries can be crossed due to "Maxwell-Stefan" effects; such boundary crossing was hitherto considered not possible in a single column arrangement. A 10-bubble cap glass distillation column has been used for experimental verification of our concept and for studying mass transfer in vapour-liquid-liquid systems. We develop generic design procedures (software) for reactive distillation columns, incorporating the Maxwell-Stefan diffusion relationships.
We use common concepts and ideas for design and scale up of multiphase reactors (bubble columns, slurry reactors, gas-solid fluidized beds, air-lift reactors, catalytically structured reactors, distillation trays, catalyst containing trays, ...). The common approach uses techniques of Computational Fluid Dynamics (CFD). For validation of our scale up procedures, we have built extensive experimental facilities: bubble column reactors, with diameter ranging from 0.05 to 0.63 m in diameter and height up to 4 m, catalytically structured packed columns consisting of criss-crossing wire-gauze channels containing 1 - 3 mm catalyst particles, sieve tray distillation columns, with horizontally disposed catalyst containing wire-gauze envelopes. Some of our experimental facilities are shown in Fig. 1.
Figure 1. Experimental facilities in reactor engineering group.
Recently we have been studying the use of low-frequency sound waves to improve gas-liquid contacting in bubble columns.
Our research activities cover a wide range of length scales, covering molecules (0.1 - 1 nm), bubbles and particles (0.1 - 50 mm), reactor internals (10 mm - 50 mm) and reactor vessels (about 1 m in diameter). A variety of computational tools and simulation techniques are used. (1) Configurational-Bias Monte Carlo (CBMC) simulations to calculate the sorption isotherms in zeolites. (2) Kinetic Monte Carlo (KMC) simulations to determine the diffusion behaviour of single component and mixtures in zeolites. (3) Volume-of-Fluid (VOF) simulations to determine bubble morphology and rise characterstics of bubbles in liquids. (4) Kinetic theory of Granular flow for modelling the hydrodynamics of gas-solid fluidized beds. (5) Eulerian simulations of bubble columns, slurry reactors, sieve trays, and structured packed columns for design and scale up purposes.
Substantive overview of results
(A) Development of the Maxwell-Stefan theory for a unified description of multicomponent diffusion
Zeolite diffusion is closely inter-twined with sorption and we use Configurational-Bias Monte Carlo (CBMC) simulation techniques to describe the sorption behaviour. We have shown that the combined CBMC - MS approach can be used to develop a novel method for separating alkane isomers relying on differences in configurational entropy.
In order to explain the principle of configurational entropy, let us consider sorption of a 50-50 mixture of normal hexane (n-C6) and 3-methyl pentane (3MP) at a temperature of 362 K CBMC simulations for the loadings in the mixture as shown in Fig. 2 (a)) for a range of pressures. It is interesting to note the maximum in the loading of 3MP at about 100 Pa . When the pressure is raised above 100 Pa the loading of 3MP reduces virtually to zero. The n-C6 molecules fit nicely into both straight and zigzag channels whereas the 3MP molecules are preferentially located at the intersections between the straight channels and the zig-zag channels. The n-C6 have a higher packing efficiency within the silicalite matrix than the 3MP molecules. It is more efficient to obtain higher loading by "replacing" the 3MP with n-C6; this configurational entropy effect is the reason behind the curious maxima in the 3MP loading in the mixture. From the mixture isotherm presented in Fig. 2 (a) it becomes clear that configurational entropy effects would manifest only at higher pressures, i.e. at high mixture loadings. In order to stress this point, we have calculated the sorption selectivity, S, as a function of the total mixture loading; the results are presented in Fig. 2 (b). The sorption selectivity increases sharply beyond a total loading of 4 molecules per unit cell, corresponding to the situation in which all the intersections are occupied.
Figure 2. (a) 50-50 mixture isotherm for nC6-3MP. (b) Sorption selectivity as a function of total mixture loading
In order to underpin the MS theory of zeolite diffusion we use Kinetic Monte Carlo (KMC) simulation techniques. The KMC simulations have helped us to provide guidelines for setting up the MS formulation for mixtures. Molecular jumps are correlated and these correlations can be modelled within the MS framework by the inclusion of exchange coefficients. For simple geometries like zeolite membranes, we have also shown how the membrane permeation rates can be estimated a priori from Lenndard-Jones parameters and force-fields, using a hierarchy of approaches (CBMC, KMC, MD, MS). In a collaborative STW project with Bliek and Smit, the procedure will be validated for separation of alkane isomers.
Figure 3. Transient uptake of N2 (component 1) and CH4
(component 2) into zeolite 4A. (a, top figure) The molar loadings of the two components within the spherical zeolite crystal.
(b, bottom figure) The average loading of the two components are plotted as a function of the square root of the time.
The MS formulation for intraparticle diffusion and interphase mass transfer, has been incorporated into
a generic design software for reactive distillation (RD) columns. We were the first to show the existence
of multiple steady-states in RD in a 1993 paper and our recently developed dynamic model has shown that
small perturbations can trigger transitions from one steady-state to another. We have validated this transition
phenomena with experimental data from Clausthal; see Fig. 4.
For the separation of ethanol-water azeotrope, for example, benzene, cyclohexane or acetone can be used
as entrainers. Conventional design of distillation columns uses equilibrium stage models. Use of such models
show that distillation boundaries cannot be crossed. We have recently shown that for real columns (tray or
packed), differences in constituent transfer efficiencies (which follows from the Maxwell-Stefan formulation)
causes the actual column trajectories to deviate significantly from those predicted by the equilibrium stage
model. Such deviations can also cause the crossing of distillation boundaries; see Fig. 5. Our recent
experiments with the system ethanol-water-acetone have confirmed the boundary crossing effect and have
shown that only the rigorous MS formulation is able to model this effect. There are important practical implications
of this result. It is possible, for example, with clever tray designs enhancing mass transfer effects selectively
on a few trays, to separate effect azeotrope separation in a one-column arrangement.
Development of generic scaling up methodologies for multiphase reactors
We are developing the science of multiphase reactor scale up. We have concentrated on a few important
reactor types for detailed study using experiments and CFD techniques: (1) bubble column slurry reactors,
(2) gas-solid bubbling fluidized beds and (3) structured catalytically packed reactors.
The scale up philosophy for bubble columns relies on the use of a hierarchy of concepts and models.
Volume-of-fluid (VOF) techniques are used for a
priori prediction of the morphology and rise characteristics of single bubbles and for studying bubble-bubble
interactions. This knowledge is encapsulated in the form of appropriate drag relations for bubble-liquid
(slurry) momentum exchange. Fully 3D Eulerian simulations are used for simulating the detailed velocity
profiles, holdups and mixing characteristics of the constituent phases. We have validated all aspects of
the CFD models with a comprehensive program of experimental work. Figure 6 compares the experimental values
of the liquid velocity at the center of the column, VL(0), with CFD simulation results.
We note that the CFD simulations properly capture the strong scale effects. Experimental work to develop the
scale up rules for a multi-stage bubble column reactor with partition plates
is in progress; see Fig. 7. A collaborative program of research is now underway within Institut Francais du Petrole,
Lyon in which the slurry reactor scale up methodology will be tested in a large scale diameter slurry bubble
column in which cooling tubes are installed. A PhD student (Forret) has started this work.
The hydrodynamics of bubbling gas-solid fluidized beds is precisely analogous.
We have used the Kinetic Theory of Granular Flow (KTGF), in both the Eulerian and Langrangian formulations, to describe
fluid bed hydrodynamics. The simulations have been validated by comparison with experimental data. The analogies between
bubble formation in a powder and in a liquid have been highlighted by VOF and KTGF simulations; see Fig. 8.
Furthermore, in a recent paper we have shown that the bubble-bubble interactions in gas-liquid and gas-solid
systems share the same physics and can be modelled in the same manner. Consequently, the scale up model
(Eulerian frame) is identical for these two reactor types.
Counter-current gas-liquid contacting for a catalytic process, such as reactive distillation, requires
reactor configurations that meet with the following requirements: (1) the pressure drop should be low,
(2) there should be no danger of flooding for a wide range of gas and liquid flows using catalyst
particles in the millimeter range, (3) we should aim for plug flow of both vapour and liquid phases,
(4) Gas-liquid and liquid-catalyst mass transfer should be high enough so as not to be limiting
(ideally the chemical reaction rate should be the rate limiting factor). The conventional trickle
bed reactor cannot be used for counter-current contacting because of the danger of flooding when
using small catalyst particles. We are studying a promising reactor configuration incorporating KATAPAK-S
packing of Sulzer wherein the catalyst is enveloped within wire gauze structures. Two types of operation
have been studied: gas continuous and liquid continuous
(i.e. packed bubble column operation). We have modelled the flow of liquid within the criss-crossing
structure of KATAPAK-S using CFD, see also
here. The simulations clearly show the excellent
radial dispersion (see Fig. 9), which is desirable from a practical point of view.
Our simulations of the dispersion are in excellent agreement with experiment and show that the radial
dispersion coefficient in KATAPAK-S is about one order of magnitude higher than in conventional trickle beds.
The next, much more difficult step, is to model the gas-liquid flow within the open channels.
In collaborative project with the University of Mexico, funded partly by the Government of Mexico,
a PhD student (Ojeda Nava) is developing the
counter-current reactor
configuration for hydrotreating and hydrodesulphurisation processes.
We are studying an alternative hardware for catalytic distillation in a tray column
in which catalyst containing envelopes are disposed on conventional sieve trays along the liquid flow path; see Fig. 10.
This structure has considerable promise. CFD simulations of the gas-liquid flow on the trays show excellent
agreement with experiment. In view of the success of our CFD approach we suggest that CFD could be used
to test out novel hardware configurations.
In a newly started project we have tried to improve gas-liquid contacting and mass transfer in bubble
columns by applying low-frequency vibrations (in the 50 - 200 Hz range).
Figure 11 shows a snapshot of bubbles emerging from a single nozzle at 0 Hz (no vibration) and when the
liquid phase is vibrated at 130 Hz. The improvement in mass transfer is of the order of 100% with minimal
energy input. We make subtle use of bubble resonance phenomena. The principle has been demonstrated using a
"loudspeaker". Scale up studies are under way and we have purchased a membrane diaphragm
pump for purposes of inducing the vibrations. We see potential applications in hydrogenation reactions
in slurry reactors where improvement in hydrogen transfer can be achieved without the use of mechanical
stirrers.
Programme development
Work on sorption and diffusion in zeolites will continue with the focus being on studying the influence
of different zeolite structures (MFI, NaX, ..). KMC simulations will be used to study mixture diffusion
and topology dependencies. The incorporation of topological effects into the continuum MS description
is an area which needs particular attention. The concept of exploiting configurational entropy effects to
effect separation of alkane isomers using silicalite will be demonstated experimentally (in collaboration
with Bliek and Smit). Our earlier work has shown that separation selectivity is highest when the loading
inside the silicalite crystals exceeds 4 molecules per unit cell. For optimum exploitation, therefore, we
should pre-saturate the zeolite crystals and maintain high loadings (> 4 molecules) during the
whole sorption process in say a packed bed adsorber.
We will venture into the area of biotechnology with the Maxwell-Stefan theory and a PhD student
(Cvetkovic, in a collaborative project with Prof van der Wielen, TU Delft) will undertake a study of the
diffusion of small biomolecules, such as peptides within protein
crystals. There are similarities with diffusion within zeolites, which we will exploit. The project offers
us the opportunity of bringing our mass transfer arsenal to bear on the description of transport
phenomena within living cells.
Experimental and modelling work on azeotropic distillation will continue to demonstrate that
simplifications in flowsheets, with concomitant energy saving, can accrue from boundary crossing due
to mass transfer effects. Future work will also consider heterogeneous azeotropic distillation where
liquid-liquid phase splitting occurs on some of the trays in the column. The description of the
interphase mass transfer process is particularly challenging and there are no design guidelines
in the literature.
The work on scale up of bubble column slurry reactors will proceed on the large scale at
IFP, Lyon.
Our work on structured catalytically packed bed reactors will continue with the objective of trying
to develop this as an alternative to a conventional slurry bubble column. The scientific challenge is
to see how far CFD models will be successful in describing the gas-liquid flow in such units.
The PhD student Ojeda Nava will use this reactor configuration for hydrotreating and hydrodesulphurisation.
The know-how we have generated in the area of CFD modelling of distillation sieve trays is of
interest to industry and it is the intention to transfer this know-how. We will proceed further with
the scale up aspects of our novel contactors involving catalyst containing envelopes disposed on
sieve trays along the liquid flow path.
We will be studying the scale up aspects of reactors in which the liquid phase is subjected to
low-frequency vibrations. The experimental set-up is already functional and the initial studies have
started. We will seek collaboration of an industrial partner for utilization of this concept for
hydrogenation processes.
Figure 4. Vapor and liquid temperature trajectories of low to high conversion steady state transition when switching the feed to pure TAME for one hour. (a) Temperatures at the bottom of the catalytic section. (b) and (c) Temperatures at the bottom of the inert section
Figure 5: (a) Residue curves for the system water (1) - ethanol (2) - acetone (3). (b) Compares EQ and NEQ distillation trajectories.
In this simulation the total number of stages is 12. The column is operated at total reflux with no feed- and product-streams.
The initial liquid composition is x1 = 0.035, x2 = 0.3 and fixed on stage 1 (condenser).
Figure 6: Comparison of experimental values of centre-line velocity VL(0) with 2D Eulerian simulations with experimental data with air - water. Also shown for comparison purposes is the Riquarts (1981) correlation
Figure 7: Introduction of partition plates in a bubble column for reducing the liquid phase backmixing
Figure 8. Snapshots of the simulation of a formation of a gas bubble in a gas-solid fluidised bed of powder and water carried out respectively with the granular theory and VOF technique. The bubble contours are drawn at various time steps.
Note the change in the scales between the upper and bottom rows. For a gas-solid system the contours of a bubble are defined by the requirement that the voidage is greater than 0.85. Animations of these simulations can be viewed on our web site
http://ct-cr4.chem.uva.nl/analogies/
Figure 9: (a) Criss-crossing structure of KATAPAK-S. (b) CFD simulation of tracer
Figure 10: Catalyst containing envelopes disposed along the liquid flow path of a sieve tray. The construction shown above has been simulated using CFD
Figure 11: Typical video snapshots taken at two different vibration frequencies for the air-water system using a single nozzle injection device. The actual video images (placed on our website: http://ct-cr4.chem.uva.nl/sonication/) have been retraced.