ADSORPTION/ DESORPTION

OF WATER AND ETHANOL ON

3A ZEOLITE IN
NEAR-ADIABATIC
FIXED BED

Innovation Driven Solutions Focused

 Ind. Eng. Chem. Res. 2009, 48, 9247–9260 9247

Adsorption/Desorption of Water and Ethanol on 3A Zeolite in Near-Adiabatic

Fixed Bed
Marian Simo,* Siddharth Sivashanmugam, Christopher J. Brown,† and Vladimir Hlavacek
Department of Chemical and Biological Engineering, UniVersity at Buffalo, Buffalo, New York 14260

A pilot scale adsorber apparatus was designed and constructed to investigate water and ethanol adsorption/
desorption kinetics on 3A zeolite for the design purposes of a fuel ethanol dehydration pressure swing adsorption
(PSA) process. Equilibrium studies have shown that 3A zeolite adsorbed a significant amount of water while
very weak ethanol adsorption was observed. The breakthrough curves were utilized to study the effects of
column pressure, temperature, flow rate, pellet size, and adsorbate concentration on the overall mass transfer
resistance. Based on experimentally observed trends, both macropore and micropore diffusion were identified
as relevant mass transfer mechanisms. A mathematical model for a bench scale adsorption bed included the
linear driving force (LDF) adsorption rate model and the variation of axial velocity. A detailed heat transfer
model was a necessity since the bed dynamics was affected by heat transfer in the bed wall. The model was
used to analyze the experimental data and extract values of pertaining diffusion coefficients.

Introduction the diffusion in a 3A zeolite pellet. The unknown parameters

were estimated by fitting experimental data to the postulated
The use of ethanol for car fuel purposes has gained a wide kinetic model. It was found that both macropore and micropore
popularity because of the fact that the raw material is renewable. diffusion mechanisms are the controlling diffusion mechanisms.
A significant cost involved in the ethanol production process is
the energy required for the product purification. The ethanol
content of a fermentation broth is usually about 6-10 wt %. Theory
By simple distillation, the ethanol-water mixture can be Commercial zeolite adsorbents are made by compressing
enriched up to a maximum of 95 wt %. Further enrichment of zeolite crystals (∼1 µm in diameter) into pellets (a few
ethanol must obviate the azeotropic point in order to deliver a millimeters in diameter) with the aid of a binder. In general,
fuel grade ethanol (g99.5 wt %). Conventionally, the final the pellet uptake can be controlled by up to four distinct mass
purification was done by azeotropic distillation.1 With the transfer resistances:5 (1) An adsorbent particle is always
development of adsorption processes and invention of molecular enclosed by a laminar film separating the particle from the bulk
sieves, pressure swing adsorption (PSA) replaced the azeotropic fluid. (2) Macropores (voids in the pellet) act as a conduit to
distillation process in the late 1980s.2 The large scale process transport the gas molecules from the pellet surface to the particle
utilizes a 3A zeolite that preferentially adsorbs water while interior. (3) Molecules are adsorbed on the surface of zeolite
ethanol molecules are excluded. The PSA process is attractive crystals (pore mouth), also known as the surface barrier
due to the low energy consumption, its capability of producing resistance. (4) Eventually molecules diffuse within the zeolite
very dry product, and its proven technical record. crystal by a mechanism known as activated micropore diffusion.
In spite of the widespread application of this technology, a A combination of two diffusion mechanisms, Knudsen diffusion
detailed numerical study on the operation and performance of and the molecular (bulk) diffusion, is possible in the macropore
this PSA process is not available. The quality of a PSA model region depending on the size of the pore and diffusing molecule.5
largely depends on the accuracy of submodels used for the Each of the above resistances exhibits a unique dependence on
description of adsorption equilibrium, heat of adsorption, the operating conditions, and thus a carefully planned experiment
kinetics, and heat transfer dynamics. The coupling of these can provide information on the rate-controlling mechanism.
effects can be quite complicated in a real PSA process due to There are only several citations in the open literature on the
the fact that the operation is inherently transient. adsorption kinetics of water and ethanol on a 3A zeolite relevant
The objective of this work was to study the adsorption/ to the ethanol-water PSA process. Several purely experimental
desorption kinetics of water and ethanol on a 3A zeolite in the studies have demonstrated the ability of a 3A zeolite to produce
range of operating conditions corresponding to the industrial fuel grade ethanol in either liquid6 or gaseous phase.7,8 It was
ethanol dehydration PSA process. Dynamic breakthrough also concluded that the extent of the ethanol coadsorption is
measurements in nonadiabatic fixed beds are a proven technique minimal; however, this claim was not supported quantitatively.
to assess the kinetic parameters for adsorption and desorption Sowerby and Crittenden developed a kinetic model for ethanol
processes.3,4 Adsorption breakthrough and desorption curves drying in small columns for the purpose of a temperature swing
were measured at different operating conditions by varying the adsorption (TSA) process design.8 In their work, an ethanol-water
temperature, pressure, bed velocity, inlet concentration, and vapor feed was used in dynamic nonisothermal breakthrough
particle size. The effects of operating parameters were consid- measurements. External fluid and solid film were considered
ered to formulate the adsorption/desorption kinetic model for as the relevant mass transfer resistances in the kinetic model.
The range of experimental conditions was quite narrow,
* To whom correspondence should be addressed. E-mail:
marian.simo@gmail.com. especially in terms of temperatures (105-120 °C), and the effect
†
Thermal Kinetics Engineering PLLC, 2420 Sweet Home Road, of pressure was not considered at all. The solid film coefficient,
Amherst, NY, 14228. used to lump the adsorption rate inside the pellet, indicated the
10.1021/ie900446v CCC: $40.75  2009 American Chemical Society
Published on Web 09/25/2009
9248 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009

liquids were stored in two 4-gal tanks. An overpressure was

created by compressed nitrogen in these tanks to force the flow
of liquid toward the controlled evaporator mixer (CEM) unit.
Any oscillations that would be normally introduced by a pump
were thus eliminated. After the desired flows were set, liquid
and gaseous streams were mixed in the CEM unit. The complete
vapor generating system was supplied by Bronkhorst: water and
ethanol MFCs LIQUI-FLOW L2302 series 10-500 g/h, nitro-
gen MFC EL-FLOW series 1-50 SLM (standard liters per
minute), CEM unit (20-200 °C), and digital readout/control
unit. The accuracy of MFCs was (1% according to the
manufacturer data. The temperature of the CEM generated vapor
stream could be further adjusted by super heater 1. The three-
way valves V5 and V6 were used to divert the stream to the
bypass line or to the adsorber. A bypass line was used for gas
chromatograph (GC) calibration or during a start-up of an
experiment.
Figure 1. Simplified process flow sheet diagram of the kinetic apparatus.
MFC, mass flow controller; TIC, temperature indicating controller; PRV, (2) Adsorber. A Swagelok 500 cm3 cylinder was used as
pressure regulating valve; PI, pressure indicator; NV, needle valve; 6, six the adsorber bed. The material of construction was 316 L
port GC sampling valve. stainless steel, the bed was 597 mm long, and the outside
diameter was 48.2 mm with a 6.1 mm thick wall. At the top
importance of activated micropore diffusion since the system and at the bottom of the bed, 3.6 mm glass beads were used to
showed a strong temperature and concentration dependence. ensure a proper gas distribution and stabilization of the zeolite
However, no comment on the actual mass transfer mechanism layer. To ensure constant axial temperature profiles in the bed
was made. prior to an experimental run, the bed was equipped with six
An isothermal breakthrough study for the ethanol-water band heaters (∼200 W each) 76 mm long and 5 mm thick. A
mixture was performed by Kupiec et al.7 Micropore diffusion 110 mm thick layer of ceramic insulation was applied next to
was assumed to be the rate-limiting step; unfortunately diffusion
minimize the heat loss to the environment. Four thermocouple
coefficients were reported only at 100 °C. Studies of Sowerby
inserts were constructed to monitor axial temperatures at the
and Kupiec used zeolite pellets of several millimeters in
bed positions: inlet, L/3, 2/3L, and outlet. Each insert could hold
diameter.
up to three thermocouples (TC), so the temperatures in the radial
Tian et al.9 performed gravimetric temperature programmed
direction could be tracked as well. The total number of TCs
desorption experiments with 3A zeolite particles smaller than
was eight due to the limitations of the data acquisition (DAQ)
0.1 mm in diameter. It was assumed that the macropore and
system. Omega K type TCs were used. W. R. Grace 3A zeolites
external film resistances were negligible for such small pellets.
Micropore diffusion was considered as the sole mass transfer 562ET (3.6 mm pellets) and 564ET (1.8 mm pellets) were used
phenomenon governing the particle uptake. Data confirmed that in the present study.
the concentration and temperature dependences of the crystal (3) Gas Chromatograph. A Shimadzu GC 2014 with a TCD
diffusion coefficient conform to the Darken equation. detector was used for the monitoring of water, ethanol, and
The Eigenberger group studied the breakthrough curves for nitrogen concentrations. A necessary amount of vapor was
water on W. R. Grace 4A zeolite.10 They concluded that the bypassed into the GC sample loop to guarantee a stable and
macropore diffusion was the controlling mechanism. Experi- robust analysis. The flow rate was adjusted by a needle valve
ments were conducted for temperatures from 25 to 80 °C and and was not measured. Two packed columns in series, Haysep
pressures from 2 to 5 bar. Both 3A and 4A pellets should be D and Haysep Q, were used to obtain efficient and fast
very similar in terms of their macropore structure; i.e., the same separation. The elution times at the GC column temperature
macropore diffusion mechanism should apply for both materials. 180 °C and helium (GC carrier gas) flow 50 mL/min were
However, it was not reported or observed experimentally that approximately 0.6, 1.2, and 3.5 min for nitrogen, water, and
the macropore diffusion mechanism was relevant for adsorption ethanol, respectively. The GC column lifetime was approxi-
uptake on a 3A zeolite. mately 6 months for a constant operation.
(4) Data Acquisition. Constant monitoring of bed temper-
Experimental Section atures and pressure was achieved through a custom-made DAQ
The dynamic breakthrough runs were performed to measure system. Labview was used to program necessary data acquisition
the equilibrium and kinetic parameters for water and ethanol and processing steps. A National Instruments Fieldpoint system
on W. R. Grace 3A zeolite. The process flow sheet of the was used for conditioning and digitalization of signals.
experimental apparatus is depicted in Figure 1. There are four A Rosemount (high temperature) pressure indicator was
elements that comprise the kinetic apparatus: (1) mixture installed right after valve V6 in Figure 1. A Swagelok pressure
preparation, (2) adsorber, (3) gas chromatograph, and (4) data regulator was used to maintain isobaric operation during the
acquisition. experiment. In the next section of the apparatus, vapor was
(1) Mixture Preparation. Nitrogen, used as an inert carrier condensed and the liquid nitrogen cold trap was installed to
gas, was supplied from a pressure cylinder. The mass flow prevent any contamination of the oil in the vacuum pump.
controller 1 (MFC) was used to control the nitrogen flow rate Kinetic apparatus operating conditions were 100-200 °C and
during adsorption and desorption experiments. The MFC 4 was 2-6.7 bar pressure. All lines had to be isolated to prevent any
used only for the bed regeneration. The MFCs 2 and 3 were vapor condensation. Electric heat tapes from Omega and ceramic
used to control the flow of water and ethanol, respectively. Both insulation were used for this purpose.
 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009 9249
A detailed derivation of the model can be found elsewhere.11,12
An axially dispersed plug flow model was assumed for both
mass and heat transfer processes in the bed and the absence of
radial bed profiles since L/D > 10. Experiments were conducted
at low pressures; thus ideal gas behavior was assumed. The fixed
bed adsorber was equipped with a pressure controller at the bed
outlet; as a result, an isobaric operation was assumed since the
pressure drop is not significant for a short bed. Variation of the
axial velocity was accounted for through the overall mass
balance.
A separate energy balance for the gas and solid phase was
considered. Temperature gradients within the pellet were
neglected; heat transfer from the pellet was entirely due to the
fluid film resistance.11 The adsorber wall energy balance was
also included in the model since it was experimentally observed
that the accumulation of heat in the bed wall affected the column
Figure 2. Experimental run w9. Temperature readings TC1 and TC2
dynamics.
correspond to the bed inlet, centerline, and wall values, respectively; TC3 Axial dispersion terms for component mass balance and
corresponds to the centerline temperature at L/3 distance from the inlet; gas phase heat balance had little effect on the concentration
TC4, TC5, and TC6 correspond to the axial position at 2/3L, centerline, and temperature history. Since the isotherm was highly
midpoint, and wall, respectively; TC7 corresponds to the bed end centerline nonlinear for our system, steep temperature and concentration
temperature.
profiles were expected. In order to make the numerical
Zeolite Regeneration. The zeolite bed was regenerated (or scheme more stable, axial dispersion terms were fully
activated) for 12-20 h prior to each experiment. The regenera- accounted for, even though the corresponding Peclet numbers
tion conditions were the following: temperature in the range were in the range 300-600.
220-240 °C, absolute pressure of 6-10 kPa, and nitrogen purge The zeolite pellet uptake model (adsorption rate) can be
of 200 mL/min. Regeneration temperatures over 270 °C lead substantially simplified by introducing the linear driving force
to an irreversible damage of the zeolite and a loss of the (LDF) approximation as suggested by Glueckauf.13 Two partial
adsorbing capacity. Temperature in the bed was raised by differential equations describing the mass transfer rates in the
changing the voltage of the band heaters using a variable pellet and zeolite crystals are replaced by a single ordinary
transformer. After the activation was over, the power output differential equation. The success and wide application of the
was changed according to the desired experimental temperature. LDF model in the adsorption modeling is due to its physical
After the experimental pressure and nitrogen flow were adjusted, consistency and tremendous savings in computational time
it took 4-6 h to obtain satisfactory initial bed temperature compared to detailed pore models.14
profiles. For particles with bidispersed pore structure, the overall mass
Experiment. Each experimental run started with an acti- transfer coefficient kLDF can be approximated by the series
vated bed that was exposed to a step change in the adsorbate combination of resistances as
concentration at time 0 h. The inlet concentration was kept
constant by the Bronkhorst system. The effluent concentra- 1 ΛRP ΛRP2 rc2
) + + (7)
tion, the bed pressure, and temperatures were constantly kLDF 3(1 - εB)kf 15(1 - εB)εPDp 15Dc
monitored until the bed was fully saturated with adsorbate
and the initial temperature profiles were recovered. As the Thus the overall resistance is composed of contributions from
last step, the desorption run was started from the saturated the external film, macropore region, and micropore region,
bed condition. respectively.3,10 The partition ratio Λ is defined by eq 8.
A typical experimental run took approximately 15-30 h. It
was inevitable to have a fully automatic DAQ and control FBqref
Λ) (8)
system. First, the adsorption step was evaluated through the cref
adsorbed component mass balance to obtain the corresponding
equilibrium loading. Next, the adsorption breakthrough curve The micropore diffusion coefficient shows a significant
and the desorption curve were used in the evaluation of kinetic temperature and concentration dependence usually expressed
parameters. The raw experimental data for a typical experimental by the Darken equation, eq 9. The concentration dependent term
run are shown in Figure 2. can be evaluated from an equilibrium isotherm. The temperature
dependence of self-diffusion coefficient (or corrected diffusivity,
Mathematical Model D0) is usually correlated through eq 10.5
Mathematical models for adsorption column dynamics of d ln P
different complexities have been introduced.11 The simplest Dc ) D0 (9)
d ln q T
breakthrough model is an isothermal plug flow model with trace
level concentration of adsorbing component and linear isotherm.
For adsorption of water on a 3A zeolite a more complex model D0 ) D∞ exp - ( ) Ea
RT
(10)
is necessary; see Table 1. A nonisothermal model is required
because of the high heat of adsorption. In addition, a high water In order to maintain constant initial temperature profiles, the
concentration (g10%) is required to treat the problem as a bulk bed was equipped with band heaters and only then a layer of
separation. Heat effects and the variation of the bed velocity insulation was applied. Thus the temperature of the wall was
must be considered. locally affected by heat input from the band heaters, conduction
9250 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009

Table 1. Mathematical Model of Isobaric Nonisothermal Fixed Bed Adsorber with Initial and Boundary Conditions

in the bed wall, heat transfer through the bed wall from the The fluid phase axial thermal conductivity, λL, was estimated
zeolite bed, and heat transfer through the insulation to the by the correlation given by Dixon:17
environment. By considering such a detailed model, two
additional parameters were introducedsheat transfer coefficient 1 0.73εB 0.5
at the wall hw and the external heat transfer coefficient ) + (12)
PeAF Re Pr 9.7εB
heslumping the thermal resistance of the insulation and the 1+
Re Pr
thermal resistance due to the natural convection heat transfer
outside the adsorber.
The fluid film mass transfer coefficient was estimated from
The set of partial differential equations reported in Table 1
the correlation (13) given by Wakao and Funazkri.18 The
was solved by the method of lines using standard initial and
correlation is valid for a wide range of Reynolds numbers in
boundary conditions. Initially an adsorbate free bed was assumed
the range 3 < ReP < 104. The fluid-solid (particle) heat transfer
for adsorption, while the saturated bed initial conditions for gas
coefficient, hfs, was obtained by assuming an analogy between
and solid concentrations were used for the desorption modeling.
mass and heat transfer; the corresponding coefficient was
The discretization approximation in the axial direction by five
estimated from eq 14. The procedure was used with success in
point upwind finite differences15 resulted in the system of
the adsorption modeling at high Reynolds numbers.11
ordinary differential/algebraic equations (DAEs). Fifty axial
points were required to obtain a reasonable level of accuracy
and stability. The maximum time step of 1 s was used in all Sh ) 2.0 + 1.1Sc1/3Re0.6 (13)
calculations. The modeling of adsorption systems with a
nonlinear isotherm produces a set of stiff DAEs; hence, they Nu ) 2.0 + 1.1Pr1/3Re0.6 (14)
were solved by the DASSL code designed by Petzold16 using
an implicit backward differentiation formula. The FORTRAN Physical and transport properties such as gas viscosity,
IMSL Library subroutine DDASPG, based on DASSL, was used thermal conductivity, and specific heat capacity were calculated
to solve the problem in double precision arithmetic on a standard for a mixture of water or ethanol in the nitrogen carrier according
laptop 1.6 GHz computer. to generally accepted formulas and property correlations.19 All
important system characteristics are summarized in Table 2.
Estimation of Model Parameters The value of the volumetric heat source output QL in eq 6
The parameters in eqs 1-7 were either measured or deter- was evaluated from the initial conditions. Since the power output
mined from empirical correlations. The axial dispersion coef- of all band heaters was kept constant throughout the experiment
ficient, DL, was determined by the correlation suggested by and all other parameters were already known, the value for QL
Wakao and recommended by Ruthven:11 was estimated from the initial temperature profiles in the bed
with the aid of mathematical model.
DL The external heat transfer coefficient, he, was introduced by
20 1
) + (11) considering a rigorous energy balance for the adsorber wall.
udP Sc Re 2
The value of the coefficient he was determined experimentally
 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009 9251
Table 2. Column and Adsorbent Characteristics

Column
adsorbent layer length, L 0.518 m
internal diameter, D1 0.034 m
outer diameter, D2 0.058 m
external diameter, D3 0.165 m
bed void fraction, εB 0.36
bulk density, FB 770 kg/m3
Wall (316 SS)
density, Fw 8000 kg/m3
heat capacity, cp,w 500 J · kg-1 · K-1
thermal conductivity, λw 17 W · m-1 · K-1
Adsorbent: W. R.Grace 3A Zeolite
pellet density, FP 1199.3 kg/m3
adsorbent particle radius, RP 1.785 × 10-3 m
pellet porosity, εP 0.37
BET surface area 45 m2/g
pellet thermal conductivity, λP 0.12 W · m-1 · K-1
pellet heat capacity, cp,s 1045 J · kg-1 · K-1
Figure 4. Equilibrium isotherms of water vapor on 3A zeolite. Data were
by heating the initially cold bed with the carrier gas (band measured at 100, 146, 167, and 200 °C for 3.6 mm pellets and at 167 °C
heaters were turned off) until steady state was reached. for 1.8 mm pellets. Solid lines represent the fit with the Langmuir isotherm
model.
Nuw ) 0.2Pr1/3Re0.8 (15) Table 3. Langmuir Equilibrium Model and Parameters for Water
on 3A Zeolite
Small radial temperature gradients in the bed (1-2 °C)
indicated that the major heat transfer resistance was in the b(T)Pw b∞ ) 5.3126 × 10-10 K0.5 Pa-1;
qw* ) qs,w(T) γ ) 23.235 ) Qst/R/T0; q0,w )
insulation and thus only an order of magnitude estimate for hw 1 + b(T)Pw
10.7446 mol/kg; δ ) 0.687 92;
was necessary. The value of the wall heat transfer coefficient, T0 ) 300 K
hw, did not have a significant effect on the generated temperature
curves. This was expected because of the large difference in
thermal conductivities of the steel wall and ceramic wool
b(T) )
b∞
√T
( )
exp γ
T0
T
insulation and turbulent flow conditions inside the bed. The
correlation given by Dixon, eq 15, was used to obtain an
estimate for the wall heat transfer coefficient.17 The developed
model was used to estimate the heat transfer coefficient for
((
qs,w(T) ) q0,w exp δ 1 -
T
T0 ))
various flows and temperatures. The values of the parameter he
showed only a weak dependence on the temperature and flow
rate, as can be seen from Figure 3. concentration, and after reaching the equilibrium, the adsorbate
uptake was evaluated through the overall mass balance.
Results and Discussion Figure 4 shows adsorption isotherms for water vapor obtained
at 100, 146, 167, and 200 °C on a 3A zeolite for particle diameters
Equilibrium Data. The adsorption breakthrough runs were 3.6 and 1.8 mm, respectively. The parameters of the equilibrium
used to obtain equilibrium data for adsorbing components. model were obtained by a nonlinear least-squares optimizer from
Initially clean bed was exposed to a step input of the adsorbate MATLAB. The Langmuir model with the temperature dependent
saturation capacity qs proved adequate. The values of the model
parameters are summarized in Table 3.
The isosteric heat of adsorption for the system water-3A
zeolite was evaluated from the value of parameter γ (see Table
3) to be 57.95 kJ/mol. Sowerby8 and Carmo20 used the values
50 and 43 kJ/mol. Lalik et al. have measured the heat of
adsorption of water on a 3A zeolite in the range 57-72 kJ/mol
in a microcalorimetric study.21 Gorbach’s work with a 4A zeolite
assumed 54.9 kJ/mol.10
The ethanol uptake by a 3A zeolite was studied at temper-
atures 100, 146, and 167 °C, respectively. The experimental
run carried out at 167 °C is shown in Figure 5. An immediate
breakthrough was observed, indicating a very low bed capacity
for ethanol.
Equilibrium loading of 0.03 mol/kg was obtained for the first
step in the experimental run E3. A very low ethanol loading
can be predicted from the temperature profiles as well. A slight
temperature increase can be observed for the first step only,
Figure 3. Comparison of experimental (points) and calculated (lines)
indicating that all sites available for ethanol adsorption were
temperature profiles used for estimation of external heat transfer coefficient occupied and no further adsorption takes place in the following
he for temperatures 100 and 167 °C and flows of 7.3, 15, and 30 SLM. steps. The fact that the small exothermic spike for the first step
9252 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009

Table 4. Experimental Conditions, Fitted kLDF Values, and Mass Transfer Resistances
resistance [s]
run no. P [kPa] T [°C] us [m/s] PH2O [kPa] FN2 [SLM] FH2O [g/h] dP [mm] 103kLDF [s-1] external film macropore micropore
0.078 8.2 15 13.5 3.57 0.90 51 376 685
w1 448 146
0.083 34.3 15 60 3.57 0.90 15 13 1084
0.083 8.2 15 13.5 3.57 0.55 73 605 1140
w2 448 100
0.083 23.5 15 40 3.57 0.55 N/A N/A N/A
0.088 8.2 15 13.5 3.57 8.50 18 100 0
w3 448 200
0.086 23.5 15 40 3.57 8.50 11 58 48
0.083 8.2 15 13.5 3.57 1.90 37 259 230
w4 448 167
0.082 23.5 15 40 3.57 1.90 17 67 442
0.083 8.2 7.35 13.5 3.57 2.60 26 130 229
w5 224 167
0.08 8.2 7.35 13.5 3.57 2.60 26 96 262
0.083 8.2 23.2 13.5 3.57 1.50 46 399 222
w6 689 167
0.082 8.2 23.2 13.5 3.57 1.50 46 297 324
0.083 12.3 15 20.5 3.57 2.00 28 202 269
w7 448 167
0.082 12.3 15 20.5 3.57 2.00 28 136 336
0.085 16.4 15 27.5 3.57 2.30 23 167 245
w8 448 167
0.082 16.4 15 27.5 3.57 2.30 23 102 310
0.167 8.2 30 27 3.57 2.20 26 259 169
w9 448 167
0.163 8.2 30 27 3.57 2.20 26 193 236
0.218 3.1 40 13.5 3.57 1.95 36 399 77
w10 448 167
0.221 8.2 40 36 3.57 1.90 22 193 311
0.083 8.2 15 13.5 1.785 9.5 12 65 28
w11 448 167
0.082 8.2 15 13.5 1.785 9.5 12 48 45
0.083 8.2 23.5 13.5 1.785 5.5 15 102 65
w12 689 167
0.082 8.2 23.5 13.5 1.785 5.8 15 74 83
0.083 12.3 15 20.5 1.785 10.5 9 51 35
w13 448 167
0.082 12.3 15 20.5 1.785 10.5 9 34 52
0.085 16.4 15 27.5 1.785 11.5 7 42 38
w14 448 167
0.082 16.4 15 27.5 1.785 11.5 7 25 54
0.167 8.2 30 27 1.785 8.5 9 68 41
w15 448 167
0.163 8.2 30 27 1.785 8.3 9 48 63
can be compared with the endothermic peak accompanying final A typical adsorption run is depicted in Figure 6, where the
desorption stage leads to the same conclusion. water breakthrough occurred after 1.25 h. The slope of the
The adsorbent selectivity for the ethanol-water mixture can breakthrough curve, especially in the early stage, is proportional
be now evaluated for the PSA feed stream conditionss92 wt to the adsorption rate. The steeper the curve the higher is the
% ethanol, 167 °C, 448.2 kPasusing the experimental value value of the mass transfer coefficient. An infinite value of the
for the ethanol and water equilibrium model from Table 3. The mass transfer coefficient would lead to a shock front and
value of ∼900 for the adsorbent selectivity (analogue to relative maximum bed utilization (dashed line in Figure 6). Mass transfer
volatility in distillation) illustrates the molecular sieving mech- has thus a negative dispersive effect on the column performance.
anism for ethanol-water separation by the PSA process. The shape of the concentration breakthrough curve is point-
Water Adsorption Kinetics Study. Breakthrough experiments symmetric to the center of the front for an isothermal system
were carried out to study the effect of the temperature, pressure, with a linear equilibrium isotherm; both conditions are strongly
water concentration, flow rate, and pellet size on the bed perfor- violated in the case of water adsorption on a 3A zeolite. The
mance and on the adsorption/desorption kinetics. The experimental effect of the isotherm shape was positive for adsorption since
apparatus and procedure were discussed earlier. Adsorption and for a favorable isotherm a fixed bed transition approaches a
desorption steps were performed in sequence for each experimental
run. Table 4 summarizes all performed experiments; here conditions
for adsorption are always shown first.

Figure 6. Effluent concentration history (right axis) and bed temperature

profiles (left axis) for experimental run w4, adsorption step. Temperature
readings at four axial positions: at z ) 0 TC 1& 2 correspond to centerline
and wall values, respectively; TC 3 corresponds to the centerline tempera-
Figure 5. Ethanol breakthrough run E3. Operating conditions: 167 °C, ture at L/3; at z ) 2/3 L TC 4, 5 & 6 correspond to centerline, midpoint,
448.2 kPa, FN2 ) 20 SLM, ethanol flow rate 13.5, 40, and 80 g/h, and wall values, respectively; TC 7 corresponds to the bed end centerline
respectively. temperature.
 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009 9253

Figure 7. Effluent concentration history (right axis) and bed temperature

profiles (left axis) for experimental run w4, desorption step. Temperature
readings at four axial positions: at z ) 0 TC 1 & 2 correspond to centerline
and wall values, respectively; TC 3 corresponds to the centerline tempera-
ture at L/3; at z ) 2/3L TC 4, 5 & 6 correspond to centerline, midpoint and
wall values, respectively; TC 7 corresponds to the bed end centerline
temperature.
constant pattern profile.22 The symmetry of the breakthrough
curve was disturbed mostly by the heat effects. An increase of
the temperature decreases the solid phase equilibrium loading,
and even after the front had passed, the bed remained hot. As
a result, the approach to the equilibrium in the later part of the
breakthrough curve was controlled by heat transfer, i.e., by
cooling of the bed. The cooling of the bed is usually a slower
process. It can be due to the heat transfer rate in the laminar
film surrounding the particle or due to the magnitude of heat
convection in the bed given by the carrier gas and adsorbent Figure 8. Adsorption breakthrough curves of water on 3A zeolite at 200,
167, 146, and 100 °C, respectively (top). Temperature profiles at four axial
heat capacities. The bed always heats up faster than it cools positions at the bed inlet, L/3, 2/3L, and the bed end (bottom).
down due to the fact that the heat is generated inside the particle
while the cooling rate is governed by phenomena occurring 8 depicts the adsorption breakthrough curves and the temperature
outside the particle. profiles for bed temperatures of 200, 167, 146, and 100 °C,
As can be seen in Figure 6, the temperature profiles were respectively. All other parameters were kept constant; see
not developed immediately and it took approximately 2/3 of Table 4.
the bed length. It is important to mention here that any flow The breakthrough time decreased with the increasing tem-
nonidealities could be responsible for further complications. A perature owing to the decrease in the bed capacity. The
detailed analysis of dynamic interactions due to the adsorption breakthrough times of 0.65, 1.18, 1.68, and 3.1 h were observed
kinetics as well as the coupling of heat and mass transfer will as the temperature decreased. An increasing trend of the slope
be addressed separately in the modeling section. of the breakthrough curve can be observed with an increase of
In the case of desorption, the effect of isotherm was undesirable the temperature. This strong temperature dependence can be
because an unfavorable isotherm generates a spreading or dispersive explained by the presence of the micropore diffusion mecha-
profile. A typical desorption run with the characteristic spreading nism. While the diffusion coefficients for pore diffusion (bulk
concentration profile is depicted in Figure 7. In the early stage, and Knudsen) have only a weak temperature dependence, a
the desorption process was fast because the water concentration strong temperature dependence is typical for activated micropore
was high and so was the bed temperature. The rate of desorption diffusion inside the zeolite crystals.
depends on the driving force and the mass transfer coefficient and The amount of water adsorbed increased when the temperature
both are decreasing as the desorption progresses; in addition the decreased; as a consequence, more heat was generated during the
isotherm becomes more unfavorable. adsorption. This fact can be extracted from the evolution of
The heat effects were smaller in the desorption operation temperature profiles. In other words, the temperature rise ∆T )
compared to the adsorption due to the dispersive effect of an (Tmax - Tinlet) or hot spot increased with a decrease of the
unfavorable isotherm. Since desorption is an endothermic temperature. The temperature rise in the bed at 200 °C was only
process, the temperature should be increased in order to speed 15 K, while for 100 °C a temperature rise of 35 K was observed.
up the process. Heat was supplied to the system only by the Similar trends are anticipated during the desorption process; see
feed stream, and nitrogen has a low thermal capacity compared Figure 9. At 200 °C, the water amount adsorbed was low and, as
to the heat capacity of the solid matrix. Several experimental a result, the bed was regenerated in less than 4 h. The equilibrium
studies have pointed out that the increase in the carrier gas flow isotherm was almost linear at elevated temperatures (Figure 4),
rate had a positive effect on the course of desorption.23 This and thus the isotherm effect did not play any significant role. At
can be explained by improved mass and heat transfer coefficients lower temperatures, the zeolite adsorption capacity for water
as well as by increased heat supply to the system. increased; the isotherm became more unfavorable and desorption
Effect of Temperature. The effect of the bed temperature was profiles were more dispersed. The water concentration dropped
investigated in the experimental runs w1, w2, w3, and w4. Figure below a detectable limit after ∼5 h for the experimental run at
9254 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009

kT
λ) (16)
Pπdmolecule2√2

Now, if the mean free path is larger than the pore diameter
the Knudsen diffusion is the governing mechanism; if the pore
diameter is much bigger than λ the bulk diffusion regime
dominates. Consequently, a transition from bulk diffusion to
Knudsen diffusion is expected as the pressure decreases since
the mean free path increases. However, this transition is
observed only for diffusion in small pores or for diffusion of
larger molecules. If this was the case here, we would expect a
different effect of pressure on the slope of a breakthrough curve.
The Knudsen diffusion coefficient itself is not a function of
pressure, and thus small or no effect would be observed.
The effect of pressure on the adsorption kinetics was studied
Figure 9. Temperature and concentration profiles for water desorption on for both 3.6 and 1.8 mm pellets, and the same trends were
3A zeolite for bed temperatures of 200, 167, and 100 °C, respectively. observed. It is therefore concluded that the macropores make a
Temperatures were measured at four different locations: the bed inlet, L/3, significant contribution to the overall mass transfer resistance
2/3L, and the bed end.
and that macropore diffusion is controlled by the molecular
diffusion mechanism alone.
For desorption runs (see Figure 11) the effect of pressure on
the rate of adsorbent regeneration was not as significant as for
the adsorption. It seems that the desorption rate might be
controlled by other factors such as diffusion in micropores,
thermal effects, and isotherm effects.
Effect of Water Concentration. The effect of water partial
pressure in the feed stream was studied at temperature 167 °C,
pressure 448 kPa, and nitrogen flow rate 15 SLM for two
different pellet sizes; see experimental runs w4, w7, w8, w11,
w13, and w14. In both cases, water partial pressures studied
were 8.2, 12.3, and 16.4 kPa.
The stoichiometric breakthrough time θs decreased as the
water partial pressure increased; see Figure 12. This was
expected because the bed is saturated faster if there is more
water in the feed stream. Experimental breakthrough times for
Figure 10. Temperature and concentration profiles for water adsorption on 8.2 and 16.4 kPa were approximately 0.65 and 1.2 h, respec-
3A zeolite for pressures 224, 448, and 689 kPa, respectively. Temperature
tively. The corresponding equilibrium loadings q* were 4.76
at axial position of 2/3L is shown. Pellet size was 3.6 mm.
and 5.91 mol/kg, respectively. The values of the stoichiometric
breakthrough time θs were evaluated for experimental runs w4
167 °C; however, for the experiment at 100 °C water was still and w8 to 2.33 and 1.4 h, respectively. As mentioned earlier,
present even after 6 h of regeneration. θs corresponds to a shock breakthrough curve in the absence of
Effect of Pressure. The individual mass transfer resistances any mass transfer resistance; the ratio of θb/θs ) 1. It follows
for an adsorbent with bidisperse pore structure such as zeolite that the closer the breakthrough time ratio to a unity is the
are external laminar film, macropore resistance, and diffusion smaller the mass transfer resistance is. By inspection of the
in the zeolite crystals (micropores). Among these only the rate experimental and stoichiometric breakthrough times it is evident
of diffusion in macropores is affected by the pressure. Experi- that the system approaches infinite mass transfer as the feed
mental runs w4, w5, and w6 were carried out to confirm or concentration increases. Hence, it was concluded that, with
contradict the significance of the macropore resistance. Figure increasing partial pressures, sharper concentration profiles were
10 depicts the measured concentration and temperature profiles. observed as a consequence of the favorable isotherm affecting
the kinetic parameters. The same information can be obtained
It is evident that the slope of the breakthrough curve increased
from the temperature profiles, where an increase in the water
as the pressure decreased. An increase of the rate of adsorption
concentration leads to higher and steeper temperature curves.
can be realized from the temperature profiles as well. The hot
At the end of adsorption, the desorption step that followed
spot increased for lower pressures. It follows that the mass
maintained the same operating conditions; see Figure 13. The
transfer into the pellet must be affected by the molecular
slow process was due to the fact that desorption was controlled
diffusion mechanism since the mass transfer rate is inversely
by the unfavorable equilibrium isotherm and the situation got
proportional to the pressure. The same dependence on the worse as the concentration decreased.
pressure is observed for the molecular (bulk) diffusion coef-
Effect of Carrier Gas Flow Rate. Experimental runs w11
ficient in the Chapman-Engskog or Fuller equation.19 and w15 were designated to show the effect of the carrier gas flow
The diffusion mechanism, where the interactions molecule- rate on the adsorption and desorption performance. Experimental
wall are more frequent than the interactions among molecules, is conditions are summarized in Table 4. Measured concentration and
known as the Knudsen diffusion. From the kinetic theory of gases, temperature profiles for adsorption runs are depicted in Figure 14.
the mean free path (λ), the distance a molecule travels between Concentration and temperature profiles for run w15 looked almost
two collisions, is given by the following expression (16):5 identical to those from run w11. An increase of the flow rate had
 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009 9255

Figure 11. Temperature and concentration profiles for water desorption on Figure 14. Adsorption runs w11 and w15. Temperature readings at axial
3A zeolite for pressures 224 and 448 kPa, respectively. Temperature at position z ) 2/3L and concentration profiles for carrier gas flow rate of 15
axial position of 2/3L is shown. Pellet size was 3.6 mm. and 30 SLM, respectively. Experimental conditions: 167 °C, 448 kPa, 8.2
kPa water partial pressure, and pellet size 1.8 mm.

Figure 12. Temperature readings at axial position z ) 2/3L and concentra-

tion profiles for water partial pressures 8.2, 12.3 and 16.4 kPa, respectively. Figure 15. Desorption runs w11 and w15. Temperature readings at axial
Experimental conditions for adsorption runs: 167 °C, 448 kPa, 15 SLM N2 position z ) 2/3L and concentration profiles for carrier gas flow rate of 15
flow, and pellet size 3.6 mm. and 30 SLM, respectively. Experimental conditions: 167 °C, 448 kPa, 8.2
kPa water partial pressure, and pellet size 1.8 mm.

The results for desorption runs w11 and w15 are shown in Figure
15. The regeneration time was halved when the flow rate was
doubled. The breakthrough time for adsorption with flow rate 15
SLM was less than 2 h; see Figure 14. At the same conditions the
regeneration took almost 6 h; however, when the purge flow rate
was doubled the regeneration process took only 3 h.
Effect of Particle Size. The size of the pellet affects many
system parameters such as film coefficients, dispersion in packed
bed, bed porosity, and intraparticle transport processes. The
effect of the first three is usually not significant in an industrial
adsorber; however, the last one can be detrimental if the
macropore diffusion is controlling the mass transfer rate. In other
words, the change of the pellet size will have no effect on the
breakthrough curve if the macropore diffusion is negligible.
Figure 13. Temperature readings at axial position z ) 2/3L and concentra- Experimental runs w9 and w15 were designated to address
tion profiles for water partial pressures 8.2, 12.3, and 16.4 kPa, respectively. this issue. Concentration and temperature profiles are plotted
Experimental conditions for desorption runs: 167 °C, 448 kPa, 15 SLM N2
flow, and pellet size 3.6 mm. in Figure 16. The breakthrough times for 3.6 and 1.8 mm pellets
were 0.44 and 0.89 h, respectively. This significant improvement
of the adsorber performance was due to the fact that the
a positive effect on the film mass transfer and heat transfer macropore resistance decreased by a factor of 4 since the
coefficients. The Reynolds number increased from 27 to 54. The diffusion time scale for macropores depends on the square of
profiles at a higher flow rate were slightly steeper, suggesting that the particle radius. This kinetic effect can be followed from the
film resistance was present but its contribution to the overall mass temperature profiles as well. The temperature peaks were steeper
transfer resistance was only minimal. and narrower due to an increase in the mass transfer rate.
9256 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009

Figure 16. Adsorption runs w9 and w15. Temperature readings at axial

position z ) 2/3L and concentration profiles for pellet size of 3.6 and 1.8
mm, respectively. Experimental conditions: 167 °C, 448 kPa, 8.2 kPa water
partial pressure, and nitrogen flow 30 SLM.

Figure 18. Modeling of effluent and temperature profiles for adsorption

run w4 with simplified model considering constant wall temperature (top)
and rigorous model (bottom). Full lines correspond to experimental data
and dashed lines to predicted data for temperature profiles.
Figure 17. Desorption runs w9 and w15. Temperature readings at axial Next, we will compare the quality of the fit for the
position z ) 2/3L and concentration profiles for pellet size of 3.6 and 1.8
mm, respectively. Experimental conditions: 167 °C, 448 kPa, 8.2 kPa water
experimental run w4 for two mathematical models. Figure 18
partial pressure, and nitrogen flow 30 SLM. (top) shows the results for the situation where the wall
temperature is considered constant (eq 6 is neglected in the
The effect of the pellet size on the course of desorption is model), and Figure 18 (bottom) depicts the results of the full
depicted in Figure 17. Almost identical concentration profiles rigorous model summarized in Table 1.
were obtained. The temperature curves were slightly affected. The correct value of the kLDF coefficient was used in both
It seems that the micropore diffusion regime is the dominant simulations to show the discrepancies, and the value of hw was
one for the desorption process. the remaining free parameter. The rigorous model predicted both
Modeling of Water Breakthrough Experiments. An ex- concentration and temperature curves very well. The temperature
perimental parametric study has investigated the effects of of the hot spot increased as the adsorption front progressed down
operating parameters on the breakthrough curves, thus giving the bed. It is evident that the heat exchange “zeolite bed-adsorber
more insight into the mass transfer mechanism governing the wall” was responsible for these phenomena observed in all
adsorption and desorption processes. It was found that both performed experiments.
macropore resistance and micropore resistance were relevant The temperature of the pellet and ambient gas increased upon
for adsorption with a minor effect of laminar film resistance. the exothermic adsorption of water. Only a portion of the
For desorption, macropore diffusion seemed to play only a minor generated heat was transferred axially by convection in the bed
role, leaving micropores as the controlling mechanism. These because of a relatively low volumetric thermal capacity of
clues provided valuable information in the formulation of the nitrogen gas. Since the heat capacity of the wall was comparable
kinetic model. to the heat capacity of adsorber it acted as a heat sink, so the
Heat Transfer in Column Wall. Many nonisothermal remaining heat was transferred to the column wall. As a result,
breakthrough models tacitly assume that the temperature of the the temperature of the wall locally increased and that created a
adsorber wall is constant during the course of an experiment. driving force for the axial heat conduction in the wall and radial
Such an assumption is valid for systems where the adsorber heat transfer through the insulation to the environment. By
column is not isolated, the column operates at room temperature, comparing the magnitudes of particular terms in eq 6, it was
and thermal effects are only moderate as for example trace found that the axial conduction in the wall had only a marginal
component separations. Otherwise, the heat balance for the effect. The rate of heat transfer through the insulation was the
adsorber wall is required in the model in order to capture the slowest process. For illustration, the transient bed and wall
column dynamics properly. temperature profiles are plotted in Figure 19.
 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009 9257

Figure 19. Transient adsorber bed and wall temperature profiles as a function
of axial coordinate. Particular time values in hours can be found in the
legend. Calculated data correspond to the fit for experimental run w4
depicted in Figure 18, bottom.

Analysis of Overall Mass Transfer Resistance. Experi-

mental adsorption and desorption runs were analyzed with the
developed mathematical model and the values of the two
remaining parameterssoverall mass transfer rate coefficient kLDF
and wall heat transfer coefficient hwswere determined. The
values of the wall heat transfer coefficient fell in the range from Figure 20. Individual mass transfer resistances for adsorption experiments
w1-w15 in percentage of overall resistance (top) and absolute mass transfer
30 to 70 W · m-2 · K-1. resistance values (bottom).
Next, the overall mass transfer resistance was evaluated as
the inverse of the kLDF coefficient for all runs. Three mass Since the isotherm is more favorable at lower temperatures,
transfer resistances in series were considered for both adsorption sharper breakthrough curves were observed with decreasing
and desorption runs according to general trends observed in the temperature as reported by Raghavan for the system “water-4A
experimental parametric study: external film, macropore region, zeolite” at room temperature.4 However, the pellet uptake was
and micropore region. controlled entirely by macropore diffusion. It follows that at
The experimental conditions along with the obtained kLDF values room temperature a sharper water breakthrough will be observed
are summarized in Table 4. By inspection of the resistance values, for a 4A zeolite since there is no additional resistance from
it is clear that the contribution of the external film resistance was micropores, which is the governing mechanism in a 3A zeolite
small (less than 10%). This was expected since the values of at a low temperature. Consequently, a 4A zeolite is the preferred
Reynolds and Biot numbers were in the ranges 19-101 and 22-51, adsorbent for air drying applications.
respectively. According to Do,24 the external mass transfer mech- The effect of pressure on the mass transfer resistance was studied
anism can be neglected for the Biot number larger than 50, and in runs w4, w5, and w6 performed at 448, 224, and 689 kPa,
thus we had to include it in our study. respectively. At lower pressures, micropore diffusion controlled
Several studies identified the micropore diffusion as the govern- the pellet uptake. Approximately 60% of the resistance was in the
ing mechanism.7,8 Our experimental results indicate that the kinetics micropore region. On the other hand, for higher pressures,
was dominated by the resistance in both macropores and mi- micropores were responsible for less than 30% of the overall
cropores. The major mechanism responsible for the molecular resistance; see run w6 in Figure 20. Transition toward the
transport in macropores was the molecular diffusion as was macropore diffusion control was observed as the pressure increased.
confirmed experimentally by varying the overall pressure and the Experimental runs w10, w4, w7, and w8 were designated to
pellet size. In order to evaluate the macropore resistance, the show the effect of the water concentration. By comparing
equation of Fuller et al.19 was used to evaluate the bulk diffusion conditions in experiments w10, w4, w7, and w8, the corre-
coefficient for water-nitrogen mixture and a tortuosity factor of 2 sponding water partial pressures were 3.11, 8.2, 12.3, and 16.4
was used according to work by Teo.6 After evaluating the resistance kPa and the equilibrium solid phase loadings were 2.91, 4.76,
of the external film and macropores, the resistance in the micropore 5.47, and 5.9 mol/kg, respectively. As a result, the overall mass
region could be evaluated using the experimental value of the transfer coefficients increased with the increasing water con-
overall mass transfer resistance; see Table 4. centration. Figure 20 shows that the contribution of the
It was pointed out in the previous section that the contribution macropore resistance decreased (76%, 48%, 40%, and 37%)
of a particular mechanism toward the overall resistance depends while the contribution of the micropore resistance increased
on the operating conditions. Trends in the adsorption kinetics (16%, 44%, 54% and 57%) as the water concentration increased.
for all experiments are summarized in Figure 20. A strong These opposite trends can be interpreted by lower values of
temperature dependence of the kLDF coefficient is evident for the partition coefficient Λ as the water concentration increased;
experiments w1-w4 carried out at 146, 100, 200, and 167 °C, i.e., the macropore resistance is smaller. Diffusion in micropores
respectively. At 100 °C more than 60% of the mass transfer was actually faster at higher concentrations; however, since both
resistance was due to the micropore region, while at 200 °C mechanisms were present the decrease in partition ratio had a
this contribution was negligible. Transition toward the macropore stronger effect. Generally, transition toward the micropore
diffusion control was observed as the temperature increased. diffusion control was observed as the concentration increased.
9258 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009

Table 5. Corrected Diffusion Coefficients

T [°C] 103(1/T) [K-1] D0/rc2 [s-1] ln(D0/rc2)
100 2.68 3.5 × 10-5 -10.26
146 2.39 7 × 10-5 -9.57
167 2.27 1.5 × 10-4 -8.80

Arrhenius dependence. A linear plot of ln D0/rc2 vs 1/T was

used to estimate the activation energy for diffusion of water in
a zeolite crystal. The experiment performed at 200 °C was not
included since the contribution of micropore resistance was
negligible. The value of the activation energy Ea was 27.6 kJ/
mol, and the diffusion coefficient at infinite temperature divided
by rc2, D0∞/r02, extrapolated to 0.2553 s-1. For experiments with
smaller pellets D0/rc2 equals 8.5 × 10-4 s-1 measured at 167
°C, which was approximately 4 times higher than the corre-
sponding value for larger pellets.
Diffusion in zeolites was widely studied.5 While different
zeolite samples of the same zeolite often show large differences
in diffusivities, there is generally little variation in the activation
energy. The activation energies correlate very well with the
molecular size. A significant amount of diffusivity data is
available in the literature for A-type zeolites.5 Typical values
of the activation energy fall in the range from 19 to 60 kJ/mol.
Tian et al. reported an activation energy of 30.2 kJ/mol.9 They
measured the micropore diffusion coefficients for water on a
Figure 21. Individual mass transfer resistances for desorption experiments 3A zeolite directly by using gravimetric temperature pro-
w1-w15 in percentage of overall resistance (top) and absolute mass transfer
resistance values (bottom). grammed desorption experiments. Their value is in excellent
agreement with our result, taking into account the fact that
Experimental runs w11-w15 were performed with smaller entirely different techniques were used.
pellets. The higher kLDF values can be explained through the
quadratic dependence of the macropore resistance on the particle Conclusions
radius; approximately 4 times smaller resistance was observed. The pilot scale near-adiabatic fixed bed apparatus was
The situation in micropores also unexpectedly improved. Since designed and constructed to investigate the water and ethanol
both zeolites were from the same manufacturer the diffusion adsorption behaviors on a W. R. Grace 3A zeolite. Adsorption
coefficients should be similar; however, it is possible that smaller breakthrough experiments were used to study both equilibrium
pellets were prepared from smaller crystals, accounting for and kinetics in the range relevant to the operating conditions
4-fold decrease in the micropore resistance as well. of a commercial ethanol-water PSA process.
The mass transfer resistance summary for desorption runs is Water equilibrium data conformed to the Langmuir isotherm
depicted in Figure 21. Overall, the kLDF values are similar to those model with the temperature dependent saturation capacity. A very
obtained for adsorption; however, the relative contributions of low ethanol uptake of 0.03 mol/kg was observed at 167 °C. The
macropore resistance and micropore resistance seem to be shifted value of the adsorbent selectivity for “ethanol-water” PSA process
toward the micropore control. The observed trend can be explained feed stream conditions was evaluated to be ≈900. The adsorbent
by the effect of an unfavorable isotherm for desorption affecting selectivity is analogous to the relative volatility in distillation; hence,
the micropore diffusion coefficient through the concentration a very high value of selectivity demonstrates the molecular sieving
dependent term in the Darken equation. This also explains why mechanism for this separation process.
the change of the particle size and pressure did not have any The adsorption-desorption kinetics study investigated the
significant effect on the course of desorption step, i.e., lower effects of the pressure, temperature, water concentration, bed
contribution of macropores toward the overall mass transfer velocity, and pellet size on the shape of the breakthrough curves
resistance. The trends in the concentration and pellet size follow to identify the relevant mass transfer mechanisms. Strong
the explanation stated above for adsorption. The sole micropore temperature dependence was revealed; the slope of the break-
control was observed for low temperature run w1 while at high through curves increased with the increasing temperature, thus
temperature both mechanisms contribute equally. confirming the activated micropore diffusion mechanism. The
Water Adsorption and Desorption Kinetics Model. The experiments at different pressures and pellet sizes identified that
experimental values for the resistance in micropores could be the macropore diffusion mechanism had a considerable effect
used to directly evaluate the crystal diffusion coefficient and as well. The gas velocity had only a weak effect on the
the self-diffusion coefficient. The problem is the proper integral breakthrough curves since the experiments were performed at
value of the concentration dependent term in the Darken high Reynolds numbers. As a result, three resistances in the
equation relating these two diffusivities. series model were proposed to describe the adsorption and
The experimental data were fitted again utilizing eqs 9 and desorption kinetics of water on a 3A zeolite.
10 to obtain the values of D0/rc2 at different temperatures directly It was observed that mass transfer was controlled by the
instead of estimating the lumped kLDF parameter. A satisfactory diffusion in micropores as the pressure decreased and the water
fit was obtained by using the same value of corrected diffusivity concentration increased; on the other hand, transition to the
for both adsorption and desorption; see Table 5. The temperature macropore diffusion mechanism was observed for experiments
dependence of corrected diffusivity is usually correlated by the at higher temperatures. Approximately 60% of the overall mass
 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009 9259
-1
transfer resistance was in the micropore region for desorption, Qst ) isosteric heat of adsorption () -∆Hst) [J · mol ]
while for adsorption this contribution was close to 40%. It is R ) universal gas constant [J · mol-1 · K-1]
obvious that the effect of the isotherm nonlinearity was more rc ) zeolite crystal radius [m]
pronounced during desorption through the concentration de- RP ) adsorbent particle (pellet) radius [m]
pendence of the micropore diffusion coefficient. Hence, the fact t ) time variable [s]
that the isotherm was unfavorable for desorption further explains T ) temperature [K]
the slow regeneration process of a 3A zeolite. us ) superficial velocity [m · s-1]
The proposed kinetic model was used as an input to a Y ) fluid phase molar fraction
mathematical model developed for the analysis of the dynamic z ) bed spatial coordinate [m]
column response. The axially dispersed plug flow model
accounted for the variation of axial bed velocity. Heat effects Dimensionless Parameters
had a significant effect on the column performance because of Bif ) fluid phase Biot mass transfer number ()kfRP/εPDp)
the high heat of adsorption and near-adiabatic operation of the Nu ) Nusselt number ()hdP/λg)
laboratory column. A detailed heat transfer model had to PeAF ) axial fluid phase Peclet number ()usFgcp,gdP/λL)
consider the energy balance for a gas phase, solid phase, and Pr ) Prandtl number ()cp,gµg/λg)
bed wall in order to reproduce the measured temperature profiles. ReP ) particle Reynolds number ()usdPFg/µg)
A mathematical model was employed to estimate the micropore Sc ) Schmidt number ()µg/FgDM)
diffusion coefficients. The Darken equation was used to account Sh ) Sherwood number ()kfdP/DM)
for the temperature and concentration dependence of micropore
diffusivity. The activation energy for water diffusion in W. R. Greek Symbols
Grace 562ET zeolite crystal was evaluated to be 27.6 kJ/mol.
All experimental profiles measured in the range of pressures Λ ) partition ratio
from 224 to 689 kPa and temperatures 100-200 °C were ε ) void fraction
successfully reproduced using the proposed adsorption- λ ) molecule mean free path [m]
desorption kinetic model comprised of the following mass λL ) axial thermal conductivity of the fluid [W · m-2 · K-1]
transfer mechanisms: external film, molecular diffusion in λw ) thermal conductivity of bed wall [W · m-1 · K-1]
macropores, and micropore diffusion in zeolite crystals. The µ ) viscosity [Pa · s]
formulated kinetic model will be used to study a commercial θs ) stoichiometric (shock front) breakthrough time ()mzqref/Ffeed)
ethanol dehydration PSA process. [s]
F ) density [kg · m-3]
τ ) tortuosity
Acknowledgment
Subscripts
The authors gratefully acknowledge the financial support of
Thermal Kinetics Engineering, PLLC. B ) bulk or bed
c ) zeolite crystal
F ) feed (inlet) conditions
Notation
g ) gas (fluid) phase
b ) isotherm equilibrium constant (adsorption affinity) [Pa-1] P ) particle, pellet
c ) fluid phase molar concentration [mol · m-3] ref ) reference conditions (usually feed stream)
cp ) isobaric specific heat [J · kg-1 · K-1] s ) solid (adsorbed) phase
D, D1 ) internal bed diameter [m] w ) wall
D2, D3 ) outer bed diameter, external bed diameter including
insulation [m]
Dc ) micropore (zeolite crystal) diffusion coefficient [m2 · s-1] Literature Cited
D0∞ ) corrected diffusion coefficient at infinite temperature [m2 · s-1] (1) Katzen, R.; Moon, G. D., Jr.; Kumana, J. D. Distillation method and
Deff ) effective diffusion coefficient ()εPDp/τ) [m2 · s-1] apparatus for making motor fuel grade anhydrous ethanol. EP 11147, 1980.
DL ) axial effective dispersion coefficient [m2 · s-1] (2) Tindall, B. M.; Natarajan, R. S. Production of anhydrous ethanol by pressure
DM ) molecular (bulk) diffusion coefficient [m2 · s-1] swing adsorption. In AIChE: New York, NY, Minneapolis, MN, 1987.
(3) Malek, A.; Farooq, S. Kinetics of hydrocarbon adsorption on
dP ) pellet diameter [m] activated carbon and silica gel. AIChE J. 1997, 43 (3), 761–776.
Dp ) macropore diffusion coefficient [m2 · s-1] (4) Raghavan, N. S.; Ruthven, D. M. Dynamic behavior of an adiabatic
Ea ) activation energy for diffusion in zeolite crystal [J/mol] adsorption columnsII. Numerical simulation and analysis of experimental
F ) volumetric or mass flow rate [SLM] or [g/h] data. Chem. Eng. Sci. 1984, 39 (7-8), 1201–1212.
(5) Karger, J.; Ruthven, D. M. Diffusion in Zeolites: and Other
he ) external heat transfer coefficient [W · m-2 · K-1]
Microporous Solids; Wiley-Interscience: New York, 1992.
hfs ) fluid-solid heat transfer coefficient [W · m-2 · K-1] (6) Teo, W. K.; Ruthven, D. M. Adsorption of water from aqueous
hw ) heat transfer coefficient at the bed wall [W · m-2 · K-1] ethanol using 3-Å molecular sieves. Ind. Eng. Chem. Process Des. DeV.
k ) Boltzmann constant ()1.38 × 10-23 J · K-1) 1986, 25 (1), 17–21.
kf ) external/fluid mass transfer coefficient [m · s-1] (7) Kupiec, K.; Rakoczy, J.; Mirek, R.; Georgiou, A.; Zielinski, L. Determi-
nation and analysis of experimental breakthrough curves for ethanol dehydration
kLDF ) linear driving force (overall) mass transfer coefficient [s-1] by vapor adsorption on zeolites. Inz. Chem. Proces. 2003, 24 (2), 293–310.
mz ) amount of adsorbent (zeolite) in packed bed [kg] (8) Sowerby, B.; Crittenden, B. D. A vapor-phase adsorption and
P ) pressure [Pa] desorption model for drying the ethanol-water azeotrope in small columns.
q* ) equilibrium sorption capacity [mol · kg-1] Chem. Eng. Res. Des. 1991, 69 (A1), 3–13.
qs ) saturation loading capacity [mol · kg-1] (9) Tian, Y.; Zhang, M.; Dong, X. Adsorption and diffusion equilibrium
of water on 3A molecular sieve. Shiyou Huagong 2004, 33 (10), 932–936.
qj ) volume averaged pellet adsorbate loading, qj ) (3/R3)∫R0 q(r) r2 (10) Gorbach, A.; Stegmaier, M.; Eigenberger, G. Measurement and
dr [mol · kg-1] Modeling of Water Vapor Adsorption on Zeolite 4A-Equilibria and Kinetics.
QL ) band heater power output per unit bed length [W · m-1] Adsorption 2004, 10 (1), 29–46.
9260 Ind. Eng. Chem. Res., Vol. 48, No. 20, 2009

(11) Ruthven, D. M. Principles of Adsorption and Adsorption Processes; (20) Carmo, M. J.; Gubulin, J. C. Ethanol-water adsorption on com-
Wiley-Interscience: New York, 1984. mercial 3A zeolites: kinetic and thermodynamic data. Braz. J. Chem. Eng.
(12) Simo, M.; Brown, C. J.; Hlavacek, V. Simulation of pressure swing 1997, 14 (3), 217–224.
adsorption in fuel ethanol production process. Comput. Chem. Eng. 2008, (21) Lalik, E.; Mirek, R.; Rakoczy, J.; Groszek, A. Microcalorimetric
32 (7), 1635–1649. study of sorption of water and ethanol in zeolites 3A and 5A. Catal. Today
(13) Glueckauf, E.; Coates, J. E. Theory of chromatography. J. Chem. 2006, 114 (2-3), 242–247.
Soc. 1947, 1315. (22) Ruthven, D. M.; Farooq, S.; Knaebel, K. S. Pressure Swing
(14) Sircar, S.; Hufton, J. R. Why Does the Linear Driving Force Model Adsorption; VCH Publishers, Inc.: New York, 1994.
for Adsorption Kinetics Work. Adsorption 2000, 6, 137–147. (23) Silva, J. A. C.; Rodrigues, A. E. Fixed-Bed Adsorption of n-Pentane/
(15) Schiesser, W. E. The Numerical Method of Lines, 1st ed.; Academic Isopentane Mixtures in Pellets of 5A Zeolite. Ind. Eng. Chem. Res. 1997,
Press: San Diego, CA, 1991. 36 (9), 3769–3777.
(16) Petzold, L. A description of DASSL: a differential algebraic system
solver. In IMACS Trans. Scientific Computing Vol. 1; Stepleman, R. S., (24) Do, D. D. Analysis of Adsorption Kinetics in a Zeolite Particle. In
Ed.; North-Holland: Amsterdam, 1993; pp 65-68. Adsorption Analysis: Equilibria and Kinetics; Imperial College Press: London,
(17) Dixon, A. G.; Cresswell, D. L. Theoretical prediction of effective 1998.
heat transfer parameters in packed beds. AIChE J. 1979, 25 (4), 663–676.
(18) Wakao, N.; Funazkri, T. Effect of fluid dispersion coefficients on ReceiVed for reView March 18, 2009
particle-to-fluid mass transfer coefficients in packed beds. Correlation of ReVised manuscript receiVed August 14, 2009
Sherwood numbers. Chem. Eng. Sci. 1978, 33 (10), 1375–1384. Accepted August 28, 2009
(19) Reid, R. C.; Prausnitz, J. M.; Poling, B. E. Properties of Gases
and Liquids, 4th ed.; McGraw-Hill: New York, 1987. IE900446V
ABOUT THERMAL KINETICS
At the forefront of new techniques and technologies, Thermal Kinetics has
been providing full-service process equipment engineering, development,
and design services since 1999. We offer advanced energy-saving
solutions, patented technological integration, and sophisticated process
plant development — proudly matching each of our customers’ specific
needs with optimum productivity and profitability.

With an emphasis on respect, teamwork, and timeliness, our innova-

tion-driven and solutions-focused team has the industry insight and
experience to guide your project from start to finish.

Learn More VIEW OUR INFOGRAPH

on the Simulation on the Modeling of PSA

of the PSA Process Process in Fuel Ethanol
Production.

Learn More About

Thermal Kinetics
CONTACT US TODAY!

Thermal Kinetics Engineering

716 691-3291 | Fax: 716 691-3294
Innovation Driven Solutions Focused 85 Northpointe Parkway, STE. 2, Amherst, NY 14228
www.thermalkinetics.net | info@thermalkinetics.net