Energy & Fuels 2005, 19, 2545-2555 2545

Upgrade of Methane from Landfill Gas by Pressure

Swing Adsorption
Simone Cavenati, Carlos A. Grande,* and Alı́rio E. Rodrigues
Laboratory of Separation and Reaction Engineering (LSRE), Department of Chemical
Engineering, Faculty of Engineering, University of Porto, Rua Dr. Roberto Frias s/n,
4200-465, Porto, Portugal

Received March 22, 2005. Revised Manuscript Received July 19, 2005

This work focuses on the production of pipeline grade methane from landfill gas (LFG). Vacuum
pressure swing adsorption technology using a kinetic adsorbent, Carbon Molecular Sieve 3K
(Takeda), was employed for the separation of methane-carbon dioxide mixture. Adsorption
equilibrium and kinetics of methane and carbon dioxide are reported at 298, 308, and 323 to
model the adsorption-based process. A four-step Skarstrom-type cycle was employed comprising
pressurization, feed, counter-current blowdown, and counter-current purge with product. Co-
current pressurization with feed stream and counter-current pressurization with product were
evaluated. The separation of a mixture of CH4 (55%)-CO2 (45%) was tested using two different
four-step cycles: pressurization with feed stream, feed, blowdown, and purge with product and
pressurization with product, feed, blowdown, and purge with product. The results indicate that
purity of methane higher than 96% can be obtained with recovery higher than 75%. The difference
of the performance in the non-adiabatic and adiabatic cases was also studied. It was observed
that the temperature in the column increases allowing a faster and more important desorption
of carbon dioxide in the blowdown step, retaining more CO2 in the column and improving methane
purity and recovery.

Introduction LFG is a multicomponent mixture containing mainly

Accumulation of greenhouse gases (GHG) in the
atmosphere is responsible for increased global warming
of our planet. In developed countries, carbon dioxide
accounts for 83% of the GHG emissions, while methane
is the most important non-CO2 GHG responsible for
global warming with nearly 10% of total GHG emis-
sions.1 Methane emissions are even more important in
developing countries where its percentage can be higher
than 20%.2 Despite the much smaller amounts of
methane released to atmosphere, the greenhouse warm-
ing potential (GWP) of this gas is 21% higher than
carbon dioxide.3 Also, the life of methane molecules in
the atmosphere is 10 times smaller than carbon dioxide,
so any reduction in methane emissions is much more
important in the short- and medium-term atmosphere
reconstruction.
Landfill gas (LFG) constitutes one of the main sources
of methane emissions in many countries. By LFG we
understand the gas released from aerobic decomposition
of organic residues in municipal solid wastes (MSW).
* To whom correspondence should be addressed: Phone: +351 22
508 1669; Fax: +351 22 508 1674. E-mail: cgrande@fe.up.pt.
(1) Pritchard, C. L. Chasing the Dragon’s Tail: Life Cycle Carbon
Analysis Applied to CO2 Sequestration; 7th International Conference
on Greenhouse Gas Control Technologies. Vancouver, Canada, Sep-
tember 5-9, 2004.
(2) Baumert, K.; Pershing, J.; Herzog, T.; Markoff, M. Climate
Data: Insights and Observations; Pew Center on Global Climate
Change: 2004; pp 5-6.
(3) U.S. EPA. Technical and Economic Assessment of Potential to
Upgrade Gob Gas to Pipeline Quality. Document 430-R-97-012;
December 1997; www.epa.gov, accessed November 2004.
methane and carbon dioxide while many other gases
(contaminants) are present (in ppm order) almost
always below 1%. The molar fraction of methane in LFG
is between 0.45 and 0.65,4-6 balanced by carbon dioxide
and contaminants. The amount and nature of these
contaminants (sulfur-bearing, aromatic, and chlorinated
compounds) depend strongly on the waste composition
and aging of the landfill.7 Contaminant removal from
LFG was studied by different techniques including
adsorption and catalytic/sorption hybrid process.7,8 It
was also reported that important amounts of nitrogen
(up to 20%) appear when LFG is collected at pressures
below atmospheric to reach higher methane yields.8,9
In Portugal as well as in many other countries, which
have to import natural gas or fossil fuels, LFG can be
treated as an important national resource of directly
available methane. This reason, together with a tighter
control in methane emissions to meet Kyoto Protocol
targets, puts LFG into consideration for energy produc-
tion. According to present alternatives, there are several
paths to recover energy from LFG:10 heat recovery from
actual flares, energy and/or combined heat and power
(4) Sircar, S. Sep. Sci Technol. 1988, 23, 519-529.
(5) Brosseau, J.; Heitz, M. Atmos. Environ. 1994, 28, 285-293.
(6) Kromann, A.; Christensen, T. H. Waste Manage. Res. 1998, 16,
437-445.
(7) Shin, H.-C.; Park, J.-W.; Song, H.-C. Environ. Pollut. 2002, 119,
227-236.
(8) He, C.; Herman, D. J.; Minet, R. G.; Tsotsis, T. T. Ind. Eng.
Chem. Res. 1997, 36, 4100-4107.
(9) Knaebel, K. S.; Reinhold, H. E. Adsorption 2003, 9, 87-94.
(10) Goossens, M. A. Renewable Energy 1996, 9, 1015-1018.

10.1021/ef050072h CCC: $30.25 © 2005 American Chemical Society

Published on Web 08/25/2005
2546 Energy & Fuels, Vol. 19, No. 6, 2005
(CHP) applications, purification for further incorpora-
tion in methane pipeline network, and reaction to
produce chemical feedstock (i.e., methanol and diesel
fuel). Many catalytic uses were studied in the last years
for a profitable use of LFG, like dry methane reform-
ing,11 acetic acid, or methanol transformation among
others.
In this work, we focused on LFG enrichment to
produce methane that meets pipeline grade specifica-
tions: maximum content of 2% of CO2 to avoid pipeline
and equipment corrosion. We evaluated the use of
Vacuum pressure swing adsorption (VSA-PSA) technol-
ogy. The adsorbent employed for this separation was
Takeda Carbon Molecular Sieve 3K (CMS3K) as a
kinetic adsorbent to perform this separation. Previous
studies by Kapoor and Yang12 used Bergbau-Forchung
CMS adsorbents to perform this separation obtaining
a methane recovery of 93.4% with a purity of of 93.3%.
VSA-PSA benchmark experiments were performed at
ambient temperature (304-308 K), feed pressure of 320
kPa, and methane content of 55% balanced by carbon
dioxide. To model the VSA-PSA process, adsorption
equilibrium and kinetics of pure gases in CMS3K were
measured at 298, 308, and 323 K. The VSA-PSA cycle
is a Skarstrom-type comprising pressurization, feed,
counter-current blowdown, and counter-current purge
with methane. Co-current pressurization with feed and
counter-current pressurization with product were com-
pared in terms of purity and recovery of methane. Also,
the performance of the non-adiabatic process was evalu-
ated under adiabatic conditions (thermally isolated
column) for comparison.
Experimental Section
Carbon molecular sieve 3K-161 was kindly provided by
Takeda Corporation (Tokyo, Japan). This adsorbent has the
same average micropore diameter as CMS3A, but only with
different pellet diameter. Adsorption equilibrium of pure gases
was performed in a Rubotherm magnetic suspension microbal-
ance (Siegen, Germany) operated in a closed system. The same
sample (around 3.6 g) was used in all the experiments. The
pressure transducer used for pressure measurements has an
error of ( 0.10 kPa that together with the error of the
microbalance (( 0.00002 g) gave accurate measurements in
the whole pressure range studied. The activation of the sample
was carried out under vacuum at 523 K overnight. The heating
rate to reach this temperature was 1 K/min. Isotherms were
measured at 298, 308, and 323 K from 0 to 500 kPa. Adsorption
and desorption measurements were performed, and all the
isotherms were reversible.
When the initial point of an equilibrium isotherm was
measured, an initial pressure lower than 10 kPa was always
established. The adsorption uptake was measured by recording
weight in predefined fixed intervals of time (10 s for carbon
dioxide and 2 min for methane) with MessPro software
(supplied with the Rubotherm microbalance equipment). The
adsorption uptake at the different temperatures was used to
determine micropore diffusivity coefficients at infinite dilution
of methane and carbon dioxide.
The VSA-PSA experimental setup is composed of a single-
column unit that operates under unsteady-state conditions
simulating the real operation of a single column in a continu-
ous multi-column VSA-PSA. The equipment allows different
(11) Juan-Juan, J.; Román-Martı́nez, M. C.; Illán-Gómez, M. J. Appl.
Catal. A 2004, 264, 169-174.
(12) Kapoor, A.; Yang, R. T. Chem. Eng. Sci. 1989, 44, 1723-1733.
Cavenati et al.
cycle configurations and can actually work with a generic
ternary mixture where single gas flow rates are individually
controlled within a pressure range from 0 to 700 kPa and from
ambient temperature to 673 K. Temperature inside the column
is measured in three different points: 0.17, 0.43, and 0.68 m
from the feed inlet. The complete equipment setup was
reported elsewhere.13 The molar flow rates of gases exiting the
column are measured in a gas chromatograph (Chrompack
9000, The Netherlands) with thermal conductivity and flame
ionization detectors in series. The column used in the GC was
a CP-Poraplot Q (Varian, The Netherlands) with a flow rate
of 7.0 mL/min of helium (used as carrier gas and as TCD
reference gas) at constant temperature of 313 K.
All gases used in this paper were provided by Air Liquide
and used as received: methane N35, carbon dioxide N24, and
helium N50 (purity greater than 99.95, 99.4, and 99.999%,
respectively).
Adsorption of Methane and Carbon Dioxide in
CMS 3K
Adsorption Equilibrium of Pure Gases. The ad-
sorption equilibrium expresses the degree of interaction
of the adsorbate (gas) with the surface of the adsorbent
by specific physical and/or chemical forces. We have
used the multisite Langmuir model14 to fit the adsorp-
tion equilibrium of pure gases and predict the behavior
of multicomponent mixtures based on pure component
parameters. If we neglect adsorbate-adsorbate interac-
tions, the multicomponent extension of this model can
be expressed as
( ) [ ∑( )]
qi* qi* ai
) KiPyi 1 - (1)
qmax,i i qmax,i
where qmax,i is the maximum amount adsorbed of
component i, ai is the number of neighboring sites
occupied by a molecule of component i, and Ki is the
adsorption constant that has an Arrhenius exponential
temperature dependence as:
Ki ) Koi exp(-∆Hi/RgTs) (2)
where Koi is the adsorption constant of component i at
the limit of T f ∞, and (-∆Hi) is the isosteric heat of
adsorption of component i at zero coverage. The satura-
tion capacity of each component is imposed by the
thermodynamic constraint aiqmax,i ) constant.
Adsorption equilibrium of pure methane and pure
carbon dioxide was obtained at 298, 308, and 323 K in
the pressure range of 0-500 kPa. Experimental data
obtained together with the fitting of the multisite
Langmuir model are presented in Figure 1. All the
isotherms reported showed complete reversibility. The
CMS3K sample adsorbs more carbon dioxide than
methane, particularly at low pressures. The plot in the
Virial domain for the isotherms at 308 K is shown in
Figure 2 as an example of how the equilibrium selectiv-
ity decreases with pressure and the good fitting of the
model at low pressures, fundamental for good predic-
tions of binary adsorption equilibrium.15 All fitting
(13) Da Silva, F. A. Cyclic Adsorption Processes: Application to
Propane/Propylene Separation. Ph.D. Dissertation, University of Porto,
Portugal, 1999.
(14) Nitta, T.; Shigetomi, T.; Kuro-Oka, M.; Katayama, T. J. Chem.
Eng. Jpn. 1984, 17, 39-45.
(15) Talu, O. Adv. Colloid Interface Sci. 1998, 76-77, 227-269.
Upgrade of Methane from Landfill Gas Energy & Fuels, Vol. 19, No. 6, 2005 2547

Adsorption equilibrium of methane on CMS3A in this

temperature range was reported in previous works.12,16-19
The isotherms presented in this work are in good
agreement with the data of Ma et al.16 The data also
agree with those obtained in CMS membranes in the
0-100 kPa pressure range.20
Carbon dioxide in CMS3A from Takeda was reported
previously,17-19,21 and data reported in this work is in
good agreement with all of these works as well as with
other CMS samples from other customers.22
Adsorption Kinetics of Pure Gases. Carbon mo-
lecular sieves are bidisperse adsorbents, although for
gas adsorption the controlling resistance is almost
always located in the micropores.23 According to the
preparation of the CMS material, the constriction of the
pores may be located in the micropore mouths or
distributed inside them. Both materials will exhibit
micropore resistance, although when a restriction is
located at the pore mouth, a surface barrier resistance
may share or even control the diffusion process.24
Successful description of diffusion in pores with mouth
restriction was achieved with a dual-resistance model:
barrier mass transfer resistance at the mouth of the
micropore and micropore resistance18,25 and also using
a lumped model.23,26 It has to be pointed out that both
models require experimental data for constant deter-
mination: linear driving force (LDF) constant in the
Figure 1. Adsorption equilibrium of methane (a) and carbon lumped model plus micropore diffusivity and barrier
dioxide (b) on Takeda CMS3K at 298, 308, and 323 K. Solid mass transfer coefficient in the distributed model.
lines are fitting with multisite Langmuir model. In this work, the adsorption kinetics was determined
using the uptake curves of adsorption at low partial
pressures where linear adsorption equilibrium can be
assumed in the analysis. The results were fitted using
the dual-resistance model with micropore diffusivity at
∞
infinite dilution (Dµ,i ) and surface barrier (kb,i) resis-
tances. The micropore diffusivity depends strongly on
the adsorbed phase concentration as will be explained
later for VSA-PSA applications. Complete modeling of
uptake curves with the dual-resistance model can be
found elsewhere18 and is summarized in Table 2. The
∞
values of the constants (Dµ,i and kb,i) that best fitted
the curves were obtained by minimization of least
squares error function using the fmins subroutine in
MATLAB 6.0 (The Mathworks Inc). Micropore and
surface barrier mass transfer coefficients have an
Figure 2. Virial plot of adsorption equilibrium of methane exponential dependence with temperature described by
and carbon dioxide at 308 K. Solid lines are the fitting of the
multisite Langmuir model. ∞ 0
Dµ,i ) Dµ,i exp(-Ea,i/RgTs) (3)
Table 1. Multisite Langmuir Adsorption Equilibrium
0
Parameters of Methane and Carbon Dioxide on CMS3K kb,i ) kb,i exp(-Eb,i/RgTs) (4)
(Takeda)a
0 0
qmax,i Koi ai -∆Hi where Dµ,i and kb,i are the preexponential terms and
gas (mol/kg) (kPa-1) (-) (kJ/mol)
CH4 11.797 2.48 × 10-10 6.303 33.674 (16) Ma, Y. H.; Sun, W.; Bhandarkar, M.; Wang J.; Miller, G. W.
Sep. Technol. 1991, 1, 90-98.
CO2 8.974 1.73 × 10-8 8.287 38.947 (17) Jayaraman, A.; Chiao, A. S.; Padin, J.; Yang, R. T.; Munson C.
a These parameters are valid from 298 to 323 K and from 0 to L. Sep. Sci. Technol. 2002, 37, 2505-2528.
(18) Qinglin, H.; Sundaram, S. M.; Farooq, S. Langmuir 2003, 19,
500 kPa.
393-405.
(19) Bae, Y. S.; Lee, C. H. Carbon 2005, 43, 95-107.
parameters of the multisite Langmuir model for pure (20) Chen; Y. D.; Yang, R. T. Ind. Eng. Chem. Res. 1994, 33, 3146-
methane and carbon dioxide are shown in Table 1. The 3153.
(21) Rutherford, S. W.; Nguyen, J. E.; Coons, J. E.; Do, D. D.
parameters given in Table 1 were the ones that best Langmuir 2003, 19, 8335-8342.
fitted the data (satisfying thermodynamic constrains) (22) Reid, C. R.; Thomas, K. M. Langmuir 1999, 15, 3206-3218.
but lack physical meaning, the reason they should be (23) Farooq, S.; Qinglin, H.; Karimi, I. A. Ind. Eng. Chem. Res. 2002,
41, 1098-1106.
used only for the pressure and temperature ranges (24) Srinivasan, R.; Auvil, S. R.; Schork, J. M. Chem. Eng. J. 1995,
reported in this work. 57, 137-144.
2548 Energy & Fuels, Vol. 19, No. 6, 2005 Cavenati et al.

Table 2. Mathematical Model for Determination of Kinetic Parameters from Uptake Curves
mass balance in gravimetric cell
C(t) ) C0 - [ ms
Vc
〈q
j (t)〉 ]
2
∫
Rp
average amount adsorbed in extrudates 〈q
j〉 ) j R dR
q
Rp2 0

( )
mass balance in the extrudates ∂Cp ∂qj  pD p ∂ ∂Cp
(with boundary and initial conditions) p + Fp ) R
∂t ∂t R ∂R ∂R

pDp
∂C
∂R (t,Rp) |
) kf(C - Cp|(t,Rp))

∂Cp
∂R | (0,t)
)0

C(R,t)0) ) 0
3
∫
rµ
average amount adsorbed in micropores j)
q qr2 dr
rµ3 0

Fick law diffusion in micropores ∂q 1 ∞ ∂ 2∂q

(with boundary and initial conditions) ) D
∂t r2 µ ∂r
r
∂r [ ]
3 ∞∂q
D
rc µ ∂r | r)rµ
) kb(q* - q)|r)rµ

∂q
∂r | (0,t)
)0

q(r,t)0) ) 0

Ea,i and Eb,i are the activation energy of micropore
diffusion and the activation energy of surface barrier
resistance for component i, respectively.
The kinetics of adsorption of both gases at 308 K is
presented in Figure 3. The pure-gas uptakes are pre-
sented as a function of the square root of time to
enhance the effect of the surface barrier at the mouth
of the micropore,24 observable by a small slope in the
initial moments. Note that almost 3 d of contact between
adsorbent and gas are required to reach methane
equilibrium. This behavior was also reported in this
kind of microporous adsorbents.16,17,19,27 In the adsor-
bent studied here, micropore resistance and surface
barrier at the mouth of the micropore share control of
the diffusion of methane. The order of magnitude of Dµ/
rµ2 reported in all these works is comparable to the one
obtained here, but in this paper we are also including
the surface barrier resistance. According to this dual-
resistance model, the parameters are in agreement to
the ones previously reported for CMS3A Takeda adsor-
bents.18
Figure 3. Adsorption kinetics of methane and carbon dioxide
into Takeda CMS3K at 308 K. Solid lines are the fitting of
the dual resistance model (micropore and surface barrier at
the mouth of the micropores resistances).
Diffusion of carbon dioxide on CMS3A was also
studied in the literature.18,19,22,28 Carbon dioxide is the
smaller molecule studied in this work (with a kinetic
diameter of 3.3 Å against 3.8 Å of methane36), and a
faster diffusion of this gas was expected when compared
to methane. In the uptake of CO2 (see Figure 3), the
initial difficulty to diffuse due to the surface barrier
resistance was not observed. Using only the micropore
resistance, we could fit the curves without requirement
of the surface barrier resistance at the mouth of the
micropore. The energy of activation of the micropore
diffusivity is in agreement with other works with an Ea
) 25.5 kJ/mol as compared with 26.6 kJ/mol,22 but in
our work the surface barrier resistance was not em-
ployed in contrast to other work.18 The absence of
surface barrier resistance for carbon dioxide diffusion
may be related to the activation protocol performed at
higher temperature than in the other mentioned works.
Solid lines in Figure 3 correspond to the fitting of the
dual-resistance model for methane and only micropore
diffusion for carbon dioxide. The parameters to describe
diffusion of the pure gases on Takeda CMS3K are
reported in Table 3.
(25) Loughlin, K. F.; Hassan, M. M.; Fatehi, A. I.; Zahur, M. Gas
Sep. Purif. 1993, 7, 264-273.
(26) Chagger, H. K.; Ndaji, F. E.; Sykes, M. L.; Thomas, K. M.
Carbon 1995, 33, 1405-1411.
(27) Ackley, M. W.; Yang, R. T. AIChE J. 1990, 36, 1229-1238.
(28) Rutherford, S. W.; Coons, J. E. Carbon 2003, 41, 405-411.
(29) Da Silva, F. A.; Rodrigues A. E. Ind. Eng. Chem. Res. 2001,
40, 5758-5774.
(30) Grande, C. A.; Rodrigues, A. E. Ind. Eng. Chem. Res. 43, 2004,
8057-8065.
(31) Ruthven, D. M.; Farooq, S.; Knaebel, K. S. Pressure Swing
Adsorption; VCH Publishers: New York, 1994.
(32) Humphrey, J. L.; Keller, G. E., II. Separation Process Technol-
ogy; McGraw-Hill: New York, 1997.
(33) Yang, R. T. Gas Separation by Adsorption Processes; Butter-
worths: Boston, 1987.
(34) Ruthven, D. M. Principles of Adsorption and Adsorption
Processes; Wiley: New York, 1984.
(35) Incropera, F. P.; Witt, D. P. D. Fundamentals of Heat and Mass
Transfer, 4th ed.; John Wiley & Sons: New York, 1996.
(36) Breck, W. D. Zeolite Molecular Sieves; Wiley: New York, 1974.
Upgrade of Methane from Landfill Gas Energy & Fuels, Vol. 19, No. 6, 2005 2549

Table 3. Adsorption Kinetic Parameters of Methane and It was previously reported that, in CMS adsorbents,
Carbon Dioxide Diffusion in CMS3K (Takeda) there is strong dependence of the micropore diffusivity
D°µ,i Ea,i k°b,i Eb,i with concentration.18 To correct the micropore diffusion
gas (s-1) (kJ/mol) (s-1) (kJ/mol) coefficient, we have used Darken’s law:33
CH4 2.77 35.271 268.7 41.212
CO2 22.12 25.551
∞
d ln(Pi)
Dµ,i ) Dµ,i (7)
d ln(qi)
Separation of Methane-Carbon Dioxide by
VSA-PSA
For this specific separation, energetic effects (non-
Binary Fixed Bed Behavior. Once adsorption equi- isothermal operation) are expected due to an important
librium and kinetics of adsorption of pure gases were release of heat when adsorption of carbon dioxide takes
measured, the VSA-PSA behavior can be modeled. As place. The mathematical model used to describe the
mentioned before, in some CMS adsorbents, a lumped fixed-bed and VSA-PSA behavior has the following
parameter model using a LDF simplification could also assumptions:
describe the kinetics of diffusion. We have used this (i) Ideal gas behavior.
simplification to diminish the computational time re- (ii) The void fraction is uniform in the entire column.
quired for simulations of the VSA-PSA process. The bi-
(iii) Heat, mass, and momentum transport in the
LDF constants applicable for bidisperse adsorbents were
radial direction of the column are neglected, and only
calculated using the parameters determined from up-
axial coordinate is considered.
take curves as
(iv) Meso-macropore resistances are lumped in a
single parameter, Kp,i. The macropore diffusivity is
1
Kµ,i ) (5) obtained relating the Knudsen and molecular diffusion
1 rµ2 via the Bosanquet equation.
+
kb,i 15D∞ (v) The micropore resistance and surface barrier at
µ,i
the mouth of the micropore are lumped in a single
15Dp,i parameter, Kµ,i.
Kp,i ) (6) (vi) Macropore and micropore diffusion equations are
Rp2 described by the LDF (linear driving force) model
instead of the mass balances in the macropores and in
where Kµ,i is the LDF constant for mass transfer in the the micropores. This assumption has a large impact in
micropores and Kp,i is the one for macropores, both for computational time.
component i. When there is no surface barrier resistance (vii) Macropore diffusivity and either surface barrier
in the mouth of the micropores, the first term in the resistance at the mouth of the micropore are not a
denominator of eq 5 vanishes. function of the adsorbed phase concentration. The

Table 4. Mathematical Model for Fixed Bed and VSA-PSA Behavior Used for Methane-Carbon Dioxide Separation with
CMS3K (Takeda)
component mass balance on the column
c
∂Ci
∂t
∂
∂z(
) c DaxCT
∂yi
∂t
-
∂(uiCi)
∂z )- (1 - c)
a′kfi
(C - ci)
Bii + 1 i
2
Ergun equation ∂P 150µg(1 - c) 1.75(1 - c)Fg
)- 3 2
ui + |ui|ui
∂z  d c p  3d c p

LDF equation for the macropores ∂ci ∂〈qi〉 Bii

p + Fp ) pKp,i (C - ci)
∂t ∂t Bii + 1 i
LDF equation for the micropores ∂〈qi〉
) Kµ,i(qi* - 〈qi〉)
∂t
gas-phase energy balance
cCtC̃v
∂Tg
∂t
)
∂ ∂Tg
λ
∂z ∂z ( )
- uiCtC̃p
∂Tg
∂z
+ cRgTg
∂CT
∂t
-
2hw
(1 - c)a′hf(Tg - Ts) - (T - Tw)
Rw g
n n ∂Ts
∑ ∑
solid-phase energy balance
(1 - c)[p ci C̃vi + Fpwc 〈qi〉C̃v,ads,i + FpC̃ps] )
i)1 i)1 ∂t
∂ci n ∂〈qi〉
(1 - c)pRgTs + Fb
∂T i)1
(-∆Hi) ∑
∂t
+ (1 - c)a′hf(Tg - Ts)

wall energy balance ∂Tw

FwC̃pw ) Rwhw(Tg - Tw) - RwlU(Tw - T∞)
∂t
Dw 1

( )
Rw ) Rwl )
e(Dw + e) Dw + e
(Dw + e) ln
Dw
Bosanquet equation 1
Dp,i
) τp ( 1
+
1
Dm,i Dk,i )
2550 Energy & Fuels, Vol. 19, No. 6, 2005 Cavenati et al.

Table 5. Details of Equipment and Adsorbent Properties

Used for CH4-CO2 Separation Using CMS3K Extrudates
(Takeda)
column length, m 0.83
column porosity 0.33
wall density, kg/m3 8238
wall heat transfer coeff, W/m2‚K 35
overall heat transfer coeff, W/m2K 25
pressurization flow rate, SLPM 1.50
feed flow rate, SLPM 1.50
purge flow rate, SLPM 0.60
column radius, m 0.0105
bulk density, kg/m3 715
wall specific heat, J/kg‚K 500
pellet radius, m 9 × 10-4
pellet density, kg/m3 1060
pellet porosity 0.46
pellet tortuosity 2.0
solid specific heat, J/kg‚K 880

Darken law describes Micropore diffusivity dependence

with concentration.
The equations of the model proposed are detailed in
Table 4. This model was already reported in the
literature and was validated for propane-propylene
separation using different microporous adsorbents.29,30
To use the mathematical model proposed, mass and heat
transport parameters were estimated according to cor-
relations existing in the literature.33-35
To validate the results obtained from simulations and
test the multicomponent adsorption equilibrium predic-
tions of the multisite Langmuir model and the bi-LDF
simplification for diffusion, a fixed-bed experiment of
the binary mixture was performed.
A binary fixed-bed experiment was performed at
ambient temperature (303 K) using a total pressure of
320 kPa with a mixture containing 55% of methane
balanced by carbon dioxide. The column used in this
experiment was the same one employed for VSA-PSA
experiments. Column details required for modeling as
well as some properties of the adsorbent are reported
in Table 5. The results of the binary breakthrough curve
of CH4-CO2 using a total inlet flow rate of 1.00 SLPM
(standard liter per minute) are shown in Figure 4.
Experimental data obtained in the laboratory unit are
molar flow rate of the gases exiting the column and
temperature profiles at three different positions of the
column (0.17, 0.43, and 0.68 m from feed inlet). As can
be seen in Figure 4a (flow rate exiting the column),
methane is almost not adsorbed in the column (due to
the very slow diffusion in the micropore network) while
carbon dioxide is much more retained, confirming the
high kinetic selectivity of the adsorbent toward carbon
dioxide. It can also be noticed from Figure 4b that
temperature variations due to carbon dioxide adsorption
are important (higher than 20 K) and cannot be ignored.
Temperature increase due to adsorption of methane is
negligible. It is also important to mention that the
velocity of the thermal wave is close to the velocity of
carbon dioxide concentration wave. The solid lines in
Figure 4 represent the predictions of the mathematical
model using pure component adsorption equilibrium
and kinetic parameters as well as transport parameters
reported in Table 5. The only fitting parameter of the
model was the overall heat transfer parameter (shown
in Table 5) that will be used in further predictions of
VSA-PSA process. The mathematical model proposed
was able to predict well the behavior of the binary
Figure 4. Binary methane-carbon dioxide (55% of CH4
balanced by CO2) breakthrough curve at 303 K and 320 kPa
total pressure. (a) Molar flow rate of each gas exiting the
column, and (b) temperature profiles inside the column at 0.17,
0.43, and 0.68 m from feed inlet.
mixture in a fixed bed. At least in the conditions used
in this work, Darken’s law provided a successful cor-
rection of the micropore diffusion coefficients in the
nonlinear regions of the isotherms, particularly for the
case of carbon dioxide.
VSA-PSA Results with Co-current Pressuriza-
tion with Feed Stream. It was mentioned before that
LFG is a multicomponent mixture where methane and
carbon dioxide are the most important gases. Several
contaminants (less than 1%) are also present, and the
stream is saturated with moisture. The VSA-PSA
process presented here for methane purification has to
be located after a cleanup process to remove trace
contaminants that may adsorb strongly or irreversibly
in the CMS adsorbent and after a TSA or membrane
process to remove moisture9 to satisfy pipeline grade
specifications.
The basic cycle configuration to recover the less
adsorbed compound (methane) in the feed step (at high
pressure) is a four-step Skarstrom-type cycle with
pressurization and feed with CH4/CO2 mixture (process-
ing steps) and counter-current blowdown and purge
with product (adsorbent regeneration steps). The ex-
periments in this work were performed at ambient
temperature and feed pressure of 320 kPa. To regener-
ate the adsorbent, the blowdown step was performed
at a pressure of 10 kPa because of the steep curvature
of CO2 isotherms at low pressure. The length of the bed
as well as the total flow rate and composition of the
steps were previously specified, and we have studied the
effect of the different step times (i.e., the residence time
of gases in the PSA column) in the performance of the
unit in terms of purity and recovery of methane.
Upgrade of Methane from Landfill Gas Energy & Fuels, Vol. 19, No. 6, 2005 2551

Figure 5. VSA-PSA experiment using co-current pressurization, feed, blowdown, and purge with product cycle: (a) pressure at
the end of the column in one cycle; (b and c) molar flow rate of CH4 and CO2 exiting the column in cycles 1 and 46, respectively;
(d and e) temperature history at 0.17 and 0.68 m from feed inlet, respectively; (f) temperature profile inside the column at the
end of each step; (g and h) simulated profiles of amount adsorbed of methane and carbon dioxide, respectively, at the end of each
step. Experimental conditions are reported in Table 4 (run 1), and feed temperature is 306 K.

At the temperature and pressure conditions used in
the experiments, we have experimentally measured that
operating at constant flow rate the pressurization time
(tpres) cannot take less than 70 s. For that reason, three
process variables can be studied: feed, blowdown, and
purge step times (tfeed, tblow, and tpurge, respectively).
An example of a laboratory experiment can be ob-
served in Figure 5. The experimental variables shown
are the molar flow rate of gases exiting the column in
cycles 1 and 46, pressure at the exit of the column, and
temperature histories at three different positions of the
column (0.17, 0.48, and 0.63 m from feed inlet). In the
2552 Energy & Fuels, Vol. 19, No. 6, 2005 Cavenati et al.

Table 6. VSA-PSA Results for Methane Upgrading from Table 7. VSA-PSA Results for Methane Upgrading from
LFG with Four-Step Cycle with Co-current LFG with Four-Step Cycle with Counter-current
Pressurization with Feed at 306 K Pressurization with Product at 306 K
tpres, tfeed, tblow, tpurge, Phigh, purity, recovery, tpres, tfeed, tblow, tpurge, Phigh, purity, recovery,
run s s s s kPa % % run s s s s kPa % %
1a 70 100 100 70 320 97.1 79.4 1 40 170 100 20 320 93.5 82.6
2 70 120 120 70 320 96.1 81.9 2 40 150 100 20 320 96.7 78.1
3 70 100 120 70 320 96.0 78.5 3 40 150 100 50 320 96.9 77.8
4 80 100 120 50 400 96.2 70.0 4a 40 150 160 50 320 97.4 77.0
5 80 100 100 90 400 97.9 68.1 5 55 150 160 50 700 98.7 63.9
a Experimental run. a Experimental run.

images, the solid points correspond to experimental data
while the solid line corresponds to the prediction of the
mathematical model. Although, thermal and pressure
cycle variations seem to reach cyclic steady state (CSS)
after 20 cycles, the CSS for all variables was reached
only after 100 cycles, where the amount adsorbed of
methane reached a constant value. Performance pa-
rameters, from cycle 35 to cycle 100, change less than
1%. This delay in achieving CSS is due to the very slow
diffusion of methane in the CMS3K micropores. In this
figure, we have also plotted the simulated temperature
profiles inside the column after the end of each step
when CSS was reached as well as the amount adsorbed
of each gas. Again, due to the very slow diffusion of
methane, it can be observed that the net adsorption-
desorption of methane per cycle (given by the area
between feed and blowdown steps) is very small. It can
be also observed that the higher temperature variations
are located in the initial portion of the column, which
is the most intensively used for CO2 adsorption. Diffu-
sion of CO2 is much faster than that of CH4. Despite
that, the rate of uptake is slow enough to produce a
large mass transfer zone (MTZ), decreasing the produc-
tivity of the VSA-PSA process.32
After confirmation of the validity of mathematical
model to the prediction of the experimental data, some
results from several simulations varying step times are
presented in Table 6. Note that the performance values
from run 1 in Table 6 correspond to the experiment
shown in Figure 5. From these results it can be observed
that there is a net reduction in product purity when the
purge step is decreased. The purity of methane is almost
not affected by a blowdown step higher than 100 s. It
was also observed that if the pressure of the feed step
increases, the purity of methane can be higher, almost
reaching 98% as desired. Purity higher than 98% could
be obtained with a purge step higher than 100 s and
with a recovery around 65%. The purge can displace
carbon dioxide from the product end only partially, but
an increase in the purge step always diminishes recov-
ery of methane.
Counter-current Pressurization with Product.
Another way to displace carbon dioxide from the product
end can be achieved by performing the pressurization
step with methane (less adsorbed gas), counter-current
to the feed direction (i.e., by recycling part of the product
stream). It was demonstrated that this methodology
might effectively enhance the purity of the less adsorbed
compound.31
The procedure to assess the importance of the differ-
ent steps was similar to the one used in the last section.
An experimental run was performed to verify the
validity of the mathematical model, and then simulation
results were employed to study the effect of step times
of this process. Results varying the tfeed, tblow, and tpurge
are presented in Table 7. In this table, a result using a
different (higher) feed pressure is also presented. Clearly
increasing the total pressure of the column, there is a
decrease in the methane recovery counter-balanced by
enhancement of product purity over 98% as desired with
recoveries over 65%, although the increase of product
purity when increasing feed pressure was also observed
with the co-current pressurization step. Also, this
recovery can be improved by compressing and recycling
the purge stream into the feed step. As mentioned
before, the validity of the mathematical model was also
tested by performing run 4 (see Table 7) in the labora-
tory unit. The molar flow rate and temperature histories
of the three thermocouples inside the column in cycle
49 are shown in Figure 6.
If we compare the product purity and recovery shown
in Tables 6 and 7, we can see that the VSA-PSA

Figure 6. VSA-PSA experiment using counter-current pressurization with product, feed, blowdown, and purge with product
cycle: (a) molar flow rate of CH4 and CO2 exiting the column in cycle 49; (b) temperature history at 0.17, 0.43, and 0.68 m from
feed inlet for one cycle when CSS was reached. Experimental conditions are reported in Table 5 (run 4), and feed temperature is
306 K.
Upgrade of Methane from Landfill Gas Energy & Fuels, Vol. 19, No. 6, 2005 2553

Figure 7. Adiabatic VSA-PSA experiment using counter-current pressurization with product, feed, blowdown, and purge with
product cycle: (a) molar flow rate of CH4 and CO2 exiting the column in cycle 49; (b) temperature history at 0.17, 0.43, and 0.68
m from feed inlet for one cycle when CSS was reached. Experimental conditions are similar to run 4 in Table 5, and feed temperature
is 306 K.

Figure 8. Simulated results of non-adiabatic and adiabatic VSA-PSA experiment using counter-current pressurization with
product: (a and b) temperature history at 0.17, 0.43, and 0.68 m from feed inlet for one cycle when CSS was reached and amount
adsorbed of CO2 at the end of each step for non-adiabatic run; (c and d) temperature history at 0.17, 0.43, and 0.68 m from feed
inlet and amount adsorbed of CO2 at the end of each step for adiabatic run. Numbers represent the following: 1, pressurization;
2, feed; 3, blowdown; and 4, purge with product.

performance obtained with co-current or counter-current
pressurization are similar. It can be observed that the
displacement of carbon dioxide from the product end
using a counter-current pressurization with product was
not much different than using a co-current pressuriza-
tion step. In fact, the slow diffusion of methane in the
CMS3K structure cannot displace the CO2 adsorbed. As
the diffusivity of CO2 is also small (when compared to
other PSA processes actually used in industry), its
desorption is almost independent of the direction of the
regeneration flow rate.
Adiabatic Behavior. Once benchmark experiments
and simulations were performed to study the possibility
of obtaining pipeline grade methane from LFG, the next
step is to scale up the process. In previous sections we
have shown considerable temperature variations due to
CO2 adsorption-desorption even when an important
flux of heat can be exchanged with the environment
(column with small diameter in a stirred air medium).
In an industrial-scale process, the diameter of the
column will be larger and not immersed in a stirred
medium, reducing heat transfer and operating close to
adiabatic behavior.
To approximate the adiabatic situation and to study
the importance of heat effects in scaling up of methane
upgrade from LFG, we have covered the VSA-PSA
column with an thermal insulation (wood glass mate-
rial) and performed the same experiment shown in
Figure 6 with counter-current pressurization (run 4 in
Table 7). The molar flow rate and temperature histories
of the three thermocouples inside the column in cycle
47 are shown in Figure 7. Only slight differences can
be observed in the molar flow rate of both gases exiting
the column (more CO2 exiting in the blowdown step of
2554 Energy & Fuels, Vol. 19, No. 6, 2005
the adiabatic run), but the temperature profiles in the
column are completely different. In the adiabatic experi-
ment, the temperature in the middle of the column
remains at a higher temperature than in other positions
of the column in all the experiment, something that was
not observed in the nonadiabatic run.
To compare the behavior of the adiabatic and non-
adiabatic experiment, the simulations of the amount
adsorbed of carbon dioxide and the temperature in the
column at the end of each step when CSS was reached
are shown in Figure 8. As can be seen, as expected, the
temperature variations in the adiabatic run are larger
and temperatures up to 330 K are achieved near the
middle of the column. The simulated profiles of amount
adsorbed of carbon dioxide on both runs are also similar,
but it can be observed that the amount of CO2 processed
(area between the feed step and the blowdown step) in
the adiabatic run is higher than in the non-adiabatic
experiment. This result was confirmed with a CH4
experimental purity of 98.1% and recovery of 77.8%,
0.8% higher than in the non-adiabatic case. It is well-
known that purity or recovery of the less adsorbed
compound (methane) should decrease if the operation
is carried out at a given throughput. In these experi-
ments, we were operating at constant pressures, and
the throughputs of the steps were not kept constant.
The molar flow rate of methane in the feed step is nearly
the same in both (non-adiabatic and adiabatic experi-
ments) experiments. Even though, it can be observed
that the flow rate of carbon dioxide in the blowdown
step in the adiabatic experiment is higher. This is in
agreement with the simulated temperature profiles that
can be observed in Figure 8. For this reason, when this
kinetically controlled system is operated adiabatically,
the decrease in capacity of carbon dioxide due to higher
temperatures in the column is counterbalanced by the
increase of carbon dioxide diffusion. This allows a faster
and more important desorption of carbon dioxide in the
blowdown step, improving methane purity and recovery.
Conclusions
VSA-PSA technology was evaluated for methane
upgrade from LFG sources. Adsorption equilibrium of
methane and carbon dioxide were evaluated in the
adsorbent used, CMS3K (Takeda). Adsorption equilib-
rium of carbon dioxide was higher than for methane in
the range studied (298-323 K). Pure component ad-
sorption equilibrium was well-described by the multisite
Langmuir model. In this adsorbent, the diffusion of
methane into the micropore structure is very slow,
offering a kinetic selectivity of the adsorbent against
CO2. The kinetics of methane diffusion was controlled
by a mixture of micropore diffusion and surface barrier
at the entrance of the micropores, while for carbon
dioxide only micropore diffusion was observed.
The separation of a mixture CH4 (55%)-CO2 (45%)
was tested using a four-step cycle: pressurization with
feed stream, feed, blowdown, and purge with product.
The results indicate that purity higher than 96% can
be obtained with recovery higher than 75%, but to
obtain purity higher than 98% feed pressure has to
increase up to 400 kPa and recovery is around 65%.
Trying to improve methane purity, we have tested
another cycle configuration: counter-current pressur-
Cavenati et al.
ization with product, feed, blowdown and purge with
product. It was observed that with this cycle configu-
ration the displacement of carbon dioxide from the end
of the column was not much different than using
pressurization with feed because methane cannot adsorb
in the microporous structure of CMS and dislocate
carbon dioxide. Using this cycle, the difference of the
performance in the non-adiabatic and adiabatic cases
was also studied. It was observed that the temperature
in the column increases to higher temperatures, allow-
ing a faster and more important desorption of carbon
dioxide in the blowdown step, retaining more CO2 in
the column, and improving methane purity and recov-
ery.
Acknowledgment. The authors are thankful for
financial support from the Foundation for Science and
Technology (FCT) Project POCI/EQU/59330/2004 and
the gift of CMS 4A adsorbent by Takeda Corporation.
C.A.G. acknowledges FCT Grant SFRH/BPD/18888/
2004.
Nomenclature
a′ ) area to volume ratio (m-1)
ai ) number of neighboring sites occupied by component i
Bii ) Biot number of component i
Co ) initial hydrocarbon concentration in uptake measure-
ments (mol/m3)
Ci ) concentration of component i in the gas phase (mol/m3)
ci ) averaged concentration in the macropores for component
i (mol/m3)
Cp ) hydrocarbon concentration in the pores of the adsorbent
(mol/m3)
C̃p ) molar constant pressure specific heat of the gas mixture
(J/mol‚K)
C̃ps ) constant pressure specific heat of the adsorbent (J/kg‚
K)
C̃pw ) specific heat of the fixed bed wall (J/kg‚K)
CT ) total gas concentration (mol/m3)
C̃v ) molar constant volumetric specific heat of the gas mixture
(J/mol‚K)
C̃v,ads,i ) molar constant volumetric specific heat of component
i (J/mol‚K)
Dax ) axial dispersion coefficient (m2/s)
Dk,i ) Knudsen diffusivity of component i (m2/s)
Dm,i ) molecular diffusivity of component i (m2/s)
Dp,i ) macropore diffusivity of component i (m2/s)
Dw ) internal diameter of the column (m)
Dµ,i ) micropore diffusivity of component i (m2/s)
∞
Dµ,i ) micropore diffusivity of component i at infinite dilution
(m2/s)
0
Dµ,i ) diffusivity of component i at infinite temperature (m2/
s)
dp ) pellet diameter (m)
Ea,i ) activation energy of micropore diffusion of component i
(kJ/mol)
Eb,i ) activation energy for micropore barrier resistance of
component i (kJ/mol)
e ) wall thickness (m)
kb,i ) barrier mass transfer coefficient for component i (s-1)
0
kb,i ) micropore barrier resistance at infinite temperature of
component i (s-1)
Ki ) equilibrium constant of component i (kPa-1)
Koi ) equilibrium constant at the limit T f ∞ of component i
(kPa-1)
Kp,i ) linear driving force constant for macropore diffusion of
component i (s-1)
Upgrade of Methane from Landfill Gas Energy & Fuels, Vol. 19, No. 6, 2005 2555

Kµ,i ) linear driving Force constant for micropore diffusion of T ) temperature (K)
component i (s-1) Tg ) temperature of gas phase (K)
hf ) film heat transfer coefficient between gas and solid phase Ts ) solid (extrudate) temperature (K)
(W/m2‚K) Tw ) wall temperature (K)
hw ) film heat transfer coefficient between gas phase and T∞ ) infinite temperature (K)
column wall (W/m2K) U ) global external heat transfer coefficient (W/m2‚K)
kfi ) film mass transfer resistance (m/s) ui ) superficial velocity (m/s)
ms ) mass of adsorbent connected to the microbalance (kg) Vc ) entire microbalance cell volume (m3)
P ) total pressure (kPa) yi ) molar fraction of component i in gas phase
qi ) adsorbed phase concentration of component i (mol/kg) z ) axial distance along the column (m)
qi* ) equilibrium adsorbed phase concentration of component Greek Letters
i (mol/kg)
〈qi〉 ) extrudate averaged adsorbed phase concentration (mol/ Rw ) ratio of the internal surface area to the volume of column
kg) wall (m-1)
qi ) crystal averaged adsorbed phase concentration (mol/kg) Rwl ) ratio of the logarithmic mean surface area of the column
qmax,i ) maximum adsorbed phase concentration of component shell to the volume of column wall (m-1)
i (mol/kg) Fb ) bulk density (kg/m3)
Fg ) gas density (kg/m3)
R ) distance along macroparticles radius (m)
Fp ) adsorbent density (kg/m3)
Rg ) ideal gas constant (8.314 kJ/mol‚K)
Fw ) wall density (kg/m3)
Rp ) pellet radius (m)
λ ) axial heat dispersion (W/m2‚K)
Rw ) radius of the static-batch column (m)
(-∆Hi) ) isosteric heat of adsorption of component i (multisite
r ) distance along microparticles radius (m)
Langmuir model) (kJ/mol)
rµ ) radius of the microparticles (m)
c ) column void fraction
t ) time (s)
p ) porosity of the pellet
tpress ) pressurization step time (s)
τp ) pore tortuosity
tblow ) blowdown step time (s)
µg ) gas viscosity (Pa‚s)
tfeed ) feed step time (s)
tpurge ) purge step time (s) EF050072H