nanomaterials

Article
Application of Nanosize Zeolite Molecular Sieves for
Medical Oxygen Concentration
Mingfei Pan, Hecham M. Omar and Sohrab Rohani *
Department of Chemical and Biochemical Engineering, Faculty of Engineering, University of Western Ontario,
London, ON N6A 5B9, Canada; pmf19516@gmail.com (M.P.); homar7@uwo.ca (H.M.O.)
* Correspondence: srohani@uwo.ca; Tel.: +1-519-661-4116

Received: 19 June 2017; Accepted: 21 July 2017; Published: 25 July 2017

Abstract: The development of a portable oxygen concentrator is of prime significance for patients
with respiratory problems. This paper presents a portable concentrator prototype design using the
pressure/vacuum swing adsorption (PVSA) cycle with a deep evacuation step (−0.82 barg) instead
of desorption with purge flow to simplify the oxygen production process. The output of the oxygen
concentrator is a ~90 vol % enriched oxygen stream in a continuous adsorption and desorption
cycle (cycle time ~90 s). The size of the adsorption column is 3 cm in diameter and 20 cm in length.
A Li+ exchanged 13X nanosize zeolite is used as the adsorbent to selectively adsorb nitrogen from
air. A dynamic model of the pressure and vacuum swing adsorption units was developed to study
the pressurization and depressurization process inside the microporous area of nanosized zeolites.
The describing equations were solved using COMSOL Multiphysics Chemical Engineering module.
The output flow rate and oxygen concentration results from the simulation model were compared
with the experimental data. Velocity and concentration profiles were obtained to study the adsorption
process and optimize the operational parameters.

Keywords: nanosize zeolite; microporous adsorption; bidisperison dynamic model; portable oxygen
concentrator; COMSOL Multiphysics

1. Background
Air pollution is becoming a worldwide problem particularly in heavily populated cities like
Beijing, Los Angeles, and Mexico City. Air pollution leads to acute and chronic respiratory problems
that are becoming a growing concern from both global and individual levels. Wearing disposable face
masks has become a popular method to get some protection from exposure to pollutants. However, for
people with poor autonomous respiration ability, like chronic obstructive pulmonary disease (COPD)
patients, it is hard to breathe through disposable face masks. As a result, there is a growing need to
address the health and quality of life through a light-weight and portable oxygen concentrator with
a medical grade oxygen supply (O2 concentration: ~88–92 vol %). Currently, oxygen concentration
techniques are based on air separation processes such as cryogenic technique, membrane separation
and pressure swing adsorption (PSA) that can effectively produce enriched oxygen using ambient
air as the source. Although cryogenic distillation acts as the leading process for air separation in
an industrial oxygen plant, the pressure swing adsorption (PSA) technique has the advantage of
reducing energy consumption for small production scales. Compared to the membrane separation
technique, which requires higher pressure and expensive selective permeable materials, pressure
swing adsorption using nanosize zeolite adsorbents is a more feasible solution for portable oxygen
concentrator development. Due to the large microporous surface area inside the zeolite adsorbents,
the portable oxygen concentrator using the PSA technique can adsorb nitrogen from air and output
an enriched oxygen stream under high-pressure conditions. The concentrator can be regenerated by

Nanomaterials 2017, 7, 195; doi:10.3390/nano7080195 www.mdpi.com/journal/nanomaterials

Nanomaterials 2017, 7, 195 2 of 19

decreasing the pressure to release the adsorbed nitrogen. In this way, a continuous oxygen stream
supply can be generated to assist an individual and increase the fraction of inspired oxygen. The
oxygen concentration system is usually coupled with a filtration system to remove PM 2.5 and improve
the quality of product gas stream.
The pressure swing adsorption process was patented by Finlayson and Sharp in 1932
(GB Patent 365092) describing an air separation process with pressurization adsorption and
depressurization desorption steps [1]. In 1958, two different pressure swing adsorption units were
patented: one using fractional product purge flow for desorption step (Skarstrom, 1958) and the other
using pressure vacuum adsorption (PVSA) step (Montgareuil and Domine, 1958) [2,3].
The selectivity for zeolite adsorbents to adsorb nitrogen compared to oxygen is due to the
interaction between electrostatic field of the cationic zeolite and the quadrupole moment of the
nitrogen and oxygen. The quadrupole of nitrogen is three times higher than that of oxygen, which
leads to a selective adsorption onto the zeolite surface [4]. The most common type of commercial
zeolite for oxygen concentration process is zeolite 13X due to its outstanding nitrogen to oxygen
adsorption selectivity. However, 13X zeolite modified with Li+ exchange method exhibits a higher
nitrogen adsorption capacity at the active cation sites of the zeolite framework [5–7]. JLOX-101 LiX
zeolite provided by LUOYANG JIANLONG Chemical is designed specifically for a PVSA process with
high oxygen selectivity [8].
The PSA system was first installed in industrial plants as a replacement for cryogenic separation
plants due to lower capital and operating costs. Portable pressure swing adsorption oxygen
concentrators have been designed and developed in the past two decades due to the synthesis of more
effective zeolites with higher gas adsorption capacity and selectivity. The PSA oxygen concentrator is
able to provide a continuous oxygen flow using the ambient air as the gas source. Two-bed or multi-bed
PSA systems are utilized alternating adsorption and desorption operations of the zeolite columns
to achieve continuous oxygen separation. For on-site medical grade oxygen production, a cylinder
gas source or several powerful air compressors are usually installed for the rapid pressure swing
adsorption process to maintain the productivity of high purity oxygen stream [9–11]. Considering the
mobility of small-scale concentrators, a variety of simplified designs would be added to the pressure
swing adsorption process for system miniaturization.
In this paper, we compare the performance of a portable miniature PVSA oxygen concentrator
unit with two adsorption columns, an air compressor, four solenoid valves and a back-pressure
regulator (total weight less than 2 kg) built in our laboratory with the simulation results under
modified Montgareuil and Domine cycle using COMSOL Multiphysics Chemical Engineering module.
Compared to the conventional PSA cycle using purge flow with intermediate evacuation pressure
(−0.6 barg) to regenerate the adsorption column, the deep vacuum (−0.82 barg) desorption step
of the PVSA cycle reduces the column and compressor size from the perspective of industrial
prototype design. A dual-head air compressor is selected to generate pressure and vacuum condition
simultaneously for the dual-column adsorption unit, which simplifies the gas source generation.
The oxygen concentrator unit can provide a ~90 vol % enriched oxygen stream to meet the concentration
requirement of medical-grade oxygen.

2. Materials and Methods

Figure 1 shows the schematic of the oxygen concentrator using a two-column PVSA
cycle. The concentrator consists of two adsorption zeolite columns, a dual-head air compressor
(Parker Hannifin Corp., TTC IIS Miniature Diaphragm Pump, Hollis, NH, USA), five three-way
solenoid valves (Parker Hannifin Corp., X-Valve, Hollis, NH, USA), a back-pressure regulator, an
oxygen sensor (Teledyne GB300 Oxygen analyzer, Thousand Oaks, CA, USA), a surge tank, a pressure
sensor (Freescale Semiconductor, MPXV7002DPT1CTND, Austin, TX, USA) and a cooling fan (avoiding
the temperature rise of the air compressor) [12–15]. The oxygen sensor is an electrochemical transducer
named Micro-fuel Cell calibrated by the Faraday type calibration method based on the product gas
Nanomaterials 2017, 7, 195 3 of 19

from electrolysis of water [16]. The valve sequence is shown in Table 1. The concentrator operation
consists of four different steps. In step 1, column 1 is pressurized by the compressor while column 2 is
connected to ambient air to decrease the pressure. In step 2, the pressure inside column 1 reaches the
working pressure (1.79 barg) with enriched oxygen flowing out through the backpressure regulator
(set at 1.79 barg) to the surge tank. Meanwhile the pressure in column 2 decreases to vacuum (set
Nanomaterials 2017, 7, 195 3 of 19
at −0.82 barg) via the air compressor to regenerate the zeolite. In steps 3 and 4, the adsorption
and desorption
the working positions
pressureof(1.79column
barg)1with
and enriched
2 are reversed
oxygen to regenerate
flowing the saturated
out through column 1 and
the backpressure
regulator (set at 1.79 barg) to the surge tank. Meanwhile the pressure in
pressurize the regenerated column 2. By switching the valve sequence (step 1 → 4 → 1), a continuous column 2 decreases to
vacuum (set at −0.82 barg) via the air compressor to regenerate the zeolite. In steps 3 and 4, the
enriched oxygen stream can be obtained from the surge tank. The flow directions of the two different
adsorption and desorption positions of column 1 and 2 are reversed to regenerate the saturated
types ofcolumn
valves1(normally open and distributor three-way solenoid valves) are shown in Figure 2 with
and pressurize the regenerated column 2. By switching the valve sequence (step 1 → 4 → 1),
two different roles enriched
a continuous in the pressure swing
oxygen stream operations.
can Valve
be obtained from the1 surge
and 2tank.
are The
normally open 3-way
flow directions of the valves
directing thedifferent
two pressurizing
types offlow
valves and countercurrent
(normally open and blowdown flow while
distributor three-way valvevalves)
solenoid 3 andare 4 are distributor
shown
in Figure
3-way valves for2 with two different
pressurizing androles in the pressure swing
depressurizing port operations.
control atValve 1 and 2 are
the column end.normally open to the
According
3-way
adsorption valves directing
isotherm of zeolite theLiX
pressurizing
shown inflow and 3countercurrent
Figure (Santos et al.,blowdown flow while
2008), zeolite valve
capacity 3 and
and selectivity
4 are distributor 3-way valves for pressurizing and depressurizing port control at the column end.
of nitrogen to oxygen is higher under pressure vacuum swing range compared to conventional pressure
According to the adsorption isotherm of zeolite LiX shown in Figure 3 (Santos et al., 2008), zeolite
swing [17]. In order
capacity and to reduce the
selectivity compressor
of nitrogen size, aisdual-head
to oxygen air compressor
higher under pressure vacuumis considered to perform
swing range
pressurization
compared and vacuum operation
to conventional pressure simultaneously.
swing [17]. In order Bytosharing thecompressor
reduce the central motor,
size, a two diaphragms
dual-head
can separately provide
air compressor is pressurized
considered togas and evacuation
perform pressurization pressure.
and vacuumCompared
operation tosimultaneously.
the size and total By power
sharing the central motor, two diaphragms can separately provide pressurized
consumption of two diaphragm compressors, a dual-compressor is less bulky (25% size reduction) gas and evacuation
pressure. Compared to the size and total power consumption of two diaphragm compressors, a dual-
and more efficient (12% input power saving), which benefits the miniaturization of the prototype.
compressor is less bulky (25% size reduction) and more efficient (12% input power saving), which
Summarizing the system design, a PVSA process that uses a vacuum for desorption and pressurization
benefits the miniaturization of the prototype. Summarizing the system design, a PVSA process that
for adsorption can achieve
uses a vacuum relatively
for desorption high capacity,
and pressurization forallowing
adsorption more nitrogen
can achieve adsorption
relatively while using a
high capacity,
smaller allowing
compressor for gas supply.
more nitrogen adsorption while using a smaller compressor for gas supply.

Figure
Figure 1. Schematicsof
1. Schematics of the
the PSA
PSAoxygen
oxygenconcentrator.
concentrator.
Nanomaterials 2017, 7, 195 4 of 19

Table 1. The valve operation sequence for the PVSA cycle.

Nanomaterials
Nanomaterials2017,
2017,7,7,195
195 44ofof19
19
Column 1 Column 2
Step Stage
Table
Table1.1.The
Thevalve
valveoperation
operationsequence forStage
sequencefor the
thePVSA
PVSAcycle.
cycle.
Valve 1 Valve 3 Valve 2 Valve 4
Column 1 Column 2
Step1
Step Pressurization
Stage
Stage P1 Column 1 P1 Blowdown
Stage
Stage P2 Column 2P2
Valve
Valve 1 Valve 3 Valve
Valve22 Valve 4
2 Production P1 1 ValveP13 Vacuum P2 Valve
P1 4
11 Pressurization
Pressurization P1
P1 P1
P1P2 Blowdown
Blowdown P2 P2
3 Blowdown P2 Pressurization P1 P2 P2
P1
22 Production
Production P1
P1 P1
P1P1 Vacuum
Vacuum P2 P1
4 Vacuum P2 Production P1 P2 P1
P1
33 Blowdown
Blowdown P2
P2 P2
P2 Pressurization
Pressurization P1
P1 P1
P1
44 Vacuum P1,P2
P2 representsP1valve positionProduction
1 and 2. P1 P1
Vacuum P2 P1 Production P1 P1
P1,
P1, P2 represents valve position 1and
P2 represents valve position 1 and2.2.

Figure
Figure
Figure 2.2.Flow
2. Flow Flow directions
directions
directions Schematics
Schematics of
Schematicsof the
thenormally
ofthe normallyopen
normally openand
open and
and distributor
distributor
distributor three-way valve
three-way
three-way under
valve
valve under
under
pressure
pressure andand
pressure vacuum
vacuum
and conditions.
conditions.
vacuum conditions.

Figure
Figure 3.3. Adsorption
Adsorption equilibrium isotherms
isotherms for
equilibriumisotherms for nitrogen and oxygen onto zeolite LiX
LiX at
at 20
20at°C
Figure 3. Adsorption equilibrium for nitrogen
nitrogenandandoxygen
oxygenonto zeolite
onto zeolite LiX °C
20 ◦ C.
Reproduced
Reproduced with
withpermission
permission from
from[17].
[17].Industrial
Industrial&& engineering
engineeringchemistry
chemistryresearch,
research,2008.
2008.
Reproduced with permission from [17]. Industrial & engineering chemistry research, 2008.
Zeolite
Zeolite JLOX-101
JLOX-101 (Ø0.4–0.8
(Ø0.4–0.8 mm)
mm) from
from JIANLONG
JIANLONG CHEMICAL
CHEMICAL (Yanshi,
(Yanshi, Henan,
Henan, China)
China) was
was
Zeolite
used
usedas JLOX-101
asthe
the (Ø0.4–0.8
adsorbent
adsorbentinside mm)
insidethe
the from JIANLONG
adsorption
adsorptioncolumn CHEMICAL
columnbased
basedon
onthe (Yanshi,
thezeolite
zeolite Henan, China)
characterization
characterization[14]. was used
[14].Zeolite
Zeolite
as the adsorbent inside the adsorption column based on the zeolite characterization [14]. Zeolite
Nanomaterials 2017, 7, 195 5 of 19
Nanomaterials 2017, 7, 195 5 of 19

JLOX-101isisa aLiLi
JLOX-101 exchangedzeolite
+ +exchanged zeolite13X13Xdesigned
designedfor forindustrial
industrialoxygen
oxygenconcentration.
concentration.Figure
Figure44
exhibits the powder XRD (PXRD) patterns of zeolite JLOX-101 with two
exhibits the powder XRD (PXRD) patterns of zeolite JLOX-101 with two types of sample zeolite 13X types of sample zeolite 13X
fromUOP
from UOP(Des (DesPlaines,
Plaines,IL,IL,USA)
USA)and andBeijing-DF
Beijing-DF(Beijing,
(Beijing,China)
China)[18,19].
[18,19].The
ThePXRD
PXRDspectra
spectraofofthe
the
collected solid forms were obtained on a 2500 diffractometer (D/MAX,
collected solid forms were obtained on a 2500 diffractometer (D/MAX, Rigaku, Japan) with a Cu Kα Rigaku, Japan) with a Cu Kα
radiation source (λ = 1.5406 Å) at 100 mA and 40 kV. The samples were
radiation source (λ = 1.5406 Å) at 100 mA and 40 kV. The samples were recorded at a scanning rate recorded at a scanning rate
ofof55° perminute
◦ per minuteoveroveraa2θ 2θrange
rangeofof5–40
5–40°.The
◦ .The XRDXRDpeak peakposition
positionand andshape
shapeofofzeolite
zeoliteJLOX101-LiX
JLOX101-LiX
are essentially the same as those of zeolite 13X, which means zeolite
are essentially the same as those of zeolite 13X, which means zeolite JLOX101-LiX maintains the JLOX101-LiX maintains the
intrinsiccrystalline
intrinsic crystallinestructure
structureofofzeolite
zeolite13X13Xafter
afterthetheLiLi + ion-exchangemethod.
+ ion-exchange method.Table
Table22shows
showsthetheBET
BET
adsorption results of zeolite JLOX-101 with the sample 13X zeolite from
adsorption results of zeolite JLOX-101 with the sample 13X zeolite from Beijing-DF through ASAP 2010 Beijing-DF through ASAP
2010 (Micromeritics
(Micromeritics Instrument
Instrument Corporation,
Corporation, Norcross,Norcross,
GA, USA). GA,
BothUSA). Bothsurface
the BET the BETareasurface area and
and micropore
micropore
volume volume zeolite
of JLOX-101 of JLOX-101 zeolite
are larger than are
those larger
of 13Xthan those
zeolite of 13X
(Table 2) andzeolite (Table 2) average
the adsorption and the
adsorption average pore diameters of zeolite JLOX-101 and 13X-BJ-DF
pore diameters of zeolite JLOX-101 and 13X-BJ-DF are 21.26 Å and 22.36 Å, respectively, which are in are 21.26 Å and 22.36 Å,
respectively, which are in agreement with the microporous nitrogen adsorption
agreement with the microporous nitrogen adsorption theory of X type zeolite. Figure 5 illustrates the theory of X type
zeolite. Figure 5 illustrates
Barrett–Joiner–Halenda the Barrett–Joiner–Halenda
adsorption cumulative pore area using adsorption
nitrogen cumulative poreadsorptive.
as the analysis area using
nitrogen as the analysis adsorptive. The nitrogen occupies most of
The nitrogen occupies most of the microporous surface area (pore diameter <2 nm) and small the microporous surface area (pore
part of
the mesoporous surface area (pore diameter 2~50 nm). Figure 6 depicts the particle size distribution of6
diameter <2 nm) and small part of the mesoporous surface area (pore diameter 2~50 nm). Figure
depicts
the zeolitethe particle
powder size distribution
samples of the zeolite 2000
through MASTERSIZER powder samples
(Malvern through MASTERSIZER
Instruments Ltd, Worcestershire,2000
UK). The diameter of the vol % peak (5.75 µm) is used as the average microparticle diameter in theas
(Malvern Instruments Ltd, Worcestershire, UK). The diameter of the vol % peak (5.75 µm) is used
the average model
bidispersion microparticle diameter in
for mathematical the bidispersion model for mathematical simulation.
simulation.

Figure 4.4.XRD
Figure XRDpattern
patternofofthree
threedifferent
differenttypes
typesofofzeolites:
zeolites:UOP-13X,
UOP-13X,Beijing-DF-13X
Beijing-DF-13Xand
andJLOX101-LiX.
JLOX101-LiX.

Table2.2.BET
Table BETadsorption
adsorptionresults
resultsofofJLOX-101
JLOX-101and
and13X-BJ-DF.
13X-BJ-DF.

Zeolite JLOX-101 13X-BJ-DF

Zeolite JLOX-101 13X-BJ-DF
BET surface area (m2/g) 574.77 522.93
BET surface area (m2 /g)
Micropore area (m2/g) 574.77 515.39 522.93
477.04
Micropore
Single area (m2 /g)
Point Adsorption Total Pore Volume (cm3/g) 515.39 0.31 477.04
0.29
Single Point Adsorption Total Pore Volume (cm 3 0.31 0.24 0.29
Micropore Volume (cm 3/g)/g) 0.22
Micropore Volume (cm /g)3 0.24 0.22
Adsorption Average Pore Diameter (Å) 21.26 22.36
Adsorption Average Pore Diameter (Å) 21.26 22.36
Nanomaterials 2017,
Nanomaterials
Nanomaterials 7,7,195
2017,
2017, 7,195
195 66ofof619
of 19
19

Figure 5.5.BJH
Figure BJHadsorption cumulative
adsorptioncumulative pore
cumulativepore area distribution
porearea
area distributionofofzeolite
zeoliteJLOX101-LiX.
JLOX101-LiX.
Figure 5. BJH adsorption distribution of zeolite JLOX101-LiX.

12
12

10
10

88
%
Volume %
Volume

66

44

22

00
0.01
0.01 0.1
0.1 11 10
10 100
100 1000
1000
ParticleSize/µm
Particle Size/µm

Figure6.6.6.Particle
Figure
Figure Particlesize
Particle sizedistribution
size distributionof
distribution ofzeolite
of zeoliteJLOX-101-LiX
zeolite JLOX-101-LiX powder
JLOX-101-LiXpowder sample.
sample.
powder sample.

Dueto
Due tothe
thenegatively
negativelycharged
chargedzeolite
zeolitesurface,
surface,thethepolar
polarcompounds
compoundsH H2O
2Oand
andCO CO22in inthe
theairairare
are
Due to the negatively
preferentially adsorbed
preferentially adsorbed and charged
and occupy zeolite
occupy the surface,
the effective the
effective zeolitepolar compounds
zeolite micropores
micropores during H 2 O
during the and
the PSA CO 2 process.air
in
PSA process. the
are preferentially
Conventionalvacuum
Conventional adsorbed
vacuumdesorption and occupy
desorptioncannot the
cannotfully effective
fullydesorb
desorbthe zeolite
theH H2O micropores
2Oand
andCO during
CO22molecules,
molecules,which the PSA
whichgradually process.
gradually
Conventional
decreasethe
decrease vacuum
thezeolite desorption
zeolitecapacity
capacityunder cannot
underthe fully
thecyclic desorb
cyclicoperation
operationof the H O and
ofadsorption
adsorption
2 COand
and molecules, which
desorptionsteps.
2 desorption steps.In gradually
Inorder
order
decrease
to the zeolite
toimprove
improve the capacity
thelifespan
lifespan of under
ofthe
the the cyclic
adsorption
adsorption operation
column,
column, thinof
aathin adsorption
layer
layer and alumina
ofactivated
of activated desorption
aluminawas was steps.
added
added Inbefore
order to
before
improve
the the layer
thezeolite
zeolite lifespan
layerasasaaof the adsorption
desiccant
desiccant column,
toprevent
to prevent zeoliteadeactivation.
zeolite thin layer ofThe
deactivation. activated
The volume
volume alumina
ratioof
ratio was
ofthe added before
thedesiccant
desiccant to
to
the zeolite
wholecolumn
whole layer
columnwas as a desiccant
was0.15–0.20
0.15–0.20(Chai to prevent
(Chaietetal., zeolite
al.,2011)
2011)using deactivation.
usingambient
ambientair The
airsource volume
source(0.04
(0.04vol ratio
vol% %CO of the
CO22and desiccant
and0.4–0.8
0.4–0.8
tovol
whole
vol column[20].
%moisture)
% moisture) was 0.15–0.20 (Chai et al., 2011) using ambient air source (0.04 vol % CO2 and
[20].
0.4–0.8 vol % moisture) [20].
Nanomaterials 2017, 7, 195 7 of 19

Table 3 shows the physical properties of the zeolite, desiccant and adsorption column. The bed
voidage and particle voidage are estimated from the voidage variation study and single point BET
adsorption total pore volume, respectively [21]. The adsorption constants for zeolite column are fitted
with the experimental data based on the single particle adsorption isotherms of LiX zeolite. The PVSA
system operates between 1.79 barg for adsorption and −0.82 barg for desorption under a compressor
flow. The time for different steps of the column test is summarized to maintain a stable oxygen flow
and achieve a complete regeneration of column in each cycle. The experimental performance of the
column would be compared to the model prediction in the following paragraphs.

Table 3. Characteristics of adsorbent, desiccant and adsorption column [17,22–25].

Adsorbent Zeolite LiX Unit

Average zeolite particle size 600 µm
Average microparticle size 5.75 µm
Zeolite bulk density 790 kg/m3
Particle voidage, ε p 0.35 1
Maximum surface excess of N2 , qeN2 3.42 mol/kg
Maximum surface excess of O2 , qOe
2
6.06 mol/kg
Adsorption constant, b N2 0.09 1/bar
Adsorption constant, bO2 0.02 1/bar
Pore tortuosity, τP 3 -
BET Surface Area 574.77 m2 /g
Micropore Volume 0.240007 cm3 /g
Adsorption Average Pore Diameter 21.2609 Å
Desiccant Type Activated Alumina
Particle density 765 kg/m3
Average desiccant particle size 600 µm
Adsorption Column - -
Length 10 cm
Inside radius 1.5 cm
Column voidage, ε c 0.36 1
Material aluminum -

3. Mathematical Model

3.1. Transport and Adsorption Equations

In order to study the performance of the adsorption column, a three-dimensional model was
developed using COMSOL Multiphysics. The equations were solved using the finite element method.
Due to the cylindrical geometry of the adsorption column, a two-dimensional axi-symmetrical
approximation was used to decrease the computational cost. The following assumptions were made in
the model equations:

(a) The air source is assumed to be a binary ideal gas with 79% N2 and 21% O2 .
(b) Uniform bed voidage and particle diameter.
(c) The stabilized gas temperature at inlet and outlet is measured to be 22.5 and 23.8 ◦ C by
thermocouple, checking the temperature dependent adsorption model (Santos et al., 2008), the
|q ◦ −q ◦ |
nitrogen and oxygen capacity difference (error = i,22.5q C ◦i,23.8 C × 100%) at 1.79 barg is calculated
i,22.5 C
to be 1.47% and 1.03%, respectively, which is negligible compared to the adsorption equilibrium
equations, so the system is assumed to be isothermal [17].
(d) Gravity effects on fluid flow are negligible.

The commercial zeolite particles are produced by aggregation of nanosize particles

(zeolite powder) using a binder. The structure of the zeolite particles is shown in Figure 7. In the
Nanomaterials 2017, 7, 195 8 of 19

The commercial
Nanomaterials zeolite particles are produced by aggregation of nanosize particles (zeolite
2017, 7, 195 8 of 19
powder) using a binder. The structure of the zeolite particles is shown in Figure 7. In the pressure
swing process, the high-pressure flow will first disperse from the column void to the macropore area
pressure
inside theswing process,
nanosize theparticles
zeolite high-pressure flow will
represented first
by the disperse from
macropore the column
dispersion voidThe
equations. to the
gas
macropore area inside the nanosize zeolite particles represented by the macropore dispersion equations.
adsorption usually happens at the micropore surface of the zeolite microparticles, which is simulated
The gas adsorption
by the adsorption usually happensequations
mass transfer at the micropore
in the surface
model. of the zeolite
Based on the microparticles,
characteristicswhich
of theis
simulated by the adsorption mass transfer equations in the model. Based on the characteristics
commercial zeolites, the mass transfer of adsorption model is controlled by a set of bidispersion of
the commercial
equations [26]. zeolites, the mass transfer of adsorption model is controlled by a set of bidispersion
equations [26].

Figure 7. Schematics
Figure 7. Schematics of
of the
the zeolite
zeolite particle
particle structure.
structure.

The velocity
The velocityprofile
profile
of of
thethe
gasgas stream
stream was developed
was developed in the in
Freetheand
Free and Media
Porous PorousFlow
Media Flow
Interface
Interface
of COMSOL of COMSOL
MultiphysicsMultiphysics using the Navier–Stokes
using the Navier–Stokes Equation to Equation
describe theto describe the flow
flow in open in open
regions, and
regions, and the Brinkman equation to describe the flow
the Brinkman equation to describe the flow in porous media. in porous media.
The flow
The flow in
in zeolite
zeolite micropore
micropore area
area is
is governed
governed by
by aa set
set of
of the
the continuity
continuity equation
equation and
and the
the
momentum equation,
momentum equation, which
which together
together form
form the
theBrinkman
Brinkmanequations
equations(COMSOL,
(COMSOL,2016)
2016)[27]:
[27]:

∂∂
((εεc ρρ))++∇( ( ρ)u=) Q
∇ρu =brQbr (1)(1)
∂t∂t
      
ρ ρ ∂u  ∂+u · ∇)
uu 
−∇ ∇ · 1 1 µ(∇u + (∇Tu) T 2− 2 µ(∇ · u)I)  − −Q
1  Qbr u (2)
 ( u = p +  −1 
+
 ∂ t + ( u ⋅ ∇ )ε cε  = −∇ p + ∇ ⋅  ε εc μ ( ∇ u + ( ∇ u ) − 3 μ3(∇ ⋅ u ) I )   −  κ μ + ε 2  u ε2c
κ µ
br
ε c ε ∂t (2)
c  c   c   c 
3 ), ε is the column voidage (–), t is the time (s), Q is the
where ρ ρis the
where is thedensity
densityof the fluid
of the (kg/m
fluid (kg/m 3),c ε is the column voidage (–), is the time (s),
c κ denotes the permeability of the porous medium
br Q
br
volumetric mass source of the fluid phase (kg/m3 /s),
is the volumetric mass viscosity
source of ofthethe
fluid phase (kg/m p isκthedenotes
·s)),/s), the permeability of the porous
3
(m 2 ), µ is the dynamic fluid (kg/(m gauge pressure of the fluid flow (Pa),
medium (m 2 ), μ is the
and u is the Darcy velocity (m/s). dynamic viscosity of the fluid (kg/(m·s)), p is the gauge pressure of the fluid
flowThe and u is the
(Pa),permeability Darcy
of the velocity
packed bed(m/s).
of spherical particles can be described by the Blake–Kozeny
equationThe (Rao
permeability of the
et al., 2009) packed bed of spherical particles can be described by the Blake–Kozeny
[22]:
equation (Rao et al., 2009) [22]: d2 εc 2
κ= ( ) (3)
1502 1 − ε c
d ε 2
where d is the average diameter of zeoliteκparticles
= ((m). c ) (3)
The mass source term Qbr is defined as the 150mass ε c rate of nitrogen and oxygen inside the
1 −deposit
adsorbent to simulate the adsorption process (Rao, 2013) [28].
where d is the average diameter of zeolite particles (m).
The mass source term Qbr 2
is defined as the mass deposit rate of nitrogen and oxygen inside
Qbr = − ∑ k i Mi (ci − c pi ) (4)
the adsorbent to simulate the adsorption process
i =1 (Rao, 2013) [28].
Nanomaterials 2017, 7, 195 9 of 19

where k i is the macropore mass transfer rate of nitrogen and oxygen into zeolite particles (1/s), Mi is
the molecular weight of N2 and O2 (kg/mol), ci and c pi are the column gas concentration and particle
concentration of nitrogen and oxygen, respectively (mol/m3 ) [29].
e
60Di,j
ki = (5)
K · d2
e are the molecular diffusivity of N and O into the zeolite pore area (m2 /s), K is the
where Di,j 2 2
dimensionless Henry’s law constant for N2 and O2 (–) [30].
The molecular gas diffusivity was defined by the Chapman–Enskog Equation [31,32].
The parameters used in the Chapman–Enskog equation are shown in Table 4.
s
e 5.953 × 10−24 T3 T3
Di,j = + (6)
pσij2 Ω D Mi Mj

where σij is the average collision diameter of components i and j (m), Ω D is the collision integral (–), Mi
is the molecular mass of component i (kg/mol) and M j is the molecular mass of component j (kg/mol).

σi + σj
σij = (7)
2
where σi is the collision diameter of component i (m) and σj is the collision diameter of components
j (m).
1.060 0.193 1.036 1.765
Ω D = 0.156 + 0.476T + 1.530T + 3.894T (8)
TN e N e N e N

where TN is the standardized temperature (–).

T
TN = q q (9)
εi εi
k b,i k b,i

where ε i is the characteristic energy of component i (J), ε j is the characteristic energy of component j (J),
k b,i is the Boltzman’s constant of component i (J/K) and k b,j is the Boltzman’s constant of component
j (J/K).
The pressure drop is given by the Ergun equation [33]

∂p 150µ (1 − ε c )2 1.75ρ (1 − ε c )
= 2 u+ u|u| (10)
∂z dp ε2c dp ε3

where z represents the coordinate axis of the gas flow (m).

Input flow rate F is a function of pressure fitted with experimental data:

F ( p) = 2.478 − 0.7251p (11)

where F is the flow rate of inlet gas (L/min).

The gas distribution is solved in the Transport of Diluted Species in Porous Media Interface of
COMSOL Multiphysics with convection and dispersion equations (COMSOL, 2016) [27].

∂ci (1 − ε c ) ∂c pi ∂q
− ∇( DD · ∇ci ) + ∇(uci ) = (ε p + ρp i ) (12)
∂t εc ∂t ∂t
Nanomaterials 2017, 7, 195 10 of 19

The total gas mass deposit rate from column void to zeolite particles are calculated from the
combination of particle concentration dispersion rate and microparticle adsorption rate.

∂c pi ∂q
εp + ρ p i = k i (ci − c pi ) (13)
∂t ∂t
∂qi
= k pi (qi∗ − qi ) (14)
∂t
(bi Pωi )1/ni
qi∗ = qie (15)
2
1/n j
1 + ∑ (b j Pω j )
j =1
e
DD = 0.7Di,j + 0.5d p u (16)

where k pi is the micropore mass transfer rate of nitrogen and oxygen (1/s), ρ p is the density of
zeolite particle (kg/m3 ), qi∗ and qi are the equilibrium and actual adsorbed species to the zeolite,
respectively (mol/kg), ωi is the volume fraction of N2 and O2 (–), qie is the maximum surface excess of
gas component onto zeolite pore area (mol/kg), bi is the adsorption constant of N2 and O2 (1/bar),
DD is the axial dispersion coefficient (m2 /s).

60Die
k pi = (17)
K · d2p
 
ε p 1
Die =  1 1
 (18)
τp De + e
Di,j
k,i

where Die is the effective intra particle diffusivity of N2 and O2 into zeolite pore area (m2 /s), Dk,i
e is

the effective Knudsen diffusivity of N2 and O2 into the zeolite pore area (m2 /s) [34–36], τp is the pore
tortuosity (–) [25].
The effective Knudsen diffusivity is given by [37]:
s
e dp 8RT
Dk,i = (19)
3 πMi

where Mi is the molecular mass of nitrogen and oxygen (kg/mol), R is the ideal gas constant
(J·K−1 ·mol−1 ), T is the column temperature (K).
Initial condition of adsorption step:
At t = 0, the initial pressure (gauge) = −0.82 (barg), the inflow is set at p = 1.79 (barg), ω N2 = 0.79;
The initial adsorbed nitrogen concentration = 0.3172 (mol/kg), initial adsorbed oxygen
concentration = 0.02564 (mol/kg);
Initial condition of desorption step:
At t = 0, the initial pressure = 1.79 (barg), the inflow is set at p (gauge) = −0.82 (barg), ω N2 = 0.79;
The initial adsorbed nitrogen concentration = 1.333 (mol/kg), the initial adsorbed oxygen
concentration = 0.1329 (mol/kg).
concentration = 0.1329 (mol/kg).

Table 4. Characteristics parameters of nitrogen and oxygen [38–40].

Adsorbent Zeolite LiX Unit Reference

Nanomaterials 2017, 7, 195
Dynamic viscosity of the fluid, μ 1.84 × 10−5 Pa·s 11 of 19
Smits et al., 2006

Dimensionless Henry’s law constant of N2, KN2 1.5 × 10−2 -

Table 4. Characteristics parameters of nitrogen and oxygen -[38–40]. Sander, 2015
Dimensionless Henry’s law constant of O2, K O2 3.2 × 10−2

Adsorbent
Average collision diameter of N2, σN Zeolite LiX3.67 Unit10−10 m Reference
2
Dynamic viscosity of the fluid, µ 1.84 × 10−5 Pa·s Smits et al., 2006
Average collision
Dimensionless diameter
Henry’s law constantofofON σ
2,2 , K N2
O
1.5 × 10−2 6.06 - 10 m
−10

Sander, 2015
Dimensionless Henry’s law constant of O2 , KO2 2 3.2 × 10−2 -
Average collision
Characteristic diameter
energy of Nof ε
2, N2 , σN2
N2
k b, N2
3.67 99.8 10 − 10 m K
Average collision diameter of O2 , σ O2 6.06 10−10 m Bird et al., 2007
Characteristic energy
Characteristic energyofofNO ε
2 ,2ε, N2 /k b,N2
N2
k b, N2
99.8 113 K K
Characteristic energy of O2 , ε N2 /k b,N2 113 K Bird et al., 2007
Molecular
MolecularWeight, MN2MN2
Weight, 28.01 28.01 10−3 10
kg/mol
−3 kg/mol

Molecular
MolecularWeight, MO2MO2
Weight, 31.99 31.99 10−3 10
kg/mol
−3 kg/mol
Gas constant, R 8.31 J/(K·mol)
Gas constant, R 8.31 J/(K·mol)

3.2.
3.2. Geometry
Geometry and
and Meshing
Meshing of
of the
the Finite
Finite Element
Element Algorithm
Algorithm
The
The geometry
geometryof ofthe
thecolumn
columnused
usedfor
forthe
themodel
model simulation was
simulation a cylinder
was with
a cylinder a height
with of 20
a height ofcm
20
and a diameter of 3 cm. The gas source entered at the bottom of the column and enriched
cm and a diameter of 3 cm. The gas source entered at the bottom of the column and enriched oxygen oxygen
stream
stream exited
exited at
at the
the top
top of
of the
the column
column shown
shown in in Figure
Figure 8a.
8a.

Figure 8. (a) Schematics of adsorption column; (b) refined mesh with axis of symmetry; (c) full
Figure 8. (a) Schematics of adsorption column; (b) refined mesh with axis of symmetry; (c) full
model geometry.
model geometry.

The computational
The computational mesh
mesh was was aa swept
swept rectangular
rectangular mesh.
mesh. The
The shape
shape of
of the
the rectangular
rectangular mesh
mesh fits
fits
both the shape of the axial gas dispersion pattern and the boundary layer condition. In order to
both the shape of the axial gas dispersion pattern and the boundary layer condition. In order to findfind
an appropriate
an appropriate mesh
mesh size,
size, aa mesh
mesh refinement
refinement study
study was
wasundertaken
undertakenwith
withthe
thedifference
differencefunction:
function:
 
ωO2 (mesh x ) − ωO2 (mesh f inest )
 
%Di f f erence = × 100%,
ωO2 (mesh f inest )

where ωO2 (mesh x ) is the oxygen vol % of outflow at the mesh size x (mm) and ωO2 (mesh f inest ) is the
oxygen vol % of outflow at the finest mesh size (mm).
Figure 9 shows the results of the mesh refinement study. The relative tolerance was set to 10−6
and largest %Difference of all the time points for each mesh size was evaluated. All the mesh sizes
below 1 mm were fine enough using the criterion of less than 1% difference. The mesh size = 1 mm was
selected and used for all the simulations. The refined mesh and geometry are shown in Figure 8b,c.
the oxygen vol % of outflow at the finest mesh size (mm).
Figure 9 shows the results of the mesh refinement study. The relative tolerance was set to 10−6
and largest %Difference of all the time points for each mesh size was evaluated. All the mesh sizes
below 1 mm were fine enough using the criterion of less than 1% difference. The mesh size = 1 mm
was selected and used for all the simulations. The refined mesh and geometry are shown in Figure
Nanomaterials 2017, 7, 195 12 of 19
8b,c.

7 6.31
6

5
%Difference 4

2 1.39
0.89
1
0.18
0
0 0.5 1 1.5 2
Mesh Size/mm

Figure 9. Mesh refinement study results.

4. Results and Discussion

4.1. Concentrator Output

The photograph
photograph ofofthe theoxygen
oxygenconcentrator
concentrator is shown
is shown in Figure
in Figure 10 with
10 with the in-house
the in-house made
made
microcontroller circuit for the continuous production. The concentrator is tested with a
microcontroller circuit for the continuous production. The concentrator is tested with
vacuum/pressure
vacuum/pressure swing swingcycle
cyclebetween −0.82barg
between−0.82 bargand
and1.79
1.79barg.
barg.The
The flowrate
flowrate of of output
output stream
stream is
is 1.128
1.128 L/min
L/min and and thetheaverage
averageoxygen
oxygenconcentration
concentrationpeak
peakisis 91.64
91.64 vol
vol %. The
The output
output oxygen
concentration is almost stable at 1.79 barg, which means the desorption under a vacuum is effective
and the desorption time is long long enough
enough toto regenerate
regenerate the
the zeolite.
zeolite. The enriched continuous oxygen
stream can be conserved in a storage tank to produce a steady enriched air stream. The cyclic oxygen
production test shows that the oxygen concentration decreased gradually to 80 vol % % after
after 300
300 cycles.
cycles.
The adsorption column could be replaced every month for commercial portable oxygen concentrator
development and and thethe deactivated
deactivatedzeolite
zeoliteadsorbent
adsorbentcould
couldbe beregenerated
regeneratedininthe
theoven
ovenatat450450◦ C°Cwith
witha
Nanomaterials
continuous 2017, 7,
nitrogen195 flow toto
remove the desorbed contaminants. 13 of 19
a continuous nitrogen flow remove the desorbed contaminants.

Figure 10. Photograph

10.Photograph of oxygen
of oxygen concentrator
concentrator prototypeprototype (a) (b)
(a) front side, front
backside, (b)Oxygen
side, (c) back side,
(c) Oxygen
analyzer. analyzer.

4.2. Model Validation

Figure 11 compares the experimental flow rate with the simulation results. The experimental
and model output flowrates are 1.226 L/min and 1.24 L/min, respectively. The difference of the
experimental data and simulation results is 1.1%, which means the modified Free and Porous Media
Flow Interface is able to represent the outlet flow condition. Before 21 s, the adsorption column is in
Nanomaterials 2017, 7, 195 13 of 19
Figure 10. Photograph of oxygen concentrator prototype (a) front side, (b) back side, (c) Oxygen
analyzer.

4.2. Model
4.2.Validation
Model Validation
Figure 11 compares
Figure the experimental
11 compares the experimental flow
flowrate
rate with thesimulation
with the simulation results.
results. The experimental
The experimental
and output
and model model output flowrates
flowrates are 1.226
are 1.226 L/minand
L/min and 1.24
1.24 L/min,
L/min, respectively. The The
respectively. difference of the of the
difference
experimental data and simulation results is 1.1%, which means the modified Free and
experimental data and simulation results is 1.1%, which means the modified Free and Porous Media Porous Media
Flow Interface is able to represent the outlet flow condition. Before 21 s, the adsorption column is in
Flow Interface is able to represent the outlet flow condition. Before 21 s, the adsorption column is in
the pressurization stage without any gas flowing out through the outlet. The flow rate of the outlet is
the pressurization stage without any gas flowing out through the outlet. The flow rate of the outlet is
maintained at 1.226 L/min at the production stage, which means the same amounts of mass flow are
maintained at 1.226
entering L/minthe
and exiting atadsorption
the production
columnstage, which
to maintain themeans the pressure
adsorption same amounts of s.mass flow are
after t = 21
entering and exiting the adsorption column to maintain the adsorption pressure after t = 21 s.

Figure 11. Comparison of the experimental and simulation results of the column outlet flow rate.
Figure 11. Comparison of the experimental and simulation results of the column outlet flow rate.

Figure 12Figure
shows12 shows the outlet
the outlet pressure
pressure profile
profile at different
at different timesduring
times duringthe
the adsorption
adsorption step
stepininthethe Free
Free and Porous Media Flow Interface. The difference between the simulation
and Porous Media Flow Interface. The difference between the simulation results and the experimental results and the
experimental data is negligible, which is significant for the adsorption function in the Transport
14 of 19ofSpecies
data is Nanomaterials
negligible,2017, 7, 195 is significant for the adsorption function in the Transport of Diluted
which
Diluted Species in Porous Media Interface. According to the outlet pressure and flowrate profile in
in Porous Media Interface. According to the outlet pressure and flowrate profile in Figures 11 and 12,
Figures 11 and 12, the model is able to represent the pressurization step flow condition of a portable
the model is able
oxygen to represent
concentrator theairpressurization
with an stepofflow
compressor instead condition
a constant of asimulation.
flowrate portable oxygen concentrator
with an air compressor instead of a constant flowrate simulation.

Figure 12. Comparison of the experimental and simulation results of the column outlet pressure.
Figure 12. Comparison of the experimental and simulation results of the column outlet pressure.

FigureFigure
13 shows the the
13 shows internal gas
internal gasvelocity profileatatdifferent
velocity profile different times
times during
during the adsorption
the adsorption step. step.
The ‘arrow’
The ‘arrow’ surface
surface represents
represents the the direction
direction ofofvelocity
velocityvector.
vector. When
Whenthe thebed
bedis is
being pressurized,
being pressurized, the
the velocity decreases from the inlet to the outlet due to the closed outlet boundary condition (Figure
13a,b). The velocity distribution at the production stage (Figure 13c) is uniform, which means the
column is saturated and concentrated gas is transported from the outlet of the column at a constant
pressure.
 Figure 12. Comparison of the experimental and simulation results of the column outlet pressure.
Nanomaterials 2017, 7, 195 14 of 19
Figure 13 shows the internal gas velocity profile at different times during the adsorption step.
The ‘arrow’ surface represents the direction of velocity vector. When the bed is being pressurized,
velocity the velocity from
decreases decreases from the
the inlet inletoutlet
to the to the due
outlet
todue
thetoclosed
the closed outlet
outlet boundary
boundary condition (Figure
condition (Figure 13a,b).
13a,b). The velocity distribution at the production stage (Figure 13c) is uniform, which means the
The velocity distribution at the production stage (Figure 13c) is uniform, which means the column is
column is saturated and concentrated gas is transported from the outlet of the column at a constant
saturated and concentrated gas is transported from the outlet of the column at a constant pressure.
pressure.

Figure 13.Figure 13. Internal

Internal velocity
velocity profile
profile of adsorption
of adsorption columnwith
column with inlet
inletdistributor
distributorat time = (a)=1(a)
at time s; (b)
1 s;10(b) 10 s;
s; (c) 30 s.
(c) 30 s.
Nanomaterials 2017, 7, 195 15 of 19

Figure 14 shows the experimental and the simulation results of the outlet oxygen concentration at
Figure 14 shows the experimental and the simulation results of the outlet oxygen concentration
adsorption pressure = 1.79 barg and desorption pressure = −0.82 barg. The enriched oxygen product
at adsorption pressure = 1.79 barg and desorption pressure = −0.82 barg. The enriched oxygen product
starts tostarts
flowtooutflowofout
theof column
the column t =t =2121s sand
at at and the outletoxygen
the outlet oxygen concentration
concentration in thein the product
product flow flow
reachesreaches
the peak the at t =at35t =s.35The
peak experimental
s. The experimental data showsthe
data shows the same
same enriched
enriched tendency
tendency of oxygen
of oxygen in in
the outflow. Compared
the outflow. Comparedto the simulation
to the simulationresult,
result, a slight
slightdifference
difference of the
of the oxygen
oxygen breakthrough
breakthrough curve curve
is observed
is observed from thefromexperimental
the experimental oxygensensor
oxygen sensor data.
data.This
Thiscould
couldbe explained by theby
be explained inaccuracy of
the inaccuracy of
the multicomponent Sips adsorption equations due to the complexity of the heterogeneous zeolite
the multicomponent Sips adsorption equations due to the complexity of the heterogeneous zeolite
porous surface.
porous surface.

14. Experimental
FigureFigure outflow oxygen concentration with simulation results.
14. Experimental outflow oxygen concentration with simulation results.

Two of
Two types types of inlet
inlet flowflow distributionwere
distribution wereinvestigated
investigated for thethe
for column adsorption
column test. Figure
adsorption test. 15a
Figure 15a
illustrates the common central flow from tubing to the column inlet without inlet distributor and
illustrates the common central flow from tubing to the column inlet without inlet distributor and
Figure 15b illustrates the distributed inflow with an inlet gas distributor. Figure 16 shows the N2
Figure 15b illustrates
concentration the distributed
profiles inflow
of two different inflowwith
types an inlet gas
at different distributor.
times Figure 16 shows
during the pressurization step. the N2
The distribution of the central flow is not uniform with a radial dispersion from column axis to the
boundary while the distributed flow is closed to an ideal plug flow with negligible radial dispersion.
Figure 17 compares the outflow concentration for distributed inflow and central flow. The results are
in accordance with the experimental results in that the pressurization time of the column with central
inflow is 1.6 s longer than that of the column with the distributed flow. The inlet gas distributor is
significant for the inflow air and decreases the pressurization time for the adsorption cycle.
 Figure 14. Experimental outflow oxygen concentration with simulation results.
Nanomaterials 2017, 7, 195 15 of 19
Two types of inlet flow distribution were investigated for the column adsorption test. Figure 15a
illustrates the common central flow from tubing to the column inlet without inlet distributor and
concentration
Figure 15b profiles of two
illustrates different inflow
the distributed inflowwith
typesan at different
inlet times during
gas distributor. Figure 16 theshows
pressurization
the N2 step.
concentrationofprofiles
The distribution of two flow
the central different inflow
is not types atwith
uniform different times dispersion
a radial during the pressurization
from column step.
axis to the
The distribution
boundary of the central flow
while the distributed flow is
is not uniform
closed to anwith a radial
ideal plugdispersion
flow with from column axis
negligible to the
radial dispersion.
boundary while the distributed flow is closed to an ideal plug flow with negligible radial dispersion.
Figure 17 compares the outflow concentration for distributed inflow and central flow. The results are
Figure 17 compares the outflow concentration for distributed inflow and central flow. The results are
in accordance withwith
in accordance the the
experimental
experimentalresults inthat
results in thatthe
the pressurization
pressurization time time
of the of the column
column with central
with central
inflowinflow
is 1.6iss 1.6
longer than
s longer that
than thatofofthe
thecolumn
column withwiththe
the distributed
distributed flow.flow. Thegas
The inlet inlet gas distributor
distributor is is
significant for the
significant for inflow airair
the inflow andanddecreases thepressurization
decreases the pressurization timetime foradsorption
for the the adsorption
cycle. cycle.

15. Schematics
FigureFigure 15. Schematics of of
inflow
inflowtype atthe
type at theinlet
inlet of the
of the adsorption
adsorption column:
column: (a)flow
(a) central central flow without
without
inlet distributor;
inlet (b)
distributor; distributed
(b)
Nanomaterials 2017, 7, 195distributedflow
flowwith
with inlet
inlet distributor.
distributor. 16 of 19

Figure 16. (a) N2 concentration profile without inlet distributor at t = 1 s; (b) N2 concentration profile
Figure 16. (a) N2 concentration profile without inlet distributor at t = 1 s; (b) N2 concentration profile
without inlet distributor at t = 3 s; (c) N2 concentration profile without inlet distributor at t = 6 s; (d)
without inlet distributor at
N2 concentration t = 3with
profile s; (c)inlet
N2 distributor
concentration
at t =profile
1 s; (e) without inlet distributor
N2 concentration t = 6 s; (d) N2
atinlet
profile with
concentration profile
distributor at twith
= 3 s; inlet
(f) N2 distributor
concentration at t = 1with
profile s; (e) Ndistributor
inlet 2 concentration
at t = 6 s. profile with inlet distributor
at t = 3 s; (f) N2 concentration profile with inlet distributor at t = 6 s.
 Figure 16. (a) N2 concentration profile without inlet distributor at t = 1 s; (b) N2 concentration profile
without inlet distributor at t = 3 s; (c) N2 concentration profile without inlet distributor at t = 6 s; (d)
N2 concentration profile with inlet distributor at t = 1 s; (e) N2 concentration profile with inlet
Nanomaterials 2017,at7,t 195
distributor = 3 s; (f) N2 concentration profile with inlet distributor at t = 6 s. 16 of 19

Nanomaterials 2017, 7, 195 17 of 19

Figure 17. Simulation results of outflow oxygen concentration for different inflow type.
Figure 17. Simulation results of outflow oxygen concentration for different inflow type.
Figure 18 shows the concentration peak of the oxygen output breakthrough curve under
Figure
different 18 shows
vacuum the range.
swing concentration peak ofcapacity
The nitrogen the oxygen output
of the breakthrough
zeolite curve underwhen
decreases drastically different
the
vacuum swing range. The nitrogen capacity of the zeolite decreases drastically when the
pressure of vacuum port increases from 18 kPa (abs) to 34 kPa (abs). The deep vacuum desorption pressure of
vacuum
step port
is able to increases from
increase the 18 kPa (abs)
productivity ofto 34adsorption
the kPa (abs). The deepThe
column. vacuum desorption
simulation resultsstep
alsoisexhibit
able to
increase the productivity of the adsorption column. The simulation results also
a good fitting precision of the experimental data with the maximum difference 0.67%. exhibit a good fitting
precision of the experimental data with the maximum difference 0.67%.

Figure 18. Output oxygen concentration peak from different vacuum swing range to 1.79 bar
Figure 18. Output oxygen concentration peak from different vacuum swing range to 1.79 bar gauge
gauge pressure.
pressure.

Figure 19
Figure 19 shows
showsthetheaverage surface
average excess
surface of theofzeolite
excess surface surface
the zeolite at the adsorption and desorption
at the adsorption and
stage. Thestage.
desorption amountTheof N2 and
amount of O adsorbed
N2 and to the to
O2 adsorbed zeolite particle
the zeolite gradually
particle decreases
gradually when
decreases whenthe
adsorption
the column
adsorption is connected
column to the to
is connected suction port of port
the suction the air
ofcompressor to create vacuum
the air compressor to createdesorption
vacuum
conditions inside
desorption the column.
conditions inside Nthe
2 and O2 are released
column. N2 and from
O2 arethe released
zeolite during
from the
theblowdown and purge
zeolite during the
stages, which
blowdown andgenerates a clean
purge stages, bed for
which the next
generates adsorption
a clean bed forcycle.
the next adsorption cycle.
 Figure 19 shows the average surface excess of the zeolite surface at the adsorption and
desorption stage. The amount of N2 and O2 adsorbed to the zeolite particle gradually decreases when
the adsorption column is connected to the suction port of the air compressor to create vacuum
desorption conditions inside the column. N2 and O2 are released from the zeolite during
Nanomaterials 2017, 7, 195
the
17 of 19
blowdown and purge stages, which generates a clean bed for the next adsorption cycle.

Figure 19. Simulation

Simulation results
results of
of average
average adsorbed
adsorbed nitrogen and oxygen amount to the zeolite particles.

5. Conclusions
A portable two-column pressure vacuum swing adsorption (PVSA) oxygen concentrator was
built in our laboratory and its performance was modeled by solving the continuity equations and the
momentum equation in COMSOL Multiphysics. The model was able to simulate the microporous
adsorption of nitrogen under a dry and CO2 free compressor flow. The velocity and oxygen
concentration profiles were investigated at every time step (∆t = 0.1 s) during the four operation
stages: pressurization, production, blowdown, and desorption. With the process optimization, it was
possible to get better operational parameters and design characteristics to improve the performance
of the concentrator units and meet the terminal requirements of oxygen supply (O2 concentration
~88–92 vol %). The valves and the compressor of the prototype are controlled by a microcontroller
program for continuous oxygen production. The concentrator is a lightweight, miniature unit suitable
for outside oxygen supplement applications.

Acknowledgments: The authors would like to acknowledge the support provided by the Natural Sciences and
Engineering Research Council of Canada (NSERC) and Ontario Centres of Excellence (OCE).
Author Contributions: Mingfei Pan performed the experiment, developed the model, ran the simulations and
wrote the manuscript. Hecham Omar helped with data analysis and writing of the manuscript. The work was
supervised by Sohrab Rohani. The draft of the manuscript was reviewed, revised and submitted by Sohrab Rohani.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Crittenden, B.; Thomas, W.J. Adsorption Technology and Design; Butterworth-Heinemann: Oxford, UK, 1998.
2. Skarstrom, C.W. Method and Apparatus for Fractionating Gaseous Mixtures by Adsorption. U.S. Patent
2,944,627, 12 July 1960.
3. Daniel, D.; Montgareuil, P.G.D. Process for Separating a Binary Gaseous Mixture by Adsorption. U.S. Patent
3,155,468, 3 November 1964.
4. Drain, L.E. Permanent electric quadrupole moments of molecules and heats of adsorption. Trans. Faraday
Soc. 1953, 49, 650–654. [CrossRef]
5. Ackley, M.W.; Stewart, A.B.; Henzler, G.W.; Leavitt, F.W.; Kane, M.S. PSA Apparatus and Process Using
Adsorbent Mixtures. U.S. Patent 6,027,548, 22 February 2000.
6. Feuerstein, M.; Accardi, R.J.; Lobo, R.F. Adsorption of nitrogen and oxygen in the zeolites LiA and LiX
investigated by 6Li and 7Li MAS NMR spectroscopy. J. Phys. Chem. B 2000, 104, 10282–10287. [CrossRef]
Nanomaterials 2017, 7, 195 18 of 19

7. Feuerstein, M.; Lobo, R.F. Influence of oxygen and nitrogen on 7Li MAS NMR spectra of zeolite LiX-1.0.
Chem. Commun. 1998, 16, 1647–1648. [CrossRef]
8. Molecular Sieve JLOX-101. Luoyang Jianlong Chemical: Yanshi, Henan, China, 2016. Available online:
http://www.psamolecularsieve.com/vpsa-molecular-sieve/molecular-sieve-jlox-100.html (accessed on
22 July 2017).
9. Rao, V.R.; Kothare, M.V.; Sircar, S. Novel design and performance of a medical oxygen concentrator using a
rapid pressure swing adsorption concept. AIChE J. 2014, 60, 3330–3335. [CrossRef]
10. Santos, J.C.; Portugal, A.F.; Magalhaes, F.D.; Mendes, A. Optimization of medical PSA units for oxygen
production. Ind. Eng. Chem. Res. 2006, 45, 1085–1096. [CrossRef]
11. Thakur, R.S.; Kaistha, N.; Rao, D.P. Novel single-bed and twin-bed pressure swing adsorption systems.
Chem. Eng. Proc. Proc. Intensif. 2015, 95, 165–174. [CrossRef]
12. TTC-IIS Series Miniature Diaphragm Pumps; Precision Fluidics Division Parker Hannifin Corporation: Hollis,
NH, USA, 2013.
13. X-Valve; Precision Fluidics Division, Parker Hannifin Corporation: Hollis, NH, USA, 2015.
14. Lauer, J. Sensors for Gas Phase Oxygen Analysis; Teledyne Analytical Instruments: Industry, CA, USA, 2003.
15. MPXV7002 Series Integrated Silicon Pressure Sensor; Freescale Semiconductor, Inc.: Austin, Texas, 2009.
16. Model GB300 Percent Portable Oxygen Analyzer; Teledyne Analytical Instruments, A Teledyne Technologies
Company: Thousand Oaks, CA, USA, 2011.
17. Santos, J.C.; Magalhaes, F.D.; Mendes, A. Contamination of zeolites used in oxygen production by PSA:
Effects of water and carbon dioxide. Ind. Eng. Chem. Res. 2008, 47, 6197–6203. [CrossRef]
18. MOLSIV 13X APG; UOP Adsorbents Honeywell UOP: Des Plaines, IL, USA, 2010.
19. Zeolite 13X; Beijing Dongfang Guansen Technology Co., Ltd.: Beijing, China, 2014.
20. Chai, S.W.; Mayuresh, V.K.; Sircar, S. Rapid pressure swing adsorption for reduction of bed size factor of a
medical oxygen concentrator. Ind. Eng. Chem. Res. 2011, 50, 8703–8710. [CrossRef]
21. De Klerk, A. Voidage variation in packed beds at small column to particle diameter ratio. AIChE J. 2003, 49,
2022–2029. [CrossRef]
22. Rao, V.; Rama, S.F.; Krantz, W.B. Design of a two-step pulsed pressure-swing adsorption-based oxygen
concentrator. AIChE J. 2010, 56, 354–370. [CrossRef]
23. Santos, J.C.; Cruz, P.; Regala, T.; Magalhães, F.D.; Mendes, A. High-purity oxygen production by pressure
swing adsorption. Ind. Eng. Chem. Res. 2007, 46, 591–599. [CrossRef]
24. Activated Alumina; Micromeritics Instrument Corporation: Norcross, GA 30093, USA, 2011.
25. Farooq, S.; Ruthven, D.M.; Boniface, H.A. Numerical simulation of a pressure swing adsorption oxygen unit.
Chem. Eng. Sci. 1989, 44, 2809–2816. [CrossRef]
26. Azevedo, D.C.S.; Alírio, E.R. Bilinear driving force approximation in the modeling of a simulated moving
bed using bidisperse adsorbents. Ind. Eng. Chem. Res. 1999, 38, 3519–3529. [CrossRef]
27. COMSOL. COMSOL Documentation, 2016. Available online: www.comsol.com (accessed on 22 July 2017).
28. Rao, V.; Rama, M.; Kothare, V.; Shivaji, S. Numerical simulation of rapid pressurization and depressurization
of a zeolite column using nitrogen. Adsorption 2014, 20, 53–60. [CrossRef]
29. Hejazi, S.A.H.; Rajendran, A.; Sawada, J.A.; Steven, M. Kuznicki Dynamic Column Breakthrough and Process
Studies of High-Purity Oxygen Production Using Silver-Exchanged Titanosilicates. Ind. Eng. Chem. Res.
2016, 55, 5993–6005. [CrossRef]
30. Rege, S.U.; Ralph, T.Y. Limits for air separation by adsorption with LiX zeolite. Ind. Eng. Chem. Res. 1997, 36,
5358–5365. [CrossRef]
31. Chapman, S.; Cowling, G.C.; Cercignani, C. The Mathematical Theory of Non-Uniform Gases: An Account of the
Kinetic Theory of Viscosity, Thermal Conduction and Diffusion in Gases; Cambridge University Press: Cambridge,
UK, 1970.
32. Omar, H.M.; Sohrab, R. Removal of CO2 from landfill gas with landfill leachate using absorption process.
Int. J. Greenh. Gas Control 2017, 58, 159–168. [CrossRef]
33. Akgiray, Ö.; Saatçı, A.M. A New Look at Filter Backwash Hydraulics. Water Sci. Technol. Water Supply 2001,
1, 65–72.
34. Satterfield, C.N. Mass Transfer in Heterogeneous Catalysis; The MIT Press: Cambridge, MA, USA, 1970.
35. Kumar, R.; Sircar, S. Skin resistance for adsorbate mass transfer into extruded adsorbent pellets.
Chem. Eng. Sci. 1986, 41, 2215–2223. [CrossRef]
Nanomaterials 2017, 7, 195 19 of 19

36. Reid, R.C.; Prausnitz, J.M.; Sherwood, T.K. The Properties of Gases and Liquids, 3rd ed.; McGraw-Hill: New York,
NY, USA, 1977; p. 548.
37. Cunningham, R.E.; Williams, R.J.J. Diffusion in Gases and Porous Media; Plenum Press: Berlin, Germany, 1980.
38. Smits, A.J.; Dussauge, J.P. Turbulent Shear Layers in Supersonic Flow, 2nd ed.; Springer Science & Business
Media: Berlin, Germany, 2006.
39. Sander, R. Compilation of Henry’s law constants (version 4.0) for water as solvent. Atmos. Chem. Phys. 2015,
15, 4399–4981. [CrossRef]
40. Bird, R.B.; Stewart, W.E.; Lightfoot, E.N. Lightfoot. In Transport Phenomena; John Wiley & Sons: Hoboken, NJ,
USA, 2007.

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).