5358
Ind. Eng. Chem. Res. 1997, 36, 5358-5365
SEPARATIONS Limits for Air Separation by Adsorption with LiX Zeolite Salil U. Rege and Ralph T. Yang* Department of Chemical Engineering, University of Michigan, Ann Arbor, Michigan 48109
In the present work the performance of two zeolite sorbents, namely LiX (Al/Si ) 1 with 100% Li+) and NaX (13X), is compared for air separation by pressure swing adsorption, with particular attention focused on the limits in the adsorption/desorption pressure ratio. The product (O2) recovery was optimized for the two sorbents at different pressure ratios (2-10), keeping the product purity and product throughput nearly constant. It was found that in the case of the LiX zeolite it is possible to operate at a pressure ratio as low as 2 with only a slight decrease in product recovery, compared to the limit of 4 for NaX. The temperature excursion in the LiX bed was 4 times that in the NaX bed. Substitution of a portion of the LiX sorbent by inert, high-heat-capacity particles (e.g., iron) in the bed decreased the temperature excursion significantly. It was shown that an optimal amount of 5-10% (v/v) of inert substitution (at the same total bed volume) could increase the O2 product recovery by 2%. Introduction Nitrogen and oxygen are, respectively, the second and third largest man-made commodity chemicals today. Since the 1920s, N2 and O2 have been produced by cryogenic distillation of air. Since the 1980s, adsorption (i.e., pressure swing adsorption or PSA) has been adopted rapidly and increasingly for air separation. Approximately 20% of air separation is presently accomplished by PSA. This is the combined result of zeolite sorbent development and adsorption process cycle development. A brief history of the sorbent and process developments is given elsewhere (Yang, 1987; Ruthven et al., 1994). Zeolites have the unique ability to adsorb N2 more strongly than O2 (i.e., by a factor of 2 or more in terms of pure component adsorption amounts). The main reason for this is the interaction between the quadrupole moment of N2 and the cation that is attached to the zeolite framework. It has been long known that Li+ is among the cations that provide the strongest interactions with N2 (McKee, 1964), due to its high polarizing power (i.e., charge/ionic radius). A major advance came only recently when it was found that (1) a threshold of approximately 70% cations must be Li+ for the strong N2 interactions, beyond which the adsorption amount steeply increases, and (2) N2 adsorption is significantly increased when the Si/Al ratio in type X zeolite approaches unity (Chao, 1989; Chao et al., 1992; Coe et al., 1992). A variety of PSA cycles are known (Yang, 1987). The dominant factor that determines the energy requirement (hence the cost of N2 and O2) of a PSA cycle is the pressure ratio of the high adsorption pressure to the low desorption pressure (Leavitt, 1991). A pressure ratio of 4 or higher has been used in industry, which has appeared to be a barrier for PSA air separation. It was first suggested by Leavitt (1991) that this barrier could possibly be lowered to 2, when LiX zeolite was used. * Author to whom correspondence should be addressed. Tel.: (313) 936-0771. Fax: (313) 743-0459. E-mail: yang@ umich.edu. S0888-5885(97)00521-6 CCC: $14.00
However, no specifics were given by Leavitt. Meanwhile, in the vast literature on PSA simulation, little or no attention has been paid to the pressure ratio, with the exception of work by Knaebel and Hill (1985) and Kayser and Knaebel (1986; 1989). In their work analytical solutions were derived with simplifying assumptions, including isothermality and the PSA cycle involving essentially two steps only (i.e., Skarstrom cycle) with complete bed utilization. Although these assumptions deviated substantially from industrial air separation, their solutions provided a qualitative insight into the process dynamics. At a fixed feed throughput and O2 product purity using 5A and 13X zeolites, their solutions correctly predicted a steep decline in product recovery as the pressure ratio was decreased to near 4. In this work, we examine the limits for air separation using the LiX zeolite (with 100% Li exchange). Particular attention was paid to the limits in the pressure ratio on the X zeolite with Si/Al ) 1. The PSA cycle used in this work was the well-developed five-step cycle that is used in industry (Yang, 1987). The cycle conditions were optimized by numerical simulation. The product throughput, expressed by the bed size factor, was within the range for industrial practice. The results of LiX were compared with that of NaX (with Si/Al being approximately 1.15), which is the sorbent being used in industry. Moreover, a means to counter the large heat effects on LiX (due to the high heat of adsorption of N2 on LiX) is suggested. Description of the PSA Cycle A five-step PSA cycle was used in this study. The steps involved in each cycle are as follows: (1) Pressurization with the feed gas (air); (2) high-pressure adsorption, i.e., feed step; (3) cocurrent depressurization; (4) countercurrent blowdown; (5) countercurrent lowpressure purge with the product (oxygen). All the above steps were of equal duration (30 s). Thus the time required for the completion of each PSA cycle was 2.5 min. The model assumed only two adsorbable components, namely O2 and N2. The lessstrongly adsorbed species like Ar, etc., were clubbed © 1997 American Chemical Society
Ind. Eng. Chem. Res., Vol. 36, No. 12, 1997 5359
purge to feed ratio (P/F) ) amount of O2 used in step 5 (3) amount of O2 fed in steps 1 and 2 Another factor called the “bed size factor” (BSF), which gives an indication of the sorbent productivity, is used in this work:
BSF ) weight of adsorbent used in the bed (kg) (4) product throughput (kg of O2/h in product) Mathematical Model Figure 1. N2 isotherms for LiX (Si/Al ) 1, 100% Li) and NaX (13X, Si/Al ∼ 1.15) zeolites at 298 K.
The model used assumes the flow of a gaseous mixture of two components in a fixed bed packed with spherical adsorbent particles of identical size and shape. The bed is considered to be adiabatic and diffusional resistance is assumed to be negligible since the diffusion of O2 and N2 in the adsorbents considered is known to be fast. Thus, there is local equilibrium between the gas and the solid phase of each gas component. Axial dispersion for mass and heat transfer is assumed, but dispersion in the radial direction is taken to be negligible. Axial pressure drop is neglected and the ideal gas law is assumed to hold since pressures involved are near atmospheric. Also the gas is assumed to have constant viscosity and heat capacity. The mass balance equation for component k in the bed is given by the axially dispersed plug flow equation (Sun et al., 1996):
Figure 2. O2 isotherms for LiX (Si/Al )1, 100% Li) and NaX (13X, Si/Al ∼ 1.15) zeolites at 298 K.
t along with O2, and it was assumed that all contaminants in air like CO2 and water vapor were removed completely prior to feeding by pretreatment beds. The cocurrent depressurization step has been shown to improve the recovery of the strongly adsorbed component by increasing the concentration of the strong adsorptive in both gas and adsorbed phases due to lowering of the pressure in the voids (Yang, 1987). In order to approach the cyclic steady state faster, the bed was pressurized initially with 90 mol % O2. In the subsequent cycles pressurization was carried out with air consisting of 22% O2 (mixture of O2 and Ar) and 78% N2. The product of each cycle comprised of a volumetric mixture of the output stream of the feed step and the cocurrent depressurization step. This product stream was partly used to purge the bed countercurrently in step (5). In order to study the performance of the two adsorbents under study, the product purity, product recovery, and product throughput were studied at various pressure ratios. The pressure ratio is defined as
pressure ratio ) high pressure during adsorption PH (1) low pressure during purge PL In this paper, the product recovery and purge to feed ratio (P/F) are defined as follows:
product recovery ) (O2 from steps 2 and 3) - (O2 used in step 5) (O2 fed in step 1 and step 2)
(2)
∂yk ∂2yk ∂(uyk) FbRT ∂qk tyk dP - Dax 2 + + + )0 ∂t ∂z P ∂t P dt ∂z (5) The overall material balance obtained is
FbRT
∂u )-
P
∂z
2
∑ k)1
∂qk
t dP
(6)
∂t
P dt
For an adiabatic bed with no heat transfer with the surroundings, the overall heat balance may be written as 2
[FgCpg + Fb(Cps +
∂T
∂T
∑ qkCpg)] ∂t + FgCpgu ∂z k)1
λL
∂2T ∂z
2
2
) Fb
∑ |∆Hj| k)1
∂qk ∂t
+
dP dt
(7)
Assuming local equilibrium, we have
∂qk ∂qk* ) ∂t ∂t
(8)
where qk* is the equilibrium amount adsorbed at the surface of the pellet. Initial and Boundary Conditions The boundary conditions employed correspond to the Dankwert’s boundary conditions for the closed-closed vessel case with no dispersion to immediate left of the z ) 0 and to the immediate right of z ) L. Here the notation z ) 0 and z ) L is used with reference to the entrance and exit points of the bed for the high-pressure
5360 Ind. Eng. Chem. Res., Vol. 36, No. 12, 1997 Table 1. Values of the Parameters in the Temperature Dependent Langmuir Isotherm of N2 and O2 for LiX and NaX (13X) Adsorbents sorbent
sorbate
NaX NaX LiX LiX
O2 N2 O2 N2
k1 (mmol/g/atm)
k2 (K)
k3 (atm-1)
k4 (K)
-∆H (kcal/mol)
Cpg (cal/mol/K)
8.21 × 10-4 6.11 × 10-4 1.11 × 10-4 1.25 × 10-3
1592.53 2168.55 1593.0 2168.6
3.51 × 10-3 1.05 × 10-3 1.03 × 10-4 2.07 × 10-4
1544.43 2012.92 2061.9 2455.5
3.16 4.31 3.16 5.60
8.27 6.50 8.27 6.50
feed step. For each of the steps having pressure dynamics, a linear pressure variation is assumed. The boundary conditions for the bed mass and temperature variables are summarized below, wherein the subscript “k” corresponds to the component index:
(Step 4) Countercurrent Depressurization Step: at z ) L, u ) 0 P ) P(t) ) PCD + (PL - PCD)(t/τcn)
| |
∂yk ∂z
z)0
∂yk ∂z
z)L
(Step 1) Pressurization Step: at z ) 0, yk ) yf,i
)
)
| |
∂T ∂z
z)0
∂T ∂z
z)L
)0 )0
(12)
at z ) L, u ) 0
(Step 5) Countercurrent Low-Pressure Purge Step:
P ) P(t) ) PL + (PH - PL)(t/τP)
at z ) L, yk ) yp,k
Dax
|
∂yk ∂z
z)0
at z ) L, u ) uL
) uH(yk|z)0 - yH,k)
P ) PL
|
∂T -λL ) FgCpguk(T|z)0 - TH) ∂z z)0
|
∂yk ∂z
) z)0
|
∂T ∂z
-Dax z)0
)0
|
∂yk ∂z
(9) -λL
z)L
at z ) 0, yk ) yf,i, u ) uf
|
∂yk ∂z
z)0
qk* )
)
z)L
|
∂T ∂z
z)L
)0
|
∂T ∂z
KkPk 2
z)L
)0
k ) 1,2
(13)
(14)
BjPj ∑ j)1
The temperature dependence of the Langmuir parameters are assumed to be as follows (Baksh et al., 1992):
|
|
)
z)L
1+
) uH(yk|z)0 - yH,k)
∂T -λL ) FgCpguH(T|z)0 - TH) ∂z z)0 ∂yk ∂z
|
∂yk ∂z
P ) PH
(uL < 0)
∂T ) FgCpguL(T|z)L - TL) ∂z z)L
(Step 2) High-Pressure Feed Step:
Dax
|
) uL(yL,k - yk|z)L)
K ) k1 exp(k2/T) (10)
Isotherm Data The equilibrium amount adsorbed on respective adsorbent was calculated using the extended mixed Langmuir isotherm:
(Step 3) Cocurrent Depressurization Step:
and
B ) k3 exp(k4/T)
(15)
The adsorbents under study were LiX and NaX (13X). It should be noted that although Baksh et al. (1992) also measured the adsorption isotherms of N2 and O2 on a commercial LiX zeolite (Si/Al ∼ 1.15), the LiX sample used in the present case has a different Si/Al ratio (of 1). Hence the isotherms obtained by Baksh et al. (1992) differ from those given in this paper. Isotherm data for NaX and LiX used in the present simulations are shown in Figures 1 and 2 and the isotherm parameters are listed in Table 1. These isotherms were measured in our laboratory using a Micromeritics ASAP 2010 apparatus.
at z ) 0, u ) 0 Numerical Method Used
P ) P(t) ) PH + (PCD - PH)(t/τco)
| |
∂yk ∂z
z)0
∂yk ∂z
z)L
)
)
| |
∂T ∂z
z)0
∂T ∂z
z)L
)0 )0
(11)
The terms in the above partial differential equations, involving spatial derivatives, were approximated by their finite difference equivalents using the CrankNicolson scheme. In each of the simulations, 100 spatial grid points were used for discretization with a relative convergence accuracy of 1 × 10- 3. The time derivative with respect to pressure was known. The time deriva-
Ind. Eng. Chem. Res., Vol. 36, No. 12, 1997 5361 Table 2. Adsorption Bed Characteristics and Operating Conditions Used in the PSA Simulations bed length diameter of adsorber bed bed external porosity bed density heat capacity of gases heat capacity of bed wall temperature feed gas composition adsorption pressure (PH) range of desorption pressure (PL) range of cocurrent depressurization pressure (PCD) range of feed interstitial velocities Pem Peh average BSF value (kg of adsorbent/kg of O2 product/h)
2.5 m 1.0 m 0.40 720 kg/m3 6.87 cal/mol/K 0.28 cal/g/K 298 K (ambient) 78% N2, 22% O2 1.0 bar 0.1-0.5 bar (depending on pressure ratio) 0.6- 0.8 bar 12-80 cm/s 150 2 37.5
tive of the average adsorbed concentration could be written as a function of time derivatives of gas phase concentrations, temperature, and pressure using the isotherm equation. Thus the partial differential equations were converted into a set of ordinary differential equations (ODEs) having first-order time derivatives of temperature and mole fractions of the gaseous components as the differential vector on the l.h.s. and a sparsebanded matrix of finite difference concentration and temperature terms on the r.h.s. Since the ODE matrix was stiff in nature, it was solved using an ODE solver, namely the LSODE (Livermore solver for ordinary differential equations) subroutine for the solution of first-order initial value problems of stiff/nonstiff systems (Hindmarsh, 1980) which is based on the Gear’s method. The PSA simulation code was written in FORTRAN and executed on a SUN-SPARC workstation. In most of the cases the cyclic steady state was reached within 30 cycles. The CPU time was found to depend upon the initial conditions and the operating parameters. The CPU time taken for computing one PSA cycle was between 18 and 25 s. Results and Discussion Comparison of LiX and NaX (13X) Zeolites as Sorbents. The PSA simulations reveal some interesting characteristics of the two zeolite sorbents under consideration. One objective of the present work was to study the oxygen product recovery at various pressure ratios. A similar study has been done analytically in the past (Kayser and Knaebel, 1989) using the binary nonlinear isotherm theory for the case of a four-step isothermal PSA cycle for air separation using zeolite 13X as the adsorbent at 0 °C. The PSA cycle specifications used in the simulations are summarized in Table 2. The dispersion coefficients and effective thermal conductivity used in the simulations were predicted using standard correlations reported in literature (Yang, 1987). The Peclet numbers mentioned in Table 2 are defined as follows:
Pem ) uL/Dax Pet ) uL/R
where R ) λLFg-1Cp-1
In the present study, by proper manipulation of feed throughput and the purge to feed ratio (P/F), the product purity and product throughput were kept constant approximately at 95.2% and 0.027 kg of O2 product/h/ kg of adsorbent, respectively, so as to provide a fair comparison of the two sorbents. In other words, the bed
Figure 3. O2 product recovery (%) at different pressure ratios (PH/PL) for LiX (Al/Si )1, 100% Li) and NaX (13X) sorbents. PH ) 1.0 atm, average bed size factor (BSF) ) 37.5 kg of adsorbent/kg of O2 product/h, O2 product purity ) 95.2%. Refer to Table 3 for operating conditions.
size factor (BSF) for these sorbents were kept constant. It should be noted that the purge to feed ratios are probably not the optimal values for each of the simulation results in themselves. However, by placing a constraint that the BSF for each of the results be the same optimization with respect to the purge to feed ratios was done. Thus the product recoveries were optimized at fixed product purity and fixed BSF. As can be seen in Figure 3, the product recovery for the NaX sorbent remained fairly constant between a pressure ratio of 5 and 10. Below a pressure ratio of 4, the product recovery started to drop drastically. In fact below a pressure ratio of 4, it was almost impossible to maintain a product purity above 95% keeping the product throughput at the constant value mentioned before. An extrapolation (see dotted line) was therefore done at pressure ratios of 2 and 3 for the NaX sorbent with data at slightly higher BSF than the other data points. The process conditions for these simulations are summarized in Table 3. In contrast, the LiX sorbent was seen to maintain its product recovery at a pressure ratio of 3. Even at a pressure ratio of 2, the product recovery is seen to have dropped to just above 50%, which is quite acceptable. At present, industrial processes employ a pressure ratio between 4 and 5. Lowering the pressure ratio implies an increase in the desorption pressure, thus providing large savings in capital and operating costs related to vacuum equipment (Leavitt, 1989). Hence there is an incentive to reduce the pressure ratio in PSA beds using LiX sorbents, provided the product throughput is acceptable. For the operating conditions in this study, the LiX sorbent was found to converge to a value of 65% at highpressure ratios while the NaX (13X) sorbent converged at about 52%. The higher recovery in the case of LiX is obvious. Effect of Pressure Ratio on Temperature Profiles. The temperature profiles in the PSA bed following the high-pressure adsorption step were studied at various pressure ratios. It is seen from Figures 4 and 5 that the peak temperatures for the LiX and NaX sorbents are 40 and 29 °C, respectively. The difference in the temperature peaks of LiX and NaX for lower pressure ratios is considerably lower than that at higher pressure ratios which is because of the higher heat of adsorption of N2 on LiX than on NaX, as can be seen from Table 1. Figures 6 and 7 show the effect of the pressure ratio on the temperature during the low-pressure purge step.
5362 Ind. Eng. Chem. Res., Vol. 36, No. 12, 1997 Table 3. PSA Simulation Operating Conditions Used in Figures 3-5, 10, and 11 pressure ratio PH/PL
PCD (atm)
feed velocity UH (m/s)
1.2 1.5 2 3 4 5 6 7 8 10
0.95 0.75 0.8 0.7 0.7 0.7 0.65 0.65 0.65 0.65
0.30 0.70 0.70 0.45 0.37 0.35 0.30 0.28 0.28 0.28
2 3 4 5 6 7 8 10
0.85 0.80 0.70 0.70 0.70 0.70 0.67 0.67
0.30 0.30 0.28 0.23 0.19 0.17 0.12 0.11
purge velocity UL (m/s)
purge to feed ratio P/F
O2 product purity (%)
LiX (Al/Si ) 1, 100% Li+) Adsorbent (PH ) 1.0 atm) 0.10 0.85 90.2 0.25 0.70 95.6 0.60 0.44 95.1 0.13 0.23 96.1 0.11 0.13 95.3 0.09 0.09 95.1 0.08 0.06 95.4 0.06 0.04 94.5 0.06 0.03 95.9 0.04 0.02 95.3 NaX (13X) Adsorbent (PH ) 1.0 atm) 0.30 0.80 96.0 0.33 0.49 93.8 0.42 0.46 95.2 0.40 0.35 95.6 0.37 0.27 95.4 0.35 0.21 95.2 0.34 0.18 95.3 0.34 0.14 95.3
Figure 4. Cyclic steady-state temperature profiles at the end of the high-pressure feed step for LiX (Al/Si ) 1, 100% Li) at BSF ) 37.5 kg of adsorbent/kg of O2 product/h. Inset figures indicate the corresponding pressure ratio (PH/PL).
Figure 5. Cyclic steady-state temperature profiles at the end of the high-pressure feed step for NaX (13X) at BSF ) 37.5 kg of adsorbent/kg of O2 product/h. Inset figures indicate the corresponding pressure ratio (PH/PL).
For both the sorbents, there was a fall in the temperature with an increase in pressure ratio. The effect, however, was more pronounced in the case of LiX, again because of its high heat of adsorption. Effect of P/F Ratio on the Product Purity and Recovery. A certain amount of the product obtained in each cycle was purged countercurrently at the lower desorption pressure through the bed to clean the bed of residual N2 impurity. The purge to feed ratio (P/F), defined earlier in eq 3, is an important parameter in deciding the adsorber performance. The effect of P/F
O2 product recovery (%)
BSF (kg of adsorbent/ kg of O2 produced/h
12.4 29.0 53.5 61.9 63.3 64.1 65.3 65.3 64.5 64.7
393 76.8 37.4 37.7 38.1 37.2 37.7 37.9 37.6 36.7
20.0 42.3 53.3 53.4 53.2 52.8 52.8 53.1
138.9 49.6 38.2 37.4 37.3 37.1 37.6 36.5
Figure 6. Cyclic steady-state temperature profiles at the end of the low-pressure purge step for LiX (Si/Al ) 1, 100% Li+) at BSF ) 37.5 kg of adsorbent/kg of O2 product/h. Inset figures indicate the corresponding pressure ratio (PH/PL).
Figure 7. Cyclic steady-state temperature profiles at the end of the low-pressure purge step for NaX (13X) at BSF ) 37.5 kg of adsorbent/kg of O2 product/h. Inset figures indicate the corresponding pressure ratio (PH/PL).
on product purity and recovery has been studied extensively (Yang, 1987; Baksh et al., 1992). In the case of the LiX zeolite, the P/F was varied from 0.19 to 0.35 at a pressure ratio of 3, keeping the feed throughput constant at 0.753 kg of O2/h/kg of adsorbent. As can be seen from Figure 8, the O2 product purity increased with an increase in the P/F ratio up to a certain value, beyond which there was no significant gain in purity. The purging of the bed with the product allowed for a sharper O2 wave front during the feed step by cleaning
Ind. Eng. Chem. Res., Vol. 36, No. 12, 1997 5363
Figure 8. O2 product purity (%) and recovery (%) versus purge to feed ratio (P/F) for LiX (Si/Al )1, 100% Li) zeolite at pressure ratio ) 3; PH ) 1.0 atm, feed throughput fixed at 0.753 kg of O2/ h/kg of adsorbent. (Note: “O2” also includes inerts like Ar.)
Figure 10. Concentration profile in NaX (13X) bed at the end of each step at pressure ratio ) 10. Refer to Table 3 for operating conditions. Inset numbers indicate step number: (1) pressurization with the feed gas (air); (2) high-pressure feed step; (3) cocurrent depressurization; (4) countercurrent blowdown; (5) countercurrent purge.
Figure 9. O2 product recovery (%) versus percentage purity for LiX (Si/Al ) 1, 100% Li) zeolite at pressure ratio ) 3; PH ) 1.0 atm, feed throughput fixed at 0.753 kg of O2/h/kg of adsorbent. (Note: “O2” also includes inerts like Ar.)
Figure 11. Concentration profile in LiX (Si/Al)1, 100% Li+) bed at the end of each step at pressure ratio ) 10. Refer to Table 3 for operating conditions. Inset figures indicate step number: (1) pressurization with the feed gas (air); (2) high-pressure feed step; (3) cocurrent depressurization; (4) countercurrent blowdown; (5) countercurrent purge.
the bed of the impurity, thus increasing the product purity. However, it is evident that, by doing so, there is a loss of O2 in the product stream, and hence there was a decline in the product recovery with an increase in the P/F ratio. Figure 9 displays the behavior of the product recovery as the product purity is increased, keeping the feed throughput constant. It should be recalled that the “oxygen” component actually was composed of a mixture of O2 and Ar. Although the purity of oxygen depicted in Figures 8 and 9 exceeded 98%, it also included about 4% Ar. Hence the actual purity of O2 that would be obtained is about 95-96%. Adsorption Bed Profiles. The cyclic steady-state N2 concentration profiles at the end of each step of the PSA cycle, for NaX and LiX adsorbents at a pressure ratio of 10, are shown in Figures 10 and 11, respectively. In the case of NaX, it is seen that the bed utilization was more since the N2 wave front extended further down the bed than it did in the case of the LiX sorbent. This result was expected since the NaX sorbent had a smaller capacity for N2 than the LiX sorbent. The temperature profiles (not shown) in the case of NaX were more flat whereas those of LiX showed some sharp peaks. The temperature variation in the bed from desorptive to adsorptive conditions was much smaller in the case of NaX (from 22 to 28 °C) than that in the case of LiX (from 15 to 40 °C). This is understandably
due to the higher heat of adsorption of the adsorbate on the LiX sorbent. A comparison of the bed concentration profiles at times half way through a step (not shown) and those at the end of the step showed a slight advancement for the feed and cocurrent depressurization steps. There was a drastic change in the profile involving the countercurrent depressurization step which was natural, considering the large pressure variation occurring during this step. Also a comparison of the concentration profiles for the two adsorbents revealed that the wave fronts were much sharper in the case of the LiX sorbent than those for the NaX sorbent. This followed from the fact that the N2 isotherm on LiX was more favorable than that on NaX, as can be seen from Figures 1 and 2. Countering the Heat Effects. Due to the combined effect of the high heat of adsorption of N2, a greater amount of adsorption and a shorter bed utilization in the case of the LiX zeolite, there was a substantial differential in the bed temperature during the adsorption and desorption steps. This can be seen from Figures 4 and 5, wherein the temperature rise in the LiX bed was about 17 °C compared to 4 °C in the NaX bed. The increase in temperature during adsorption and the consequent decrease during desorption is detrimental to the separation. Yang and Cen (1986) had suggested two methods for countering the adverse heat effects in the adsorption process, namely providing heat
5364 Ind. Eng. Chem. Res., Vol. 36, No. 12, 1997 Table 4. PSA Bed Characteristics and Process Performance in the Case of Different vol % of Iron Particle Addition (PH ) 1.0 atm, PL ) 0.1 atm, PCD ) 0.60-0.65 atm) vol % iron particles
effective specific heat Cps (cal/g/K)
adsorbent bed density Fb (kg/m3)
feed velocity UH (m/s)
purge velocity UL (m/s)
O2 product purity (%)
O2 product recovery (%)
BSF (kg of adsorbent/ kg of O2 produced/h)
0 5 10 20 30
0.28 0.38 0.40 0.56 0.75
720 684 648 576 504
0.28 0.23 0.22 0.19 0.14
0.04 0.05 0.05 0.04 0.04
95.3 95.1 95.3 95.2 95.2
64.7 66.3 65.5 62.6 62.5
36.7 36.2 36.3 36.6 36.3
exchange between adsorbers and the addition of high heat capacity additives. The former method is not very attractive since it adds to the complexity of the multibed adsorptive process. The second proposal suggests the addition of inert additives like iron particles (of nearly the same size as that of the zeolite particles) which will serve to store heat during adsorption and release it during desorption. Thus the problem of temperature excursions may be overcome. A similar approach was employed in the present case of air separation using the LiX zeolite for a pressure ratio of 10. Simulations were carried out using 0, 5, 10, 20, and 30% of inert iron particles by volume. The internal heat-transfer resistance to the iron particles was assumed to be negligible due to its high heat conductivity. The specific heat of iron (0.11 cal/g/K) is actually less than that of zeolite (0.28 cal/g/K); however since its density (7200 kg/m3) is 10 times that of zeolite (720 kg/m3) the effective heat capacity of the zeoliteiron particle system is enhanced to about 2-3 times of the value in the case of pure zeolite with 10-30% inert particle addition by volume. For the present simulations, the bed dimensions were maintained and consequently, with the addition of inerts, the adsorbent loading (as well as the bed adsorbent density) was reduced. The effective adsorbent bed density, specific heat and the bed performance for the various cases of inert particle loadings are summarized in Table 4. In all the runs, the O2 product purity and the BSF factor were kept constant. As the adsorbent loading was reduced due to inert addition, the feed throughput had to be decreased to keep the bed utilization at a constant value. It was found that the temperature deviation in the case of inert addition was reduced by almost 60% than that in the case of the pure adsorbent. For inert particle loading up to 10%, slight enhancements in product recovery by 1-2% were observed, as can be seen in Table 4. However with inert addition of more than 10%, it was observed that the product recovery actually declined. Thus there appears to be an optimum for the amount of inert addition at about 5% (vol). Limits of Operability. It has been shown in this work (Figure 3) that it is possible to operate the adsorption bed packed with LiX zeolite with a pressure ratio as low as 2 with only a slight loss in product recovery. This is possible because of the superior adsorptive properties of this adsorbent. However, an important factor in deciding the effectiveness of the PSA process is the bed size factor (BSF) which was defined previously (eq 4). The BSF factor provides an idea about the productiveness of the process and is indicative of the extent of bed utilization. Since it is desirable to have a maximum product throughput for a given amount of adsorbent, as per definition, we would like to minimize the value of BSF as much as possible. It should be noted that optimization of the PSA process is highly subjective by virtue of the presence of a large number of process variables and no well-defined algorithm exists for the same. There always exists a
Figure 12. O2 product purity (%) and O2 product recovery (%) vs bed size factor (BSF) for pressure ratio ) 2 with LiX (Si/Al)1, 100% Li+) adsorbent.
trade-off between product purity, recovery, and the BSF. In this case, we attempted to extract the minimum possible BSF, keeping the product purity and recovery within tolerable limits. The BSF value may be arbitrarily optimized by manipulating the following operating conditions: (1) time duration of each step; (2) feed throughput (kg of feed/h/kg of sorbent); (3) purge to feed ratio (P/F); (4) end pressure of cocurrent depressurization (PCD). It was decided to operate at low cycle times (125150 s/cycle) in order to sustain high-feed flow rates. The feed throughput was increased to a value just below the breakthrough threshold. It was found that for the given conditions (PH ) 1 bar, PL ) 0.5 bar) cocurrent depressurization pressure (PCD) was optimally located between 0.80 and 0.85 bar. A further decrease in PCD resulted in breakthrough during step (3). Next the purge to feed ratio (P/F) was adjusted to provide an minimum BSF value. It can be seen from Figure 12 that as the BSF value falls, the product purity is initially above 90%, but there is a drastic fall just below the BSF value of 18 kg of adsorbent/kg of O2/h. Similarly, the product recovery increases drastically below this BSF value. Thus, the optimal BSF value for the given LiX adsorbent bed system is approximately 18 kg of adsorbent/kg of O2/h under the process conditions mentioned above. Conclusions PSA simulations were made to compare the performance of LiX (Al/Si ) 1 with 100% Li+) and NaX (13X) adsorbents. A study of the dependence of O2 product recovery on the pressure ratio at a fixed product purity and product throughput reveals that the LiX sorbent gives higher recovery for PH:PL from 1 to 10. By proper choice of operating conditions it is possible to operate a PSA system with the LiX adsorbent at a pressure ratio of 2 with high product purity, recovery, and throughput.
Ind. Eng. Chem. Res., Vol. 36, No. 12, 1997 5365
The performance of the LiX sorbent is considerably better than that of the NaX sorbent as seen from the product recovery versus pressure ratio curve. The effect of the P/F ratio on product purity and recovery was studied and found to be as expected. The Bed Size Factor (BSF) is an important parameter for PSA performance evaluation. In this work, the lowest optimal BSF value for the LiX adsorbent at a pressure ratio of 2.0 was determined to be about 18 kg of adsorbent/kg of O2/h. The temperature excursions in the LiX beds were approximately 4 times higher than that in the NaX beds, due to the combined result of higher N2 heat of adsorption and a shorter bed utilization in the LiX system. One method, namely the addition of high heat capacity inerts, was used in the case of the LiX zeolite in order to overcome the large temperature effects. The method was successful in reducing the temperature deviations during adsorption and desorption, with an improvement in product recovery of 2%. It was also observed that there was an optimum for the volume of inert addition (in the present case 5-10% by volume), beyond which a drop in product recovery was observed. Acknowledgment We are grateful to Praxair, Inc. for donating the X zeolite sample (with Si/Al ) 1) and Nick Hutson for performing Li+ ion exchange and isotherm measurements. This work was supported by the NSF under Grant CTS-9520328. Nomenclature B ) Langmuir parameter (atm-1) BSF ) bed size factor (kg of adsorbent/kg of O2 produced/ h) Cp ) specific heat (kcal/g/K) Dax ) axial dispersion coefficient in adsorbent particles (m2/ s) ∆H ) heat of adsorption (kcal/mol) K ) Langmuir parameter (mmol/g/atm) k1 ) Langmuir temperature dependence constant (mmol/ g/atm) k2, k4 ) Langmuir temperature dependence constant (K) k3 ) Langmuir temperature dependence constant (atm-1) L ) total length of the adsorption bed (m) P ) total pressure (bar) Pem ) mass Peclet number Peh ) thermal Peclet number P/F ) purge to feed ratio qj ) volume-averaged adsorbed amount (mmol/g) R ) gas constant (kcal/mmol/K) t ) time (s) T ) temperature (K) u ) interstitial gas velocity (m/s) y ) mole fraction of the components in the gas phase z ) axial coordinate in the bed (m) Greek Letters R ) thermal dispersion coefficient (m2/s) ) void fraction of the packing t ) bed void fraction including macropores in particles λL ) effective thermal conductivity (W/(m‚K)) Fb ) bed density (kg/m3) Fg ) gas phase density (kg/m3)
τ ) time duration of process step (s) Subscripts CD ) intermediate pressure corresponding to the cocurrent depressurization step co ) cocurrent blowdown step cn ) countercurrent blowdown step g ) gas phase H ) corresponding to the feed step i ) species i j, k ) gas phase component index L ) low pressure corresponding to purge step p ) purge step P ) pressurization step s ) solid phase Superscripts * ) at equilibrium
Literature Cited Baksh, M. S. A.; Kikkinides, E. S.; Yang, R. T. Lithium Type X Zeolite as a Superior Sorbent for Air Separation. Sep. Sci. Technol. 1992, 27, 277. Chao, C. C. Process for separating nitrogen from mixtures thereof with less polar substances. U.S. Patent 4,859,217, 1989. Chao, C. C.; Sherman, J. D.; Mullhaupt, J. T.; Bolinger, C. M. Mixed ion-exchanged zeolites and processes for the use thereof in gas separations. U.S. Patent 5,174,979, 1992. Coe, C. G.; Kirner, J. F.; Pierantozzi, R.; White, T. R. Nitrogen adsorption with a Ca and/or Sr exchanged lithium X-zeolite. U.S. Patent 5,152,813, 1992. Hindmarsh, A. C. LSODE and LSODI: Two New Initial Value Ordinary Differential Equation Solvers. ACM-SIGNUM Newslett. 1980, 15 (4), 10. Kayser, J. C.; Knaebel, K. S. Pressure Swing Adsorption: Experimental Study of an Equilibrium Theory. Chem. Eng. Sci. 1986, 41, 2931. Kayser, J. C.; Knaebel, K. S. Pressure Swing Adsorption: Development of an Equilibrium Theory for Binary Gas Mixtures with Nonlinear Isotherms. Chem. Eng. Sci. 1989, 44, 1. Knaebel, K. S.; Hill, F. B. Pressure Swing Adsorption: Development of an Equilibrium Theory for Gas Separations. Chem. Eng. Sci. 1985, 40, 2351. Leavitt, F. W. Air separation pressure swing adsorption process. U.S. Patent 5,074,892, 1991. McKee, D. W. Separation of an oxygen-nitrogen mixture. U.S. Patent 3,140,933, 1964. Ruthven, D. M.; Farooq, S.; Knaebel, K. S. Pressure Swing Adsorption. VCH Publishers: New York, 1994. Sun, L. M.; Le Quere, P.; LeVan, M. D. Numerical Simulation of Diffusion-Limited PSA Process Models by Finite Difference Methods. Chem. Eng. Sci. 1996, 51, 5341. Yang, R. T. Gas Separation by Adsorption Processes. Butterworth: Boston, 1987; reprinted (in paperback) by Imperial College Press: London and World Scientific Publishing Co.: River Edge, NJ, 1997. Yang, R. T.; Cen, P. L. Improved Pressure Swing Adsorption Processes for Gas Separation: By Heat Exchange between Adsorbers and by High-Heat-Capacity Inert Additives. Ind. Eng. Chem. Process Des. Dev. 1986, 25, 54.
Received for review July 14, 1997 Revised manuscript received September 15, 1997 Accepted September 15, 1997X IE9705214
X Abstract published in Advance ACS Abstracts, November 1, 1997.