Production of Argon from an Oxygen− Argon Mixture by Pressure

PSA experiments were conducted to concentrate argon from a 95:5 (mole ratio) oxygen−argon mixture using a single-bed, three-step cycle operated in 0...
0 downloads 0 Views 212KB Size
Ind. Eng. Chem. Res. 2006, 45, 5775-5787

5775

Production of Argon from an Oxygen-Argon Mixture by Pressure Swing Adsorption Xu Jin, A. Malek, and S. Farooq* Department of Chemical & Biomolecular Engineering, National UniVersity of Singapore, 4 Engineering DriVe 4, Singapore 117576

Kinetic selectivity of four potential adsorbents for the separation of oxygen-argon mixture by pressure swing adsorption (PSA) was compared based on adsorption equilibrium and kinetics measured in the linear range. In the chosen adsorbent, Takeda II carbon molecular sieve (CMS), the measurements were extended up to 6 atm pressure. Appropriate models were developed that explained the measured data and gave necessary parameters for subsequent use in PSA simulation. PSA experiments were conducted to concentrate argon from a 95:5 (mole ratio) oxygen-argon mixture using a single-bed, three-step cycle operated in 0-3 atm pressure range. PSA simulations were conducted using a bidispersed adsorbent model and a dual transport resistance with strongly concentration-dependent thermodynamically corrected transport parameters in the CMS micropores. It was a kinetically controlled separation with a high proportion of faster component in the feed mixture. The performance of the single-bed, three-step vacuum cycle at different operating conditions was systematically studied using the PSA simulation model validated with experimental results. Introduction Pressure swing adsorption (PSA) processes for gas separation are known for their relatively low energy consumption and ease of operation. Air separation is one of the most important industrial applications of this technology. Industrial PSA processes for oxygen production from air use equilibriumcontrolled adsorbents that preferentially adsorb nitrogen to oxygen. The oxygen product purity in such a process is typically limited to 95% because of the presence of argon, which accounts for 1% in feed air and shows practically the same adsorption equilibrium behavior as oxygen on all adsorbents used for equilibrium-controlled PSA air separation. To meet the need for high-purity oxygen, for example, in medical uses, in ferrous metal cutting and welding, etc., further removal of argon becomes important. Since the supply of air is almost infinite, the enriched oxygen stream containing ∼5% argon is also a source for argon production. Hence, an efficient and economic method for oxygen-argon separation is of practical significance. Currently, three methods are commercially used for oxygenargon separation. (1) Cryogenic distillation: high-purity oxygen (99.5%) and argon (99.9%) can be obtained by this method.1 But since the boiling points of pure oxygen and pure argon are very close, a very high reflux ratio and a large number of theoretical stages are required, which decreases the productivity and increases capital cost. (2) Catalytic hydrogenation: in this method, oxygen is removed as water by reacting with hydrogen, and hence, oxygen recovery is not possible. Therefore, this method is only practical as a polishing step to remove trace or small concentrations of oxygen from a crude argon stream. Moreover, the addition of hydrogen increases the operating cost. (3) Cryogenic adsorption: the adsorption equilibrium selectivity of oxygen to argon on 4A zeolite becomes sufficiently large at a very low temperature (-150 °C to -190 °C) to enable adsorption separation.2 However, the process will be highly energy intensive. * To whom correspondence should be addressed. Tel.: (65)65166545. Fax: (65)6779-1936. E-mail: [email protected].

While an equilibrium-based PSA process will fail to efficiently separate oxygen and argon because of the near-unity separation factor on all known adsorbents at atmospheric temperature, kinetically controlled PSA seems to be promising. It has been shown that, in some adsorbents, oxygen diffusion is much faster than that of argon, which may be exploited to lead to the desired separation. Hayashi et al.3 carried out a twostage experimental study with direct air feed using Takeda 3A carbon molecular sieve (CMS) in the first stage to remove argon and Ca-X zeolite in the second stage to remove nitrogen. The purity of oxygen was reported to be 99% with a recovery of 50%. In an earlier study, Hayashi et al.4 described an argon process using 5A zeolite to remove nitrogen and 3A CMS to remove oxygen. In this process, argon product was 99% pure with 40% recovery. Rege and Yang5 theoretically studied the feasibility of separating a 95:5 (mole %) oxygen-argon mixture using BF CMS, which was reported to have a higher diffusivity ratio of oxygen to argon than in Takeda CMS.6 Two different modes of PSA operation, namely, a five-step cycle giving highpurity argon as product and a four-step cycle to obtain highpurity oxygen product, were suggested. In a kinetically controlled separation process, the selectivity depends on both kinetic and equilibrium effects. Hence, both kinetic and adsorption equilibrium data are required for process design. Because interest in CMS has been stimulated mainly due to its ability to separate oxygen and nitrogen, both kinetic and equilibrium data for these two sorbates in CMS have been reported in many studies (see Table 1 in ref 7 for a list of more relevant studies). On the other hand, data for argon adsorption and diffusion are very limited. Reid et al.8 measured adsorption equilibrium and kinetics of argon on a CMS sample supplied by Air Products and Chemicals, Inc., using the gravimetric method. In that study, the Henry’s law constant calculated from extrapolation of virial equation for a wide-range isotherm (0-900 kPa) was in good agreement with that obtained from low-pressure range measurements (0-9 kPa). The results also showed that the kinetics for Ar on this sample followed a linear driving force (LDF) model.

10.1021/ie060113c CCC: $33.50 © 2006 American Chemical Society Published on Web 07/06/2006

5776

Ind. Eng. Chem. Res., Vol. 45, No. 16, 2006

Table 1. Equilibrium and Kinetic Parameters of Argon and Oxygen in CMS and RS-10 Adsorbents in the Linear Rangea oxygenb

argon temp (K) Kc (Kc)DCBT D/c0/rc2 k/c0 L/5

283.15 16.9 16.2 4.29 4.68 0.72

BF CMS 293.25 12.8 13.5 6.23 8.15 0.87

308.15 8.7 9.7 12.2 16.2 0.88

Kc (Kc)DCBT D/c0/rc2 k/b0 L/5

20.6 19.6 0.42 1.56 1.99

Takeda II 15.7 13.4 0.84 2.07 2.26

10.6 10.2 1.78 5.29 1.98

293.15 11.9

oxygenb

argon Takeda I 293.15 17.2 17.2 3.54 9.36 1.76

58.4 169 1.94

283.15 21.2 21.4 2.02 6.01 1.99

308.15 11.8 12.2 8.47 15.8 1.24

293.15 14.0 108 201 1.24

13.9

4.3

3.5

2.7

4.0

84.7 98.9 0.77

2.07

2.78

3.85

99.4

RS-10

a Units: K is dimensionless, D/ /r 2 is 10-4 s-1, k/ is 10-3 s-1; T is K. b Oxygen data for the CMS samples are from Huang et al.9 and for the RS-10 c0 c b0 sample are from Farooq.11

In another study, Ma et al.6 measured the adsorption and diffusion data of argon on three different CMS samples (Takeda 5A, Takeda 3A, and BF CMS). The experiments were carried out in two gravimetric setups, one for high pressure and the other for low pressure. Henry’s law constants obtained from the two ranges were generally consistent. Contrary to the LDF model fit in the study of Reid et al.,8 the uptake was well-described by Fick’s law in the study by Ma et al.6 Hence, the mechanism of argon transport in CMS micropores is not clear from the published studies. In process development, the results from experimental study are normally more reliable, but they do not necessarily represent optimum performance. On the other hand, while simulation can be easily conducted over a wide range for a full search, such results without experimental validation remain doubtful. The best strategy is a combined approach in which first the model is verified with experimental results and then a full search is conducted using the verified simulation model. In the present study, one BF CMS, two Takeda CMS (designated as Takeda I and Takeda II for easy reference to the previous publications from this laboratory), and a modified 4A zeolite (RS-10) were investigated for oxygen-argon separation. Since extensive equilibrium and kinetic data of oxygen in all four adsorbents were measured in previous studies from this laboratory,9 only argon data were measured in the linear range for screening of the adsorbents. Complete equilibrium and kinetics of argon on the chosen adsorbent, Takeda II CMS, were then measured over a wide pressure range, and appropriate models to capture the observed behavior were established. PSA oxygen-argon separation experiments were conducted on the chosen adsorbent to concentrate argon. A PSA simulation model including appropriate descriptions for binary equilibrium and transport in the adsorbent micropores was developed to explain the experimental results. The experimentally verified simulation model was then used to study in detail the effects of different operating conditions on PSA performance. Screening of Adsorbents Adsorption Measurement in the Linear Range. The equilibrium and kinetics of the components in the linear range are normally used to carry out the initial screening of adsorbents, although many PSA processes operate outside the linear range.10 In this study, adsorption measurements were carried out using a constant-volume apparatus. The setup, experimental procedure, and data processing were practically the same as those detailed in a previous publication from this laboratory.9 The only modification was to replace the chart recorder with a data-

acquisition card (National Instruments, model AT-M10-16E10) connected to a personal computer (Intel Pentium III processor). Adsorption equilibrium and kinetics of argon in the four adsorbents were measured in the linear range at three temperatures. At each temperature, three small pressure steps, each of which was ∼0.1 bar, were given to make sure that the measurements were within the linear range where Henry’s law was applicable. As such, the above constant-volume experiments gave constant, limiting kinetic parameters. In the previous publication from this laboratory mentioned above, adsorption and diffusion of oxygen, nitrogen, carbon dioxide, and methane were measured in a BF CMS sample and in a Takeda CMS sample (Takeda I). Similar measurements were also made for oxygen and nitrogen in a second Takeda CMS sample (Takeda II). A dual-resistance model was shown to be the desirable unified approach that fitted the experimental results in the entire range covered in that study. Since the same model was applied to analyze the argon uptake, the model equations are not repeated here. Barrier coefficient and micropore diffusivity are the two unknowns in the dual-resistance model for transport in CMS micropores, which were obtained by individually optimizing the model fit to the experimentally measured differential argon uptake data in three CMS samples. The experimental argon uptake results in RS-10, which is a zeolite adsorbent, were analyzed assuming pore diffusion only. The pore model is actually an extreme case of the dual model. A large value was assigned to barrier coefficient in order to numerically reduce the dual model solution to that of the pore model. The obtained transport parameters, together with Henry’s constant, are listed in Table 1. Two representative optimized model fits to the experimental uptake curves are shown in Figure 1. Figure 1 also shows that the experimental uptake curves at different subatmospheric pressure levels (within the linear range) converged very well, which is consistent with the theoretical expectation that, in the linear range, kinetic parameters are independent of concentration.12 The reproducibility of uptake measurement is further validated by the consistency between the experimental uptake curves of argon in Takeda II CMS measured in two volumetric setups. Dynamic column breakthrough (DCBT) experiments were also carried out using the three CMS samples to validate the results from the above volumetric measurements. For the singlecomponent measurements, argon was fed by mixing with helium, which is considered inert and, therefore, did not adsorb in the adsorbents. The column pressure was controlled slightly higher than atmospheric pressure, and the concentration of Ar

Ind. Eng. Chem. Res., Vol. 45, No. 16, 2006 5777

mt qt - q0 ) m∞ q∞ - q0 where q0 is the initial concentration and qt is the average concentration in the adsorbed phase at time t:

qj )

Figure 1. Experimental argon uptakes in Takeda II CMS and RS-10 in the linear range of the isotherm at 293.15 K and corresponding model fits. Takeda II and Takeda IIa represent experiments at ∼0.1 bar and ∼0.2 bar, respectively. Takeda IIb was carried out at ∼0.1 bar in a second constantvolume apparatus.

m∞

was controlled at a low level of ∼3 mol %. The corresponding partial pressure level was well within the linear range. The Henry’s law constants obtained from the DCBT study are also listed in Table 1 for comparison with volumetric data. Henry’s law constant based on particle volume, K, is related to the mean residence time of adsorption breakthrough, ht, by the following equation:7,13

ht )

∫0∞

( ) 1-

[

ce 1 - b L dt ) 1 + K c0 V0 b

]

(1)

where L is the column length, V0 is the interstitial feed velocity at the inlet, b is the bed voidage, c0 is the constant adsorbate concentration in the feed, and ce is the changing adsorbate concentration in the exit stream measured continuously in the breakthrough experiment. K is related to the Henry’s constant based on the crystal volume, Kc, by the following equation:

K ) p + (1 - p)Kc

(2)

Figure 2 shows the agreement between a representative experimental breakthrough curve and the model prediction using transport parameters from the volumetric data. To magnify the early part of the response, results after 750 s are shown in the inset. Simulating column dynamics in the linear range of the isotherm is a limiting case of the high-pressure adsorption step in PSA simulation. The model equations will be discussed together in a later section. Relative Importance of the Two Resistances in the CMS Samples. Fractional uptake, following a step change in gasphase concentration, is defined as

∫0r qr2 dr c

For simplicity, only micropore transport is considered in this part of the discussion. Neglecting macropore resistance for the gases of our interest in the chosen adsorbents will not affect the conclusions. Also, constant transport parameters and constant gas-phase concentration are assumed in this section. Subjected to these assumptions, the analytical expressions of singlecomponent fractional uptake for different transport mechanisms have been well-established in the literature.14,15 For transport controlled by diffusion in the micropore interior, the uptake is given by

mt

Figure 2. Comparison between experimental and predicted breakthrough of argon in a Takeda II column at 293.15 K in the linear range of the isotherm: column length, 40 cm; internal column diameter, 3.8 cm; bed voidage, 0.35; interstitial feed velocity, 1.57 cm/s; feed, ∼3% argon in helium; operating pressure, slightly above atmospheric. Equilibrium and kinetic parameters obtained in the volumetric experiments were used for model prediction.

3 rc3

)1-

6





π2 n)1

e-n π Dct/rc 2 2

2

n2

(3)

where mt is the amount adsorbed up to time t, m∞ is the total amount adsorbed since the introduction of step change until new equilibrium is reached, and Dc/rc2 is the pore diffusional time constant. In the case of transport controlled by a restriction at the pore mouth rather than the pore interior, the uptake rate is given by

mt ) 1 - e-kbt m∞

(4)

where kb is the barrier coefficient. On the basis of the analysis of the first and second derivatives of eqs 3 and 4 with respect to xt, it may be concluded that the fractional uptake curve of a pore model is monotonically increasing and convex upward for the entire range, while that of a barrier model has a turning point from concave to convex shape. Such a distinction is not seen when derivatives with respect to t are compared. Hence, fractional uptake, mt/m∞, plotted against xt can clearly shed light on possible transport mechanism. It may be seen in Figure 1 that, consistent with our understanding from vast literature studies on uptake of gas in zeolite adsorbents, uptake of argon in RS-10 is pore-diffusion controlled. The S-shape in the early part of the uptakes in the CMS samples is a clear sign of the presence of barrier resistance. This observation is in line with the behavior of oxygen, nitrogen, methane, and carbon dioxide in the CMS samples reported in another study from this laboratory.9 When both pore and barrier resistances are important, fractional uptake of the dual model is given by the following equation,16

mt m∞



)1-



6L2 exp(-βn2Dct/rc2)

n)1 β

2 n

[βn2 + L(L - 1)]

(5)

where βn are the nonzero roots of the equation βn cot βn + L 1 ) 0, and L ) kb/3Dc/rc2. Loughlin et al.16 also studied the overall mass transfer coefficient koverall for the dual-resistance system and derived

5778

Ind. Eng. Chem. Res., Vol. 45, No. 16, 2006

the following relationship:

Dc 5L koverall ) 3 2 rc 5 + L

(

)

(6)

According to Glueckauf and Coates’s approximation,17 for an equivalent linear driving force representation of a porediffusional problem, the linear driving force coefficient is given by 15Dc/rc2. Therefore, on the basis of Glueckauf’s approximation, when L ) 5, barrier resistance and pore resistance have equal contribution to the overall resistance. Hence, the value of L/5 may be used as a useful measure of the relative importance of pore and barrier resistances. The L/5 values for argon and oxygen on the three CMS samples are in the range of 0.82.3, as shown in Table 1. This indicates that, in the linear range, both pore resistance and barrier resistance are important in the CMS samples. It is well-known that Glueckauf’s linear driving force representation of a pore-diffusional problem is only approximate and cannot capture the true dynamics of a porediffusion-controlled process.18 Hence, the koverall, given by eq 6 for a dual-resistance system, does not provide an alternative to the full solution, if the objective is to accurately capture the process dynamics. Kinetic Selectivity of Oxygen-Argon in the Adsorbents Studied. To estimate the efficiency of an adsorbent for gas separation, the selectivity is generally defined as

RAB )

qjA/cA qjB/cB

(7)

where A and B denote the components to be separated. For an equilibrium-controlled separation, the above equation reduces, in the linear range of isotherm or when both the components obey Langmuir isotherm, to the ratio of the Henry’s constant of the two components. For a kinetically controlled separation, eq 7 may be manipulated to the following form:

RK )

( ) ( )

mt q/A m∞ AcA mt q/B m∞ BcB

(8)

It is clear from the above equation that kinetic selectivity is time dependent and will ultimately approach the equilibrium selectivity. Assuming uncoupled equilibrium and kinetics, Ruthven et al.10 showed that, for short contact and porediffusion-controlled uptake, the kinetic selectivity is given by

(RK)pore )

x

KA KB

(Dc0)A (Dc0)B

The analogous equation for barrier-controlled uptake is

(RK)barrier )

KA (kb)A KB (kb)B

These two equations provide simple limiting forms for first screening of adsorbents. However, it is obvious that the series solution of the dual-resistance-controlled fractional uptake given by eq 5 will not reduce to a simpler expression at short time. To compare the four adsorbents for oxygen-argon separation, the complete time history of kinetic selectivity was calculated

Figure 3. Comparison of calculated kinetic selectivities of the four adsorbents for oxygen/argon separation.

using the analytical solutions for mt/m∞. A FORTRAN program was used to evaluate the series solutions corresponding to the pore model and the dual model. On the basis of the findings reported in the previous section, pore-model and dual-model solutions were used for RS-10 and CMS, respectively. The number of terms in the series solutions was increased until the difference from an additional term was