Application of High Silica Zeolite to Remove Organic Vapors from Industrial Effluent Air

Fundamentals of Adsorption Proc. IVth lnt. Conf. on Fundamentals of Adsorption, Kyoto, May 11-22, 1992 Copyright 0 1993 International Adsorption Society

Application of High Silica Zeolite to Remove Organic Vapors from Industrial Effluent Air

Y. Takeuchi', H. Iwamotol, S. Asano2 and M.Harada2 1) Department of Industrial Chemistry, Meiji University, Higashi-Mita, Tama-ku, Kawasaki 214, Japan 2) TOSOH Co.. Ltd.. 4560 Kaisei-machi, Shin-Nan-yo, Yamaguchi 746, Japan

ABSTRACT A series of small-scall laboratory tests were performed under almost the same conditions used in industry, to remove lacker-solvent vapor from air. A flow method was used to measure breakthrough curves as well as equilibrium data. Starting from single component vapor adsorption, three binary and one ternary vapor systems among 1-Butanol,pXylene and 2-Butoxyethanol, respectively, were studied, Equilibria for each single component vapor in air were obtained from the analysis of breakthrough curves, and were expressed by the Langmuir equation. Binary equilibria were obtained in the same way as above, and were correlated by Markham-Benton equation. Regarding intraparticle diffusion, it was found from the analyses of breakthrough curves that macro-pore diffusion was the rate-deterniining step. Break times could be estimated by Extended-MTZ-Method with fairly good accuracy. Furthermore, for regeneration of spent High Silica Zeolite particles, it was known that degradation of adsorption capacity did not occur when those of higher SiO1/A12O3ratio were used. INTRODUCTION Organic solvent has been used in many industrial processes such as painting, and the solvent vapors discharged from the processes generally bring about bad effects to human bodies. Therefore, it is necessary to remove them before discharging to environment, even if the organic solvent recovered is difficult to reuse. Activated carbon adsorption has been used widely for the removal or recovery of the solvent vapors. However, ignition of activated carbon bed occurred in some cases, and use of zeolite is considered. As zeolite such as 4A, 5A, 13X etc. has serious disadvantages in applying for adsorption of organics, because it adsorbs preferentially water vapor, it was thought that such types of zeolite as above are not suitable for the removal of solvent vapor from humid air. In recent years, high silica zeolite (hereafter HSZ) has been commercialized which has higher SiOz/A1203ratio and is hydrophobic or nonpolar, and adsorbs preferentially organic vapors. In this study, fixed-bed breakthrough curves were measured for 1-Butanol. pXylene and 2-Butoxyethanol (hereafter called as n-Butanol, p-Xylene and n-Butyl Cellosolve) which are main components of mixed solvent used generally for the coating process in automobile industry, and data of adsorption equilibria and rate processes are described. Also, results on regeneration of spent HSZ (desorption of solvent-laden air) are shown. FIXED-BED ADSORPTION FOR SOLVENT-LADEN AIR ExDeriments were carried out to obtain breakthrough curves in measuring concentrationsof gases at the dxit of HSZ beds by a well-known flow method. The schematichiagram of experhental apparatus is shown in Fig. 1. The air from a blower was fed to a silica gel column to make the feed air dry. A part of the dried air was fed to respective evaporators containing a certain organic solvents. Concentration of each organic solvent in air were changed by changing flow rates at each evaporator. Then, the air stream was fed to a fixed bed of HSZ chosen among those shown in Table 1. These samples (A) to (D), originally in 0.003m pellets, were used after crushing to be 16/22 mesh (mean particle diameter 0.0008388m). Organic solvents, i.e., n-Butanol, p-Xylene and n-Butyl Cellosolve. respectively, were of research grade. Experimental conditions are shown in Table 2. The amount 641

648

Y.Takeuchi, H. Iwamoto, H. Asano and M. Harada

adsorbed at equilibrium was determined by graphical integration of the breakthrough curves or the weight increase of HSZ bed obtained from each run. Both results agreed well within 10%error.

:............................................................. 1 Blowcr 4 Row Mcler

7 Adsorption Column

2 Air Drycr(SilicaGel Column)

:

3 Mars Flow ConaoUei 6 Vaporizer

5 Constanr TcmperaNre Bath 8 Gas Chromatograph

Fig. 1 Schematic diagram of experimental apparatus for Adsorption expperiment.

Table 1 Physical Properties of High Silica Zeolite Sample (4 (B) (C) Diameter [ml 8.39~10-~ 8 . 3 9 ~ 1 0 ~ 8 . 3 9 104 ~ 4.97~105 6.01~101 4.62~10s Surface m a [m2n(g] Pore volume [m3/kg] 6 . 6 9 ~ 1 0 - ~ 7.24~10" 5.81~10~ 2.30~103 2.23~103 2.17~103 True density [kg/mS] Particle density [kg/m3] 9.06~102 8.51~102 9.07~102 0.558 Poreratio [-3 0.606 0.617 Type of Zeolite Y Y Madenite 460 SiOZ/A1203 [-I 14.5 220 note : All samples were produced by Tosoh Co.. Ltd.Japan

@)

8.39~10~ 3.35~105 3.29~10~ 2.42~103 1.35~103 0.444

ZSM-5 2200

Table 2 Experimental Condition High Silica zeolite, Sample (A) - (D) (Produced by TOSOH Co., Ltd.) Adsorbate n-Butanol, p-Xylene, n-Butyl Cellosolve Temperature 298.2 [K] Linear flow rate 0.17 - 0.22 [m/Sec.] Influent concentration 100 - 500 [ppm(vol.)] Bed density 460 - 560 [kg/m3] Bed length 0.04 - 0.15 [m] Adsorbent

RESULTS AND DISCUSSION (1) Single component adsorption Adsorption isotherms of n-Butanol for Sample (A) to (D) were determined as shown in Fig.2. Both Samples (A) and (B) had the same structure. but the latter showed higher amount adsorbed due to higher SiOZ/Al2O3ratio. Adsorption isotherms for each component - Sample (A) system are shown in Fig.3. These data could be correlated with the Langmuir equation [4], Eq.(l), and two constants, K and qjd. were determined by regression of the experimental data, as listed in Table 3.

2.0 1.O

0.01 0.0

1.O

C x lb [movm'] Fig. 2 Adsorpfion isotherms for n-Butanol vaper - HSZ systems. (298.2K)

I

2.0

High Silica Zeolite to Remove Organic Vapors

2.0

1.0

0.0

C

0.0

x 10* [moYm']

0.1

PIP0

Fig. 3 Adsorption isotherms for single - component solvent vaper - Sarnple(A) system. (298.2K)

1-1

649

0.2

Fig. 4 Adsorption isotherms for single - component

solvent vaper - Sanple(A) system. (298.2K)

Table 3 Langmuir Constants for single component Systems (298.2K) n-Butanol p-Xylene n-Butyl Cellosolve qinf. K qir. K qinf. K [moWgl [m3/mol] [moVkgl [m3/mol] [moVkgl [m3/moll 1.61 693 1.16 1680 1.38 3120 ------2.17 555 1.40 1306 _-------_---0.98 2090 ---_-_------1.31 2760 7 1 7 r 1.61 59 (A)* 1.16 77 1.38 1290 (A)**

Solvent vapor Sample (A) (B) 02)

03

for binarv svstems * n-Bu&h - n-Butyl Cellosolve - Sample (A) system ** p-Xylene - n-Butyl Cellosolve Sample (A) system

-

Also ,the above results are represented in Fig.4 in terms of the liquid volume of each adsorbate vs their relative pressure. It was known that the concentrations of n-Butyl Cellosolve vapor chosen in the study were near its saturated pressure but others were not. However, the data were located in the same line and it was considered that solvents were adsorbed in liquid state. Further, concentrations which bring about capillary condensation were calculated by the Kelvin equation, taking Kelvin radius to be the micro-pore radius of each sample. The contact angle at the capillaries was assumed 0'. Though the values calculated may include a certain errors, capillary condensation seems to occur at about 5 ppm for p-Xylene and n-Butyl Cellosolve, and about 110 ppm for n-Butanol, respectively. As a result, the assumption adopted in the correlation shown in Fig.4, i.e., solvents adsorbed are in liquid state, seems to be acceptable. For Langmuir-type equation systems, it is known to obtain constant patterns in breakthrough curves, and the values of intraparticle volumetric mass transfer coefficient(pk,a,) can be obtained from breakthrough cuwes using the r-cmethod [3], where the value of fluid film volumetric mass transfer coefficient(kP,) can be estimated, e.g., by Yoshida et al's. equation.rl31. Using the well known procedure, the values of Pksavre.g., for n-Butanol - Sample (A) system, were obtained and plotted against the slope of operational line p ,adsorption coefficient, as shown in Fig.5. The values show some deviations but are constant irrespective of p. Therefore, the effective intraparticle diffusivity @p) was calculated from the average values by Glueckauf s equation [2], Eq.(2).

DP= Pksavd; 60(1-4

( 2)

The ratio of the effective diffusivity to molecular diffusivity(Dm) were calculated as aboutO.l. It can be concluded that the intraparticle diffusion is determined by the transpa within mampoxe,

650

Y.Takeuchi, H. Iwamoto, H. Asano and M. Harada

€a2,

because the ratio, 0.1 coincides with the value when macro porosity ~,=0.3,and the tortuosity factor k2=3 are used. Therefore, the adsorption equilibria and the values of Pksav calculated assuming macro pore diffusion controlling and , times of each taking D P / D A B = O . ~break component can be estimated.

(2) Multi-component adsorption An example of breakthrough curves obtained for ternary vapor - Sample (A) system are shown in Fig.6. n-Butyl Cellosolve was always the most adsorbable component even though concentration was low, and, when the concentration at the outlet of the column became the same to that at inlet, it was known that almost all the amount adsorbed consisted almost only of n-Butyl Cellosolve.Other two components showed almost the same adsorbability, and the order of breakthrough or appearance at the exit of the column with respect to both components, i.e., n-Butanol and p-Xylene, changed depending on their initial concentrations. This phenomenon could be interpreted from the types of breakthrough curves for n-Butanol - pXylene - Sample (A) system, as shown in Fig. 7. Three types of breakthrough curves were observed depending on the initial concentrations expressed in terms of mole fraction (n-Butanol / (n-Butanol + p-Xylene)). The point where reversal of the order of breakthrough occurred was 0.4 mole fraction which corresponded to azeotropic point of vapor - liquid equilibrium for the system. This phenomenon is the same as the result observed by Basmadjian et al. [l], for Ethylene - Carbon Dioxide - 5A molecular sieves system. Also, when both components had the same adsorbability, the two components behaved as if one component. Experiments were performed for n-Butanol - n-Butyl Cellosolve - Sample (A) system and p-Xylene - n-Butyl Cellosolve Sample (A) system, respectively. Regarding these systems, breakthrough curves showed the constant pattern for each component. The adsorption equilibria are shown in Fig.8 in X-Y diagram, which could be expressed by Markham Benton equation [5].Eq.(3), with constant values determined from the comparison of experimental and estimated amounts adsorbed. [6,8]

200

100

0

P

300

[m'kgl

Fig. 5 Dependence of pksav on p for n-Butanol -Sample(A) system.

2.0

.

0 n-Butanol

COI = 413.5 ppm Coz = 134.0 ppm Coi = 94.2 ppm

A pXylcnc

n-Bury1 Cc~~arolvc

.

u

nn 400

0

800

I200

Time Elapsed [min.]

(EX.1)

-

..... g.......:""""""...........::...... 0 =

2

7 ,

,,o

1

i :

...4

u IU miscc. 0.072m

I

0.0 0

-:* I

200

4M1

600

800

.

IMX)

Time Elapsed [min.]

(EX. 2) Fig. 6 Example of experimental breakthrough curves for the ternary solvent vaper - Sample(A) system.

The most suitable set of constant values of Ki determined by use of mal and error method are listed in Table 3. On the other hand, values of qd.are assumed to be the same as that of single component

High Silica Zeolite to Remove Organic Vapors

systems. Also. almost the same way as the r-4 method [3] can be used to determine pksavor intraparticle diffusivities, from breakthrough curves for single component adsorption. The values of pk,a,for binary systems are plotted against p as shown in Fig.9. for an example of nButanol - n-Butyl Cellosolve system. The values of pksav are almost constant irrespective of p in analogy with the single component system, and it can be seen that the intraparticle diffusion is determined by macro-pore diffusion, again. Further, values of Dp calculated from pksav by Glueckaufs equation were again about one tenth of the values of DAB. By use of adsorption equilibria equation and pksav calculated taking Dp/Dm = 0.1, break times of each component were estimated from the equation included in the Extended-MTZ-Method [6- 121, as shown in Table 4. It became clear from Table 4 that the break times estimated agreed fairly well with experimental data.

*

0.4

0.2

1 I J1

I

I

1

I

I .o

.bUt)0.4 0.5

0.0 ......................

.

:

u 00

....................... T IEhpd ............................................... ..-.

?-Fa lapid

Fig. 7 Three cases of breakthrough curves for n-Butanol - pXyIene - Sample(A) system.

B

....

0.0 . 0.0

mole fraction Cnrrav-! I (Cnna.d + Cnpxyicm)

n-Butyl Ccllorolvc

.,,,

,

651

1

4 a d A.

0.2

0.4

0.6

x [-I

0.8

1.0

P pxy-

x,lml *..$C.IMrlYrml -lu~y~&~.e~~Ru-. nBaylCollrmrrsyiml CdIWd" lPX,k.. *B.Rl C a M n .wan)

0 .B"RI

Fig. 8 X - Y diagram represented by adsorption equilibria for n-Butanul - n-Butyl Cellosolve - Sample(A) system and p-Xylenc - n-Butyl Cellosolve . Sample(A) system.

0

j0

LOO

I50

200

P [m'kgl

250

300

350

Fig. 9 Dependence of pklav on p for each component of n-Butanol - n-Butyl Cellosolve - Sample(A) system.

Table 4 Comparison of break times for binary vapor adsorption obtained from experimental breakthrough curves with those predicted by the Extended-MTZ-Method Run t g l [min.] tg2 [min.] No. exp. calc.* calc.** exp. calc.* calc.** 1 158.3 187.5 161.7 561.3 649.3 598.4 2 288.3 262.9 247.5 550.7 516.2 515.2 3 207.0 194.6 181.4 783.7 754.6 770.1 4 186.6 184.9 179.1 527.3 548.3 548.0 5 222.2 200.2 200.8 739.4 733.9 750.2 501.5 6 107.6 91.5 119.3 612.1 7 303.3 262.7 275.0 655.7 7 14.7 679.0 257.7 250.9 8 271.1 549.6 586.0 592.5 163.0 138.0 9 148.3 357.8 346.1 407.1 109.1 97.8 10 119.4

2%

-"zE---

From Run 1 OI 5 : n-Butanol - n-Butyl Cellosolve - Sample (A) system From Run 6 to 10 : p-Xylene - n-Butyl Cellosolve - Sample (A) system * from the Extended-MTZ-Method by use of the experimental pksav ** from the Extended-MTZ-Method by use of lhe P&av estimated by use of Fig.2, taking DPm=O.I

652

Y.Takeuchi, H. Iwamoto, H. Asano and M. Harada

ON THE DEGRADATION OF HIGH SILICA ZEOLITE BY REPEATED USE Experimental II Experiments were carried out ,repeating consecutively adsorption of vapor from vapor-laden air and desorption by heated air. The adsorption step was carried out by a flow method described above and the experimental conditions are listed in Table 5 . The desorption step was performed by feeding heated air (473.210 to the HSZ bed used for adsorption. The schematic diagram of experimental apparatus is shown in Fig. 10.

I Blower 4 Re

Ham

7 Sampling Bag

2 hir Dryer (SilicaGel Column) 5 Zeoliic Bed

3 M a s now Cmmun 6 E l d c Furnace

Fig.10 Schematic diagram of expenmental apparatus for desorption experiment.

The air from a blower was fed to silica gel column to make the feed air dry. Then, the air was heated from room temperature to 473.2K by preheater, before supplying to HSZ bed. HSZ used here were Sample (A) and (B),respectively, and the organic solvents used were p-Xylene and n-Butyl Cellosolve. For the samples subjected to an adsorption and desqtion step, i.e. one cycle, the degree of degradation, was determined, based on the amount of p-Xylene adsorbed onto the HSZ bed, in equilibrium with 400 ppm of single vapor of p-Xylene in air at 298.2K. Also, desorption curves were obtained by measuring concentrations of vapors at the exit of HSZ bed. The total amount desorbed was determined by graphical integration of the desorption curves obtained for each cycle, and the ratio of desorption was calculated as the ratio of the amount desorbed to the amount adsorbed. Table 5 Experimental conditions High Silica Zeolite, Sample (A) & (B) Adsorbent No. of cycle repeated 2, 4,8, 16, 32 [cycle] 0.07 [m] Bed length Adsorption step Desorption step Air Influent gas p-Xylene = 350 [ppm] Influent gas n-Butyl Cellosolve = 150 [ppm] Linear flow rate 0.2 [dsec.] Linear flow rate 0.2 [m/sec.] Adsorption temperature 298.2 [K] Desorption temperatwe 473.2 [K] 40 [min.] Adsorption time Desorption time 40 [min.]

of HSZ

The amount of p-Xylene adsorbed was plotted against the cycle number n as shown in Fig.11. Regarding Sample (A), amount of p-Xylene adsorbed were decreased until 8 cycles, but in subsequent runs, the values were constant. Regarding Sample (B), the amount of p-Xylene adsorbed were almost constant irrespective of the number n, in other words, no degradation was observed. This is because Sample (B) has higher SiOdA1203 ratio and has less catalytic effect to the solvent. Therefore, it can be concluded that degradation of adsorption capacity did not occur when higher SiOz/Alz@ ratio was taken. Next, how the ratio of desorption changed with for p-Xylene n was considered. As shown in Fig. 12, the data of Sample (A) and (B) for p-Xylene are almost constant irrespective of the cycle

High Silica Zeolite to Remove Organic Vapors

653

number n and always higher than the values for sample(B). However, for n-Butyl Cellosolve, the degree of desorption for Sample (B) was almost constant irrespective of n, and those for Sample (B) changed until 8 cycles. From these results, it can be seen that the cause of degradation for Sample (A) is the accumulation of n-Butyl Cellosolve not desorbed or the change in the structure of micro pores. From identification of residual substances, it was known that they were n-Butyl Cellosolve and Diethylene Glycol Diethyl Ether. 0.15

'

.

I

.

,

.

I

.

,

.

,

'--CL0-0

-

5

I

014

0.13

C

0.12

.

~ I

-

70

t

A

C

; . 6o 50

:k: Sample A

011

. .

"

'

'

I

'

'

0 Sample 8

.

5

0

10

20

15

No. of Cycles [-I

25

35

30

Fig.12 Dependence of ratio of desorption on Cycle No. n.

In industrial operations, solvent vapor-laden air is not always dry,therefore, it is necessary to know the effect of humidity on adsorption capacity of beds. For n-Butanol - Sample (A) system, as example, the amount adsorbed obtained from a series of experiments is plotted against relative humidity as shown in Fig.13. Where, experimental conditions, other than relative humidity, were kept constant. It became clear that adsorption equilibria and adsorption kinetics were not affected by relative humidity R.H.

60%. AS the air used in industrial processes shows generally 40 to 608v the effect of relative humidity on the amount adsorbed seems to tie not so large.

CONCLUSIONS

0.2

3=

,

,

'

I

.

,

.

,

.

L/.--o-o'

0.1

'

-07

0 n.~utma~ 0 W.er

0 Total

00 u

.

". LU

.,so;, '

n 4u

w

R H [%I

au

IW

Fig 13 Change of the amount adsorbed with relative humidty for n- Butanol - Sample(A) system

Adsorption and desorption of organic solvent vapor on and from HSZ beds were studied. Adsorption equilibria could be expressed by Langmuir Equation 141 for single component systems and by Markham-Benton equation [5] for multi-component systems. Also, break times for fixed bed adsorption could be estimated by the Extended-MTZ-Method[6-121 with fairly good accuracy from 1 and from intraparticle volumetric mass transfer coefficients (pksa,) calculated taking D~/DAB=O. Gluckauf s equation. Furthermore, for regeneration of spent HSZ particles, it was known that degradation of adsorption capacity did not occur when those of higher SiOdA1203 ratio were used. From these results, it is clear that HSZ studied here can be used satisfactorily to remove organic solvent vapors from industrial effluent air.

NOMENCLATURE

av = specific surface area in the bed [m2/m3]

C = concentration of solvent vapor in air [moVm3] DP= effective intraparticle diffusivity [m%] dp = diameter of particle [m] DAB= molecular diffusivity [m2/s]

654

Y.Takeuchi, H. Iwamoto, H. Asano and M. Harada

K = Langmuir constant [m3/mol] kfav = fluid film volumetric mass transfer coefficient [ l/s] ksav = intraparticle volumetric mass transfer coefficient based on the difference in amount: adsorbed [kg/m3sl q = amount of solvent vapor adsorbed [moVkg] qd.= amount adsorbed at monolayer saturation [moVkg] t = time elapsed [rnin.] tBi = break time with respect to i-component [min.] P = vapor pressure [mmHg] PO= saturation vapor pressure [mmHg] R.H. = relative humidity [%] u = linear flow rate of gas [ d s ] X = mole fraction of gas-phase [-I Y = mole fraction of adsorbate-phase [-3 Z = bed length [m] p = adsorption coefficient (slop of operating line) [m3/kg] E = void fraction of the bed [-I 0 = initial or inlet i = refer to component exp. = experimental value calc. = calculated value ACKNOWLEDGMENT The authors would like to express their deepest thanks to Messrs. A.Shigeta, H.Kuwabara, K.Nishinaka, K.Miyamoto, O.Kohyama, T.Sugiyama and N.Miyata for their assistance in experimental works. LITERATURE CITED [I] Basmadjian, D. et al., Can. J. Chem. Eng., 58 (1980) 185 34 (1955) 1540 [2] Glueckauf,E., Trans. Faraday Sw., [3] Kawazoe, K. and Y. Fukuda, Kagaku-Kogaku, 29 (1965) 374 [4] Langmuir, I., J. Am, Chem. SOC.,40 (1918) 1361 [ 5 ] Markham, E. C., and A. F. Benton, J. Am. Chem. Soc.,53 (1931) 497 [6] Takeuchi, Y. and E. Furuya, Roc. 1st Int. Conf. on Fundamentals of Adsorption, Engineering Foudation, New York, 1984.p.629 [7] Takeuchi, Y., E. Furuya and Y.Suzuki, Proc. 2nd Int. Conf. on Fundamentals of Adsorption, Engineering Foudation, New York, 1987,p.547 [8] Takeuchi, Y. and E. Furuya, Proc. 3rd Int. Conf. on Fundamentals of Adsorption, Engineering Foudation, New York, 1991, p.889 [9] Takeuchi. Y. and A. Shigeta, J. Chem. Eng. Japan, 24 (1991) 41 1 [ 101 Takeuchi, Y., T. Wasai and SSuginaka, J. Chem. Eng. Japan, 11 (1978) 458 [ll] Takeuchi, Y., Y.Suzuki and E. Furuya, J. Chem. Eng. Japan, 12 (1979) 486 [12] Takeuchi, Y. and E. Furuya, J. Chem. Eng. Japan, 13 (1980) 500 [13] Yoshida, F. et al., A. I. Ch. E. J., 8 (1962) 5