Frederic Heymes; Treatment of Toluene by Absorptio_2007

Chemical Engineering Science 62 (2007) 25762589www.elsevier.com/locate/ces

Treatment of gas containing hydrophobic VOCs by a hybridabsorptionpervaporation process: The case of toluene

Frederic Heymesa, Peggy Manno Demoustierb, Franoise Charbitb, Jean-Louis Fanloa,Philippe Moulinb,

aLGEI, Ecole des Mines dAls, 6 avenue de Clavires, Als 30319, FrancebLPPE, Universit Paul Czanne, Batiment Laennec, Hall C, Europole de larbois, BP 80, Aix en provence 13545, France

Received 12 April 2006; received in revised form 1 February 2007; accepted 2 February 2007Available online 20 February 2007

Abstract

The work focuses on a hybrid process for treating air charged with a hydrophobic volatile organic compound (VOC), coupling an absorptionprocess with membrane pervaporation in order to reuse the absorbent. Toluene was chosen as the target VOC. Four topics were investigated:choice of the absorbent, hydrodynamics and mass transfer in a packed column, regeneration by pervaporation and nally analysis of the couplingof the two processes. In a previous study, 7 absorbents were compared with regard to experimental data (gasliquid equilibrium constants,viscosity) and data from the literature. Di(2-ethylhexyl) adipate (DEHA) was shown to be the most suitable absorbent. In the rst part of thiswork, experiments in a packed column showed that the viscosity of DEHA led to an increase in pressure drop, which nevertheless remained ata reasonable level. Mass transfer experiments were performed and kinetic constants (KLa) calculated. It was proven than washing with DEHAis highly efcient for toluene absorption. The most innovative part of the work is the regeneration of used absorbent by pervaporation. PDMSwas chosen as the active membrane layer. Pervaporation ow rates of toluene were measured for the DEHAtoluene solutions correspondingto column foot concentrations. Transfer resistance is mainly controlled by the liquid boundary layer close to the membrane. The system wasmodelled and several interesting conclusions deduced. Solving the equations by means of a numerical method enabled calculation of the columnheight and membrane surface area required to treat a 10 000m3 h1 gas ow charged with 5 gm3 of toluene. 2006 Elsevier Ltd. All rights reserved.

Keywords: Absorption; Chemical processes; Environment; Mass transfer; Membranes; Packed bed

1. Introduction

Atmospheric volatile organic compounds (VOCs) are in-volved in photochemical reactions creating ground-levelozone (urban smog) and entail environmental threats suchas global warming, acid rain and air toxins. Although VOCsare emitted naturally, human sources are predominant in ur-ban/industrialized areas. In order to reduce anthropic emis-sions, legislation such as the Clean Air Act Amendments(EPA, 1998) and EEC Directives (EEC, 1999, 2001) specifymaximum concentration levels for VOC emissions into theatmosphere. When no upstream process improvement can lead

Corresponding author. Tel.: +33 4 42 90 85 01; fax: +33 4 42 90 85 15.E-mail address: [email protected] (P. Moulin).

0009-2509/$ - see front matter 2006 Elsevier Ltd. All rights reserved.doi:10.1016/j.ces.2007.02.001

to a decrease in VOC emissions, it is necessary to set up anair treatment process. Currently, various processes are availablefor VOC abatement, such as thermal or catalytic oxidation, ad-sorption, condensation, absorption, membrane permeation andbiological treatments (Le Cloirec, 1998). Evaluation and selec-tion of an appropriate VOC abatement technology depends onthe concentration and nature of the compounds, the ow rateof the vent gas and other factors such as safety or economicconsiderations. However, no single method can be used in allcases: most of the methods are specic in nature.

The aim of this research work is to study a hybrid processcoupling an absorption tower with a membrane separation. Theprocess ow sheet is given in Fig. 1. The loaded vent gas owsthrough a packed column in order to transfer the VOCs tothe liquid phase. Consequently, the loaded air is puried. Theliquid collected at the bottom of the column and enriched in

F. Heymes et al. / Chemical Engineering Science 62 (2007) 25762589 2577

Fig. 1. Process ow sheet of the hybrid air treatment process.

Fig. 2. Overall mass transfer system.

VOCs is pumped to the membrane device. Thanks to a vacuumpump, the VOCs are extracted from the absorbent through themembrane and the latter one can be reused in the absorber.

High-boiling organic absorbents were shown to be highlyefcient to trap hydrophobic VOCs in polluted air. However,these absorbents are not used in industry because of the prob-lem of the used absorbent. The organic absorbents may not bereleased in the nature nor in a water treatment unit and requirea specic treatment. Since the absorbent will ow in a closedloop in the considered process, high-boiling organic absorbentscan be used in the hybrid process.

The analysis of the system shows that three phases are in-volved: the gas phase of the exhaust air and the vacuum outlet,the liquid phase of the absorbent and a pseudosolid phase of themembrane material. However, the polymers used in pervapora-tion behave likely as a supplementary liquid phase than a solidporous phase. Fig. 2 represents all phases that the VOCs willhave to cross. Since the process involves many phase transfers,absorption equilibriums and boundary layers will play a major

role. Therefore, the following points have to be considered:

The thermodynamic equilibriums between the VOCs and theliquid phase (1), between the liquid phase and the membranematerial (2) and between the membrane material and thevacuum gas phase (3).

The mass transfer kinetics between the air and the absorbent(A and B), between the liquid and the membrane material(C) and inside the membrane (D).

It has to be noted that contrary to the scheme in Fig. 2, theconcentration into the liquid vary from point to point in the pro-cess and is not the same in the absorber and the pervaporationunit.

In this paper, we shall present the main experimental resultsobtained from absorption and pervaporation studies carried outin a pilot unit. Toluene was chosen because it is a commonand widely used compound, highly hydrophobic. In the Discus-sion section we shall study the coupling of the two processesinvolved.

2. Experimental results

2.1. Thermodynamic equilibriums

The key point of the process is the choice of the absorbent.The aim of a previous paper (Heymes et al., 2006a) was to se-lect the more suitable absorbent. A target hydrophobic VOCwas chosen (toluene). Seven absorbents were experimentallyinvestigated and theoretical considerations led to the selectionof di(2-ethylhexyl) adipate (DEHA) as the most attractive ab-sorbent for this study. Chemical properties of pure DEHA, puretoluene and DEHAtoluene mixtures are reported in Table 1.DEHA was chosen because (1) absorption equilibrium betweengaseous toluene and liquid DEHA promotes toluene absorp-tion, (2) vapor pressure of DEHA is very low, (3) DEHA is lessviscous than other organic absorbents previously studied, (4)DEHA is currently not considered as toxic nor explosive. Theequilibrium between toluene and DEHA was found to followHenrys law:

P = Hx. (1)A previous investigation gave Henrys constants at different

temperatures. Finally, the constant can be calculated accordingto the equation

lnH = 26.925 5772T

. (2)In dilute phase, Henrys law can be written in order to involveconcentrations:

CGRT = H CLCL + DEHA/MDEHA

, (3)

CG HRT

MDEHADEHA

CL = H CL. (4)

This dimensionless Henrys constantH related to concentra-tions will be used in modelling. Absorption of gaseous toluene

2578 F. Heymes et al. / Chemical Engineering Science 62 (2007) 25762589

Table 1Selection of physicochemical properties of the system

Pure component properties Binary properties

Pure DEHA Pure toluene TolueneDEHA

Molar weight 350.8 92.14 Activity coefcient (gmol1) (25 C) 0.52Density (25 C) 910 864 Henrys law constant H(kgm3) (25 C) (kPa) 1.93Fusion 76 95 Diffusion coefcient DLtemperature (C) (25 C) (m2 s1) 8.65 1010

Vaporization 210 110 ToluenePDMStemperature (C)Viscosity (25 C) 11.2 Partition coefcient S(mPa s) (25 C) 1018Vapor pressure 7.48 106 3804 Diffusion coefcient DM(20 C) (Pa) (25 C) (m2 s1) 1.10 1010

Fig. 3. Pressure drops and liquid retention in the packed column in function of gas velocity and liquid ow rate.

into polydimethyl siloxane (PDMS) was determined by De Boet al. (2002). According to this author, the equilibrium betweenthe gas phase and the polymer phase can be described by thelaw CM = SCG with S given as the following:

log S = 0.58 + 0.91(

591T

)2. (5)

The equilibrium between liquid and gas phases can be de-duced from the previous data by considering a pseudogas phase.The equilibrium is represented by the equation

CM S HRT

MDEHADEHA

CL. (6)


2.2. Mass transfer in the absorber

The aim of this section is to study hydrodynamics and masstransfer behavior in a counter-current packed column. Thepacked column was chosen because of low pressure drop resis-tance and high interfacial area. Since DEHA has a low vaporpressure, DEHA stripping should occur at a low rate.

A rst point was to investigate hydrodynamic parameters(pressure drop, liquid retention) in counter-current operatingmode. A second point was to study the mass transfer efciencyand kinetics of toluene absorption by DEHA owing in a closedloop. The experimental setup is detailed in a previous paper(Heymes et al., 2006b). The experimental conditions were cho-sen as follows:

air temperature between 22 and 31 C, toluene concentrations in the gas: 500; 1000; 5000mgm3, liquid ow rates: 2.96, 4.28, 6.58, 9.87, 13.16 kgm2 s1, gas velocities: [0.55m s1].

These data were chosen according to industrial concerns (LeCloirec, 1998). The wet column pressure drop and liquid reten-tion were measured at constant liquid ow rate, increasing thegas ow incrementally. Any sudden increase in the gas ow per-turbs the system, which then gradually returns to a state of hy-drodynamic equilibrium in which liquid retention and pressuredrop are constant. All the results presented below were there-fore measured after stabilization of the system. The plots ofthe pressure drop and liquid retention curves, at constant liquidow, are shown in Fig. 3. Pressure drop [A] and liquid hold-up[B] are given in loglog scale. Flooding data were determinedand are reported in the chart [C]. Concerning liquid hold-up,the resulting curves correspond to the classical behavior as de-scribed in the literature. Concerning modelling pressure dropand liquid retention, Mackowiak (1990, 1991) and Billets for-mulae are sufciently precise and can be used in order to setup an industrial absorption column (Heymes et al., 2006b).

Fig. 4 shows the experimental results for toluene recoveryperformance with three different gas ow rates. The toluene re-covery performance is dened as the ratio of outlet toluene con-centration abatement on the inlet toluene concentration. Hun-dred percent means a complete absorption. It appears that theranges of ow rates and concentrations investigated give aninitial toluene removal performance close to 100%, whateverthe conguration. This conrms the interest of DEHA as anefcient absorbent for treating a gas loaded with toluene, withconcentrations varying between 500 and 5000mgm3. It isobvious to note that the removal performance decreases con-tinuously because of the absorbent enrichment with toluene,justifying the regeneration of the absorbent.

Mass transfer constants KLa and KGa are required to designan absorption tower. Values were calculated from experimentaldata and are reported in Table 2. Considering experimental data,it appears that KLa varies considerably depending on the hy-drodynamics. A difference of a factor by 10 is observed by com-paring KLa inside and below the loading zone (Heymes et al.,2006b). This conrms the interest to operate in the loading zone.

Fig. 4. Inuence of gas velocity on efciency curves in function of time(L = 6.58 kg s1 m2).

Table 2Experimental results of mass transfer in the packed column

Experiment [Toluene] L G KLa KGanumber (mgm3) (kgm2 s1) (kgm2 s1) (s1) (s1)

1 4990 13.16 0.51 2.52 104 2.752 1205 13.16 0.51 2.51 104 0.793 1193 9.87 0.51 3.69 104 1.894 5038 6.58 0.51 3.23 104 1.345 520 6.58 1.16 2.28 103 13.56 1005 6.58 1.21 2.38 103 13.87 1005 6.58 1.81 2.43 103 12.1

The experimental KLa values can be compared with datafrom the literature. We observe that with the same column,packing and ow rates, DEHA [KLa: 2.527 104 m s1]occupies an intermediate position between PEG 400 [KLa:1.45.5 104 m s1] and water [KLa: 7883 104 m s1](Heymes et al., 2006b). DEHA is thus more effective thanPEG 400 for absorbing toluene. This is not surprising if wecompare the Schmidt numbers usually used in mass transferstudies: Sc(water) = 14 103, Sc(DEHA) = 14 104 andSc(PEG400) = 8.7 105. The Schmidt number is a determin-ing factor to the mass transfer velocity in a liquid. We notea difference in the order of 10 between the Schmidt numbersof PEG 400 and DEHA, which reects the greater aptitude ofDEHA than PEG 400 for toluene absorption.

Correlations from two authors were tested: Onda et al. (1968)and Billet and Schultes (1993, 1999). The two authors ap-proaches are similar in terms of modelling. Nonetheless, thepredictions of these authors differ considerably with regard tothe true wetted interface area. This large difference in fact re-ects the developments in packing materials that took place be-tween 1968 and 1993. A high void fraction and improved shapesenable modern packing materials to provide much greater in-terface areas by means of a large number of droplets andtrickles. Billets correlation, established using more modernpacking materials, therefore seems more appropriate to the


packing used in this study. As expected, simply using the cor-relations of Onda et al. (1968) gives values very far from theexperimental results. Billets more recent correlation does givepredictions closer to reality, but still very far from experimen-tal values (overestimating KL by a factor of 1050), which aretherefore invalid for an industrial utilization.

Neither correlation from the literature is therefore satisfac-tory for modelling our system. There is currently a lack ofknowledge concerning mass transfer in a packed column withviscous uids. We shall therefore propose a correlation enablingthe prediction of our results with a view to the industrial scale-up of an absorption column. In the framework of this study,we propose an empirical correlation. The form followed by ourexperimental results is

ShL = KLaDL

= aawet

AScL ReBG Re

CL . (7)

A statistic tting of our data gives A = 1.67 108, B = 1.70and C =0.048. With these values, the mean relative absoluteerror is 13%.

To conclude this absorption section, the hydrodynamic studyshowed that the viscosity of DEHA was not a technical obsta-cle to its implementation in an industrial column. The pressuredrops caused remain acceptable. In case of mass transfer con-stants modelling, we established an empirical correlation accu-rate only in the investigated system. These two formulae wereemployed in Section 2.

2.3. Mass transfer in the pervaporation device

Pervaporation is a membrane-based separation process that isuseful for fractionating certain difcult-to-separate liquid mix-tures. The mechanism involved in the separation can obtainhigh separation selectivities, but with relatively low ow rates,which connes this technique to specic applications. A mem-brane process was chosen because thermal decomposition ofheavy absorbents by distillation makes this process unwise toemploy.

The chemical structures of toluene and DEHA are very dif-ferent. A molar weight difference and a sharp boiling tempera-ture difference (105.6 C) between toluene and DEHA can benoticed (Table 1). The vapor pressures of toluene and DEHAare signicantly different: toluene shall be easily extracted fromDEHA by pervaporation. The membrane was therefore chosento extract with a high ux toluene from the liquid phase. A liter-ature review showed that PDMS is very permeable for tolueneand is an interesting choice from an industrial point of viewsince it is a stable polymer, produced by a well-established lmmanufacture technique, and is already successfully used in in-dustrial applications.

PDMS swells considerably on contact with organic solutionsdue to the solubilization of the molecules in the polymer. Thisinduces several problems for an industrial use: the degrada-tion of the polymer selectivity and a possible loss of the mem-brane integrity consequent to a non-homogenous and importantswelling of the active layer of the membrane. A at commercial

Table 3Toluene ow measured during pervaporation of pure toluene (T = 48 C)Toluene concen-tration in feed

Liquid velocity(m s1)

Average pervaporation ow

(gL1) rate (kgm2 h1)

4.6 0.34 3.84.6 0.60 5.14.6 0.95 6.26.3 0.34 5.06.3 0.60 6.86.3 0.95 8.610.7 0.34 8.410.7 0.60 11.110.7 1.30 16.9

PDMS membrane (PERVAP 1060) was used in order to char-acterize the pervaporation of toluene from DEHA mixtures.

A previous study (Heymes et al., 2006c) aimed to examinethe feasibility to regenerate the used DEHA by pervaporationwith a PDMS membrane (PERVAP 1060). Toluene concen-trations into liquid were chosen in accordance of a real ab-sorbent solution concentration at the bottom of the absorptioncolumn. Theoretically, the absorption tower treating an exhaustgas (25 C) charged with toluene at 5 gm3 will provide a max-imum concentration of 17 gL1 (DEHA). Since the thermody-namic equilibrium will never be reached in an absorption tower,the liquid concentration range was chosen as [110 gL1] forthe experimental investigations.

The points of interest are the selectivity of the separationand the toluene ow rate extracted from the liquid mixture.Table 3 gives experimental toluene ow rates recovered in coldtraps at different concentrations and liquid velocities. All ex-periments were performed at 48 C. Results show that toluenecan be extracted from the tolueneDEHA mixtures even at lowconcentrations. Infrared spectrometry of the permeates revealedno DEHA in the liquid samples.

However, ow rates were very low (< 20 gm2 h1) andwill require a large membrane area to regenerate sufciently theabsorbent. This negative conclusion is an evident weakness ofthe hybrid process and it was necessary to identify the reasonof that low level of ow rate. Experiments led with differentmembrane thicknesses, different support layer and different liq-uid velocities proved that the main resistance to mass transferis located in the liquid boundary layer at the membrane surface(Heymes et al., 2006c). Hydrodynamics of the liquid upstreamof the membrane is the determining parameter of the separation.

The experiments of this piece of work were performed in asmall laboratory device (S=55 cm2), designed to test differentmembranes but not to characterize hydrodynamics in a generalmanner. Therefore, an empirical correlation based on the resultsof this piece of work would not be valid to design an industrialpervaporation unit. Correlations from literature will be used inthe modelling part.

3. Discussion

In this part, the experimental results and conclusions willbe used in order to study the hybrid process. The classical


Fig. 5. Evolution of toluene concentration into the liquid during the rst three loops.

equations governing the absorption are as follows:

The mass balance:= QG(CG,in CG,out) = QL(CL+ CL). (8)

The mass transfer equation:= QG(CG,in CG,out) = KabsL aVCMLL,abs, (9)

where

CMLL,abs =(CG,in/H CL+) (CG,out/H CL)

ln((CG,in/H CL+)/(CG,out/H CL)).

(10)

The equations governing the pervaporation are:

The mass balance:= QL(CL+ CL). (11)

The mass transfer kinetics:= QL(CL+ CL) = KpervL SCMLL,perv, (12)

where

CMLL,perv =CL+ CL

ln(CL+/CL). (13)

Considering all these equations, it appears that the liquid owrate QL, the liquid concentrations out from the absorber CL+and out from the membrane device CL and the temperature(included in Henrys law constant) are coupling values.

The discussion part is divided into three parts: analysis ofthese coupling values on each side of the process, theoreticaldirect coupling as represented in Fig. 1 (the outlet of one sidefeeds directly the inlet of the other one) and nally pseudocou-pling in the case that one at least of the three coupling valuesis not directly coupled between both parts of the process.

3.1. Analysis of coupling values

3.1.1. Liquid concentrationOut of the three coupling values, the pollutant concentration

of the liquid is the only one which is not a directly adjustableparameter. It is the physical result of all the transfers takingplace in the hybrid system. It is not constant throughout itscirculation loop: it is at its maximum at the column outlet (CL+)and minimum at the outlet of the pervaporation module (CL).

In the ideal case of an innitely high column and innitemembrane surface area, we would have both thermodynamicequilibriums between the liquid phases at the column outlet andthe gas phases at the column inlet, and a zero concentration oftoluene in the liquid at the pervaporation module outlet:

CL+ RT

H

DEHAMDEHA

CG,E and CL = 0.

However, the concentration CL+ is lower than this equilib-rium value and CL is not zero. The values of this concentra-tion depend on the efciency and scaling of the apparatus used.Solving previous equations independently and successively al-lows drawing the concentration into the liquid loop (Fig. 5).

3.1.2. Inuence of temperatureThe temperature is a coupling value acting differently on

each side of the process. From the hydrodynamics point ofview, an increase of temperature will decrease the viscosity ofthe absorbent and increase the Reynolds number of the liquid.The boundary layer lm will be thinner and the mass transferwill be increased. However, temperature will modify the ther-modynamic equilibriums. Table 4 summarizes the inuence oftemperature on the Reynolds number, Henrys law constant H and the air-membrane solubility S in case of a ow in a tube(5m s1, 1mm). As expected, the Reynolds number increasesrapidly with the temperature. Operating at 60 C instead of20 C will increase the Reynolds number by a factor of four


but will decrease the solubility in the liquid by a factor of 10and by a factor of seven in the membrane.

Because of these antagonist effects, it is not obvious to modelthe inuence of the temperature. On the absorption process,the temperature increases the mass transfer mechanisms butdecreases strongly the absorption capacity of the liquid. Onthe membrane process, a temperature increase will increase themass transfer in the liquid lm and through the polymer butwill decrease the absorption capacity of the polymer.

Experimentally, results show that rising the temperature de-creases the overall efciency of absorption but increases the ef-ciency of pervaporation. The usual law proposed for the latterone is

J = J 0 exp[E

R

(1T

1T 0

)](14)

with E as the activation energy found to be equal to23.024 kJmol1 in case of DEHA mixture enriched at 9 gL1of toluene (Heymes et al., 2006c). Fig. 6 shows the re-sults calculated about an industrial case (10 000m3 h1,toluene concentration = 5 gm3). The absorption column waschosen as supplied with pure absorbent and the pervaporationunit was chosen as supplied with a 10 gL1 tolueneDEHAmixture. This gure shows that increasing the temperature upto 60 C will divide the absorption rate by a factor of six andwill multiply the regeneration rate by a factor of three. Increas-ing the process efciency by modifying the temperature will in

Table 4Inuence of temperature on Reynolds number and thermodynamic equilibri-ums

Temperature (C) 10 20 30 40 50 60

Reynolds number 217 334 484 667 880 1118Henrys constant H (106) 118 231 431 772 1334 2231Air-membrane solubility constant S 2500 1352 777 470 298 197

Fig. 6. Inuence of temperature on the absorption and the pervaporation processes.

all cases promote one side and harm the other side. A besttemperature, ideal for the overall process, should be determined.

Fig. 6 cannot be used to calculate that ideal temperature sinceboth processes are coupled. In the previous considerations, theliquid was considered as pure at the inlet of the absorber and en-riched with 10 gL1 at the inlet of the membrane device. Thiswould not be the case in the coupled process and complicatesthe understanding of the unit. Fig. 6 shows that increasing thetemperature will decrease the absorption efciency, and con-sequently the liquid enrichment at the bottom of the absorber.For example, raising the temperature from 20 to 60 C willlower the bottom concentration from 1.4 to 0.22 gL1. Conse-quently, the regeneration will be harder, and nally increasingthe temperature will also reduce the pervaporation efciency.

3.1.3. Inuence of liquid ow rateThe liquid ow rate has a strong effect on mass transfer due to

boundary layers effects and diffusion resistances. Moreover, inthe packed column, the liquid ow rate has to be high enough toabsorb sufciently the toluene. This point can be checked by theabsorption rate dened asR=L/GH . The higher theR will be,the better the absorption will be. But the ow rates cannot varywidely in a packed column since the hydrodynamics has to belocated in the loading zone conditions. It is always desirable tobe situated in a domain close to 60% of ooding, since this zonecorresponds to the best contact of gas and liquid (maximum in-terface area). Decreasing the liquid ow rate will lead below theloading zone with poor mass transfer efciency, rising up theliquid ow rate can lead to ooding. In pervaporation, increas-ing the liquid ow rate will promote turbulence and increasemass transfer. Two equations were used to check this point:

Mass transfer in the packed column (Eq. (7)):

Sh = KabsL

aDL= 1.67 108 a

awetScL Re

1.7G Re

0.048L .


Fig. 7. Inuence of temperature on the absorption and the pervaporation processes.

Mass transfer in a tubular membrane unit (Nel, 1997):

Sh = KpervL d

D= 1.85

[Re Sc

(d

L

)]1/3. (15)

Results are illustrated in Fig. 7. The black symbols representthe VOC ux absorbed in the column or extracted in the mem-brane device. Both uxes increase with the liquid ow rate,what is due to higher mass transfer.

However, it should be observed that an increase in ow rateleads to a lower enrichment or regeneration of the liquid. Thisis shown with the white symbols in Fig. 7. At a liquid owrate of 4.6m3 h1, the liquid is enriched with 13 gL1 in theabsorber and loses 1.9 gL1 of toluene in the membrane de-vice. At a higher ow rate of 30m3 h1, the liquid is enrichedonly with 1 gL1 in the absorber and loses only 0.6 gL1 of

toluene in the membrane device. This can be understood as aloss of transferred matter due to a lower time spent in the device.In case of the coupled process, this could play a limiting roleon the overall efciency.

In conclusion, the study of the coupling values shows thatthe hybrid process depends on both temperature and liquidow, whose effects can have opposite consequences on theperformance of each operation. The aim of the following sectionis to model the system in order to determine that optimum.

3.2. Formulation of equations for the coupled process

Considering the case of a continuous process such as thatshown in Fig. 1, the hybrid system can be modelled from basisequations (8)(13). The problem therefore consists in resolvingthe following system of equations:

KabsL aV((CG,in/H ) CL+) ((CG,out/H ) CL)

ln(((CG,in/H ) CL+)/((CG,out/H ) CL))= QG(CG,in CG,out)

QL(CL+ CL) = QG(CG,in CG,out)

ln(CL+CL

)= K

ML S

QL

. (16)

This system can be modied as the following:

KabsL aV

H (1 H CL+/CG,in) (CG,out/CG,in)(1 H CL/CG,out)

ln((CG,in/CG,out)(1 H CL+/CG,in)/(1 H CL/CG,out))= QG

(1 CG,out

CG,in

)

QL

H

(H CL+CG,in

CG,outCG,in

H CLCG,out

)= QG

(1 CG,out

CG,in

)

ln(

H CL+/CG,inH CL/CG,out

CG,inCG,out

)= K

ML S

QL

. (17)


At this point we introduce the dimensionless values X+ andX:

X+ = HCL+

CG,inand X = H

CLCG,out

.

X+ and X represent the ratio of the hypothetic gas con-centration in equilibrium with the liquid with the real gas con-centration. In all cases X+ and X would be as 0


Fig. 8. Resolution of the system of equations: A, X+ and X vs P .

X+ and P , the resolution of the system gives the followingequalities:

X = RSK R + 1 SRS(K 1)

X+ = RS exp(K) R + K(1 S)R(K 1)

ln(P ) = 1 + [K 1](1 S)R(SK 1) + 1 S

. (22)

To determine X, X+ and A, the resolution of the system givesthe following equalities:

X = (1 S)RS[exp(P ) 1]

X+ = (1 S) exp(P )R[exp(P ) 1]

A = R 1R ln[(exp(P )[R 1 + S] R)/(RS exp(P ) RS 1 + S)]

. (23)

In the other two cases, the analytical resolution is complexand the solutions are given by a numerical resolution.

The system was solved as the example in Fig. 8. The tolueneremoval efciency was taken as E = 0.9, meaning that 90%of the initial amount of toluene will be removed from the ex-haust air. An average absorption rate was chosen to be R = 2.Fig. 8 shows on the rst vertical axis the A value necessary tofulll an abatement S as a function of P . This means that therst curve shows the necessary KLaV value to treat 90% of apolluted air stream as a function of the pervaporation deviceefciency. This curve shows that for a decient pervaporationunit (P>1), a large column is needed (A?1). On the otherhand, if the column is not very efcient (A>1), a large per-vaporation device will be needed (P?1) to fulll the require-ments. These remarks can be understood while considering X+and X values in the gure. In case of a very efcient absorp-tion device (A?1), X+ is very close to unity (absorbent quite

completely saturated). The pervaporation conditions are verygood and membrane device can be quite small. In the case of avery efcient pervaporation device (P?1), the liquid is quitecompletely regenerated (X close to zero). A small absorptiontower is required to absorb toluene with quite pure absorbent.

3.3. Uncoupling the process

The above points have shown that the coupling values canhave opposite inuences on the efciency of either operationin the process. It may therefore be interesting to uncouple

these values in an industrial process.Temperature can be uncoupled by utilizing heat exchangers,

one to cool the uid before the column, the other to reheat itbefore the pervaporation module (Fig. 9). In addition, a heatrecovery system could be used to optimize the global yield. InFig. 10, we observe that the ux of toluene transferred will bemaximized for each step. While the gain in terms of facilitysize is obvious, the cost of additional heat exchangers shouldnonetheless be taken into account, together with their operatingcosts.

The liquid ow can be uncoupled by using two additionalpumps and a container to act as buffer volume (Fig. 11). By thisway, the liquid ow in the column could be adjusted in func-tion of the hydrodynamic conditions in the packing, whereasthe liquid ow in the pervaporation module could remain con-stant. Note that the pump at the foot of the column could beeliminated if conditions enabled gravitational ow. In addition,


Fig. 9. Diagram of the principle of the dithermal process.

Fig. 10. Flow sheet of dithermal operation of the process.

the utilization of the buffer volume would make it possible toconsider an interesting sequential operating system for the hy-brid process. During stoppage of emission of gaseous efuentsto be treated, the absorption column could be kept on stand-bywhile the pervaporation module could continue the regenera-tion process, in order to purify more washing liquid or regener-ate it at times of lower operating costs (for example, at night).Direct uncoupling of the concentrations would occur during un-coupling of the liquid ow rates. Nonetheless, this uncouplingwould not be favorable for the user. The dilution effect of thebuffer volume on the overall operation would be negative. Asimple mass balance equation shows that the concentration of

Fig. 11. Diagram of the principle of the process using a buffer container.

the buffer volume would be both lower than that at the foot ofthe column and higher than that of the liquid regenerated bythe pervaporation module:

Q1 CL+ + Q3 CL = Q3CM + Q2CM +dV CM

dtQ1 + Q3 = Q2 + Q3 + dVdt

.

(24)In the case of a permanent regime, we obtain{CM = Q1

Q2 + Q3CL+ +Q3

Q2 + Q3CLQ1 = Q2

}, (25)

and thus,

CM = Q2Q2 + Q3CL+ +

Q3Q2 + Q3CL . (26)

Since all the terms in this equation are positive, CLCMCL+ . The two facilities would therefore operate under lessfavorable conditions for mass transfer.

3.4. Application: scale-up of an industrial plant

The case of a continuous emission of toluene-loaded air willbe used as the basis for the scale-up. It could, for example, be avarnishing workshop. The efuent of air to be treated was cho-sen to present the following specicities: gaseous ow rate =10 000m3 h1, toluene concentration at factory outlet 5 gm3and temperature 25 C. The constraints of the specicationsare as follows: (i) in compliance with the decree regulatinggaseous emissions, which xes emission limits at 110mgm3of total organic carbon, the maximum acceptable toluene con-centration in the gas ow is xed at 120mgm3. (ii) Since thevarnishing workshop does not possess a heating utility, the pro-cess is chosen as mono-thermal. Varnishing being continuous,the utilization of a buffer tank is considered superuous and


the simplest version of the process is chosen. (iii) The absorp-tion column is a packed column lled with Hiow rings. (iv)The pervaporation process is carried out in a 1mbar vacuum.Based on the results of this work, it was decided to use a planarpervaporation module tted with PDMS membranes. This is adistinctly optimistic hypothesis, since industrial plants operateat downstream pressures of 100mbar. Nonetheless, this down-stream pressure of 1mbar will be retained so as to be able touse the results of this study.

The quantity of toluene to be recovered per hour is48 800 g h1. In order to determine the diameter of the column,we choose an arbitrary liquid ow rate R = 5, which is a me-dian value. For 60% ooding, the calculated column diameteris 130 cm. The empty column gas speed would therefore be2.09m s1, and the surface liquid ow from 3.03103 m s1.With these values, the KLa in the column is 7.6 103 s1.

The value of the parameter S is xed at 0.0024 (CG,in =5 gm3 and CG,out = 0.12 gm3). R is chosen as parameter, Ais the variable and X, X+ and P are the solutions determinedusing the equations given above. The results are presented inFig. 12. A good deal of information can be concluded from thatgure:

An increase in membrane surface area enables the columnheight to be reduced (and vice versa).

For a given liquid ow rate, there is a membrane surfacearea above which any additional increase in surface area issuperuous. The process is limited by absorption.

For a given liquid ow rate, there is a column height areaabove which any additional increase in column height issuperuous. The process is limited by pervaporation.

For a given membrane surface area, the column height re-quired for purication of the polluted efuents increases ifthe liquid ow rate increases.

For a given column height, the membrane surface area re-quired for regeneration increases if the liquid ow rate in-creases.

These observations are easily explained by the fact that ifregeneration is very effective (high membrane surface area),then the liquid is highly depleted and the column operates un-der more favorable conditions, hence the lower column heightrequired. The inverse can be explained by the fact that a highcolumn leads to an enrichment of the liquid, which is then eas-ier to regenerate.

Considering Fig. 12, it appears that the most economic plantto treat 10 000m3 h1 of toluene enriched at 5 gm3 shouldbe a 3.6m high column (diameter 1.3m) coupled with a 20m2membrane unit.

Certain hypotheses have been made that should be under-lined. In the rst place, it was assumed that the extrapolationsof the mass transfer equations in the column and pervaporationmodule remain valid on the industrial scale. Next, it was as-sumed that the hydrodynamic behavior in the column (scaledfor an absorption rate R = 5) would not change when the ab-sorption rate varied in the range [130], whereas ooding islikely to occur with R values so far from the initial value. Fi-

nally, the hypothesis of an industrial vacuum of 1mbar is notrealistic and a value of 100mbar should be considered. A lossof permeate ow is to be expected. Nonetheless, the interest ofthis modelling study was to show the feasibility of the hybridprocess. The feasible column and membrane sizes calculatedin this section conrm that the hybrid process is industriallyviable. Furthermore, it has been shown that there are innitepossibilities for dealing with a given problem: column size,membrane surface area, several compromises are possible. Tochoose the best compromise, an economic analysis is required.That is beyond the scope of this work due to lack of infor-mation. The remainder of this paper nonetheless positions theprocess with respect to other processes, without including anyeconomic analysis.

3.5. Positioning of the hybrid process: comparison with otherair treatment processes

The strengths and weaknesses of the process can be high-lighted. The strengths are with respect to the limitations of otherair treatment processes:

exibility of utilization, possibility of sequential or continu-ous operation,

no by-products generated apart from pervaporates, relatively low temperature, therefore low risk of explosion

or re, possibility of operating under humid conditions, possible reuse of pervaporates.

Nonetheless, certain weaknesses should be presented:

Unknown service-life of membranes, which may consider-ably increase the cost of the plant.

The need to study the behavior of the membrane with eachnew pollutant and take into account the coupling effects.

If dust or solid matter is present in the gas to be treated, thereis risk of accumulation in the circulation loop. However, thisrisk can be avoided by upstream dedusting.

It should also be noted that distillation under vacuum is apossible alternative technological solution instead of pervapo-ration for regenerating the absorbent. If we assume that the liq-uid and gas phases are ideal, the equilibrium curves at constantpressure can be traced from the saturated vapor pressures. Thepacked column absorption experiments showed that the con-centrations at the foot of the column do not exceed 20 gL1,i.e., a molar fraction of toluene of 0.083. The minimum numberof theoretical plateaux can be calculated by means of Fenskesequation. If we assume a toluene molar fraction at the head ofthe column greater or equal to 0.999 and an absorbent puri-cation rate below or equal to 0.001 (0.23 gL1), then Fenskesformula gives Nmin = 2.42.

Regeneration of the DEHA by distillation would thereforebe entirely feasible, provided that the DEHA is thermally sta-ble. Preliminary heating experiments in which the DEHA was


Fig. 12. Relation between column height and membrane surface area.

maintained at 100 C for 30 days showed that the refraction in-dex and viscosity of the DEHA were unchanged. Nonetheless,further studies at higher temperature or lower pressure are in-dispensable in order to decide whether distillation should beused for the regeneration process. The comparison between dis-tillation and pervaporation requires a technical-economic anal-ysis which is beyond the scope of this paper for reason similarto those mentioned above. Nonetheless, an essential point inthe comparison is the operating mode of the plant for reasonsof thermal inertia; distillation is more cost-effective in contin-uous operation, whereas pervaporation is potentially suited tocontinuous or sequential regimes.

4. Conclusion

VOCs are chemical species involved in the problems of at-mospheric pollution. They are currently the target of regulationintended to reduce their emission levels. This measure musttherefore be accompanied by the development of processesable to satisfy the new requirement in terms of emissions. Thisstudy falls within the scope of that general context. The aim ofthis research was to develop a treatment process for air loadedwith hydrophobic compounds by coupling absorption and insitu membrane-based regeneration. The approach can be splitinto several steps. The rst consisted in nding an absorbentlikely to be suitable for an industrial application: DEHA. Thesecond concerned the characterization and validation of the hy-drophobic absorption by means of a pilot-scale apparatus. Thenext was devoted to studying tolueneabsorbent separation us-ing a dense membrane. Finally, the work was completed witha study of the coupling of the absorption step and regenerationby pervaporation.

DEHA was used in a pilot-scale packed absorption column.High efciencies were observed when the column was fed withpure liquid. These efciencies fell sharply when the absorbent

became loaded with toluene, hence the need to regenerate theabsorbent. The experiments showed that resistance to masstransfer was situated on the liquid side and also enabled calcu-lation of the liquid-side mass transfer constant. It was shownthat the correlations proposed in the literature do not correctlypredict the experimental results, probably because of the higherviscosity of DEHA compared to the liquids used by the authors.

Organic separation by pervaporation is a subject of currentimportance that has given rise to many studies. Authors proposevarious selective membranes made of complex materials, butPDMS was nally chosen as the membrane material for sepa-rating toluene from DEHA. The inuence of various operatingparameters (temperature, downstream pressure, ow rate) wasinvestigated, as were models from the literature. It was shownthat the limiting phenomenon in the process is boundary layerresistance, much higher than resistance inside the membranematerial. The hydrodynamics of the upstream part of the mod-ule is thus the key to the regeneration process.

The hybrid process was coupled in function of three val-ues: ow rate, temperature and concentration. The concentra-tion of toluene in the absorbent has a determining effect onthe efciency of the overall process. However, it is not a di-rectly adjustable parameter, unlike temperature and ow rate.Temperature has opposite effects on the efciency of the twosteps in the process, which means either nding an optimumcompromise or dithermal operation (different temperatures forthe absorption column and pervaporation module). Importantinformation was provided by formulating the equations of thesystem and using the correlations determined in the previoussections. It was shown that there were multiple possible mem-brane surface area/column height compromises for treating agiven gaseous efuent. Any increase in membrane surface areacan be compensated by a decrease in column height (operatingat low global toluene concentrations in the liquid phase). In-versely, a high column enables a lower membrane surface area


to be used (operating at high global toluene concentrations inthe liquid phase). Calculating the initial investment and oper-ating costs will enable the user to choose the best compromisebetween column height and membrane surface area.

The eld of this study is vast and research has been deliber-ately restricted. For the absorption part, the problems of VOCabsorption by viscous uids have not been much studied in theliterature. The existing correlations did not prove sufcientlyaccurate for predicting mass transfer constants, so additionalwork in this area would be useful. Similarly, the hydrodynam-ics of the boundary layer upstream of the membrane requiresfurther investigation in order to determine the diffusion behav-ior and validate the appropriate correlations.

These two limitations are naturally also seen in the extrap-olation of results obtained in the laboratory to the scale-up ofan industrial unit. While it is always hard to extrapolate resultsobtained on a small scale to an industrial application, this ismade even more difcult by the fact that the hydrodynamics ofa laboratory pervaporation module is very different from thoseof an industrial facility. Appropriate studies are therefore in-dispensable in order to specify these factors. The work endedwith an analysis of the coupling parameters and did not attemptan economic assessment of investment and operating costs. Weconsider that the scale-up results have not been sufciently val-idated for costs to be estimated that would situate the projectfrom an economic point of view.

Acknowledgment

This work was accomplished with the nancial support ofAgence de lEnvironnement et de la Maitrise de lEnergie,(ADEME) France and the Europe Environnement Company(Vieux Thann, France).

References

Billet, R., Schultes, M., 1993. Predicting mass transfer in packed columns.Chemical Engineering Technology, 161169.

Billet, R., Schultes, M., 1999. Prediction of mass transfer columns withdumped and arranged packings. Transactions of the IChemE 77, 498504.

De Bo, I., et al., 2002. Application of vapour phase calibration methodfor determination of sorption of gases and VOC in polydimethylsiloxanemembranes. Journal of Membrane Science 209, 3952.

EEC, 1999, 2001. European Community Parliament, 1999/13/CE Directive,March 11th, 1999. 2001/81/CE Directive, October 23rd, 2001. (modiedin 2001).

EPA, 1998. National Air Quality and Emissions Trends Report, USA.Heymes, F., et al., 2006a. A new efcient absorption liquid to treat

hydrophobic VOC polluted air. Chemical Engineering Journal 115,225231.

Heymes, F., et al., 2006b. Hydrodynamics and mass transfer in a packedcolumn: case of toluene absorption with a viscous absorbent. ChemicalEngineering Science 61, 50945106.

Heymes, F., et al., 2006c. Recovery of toluene from temperature boilingabsorbents by pervaporation. Journal of Membrane Science 284, 145154.

Le Cloirec, P., 1998. Les composs organiques volatils (COV) danslenvironnement. Lavoisier, Paris, France.

Mackowiak, J., 1990. Determination of ooding gas velocity and liquidhold-up at ooding in packed columns for gas/liquid systems. ChemicalEngineering Technology 13, 184196.

Mackowiak, J., 1991. Pressure drop in irrigated packed columns. ChemicalEngineering Program 29, 93105.

Nel, J., 1997. Pervaporation. Lavoisier, Paris, France. (ISBN: 2-7430-0213-1).

Onda, U., et al., 1968. Mass transfer coefcients between gas and liquidphases in packed columns. Journal of Chemical Engineering of Japan1, 56.

Treatment of gas containing hydrophobic VOCs by a hybrid absorption--pervaporation process: The case of tolueneIntroductionExperimental resultsThermodynamic equilibriumsMass transfer in the absorberMass transfer in the pervaporation device

DiscussionAnalysis of coupling valuesLiquid concentrationInfluence of temperatureInfluence of liquid flow rate

Formulation of equations for the coupled processUncoupling the processApplication: scale-up of an industrial plantPositioning of the hybrid process: comparison with other air treatment processes

ConclusionAcknowledgmentReferences

Frederic Heymes; Treatment of Toluene by Absorptio_2007

Documents

Transcript of Frederic Heymes; Treatment of Toluene by Absorptio_2007