Author’s Accepted Manuscript

Potentials of pervaporation to assist VOCs recovery

by liquid absorption

Denis Roizard, François Lapicque, Eric Favre, Christine

Roizard

PII: S0009-2509(09)00026-8
DOI: doi:10.1016/j.ces.2009.01.014
Reference: CES 8364
www.elsevier.com/locate/ces

To appear in: Chemical Engineering Science

Received date: 23 September 2008

Revised date: 5 January 2009
Accepted date: 8 January 2009

Cite this article as: Denis Roizard, François Lapicque, Eric Favre and Christine Roizard,
Potentials of pervaporation to assist VOCs recovery by liquid absorption, Chemical Engi-
neering Science (2009), doi:10.1016/j.ces.2009.01.014

This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting galley proof
before it is published in its final citable form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply
to the journal pertain.
 Potentials of pervaporation to assist VOCs recovery by liquid absorption

Denis Roizard*, François Lapicque, Eric Favre and Christine Roizard

Laboratoire des Sciences du Génie Chimique, CNRS 6811

ENSIC, Nancy-University, 1 rue Grandville, BP 20451, F-54001 Nancy, France

Abstract

Gas treatment by liquid absorption is a well-known process to remove volatile organic

p t
r i
compounds (VOCs) from industrial waste gases. Usually the liquid is an organic solvent of

s c
high boiling point; however after VOCs absorption it must be regenerated for the possible

u
reuse and this step is classically achieved by heating the liquid. The paper presents the work

n
a
directed to investigate an alternative regeneration step based on a liquid-vapour membrane

separation, i.e. pervaporation. Because most of the energy required in pervaporation processes

m
is consumed to remove the minor component from the initial mixture by selective permeation

e d
through the membrane, one can expect a significant energy cut in the operational costs linked

p t
to the regeneration of the liquid if the pervaporation step can substitute the heating one. The

e
results reported here show that the technological possibility to use pervaporation is first

c
c
governed by the stability of the membrane in the absorption liquid. The viability of the overall

A
process is actually controlled by the mutual affinity between the VOCs, the solvent phase and

the polymeric material. As a matter of fact, whereas VOCs have to exhibit strong affinities to

both the solvent and the membrane material, the polymer has to be well resistant and even

repellent to the solvent to avoid the possible sorption in the membrane that would drastically

depress the pervaporation efficiency. In other words the membrane transport properties must

be specific for the VOCs. This goal was reached following several experimental approaches,

*
Dr. D. Roizard +33 0 383 175 217 Denis.Roizard@ensic.inpl-nancy.fr

Page 1
going from membrane modifications to the selection of suitable heavy protic solvents. Hence

it has been shown for the case of dichloromethane (DCM) that low molecular weights

polyalcohols (e.g. glycols) appeared to be suitable media to allow in particular the specific

transport of DCM. On the other hand, polydimethylsiloxane based membranes (PDMS) were

selected for their stability in these polyglycols and for their marked affinity for DCM. The

simulation of the hybrid gas treatment process at pilot-scale was also achieved by a simple

model relying on experimental data for both vapour liquid equilibria and permeation flux. A

t
simple comparison of the energy needed to regenerate the heavy solvent by each possible step

p
has also been made.

r i
c
Keywords: Absorption, Separation, Membrane, Environment, Gas treatment, Pervaporation

s
Introduction

n u
a
Emissions of volatile organic compounds (VOCs) are a significant drawback of industrial

m
activity. Since their intensive use from the 1950’s, they have often been neglected in waste

d
treatment because treatment of diluted species represent a tedious, costly task, and with little

e
t
added value for the case of most VOCs. Under the pressure of political incentives and

p
e
incoming stringent regulations, chemical industry must now face environmental issues, and

c c
there is a need for study and development of efficient technology to recover VOCs from

polluted air with acceptable costs.

A
Various regenerative processes can be considered for VOCs recovery (Khan and Ghoshal,

2000). Among them, cryogenic techniques and direct membrane processes are restricted to

concentrated waste gases (Baker and Wijmans, 1994; Néel, 1997). Absorption of VOCs in

selective heavy solvents has been investigated as an alternative technique allowing both the

cleaning of waste air and the possible recovery of the chemicals (Schmidt et al., 1990;

Weisweiller et al. 1995; Hadjoudj et al., 2004), as it has been suggested for the regenerative

Page 2
absorption of other compounds such as sulphur dioxide (Sciamanna et al., 1988; van Dam et

al., 1997; de Kermadec et al., 2002). Selection of suitable solvents and achievement of the

gas/liquid (G/L) absorption has been extensively investigated, but the management of the

solvent phase recovered at the bottom of the G/L contactor still requires further development.

Due to poorly active interactions between most VOCs and the usual solvents considered e.g.

mineral oils, polyethers, esters, etc., the organic phases recovered are moderately

concentrated, which increases the operating costs of the overall treatment process. In

t
particular, this process supposes at least the use of 2 columns in parallel, one for the

p
r i
absorption step and the other for the regeneration step if a continuous treatment is wanted;

c
moreover, a thermal regeneration implies a lot of energy to heat the whole liquid phase and to

s
u
get the VOC vaporized. Amongst the various techniques available, vacuum operation was

n
developed some years ago up to pilot scale by CTP in Italy (Pirucci et al. 2003) and Astra in

a
Sweden (Uddelholm et al. 1987). Distillation has recently been suggested for the separation of

m
compounds of very different boiling points. Besides stripping of the organic liquid by vapour

d
or nitrogen allows the production of concentrated gaseous streams to be treated afterwards

e
t
(Wang et al., 2001; Poddar et al., 1996, 1997; Xia et al. 1999, Majumdar et al., 2001; Roizard

p
e
et al., 2004), in particular by membrane techniques. Reduced pressure favours the efficiency

c c
of the stripping operation. However stripping is little selective and low impurities contained in

the heavy solvent may also be entrained by the carrier gas.

A
It is known that VOCs can be removed from a liquid mixture by pervaporation as reported in

the literature (Peng et al. 2003), but one should note that all of these studies involved water-

containing mixtures. In fact very low attention has been paid to the removal of VOCs from

organic liquid mixtures especially with liquids of low volatility as it supposed to be the case

in the present hybrid process.

Page 3
Using pervaporation is an appealing strategy since it offers the possibility to achieve the

continuous regeneration of the liquid solvent by a membrane-assisted pressure decrease. A

pressure change remains indeed the preferred regeneration mode for physical solvents.

Moreover, the fact that pressure reduction is exerted to the VOC only (due to the postulated

membrane selectivity), is to offer promising performance of the process in terms of energy

consumption required for regeneration. In fact, the liquid solvent would not have, in that case,

to be recompressed when it is recycled back to the absorption column.

t
To our knowledge, only three teams have yet considered the pervaporation strategy, i.e.

p
r i
Philippe, 1998, Heymes et al., 2007 and us (through the work, initiated in 2005). Philippe and

c
Heymes have both pointed out the fact that the solvent, an ester in both cases, was observed to

s
u
be accumulated significantly in the membrane. As a result, the first team has discontinued its

a n
investigation whereas the second one has evaluated the hybrid process without taking into

account the real meaning of the solvent permeation. Indeed, as soon as the solvent

m
accumulates in the membrane and thus into the membrane support, the permeability of the

e d
membrane is irreversibly and drastically decreased leading to the shut-down of the operation.

p t
The present study intends to explore to what extent this major drawback could be

e
circumvented. It aims at designing a hybrid process suitable for long term treatment of VOC-

c
c
containing gases consisting of absorption into a heavy inert solvent and the subsequent

A
regeneration of the solvent by selective VOC permeation through a thin, dense membrane.

The case of two chlorinated VOCs namely dichloromethane (DCM) and tetrachlorethylene

(PCE) was considered here. After presentation of the methodology for designing such a

hybrid process, the solubility values of the considered VOCs in various solvents are given.

Then the compatibility between several liquid solvents and membranes was considered and

the VOC permeation selectivity through a PDMS-based membrane was investigated, which at

Page 4
last allowed the hybrid process to be simulated for the case of DCM-containing air. Finally

the energy demand needed for the solvent regeneration was evaluated for each method.

Methodology of the investigation

The process combines the absorption of gas-phase VOC into the heavy solvent and

pervaporation of VOC through the chosen polymeric membranes, the two operations being

t
achieved at temperature close to the ambient. Two separation units have to be designed, and

p
the chemical engineering techniques required for this purpose have been developed for years

(Dankwerts, 1970; Baker, 2004).

r i
Importance of the chemical nature of the various species
s c
n u
In spite of the importance of the process design, we preferred in the present study to

a
emphasise the physico-chemical aspects of the process. Three families of chemicals are

m
involved in the process: the VOC to be removed and concentrated, the solvent, and the

d
polymer forming the membrane. VOCs to be treated can be of various chemical natures from

t e
hydrophobic molecules e.g. hydrocarbons or chlorinated compounds to more polar

p
oxygenated molecules. The rules for selection of the absorption solvent have been given in a

e
c
number of papers, with particular mention of very low vapour pressure, non-toxicity,

A c
chemical stability and high capacity for the VOC (Hadjoudj et al.; 2004, Van Dam et al,

1997). This later point is the consequence of positive interactions between the VOC and the

solvent.

The presence of the membrane has rarely been considered in the selection of the chemicals.

As shown in Figure 1, VOCs and solvents have to exhibit very different interactions for

selective separation of the VOC from the solvent phase by pervaporation. As a matter of fact,

VOCs have to possess a strong affinity for the polymer membrane, which has in turn to be

repellent to the solvent.

Page 5
 Methods to be employed

The suitable solvents for the two VOCs considered were searched following the above

criteria, in particular with concern to the high solubility. The permeation flux of VOCs

through the membrane has to be measured, depending on their concentration in the solvent.

Also, the physical and chemical stability of the polymer towards the absorption solvents has

to be evaluated. Previous works related to the removal of chlorinated compounds from waste

p t
gases had led to the selection of ethers such as tetraethylene glycol dimethyl ether

r i
(TEGDME), and two heavy esters di(2-ethylhexyl) phthalate (DEHP) and di(2-ethylhexyl)

s c
adipate (DEHA) (Hadjoudj et al., (2004)). Due to the low-polar nature of the VOCs

u
investigated, polydimethylsiloxane (PDMS) was tested as membrane material. However, after

a n
clear evidence that the known heavy solvents could permeate through the membrane, further

work had to be done for identification of more suitable solvents and membrane polymers, to

fulfil the conditions shown in Figure 1.

m
e d
Chemicals products
p t
c e
All chemicals were of analytical grade. DCM and PCE were purchased from Merck, France.

A c
Their physical properties are reported in Table 1. DEHA and DEHP, respectively aliphatic

and aromatic esters, were obtained from Acros Organics, France. Genosorb® solvents, i.e.

monomethyl ether solvents, were Clariant products, France. Polyethylene glycol 400 (PEG

400), tetraethylene glycol (TEG), polypropylene glycol 425 (PPG 425) and polycaprolactone

triol 300 (PCLT 300) were purchased also from Acros Organics. Physical properties of the

various solvents are summarised in Table 2.

Page 6
Pure polydimethylsiloxane (PDMS) films were prepared using the standard hydrosylation

cross-linking procedure using low molecular weight silane and siloxane vinyl precursors,

RTVA® and RTVB®, both from Rhodia, France.

Membrane samples used for pervaporation experiments were commercial Silastic® PDMS

flat sheets (thickness 1.5 10-4 m), from Dow Corning, France.

t
Determination of the vapour-liquid equilibria

i p
The solubility of VOC contained in an air stream in the selected solvent was determined by

r
measuring the partial pressure of VOC over a binary VOC-solvent liquid mixture, using the

s c
headspace gas chromatography technique (Hadjoudj et al, 2004). The chamber containing the

u
glass vial for establishment of VLE could not be regulated below 30°C and measurements

n
a
were carried out at this temperature.

m
Henry’s constant, H, was deduced from the slope of the variation of VOC pressure with the

d
mole fraction in the liquid phase. The activity coefficient at infinite dilution of the VOC in the

t e
solvent was calculated from H and the vapour pressure of the VOC at the considered

p
temperature, saturated pressure (Psat):

e
γ∞ =
H

c
P sat
c (1)

A
Table 3 shows that most of the solvents tested exhibit positive affinity to DCM since its

activity coefficient in these solvents is noticeably below 1. PEG 400 seems to be the most

promising solvent with respect to its capacity for DCM. Activity coefficient of PCE is closer

to unity as shown in Table 3: this could express the fact that PCE does not possess active H

sites which could interact with the acceptor oxygen site of the various solvents tests.

Nevertheless, PCE possesses a high solubility in the various solvents because of its moderate

vapour pressure at temperature close to ambient.

Page 7
The activity coefficients obtained with TEGDME and DEHA, in comparison with formerly

published data or calculated values using modified Uniquac methods, were discussed in a

previous paper (Hadjoudj et al., 2004). The data obtained with the two Genosorb® were

found to be in fair agreement with the data published by Clariant.

Activity coefficients are little dependent on temperature in a narrow range (Hadjoudj et al.,

2004): the data reported in Table 3 were therefore considered as valid at ambient temperature

for the design of the gas treatment process.

p t
Besides the activity coefficient as a criterion for absorption solvent selection, one has also to

r i
take into account the solvent viscosity (Shah, 1990; Heymes et al., 2007). With respect to the

s c
two separation methods involved in the studied hybrid process, i.e. gas liquid contactor and

u
membrane permeation, it is clear that a high solvent viscosity could be a drawback for each

a n
step. The viscosity data gathered in the Table 2 show a two orders of magnitude range for the

tested solvents; obviously the solvents of higher viscosity, i.e. above 100 mPa.s, will be the

less promising media at room temperature.

m
e d
p t
Investigation of membrane-solvent pairs for separation of VOC-solvent mixtures

c e
Membrane preparation

A c
Membrane samples used for sorption tests were prepared from polymer solutions at 15% in

THF, by casting in stainless steel moulds. After evaporation of the solvent, films were dried at

60°C under vacuum until constant weight was reached. The membrane thickness was

accurately measured with an electronic gauge, and ranged from 100 to 200 micrometers

depending on the sample prepared.

Some films of polymers contained inorganic fillers were also produced. Particles of graphite

(SFG 6 and SFG75, Carbone Lorraine, France) with diameters near 10-6 m were used for this

Page 8
purpose. Alternatively, the polymer was loaded with silica particles (Spherotech, Polysciences

inc.) of two different size ranges, namely 0.4-0.6 10-6 m, and 1.5-1.9 10-6 m. The defined

amount of particles was incorporated to RTVA® and mixed for 15 hours at ambient

temperature; aliquots of chloroform were added to reduce the viscosity to an acceptable level.

RTVB® was thereafter added and the mixture thoroughly stirred before casting. Due to the

high viscosity of the bulk, solvent evaporation from the viscous mixture had to be carried out

carefully to avoid formation of bubbles at the membrane surface. Thickness of filled

membranes usually exceeded greatly 10-4 m.

p t
Solvent sorption on polymers

r i
c
A sample of polymer with a weight close to one gram was immersed in 50 cm3 solvent at

s
u
80°C, a temperature level far above the scheduled permeation temperature to magnify the

a n
undesired phenomenon. Some tests were also made at 40°C for comparison. The weight of the

sample was controlled every day and the test was continued until steady state of the sample

m
could be attained, i.e. from 2 to 7 days according to the samples. The swelling extent was

e d
deduced on the basis of the initial weight of the polymer sample. The relative error in the

p t
determination was estimated to be below 10%. Additional tests were also carried out with

e
another elastomer, namely, poly (ethylene-co-propylene) (PEPP) for comparison.

c
c
The two polymers exhibit comparable behaviour as shown in Table 4, although with higher

A
swelling of PEPP for most of the tested solvents. As expected, more significant swelling was

observed at high temperature. Solvents whose ether or ester groups have a significant

hydrocarbon chain e.g. Genosorb 1843 and DEHA, exhibit appreciable affinity for the

polymer, whose aspect was dramatically changed after long-term immersion. In case that the

ether function possesses only a methyl group, the affinity between the elastomer and the

solvent is reduced resulting in lower swelling of the polymer. The presence of alcohol

functions in α and ω positions of the solvents alters drastically the affinity of the polymer for

Page 9
the solvent, which is nearly not sorbed anymore by the polymer. Tests carried out with

α−ω diols of higher molecular weights such as PPG or PCLT yielded swelling extent near 1%

or less.

Although PDMS was shown to be chemically inert only in α−ω diols, additional sorption tests

were made with PDMS filled with graphite or silica particles. As reported in previous papers

(Duval et al. 1994; Ji et al., 1995, 1996), incorporation of inorganic charges in the polymer

favours the selectivity toward hydrophobic organic compounds, such as the chlorinated

p t
compounds of interest. Additionally, this can also improve the mechanical properties of the

r i
solid film (Robb, W.L., 1968). Experiments nevertheless revealed the moderate effect of

s c
polymer filling: as a matter of fact, the dispersion of silica or graphite particles allowed the

u
swelling extent to be reduced up to 30% with DEHA or TEGDME. Although noticeable, this

a n
effect is not sufficient to consider ethers or esters as potential solvents for the hybrid

absorption-pervaporation process. Besides, the little significant sorption of solvents exhibiting

m
alcohol groups at their extremities was not changed in a significant manner by filling the

polymer structure.

e d
p t
Polymer filling does not allow sufficient repelling of the solvents. In addition, higher contents

e
of particles would result in reduced permeation flux due to increased tortuosity effects

c
c
because only non-porous particles (i.e. silica, graphite) were used. Therefore amongst the

A
various solvents investigated only alcohol group-containing solvents can be considered in the

design of the hybrid process relying upon PDMS membranes.

Pervaporation experiments

All the pervaporation tests were carried out at 30°C using a double walled cell with a 10-4m3

reservoir (Krea et al., 2004). PDMS commercial membranes were used (thickness≈1.5 10-4m,

area≈1 10-3m2), supported by a sheet of porous sintered metal and clamped onto the test cell

Page 10
with a viton ‘O’ ring arrangement forming a leak-free seal. The feed reservoir was vigorously

stirred with a magnetic barrel (500rpm) to limit as far as possible the liquid boundary

resistance. Pervaporation conditions were achieved using a vacuum pump (downstream

pressure range: 100Pa) while the permeate was continuously condensed in a glass trap

immersed in liquid nitrogen. The measurements showed a good reproducibility (within 10%

accuracy) and the pervaporation flux was calculated from the trapped permeate as follows:

J = mass (kg) /(time(h) . area (m2))

p t
For pervaporation of VOC-solvent mixtures, the results given are related to steady state

r i
conditions usually reached after one hour or less. The compositions of the feed and of the

s c
permeate were determined by gas chromatography analysis several times per day, taking into

u
account the permeate flux rate and the VOCs content. For experiments with low amounts of

a n
VOCs, the feed composition was additionally checked and adjusted when necessary to keep

constant the driving force (GC control of the feed concentration).

m
Preliminary experiments have been made with VOC-free solvents for several days at 80°C.

e d
This relatively high temperature was used to enhance solvent mass transfer if any. TEG, PEG

p t
400 and PCLT were tested: no solvent was found either in the cold traps or in any

e
downstream part of the module (in pipes,..); only slight fractions of water were recovered

c
c
after six days, due to some unsignificant air leakage in the pervaporation system. In fact, leaks

A
and dissolved gases permeation cannot be totally annihilated when pervaporation experiments

are performed for a long time (Vallières et al., 2001). This absence of permeation is due to the

combination of several physical features of these solvents such as their low volatility, high

molecular weight and polar property (glycol type), that finally inhibit their sorption and their

diffusion of these particular solvents in the hydrophobic PDMS membrane. By comparison

with the properties of the other solvents (Tab.2), it seems that the glycol feature was the

prevailing parameter. This result showing that the permeation could be totally prevented was

Page 11
the milestone of our work; they were fully in agreement with the extremely low sorption value

calculated from the swelling experiments (Table 4).

Solvents with 10 wt. % DCM were thereafter tested at 30°C. Under these working conditions,

the first result that must be underlined is that in any case the permeate consisted exclusively

of DCM. Again it confirmed the above results with mixtures containing 10wt% of VOCs, that

actually established the potential of this membrane method in a hybrid absorption –

pervaporation process to recover VOCs. Significant DCM fluxes were obtained, in particular

with TEG and PEG 400 (Table 5); the data were expressed with reference to a 100

p t
r i
micrometers membrane, assuming that the flux varied with the reciprocal of the membrane

c
thickness, as a result of a solution-diffusion process (Baker and Wijmans, 1995). With these

s
u
solvents the weight fraction of DCM in the liquid reservoir decreased quite rapidly with time,

a n
due to significant pervaporation mass transfer: thus the feed concentration had to be regularly

adjusted to the initial value so that the concentration gradient did not vary too much (i.e.

m
adjustment each hour or so); nevertheless the values reported in Table 5 being an average over

e d
time are likely underestimated. This phenomenon was less significant for PCLT because of

t
the lower permeation flux measured: the overall permeation of DCM is to be partly controlled

p
e
by mass transfer in the liquid to the membrane surface – polarisation phenomenon – and the

c
high viscosity of PCLT likely hinders to some extent this transfer occurring prior to transport

c
A
through the membrane.

Therefore a particular attention was put on TEG regeneration and dedicated pervaporation

experiments were carried out either with pure TEG or with TEG-DCM to determine the DCM

flux according to its feed concentration ranging from 10 to 0.5 wt.% in TEG. With pure TEG,

after three days of pervaporation no TEG could be detected on the downstream side of the

membrane, confirming once more the absence of affinity of TEG for the silicone with the

VOCs mixtures. With these binary feed mixtures, long-term experiments were continuously

Page 12
run at 30°C, going from one to three days for each concentration (i.e. 0.5 to 10 wt%, Figure

2). The results recorded exhibited a linear flux dependance with the DCM feed concentration.

As it is known from literature [Bhattacharya, S., 1997; Charbit, G., 1997], an additional liquid

boundary layer resistance as well as temperature polarization effects (Favre, 1999) could

occur in pervaporation under the combination of some specific experimental conditions, in

particular:

- low concentrated species,

- poor stirring conditions or low feed velocity,

p t
-

r i
when high permeate fluxes are obtained, due to high temperature or high diffusing

species,

s c
u
- and with very thin dense membranes.

a n
Hence, it is likely that the observed linearity indicates that the PV mass transfer under the

selected operating conditions was not affected by a concentration polarization in the liquid

phase at the upstream side of the membrane.

m
e d
The flux obtained at 10wt.% was significantly higher than the fluxes recorded with all the

p t
other solvents tested (Table 5), suggesting that TEG is the most promising candidate.

e
Furthermore the most important result was, once again, the fact that no permeation of TEG

c
c
could be detected in the downstream side of the pervaporation cell expressing the perfect

A
selectivity of the separation VOC- solvent.

Modelling of the hybrid absorption-permeation system

Development of the model

The model for the hybrid system shown in Figure 3 was a simple approach for the design

developed after a conventional method, as done by Shah et al. (2004) for another hybrid

Page 13
membrane process. Fundamentals related to liquid absorption and to membrane separation

can be respectively found in Dankwerts (1970) and Hwang and Kammermeyer (1975) The

specifications of the process were a given VOC content in the waste gas, the abatement

efficiency of the gas treatment process, the flow rates of gas and liquid phases, and the

number of transfer units of the column. Unknowns to be determined were the compositions of

the liquid organic phase at the inlet and the outlet of the column, and the membrane area (S)

required for sufficient regeneration of the solvent.

The following assumptions were made.

p t
1. Solutions and gases are ideal phases.
r i
s c
2. Both absorption in the solvent and permeation are isothermal and carried out at 20°C.

u
Although the permeation flux had been determined at 30°C, the value reported in

n
a
Table 5 was taken for the calculations.

m
3. The pressure is kept constant at the ambient level: pressure drops are neglected in the

d
simple approach.

t e
4. The G/L column is operated under counter-current conditions, with axial flow of the

two phases.

e p
c
5. The permeate pressure in the separation module is maintained at a low value.

c
A
6. The permeation module is designed in such a way that there is axial flow in the feed

side and complete mixing in the permeate side, as done in most cases (Hwang and

Kammermeyer, 1975).

Absorption tower

The number of transfer units, NTU, was calculated from the abatement of VOC in the column

on the basis of mole fraction defined with respect to the inert gas, Y:

Page 14
 Yout
dY
NTU = ∫ Y −Y
Yin
*
(2)

Y is linked to mole fraction y as:

y
Y= (3)
1− y

Y* is the gas phase mole fraction of VOC in equilibrium with the actual liquid phase at the

considered location in the column. Equilibrium between the two phases was expressed using

Henry’s law:

p t
y* = H ' x (4)
r i
Differential mass balance in the column was written as:
s c
GdY = LdX

n u (5)

where X =
x
a (6)

m
1− x

d
Integration of balance (5) between the bottom of the column and a given location yields:

e
G (Yin − Y ) = L( X out − X )

p t (7)

e
In particular, the overall mass balance in the column is written as

c
c
G (Yin − Yout ) = L( X out − X in ) (7’)

A
From rel. (4) and (7) Y* could be expressed as a function of Y, depending on mole fractions

Xout and Yin. The obtained expression was incorporated in rel. (2). Since NTU, Yin and the

VOC abatement defined as (1-Yout/Yin) are specified in the model, integration of rel. (2) yields

an equation of single variable, Xout, which could be thereafter determined. Mole fraction Xin

was deduced from Xout using the overall balance (7’).

Page 15
Permeation module

The calculations presented below were made according to the usual pervaporation

assumptions (Kammermeyer, 1975). The membrane thickness chosen for the calculation was

10 micrometers assuming the validity of the inverse flux-thickness relationship could hold,

i.e.assuming that no severe additional liquid layer resistance would occur. As a matter of fact,

it can be observed that a 10micrometers active layer is not such a thin one from the

manufacture point of view; indeed membrane with active layer of one or two orders of

p t
magnitude lower have already been manufactured in industry. One the other hand, the flux

r i
one can expect at 30°C with a 10micrometers active layer is not very high under the

c
postulated conditions (Fig.3: J<0.25kg/(h.m2)), i.e. the possible deviation due to a polarization

s
u
at the upstream side of the membrane should remain limited.

a n
As assumed above, plug-flow operation in the liquid phase was considered. Both the overall

liquid flow rate, q (mol.s-1), and the mole fraction of VOC, x, varied along the membrane

m
surface, S. At the inlet of the module, x=xout; the local low rate, qout was calculated from the

e d
flow rate of pure solvent and the amount of absorbed VOC, (L.Xout), taking into account the

p t
molecular weights and the density of the considered VOC.

e
Mass conservation over a differential area dS leads to the balance:

c
A c
d(qx) = -KmxdS = ymdq (8)

where y is the mole fraction of VOC in the gas drawn through the membrane. The overall

permeation constant Km, in mol.m-2 s-1, is often related to the membrane permeability Q used

in relevant investigations, and to the membrane thickness l:

Q
Km = (9)
l

Page 16
Since the membrane can be considered as perfectly selective with respect to the VOC, mole

fraction y is equal to unity, which leads to:

d(qx) = dq (10)

Developing the differential expression d(qx) of rel. (8) yields a differential equation linking x

and q. Integration of the obtained equation between the module inlet and a given location on

the membrane surface leads to:

(1 − xout )
q = qout
(1 − x )
(11)

p t
r
The elementary variation of the flow rate, dq, is equated to the amount of VOC permeated,
i
yielding the following differential equation:

s c
q out
(1 − xout ) dx = − K
(1 − x) 2 m x.dS

nu (12)

a
m
Integration of rel. (12) between the inlet and the outlet of the permeation module leads to the

d
expression of the surface required for the specified regeneration of the solvent phase:

S=
q out (
t e ) (
(1 − xout ) ln xout 1 − xin + qout xout − xin
xin (1 − x out ) K m (1 − xin )
)
(13)

p
Km

c e
A c
Application of the model to DCM absorption in TEG

The model was applied to the design of a pilot plant for the treatment of DCM-containing

gaseous wastes in TEG with the specifications given in Table 6. DCM absorption was carried

out in a G/L column provided with a Sulzer BX structured packing. Permeation of DCM was

carried out using a non-supported PDMS membrane.

Hydrodynamics and G/L mass transfer rates in such columns with comparable (VOC-solvent)

systems, were investigated formerly (Hadjoudj, 2004), which made it possible to estimate for

Page 17
the liquid flow rate, the column diameter and the height of transfer unit, HTU. The simulation

was carried out for a column assuming that NTU=6 (Table 6), corresponding to an overall

column height of approx. two meters.

Henry constant, H, was deduced from the activity coefficient of DCM in the solvent after:

H = γ DCM / PEG P sat (14)

The value for mass transfer coefficient Km was calculated from the permeation data, J,

t
recorded from low concentrated DCM in TEG (Fig.2), taking into account the molecular

i p
weight and the density of the VOC and TEG. For this work, it was preferred to calculate the

r
c
coefficient Km for a 10 micrometer thick membrane (a more industrial realistic thickness

value), leading to Km= 4.63 10-2 mol.m-2.s-1.

u s
n
The VOC abatement XVOC defined above was varied from 90 to 99% and the membrane area

a
required to fulfil the gas scrubbing condition was calculated depending on XVOC. As expected,

m
the variation of the membrane area needed for the VOC removal is not a linear relationship

d
and larger membrane area is required for high efficiencies, i.e. for small values of Yout/Yin

t e
(Figure 4). The abatement of the membrane separation step, θ, was also calculated:

θ = 1−
xin

e p (15)

c
xout

A c
Note that the permeation module is fed with the liquid at xout. The separation cut is shown to

be improved by a large membrane area, to be used for highly efficient gas treatment.

The values for S may appear as fairly large considering the modest flow rate of the air stream

to be treated. However, the existing technology for pervaporation (Néel, 1997; Scott, 1998),

with for instance spiral wound membrane modules, would allow compact systems of an

acceptable volume. Taking 1000 m2/m-3 as a realistic specific surface area for spiral wound

modules (Baker R.W., 2004) this would correspond to into a maximum volume of the

Page 18
membrane module close to 0.25 m3, for the range of operating conditions tested here. (NB:

this figure should be ideally compared to the volume of the absorption column). Further

decrease in the required membrane area could be obtained providing that a thinner membrane

active layer is used. Some available PDMS composite membranes show indeed active layer

thicknesses in the 10micrometer range (Research Institute, Geesthacht, Germany - GKSS); of

course in such a case the issue of the liquid boundary layer will have to be checked with the

appropriate module. However, should the absorbent solvent flux through the silicone rubber

t
be maintained at zero, as observed in this study, a viable hybrid process would be obtained.

p
r i
Comparative process analysis

s c
u
It has been shown in the previous section that is was possible to regenerate the absorption

n
a
solvent by pervaporation instead of using the conventional heating of the solvent. We now

m
report in the two following paragraphs the calculations of the energy demand corresponding

to each technique. Of course these calculations were made considering the same operating

e d
parameters of the absorptive solvent after the VOC absorption.

p t
Separation of DCM from the organic phase by flash distillation

c e
Flash distillation can be considered as a first approach for separation of two compounds of

A c
different boiling points. From the column specification and the regeneration conditions

deduced from the overall process, the following values have been considered

Solvent (TEG) 0.4804 mol s-1, i.e. 93.34 g s-1

xin, flash 0.00936

xout, flash 0.00297

The inlet flux is to be heated to a defined temperature and the liquid and vapour phases

recovered are assumed to be in equilibrium: the temperature to be applied in the flash is to

allow DCM be in equilibrium with DCM vapour at the considered pressure, Pflash. Because of

Page 19
the low volatility of the solvent, the vapour is assumed to consist exclusively of DCM

whereas the liquid contains DCM at mole fraction xout, flash. Phase equilibrium between the

two phases at the outlet of the flash can be expressed as:

PDCM = Pflash = γ DCM / TEG .xout , flash .P sat (T flash ) (16)

The saturation pressure of pure DCM at Tflash is deduced straightforward:

Pflash
P sat (T flash ) = (17)
γ DCM ,TEG .xout , flash

p t
i
Taking into account the value for the activity coefficients and the process specifications, rel.

(17) yields:

c r
P sat (T flash ) = 886.94 Pflash

u s (18)

B
a n
The saturation pressure of DCM can be estimated by using Antoine’s expression:

m
sat
log P (T ) = A − (19)
T +C

d
With A = 7.0803, B = 1138.91 K and C=-41.7 K, where P is in Torrs. More accurate relation

e
t
can be obtained using Chebychev correlation included in PRO II package for instance, as

p
e
explained in a previous paper (Hadjoudj et al., 2004). Calculations have been made using the

c
second technique for temperature ranging from 80 to 320°C (Figure 5).

c
A
As shown in Figure 5, because of the very low DCM content in the outlet liquid, very high

temperatures have to be applied so that flash distillation can be operated with acceptable

pressures. However, vaporization of the solvent whose boiling point is near 322°C at ambient

pressure, is to severely compete with that of very dilute DCM: contrary to the above

assumption the vapour collected at the outlet can be expected to have a significant mole

fraction of solvent. Therefore, because of the very low DCM fraction in the organic liquid,

separation by heating the organic mixture cannot be carried out with a high efficiency.

Page 20
Insert Figure 5

Nevertheless, the energy required to heat the organic phase and to allow DCM vaporization

have been estimated considering a flash temperature of 200°C, corresponding to a flash

pressure at 0.041 bar. The specific heat of TEG has been estimated depending on temperature

t
from published data for 1, 2- ethanediol and glycerol: as a matter of fact thermal properties of

p
organic molecules are more governed by their chemical formula and the nature of the

r i
functional groups, than by the molecular weights: as a matter of fact, the two emulation

s c
solvents indeed exhibit fairly similar specific heat (Figure 6) and the following law has been

considered for TEG:

n u
a
Cp (kJ kg-1 K-1) = 2.4 + 0.004 T (20)

m
where T is in °C.

e d
Insert Figure 6

p t
c e
c
Taking into account the flux of solvent leaving the absorption column and neglecting the

A
contribution of dilute DCM yields the estimate for the heat required to heat up the liquid from

20°C to 200°C:

QHeating solvent = 47.72 kW

In addition, from the vaporization enthalpy of DCM, taken at 377 kJ kg-1 (Handbook of

Chemistry and Physics), around 98 W are required for vaporization of the specified DCM

flux. Besides, the energy consumed to reduce the pressure from the ambient level to the actual

pressure has to be calculated and added to the two above contributions

Page 21
To conclude, because of the low content of VOC in the organic phase issued from the

absorption column, regeneration of the heavy solvent by simple heating cannot be easily

carried out because of the appreciable energy consumption (even though heat exchangers

could be designed for partial heat recovery) and mainly because of the significant,

competitive vaporization of the solvent.

Separation of DCM from the organic phase by pervaporation

Pervaporation is now the method considered to regenerate the absorption solvent. Conversely

p t
to the previous method, the energy demand is not directly related to the liquid-vapour

r i
equilibrium of DCM and TEG but to the removal of the DCM vapour sorbed within the

s c
membrane. For this purpose, we shall consider using a vacuum pump at the downstream side

u
of the membrane as the main utility, taking the operating parameters identical to those of the

n
a
previous case, i.e. for DCM xin =0.00936 at the inlet of the membrane module and and xout =

0.00297 at the outlet, hence:

m
d
DCM flux through the membrane = L(xin – xout) = 3.07 x10-3 mol s-1

t e
with the liquid feed flow L=0.4804 mol s-1

e p
Given these data, the specific work of a vacuum pump in J per mol can be calculated as:

c
⎛ 1− γ1
⎜ψ
⎜
⎝
c − 1⎟
⎞
⎟
⎠

A
W = γ .R.T (21)
η .(γ − 1)

with T: temperature (K), γ: adiabatic expansion coefficient of DCM (ca 1.4), ψ:

pressure ratio (vacuum pressure/discharge pressure), η: isothermal efficiency (ca 0.75)

(Häring H.W., 2008).

So, taking 1 Bar as discharge pressure and 0.1 to 0.001 Bar as vacuum pressure, a pressure

ratio between 0.1 and 0.001 can be assumed (NB: for a liquid ring vacuum pump such as

Page 22
often proposed for pervaporation applications, a 0.01 Bar vacuum pressure can easily be

achieved). The corresponding specific work of the vacuum pump is about 80 kJ.mol-1, which

translates to an overall energy requirement of ~250 W. For a more reliable estimation, one can

also take into account the second thermodynamic law, i.e. a factor 3 to 4 for possible

comparison of the heat duty of flash distillation to electrical power needed by the vacuum

pump, meaning that pervaporation will require about 1 kW.

Compared to the energy demand calculated for the flash distillation process, the difference is

p t
striking, showing that pervaporation remains also more interesting also from the energy point

of view.

r i
s c
n u
a
Conclusion

m
The paper presents the results of an investigation of regenerative processes for treatment of

d
VOC-containing air relying upon the combined selective absorption and solvent regeneration

t e
by membrane techniques. This case study has explored the removal of chlorinated compounds

p
from air. Emphasis was put on compatibility of their physico-chemical properties of the VOC,

e
c
the absorption solvent and the membrane material, in particular on their mutual interactions:

A c
VOCs have to exhibit a high affinity to both the solvent and the membrane material, whereas

the polymer has to be repellent to the absorption solvent. Obviously a conflicting situation

existed and a compromise had to be reached for the choice of the solvent and of the

membrane.

It was found that using aliphatic esters (i.e. DEHA or DEHP) as absorption liquids with

polymeric membranes being homogeneous or modified with inorganic fillers (either carbon or

silica), would not lead to a viable hybrid process because the performance of the membrane

will be rapidly altered by the fast sorption of the low polar liquid into the membrane layer.

Page 23
Conversely, it was evidenced that using protic liquids like polyglycols did not exhibit the

above dramatic drawback for membrane integrity. Indeed tetraethyleneglycol and

polyethylene glycol 400 were found to be suitable for the considered VOC and PDMS

membranes. It must be also underlined that using polyols and PDMS allow reaching infinite

selectivity for VOC permeation. As a matter of fact, even with relatively high levels of VOC

in the mixtures (from 1 to 10 wt.%), only VOC could be collected at the downstream side of

the membrane. Hence it could be concluded that coupling liquid absorption and membrane

separation units is technically feasible using already commercial liquids and PDMS

p t
r i
membrane. As a promising fact, one should note that liquid regeneration is thus carried out

c
with infinite selectivity. Last but not least, it was shown that the energetic demand evaluated

s
u
for the regeneration of the solvent was much smaller by pervaporation.

a n
To go ahead with the industrial evaluation of a hybrid absorption-pervaporation process, first

molecular simulation could be used to refine the selection of protic liquid absorbents in order

m
to optimize VOCs sorption with liquids of low viscosity. Secondly, pilot-scale experiments

e d
with an industrial pervaporation module will be required to allow more accurate predictions of

t
its treatment capacities and potential energy saving. In particular these experiments will help

p
e
to take into account the pressure loss in the module and the exact significance of the

c
polarization concentration linked to pervaporation operating conditions, i.e. mainly VOC

c
A
concentration, membrane thickness, used temperature and module hydrodynamics.

Acknowledgments

The programme was funded by the Incitative Coordinated Action (ACI) programme Non-

Pollution, Depollution of CNRS-DGA and Ademe in France.

Nomenclature

Page 24
DCM dichloromethane

DEHA di(2-ethylhexyl) adipate

DEHP di(2-ethylhexyl) phthalate

Genosorb monomethyl ether solvents (Clariant)

H Henry’s constant (Pa)

H’ Henry’s constant defined on the mole fractions

G gas flow rate (mol.s-1)

p t
Km permeation constant (mol.m-2 s-1)

r i
L liquid flow rate (mol.s-1)

sc
PCLT polycaprolactone triol

n u
a
PDMS polydimethylsiloxane based membranes

m
PEG polyethyleneglycol

d
PEPP poly(ethylene-co-propylene)

PPG polypropylene glycol

t e
Psat

e
saturated pressure
p
PV
c
pervaporation

c
A
Q membrane permeability (mol.m.s-1)

q liquid flow rate (mol.s-1)

NTU number of transfer units

RTV room temperature vulcanization

PCE tetrachloroethylene

Psat saturation vapour pressure (Pa)

Page 25
S membrane surface (m2)

TEG tetraethylene glycol

TEGDME tetraethylene glycol dimethylether

VLE vapour liquid equilibria

VOCs volatile organic compounds

x mole fraction of VOC in the liquid

y mole fraction of VOC in the gas

p t
ym mole fraction of VOC in the gas drawn through the membrane

r i
Y
c
mole fraction of VOC in the gas defined on the basis of inerts

s
Y*
u
gas phase mole fraction of VOC in equilibrium with the liquid

n
a
γ activity coefficient

m
θ abatement of the membrane separation step

e d
p t
c e
A c

Page 26
References

Baker, R.W. and Wijmans, J.G., (1995), Process for recovering organic components from

liquid streams, US Patent 5, 169, 533.

Baker, R.W. (2004), Membrane technology and applications, 2nd Ed, John Wiley&Sons,

Chischester, England

Bhattacharya, S., and Hwang, S. T., (1997), Concentration polarization, separation factor, and

t
Peclet number in membrane processes: Journal of Membrane Science, 132, 73-90

i p
Charbit, G., Charbit, F., and Molina, C., 1997, Study of mass transfer limitations in the

r
c
deterpenation of waste waters by pervaporation: Journal of Chemical Engineering of

Japan, v. 3, p. 382-387

u s
n
Clariant, Genosorb brochure, available at

a
http://fun.clariant.com/fun/internet.nsf/vwWebSpider/4437E6CED995464CC1256E29003

865C9
m
d
Dankwerts, P.V., “Gas-liquid reactions” McGraw Hill, New York (1970)

e
t
De Kermadec, R., Lapicque, F., Roizard, D. and Roizard, C., (2002), Characterisation of the

p
e
SO2-N-formylmorpholine complex: application to a regenerative process for gas

c
scrubbing, Ind. Eng. Chem. Res. 41, 153-163.

c
A
Duval, J.-M., Kemperman, A.J.B., Folkers, B., Mulder, M.H.V., Desgrandchamps, G.,

Smolders, C.A., (1994), Preparation of zeolite filled glassy polymer membranes, Journal

of Applied Polymer Science 54 (4), 409-418

Hadjoudj, R, Ph.D. Dissertation, INPL Nancy (2004) (in French)

Hadjoudj, R., H. Monnier, Roizard, C. and Lapicque, F., (2004), Absorption of chlorinated

VOCs in high-boiling solvents: determination of Henry’s law constants and infinite

dilution activity coefficients, Ind. Eng. Chem. Res. 43, 2238-2246.

Page 27
Häring H.W., (2008) Industrial gases processing. Wiley-VCH Verlag, Weinheim

Heymes, F., Demoustier, P.M., Charbit, F., Fanlo, J.-L., Moulin, P., (2007), Treatment of gas

containing hydrophobic VOCs by a hybrid absorption-pervaporation process: The case of

toluene, Chemical Engineering Science 62 (9), 2576-2589

Hillaire, A., Favre, E., (1999), Isothermal and nonisothermal permeation of an organic vapor

through a dense polymer membrane, Industrial and Engineering Chemistry Research 38

(1), 211-217

Hwang, S-K. and Kammermeyer, K., (1975), Membranes in Separation, Technique of

p t
Chemistry, Vol. VII, Wiley and Sons, New-York.
r i
s c
Ji, W. and Sidkar, S.K., (1996), Pervaporation using adsorbent-based membranes, Ind. Eng.

Chem. Res., 35, 1124-1132.

n u
a
Ji, W., Sikdar, S.K., S.-T. Hwang, (1995), Sorption, diffusion and permeation of 1,1,1,-

m
trichloroethane through adsorbent-filled polymeric membranes, J. Membrane Sci., 103,

d
243-255.

t e
Khan, F.I., and Ghoshal, A.K., (2000), Removal of volatile organic compounds from polluted

p
air, J. Loss Prevention in the Process Industries, 13, 527-545.

e
c c
Krea, M., Roizard, D., Moulai-Mostefa, N., and Sacco, D., (2004), New copolyimide

membranes with high siloxane content designed to remove polar organics from water by

A
pervaporation, J. Membrane Sci., 241, 55-64.

Lide, D. R. (2002) Handbook of Chemistry and Physics (2002), 83th Ed., CRC Press, 2002,

ISBN 0-8493-0483-0

Majumdar, S., Bhaumik, D., Sirkar, K.K., and Simes, G., (2001), A pilot-scale demonstration

of a membrane based absorption stripping process for removal and recovery of volatile

organic compounds, Environ. Progress, 20, 27-36

Page 28
Néel, J., Pervaporation, Tech & Doc. Lavoisier, Paris (1997) (in French).

Peng, M., Vane, L. M., and Liu, S. X. (2003) Recent advances in VOCs removal from water

by pervaporation, Journal of Hazardous Materials, 98, 69-90.

Philippe, J.L. (1997), Lösungsmittelrückgewinnung aus Abluftströmen durch Absorption und

Pervaporation, Fortschrittberichte VDI Reihe 15, Nr. 192, VDI-Verlag Düsseldorf (ISBN

3-18-319215-2)

Pirruci, S., Bombardi, D., Concu, A., and Lugli, G., (2003), Modelling, design and

commissioning of a sustainable process for VOCs recovery from spray paint booths,

p t
Computer-aided Chemical Engineering, 14, 251.
r i
s c
Poddar, T.K, and Sirkar, K.K., (1997), A hybrid of vapour permeation and membrane-based

u
absorption-stripping for VOC removal and recovery from gaseous emissions, J.

n
a
Membrane. Sci., 132, 229-233.

m
Poddar, T.K., Majumdar, S., and Sirkar, K.K., (1996), Membrane-based absorption of VOCs

d
from a gas stream, AIChE J., 42 3267-3280.

t e
Riddick, J. A., (1986), Organic Solvents: Physical Properties and Methods of Purification, 4th

p
Ed. (1986), John Wiley, vol. 002, ISBN 0-471-08467-0

e
c
Roizard, D., Teplyakov, V., Favre, E. , Fefilatiev, L., Lagunstsov, N., Khotimsky, V., (2004),

A c
VOC’s removal from water with a hybrid system coupling a PTMSP membrane module

with a stripper Desalination 162: 41-46.

Robb, W.L., (1968), Thin silicone membranes. Their permeation properties and some

applications. Annals of the NewYork Academy of Sciences, 146, 119-137.

Shah, Y. T., (1990), The effect of solid wettability on gas-liquid mass transfer in a slurry

bubble column. CES, 45 (12) 3593-3598.

Page 29
Schmidt, A. and Uhlrich, M., (1990), Absorptionmittel zur Reiniging Lösemitelhaltiger

Abluft (Absorption Media for purification of solute-containg waste gases), Chem.-Ing.

Technik, 62, 43.

Sciamanna, S.F. and Lynn, S., (1988), Solubility of hydrogen sulphide, sulphur dioxide,

carbon dioxide, propane, and n-butane in poly-(glycol ethers), Ind. Eng. Chem. Res. 27,

492-499.

Scott, K, Handbook of Industrial Membranes, (1998), 2nd edition, Elsevier Advanced

Technology, Oxford.

p t
r i
Shah, M.R., Noble, R.D., and Clough, D.E, (2004), Pervaporation-air stripping hybrid process

s
for removal of VOCs from groundwater, J. Membrane Sci., 241, 257-263.
c
u
Uddelholm, H., Armand, B., and Vikström, P., (1987),Verfahren und Vorrichtung zur

n
a
Reinigung von schadstoffhaltigen Gasen, German Patent DE 3618511.

m
Vallières, C., Favre, E., Roizard, D., Bindelle, J., Sacco, D., (2001) New insights into

d
pervaporation mass transport under increasing downstream pressure conditions: Critical

t e
role of inert gas entrance, Industrial and Engineering Chemistry Research 40 (6), 1559-

1565

e p
c c
Van Dam, M.H.H., Lamine, A.S., Roizard, D., Lochon, P., and Roizard, C., (1997), Selective

sulphur dioxide removal using organic solvents, Ind. Eng. Chem. Res. 36, 4628-4637.

A
Wang, X., Daniels, R., and Baker, R.W., (2001), Recovery of VOCs from high-volume low-

VOV-concentration air streams, AIChE J., 47, 1094-1100.

Weisweiller, W., Pinter, A. and Winterbauer, H., (1995), Neuartige

Waschflüssigkeitssysteme zur Absorption Flüchtiger Organischer Kohlenwasserstoff

(VOC), Final report, FZKA-PEF 137 of the PEF Project Europäisches Forschungszentrum

für Massnahmen zur Luftreinhaltung, Universität Karlsruhe (in German).

Page 30
Xia, B., Majumdar, S., Sirkar, K.K., (1999), Regenerative oil scrubbing of volatile organic

compounds from a gas stream in hollow-fiber membrane device, Ind. Eng. Chem. Res. 38,

3462-3472.

p t
r i
s c
n u
a
m
e d
p t
c e
A c

Page 31
 Chlorinated VOCs

Strong affinity Strong affinity

by the presence by hydrophobic
of apolar groups groups in the polymer

No mutual affinity
Heavy solvent Membrane polymer
Presence of
polar groups
t
Limit the solvent permeation

p
i
* Higher Tg with PPEP
e.g. OH functions

r
* Filling with carbon or silica

s c
n u
a
m
Figure 1: Constraints for the various compounds involved in a hybrid absorption-
pervaporation process for treatment of VOC containing gases.

e d
p t
c e
A c

Page 32
 4000
3500
Normalized flux (g/h.m )
2

3000
2500
2000
1500
1000
500

p t
ri
0
0 1 2 3 4 5 6 7 8 9 10

c
C, DCM (wt%)

u s
Figure 2: Regeneration of the solvent (TEG) by specific removal of DCM at 30°C. Effect of

n
DCM feed concentration on pervaporation normalized flux (10micrometer thickness).

a
m
e d
p t
c e
A c

Page 33
 Treated gas (yout)
Regenerated solvent
with VOC (xin)

Gas-liquid Solvent flow rate L

column
Pervaporation chamber

x
Air flow rate G Vacuum pump

t
Red. pressure

r i p
c
Inlet waste gas VOC-Solvent Pure VOC

s
VOC at yin (xout)

containing gases.

n u
Figure 3: Schematic view of a hybrid absorption-pervaporation process for treatment of VOC

a
m
e d
p t
c e
A c

Page 34
 35 1

30
0.8

Membrane abatement θ
Membrane area (m2)
2

25

20 0.6

15 0.4
10 t
u
C
ran
b
em
M

0.2

t
5
/
ran
b
em
M

(2)
ran
b
em
M

0
0 0.02 0.04 0.06 0.08 0.1

r i
0
p
Yout/Y in

sc
n u
a
Figure 4: Membrane area of the pervaporation chamber and membrane cut versus the (out/in)

m
concentration ratio for treatment of 30 m3 h-1 DCM-containing gas by absorption in
TEG and separation through a 10 micrometer PDMS membrane (see Table 6).

e d
p t
c e
A c

Page 35
 0.25
Pressure in the flash (bar)

0.2

0.15

t
0.1

0.05

r i p
c
0

s
0 50 100 150 200 250 300 350

u
Flash temperature (°C)

a n
m
d
Figure 5: Variation of the flash pressure with the temperature for DCM vaporization

t e
e p
c c
A

Page 36
 3.6

3.4

Specific heat (kJ kg-1 K-1)

3.2

2.8

2.6 Ethanediol
2.4 Glycerol

2.2 Post. Law for TEG

t
0 50 100 150 200 250

i p
Temperature (°C)

c r
u s
n
Figure 6: Specific heat of 1,2- ethanol diol and glycerol and postulated law for TEG.

a
m
e d
p t
c e
A c

Page 37
 VOC Molecular weight Density Boiling point at 1 Atm. Psat at 20°C
(kg mol-1) (kg m-3) (°C) (kPa)
DCM 0.0849 1325 39.6 47.74
TCE 0.1658 1623 121.1 1.81

Table 1: Physico-chemical properties of the two VOCs

p t
r i
s c
n u
a
m
e d
p t
c e
A c

Page 38
Solvent Type Molecular weight Density Boiling point Psat at Viscosity at
at 1 Atm 20°C 20°C
(g mol-1) (kg m-3) (°C) (Pa) (mPa.s)

TEGDME Ether 222.3 1011 275 0.25 4.0

DEHA Ester 370.6 920 417 < 0.1 12.3
DEHP Ester 390.6 980 384 < 0.1 56.5
Genosorb 300 Ether 300 (3 <n< 8) * 1030 > 250 <1 7.8
Genosorb 1843 Ether 1843 (3 <n< 8) * 920 > 250 < 0.1 4
TEG Glycol 194.3 1120 327 42

t
PEG 400 Glycol approx. 400 1127 <1 120
PPG 425 Glycol approx. 425 1000 > 100
PCLT 300 Glycol approx. 300 1070 > 310 <1

r i p > 200

c
Table 2: Physicochemical properties of the solvents tested. Genosorb 300® and Genosorb

s
1843® are respectively dimethyl- and dibutyl ethers of polyethylene glycol –(OEt)n-
(*): average number of repeating unit

n u
a
m
e d
p t
c e
A c

Page 39
 Genosorb 300® TEGDME DEHA TEG PEG 400 PPG 425
Ether or Ester type of solvent α,ω hydroxyl type of solvent
DCM 0.20 0.23 0.37 0.38 0.22 0.32
TCE 1.49 1.62 0.71 - - -

Table 3: Activity coefficient at infinite dilution at 30°C of the two VOCs in the studied
solvents

p t
r i
s c
n u
a
m
e d
p t
c e
A c

Page 40
 Genosorb Genosorb TEGDME DEHA PCLT PEG 400 PPG TEG
300® 1843® 425
Ether or Ester type of solvent α,ω−hydroxyl type of solvent
PDMS 4.2 36.2 6.6 23.3 0 0.2 1.9 0.2
(2.8) (28.7) (4.4) (15.2)
PEPP - 9 76.5 - 0.2 -

t
Table 4: Swelling extent (in weight percents) of polymers in various solvents at steady state.
Data in parenthesis were obtained at 40°C, the others at 80°C.

r i p
s c
n u
a
m
e d
p t
c e
A c

Page 41
Solvent DCM Permeation flux
(kg m-2 h-1)
TEG 0.31
PEG 400 0.18
PCLT 300 0.07

Table 5: DCM flux measured at 30°C through the PDMS membrane, with reference to a
100µm membrane. DCM: 10weight % in the feed mixtures.

p t
r i
s c
n u
a
m
e d
p t
c e
A c

Page 42
 Features Value
Temperature (°C) 20
Air flow rate (m3 h-1 at 20°C) / (mol s-1) 30 / 0.3464
Solvent flow rate (m3 h-1 / mol/s) 0.30 / 0.4803
VOC abatement 90 – 99%
PDMS active layer thickness (µm) 10
Diameter of the column (m) 0.15
NTU of the column 6

p t
i
Membrane area for 90% and 99% abatement (m2) 8 – 32.8

c r
s
Table 6: Features of the hybrid process considered for DCM abatement

u
a n
m
e d
p t
c e
A c

Page 43