Colloids and Surfaces

A: Physicochemical and Engineering Aspects 163 (2000) 151 – 164

www.elsevier.nl/locate/colsurfa

Separation of emulsions using fibrous fabric

B.J. Briscoe *, P.F. Luckham, J.N. Jayarajah, T. Akeju
Department of Chemical Engineering and Chemical Technology, Particle Technology Group, Imperial College of Science,
Technology and Science, Prince Consort Road, London, SW 7 2BY, UK

Received 30 November 1998; accepted 22 June 1999

Abstract

The separation of emulsions is a difficult problem which is frequently encountered in industry. Physical treatment
processes involving the use of membranes are commonly used to separate one phase from another. The study to be
described involves the determination of the efficiency of an emulsion separation process, using a fibrous membrane.
An emulsion consisting of oil (di-n-butyl phthalate) droplets in water was made to flow, at various rates through a
membrane comprising of a mat of aramid (du Pont, USA, Kevlar™ 49) fibre. An increase in the oil phase
concentration of the emulsion, causes an increase in the separation efficiency and an optimum concentration that
provides the maximum separation is identified. Increasing the flow rate and the fabric thickness, obtained by using
multiple layers of fabric, were both found to lower the separation efficiency. It was observed that the Kevlar™ fabric
had a natural saturation oil phase limit, which was reached more rapidly for the thicker fabrics. The critical oil
exposure volume (COEV), (saturation volume) was calculated to be between 0.66 – 2.82 × 10 − 3 m3 of oil per m2 of
flux area of fabric per fabric layer for different fabric thickness. However, the values are not the absolute natural
saturation limits. Calculations performed using the various separation rates, determined from the experimental
results, for each fabric thickness suggest that distinctly separated single Kevlar™ fabric units have a significantly
higher separating efficiency than multiple adjacently arranged Kevlar™ fabric units. © 2000 Elsevier Science B.V. All
rights reserved.

Keywords: Liquid–liquid separation; Emulsion droplets; Fibrous fabric; Droplet capture; Demulsification

1. Introduction sion separation is encountered in practice and is

of both industrial and academic interest [1–3].
The paper reports a study of the efficiency of The separation of emulsions has been an issue
oil droplet capture on a fabric surface, as a means of concern in industry for many years. Effluent oil
of emulsion separation. This technique of emul- removal from water, which is frequently encoun-
tered in the management of wastewater, is a very
* Corresponding author. Tel.: + 44-171-5945559; fax: +44-
difficult problem to effectively and economically
171-5895111. resolve. The problems associated with oily-waters
E-mail address: b.briscoe@ic.ac.uk (B.J. Briscoe) are mainly due to the highly variable physical and

0927-7757/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 7 - 7 7 5 7 ( 9 9 ) 0 0 3 0 9 - X
152 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164

chemical characteristics of the oil in water dis- average pore size and configuration are estab-
persion. Usually, oily wastes are difficult to han- lished from the size and form of individual ele-
dle and create problems in secondary biological ments from which the filtration medium is
treatment facilities. On offshore platforms, water manufactured. On the average, pore sizes are
separated from gas – oil mixtures contains dis- greater for the larger medium elements and tend
persed oil droplets, which must be separated to be more uniform. The fabrication method of
from the water before discharge to the sea. The the filter also affects the average pore size and
most common commercial types of oil and wa- form. Pore characteristics depend upon the
ter separators rely primarily upon the action of properties of fibres in woven fabrics, as well as
gravity, as these units are effective and also eco- upon the exact methods of sintering for glass
nomical [4,5]. The main constraint is to provide and metal powders. In addition, pore character-
sufficient residence time for the less dense oil to istics are greatly influenced by the separation
float on top of the water surface. This method processes occurring within the pore passages, as
becomes very much less effective, however, when this leads to a decrease in effective pore size and
dealing with emulsions containing very small oil consequently an increase in the hydraulic flow
droplets, and other methods must be employed resistance [13,17]. This results from particle pen-
to remove the smaller drops [5,6]. Fibrous mem- etration and retention within the pores of the
branes have been used successfully to trap and filtration medium.
coalesce such droplets and separate them from This paper describes the separation of oil dis-
water through size enlargement processes [7–13]. persions from water using a fibrous membrane;
The selection of appropriate materials for the oil and the aqueous phase were chosen to be
membrane construction in filtration units is a of a comparable density (this would not gener-
significant challenge since this component largely ally be the case in practice — see later); an
determines the overall separation efficiency, par- emulsion consisting of a dispersed phase of an
ticularly during the initial operating period of a oil in water was pumped through a flow cell,
filter, as well as determining the flux-pressure separated into two parts by a fibrous membrane.
drop relationship and also the operating life of The droplets of oil adhere to the surface of the
a filter [14]. A good filtration medium should fibre, and the efficiency of the process was de-
have the following characteristics in varying de- termined after several passes of the emulsion
grees, depending on the specific problem: it through the membrane. Then, using the experi-
should offer a minimum hydraulic resistance to mental results, calculations were performed to
the filtrate flow, allow the easy discharge of discover an arrangement of fabrics that gener-
residue; be resistant to cake formation; should ates the greatest separation within the filter ar-
not undergo swelling when in contact with rangement.
filtrate and washing liquid; remain heat resistant
within the ranges of filtration; have sufficient
strength to withstand the filtering pressure and 2. Equipment and materials used
the mechanical wear and erosion; and finally
must also be capable of suppressing the exces- The fabric materials used for were Kevlar™
sive inclusion of particles into its pores [14–18]. 49 fabrics [19] (Du Pont Aramid, USA; com-
Different filter media are employed depending monly used for ballistic projectile protection ar-
upon the particular technique and the intended mour) dibutyl phthalate (Aldrich, 99%),
application [17]. Examples of common filtration waxoline Red MP, acetone (BDH, 99.5%) and
media are cottons, wools, linens, glass fibres, de-ionised water (Purite Ltd.). A single fabric
porous carbon, sand, diatomite, coal, quartz and was of approximately 250 mm in thickness; the
metals. Such a classification is convenient for filaments with the fabric were ca. 7 mm in di-
selection purposes, especially when resistance to ameter. This fabric was chosen primarily be-
aggressive suspensions is a consideration. The cause of its small constituent filament diameter.
 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164 153

Fig. 1. The equipment used for the separation of di-n-butyl phthalate dispersions in water. The flow rate of the emulsion in the inlet
of the cell was controlled by the height of the capillary head.

The relative low environmental sensitivity and 3. Experimental procedure

high mechanical strength were useful features.
A flow cell constructed from glass was made up The principal steps involved in the experimental
of two similar halves and was joined together procedure were as follows.
bymetal and PTFE gaskets, which also supported
the Kevlar™ fabric (Fig. 1). The horizontal length 3.1. Emulsion preparation
of the glass tube when joined together was 313
mm. The glass tubes had an internal diameter of The emulsion was prepared by mixing 5 g of
di-n-butylphthalate in 1 l of distilled water for
26 mm and were 450 mm in height. Capillary
about 10 min with the aid of a laboratory mixer
tubes of 6 mm in diameter were placed inside the
emulsifier [Silverson Machines Ltd., model
inlet (upstream) and the outlet (downstream) of L2AIR]. The dibutyl phthalate contained waxo-
the flow cell. The flow of the emulsion through line red MP in the ratio of 1 g of dye to 50 g of
the fabric was controlled by varying the head oil. The oil was stained red so as to enable the
height of the inlet capillary tube with respect to amount of oil in the emulsion to be measured by
the capillary head at the outlet; which was at a the aid of an ultraviolet visible photospectrometer
fixed reference (zero) position (see Fig. 1). (UV-VIS 554). Although the indicator certainly
154 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164

modified the interfacial properties of the oil, this centration of oil remaining in the emulsion after
aspect was not investigated further since the each cycle. A wavelength value of 530 nm for
correlation of the dye was kept constant. repetitive scanning was used; the instrument was
calibrated against emulsions of various oil phase
3.2. Insertion of fabric composition and thus the concentration of oil in
the test emulsion could be readily computed.
A new piece of as-received Kevlar™ fabric was
inserted into the cell, and constrained using the
metal gaskets, at the commencement of each ex- 4. Results and discussion
periment. On some occasions, more than one
fabric layer was inserted to increase the thickness The results acquired by varying different
of the fabric filter. The active flux area of the parameters are discussed below. Firstly, the effi-
fabric was 5.31× 10 − 4 m2. ciency of the separation process for a given con-
centration of the emulsion, using a single
3.3. Experiment Kevlar™ fabric layer, is described. In the subse-
quent sections, the effects of varying the initial oil
The prepared emulsion was poured into a flask, phase concentrations, the flow rates and the
which was placed on a magnetic stirrer and was Kevlar™ fabric thickness on the efficiency of the
stirred constantly, and pumped into the flow cell separation process are considered. In the latter
via the peristaltic pump. (The emulsion in the feed part of this section, the relative separation effi-
was constantly stirred so that the oil did not ciency of single and multiple Kevlar™ fabric ar-
separate and also so that the oil drop size distri- rangements (units) are estimated using the
bution may be preserved.) The emulsion was then experimental data.
flowed through the cell (fabric) and the residue
was drained at the other end (downstream) of the 4.1. Efficiency of the separation process; general
cell, into a beaker. Samples (10 ml) of this residue effects
emulsion, for the first pass, were taken for analy-
sis. The residue emulsion was then re-emulsified Fig. 2 shows a graph of the percentage (mass)
and pumped back into the cell and samples from separation of oil against the number of passes of
the downstream of this pass were collected for the emulsion (initial concentration of 5 g l − 1 of
analysis. This process was repeated several times
and samples were acquired, from the downstream,
after each pass.
When the fabric had reached the saturation
point, i.e. no flow of emulsion occurred across the
fabric, a vacuum pump was used at the down-
stream to enhance the flow. The vacuum pump
was operated for a short period; and unblocked
some of the pores in the fabric by releasing the
captured oil into the effluent stream.

3.4. Analysis

To 5 ml of each sample were added 2 ml of

acetone to solubilise the emulsion thus forming
one phase, which was pinkish red in colour (due Fig. 2. The separation achieved for a 5 g l − 1 of emulsion at a
to the dye in the oil phase). The spectrophotome- constant flow rate at capillary head height of 100 mm after
ter (UV-VIS 554) was used to determine the con- each pass through a single fabric.
 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164 155

tion 4.2). It should be noted that the dibutyl

phthalate (density of 1044 Kg m − 3 at 20°C) has a
similar density to that of the water (1000 Kg m − 3
at 20°C) and hence there is likely to be little or no
gravity separation occurs within the filter. In
many other instances, the oil would be less dense
and migrate out of the filter to the top of the cell,
thus increasing the separation; this would natu-
rally be an important component of the overall
separation process.

4.2. The effect of 6arying oil phase concentration

on separation

In these experiments, the total separation of oil,

after each pass was monitored for different initial
concentrations of the emulsion. This included the
oil entrapped on the fabric as well as the filtered
oil retained in the upstream. Thus to determine
the total separation in the flow cell, the residue (in
downstream) emulsion was collected and
emulsified before recycling back into the up-
stream, whereas the filtered oil in the inlet re-
mained at the upstream position.
Fig. 3a and b show the effects of using different
initial concentrations of the emulsion upon the
total efficiency of the separation process. As may
be seen, the curves are either linear or linear
Fig. 3. (a) The mass of oil removed for various initial concen-
trations of oil, at a constant flow rate at capillary head height
initially and then follow a power law relationship,
of 100 mm, after each pass through a single fabric. (b) The with the index being less than one.
amount of oil separated (mass percentage) for various initial Further, plotting the percentage separation (af-
concentrations of oil, at a constant flow rate at capillary head ter nine passes) as a function of the initial concen-
height of 100 mm, after each pass through a single fabric. tration of the oil (Fig. 4) indicates that the
separation efficiency increases with the initial con-
water) across the fabric at a constant flow rate.
centration of the oil (0–5 g l − 1) until it achieves
This graph indicates that there was an increase in
an optimum value (ca. 5 g l − 1), after which the
the amount of oil separated after each pass due to separation efficiency starts to decrease (5–10
the capture of oil droplets onto the fabric surface. g l − 1). This trend is probably due to the establish-
At the end of the ninth pass, 61% oil separation ment of a critical oil exposure volume (COEV) of
had taken place. the fabric (COEV is discussed in greater detail
As can be seen from these data, the separation later). The COEV is naturally achieved much
increases linearly with the increase in the number more rapidly for the higher initial concentrations
of passes; however, the first pass produced the of oil than the lower ones (Fig. 3a). With the 2.5
most marked separation. After a certain number g dibutyl phthalate/litre of water emulsion, the
of passes, there would probably be no increase in amount of oil separated after the ninth pass was
the amount of oil absorbed by the Kevlar™ fab- 0.82 g. When the concentration was doubled, i.e.
ric. Fig. 3 illustrates this saturation phenomenon 5 g l − 1 emulsion, the amount of oil separated
and the origin of which is discussed below (Sec- after nine passes was almost four times that
156 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164

Fig. 6. Fraction of oil separation after each pass through a

single fabric for an initial concentration of emulsion of 5 g l − 1
Fig. 4. The amount of oil separation (%) after nine passes at various capillary head height (increasing flow rates) of 70,
through a single fabric for various initial concentrations of 100 and 130 mm.
emulsion at constant flow with capillary head height at 100
mm. an increase in the amount of oil separated. This
trend follows power law relation. From Fig. 5,
amount (3.06 g), while that for the 10 g l − 1 however, the actual figures in terms of the total
emulsion was just 1.6 times (4.8 g) the amount of mass separated were 3.06 and 4.8 g for 5 and 10 g
oil separated for the 5 g l − 1 emulsion. Hence, the l − 1, respectively. This suggests that even though
separation efficiency starts to increase a great deal the separation efficiency had fallen, the amount of
as the initial concentration increases, and then oil being separated was still increasing, but by a
gradually starts to decrease as the emulsion be- lesser amount. This could be due to the release of
comes more concentrated. the trapped oil in the fabric into the effluent
Fig. 4 indicates that the separation efficiency stream (downstream) as a result of saturation of
decreases (5 – 10 g l − 1) after achieving a maximum the fabric (see later for explanation of COEV).
(5 g l − 1); the percentage separation after nine
passes for the 5 g l − 1 emulsion and 10 g l − 1 4.3. Efficiency of separation as a function of flow
emulsion was 61 and 48%, respectively. Fig. 5, on rate
the other hand, shows that with an increase in the
The effect of varying the emulsion flow rate
initial concentration of the emulsion, there is still
upon the separation efficiency with a 5 g l − 1
emulsion is illustrated in Fig. 6. It appears that
the greatest amount of separation was achieved
with the lowest flow rate (70 mm of capillary head
height). At the higher flow rates (i.e. higher capil-
lary heads) the separation is less but increases
approximately linearly with increasing number of
passes. However, for the lowest flow rate the
separation increases in a power law manner, with
an index less than one. At the flow rate of 70 mm
of capillary head height, a separation of about
83% is achieved after nine passes, while at a flow
rate of 130 mm of capillary head height, the
Fig. 5. Mass of oil removed after nine passes through a single separation is about 43% (a factor of two less).
fabric for various initial concentrations of emulsion at con- This is likely to be because, at the slower flow
stant flow at capillary head height of 100 mm. rates, the emulsion droplets had more time within
 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164 157

Fig. 7. Total amount of oil separated in the flow cell after nine passes through a single fabric for an initial concentration of emulsion
of 5 g l − 1 at various capillary head height flow rates.

the fibres to deposit and coalesce, thereby forming 4.4. The effects of increasing fabric thickness
larger captured droplets. These larger droplets,
due to their large size become absorbed within The effects of increasing the thickness of the
thepores of the Kevlar™ fabric, enabling a greater fabric, on the separation, were also investigated.
separation to be attained. With the lowest flow The experiments were performed by binding two,
rate, the first four passes produced about 72% three and four Kevlar™ fabrics together, to in-
separation while the next five produced only a crease the thickness of the total fabric within the
further 11% separation in total. There was only flow cell. Also, the number of passes was in-
about a 2.6% additional separation attained, be- creased to 18, instead of nine and a vacuum pump
tween the passes 6 and 9. These values suggest was used to enable the flow of emulsion through
that by the sixth pass, the fabric had become the fabric when there was no (or very little) flow.
almost fully saturated with oil and an increase in
The vacuum pump was only used for a fraction of
the number of passes thereafter would not yield
a second and only once for each fabric thickness.
an appreciable additional amount of separation.
It was deduced, from Fig. 8, that after a certain
This is also the case for the experiments per-
number of passes, the pores of the fabric became
formed with higher flow rates. Also at higher flow
blocked, thereby restricting the passage of emul-
rates the droplets tend to detach from the fabric.
The total amount of oil separated (g) after nine sion from one side of the fabric to the other. This
passes against the flow rate, i.e. the height of condition is termed here as the ‘critical oil expo-
capillary head (mm), is plotted in Fig. 7. This sure 6olume (COEV)’ of the fabric. This COEV is
figure shows an increase in the flow rate results in defined as the maximum volume of oil retained
a decrease in the separation, which is of a linear per unit area of the flux (i.e. surface area of the
nature. exposed fabric at the upstream) per layer of fab-
158 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164

ric. Following the attainment of the COEV limit, of the fabric may be determined.
the fabric is unable to accommodate any more oil Once the fabric attains the COEV limit, some
and due to the constant pumping of the emulsion of the already captured oil on the fabric is re-
across the fabrics, the oil already captured is now leased into the effluent stream thus significantly
extensively released into the effluent stream, reducing the separation efficiency. This phe-
partly due to the flow direction of emulsion, thus nomenon was also noted by Okada and Akagi
contaminating the downstream and lowering the [20] during their experiments, which involved the
overall separation efficiency on the filter. It should separation of oil from oil-in-water mixtures by
be noted that although the vacuum pump did flow through a bed of glass beads. They observed
release some of the oil into the effluent stream, the coalescence behaviour of oil droplets in the
this effect was minimal, since the separation con- bed photographically. During the initial period
tinues to decrease with the increase in the number there was very little separation but after 30 min
of passes even after the cessation of the use of the 80% of the oil had been removed by the bed. At
vacuum pump as can be noted from the Fig. 8. this juncture, a thin layer of oil was observed in
A graph of the separation (% mass) achieved the downstream and after 150 min, this thin layer
versus the filter thickness is depicted in Fig. 8. It of oil was pushed out further downstream, and it
shows that the separation increases almost lin- migrated on to the upper surface of the down-
early with increasing number of passes of emul- stream wall; the oil they used was less dense than
sion across the fabric until it reaches a maximum. the continuous fluid. The form of their data for
From this maximum value the average amount of the separation efficiency versus time was similar
oil retained, per unit area of the flux, in each layer to that shown in Fig. 8.
Using a single layer of fabric, for an emulsion
of 5 g l − 1 of water, the critical decrease in the
separation efficiency occurred by the 13th pass
and the efficiency decreased even further until the
last pass, to an efficiency of about 12.5%. The
highest separation was achieved at the 12th pass,
with a 31% recovery. With two, three and four
fabrics this maximum was achieved after passes 9,
6 and 3, respectively, yet the separation in all
cases was between 1.47 and 1.68 g of oil (Table 1).
This shows that an increase in the fabric thickness
does not markedly influence the maximum
amount of oil retained within the fabrics.
Correspondingly, the value of the critical oil
exposure volume (COEV) decreases with the in-
crease in the fabric thickness, as is depicted in Fig.
Fig. 8. Effect of increasing thickness of Kevlar fabric on 9. This shows a power law relationship, with a
separation for a 5 g l − 1 emulsion at constant flow rate of 150 negative index of approximately unity. This result
mm of capillary head height. The vacuum pump (connected is quite unexpected because an increase in the
downstream) was turned on for a short period (fraction of a
second) when there was no flow through the fabrics; this was
fabric thickness, might be expected to yield a
at pass number 12, 9, 6 and 3 for one, two, three and four greater capture of oil on the fibre surfaces thus
fabric layers, respectively. This action unblocked some of the increasing the separation efficiency, i.e. four layers
pores in the fabric by removing some of the oil into the of fabric should perhaps, in theory, separate four
downstream. The vacuum pump was used at the peak points. times as much oil as one fabric layer; the current
The maxima on each of the graph represents the COEV limit.
The COEV was achieved at passes 3, 6, 9 and 12 for fabric
results show otherwise (Table 1 and Fig. 9).
thickness four, three, two and one, respectively. The COEV is This phenomenon is clearly due to the increase
in the range of 0.66–2.82 m3 (m2 per fabric layer) − 1. in the axial (with respect to the flow direction)
 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164 159

Table 1
Critical oil exposure volume (COEV) for different thickness of fabricsa

No. of fabric Total mass of oil recov- Total volume of oil re- Volume of oil retained COEV (m3 of oil per m2
(thickness) ered (g) in fabrics tained in fabrics (m3) per fabric layer fabric flux per fabric
(m3 fabric layer−1) layer)
(a) (b) (c) =[1.044/(a)]×10−6 (d) = (c)/(a) (e) = (d)/5.309×10(4
(×10−6) (×10−6) (×10(3)

1 1.563 1.499 1.499 2.82

2 1.597 1.531 0.766 1.44
3 1.679 1.627 0.542 1.02
4 1.469 1.408 0.352 0.66

a
Density of DBP = 1.044 g cm−3; Flux diameter (diameter of inlet of cell) =26 mm; Flux surface area = p*d 2/4 =5.309×10−4
2
m.

Fig. 9. Critical oil exposure volume (m3 of oil retained in fabrics per m2 of flux area of fabric per layer of fabric).

surface provided by the addition of extra fabric Kevlar™ fabric.) When the emulsion is passed
layers. This additional surface area may attract across the fabric, some of the initial micro-
more oil droplets on the fibre surfaces, and an oil droplets adhere to the surfaces of the fibre
drop moving axially envelopes these smaller whereas others move into the fabric through the
droplets and grows in size. As illustrated schemat- pores and are captured by the fibres within the
ically in Fig. 10, the released droplet from the fabric matrix. The droplets that have adhered to
four fabric layers is much larger than that of a the fibre facing the inlet stream grow by coalesc-
single fabric layer. This means that a larger vol- ing with either the neighbouring droplets or with
ume of oil is released into the effluent stream from the micro-droplets from the emulsion, when the
the four fabric layer arrangement and this in turn peripheries of the droplets touch and after the
reduces the separation efficiency, Fig. 9. thinning of the water film between these droplets.
The mechanism by which the droplet grows in As the droplet grows, the capillary forces that
the fabric is probably as follows. (Note: di-n-butyl hold the droplet to the fibre decreases and conse-
phthalate (DBP) does not readily wet the quently the body force of the droplet increases.
160 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164

Since the flow is in the axial direction, this enlarged size of the droplet released into the effluent stream
droplet is either released into the inlet stream or would be larger as is shown schematically in Fig.
forced into the fabric due the axial pressure exerted 10. This therefore means that the saturation limit
by the flow. When this droplet enters the fabric observed from the plots are not the true or ideal
matrix, it envelopes those droplets already adhered saturation of the fabrics. Basically, the drops pass-
onto the fibre which are in its path and coalesces ing through the large thickness fabric scavenge the
with them and grows larger. This process continues resident drops.
until the drop reaches the end of the fabric where
it is released into the effluent stream thus decreas- 4.5. Analysis of separation efficiency and the
ing the separation efficiency of the filter. If the axial configuration of the fabric (Table 2)
distance of the fabric is increased, i.e. by increasing
the fabric thickness, then it is more likely that the The percentage separation rate per pass was

Fig. 10. The above diagram shows the growth of a small droplet as it passes through the fabric. The growth is due to the coalescence
with other droplets on fabric. With the four fabric layers the eventual size of the droplet released into effluent stream is very much
larger than that of a similar droplet in a single fabric.
 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164 161

separation rates thus obtained is shown in Table

2, and represented in Fig. 11 which shows that
increasing the thickness of the fabric yields an
increase in the separation rate. The percentage
separation rate per pass increases steeply as the
thickness is increased. It should be noted that this
section deals with the percentage separation rate
per pass whereas the previous section dealt with
the absolute value of separation of oil per fabric
layer. Also, for the different configurations stated
below, the same experimental conditions of Fig. 8
is assumed.
The separation rates calculated can now be
used to infer the separation efficiency for various
Fig. 11. Percentage separation per pass against fabric thickness configurations of the fabric arrangements, as is
for a 5 g l − 1 emulsion at constant flow rate. These were shown schematically in Figs. 12–14. Fig. 12
determined by the slopes of the curves in Fig. 8, between the shows two potential fabric arrangements in the
initial and the first maximum points.
flow cell; for the case where two single fabrics are
obtained by taking the slope between the initial separated by a distinct gap, so that the emulsion
pass and the peak of each curve, from Fig. 8, for is separated twice in one pass, (for a 5 g of oil l − 1
the respective fabric thickness. A table of the of water emulsion 2.38% (Table 2) is separated

Fig. 12. Comparison of calculated separation efficiency between two single Kevlar fabric unit and one two layered Kevlar fabric
units using the data from Table 2 at various stages (passes) of the experiment. The first illustration shows the concentration of the
oil when passed through two single fabrics spaced apart (the results of the single fabric from Fig. 8 and Table 2 are used); and the
latter shows the relative effect of two fabrics bound together (the results of two fabrics from Fig. 8 and Table 2 is used).

Fig. 13. Comparison of calculated separation efficiency between three single Kevlar fabric unit and one three layered Kevlar fabric
units. The first illustration shows the concentration of the oil when passed through three fabrics spaced equidistant apart (the results
of the single fabric from Fig. 8 and Table 2 are used); and the latter shows the relative effect when three fabrics are bound together
(the results of three fabrics from Fig. 8 and Table 2 is used).
162 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164

Fig. 14. Comparison of calculated separation efficiency between four single Kevlar fabric unit and one four layered Kevlar fabric
units. The first illustration shows the concentration of the oil when passed through four fabrics spaced equidistant apart (the results
of the single fabric from Fig. 8 and Table 2 are used); and the latter shows the relative effect when four fabrics are bound together
(the results of four fabrics from Fig. 8 and Table 2 is used).

Table 2 passes (Fig. 8). Also, in comparison to the two

Separation rates for different Kevlar fabric thicknessa
single fabric system (Fig. 12), the three single fabric
Number of Kevlar fabrics 1 2 3 4 system achieves a much higher separation; the
separation after one and 12 passes are, respectively,
Mass separation (%) per 2.34 3.689 5.942 9.234 4.6 and 44% for the two single fabric system and
pass
7 and 58% for the three single fabric arrangement.
a
The mass % separation per pass was deduced from the Fig. 14 compares a four single fabric arrange-
corresponding slopes for different fabric thickness from Fig. 8. ment with that of a single four layered system.
A linear relationship was assumed between the initial and the Interestingly, both systems separate the same
first maximum point. amounts after each pass, as a result of which both
when it flows through the first fabric, and the same systems are identical in their behaviour. Since the
percentage is recovered when it passes through the four layered arrangement reaches the COEV limit
second) and therefore the separation efficiency is much quicker, hence this system is inefficient since
twice that of one single fabric system. Fig. 12 also the separation efficiency decreases following the
shows the concentration of the emulsion leaving COEV limit as explained earlier. The separation
the filter system after passes 2, 9 and 12; the efficiency of the four single fabric system is 9 and
concentration of the emulsion are 4.54, 3.24 and 69% after one and twelve passes, respectively.
2.80 g l − 1 respectively, at which point the COEV Again comparing this with the separation effi-
is attained. The second arrangement shows the ciency of the other systems described in Figs. 12
same elements as described above but instead of and 13 shows that the four single fabric system is
two single fabrics, one double layered fabric is overall the most efficient system. Finally, it can
used. In this arrangement, the respective concen- be concluded that greater separation can be
tration of the emulsion leaving the cell after passes achieved by placing many fabrics apart, in the flow
2 and 9 are 4.63 and 3.56 g l − 1, which are greater cell, rather than binding them together. It may be,
than the corresponding values for two single fabric however, that the system with the adjacent fabrics
system, hence the latter system is more efficient will produce large drops in the effluent and hence
than the former. produce more extensive gravitational separation
In Fig. 13, three single fabric units and one three outside the filter.
layered fabric units are compared. Again, the three
single fabric units separates more than the three
layered unit, after passes 1 and 6. The former 5. Conclusions
arrangement reaches the COEV after the 12th pass
whereas the latter reaches it in half the number of An experimental study of the separation of an
 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164 163

oil/water emulsion using fibrous membranes is tion limit of fabric, was calculated to be
described. It was observed that the separation of between 0.66 and 2.82 m3 per m2 of flux area of
the oil increases almost linearly with the in- fabric per layer of fabric of oil for the different
crease in the number of passes until a maximum Kevlar™ fabric thickness and it increased in
was reached (Figs. 2, 3a and b). The greatest value with the increasing of the fabric thickness.
amount of oil recovered with the filter, after The critical oil exposure volume (COEV) de-
nine passes, was from the emulsion that had the creases with the increase in the fabric thickness,
highest initial concentration of oil, as is illus- as is depicted in Fig. 9. It is a power law rela-
trated in Fig. 3a. From this figure, it is apparent tionship with a negative index, in Fig. 9 this
that the mass of oil separated increases with the index is approximately equal to − 1. This result
increasing of the initial emulsion concentration. is unexpected as an increase in the fabric thick-
Further, the separation efficiency increases ness, or rather in the fibre surface, should yield
with the initial concentration of the oil (0–5 g a greater capture of oil on the fibres thus in-
l − 1) until it achieves an optimum value (5 g creasing the separation efficiency, i.e. one would
l − 1), after which the separation efficiency starts expect four layers of fabric to absorb four times
to decrease (5 – 10 g l − 1). This trend is probably as much oil as one fabric layer and therefore
due to the existence of a critical oil exposure increase the separation. But the current results
volume (COEV) for the fabric. show otherwise (Table 1 and Fig. 9). This phe-
The separation in the filter increases markedly nomenon may be attributed to the increase in
with the decreasing of the flow rates (Fig. 6), the axially positioned capture surface provided
and in a linear pattern (Fig. 7). At a flow rate by the addition of extra fabric layers as is
of 70 mm of capillary head height, a separation shown in Fig. 10. This therefore indicates that
of about 83% was achieved after nine passes, the maximum saturation values of the various
while at a flow rate of 130 mm of capillary fabric configurations are a function of the oper-
head height, the separation was about 43% (a ating conditions.
factor of two less). This was attributed to the Calculations were also performed to determine
fact that, at slower flow rates, the emulsion has the separation rates (mass separation % per
more time to settle, coalesce and migrate, pass) for different fabric thickness before the
thereby forming larger droplets. These droplets, COEV limit. These were obtained as 2.4, 3.7,
because of their large size become absorbed 5.9 and 9.2% separation/pass for one, two, three
within the pores of the Kevlar™ fabric enabling and four contacting fabric layers, respectively
a greater separation to be attained. (Table 2). Finally, different fabric configura-
Increasing the thickness of the fabric, by us- tions, on the filter, in the flow cell have been
ing multiple adjacent layers, caused lower but considered and the best configuration for maxi-
more rapid separations. After a certain number mum separation was determined using the mass
of passes, the fabric pores become saturated percentage separation per pass values above and
with oil, and an increasing flow of emulsion also the COEV limit. The values deduced
through it causes the oil already adhered to the showed that it would be more effective to adopt
fabric to be forced out into the effluent stream, a single Kevlar™ fabric unit instead of two or
thereby lowering the separation efficiency in the three fabric layer combined units (Figs. 12 and
filter. The number of passes for this situation to 13). The single and the four fabric units
occur was found to be different for the different achieved similar separation (Fig. 14). Yet, the
fabric thickness. For the single Kevlar™ fabric single fabric unit would be the best option be-
unit, this was at the 12th pass, while for two, cause the COEV is achieved during the 12th
three and four Kevlar™ fabric units, the num- pass whereas for the four fabric layers this was
ber of passes was found to be nine, six and reached at the third pass; the separation effi-
three, respectively. The critical oil exposure vol- ciency, on the filter, decreases significantly after
ume (COEV), the volume of oil at the satura- the separation reaches the COEV limit.
164 B.J. Briscoe et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 163 (2000) 151–164

References [10] D.B. Chambers, Chem. Ind. Nov. (1978) 35.

[11] D.F. Sherony, R.C. Kintner, Can. J. Chem. Eng. 49
[1] K.J. Lissant, Demulsification: industrial applications, (1971) 314.
Surf. Sci. Ser. 13 (1983) 63. [12] B.J. Briscoe, J.N. Jayarajah, P.F. Luckham, Wetting of
[2] H.J. Schmidt-Holthausen, Filtr. Sep. 15 (1978) 350. fibres: capture and growth of oil droplets on a single fibre,
in: Proceedings of the Third World Congress on Particle
[3] L.A. Spielman, S.L. Goren, Ind. Eng. Chem. Fundam. 11
Technology, Brighton, UK, 1998 No. 286 (CD-ROM),
(1972) 73.
pp. 21528 – 21598.
[4] N.P. Cheremisinoff, P.N. Cheremisinoff, Filtration
[13] G.A. Davies, G.V. Jefferies, Ind. Eng. Chem. Fundam. 9
Equipment for Wastewater Treatment, Prentice-Hall,
(1970) 519.
New York, 1993.
[14] A.R.N. Fairclough, K.L. Chan, G.A. Davies, Chem. Eng.
[5] D.B. Purchas, Industrial Filtration of Liquids, 2nd, Leon- Res. Des. 65 (1987) 396.
ard Hill Books, 1971. [15] G.A. Davies, G.V. Jefferies, Filtr. Sep. 6 (1969) 349.
[6] G.D. Dickey, Filtration, Reinhold, New York, 1961. [16] R.E. Treybal, Liquid Extraction, 2nd, McGraw-Hill, New
[7] S.S. Sareen, P.M. Rose, R.C. Gudesen, R.C. Kintner, York, 1963.
AIChE J. 12 (1966) 1045. [17] G.D. Dickey, Filtration, Reinhold, New York, 1961.
[8] B.J. Briscoe, K.P. Galvin, P.F. Luckham, A.M. Saied, [18] C. Defromont, F. Douard, Filtr. Sep. 2 (1965) 303.
Colloids Surf. A 56 (1991) 301. [19] H.H. Yang, Kevlar Aramid Fiber, Wiley, New York,
[9] A.M. Saeid, Separation of Oil and Water Using Modified 1993.
Fibre, MSc Thesis, Imperial College, London, 1990. [20] K. Okada, Y. Akagi, J. Chem. Eng. Jpn. 18 (1985) 509.