Analysis and design of Mass Transfer for Continuous Liquid-liquid

extraction with Static Mixers: a practical approach.

Lorenzo Milani, Trevor Murray and Andrea Adamo
Zaiput Flow Technologies, MA-USA

Introduction

Liquid-liquid extraction (LLE) is an extremely popular chemical processing technique used to purify
compounds of interest partly due to its large capacity at low energy requirements. LLE consists of two
primary steps – mass transfer and phase separation, which, when effectively coupled, provide a
“theoretical stage” of extraction. In traditional extraction techniques, two phases are extensively mixed
through overhead mixers in a batch vessel before gravitational forces cause the two phases to settle out.
This methodology provides inherent disadvantages, from more obvious drawbacks like the time it takes
to mix the two phases and allow them to settle, to more subtle inefficiencies like conducting extractions
in reactor vessels which reduces a plant’s productivity by the extractive ratio. A change in extraction
paradigm is needed to improve this antiquated method of LLE.
Zaiput Flow Technologies has taken the first step in this effort by developing a continuous liquid-liquid
separator. This scalable, membrane driven separator allows for the instantaneous separation of biphasic
systems and can even separate emulsified systems due to its lack of reliance on gravity as a driving force
of separation. This separator can be hooked up to a batch vessel after mixing to reduce settling times,
allowing for a degree of plant efficiency improvement. Further advantage can be gained, however, if the
mixing between phases can also be brought in to flow. This not only eliminates the waiting time for mixing,
but also opens up capacity in the upstream reactor since there is no longer dead space in the reactor left
for the extractive solvent – allowing for reactor output improvements of 2x or more. Unfortunately, there
is not a well-defined predictive model for determining the mixing type and length for this inline mass
transfer.

Figure 1. Future extraction setup leveraging full reactor capacity and inline extraction.
In this paper we attempt to simplify and model mixing and mass transfer phenomena using static mixers
in a tube to roughly estimate the time required to reach chemical equilibrium in a given liquid-liquid
extraction process. To the best of our knowledge, there has been little or no effort in analyzing mass
transfer over time during a continuous liquid-liquid extraction process such as this one.
In order to estimate the equilibrium times, we simplify our mass transfer problem as a diffusion one. In
the next sections, we explain our approach on how we modelled diffusion of a solute in two immiscible
phases, and how we estimated the effective diffusivity constant for the liquid-liquid extraction process,
as well as the characteristic diffusion length given by the mean droplet radius. In the end, we then present
our model results with some experimental validation.

Mass transfer Approach

In liquid-liquid extraction, mass transfer is driven by a difference in chemical potential in towards a more
stable condition with lower free energy (chemical equilibrium). The presence of two immiscible phases
introduces mass transfer limitations from a thermodynamic standpoint compared to a natural diffusion
mass transfer process with only one phase. Mass transfer will happen until equilibrium is reached rather
than when the concentration gradient is zero. Although convection also plays a role, diffusion of solutes
in liquids is still the most important principle even in liquid-liquid extraction. Therefore, in our model, we
treat the liquid-liquid extraction process as a regular molecular diffusion in liquids with its governing
conservation of mass equations and diffusivity as the proportionality constant (see next section for more
details). This requires an estimation of the effective diffusion constant for our given liquid-liquid extraction
process, which is explained later on in this paper.
Static mixers are implemented to carry out many different mixing operations and their main principles are
well documented (Grace, 1971; Mutsakis, 1986; Cybulski, 1986; Myers, 1997; Thakur, 2003). Static mixers
can be very helpful in continuous liquid-liquid extraction processes as well because they allow the two
phases involved to be brought into intimate contact with a high degree of turbulence in order to obtain
high mass-transfer rates. Although there has been plenty of work, both experimentally and theoretically,
on the characterization and effects of implementing static mixers in terms of degree of mixing or particle
size distribution (Middleman, 1974; Chen, 1978; Haas, 1987; Berkman, 1988; Streiff, 1997; Legrand, 2001;
Theron, 2011) there has been little or no effort in analyzing mass transfer while using static mixers (Streiff,
1977). Furthermore, to the best of our knowledge, no one has studied static mixers from an extraction
perspective. This is most likely due to the lack of theoretical studies, complexity in experimental work,
multiplicity of factors that may play a role, and scarcity of continuous equipment capable of carrying out
liquid-liquid extraction in a tube.
In our liquid-liquid extraction setup, the static mixer causes droplet breakup of the two immiscible phases.
For our purposes, this can be seen as an increase in mass transfer due to faster diffusion because of a
reduced length scale and increased interfacial area. If we model the original feed phase after mixing as a
system of droplets containing the solute, the smaller the droplet the less time it would take for the solute
to diffuse out and leave that droplet and be extracted by the adjacent phase. In transient diffusion
problems, the diffusion time is indeed directly proportional to the square of the characteristic length,
which for our case, if we simplify the droplets to be of spherical shape, would be the radius of the droplet.
Because of their strong impact on diffusion time to reach equilibrium, droplet size obtained after the static
mixer were therefore investigated and presented later in this paper.
In conclusion, our idea was to model the complex mass transfer problem in continuous liquid-liquid
extraction processes by assuming a transient diffusion problem happening inside a droplet. This applies
whether the solute is “entering” the droplet (droplet is the extraction solvent) or “leaving” it (the droplet
is the raffinate phase). Details on the model are presented in the next section.

Mass Transfer Model

In our setup, the continuous and dispersed phase are travelling together inside a tube. As mentioned
earlier, we modelled our mass transfer phenomena as a transient diffusion problem happening inside of
a sphere, therefore using spherical coordinates. The idea behind this is to simplify the droplet of the
dispersed phase as sphere particles surrounded by the continuous phase, as depicted in Figure 2. The
model is valid whether the solute is originally in the dispersed phase (as in Figure 2) or in the continuous
phase since the mathematical problem is the same in both cases.

Figure 2. Representation of the spherical mass transfer model for the concentration of solute A (intensity of blue
color represents concentration) in the two phases initially and at equilibrium. Phase B is the dispersed phase, while
phase C is the continuous phase. In this case, the solute A is originally in phase B.

The complete conservation of mass equation in terms of concentration of a species i in spherical

coordinates is the following (Deen, 2011):

𝜕𝐶𝑖 𝜕𝐶𝑖 𝑣𝜃 𝜕𝐶𝑖 𝑣𝜑 𝜕𝐶𝑖 1 𝜕 𝜕𝐶𝑖 1 𝜕 𝜕𝐶𝑖 1 𝜕2 𝐶𝑖

+ 𝑣𝑟 + + = 𝐷𝑖 [ (𝑟 2 )+ (𝑠𝑖𝑛𝜃 )+ ] + 𝑅𝑉𝑖 Eq. 1
𝜕𝑡 𝜕𝑟 𝑟 𝜕𝜃 𝑟 𝑠𝑖𝑛𝜃 𝜕𝜑 𝑟 2 𝜕𝑟 𝜕𝑟 𝑟 2 𝑠𝑖𝑛𝜃 𝜕𝜃 𝜕𝜃 𝑟 𝑠𝑖𝑛 𝜃 𝜕𝜑2
2 2

In our system, the presence of two immiscible phases B and C leads to a coupled system of two equations,
one for the concentration of the solute A in the continuous phase C and one in the dispersed phase B,
which both need to be solved simultaneously. This is because the concentration of the solute in the
continuous phase depends on the dispersed phase, and vice versa. However, to simplify the mathematical
problem, we can decouple the system and solve Eq. 1 only for the concentration profile of the solute A
inside the inner sphere (phase B) in the model. The effect of the continuous phase C can be included in
this case by switching the diffusivity constant DAB to an effective diffusivity DABC (see next section for
explanation). For our purposes, this effective diffusivity can be treated similar to an overall mass transfer
coefficient for our liquid-liquid extraction process. The role of the phase ratio between the two phases
has an impact on the effective diffusivity of the system (see next section for more details) but does not
change the mathematical solution to Eq. 1 when considering the concentration of A inside the sphere
because there is no radial dependence outside the dispersed phase.
Furthermore, by assuming that the dispersed phase and the continuous phase after the static mixer are
travelling at roughly the same speed and there are no substantial disruption phenomena happening inside
the sphere of liquid, it is possible to neglect the convection terms of Equation 1. Furthermore, by making
the fair assumption of symmetry and homogeneity in the small spherical droplet, only diffusion in the
radial direction can be considered, therefore dropping the two angular diffusion terms. The reaction term
can also be dropped due to this being an extraction process in which the amount of molecules of solute
does not change during this step because there are no chemical reactions. This is also valid for acid-base
extractions because, although the solute might react and change into its conjugate, the total number of
molecules between original solute and conjugate is unchanged. The final assumption, required to solve
the partial differential equation analytically by providing one more boundary condition, is to consider the
concentration of the solute in the two phases to be already in equilibrium at the interface right from the
beginning. This is a reasonable assumption since mass transfer of the solute A from phase B to phase C
starts in the static mixer during the droplet breakup process. All of these considerations lead to the
following simplified equation:

𝜕𝐶𝐴 1 𝜕 𝜕𝐶𝑖
= 𝐷𝐴𝐵𝐶 [ (𝑟 2 )] Eq. 2
𝜕𝑡 𝑟 2 𝜕𝑟 𝜕𝑟

where 𝐷𝐴𝐵𝐶 is the effective diffusivity of the LLE system described in details in the next section. By making
the substitution U=CAr (common substitution done in many spherical coordinates problems), the problem
can be rewritten as:

𝜕𝑈 𝜕2 𝑈
𝜕𝑡
= 𝐷𝐴𝐵𝐶 [
𝜕𝑟 2
] Eq. 3

with the following boundary conditions:

𝑈=0 𝑎𝑡 𝑟 = 0 𝑎𝑡 𝑎𝑛𝑦 𝑡 > 0 Eq. 4
𝑈 = 𝑟𝐶𝐴0 𝑎𝑡 𝑡 = 0 𝑎𝑡 0 < 𝑟 < 𝑅 Eq. 5
𝑈 = 𝑅𝐶𝐴,𝑒𝑞 𝑎𝑡 𝑟 = 𝑅 𝑎𝑡 𝑎𝑛𝑦 𝑡 > 0 Eq. 6

where CA,eq and CA0 are the equilibrium and the initial concentration, respectively, of the solute in the
dispersed phase, while R is the radius of the spherical droplet of the dispersed phase. The initial
concentration is uniform throughout the sphere (Eq. 5), except at the interphase where it is assumed to
be already in equilibrium (Eq. 6).

The solution of this problem leads to the following dimensionless concentration profile (Williams and
Crank, 1975):

𝐶𝐴 −𝐶𝐴0 2𝑅 (−1)𝑛 𝑛𝜋𝑟

𝑍(𝑟, 𝑡) = 𝐶 = 1 + 𝜋𝑟 ∑∞
𝑛=1 𝑠𝑖𝑛 𝑅 𝑒𝑥𝑝(−𝐷𝐴𝐵𝐶 𝑛2 𝜋 2 𝑡/𝑅 2 ) Eq. 7
𝐴,𝑒𝑞 −𝐶𝐴0 𝑛

thus, the equilibrium time can be estimated by extrapolating the time required to make Z approximately
equal to 1.
Effective Diffusivity Estimation

While the diffusivity of a diluted gas mixture (as long as it is at low pressure) can be derived using the
kinetic theory of gases, this theory does not hold for liquids since only gases can be assumed to be
spherical particles that are completely elastic upon collision with each other. Since such derivation does
not apply for liquids, and no other theory for diffusion in liquids is yet well established, only semi-empirical
equations have been used to estimate diffusivities of solutes in liquids (Geankoplis, 2003).
The only theoretical derivation that predicts diffusivity of a solute (A) in a liquid (B) accurately is for very
large solutes only, and it was obtained starting from the Stokes-Einstein equation and making few
assumptions, leading to the following (Geankoplis, 2003):

9.96 ∙ 𝑇 ∙ 10−16
𝐷𝐴𝐵 = 1/3 Eq. 8
𝜇 ∙ 𝑉𝐴

where DAB is diffusivity in m2/s of solute A in liquid B, μ is dynamic viscosity of the solution in Pa.s, T is
temperature in K, and VA is the solute molar volume at its normal boiling point in m3/kg mol. This equation
is not accurate for smaller solutes that have a molecular weight of less than 1000 g/mol or a molar volume
less than 0.5 m3/kg mol.
One of the most accurate and robust semi-empirical equation is the Wilke-Chang correlation:

1.173 ∙ 𝑇 ∙ 10−16 ∙ (𝜃 ∙ 𝑀𝐵 )0.5

𝐷𝐴𝐵 = Eq. 9
𝜇𝐵 ∙ 𝑉𝐴0.6

where MB is the molecular weight and θ is an “association parameter”, which is 2.6 for water and 1.0 for
a typical organic solvent (Geankoplis, 2003).
Since the Wilke-Chang correlation predicts diffusivities quite accurately for most cases around room
temperature, it is therefore our default equation to estimate the diffusivity for the system involved in the
liquid-liquid extraction process, unless the solutes are very large molecules or high temperatures, in which
we would use Equation 8.
It is important to mention that the diffusivity constant is pairwise in a multi-species system, so if we
consider molecular diffusion in liquid-liquid extraction, there are two diffusivity constants that play a role:
the diffusivity of the solute in the original feed phase (DAB) and the diffusivity of the solute in the extracting
phase (DAC). The latter is also important because the rate of diffusion of the solute in the extracting phase
also plays a role in how fast the solute leaves the original feed phase. In fact, if the solute diffuses very
slowly in the extracting phase close to the interface between the two phases because of a low diffusivity
in the extracting phase, the concentration gradient (which drives diffusion in a regular mass transfer
process) at the interface would take more time to reach the same level it would if the diffusivity was
higher, therefore slowing the entire mass transfer process down. In other words, if the extracted solute
takes more time to “leave” the area close to the interface and diffuse in the bulk extracting phase, the
solute gradient concentration at the interface which drives mass transfer is less sharp/efficient, leading
to an increase in time required for more solute to leave the original phase, therefore requiring more time
to achieve chemical equilibrium.
In sophisticated diffusion problems such as this one, the regular molecular diffusivity is usually substituted
by a dispersion coefficient or effective diffusivity, in which factors such as turbulence, dispersion, and
tortuosity are included to accurately determine the effective diffusivity for such system. Although it would
be very complicated to model in our case, it is clear that both diffusivities contribute to the overall
effective diffusivity associated with a given extraction system. For our practical purposes and
simplifications during the estimation of the time required to achieve equilibrium in a liquid-liquid
extraction process, we defined the effective diffusivity (DABC) of the extraction system to be a linear
combination of the two estimated diffusivities (DAB and DAC), meaning that the effective diffusivity lies
somewhere between the two. A good first prediction would be to use the weighted average diffusivity
based on the two phase flowrates in the system. In the case of a 1:1 flowrate ratio between the phases,
the estimated effective diffusivity (DABC) would then be simply the average of the two estimated
diffusivities (DAB and DAC). It is important to mention that a weighted average diffusivity is a reasonable
and fair estimation for mass transfer sizing purposes, although it might not be the most accurate
estimation method for certain systems.
With those simplifications and by neglecting other phenomena such as dispersion and tortuosity, it is thus
possible to approximately estimate the effective diffusivity in liquid-liquid extraction in an easy way.
Later on in this paper, we provide a simple tool to estimate the effective diffusivity value for any liquid-
liquid extraction system.

Static Mixer Particle Size estimation

In our studies, we decided to focus our attention on the Kenics static mixer which is very popular and the
most investigated mixer for liquid-liquid emulsification (Middleman, 1974; Chen 1978; Haas 1987;
Berkman, 1988; Yamamoto, 2007;).
The mean particle size in Kenics static mixers, usually defined as the mean Sauter Diameter (D32) which is
the ratio of volume and surface area, can be predicted using available semi-empirical correlations which
the majority have been based on Kolmogoroff’s theory of turbulence (Hinze 1955, 1959). One of the
limitations of those correlations is the fact that this theory does not apply well with static mixers because
it assumes a flow field that should be both homogenous and isotropic. However, starting from this theory
and taking into consideration the Reynolds and Weber numbers, Middleman (1974) proposed a
correlation to estimate the Sauter mean diameter of a droplet in a Kenics mixer. After him, different
authors have proposed similar correlations, and those are reported in Table 1 (see Author’s article for
more details and variables explanation).

Table 1. Correlations for the prediction of droplet mean diameter (D32) resulting from emulsification in Kenics static
mixers with their range of applicability in terms of Reynolds and Weber number. The tube diameter range used in
their experiments is also included.
Authors Correlation We Re D (cm)
Middleman 𝐷32
= 𝐾𝑊𝑒 −0.6 𝑅𝑒 0.1 ~5-300 ~103-104 ~1.22-2.54
(1974) 𝐷
Chen and Libby 𝐷32 𝜇𝑑 0.18
= 1.14𝑊𝑒 −0.75 ( ) NA NA NA
(1978) 𝐷 𝜇𝑐
𝐷32 𝜇𝑑 0.5
Haas (1987) = 1.2𝑊𝑒 −0.65 𝑅𝑒 −0.2 ( ) 5-236 184-8090 0.335-0.794
𝐷 𝜇𝑐
Berkman and 𝐷32 𝐷32 0.33 0.6 144-
= 0.49𝑊𝑒 −0.6 (1 + 1.38𝑉𝑖 ( ) ) >12000 1.91
Calabrese (1988) 𝐷 𝐷 665
Model Results

In this study, we investigated the liquid-liquid extraction of acetone from water using toluene with a 1:1
volume ratio between the two phases. The diffusivity of acetone in water at 25 ᵒC was estimated using
Eq. 2 to be 1.28*10^-9 m2/s, while the diffusivity in toluene was 2.86*10^-9 m2/s. Therefore, the effective
diffusivity of this LLE system was taken as 2.07*10^-9 m2/s (average of the two due to 1:1 volume ratio).
In our study, we used the correlation developed by Haas (1987) in Table 1 to predict the mean Sauter
diameter of our droplets in order to have an idea on the mean radius, which is considered to be our
characteristic diffusion length in our mass transfer model. The reason we picked this correlation was
because our range of Weber and Reynolds number, as well as tube diameter lies the best within its range
of applicability compared to the other correlations (Table 1). Droplet size mean diameters were predicted
for two different tube diameters of interest (1/8” and 3/8”) at different flowrates (from 20 ml/min to 3000
ml/min), and the results are shown in Table 2. Water was considered to be the continuous phase, while
toluene was the dispersed phase as suggested by Ribeiro (2006) and others. Predicted droplet diameters
inside the range of applicability of the correlation falls within the range obtained by Haas (1987) in his
experiments (40-640 μm).

Table 2. Predicted mean droplet size diameter using Kenics static mixer in two different tubes at different flowrates
using the correlation developed by Haas (1987). N.A. stands for “not applicable” and it was used if either Re or We
were outside the range of applicability of the correlation (Table 1).
Tube Inside
Flowrate Diameter Re We D32 (μm) Notes
(ml/min) (inches)
20 1/8" 150 0.1 3845 NA
40 1/8" 300 0.6 1359 NA
80 1/8" 600 2.4 481 NA
120 1/8" 899 5.3 262
150 1/8" 1124 8.3 187
200 1/8" 1499 14.8 122
300 3/8" 749 1.2 2107 NA
500 3/8" 1249 3.4 979 NA
1000 3/8" 2498 13.7 346
1500 3/8" 3747 30.8 188
2000 3/8" 4996 54.7 122
2500 3/8" 6246 85.5 88
3000 3/8" 7495 123.2 67

Concentration profiles inside the spherical droplet and equilibrium time were obtained for this system by
applying our mathematical model described earlier, using the effective diffusivity and varying droplet size
diameter. Equilibrium time was calculated based on the time required to have a concentration above
99.9% of the equilibrium concentration at every radial position throughout the sphere. For example, for
a 100 μm droplet, it roughly takes 1 second to achieve equilibrium throughout the sphere as shown in
Figure 3.
The effect of droplet size and diffusivity was obtained by solving the concentration profiles with different
conditions within a reasonable droplet size and diffusivity range of values (less than 200 μm droplets are
typical when using static mixers, and diffusivity in liquids is usually between 0.5*10-9 and 4*10-9). As it was
done earlier, the equilibrium time was calculated based on the time required to achieve a uniform
concentration above 99.9% of the equilibrium concentration at any radial position within the droplet.
These results are shown in Figure 4.

Figure 3. Dimensionless concentration profiles as a function of dimensionless radius inside a 100 μm sphere after
different time intervals. Equilibrium is reached when the value of Z (x-axis) is above 0.999 at any radial position.

Figure 4. Estimated equilibrium time as a function of effective diffusivity and droplet size diameter in a continuous
LLE process. The white dot corresponds to the studied system (acetone-water-toluene) with a droplet diameter of
100 μm, leading to an equilibrium time of 1.0 sec.
With our model, it is thus possible to rapidly approximate the time required to achieve equilibrium in any
continuous liquid-liquid extraction process happening in a tube. This provides a quick approximation when
setting up a continuous liquid-liquid extraction process by giving an estimate of the tube length required
to achieve equilibrium at a given flowrate and tube diameter for any solvent system. For example, in our
studied system, times and tube length requirement at the investigated flowrates are reported in Table 3
(only for the cases in which the correlation was applicable).

Table 3. Equilibrium times and tube length requirement for a water-acetone-toluene extraction system using Kenics
static mixers at different flowrates for two different tubes.

Tube Inside Diameter Model Eq. Tube Length

Flowrate D32 (μm)
(inches) Time (s) Required (cm)
(ml/min)
120 1/8" 262 6.4 162
150 1/8" 187 3.4 107
200 1/8" 122 1.4 59
1000 3/8" 346 11.2 262
1500 3/8" 188 3.4 119
2000 3/8" 122 1.4 65
2500 3/8" 88 0.8 47
3000 3/8" 67 0.6 42

Experimental Validation

In our experimental setup, we used a 1:40 and 1:10 acetone/water solution by volume and toluene as our
extracting solvent. The two solutions were pumped at the same flowrate and merged into a single tube
containing Kenics static mixer followed by an empty tube. At the end of the tube, the aqueous and the
organic phases were separated using a Zaiput membrane based separator. A sample from both phases
was collected at the outlets of the separator and analyzed using HPLC. The goal of this experimental study
was to find out the residence time required to achieve equilibrium before separation. Multiple trials were
performed by varying the length of the empty tube (residence time) as well as the flowrate of the two
phases. The HPLC collected samples were used to calculate the partition coefficient of acetone in the two
phases, and this was compared to the equilibrium partition coefficient (KD,EQ = 0.65 when 1:40 acetone-
water ratio and 0.70 when 1:10 ratio) which was obtained by a traditional batch experiment (manually
mixed the system by strong shaking in a vial and allowed time to settle by gravity, then HPLC on both
layers).
The results of these experiments are shown in Table 4 (ID: 3/8”) and in Table 5 (ID: 1/8”), and the
comparison between experimental and model prediction is summarized in Table 6.
Table 4. Summary of the experiments performed with the acetone-water-toluene LLE system, using 3/8” ID tubing
with 32” of Kenics static mixer followed by empty tubing.
Toluene Tube
Water/Acetone Water/Acetone Residence KD
Flowrate Length
Ratio Flowrate (ml/min) Time (s) (O/A)
(ml/min) (cm)
40:1 250 250 180 12.8 0.6
40:1 500 500 180 7.7 0.63

Table 5. Summary of the experiments performed with the acetone-water-toluene LLE system, using 1/8” ID tubing
with 12” of Kenics static mixer followed by empty tubing.
Toluene Tube
Water/Acetone Water/Acetone Residence KD
Flowrate Length
Ratio Flowrate (ml/min) Time (s) (O/A)
(ml/min) (cm)
40:1 20 20 30 3.6 0.59
40:1 80 80 30 0.9 0.65
40:1 20 20 150 17.8 0.65
40:1 80 80 150 4.5 0.65
40:1 50 50 30 1.4 0.64
10:1 20 20 30 3.6 0.63
10:1 40 40 30 1.8 0.66
10:1 60 60 30 1.2 0.65
10:1 80 80 30 0.9 0.54
10:1 100 100 30 0.7 0.55
10:1 20 20 100 11.9 0.65
10:1 40 40 100 5.9 0.70
10:1 70 70 100 3.4 0.70
10:1 80 80 100 3 0.68
10:1 100 100 100 2.4 0.67
Table 6. Comparison of model predicted to experimental equilibrium time for water-acetone-toluene L-L extraction
in a tube (ID: 1/8” until 200 ml/min, 3/8” above 200 ml/min) using Kenics static mixers. N.A. stands for “not
applicable” and it was used if either Re or We were outside the range of applicability of the correlation (Table 1).

Experimental
Flowrate Model Eq. Time (s)
Eq. Time (s)
(ml/min)
40 NA 17.8
80 NA 5.9
120 6.4 4
150 3.4 ~1-3.4
200 1.4 ~1-3
500 NA > 13
1000 11.2 ~8-9
1500 3.4 NA
2000 1.4 NA
2500 0.8 NA
3000 0.6 NA

Recommendations and tube length selection charts

The goal of this section is to use engineering judgment combined with the findings from this study to come
up with easy to consult charts for the selection of the length of tubing required to successfully carry out
any continuous LLE using our proposed equipment system.
A typical unit of Kenics static mixer usually consists of 10 to 12 elements. Although Middleman (1974)
shows that a unit of 10 elements of Kenics static mixer is sufficient to reach the terminal mean droplet
size distribution, Haas (1987) and Berkman (1988) used in their experiments units of 21 and 24 elements,
respectively, because they suggest higher viscosities require more time to reach such distribution. A
longer static mixer portion would increase the chances of reaching terminal droplet size distribution,
increase mixing and contact of the two phases, even for very challenging or viscous systems, as well as
slowing down the coalescence rate in the empty tubing portion. When considering mass transfer during
the LLE process, coalescence after the mixer strongly reduces mass transfer because droplet size has a
quadratic effect on equilibrium time. This leads to the proposition for alternating static mixers units and
empty tubing to allow less time for coalescence due to repetitive mixing, thus allowing the system to be
closer to the terminal droplet size distribution throughout the tubing which enhances mass transfer
(smaller droplets). It is for all these reasons that we suggest using multiple units of Kenics static mixer for
LLE application, especially at low diameters.
For flowrates up to 200 ml/min, we recommend using our 2’ tubing unit of 1/8” or 5/32” ID tubing (OD:
1/4”). The tubing unit contains roughly 4-5” of Kenics static mixers (StamixCo: 4 units of HT-40-3.18-12-‘
[PP or HT-40-3.18-12-AC, or 3 units of HT-40-3.75-10-PTFE) followed by roughly 19-20” of empty tubing
before entering into the membrane based separator (Zaiput SEP-200). For more challenging extractions,
especially when the effective diffusivity is low, we recommend using multiple units of this tubing in order
to achieve equilibrium (see Figure 5 for recommended number of tubing units).
Above 200 ml/min until 3000 ml/min, we recommend using our 3’ tubing unit of 3/8” ID tubing (OD: 1/2”).
The tubing unit contains roughly 8” of Kenics static mixers (StamixCo: 1 unit of HT-40-9.47-24-PP or HT-
40-9.47-24-AC, or HT-40-9.30-20-PTFE) followed by roughly 28” of empty tubing before entering into the
membrane based separator (Zaiput SEP-3000). As it was for flowrates below 200 ml/min, we recommend
using multiple units of this this tubing in order to achieve equilibrium (see Figure 6 for recommended
number of tubing units).
With those recommendations, the following charts (Figure 3-8) represents easy to consult tools for an
approximation of the length of 1/8” ID, 5/32” or 3/8” ID tubing required to achieve equilibrium for any
LLE system given effective diffusivity (see estimation section if needed) and flowrate (Figure 5 - 6).
When developing those charts, a safety factor was added to the length required in order to increase the
approximation’s robustness, making it more likely to be applicable to a large number of systems. It is
important to mention that the below charts represent a great guideline for sizing a continuous liquid-
liquid extraction system such as the one we propose. However, tubing length could be adjusted after
experimentation with the suggested length in order to lower the dead volume (if equilibrium is reached
with the suggested length) or to improve extraction (if residence time was not enough to reach
equilibrium).

Figure 5. Approximate tubing length (ID=1/8” or 5/32”) required to achieve equilibrium, depending on flowrate (20-
200 ml/min) and effective diffusivity (DABC) range, using Zaiput SEP-200. (If needed, consult the following diffusivity
estimator section to obtain effective diffusivity).
 Figure 6. Approximate tubing length (ID=3/8”) required to achieve equilibrium, depending on flowrate (200-3000
ml/min) and effective diffusivity (DABC) range, using Zaiput SEP-3000. (If needed, consult diffusivity estimation
section to obtain effective diffusivity).

Effective Diffusivity Estimator

The goal of this section is to guide in the estimate of the effective diffusivity for any liquid-liquid extraction
system by providing easy to consult charts. As a reminder, the effective diffusivity (DABC) is the weighted
average (based on flowrate ratio) of the diffusivity of the solute in solvent B (D AB) and the one in solvent
C (DAC), therefore both need to be estimated first. To start, we can estimate the role of the molar mass of
the solvent B (or C) and the molar volume of the solute A by extracting two values of K (dimensionless
variable) from Figure 7: one using the molar mass of solvent B and one using the one for solvent C. Once
K values are known, they can be used with the viscosity of the corresponding solvent (B or C) to extract
the two values of DAB and DAC from Figure 8. If the value of K is outside the chart (Figure 8) for a given
viscosity, that means the diffusivity is very high above 5*10-9 m2/s. Also, it is important to keep in mind
that if the extraction does not happen at room temperature (298 K), the two values of diffusivities should
be adjusted by multiplying them by the ratio of the actual temperature (in Kelvin) and room temperature
(298 K). Furthermore, if the solvent is water, the approximated diffusivity for the solute in water should
be multiplied by 2.6.
At this stage, after adjusting for temperature and for water, the only thing left is to calculate the effective
diffusivity (DABC) by computing the weighted average (based on flowrate ratio) of the approximated values
of DAB and DAC.
 Figure 7. Approximation of the value of K as a function of the molar mass of the solvent B or C and the molar
volume of the solute A. Molar volume is the ratio between molar mass and density.

Figure 8. Approximation of the value of the diffusivity as a function of the previous obtained K value and the
viscosity for Solvent B or C. The system effective diffusivity can be obtained by the weighted average of the two
extracted values of diffusivity (DAB and DAC).
Further Considerations and Conclusions

The ability to accurately model mass transfer for in-line mixing solves one of the key barriers to the
accurate design of continuous extractive systems. With these models in place, which led to the
development of easy to use consultation charts for sizing mass transfer in order to reach equilibrium, and
the before mentioned continuous membrane separators from Zaiput Flow Technologies, continuous
extractive platforms, which have several advantages compared to more traditional approaches, can be
easily constructed.
In this work, we have shown that this model is not only a powerful tool, but it is also easily scalable across
a range of flow rates and tubing diameters, and applicable to nearly all extractive systems. Future work
can be done to further validate our model and apply the theory at even larger flow rates.

References

Berkman, P. D., & Calabrese, R. V. (1988). Dispersion of viscous liquids by turbulent flow in a static
mixer. AIChE Journal, 34(4), 602-609.
Chen, S. J., Libby, D. R. (1978). Gas-liquid and liquid-liquid dispersions in a Kenics mixer. 71st Annual AIChe
Meeting
Cybulski, A., & Werner, K. (1986). Static mixers-criteria for applications and selection. Int. Chem.
Eng, 26(1), 171-180.
Deen, W. M. (2011). Analysis of transport phenomena. Oxford university Press, 2nd edition
Genkoplis, C. J. (2003). Transport processes and separation process principles. Pearson, 4th edition
Grace, C. D. (1971). Static mixing and heat transfer. Chemical and Process Engineering, 52(7), 57.
Haas, P. A. (1987). Turbulent dispersion of aqueous drops in organic liquids. AIChE journal, 33(6), 987-995.
Hinze, J. O. (1955). Fundamentals of the hydrodynamic mechanism of splitting in dispersion
processes. AIChE journal, 1(3), 289-295.
Hinze, J. O. (1959). Turbulence McGraw-Hill. New York, 421, 3.
Legrand, J., Morançais, P., & Carnelle, G. (2001). Liquid-liquid dispersion in an SMX-Sulzer static
mixer. Chemical Engineering Research and Design, 79(8), 949-956.
Middleman, S. (1974). Drop size distributions produced by turbulent pipe flow of immiscible fluids through
a static mixer. Industrial & Engineering Chemistry Process Design and Development, 13(1), 78-83.
Mutsakis, M. (1986). Advances in static mixing technology. Chemical engineering progress, 82(7), 42-48.
Myers, K. J., Bakker, A., & Ryan, D. (1997). Avoid agitation by selecting static mixers. Chemical engineering
progress, 93(6), 28-38.
Ribeiro, M. M. M., Gonçalves, C., Regueiras, P. F., Guimaraes, M. M. L., & Pinto, J. C. (2006). Measurements
of toluene–water dispersions hold-up using a non-invasive ultrasonic technique. Chemical
Engineering Journal, 118(1-2), 47-54.
Streiff, F. A. (1977). In-line dispersion and mass transfer using static mixing equipment. Sulzer Technical
Review, 3, 108-114.
Streiff, F. A., Mathys, P., & Fischer, T. U. (1997). New fundamentals for liquid–liquid dispersion using static
mixers. Récents progrés en Génie des procédés, 11(51), 307-314.
Thakur, R. K., Vial, C., Nigam, K. D. P., Nauman, E. B., & Djelveh, G. (2003). Static mixers in the process
industries—a review. Chemical engineering research and design, 81(7), 787-826.
Theron, F., & Le Sauze, N. (2011). Comparison between three static mixers for emulsification in turbulent
flow. International journal of multiphase flow, 37(5), 488-500.
Williams, M. M. R. (1975). The mathematics of diffusion: By J. Crank. Clarendon Press, Oxford, £ 13.50,
1975, 414 pp.
Yamamoto, T., Kawasaki, H., & Kumazawa, H. (2007). Relationship between the dispersed droplet
diameter and the mean power input for emulsification in three different types of motionless
mixers. Journal of Chemical Engineering of Japan, 40(8), 673-678.