The separation of hexamethylenediamine from aqueous

solution by solvent extraction with butanol

Michael J. Sargent a, Nigel K. H. Slater a, J. S. Dennis a,*, Gary J. Smith b, Paul S. Pearlman b
a Department of Chemical Engineering and Biotechnology, University of Cambridge, West Cambridge Site,

Cambridge, UK.
b INVISTA Textiles (U.K.) Limited, Redcar, UK

* Corresponding author.

Email address: [email protected] (J. S. Dennis)

Abstract

This paper details an investigation into the separation of hexamethylenediamine (HMD) from dilute aqueous solutions

by solvent extraction. HMD is a monomer used in the production of Nylon-6,6 and as such this separation is potentially

useful in reclaiming HMD from waste generated in the production of nylon. This separation is potentially also

important in extracting HMD produced from biological sources or reclaimed from waste emanating from the

depolymerization of nylon during recycling. Butanol was identified as a suitable solvent for the preferential

dissolution of HMD from an aqueous solution, and it was found that the addition of sodium hydroxide resulted in

improved separation. A model was derived to predict the partitioning of HMD between the aqueous and organic

phases, and this was used to investigate the operation of a simple countercurrent solvent extraction process.

Keywords

Amine

Solvent extraction

Butanol

Hexamethylenediamine

Electronic copy available at: https://ssrn.com/abstract=3979466

1. Introduction

Hexamethylenediamine, hereinafter referred to as HMD, is an unbranched 6 carbon chain, terminated at each end by

an amine group. It is used in the production of nylon polymers, most commonly Nylon-6,6, in which it is reacted with

adipic acid. Due to the relative monetary values of HMD and adipic acid, this reaction is done with a stoichiometric

excess of HMD, resulting in an aqueous waste which contains a low concentration of HMD [1, 2].

In the interests of reducing the waste associated with the production of nylon polymers, methods for the recovery and

reuse of low concentrations of HMD in aqueous solution were investigated. The nature of HMD as an amine allows

for methods of separation exploiting the variable charge on the molecule: depending on the pH of the solvent, HMD

can be either highly charged and hydrophilic, or uncharged and hydrophobic. This paper builds upon previous work

which investigated methods of extracting HMD using pH and salt to cause a secondary phase to be generated [3]. The

present work includes the addition of an organic solvent, which is immiscible with water, into which the HMD is

preferentially extracted.

The study of amines in the field of solvent extraction is largely concerned with using the amine as the secondary

solvent with which to extract a solute from a primary solvent. Long chain amines with low water solubility have been

used for a wide range of solutes including rare earth mineral ions [4-6] and organic acids [7-9]. However, the extraction

of amines as the solute has been much less widely studied, although research has been undertaken in which a

membrane has been used to separate the liquid phases [10], or a dye has been reacted with the amine prior to extraction

[11].

This paper details an experimental campaign which used solvents to extract HMD from a dilute aqueous mixture, to

understand the thermodynamics of the phase equilibrium. The resulting thermodynamic data were used to construct a

model which could be used to predict the performance of single-stage or multi-stage solvent extraction.

2. Materials and Methods

As a diamine, HMD can exist in three different valence states, depending on the pH of the solution. Figure 1 shows

the fraction of HMD present in each of these states, using the following values of the dissociation constants: pKa,1 =

10.7 and pKa,2 = 11.02 [12].

Electronic copy available at: https://ssrn.com/abstract=3979466

 Fig. 1. Fraction of HMD in each charge valence state as a function of pH.

Under neutral to mildly basic conditions, HMD is divalent and highly hydrophilic. However, if the pH is raised high

enough, the HMD loses its charge and becomes nonvalent, increasing its affinity for hydrophobic solvents. The pH

can be raised by the addition of basic compounds, of which sodium hydroxide was primarily used in this work.

However, as pH is a measure of the activity of H+ ions in solution, the pH is also affected by what percentage of the

solvent is water. Pure water has a self ionisation constant of pKw = 14 at 25°C for the reaction:

H2O ⇌ H + + OH -

(1)

This means that pure water will naturally dissociate and equilibrate at pH 7. However, any other species present in

sufficient concentration will alter this pH, even if it does not itself dissociate, by the fact that it will lower the overall

molar concentration of water [13]. Thus, a solvent which is partially water and partially organic will naturally have a

pH higher than 7, owing to the reduced availability of water molecules which are then able to dissociate.

2.1. Sample Preparation

Solvent extraction is fundamentally about the equilibrium between concentrations of the same species in multiple

phases. Each component present in the system is capable of diffusing between the different phases and dissolving until

its chemical potential within each phase is the same, and there is no more thermodynamic driving force for mass

transfer. Whilst solvent extraction can be performed without waiting for complete equilibrium to be reached, this work

will focus on the equilibrium, as it sets an upper bound for the possible transfer of HMD from an aqueous phase to an

organic solvent.

Electronic copy available at: https://ssrn.com/abstract=3979466

The Gibbs phase rule indicates that in a system with 3 components (C = 3) and two phases (P = 2), there are F = C -

P + 2 = 3 degrees of freedom. Assuming an isothermal and isobaric system, this reduces to a single degree of freedom.

Therefore, if the mass fraction of any one component in any one phase changes, all the other mass fractions must also

change if equilibrium is to be maintained. The total mass fraction of HMD was the variable altered, and so the

compositions of both phases resulting from the equilibrium depends on this.

To this end, samples were made which contained different concentrations of HMD, water, and solvent, as well as

NaOH. The NaOH was added as a further method of controlling the speciation of HMD, and was included as a number

of molar equivalents to the HMD. At 1 equivalent, there was 1 mole of NaOH for every mole of HMD present,

regardless of the overall concentration of either compound. This method was chosen because at 2 equivalents, the

amount of OH- produced by the dissociation of the NaOH exactly matches the amount of H+ produced by the

conversion from the divalent speciation of the HMD into its non-valent state. Therefore, this relative method of

quantifying NaOH gives an indication of the speciation state of the HMD present, although the actual speciation will

be affected by all the components in the system.

To control accurately the total concentrations of all components present in the system, stock solutions were prepared

with carefully controlled concentrations. By pipetting known volumes of these stock solutions together, the overall

composition of a two-phase mixture was determined. This information could then be used to improve the precision of

the concentration measurements for each individual phase. All chemicals used for these solutions were from Sigma

Aldrich.

Four stock HMD solutions were prepared, each using 101±1 g of 99% purity HMD, dissolved into deionized water in

a 500±1 mL volumetric flask, resulting in HMD concentrations of 200±1.4 g L-1. Each of these solutions contained 0,

1, 2, or 3 equivalents of NaOH, by the addition of 35±1 g of 99% purity NaOH per equivalent to the volumetric flask

prior to the addition of the deionised water. The aqueous solutions used to calculate the solvent extraction equilibria

were made by diluting these stock solutions with deionised water to produce HMD concentrations ranging from 20 to

200 g L-1, while maintaining the number of NaOH equivalents.

The aqueous solutions used in each determination of phase separation contained all the HMD and NaOH which would

be present, and so the organic solutions were pure solvent. Generally, 5±0.01 mL of the aqueous solution was

combined with 5±0.01 mL of the solvent, because these quantities were found to be large enough for accurate

Electronic copy available at: https://ssrn.com/abstract=3979466

measurement of the volumes of the resulting phases. It was found that mixing the solutions by shaking them manually

for a few seconds was sufficient to ensure equilibrium was reached, as the results for any sample were identical

between manual mixing and the use of a mechanical shaker for several minutes. Similarly, the two resulting phases

spontaneously separated in under 30 seconds after being left to stand without the need for centrifugation, and the use

of a centrifuge did not alter any of the resulting measurements. The two phases were then separated into different

containers for independent measurements.

Since the phase separation is governed by thermodynamics, it would be expected that the temperature at which the

mixing and separation takes place would influence the compositions in each phase. Two temperatures were

investigated, 22°C and 37°C; however, no significant differences were detected in the phase compositions between

these two temperatures. Therefore in this work, both sets of results will be presented as temperature independent over

the range 22 to 37°C.

2.2. Measurements

The method of sample preparation allowed for the total volume and the total concentration of each component in the

mixture to be accurately known. After the two phases were mixed and settled, this same information was required for

each of the resulting phases. The volume of each phase was accurately measured using image analysis of the position

of the meniscus in the container holding the phases. This allowed volume measurements with an accuracy of 0.1 mL.

The mass concentrations of HMD and the solvents were determined using an Agilent 6850 series gas chromatograph,

fitted with a CP-Sil 8 CB for Amines column and an FID for detection. The internal standard used for concentration

calibration was composed of 10 mL of hexylamine and 40 g of NaOH, dissolved in 1 L of an equal volume mixture

of water and methanol. The hexylamine provided the standard peak, and the NaOH served to ensure that the HMD

present was pushed into its non-valent state, because the various valent states of HMD have inconsistent evaporation

behaviour within the gas chromatograph. The use of the mixture of water and methanol as the solvent for the internal

standard allowed it to be used for both aqueous and organic samples.

HMD and all the solvents used were each given a multipoint calibration curve using this single internal standard. A

peculiarity in the measurement of HMD meant that concentrations of less than 10 g L-1 HMD showed no signal.

However, this was only relevant for a few of the aqueous phases, and values for the HMD concentration in these

Electronic copy available at: https://ssrn.com/abstract=3979466

phases could be calculated by a mass balance from the total HMD present in the system and the measured

concentration in the corresponding organic phase.

The concentration of NaOH was first measured using a Mettler Toledo PerfectION™ sodium-selective ion probe.

This technically measures the sodium ion concentration, the only sodium present in the system must come from NaOH,

and so the two measurements are synonymous. However, it was found that for all the organic phases measured, the

NaOH concentration was below 1 mg L-1. Therefore, the simplification was made that the organic phase completely

rejected the NaOH, so that it all remained in the aqueous phase.

The final determination of concentration required was that of water in each of the phases. Whilst it could not be

measured directly, the concentration of water was calculated from knowing the concentrations of the other 3

components in each of the phases. So far in this work, the amount of each component in each phase has been expressed

in terms of mass concentrations, measured in g L-1. This has been used because all the sample preparation and

measurements utilised known volumes of the solutions. In the analysis of the phase equilibria, it becomes useful to

switch from using mass concentrations to mass fractions and the total mass of each phase, as mass is conserved

whereas volume is not. Calculating the mass fractions from the mass concentrations required the density of each

component as a pure phase, and the assumption that the volume change on mixing is negligible.

3. Experimental results

3.1. Solvent selection

The first experimental step was to determine which solvent should be used for the most efficient extraction of HMD

from the aqueous phase. To this end, preliminary experiments were performed with a range of solvents. Each solvent

was mixed in a 1:1 volumetric ratio with four different aqueous solutions, containing either 50 or 90 g L-1 HMD, and

either 0 or 3 equivalents of NaOH. For these preliminary experiments, the characteristic by which the solvents were

measured was the partitioning coefficient, taken as the ratio of the mass fraction in the organic phase to that in the

aqueous phase. The results are shown in Fig. 2.

Electronic copy available at: https://ssrn.com/abstract=3979466

Fig. 2. Five solvents compared for their effectiveness in extracting HMD from aqueous solution. Partitioning

coefficient is the ratio of mass fractions in the organic phase to that in the aqueous phase.

The candidate solvents were divided into two types, polar hydrocarbons and HMD analogues. The unbranched

alcohols n-butanol, n-pentanol, and n-hexanol were investigated because of their polar nature being compatible with

that of the amine groups in HMD [14]. A trend in alcohol chain length is a reduced solubility in water [15], which is

beneficial for solvent extraction as it results in less solvent present in the aqueous phase, and vice versa. However,

increasing the chain length of the alcohol is seen from Fig. 2 to give a decreasing partitioning coefficient, meaning

that HMD is also less soluble in longer-chain alcohols.

The HMD-analogues n-hexene and 1,5-hexadiene were examined because of the similarity of their molecular size

and functional group location to HMD. It was hoped that the similar structure would result in an increased solubility

of HMD in these solvents, just as n-butene has been shown to be a good solvent for the extraction of aqueous n-

butanol [16]. However, these analogues had a far lower partitioning coefficient than the alcohols, and so were

rejected.

The solvent chosen for further investigation was n-butanol, because of its very high partitioning coefficient in the

presence of large equivalents of NaOH. The maximum solubility of HMD in butanol was determined by dissolving

solid 99.9 wt% HMD in pure butanol, and was found to be 78 wt%. In passing, the trend in solubility with

Electronic copy available at: https://ssrn.com/abstract=3979466

decreasing length of chain would imply that n-propanol should have a higher partitioning coefficient, but it is fully

miscible with water at temperatures above -1.7°C [15], and so cannot be used for solvent extraction.

3.2. Ternary data

In a generic system with 3 components which is not necessarily in equilibrium, the mass fractions of any 2 of the

components can be altered, and the mass fraction of the third component must change accordingly to keep the total of

all the mass fractions at 100%. This means that the composition of such a mixture can be expressed in a ternary

diagram, as shown in Fig. 3.

Fig. 3. Ternary diagram at zero equivalents. A three-component system of water, butanol, and HMD. Solutions

which are in equilibrium with one another are linked by tie lines. The uncertainties in the mass fractions are

smaller than the marks.

Figure 3 shows a ternary plot for the system containing water, butanol, and HMD. It contains results for the phase

equilibrium of six samples, which did not include any NaOH, and so corresponds to zero equivalents. For each sample

there are two marks on the figure, corresponding to the compositions of the aqueous and organic phases in equilibrium

with one another, each linked by a tie line. The lower left corner of the figure corresponds to pure water, and so the

marks at the left ends of the tie lines are the aqueous phases, with the marks at the right ends of the tie lines being the

organic phases. The uncertainties in the mass fractions are smaller than the marks.

Electronic copy available at: https://ssrn.com/abstract=3979466

The first sample, shown by the red squares, is the mutual miscibility of water and butanol in the absence of any HMD.

This miscibility lies at 7.3 wt% of butanol in water, and 18.0 wt% of water in butanol [15]. The next 3 samples, shown

by black circles, were produced with overall HMD contents of approximately 3 wt%, 6 wt%, and 9 wt%. These

samples also show incomplete miscibility, and separate into two phases, each of which contains all three components

in various quantities. As a result of increasing the amount of HMD present, the mutual miscibility of the solvents

increased, as shown by the tie lines getting shorter as their vertical position increases.

The area covered by the tie lines for these samples corresponds to the phase gap region, the mixtures which will

separate into two phases. The last 2 samples shown in Fig. 3, with HMD content of above 11 wt%, show complete

miscibility; they did not form a two-phase mixture and instead the presence of the HMD served to allow the water and

butanol to mix. In the absence of any NaOH to alter the pH of the samples, the nature of the HMD as an organic, yet

highly-charged, molecule serves to bridge the gap between the two solvents.

Figure 4 shows the same axes as Fig. 3, with results for samples made with two equivalents of NaOH. Therefore, there

are four components present in the system, and so the mass fraction of NaOH is included in the axis for the mass

fraction of water. The highest mass fraction of NaOH in any of the aqueous phases was 3.6 wt%, which is a similar

magnitude to the uncertainty in the mass fraction of the water mass itself. As such, it is not unreasonable to combine

these two mass fractions so as to enable the results to be plotted on these ternary axes.

By comparing Figs. 3 and 4, it can be seen that by including NaOH with the HMD, the bridging effect of the HMD is

removed. As the HMD content is increased, the tie lines stay the same length. The water content in the organic phase

remains at 15 - 20 wt%, and the butanol content in the aqueous phase falls to below 5 wt%. In addition, the partitioning

of HMD between the two phases becomes more favourable, with less of the HMD remaining in the aqueous phase.

Samples with higher HMD content were not made, because it is the phase separation behaviour for these dilute HMD

systems which was of most importance.

Electronic copy available at: https://ssrn.com/abstract=3979466

Fig. 4. Ternary diagram of mass fractions, at two equivalents. A four-component system of water, butanol, and

HMD, with two moles of NaOH present for every mole of HMD. Solutions in equilibrium with one another are

linked by tie lines. The uncertainties in the mass fractions are smaller than the marks.

4. Modelling the Equilibria

This experimental results give an understanding of how the 4-component system reaches equilibrium between aqueous

and organic phases. These results were then used to construct a model of the system, which could be used to investigate

how butanol might be used practically to extract dilute HMD from its aqueous phase.

Consider a single mixer settler as an equilibrium stage, as shown in Fig. 5. There will be a total combined feed mass

flowrate 𝐹, which will have mass fractions 𝑓𝑤𝑎𝑡, 𝑓𝑏𝑢𝑡, 𝑓𝐻𝑀𝐷, and 𝑓𝑁𝑎𝑂𝐻 for water, butanol, HMD, and NaOH,

respectively. If these variables are known, then only the equilibrium of the phase separation will determine the

outflows. There will be an aqueous phase with 4 unknown mass fractions 𝑥𝑤𝑎𝑡, 𝑥𝑏𝑢𝑡, 𝑥𝐻𝑀𝐷, and 𝑥𝑁𝑎𝑂𝐻, as well as an

unknown total mass flowrate 𝐴. There will also be an organic phase with its own separate mass fractions 𝑦𝑤𝑎𝑡, 𝑦𝑏𝑢𝑡,

𝑦𝐻𝑀𝐷, and 𝑦𝑁𝑎𝑂𝐻, and mass flowrate 𝑆.

10

Electronic copy available at: https://ssrn.com/abstract=3979466

 Fig. 5. A single mixer settler unit fed by a combined feed stream, with aqueous and organic outflows.

Thus, there are 10 unknown variables describing the system, and so 10 independent equations are required relating

these variables. The first four equations are the mass balances on each component, conserving the mass of each species,

and following the format:

𝐹 𝑓𝑖 = 𝐴 𝑥𝑖 + 𝑆 𝑦𝑖

(2)

Because the individual mass balances are used, the total mass balance 𝐹 = 𝐴 + 𝑆 is not independent. The next two

equations are the summation equations which require that the total of the mass fractions in each phase are 100%:

∑𝑦 𝑖 = 1

(3)

∑𝑥 𝑖 = 1

(4)

This leaves four remaining independent equations to solve this system of unknown variables, derived from the

equilibrium between the two phases. The first of these equilibrium equations is given simply by the rejection of NaOH

from the organic phase, so that:

𝑦𝑁𝑎𝑂𝐻 = 0

(5)

Therefore, there are three pieces of equilibrium information which must be expressed as equations in these variables

for a solution to be found. These were selected to be the two mutual miscibilities of the solvents, being the mass

11

Electronic copy available at: https://ssrn.com/abstract=3979466

fraction of water in the organic phase and the mass fraction of butanol in the aqueous phase, as well as the partitioning

coefficient.

4.1. Water in the organic phase

In the absence of any HMD or NaOH, the equilibrium between water and butanol allows for up to 18.0 wt% of water

in butanol before a two-phase system is created [15]. It has already been shown in section 3.2 that 𝑦𝑤𝑎𝑡 will increase

with HMD content in the absence of NaOH, yet, is unchanging with HMD content when there is sufficient NaOH

present.

Fig. 6. Water content in the organic phase, 𝒚𝒘𝒂𝒕. The dependence of the mass fraction of water in the organic

phase, 𝒚𝒘𝒂𝒕, on the total mass fractions of HMD, 𝒇𝑯𝑴𝑫, and NaOH, 𝒇𝑵𝒂𝑶𝑯. The experimental results were fitted

to a function which is linearly increasing in 𝒇𝑯𝑴𝑫, and exponentially decreasing in 𝒇𝑵𝒂𝑶𝑯.

Figure 6 contains the results of 50 samples prepared at a range of values of HMD content, as well as number of NaOH

equivalents, showing 𝑦𝑤𝑎𝑡 as a function of both 𝑓𝐻𝑀𝐷 and 𝑓𝑁𝑎𝑂𝐻. It can easily be seen that in the absence of NaOH,

and thus zero equivalents, there is a linear increase in 𝑦𝑤𝑎𝑡 as 𝑓𝐻𝑀𝐷 increases. However, at higher values of 𝑓𝑁𝑎𝑂𝐻,

𝑦𝑤𝑎𝑡 remains unchanged with 𝑓𝐻𝑀𝐷. This dependence is captured by the fitted function:

𝑦𝑤𝑎𝑡 = 0.1523 + (2.066exp ( ‒ 83.64 𝑓𝑁𝑎𝑂𝐻)) × 𝑓𝐻𝑀𝐷

(6)

This function has a linear dependence of 𝑦𝑤𝑎𝑡 on 𝑓𝐻𝑀𝐷; however, the gradient is exponentially decaying with 𝑓𝑁𝑎𝑂𝐻.

The fitted values imply that, in the absence of any NaOH, ywat will always be 15.23 wt%, regardless of the amount of

12

Electronic copy available at: https://ssrn.com/abstract=3979466

HMD present. This does not match the value of 18.0 wt% observed for the two-component mixture of butanol and

water, but is within the confidence bounds for this fit, which is ±3.5 wt%. The interpretation of this dependence is that

valent HMD present in the organic phase allows it to contain more water. However, when the HMD is pushed into its

non-valent state by the presence of NaOH, this effect is removed.

4.2. Butanol in the aqueous phase

In the absence of any HMD or NaOH, the equilibrium between water and butanol allows for up to 7.3 wt% of water

in butanol before a two-phase system is created, at 22 °C [15]. The trends observed in section 3.2 for the dependence

of 𝑥𝑏𝑢𝑡 were that as the number of equivalents of NaOH increased, 𝑥𝑏𝑢𝑡 decreased.

Fig. 7. Butanol content in the aqueous phase, 𝒙𝒃𝒖𝒕 The dependence of the butanol mass fraction in the aqueous

phase, 𝒙𝒃𝒖𝒕 , on the total mass fraction of NaOH, 𝒇𝑵𝒂𝑶𝑯. The experimental data are fitted to a function which

is decreasing in 𝒇𝑵𝒂𝑶𝑯, and independent of the total mass fraction of HMD, 𝒇𝑯𝑴𝑫.

Figure 7 shows the results of the same 50 samples as shown in Fig. 6, this time showing 𝑥𝑏𝑢𝑡 as a function of 𝑓𝑁𝑎𝑂𝐻.

Although the ternary plot in Fig. 3 showed that there was a trend for 𝑥𝑏𝑢𝑡 depending on 𝑓𝐻𝑀𝐷, this was found to be of

a much lower confidence than the dependence purely on 𝑓𝑁𝑎𝑂𝐻. Instead, there is a strong correlation between 𝑥𝑏𝑢𝑡 and

𝑓𝑁𝑎𝑂𝐻, to which an exponentially decaying function has been fitted, with equation:

𝑥𝑏𝑢𝑡 = 0.0113 + 0.0864exp ( ‒ 35.11 𝑓𝑁𝑎𝑂𝐻)

(7)

13

Electronic copy available at: https://ssrn.com/abstract=3979466

The format of this function implies an asymptote at 𝑥𝑏𝑢𝑡 = 1.13 wt% for sufficiently high values of 𝑓𝑁𝑎𝑂𝐻. There is

also a disagreement with the observed value of 𝑥𝑏𝑢𝑡 = 7.3 wt% in the two-component system of butanol and water,

because this model gives a value of 𝑥𝑏𝑢𝑡 = 9.77 wt%, even when including the confidence bound for this fit of ±1.9

wt%. The interpretation of this dependence is that increasing the ionic strength of the aqueous phase reduces its

capacity to hold the uncharged butanol in solution.

4.3. Partitioning of HMD between phases

The partitioning coefficient, Φ, is the ratio of the HMD mass fractions between the organic and aqueous phases:

𝑦𝐻𝑀𝐷
Φ =
𝑥𝐻𝑀𝐷

(8)

Extraction of the HMD from the aqueous phase into the organic phase is improved at higher values of Φ, as the amount

of HMD remaining in the aqueous phase will be reduced. Section 3.2 showed that the value of 𝑥𝐻𝑀𝐷 decreased greatly

when the number of equivalents of NaOH was increased from 0 to 2, resulting in an increased value of Φ.

Figure 8 shows the partitioning coefficients for the 50 samples shown previously, as a function of 𝑓𝐻𝑀𝐷. Rather than

showing the dependence on the absolute amount of NaOH, fNaOH , the values have been shown for different numbers

of equivalents, N, because the partitioning behaviour depends strongly on the speciation of the HMD.. Figure 8 shows

that the partitioning of HMD is more effective at lower overall HMD content. In addition, increasing the number of

equivalents serves to increase the partitioning coefficient, in agreement with the results seen in Fig. 2 and section 3.2.

To this end, the partitioning coefficient was fitted to a bi-exponential function:

Φ = 1.933 + 24.26exp ( ‒ 133.1 𝑓𝐻𝑀𝐷) + 4.955 (1 ‒ exp ( ‒ 1.066 𝑁))

(9)

14

Electronic copy available at: https://ssrn.com/abstract=3979466

Fig. 8. The dependence of the partitioning coefficient, Φ, on the total mass fraction of HMD, 𝒇𝑯𝑴𝑫, for different

number of equivalents of NaOH, 𝑵. The experimental results are fitted to a bi-exponential function in both

𝒇𝑯𝑴𝑫 and 𝑵.

This function reaches an asymptotic value of Φ = 6.92 when both 𝑁 and 𝑓𝐻𝑀𝐷 are large, and has a confidence interval

of ±2.3. The exponential dependence on 𝑁, in particular, shows that using more than three equivalents of NaOH makes

very little improvement on the partitioning behaviour, which is also true for HMD mass fractions above 4 wt%.

5. Multistage extraction

The experimentally-derived equilibrium equations (6), (7) and (9) can be included in the model for a single mixer-

settler given at the start of Section 4. The model is now closed, as it contains 10 independent equations for the 10

unknown variables, and as such the equilibrium separation of such a mixer-settler unit can be predicted. The model

can therefore be incorporated into a model of a multi-stage extraction process.

Fig. 9. An assembly of three, countercurrent equilibrium stages for the extraction of HMD from an aqueous

stream to an organic stream.

15

Electronic copy available at: https://ssrn.com/abstract=3979466

Figure 9 shows a schematic diagram of a three-stage countercurrent solvent extraction process. The complete system

contains 30 unknown variables, and the total inflow to each stage is the combination of the aqueous outflow from the

stage to its left, or the aqueous feed for stage 1, and the organic outflow from the stage to its right, or the organic feed

for stage 3.

5.1. Calculation procedure

This system was modelled using the phase equilibrium results presented earlier, as an investigation into how such a

system might operate. The model was written in MATLAB and solved as a system of 30 simultaneous equations using

the lsqnonlin function. The control parameters for the simulation are shown in Table 1.

This simulation is for a very dilute HMD solution at 1 wt%. The amount of NaOH present corresponds to 2 equivalents

in the feed. The flow rate of the organic phase is a half that of the aqueous phase, which means that any HMD extracted

into the organic phase would have its concentration approximately doubled.

Table 1. Control parameters for the simulation of a multistage solvent extraction process.

Parameter Value
Number of stages 3
Connectivity Countercurrent
Aqueous feed mass flow rate 10.0 g s-1
Aqueous feed water content 98.31 wt%
Aqueous feed butanol content 0.00 wt%
Aqueous feed HMD content 1.00 wt%
Aqueous feed NaOH content 0.69 wt%
Organic feed mass flow rate 5.0 g s-1
Organic feed water content 0.00 wt%
Organic feed butanol content 100.00 wt%
Organic feed HMD content 0.00 wt%
Organic feed NaOH content 0.00 wt%

5.2. Simulation results

The results of a converged simulation are the mass fractions of each component in each aqueous and organic stream

between each stage, as well as the total mass flow rate in each stream.

16

Electronic copy available at: https://ssrn.com/abstract=3979466

Fig. 10. The mass flow rates and mass fractions of dissolved solvents in the streams between each equilibrium

stage in a system of 3 mixer-settler units operating in countercurrent flow. Circles mark the values, arrows

show the direction of each stream flow. The aqueous streams flow left to right, and the organic streams right

to left.

Figure 10 shows the total mass flow rates of both the aqueous and organic streams as they pass through the system, as

well as the mass fraction of each solvent in its opposing phase. The aqueous stream starts at 10 g L-1 and flows left to

right, whereas the organic stream starts at 5 g L-1 and flows right to left. Across stage 1, the aqueous stream increases

in total mass flow rate as it picks up butanol, and across stage 3 it decreases in total mass flow rate as it loses water.

This is mirrored in the behaviour of the organic phase, which picks up water across its first stage, stage 3, and loses

butanol across its final stage, stage 1. In total the net mass transfer is 0.08 g L-1 from the aqueous phase to the organic

phase, as more water can dissolve in the butanol than vice versa, and there is more water present overall.

17

Electronic copy available at: https://ssrn.com/abstract=3979466

Fig. 11. The mass fractions of HMD and NaOH in the organic and aqueous phases between each equilibrium

stage in a system of 3 mixer-settler units operating in countercurrent flow. Circles mark the values, arrows

show the direction of each stream flow. The aqueous streams flow left to right, and the organic streams right

to left.

Figure 11 shows the mass fractions of HMD and NaOH of both the aqueous and organic streams as they pass through

the system. As the aqueous phase passes through the system, xHMD decreases as the HMD is progressively extracted

into the organic phase, and, in total, 99.95% of the HMD is extracted. This extraction occurs by 86.14% across stage

1, 12.98% across stage 2, and 0.84% across stage 3. At each stage, approximately 90% of the HMD remaining in the

aqueous phase is transferred to the organic phase.

Although the organic stream cannot extract any of the NaOH present in the aqueous stream, the mass fraction 𝑥𝑁𝑎𝑂𝐻

does vary through the system as the total mass of the aqueous stream changes. This results in a slightly lower value of

𝑥𝑁𝑎𝑂𝐻 around the central stage because the aqueous stream has picked up butanol but not yet lost water. With the

NaOH content in the aqueous stream remaining approximately constant, but the HMD content decreasing as it is

extracted into the organic phase, there is a higher effective number of equivalents in later stages, making the separation

more effective.

Because the total mass flow rate of the organic phase is approximately half that of the aqueous phase, this results in

values of 𝑦𝐻𝑀𝐷 approximately double those of 𝑥𝐻𝑀𝐷 at the same location. Accounting for the transfer of water and

18

Electronic copy available at: https://ssrn.com/abstract=3979466

butanol between phases, this results in an organic product with 1.97 wt% HMD and 16.26 wt% water at a flow rate of

5.08 g s-1 in this example

6. Conclusions

This work has investigated the extraction of dilute aqueous solutions of HMD using organic solvents. A preliminary

investigation showed that butanol would be a promising solvent, owing to its polar and organic nature, which allows

it to dissolve HMD readily, with a maximum solubility for HMD of 78 wt%. Samples containing HMD, water, and

butanol at various concentrations were produced, and the mass fractions of all species in each of the aqueous and

organic phases were measured by gas chromatography.

It was found that HMD has a bridging effect on the partial miscibility of water and butanol, and approximately 10

wt% of HMD was sufficient to allow for complete miscibility and the avoidance of phase separation. However, the

addition of NaOH to the system reversed this bridging effect and reduced the solubility of butanol in the aqueous

phase whilst also increasing the partitioning of HMD between the organic and aqueous phase.

A model for an equilibrium mixer-settler unit was developed with three important findings emerging. Firstly, the

amount of water which could be dissolved in the organic phase was found to increase with the amount of HMD present,

although this effect was neutralised by the presence of sufficient NaOH. Secondly, the amount of butanol which could

be dissolved in the aqueous phase was found to decrease as the amount of NaOH present increased, but was largely

independent of the amount of HMD. Lastly, the partitioning coefficient of HMD was found to decrease as the amount

of HMD increased, but increased as the molar ratio of NaOH to HMD was raised. Fitted equations for these three

findings allowed closure to be achieved for the model.

This model for the solvent extraction of HMD by means of butanol was investigated in a simple 3-stage countercurrent

system. Operating such a system with a lower organic flow rate than aqueous flow rate can effectively increase the

HMD concentration in the organic product relative to the aqueous feed, but results in a net transfer of mass to the

organic phase. As the aqueous stream passes through the equilibrium stages, the separation becomes more efficient as

the HMD mass fraction decreases, which also increases the effective number of equivalents of NaOH. Using only

three equilibrium stages, this model has predicted a set of operating conditions which allows for 99.95% extraction of

a very dilute aqueous HMD solution, including a doubling of the HMD concentration in the organic product.

19

Electronic copy available at: https://ssrn.com/abstract=3979466

Acknowledgements

The authors would like to thank INVISTA Textiles (U.K.) Limited funding for this work.

References

[1] Intratec Solutions LLC. Nylon 6, 6 Production - Cost Analysis - Nylon 66 E12A, Intratec, 2019.

[2] P.L. Morse, Process Economics Program Report 31: Hexamethylenediamine, Stanford Research Institute, Menlo

Park, California, 1967.

[3] M.J. Sargent, N.K.H. Slater, J.S. Dennis, G.J. Smith and P.S. Pearlman, The extraction of hexamethylenediamine

from aqueous solution by pH control and salt addition, Chemical Engineering Journal, 419 (2021) 129428.

[4] J.D. Miller and M.B. Mooiman, A review of new developments in amine solvent extraction systems for

hydrometallurgy, Separation Science and Technology, 19 (1984) 895–909.

[5] T.H. Nguyen and M.S. Lee, A review on the separation of molybdenum, tungsten, and vanadium from leach liquors

of diverse resources by solvent extraction, Geosystem Engineering, 19 (2016) 247–259.

[6] Z. Zhu and C.Y. Cheng, Solvent extraction technology for the separation and purification of niobium and tantalum:

A review, Hydrometallurgy, 107 (2011) 1–12.

[7] W.P. Ratchford, E.H. Harris, C.H. Fisher and C.O. Willits, Extraction of lactic acid from water solution by amine-

solvent mixtures, Industrial & Engineering Chemistry, 43 (1951) 778–781.

[8] A. Senol, Effect of diluent on amine extraction of acetic acid: modeling considerations, Industrial and Engineering

Chemistry Research, 43 (2004) 6496–6506.

[9] H. Uslu, Reactive extraction of formic acid by using Tri Octyl Amine (TOA), Separation Science and Technology,

44 (2009) 1784–1798.

[10] G. Audunsson, Aqueous/aqueous extraction by means of a liquid membrane for sample cleanup and

preconcentration of amines in a flow system, Analytical Chemistry, 58 (1986) 2714–2723.

[11] D.H. Rosenblatt, P. Hlinka and J. Epstein, Use of 1,2-naphthoquinone-4-sulfonate for the estimation of

ethylenimine and primary amines, Analytical Chemistry, 27 (1955) 1290–1293.

20

Electronic copy available at: https://ssrn.com/abstract=3979466

[12] National Center for Biotechnology Information. PubChem Compound Summary for CID 16402, 1,6-

Hexanediamine, 2020.

[13] Y. Marcus, Solvent Mixtures: Properties and Selective Solvation, Taylor & Francis, 2002.

[14] S.K. Bhasin and R. Gupta, Pharmaceutical Organic Chemistry -E-Book, Elsevier Health Sciences, 2013.

[15] A.F.M. Barton, (Ed.), Alcohols with Water, IUPAC Solubility Data Series Volume 15, 1984.

[16] P.J. Hodgson, A Novel Technique for the Separation of Dilute Butanol from Aqueous Fermentation Broths, PhD

Dissertation, University of Cambridge, 2018.

21

Electronic copy available at: https://ssrn.com/abstract=3979466