New insights in reactive extraction mechanisms of organic acids: An experimental approach for 3-hydroxypropionic acid extraction with tri-n-octylamine

A detailed study of 3-hydroxypropionic acid (3-HP) reactive extraction with tri- n -octylamine (TOA) is 10 proposed for the first time. It aims at uncovering some solvent-solutes interactions and providing 11 global mechanisms to better understand and design the reactive liquid-liquid extraction of 3-HP in a 12 biotechnological process. Eleven solvents of similar molecular sizes and several chemical types 13 (alcohols, esters and alkanes) were investigated to understand their role on the extraction ability. 14 Alcohols were found to be the best solvents thanks to their H-bond donor characteristic and water 15 loading that allowed good solvation of the acid-amine complexes. Further investigations were then 16 undertaken, for n -decanol and oleyl alcohol as solvents, varying acid (0.0028 – 0.56 mol/L 17 corresponding to 0.25 – 50 g/L) and amine (0 – 2.3 mol/L corresponding to 0 – 100% v/v) 18 concentrations. At 0.011 mol/L (1 g/L) of 3-HP, maximum extraction yields of 77% for n -decanol

It is now recognized that 3-hydroxypropionic acid (3-HP) is a valuable platform molecule that can be 2 used as a building-block for the production of new polymers like polyesters and acrylic acid 3 derivatives such as acrylates or acrylamides [1]. This molecule is currently not widely used due to its 4 exclusive and difficult synthesis via chemical pathways including transformation of petrochemical 5 products such as propylene but is expected to exhibit a significant growth in the next years. Indeed, it 6 has been referred as one of the Top 10 "chemical opportunities from biorefinery carbohydrates" [2]. 7 Lately, research has focused on the production of 3-HP using biotechnologies with natural [3,4] and 8 modified [5,6] micro-organisms able to convert glycerol or glucose into 3-HP [7]. Other substrates 9 like CO 2 [8,9] and C5 sugars [10,11] are also currently studied. Biotechnology is considered as a 10 promising route for the sustainable and efficient production of 3HP but one of the main drawbacks of 11 this approach is the inhibitory effect the products can have on the micro-organisms, limiting 12 productivity and final concentrations to quite low amounts, which is particularly true for the 3-HP 13 production [7]. In this context, in-situ product recovery (ISPR) seems to be a promising approach 14 because the simultaneous removal of the product from the bioconversion medium could favor higher 15 yields, productivity and concentrations. Liquid-liquid reactive extraction for organic acids' recovery is 16 known to be particularly efficient and selective for dilute media as an ISPR technique in 17 biotechnological processes and can prevent toxic metabolites and/or products from accumulating 18 [12,13]. 19 Current concentrations of 3-HP in bioconversion media for the best current biotechnological 20 performances found in the literature do not exceed about 70 g/L [6,14]. For short chain carboxylic 21 acids in dilute streams such as 3-HP, their hydrophilic character prevents their direct recovery in an 22 organic solvent and a reagent is needed in the organic phase to trap the compound in a rather 23 hydrophobic complex. In the case of hydrophilic carboxylic acids, these reagents can be hydrophobic 24 basic molecules being able to interact with acidic groups [ short chain amines (tributylamine, N,N-dimethyloctylamine) and primary amines (n-octylamine) had 10 the lowest performances. On the contrary, long chain secondary and tertiary amines proved to be 11 very efficient with very high extraction yields, probably due to their higher basicity and very low 12 solubility in water. Secondary amines (di-n-octylamine) had better performances than tertiary 13 amines (tri-n-octylamine) [37], and this could be explained by the lesser extent of the steric 14 hindrance around the reactive amine group, their higher basicity and higher density of amine groups. 15 Similar results were found in a previous study which also demonstrated that primary amines were 16 not to be used in reactive extraction due to moderate extraction efficiency, high water solubility and 17 high interfacial activity causing emulsion [38]. 18 The structure of the aliphatic amines is crucial for their role as extractants, but they cannot be used 19 In order to further investigate the influence of TOA purity on extraction efficiency, three different 4 organic phases have been made with three different TOA purities. Two TOA solutions were 5 purchased from the same commercial reference but had two lot numbers with different purity levels 6 corresponding to the variations of production quality: 98.1%, solution 1, and 99.6%, solution 2, 7 according to the certificates of analysis given by the supplier (Sigma-Aldrich). Solution 3 was obtained 8 by purification of solution 2 as detailed below. 9

TOA purification 10
Purified TOA (solution 3) was produced as follows. The 99.6% TOA was washed using same volumes 11 of H 2 SO 4 0.1 M. After centrifugation, 3 homogeneous phases were observed. Only the top phase was 12 recovered and further washed with NaOH 0.3 M. Indeed, the phase in-between is suspected to be 13 loaded in acid-amine complexes [23] and to be sure that a maximum of impurities were removed 14 from the TOA phase, only the top phase was recovered. These operations were repeated twice and 15 the last TOA top phase was finally washed with deionized water. The pH of the final aqueous phase 16 has been checked to be neutral, proving that no inorganic salts were left in purified TOA and this 17 latter is called solution 3. 18

Experiments 19
Several experiments of reactive liquid-liquid extractions were performed for different initial 20 concentrations of TOA and 3-HP. Unless specified, TOA of solution 2 (99.6% purity) was used. For 21 each extraction, 10 mL of the aqueous phase and 10 mL of the organic phase were mixed in 22 centrifuge tubes which were then manually shaken for 3 minutes. The stable emulsion formed was 23 then left at 25 °C for one day before being shaken again for 3 minutes and centrifuged at 15557 g for 24 1 hour. Once the aqueous phase recovered, it was analyzed using High Performance Liquid 1 Chromatography (HPLC) and the extraction yield was calculated as follows: 2 with [ ] being the total initial acid concentration in the aqueous phase according to HPLC 3 analysis and [ ] the same at equilibrium. 4 pH of the resulting aqueous phases was measured and compared to the theoretical pH value, based 5 on the 3-HP dissociation (reaction 3) and calculated as follows: 6 All the extractions were made in duplicates. For some experiments, organic phases have been back 7 extracted with NaOH solutions (0.1-0.5 M) in order to confirm that mass balance was satisfied within 8 3% or less. 9 Experiments were centered on 0.011 mol/L (1 g/L) 3-HP and the optimal extractant concentration for

Analytical methods 16
The 3-HP concentration in aqueous phases was determined using HPLC containing an Aminex HPX-17 87H H + exchange column (300x7.8 mm, Biorad, USA) and an H + Micro-Guard column (30mmx4.6mm, 18 Biorad, USA). The mobile phase consisted in a 0.01N H 2 SO 4 solution circulating at a flowrate of 0.6 1 mL/min and the column temperature was set to 50 °C. An internal standard consisting in 10 g/L citric 2 acid in water was added (50% v/v) to the samples and 20 µL of each sample were used for injection. 3 Signals were obtained using a UV detector with the specific wavelength of 210 nm for acids. The 4 overall HPLC analysis uncertainty, defined as the coefficient of variation of the internal standard peak 5 area, was evaluated to 1%. 6 Impurities identification and titration in the aqueous phase were performed using Ultra High 7 Performance Liquid Chromatography coupled with High Resolution Mass Spectroscopy (UHPLC-8 HRMS). Separations were performed using a Hypersil Gold C18 column (50X2.1 mM; 1.9 µm particle 9 size; Thermo Fisher Scientific, USA) with an Ultimate 3000 pump and Thermo 3000 RS autosampler 10 (Thermo Fisher Scientific, USA). The oven temperature was set to 25°C. The mobile phase was a 11 mixture of LC/MS grade acetonitrile and water loaded with 0.1% formic acid (Optima, Thermo Fischer 12 Scientific, USA) flowing at 250 µL/min. In a run, it first consisted in 2% acetonitrile for 4 minutes. 13 Then a gradient was applied to reach 98% acetonitrile in 6 minutes and the mixture was then 14 stabilized at 98% acetonitrile for 5 minutes. The mobile phase was finally set back to initial conditions 15 (2% acetonitrile) with a 2 minutes gradient. 16 Samples were diluted 100-fold in water (Optima LC/MS grade, Thermo Fischer Scientific, USA) and 17 filtered before assay with syringe filters (nylon membrane, 0.2 μm porosity and 13 mm diameter, 18 Merck Millipore, USA). 19

MS analysis was carried out with a Q-EXACTIVE Orbitrap mass spectrometer (Thermo Fisher 20
Scientific, USA) in positive mode with a Heated ElectroSpray Ionization (HESI) probe using a high 21 resolution full scan. The MS resolution was 70000 in a full MS mass range (m/z) from 50 to 700 AMU. 22 The ion transfer tube and vaporizer temperatures were set to 300°C. The electrospray voltage was 23 set to 3 kV and the sheath, auxiliary and sweep gas (nitrogen) flow rates were set to 35, 10 and 0 24 arbitrary units, respectively. 25 IR analyses were performed using a Cary 630 Fourier Transform InfraRed spectrometer of Agilent 1 Technologies (California, USA). 2 pH measurements were obtained from a SevenCompact pHmeter equipped with the Inlab Micro Pro-3 ISM combined pH electrode with temperature probe (Mettler Toledo, Switzerland). 4 Water concentration in the organic phase was determined using Karl-Fischer method with a 756 KF 5 Coulometer (Metrohm, Switzerland). Methanol was the solvent for titration. Organic phase samples 6 were analyzed in triplicate. 7 Statistical significance of differences between data was assessed using ANOVA tests at 0.05 level. 8

13
First, a set of 11 solvents were selected for a screening regarding the reactive extraction of 0.011 14 mol/L (1 g/L) of 3-HP in aqueous phase using 0.46 mol/L of purified TOA (20% v/v) in the organic 15 phase. Figure 1 provides the extraction yields for these solvents. 16 Three kinds of solvents were tested: alcohols, esters and alkanes with different lengths of alkyl chain 1 and numbers of carbon atoms. They can be compared according to their extraction ability and 2 physicochemical properties like polarity (dipole moment and relative permittivity) and H-bond ability. 3 4 Table 1: Physicochemical properties of the studied solvents [47][48][49][50]

10
The immediate conclusion from the observation of Figure 1 is that alcohols have the highest 11 extraction yields. Among alcohols, the shorter the chain, the higher the extraction yield: 91% for the 12 C 6 hexanol, 53% for the C 18 oleyl alcohol. For alkanes and esters, the yield is independent of the chain 13 length (ANOVA test with p=0.6 for both), which means that the chain length cannot be used as a 14 good parameter to discriminate solvents. If all the n-alcohols have approximately the same dipole 15 moment (i.e "molecular polarity") and H-bond ability due to the same primary hydroxyl groups (Table  16 1), they differ in their relative permittivity that is more related to the "bulk polarity" and hence to the 17 molecular size and organization. In this case, the difference in solvent activity is given by their 18 relative permittivity so that the smaller the n-alcohol the higher the yield. It is obvious that totally 19 inert solvents like alkanes (apolar and without H-bond ability) are unable to extract the acid 20 efficiently. n-hexane, n-decane and n-dodecane having the same values of relevant parameters 1 (dipole moment, relative permittivity and H-bond ability), all provide the same efficiency as expected 2 (5% yield). The same observations apply to the esters: both esters having the same relevant 3 properties, they show equivalent performances (around 8% yield). This slightly, but significantly 4 higher yield (ANOVA test with p<0.02), can be explained by the slightly higher polarity of the two 5 esters compared to the three alkanes, which makes them a bit more active diluents. It is also 6 interesting to compare esters and alcohols with similar polarities but different H-bond abilities like 7 butylhexanoate and ethyloctanoate vs. oleyl alcohol and n-dodecanol. A much higher yield is 8 obtained for both alcohols (7 times higher for oleyl alcohol than for the esters), which leads to the 9 conclusion that the H-bond donor character of the solvent plays a key role, implying that alcohols 10 stabilize the reactive system through H-bonding interactions. Ricker [43] suggests that the acidic 11 hydrogen of the carboxylic acid being involved in the acid-amine bond, carboxyl becomes a strictly Based on these results, n-decanol and oleyl alcohol were chosen for further investigations. n-hexanol 17 and octanol were ruled out due to their excessive solubility in water that is expected to be toxic for 18 microorganisms under integrated extractive bioconversion (ISPR) [51,52]. According to the literature 19 review, some biocompatible vegetable oils, thanks to their low polarity and solubility in water, have 20 been shown to be quite efficient for the removal of similar acids (propionic, pyruvic, lactic) from 21 aqueous phases [47][48][49]. But in our present results, sunflower oil was found to be inefficient for 3-HP 22 extraction with TOA (less than 7% yield) whatever the proportions used because of its weak solvation 23 properties. 24

Physical extraction 25
For physical extraction experiments, concentration of 3-HP were varied from 0.0028 to 0.56 mol/L 1 (0.25 to 50 g/L) with no TOA in the organic phase. In n-decanol, a linear dependence (R²=0.96) of the 2 equilibrium acid concentration in the organic phase with respect to that in the aqueous phase was 3 found (results not shown). The slope, i.e. the partition coefficient (defined as the ratio of the molar 4 concentrations at equilibrium), was evaluated to 0.020±0.001 by linear regression, meaning that 5 physical extraction yield was around 2%. This is in line with Kertes and King's statement that the acid 6 partition coefficient is independent on its concentration in the case of extraction using alcohols [15]. 7 In oleyl alcohol, 3-HP partitioning was not detectable. Physical extraction yields were thus too low for 8 practical purposes, confirming that the efficient liquid-liquid extraction of 3-HP needs to be driven by 9 a chemical reaction. Alcohols, with polar and H-bond donor hydroxyl groups, are thus able to stabilize the hydrophilic 1 anionic conjugate in the organic phase. 2 As a consequence, the extraction yield of 3-HP decreases at high amine concentrations, a behavior 3 that aligns well with reported data for acetic [43],  Moreover, water solubility appears to play a crucial role. As shown in Figure 4, the more the organic 13 phase is loaded with TOA, the less water is soluble, water being practically insoluble in TOA. 14 Increasing the amount of TOA leads to a decrease of water content in the organic phase which could 15 lead to a lesser stabilization of the complex and hence a lesser extraction efficiency. Tamada et al 16 [57] reported that the amount of acid extracted and the stoichiometry of the complex in the organic 17 phase depended on the water content in the organic phase. The amount of water soluble in the 18 organic phase is supposed to stabilize the hydrophilic part of the complex in a rather apolar 19 environment. Increasing the concentration of alcohol in the organic phase increases the water 1 solubility and therefore the solvation effect of alcohols could be in fact partially due to the solvation 2 of water. The balance between the respective contributions in complex stabilisation cannot be 3 assessed based on the present results because the amounts of water and alcohol are correlated.

8
Concerning the influence of the acid concentration, Figure 5 shows that the plot of the extraction 9 yield against initial acid concentration is not monotonic. If we consider only the partitioning of the 10 acid and the complex formation, such a profile is not expected. Indeed, the more diluted is the acid 11 and the more in excess are the amines; this should lead to a higher extent of reaction 2 (acid-amine 12 complex formation) and therefore the increase of extraction yield with the decrease of 3-HP 13 concentration. This behavior was indeed observed for initial acid concentration levels between 0.11 14 and 0.56 mol/L (10 -50 g/L). But for the weakest concentration levels, a sharp decrease is observed 15 and other phenomena are needed to explain such results. For example, in oleyl alcohol the 16 extraction yield decreased from 51% to 40% and in n-decanol from 72% to 51% between 0.

4
A further look at the water content of the organic phase after extraction proves that water is 5 extracted in the organic phase jointly with the acid (Figure 6). Indeed, the more acid extracted, the 6 more water in the organic phase. Indeed, water concentration reached higher amounts than its 7 solubility in n-decanol with 20%v/v TOA (1.14 mol H 2 O/L, Figure 4) and in oleyl alcohol with 20%v/v 8 TOA (0.50 mol H 2 O/L). Figure 6 shows that the amount of water in the organic phase varies linearly 9 with the acid concentration. This tends to prove that, in the range studied, each 3-HP molecule 10 carries the same average number of water molecules in the organic phase. This hydration number 11 (given by the slopes of the regression lines in Figure 6) is around 2 for n-decanol and oleyl alcohol as 12 solvents. This tendency of 3-HP to carry water molecules while being extracted in the organic phase 13 could also partly explain why the extraction yield was very low with alkanes (section 3.1). 14 15 16 17 18 3.5. Influence of TOA impurities on extraction performances 1 3.5.1. pH at equilibrium 2 ), it appears that for the lowest acid concentrations the experimental data are above the 1 calculated values from acid concentrations at equilibrium alone ( ℎ , reaction 3, equation 2). This 2 means that higher amounts of 3-HP molecules are deprotonated compared to what is expected with 3 only 3-HP in water. When performing pH titrations of aqueous phases after acetic acid extraction by 4 TOA in chloroform, Wardell and King [58] also reported pH deviations. They attributed this 5 phenomenon to the transfer of tri-n-octylamine in the aqueous phase in the presence of acid. The 6 same was found for lactic acid extraction at low concentration with Alamine 336 [59]. However, 7 Ricker [43] demonstrated that, even in 10% wt acetic acid in water, TOA solubility would not exceed 8 10 ppm (i.e. < 3 10 -5 mol/L) and suggested that pH deviation may have been due to impurities in 9 commercial products like primary and secondary amines, lower-molecular-weight tertiary amines 10 and starting materials for the amine synthesis. 11 When sodium hydroxide was added to our aqueous samples after extraction, a condensation of 12 organic species occurred and the solution became cloudy as already reported by some authors 13 [57,58]. On the contrary, pure distilled water contacted with commercial TOA solution resulted in a 14 neutral pH without any noticeable visual change when NaOH was added. This illustrates the fact that 15 impurities are extracted in the aqueous phase under protonated form through reactive extraction 16 with acidic species (see reaction 4) and that their partitioning as neutral forms can be neglected: 17

Identification of TOA impurities 18
Aqueous samples after extraction were analyzed with mass spectroscopy. Two main peaks appeared 19 at m/z=130 and m/z=242 corresponding to both peaks of protonated n-octylamine (129 + 1 = 130 20 g/mol) and protonated di-n-octylamine (241 + 1 = 242 g/mol). Commercial samples of n-octylamine 21 (OA) and di-n-octylamine (DOA) were also analyzed to confirm species identification. Several 22 manufacturing processes exist for the production of fatty amines like nitrile hydrogenation or 23 alcohols amination. For example, TOA has been produced by BASF using n-octanol amination with 24 ammonia and OA leading to impurities of around 2% DOA and 1% OA [60]. This tends to confirm the 1 presence of these impurities in the products in our study. 2 When acidic species are present in the aqueous phase, these TOA impurities (OA and DOA) undergo 3 protonation and both deprotonated acid molecules and protonated amine molecules are released in 4 the aqueous phase increasing the pH ( Table 2, solutions 1 and 2). This phenomenon can be neglected 5 at high acid concentration because impurities amount remains relatively low and the protonated 6 form of TOA is practically insoluble in water [43]  Analyses of aqueous phases after extraction proved that no significant TOA were released in the 23 aqueous phases. However, OA and DOA were identified and quantified, with the former as the main 24 amine in the aqueous phase (more than 90% of total amines concentration). Hence, total amount of 1 amines (OA+DOA) in the aqueous phase is assimilated to OA to simplify further interpretation. The 2 amount of OA varied between 0.031 and 9.1 mmol/L (Table 2), depending on the following operating 3 conditions: TOA initial concentration in the organic phase, purity of TOA and 3-HP initial 4 concentration in the aqueous phase. 5 Amine transfer into the aqueous phase increases the pH. Accordingly, an amended formula should be 6 used to calculate the pH by taking into account the amines. Knowing the total concentration of 7 amines in the aqueous phase and their dissociation constants, the equilibrium pH of 3-HP solutions 8 can then be calculated. If we consider that amines in the aqueous phase mainly consist in OA, the 9 theoretical pH calculation becomes: 10 ℎ, [H + ] being the physical solution of the following equation (4), based on 3-HP dissociation (reaction 3), 11 n-octylammonium dissociation (reaction 5), water autoprotolysis (reaction 6) and the 12 As can be seen in table 2, values of pH th,OA are much closer to the pH measured experimentally. 14 Accordingly, the deviations observed between pH eq and pH th are directly related to the impurities in 15 commercial TOA. Moreover, pH measurement can give, as a first approximation, reliable information 16 about the amount of impurities (OA) transferred into the aqueous phase. 17 Since only non-dissociated 3-HP can be extracted by TOA in the organic phase, amines transfer into 18 the aqueous phase and the consequent increase in pH eq reduces the amount of protonated 3-HP 19 available for extraction. The final outcome of this phenomenon is the decrease of the extraction 1 yield. 2 3  Yield reduction caused by TOA impurities is illustrated in figures 5 and 7A for low acid concentrations. 8 In Figure 7A, the plot is given in logarithmic scale for initial 3-HP concentration in order to clearly 9 observe the discrepancies for the small acid concentrations and the similarities for high acid 10 concentrations. For a given TOA concentration in the organic phase (0.46 mol/L = 20% v/v), the three 11 solutions provide equivalent yields at high acid concentration (>0.11 mol/L = 10 g/L). This was 12 expected since the amount of amines released in the aqueous phase is small compared to the acid 13 concentration, and its influence on the yield becomes negligible. However, for low acid 14 concentrations, OA and DOA impurities present in TOA solutions have a strong impact. For example, 15 at around 5mmol/L (0.5 g/L) 3-HP initially, the extraction yield is 23% for solution 1 (2.5 mmol/L OA), 16 58% for solution 2 (1.0 mmol/L OA) and 74% for solution 3 (0.0058 mmol/L OA). The yield decrease 17 at low acid concentration with commercial TOA (solutions 1 and 2) is dramatic whereas with purified 18 TOA (solution 3) the decrease is much smaller. Nevertheless, it still denotes the release of weak 19 amounts of impurities, but with much lower importance than from unpurified TOA. In figure 7B, one 20 can see a much smaller yield decrease for purified TOA (solution 3) than for solution 1 along with the 1 increase of amine concentration, up to about 50% v/v TOA. This steeper decrease for unpurified TOA 2 is due to impurities release, whereas for purified TOA only the solvating effect is noticeable. 3 In the case of purified TOA, results shown in Table 2 confirm that the purification of TOA was 4 effective, dividing the amount of amines released in the aqueous phase by a factor 30 when 5 compared to solution 2 and a factor 65 when compared to solution 1. This decrease makes the 6 amount of amines practically negligible in the aqueous phase, as confirmed by pH measurements. 7 Besides the effects described above on the extraction chemistry, the diminution of the released 8 amines by means of the purification step is particularly recommended if microorganisms are used in 9 a biological process due to expected amine toxicity. For example, a similar lactate -long chains alkyl 10 ammonium salt in an aqueous phase proved to be very toxic for several kinds of microorganisms 11 [61]. Purification of fatty tertiary amines can be reached for example with an adsorption process as 12 described in a US patent [62]. In continuous chemical processes, however, the presence of impurities 13 would have less impact because the recycled organic phase would be washed by the continuous 14 fresh aqueous flow and small amines would be readily depleted [43]. 15

16
This paper reports a comprehensive study of the reactive extraction of 3-HP, a valuable platform 17 molecule. The experimental approach was designed in line with a potential application of reactive 18 extraction for the in situ product recovery of this acid. It appeared from the solvent screening 19 experiments, performed for the first time on 3-HP, that the type of solvent has a great influence on 20 the extraction yield, with H-bond donor characteristic and polarity being the most important 21 properties to increase the yield. This highlights strong interactions between the solvent and the 22 complexes formed between acids and amines. Water solubility in the organic phase also seems to be 23 a major factor in complex stabilization. In active diluents like n-decanol and oleyl alcohol, a bell-24 shaped profile of the extraction yield as a function of amine concentration is observed with very low 25 extraction yields at high amine concentrations. This proves the major role of the solvent in the 1 extraction process. In these solvents, the increase in initial acid concentration from 0.003 up to 0.1 2 mol/L (0.25 -10 g/L) led to an increase in extraction yield while a further increase from 0.1 to 0.56 3 mol/L (10 -50 g/L) led to the reduction of the extraction yield. This phenomenon has been shown to 4 be due to amines transfer in the aqueous phase, with significant impact for low acid concentrations. 5 Deviation of equilibrium pH from what should be expected with the remaining acid in the aqueous 6 phase demonstrated the presence of impurities. We identified the release of n-octylamine and di-n-7 octylamine coming from commercial TOA into the aqueous phase even with 99.6% purity. This 8 transfer reduces extraction yields at low initial acid concentrations and increases equilibrium pH 9 which is, accordingly, a reliable indicator of amines concentration in water. Given the extraction yield 10 reduction and the suspected toxicity for microorganisms performing bioconversions, purification of 11 TOA before use could be necessary. These results give insights in extraction mechanisms of organic 12 acids, further clarifying the role of the organic solvent and water in the stabilization of the acid-amine 13 complex and pointing out the role of the amine impurities in extraction yield reduction and 14 biocompatibility of the extraction process. All this understanding is essential for further mechanistic 15 and predictive modeling. 16 17