Recovery of Gallium-68 and Zinc from HNO3-Based Solution by Liquid–Liquid Extraction with Arylamino Phosphonates

The cyclotron production of gallium-68 via the 68Zn(p,n)68Ga nuclear reaction in liquid targets is gaining significant traction in clinics. This work describes (1) the synthesis of new arylamino phosphonates via the Kabachnik–Fields reaction, (2) their use for liquid–liquid extraction of 68Ga from 1 M Zn(NO3)2/0.01 M HNO3 in batch and continuous flow, and (3) the use of Raman spectroscopy as a process analytical technology (PAT) tool for in-line measurement of 68Zn. The highest extraction efficiencies were obtained with the extractants functionalized with trifluoromethyl substituents and ethylene glycol ponytails, which were able to extract up to 90% of gallium-68 in batch and 80% in flow. Only ppm amounts of zinc were co-extracted. The extraction efficiency was a function of pKa and the aqueous solubility of the extractant and showed marked concentration, solvent, and temperature dependence. Raman spectroscopy was found to be a promising PAT tool for the continuous production of gallium-68.


Introduction
Gallium-68 radiopharmaceuticals remain one of the cornerstones of positron emission tomography (PET). New 68 Ga PET tracers significantly improve patients' clinical outcomes, and the number of clinical trials and publications involving gallium-68 continues to grow. 68 Ga-PSMA and 68 Ga-NETSPOT™ ( 68 Ga-SOMAkit-TOC) are becoming the gold standard for prostate cancer and neuroendocrine tumor diagnostics, [1] and Ga-FAPI is emerging as a new general PET tracer [2].
The clinical success of 68 Ga radiotracers drives the soaring demand for 68 Ga radionuclide. Most 68 Ga is currently supplied by germanium-68 generators, which are convenient to use but expensive and in limited supply. The short life of germanium/gallium generators and their decreasing elution yield due to the decay of parent isotopes cause the total cost of generator ownership to be very high compared to the amount of 68 Ga used for scans. An alternative technology that can potentially solve the 68 Ga shortage has emerged. It is based on irradiation of the stable 68 Zn isotope using medical cyclotrons. The 68 Zn(p,n) 68 Ga nuclear reaction has a large cross-section and can provide 68 Ga with high radionuclidic purity [3]. Traditionally, proton bombardment of solid targets has been used as a primary method of producing PET radiometals. The production of multi-curie amounts of 68 Ga has been reported by using electrodeposited [4], pressed [5], and fused [6,7] 68 Zn solid targets. However, this requires specialized and costly equipment for target preparation, irradiation, cooling, transportation, and post-irradiation target dissolution [8]. In 2011, Jensen and Clark showed that irradiation of a 18 F liquid target charged with a concentrated solution of 68 ZnCl 2 can produce clinically-relevant yields of 68 Ga [9]. It was subsequently recognized that liquid targets have a number of advantages over solid targets. Most importantly, liquid target-based production could be readily deployed within the existing alytical technology tools (PAT) is an opportunity to implement the FDA's Quality by Design approach to radiopharmaceutical production. In terms of process integration and automatization, LLEF is fully compatible with downstream SPE-based processing, if additional steps are required. Setting the stage for future research, the new liquid extractants can also be used for transferring the solution chemistry to an SPE platform by grafting the extractant onto a solid support. Figure 1. Conceptual schematic of the continuous production of 68 Ga using a cyclotron solution target (T). Phase separation is performed in module S using a membrane separator; the 68 Ga is backextracted from the organic phase with 0.1 M HCl and is sent directly into the hot cell. The organic phase is discarded.
Central to this conceptual design is a liquid-liquid extraction (LLE)-based separation module. We previously reported near-quantitative LLE of titanium-45 [22,23] and radiogallium [21] from concentrated HCl using a membrane separator with integrated pressure control. The same system could be implemented here with one important caveat: no compound capable of efficient and selective extraction of gallium from nitric acid in the presence of zinc into an organic phase has been reported. Phosphine oxides (Cyanex 923 [24], Cyanex 925 [25]) and phosphoric, phosphonic [26], and phosphinic acids (Cyanex 301 [27]) have been previously tested but extraction was poor.
This study had three objectives. First, we set out to design, synthesize, and test a new family of extractants able to efficiently perform LLE and separation of 68 Ga from zinc nitrate/nitric acid solutions in batch and in flow. Second, we explored the possibility of implementing Raman spectroscopy as a PAT tool and used the Design of Experiment (DoE) technique for LLE optimization. Lastly, we sought to rationalize the results in the context of the extractant's structure and its physical properties. Conceptual schematic of the continuous production of 68 Ga using a cyclotron solution target (T). Phase separation is performed in module S using a membrane separator; the 68 Ga is back-extracted from the organic phase with 0.1 M HCl and is sent directly into the hot cell. The organic phase is discarded.
Central to this conceptual design is a liquid-liquid extraction (LLE)-based separation module. We previously reported near-quantitative LLE of titanium-45 [22,23] and radiogallium [21] from concentrated HCl using a membrane separator with integrated pressure control. The same system could be implemented here with one important caveat: no compound capable of efficient and selective extraction of gallium from nitric acid in the presence of zinc into an organic phase has been reported. Phosphine oxides (Cyanex 923 [24], Cyanex 925 [25]) and phosphoric, phosphonic [26], and phosphinic acids (Cyanex 301 [27]) have been previously tested but extraction was poor.
This study had three objectives. First, we set out to design, synthesize, and test a new family of extractants able to efficiently perform LLE and separation of 68 Ga from zinc nitrate/nitric acid solutions in batch and in flow. Second, we explored the possibility of implementing Raman spectroscopy as a PAT tool and used the Design of Experiment (DoE) technique for LLE optimization. Lastly, we sought to rationalize the results in the context of the extractant's structure and its physical properties.

System Design
The cyclotron target solutions used in clinics can be prepared in a variety of nitric acid concentrations: from 0.01 M to 1.5 M [15,18,28,29]. From an extraction standpoint, the lowest acidity is preferred. Since 1 M solution of 68 Zn(NO 3 ) 2 in 0.01 M HNO 3 is also commercially available from Fluidomica for clinical production of 68 Ga, it became our choice of aqueous phase for LLE. In the 1960s, Jagodić reported that arylamino phosphonic acid 1 was able to extract a broad range of metals from various acids [30]. Therefore, we have chosen the arylamino phosphonic acid scaffold for further development, recognizing that it can function as a chelator due to the presence of arylamino moieties ( Figure 2). Importantly, the acidity and basicity of NH, and hence the chelation capacity of the arylamino phosphonic acid extractant, can be controlled by the judicious choice of substituents on the aryl rings Ar 1 and Ar 2 . Organic and aqueous solubility can be further tuned by controlling the hydrophilicity of the phosphonic acid monoester moiety.

System Design
The cyclotron target solutions used in clinics can be prepared in a variety of nitric acid concentrations: from 0.01 M to 1.5 M [15,18,28,29]. From an extraction standpoint, the lowest acidity is preferred. Since 1 M solution of 68 Zn(NO3)2 in 0.01 M HNO3 is also commercially available from Fluidomica for clinical production of 68 Ga, it became our choice of aqueous phase for LLE. In the 1960s, Jagodić reported that arylamino phosphonic acid 1 was able to extract a broad range of metals from various acids [30]. Therefore, we have chosen the arylamino phosphonic acid scaffold for further development, recognizing that it can function as a chelator due to the presence of arylamino moieties ( Figure 2). Importantly, the acidity and basicity of NH, and hence the chelation capacity of the arylamino phosphonic acid extractant, can be controlled by the judicious choice of substituents on the aryl rings Ar1 and Ar2. Organic and aqueous solubility can be further tuned by controlling the hydrophilicity of the phosphonic acid monoester moiety.

LLE of 68 Ga Using Extractants 1-9 in Batch
The results of LLE of 68 Ga in batch and the calculated pKa and aqueous solubility of extractants 1-9 in different solvents at either room temperature or 50 • C are presented in Table 1.

Estimation of pKa Using COSMO-RS
The pKa for compounds 1-9 was computationally estimated using the conductorlike screening model for real solvents (COSMO-RS). This computational technique uses density functional theory to calculate molecular screening charge densities and then applies statistical thermodynamics to yield chemical potentials [32]. pKa can be calculated from the Gibbs free energies of the neutral and ionic compounds [33]. Table 1 shows that the calculated pKa value varies from 1.5 (1) to 0.4 (9). A significant increase in acidity was noted for all extractants functionalized with the CF 3 groups (Table 1, entries 6-9).

Batch LLE Optimization Studies
Compounds 1-9 demonstrated significant variability in aqueous solubility, pKa calculations, and extraction efficiency (EE) ( Table 1). The compounds where both Ar 1 and Ar 2 were functionalized with trifluoromethyl groups showed the highest EE. A 7-17% increase in extraction was observed when the concentration of the extractants increased from 10 mM to 30 mM ( Figure S1). A marked solvent dependence was also noted. For most extractants, the best results were obtained in TFT, heptane, and Bu 2 O. Across all tested compounds, an increase in temperature from RT to 50 • C led to a significant increase in EE ( Figure 3). In all cases, clear phase separation between the aqueous and organic phases was observed with no detectable extraction of zinc into the organic phase as evidenced by 65 Zn activity measurements. Preliminary experiments indicated that, for an extractant in a given solvent, the concentration of the extractant and the extraction temperature were the main factors affecting extraction efficiency. A traditional approach to optimization is to change one factor (concentration, temperature) at a time. This approach, however, is inefficient as it does not necessarily lead to the optimal experimental conditions, especially when there is interaction between the factors. A better approach is to use a statistically-driven experimental design, called the Design of Experiment (DoE) approach [34]. Guided by a software algorithm, DoE allows one to define a design space and systematically optimize the response (EE) while taking into account the interaction between the factors. To optimize the yield of LLE, we used a central composite face-centered design algorithm implemented in the software package MODDE 9.1.1 ( Table 2). Concentration and temperature were varied between 10-30 mM and 25-50 • C, correspondingly. These limits defined the center point (20 mM, 37.5 • C) around which four corner and four median experiments were constructed (see Figure S2 for a graphical representation of the design). The reproducibility of the design was estimated by running the center points in triplicate. optimization is to change one factor (concentration, temperature) at a time. This approach, however, is inefficient as it does not necessarily lead to the optimal experimental conditions, especially when there is interaction between the factors. A better approach is to use a statistically-driven experimental design, called the Design of Experiment (DoE) approach [34]. Guided by a software algorithm, DoE allows one to define a design space and systematically optimize the response (EE) while taking into account the interaction between the factors. To optimize the yield of LLE, we used a central composite face-centered design algorithm implemented in the software package MODDE 9.1.1 ( Table 2). Concentration and temperature were varied between 10-30 mM and 25-50 °C, correspondingly. These limits defined the center point (20 mM, 37.5 °C) around which four corner and four median experiments were constructed (see Figure S2 for a graphical representation of the design). The reproducibility of the design was estimated by running the center points in triplicate. Figure 3. LLE of 68 Ga using extractants 1-9 in batch. In each case, the solvent system was chosen in such a way as to provide the best solubility and phase separation for a given extractant. The results of 11 runs were fitted with the partial least squares (PLS) algorithm, producing a quadratic model of excellent statistical quality (Figure 4). Both concentration and temperature were positively correlated with LLE efficiency, but the square of concentra- Figure 3. LLE of 68 Ga using extractants 1-9 in batch. In each case, the solvent system was chosen in such a way as to provide the best solubility and phase separation for a given extractant.
The results of 11 runs were fitted with the partial least squares (PLS) algorithm, producing a quadratic model of excellent statistical quality (Figure 4). Both concentration and temperature were positively correlated with LLE efficiency, but the square of concentration was negatively correlated.  The optimization studies indicated that the extractions performed at and above 20 mM and 37.5 • C consistently produced EE > 83% (Table 2). The optimization studies indicated that the extractions performed at and above 20 mM and 37.5 °C consistently produced EE > 83% (Table 2).

LLE of 68 Ga using Extractants 8 and 9 in Continuous Flow
Having established extractants 8 and 9 as the best performers in batch, we translated batch into flow ( Figure 5). The Zaiput membrane separator provided a clean phase separation with no phase breakthrough at both the extraction and stripping stages. At the extraction stage, LLE was on average 10% less efficient in flow. As with the batch experiments, raising the temperature to 50 °C significantly improved EE. Stripping in 2 M HCl was quantitative. ICP analysis of the stripped solution indicated the presence of 11 ppm of zinc.

LLE of 68 Ga using Extractants 8 and 9 in Continuous Flow
Having established extractants 8 and 9 as the best performers in batch, we translated batch into flow ( Figure 5). The Zaiput membrane separator provided a clean phase separation with no phase breakthrough at both the extraction and stripping stages. At the extraction stage, LLE was on average 10% less efficient in flow. As with the batch experiments, raising the temperature to 50 • C significantly improved EE. Stripping in 2 M HCl was quantitative. ICP analysis of the stripped solution indicated the presence of 11 ppm of zinc.  The optimization studies indicated that the extractions performed at and above 20 mM and 37.5 °C consistently produced EE > 83% (Table 2).

LLE of 68 Ga using Extractants 8 and 9 in Continuous Flow
Having established extractants 8 and 9 as the best performers in batch, we translated batch into flow ( Figure 5). The Zaiput membrane separator provided a clean phase separation with no phase breakthrough at both the extraction and stripping stages. At the extraction stage, LLE was on average 10% less efficient in flow. As with the batch experiments, raising the temperature to 50 °C significantly improved EE. Stripping in 2 M HCl was quantitative. ICP analysis of the stripped solution indicated the presence of 11 ppm of zinc.

Zinc Nitrate Quantification Using Raman Spectra
Twelve solutions of zinc nitrate in 0.01 M HNO 3 were prepared (with the concentration of zinc varying in the range of 0.3-1.2 M) and the Raman spectra were acquired. A multivariate analysis of the spectra using SIMCA yielded a four-principal component PLS  Figure 6). The model showed excellent observed vs. predicted linearity (R 2 = 0.99) in the whole concentration range and was validated using independently prepared samples. This multivariate calibration was subsequently used to quantify the concentration of zinc before and after LLE. The analysis showed that the overall depletion of zinc in the aqueous phase after LLE was less than 10%. This was independently confirmed by the mass balance measurement and 65 Zn radiotracing.

Discussion
The substituents on Ar 1 , Ar 2 , and the phosphonic acid affect the properties of compounds 1-9 in two major ways. The electron-withdrawing CF 3 groups increase the acidity of the phosphonic acid moiety, whereas electron-releasing tBu and OMe retard dissociation. This is reflected in the calculated pKa value, which varies by more than 1 pKa unit across the series. Experimentally, we found that the increase in the strengths of the phosphonic acid moiety due to CF 3 substitution resulted in compounds 8 and 9 being stronger extractants. This is in line with the general observation that the potassium salts of compounds 1-9 proved to be better extractants than the corresponding acids. The effect was particularly pronounced in 4: EE increased from 8% for the phosphonic acid to 75% for its potassium salt. A similar effect was previously observed by Jagodić and attributed to an increase in the acidity of the solution due to the release of H + during complexation [30]. Although this explanation is not applicable to the present case due to the picomolar amount of 68 Ga, the importance of phosphonic acid dissociation is further highlighted by the critical role aqueous phase acidity played in extraction. When 1 was used as an extractant, EE dropped from 30% in 0.01 M HNO 3 to 4.2% in 0.2 M HNO 3 , and then to 1.2% in 1 M HNO 3 . Despite a qualitative correlation between pKa and EE, no statistically significant model emerged from the data. The effect of substituents in Ar 1 and Ar 2 on the solubility of compounds 1-9 is even more subtle, as both electronic and steric effects play a role. The substitution of an octyl group for a hydrophilic diethylene glycol significantly increased the aqueous solubility of 8 and 9. Although no model was able to correlate solubility and EE, the combination of COSMO-RS-calculated acidity and solubility yielded a statistically significant PLS model ( Figure S3). The combination of higher acidity and higher aqueous solubility favored extraction. A marked solvent, temperature, and concentration dependency can be further rationalized in terms of extractant dimerization, which in the case of alkyl and aryl phosphonic acids has been observed in solvents of low polarity [35]. The negative correlation between EE and the square of the concentration of the extractant we found during batch optimization ( Figure 4) suggests that dimerization of extractant 8 competes with extraction. Under this scenario, the rate of dimerization would be proportional to the square of the concentration and would lead to a decrease in EE due to the deactivation of the extractant. Increased temperature is expected to favor the dissociation of the dimer, leading to higher EE. The higher EE of the potassium salts of 1-9 we observed in the preliminary experiments also supports the dimerization hypothesis because the deprotonated species are unable to form aggregates.
Aqueous solutions are essentially transparent to Raman scattering. In our continuous flow design, the laser light was delivered to the flow cell via fiber optics, making Raman spectroscopy an ideal tool for remotely controlled in-line analysis of radioactive mixtures under continuous flow conditions. The Raman spectra of 0.3-1.2 M Zn(NO 3 ) 2 prepared in 0.01 M HNO 3 solutions were dominated by the symmetric stretching band of nitrate anion centered at 1000 cm −1 (Figure 6). The broad peak spanning 305-399 cm −1 was assigned to the hexaaquazinc(II) ion [Zn(H 2 O) 6 ] 2+ (390 cm −1 ) [36] and the lower-frequency mode corresponding to the nitrate-associated [Zn(H 2 O) x NO 3 ] + , x < 6 [37]. We found that inclusion of the entire spectral region (3-3600 cm −1 ) and autoscaling the variables to unit variance were essential for obtaining calibrations with the best possible statistical quality.

Instrumentation and Methods
The NMR spectra were recorded on an Agilent 400 MR spectrometer (Agilent, Santa Clara, CA, USA) operating at 400.445 MHz (1H). The Raman spectra were obtained using an Avantes AvaSpec-ULS-RS-TEC spectrometer with 788 nm laser excitation. Zinc-65 and gallium-68 were quantified by gamma spectroscopy using a Princeton Gammatech LGC 5 or Ortec GMX 35195-P germanium detector, calibrated using certified barium-133 and europium-152 sources. Zinc at natural abundance was quantified using a Thermo Scientific iCAP 6000 Series ICP Optical Emission Spectrometer. An Eppendorf 5702 centrifuge was used to assist in phase separation. All experiments used 0.2 µm membrane pore size, a 0.002" (0.051 mm) diaphragm, two 10-element static mixers, and a 108 cm mixing tube. The solutions for the continuous membrane-based separation were pumped using KDS 100 Legacy Syringe pumps. For batch experiments, phase mixing was performed using an IKA ROCKER 3D digital shaker.

Batch LLE Extractions
Initially, 1 M zinc nitrate prepared in 0.01 M HNO 3 was used as the aqueous phase for the batch LLE extraction. The organic phase was prepared by dissolving extractants 1-9 in different organic solvents or a mixture of solvents to achieve a final concentration of 10 mM, 20 mM, and 30 mM. A 1 mL aliquot of the aqueous phase was then transferred into a 15 mL centrifuge tube and mixed with 3 mL of the organic phase. The tube was shaken at 80 rpm for 15 min at either 25 • C, 37.5 • C, or 50 • C. After centrifugation for 15 min, the phases where separated and the activity of 65 Zn and 68 Ga in the aqueous and organic phases were quantified using gamma spectroscopy. 68 Ga extraction efficiency was calculated according to the following equation: EE(%) = ( 68 Ga org × 100%)/( 68 Ga org + 68 Ga aq )

Continuous Membrane-Based LLE
LLE in flow was performed in two stages. At first, the extraction stage used the same aqueous phase as was used in batch LLE. Extractants 8 and 9, which showed the best performance in batch extraction, were used as 10 mM TFT solutions for the organic phase. The extraction was performed at 50 • C. The fluidics were driven by two separate syringe pumps equipped with Hamilton glass syringes, which were filled with the organic (9 mL) and the aqueous (3 mL) phases. The aqueous flow rate was 15 mL/h and the organic flow rate was 45 mL/h for all experiments. The aqueous to organic ratio was maintained at 1:3 (v/v) at all times. For the extraction process, two phases passed through PFA tubing (1/16" OD, 0.03" ID) and entered the Syriss Chip Climate controller mixer set at 50 • C. The organic and aqueous phases were mixed inside the chip. The two phases were further mixed in a 100 cm long PFA tubing (1/16" OD, 0.03" ID) mixing loop by steady slug flow and then passed into the membrane separator. In the membrane separator, the organic phase permeated the hydrophobic membrane (PTFE/PP, 0.2 µm pore size and 139 µm thickness) and passed through the permeate outlet, while the aqueous phase was retained and passed through the retentate outlet. The 0.002" PFA diaphragm in the membrane separator worked as a form of integrated pressure control, and complete phase separation between the aqueous and the organic phase was obtained with the chosen membrane and diaphragm. In the second stage (stripping), 68 Ga was back-extracted from the organic phase into the aqueous phase following the experimental protocol described above. This time, however, the organic and aqueous phases were mixed by two PTFE static mixers at room temperature. Again, the aqueous to organic ratio was maintained at 1:3 (v/v), where 9 mL of organic phase containing 68 Ga was stripped with 3 mL of 2 M HCl.

General Comments
NMR spectra were recorded at ambient probe temperatures and referenced as follows (δ, ppm): 1 H, residual internal DMSO-d5 (2.50); 13 C{ 1 H} internal DMSO-d6 (39.52). The partial structural assignment was performed on a basis of one-dimensional nuclear Overhauser effect spectroscopy (1D NOESY), Heteronuclear Multiple Bond Correlation with adiabatic pulses (HSQCAD), and Gradient-Selected Correlation Spectroscopy (gCOSY) experiments. qNMR studies were performed at ambient temperature using the following acquisition parameters: 1 H NMR-wet1D pulse sequence with 90 o pulse, acquisition time 4 s, relaxation delay 60 s, zero-filling to 256 k, exponential multiplication with 0.3 Hz line broadening, manual phasing, and 5th degree polynomial baseline correction. 31 P NMR-inverse gated decoupling pulse sequence with 90 • pulse, acquisition time 1 s, relaxation delay 30 s, zerofilling to 128 k, exponential multiplication with 3 Hz line broadening, manual phasing, and 5th degree polynomial baseline correction.

Determination of the Aqueous Solubility of Extractants 1-9
A 1 mL volumetric flask was charged with 10-15 mg of the extractant and 1 mL of D 2 O. The resulting suspension was sonicated for 30 min and the content was centrifuged, filtered through a 0.45 µm syringe filter, and transferred into an NMR tube. A sealed capillary containing an external calibrant (acetanilide for 1 H qNMR and triphenylphosphine for 31 P qNMR) was inserted into the NMR tube and the solubility of extractants 1-9 was determined from their 1 H or 31 P NMR spectra using the formula: C x = (I x /I cal ) × (N cal /N x ) × C cal , where I, N, and C are the integral area, number of nuclei, and the concentration of the extractant (x) and the calibrant (cal), respectively.

Multivariate Analysis
The Design of Experiment (DoE) studies were performed using MODDE 9.1.1 (Umetrics AB), using the central composite face-centered design. Multivariate calibration of the zinc nitrate concentrations was performed by acquiring the corresponding Raman spectra using 788 nm laser excitation with 5 s acquisition (10 average) and exporting the spectra as GRAMS SPC files into SIMCA 17.0.0 (Sartorius Stedim Data Analytics AB). The dataset (3-3600 cm −1 ) was scaled to unit variance and processed using PLS. Five independently prepared samples were used for validation of the calibration model.

Computational Methods
All gas-phase and COSMO calculations were performed using the TURBOMOLE 7.5.1 suite of programs using resolution of identity approximation (RI) [39]. The gas-phase structures were optimized at the RI BP/def2-TZVPD level and convergence to the ground state was verified by running analytical frequency calculations. The single-point gas-phase energies were then re-evaluated at the RI MP2/def2-TZVPP level. The COSMO files were obtained at the same theory level in the COSMO phase with a smooth radii-based isosurface cavity, and convergence to the ground state was verified by running numerical frequency calculations. The resulting COSMO files were used to perform solution thermodynamics calculations using COSMOtherm, Version 20.0.0 (Dassault Systèmes), yielding the free energies of solvation, sigma surfaces, and sigma profiles [40].

General Procedure for the Synthesis of Arylaminophosphonic Acids 1-9
A 100 mL round-bottom flask equipped with the Dean-Stark trap was charged with an equimolar (5 mmol) amount of a dialkylphosphite, an aniline, an aldehyde, and a catalytic amount of pTSA dissolved in 40 mL of toluene. The reaction mixture was refluxed overnight. Toluene was removed under reduced pressure, the reaction mixture was redissolved in 50 mL of ethanol, and a solution of 10 mmol of KOH in 4 mL of water was added. The reaction mixture was refluxed overnight, cooled to room temperature, acidified with 3 eq (15 mmol) of 6 M HCl, and purified on silica (gradient elution: CHCl 3 then CHCl 3 /CH 3 OH 9/1 v/v). Analytical purity was confirmed by qNMR as described above.

Conclusions
A new family of arylaminophosphonic acids (1-9) was synthesized. All compounds selectively extracted gallium-68 in the presence of zinc from a solution containing 1 M Zn(NO 3 ) 2 in 0.01 M HNO 3 . Extractants 8 and 9 were able to extract up to 90% of gallium in batch and up to 80% in flow. Only ppm amounts of zinc were co-extracted. Extraction efficiency correlated with pKa and the aqueous solubility of the extractant. Raman spectroscopy was found to be well suited as a PAT tool for continuous flow production of gallium-68. We are currently evaluating the suitability of cyclotron-produced and LLEFpurified gallium-68 for DOTATATE radiolabeling.
Supplementary Materials: The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/molecules27238377/s1. Figure S1: The extraction efficiencies for compound 8, performed at 10 mM and 30 mM in various solvents at RT and 50 • C. The relative increase in EE upon the increase in concentration of compound 8 from 10 to 30 mM is shown on the X axis.  Figure S2: The design region for the central composite face-centered design shown in Table S1. The numbers in the circles correspond to the experiment numbers in Table S1. The center point was run in triplicate (experiments 9, 10, 11). Figure