Membrane-based TBADT recovery as a strategy to increase the sustainability of continuous-flow photocatalytic HAT transformations

Photocatalytic hydrogen atom transfer (HAT) processes have been the object of numerous studies showcasing the potential of the homogeneous photocatalyst tetrabutylammonium decatungstate (TBADT) for the functionalization of C(sp3)–H bonds. However, to translate these studies into large-scale industrial processes, careful considerations of catalyst loading, cost, and removal are required. This work presents organic solvent nanofiltration (OSN) as an answer to reduce TBADT consumption, increase its turnover number and lower its concentration in the product solution, thus enabling large-scale photocatalytic HAT-based transformations. The operating parameters for a suitable membrane for TBADT recovery in acetonitrile were optimized. Continuous photocatalytic C(sp3)-H alkylation and amination reactions were carried out with in-line TBADT recovery via two OSN steps. Promisingly, the observed product yields for the reactions with in-line catalyst recycling are comparable to those of reactions performed with pristine TBADT, therefore highlighting that not only catalyst recovery (>99%, TON > 8400) is a possibility, but also that it does not happen at the expense of reaction performance.

6.3 (Ir[(dF(CF 3 )ppy)] 2 (dtbpy))PF 6  In the NMR spectra the following abbreviations were used to describe the multiplicity: s = singlet, d = doublet, t = triplet, q = quartet, p = pentet, h = hextet, hept = heptet, m = multiplet, dd = double doublet, td = triple doublet. NMR data was processed using the MestReNova 12.0 software package. Known products were characterized by comparing to the corresponding 1 H NMR and 13 C NMR from literature. GC-MS data was recorded on a Shimadzu GC2010 Plus gas chromatograph coupled with GCMS-QP2010-SE mass spectrometer using a SH-Rtx-5 Amine column (30 m × 0.25 mm × 0.25 μm). High-resolution mass spectra (HRMS) were collected on an AccuTOF GC v 4g, JMS-T100GCV Mass spectrometer (JEOL, Japan). UV-Vis spectra were recorded on an inline UV-Vis spectrometers from AvaSpec StarLine family of Avantes. The names of all products were generated using the PerkinElmer ChemDraw 20.1 software package.

General procedure for membrane screening
In this paper, five organic solvent nanofiltration membranes were selected to perform the TBADT recovery. The general procedure for membrane screening was as shown in

Sample preparation for nanofiltration membranes
The processing of each nanofiltration membrane follows those steps below.
1. Always place the membrane between two protecting papers before you cut it.
2. Mark the area that you want to cut on one of the papers before you cover the membrane with it. 7. Pump the CH 3 CN with HPLC pump to separator at the needed flow rate until reaching the desired pressure.

Definitions
Here is a list of the definitions of the items used in this paper. These definitions apply for both the main text and the Supporting Information.

Permeate flux 2
Permeate flux equals to the volume of liquid that permeates the membrane per unit of time, per membrane area, as shown in Equation S1: where V stands for volume (mL).

Permeate/retentate ratio (P/R)
The permeate/retentate ratio (P/R ratio) is defined as the mass of the permeate solution over the mass of retentate solution as shown in Equation S2: where m stands for mass (g).

TBADT recovery
TBADT recovery equals to the amount of TBADT in retentate divided the input amount of TBADT, as shown in Equation S3: where R stands for retentate, Q in is the inlet flow rate of TBADT solution (mL/min); t is the operation time (min).

Product recovery
Product recovery equals to the amount of product in permeate divided the input amount of product, as shown in Equation S4: where P stands for permeate, , is product concentration of fraction i of permeate solution (M); , is volume of the fraction i of permeate solution (mL); 0 is the product concentration of the inlet solution pumped into the separator (M).

Fraction yield in permeate 3
Fraction yield in permeate equals to the moles of product in permeate solution over the input moles of starting materials, as shown in Equation S5: where P stands for permeate, , is product concentration of fraction i (M); , is volume of the permeate of fraction i (mL).
Turnover number in permeate (TON P ) 3 Turnover number in permeate equals to the moles of desired product formed in permeate over the moles of TBADT used as shown in Equation S6: Total turnover number (TON total ) 3 Total turnover number equals to the moles of desired product formed in the whole process over the moles of TBADT used as shown in Equation S7: where is the concentration of product in the solution remaining in the system (M); is the volume of the solution remaining in the system (mL).

Turnover frequency (TOF) 4
Turnover frequency equals to the moles of desired product formed in 1 hour over the moles of TBADT used as shown in Equation S8: where ∆ is the time interval for each sample.

Membrane screening
Follow the general procedure of membrane screening mentioned above (Section 2. Given the poor solubility of TBADT in the most common organic solvents, the only suitable solvent was Acetonitrile, which has limited compatibility with most commercial OSN membranes. 5, 6 We started with PuraMem ® serial membrane first. Followed with two-hour pre-processing with CH 3 CN, we tested those three type membranes with 0.15 mL/min input flow rate. Unfortunately, none of them could reach the desired pressure (14 bar). With the increasing input flow rate to 0.5 mL/min, only the experiment of PuraMem ® selective membrane can maintain 5 bar, which was still lower than the desired pressure. As shown in

High-pressure membrane separator design
All the nanofiltration experiments with higher pressure (>20 bar) and inline nanofiltration were performed with our high-pressure membrane separator (Supplementary Figure 13),

Input flow rate screening
With the operating pressure of 35 bar settled, a series of input flows rates from 0.1 mL/min to 0.3 mL/min were evaluated (see Supplementary Figure 19). A lower input flow rate equals to a longer residence time in the separator, leading to a higher permeate-retentate volumetric ratio, which also minimized the amount of product coming back to the photoreactor. However, a decreased flux was observed for the lowest flow rate, which can be explained by reduced mixing in the cell and likely concentration polarization across the membrane. Meanwhile, although slightly higher input flow rate resulted more permeate flux, the amount of product coming back to the reactor also increased in this case. Therefore, the flow rate for catalyst recovery was fixed at 0.15 mL/min.

TBADT stability test
The potential negative impact on catalyst activity of the recycling process was taken into consideration and tested by performing alkylation reactions with TBADT that was obtained either via OSN or via non-solvent extraction and filtration. The C(sp 3 )-H alkylation reaction was shown below. Firstly, the TBADT was filtered by drying the reaction solution in vacuum and then adding ethyl acetate. The recycled TBADT was tested 5 times without any noticeable loss of photocatalytic activity (Supplementary Figure 20). Besides, another experiment was done with the recycled TBADT via OSN. As presented in Supplementary Table 2, the product yield is comparable to the one achieved with pristine TBADT, which confirmed that the recycle process has no negative effect on TBADT activity.

Potentially undesired effect of membrane-reactant interactions
As the screening experiments performed above were carried out with single TBADT solution, we still need to see if the reactants may affect the nanofiltration process or not. Different

Estimation of how many NF stages needed to reach 95% product recovery
Considering the concentrations of product in permeate and retentate solution are almost the same, we only need to calculate the total permeate volume over the inlet solution volume.
Let's assume that the P/R ratio is constant and take the P/R ratio we obtained at the first stage for the derivation of the number of stages required to achieve 95% product recovery. The P/R ratio is 0.86. After first stage, we obtain 0.86 of permeate and 1 of retentate. And then we will obtain1 * (1 (1 + 0.86)) ⁄ −1 of retentate and 0.86 * (1 (1 + 0.86)) ⁄ −1 of permeate at the Nth stage. We want that ∑ 0.86 * (1 (1 + 0.86)) ⁄ −1 1 > 0.95 * 1.86, so N should be larger than 5, which means we need 5 stage recirculation to obtain 95% recovery of product.
In reality, the number may be lower given the increase in P/R for subsequent stages.
Next, CH 3 CN was added to acquire a total volume of 5 mL. The liquid was taken up with a syringe, which was then mounted on a syringe pump. The syringe was directly connected to a 3.06 mL Vapourtec Reactor (PFA capillary tubing, 750 μm inner diameter). Stock solution feed was pumped into the flow reactor with 0.153 mL/min flow rate, corresponding to 20 minutes residence time. When the syringe was fully empty, again CH 3 CN was loaded into another syringe and injected to collect all product at the end of the reactor in a flask. The solvent was removed under reduced pressure and purified by flash column chromatography on a Biotage® Isolera Four system affording the product, which was characterized by 1 H NMR and 13 C NMR.

General procedure 2 (GP2) for single run of C(sp 3 )-H amination:
A 5 mL volumetric flask was charged with diisopropyl azodicarboxylate (0.2 M, 1 mmol, 1 equiv.), tetrahydrofuran (2 M, 10 mmol, 10 equiv.) and tetrabutylammonium decatungstate (TBADT) (0.8 mM, 4 μmol, 0.4 mol%). Next, CH 3 CN was added to acquire a total volume of 5 mL. The liquid was taken up with a syringe, which was then mounted on a syringe pump. The syringe was directly connected to a 3.06 mL Vapourtec Reactor (PFA capillary tubing, 750 μm inner diameter). Stock solution feed was pumped into the flow reactor with 0.153 mL/min flow rate, corresponding to 20 minutes residence time. When the syringe was fully empty, again CH 3 CN was loaded into another syringe and injected to collect all product at the end of the reactor in a flask. The solvent was removed under reduced pressure and purified by flash column chromatography on a Biotage® Isolera Four system affording the product, which was characterized by 1 H NMR and 13 C NMR.

General procedure 3 (GP3) for inline nanofiltration:
The general procedure for inline nanofiltration experiment follows several steps below: 1. Assembly the membrane separator unit.
2. Integrate the photoreactor with the membrane separator unit.
3. Pump pure CH 3 CN to the system until the desired pressure (35 bar) was reached.

General experimental procedure for screening alkylation reaction:
The optimization of the reaction conditions was carried out by studying the alkylation of cyclohexane onto dimethyl maleate to give dimethyl 2-cyclohexylsuccinate on a 0.5 mmol scale.

TBADT loading screening
Supplementary

Cyclohexane equivalent screening
Supplementary  Supplementary table 6

General experimental procedure for screening C(sp 3 )-H amination reaction:
The optimization of the reaction conditions was carried out by studying the amination of tetrahydrofuran (THF) onto diisopropyl azodicarboxylate to give corresponded hydrazine on a 0.5-1 mmol scale.  Supplementary table 7  which provided an excellent TBADT recovery. Additionally, a decreasing permeate flux was also observed in this long-term experiment, which leads a decreased TOF over time (Supplementary Figure 29). One assumption for this is the membrane fouling happened over time, which was confirmed by checking the membrane after the experiment ( Supplementary   Figure 30(A)). Interestingly, the fouling membrane can be easily cleaned with pure acetonitrile (Supplementary Figure 30(B)

6.Preliminary investigations of other photocatalysts recovery with Nanofiltration
To further highlight the potential of this technology, three commonly used photocatalysts were selected to see if it is practical to perform recovery with the nanofiltration process.
The basic information about those photocatalysts is listed below (Supplementary table 8     The experiment with inline TBADT recovery was preformed following the general procedure The spectroscopic data are in agreement with the literature. 10 The experiment with inline TBADT recovery was preformed following the general procedure