Kinetic Treatments for Catalyst Activation and Deactivation Processes based on Variable Time Normalization Analysis

Abstract Progress reaction profiles are affected by both catalyst activation and deactivation processes occurring alongside the main reaction. These processes complicate the kinetic analysis of reactions, often directing researchers toward incorrect conclusions. We report the application of two kinetic treatments, based on variable time normalization analysis, to reactions involving catalyst activation and deactivation processes. The first kinetic treatment allows the removal of induction periods or the effect of rate perturbations associated with catalyst deactivation from kinetic profiles when the quantity of active catalyst can be measured. The second treatment allows the estimation of the activation or deactivation profile of the catalyst when the order of the reactants for the main reaction is known. Both treatments facilitate kinetic analysis of reactions suffering catalyst activation or deactivation processes.

HPLC were recorded on a Thermo Scientific UHPLC Dionex Ultimate 3000. Mass Spec were recroded in an Exactive Plus EMR (Extended Mass Range) Orbitrap using HESI method in positive mode.

Experimental setup
Hydroformylation reactions were carried out in a 25 mL 316SS autoclave from Berghof equipped with inlet and outlet sample lines. Sample circulation was carried out using a HPLC pump (JASCO-PU-1580). The temperature of the reaction mixture was stabilized and kept constant with a Cole-Parmer Polystat Refrigerated Circulating Bath. The hydroformylation mixture was circulated into the spectrometer using an InsightMR flow tube (4 meters) from Bruker. The syngas cylinder (CO/H2 1:1 ratio) and the pressure regulator was purchased from BOC Group. Fluorinated ethylene propylene (FEP) tubing was used ((IDEX-Upchurch Product #1692: FEP Capillary 1/32" OD (0.80 mm) x 0.016" ID (0.405 mm))) and the length of tubing connecting the reactor and pump to the spectrometer was minimized. The FEP tubing is a non-reactive material and offers chemical and mechanical stability (up to 50 °C, up to 121 bar) along with low gas permeability and good flexibility. Fittings used in the system were stainless steel from Swagelok or Upchurch PEEK (polyetherether ketone) from Tecknocroma.
All the equipment was positioned 0.5 m outside the 5 Gauss line of the spectrometer's magnet.
The high-pressure flow system configuration used for hydroformylation reactions is shown in Fig. S1. A 25 mL Berghof autoclave equipped with inlet and outlet sample lines (blue lines) was used. The reaction mixture was circulated from the reactor to the HPLC pump, then to the InsightMR flow tube and back to the reactor employing FEP tubing as sample lines. The temperature was kept constant throughout the whole system (i.e., reactor, spectrometer and InsightMR flow tube). The reactor was connected to the high-pressure cylinder using stainless tubing (black lines), Swagelok fittings and valves. The design of the system allowed both the pressurization with different gases (N2 and syngas) and depressurization by connecting the system to an exhaust. Due to the toxicity and flammability of syngas, CO and H2 detectors were placed in the vicinity of the reactor, syngas cylinder and exhaust. Figure S1. Experimental setup for on-line monitoring of high-pressure reactions.
For the hydroformylation reactions, the corresponding temperature value (40 °C) was set in the thermostat, NMR spectrometer and heating plate. The system was purged three times with 2 bar of N2. With a positive flow of N2 (1 bar) in the reactor, the sample inlet valve was opened and solution A was injected followed by solution B. The stirring rate was set at 800 rpm. The reaction mixture was circulated under nitrogen at 5 mL/min for 5 min. NMR spectra were recorded while flowing under nitrogen atmosphere. The system was purged three times with 10 bar of syngas (CO/H2 1:1 ratio) and the syngas cylinder was left open during the reaction monitoring. Once the experiment was finished, the syngas cylinder was closed and the reactor was depressurized. The thermostat, NMR spectrometer and heating plate were switched off, the crude mixture was removed from the reactor and the sample lines were cleaned. For sample lines cleaning, acetone was injected at 5 mL/min followed by air flushing until complete dryness.
Enantioselectivity was determined by GC analysis, performed on an Agilent 6890N equipped with a FID detector using chiral stationary phases. GC analysis of the crude mixture confirmed the already reported regio-and enantioselectivity. [1] S-4      Table S1. Figure 2c shows the reaction profile corrected by the amount of active catalyst measured as a hydride of the supramolecular rhodium complex at each time point. The time values shown in Table S1 have been calculated using the VTNA formula:  Figure 4a shows the original temporal profile of the concentration of product obtained from the integration of consecutive 1 H-NMR spectra and the corresponding data is shown in Table S2. Figure 4b shows the linearized VTNA profile resulting from the normalization of the time scale by the concentration of starting material 1 (vinyl acetate) and the concentration of estimated active catalyst. All the corresponding numerical data is shown in Table S2.

NMR spectra
S-14 Figure 4c shows the comparison of the estimated temporal profile of percentage of catalyst and the original concentration measured from the 1 H-NMR spectra. The values for the estimated concentration profile were found using Solver. In this case, we imposed the condition that the concentration of catalyst could not decrease during the reaction. To do so in Excel, the concentration of catalyst at a given time was defined as the concentration in the previous time plus a given increment. Specifically, these increment values represented the 'Variable Cells' changed by Solver in order to maximize the R 2 (SQR function: square of the Pearson product moment correlation coefficient through data points (concentration of product, normalized time)) of the resulting VTNA profile shown in Figure 4b.

Experimental setup
The Michael reaction was setup at room temperature in standard NMR tubes from stock solutions prior to data acquisition within the NMR spectrometer. To a sample of freshly vacuum distilled propanal (3)  The sample tube was then shaken thoroughly and the reference capillary returned to the tube before loading into the NMR spectrometer and the kinetic data collected using an automated shim-acquisition sequence. The timing between the catalyst addition and the start of the acquisition sequence was noted for use in the kinetic calculations. Spectra were collated and product 6 and catalyst concentrations measured by peak integration within the limits of the data (see NMR spectra below).
S-20     Table S3.    Figure 5a shows the original temporal profile of the concentration of product obtained from the integration of consecutive 1 H-NMR spectra and the corresponding data is shown in Table S4. Figure 5b shows the linearized VTNA profile resulting from the normalization of the time scale by the concentration of estimated active catalyst. All the corresponding numerical data is shown in Table S4. Figure 5c shows the comparison of the estimated temporal profile of percentage of catalyst and the original concentration measured from the 1 H-NMR spectra. The values for the estimated concentration profile were found using Solver. In this case, we imposed the condition that the concentration of catalyst could not increase during the reaction. To do so in Excel, the concentration of catalyst at a given time was defined as the concentration in the previous time minus a given increment. Specifically, these increment values represented the 'Variable Cells' changed by Solver in order to maximize the R 2 (SQR function: square of the Pearson product moment correlation coefficient through data points (concentration of product, normalized time)) of the resulting VTNA profile shown in Figure 5c.

Study of catalyst deactivation pathways
Michael addition of propanal (3) to trans-β-nitrostyrene (4) is generally performed with 10-20 mol% catalyst loading due to catalyst deactivation processes. When the reaction was run at 0.5 mol% of catalyst 5, the reaction failed to reach completion, which could be due to catalyst deactivation.

Proof of catalyst deactivation
In order to prove that catalyst deactivation was the reason for the uncompleted reaction, we performed three different experiments: same excess, same excess with product added and extra addition of fresh catalyst. The same excess reaction with product added is not just quicker than the standard reaction when [3] = 0.6 M, but it is also quicker than the same excess reaction. This proves that the reaction, rather than suffering product inhibition, is accelerated by the presence of product. Therefore, the reason why the standard reaction at [3] = 0.6 M was slower than the same excess reaction should be catalyst deactivation.
We also confirmed catalyst deactivation by adding 0.5 mol% of fresh catalyst to the standard reaction once it was not progressing anymore. The reaction immediately recovered a good rate until the complete consumption of starting material (catalyst added).

Deactivation of catalyst 5 by desilylation
A well-known mechanism of deactivation of the Jørgensen-Hayashi catalyst 5 is the cleavage of O-Si bond. [3] The desilylation of the catalyst would yield aminoalcohol 12, which could be present at the end of the reaction as it is or as a product of reaction with propanal (3), trans-βnitrostyrene (4) or product 6. The Michael reaction is run under acidic conditions ([AcOH]o = 0.1 M) and becomes more acidic over time due to the formation of product 6. Therefore, if aminoalcohol 12 is present at the end of the reaction it could be protonated. However, the presence of aminoalcohol 12 in significant amounts in the reaction crude was discarded based on the absence of the characteristic peaks of H2 and H5 in the HSQC of the crude of the reaction ( Figure S20).
The presence of significant amounts of the oxazolidinones 13 and 15 in the reaction crude after catalyst deactivation was discarded based on the comparison of NMR spectroscopic data of the oxazolidinones and the crude of the reaction. The main indicator was the absence of the characteristic H1, H2 and H5 peaks of the oxazolidinones 13 and 15 in the HSQC spectrum of the reaction mixture containing the deactivated catalyst ( Figure S21).
In the presence of AcOH, only protonated aminoalcohol 12 was observed ( Figure S22).
In the absence of AcOH, the addition product 14 was formed ( Figure S23).

S-35
Characterisation of products formed by reacting aminoalcohol 12 with trans-β-nitrostyrene (4) Reaction of aminoalcohol 12 with trans-β-nitrostyrene (4) in toluene-d8 resulted in the formation of new signals in the 1 H NMR spectrum. The signals of the newly formed species were characterised by 2D NMR spectroscopy including COSY, HSQC and HMBC. The HMBC cross-peaks H6-C5 and H6-C2 imply that the addition to trans-β-nitrostyrene (4) proceeds through nucleophilic attack by the N atom of the aminoalcohol 12.
When extra aminoalcohol 12 (1.52 mg, 0.5 mol%) was added to the crude reaction (with the deactivated catalytic species) none of the peaks observed in the crude reaction mixture increased and two new multiplets appeared. These two facts discarded the desilylation as a significant pathway of catalyst deactivation.

Deactivation pathways without desilylation of the catalyst
The catalyst can get deactivated by forming inactive, stable out-cycle species. These species could be formed by reaction of any on-cycle catalytic species with the wrong reaction partner. These could be due to the reaction of: a) Catalyst 5 with trans-β-nitrostyrene (4).
We have investigated these possible deactivation reactions by pre-mixing the corresponding reaction components and, in the case of appearance of new species, by testing their catalytic activity under reaction conditions.
In the presence of AcOH, only protonated catalyst 5 was observed ( Figure S29).
In the absence of AcOH, no reaction occurred ( Figure S30). These two reaction components did not react in toluene-d8 either in the absence or the presence of AcOH. Therefore, this is discarded as a deactivation pathway of the reaction. Over time, products derived from the self-aldol reaction 16, 17 and 10 were observed ( Figure  S31). The formation of 10 as a side-product of the aldol reaction was already mentioned by MacMillan et al. [5] (Check page S-60 for the spectroscopic partial characterisation of 16, 17 and 10). After 5 h of pre-mixing, the catalyst has been modified and new products are present (new TMS peaks increased over time Figure S32). Therefore, we checked the reactivity of the catalytic species under these reaction conditions by adding trans-β-nitrostyrene (4) (53.64 mg, 0.36 mmol) and the amount of propanal (3) consumed during self-condesation (10.30 mg, 0.18 mmol). The same excess reaction with pre-mixing propanal (3) with catalyst 5 did not deactivate the catalyst due to curve overlap with the same excess reaction ( Figure S33). Therefore, self-condensation side-products formation were discarded as a significant pathway of catalyst deactivation.   After 5 h of pre-mixing, the catalyst has been modified and new products are present (new TMS peaks increased over time Figure S35). Therefore, we checked the reactivity of the catalytic species under these reaction conditions by adding trans-β-nitrostyrene (4) (53.64 mg, 0.36 mmol). The rate of the reaction after pre-mixing catalyst 5, propanal (3) and product 6 was clearly slower than the rate for the same excess reaction with product added 6, but quicker than the standard reaction at [3] = 0.6 M. These facts are consistent with a partial deactivation of the catalyst, which may be relevant during the reaction ( Figure S36). Unfortunately, we were unable to characterise the inactive catalytic species by NMR spectroscopy.

S-48
After 18 h of pre-mixing, it is clear that the catalyst has been modified and new products are present (new TMS peaks increased over time Figure S38). Therefore, we checked the reactivity of the catalytic species under these reaction conditions by adding propanal (3) (20.88 mg, 0.36 mmol). The rate of the reaction after pre-mixing the catalyst 5 with product 6 and trans-β-nitrostyrene (4) overlays with the rate of the reaction for the same excess reaction with product added ( Figure S39). Therefore, the formation of side-products from the reaction of the enamine of product 6 with trans-β-nitrostyrene (4) or the reaction of the enamine of product 6 with product 6 were discarded as a significant pathway of catalyst deactivation.

S-49
Pre-mixing catalyst 5 with all the possible combinations of reaction components for 5 h did not deactivate the catalyst to the same degree as observed in the reaction. This observation suggests that the major pathways of deactivation, operating during the reaction under standard conditions, may also involve a catalytic intermediate different from the free catalyst, enamine of the propanal, and enamine of the product.
Another important catalytic intermediate in this reaction is the cyclobutane 18, which is in equilibrium with the putative zwitterionic iminium nitronate 7. [6] Intermediate 7 can only exist as a transient species during the reaction and it is very unlikely to be formed by mixing catalyst 5 and product 6. Intermediate 7 could react with propanal (3), trans-β-nitrostyrene (4) and product 6 forming very stable 6-member ring cycles (8, 9 and 19), which would trap the catalyst. Due to the small amount of catalyst used in the standard reaction and the variety of possible diastereoisomers of products 8, 9 and 19 we have been unable to confirm the presence of these intermediates by NMR (spectra were recorded on a Bruker AVIII 800 spectrometer equipped with a TCI helium-cooled cryoprobe). Therefore, we decided to monitor the reaction by accurate MS. S-50