LED-NMR Monitoring of an Enantioselective Catalytic [2+2] Photocycloaddition

Abstract

We report that an NMR spectrometer equipped with a high-power LED light source can be used to study a fast enantioselective photocatalytic [2+2] cycloaddition. While traditional ex situ applications of NMR provide considerable information on reaction mechanisms, they are often ineffective for observing fast reactions. Recently, motivated by renewed interest in organic photochemistry, several approaches have been reported for in situ monitoring of photochemical reactions. These previously disclosed methods, however, have rarely been applied to rapid (<5 min) photochemical reactions. Furthermore, these approaches have not previously been used to interrogate the mechanisms of photocatalytic energy-transfer reactions. In the present work, we describe our experimental setup and demonstrate its utility by determining a phenomenological rate law for a model photocatalytic energy-transfer cycloaddition reaction.

Graphical Abstract

Integration of LED illumination and NMR spectroscopy facilitate the mechanistic investigation of a rapid photocatalyzed [2+2] photocycloaddition.

Introduction

NMR spectroscopy is among the most powerful analytical tools available to modern chemists. This technique is especially information-rich compared to other spectroscopies and can simultaneously provide both structural information and quantification of multiple species in complex reaction mixtures. Consequently, NMR techniques have been frequently applied to provide detailed insight into the reaction mechanisms and kinetics of complex organic transformations. However, monitoring rapid reactions can be challenging for several reasons.^[1] In a typical ex situ experiment, the rate of data collection is usually limited by the need to quench aliquots and then prepare and shim NMR samples for each time point. For in situ NMR spectroscopy, the challenges include consistent and controllable initiation of the reaction such that data collection can be conducted in a reproducible manner. Photochemical reactions are a natural fit for in situ NMR, as the ability to turn the light source on or off provides temporal control over the reaction. However, such approaches have been largely neglected until recently, primarily due to the technical challenges associated with introducing light into the sample during an experiment.

In the Center for Laser-Assisted NMR,^[2] high-power argon-ion and excimer lasers have been frequently utilized for in situ NMR studies. One of the most prevalent applications of these in situ NMR experiments is Photo-CIDNP (Chemically Induced Dynamic Nuclear Polarization), a technique which has found considerable use over the last several decades to investigate organic radical reactions and protein folding.^[3] Recently, high-powered light-emitting diodes (LEDs) have proved suitable, cost-effective replacements for laser systems.^[4] These in situ illumination techniques have been used to study other photochemical reactions such as RNA or protein folding,^[5] photoisomerization,^[6,7] environmental photochemistry,^[8] polymerization,^[9] and organometallic reactions;^[10] however, the application of light-assisted NMR has been relatively limited in synthetic organic photochemistry.

Over the past decade, there has been a resurgence of interest in organic photochemistry and photocatalysis,^[11] resulting in a wide variety of powerful new synthetic methods.^[12] New reactions, however, pose new mechanistic questions, and NMR spectroscopy has the potential to provide distinctive insights. Gschwind and co-workers pioneered the use of light-coupled in situ NMR (LED-NMR), to study a variety of photoinduced electron-transfer reactions,^[13] including oxidations,^[14] reductions,^[15] and C–C bond formations.^[16] From this pioneering work, other studies have shown that the LED-NMR technique is valuable in the interrogation of both homogenous^[17] and heterogenous photocatalytic reaction mechanisms.^[18]

Despite these advances, there are still several major questions that we sought to address. First, all research to date has focused on either the direct excitation of the organic compound undergoing the reaction^[6–10] or photoredox reactions where an excited-state photocatalyst reduces or oxidizes the organic substrate.^[14–18] To our knowledge, there are no reports of photocatalytic energy transfer reactions that have been studied by in situ LED-NMR. This important class of photoreactions involves promotion of the substrate to an electronically excited state via triplet energy transfer from the photocatalyst and allows efficient access to the distinctive reactivity of electronically excited organic compounds. Second, the reactions previously studied by light-assisted NMR spectroscopy have been primarily limited to those that complete on the order of hours, rather than minutes.^[19] Some of the photocatalytic reactions being developed in our laboratory are quite fast, and thus traditional ex situ techniques do not have the resolution to capture their kinetic profiles with sufficient precision.

In this report, we detail the setup and NMR method for the light-coupled in situ kinetic investigation of an asymmetric photosensitized [2+2] reaction recently developed in our laboratory.^[20] The systematic variation of substrate concentration, photocatalyst concentration, and light intensity provided the kinetic information needed to determine the rate law of the reaction. Importantly, single NMR spectra could be acquired in 4 seconds with a time resolution of 10 seconds. This allowed kinetic information to be collected on reactions that completed in as little as 2 minutes. In contrast to traditional mechanistic investigations by ex situ NMR spectroscopy, this approach is limited only by the relaxation times of the nuclei of interest, rather than experimental considerations such as aliquot quenching, sample preparation, or shimming; with judicious choice of reaction conditions and materials, data may be acquired in less than 5 seconds. Overall, this study demonstrates that LED-NMR techniques can also be used to study rapid photocatalytic excited-state reactions.

Light-Assisted UHP-LED-NMR Design

There have been multiple approaches previously developed to introduce light into an NMR probe. Historically, light from a laser was directed through the base of the NMR probe to provide illumination from the bottom of the NMR tube.^[21] Alternatively, the addition of a quartz rod and cylindrical mirror can provide illumination from the side of the NMR tube directly within the detection region of the probe.^[22] These methods required significant modification to costly NMR probes and were restricted to single field strengths.^[13a] In recent years, fiber optic cables have been used to introduce light through the top of the NMR magnet bore.^[4b,13a] Inserting the optical fiber into the NMR tube allows the edge of the fiber to be placed directly above the detection region. Both bottom and top irradiation share the complication of ensuring the NMR sample is irradiated uniformly within the detection region. Hore and coworkers applied a stepwise taper to a fiber optic cable to provide direct irradiation within the detection region.^[23] Gschwind and coworkers achieved similar results by roughening the last few centimeters of the fiber optic through sandblasting.^[4b] These LED-NMR setups provided a significant improvement in illumination intensity within the detection region in reactions where the solutions are optically dense and absorb nearly all the light.

The system we developed directs light from an ultra high power light emitting diode (UHP-LED) through a polymer fiber optic cable directly into a 5 mm NMR tube sitting in the magnet and probe, similar to the previously described setups (Figure 1). The fiber optic cable is coupled directly to the UHP-LED, providing a reproducible, variable output power between 10–600 mW at the tip of the cable (Figure 1C). The end of the fiber optic cable is shrouded in a 4 mm glass tube to protect it from the NMR solvent and is positioned directly above the detection region of the probe. This ensures that the fiber optic does not interfere with tuning, shimming, or data acquisition. The solutions in this study were not optically dense, typically absorbing <20% of the incident light. However, the above-mentioned methods for optically dense solutions could be readily integrated into our system. All standard liquid NMR experiments can be performed in the presence of the fiber optic cable with the LED turned on or off. Thus, this in situ approach eliminates dead-time issues and enables direct monitoring of photocatalytic reactions in real time.

Figure 1.

A) UHP-LED setup, B) NMR sample with optical fiber and spinner, and C) LED power output measured at various intensities directly at the tip of the optical fiber enclosed in a 4 mm glass tube.

The standard Bruker 1D pulse program “zg30” was modified to include statements that allowed the use of TTL ports to control irradiation. Dedicated dark and kinetic pulse sequences allow for precise LED control. A typical experiment begins with the acquisition of two dark spectra in order to homogenize the protic spins in solution and accurately designate the point at which the LED is initiated. These dark experiments are immediately followed by sequential light experiments in which a high and a low TTL statement are added to the zg30 pulse program to facilitate turning on and off the LED, respectively. The pulse program (Figure 2) consists of a loop in which data are acquired and written to the disk while the LED is illuminated. In between light experiments, the low TTL command shuts the LED off, but is followed near instantaneously by the high TTL statement of the next light experiment turning back on the LED. The time the LED is off in between light experiments (1 μs) is negligible compared to the time scales of the data acquisition (10 s). The precise control of timing provided by the TTL control could be utilized to modify other 1D and 2D pulse sequences, with a range of NMR experiments available that could be coupled to in situ illumination.

Figure 2.

Modified pulse program for kinetic studies.

LED-NMR Interrogation of a Photocatalytic Energy-Transfer [2+2] Cycloaddition

Recently, we reported a highly enantioselective [2+2] photocycloaddition catalyzed by a functionalized chiral iridium photocatalyst (Figure 3A).^[20] We proposed a mechanism involving photocatalytic energy transfer (Figure 3B) in which a ground-state hydrogen bond between the photocatalyst (PC) and quinolone substrate (sub) is critical for orienting the quinolone in close proximity to the chiral iridium photocatalyst throughout the transformation. The association constant for the hydrogen-bonded complex was determined to be K_EQ = 19000 at room temperature. Thus at 1 mol% catalyst loading, more than 99% of the photocatalyst is bound to a substrate at initial reaction conditions. Upon illumination, Dexter energy transfer within the hydrogen-bonded complex is facile, with a large Stern–Volmer quenching constant (K_SV = 4700 M⁻¹ in CH₂Cl₂) leading to selective triplet sensitization of the bound quinolone. The substrate then undergoes an intramolecular [2+2] cycloaddition within the chiral environment of the photocatalyst. Finally, displacement of the chiral cyclobutane by starting material completes the catalytic cycle. Importantly, control reactions in which the N–H of the quinolone is replaced by N–Me showed no enantioselectivity, consistent with the need for strong hydrogen-bonding interactions. As the mechanism of this reaction was reasonably supported by Stern–Volmer analysis, control experiments, and computation, we identified it as an ideal model reaction to study using the in situ NMR system described above. Moreover, this would be the first photocatalytic energy-transfer reaction investigated by in situ LED-NMR. We anticipated that this technique could provide complementary information to traditional photophysical experiments such as transient absorption spectroscopy, with a considerably lower cost of implementation.^[24] Finally, we hoped that validation of the proposed mechanism could provide a basis for future investigations of more complex photocatalytic reactions.

Figure 3.

A) Triplet sensitized asymmetric [2+2] reaction. B) Proposed reaction mechanism. TET = triplet energy transfer. Asterisk denotes the excited state. ^a Yield was determined by ¹H-NMR with an internal naphthalene standard. ^b Reference 20. ^c Isolated yield.

We began by investigating the aforementioned reaction using traditional ex situ NMR spectroscopy. We chose to conduct experiments at room temperature for both convenience and to facilitate rapid data acquisition. After irradiation for 24 h under these conditions, the reaction had cleanly formed the cyclobutane product in an 85% spectroscopic yield (55% ee) with full conversion of starting material.^[25]

Figure 4 shows representative time course data for the [2+2] reaction. To match the reaction conditions as closely as possible, we performed the reaction in a 1:1 pentane:CH₂Cl₂ solvent mixture spiked with 10% CD₂Cl₂. Under these conditions, with in situ irradiation by the 450 nm UHP-LED, the reaction was complete in less than 3 minutes. To ensure quantitative data interpretation, we separately measured T1 relaxation times for the ¹H resonances of both the substrate and product (SI Figure S1). The alkyl protons on the n-butene chain of the starting material (ppm range ~2.5–4.5) had T1 relaxation times of <2.5 seconds in CD₂Cl₂. With the choice of dimethyl terephthalate as an internal standard, quantitative data were acquired with a time resolution of 10 seconds (Figure 4B).

Figure 4.

A) Selected NMR spectra for a typical in situ ¹H NMR experiment with substrate (a) and product (b) resonances denoted (5 mM substrate, 2 mol% photocatalyst, and 150 mW LED intensity). The internal standard is the singlet at 4.0 ppm. B) Concentration profile as it evolves over time for the reaction in A. Dashed line indicates when the LED was turned on, t=0.

Representative spectra acquired during the time course are shown in Figure 4A. The alkyl methylene peaks of the substrate (a) decay concomitant with the rise of signals for the cyclobutane protons (b) under constant illumination. The concentration profiles of both the substrate and product over the course of the reaction are shown in Figure 4B. As expected for a reaction with a zero-order dependence in substrate, no saturation of the rate was detected over 80% of the reaction. The reaction proceeded cleanly with good mass balance; the decay of starting material aligned well with the growth of product, and no other species were detected during the course of the reaction. Therefore, we monitored the starting material decay to facilitate accurate initial rate measurements.

We investigated three aspects of the reaction that may affect the rate: photocatalyst concentration, substrate concentration, and LED intensity. As a representative example, Figure 5 shows data from the experiments probing variation of photocatalyst concentration between 0.006–0.15 mM. This method was highly reproducible across three different trials, which were prepared and conducted on separate days. As the substrate loss was linear with time, the data could typically be fit with a linear regression through ~50% of the substrate decay without a significant change in the measured slope. Even at extremely low catalyst loadings, the reactions were still nearly complete within 10 minutes, and at higher catalyst concentrations were complete within 2 minutes. Despite these short time scales, the initial rate data extracted from these plots were highly consistent. A plot of the initial rates vs. photocatalyst concentration (Figure 6A) was linear, indicating a first-order reaction with respect to the photocatalyst concentration.

Figure 5.

A representative set of kinetic experiments in triplicate, in which the photocatalyst concentration was varied over the concentrations indicated (5 mM substrate; 150 mW LED power). The initial rates were measured through a linear regression of the initial ~50% of the reaction time course (black line).

Figure 6.

Initial rates kinetic analysis for A) photocatalyst concentration (5 mM substrate; 150 mW LED power), B) substrate concentration (0.05 mM photocatalyst; 150 mW LED power), and C) LED intensity (5 mM substrate; 0.05 mM, 1 mol% photocatalyst). Values represent the average of three trials and error bars represent the standard deviation.

We next performed similar experiments varying the initial substrate concentration and observed that the initial rates were insensitive over almost an order of magnitude of concentration change (Figure 6B). This finding aligns well with our proposed mechanism, where the photocatalyst and substrate are preorganized in a hydrogen-bonded complex prior to photoexcitation. As noted previously, the strong binding interaction ensures that at 1 mol% catalyst loading, essentially all the photocatalyst in solution is bound in this complex at initial reaction concentrations. As the substrate is present in a large excess relative to catalyst, the concentration of the [photocatalyst-substrate] complex remains effectively constant over the majority of the reaction time course. Thus, changes to the initial substrate concentration do not change the concentration of the photoactive species nor the rate of reaction.

Finally, the initial rates showed a linear dependence on LED intensity over a 300 mW range (Figure 6C). We were unable to collect meaningful data at higher intensities because the reaction proceeded to completion in less than 40 seconds. Many photocatalytic reactions conducted on preparative scale exhibit a linear relationship between photon flux and reaction rate because the solutions are optically dense, absorbing >99% of the incident light.^[13] In such a photon-limited regime, photocatalytic reaction rates would be controlled solely by the absorption of photons and should thus be zero-order in the photocatalyst.^[13b,26] This expectation is not consistent with the data in our reaction, as we observe a first-order dependence on photocatalyst concentration under identical conditions. We conclude therefore that these experiments fall outside of the photon-limited regime,^[13b,26] and thus both photon flux and photocatalyst concentration have an influence on the reaction rate law. Indeed, even at the highest photocatalyst concentration, a simple Beer’s law calculation shows that only 20% of the photons will be absorbed over a 3 cm pathlength within the NMR tube (see SI, Table S1).

The use of a controlled light source enables the rapid estimation of the quantum yield of the reaction. Recent work has provided convenient methods based on in situ, light-coupled NMR spectroscopy.^[27] These methods are typically applied under photon-limited conditions, allowing a single point initial rate measurement to be used to calculate quantum yield.^[13b] Because we were unable to reach a photon-limited regime, we combined our experimental initial rates with a correction for the fraction of light absorbed (see SI), and were able to estimate the quantum yields of the [2+2] cycloaddition at the different LED intensities shown in Figure 6C.^[28] The quantum yields were consistent across the entire illumination intensity range and provided an average quantum yield of 0.31(±0.05).

Conclusion

In conclusion, in situ LED-NMR spectroscopy was used to study the mechanism of a photocatalytic energy-transfer reaction for the first time. The reaction rate law for the model asymmetric [2+2] photocycloaddition shows a first-order dependence on photocatalyst concentration and light intensity and a zero-order dependence on substrate concentration. The results of this analysis align well with the mechanism we proposed previously involving a hydrogen-bonded substrate-photocatalyst resting state.^[20,29] Importantly, by having spectrometer control over the LED, we were able to monitor rapid reaction kinetics with precise control over reaction initiation and timing. This approach allowed time-dependent data to be collected on reactions that are complete in under 2 minutes. Overall, we have shown that in situ LED-NMR is a powerful tool for the study of photocatalytic energy transfer reactions. We expect that methods similar to the one developed here will be of great utility in the future study and development of photocatalytic and photosynthetic reactions.