Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2026 Apr 5;64(7):621–628. doi: 10.1002/mrc.70103

Integrated Reaction Monitoring and In Situ Structural Elucidation of Reactive Intermediates With Time‐Resolved NOAH NMR Supersequences

Sara M Guzman 1, Yael Ben‐Tal 1,✉, Jiayu Zhang 1, Mark Dixon 2, Jason E Hein 1,3,4,✉
PMCID: PMC13238404  PMID: 41936452

ABSTRACT

Structural elucidation of reaction intermediates and side products is often key to mechanistic understanding. However, individual isolation of these species is time‐consuming, labor‐intensive, and often technically challenging, if not impossible. Here, we present an NMR experimental workflow involving the time‐resolved acquisition and processing of multiple types of two‐dimensional NMR spectra acquired in tandem using NOAH supersequences on an active reaction. This time‐resolved NOAH nonuniform sampling method (TR‐NOAH‐NUS) enables the generation of robust concentration–time plots for kinetic analysis in parallel to structural characterization of reaction components. Here, we demonstrate the application of TR‐NOAH‐NUS on an organo‐catalyzed acylation of alcohols. Despite the short (30 min) timespan of the reaction, we are able to obtain high‐quality HSQC, HMBC, and COSY spectra with excellent (less than 15 s) temporal resolution. From these data, we can observe both an on‐cycle catalytic intermediate and an off‐cycle catalyst degradation product and can directly assign the structures of both without need for further isolation.

TR‐NOAH‐NUS enables real‐time reaction monitoring and direct structural characterization of transient intermediates. Integration of NOAH supersequences with t0‐filling minimizes reaction monitoring deadtime and provides high temporal resolution for kinetic analysis.

graphic file with name MRC-64-621-g004.jpg

Abbreviations

BTM

(−)‐benzotetramisole

COSY

correlation spectroscopy

FID

free induction decay

HMBC

heteronuclear multiple‐bond correlation

HPLC

high‐performance liquid chromatography

HSQC

heteronuclear single‐quantum correlation

MDD

multidimensional decomposition

MS

mass spectrometry

NMR

nuclear magnetic resonance

NOAH

NMR by ordered acquisition using 1H detection

NUS

nonuniform sampling

SNR

signal‐to‐noise ratio

TOCSY

total correlation spectroscopy

TR

time‐resolved

1. Introduction

Structural characterization of reactive intermediates and side products is often key to fully understanding reactions and to allow for different potential reaction pathways to be distinguished. Resultantly, NMR is a powerful reaction‐monitoring technique as it can provide quantitative kinetic information with good temporal resolution as well as nuanced structural information [1, 2]. Most reaction monitoring with NMR utilizes one‐dimensional (1D) techniques, typically 1H and 19F monitoring. Although these methods provide excellent kinetic information, they often provide insufficient structural details (due to spectral overlap of reagents, products, and intermediates with a common structural core in the 1H spectrum or due to insufficient nuclei of interest in the 19F spectra). Although some side products can be isolated and characterized independently, this process is often time‐consuming and not always straightforward; in the case of reactive intermediates, isolation is often impossible.

Two‐dimensional (2D) NMR is the gold standard technique for structural characterization [3]. However, 2D experiments are significantly more time‐consuming than 1D NMR methodologies. Several innovations over the years have enabled a reduction in this experiment time, from the original development of pulsed‐field gradient methodologies to the implementation of nonuniform sampling (NUS) [4, 5, 6] to Frydman's more recent ultrafast 2D techniques [7, 8, 9]. One important advance is the recent development of the NMR by ordered acquisition using 1H detection (NOAH) supersequences by Kupče and coworkers [10, 11, 12]. This has enabled a dramatic reduction in the time required to perform a full characterization suite through magnetization sharing between the individual spectral modules within the supersequence, resulting in the collection of up to five separate spectra in the time it usually would have taken to acquire one.

Despite their additional analytical power, implementation of 2D techniques on in situ reactions remains uncommon due to the longer timescales required for acquisition, alongside related sensitivity challenges. In 2014, Mayzel and coworkers first reported the application of time‐resolved 2D NMR to monitor dynamic systems using multidimensional decomposition (mdd) coprocessing [13]. Soon after, in 2015 Dass and coworkers reported a similar approach using compressed sensing for individual processing of multiple overlapping spectra [14]. Both these methodologies take advantage of NUS to capture a series of 2D spectra in a time‐resolved rolling‐average fashion (TR‐NUS) [13, 14, 15]. By acquiring a single 2D spectrum utilizing NUS during the entire course of a reaction and then processing a subset of the acquired data, changing the position of the processing window each time, they were able to collect a series of time‐resolved high‐resolution 2D spectra over the course of the reaction. Subsequently, a handful of mechanistic studies [16, 17, 18, 19, 20, 21] have utilized the TR‐NUS technique, for example to characterize photopolymerizations [22].

Existing reported applications of the TR‐NUS method, including novel uses of NMR supersequences in reaction monitoring, have only utilized one 2D technique at a time [23]. However, in many cases, full structural elucidation requires a combination of the information obtained from multiple 2D experiments. As there exist variants of the NOAH supersequences that are compatible with NUS [24], we reasoned that it should be possible to apply the TR‐NUS processing technique onto the data that had been collected using NOAH‐NUS to gain a series of several different types of 2D spectra, each resolved in time as the reaction progresses. Here, we demonstrate both the first application of and the utility of this combined TR‐NOAH‐NUS technique by elucidating the identity of intermediates within an organo‐catalyzed acylation reaction.

2. Results and Discussion

2.1. Enantioselective Acylation—Identification of Intermediates and Off‐Cycle Species

The enantioselective acylation of alcohols catalyzed by (−)‐benzotetramisole (BTM, 4) has been an ongoing object of study in our lab [25, 26]. Online kinetic monitoring of this reaction by HPLC‐MS revealed the formation of several distinct catalytically active (5) and inhibitory species (6), which appeared by mass balance to be associated with the starting BTM catalyst. However, we had little structural information about the species beyond what was provided by mass spectrometry, and isolation of any of these species proved nontrivial. Therefore, we turned to NMR to aid with structural characterization. Analysis of the one‐dimensional spectra and associated reaction time courses provided insufficient and ambiguous structural information. We envisioned that time‐resolved 2D data, highlighting multiple couplings, would allow us to identify these species.

2.2. Considerations for Implementation of NOAH in Reaction Monitoring

NOAH enables the simultaneous acquisition of multiple 1H‐detected 2D NMR experiments as a series of “modules” acquired using a shared recovery delay (D1), with careful preservation of magnetization and unused coherences between each module [12]. This dramatically reduces total experiment time while maintaining compatibility with conventional NMR instrumentation and processing methods, including NUS techniques. To get meaningful information from the TR‐NOAH‐NUS data, we carefully selected the modules within our NOAH experiment, as well as the order in which the modules were acquired. For full structural elucidation, we required at a minimum the connectivity information provided by both HSQC and HMBC, as well as COSY or TOCSY to allow us to differentiate between spin systems that could not be distinguished on the basis of kinetic behavior. Furthermore, we wanted to be able to plot accurate concentration–time data. We did this by taking integrals from the HSQC spectra; therefore, the NOAH sequence was initially designed to ensure that the HSQC was the first module to avoid the potential for uneven consumption of magnetization by other modules, and 1JHC was carefully chosen (145 Hz) by inspection of those 1JCH values measurable a priori to ensure it would be semiquantitative for this system (see Supporting Information Section S2, Figure S4). This assumption was later validated by mass balance and inspection of the 2D spectra acquired during the reaction.

2.3. Considerations for Implementation of TR‐NUS for Reaction Monitoring

Traditional 2D NMR acquires a uniform array of datasets along the indirect dimension, with the total number of sampled points (n = t 1 increments) determining the overall acquisition time, digital resolution, and sensitivity. These are collected sequentially and are limited by the maximum number of points to acquire (t 1max). However, if instead of a stepwise linear acquisition of t 1 increments, a sparse random subset of those points is acquired, the missing data can later be reconstructed to generate a full spectrum that has the same t 1max, and therefore digital resolution, but which was acquired in a fraction of the time. This technique is commonly known as NUS and is widely implemented for conventional NMR experiments [4, 5, 6, 27].

Time‐resolved NUS methods arise directly from this ability to reconstruct a full spectrum from a limited and unevenly distributed dataset. Any subset of points from a large randomly ordered dataset (e.g., one acquired over the course of an entire reaction) can be treated as an independent smaller sparsely distributed dataset instead (e.g., a specific time window within the reaction) and thus be reconstructed into its own spectrum, assuming Nyquist–Shannon sampling requirements are met for each subset [14]. This generates a series of time‐resolved 2D spectra reconstructed from the same number of points throughout a single reaction, each temporally resolved by the time required to acquire each t 1 datapoint [13, 14].

A number of different methods exist for determining both the NUS sampling schedule and the resultant reconstruction methods, each with different strengths and weaknesses [6, 13, 27]. Most conventional methods for creating this list generate a random subset consisting of any t 1 increment within the range provided by the number of points defined in the indirect dimension (t 1max). Since the experiment acquisition time is dependent on the total number of t 1 increments sampled, in the case of TR‐NUS performed on a dynamically changing sample where the 2D spectral acquisition must continue over the entire time course of the reaction, the t 1max increases far beyond what would be required for a static 2D spectrum. Furthermore, the acquired data are later processed as a series of separate spectra, each of which only uses a subset of the acquired data. An increase of t 1max increases the size of the total Nyquist grid and thus also the minimum number of NUS points required to reconstruct the space, limiting the temporal resolution [14]. Additionally, signal‐to‐noise ratio (SNR) is significantly higher for data points collected with smaller t 1 values. Thus, in order to balance signal and resolution, an artificial limit on the value of t 1max was imposed while keeping the number of acquired points high [14]. This resulted in a long NUS list containing multiple repetitions that allowed time‐resolved reconstruction of short, overlapping time windows. Indeed, we were able to observe a notable difference in SNR between different time courses recorded with different values of t 1max but otherwise processed identically (see Supporting Information Section S3.1, Figure S5).

Although some thought must be put into selecting the NUS sampling schedule ahead of time, TR‐NUS is primarily a postprocessing method, which enables the optimization of analysis without the requirement to perform additional experiments. A key feature here is in the selection of processing window size; i.e., how many NUS points are used to reconstruct each time‐resolved 2D spectrum. Although larger window sizes can increase the spectral resolution and SNR, this comes at the cost of temporal resolution of the kinetic profile [14]. In particular, the measurement “dead time”—the time at the beginning of the reaction which cannot be observed—varies dramatically with window size. As the first points in a reaction are typically those containing the most meaningful kinetic data, a delicate balance must be struck between SNR and temporal resolution. Additionally, exceedingly large window sizes may also lead to overaveraging, which can result in inaccurate representation of the reaction kinetics (see Supporting Information Section S3.2, Figure S6). Just as each reconstruction of the 2D spectra require an appropriate number of datapoints (dependent on the number of peaks and size of the Nyquist grid), defining the upper limit for the window size is essential to obtain reliable kinetic information.

2.4. Kinetic Monitoring by TR‐NOAH‐NUS and Elimination of Reaction Monitoring “Dead Time”

The specific reaction and conditions studied are illustrated in Scheme 1. Although this reaction is much shorter than any on which the TR‐NUS concept has previously been demonstrated (previous reactions have been in the order of hours to days) [13, 14, 18, 20, 28], we reasoned that we could nonetheless take advantage of the much shorter acquisition times required for quantitative HSQC relative to 1D acquisition [29] to enable collection of sufficiently high‐quality data.

SCHEME 1.

SCHEME 1

Open in a new tab

Acylation of chiral alcohols catalyzed by BTM (4), alongside the specific conditions used in this study.

We initially collected a two‐module NOAH (13C‐1H HSQC‐HMBC) supersequence utilizing NUS. After optimizing the postprocessing parameters, we were able to collect a series of 771 HSQC spectra, with an average SNR of 284 in the first 10 timepoints (calculated from starting Material 1) and an average SNR of 490 at the end of the reaction (calculated from Product 3, last 10 timepoints). The concentration–time plot obtained by integrating selected peaks from the HSQC showed excellent agreement with that obtained by conventional 1D kinetic monitoring of a separate reaction with the same conditions (Figure 1), including the ability to observe all three catalyst‐linked species (see Supporting Information Section 4.1, Figure S8).

FIGURE 1.

FIGURE 1

Open in a new tab

Overlaid concentration–time plots of two separate reactions monitored with either conventional 1D NMR (dark circles) or the TR‐NOAH‐NUS technique (light circles; concentration values calculated by integration of HSQC spectra).

However, even with these optimized conditions, the initial moments of the reaction, lost to both reaction assembly and to the time required to collect sufficient transients to reconstruct a 2D spectrum, are on the order of 5 min (c.f., the 2 min of “dead time” arising solely from reaction assembly in the single‐scan 1D case). As the reaction has reached completion within 40 min, this amounts to missing the first eighth of the reaction entirely.

To minimize this monitoring “dead time” we proposed to incorporate timepoints from prior to reaction initiation into our NUS spectral reconstruction (Figure 2). We acquired a 25‐min, two‐module NOAH‐NUS experiment (128 points in the indirect dimension) on a sample of the reaction solution without propionic anhydride 2; following the addition of 2, we obtained a second two‐module NOAH‐NUS experiment with 1600 points in the indirect dimension, utilizing otherwise identical parameters. These experiments (and NUS lists) were then combined to create a single 2D dataset, which could then be analyzed in the same TR‐NUS fashion as previously (Figure 3a). This novel reconstruction procedure provided a drastically decreased reaction‐monitoring dead time without sacrificing SNR or deviating time‐course “ground truth,” which improved the quality of the kinetic trends for the transient catalytic species (Figure 3b). A similar approach, utilizing signal averaging [30] rather than NUS reconstruction, should also be applicable to reduce effective dead time in 1D kinetics experiments (see Supporting Information Section S4.3, Figure S10).

FIGURE 2.

FIGURE 2

Open in a new tab

An illustration of the process of reconstructing a single reaction time course with a merged dataset comprising of data obtained from before and after reaction initiation. The sizes of the datasets have been reduced from reality to aid reader clarity.

FIGURE 3.

FIGURE 3

Open in a new tab

(a) Comparison of the time course obtained from the initial TR‐NOAH‐NUS experiment to that acquired on a separate reaction prepared similarly, utilizing the merging of an initial (20 min) NOH‐NUS experiment recorded prior to reaction initiation with that recorded after the reaction had begun. The initial experiment utilized a two‐module NOAH HSQC‐HMBC sequence with 4 scans for each transient, whereas the second experiment utilized a two‐module NOAH HMBC‐HSQC sequence with 16 scans for each transient. Trends for both reactions were obtained from integration of the time‐resolved HSQC spectra. (b) Time‐course trends for each of the three catalyst‐linked species, obtained from the merged NOAH HMBC‐HSQC experiment. (c) Elucidation of the structures of transient intermediate 5 (i, i.i) and off‐cycle Species 6 (ii, ii.i) from HMBC spectra reconstructed from distinct regions of the reaction time course. The temporal regions used to reconstruct each spectrum are illustrated in (b). Note that the signals arising from 5 cannot be distinguished from noise in the case where a single HMBC is reconstructed from the entire reaction (Figure S11). Only the key correlations are highlighted here; full details of the structural elucidation, including additional zoomed‐in perspectives on the spectra, are provided in the Supporting Information (Section S5.1, Figure S11 and Section S5.2).

As this methodology allowed for improvements in both SNR and temporal resolution, we were able to switch the order of the modules in the NOAH sequence and increase the number of scans to ensure the highest possible spectral resolution for the HMBC, which was necessary for the structural elucidation of the intermediate species (Figure 3c; Section 2.5).

It is important to note that this is a data reconstruction method; although we anticipate that it will provide excellent early‐reaction data in most instances, it is important to remember that it will be unable to provide insight into processes that begin upon reaction initiation but are complete prior to the acquisition of the first data point in the 2D FID, such as catalyst activation processes. Thus, caution and good judgment must be used when implementing this technique.

2.5. Structural Elucidation of Catalyst‐Linked Species

After determining from the HSQC that we were able to monitor all three previously observed catalyst‐linked species by 2D NMR, we turned to the HMBC module to aid us in determining the structure of all three species. Kinetic monitoring shows an initial rapid conversion of BTM Catalyst 4 to an Intermediate 5, which slowly reverts back to 4 as the reaction progresses. In tandem, we observed formation of a related off‐cycle Species 6 (Figure 3b).

Analysis of the time‐resolved HMBC spectra allowed for structural elucidation of Species 5 and 6. The resonances corresponding to each species could be distinguished from each other not only by following coupling correlations but also by each species' distinct time‐course profile. From the HSQC, HMBC, and m/z data, Species 5 was quickly identified as the acylated form of 4 (see Figure 3c, Figure 4, and Supporting Information Section S5.1, Figure S11), which is in line with previous mechanistic studies conducted on this and similar reactions [31, 32, 33].

FIGURE 4.

FIGURE 4

Open in a new tab

Proposed simplified mechanistic cycle, including the observed intermediate Species 5 and the off‐cycle Species 6. The persistence of 5 is in line with previous kinetic studies on this system [31].

Analysis of the HMBC data corresponding to off‐cycle Species 6 revealed the presence of three distinct carbonyl groups, including one at 201 ppm. Although unusual for species in this catalytic system, a peak at this chemical shift is extremely diagnostic of a thioester [34]. Further analysis of a more‐resolved spectrum obtained through the combined processing of all regions of the acquired data containing a large amount of 6 allowed for further correlations to be drawn. Final confirmation of the structure of 6 (see Figure 3c, and Supporting Information Section S5) was provided by performing a three‐module NOAH experiment containing an additional COSY module. Previous work had indicated reaction inhibition in the presence of water [33, 35]; indeed, addition of water to a 1:3 mixture of BTM (3) and propionic anhydride (2) allowed us to clearly track formation of 6 and to unambiguously assign its structure using a combination of HSQC, HMBC, and COSY (see Supporting Information Section S5.2 for full details). These results are in line with similar observations previously reported by Li and coworkers [33].

3. Conclusions

We here present the first application of the TR‐NOAH‐NUS technique, allowing HSQC, HMBC, and COSY spectra to be acquired in tandem, with good temporal resolution throughout the course of a live reaction. Analysis of these resultant spectra allowed for the identification and full structural elucidation of a true acylated intermediate species, as well as the catalyst decomposition species, without further isolation or the need for further experimentation. This structural analysis, particularly in the latter case, required the combination of different 2D experiments obtained by this method. In this case study, careful choice of 1JCH value enabled NOAH‐HSQC spectra to be acquired in a semiquantitative value, enabling the generation of time‐course data showing excellent agreement with kinetic data acquired using conventional 1D measurement. Recent work has demonstrated the development of fully quantitative HSQC spectra [36], and work is currently ongoing in our group to incorporate qHSQC techniques into the TR‐NOAH‐NUS workflow.

4. Experimental

(S)‐1‐phenylethan‐1‐ol (1) (124 μmol, 15 μL) was added to a solution of (−)‐benzotetramisole (BTM, 4) (21.4 μmol, 5.4 mg) in CDCl3 (0.5 mL). The homogenous solution was transferred to a 5 mm NMR tube. A 1H spectrum was acquired after locking and shimming, followed by a 25‐min, two‐module NOAH‐NUS (HMBC, HSQC) experiment. The sample was ejected from the spectrometer, and the reaction was initiated by addition of propionic anhydride (2) (116 μmol, 15 μL). The tube was mixed well and immediately returned to the spectrometer, where a 7‐h, two‐module NOAH‐NUS (HMBC, HSQC) experiment was started without further shimming. The NUS list was generated using mddNMR 2.7 software [6], with NImax = 200 for both experiments and NI = 32 for the first NOAH experiment and NI = 400 for the second. All TR‐NUS processing was performed using the mdd‐NMR software. Further details can be found in the Supporting Information (see Section S1, Figure S1).

Author Contributions

M.D., Y.B.T., and J.E.H. conceptualized the project. S.M.G. devised and executed all experimental work and analysis. J.Z. helped with acquiring HPLC‐MS data and established the reaction conditions. Y.B.T. and S.M.G. contributed to the preparation of the manuscript and the Supporting Information. All authors have given approval to the final version of the manuscript.

Funding

Financial support for this work was provided by the Canada Foundation for Innovation, CFI‐44843, CFI‐35833; the Natural Sciences and Engineering Research Council of Canada, RGPIN‐2021‐03168; and The University of British Columbia (UBC). This research was undertaken thanks in part to funding provided to the University of Toronto's Acceleration Consortium from the Canada First Research Excellence Fund, CFREF‐2022‐00042.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Flowchart for the set‐up and processing of time‐resolved NOAH experiments using mddNMR. Refer to mddNMR manual for more details about the software. (a) For discussion of considerations underlying parameter choice see section 2.3 of manuscript. (b) For discussion of reaction‐monitoring “dead time,” see Section 2.4 of manuscript. (c) For t 0 filling, run an experiment with at least the same number of TD1 points as the maximum window size used during post processing. This guarantees at least one “true t 0” point.

Figure S2. (A) Proposed simplified mechanistic cycle, including the observed intermediate Species 5 and the off‐cycle Species 6. The persistence of 5 is in line with previous kinetic studies on this system.4 (B) 1H–13C HSQC experiment resulting from the overall reaction‐monitoring time. Regions used to get integrals of each compound of interest are highlighted. (C) Zoomed‐in region of 1H–13C HSQC showing integral used for 5. (D) Zoomed‐in region of 1H–13C HSQC showing integral used for 6. NOAH 2‐ SpB: d1 = 1.5 s, NS = 4, full TD1 per module without NUS = 1600. TD1 per module with NUS = 800, NI = 400, NImax = 200.

Figure S3: 1H NMR of the reaction with zoomed‐in regions. In purple starting materials and product (1, 2, 3). In blue, a zoomed‐in spectra of the catalytic species of interest (4, 5, 6) and other low concentration side products. (a) Zoomed‐in region from 6 to 8.5 ppm. (b) Zoomed‐in region from 3.25 to 6 ppm. (c) Zoomed‐in region from 0.75 to 3.25 ppm.

Figure S4: Zoomed in 1H NMR spectra of resonances of interest showing J‐coupling trees of the 13C satellites. (a) Starting material (1), (b) BTM catalyst (4). (c) deactivated catalyst (6). The J‐coupling chosen for the NOAH experiment (1JHC = 145 Hz) was the average between the 1‐bond H‐C couplings observed in (a) (143.5 Hz) and (b) (146.6 Hz). Spectra for 1 and 4 are from purchased chemicals, spectra for 6 was obtained at the end of the reaction described in S5.2.

Figure S5: Overlay of the reaction trends obtained from experiments run using the same numbers of acquired hypercomplex points (NI), but different full indirect dimension sizes (NIMAX, referred to as t1max in the main text). The full circles are from NIMAX 200, and the empty circles are from NIMAX 400. NOAH 2‐ SpB: d1 = 1.5 s, NS = 4, full TD1 per module without NUS = 1600. TD1 per module with NUS = 800, NI = 400, NImax = varies.

Figure S6. Overlay of the reaction trends obtained from the same experiment but processed using different window sizes. The circles are from standard window size of 30, Step 1, Overlap 29. The squares are from window size of 60, Step 1, Overlap 59. The X’s are from window size of 120, Step 1, Overlap 119. (a) Reaction trends of starting Material 1 and Product 3. (b) Reaction trends of BTM Catalyst 4. Experiment ran was a NOAH 2‐BSp: d1 = 1.5 s, NS = 16, full TD1 per module without NUS = 1600. TD1 per module with NUS = 800, NI = 400, NImax = 200.

Figure S7: Overlay of the reaction trends obtained from the same experiment but processed using different window sizes. The full circles are from window size of 15, Step 1, Overlap 14. The empty circles are from standard window size of 30, Step 1, Overlap 29. Experiment ran was a NOAH 2‐BSp: d1 = 1.5 s, NS = 16, full TD1 per module without NUS = 1600. TD1 per module with NUS = 800, NI = 400, NImax = 200 with t0 appending postacquisition processing.

Figure S8. Overlay of the reaction trends obtained from experiments run using the same numbers of acquired hypercomplex points (NI = 400), same indirect dimension sizes (NIMAX = 200), same relaxation delay (d1 = 1.5 s) but different order in NOAH modules and different number of scans (NS). The full circles are from NOAH 2‐ SpB (4 scans for each transient), the empty circles are from NOAH 2‐ BSp (16 scans for each transient) with t 0 appending postacquisition processing. Trends for both reactions were obtained from integration of the time‐resolved HSQC spectra. (A) Time‐course trends for the starting Material 1, Product 3, and Catalyst 4. (B) Isolated time‐course trends for the Catalyst 4. (C) Isolated time‐course trends for the catalyst intermediate 5. (D) Isolated time‐course trends for the catalyst degradation Product 6.

Figure S9: Overlay of the reaction trends obtained from monitoring the standard reaction using 1D and 2D NMR experiments. The 2D experiment used was a two‐module NOAH BSp sequence with 16 scans for each transient, with t0 appending postacquisition processing. Trends were obtained from integration of the time‐resolved HSQC spectra. Quantitative 1D 1H NMR experiments were run with 1 scan, 60‐s delay, and 90° pulse angle. (A) Overlay with 1H NMR experiments as acquired. (B) Overlay with 1H NMR experiments resulting from merging multiple t0 experiments recorded prior to reaction initiation with those recorded after the reaction had begun using postacquisition signal averaging of continuous 4‐scan windows.5

Figure S10: Overlay of the reaction trends obtained from monitoring the standard reaction using 1D 1H NMR experiments before (filled circles) and after (empty circles) merging multiple t0 experiments recorded prior to reaction initiation with those recorded after the reaction had begun using postacquisition signal averaging of continuous 4‐scan windows.5

Figure S11a: 1H–13C HMBC resulting from the whole acquisition throughout the reaction, with highlighted correlations for the compounds of interest. (b) Zoomed‐in region of 1H–13C HMBC showing correlations for 4 and 6. The structures show only the main correlations characteristic of the compounds of interest. NOAH 2‐ BSp: d1 = 1.5 s, NS = 16, full TD1 per module without NUS = 1600. TD1 per module with NUS = 800, NI = 400, NImax = 200.

Figure S12. Time‐course trends for each of the three catalyst‐linked species, obtained from integrating the HSQC spectra obtained from the merged NOAH experiments. Showing the time of the addition of water and highlighting the reaction times used to generate 2D spectra used for the elucidation of each species of interest. (A) Time region 15–38 min used for Species 5 elucidation. (B) Time region 50–85 min used for Species 6 elucidation.

Figure S13: (A) 1H–13C HSQC of isolated system obtained from the reaction time where 5 is at its peak conc (Figure 8A). (B) Zoomed‐in region of 1H–13C HSQC showing 5–1 cross peaks. (C) Zoomed‐in region of 1H–13C HSQC showing 5–2 cross peaks. (D) Zoomed‐in region of 1H–13C HSQC showing 5–6 cross peaks. (E) Zoomed‐in region of 1H–13C HSQC showing cross peaks in the aromatic region. Anhydride 2 correlations are boxed in gray.

Figure S14: (A) 1H–13C HMBC of isolated system obtained from the reaction time where 5 is at its peak conc (Figure 8A). (B) Zoomed‐in region of 1H–13C HMBC showing 5–6 and 5–2 correlations. (C) Zoomed‐in region of 1H–13C HMBC showing correlations from 5–6 to 5–7 and 5–4. (D) Zoomed‐in region of 1H–13C HMBC showing correlations in the aromatic region. Anhydride 2 correlations are boxed in gray.

Figure S15. (A) 1H COSY of isolated system obtained from the reaction time where 5 is at its peak conc (Figure 8A). (B) Zoomed‐in region of 1H COSY showing 5–6 correlations to 5–5 and 5–6. (C) Zoomed‐in region of 1H COSY showing 5–12 correlations to 5–11 and 5–13, as well as 5–13 correlation to 5–14. Anhydride 2 correlations are boxed in gray.

Figure S16: (A) 1H–13C HSQC of isolated system obtained from the reaction time where 6 is at its peak conc (Figure 8B). (B) Zoomed‐in region of 1H–13C HSQC showing 6–2 and 6–18 cross peaks. (C) Zoomed‐in region of 1H–13C HSQC showing 6–5 and 6–6 cross peaks. (D) Zoomed‐in region of 1H–13C HSQC showing cross peaks in the aromatic region. Anhydride 2 correlations are boxed in gray.

Figure S17: (A) 1H–13C HMBC of isolated system obtained from the reaction time where 6 is at its peak conc 7 (Figure 8B). (B) Zoomed‐in region of 1H–13C HMBC showing 6–19 correlations to 6–18, as well as 6–2 correlations to 6–1. (C) Zoomed‐in region of 1H–13C HMBC showing correlations from 6–6 to 6–4, 6–5, and 6–7, as well as 6–5 correlations to 6–8. (D) Zoomed‐in region of 1H–13C HMBC showing correlations from 6–3 to 6–1 and 6–2, as well as 6–17 correlations to 6–18 and 6–19. (E) Zoomed‐in region of 1H–13C HMBC showing correlations in the aromatic region. Anhydride 2 correlations are boxed in gray.

Figure S18: (A) 1H COSY of isolated system obtained from the reaction time where 6 is at its peak conc (Figure 8B). (B) Zoomed‐in region of 1H COSY showing 6–2 correlations to 6–1, as well as 6–18 correlation to 6–19. Anhydride 2 correlations are boxed in gray.

MRC-64-621-s001.docx (10MB, docx)

Contributor Information

Yael Ben‐Tal, Email: ybental@chem.ubc.ca.

Jason E. Hein, Email: jhein@chem.ubc.ca.

Data Availability Statement

The data that support the findings of this study are openly available in GITlab code base at https://gitlab.com/sguzmanc/2D_fid_append.

References

  • 1. Ben‐Tal Y., Boaler P. J., Dale H. J. A., et al., “Mechanistic Analysis by NMR Spectroscopy: A User's Guide,” Progress in Nuclear Magnetic Resonance Spectroscopy 129 (2022): 28–106. [DOI] [PubMed] [Google Scholar]
  • 2. Carreras J., Caputo M., Colasurdo D., Pila M., Ruiz D., and Laurella S., “Elucidation of Organic Reaction Mechanisms Using Nuclear Magnetic Resonance: A Review,” Applied Magnetic Resonance 55 (2024): 1335–1376. [Google Scholar]
  • 3. Elyashberg M., “Identification and Structure Elucidation by NMR Spectroscopy,” Trends in Analytical Chemistry 69 (2015): 88–97. [Google Scholar]
  • 4. Kaur M., Lewis C. M., Chronister A., Phun G. S., and Mueller L. J., “Nonuniform Sampling in NMR Spectroscopy and the Preservation of Spectral Knowledge in the Time and Frequency Domains,” Journal of Physical Chemistry. A 124 (2020): 5474–5486. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Delaglio F., Walker G. S., Farley K. A., et al., “Nonuniform Sampling for All: More NMR Spectral Quality, Less Measurement Time,” American Pharmaceutical Review 20, no. 4 (2017): 339681. [PMC free article] [PubMed] [Google Scholar]
  • 6. Orekhov V. Y. and Jaravine V. A., “Analysis of Nonuniformly Sampled Spectra With Multidimensional Decomposition,” Progress in Nuclear Magnetic Resonance Spectroscopy 59 (2011): 271–292. [DOI] [PubMed] [Google Scholar]
  • 7. Giraudeau P. and Frydman L., “Ultrafast 2D NMR: An Emerging Tool in Analytical Spectroscopy,” Annual Review of Analytical Chemistry 7 (2014): 129–161. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Tal A. and Frydman L., “Single‐Scan Multidimensional Magnetic Resonance,” Progress in Nuclear Magnetic Resonance Spectroscopy 57 (2010): 241–292. [DOI] [PubMed] [Google Scholar]
  • 9. Frydman L., Scherf T., and Lupulescu A., “The Acquisition of Multidimensional NMR Spectra Within a Single Scan,” Proceedings. National Academy of Sciences. United States of America 99 (2002): 15858–15862. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Kupče E., Yong J. R. J., Widmalm G., and Claridge T. D. W., “Parallel NMR Supersequences: Ten Spectra in a Single Measurement,” JACS Au 1 (2021): 1892–1897. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Kupče E., Mote K. R., Webb A., Madhu P. K., and Claridge T. D. W., “Multiplexing Experiments in NMR and Multinuclear MRI,” Progress in Nuclear Magnetic Resonance Spectroscopy 124–125 (2021): 1–56. [DOI] [PubMed] [Google Scholar]
  • 12. Kupče E. and Claridge T. D. W., “NOAH: NMR Supersequences for Small Molecule Analysis and Structure Elucidation,” Angewandte Chemie 129 (2017): 11941–11945. [DOI] [PubMed] [Google Scholar]
  • 13. Mayzel M., Rosenlöw J., Isaksson L., and Orekhov V. Y., “Time‐Resolved Multidimensional NMR With Nonuniform Sampling,” Journal of Biomolecular NMR 58 (2014): 129–139. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Dass R., Kozminski W., and Kazimierczuk K., “Analysis of Complex Reacting Mixtures by Time‐Resolved 2D NMR,” Analytical Chemistry 87 (2015): 1337–1343. [DOI] [PubMed] [Google Scholar]
  • 15. Gołowicz D., Kasprzak P., Orekhov V. Y., and Kazimierczuk K., “Fast Time‐Resolved NMR With Nonuniform Sampling,” Progress in Nuclear Magnetic Resonance Spectroscopy 116 (2020): 40–55. [DOI] [PubMed] [Google Scholar]
  • 16. Gołowicz D., Kaźmierczak M., and Kazimierczuk K., “Benefits of Time‐Resolved Nonuniform Sampling in Reaction Monitoring,” Magnetic Resonance in Chemistry 59 (2021): 213–220. [DOI] [PubMed] [Google Scholar]
  • 17. Nawrocka E. K., Kasprzak P., Zawada K., et al., “Nonstationary Two‐Dimensional Nuclear Magnetic Resonance: A Method for Studying Reaction Mechanisms In Situ,” Analytical Chemistry 91 (2019): 11306–11315. [DOI] [PubMed] [Google Scholar]
  • 18. Karamanos T. K., Pashley C. L., Kalverda A. P., et al., “A Population Shift Between Sparsely Populated Folding Intermediates Determines Amyloidogenicity,” Journal of the American Chemical Society 138 (2016): 6271–6280. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Dass R., Grudziaz K., Ishikawa T., Nowakowski M., Dębowska R., and Kazimierczuk K., “Fast 2D NMR Spectroscopy for In Vivo Monitoring of Bacterial Metabolism in Complex Mixtures,” Frontiers in Microbiology 8 (2017): 1306. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Louša P., Nedozrálová H., Župa E., Nováček J., and Hritz J., “Phosphorylation of the Regulatory Domain of Human Tyrosine Hydroxylase 1 Monitored Using Nonuniformly Sampled NMR,” Biophysical Chemistry 223 (2017): 25–29. [DOI] [PubMed] [Google Scholar]
  • 21. Lane D., Bermel W., Ning P., et al., “Time‐Resolved In Vivo 2D NMR During a Concentration Ramp,” Analytical Chemistry 92 (2020): 9856–9865. [DOI] [PubMed] [Google Scholar]
  • 22. Kristinaityte K., Mames A., Pietrzak M., et al., “Deeper Insight Into Photopolymerization by Time‐Resolved Nonuniform Sampling and Diffusion NMR,” Journal of the American Chemical Society 144 (2022): 13938–13945. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Wolff W. W., Bermel W., Steiner K., et al., “Signal and Time Resolved Information Experiment (SATIRE): An NMR Supersequence for Monitoring Complex Environmental and Biological Processes Without 13C Enrichment,” Analytical Chemistry 98 (2026): 300–308. [DOI] [PubMed] [Google Scholar]
  • 24. Claridge T. D. W., Mayzel M., and Kupče E., “Triplet NOAH Supersequences Optimized for Small‐Molecule Structure Characterization,” Magnetic Resonance in Chemistry 57 (2019): 946–952. [DOI] [PubMed] [Google Scholar]
  • 25. Birman V. B. and Li X., “Benzotetramisole: A Remarkably Enantioselective Acyl Transfer Catalyst,” Organic Letters 8 (2006): 1351–1354. [DOI] [PubMed] [Google Scholar]
  • 26. Zhang J. and Hein J. E., “Improving the Univariate Calibration Approach With Bayesian Modeling for IR Reaction Monitoring,” Reaction Chemistry & Engineering (2026). [Google Scholar]
  • 27. Kazimierczuk K. and Orekhov V. Y., “Accelerated NMR Spectroscopy by Using Compressed Sensing,” Angewandte Chemie, International Edition 50 (2011): 5556–5559. [DOI] [PubMed] [Google Scholar]
  • 28. Wu Y., D'Agostino C., Holland D. J., and Gladden L. F., “In Situ Study of Reaction Kinetics Using Compressed Sensing NMR,” Chemical Communications 50 (2014): 14137–14140. [DOI] [PubMed] [Google Scholar]
  • 29. Heikkinen S., Toikka M. M., Karhunen P. T., and Kilpelä I. A., “Quantitative 2D HSQC via Suppression of J‐Dependence of Polarization Transfer,” Journal of the American Chemical Society 125 (2003): 4362–4367. [DOI] [PubMed] [Google Scholar]
  • 30. Flook A. and Lloyd‐Jones G. C., “Simple Parameters and Data Processing for Better Signal‐to‐Noise and Temporal Resolution in In Situ 1D NMR Reaction Monitoring,” Journal of Organic Chemistry 89 (2024): 16586–16593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Wagner A. J. and Rychnovsky S. D., “Kinetic Analysis of the HBTM‐Catalyzed Esterification of an Enantiopure Secondary Alcohol,” Organic Letters 15 (2013): 5504–5507. [DOI] [PubMed] [Google Scholar]
  • 32. Dale H. J. A., Hodges G. R., and Lloyd‐Jones G. C., “Kinetics and Mechanism of Azole N–π*‐Catalyzed Amine Acylation,” Journal of the American Chemical Society 145 (2023): 18126–18140. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Li X., Jiang H., Uffman E. W., et al., “Kinetic Resolution of Secondary Alcohols Using Amidine‐Based Catalysts,” Journal of Organic Chemistry 77 (2012): 1722–1737. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Hall C. M. and Wemple J. A., “Carbon‐13 Nuclear Magnetic Resonance Study of Thiol Esters,” Journal of Organic Chemistry 42 (1977): 2118–2123. [Google Scholar]
  • 35. Zhao F., Shu C., Young C. M., Carpenter‐Warren C., Slawin A. M. Z., and Smith A. D., “Enantioselective Synthesis of α‐Aryl‐β2‐Amino Esters by Cooperative Isothiourea and Brønsted Acid Catalysis,” Angewandte Chemie, International Edition 60 (2021): 11892–11900. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Mäkelä V., Helminen J., Kilpeläinen I., and Heikkinen S., “Quantitative Equal‐Carbon‐Response HSQC Experiment (QEC‐HSQC),” Journal of Magnetic Resonance 271 (2016): 34–39. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Figure S1: Flowchart for the set‐up and processing of time‐resolved NOAH experiments using mddNMR. Refer to mddNMR manual for more details about the software. (a) For discussion of considerations underlying parameter choice see section 2.3 of manuscript. (b) For discussion of reaction‐monitoring “dead time,” see Section 2.4 of manuscript. (c) For t 0 filling, run an experiment with at least the same number of TD1 points as the maximum window size used during post processing. This guarantees at least one “true t 0” point.

Figure S2. (A) Proposed simplified mechanistic cycle, including the observed intermediate Species 5 and the off‐cycle Species 6. The persistence of 5 is in line with previous kinetic studies on this system.4 (B) 1H–13C HSQC experiment resulting from the overall reaction‐monitoring time. Regions used to get integrals of each compound of interest are highlighted. (C) Zoomed‐in region of 1H–13C HSQC showing integral used for 5. (D) Zoomed‐in region of 1H–13C HSQC showing integral used for 6. NOAH 2‐ SpB: d1 = 1.5 s, NS = 4, full TD1 per module without NUS = 1600. TD1 per module with NUS = 800, NI = 400, NImax = 200.

Figure S3: 1H NMR of the reaction with zoomed‐in regions. In purple starting materials and product (1, 2, 3). In blue, a zoomed‐in spectra of the catalytic species of interest (4, 5, 6) and other low concentration side products. (a) Zoomed‐in region from 6 to 8.5 ppm. (b) Zoomed‐in region from 3.25 to 6 ppm. (c) Zoomed‐in region from 0.75 to 3.25 ppm.

Figure S4: Zoomed in 1H NMR spectra of resonances of interest showing J‐coupling trees of the 13C satellites. (a) Starting material (1), (b) BTM catalyst (4). (c) deactivated catalyst (6). The J‐coupling chosen for the NOAH experiment (1JHC = 145 Hz) was the average between the 1‐bond H‐C couplings observed in (a) (143.5 Hz) and (b) (146.6 Hz). Spectra for 1 and 4 are from purchased chemicals, spectra for 6 was obtained at the end of the reaction described in S5.2.

Figure S5: Overlay of the reaction trends obtained from experiments run using the same numbers of acquired hypercomplex points (NI), but different full indirect dimension sizes (NIMAX, referred to as t1max in the main text). The full circles are from NIMAX 200, and the empty circles are from NIMAX 400. NOAH 2‐ SpB: d1 = 1.5 s, NS = 4, full TD1 per module without NUS = 1600. TD1 per module with NUS = 800, NI = 400, NImax = varies.

Figure S6. Overlay of the reaction trends obtained from the same experiment but processed using different window sizes. The circles are from standard window size of 30, Step 1, Overlap 29. The squares are from window size of 60, Step 1, Overlap 59. The X’s are from window size of 120, Step 1, Overlap 119. (a) Reaction trends of starting Material 1 and Product 3. (b) Reaction trends of BTM Catalyst 4. Experiment ran was a NOAH 2‐BSp: d1 = 1.5 s, NS = 16, full TD1 per module without NUS = 1600. TD1 per module with NUS = 800, NI = 400, NImax = 200.

Figure S7: Overlay of the reaction trends obtained from the same experiment but processed using different window sizes. The full circles are from window size of 15, Step 1, Overlap 14. The empty circles are from standard window size of 30, Step 1, Overlap 29. Experiment ran was a NOAH 2‐BSp: d1 = 1.5 s, NS = 16, full TD1 per module without NUS = 1600. TD1 per module with NUS = 800, NI = 400, NImax = 200 with t0 appending postacquisition processing.

Figure S8. Overlay of the reaction trends obtained from experiments run using the same numbers of acquired hypercomplex points (NI = 400), same indirect dimension sizes (NIMAX = 200), same relaxation delay (d1 = 1.5 s) but different order in NOAH modules and different number of scans (NS). The full circles are from NOAH 2‐ SpB (4 scans for each transient), the empty circles are from NOAH 2‐ BSp (16 scans for each transient) with t 0 appending postacquisition processing. Trends for both reactions were obtained from integration of the time‐resolved HSQC spectra. (A) Time‐course trends for the starting Material 1, Product 3, and Catalyst 4. (B) Isolated time‐course trends for the Catalyst 4. (C) Isolated time‐course trends for the catalyst intermediate 5. (D) Isolated time‐course trends for the catalyst degradation Product 6.

Figure S9: Overlay of the reaction trends obtained from monitoring the standard reaction using 1D and 2D NMR experiments. The 2D experiment used was a two‐module NOAH BSp sequence with 16 scans for each transient, with t0 appending postacquisition processing. Trends were obtained from integration of the time‐resolved HSQC spectra. Quantitative 1D 1H NMR experiments were run with 1 scan, 60‐s delay, and 90° pulse angle. (A) Overlay with 1H NMR experiments as acquired. (B) Overlay with 1H NMR experiments resulting from merging multiple t0 experiments recorded prior to reaction initiation with those recorded after the reaction had begun using postacquisition signal averaging of continuous 4‐scan windows.5

Figure S10: Overlay of the reaction trends obtained from monitoring the standard reaction using 1D 1H NMR experiments before (filled circles) and after (empty circles) merging multiple t0 experiments recorded prior to reaction initiation with those recorded after the reaction had begun using postacquisition signal averaging of continuous 4‐scan windows.5

Figure S11a: 1H–13C HMBC resulting from the whole acquisition throughout the reaction, with highlighted correlations for the compounds of interest. (b) Zoomed‐in region of 1H–13C HMBC showing correlations for 4 and 6. The structures show only the main correlations characteristic of the compounds of interest. NOAH 2‐ BSp: d1 = 1.5 s, NS = 16, full TD1 per module without NUS = 1600. TD1 per module with NUS = 800, NI = 400, NImax = 200.

Figure S12. Time‐course trends for each of the three catalyst‐linked species, obtained from integrating the HSQC spectra obtained from the merged NOAH experiments. Showing the time of the addition of water and highlighting the reaction times used to generate 2D spectra used for the elucidation of each species of interest. (A) Time region 15–38 min used for Species 5 elucidation. (B) Time region 50–85 min used for Species 6 elucidation.

Figure S13: (A) 1H–13C HSQC of isolated system obtained from the reaction time where 5 is at its peak conc (Figure 8A). (B) Zoomed‐in region of 1H–13C HSQC showing 5–1 cross peaks. (C) Zoomed‐in region of 1H–13C HSQC showing 5–2 cross peaks. (D) Zoomed‐in region of 1H–13C HSQC showing 5–6 cross peaks. (E) Zoomed‐in region of 1H–13C HSQC showing cross peaks in the aromatic region. Anhydride 2 correlations are boxed in gray.

Figure S14: (A) 1H–13C HMBC of isolated system obtained from the reaction time where 5 is at its peak conc (Figure 8A). (B) Zoomed‐in region of 1H–13C HMBC showing 5–6 and 5–2 correlations. (C) Zoomed‐in region of 1H–13C HMBC showing correlations from 5–6 to 5–7 and 5–4. (D) Zoomed‐in region of 1H–13C HMBC showing correlations in the aromatic region. Anhydride 2 correlations are boxed in gray.

Figure S15. (A) 1H COSY of isolated system obtained from the reaction time where 5 is at its peak conc (Figure 8A). (B) Zoomed‐in region of 1H COSY showing 5–6 correlations to 5–5 and 5–6. (C) Zoomed‐in region of 1H COSY showing 5–12 correlations to 5–11 and 5–13, as well as 5–13 correlation to 5–14. Anhydride 2 correlations are boxed in gray.

Figure S16: (A) 1H–13C HSQC of isolated system obtained from the reaction time where 6 is at its peak conc (Figure 8B). (B) Zoomed‐in region of 1H–13C HSQC showing 6–2 and 6–18 cross peaks. (C) Zoomed‐in region of 1H–13C HSQC showing 6–5 and 6–6 cross peaks. (D) Zoomed‐in region of 1H–13C HSQC showing cross peaks in the aromatic region. Anhydride 2 correlations are boxed in gray.

Figure S17: (A) 1H–13C HMBC of isolated system obtained from the reaction time where 6 is at its peak conc 7 (Figure 8B). (B) Zoomed‐in region of 1H–13C HMBC showing 6–19 correlations to 6–18, as well as 6–2 correlations to 6–1. (C) Zoomed‐in region of 1H–13C HMBC showing correlations from 6–6 to 6–4, 6–5, and 6–7, as well as 6–5 correlations to 6–8. (D) Zoomed‐in region of 1H–13C HMBC showing correlations from 6–3 to 6–1 and 6–2, as well as 6–17 correlations to 6–18 and 6–19. (E) Zoomed‐in region of 1H–13C HMBC showing correlations in the aromatic region. Anhydride 2 correlations are boxed in gray.

Figure S18: (A) 1H COSY of isolated system obtained from the reaction time where 6 is at its peak conc (Figure 8B). (B) Zoomed‐in region of 1H COSY showing 6–2 correlations to 6–1, as well as 6–18 correlation to 6–19. Anhydride 2 correlations are boxed in gray.

MRC-64-621-s001.docx (10MB, docx)

Data Availability Statement

The data that support the findings of this study are openly available in GITlab code base at https://gitlab.com/sguzmanc/2D_fid_append.

Articles from Magnetic Resonance in Chemistry are provided here courtesy of Wiley

RESOURCES