We report the results of an attempt to reproduce a reported cavity catalysis of the ester hydrolysis of para-nitrophenyl acetate due to vibrational strong coupling. While we achieved the same light–matter coupling strength and detuning, we did not observe the reported ten-fold increase in the reaction rate constant. Furthermore, no obvious detuning dependence was observed. The inconsistency with the reported literature suggests that cavity catalysis is sensitive to experimental details beyond the onset of vibrational strong coupling. This indicates that other important factors are involved and have been overlooked so far. We find that more investigation into the limits, key factors, and mechanisms to reliably actualize cavity modified reactions is needed.
Reaction modification by vibrational strong coupling (VSC) is an intriguing idea with the potential to reshape the field of chemical synthesis. By placing a molecular system inside a confined optical environment, such as those present within a Fabry–Pérot cavity, VSC between the molecular vibrational and cavity modes is achieved. The resulting strong coupling between photons and molecular vibrations is called molecular vibrational polaritons.1–8 Concomitantly, when strongly coupled molecules participate in chemical reactions, the corresponding reaction rate constants have been shown to be significantly modified owing to VSC.9–18
One fascinating example is the cavity catalysis enabled by cooperative VSC. Lather et al.9 observed modification of the hydrolysis of para-nitrophenyl acetate (PNPA) by tetrabutylammonium fluoride (TBAF) in the ethyl acetate (EtOAc) solvent [Fig. 1(a)]. It was reported that when the cavity mode was tuned to the vibrational frequency of EtOAc at 1740 cm−1, the CO modes of EtOAc and PNPA together strongly coupled to the cavity mode, and the reaction accelerated by an order of magnitude. The preliminary mechanism that the reaction proceeds by is a two-step nucleophilic substitution (BAc2) base-catalyzed acyl hydrolysis. F− ions perform a nucleophilic attack on the carbonyl of the PNPA reactant. The intermediate state then dissociates into acetyl fluoride and para-nitrophenoxide (PNP−). Due to the basic condition provided by the TBAF, the PNP− remains as the solvated counter ion to the tetrabutylammonium [Fig. 1(b)]. In this mechanism, the rate limiting step is the nucleophilic attack. The ability of F− ions to access the carbonyl group of the PNPA is affected by the coordination with neighboring solvent molecules. It was postulated that, by strongly coupling with CO modes of the solvent EtOAc molecules and reactant PNPA, cooperatively, CO modes in both solvents and reactants reached VSC. This cooperative VSC modulated the reaction rate by a factor of 10, possibly through external vibrational energy transfer. This was remarkable considering the reactants by itself could not reach VSC due to its low concentration. A notable finding was that the modification of the reaction rate was highly sensitive to the detuning of the cavity and molecular mode frequencies. A similar work reported enhancement in the ester hydrolysis of PNPA by α-chymotrypsin (α-CT) when strongly coupled to O–H stretching modes of solvent water and the enzyme itself.12 The active site of the α-CT is a serine-histadine-aspartate catalytic triad that acts as a charge-relay system. A seven times enhancement in catalytic efficiency was observed when the hydrolysis was performed under VSC. Similarly, the cooperative VSC between the OH of enzyme and of water was proposed to affect this reaction.
These innovative findings make clear that mechanistic understandings are necessary for researchers to rationally design cavities to manipulate reactions. While many experimental and theoretical works have been attempted, the pursuit of a consistent picture continues.19–34 Intrigued by the results, we aimed to study ultrafast dynamics35–41 of the VSC-modified hydrolysis reaction in Ref. 9. As a first step, we attempted to replicate the findings of Ref. 9. As shown below, by carefully repeating the experimental methods and controlling pertinent parameters, we were able to reproduce VSC between cavity, and CO modes of EtOAC and PNPA, and to accurately determine and choose the detuning conditions, as shown in Fig. 1(c). When the cavity modes were tuned to overlap with the CO absorption of EtOAC (blue dashed line), the so-called zero-detuning, the cavity modes split into the upper and lower polaritons (UP and LP in the black curve), with nearly 150 cm−1 coupling strength. However, at the zero-detuning conditions, we did not observe the reported ten times reaction rate constant acceleration.9 Indeed, there was no clear indication of reaction enhancement observed. The contrast between our attempt and the literature result may be due to a mismatch of some other crucial factors that were overlooked and could imply a particular sensitivity of VSC-modified reactions to experimental conditions.
To initiate and measure the reaction, we encapsulated a binary solution of EtOAc-MeOH with 9:1 ratio. We then combined 90 µl of PNPA (0.1M) solution in EtOAc and 10 µl TBAF (0.1M) solution in MeOH and immediately injected the mixture into the cavity or regular sample cells composed by two transparent optical windows. We monitored the reaction by the UV–Vis absorption band near 400 nm—PNP− in its unprotonated state. The full spectra were taken at 1.5 s intervals from 0 to 600 and 60 s intervals from 600 s to reaction completion (∼3600 s).
The reaction was first performed in the cell using transparent CaF2 windows. The non-cavity reactions showed a peak growth near 400 nm [Fig. 2(a)]. By fitting the feature integrated between 375 and 425 nm [Fig. 2(b)], we found that the reaction rate constant outside the cavity was 1.5 × 10−3 ∓ 2.9 × 10−4 s−1.42 This rate agreed reasonably well with the out of cavity rate constant found in Ref. 9 of 2.0 × 10−3 s−1. The reactions were run five times and showed high reproducibility (see Sec. S6 of the supplementary material).
After ensuring our ability to use this microfluidic cell method to quantitatively measure the reaction rate, we turned to the VSC-modified reaction. Following Ref. 9, to create the VSC condition, we used a Fabry–Pérot cavity which consisted of two gold-coated CaF2 mirrors (97.5% reflectivity at 1740 cm−1) separated by an 18 µm Teflon spacer (Fig. 3). The mirrors were prepared by sputtering 10 nm Au on one face and 100 nm SiO2 on top to insulate chemical samples and protect the Au layer. The cavity was contained in a stainless-steel sample cell (Specac, Inc.) with two channels for passing chemical samples. The FSR was determined by peak separation of higher order cavity modes that are completely uncoupled with the vibrational modes. This was necessary because the neighboring cavity modes to the polariton peaks were also weakly coupled to the vibrational modes, rendering them being pushed outwards relative to their uncoupled positions. We noted that the free spectral range (FSR) varied from 154 cm−1 to 182 cm−1 among different experimental trials, because it was very sensitive to the cavity thickness. Still, our FSR remained close to the 175 cm−1 FSR reported in Ref. 9. The cavity had a full width half max of 16 cm−1 (vs ∼20 cm−1 in Ref. 9), and a Q-factor of 125 when filled with pure methanol (vs ∼96 in Ref. 9). To precisely control the coupling strength and detuning, we gradually injected solutions of increasing EtOAc concentration until the targeted mixture ratio was achieved. Once the EtOAc–MeOH concentrations reached a 9:1 ratio (the reaction condition), we let the spacer equilibrate for 30 min and then recorded the final VSC condition by FTIR (Nicolet iS50).
Due to the Rabi splitting at reaction concentration of EtOAc (∼145 cm−1) being nearly that of the free spectral range (from 154 cm−1 to 182 cm−1) of the cavity, it was challenging to distinguish the exact positions of the UP and LP peaks, and thus the precise detuning conditions. To characterize the detuning and coupling strength of the polaritons, we measured their dispersion curve by tuning the cavity thickness and, therefore, the cavity resonance frequency. The experimental results [Fig. 4(a)] showed two curves in the middle of plot (i.e., from 1650 to 1800 cm−1 at zero detuning) representing the polaritons, whereas the rest of the curves were the neighboring cavity modes far detuned from and weakly coupled to the CO vibrational transitions (ωmol). The dispersion curve was then simulated by a 2-by-2 Hamiltonian to extract the coupling strength and cavity frequencies (ωcav), which further determined the detuning using δ = ωcav − ωmol. The simulated dispersion curve overlapped [dashed lines in Fig. 4(a)] with the measured one well. We further simulated the full dispersion spectra using the transfer matrix model [TMM, Fig. 4(b); simulation code in Sec. S7 of the supplementary material], which reproduced the experimental dispersion.43 The agreement between experimental data and both theoretical models suggested that the extracted parameters, such as detuning and coupling strength, accurately reflected the experimental conditions.
We thereby used the dispersion curve as a reference to prepared polaritons at specific detuning and ran the reactions. For all reactions, the coupling strength was kept at ∼145 cm−1, similar to Ref. 9. We injected the reaction mixture into the cavity and monitored the reaction using UV–Vis spectroscopy in a similar manner as it was carried out as outside the cavity (see Sec. S5 of the supplementary material). We calibrated the effective pathlength of the cavity44 and transparent cell using ferrocene and found out that the effective pathlength difference in the UV region was negligible between the cavity and transparent cell (see Sec. S3 and Fig S3 of the supplementary material). Nevertheless, it was important to account for any effect of the cavity on the pathlength.
The reaction in the cavity showed a similar growth of the 400 nm peak. Fringe patterns were observed, as the cavity still had a moderately high Q in the UV regime. At zero detuning (ωcav = ωvib = 1740 cm-1), we observed the reaction rate to be 1.5 × 10−3 ∓ 1.0 × 10−4 s−1 (determined by three measurements, vs 2.5 × 10−2 s−1 in Ref. 9). Thus, there was no noticeable enhancement at zero-detuning in our experiment, relative to the rate outside of cavity. We noted that in Ref. 9, the CO mode was reported to be 1750 cm−1. Thus, we also tuned the cavity to 1750 cm−1 and still found the rate to be 1.75 × 10−3 s−1 (green hex in Fig. 5).
We then further conducted reactions at detuning ranging from Δ = ∓∼80 cm−1. All detailed reaction results are included in Secs. S5 and S6 of the supplementary material . Based on single reactions (blue dots), we observed the reaction rate fluctuating from a minimum of 1.3 × 10−3 s−1 to a maximum of 2.5 × 10−3 s−1 without a clear trend. We further binned data with close detuning and presented the average rate and the uncertainty for several detuning ranges. Still, there was no clear reaction rate enhancement at the level reported by Ref. 9. It was plausible that there was a 60% enhancement at +50 cm−1 detuning, but there was also a large uncertainty due to the sparsity of data at that region of detuning. The small volume of reaction media could make precise control of reactant concentration challenging, leading to the reaction rate constant to have a standard deviation of 20%. However, should the ten-times rate enhancement and strong cavity detuning dependence exist in our experiment, they should be distinctly observable.
To summarize, a negligible reaction rate enhancement was observed when the C=O vibrational modes of the reactants and solvents were strongly coupled to the cavity mode; nor the resonant dependence reported in Ref. 9 were reproduced. To the best of our knowledge, we have carefully kept all critical parameters of this reaction as close to the literature values. Thus, it remains a mystery to us what factors caused the experimental discrepancy.
A major difference between our work and Lather et al.9 was the carbonyl absorption frequencies of EtOAc. In Ref. 9, the vibrational frequency of EtOAc was 1750 cm−1, measured by FTIR 10% (v/v) in hexanes. In our current results, the frequency was observed at 1742 cm−1 for EtOAc:MeOH using transmission FTIR (Fig.S9), which was the vibrational frequency of the solvent systems that VSC occurred for the reactions. Indeed, the carbonyl mode of EtOAc may vary by up to 38 cm−1, depending on the solvent,45 which was also verified here [see Fig.S8]. Furthermore, EtOAc CO modes actually have three distinct frequencies when mixed with MeOH due to different H-bond environments.46 Thus, it was important to quantify the frequency of the solvent as the same as the reaction condition. We observed the carbonyl mode of PNPA at 1758 cm−1 by ATR-FTIR, which agreed with Ref. 9, as the PNPA peak in their spectrum also blue shifted from 1750 cm−1.
To investigate how critical the relative peak positions of PNPA and EtOAc were to the cooperative coupling effect, we calculated the composition of polaritons (namely, Hopfield coefficients) using the peak positions observed in the current experiment (ωPNPA = 1758 cm−1, ωEtOAc = 1742 cm−1) and the idealized peak positions (ωPNPA = ωEtOAc = 1750 cm−1).9 We constructed a three-oscillator model to observe any impact on the Hopfield coefficients due to the small relative peak difference.47–49 We found that in either case, the contribution of PNPA to the UP/LP states was very low, and MP was composed by a localized PNPA CO modes with negligible compositions from EtOAc and cavity modes (Sec. S8 of the supplementary material). Thus, we concluded the CO modes of PNPA remained mostly a localized vibrational mode with only a tiny fraction involved in the delocalized polariton wavefunctions, and the peak position shift did not introduce any practical difference in the polariton compositions.
Further possible reasons for the discrepancy could be the following: Differences in cavity cell sealing resulting in loss of solvent could artificially change the concentration of the solutes; another noted difference from the reported data was the cavity quality factor, which may play a role in reaction modification. We note, in Ref. 9, the cavity volume was 3 µl, and 300 µl of reaction volume was injected into the cavity for each reaction. The cavity volume given agrees with our own measurements of the cavity. We used 100 µl reaction volume to minimize the effect of the uncoupled reaction mixture diffusing into the cavity and reducing the rate of reaction. Despite this, there was no effect on the reaction rate from reducing the total reaction volume injected into the cavity. Additionally, the method of data processing may result in error in the rate. Obtaining kinetic traces by integration of the PNP− absorption peak across all collected spectra was compared to the use of following absorption at a single frequency (Secs. S4–S6 of the supplementary material). The low total absorption of the system, the presence of fringes in UV–vis spectra inherent to the cavity, and variations in cavity parallelism and mounting produced much greater error in the calculated rate when following a single frequency trace (Fig. S4). Time resolution also appears to be crucial in following cavity effects in the early time reaction kinetics. In an initial attempt, an instrument with slow scanning speed resulted in a low number of data points within the first 600 s of the reactions. Additionally, it was unable to scan the equilibrium absorbance at a preset time. These issues caused a systematic error that made all measured rate constants larger. Even in that case, no resonant enhancement was observed.
Our results, which measure reaction rate under the influence of collective vibrational strong coupling (i.e., many molecules strongly couple to the cavity together), agree with theoretical works that simulated the reaction in the collective coupling regime.20,22,34,50 However, our findings differ from the predictions from theoretical works where VSC under a single molecule regime is simulated.23–25,51 These simulated conditions differ considerably from the collective strong coupling condition in our experiments. The resulting differences indicate that VSC-modified reactions are very delicate and sensitive to details. For example, other key factors, instead of VSC coupling strength, could play a significant role but so far were overlooked. Considering the potentially disruptive influence on the field of chemistry, this and other conflicting results29,52 demand more investigations on VSC-modified reactions in both theory and experiments to reach a consensus on to what extent VSC can modify reactions. Other frontier phenomena with a substantial ability to change the landscape of chemical research and beyond have been flashpoints for debate into the underlying mechanisms at play.53,54 Polariton chemistry is another quintessential example, which will undoubtedly benefit from continuing development and investigations.
The supplementary material includes experimental details, additional experimental data, and theoretical models used in this research.
W.X. acknowledges the support from the Alfred P. Sloan Research Fellowship (Grant No. FG-2020-12845). G.D.W. was supported by the NSF Grant No. DMR-1848215. The authors thank Zimo Yang, Bo Xiang, Joseph Palasz, Matthew Du, and Jorge Campos-Gonzalez-Angulo for helpful advice and discussions. We acknowledge Professor Nathen Romero for his generous support of instrumentation. We also acknowledge the Kubiak Group for technical support and generosity in instrument use.
Conflict of Interest
The authors have no conflict to disclose.
The data that support the findings of this study are available from the corresponding author upon reasonable request.