Structural analysis of transient reaction intermediate in formic acid dehydrogenation catalysis using two-dimensional IR spectroscopy
- aCollege of Chemistry and Molecular Engineering, Beijing National Laboratory for Molecular Sciences, Peking University, 100871 Beijing, China;
- bDepartment of Chemistry, Rice University, Houston, TX 77005;
- cKing Abdullah University of Science and Technology Catalysis Center, King Abdullah University of Science and Technology, 23955-6900 Thuwal, Saudi Arabia;
- dDivision of Physical Science and Engineering, King Abdullah University of Science and Technology , 23955-6900 Thuwal, Saudi Arabia
See allHide authors and affiliations
Edited by Michael D. Fayer, Stanford University, Stanford, CA, and approved October 29, 2018 (received for review May 31, 2018)

Significance
We demonstrate that the transient intermediate structure, which is too fast for traditional techniques, e.g. NMR or X-ray diffraction, to resolve in a thermally driven reaction can be determined in situ with the vibrational cross-angle method. The principle of the method illustrated in the work is general for other organic reactions and the method is expected to bring profound impacts for the studies of chemical reactions. The reaction mechanism of a well-defined catalytic system with performance to generate H2 from formic acid is elucidated.
Abstract
The molecular structure of a catalytically active key intermediate is determined in solution by employing 2D IR spectroscopy measuring vibrational cross-angles. The formate intermediate (2) in the formic acid dehydrogenation reaction catalyzed by a phosphorus–nitrogen PN3P–Ru catalyst is elucidated. Our spectroscopic studies show that the complex features a formate ion directly attached to the Ru center as a ligand, and a proton added to the imine arm of the dearomatized PN3P* ligand. During the catalytic process, the imine arms are not only reversibly protonated and deprotonated, but also interacting with the protic substrate molecules, effectively serving as the local proton buffer to offer remarkable stability with a turnover number (TON) over one million.
Identifying molecular structures of reaction intermediates is often the key to elucidate reaction mechanisms, which are indispensable for rational designs and optimization of catalytic systems in broad areas, including medicinal chemistry, industrial processes, and material science (1, 2). One of the most ideal approaches to study reaction intermediates is to first capture and then isolate them. In practice, however, reaction intermediates are often metastable and short-lived, making them very difficult or even impossible to isolate. It is challenging to resolve the molecular structures of these transient species in situ with traditional structural tools like X-ray crystallography and NMR techniques, due to their inherently low temporal resolutions. Two-dimensional IR (3⇓⇓–6), an ultrafast analog of 2D NMR, holds great promise to resolve structures of these transient species. It has been demonstrated that ultrafast dynamic and structural information can be obtained from 2D IR measurements for equilibrium systems and phototriggered reactions (7⇓⇓⇓⇓⇓⇓⇓–15). However, the majority of chemical reactions on which most organic chemists work on a daily basis are thermally driven reactions. Their transient intermediates are yet to be explored by ultrafast laser spectroscopy. In this work, we apply a specially designed multiple-mode 2D IR technique that provides sufficiently high excitation power for resolving very weak vibrational cross-angle signals in reacting systems, to solve in situ the fast fluctuating conformation of a transient intermediate in a thermally driven reaction. The method relies on the fact that the ultrafast (femtosecond) resolution of the technique allows the determination of individual rapid-switching conformation rather than the average structure offered by a technique with temporal resolution slower than the switching dynamics.
The selective hydrogen generation reaction from the decomposition of formic acid catalyzed by a PN3P*–ruthenium pincer complex is chosen as the model reaction system (Fig. 1A) (16). Although hydrogen gas is considered to be an ideal fuel (17⇓⇓⇓⇓–22), using high-pressure gas tanks, the currently most practical method to store H2, raises serious safety concerns in addition to low volumetric capacity. Formic acid (FA), a nontoxic and nonflammable liquid that could decompose reversibly into H2 and CO2 (CO2 + H2 = HCOOH), has been proposed as a hydrogen energy carrier (23⇓⇓–26). The forward and backward reactions together form a cycle of hydrogen storage and generation: In H2 generation plants H2 can be converted into FA for storage and transportation, and in places where H2 is needed, FA can be decomposed into H2 (27, 28). Notably, FA has a relatively high volumetric hydrogen capacity: 53 g H2/L, corresponding with an energy density of 1.77 kW·h/L, making it a promising on-board storage solution for vehicles (29⇓⇓⇓–33). Unsurprisingly, many research efforts have been devoted in developing both homogeneous and heterogeneous catalyst systems for the generation of H2 from FA (24, 25, 33⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓–58). However, when considering industry feasibility, many current catalytic systems face several challenges. Firstly, FA can decompose into either H2 and CO2 or H2O and CO. The generation of CO cannot be avoided in many catalytic systems, and even a trace amount of CO generated could be fatal to the hydrogen fuel cells downstream. Secondly, the typical turnover numbers (TONs) achieved by those catalysts are not sufficiently large, often due to their sensitivity to air, water, and/or other impurities. Only a handful of examples of high TONs > 1,000,000 were reported (55, 58⇓⇓–61).
(A) The decomposition of FA catalyzed by a PN3P*–pincer catalyst (1), the reactions going through a key intermediate (2); (B) 1H NMR spectra of the catalyst 1 (Upper) and a mixture of the catalyst and FA (Lower), both in the solution of DMSO-d6.
Our PN3P*–ruthenium pincer system can selectively generate hydrogen under mild conditions with no detectible CO. The highest turnover frequency (TOF) of >38,000 h−1 and TON of more than one million are achieved. Furthermore, the catalyst is air-stable and no noticeable catalytic activity loss was observed after aging in air for 1 mo (16). These properties are highly desirable in practical applications. In an effort to gain further insights into the unprecedented stability, we utilize 2D IR spectroscopy as a powerful tool to identify the molecular structure of an important intermediate in this highly efficient and promising catalytic reaction. The structure and conformation of the reaction intermediate (2) is determined, and the stability of our catalyst is rationalized together with density-functional (DFT) calculations.
Results and Discussion
NMR and FTIR Spectroscopies Support the Formation of a Ru Formate Complex (2).
The catalytic system is first investigated with both NMR and FTIR spectroscopies. As we reported earlier (16), the comparison of the 1H NMR spectra before and after addition of FA shows the generation of a species (Fig. 1B) that we rationalize to be a ruthenium formate complex. The three chemical shifts at δ 5.58 (d), 5.95 (d), and 6.86 (t) ppm for the three hydrogen atoms on the 3, 5, 4 positions of the pyridine ring, respectively, are typical for an unsymmetrical, dearomatizated pyridine ring. After adding FA (Fig. 1B, Lower), the proton signals shift downfield to δ 6.33 and 7.27 ppm. The two peaks at δ 5.95 ppm and 5.58 ppm in Fig. 1B (Upper) merge into the one that appears at δ 6.33 ppm. The area of this peak from these protons is now twice that of the peak at δ 7.27 ppm from the third proton on the same pyridine ring. In other words, two of the three protons on the pyridine ring become equivalent, indicating a Cs symmetry. Furthermore, the integration of the peak at δ 8.18 ppm for the NH groups is now twice that of a single H, clearly indicating that the imine group is protonated, resulting in a symmetrical and aromatized pyridine ring. Such a dearomatization/rearomatization process of the central pyridine ring through deprotonation/reprotonation of one of the N-H arms of the pincer complexes has been demonstrated to be a key step in the metal–ligand cooperative catalysis for a class of PN3(P)–pincer systems (62).
The FTIR spectra shown in Fig. 2 feature five groups of bands between 1,350 and 2,250 cm−1, each identified and attributed to five vibrational modes in the reactant or the catalyst, denoted as v1 to v5, and the corresponding modes in the additive complex, denoted as v1′ to v5′. The assignments of these vibrational modes are listed in SI Appendix, Table S1. Herein, each mode will be briefly discussed, and they will be further interrogated in 2D IR spectroscopy to determine the structure and conformation of the catalyst/HCOOH complex, as elaborated in the next section.
FTIR spectra of the catalyst (red), the reactant (blue), and the mixture (black) in the frequency range from 1,200 to 2,400 cm−1. (Inset) Zoomed-in FTIR spectra for the vibrational mode of Ru–H stretch in the catalyst and the reaction mixture (v1 and v1′), respectively.
First, the small absorption band at 2,113 cm−1 in Fig. 2 (Inset) (v1) belongs to the Ru–H stretch. As expected, such a feature is absent in pure HCOOH. In the catalyst/HCOOH mixture, however, its frequency decreases to 2,063 cm−1 and became v1′. The large redshift indicates that the Ru–H bond becomes weaker, which is expected when adding another trans ligand to the Ru center.
Second, the major absorption band at 1,940 cm−1 (v2) is attributed to the carbonyl ligand that is directly attached to Ru. With addition of HCOOH, its corresponding frequency in v2′ increases for ∼15 cm−1. Such a blueshift indicates a modest strengthening of the C≡O bond and therefore a weakening of the metal–ligand interaction, which is again consistent with the notion of another ligand attaching to the Ru center.
Third, the major absorption band at 1,720 cm−1 (v3) belongs to the C=O stretch from HCOOH. As expected, no such feature can be found on the catalyst itself. Interestingly, its corresponding frequency (v3′) remains virtually the same upon the formation of the complex. Such a frequency is very different from the carbonyl stretch in a free formate, where the C=O stretch mode would be completely delocalized on two oxygens and move to a much lower frequency (e.g., ∼1,620 cm−1 in sodium formate). The observed little-to-none frequency shift is consistent with formate binding as a ligand through a single Ru–oxygen bond, so that C=O remains largely localized. The lack of the frequency change further supports our conclusion that formate ion, rather than FA, serves as the ligand. If FA binds to Ru, it should be the oxygen of the C=O bond rather than the OH bond that attaches to Ru, which would immediately change the C=O vibrational frequency.
Lastly, the absorption bands at 1,450 cm−1 (v4) and 1,615 cm−1 (v5) undoubtedly belong to the symmetric and antisymmetric vibrational modes of the pyridine ring on the pincer ligand, respectively. Both v4 and v4′, and v5 and v5′ strongly overlap with each other. Apparently, they are not significantly affected by the formation of the complex. Consistent with the NMR results, FTIR spectra suggest that a formate ion, instead of a free FA, is bound to the Ru center as a ligand in the complex.
Although the above experimental observations are in good agreement with the formation of complex 2 with two NH arms and a formate ligand attached to Ru, they are not definitive and they do not provide information on the geometry which is critical for the understanding of reaction mechanism but fast fluctuates at the time scale of a few picoseconds (manifested by the anisotropy decay in Fig. 3C). During the catalytic reaction, 2 is observed as a transient species, and thus attempts to isolate 2 are in vain. However, as 2 is ligated by a well-defined and rigid PN3P–pincer ligand, and the IR bands for the rest of three ligands, namely Ru–H, R–CO, and formate, are well distinguishable, 2D IR technique becomes suitable to elucidate the molecular structure of 2.
Cross-peak measurements for vibrational mode pairs. (A) Illustrations of the cross-angle vibrational coupling measurement method for vibrational modes carbonyl stretch (v2′) and C=O stretch (v3′). The t-butyl groups have been omitted for clarity. (B) The 2D spectrum of the cross-peak for v2′ and v3′ at a waiting time of 0.1 ps. (C) Plots of the polarization-dependent cross-peak intensities with the polarization of the excitation beam parallel (//) and perpendicular (⊥) to the polarization of the detection beam, respectively. (Lower) Enlarged figures of the early waiting time signals labeled on the upper panels.
Molecular Structure of Complex 2 Determined by 2D IR.
In principle, the relative cross-angle between any two vibrational modes can be determined by measuring the anisotropy of the vibrational coupling between the two modes, i.e., the cross-peak in 2D IR spectroscopy if the instrument sensitivity is high enough. Briefly, the cross-peak measurement is illustrated in measurement schematic Fig. 3A. For example, the carbonyl stretch mode (v2′) on the catalyst is resonantly excited by a linearly polarized narrow-band IR pulse at about 1,955 cm−1 (
More importantly, the signal amplitudes of the cross-peaks in 2D IR depend on the relative polarizations of the excitation and detection pulses and the directions of transition dipole moments of the involved vibrational modes (66, 71). For an isotropically distributed sample within the laser focus spot (diameter ∼ 150 μm), the vibrational cross-angle between the transition dipole moments of the involved vibrational modes can be directly determined using the equation (72, 73)
Vibrational cross-angles experimentally determined
It is important to note that some vibrational modes involved in the above measurements are delocalized in the complex; the direction of the transition dipole moments of the vibrational modes are closely related to, but not necessarily the same as, those of the chemical bonds. There is no direct method to identify a conformation from a set of vibrational angles. However, the reverse can be achieved; each conformation must feature a unique set of vibrational cross-angles, which can be calculated individually using ab initio method (66). We therefore define the most likely conformation as the one that minimizes the overall difference between the experimental and calculated vibrational cross-angles, according to the following equation:
For this complex particularly, we explore the conformational space by rotating the formate along the two single bonds: the Ru–O1 bond that connects the metal center and ligand, and the C2–O2 bond next to the Ru–O bond. These two degrees of freedom are essentially the tilt angle and twist angle of the formate ion relative to the pyridine ring. In total, we search 576 possible conformations by varying the dihedral angle C1(CO)–Ru–O1–C2 from −5° to 175° for every 10° and Ru–O1–C2–O2 from 1° to 351° for every 10°. The vibrational cross-angles for each conformation are calculated and then compared with the measured one according to Eq. 2. The Er values are plotted in Fig. 4A. The minimum Er value is found at the dihedral angles (Ru–O1–C2–O2, C1–Ru–O1–C2) (321°, 155°) at the northeastern corner of the map. The corresponding conformation of the catalyst/formate complex is displayed in Fig. 4B.
Determine the conformation of complex 2. (A) The difference (Er) between experimental and calculated vibrational cross-angles for 576 possible conformations of with different Ru–O1–C2–O2 and C1–Ru–O1–C2 dihedral angles. The z axis is the value of Er shown in a color map. (B) The most stable conformation (red circle in Fig. 4A) from different views. The butyl groups are omitted for clarity.
Insights from the Conformational Analysis.
The molecular structure of formate species 2 determined by the 2D IR is very intriguing, because DFT calculations indeed suggest a different structure as the global minimum (2′) among all conformers (Fig. 5). Both structures show similar PN3P–pincer framework but different orientation of the coordinated formate group. Structure 2′ was computed to be 1.2 (3.8) kcal/mol more stable than 2 [energy in the Polarizable Continuum Model (PCM) in parentheses; see SI Appendix for details], presumably due to the relief from the electron–electron repulsion between the formate ligand and the pyridine ring. These observations strongly suggest the existence of certain driving force to allow complex 2 to stay in an apparently more strained conformation.
Comparison of the global minimum (2′) and “optimized” 2D-IR (2) structures of complex 2. Key geometrical parameters (Å and °): 2: Ru–C1 1.847, Ru–O1 2.259, Ru–P1 2.330, C1–Ru–O1–C2 147.53; 2′: Ru–C1 1.858, Ru–O1 2.235, Ru–P1 2.341, C1–Ru–O1–C2 −3.50. The butyl groups are omitted for clarity.
In light of the 2D IR structure, our hypothesis is that there may be a protic molecule (H2O or HCOOH) bridging the carbonyl oxygen of the formate ligand and the proton of one of the N-H arms. To our delight, the fully optimized structure of complex 2 containing either one water molecule (2-W) or one FA (2-FA) shows a very similar geometry of the formate group to that of 2 (Fig. 6). Moreover, the formation of 2-W and 2-FA is found energetically favorable by 6.5 (8.6) (compared with 2′ and H2O, energy in the PCM in parentheses; see SI Appendix for details) and 4.8 (9.6) kcal/mol (compared with 2′ and HCOOH), respectively. These findings imply that the N-H arms of the PN3P–Ru catalyst can play a role to interact with the protic molecules in the reaction mixture in addition to mediating the dearomatization/rearomatization of the central pyridine ring for FA decomposition and hydrogen generation processes. The formate–water (FA)–NH ring structure (2-W and 2-FA) is very interesting and may be one important reason responsible for the excellent stability of the catalyst. It forms an effective polar screen on the top of the N1 and P1 atoms, which can reduce their probability of being attacked by ions or polar molecules, e.g., protonated water, and thus minimize the decompositions of the N–P and P–Ru bonds. We want to emphasize that the formate–water–NH hydrogen and formate–FA–NH hydrogen bonds proposed here are only representative examples to form such a ring structure. The water or HCOOH molecule can be any protic species in the reaction.
Computed structures of 2·H2O (2-W) and 2·HCOOH (2-FA). Key geometrical parameters (Å and °): 2-W Ru–C1 1.848, Ru–O1 2.296, Ru–P1 2.339, O2–H1 1.771, O3–H2 1.998, C1–Ru–O1–C2 152.96; 2-FA Ru–C1 1.850, Ru–O1 2.302, Ru–P1 2.338, O2–H1 1.613, O3–H2 2.302, C1–Ru–O1–C2 156.65. The butyl groups are omitted for clarity.
Concluding Remarks
In summary, we have demonstrated that the structure of a key reactive intermediate in a thermally driven hydrogen generation reaction can be resolved by 2D IR spectroscopy using the ultrafast vibrational cross-angle method. The results from the ultrafast vibrational measurements and NMR results suggest that the NH arms of the PN3P–pincer Ru catalyst indeed interact with the protic environments, namely playing a role as the local buffer. This may facilitate the reaction by stabilizing the additive complex and also significantly reduce the chance of free proton directly attacking the metal center and other coordination atoms, minimizing decomposition of the catalyst. These are likely the contributing factors for this catalytic system with high reactivity and stability, as evident in its high TOF and TON values. Very importantly, our work demonstrates that the high-powered ultrafast vibrational cross-angle measurement holds great promise to resolve the transient intermediate structures in situ, which are difficult to study with traditional techniques due to their fast dynamics, and this technique holds a great promise for mechanistic study in catalysis.
Materials and Methods
Synthesis and Sample Preparation.
RuPN3P was synthesized by adding a solution of Ru1 (47 mg, 0.05 mmol) in THF (10 mL) dropwise to a ligand tBu2PN3P (20 mg, 0.05 mmol) in THF (4 mL) and heated at 65 °C for 12 h. RuPN3P (281.6 mg, 0.5 mmol) was dissolved in THF (4 mL). A solution of KOtBu (56.0 mg, 0.5 mmol) in THF (4 mL) was added to it and the reaction mixture was stirred at room temperature for 2 h. The solvent was then removed under vacuum and washed with hexane (1 mL × 3) and dried under vacuum to obtain the catalyst. All reactions were conducted under argon protection. FA was purchased from Sigma-Aldrich and used as received. For the FTIR and vibrational cross-angle measurements, catalyst 1 and FA were mixed with a molar ratio of 1:4 and sealed in a cell with CaF2 windows. During the ultrafast spectroscopic measurements, hydrogen was constantly generated from the decomposition of FA. More experimental details are provided in SI Appendix.
NMR Experiments.
For 1H-NMR and NOESY experiments, HCOOH (1.62 mg) was added into a DMSO-d6 (0.5 mL) solution of 1 (20.2 mg) in a screw-capped NMR tube at room temperature. Reaction was observed by visual inspection.
Optical Experiments.
The optical setup used is described in detail in our previous publication (66). The oscillator provides the same seed pulse for a picosecond (ps) amplifier and a femtosecond (fs) amplifier (UpTek Solution Corp.) which are synchronized. The ps amplifier pumps an optical parametric amplifier to produce ∼0.8 ps (would vary from 0.7 ∼0.9 ps in different frequencies) mid-IR pulses with a bandwidth 10 ∼35 cm−1 in a tunable frequency range from 400 to 4,000 cm−1 with energy 1∼40 μJ/pulse (1∼10-μJ/pulse for 400–900 cm−1 and >10-μJ/pulse for higher frequencies) at 1 kHz. The bandwidth is narrower at a longer wavelength. At 2,000 cm−1, the bandwidth is about 18 cm−1. The detailed characterization of the outputs of the laser systems can be found in our previous work (73). This ps IR pulse is used as excited pulse in the cross-coupling experiments. The fs amplifier is used to generate a high-intensity mid-IR and terahertz supercontinuum pulse with a pulse duration around 110 fs in the frequency range from <20 to >3,500 cm−1 at 1 kHz, and the shot-to-shot fluctuation is less than 1% in most of the spectral region. This supercontinuum pulse is used as the detection pulse, which is frequency-resolved by a spectrograph. More details are provided in SI Appendix.
Theoretical Computations.
Computations on nonmetal elements were performed by employing the program suite Gaussian 09, using DFT with Truhlar’s hybrid metaexchange-correlation functional M06 (76) in conjunction with a 6–31G(d,p) basis set. For Ru atom, Stuttgart–Dresden effective core potentials (77) were used. To generate all of the possible structures, we preset 576 possible conformations for intermediate 2 by varying the dihedral angle C1–Ru–O1–C2 from −5° to 175° for every 10° and Ru–O1–C2–O2 from 1° to 351° for every 10°. Each structure with different pair of dihedral angles fixed is optimized under the computational method described above. The optimized structures of 2, 2′, 2-W, and 2-FA were characterized by frequency calculations. Solvent effects were estimated by performing a single point calculation at the same level of theory using the PCM with the integral equation formalism variant (78) in H2O (a continuum solvent dielectric of ε = 78.3553) and DMSO (a continuum solvent dielectric of ε = 46.826).
Acknowledgments
Y.Z. acknowledges a Schlumberger future faculty award. This work is supported by National Natural Science Foundation of China (NSFC) (NSFC-21627805, 21673004, and 21821004), NSF (CHE-1503865), and Ministry of Science and Technology of the People’s Republic of China (2017YFA0204702) (to J.Z.), and by King Abdullah University of Science and Technology (to K.-W.H.).
Footnotes
↵1Y.Z. and X.C. contributed equally to this work.
↵2Present address: Beijing National Laboratory for Condensed Matter Physics, CAS Key Laboratory of Soft Matter Physics, Institute of Physics, Chinese Academy of Sciences, 100190 Beijing, China.
- ↵3To whom correspondence may be addressed. Email: hkw{at}kaust.edu.sa or junrong{at}pku.edu.cn.
Author contributions: Y.Z., K.-W.H., and J.Z. designed research; Y.Z., X.G., H.C., and J.Z. performed research; B.Z., Y.P., H.L., S.M., C.G., and K.-W.H. contributed new reagents/analytic tools; Y.Z., X.C., H.C., K.-W.H., and J.Z. analyzed data; and Y.Z., X.C., K.-W.H., and J.Z. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1809342115/-/DCSupplemental.
Published under the PNAS license.
References
- ↵
- Milo A,
- Neel AJ,
- Toste FD,
- Sigman MS
- ↵
- Ananikov VP
- ↵
- Mukamel S
- ↵
- ↵
- ↵
- ↵
- van Wilderen LJ,
- Messmer AT,
- Bredenbeck J
- ↵
- ↵
- Donaldson PM,
- Hamm P
- ↵
- Kurochkin DV,
- Naraharisetty SRG,
- Rubtsov IV
- ↵
- ↵
- Slenkamp KM, et al.
- ↵
- ↵
- ↵
- ↵
- Pan Y, et al.
- ↵
- Lewis NS,
- Nocera DG
- ↵
- Whitesides GM,
- Crabtree GW
- ↵
- Moriarty P,
- Honnery D
- ↵
- Moriarty P,
- Honnery D
- ↵
- Turner JA
- ↵
- ↵
- ↵
- ↵
- ↵
- Eppinger J,
- Huang KW
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- Zhou X, et al.
- ↵
- Boddien A, et al.
- ↵
- ↵
- Ting S-W,
- Cheng S,
- Tsang K-Y,
- van der Laak N,
- Chan K-Y
- ↵
- Himeda Y
- ↵
- Bulushev DA,
- Beloshapkin S,
- Ross JRH
- ↵
- ↵
- Scholten JD,
- Prechtl MHG,
- Dupont J
- ↵
- ↵
- ↵
- Boddien A, et al.
- ↵
- Solymosi F,
- Koos A,
- Liliom N,
- Ugrai I
- ↵
- Tanaka R,
- Yamashita M,
- Chung LW,
- Morokuma K,
- Nozaki K
- ↵
- Luo Q,
- Feng G,
- Beller M,
- Jiao H
- ↵
- Mori K,
- Dojo M,
- Yamashita H
- ↵
- Gan W,
- Snelders DJM,
- Dyson PJ,
- Laurenczy G
- ↵
- Wang Z-L, et al.
- ↵
- Wang Z-L,
- Yan J-M,
- Wang H-L,
- Ping Y,
- Jiang Q
- ↵
- Zell T,
- Butschke B,
- Ben-David Y,
- Milstein D
- ↵
- Oldenhof S, et al.
- ↵
- Sponholz P,
- Mellmann D,
- Junge H,
- Beller M
- ↵
- Filonenko GA, et al.
- ↵
- Bielinski EA, et al.
- ↵
- Wang Z,
- Lu SM,
- Li J,
- Wang J,
- Li C
- ↵
- Onishi N, et al.
- ↵
- Iguchi M,
- Himeda Y,
- Manaka Y,
- Matsuoka K,
- Kawanami H
- ↵
- Celaje JJA, et al.
- ↵
- Li H,
- Zheng B,
- Huang K-W
- ↵
- ↵
- ↵
- ↵
- Chen H, et al.
- ↵
- ↵
- Naraharisetty SRG,
- Kurochkin DV,
- Rubtsov IV
- ↵
- ↵
- ↵
- ↵
- Lakowicz J
- ↵
- Chen HL,
- Bian HT,
- Li JB,
- Wen XW,
- Zheng JR
- ↵
- Chen H, et al.
- ↵
- Chen H,
- Bian H,
- Li J,
- Wen X,
- Zheng J
- ↵
- Zhao Y,
- Truhlar DG
- ↵
- ↵
Citation Manager Formats
Article Classifications
- Physical Sciences
- Chemistry