Aggregation of lactic acid in cold rare-gas matrices and the link to solution: a matrix isolation-vibrational circular dichroism study

Angelo Shehan Perera , Joseph Cheramy , Mohammad Reza Poopari and Yunjie Xu *
Chemistry Department, The University of Alberta, Edmonton, Alberta T6G 2G2, Canada. E-mail: yunjie.xu@ualberta.ca

Received 26th July 2018 , Accepted 18th September 2018

First published on 18th September 2018


The matrix isolation (MI) technique has been utilized with vibrational circular dichroism (VCD) spectroscopy to obtain MI-VCD spectra of lactic acid (LA) in cold argon matrices, in addition to their MI-IR spectra. The experiments have been done at three different deposition temperatures (10 K, 16 K and 24 K) under different Ar flow rates so that different degrees of LA self-aggregation occur. The structural and spectral investigations of the LA monomer and the larger (LA)2,3,4 aggregates have been undertaken at three levels of theory (B3LYP/6-311++G(2d,p), B3LYP-D3BJ/6-311++G(2d,p) and B3LYP-D3BJ/def2-TZVPD) to evaluate the effects of dispersion correction and basis sets on optimized structures, relative conformer energies, and IR/VCD spectral features. Interestingly, the relative conformer energies vary considerably with and without dispersion correction, especially when the molecule gets larger and when it is placed in solution. Such uncertainties in the relative energies and in the vibrational band positions and IR/VCD intensities highlight the challenges in interpreting experimental spectroscopic data, especially those obtained in solution. With the narrow MI-IR band width and highly characteristic MI-VCD spectral features and the trend observed at three temperatures, we have been able to correlate the spectral features confidently to those of the LA monomer and the larger (LA)2,3,4 aggregates, with the aid of theoretical modeling. Finally, by noting the similarity of MI-IR and especially MI-VCD features obtained at 24 K with those of the 0.2 M solution, and with the aid of spectral simulation at the B3LYP-D3BJ/def2-TZVPD level, a composition of LA aggregates dominated by the LA tetramer and trimer has been identified. This conclusion differs from the previous reports where the LA dimer was identified as the main species at even higher concentration in CDCl3. The present work showcases the power of MI-VCD spectroscopy in aiding solution spectral assignment and in providing insight into the complex self-aggregation behavior of LA in solution.


Introduction

Vibrational circular dichroism (VCD) spectroscopy1 is nowadays used extensively in the determination of absolute configurations and conformations of chiral molecules directly in solution.2–5 In recent years, this technique has emerged as an important tool for probing non-covalent interactions, such as solute–solute and solute–solvent interactions of chiral molecules in solution.6–10 In particular, such non-covalent interactions between a chiral solute and achiral solvent molecules may induce significant VCD intensities in the vibrational modes of the achiral solvent molecules, offering significant insight into solvent effects.11–14

On the other hand, the existence of multiple conformers, in addition to many more species generated through non-covalent interactions such as hydrogen (H)-bonding and halogen-bonding of solute–solute and solute–solvent molecules in solution, tend to generate broader and weaker IR and VCD bands in the solution, rendering concrete, detailed spectral assignments difficult. It is therefore highly desirable to minimize some of these line broadening factors. The combined approach of matrix-isolation (MI) and VCD spectroscopy, i.e. MI-VCD spectroscopy, offers a powerful new way to achieve this goal. Generally, the replacement of an interactive solvent environment by a cold inert-gas matrix environment removes the severe solute–solvent interactions encountered in solution, thus reducing the number of possible species containing the chiral solute of interest. This reduction, in addition to the low temperature environment, generates much better resolved MI-IR and especially MI-VCD features, which can be critical in clarifying the ambiguous band assignments in solution IR and VCD spectra of chiral molecules.15,16 Furthermore, by varying the deposition matrix temperature and sample composition,17–19 one can alter the degree of self-aggregation of chiral solutes or the complexation of the chiral solute with another molecule of interest. Since VCD intensities are typically only in the range of 10−4 to 10−5 times that of the corresponding parent IR intensities, this places stringent requirements on the quality of the prepared cold rare gas matrices. It is also highly desirable to measure MI-VCD spectra of both enantiomers to verify the reliability of the bands observed. Such measurements may be difficult in practice because one enantiomer may be quite rare naturally, difficult to synthesize, or can contain different impurities that may interfere with the final measurements. Nevertheless, very good quality mirror image MI-VCD spectra have been demonstrated.17,18

The system of interest in the current study is the lactic acid (LA) monomer and its aggregates. LA belongs to the group of chiral alpha hydroxy acids. Biologically, LA is considered to be important mainly because of its involvement in the biological reactions as a metabolite. Also, the biodegradable character of LA based polymer derivatives has extended their applications to a range of different fields.20,21 Because of its extensive applications in many fields, LA has been the focus of many experimental and theoretical investigations. The most stable monomer of LA, M1, and one higher energy conformer, M4, were identified using microwave spectroscopy in the gas phase.22–24 M1–4 are the four most stable LA monomer conformations predicted in the gas phase which will be discussed in the main section in detail. Borda et al. carried out an MI-IR study of LA and identified not only the most stable conformer, but also M2 and M4 in their MI-IR spectrum aided by DFT calculations.25 Because of its propensity to form larger aggregates, the aggregation behaviour of LA in solution and solid-state has been the subject of interest in many prior investigations. Shouten et al. investigated the crystalline structure of LA at 100 K using the X-ray diffraction technique.26 The aggregation of LA in water was studied using IR and Raman spectroscopic methods,27,28 where possible intermolecular association compounds connected by H-bonds between the hydroxyl groups were tentatively proposed in the early 90s, although no specific species were identified. The dominant existence of the cyclic OH⋯O eight-membered ring LA dimer as the main species in water was suggested by Fekete et al. using both IR spectroscopy and ab initio calculations.29 Later on, Losada et al. applied both IR and VCD spectroscopy with ab initio calculations to investigate the aggregation of LA in CDCl3, water and CH3OH.30 In the interpretation of IR and VCD spectra of LA in CDCl3, they considered the contribution of LA up to the binary size, although possible contribution from even larger aggregates were only mentioned briefly since further assignment was hampered by the broad spectral width and the potential existence of many species.

In this study, we take advantage of the much narrower linewidth and the better sample control afforded by the MI technique to unravel the aggregation behaviour of LA. We first aim to obtain high quality MI-VCD spectrum of the LA monomer, which is extremely difficult to do in solution because of severe aggregation even at low LA concentration. Secondly, we will utilize the MI technique to assist the selective formation of different sizes of LA aggregates with controlled sample conditions in a cold rare-gas matrix and obtain their corresponding MI-IR and MI-VCD spectra. By comparing the MI-IR and especially the MI-VCD features obtained under controlled conditions to those obtained in solution, we aim to achieve concrete assignment of the solution species with the aid of high level theoretical calculations.

Methods

Experimental details

Both L-(+)-LA (≥98%) and D-(−)-LA (≥90%) were obtained from Sigma Aldrich and used without further purification. One main difference in their purity seems to be the amount of water in the sample. The solid powder samples were placed in a stainless-steel sample reservoir situated right before the sample injection tube mounted at the cold head and were kept at room temperature (25 °C). The stainless-steel reservoir was evacuated for one or two hours to remove water trapped in the samples. The flow rate of the Ar backing gas was held constant at 14 sccm and 1.5 sccm (sccm = standard cubic centimeters per minute) for the MI experiments of the LA monomer and the LA aggregates using a flow controller (MKS 1179A), respectively. The matrix isolation experimental setup contains a closed-cycle helium cryostat from Advanced Research Systems, Inc. (ARS 4HW compressor with a DE 204SI expander) and a stainless-steel vacuum line. Several deposition temperatures (10 K, 16 K and 24 K) were used. At a 10 K deposition temperature where the flow rate was set to 14 sccm, the deposition was carried out for one hour and at 16 K and 24 K where the flow rate was set to 1.5 sccm, the deposition was performed for three hours and 30 minutes. The high Ar flow rate at 10 K was used to optimize the generation of the LA monomer in the matrix, while the about ten-times slower flow rate at 16 K and 24 K was to aid the formation of the LA aggregates.

The experimental IR and VCD spectra were measured using a Bruker Vertex 70 supplemented with a PMA 50 module for polarization modulated measurements. All MI-IR and MI-VCD spectra were measured with a spectral resolution of 2 cm−1. The MI-VCD spectra were collected with an approximately 30[thin space (1/6-em)]000 scans (12 h acquisition time) which were carried out as four individual sets, where each set corresponds to a 3 h acquisition time. The photoelastic modulator (PEM) was set to 1700 cm−1, which improves the reliability of the MI-VCD features obtained in the carbonyl stretching absorption region. Measurements done with the PEM at 1400 cm−1 show little difference for all the lower wavenumber bands. Since the purity of the L-(+)-LA and D-(−)-LA samples was much higher than those reported in ref. 30 where L-(+)-LA (90%) and racemic LA (85%) were used, we also re-measured the 0.1 M and 0.2 M solution IR and VCD spectra of LA in CDCl3. For solution preparation, chloroform-d, 99.8% D from Sigma Aldrich was used. The final MI-VCD and solution VCD spectra were produced using the standard (S-R)/2 procedure, while the final solution IR spectra were solvent subtracted and the raw VCD data are provided in Fig. S1, ESI.

Computational details

All geometry optimizations, harmonic frequency calculations, and IR and VCD intensity calculations were performed using the Gaussian 1631 program package. All computations were undertaken using the Becke, three-parameter, Lee–Yang–Parr (B3LYP)32 functional and the 6-311++G(2d,p)33 basis set initially. The B3LYP hybrid functional was selected mainly due to its reasonable accuracy in simulating VCD intensities.34,35 More recently, it was recognized that the inclusion of the dispersion-correction36 with Becke–Johnson damping37 (B3LYP-D3BJ) provides much more accurate conformational geometries and energies than without,38 as also demonstrated by many rotational spectroscopic studies of monomers and H-bonded molecular systems, for example the trifluoroethanol trimer.39 Therefore, calculations with the B3LYP-D3BJ functional were carried out. Furthermore, since diffuse functions are considered important for chiroptical calculations,34 we also performed the calculations with 6-311++G(2d,p) and def2-TZVPD basis sets for comparison.40 The theoretical IR and VCD spectra were simulated using a Lorentzian line shape with half-width at half maximum (HWHM) of 4 cm−1 for the comparison with the experimental MI-IR and MI-VCD spectra. No frequency scaling was applied. Please note that we use the S-enantiomer which is L-(+)-LA for all the calculations and comparison with the experimental data throughout this paper unless otherwise specified explicitly.

Results and discussion

Experimental MI-IR and MI-VCD spectra

In this study, we aim to obtain MI-IR and MI-VCD spectra corresponding to the LA monomer and larger aggregates. The goal is to obtain well separated and distinctive IR and VCD spectral features for these species experimentally in order to ultimately achieve concrete assignment of solution spectra. After many trialed experiments at several source temperatures, it was recognized that to achieve good quality MI-VCD spectra it is best to leave the sample at 25 °C and evacuate the sample reservoir for one to two hours to remove residual water. This step seems to be crucial to obtain good mirror-imaged MI-VCD spectra of D- and L-LA enantiomers, possibly because the D- and L-samples purchased came with much different amount of water, and excessive heating and less pumping before the start of deposition tended to cause noticeable variation in the exact species composition in the D- versusL-experiments.

The experimental MI-IR and MI-VCD spectra of LA obtained are presented in Fig. 1. The MI-IR features recorded at 10 K are sharp in nature and agree generally with the MI-IR results recorded in the Ar matrix at 9 K by Borda et al.25 Please note that there is a gap from 1700–1600 cm−1 in Fig. 4 of ref. 25. The raw experimental MI-VCD spectra of L- and D-forms of LA reveal clear mirror images for the sharp bands and the MI-VCD spectra in Fig. 1 show in general good baseline.


image file: c8cp04748k-f1.tif
Fig. 1 Comparison of the experimental MI-IR and MI-VCD spectra of LA obtained at 10 K, 16 K and 24 K. The dashed lines indicate the contribution from the LA monomer while the dashed-dotted lines indicate the contribution from the LA aggregates such as the LA dimer, trimer and tetramer. Please see text for discussion. The broad feature indicated with * is artifact due to some minor baseline variation in the MI-measurements.

The strong MI-IR peak obtained at 10 K at 1767.3 cm−1 loses a significant portion of its IR intensity at 16 K and even further at 24 K. In the previous MI-IR study, this band at 1767.3 cm−1 at 10 K was assigned to the C[double bond, length as m-dash]O stretching of M1 and the weak bands observed at higher wavenumber were assigned to M2 and M4.25 The three weak MI-IR bands observed at 10 K at lower wavenumber than 1767.3 cm−1 seem to line-up well with the corresponding bands at the 16 K spectrum, i.e. at 1750.7 cm−1 (shoulder), 1729.2 cm−1 and 1713.2 cm−1, with noticeable increase in their relative intensity to 1767.3 cm−1 when compared to those at 10 K. At 24 K, the relative intensity of the bands below 1740 cm−1 increases further at the expense of those above 1740 cm−1. Based on the above comparison, we hypothesize that the three lower wavenumber MI-IR peaks in the 1700–1800 cm−1 region are likely associated with larger LA aggregates, i.e. LA dimer, trimer, and tetramer, etc. The appearance of larger aggregates can also be seen in the IR bands below 1200 cm−1 where the three sharp peaks at 10 K become broader and with obvious shoulders at 16 K and 24 K. In the 1200–1300 cm−1 region, we observe significant relative intensity increase going from 10 K to 16/24 K. In contrast, the two bands at ∼1327.5 and 1320.4 cm−1 exhibit a noticeable decrease in their relative intensity, while those centred at 1378.7 cm−1 show noticeable broadening.

To assist the discussion, we use dashed lines to mark the VCD features at 10 K and extend them to all IR and VCD spectra at the three temperatures in order to correlate the corresponding features, a task made easy by many sharp MI-VCD features observed at 10 K. Interestingly, a close examination shows all these sharp VCD features have their corresponding sharp IR features in the 10 K spectrum. Furthermore, these bands generally become relatively less intense as one moves to higher temperatures, supporting the aforementioned assignment that they belong to the LA monomer. A further interesting observation is that the MI-IR and MI-VCD features at 16 K and 24 K have broader bands and look fairly alike, suggesting that the dominant species at these two temperatures may be similar. Furthermore, the overall VCD/IR intensity ratios at 16 K and 24 K are similar to each other and are much larger than that at 10 K. To assign the observed spectral features, we need to simulate IR and VCD spectra for the LA monomer and its larger aggregates. The associated calculations are discussed in the next section.

Energetics and geometries of the LA monomer and (LA)nn = 2, 3, 4 aggregates

Here we examine the possible monomeric, binary, ternary and quaternary conformations of LA. The structural aspects of the LA monomer were discussed in several previous investigations.7,25,41,42 The four energetically relevant conformations at 298 K identified before7,25 have been reoptimized at the B3LYP-D3BJ/6-311++G (2d,p) and B3LYP-D3BJ/def2-TZVPD levels of theory. Similar geometries were obtained with the inclusion of D3BJ and are summarized in Fig. 2. M1, M2/M3, and M4 exhibit an OHalcohol⋯Ocarbonyl, OHalcohol⋯Oacid and OHacid⋯Oalcohol intramolecular H-bond, respectively. The subtle difference in M2 and M3 conformers comes from which electronic lone pair of Oacid is used for the intramolecular H-bond. Please note that M1, M2, M3 and M4 correspond to SsC, GskC, GskC, and AaT in ref. 25, respectively, and are shown in order of decreasing stability according to calculations at the B3LYP/6-311++G(d,p) level of theory. Please note that this relative stability order may alter depending on the levels of theory and whether the molecule is in the gas phase or in solution. Their relative free energies calculated at the three levels of theory at 298 K in the gas phase and with the PCM of CDCl3 are compared in Table 1, as are the associated Boltzmann factors. Other additional higher energy conformers are without the intramolecular H-bonds shown above and make negligible contribution at 298 K. We note that the changes in the relative free energies among the three levels of theory are small in the gas phase, whereas the addition of the PCM of CDCl3 already alters the stability ordering of the higher energy conformers and the Boltzmann factors noticeably.
image file: c8cp04748k-f2.tif
Fig. 2 The optimized geometries of the four most stable conformers of the LA monomer in the gas phase are shown in the first row where the name “M” stands for monomer. The names in the brackets are the corresponding names used in ref. 25. The optimized geometries of the most stable dimer (D) and the representative trimer (T) and tetramer (Te) conformers of LA in the gas phase are given in the subsequent rows. The intra- and intermolecular H-bond lengths (in Å) are indicated with dashed lines for each structure, as well as the OH⋯O H-bond angles (θs).
Table 1 Comparison of the relative free energies (ΔG in kJ mol−1) and Boltzmann factor (Bf in %) at 298 K of the LA monomer conformers computed at the B3LYP and B3LYP-D3BJ/6-311++G(2d,p) and the B3LYP-D3BJ/def2-TZVPD levels of theory
Conf. B3LYP/6-311++g(2d,p) B3LYP-D3BJ/6-311++G(2d,p) B3LYP-D3BJ/def2-TZVPD
ΔG Bf ΔG Bf ΔG Bf
a The values in brackets are obtained with the PCM of CDCl3 added to the calculations.
M1 0.0 92.9 0.0 92.5 0.0 92.5
(0.0)a (80.4) (0.0) (77.7) (0.0) (83.7)
M2 8.2 3.3 8.4 3.1 8.5 3.0
(6.0) (7.3) (5.9) (7.0) (6.3) (6.5)
M3 9.1 2.4 8.9 2.5 9.0 2.5
(6.9) (4.9) (7.0) (4.6) (7.0) (4.9)
M4 10.4 1.4 9.6 1.9 9.4 2.0
(7.4) (7.4) (4.9) (10.7) (7.0) (4.9)


The formation of LA aggregates is readily facilitated by the availability of aliphatic hydroxyl and carboxylic acid functional groups in LA which can serve both as H-bond acceptors and donors. An extensive set of LA dimer conformations can be generated from these H-bond donor and acceptors sites utilizing different intermolecular H-bonding topologies. The preliminary calculations by Fekete et al. at the HF level identified the cyclic carboxylic–carboxylic binary conformations (Fig. 2) as by far the most stable ones whereas all the other binding topologies resulted in structures which are about 30 kJ mol−1 less stable.29 For this reason, we may expect six cyclic carboxylic–carboxylic LA dimers, i.e. M1M1, M1M2, M1M3, M2M2, M2M3 and M3M3 which utilize the carboxylic acid functional groups for the intermolecular H-bonds. These structures were reported previously by Losada et al.30 and are re-optimized in this study with the inclusion of dispersion correction with damping factor. The geometries of the six stable LA dimer conformations are shown in Fig. 2 and the corresponding relative energies and Boltzmann factors of these stable dimers of LA in the gas phase and with the PCM of CDCl3 are compared at the three levels of theory in Table 2.

Table 2 Comparison of the relative free energies (ΔG in kJ mol−1) and Boltzmann factor (Bf in %) at 298 Ka of the LA dimer conformers computed at the B3LYP and B3LYP-D3BJ/6-311++G(2d,p) and the B3LYP-D3BJ/def2-TZVPD levels of theory
Dimer conf. B3LYP/6-311++g(2d,p) B3LYP-D3BJ/6-311++G(2d,p) B3LYP-D3BJ/def2-TZVPD
ΔG Bf ΔG Bf ΔG Bf
a For the relevant cold matrix temperatures of 10, 16 and 24 K, only the most stable conformer D1 contributes (∼100%) to the experimental spectra. See text for discussion. b The values in brackets are obtained with the PCM of CDCl3 added to the calculations.
D1 0.0 86.4 0.0 86.2 0.0 83.7
(0.0)b (92.0) (0.0) (78.3) (0.0) (72.1)
D2 6.4 6.5 6.5 6.4 6.0 7.5
(8.1) (3.5) (5.0) (10.5) (4.6) (11.3)
D3 6.8 5.6 6.6 6.0 6.3 6.7
(8.2) (3.4) (5.4) (8.9) (4.3) (12.7)
D4 12.6 0.5 12.7 0.5 11.8 0.7
(13.9) (0.3) (10.9) (0.9) (9.3) (1.7)
D5 13.2 0.4 12.8 0.5 12.0 0.7
(14.6) (0.4) (11.6) (0.7) (9.8) (1.4)
D6 12.9 0.5 12.7 0.5 11.8 0.7
(13.2) (0.3) (12.0) (0.6) (10.8) (0.9)


Again, the relative energy ordering of the LA dimer conformers and their Boltzmann factors in the gas phase appear very consistently among the three levels of theory. Much more noticeable differences are observed among the three levels of theory when the PCM of CDCl3 is introduced. It is interesting to note that the relative dimer stability hinges largely on the respective monomeric subunits. For example, D1 which consists of two M1 (by far the most stable monomeric unit), is by far the most stable dimer, followed by D2 and D3 which contain one M1 each. This is perhaps not surprising because the central intermolecular H-bonding topologies are essentially the same among all these dimers.

We have also explored larger LA aggregates in light of the experimental data presented in the previous section. A third LA molecule can form H-bonding interactions with the most stable M1M1 dimer in two main ways: (1) insertion of the third LA molecule into one of the intramolecular H-bonded rings of the dimer; (2) insertion of the third LA molecule into the existing intermolecular H-bonded ring. When the third M1 utilizes its carboxylic acid group to form the new intermolecular H-bonded ring, this results in T1 and T2 (Fig. 2) each of which consists of three M1 subunits. In T1, the insertion of the third M1 molecule happens above the plane of intramolecular H-bonded ring, whereas in T2, the insertion of the third M1 molecule happens below the aforementioned plane. If the third M1 molecule instead uses its C[double bond, length as m-dash]O and OH groups to form the new intermolecular H-bonded ring, the resulting trimer structures, i.e. T4 and T5, are much less stable than T1 and T2. If one replaces the third M1 with M2 or M3 in T1, two additional structures, T6 and T7, which are much less stable than T1 and T2, are obtained. Geometries of T4–T7 are given in Fig. S2, ESI. In the second scenario, the existing eight-member intermolecular H-bonded ring in the M1M1 dimer gets extended into a twelve-member intermolecular H-bonded ring, resulting in T3 (Fig. 2). Again, T3 is much less stable than T1 and T2.

In terms of tetramers, the remaining intramolecular H-bonded rings of T1 and T2 trimers could assist the insertion of the fourth M1 to form a LA tetramer. Like in the case of trimers, the most stable tetramer, Te1, is generated from the T1 trimer where the insertion of the fourth M1 also happens above the plane of intramolecular H-bonded ring. The second stable tetramer, Te2, is also generated from the most stable T1 trimer. But unlike in Te1, the insertion of a fourth M1 comes below the plane intramolecular H-bonded ring, i.e. with the third and fourth M1 molecules above and below, respectively. Te3 has the third and fourth M1 molecules approach the D1 dimer below the planes of intramolecular H-bonded rings. We further confirmed that if the third and/or fourth M1 uses instead its C[double bond, length as m-dash]O and OH groups to form the new intermolecular H-bonded ring, the resulting tetramers (Te4 and Te5 in Fig. S2) are much less stable. The geometries of the three most stable tetramer structures, Te1–Te3, are provided in Fig. 2.

For easy comparison of the computational results of the representative structures, the relative free energies and the Boltzmann factors of T1–T3 and Te1–Te3 in the gas phase and with the PCM of CDCl3 at the three levels of theory are summarized in Table 3. Here we observe very noticeable changes in the relative free energies and therefore their Boltzmann factors among the ternary (T1–T3) and quaternary (Te1–Te3) LA complexes with the inclusion of the D3BJ correction: some stability orderings are reversed in the gas phase. Such variations due to the inclusion of the D3BJ correction become even more drastic when these complexes are placed in the PCM of the solvent. While the calculations with the inclusion of D3BJ dispersion correction generally appear consistent with each other with the two different basis sets used, the discrepancies in the relative free energies are still large as the molecular system gets larger. Since weakness of calculations is known to affect the entropic component more severely, we also listed the corresponding relative energies in Tables S1–S3, ESI. The trend observed for ΔG is still there for ΔE, although the corresponding discrepancies are smaller for ΔE than ΔG. All these observations highlight the difficulties one may face in modelling larger aggregates in solution. The sensitivity of the IR and VCD spectral features to the D3BJ correction and the inclusion of PCM will be discussed later on. For consistency, all theoretical interpretations are based on the calculations at the B3LYP-D3BJ/def2-TZVPD level, either performed in the gas phase or with a PCM of CDCl3 for the remainder of the paper.

Table 3 Comparison of the relative free energies (ΔG in kJ mol−1) and Boltzmann factor (Bf in %) at 298 Ka of the LA trimers and tetramers computed at the B3LYP and B3LYP-D3BJ/6-311++G(2d,p) and the B3LYP-D3BJ/def2-TZVPD levels of theoryb
(LA)3&(LA)4 conf. B3LYP/6-311++g(2d,p) B3LYP-D3BJ/6-311++G(2d,p) B3LYP-D3BJ/def2-TZVPD
ΔG Bf ΔG Bf ΔG Bf
a For the relevant cold matrix temperatures of 10, 16 and 24 K, only the most stable conformers T1 and Te1 contribute predominantly (≥98%) to the experimental spectra. b Only the three conformers listed are included in the Boltzmann factor calculations for easy comparison among different levels of theory. Additional high energy conformers were optimized and their relative energies calculated at the B3LYP-D3BJ/def2-TZVPD level and the results are given in ESI. c The values in brackets are obtained with the PCM of CDCl3 added to the calculations.
T1 0.0 43.5 0.0 61.2 0.0 63.0
(0.0)c (50.9) (0.3) (46.8) (0.0) (67.5)
T2 0.8 31.8 1.2 38.1 1.4 35.4
(0.6) (39.2) (0.0) (52.2) (1.8) (32.0)
T3 1.4 24.7 11.2 0.7 9.1 1.6
(4.0) (9.9) (11.2) (0.6) (13.9) (0.3)
Te1 0.0 49.4 0.0 44.4 0.0 48.0
(1.4) (28.1) (0.0) (57.7) (0.0) (76.9)
Te2 2.5 18.1 0.6 35.4 1.0 32.2
(1.9) (22.6) (4.9) (8.0) (6.7) (5.1)
Te3 1.0 32.5 1.9 20.4 2.2 19.8
(0.0) (49.4) (1.3) (34.2) (3.6) (17.9)


Assignment of the MI-IR and MI-VCD spectra at 10 K

As discussed in the experimental result section, the sharp bands observed in both the MI-IR and MI-VCD spectra of LA obtained at 10 K provide a solid experimental foundation for the interpretation using the simulated IR and VCD features of the LA monomer. While the assignment of MI-IR spectrum to the LA monomer had previously been discussed in ref. 25, there were considerable uncertainties in the assignment in the 1150–1500 cm−1 region where poor agreement between the experimental MI-IR features and the calculated ones was attributed to matrix site splitting and Fermi resonance. Here, we take advantage of the unique MI-VCD features to complement the MI-IR features to nail down the assignment.

The simulated IR and VCD spectra of the four most stable LA conformers in the gas phase at the B3LYP-D3BJ/def2-TZVPD level of theory are shown in Fig. 3. For completion, the simulated IR and VCD features for the four conformers at the three levels of theory are compared in Fig. S3, ESI. For the LA monomeric conformers, the variation in the geometries, relative energies, and spectral features are all relatively small among the three levels of theory used. The previous microwave spectroscopic study24 identified that the abundance of M4 is about 2% that of M1, pretty close to the predictions by the two B3LYP-D3BJ calculations. In Fig. 3, we use the dashed lines to connect the sharp experimental VCD and IR features and then correlate them to the simulated spectral features. To assist discussions, we label all main IR and the corresponding VCD bands in M1 with a, b, c, etc. and label the assigned bands in the experimental data accordingly. Because of the positive and negative VCD features of these sharp bands, it is straightforward to assign the IR and VCD bands. Every band in M1 has its spectral features identified in the experimental MI-IR and MI-VCD spectra. Overall, the agreement between experiment and theory for both IR and VCD spectra is excellent. It is interesting to note a few differences in the current assignment and that in ref. 25. The shoulder band e at 1250.5 cm−1 can now be definitely assigned to M1 based on its distinctively large negative VCD. The split bands marked with *, which were previously assigned to M1, are mainly contributions from the larger LA aggregates, which will be discussed later on. Instead, we assign the experimental bands at 1328.6 (g) and 1323.1 cm−1 (f) approximately to the bending mode of CCalcoholH (g) and bending mode of HOCacid/COalcoholH (f) of M1 which exhibit a distinctive position/negative VCD couplet. In the 1350–1500 cm−1 region, the assignment of the three weak IR bands, i.e. i, j and k, were not certain before. Now, with the corresponding distinctive medium to strong −/+/− VCD pattern, one can identify them confidently.


image file: c8cp04748k-f3.tif
Fig. 3 Comparisons of the experimental MI-IR and MI-VCD spectra of LA obtained at 10 K with the simulated IR and VCD spectra of the monomeric LA conformers in the gas phase and also with the related population weighted spectra. The dashed lines connect the corresponding experimental IR and VCD features with the simulated ones. The thicker/thinner lines are used for the experimental VCD of S/R-LA, respectively. * indicates the contributions from the larger LA aggregates. See the main text for discussion.

How about the contribution from the less stable LA monomeric conformers? So far, the assigned MI-IR and MI-VCD spectral features are all consistent with those of M1 which has a dominant population of ∼93% of the total monomer population. Because of the interference from the LA aggregates, the only clear indication of the higher energy conformers resides in the region higher in wavenumber than the C[double bond, length as m-dash]O stretching band of M1. Since M2 and M3 are close in terms of their structures and are also close in energy, we verified their interconversion barrier, which is 1.9 kJ mol−1 from M3 to M2 at the B3LYP-D3BJ/def2-TZVPD level. We further estimated the zero-point energy corrected barrier, which is predicted at the same level of theory to be 0.7 kJ mol−1. This small barrier can be overcome at 10 K based on the Barnes relation.43 With this consideration in mind, we have the following populations: M1 (92.5%), M2 (5.5%) and M4 (2.0%) for the MI spectra at 10 K. Without consideration of conformational conversion from M3 to M2, one would have obtained a stronger negative C[double bond, length as m-dash]O stretching VCD band of M2/M3 than that at M1. This is because this M3 VCD band is much more negative than that of M2 and they are at essentially the same C[double bond, length as m-dash]O stretching wavenumber. Such a prediction would be inconsistent with the experimental VCD data where M1 has the most negative VCD intensity. The two small IR bands observed at 1788 and 1781 cm−1 can be assigned to M4 and M2/M3, respectively. While the 1788 cm−1 band is too weak and too close to the cut off of the IR filter for a good VCD measurement, the negative VCD band associated with M2 is visible in the experiment. Some weaker experimental MI-IR peaks observed in the 1700–1760, 1250–1300 and 1130–1150 cm−1 regions and marked with * have no corresponding features in the simulated IR/VCD spectra of the LA monomer. As discussed in the previous section, these are likely from the larger LA aggregates. Their detailed assignments will be discussed in the next section in relation to the MI-IR and MI-VCD spectra obtained at 16 K and 24 K.

Assignments of the 16 and 24 K MI-IR and MI-VCD spectra and the self-aggregation of LA in an Ar matrix

The gradual variations of the MI-IR and MI-VCD spectral features going from 10 K, to 16 K, and finally to 24 K have provided insights into the self-aggregation of LA. In order to evaluate the spectral features associated with the LA aggregates, the IR and VCD spectra of the binary, ternary and quaternary LA aggregates in the gas phase have been simulated and are shown in Fig. S4–S6, ESI, respectively, at all three levels of theory. Generally speaking, both the IR and VCD spectral features are fairly consistent among the three levels of theory, although with some small variations in the wavenumber positions. This provides confidence in using them to interpret the MI data. The only exception is T3 which shows much different VCD features in the C[double bond, length as m-dash]O stretching region at the B3LYP level versus those at the B3LYP-D3BJ level. This VCD alternation is associated with the obvious change in the optimized T3 geometry when D3BJ is included (see Fig. S7, ESI, for a comparison). The two different basis sets, 6-311++G(2d,p) and def2-TZVPD, using the B3LYP-D3BJ functional provide very consistent results across the board.

Since the depositions were done at the low temperatures of 16 K and 24 K over long period of time, we assume that the conformational temperatures of the H-bonded LA aggregates are the same as the matrix temperatures. At such low temperatures, only the most stable conformer is dominantly populated (≥98%) for every LA aggregate size. Even if the conformational temperature is slightly higher than the one assumed, the resulting MI-IR and MI-VCD spectra are still expected to be dominated by the most stable conformer of each species. For this reason, the simulated IR and VCD spectra of the most stable conformer of LA dimer, trimer and tetramer are shown in Fig. 4, together with the experimental MI-IR and MI-VCD spectra obtained at 10, 16, and 24 K. For the monomer, the same population weighted IR and VCD spectra used before are depicted in Fig. 4. Based on the experimental analysis presented above and the simulated IR and especially VCD features, we can assign the IR and VCD bands. We label the IR/VCD bands of the LA dimer with a′, b′ and c′, etc., while those of the LA trimer/tetramer are with a′′/a′′′, b′′/b′′′ and c′′/c′′′, etc., respectively. For the LA trimer/tetramer, only relevant bands are labelled for simplicity.


image file: c8cp04748k-f4.tif
Fig. 4 Comparisons of the experimental MI-IR and MI-VCD spectra of LA obtained at 10 K, 16 K and 24 K with the simulated IR and VCD spectra of the monomeric, binary, ternary and quaternary LA species in the gas phase. The band assignment made for the 10 K spectra in Fig. 3 is used as the base to recognize the new features at 16 K and 24 K. The largest contribution to the 16 K spectra is from the LA dimer, identified with bands labelled with a′, b′, etc. The additional contribution to the 24 K spectra from the LA trimer are identified with bands labelled with a′′, b′′, and so on. The empirically population weighted spectra at 16 K and 24 K are also presented. () indicates an assignment is only tentative. Please see the main text for the detailed discussions.

At 16 K, the three IR bands to the red of the C[double bond, length as m-dash]O stretching band of M1, visible at 10 K, become much stronger relative to that of M1. These bands can be tentatively assigned to the LA dimer, trimer and tetramer based on the comparison to the simulated spectra of each aggregated species. While the furthest red shifted band, l′, can be assigned to the LA dimer, the positions of the trimer and tetramer are predicted to be essentially overlapped in contrast to the separated bands observed. This small although crucial deviation in the predicted versus experimental band positions could have dramatic consequence for the appearance of the VCD signatures in this region because of the multi-signate nature of the associated VCD bands. For the example, the +/− VCD bands underneath l′′ may overlap more to produce just a negative band.

The contribution of the LA monomer can also be recognized from the distinctive −/+/− VCD feature associated with a/b/c bands, along with some other similarities to the 10 K spectral features. With the availability of the distinctive experimental MI-IR and MI-VCD spectral features at 10 K, we can estimate the contribution from the LA monomer to the −/+/− feature to be about 10% of its total intensity. Please note that the experimental VCD intensity at 10 K is on the right side and if it were plotted on the same scale as those at 16 K and 24 K, it would be about half the height as currently depicted. The calculated LA monomer and dimer VCD spectra are amplified by a factor of 2 for easy visualization of the spectral features. Interestingly, the distinctive −/+/− feature mentioned above is predicted not only for the LA monomer but also for its aggregates with only minor variation in the band positions, although the relative intensity of the positive centre band versus the two negative side bands drops moving from the monomer to the dimer and the trimer. Therefore, it is not surprising this distinctive feature remains very much the same at different deposition temperatures. On the other hand, the observed broadening of the bands and the drop in the relative intensity of the positive centre band versus the two negative side bands points to the contributions by the LA dimer and also the trimer. A further detailed examination of the IR and VCD features in the 1500–1200 cm−1 region indicates that the overall IR and especially the VCD features can be largely attributed to the LA dimer with some contribution from the trimer and monomer. For example, the experimental VCD features, e′ and e′′ at 16 K/24 K are blue shifted from those at 10 K, consistent with the predicted positions of the e′ of the LA dimer and e′′ of the LA trimer. Similarly, the very noticeable blue shift of the positive j′ at 16 K/24 K from j at 10 K is captured by the prediction of the blue shift j’ of the dimer with respect to J of the monomer.

The detailed band assignments are indicated in Fig. 4 for the 16 K spectra. The 24 K spectra are very similar to those at 16 K, except they contain a bit more relative contribution from the larger LA aggregates and less monomer than at 16 K. For conciseness, in the 24 K spectra, we label only the bands with more obvious contribution from the LA trimer at 24 K than at 16 K. The empirically population weighted IR and VCD spectra at 16 K with 31% monomer, 60% dimer, 8.5% trimer and 0.5% tetramer and those at 24 K with 20% monomer, 65% dimer, 12% trimer and 3% tetramer are also included in Fig. 4 for comparison. Overall, the agreement with the experimental data is very good, except in the C[double bond, length as m-dash]O stretching region in part due to the inaccuracy in the predicted relative band positions for larger aggregates and the cancellation of the multi-signate VCD features.

From Fig. 4, one may consider assigning the 10 K spectra to the LA trimer even though the agreement is worse for some VCD bands. For example, the experimental a, b and c VCD bands are very sharp, whereas (LA)3 has additional VCD bands predicted close to a′′, and its c′′ band shows splitting, inconsistent with the experiment. The d VCD band observed has no corresponding feature in (LA)3. Furthermore, at a deposition temperature of 10 K, the main species trapped in the matrix are the species already present in the gas phase, based on our own and others’ experience.17,18,25 The rotational spectroscopic (gas phase) studies show that the LA monomer dominates the experimental rotational spectrum.24 It was clear that we had mostly the LA monomer in the gas phase before deposition.

We recognize that it is still quite challenging to obtain accurate theoretical prediction of band position and intensity, especially for larger aggregates. Such deficiency makes it very difficult to reproduce experimental data with many different species since minor frequency shifts and intensity variation can generate very different final VCD patterns. This deficiency is particularly acute in the current case in the C[double bond, length as m-dash]O stretching region where the trimer and tetramer exhibit extremely strong VCD intensity due to exciton coupling,44–46 a concept recently illustrated using bicamphor molecules by Abbate and co-workers.47 A minor inaccuracy in the band positions and/or intensities of these extremely strong VCD couplets may change the pattern and appearance completely. We caution against over interpreting the VCD signatures at the C[double bond, length as m-dash]O region under such condition. The availability of the MI-IR and MI-VCD at multiple temperatures and experimental conditions allows one to follow the aggregation process sequentially and makes the assignment conclusive.

Re-examination of self-aggregation of LA in CDCl3

Self-aggregation of LA in CDCl3 is considerably more complicated than in a cold rare-gas matrix because at 298 K many more conformers of each species are populated and because of solvent effects. It is also clear from the discussion in the theoretical modelling section, that the relative free energies and their Boltzmann factors at 298 K for the larger LA aggregates in solution change very noticeably from the gas phase to solution (see Table 3). To further complicate the interpretation, the aforementioned variations strongly depend on the levels of theory used. We therefore aim to use the experimental MI-IR and MI-VCD spectra obtained at multiple temperatures to aid the solution assignment. In Fig. 5, the solution IR and VCD obtained at 0.1 M and 0.2 M are compared with the experimental MI data at 24 K.
image file: c8cp04748k-f5.tif
Fig. 5 Top: Comparisons of the solution IR and VCD spectra of LA in CDCl3 recorded at 0.1 M and 0.2 M with the MI-IR and MI-VCD spectra of LA obtained at 24 K. The 0.1 M trace is also rescaled and superimposed on the 0.2 M trace for easy comparison. Bottom: The population weighted IR and VCD spectra of the LA monomer, dimer, trimer and tetramer in CDCl3 at 298 K. The peak assignments are indicated by the dashed lines.

Also included in Fig. 5 are the population weighted IR and VCD spectra of the LA monomer, dimer, trimer and tetramer species at 298 K. We use the predicted Boltzmann population factors at the B3LYP-D3BJ/def2-TZVPD level of theory because of its reliability in terms of conformational geometries and relative energies.38 The individual conformer IR and VCD spectra of the LA monomer, dimer, trimer and tetramer, and their population weighted spectra at 298 K are provided in Fig. S8–S11, ESI for comparison. It is interesting to note that with the PCM of CDCl3, the negative C[double bond, length as m-dash]O VCD band of M1 now becomes positive. Since the sign of a VCD band is determined by the sign of cos[thin space (1/6-em)]α where α is the angle between the electric and magnetic dipole transition moment vectors of the VCD mode, one labels such a mode non-robust if α is near 90°.48 The α values of the C[double bond, length as m-dash]O stretching modes are listed in Table S4, ESI, for the main conformers of each LA species calculated. For M1 and some angles in D1 and T1, these α values are in the range of 82.1–90.2°, indicating that these are the non-robust modes. Another intriguing observation is that the C[double bond, length as m-dash]O VCD features for the most stable and the second most stable conformers of the dimer, trimer and tetramer are essentially opposite in signs in every aggregate size. Any change in the α values and in the population factors for these two most stable conformers can alter the final appearance of the VCD features in this region. A similar issue associated with non-robust conformer population was also reported recently.49 Again, this highlights the challenge one faces to correctly predict the C[double bond, length as m-dash]O stretching VCD signatures.

Overall, the IR and VCD spectra obtained at 0.1 M and 0.2 M share similar spectral features and are also similar to the MI-IR and MI-VCD spectra obtained at 24 K. There are, however, some small yet informative changes, especially in the VCD spectra. The broadening of the VCD bands at 0.2 M compared to those at 0.1 M suggest that larger LA aggregates become more dominant at higher concentration. The VCD features in the C[double bond, length as m-dash]O stretching region become more intense relative to the VCD features in the lower wavenumber region in the 0.2 M solution versus the 0.1 M solution. This would strongly suggest the contribution from the LA tetramer since this is the only species that could generate such an increase based on the theoretical modelling. A detailed analysis allows one to assign most of the observed VCD bands to the tetramer, indicated by the dashed lines in Fig. 5, with some small contributions from the LA dimer and trimer. The observed positive VCD band marked with 3 is reproduced by the LA tetramer. A minor shift in the predicted C[double bond, length as m-dash]O band positions marked with 1 and 2 may generate the negative band observed, although the current prediction shows a +/− VCD pattern for 1/2. Based on the gradual evolution of the IR and VCD spectral features from the 10 K, to 16 K, 24 K, to 0.1 M and finally to 0.2 M solution conditions, we can confidently conclude that even for the 0.2 M solution there is significant contribution to the observed IR and VCD spectra from the large LA aggregates such as LA tetramer. The simulated IR and VCD spectra with empirical Boltzmann factors of 20% LA dimer, 30% LA trimer and 50% LA tetramer are compared with the experimental data obtained at 24 K and in 0.1 M and 0.2 M solution in Fig. S12, ESI, showing very good agreement with the 0.2 M experimental data.

Conclusions

In this study, we have undertaken MI-IR and MI-VCD measurements of LA at 10 K, 16 K and 24 K temperatures and obtained spectra of LA dominated by the monomer and by larger aggregates sequentially. The sharp and characteristic MI-VCD spectral features at 10 K provide the essential experimental tool to conclusively assign the IR bands belonging to the LA monomer which show excellent agreement with the theoretical calculations. By following the experimental trend observed in the MI-IR and especially MI-VCD spectral features at the higher deposition temperatures, and aided with the B3LYP-D3BJ/def2-TZVPD calculations, we have been able to confidently identify the LA dimer as the main species at 16 K and 24 K. Finally, through the comparison of the experimental MI-VCD spectral features and those in the 0.2 M solution, in conjunction with the theoretical simulations, the main carriers have been identified as the LA tetramer and trimer in solution. This is different from the previous studies which identify the LA dimer as the main species at even higher concentrations. It is worth emphasizing that the characteristic VCD features are essential in achieving the conclusive assignment since IR features alone do not allow a clear assignment. The experimental data obtained with different degrees of LA self-aggregation are also crucial in facilitating the spectral assignment since the theoretical uncertainties in the vibrational band positions and the IR/VCD intensities, as well as the relative abundance of conformers become larger as one moves to larger aggregates. In addition, we show that the B3LYP functional (without the D3BJ dispersion correction) commonly used in VCD research gives very different relative energies as the molecular systems get larger and are placed in solution when compared to those with D3BJ correction. The current work highlights the importance of utilizing extensive, controlled experimental data to aid the IR and VCD spectroscopic interpretation, in addition to the high level DFT calculations.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was funded by the Natural Sciences and Engineering Research Council of Canada and by the University of Alberta. We thank Dr Javix Thomas for his involvement at the early stage of the matrix isolation experiments. We gratefully acknowledge access to the computing facilities by the Shared Hierarchical Academic Research Computing Network, the Western Canada Research Grid (Westgrid), and Compute/Calcul Canada. YX is a Tier I Canada Research Chair in Chirality and Chirality Recognition.

References

  1. L. A. Nafie, Vibrational Optical Activity: Principals and Applications, Wiley, Chichester, 2011 Search PubMed.
  2. J. M. Batista Jr., E. W. Blanch and V. da Silva Bolzan, Nat. Prod. Rep., 2015, 32, 1280–1302 RSC.
  3. A.-C. Chamayou, G. Makhloufi, L. A. Nafie, C. Janiak and S. Lüdeke, Inorg. Chem., 2015, 54, 2193–2203 CrossRef CAS.
  4. Y. Zhang, M. R. Poopari, X. Cai, A. Savin, Z. Dezhahang, J. Cheramy and Y. Xu, J. Nat. Prod., 2016, 79, 1012–1023 CrossRef CAS.
  5. T. Wu, X.-X. You and P. Bouř, Coord. Chem. Rev., 2015, 284, 1–18 CrossRef CAS.
  6. C. Merten, C. J. Berger, R. McDonald and Y. Xu, Angew. Chem., Int. Ed., 2014, 53, 9940–9943 CrossRef CAS.
  7. M. Passarello, S. Abbate, G. Longhi, S. Lepri, R. Ruzziconi and V. P. Nicu, J. Phys. Chem. A, 2014, 118, 4339–4350 CrossRef CAS PubMed.
  8. J. E. Rode, J. C. Dobrowolski, K. Lyczko, A. Wasiewicz, D. Kaczorek, R. Kawęcki, G. Zając and M. Baranska, J. Phys. Chem. A, 2017, 121, 6713–6726 CrossRef CAS.
  9. C. Merten, Phys. Chem. Chem. Phys., 2017, 19, 18803–18812 RSC.
  10. G. Yang and Y. Xu, Top. Curr. Chem., 2011, 298, 189–236 CrossRef CAS.
  11. J. Sadlej, J. C. Dobrowolski and J. E. Rode, Chem. Soc. Rev., 2010, 39, 1478–1488 RSC.
  12. M. R. Poopari, Z. Dezhahang, G. Yang and Y. Xu, ChemPhysChem, 2012, 13, 2310–2321 CrossRef CAS PubMed.
  13. A. S. Perera, J. Thomas, M. R. Poopari and Y. Xu, Front. Chem., 2016, 4, 1–17 Search PubMed.
  14. A. S. Perera, J. Cheramy, C. Merten, J. Thomas and Y. Xu, ChemPhysChem, 2018, 19, 2234–2242 CrossRef CAS.
  15. G. Tarczay, G. Magyarfalvi and E. Vass, Angew. Chem., Int. Ed., 2006, 45, 1775–1777 CrossRef CAS.
  16. C. H. Pollok and C. Merten, Phys. Chem. Chem. Phys., 2016, 18, 13496–13502 RSC.
  17. C. Merten and Y. Xu, ChemPhysChem, 2013, 14, 213–219 CrossRef CAS PubMed.
  18. C. Merten and Y. Xu, Angew. Chem., Int. Ed., 2013, 52, 2073–2076 CrossRef CAS.
  19. W. Sander and M. Gantenberg, Spectrochim. Acta, Part A, 2005, 62, 902–909 CrossRef.
  20. M. Kleerebezem and J. Hugenholtz, Curr. Opin. Biotechnol., 2003, 14, 232–237 CrossRef CAS.
  21. A. Narladkar, E. Balnois and Y. Grohens, Macromol. Symp., 2006, 241, 34–44 CrossRef CAS.
  22. B. P. van Eijck, J. Mol. Spectrosc., 1983, 101, 133–138 CrossRef CAS.
  23. L. Pszczółkowski, E. Białkowska-Jaworska and Z. Kisiel, J. Mol. Spectrosc., 2005, 234, 106–112 CrossRef.
  24. Z. Kisiel, E. Białkowska-Jaworska, D. P. Zaleski, J. L. Neill, A. L. Steber and B. H. Pate, WH10, 66th Int. Sym. on Mol. Spectrosc., Columbus, Ohio, 2011.
  25. A. Borda, A. Gomez-Zavaglia, L. Iapinski and R. Fausto, Phys. Chem. Chem. Phys., 2004, 6, 2101–2108 RSC.
  26. A. Shouten, J. Kanters and J. van Krieken, J. Mol. Struct., 1994, 323, 165–168 CrossRef.
  27. G. Cassanas, M. Morssli, E. Fabregue and L. Bardet, J. Raman Spectrosc., 1991, 22, 409–413 CrossRef CAS.
  28. G. Cassanas, G. Kister, E. Fabregue, M. Morssli and L. Bardet, Spectrochim. Acta, Part A, 1993, 49, 271–279 CrossRef.
  29. Z. Fekete, T. Kortvelyesi, J. Andor and I. Palinko, J. Mol. Struct.: THEOCHEM, 2003, 666, 159–162 CrossRef.
  30. M. Losada, H. Tran and Y. Xu, J. Chem. Phys., 2008, 128, 014508 CrossRef PubMed.
  31. M. J. Frisch, et al., Gaussian 16 (Revision B. 01) Search PubMed, full reference is included in the ESI.
  32. A. D. Becke, J. Chem. Phys., 1993, 98, 5648–5652 CrossRef CAS.
  33. R. Krishnan, J. S. Binkley, R. Seeger and J. A. Pople, J. Chem. Phys., 1980, 72, 650–654 CrossRef CAS.
  34. J. Cheeseman and M. Frisch, J. Chem. Theory Comput., 2007, 7, 3323–3334 CrossRef PubMed.
  35. T. Kuppens, K. Vandyck, J. van der Eycken, W. Herrebout, J. vander Veken and P. Bultinck, J. Org. Chem., 2005, 70, 9103–9114 CrossRef CAS.
  36. S. Grimme, S. Ehrlich and L. Goerigk, J. Comput. Chem., 2011, 32, 1456–1465 CrossRef CAS.
  37. A. D. Becke and E. R. Johnson, J. Chem. Phys., 2005, 123, 154101 CrossRef PubMed.
  38. M. K. Kesharwani, A. Karton and J. M. L. Martin, J. Chem. Theory Comput., 2016, 12, 444–454 CrossRef CAS PubMed.
  39. J. Thomas, N. A. Seifert, W. Jäger and Y. Xu, Angew. Chem., Int. Ed., 2017, 56, 6289–6293 ( Angew. Chem. , 2017 , 129 , 6386–6390 ) CrossRef CAS PubMed.
  40. D. Rappoport and F. Furche, J. Chem. Phys., 2010, 133, 134105 CrossRef PubMed.
  41. J. Sadlej, J. Dobrowolski, J. Rode and M. Jamroz, Phys. Chem. Chem. Phys., 2006, 8, 101–113 RSC.
  42. M. Pecul, A. Rizzo and J. Leszczynski, J. Phys. Chem. A, 2002, 106, 11008–11016 CrossRef CAS.
  43. A. Barnes, J. Mol. Struct., 1984, 113, 161–174 CrossRef CAS.
  44. G. Holzwarth and I. Chabay, J. Chem. Phys., 1972, 57, 1632–1635 CrossRef CAS.
  45. V. Setnička, M. Urbanová, P. Bouř, V. Král and K. Volka, J. Phys. Chem. A, 2001, 105, 8931–8938 CrossRef.
  46. T. Taniguchi and K. Monde, J. Am. Chem. Soc., 2012, 134, 3695–3698 CrossRef CAS.
  47. S. Abbate, G. Mazzeo, S. Meneghini, G. Longhi, S. E. Boiadjiev and D. A. Lightner, J. Phys. Chem. A, 2015, 119, 4261–4267 CrossRef CAS.
  48. V. P. Nicu and E. J. Baerends, Phys. Chem. Chem. Phys., 2009, 11, 6107–6118 RSC.
  49. G. Longhi, M. Tommasini, S. Abbate and P. L. Polavarapu, Chem. Phys. Lett., 2015, 639, 320–325 CrossRef CAS.

Footnote

Electronic supplementary information (ESI) available. See DOI: 10.1039/c8cp04748k

This journal is © the Owner Societies 2019