Using diode laser vibration-rotation-tunneling spectroscopy near 15 Thz (500 cm−1), we have measured and assigned 142 transitions to three a-type librational subbands of the water hexamer-d12 prism. These subbands reveal dramatically enhanced (ca. 1000×) tunneling splittings relative to the ground state. This enhancement is in agreement with that observed for the water dimer, trimer, and pentamer in this same frequency region. The water prism tunneling motion has been predicted to potentially describe the motions of water in interfacial and confined environments; hence, the results presented here indicate that excitation of librational vibrations has a significant impact on the hydrogen bond dynamics in these macroscopic environments.
INTRODUCTION
The challenge in developing a general predictive molecular-scale description of water is essentially that of correctly describing its extended and dynamic hydrogen-bonded network.1–9 High-resolution spectroscopy of water clusters provides a means of systematically untangling its complexities. In this context, the water hexamer is of particular interest, as it is the smallest cluster exhibiting a 3-D hydrogen bond network.4,10–23 Previous studies have revealed the existence of low energy cage, prism, and book hexamer isomers within a supersonic expansion.12,24–27 It has been established that the cage isomeric structure is the global minimum energy structure for the all-H2O isomer,12,28 although the prism isomer lies in close energetic proximity, and is actually the lowest energy isomer for the all-D2O isomer.4,25,28–36 Additionally, cyclic hexamer clusters have been observed in frozen rare gas matrices.37,38
The first detailed experimental study of a water hexamer cluster was reported by Liu et al., who observed and characterized terahertz spectra of the cage isomer in a supersonic expansion.12,24 More recently, Perez et al. observed the cage, prism, and book isomers (in addition to heptamer and nonamer clusters) in a supersonic expansion using broadband microwave spectroscopy.25,39 This study characterized the ground state structures of these isomers at high resolution and provided critical insights into the associated hydrogen bond network tunneling motions. A subsequent study by Richardson et al. provided an in-depth description of the tunneling motions of the prism, revealing a pathway which simultaneously breaks two hydrogen bonds.26 The structure of the water prism is shown in Fig. 1.
The lowest energy structure of the water hexamer prism. Oxygens are labeled as A–F, and hydrogens are labeled as 1–12. Reproduced with permission from Richardson et al., Science 351, 1310 (2016). Copyright 2016 AAAS.
The lowest energy structure of the water hexamer prism. Oxygens are labeled as A–F, and hydrogens are labeled as 1–12. Reproduced with permission from Richardson et al., Science 351, 1310 (2016). Copyright 2016 AAAS.
Here we report the measurement of a parallel vibration-rotation-tunneling (VRT) librational band of a (D2O)6 cluster centered in the 15 THz (510 cm−1) region. The experimental rotational constants describing the mass distribution in the cluster via the principal moments of inertia extracted from the present VRT spectra agree most closely with those predicted for the prism structure. We find a dramatic librational enhancement in the bifurcation tunneling splittings of the D2O cluster relative to the ground state, in analogy with other recent cluster studies in this spectral region.40–44
EXPERIMENTAL
Our previous investigations of water dimer, trimer, and pentamer librational motions in the 15 THz region prompted us to search for similar transitions in larger clusters.43,45 The Berkeley diode laser/supersonic beam spectrometer used in this study has been described in detail elsewhere and only a short description is provided here.46,47
A helium-cooled spectrometer (Spectra Physics) using lead-salt diodes (Laser Photonics) was used to produce infrared radiation from 509 to 514 cm−1. The beam was multipassed 18-22 times through a pulsed planar supersonic expansion of a mixture of D2O and He using a Herriott cell and detected using a helium-cooled (Si:B) photoconductive detector (IR Labs). The supersonic expansion was produced by bubbling pure He gas, with a backing pressure of 1-2 atm through liquid D2O (Cambridge Labs, 99.96% purity), and then expanding through a 101.6 mm long slit at a repetition rate of 35 Hz into a vacuum chamber maintained at ∼200 mTorr by a Roots blower (Edwards 4200) backed by two rotary pumps (E2M 275).48 Simultaneously, the fringe spacing of a vacuum-spaced etalon and carbonyl sulfide (OCS) reference gas spectra was detected with a liquid He cooled (Cu:Ge) detector (Santa Barbara Research Center) and recorded to enable precise frequency calibration. The observed linewidths of ∼30-40 MHz full-width half maximum (FWHM) are somewhat larger than the Doppler-limited linewidths extrapolated from earlier experiments using argon expansions. Typical frequency measurement accuracy is 10-20 MHz, limited by both linewidths of the cluster absorptions and laser drift. Spectra were detected in direct absorption using a time-gated phase sensitive signal processing approach.48
Accessing the 15 THz (500 cm−1) region of the electromagnetic spectrum has generally been notoriously difficult. The spectra reported here required the use of 10 separate laser diodes, each scanned across several modes to cover the specified spectral range. Moreover, large laser gaps are present in the spectra, which causes considerable difficulty in the assignment. Additionally, the spectrum reflects several distinct laser intensity fluctuations across different devices that are apparent in the complete spectra shown in Fig. 2. Specifically, between 512.4 cm−1 and 513.2 cm−1 the laser intensity is enhanced (as was found in a previous study41) and between 509.5 cm−1 and 510 cm−1 the laser intensity is depressed.
(a) All 188 transitions observed in the experimental range examined are shown in black. The 46 remaining unassigned transitions are shown in red. (b) Transitions assigned to subband 3, showing the typical pattern observed in these experiments. Laser gaps prohibited further assignment of the R and P branches.
(a) All 188 transitions observed in the experimental range examined are shown in black. The 46 remaining unassigned transitions are shown in red. (b) Transitions assigned to subband 3, showing the typical pattern observed in these experiments. Laser gaps prohibited further assignment of the R and P branches.
RESULTS AND ANALYSIS
Assignment
We have assigned 142 of the 188 observed transitions in the studied region using the PGOPHER platform.49 The transitions belong to three distinct a-type (ΔKa = 0) subbands, which are assigned to different tunneling levels of the water hexamer-d12 prism isomer. The observed transitions were fit to an S-reduced Watson Hamiltonian. For the assignment, we fixed the ground state constants to those obtained in Ref. 44. Correlation matrices of the fit and a list of all assigned transitions are given in the supplementary material.
Anharmonic excited state vibrational calculations
To elucidate the nature of the observed vibration, anharmonic frequencies and rotational constants were calculated by applying generalized second-order vibrational perturbation theory (GVPT2)50,51 to the second-order Møller–Plesset (MP2) perturbation theory potential energy surface using the Gaussian 09 D.01 package.52 The geometry of the lowest energy D2O prism was optimized at the MP2/aug-cc-pVDZ level of theory enforcing “very tight” convergence criteria, following which a GVPT2 calculation was performed at the same level of theory using the parameters recommended by Temelso and Shields.23 The Gaussian input file, the full list of calculated harmonic and anharmonic frequencies, and the rotational constants are listed in the supplementary material.
Fit analysis
We obtained a good quality fit of the observed transitions; the average RMS of the fit is ∼37 MHz as a result of the wavelength accuracy of <20 MHz and the observed linewidths of 30-40 MHz (Table I). We note that the A, B, and C constants of all three bands show good agreement with one another, providing evidence that these bands originate from tunneling sublevels of a common excited state. Additionally, a fit with floating ground state constants yielded good agreement with the observed values from Ref. 44, although we have chosen to keep those values fixed in the reported fit.
Molecular constants for the asymmetric top fit with fixed ground state constants. The uncertainty in the final two digits for each constant is given in parentheses. The values of the subband origin, A, B, C, and RMS error are given in MHz; all other values are in kHz. Ground state constants were kept fixed in the fit and obtained from Ref. 44.
. | Fit constants . | . | ||
---|---|---|---|---|
. | Subband 1 . | Subband 2 . | Subband 3 . | Ground state . |
Origin | 152 774 46(17) | 152 811 36(17) | 152 843 28(16) | 0 |
A | 1599.33(20) | 1601.84(28) | 1594.04(24) | 1493.9052(12) |
B | 1257.40(25) | 1251.51(56) | 1250.01(41) | 1218.3566(12) |
C | 1212.30(26) | 1203.86(58) | 1214.47(38) | 1185.6460(11) |
DK | 0.724(15) | −0.466(22) | −0.312(134) | −0.961(63) × 103 |
DJK | −0.697(11) | 0.307(27) | −0.045(20) | 1.662(58) × 103 |
DJ | 0.2245(44) | 0.0103(87) | 0.0476(82) | 0.4094(94) × 103 |
δK | 0.593(32) | 3.005(25) | 2.521(97) | −2.92(27) × 103 |
δJ | 0.0295(15) | 0.0140(97) | −0.0664(46) | 0.0511(63) × 103 |
HK | 0.0210(27) | 0.1283(20) | 0.0769(95) | 0 |
HKJ | −0.0136(30) | −0.1843(30) | −0.1224(14) | 0 |
HJK | −7.7(10) × 104 | 0.059(12) | 0.0452(54) | 0 |
HJ | 1.72(31) × 103 | −9.1(11) × 104 | −2.19(86) × 103 | 0 |
RMS (MHz) | 39.2 | 31.0 | 40.3 | |
Number of transitions | 51 | 38 | 53 |
. | Fit constants . | . | ||
---|---|---|---|---|
. | Subband 1 . | Subband 2 . | Subband 3 . | Ground state . |
Origin | 152 774 46(17) | 152 811 36(17) | 152 843 28(16) | 0 |
A | 1599.33(20) | 1601.84(28) | 1594.04(24) | 1493.9052(12) |
B | 1257.40(25) | 1251.51(56) | 1250.01(41) | 1218.3566(12) |
C | 1212.30(26) | 1203.86(58) | 1214.47(38) | 1185.6460(11) |
DK | 0.724(15) | −0.466(22) | −0.312(134) | −0.961(63) × 103 |
DJK | −0.697(11) | 0.307(27) | −0.045(20) | 1.662(58) × 103 |
DJ | 0.2245(44) | 0.0103(87) | 0.0476(82) | 0.4094(94) × 103 |
δK | 0.593(32) | 3.005(25) | 2.521(97) | −2.92(27) × 103 |
δJ | 0.0295(15) | 0.0140(97) | −0.0664(46) | 0.0511(63) × 103 |
HK | 0.0210(27) | 0.1283(20) | 0.0769(95) | 0 |
HKJ | −0.0136(30) | −0.1843(30) | −0.1224(14) | 0 |
HJK | −7.7(10) × 104 | 0.059(12) | 0.0452(54) | 0 |
HJ | 1.72(31) × 103 | −9.1(11) × 104 | −2.19(86) × 103 | 0 |
RMS (MHz) | 39.2 | 31.0 | 40.3 | |
Number of transitions | 51 | 38 | 53 |
Tunneling dynamics
We refer the reader to the recent work of Richardson et al. for a detailed treatment of the water hexamer prism tunneling dynamics and instead focus only on the relevant considerations for the fully deuterated cluster.26 The feasible prism tunneling motions can be described by the complete nuclear permutation inversion (CNPI) subgroup isomorphic to point group D2d. Previous work has shown that the water hexamer prism exhibits two feasible tunneling motions (here we define “feasible” as “experimentally observed”53), which are referred to as Pa and Pg. In the CNPI notation, Pa = (A D)(B F)(C E)(1 7)(2 8)(3 11)(4 12)(5 9)(6 10), where the labels correspond to the structure in Fig. 1. The motion associated with this element involves a double flip of the free hydrogens, resulting in breaking a single hydrogen bond. Likewise, in the CNPI notation, Pg = (A D)(B F)(C E)(1 8 2 7)(3 11)(4 12)(5 9)(6 10), and the motion is described as a double flip accompanied by a bifurcation, which breaks two hydrogen bonds. The character table for this group is given in Table II.
Character table for the point group D2d; group elements correspond to CNPI operations.
D2d . | E . | 2Pg . | (12)(78) . | 2Pa . | 2(12) . |
---|---|---|---|---|---|
A1 | 1 | 1 | 1 | 1 | 1 |
A2 | 1 | 1 | 1 | −1 | −1 |
B1 | 1 | −1 | 1 | 1 | −1 |
B2 | 1 | −1 | 1 | −1 | 1 |
E | 2 | 0 | −2 | 0 | 0 |
Γns | 531 441 | 243 | 59 049 | 729 | 177 147 |
Γdip | 3 | −1 | 3 | −1 | 3 |
D2d . | E . | 2Pg . | (12)(78) . | 2Pa . | 2(12) . |
---|---|---|---|---|---|
A1 | 1 | 1 | 1 | 1 | 1 |
A2 | 1 | 1 | 1 | −1 | −1 |
B1 | 1 | −1 | 1 | 1 | −1 |
B2 | 1 | −1 | 1 | −1 | 1 |
E | 2 | 0 | −2 | 0 | 0 |
Γns | 531 441 | 243 | 59 049 | 729 | 177 147 |
Γdip | 3 | −1 | 3 | −1 | 3 |
The character representation of the nuclear spin wavefunction (Γns) and the electric dipole moment (Γdip) is given at the bottom of Table II. We can reduce Γns to its irreducible representation: .
For the fully deuterated hexamer prism, the total spin wavefunction must transform as A1 which leads to the spin statistical weights: A1:A2:B1:B2:E of 487:121:122:485:486. The irreducible representation of the electric dipole moment is and for a-type (ΔKa = 0) transitions, the dipole transforms as B2. Thus, the expected selection rules are A1 <-> B2, A2 <-> B1, and E <-> E. These selection rules lead to a “doublet of triplets” pattern with a central doublet, as shown in Fig. 3. Richardson et al. established that the energy level pattern of the hexamer prism is represented by the eigenvalue of the tunneling matrix shown in Fig. 3(a).26
(a) Tunneling Hamiltonian of the water hexamer prism. The term v represents the band origin. The terms hg and ha represent the tunneling splittings associated with the group elements Pg and Pa, respectively. (b) Energy level diagram resulting from the two feasible tunneling motions present in the water prism hexamer [n.b. the subscript on the E symmetry labels merely corresponds to whether the level belongs to the upper (2) or lower (1) triplet]. (c) Energies of the selection rule-allowed tunneling transitions for the prism hexamer. E(J, K) represents the typical rigid rotor energy level. (N.B. E1 and E2 energy levels refer to whether the level belongs to the upper or lower energy triplet.) (d) Energy level diagram depicting the level ordering shown in (b). The diagram assumes that all tunneling elements are positive and hg′ < hg″, which was an arbitrary choice for the figure.
(a) Tunneling Hamiltonian of the water hexamer prism. The term v represents the band origin. The terms hg and ha represent the tunneling splittings associated with the group elements Pg and Pa, respectively. (b) Energy level diagram resulting from the two feasible tunneling motions present in the water prism hexamer [n.b. the subscript on the E symmetry labels merely corresponds to whether the level belongs to the upper (2) or lower (1) triplet]. (c) Energies of the selection rule-allowed tunneling transitions for the prism hexamer. E(J, K) represents the typical rigid rotor energy level. (N.B. E1 and E2 energy levels refer to whether the level belongs to the upper or lower energy triplet.) (d) Energy level diagram depicting the level ordering shown in (b). The diagram assumes that all tunneling elements are positive and hg′ < hg″, which was an arbitrary choice for the figure.
Based on these considerations, we assign the B2 → A1, E2 → E1, and A2 → B1 transitions to the subbands 1, 2, and 3, respectively. This assignment is based on the equal spacing between three subbands and the intensity ratio of 1:1:0.85. While this intensity ratio does not agree quantitatively with the expected ratio (based on spin statistics above), given the large intensity fluctuations of the lasers used, we consider this assignment as the most likely of those possible under parallel selection rules. Based on this assignment, we can estimate the tunneling matrix element, hg. The separation between subband 1 and subband 2 corresponds to the quantity 2hg′ + 2hg″, wherein the prime and double prime represent the values in the excited and ground state, respectively. The separation between subbands 2 and 3 corresponds to the same quantity. Given the measured values in the ground state (which are <1 MHz), we can explicitly calculate the value for the excited state (assuming these transitions originate in the ground vibrational state). Based on our fitted values of the band origins, we find the average value of hg′ to be ∼1720 MHz, which corresponds to a ca. 1000× enhancement of that tunneling splitting relative to the value for the ground state. This enhancement would be consistent with the enhancement of tunneling splitting observed for the water dimer, trimer, and pentamer upon excitation of a single quantum of 15 Thz librational vibrations. We note that using the intensity ratio to assign the observed subbands to the lower triplet is in some ways arbitrary due to the power fluctuations of the diode laser. It can be argued that the subbands could be assigned to the upper triplet just as easily; however, that assignment does not change the analysis of the tunneling enhancement, as both triplets are symmetric. Further experiments are needed to definitively establish which triplet these subbands represent.
DISCUSSION
Vibrational origin
These experiments were conducted in a He supersonic expansion, which cools the rotational temperature to approximately 10 K and the vibrational temperature to ca. 100 K. The most likely vibrational origin of the transitions is the ground state. Based on the calculated anharmonic vibrational frequencies and assuming that the transitions observed originate from the absolute ground state, we attempted to assign a vibrational mode to the experimental transitions; however, we note several interesting observations. We find two likely candidates for anharmonic vibrational modes with transition frequencies from the absolute ground state of 505.693 and 539.359 cm−1, both of which lie close to the observed average band origin of 509.7183 cm−1. However, we stress that this experimental band origin corresponds to the average origin of individual tunneling bands and is not the true vibrational band origin. In Table III, we show a comparison between the observed rotational constants and the predicted values for the 4 closest vibrational modes.
Comparison of observed and calculated values of D2O hexamer prism rotational constants. All values are reported in cm−1.
. | Origin . | A . | B . | C . | ΔA . | ΔB . | ΔC . |
---|---|---|---|---|---|---|---|
Observed | 509.718 | 0.049 831 | 0.041 795 | 0.040 368 | 0.003 486 | 0.001 155 | 0.000 819 |
Calculated | 635.026 | 0.049 008 | 0.040 504 | 0.039 004 | −0.000 141 | −0.000 202 | −0.000 125 |
539.359 | 0.049 009 | 0.040 609 | 0.038 967 | −0.000 140 | −0.000 097 | −0.000 162 | |
505.693 | 0.049 097 | 0.040 533 | 0.039 039 | −0.000 052 | −0.000 173 | −0.000 090 | |
426.105 | 0.049 251 | 0.040 520 | 0.039 026 | 0.000 102 | −0.000 186 | −0.000 103 |
. | Origin . | A . | B . | C . | ΔA . | ΔB . | ΔC . |
---|---|---|---|---|---|---|---|
Observed | 509.718 | 0.049 831 | 0.041 795 | 0.040 368 | 0.003 486 | 0.001 155 | 0.000 819 |
Calculated | 635.026 | 0.049 008 | 0.040 504 | 0.039 004 | −0.000 141 | −0.000 202 | −0.000 125 |
539.359 | 0.049 009 | 0.040 609 | 0.038 967 | −0.000 140 | −0.000 097 | −0.000 162 | |
505.693 | 0.049 097 | 0.040 533 | 0.039 039 | −0.000 052 | −0.000 173 | −0.000 090 | |
426.105 | 0.049 251 | 0.040 520 | 0.039 026 | 0.000 102 | −0.000 186 | −0.000 103 |
From the table we can see that the observed values are all larger in magnitude than the predicted values. The most striking observation in the calculations is that the calculated values of ΔM (where M is A, B, or C) are opposite in sign and about an order of magnitude smaller than what is observed experimentally. Some of that disagreement can be attributed to the fact that there are large perturbations to the excited state, shown by the large higher order terms in the fit results, indicating a very “floppy” excited state. We have previously observed dramatic fluctuations of the centrifugal distortion constants in this experimental region.41,43,45 Another possibility would be that the ground state of these transitions is not actually the absolute ground state of this cluster, but rather a “hot band.” While this is unlikely, given the strong cooling in a supersonic expansion, there are several vibrational states located within 10 cm−1 of the absolute ground state. However, further experimental studies and calculations will be needed to determine a definitive resolution of this disagreement. We also stress that these values represent only the lower triplet of the tunneling pattern, which should be fairly representative of the excited state.
Librational enhancement of tunneling
We have previously studied the water dimer, trimer, and pentamer in the 15 Thz librational region, with the most salient observation being the dramatic enhancement of the tunneling splittings in these clusters.40,41,43 We observe a similar effect here; however, we can only determine the enhancement for the hg pathway. As described by Richardson et al.,26 the hg pathway motions are characterized by a double flip accompanied by a bifurcation, which involves the breaking of two hydrogen bonds. Studies of the water dimer, trimer, and pentamer tunneling in the 500 cm−1 region involve motions which only break a single hydrogen bond.40,41,43 We observe a 1000× enhancement relative to the observed ground state tunneling for the (D2O)6 prism from microwave experiments.26 This observation indicates that the observed enhancement in the librational region does not seem to be influenced by the number of hydrogen bonds broken in the tunneling pathway. Additionally, the reported enhancement is with respect to the observed tunneling in the (H2O)6 prism reported by Richardson et al.26 As observations of the tunneling in the fully deuterated ground state do not presently exist, an accurate measure of the enhancement cannot be obtained; however, due to the larger mass-weighted path for the deuterated species, we can consider the observed 1000× enhancement as a lower bound.
Due to the absence of the higher energy triplet [Fig. 3(c)] in our measured spectra, we cannot presently characterize the ha tunneling pathway in this excited librational state. The absence of this triplet is most likely a result of the large laser gaps present in the experiment, coupled with the fact that the observed transitions occur near the edge of the available laser coverage.
The results reported here are significant, as the water hexamer represents a transition of the minimum energy structure of water clusters from ring-like forms to a fully 3-D structure. The water prism tunneling motion has been predicted to potentially describe the motions of water in interfacial and confined environments; hence, the results presented here indicate that excitation of librational vibrations has a significant impact on the hydrogen bond dynamics in these macroscopic environments.
CONCLUSIONS
We have measured 3 a-type subbands belonging to a common librational vibration of the (D2O)6 prism cluster. Assigning the transitions to tunneling sublevels allows us to extract a value of ∼1720 MHz for the hg tunneling motion, representing a ca. 1000× enhancement of the tunneling splitting, relative to the ground state splitting observed for the (H2O)6 cluster. This enhancement is consistent with the dramatic tunneling enhancement observed previously for the water dimer, trimer, and pentamer in the same (15 THz) experimental region.
From comparison to theoretical calculations, we find the observed change in the all-rotational constants to be dramatically larger than predicted. This large change indicates a very “floppy” excited state, which would be consistent with the large tunneling splittings observed and the breaking of one or two hydrogen bonds in the tunneling pathway.
SUPPLEMENTARY MATERIAL
See supplementary material for a list of all the assigned transitions in Table S1 and correlation matrices for the fits in Table S2. Calculated anharmonic vibrational frequencies are given in Table S3. The Gaussian input file for calculations is given in Fig. S1.
ACKNOWLEDGMENTS
The Berkeley Terahertz project was previously supported by the Chemical Structure, Dynamics, and Mechanisms-A Division of the National Science Foundation under Grant No. 1300723. This project is currently supported by the CALSOLV collaboration, an affiliate program of RESOLV (Ruhr-Unversitaet Bochum). D.J.W. gratefully acknowledges financial support from the EPSRC.